Handbook of Linear Partial Differential Equations For Engineers and Scientists-Andrei D. Polyanin, Vladimir E. Nazaikinskii PDF

Mathematics, Engineering, Physics
SECOND EDITION
ENGINEERS AND SCIENTISTS
DIFFERENTIAL EQUATIONS FOR
LINEAR PARTIAL
HANDBOOK OF
HANDBOOK OF
LINEAR PARTIAL
SECOND EDITION
UPDATED, REVISED AND EXTENDED
Features
• Includes nearly 4,000 linear partial differential equations (PDEs) with solutions
• Presents solutions of numerous problems relevant to heat and mass
transfer, wave theory, hydrodynamics, aerodynamics, elasticity, acoustics,
electrodynamics, diffraction theory, quantum mechanics, chemical engineering
sciences, electrical engineering, and other fields
• Outlines basic methods for solving various problems in science and
engineering
• Contains many more linear equations, problems, and solutions than any other
book currently available
• Provides a database of test problems for numerical and approximate analytical
methods for solving linear PDEs and systems of coupled PDEs
New to the Second Edition

• More than 700 pages with 1,500+ new first-, second-, third-, fourth-, and
higher-order linear equations with solutions
• Systems of coupled PDEs with solutions
• Some analytical methods, including decomposition methods and their
applications
• Symbolic and numerical methods for solving linear PDEs with Maple,
Mathematica, and MATLAB®
• Many new problems, illustrative examples, tables, and figures
To accommodate different mathematical backgrounds, the authors avoid wherever Polyanin

possible the use of special terminology, outline some of the methods in a schematic,
simplified manner, and arrange the material in increasing order of complexity.
Nazaikinskii
K18874
w w w. c rc p r e s s . c o m
K18874_cover.indd 1 12/8/15 9:10 AM

HANDBOOK OF
LINEAR PARTIAL
SECOND EDITION
HANDBOOK OF
LINEAR PARTIAL
SECOND EDITION
Andrei D. Polyanin
Vladimir E. Nazaikinskii
MATLAB® is a trademark of The MathWorks, Inc. and is used with permission. The MathWorks does not warrant the
accuracy of the text or exercises in this book. This book’s use or discussion of MATLAB® software or related products
does not constitute endorsement or sponsorship by The MathWorks of a particular pedagogical approach or particular
use of the MATLAB® software.
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2016 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business
No claim to original U.S. Government works

Version Date: 20151119
International Standard Book Number-13: 978-1-4665-8149-4 (eBook - PDF)
This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been
made to publish reliable data and information, but the author and publisher cannot assume responsibility for the valid-
ity of all materials or the consequences of their use. The authors and publishers have attempted to trace the copyright
holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this
form has not been obtained. If any copyright material has not been acknowledged please write and let us know so we may
rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or uti-
lized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopy-
ing, microfilming, and recording, or in any information storage or retrieval system, without written permission from the
publishers.
For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://
www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923,
978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For
organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for
identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
CONTENTS
Preface to the Second Edition xxv

Preface to the First Edition xxvi
Authors xxix
Notation xxxi
Part I Exact Solutions 1

1 First-Order Equations with Two Independent Variables 3
1.1 Equations of the Form f (x, y) ∂w
∂x + g(x, y) ∂w
=0 .........................
∂y 3
1.1.1 Equations Containing Power-Law Functions . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.2 Equations Containing Exponential Functions . . . . . . . . . . . . . . . . . . . . . . . . 22
1.1.3 Equations Containing Hyperbolic Functions . . . . . . . . . . . . . . . . . . . . . . . . 29
1.1.4 Equations Containing Logarithmic Functions . . . . . . . . . . . . . . . . . . . . . . . 33
1.1.5 Equations Containing Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . 36
1.1.6 Equations Containing Inverse Trigonometric Functions . . . . . . . . . . . . . . . 45
1.1.7 Equations Containing Arbitrary Functions of x . . . . . . . . . . . . . . . . . . . . . . 51
1.1.8 Equations Containing Arbitrary Functions of Different Arguments . . . . . 59
1.2 Equations of the Form f (x, y) ∂w ∂w
∂x + g(x, y) ∂y = h(x, y) . . . . . . . . . . . . . . . . . . . 66
1.2.7 Equations Containing Arbitrary Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 85
∂x + g(x, y) ∂y = h(x, y)w . . . . . . . . . . . . . . . . . . 91
∂x + g(x, y) ∂y = h1 (x, y)w + h0 (x, y) . . . . . . . 114
v
vi C ONTENTS
2 First-Order Equations with Three or More Independent Variables 139

2.1 Equations of the Form f (x, y, z) ∂w
∂x + g(x, y, z) ∂w
∂y + h(x, y, z) ∂w
=0 .......
∂z 139
2.2 Equations of the Form f1 ∂w ∂w ∂w
∂x + f2 ∂y + f3 ∂z = f4 , fn = fn (x, y, z) . . . . . . . . . 170
∂x + f2 ∂y + f3 ∂z = f4 w, fn = fn (x, y, z) . . . . . . . 196
∂x + f2 ∂y + f3 ∂z = f4 w + f5 , fn = fn (x, y, z) . . . 222
2.4.8 Underdetermined Equations Containing Operator div . . . . . . . . . . . . . . . . 248
2.4.9 Equations with Four or More Independent Variables . . . . . . . . . . . . . . . . . 251
3 Second-Order Parabolic Equations with One Space Variable 261
3.1 Constant Coefficient Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
∂2w
3.1.1 Heat Equation ∂w∂t = a ∂x2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
∂w 2
3.1.2 Equation of the Form ∂t = a ∂∂xw2 + Φ(x, t) . . . . . . . . . . . . . . . . . . . . . . . . . 270
∂w 2
3.1.3 Equation of the Form ∂t = a ∂∂xw2 + bw + Φ(x, t) . . . . . . . . . . . . . . . . . . . . 275
∂w 2
3.1.4 Equation of the Form ∂t = a ∂∂xw2 + b ∂w
∂x + Φ(x, t) . . . . . . . . . . . . . . . . . . . 280
∂w 2
3.1.5 Equation of the Form ∂t = a ∂∂xw2 + b ∂w
∂x + cw + Φ(x, t) . . . . . . . . . . . . . . 284
C ONTENTS vii
3.2 Heat Equation with Axial or Central Symmetry and Related Equations . . . . . . . 288
∂2w

3.2.1 Equation of the Form ∂w 1 ∂w
∂t = a ∂r 2 + r ∂r . . . . . . . . . . . . . . . . . . . . . . . . . 288
∂2w
∂t = a ∂r 2 + r ∂r + Φ(r, t) . . . . . . . . . . . . . . . . . 294
∂2w
∂t = a ∂r 2 + r ∂r . . . . . . . . . . . . . . . . . . . . . . . . . 298
∂2w
∂t = a ∂r 2 + r ∂r + Φ(r, t) . . . . . . . . . . . . . . . . . 305
∂2w 1−2β ∂w
3.2.5 Equation of the Form ∂w
∂t = ∂x2 + x ∂x . . . . . . . . . . . . . . . . . . . . . . . . . 308
∂2w 1−2β ∂w
3.2.6 Equation of the Form ∂w
∂t = ∂x2 + x ∂x + Φ(x, t) . . . . . . . . . . . . . . . . 311
3.3 Equations Containing Power Functions and Arbitrary Parameters . . . . . . . . . . . . 312
∂2w
3.3.1 Equations of the Form ∂w
∂t = a ∂x2 + f (x, t)w . . . . . . . . . . . . . . . . . . . . . . . 312
∂2w
3.3.2 Equations of the Form ∂w ∂w
∂t = a ∂x2 + f (x, t) ∂x . . . . . . . . . . . . . . . . . . . . . 318
2
3.3.3 Equations of the Form ∂w ∂ w ∂w
∂t = a ∂x2 + f (x, t) ∂x + g(x, t)w + h(x, t) . . . . 321
2
∂t = (ax + b) ∂x2 + f (x, t) ∂x + g(x, t)w . . . . . . 324
2
3.3.5 Equations of the Form ∂w 2 ∂ w ∂w
∂t = (ax + bx + c) ∂x2 + f (x, t) ∂x + g(x, t)w 327
2
∂t = f (x) ∂x2 + g(x, t) ∂x + h(x, t)w . . . . . . . . . 329
2
∂t = f (x, t) ∂x2 + g(x, t) ∂x + h(x, t)w . . . . . . . 334
∂2w
3.3.8 Liquid-Film Mass Transfer Equation (1 − y 2 ) ∂w ∂x = a ∂y 2 . . . . . . . . . . . . . 335
∂2w
3.3.9 Equations of the Form f (x, y) ∂w ∂w
∂x + g(x, y) ∂y = ∂y 2
+ h(x, y) . . . . . . . 338
3.4 Equations Containing Exponential Functions and Arbitrary Parameters . . . . . . . 338
∂2w
3.4.1 Equations of the Form ∂w ∂t = a ∂x2 + f (x, t)w . . . . . . . . . . . . . . . . . . . . . . . 338
∂2w
∂t = a ∂x2 + f (x, t) ∂x . . . . . . . . . . . . . . . . . . . . . 341
∂2w
∂t = a ∂x2 + f (x, t) ∂x + g(x, t)w . . . . . . . . . . . . 344
3.4.4 Equations of the Form ∂w n ∂2w ∂w
∂t = ax ∂x2 + f (x, t) ∂x + g(x, t)w . . . . . . . . . 345
3.4.5 Equations of the Form ∂w βx ∂ 2 w + f (x, t) ∂w + g(x, t)w . . . . . . . . .
∂t = ae ∂x2 ∂x 346
3.4.6 Other Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
3.5 Equations Containing Hyperbolic Functions and Arbitrary Parameters . . . . . . . . 349
3.5.1 Equations Containing a Hyperbolic Cosine . . . . . . . . . . . . . . . . . . . . . . . . . 349
3.5.2 Equations Containing a Hyperbolic Sine . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
3.5.3 Equations Containing a Hyperbolic Tangent . . . . . . . . . . . . . . . . . . . . . . . . 351
3.5.4 Equations Containing a Hyperbolic Cotangent . . . . . . . . . . . . . . . . . . . . . . 352
3.6 Equations Containing Logarithmic Functions and Arbitrary Parameters . . . . . . . 354
∂2w
∂t = a ∂x2 + f (x, t) ∂x + g(x, t)w . . . . . . . . . . . . 354
2
3.6.2 Equations of the Form ∂w k∂ w ∂w
∂t = ax ∂x2 + f (x, t) ∂x + g(x, t)w . . . . . . . . . 354
3.7 Equations Containing Trigonometric Functions and Arbitrary Parameters . . . . . 356
3.7.1 Equations Containing a Cosine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
3.7.2 Equations Containing a Sine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
3.7.3 Equations Containing a Tangent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
3.7.4 Equations Containing a Cotangent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
3.8 Equations Containing Arbitrary Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
∂2w
3.8.1 Equations of the Form ∂w ∂t = a ∂x2 + f (x, t)w . . . . . . . . . . . . . . . . . . . . . . . 360
∂2w
∂t = a ∂x2 + f (x, t) ∂x . . . . . . . . . . . . . . . . . . . . . 363
∂2w
∂t = a ∂x2 + f (x, t) ∂x + g(x, t)w . . . . . . . . . . . . 368
viii C ONTENTS
2
3.8.4 Equations of the Form ∂w n∂ w ∂w
∂t = ax ∂x2 + f (x, t) ∂x + g(x, t)w . . . . . . . . . 370
2
3.8.5 Equations of the Form ∂w ∂t = ae
βx ∂ w + f (x, t) ∂w + g(x, t)w . . . . . . . . .
∂x2 ∂x 372
∂2w
3.8.6 Equations of the Form ∂t = f (x) ∂x2 + g(x, t) ∂w
∂w
∂x + h(x, t)w . . . . . . . . . 373
∂w ∂2w ∂w
3.8.7 Equations of the Form ∂t = f (t) ∂x2 + g(x, t) ∂x + h(x, t)w . . . . . . . . . 382
∂2w
3.8.8 Equations of the Form ∂w ∂t = f (x, t) ∂x + g(x, t) ∂w + h(x, t)w . . . . . . . 385
∂w ∂
2 ∂w ∂x
3.8.9 Equations of the Form s(x) ∂t = ∂x p(x) ∂x − q(x)w + Φ(x, t) . . . . . . 388
3.9 Equations of Special Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
3.9.1 Equations of the Diffusion (Thermal) Boundary Layer . . . . . . . . . . . . . . . 393
~2 ∂ 2 w
3.9.2 One-Dimensional Schrödinger Equation i~ ∂w ∂t = − 2m ∂x2 + U (x)w . . . 396
4 Second-Order Parabolic Equations with Two Space Variables 401
4.1 Heat Equation ∂w
∂t = a∆2 w . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
4.1.1 Boundary Value Problems in Cartesian Coordinates . . . . . . . . . . . . . . . . . . 401
4.1.2 Problems in Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 416
4.1.3 Axisymmetric Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423
4.2 Heat Equation with a Source ∂w ∂t = a∆2 w + Φ(x, y, t) . . . . . . . . . . . . . . . . . . . . . . 434
4.2.1 Problems in Cartesian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434
4.3 Other Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
4.3.1 Equations Containing Arbitrary Parameters . . . . . . . . . . . . . . . . . . . . . . . . . 455
5 Second-Order Parabolic Equations with Three or More Space Variables 463
5.1 Heat Equation ∂w
∂t = a∆3 w . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463
5.1.2 Problems in Cylindrical Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 487
5.1.3 Problems in Spherical Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517
5.2 Heat Equation with Source ∂w
∂t = a∆3 w + Φ(x, y, z, t) . . . . . . . . . . . . . . . . . . . . . 522
5.3 Other Equations with Three Space Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 537
5.3.3 Equations of the Form ρ(x, y, z) ∂w ∂t =
div[a(x, y, z)∇w] − q(x, y, z)w + Φ(x, y, z, t) . . . . . . . . . . . . . . . . . . . . . . 542
5.4 Equations with n Space Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 545
5.4.1 Equations of the Form ∂w
∂t = a∆n w + Φ(x1 , . . . , xn , t) . . . . . . . . . . . . . . . 545
5.4.2 Other Equations Containing Arbitrary Parameters . . . . . . . . . . . . . . . . . . . 548
6 Second-Order Hyperbolic Equations with One Space Variable 557
6.1 Constant Coefficient Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 557
2
6.1.1 Wave Equation ∂∂tw 2 ∂2w . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2 = a ∂x2 557
2
6.1.2 Equations of the Form ∂∂tw 2 ∂2w
2 = a ∂x2 + Φ(x, t) . . . . . . . . . . . . . . . . . . . . . . 563
2
6.1.3 Equation of the Form ∂∂tw 2 ∂2w
2 = a ∂x2 − bw + Φ(x, t) . . . . . . . . . . . . . . . . . . 567
C ONTENTS ix
∂2w 2
6.1.4 Equation of the Form ∂t2
= a2 ∂∂xw2 − b ∂w
∂x + Φ(x, t) . . . . . . . . . . . . . . . . . 571
∂2w 2
= a2 ∂∂xw2 + b ∂w
∂x + cw + Φ(x, t) . . . . . . . . . . . . 574
6.2 Wave Equations with Axial or Central Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . 577
2
2 ∂ 2 w + 1 ∂w

6.2.1 Equation of the Form ∂∂tw2 =a ∂r 2 r ∂r ....................... 577
∂2w 2 ∂2w 1 ∂w

6.2.2 Equation of the Form ∂t2 = a ∂r2 + r ∂r + Φ(r, t) . . . . . . . . . . . . . . . 580
2
2 ∂ 2 w + 2 ∂w

6.2.3 Equation of the Form ∂∂tw2 =a ∂r 2 r ∂r ....................... 581
∂2w 2 ∂2w 2 ∂w

6.2.4 Equation of the Form ∂t2 = a ∂r2 + r ∂r + Φ(r, t) . . . . . . . . . . . . . . . 585
2
2 ∂ 2 w + 1 ∂w − bw + Φ(r, t) . . . . . . . . . .
6.2.5 Equation of the Form ∂∂tw2 =a ∂r 2 r ∂r 586
2
2 ∂ 2 w + 2 ∂w − bw + Φ(r, t) . . . . . . . . . .
6.2.6 Equation of the Form ∂∂tw2 =a ∂r 2 r ∂r 590
6.3 Equations Containing Power Functions and Arbitrary Parameters . . . . . . . . . . . . 593
2 ∂2w
6.3.1 Equations of the Form ∂∂tw ∂w
2 = (ax + b) ∂x2 + c ∂x + kw + Φ(x, t) . . . . . . 593
2 2
6.3.2 Equations of the Form ∂∂tw 2 ∂ w ∂w
2 = (ax + b) ∂x2 + cx ∂x + kw + Φ(x, t) . . . . 598
6.3.3 Other Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 600
6.4 Equations Containing the First Time Derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . 607
2
6.4.1 Equations of the Form ∂∂tw ∂w 2 ∂2w ∂w
2 + k ∂t = a ∂x2 + b ∂x + cw + Φ(x, t) . . . . . 607
2 2
6.4.2 Equations of the Form ∂∂tw ∂w ∂ w
2 +k ∂t =f (x) ∂x2 +g(x) ∂x +h(x)w+Φ(x, t)
∂w
616
6.4.3 Other Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
6.5 Equations Containing Arbitrary Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623
2
6.5.1 Equations of the Form s(x) ∂∂tw ∂ ∂w
2 = ∂x p(x) ∂x − q(x)w + Φ(x, t) . . . . . 623
2
6.5.2 Equations of the Form ∂∂tw 2 + a(t) ∂t =
∂w
∂ ∂w

b(t) ∂x p(x) ∂x − q(x)w + Φ(x, t) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 626
6.5.3 Other Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 628
7 Second-Order Hyperbolic Equations with Two Space Variables 633
∂2w
7.1 Wave Equation ∂t2 = a2 ∆
2w . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633
2
7.2 Nonhomogeneous Wave Equation ∂∂tw 2
2 = a ∆2 w + Φ(x, y, t) . . . . . . . . . . . . . . . 651
2
7.3 Equations of the Form ∂∂tw 2
2 = a ∆2 w − bw + Φ(x, y, t) . . . . . . . . . . . . . . . . . . . . . 659
2
7.4 Telegraph Equation ∂∂tw ∂w 2
2 + k ∂t = a ∆2 w − bw + Φ(x, y, t) . . . . . . . . . . . . . . . . . 676
7.5 Other Equations with Two Space Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693
x C ONTENTS
8 Second-Order Hyperbolic Equations with Three or More Space Variables 695

2
8.1 Wave Equation ∂∂tw 2
2 = a ∆3 w . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 695
2
8.2 Nonhomogeneous Wave Equation ∂∂tw 2
2 = a ∆3 w + Φ(x, y, z, t) . . . . . . . . . . . . . 717
2
8.3 Equations of the Form ∂∂tw 2
2 = a ∆3 w − bw + Φ(x, y, z, t) . . . . . . . . . . . . . . . . . . . 720
2
8.4 Telegraph Equation ∂∂tw ∂w 2
2 + k ∂t = a ∆3 w − bw + Φ(x, y, z, t) . . . . . . . . . . . . . . . 743
8.5 Other Equations with Three Space Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 765
2
8.5.2 Equation of the Form ρ(x, y, z) ∂∂tw 2 =
div[a(x, y, z)∇w] − q(x, y, z)w + Φ(x, y, z, t) . . . . . . . . . . . . . . . . . . . . . . 765
8.6 Equations with n Space Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 768
2
8.6.1 Wave Equation ∂∂tw 2
2 = a ∆n w . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 768
2
8.6.2 Nonhomogeneous Wave Equation ∂∂tw 2
2 = a ∆n w + Φ(x1 , . . . , xn , t) . . . 770
2
8.6.3 Equations of the Form ∂∂tw 2
2 = a ∆n w − bw + Φ(x1 , . . . , xn , t) . . . . . . . . 773
8.6.4 Equations Containing the First Time Derivative . . . . . . . . . . . . . . . . . . . . . 776
9 Second-Order Elliptic Equations with Two Space Variables 781
9.1 Laplace Equation ∆2 w = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 781
9.1.1 Problems in Cartesian Coordinate System . . . . . . . . . . . . . . . . . . . . . . . . . . 781
9.1.2 Problems in Polar Coordinate System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 787
9.1.3 Other Coordinate Systems. Conformal Mappings Method . . . . . . . . . . . . 792
9.2 Poisson Equation ∆2 w = −Φ(x) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 794
9.2.1 Preliminary Remarks. Solution Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . 794
9.2.4 Arbitrary Shape Domain. Conformal Mappings Method . . . . . . . . . . . . . . 807
9.3 Helmholtz Equation ∆2 w + λw = −Φ(x) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 809
9.3.1 General Remarks, Results, and Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . 809
9.3.4 Other Orthogonal Coordinate Systems. Elliptic Domain . . . . . . . . . . . . . . 830
9.4 Other Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 832
9.4.1 Stationary Schrödinger Equation ∆2 w = f (x, y)w . . . . . . . . . . . . . . . . . . . 832
9.4.2 Convective Heat and Mass Transfer Equations . . . . . . . . . . . . . . . . . . . . . . 835
9.4.3 Equations of Heat and Mass Transfer in Anisotropic Media . . . . . . . . . . . 843
9.4.4 Other Equations Arising in Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . 851
2 2
9.4.5 Equations of the Form a(x) ∂∂xw2 + ∂∂yw2 + b(x) ∂w ∂x + c(x)w = −Φ(x, y) . 855
C ONTENTS xi
10 Second-Order Elliptic Equations with Three or More Space Variables 859

10.1 Laplace Equation ∆3 w = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 859
10.1.1 Problems in Cartesian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 859
10.1.2 Problems in Cylindrical Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 862
10.1.3 Problems in Spherical Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 863
10.1.4 Other Orthogonal Curvilinear Systems of Coordinates . . . . . . . . . . . . . . 866
10.2 Poisson Equation ∆3 w + Φ(x) = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 866
10.2.1 Preliminary Remarks. Solution Structure . . . . . . . . . . . . . . . . . . . . . . . . . 866
10.3 Helmholtz Equation ∆3 w + λw = −Φ(x) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 892
10.3.1 Homogeneous Helmholtz Equation. Eigenvalue problems . . . . . . . . . . 892
10.3.2 Nonhomogeneous Helmholtz Equation. General Remarks, Results,
and Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 893
10.3.6 Other Orthogonal Curvilinear Coordinates . . . . . . . . . . . . . . . . . . . . . . . . 929
10.4 Other Equations with Three Space Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 931
10.4.1 Equations Containing Arbitrary Functions . . . . . . . . . . . . . . . . . . . . . . . . 931
10.4.2 Equations of the Form div[a(x, y, z)∇w] − q(x, y, z)w = −Φ(x, y, z) 934
10.5 Equations with n Space Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 936
10.5.1 Laplace Equation ∆n w = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 936
10.5.2 Other Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 937
11 Higher-Order Partial Differential Equations 941
11.1 Third-Order Partial Differential Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 941
11.1.1 One-Dimensional Equations Containing the First Derivative in t . . . . . 941
11.1.2 One-Dimensional Equations Containing the Second Derivative in t . . 944
11.1.3 One-Dimensional Equations Containing a Mixed Derivative and the
First Derivative in t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 945
11.1.4 One-Dimensional Equations Containing a Mixed Derivative and the
Second Derivative in t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 951
11.1.5 Two- and Three-Dimensional Equations . . . . . . . . . . . . . . . . . . . . . . . . . . 954
11.2 Fourth-Order One-Dimensional Nonstationary Equations . . . . . . . . . . . . . . . . . . 957
11.2.1 Equation of the Form ∂w 2 ∂4w
∂t + a ∂x4 = Φ(x, t) . . . . . . . . . . . . . . . . . . . . . . 957
2
2 + a ∂x4 = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . 960
2
2 + a ∂x4 = Φ(x, t) . . . . . . . . . . . . . . . . . . . . . 962
2
2 + a ∂x4 + kw = Φ(x, t) . . . . . . . . . . . . . . . . 965
11.2.5 Other Equations without Mixed Derivatives . . . . . . . . . . . . . . . . . . . . . . . 968
11.2.6 Equations Containing Second Derivative in x and Mixed Derivatives . 971
11.2.7 Equations Containing Fourth Derivative in x and Mixed Derivatives . . 980
11.3 Two-Dimensional Nonstationary Fourth-Order Equations . . . . . . . . . . . . . . . . . . 986

2 ∂ 4 w + ∂ 4 w = Φ(x, y, t) . . . . . . . . . . . .
11.3.1 Equation of the Form ∂w ∂t + a ∂x4 ∂y 4
986
∂2w
+ a2 ∆∆w = 0 . . . . . . . . . . . . . . . . . . . . . . . . . 988
∂2w
+ a2 ∆∆w + kw = Φ(x, y, t) . . . . . . . . . . . . 991
xii C ONTENTS
2
2 ∂ 4 w + ∂ 4 w + kw = Φ(x, y, t) . . . . . .
11.3.4 Equation of the Form ∂∂tw2 +a ∂x 4 ∂y 4 993
11.3.5 Other Two-Dimensional Nonstationary Fourth-Order Equations . . . . . 994
11.4 Three- and n-Dimensional Nonstationary Fourth-Order Equations . . . . . . . . . . 997
2
11.4.1 Equation of the Form ∂∂tw 2
2 + a ∆∆w = 0 . . . . . . . . . . . . . . . . . . . . . . . . . 997
11.4.2 Equations Containing Mixed Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . 999
11.5 Fourth-Order Stationary Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1003
11.5.1 Biharmonic Equation ∆∆w = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1003
11.5.2 Equation of the Form ∆∆w = Φ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1009
11.5.3 Equation of the Form ∆∆w − λw = Φ(x, y) . . . . . . . . . . . . . . . . . . . . . . 1012
4 4
11.5.4 Equation of the Form ∂∂xw4 + ∂∂yw4 = Φ(x, y) . . . . . . . . . . . . . . . . . . . . . . . 1014
4 4
11.5.5 Equation of the Form ∂∂xw4 + ∂∂yw4 + kw = Φ(x, y) . . . . . . . . . . . . . . . . . . 1016
11.5.6 Stokes Equation (Axisymmetric Flows of Viscous Fluids) . . . . . . . . . . 1017
11.6 Higher-Order Linear Equations with Constant Coefficients . . . . . . . . . . . . . . . . . 1020
11.6.1 Fundamental Solutions. Cauchy Problem . . . . . . . . . . . . . . . . . . . . . . . . . 1020
11.6.2 Elliptic Operators and Elliptic Equations . . . . . . . . . . . . . . . . . . . . . . . . . 1022
11.6.3 Hyperbolic Operators and Hyperbolic Equations . . . . . . . . . . . . . . . . . . 1025
11.6.4 Regular Equations. Number of Initial Conditions in the Cauchy
Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1025
11.6.5 Some Equations with Two Independent Variables Containing the First
Derivative in t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1029
11.6.6 Some Equations with Two Independent Variables Containing the
Second Derivative in t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1035
11.6.7 Other Equations with Two Independent Variables . . . . . . . . . . . . . . . . . . 1039
11.6.8 Equations with Three and More Independent Variables . . . . . . . . . . . . . 1041
11.7 Higher-Order Linear Equations with Variable Coefficients . . . . . . . . . . . . . . . . . 1045
11.7.1 Equations Containing the First Time Derivative . . . . . . . . . . . . . . . . . . . 1045
11.7.2 Equations Containing the Second Time Derivative . . . . . . . . . . . . . . . . . 1050
11.7.3 Nonstationary Problems with Many Space Variables . . . . . . . . . . . . . . . 1052
11.7.4 Some Special Equations with Variable Coefficients . . . . . . . . . . . . . . . . 1054
12 Systems of Linear Partial Differential Equations 1059
12.1 Preliminary Remarks. Some Notation and Helpful Relations . . . . . . . . . . . . . . . 1059
12.2 Systems of Two First-Order Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1059
12.3 Systems of Two Second-Order Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1063
12.3.1 Systems of Parabolic Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1063
12.3.2 Systems of Hyperbolic or Elliptic Equations . . . . . . . . . . . . . . . . . . . . . . 1064
12.4 Systems of Two Higher-Order Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1064
12.5 Simplest Systems Containing Vector Functions and Operators div and curl . . . 1066
12.5.1 Equation curl u = A(x) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1066
12.5.2 Simple Systems of Equations Containing Operators div and curl . . . . . 1067
12.5.3 Two Representations of Vector Functions . . . . . . . . . . . . . . . . . . . . . . . . . 1069
12.6 Elasticity Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1071
12.6.1 Elasticity Equations in Various Coordinate Systems . . . . . . . . . . . . . . . . 1071
12.6.2 Various Forms of Decompositions of Homogeneous Elasticity
Equations with f = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1073
12.6.3 Various Forms of Decompositions for Nonhomogeneous Elasticity
Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1075
12.6.4 Cauchy Problem and Its Solution. Fundamental Solution Matrix . . . . . 1076
C ONTENTS xiii
12.7 Stokes Equations for Viscous Incompressible Fluids . . . . . . . . . . . . . . . . . . . . . . 1077

12.7.1 Stokes Equations in Various Coordinate Systems . . . . . . . . . . . . . . . . . . 1077
12.7.2 Various Forms of Decompositions for the Stokes Equations with f = 0 1081
12.7.3 Various Forms of Decompositions for the Stokes Equations with f 6= 0 1082
12.7.4 General Solution of the Steady-State Homogeneous Stokes Equations 1084
12.7.5 Solution of the Steady-State Nonhomogeneous Stokes Equations . . . . 1085
12.7.6 Solution of the Cauchy Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1085
12.8 Oseen Equations for Viscous Incompressible Fluids . . . . . . . . . . . . . . . . . . . . . . . 1086
12.8.1 Vector Form of Oseen Equations. Some Remarks . . . . . . . . . . . . . . . . . . 1086
12.8.2 Various Forms of Decompositions for the Oseen Equations with f = 0 1087
12.8.3 Various Forms of Decompositions for the Oseen Equations with f 6= 0 1088
12.8.4 Oseen Equations with Variable Coefficients . . . . . . . . . . . . . . . . . . . . . . . 1088
12.9 Maxwell Equations for Viscoelastic Incompressible Fluids . . . . . . . . . . . . . . . . . 1089
12.9.1 Vector Form of the Maxwell Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 1089
12.9.2 Various Forms of Decompositions for the Maxwell Equations with
f=0 .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . .. .. .. .. .. .. .. .. 1089
12.9.3 Various Forms of Decompositions for the Maxwell Equations with
f 6= 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1090
12.10 Equations of Viscoelastic Incompressible Fluids (General Model) . . . . . . . . . . 1091
12.10.1 Linearized Equations of Viscoelastic Incompressible Fluids. Some
Models of Viscoelastic Fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1091
12.10.2 Various Forms of Decompositions for Equations of Viscoelastic
Incompressible Fluids with f = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1092
12.10.3 Various Forms of Decompositions for Equations of Viscoelastic
Incompressible Fluids with f 6= 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1093
12.11 Linearized Equations for Inviscid Compressible Barotropic Fluids . . . . . . . . . 1094
12.11.1 Vector Form of Equations without Mass Forces. Some Remarks . . . 1094
12.11.2 Decompositions of Equations for Inviscid Compressible Barotropic
Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1095
12.11.3 Vector Form of Equations with Mass Forces . . . . . . . . . . . . . . . . . . . . 1095
12.12 Stokes Equations for Viscous Compressible Barotropic Fluids . . . . . . . . . . . . . 1096
12.12.1 Linearized Equations of Viscous Compressible Barotropic Fluids . . 1096
12.12.2 Decompositions of Equations of Viscous Compressible Barotropic
Fluid with f = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1097
12.12.3 Decompositions of Equations of a Viscous Compressible Barotropic
Fluid with f 6= 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1098
12.12.4 Reduction to One Vector Equation and Its Decompositions . . . . . . . . 1098
12.12.5 Independent Equations for u and p . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1099
12.13 Oseen Equations for Viscous Compressible Barotropic Fluids . . . . . . . . . . . . . 1100
12.13.1 Vector Form of Equations. Some Remarks . . . . . . . . . . . . . . . . . . . . . . 1100
12.13.2 Decompositions of Equations with f = 0 . . . . . . . . . . . . . . . . . . . . . . . . 1100
12.13.3 Decomposition of Equations with f 6= 0 . . . . . . . . . . . . . . . . . . . . . . . . . 1101
12.14 Equations of Thermoelasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1101
12.14.1 Vector Form of Thermoelasticity Equations . . . . . . . . . . . . . . . . . . . . . 1101
12.14.2 Decompositions of Thermoelasticity Equations with f = 0 . . . . . . . . 1101
12.14.3 Decompositions of Thermoelasticity Equations with f 6= 0 . . . . . . . . 1102
xiv C ONTENTS
12.15 Nondissipative Thermoelasticity Equations (the Green–Naghdi Model) . . . . . 1103

12.15.1 Vector Form of the Nondissipative Thermoelasticity Equations . . . . 1103
12.15.2 Decompositions of the Nondissipative Thermoelasticity Equations
with f = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1104
12.15.3 Decompositions of Thermoelasticity Equations with f 6= 0 . . . . . . . . 1105
12.16 Viscoelasticity Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1106
12.16.1 Vector Form of Viscoelasticity Equations . . . . . . . . . . . . . . . . . . . . . . . 1106
12.16.2 Decompositions of Viscoelasticity Equations with f = 0 . . . . . . . . . . 1106
12.16.3 Various Forms of Decompositions for Viscoelasticity Equations with
f 6= 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1108
12.17 Maxwell Equations (Electromagnetic Field Equations) . . . . . . . . . . . . . . . . . . . 1108
12.17.1 Maxwell Equations in a Medium and Constitutive Relations . . . . . . 1108
12.17.2 Some Transformations and Solutions of the Maxwell Equations . . . 1109
12.18 Vector Equations of General Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1110
12.18.1 Vector Equations Containing Operators div and ∇ . . . . . . . . . . . . . . . 1110
12.18.2 Decompositions of the Homogeneous Vector Equation . . . . . . . . . . . 1111
12.18.3 Decompositions of the Nonhomogeneous Vector Equation . . . . . . . . 1112
12.18.4 Vector Equations Containing More General Operators . . . . . . . . . . . . 1113
12.19 General Systems Involving Vector and Scalar Equations: Part I . . . . . . . . . . . . 1114
12.19.1 Systems Containing Operators div and ∇ . . . . . . . . . . . . . . . . . . . . . . . 1114
12.19.2 Decompositions of Systems with Homogeneous Vector Equation . . 1115
12.19.3 Decompositions of Systems with Nonhomogeneous Vector
Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1115
12.19.4 Equations for u and p. Reduction to One Vector Equation . . . . . . . . . 1116
12.19.5 Systems Containing More General Operators . . . . . . . . . . . . . . . . . . . 1117
12.20 General Systems Involving Vector and Scalar Equations: Part II . . . . . . . . . . . 1118
12.20.1 Class of Systems Considered . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1118
12.20.2 Asymmetric Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1118
12.20.3 Symmetric Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1119
Part II Analytical Methods 1121

13 Methods for First-Order Linear PDEs 1123
13.1 Linear PDEs with Two Independent Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1123
13.1.1 Special First-Order Linear PDEs with Two Independent Variables . . . 1123
13.1.2 General First-Order Linear PDE with Two Independent Variables . . . . 1126
13.2 First-Order Linear PDEs with Three or More Independent Variables . . . . . . . . 1129
13.2.1 Characteristic System. General Solution . . . . . . . . . . . . . . . . . . . . . . . . . . 1129
13.2.2 Cauchy Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1131
14 Second-Order Linear PDEs: Classification, Problems, Particular Solutions 1133
14.1 Classification of Second-Order Linear Partial Differential Equations . . . . . . . . 1133
14.1.1 Equations with Two Independent Variables . . . . . . . . . . . . . . . . . . . . . . . 1133
14.1.2 Equations with Many Independent Variables . . . . . . . . . . . . . . . . . . . . . . 1138
14.2 Basic Problems of Mathematical Physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1140
14.2.1 Initial and Boundary Conditions. Cauchy Problem. Boundary Value
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1140
14.2.2 First, Second, Third, and Mixed Boundary Value Problems . . . . . . . . . 1142
C ONTENTS xv
14.3 Properties and Particular Solutions of Linear Equations . . . . . . . . . . . . . . . . . . . . 1144

14.3.1 Homogeneous Linear Equations. Basic Properties of Particular
Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1144
14.3.2 Separable Solutions. Solutions in the Form of Infinite Series . . . . . . . . 1147
14.3.3 Nonhomogeneous Linear Equations and Their Properties . . . . . . . . . . . 1150
14.3.4 General Solutions of Some Hyperbolic Equations . . . . . . . . . . . . . . . . . . 1150
15 Separation of Variables and Integral Transform Methods 1153
15.1 Separation of Variables (Fourier Method) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1153
15.1.1 Description of Separation of Variables. General Stage of Solution . . . 1153
15.1.2 Problems for Parabolic Equations: Final Stage of Solution . . . . . . . . . . 1157
15.1.3 Problems for Hyperbolic Equations: Final Stage of Solution . . . . . . . . 1159
15.1.4 Solution of Boundary Value Problems for Elliptic Equations . . . . . . . . 1160
15.1.5 Solution of Boundary Value Problems for Higher-Order Equations . . . 1163
15.2 Integral Transform Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1165
15.2.1 Laplace Transform and Its Application in Mathematical Physics . . . . . 1165
15.2.2 Fourier Transform and Its Application in Mathematical Physics . . . . . 1170
15.2.3 Fourier Sine and Cosine Transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1173
15.2.4 Mellin, Hankel, and Other Integral Transforms . . . . . . . . . . . . . . . . . . . . 1177
16 Cauchy Problem. Fundamental Solutions 1181
16.1 Dirac Delta Function. Fundamental Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1181
16.1.1 Dirac Delta Function and Its Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . 1181
16.1.2 Fundamental Solutions. Constructing Particular Solutions . . . . . . . . . . 1182
16.2 Representation of the Solution of the Cauchy Problem via the Fundamental
Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1185
16.2.1 Cauchy Problem for Ordinary Differential Equations . . . . . . . . . . . . . . . 1185
16.2.2 Cauchy Problem for Parabolic Equations . . . . . . . . . . . . . . . . . . . . . . . . . 1187
16.2.3 Cauchy Problem for Hyperbolic Equations . . . . . . . . . . . . . . . . . . . . . . . 1190
16.2.4 Higher-Order Linear PDEs. Generalized Cauchy Problem . . . . . . . . . . 1193
17 Boundary Value Problems. Green’s Function 1199
17.1 Boundary Value Problems for Parabolic Equations with One Space Variable.
Green’s Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1199
17.1.1 Representation of Solutions via the Green’s Function
.............. 1199
∂w ∂ ∂w
17.1.2 Problems for Equation s(x) ∂t = ∂x p(x) ∂x − q(x)w + Φ(x, t) . . . 1202
17.2 Boundary Value Problems for Hyperbolic Equations with One Space Variable.
Green’s Function. Goursat Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1205
17.2.1 Representation of Solutions via the Green’s Function . . . . . . . . . . . . . . 1205
2
17.2.2 Problems for Equation s(x) ∂∂tw ∂ ∂w
2 = ∂x p(x) ∂x − q(x)w + Φ(x, t) . . . 1207
2
17.2.3 Problems for Equation ∂∂tw 2 + a(t) ∂t =
∂w
∂
b(t) ∂x p(x) ∂w ∂x − q(x)w + Φ(x, t) . . . . . . . . . . . . . . . . . . . . . . . . . . . 1208
17.2.4 Generalized Cauchy Problem with Initial Conditions Set along a
Curve. Riemann Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1210
17.2.5 Goursat Problem (a Problem with Initial Data on Characteristics) . . . . 1212
17.3 Boundary Value Problems for Elliptic Equations with Two Space Variables . . 1214
17.3.1 Problems and the Green’s Functions for Equation
2 2
a(x) ∂∂xw2 + ∂∂yw2 + b(x) ∂w ∂x + c(x)w = −Φ(x, y) . . . . . . . . . . . . . . . . . . . . 1214
17.3.2 Representation of Solutions of Boundary Value Problems via Green’s
Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1216
xvi C ONTENTS
17.4 Boundary Value Problems with Many Space Variables. Green’s Function . . . . 1218
17.4.1 Problems for Parabolic Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1218
17.4.2 Problems for Hyperbolic Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1220
17.4.3 Problems for Elliptic Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1221
17.4.4 Comparison of the Solution Structures for Boundary Value Problems
for Equations of Various Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1222
17.5 Construction of the Green’s Functions. General Formulas and Relations . . . . . 1223
17.5.1 Green’s Functions of Boundary Value Problems for Equations of
Various Types in Bounded Domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1223
17.5.2 Green’s Functions Admitting Incomplete Separation of Variables . . . . 1224
17.5.3 Construction of Green’s Functions via Fundamental Solutions . . . . . . . 1227
18 Duhamel’s Principles. Some Transformations 1233
18.1 Duhamel’s Principles in Nonstationary Problems . . . . . . . . . . . . . . . . . . . . . . . . . 1233
18.1.1 Problems for Homogeneous Linear Equations . . . . . . . . . . . . . . . . . . . . . 1233
18.1.2 Problems for Nonhomogeneous Linear Equations . . . . . . . . . . . . . . . . . . 1235
18.2 Transformations Simplifying Initial and Boundary Conditions . . . . . . . . . . . . . . 1237
18.2.1 Transformations That Lead to Homogeneous Boundary Conditions . . 1237
18.2.2 Transformations That Lead to Homogeneous Initial and Boundary
Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1238
19 Systems of Linear Coupled PDEs. Decomposition Methods 1239
19.1 Asymmetric and Symmetric Decompositions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1239
19.1.1 Asymmetric Decomposition. Order of Decomposition . . . . . . . . . . . . . . 1239
19.1.2 Symmetric Decomposition. Invariant Transformations . . . . . . . . . . . . . 1242
19.2 First-Order Decompositions. Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1244
19.2.1 Systems of Linear PDEs without Mass Forces (f = 0) . . . . . . . . . . . . . . 1244
19.2.2 Systems of Linear PDEs with Mass Forces . . . . . . . . . . . . . . . . . . . . . . . . 1249
19.3 Higher-Order Decompositions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1251
19.3.1 Decomposition of Systems Consisting of One Vector Equation . . . . . . 1251
19.3.2 Decomposition of Systems Consisting of a Vector Equation and a
Scalar Equation (the First Approach) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1252
19.3.3 Decomposition of Systems Consisting of a Vector Equation and a
Scalar Equation (the Second Approach) . . . . . . . . . . . . . . . . . . . . . . . . . . 1253
20 Some Asymptotic Results and Formulas 1255
20.1 Regular Perturbation Theory Formulas for the Eigenvalues . . . . . . . . . . . . . . . . . 1256
20.1.1 Statement of the Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1256
20.1.2 Formulas for the Coefficients of the Expansion . . . . . . . . . . . . . . . . . . . . 1256
20.2 Singular Perturbation Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1257
20.2.1 Cauchy Problem for the Schrödinger Equation . . . . . . . . . . . . . . . . . . . . 1257
20.2.2 Stationary Phase Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1261
20.2.3 Fourier Transform with a Parameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1264
21 Elements of Theory of Generalized Functions 1265
21.1 Generalized Functions of One Variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1265
21.1.1 Important Terminological Remark . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1265
21.1.2 Test Function Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1265
21.1.3 Distribution Space. Dirac Delta Function . . . . . . . . . . . . . . . . . . . . . . . . . 1266
21.1.4 Derivatives of Distributions. Some Formulas . . . . . . . . . . . . . . . . . . . . . . 1267
21.1.5 Operations on Distributions and Some Transformations . . . . . . . . . . . . 1269
C ONTENTS xvii
21.1.6 Tempered Distributions and Fourier Transform . . . . . . . . . . . . . . . . . . . . 1270

21.1.7 Generalized Solutions of Linear Ordinary Differential Equations . . . . 1271
21.2 Generalized Functions of Several Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1271
21.2.1 Some Definitions. Partial Derivatives. Direct Product. Linear
Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1271
21.2.2 Dirac Delta Function. Generalized Solutions of Linear PDEs . . . . . . . . 1273
Part III Symbolic and Numerical Solutions with Maple, Mathematica,

and MATLAB R
1275
22 Linear Partial Differential Equations with Maple 1277
22.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1277
22.1.1 Preliminary Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1277
22.1.2 Brief Introduction to Maple . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1279
22.1.3 Maple Language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1280
22.2 Analytical Solutions and Their Visualizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1282
22.2.1 Constructing Analytical Solutions in Terms of Predefined Functions . 1282
22.2.2 Constructing General Solutions via the Method of Characteristics . . . . 1289
22.2.3 Constructing General Solutions via Transformations to Canonical
Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1291
22.2.4 Constructing Analytical Solutions of Cauchy Problems . . . . . . . . . . . . . 1293
22.2.5 Constructing Analytical Solutions of Boundary Value Problems . . . . . 1297
22.2.6 Constructing Analytical Solutions of Initial-Boundary Value Problems 1298
22.2.7 Constructing Analytical Solutions of Systems of Linear PDEs . . . . . . . 1299
22.3 Analytical Solutions of Mathematical Problems . . . . . . . . . . . . . . . . . . . . . . . . . . 1301
22.3.1 Constructing Separable Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1301
22.3.2 Constructing Analytical Solutions via Integral Transform Methods . . . 1305
22.3.3 Constructing Analytical Solutions in Terms of Green’s Functions . . . . 1306
22.4 Numerical Solutions and Their Visualizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1309
22.4.1 Constructing Numerical Solutions in Terms of Predefined Functions . 1310
22.4.2 Numerical Methods Embedded in Maple . . . . . . . . . . . . . . . . . . . . . . . . . 1312
22.4.3 Numerical Solutions of Initial-Boundary Value Problems . . . . . . . . . . . 1318
22.4.4 Numerical Solutions of Boundary Value Problems . . . . . . . . . . . . . . . . . 1322
22.4.5 Numerical Solutions of Cauchy Problems . . . . . . . . . . . . . . . . . . . . . . . . 1323
22.4.6 Numerical Solutions of Systems of Linear PDEs . . . . . . . . . . . . . . . . . . 1325
23 Linear Partial Differential Equations with Mathematica 1327
23.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1327
23.1.1 Some Notational Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1327
23.1.2 Brief Introduction to Mathematica . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1327
23.1.3 Mathematica Language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1329
23.1.4 Dynamic Computation and Visualization in Mathematica Notebook . . 1332
23.2 Analytical Solutions and Their Visualizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1333
23.2.1 Constructing Analytical Solutions in Terms of Predefined Functions . 1333
23.2.2 Constructing General Solutions via the Method of Characteristics . . . . 1335
23.2.3 Constructing General Solutions via Conversion to Canonical Forms . . 1338
23.2.4 Constructing Analytical Solutions of Cauchy Problems . . . . . . . . . . . . . 1340
23.2.5 Constructing Analytical Solutions of Boundary Value Problems . . . . . 1342
23.2.6 Constructing Analytical Solutions of Initial-Boundary Value Problems 1344
23.2.7 Constructing Analytical Solutions of Systems of Linear PDEs . . . . . . . 1345
xviii C ONTENTS
23.3 Analytical Solutions of Mathematical Problems . . . . . . . . . . . . . . . . . . . . . . . . . . 1347

23.3.1 Constructing Separable Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1347
23.3.2 Constructing Analytical Solutions via Integral Transform Methods . . . 1350
23.3.3 Constructing Analytical Solutions in Terms of Green’s Functions . . . . 1352
23.4 Numerical Solutions and Their Visualizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1356
23.4.1 Constructing Numerical Solutions in Terms of Predefined Functions . 1356
23.4.2 Numerical Methods Embedded in Mathematica . . . . . . . . . . . . . . . . . . . 1358
23.4.6 Numerical Solutions of Systems of Linear PDEs . . . . . . . . . . . . . . . . . . 1365
24 Linear Partial Differential Equations with MATLAB R
1367
24.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1367
24.1.1 Preliminary Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1367
24.1.2 Brief Introduction to MATLAB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1368
24.1.3 MATLAB Language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1371
24.2 Numerical Solutions of Linear PDEs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1374
24.2.1 Constructing Numerical Solutions via Predefined Functions . . . . . . . . . 1375
24.2.2 Numerical Methods Embedded in MATLAB . . . . . . . . . . . . . . . . . . . . . . 1383
24.3 Constructing Finite Difference Approximations . . . . . . . . . . . . . . . . . . . . . . . . . . 1392
24.3.1 Explicit Finite Difference Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1392
24.3.2 Implicit Finite Difference Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1395
24.4 Numerical Solutions of Systems of Linear PDEs . . . . . . . . . . . . . . . . . . . . . . . . . 1396
24.4.1 Linear Systems of 1D PDEs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1396
24.4.2 Linear Systems of 2D PDEs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1399
Part IV Tables and Supplements 1403

25 Elementary Functions and Their Properties 1405
25.1 Power, Exponential, and Logarithmic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 1405
25.1.1 Properties of the Power Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1405
25.1.2 Properties of the Exponential Function . . . . . . . . . . . . . . . . . . . . . . . . . . . 1405
25.1.3 Properties of the Logarithmic Function . . . . . . . . . . . . . . . . . . . . . . . . . . . 1406
25.2 Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1407
25.2.1 Simplest Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1407
25.2.2 Reduction Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1408
25.2.3 Relations between Trigonometric Functions of Single Argument . . . . . 1408
25.2.4 Addition and Subtraction of Trigonometric Functions . . . . . . . . . . . . . . 1408
25.2.5 Products of Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1409
25.2.6 Powers of Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1409
25.2.7 Addition Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1409
25.2.8 Trigonometric Functions of Multiple Arguments . . . . . . . . . . . . . . . . . . 1410
25.2.9 Trigonometric Functions of Half Argument . . . . . . . . . . . . . . . . . . . . . . . 1410
25.2.10 Differentiation Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1410
25.2.11 Integration Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1410
25.2.12 Expansion in Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1411
C ONTENTS xix
25.2.13 Representation in the Form of Infinite Products . . . . . . . . . . . . . . . . . . . 1411

25.2.14 Euler and de Moivre Formulas. Relationship with Hyperbolic
Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1411
25.3 Inverse Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1411
25.3.1 Definitions of Inverse Trigonometric Functions . . . . . . . . . . . . . . . . . . . . 1411
25.3.2 Simplest Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1412
25.3.3 Some Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1412
25.3.4 Relations between Inverse Trigonometric Functions . . . . . . . . . . . . . . . . 1413
25.3.5 Addition and Subtraction of Inverse Trigonometric Functions . . . . . . . 1413
25.3.6 Differentiation Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1414
25.3.7 Integration Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1414
25.3.8 Expansion in Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1414
25.4 Hyperbolic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1414
25.4.1 Definitions of Hyperbolic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1414
25.4.3 Relations between Hyperbolic Functions of Single Argument (x ≥ 0) . 1415
25.4.4 Addition and Subtraction of Hyperbolic Functions . . . . . . . . . . . . . . . . . 1415
25.4.5 Products of Hyperbolic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1415
25.4.6 Powers of Hyperbolic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1416
25.4.7 Addition Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1416
25.4.8 Hyperbolic Functions of Multiple Argument . . . . . . . . . . . . . . . . . . . . . . 1416
25.4.9 Hyperbolic Functions of Half Argument . . . . . . . . . . . . . . . . . . . . . . . . . . 1417
25.4.10 Differentiation Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1417
25.4.11 Integration Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1417
25.4.12 Expansion in Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1417
25.4.13 Representation in the Form of Infinite Products . . . . . . . . . . . . . . . . . . . 1417
25.4.14 Relationship with Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . 1418
25.5 Inverse Hyperbolic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1418
25.5.1 Definitions of Inverse Hyperbolic Functions . . . . . . . . . . . . . . . . . . . . . . 1418
25.5.3 Relations between Inverse Hyperbolic Functions . . . . . . . . . . . . . . . . . . 1418
25.5.4 Addition and Subtraction of Inverse Hyperbolic Functions . . . . . . . . . . 1418
25.5.5 Differentiation Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1419
25.5.6 Integration Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1419
25.5.7 Expansion in Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1419
26 Finite Sums and Infinite Series 1421
26.1 Finite Numerical Sums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1421
26.1.1 Progressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . P
.............. 1421
26.1.2 Sums of Powers of Natural Numbers Having the Form P km . . . . . . . 1421
26.1.3 Alternating Sums of Powers of Natural Numbers, (−1) km . . . . . . k 1422
26.1.4 Other Sums Containing Integers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1422
26.1.5 Sums Containing Binomial Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . 1422
26.1.6 Other Numerical Sums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1423
26.2 Finite Functional Sums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1424
26.2.1 Sums Involving Hyperbolic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1424
26.2.2 Sums Involving Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . . . . 1425
26.3 Infinite Numerical Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1426
26.3.1 Progressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1426
26.3.2 Other Numerical Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1426
xx C ONTENTS
26.4 Infinite Functional Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1428

26.4.1 Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1428
26.4.2 Trigonometric Series in One Variable Involving Sine . . . . . . . . . . . . . . . 1429
26.4.3 Trigonometric Series in One Variable Involving Cosine . . . . . . . . . . . . . 1431
26.4.4 Trigonometric Series in Two Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . 1433
27 Indefinite and Definite Integrals 1435
27.1 Indefinite Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1435
27.1.1 Integrals Involving Rational Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 1435
27.1.2 Integrals Involving Irrational Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 1439
27.1.3 Integrals Involving Exponential Functions . . . . . . . . . . . . . . . . . . . . . . . . 1442
27.1.4 Integrals Involving Hyperbolic Functions . . . . . . . . . . . . . . . . . . . . . . . . . 1443
27.1.5 Integrals Involving Logarithmic Functions . . . . . . . . . . . . . . . . . . . . . . . . 1446
27.1.6 Integrals Involving Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . 1447
27.1.7 Integrals Involving Inverse Trigonometric Functions . . . . . . . . . . . . . . . 1451
27.2 Definite Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1452
27.2.1 Integrals Involving Power-Law Functions . . . . . . . . . . . . . . . . . . . . . . . . . 1452
27.2.2 Integrals Involving Exponential Functions . . . . . . . . . . . . . . . . . . . . . . . . 1455
27.2.3 Integrals Involving Hyperbolic Functions . . . . . . . . . . . . . . . . . . . . . . . . . 1456
27.2.4 Integrals Involving Logarithmic Functions . . . . . . . . . . . . . . . . . . . . . . . . 1457
27.2.5 Integrals Involving Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . 1457
27.2.6 Integrals Involving Bessel Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1460
28 Integral Transforms 1463
28.1 Tables of Laplace Transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1463
28.1.1 General Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1463
28.1.2 Expressions with Power-Law Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 1465
28.1.3 Expressions with Exponential Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 1465
28.1.4 Expressions with Hyperbolic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 1466
28.1.5 Expressions with Logarithmic Functions . . . . . . . . . . . . . . . . . . . . . . . . . 1467
28.1.6 Expressions with Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . . . 1467
28.1.7 Expressions with Special Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1469
28.2 Tables of Inverse Laplace Transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1470
28.2.2 Expressions with Rational Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1472
28.2.3 Expressions with Square Roots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1476
28.2.4 Expressions with Arbitrary Powers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1477
28.3 Tables of Fourier Cosine Transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1482
C ONTENTS xxi
28.4 Tables of Fourier Sine Transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1488

29 Curvilinear Coordinates, Vectors, Operators, and Differential Relations 1495
29.1 Arbitrary Curvilinear Coordinate Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1495
29.1.1 General Nonorthogonal Curvilinear Coordinates . . . . . . . . . . . . . . . . . . . 1495
29.1.2 General Orthogonal Curvilinear Coordinates . . . . . . . . . . . . . . . . . . . . . . 1497
29.2 Cartesian, Cylindrical, and Spherical Coordinate Systems . . . . . . . . . . . . . . . . . . 1498
29.2.1 Cartesian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1498
29.2.2 Cylindrical Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1499
29.2.3 Spherical Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1500
29.3 Other Special Orthogonal Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1502
29.3.1 Coordinates of a Prolate Ellipsoid of Revolution . . . . . . . . . . . . . . . . . . . 1502
29.3.2 Coordinates of an Oblate Ellipsoid of Revolution . . . . . . . . . . . . . . . . . . 1503
29.3.3 Coordinates of an Elliptic Cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1504
29.3.4 Conical Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1505
29.3.5 Parabolic Cylinder Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1506
29.3.6 Parabolic Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1506
29.3.7 Bicylindrical Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1507
29.3.8 Bipolar Coordinates (in Space) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1507
29.3.9 Toroidal Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1508
30 Special Functions and Their Properties 1509
30.1 Some Coefficients, Symbols, and Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1509
30.1.1 Binomial Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1509
30.1.2 Pochhammer Symbol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1510
30.1.3 Bernoulli Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1510
30.1.4 Euler Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1511
30.2 Error Functions. Exponential and Logarithmic Integrals . . . . . . . . . . . . . . . . . . . 1512
30.2.1 Error Function and Complementary Error Function . . . . . . . . . . . . . . . . 1512
30.2.2 Exponential Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1512
30.2.3 Logarithmic Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1513
30.3 Sine Integral and Cosine Integral. Fresnel Integrals . . . . . . . . . . . . . . . . . . . . . . . 1514
30.3.1 Sine Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1514
30.3.2 Cosine Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1515
30.3.3 Fresnel Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1515
30.4 Gamma Function, Psi Function, and Beta Function . . . . . . . . . . . . . . . . . . . . . . . 1516
30.4.1 Gamma Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1516
30.4.2 Psi Function (Digamma Function) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1517
30.4.3 Beta Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1518
30.5 Incomplete Gamma and Beta Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1519
30.5.1 Incomplete Gamma Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1519
30.5.2 Incomplete Beta Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1520
xxii C ONTENTS
30.6 Bessel Functions (Cylindrical Functions) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1520

30.6.1 Definitions and Basic Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1520
30.6.2 Integral Representations and Asymptotic Expansions . . . . . . . . . . . . . . . 1522
30.6.3 Zeros and Orthogonality Properties of Bessel Functions . . . . . . . . . . . . 1524
30.6.4 Hankel Functions (Bessel Functions of the Third Kind) . . . . . . . . . . . . . 1525
30.7 Modified Bessel Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1526
30.7.1 Definitions. Basic Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1526
30.8 Airy Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1529
30.8.1 Definition and Basic Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1529
30.8.2 Power Series and Asymptotic Expansions . . . . . . . . . . . . . . . . . . . . . . . . 1529
30.9 Degenerate Hypergeometric Functions (Kummer Functions) . . . . . . . . . . . . . . . 1530
30.9.1 Definitions and Basic Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1530
30.9.3 Whittaker Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1534
30.10 Hypergeometric Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1534
30.10.1 Various Representations of the Hypergeometric Function . . . . . . . . . 1534
30.10.2 Basic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1536
30.11 Legendre Polynomials, Legendre Functions, and Associated Legendre
Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1536
30.11.1 Legendre Polynomials and Legendre Functions . . . . . . . . . . . . . . . . . . 1536
30.11.2 Associated Legendre Functions with Integer Indices and Real
Argument . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1539
30.11.3 Associated Legendre Functions. General Case . . . . . . . . . . . . . . . . . . 1539
30.12 Parabolic Cylinder Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1542
30.12.1 Definitions. Basic Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1542
30.12.2 Integral Representations, Asymptotic Expansions, and Linear
Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1543
30.13 Elliptic Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1544
30.13.1 Complete Elliptic Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1544
30.13.2 Incomplete Elliptic Integrals (Elliptic Integrals) . . . . . . . . . . . . . . . . . 1545
30.14 Elliptic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1547
30.14.1 Jacobi Elliptic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1547
30.14.2 Weierstrass Elliptic Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1551
30.15 Jacobi Theta Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1553
30.15.1 Series Representation of the Jacobi Theta Functions. Simplest
Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1553
30.15.2 Various Relations and Formulas. Connection with Jacobi Elliptic
Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1554
30.16 Mathieu Functions and Modified Mathieu Functions . . . . . . . . . . . . . . . . . . . . . 1555
30.16.1 Mathieu Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1555
30.16.2 Modified Mathieu Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1557
30.17 Orthogonal Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1557
30.17.1 Laguerre Polynomials and Generalized Laguerre Polynomials . . . . . 1558
30.17.2 Chebyshev Polynomials and Functions . . . . . . . . . . . . . . . . . . . . . . . . . 1559
30.17.3 Hermite Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1561
30.17.4 Jacobi Polynomials and Gegenbauer Polynomials . . . . . . . . . . . . . . . . 1563
C ONTENTS xxiii
30.18 Nonorthogonal Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1564

30.18.1 Bernoulli Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1564
30.18.2 Euler Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1565
References 1569
Index 1587
PREFACE TO THE SECOND EDITION
The Handbook of Linear Partial Differential Equations for Engineers and Scien-
tists, a unique reference for scientists and engineers, contains nearly 4,000 linear partial
differential equations with solutions as well as analytical, symbolic, and numerical methods
for solving linear equations. First-, second-, third-, fourth-, and higher-order linear equa-
tions and systems of coupled equations are considered. Equations of parabolic, hyperbolic,
elliptic, mixed, and other types are discussed. A number of new linear equations, exact
solutions, transformations, and methods are described. Formulas for effective construction
of solutions are given. A number of specific examples where the methods described in
the book are used are considered. Boundary value problems and eigenvalue problems are
described. Symbolic and numerical methods for solving PDEs with Maple, Mathematica,
and MATLAB R
are considered. All in all, the handbook contains many more linear partial
differential equations than any other book currently available.
In selecting the material, the authors have given highest priority to the following major
topics:
• Equations and problems that arise in various applications (heat and mass transfer theory,
wave theory, elasticity, hydrodynamics, aerodynamics, continuum mechanics, acous-
tics, electrostatics, electrodynamics, electrical engineering, diffraction theory, quantum
mechanics, chemical engineering sciences, control theory, etc.).
• Systems of coupled equations that arise in various fields of continuum mechanics and
physics.
• Analytical and symbolic methods for solving linear equations of mathematical
physics.
• Equations of general form that depend on arbitrary functions and equations that involve
many free parameters; exact solutions of such equations are of major importance for
testing numerical and approximate analytical methods.
The second edition has been substantially updated, revised, and expanded. More than
1,500 linear equations and systems with solutions, as well some methods and many exam-
ples, have been added, which amounts to over 700 pages of new material (including 250
new pages dealing with methods).
New to the second edition:
• Some second-, third-, fourth-, and higher-order linear PDEs with solutions.
• Systems of coupled partial differential equations with solutions.
• First-order linear PDEs with solutions.
• Some analytical methods including decomposition methods and their applications.
• Symbolic and numerical methods with Maple, Mathematica, and MATLAB.
• Some transformations, asymptotic formulas and solutions.
• Many new examples and figures included for illustrative purposes.
• Some long tables, including tables of various integral transforms.
• Extensive table of contents and detailed index.
xxv
xxvi P REFACE
Note that Chapters 1–12 of the book can be used as a database of test problems for
numerical, approximate analytical, and symbolic methods for solving linear partial differ-
ential equations and systems of coupled equations. To satisfy the needs of a broad au-
dience with diverse mathematical backgrounds, the authors have done their best to avoid
special terminology whenever possible. Therefore, some of the methods are outlined in
a schematic and somewhat simplified manner with necessary references made to books
where these methods are considered in more detail. Many sections are written so that they
can be read independently from each other. This allows the reader to get to the heart of the
matter quickly.
Separate sections of the book can serve as a basis for practical courses and lectures on
equations of mathematical physics and linear PDEs.
We would like to express our keen gratitude to Alexei Zhurov for fruitful discussions
and valuable remarks. We are very thankful to Inna Shingareva and Carlos Lizárraga-
Celaya, who wrote three chapters (22–24) of the book at our request.
The authors hope that the handbook will prove helpful for a wide audience of re-
searchers, university and college teachers, engineers, and students in various fields of ap-
plied mathematics, mechanics, physics, chemistry, economics, and engineering sciences.
Andrei D. Polyanin
Vladimir E. Nazaikinskii
PREFACE TO THE FIRST EDITION
Linear partial differential equations arise in various fields of science and numerous ap-
plications, e.g., heat and mass transfer theory, wave theory, hydrodynamics, aerodynamics,
elasticity, acoustics, electrostatics, electrodynamics, electrical engineering, diffraction the-
ory, quantum mechanics, control theory, chemical engineering sciences, and biomechanics.
This book presents brief statements and exact solutions of more than 2000 linear equa-
tions and problems of mathematical physics. Nonstationary and stationary equations with
constant and variable coefficients of parabolic, hyperbolic, and elliptic types are consid-
ered. A number of new solutions to linear equations and boundary value problems are
described. Special attention is paid to equations and problems of general form that depend
on arbitrary functions. Formulas for the effective construction of solutions to nonhomo-
geneous boundary value problems of various types are given. We consider second-order
and higher-order equations as well as the corresponding boundary value problems. All in
all, the handbook presents more equations and problems of mathematical physics than any
other book currently available.
For the reader’s convenience, the introduction outlines some definitions and basic equa-
tions, problems, and methods of mathematical physics. It also gives useful formulas that
enable one to express solutions to stationary and nonstationary boundary value problems
of general form in terms of the Green’s function.
Two supplements are given at the end of the book. Supplement A lists properties of
the most common special functions (the gamma function, Bessel functions, degenerate hy-
pergeometric functions, Mathieu functions, etc.). Supplement B describes the methods of
P REFACE xxvii
generalized and functional separation of variables for nonlinear partial differential equa-
tions. We give specific examples and an overview application of these methods to construct
exact solutions for various classes of second-, third-, fourth-, and higher-order equations
(in total, about 150 nonlinear equations with solutions are described). Special attention is
paid to equations of heat and mass transfer theory, wave theory, and hydrodynamics as well
as to mathematical physics equations of general form that involve arbitrary functions.
The equations in all chapters are in ascending order of complexity. Many sections
can be read independently, which facilitates working with the material. An extended table
of contents will help the reader find the desired equations and boundary value problems.
We refer to specific equations using notation like “1.8.5.2,” which means “Equation 2 in
Subsection 1.8.5.”
To extend the range of potential readers with diverse mathematical backgrounds, the
author strove to avoid the use of special terminology wherever possible. For this reason,
some results are presented schematically, in a simplified manner (without details), which
is, however, quite sufficient in most applications.
Separate sections of the book can serve as a basis for practical courses and lectures on
equations of mathematical physics.
The author thanks Alexei Zhurov for useful remarks on the manuscript.
The author hopes that the handbook will be useful for a wide range of scientists, univer-
sity teachers, engineers, and students in various areas of mathematics, physics, mechanics,
control, and engineering sciences.
Andrei D. Polyanin
MATLAB R
is a registered trademark of The MathWorks, Inc. For product information,
please contact:
The MathWorks, Inc. 3
Apple Hill Drive
Natick, MA 01760-2098 USA
Tel: 508 647 7000
Fax: 508-647-7001
E-mail: info@mathworks.com
Web: www.mathworks.com
AUTHORS
Andrei D. Polyanin, D.Sc., is an internationally renowned

scientist of broad interests and is active in various areas
of mathematics, mechanics, and chemical engineering sci-
ences. He is one of the most prominent authors in the field
of reference literature on mathematics.
Professor Polyanin graduated with honors from the Fac-
ulty of Mechanics and Mathematics at Lomonosov Moscow
State University in 1974. He received his Ph.D. in 1981 and
D.Sc. in 1986 at the Institute for Problems in Mechanics of
the Russian (former USSR) Academy of Sciences. Since
1975, Professor Polyanin has been working at the Institute
for Problems in Mechanics of the Russian Academy of Sci-
ences; he is also Professor of applied mathematics at Bau-
man Moscow State Technical University and at National Research Nuclear University
MEPhI. He is a member of the Russian National Committee on Theoretical and Applied
Mechanics and the Mathematics and Mechanics Expert Council of the Higher Certification
Committee of the Russian Federation.
Professor Polyanin has made important contributions to the theory of differential and in-
tegral equations, mathematical physics, engineering mathematics, theory of heat and mass
transfer, and chemical hydrodynamics. He has obtained exact solutions for several thousand
ordinary differential, partial differential, delay partial differential, and integral equations.
Professor Polyanin has authored more than 30 books in English, Russian, German,
and Bulgarian as well as over 170 research papers and three patents. He has written a
number of fundamental handbooks, including A. D. Polyanin and V. F. Zaitsev’s Hand-
book of Exact Solutions for Ordinary Differential Equations (CRC Press, 1995 and 2003);
A. D. Polyanin and A. V. Manzhirov’s Handbook of Integral Equations (CRC Press, 1998
and 2008); A. D. Polyanin’s Handbook of Linear Partial Differential Equations for Engi-
neers and Scientists (Chapman & Hall/CRC Press, 2002); A. D. Polyanin, V. F. Zaitsev, and
A. Moussiaux’s Handbook of First Order Partial Differential Equations (Taylor & Francis,
2002); A. D. Polyanin and V. F. Zaitsev’s Handbook of Nonlinear Partial Differential Equa-
tions (Chapman & Hall/CRC Press, 2004 and 2012); A. D. Polyanin and A. V. Manzhi-
rov’s Handbook of Mathematics for Engineers and Scientists (Chapman & Hall/CRC Press,
2007), and A. D. Polyanin and A. I. Chernoutsan’s (Eds.) A Concise Handbook of Mathe-
matics, Physics, and Engineering Sciences (Chapman & Hall/CRC Press, 2010).
Professor Polyanin is editor-in-chief of the international scientific educational website
EqWorld—The World of Mathematical Equations and he is editor of the book series Differ-
ential and Integral Equations and Their Applications (Chapman & Hall/CRC Press, Lon-
don/Boca Raton). Professor Polyanin is a member of the editorial board of the journals
Theoretical Foundations of Chemical Engineering, Mathematical Modeling and Compu-
tational Methods (in Russian), and Bulletin of the National Research Nuclear University
MEPhI (in Russian).
xxix
xxx AUTHORS
In 1991, Professor Polyanin was awarded the Chaplygin Prize of the Russian Academy
of Sciences for his research in mechanics. In 2001, he received an award from the Ministry
of Education of the Russian Federation.
Address: Institute for Problems in Mechanics, Russian Academy of Sciences, 101 Vernadsky Ave.,
Bldg 1, 119526 Moscow, Russia
Home page: http://eqworld.ipmnet.ru/polyanin-ew.htm
Vladimir E. Nazaikinskii, D.Sc., is an actively working

mathematician specializing in partial differential equations,
mathematical physics, and noncommutative analysis. He
was born in 1955 in Moscow, graduated from the Moscow
Institute of Electronic Engineering in 1977, defended his
Ph.D. in 1980 and D.Sc. in 2014, and worked at the Institute
for Automated Control Systems, Moscow Institute of Elec-
tronic Engineering, Potsdam University, and Moscow State
University. Currently he is a senior researcher at the Institute
for Problems in Mechanics, Russian Academy of Sciences.
He is the author of seven monographs (V. P. Maslov and V. E.
Nazaikinskii, Asymptotics of Operator and Pseudo-Differential Equations, Consultants Bu-
reau, New York, 1988; V. Nazaikinskii, B. Sternin, and V. Shatalov, Contact Geometry and
Linear Differential Equations, Walter de Gruyter, Berlin-New York, 1992; V. Nazaikinskii,
B. Sternin, and V. Shatalov, Methods of Noncommutative Analysis. Theory and Applica-
tions, Walter de Gruyter, Berlin-New York, 1996; V. Nazaikinskii, B.-W. Schulze, and B.
Sternin, Quantization Methods in Differential Equations, Taylor and Francis, London-New
York, 2002; V. Nazaikinskii, A. Savin, B.-W. Schulze, and B. Sternin, Elliptic Differen-
tial Equations on Singular Manifolds, CRC Press, Boca Raton, 2005; V. Nazaikinskii, A.
Savin, and B. Sternin, Elliptic Theory and Noncommutative Geometry, Birkhäuser, Basel,
2008; V. Nazaikinskii, B.-W. Schulze, and B. Sternin, The Localization Problem in Index
Theory of Elliptic Operators, Springer, Basel, 2014) and more than 90 papers on various
aspects of noncommutative analysis, asymptotic problems, and elliptic theory.
Address: Institute for Problems in Mechanics, Russian Academy of Sciences, 101 Vernadsky Ave.,
Bldg 1, 119526 Moscow, Russia
BASIC NOTATION AND REMARKS
Latin Characters
curl u curl of a vector u, sometimes also denoted by rot u

div u divergence of a vector u; div u = ∂u ∂u2
∂x + ∂y in the two-dimensional case
1
u = (u1 , u2 )
E fundamental solution of the Cauchy problem
Ee fundamental solution corresponding to an operator (or fundamental so-
lution of an equation)
grad a gradient of a scalar a, also denoted by ∇a, where ∇ is the nabla vector
differential operator
Im[A] imaginary part of a complex number A
G Green function
Rn n-dimensional Euclidean space, Rn = {−∞ < xk < ∞; k = 1, . . . , n}
Re[A] real part of a complex number pA
r, ϕ, z cylindrical coordinates, r = p x2 + y 2 with x = r cos ϕ and y = r sin ϕ
r, θ, ϕ spherical coordinates, r = x2 + y 2 + z 2 with x = r sin θ cos ϕ,
y = sin θ sin ϕ, and z = r cos θ
t time (t ≥ 0)
w unknown function (dependent variable)
x, y, z space (Cartesian) coordinates
x1 , . . . , xn Cartesian coordinates in n-dimensional space
x n-dimensional vector, x = {x1 , . . . , xn } p
|x| magnitude (length) of n-dimensional vector, |x| = x21 + x22 + · · · + x2n
y n-dimensional vector, y = {y1 , . . . , yn }
Greek Characters
∆ Laplace operator
∂2 ∂2
∆2 two-dimensional Laplace operator, ∆2 = ∂x2
+ ∂y 2
∂2 ∂2 ∂2
∆3 three-dimensional Laplace operator, ∆3 = ∂x 2 + ∂y 2
+ ∂z 2
Pn
∂2
∆n n-dimensional Laplace operator, ∆n = ∂x2k
k=1
∂ 4 ∂ 4 ∂4
∆∆ biharmonic operator; ∆∆ = ∂x 4 + 2 ∂x2 ∂y 2 + ∂y 4 in the two-dimensional
case Z a
δ(x) Dirac delta function; f (y)δ(x − y) dy = f (x), where f (x) is any con-
−a
tinuous function, a > 0, n
and −a < x < a
1 if n=m,
δnm Kronecker delta, δnm = 0 if n6=m n
1 if x>0,
ϑ(x) Heaviside unit step function, ϑ(x) = 0 if x≤0
xxxi
xxxii BASIC NOTATION AND REMARKS
Brief Notation for Derivatives

Partial derivatives:
∂w ∂w ∂2w ∂2w
wx = ∂x w = , wt = ∂t w = , wxx = ∂xx w = , wxt = ∂tx w = ,
∂x ∂t ∂x2 ∂x∂t
∂2w ∂3w ∂3w
wtt = ∂tt w = , w xxx = ∂xxx w = , w xxt = ∂xxt w = , ...
∂t2 ∂x3 ∂x2 ∂t
Ordinary derivatives for f = f (x):
df d2 f d3 f dn f
fx′ = , ′′
fxx = , ′′′
fxxx = , fx(n) = with n ≥ 4.
dx dx2 dx3 dxn
Special Functions
1
R∞ 1 3

Ai(x) = π 0 cos 3t + xt dt Airy function;
1/2
Ai(x) = π1 13 x K1/3 2 3/2
3x
∞
P
Ce2n+p (x, q) = A2n+p
2k+p cosh[(2k+p)x] even modified Mathieu functions, where
k=0 p = 0, 1; Ce2n+p (x, q) = ce2n+p (ix, q)
∞
P
ce2n (x, q) = A2n
2k cos 2kx even π-periodic Mathieu functions; these
k=0 satisfy the equation y ′′ +(a−2q cos 2x)y =
0, where a = a2n (q) are eigenvalues
∞
P
ce2n+1 (x, q) = A2n+1
2k+1 cos[(2k+1)x] even 2π-periodic Mathieu functions; these
0, where a = a2n+1 (q) are eigenvalues
Dν = Dν (x) parabolic cylinder function; it satisfies the
equation y ′′ + ν + 12 − 14 x2 y = 0
Rx
erf x = √2 exp −ξ 2 dξ error function
π 0
R∞
erfc x = √2π x exp −ξ 2 dξ complementary error function
2 dn −x2

Hn (x) = (−1)n ex dx n e Hermite polynomial
(1)
Hν (x) = Jν (x) + iYν (x) Hankel function of the first kind; i2 = −1
(2)
Hν (x) = Jν (x) − iYν (x) Hankel function of the second kind
∞
P (a)n (b)n xn
F (a, b, c; x) = 1 + (c)n n! hypergeometric function,
n=1 (a)n = a(a + 1) . . . (a + n − 1)
∞
P (x/2)ν+2n
Iν (x) = n! Γ(ν+n+1) modified Bessel function of the first kind
n=0
∞
P (−1)n (x/2)ν+2n
Jν (x) = n! Γ(ν+n+1) Bessel function of the first kind
n=0
Kν (x) = π2 I−νsin(πν)
(x)−Iν (x)
modified Bessel function of the second
kind
BASIC NOTATION AND REMARKS xxxiii
1 −s x dn

Lsn (x) = n! x e dxn x
n+s e−x generalized Laguerre polynomial
n
Pn (x) = n!12n dx
d 2
n (x − 1)
n Legendre polynomial
dm
Pnm (x) = (1 − x2 )m/2 dx m Pn (x) associated Legendre functions
∞
P 2n+p
Se2n+p (x, q) = B2k+p sinh[(2k+p)x] odd modified Mathieu functions, where
k=0 p = 0, 1; Se2n+p (x, q) = −i se2n+p (ix, q)
∞
P
se2n (x, q) = 2n sin 2kx
B2k odd π-periodic Mathieu functions; these
0, where a = b2n (q) are eigenvalues
∞
P 2n+1
se2n+1 (x, q) = B2k+1 sin[(2k+1)x] odd 2π-periodic Mathieu functions; these
0, where a = b2n+1 (q) are eigenvalues
Jν (x) cos(πν)−J−ν (x)
Yν (x) = sin(πν) Bessel function of the second kind
R x −ξ α−1
γ(α, x) = 0 e ξ dξ incomplete gamma function
R ∞ −ξ α−1
Γ(α) = 0 e ξ dξ gamma function
P∞
(a)n xn
Φ(a, b; x) = 1 + (b)n n! degenerate hypergeometric function,
n=1 (a)n = a(a + 1) . . . (a + n − 1)
Miscellaneous Remarks
1. The previous handbooks by Polyanin (2002) and Polyanin, Zaitsev, and Moussiaux
(2002) were extensively used in compiling this book; references to these sources are
often omitted.
2. The conventional abbreviations ODE and PDE stand for “ordinary differential equa-
tion” and “partial differential equation,” respectively.
3. The conventional abbreviations 2D equation and 3D equation stand for “two-dimen-
sional equation” and “three-dimensional equation,” respectively.
4. Throughout the book, unless explicitly specified otherwise, all parameters occurring
in the equations considered are assumed to be real numbers.
5. The term “exact solution” with regard to linear PDEs and systems of PDEs is used
in the following cases:
• the solution is expressible in terms of elementary functions;
• the solution is expressible via special functions, in closed form via infinite func-
tion series, and/or via definite (indefinite) integrals; the solution may depend on
arbitrary functions, which may occur in the equation itself or in the initial and
boundary conditions.
6. If a formula or a solution contains derivatives of some functions, then the functions
are assumed to be differentiable.
7. If a formula or a solution contains finite or definite integrals, then the integrals are
supposed to be convergent.
xxxiv BASIC NOTATION AND REMARKS
f (x)
8. If a formula or a solution contains an expression like a−2 , then the assumption that
a 6= 2 is implied but often not stated explicitly.
9. Equations are numbered separately within each subsection. In Chapters 1–12, when
referring to a particular equation, we use notation like 3.2.1.5, which denotes Eq. 5
in Section 3.2.1.
10. The symbol ⊙ indicates references to literature sources whenever
• at least one of the solutions was obtained in the cited source;
• the cited source provides further information on the equations in question and
their solutions.
11. The symbol ◮ marks the beginning of a small section; such sections are referred to
as paragraphs.
12. The symbol ⇒ stands for uniform convergence.
Part I
Exact Solutions
Chapter 1
First-Order Equations
with Two Independent Variables
∂w ∂w
1.1 Equations of the Form f (x, y) + g(x, y) =0
∂x ∂y
◆ For brevity, often only a principal integral
Ξ = Ξ(x, y)
of an equation will be presented in Section 1.1. The general solution of the equation is
given by
w = Φ(Ξ),
where Φ = Φ(Ξ) is an arbitrary function.
1.1.1 Equations Containing Power-Law Functions

◮ Coefficients of equations are linear in x and y.
∂w ∂w
1. a +b = 0.
∂x ∂y
General solution: w = Φ(bx − ay), where Φ is an arbitrary function.

⊙ Literature: E. Kamke (1965).
∂w ∂w
2. a + (bx + c) = 0.
∂x ∂y
Principal integral: Ξ = 12 bx2 + cx − ay.
∂w ∂w
3. + (ax + by + c) = 0.
∂x ∂y
Principal integral: Ξ = (abx + b2 y + a + bc)e−bx .
3
4 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂w ∂w
4. ax + by = 0.
∂x ∂y
For a = b, this is a conoid equation. Principal integral: Ξ = |x|b |y|−a .

∂w ∂w
5. ay + bx = 0.
∂x ∂y
Principal integral: Ξ = bx2 − ay 2 .

∂w ∂w
6. y + (y + a) = 0.
∂x ∂y
Principal integral: Ξ = x − y + a ln |y + a|.
∂w ∂w
7. (ay + bx + c) − (by + kx + s) = 0.
∂x ∂y
Principal integral: Ξ = ay 2 + kx2 + 2(bxy + cy + sx).

∂w ∂w
8. (a1 x + b1 y + c1 ) + (a2 x + b2 y + c2 ) = 0.
∂x ∂y
The principal integral is determined by solutions of the following auxiliary system of alge-
braic equations for the parameters s, λ, µ, α, β, and γ:
(a1 − s)(b2 − s) = a2 b1 , (1)

a1 λ + a2 µ = sλ, b1 λ + b2 µ = sµ, (2)
c1 α + c2 β − sγ = c1 λ + c2 µ, (3)
(a1 − s)α + a2 β = λs, b1 α + (b2 − s)β = µs. (4)
Case 1: (a1 − b2 )2 + 4a2 b1 6= 0. Equation (1) has two different roots s1 and s2 . To
these roots there correspond two sets of solutions, λ1 , µ1 and λ2 , µ2 , of system (2).
1.1. If a1 b2 − a2 b1 6= 0, then s1 6= 0 and s2 6= 0. Hence the principal integral has the
form
|s1 (λ1 x + µ1 y) + λ1 c1 + µ1 c2 |s2
Ξ= .
|s2 (λ2 x + µ2 y) + λ2 c1 + µ2 c2 |s1
1.2. If a1 b2 − a2 b1 = 0, then s1 = s = a1 + b2 and s2 = 0.
Principal integral for λ2 c1 + µ2 c2 6= 0:
λ2 x + µ2 y
Ξ=s − ln |s1 (λ1 x + µ1 y) + λ1 c1 + µ1 c2 |.
λ2 c1 + µ2 c2
Principal integral for λ2 c1 + µ2 c2 = 0:
Ξ = λ2 x + µ2 y.
1
Case 2: (a1 − b2 )2 + 4a2 b1 = 0. Equation (1) has the double root s = 2 (a1 + b2 ).
System (2) gives λ and µ not equal to zero simultaneously.
1.1. Equations of the Form f (x, y) ∂w
∂x
+ g(x, y) ∂w
∂y
=0 5
2.1. If s 6= 0, then we find γ from (3) and take nonzero α and β that satisfy relations (4).
This leads to the principal integral
s(αx + βy + γ)
Ξ = ln |s(λx + µy) + c1 λ + c2 µ| − .
s(λx + µy) + c1 λ + c2 µ
2.2. If s = 0, then b2 = −a1 . We have
Ξ = a2 x2 − 2a1 xy − b1 y 2 + 2c2 x − 2c1 y.
◮ Coefficients of equations are quadratic in x and y.
∂w ∂w
9. + (ax2 + bx + c) = 0.
∂x ∂y
Principal integral: Ξ = 13 ax3 + 12 bx2 + cx − y.
∂w ∂w
10. + (ay 2 + by + c) = 0.
∂x ∂y
1◦ . Principal integral for 4ac − b2 > 0:
2 2ay + b
Ξ=x− √ arctan √ .
4ac − b2 4ac − b2
2◦ . Principal integral for 4ac − b2 < 0:

√
2 2ay + b − b2 − 4ac
Ξ=x− √ ln √ .
b2 − 4ac 2ay + b + b2 − 4ac
∂w ∂w
11. + (ay + bx2 + cx) = 0.
∂x ∂y
This is a special case of equation 1.1.7.1 with f (x) = a and g(x) = bx2 + cx.
∂w ∂w
12. + (axy + bx2 + cx + ky + s) = 0.
∂x ∂y
Z

Principal integral: Ξ = y exp − 12 ax2 − kx − (bx2 + cx + s) exp − 12 ax2 − kx dx.
∂w ∂w
13. + (y 2 − a2 x2 + 3a) = 0.
∂x ∂y
Z
exp(ax2 ) dx
Principal integral: Ξ = + exp(ax2 ) .
x(xy − ax2 + 1) x2
∂w ∂w
14. + (y 2 − a2 x2 + a) = 0.
∂x ∂y
This is a special case of equation 1.1.1.59 with n = 1.
∂w ∂w
15. + (y 2 + axy + a) = 0.
∂x ∂y
∂w ∂w
16. + (y 2 + axy − abx − b2 ) = 0.
∂x ∂y
∂w ∂w
17. + k(ax + by + c)2 = 0.
∂x ∂y
This is a special case of equation 1.1.8.6 with f (z) = kz 2 .
∂w ∂w
18. x + (ay 2 + cx2 + y) = 0.
∂x ∂y
This is a special case of equation 1.1.1.75 with b = 1.
∂w ∂w
19. x + (ay 2 + bxy + cx2 + y) = 0.
∂x ∂y
∂w ∂w
20. (ax + c) + α(ay + bx)2 + β(ay + bx) − bx + γ = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ = ln |ax + c| − , v = ay + bx.
αv 2 + βv + γ + bc/a
∂w ∂w
21. ax2 + by 2 = 0.
∂x ∂y
1 1
Principal integral: Ξ = − .
by ax
∂w ∂w
22. (ax2 + b) − y 2 − 2xy + (1 − a)x2 − b = 0.
∂x ∂y
Z
dx 1
Principal integral: Ξ = − 2
+ .
ax + b y−x
∂w ∂w
23. (a1 x2 + b1 x + c1 ) + (a2 y 2 + b2 y + c2 ) = 0.
∂x ∂y
Z Z
dx dy
Principal integral: Ξ = 2
− 2
.
a1 x + b1 x + c1 a2 y + b2 y + c2
∂w ∂w
24. (x − a)(x − b) − y 2 + k(y + x − a)(y + x − b) = 0.
∂x ∂y
1◦ . Principal integral for a 6= b:
k
y + k(y + x − a) x−a
Ξ= , k 6= 0, k 6= −1.
y + k(y + x − b) x−b
2◦ . Principal integral for a = b:

(x − a) + y + k(y + x − a)
Ξ= , k 6= 0, k 6= −1.
y + k(y + x − a) (x − a)
∂x
+ g(x, y) ∂w
∂y
=0 7
∂w ∂w
25. (a1 y 2 + b1 y + c1 ) + (a2 x2 + b2 x + c2 ) = 0.
∂x ∂y
Principal integral: Ξ = 13 a1 y 3 + 12 b1 y 2 + c1 y − 13 a2 x3 − 12 b2 x2 − c2 x.
∂w ∂w
26. y(ax + b) + (ay 2 − cx) = 0.
∂x ∂y
(ax + b)2
Principal integral: Ξ = .
cx2 + by 2
∂w ∂w
27. (ay 2 + bx) − (cx2 + by) = 0.
∂x ∂y
Principal integral: Ξ = 13 ay 3 + 13 cx3 + bxy.
∂w ∂w
28. (ay 2 + bx2 ) + 2bx = 0.
∂x ∂y
This is a special case of equation 1.1.8.2 with f (x) = bx2 and g(y) = ay 2 .
∂w ∂w
29. (ay 2 + bx2 ) + 2bxy = 0.
∂x ∂y
bx2 − ay 2
Principal integral: Ξ = .
y
∂w ∂w
30. (ay 2 + x2 ) + (bx2 + c − 2xy) = 0.
∂x ∂y
Principal integral: Ξ = ay 3 − bx3 + 3(x2 y − cx).
∂w ∂w
31. (Ay 2 + Bx2 − a2 B) + (Cy 2 + 2Bxy) = 0.
∂x ∂y
Principal integral:
Z
E dv y
Ξ = (x − a)E + 2aB , v= ,
v(Av 2 − Cv − B) x−a
Z
(Av 2 + B) dv
where E = exp .
v(Av 2 − Cv − B)
∂w ∂w
32. (ay 2 + bx2 + cy) + 2bx = 0.
∂x ∂y
This is a special case of equation 1.1.8.2 with f (x) = bx2 and g(y) = ay 2 + cy.
∂w ∂w
33. (Axy + Bx2 + kx) + (Dy 2 + Exy + F x2 + ky) = 0.
∂x ∂y
Principal integral:
Z
V dv y
Ξ = xV + k , v= ,
(A − D)v 2
+ (B − E)v − F x
Z
(Av + B) dv
where V = exp .
(A − D)v 2 + (B − E)v − F
∂w ∂w
34. (Axy + Aky + Bx2 + Bkx) + Cy 2 + Dxy + k(D − B)y = 0.
∂x ∂y
Principal integral:
Z
E dv y
Ξ = (x + k)E + kB , v= ,
v (C − A)v + D − B x+k
Z
(Av + B) dv
where E = exp .
v (A − C)v + B − D
∂w ∂w
35. (Ay 2 + Bxy + Cx2 + kx) + (Dy 2 + Exy + F x2 + ky) = 0.
∂x ∂y
Principal integral:
Z
V dv y
Ξ = xV + k , v= ,
Av 3 + (B − D)v 2 + (C − E)v − F x
Z
(Av 2 + Bv + C) dv
where V = exp .
Av 3 + (B − D)v 2 + (C − E)v − F
∂w ∂w
36. (Ay 2 + Bxy + Cx2 ) + (Dy 2 + Exy + F x2 ) = 0.
∂x ∂y
Principal integral:
Z
(Av 2 + Bv + C) dv y
Ξ= + ln |x|, v= .
Av 3 + (B − D)v 2 + (C − E)v − F x
∂w ∂w
37. (Ay 2 + 2Bxy + Dx2 + a) − (By 2 + 2Dxy − Ex2 − b) = 0.
∂x ∂y
Principal integral: Ξ = Ay 3 − Ex3 + 3(Bxy 2 + Dx2 y + ay − bx).
∂w ∂w
38. (y 2 − 2xy + x2 + ay) + ay = 0.
∂x ∂y
a
Principal integral: Ξ = + ln |y|.
x−y
∂w ∂w
39. (xf1 − f2 ) + (yf1 − f3 ) = 0, fn = an + bn x + cn y.
∂x ∂y
Hesse’s equation. The introduction of the homogeneous coordinates x = ξ2 /ξ1 , y = ξ3 /ξ1

leads to an equation with three independent variables for w = w(ξ1 , ξ2 , ξ3 ):
∂w ∂w ∂w
g1 + g2 + g3 = 0,
∂ξ1 ∂ξ2 ∂ξ3
where gn = an ξ1 + bn ξ2 + cn ξ3 (n = 1, 2, 3). See 2.1.1.21 for the solution of this equation.

∂x
+ g(x, y) ∂w
∂y
=0 9
◮ Coefficients of equations contain integer powers of x and y.

∂w ∂w
40. + (y 2 + bx2 y − a2 − abx2 ) = 0.
∂x ∂y
This is a special case of equation 1.1.7.3 with f (x) = bx2 .
∂w ∂w
41. + (ax2 y + bx3 + c) = 0.
∂x ∂y
This is a special case of equation 1.1.7.1 with f (x) = ax2 and g(x) = bx3 + c.
∂w ∂w
42. + (ax2 y + by 3 ) = 0.
∂x ∂y
This is a special case of equation 1.1.7.2 with k = 3, f (x) = ax2 , and g(x) = b.
∂w ∂w
43. + (axy + b)y 2 = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − ln |x|, v = xy.
v(av 2 + bv + 1)
∂w ∂w
44. + A(ax + by + c)3 = 0.
∂x ∂y
This is a special case of equation 1.1.8.6 with f (z) = Az 3 .
∂w ∂w
45. x + ax4 y 3 + (bx2 − 1)y + cx = 0.
∂x ∂y
Principal integral: Z
dv x2
Ξ= − , v = xy.
av 3 + bv + c 2
∂w ∂w
46. x2 + (ax2 y 2 + bxy + c) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − ln |x|, v = xy.
av 2 + (b + 1)v + c
∂w ∂w
47. (ax2 y + b) − (axy 2 + c) = 0.
∂x ∂y
Principal integral: Ξ = 12 ax2 y 2 + by + cx.
∂w ∂w
48. (ax + by 3 ) − (cx3 + ay) = 0.
∂x ∂y
Principal integral: Ξ = axy + 14 by 4 + 14 cx4 .

◆ See also equations 1.1.1.56–1.1.1.111 for integer values of exponents.
◮ Coefficients of equations contain fractional powers.
∂w √ ∂w
49. + (a x y + b) = 0.
∂x ∂y
Z

Principal integral: Ξ = y exp − 23 ax3/2 − b exp − 23 ax3/2 dx.
∂w √ √ ∂w
50. + (a x y + b y ) = 0.
∂x ∂y
√
This is a special case of equation 1.1.7.2 with k = 12 , f (x) = a x, and g(x) = b.
∂w √ √ ∂w
51. + (a x y + bx y ) = 0.
∂x ∂y
√
This is a special case of equation 1.1.7.2 with k = 12 , f (x) = a x, and g(x) = bx.
∂w p ∂w
52. + A ax + by + c = 0.
∂x ∂y
√
This is a special case of equation 1.1.8.6 with f (z) = A z.
∂w p ∂w
53. x + ay + b y 2 + cx2 = 0.
∂x ∂y
1◦ . Principal integral for a 6= 1:

Z
du y
Ξ = ln |x| − √ , u= .
(a − 1)u + b u2 + c x
2◦ . Principal integral for a = 1:

p
Ξ = |x|−b−1 y + y 2 + cx2 .
√ ∂w √ ∂w
54. (ax + b y ) − (c x + ay) = 0.
∂x ∂y
Principal integral: Ξ = axy + 23 by 3/2 + 23 cx3/2 .
p p 4
X
∂w ∂w
55. f (x) + f (y) = 0, f (t) = a ν tν .
∂x ∂y
ν=0
p p 2
f (x) + f (y)
Principal integral: Ξ = − a4 (x + y)2 − a3 (x + y).
x−y
◆ See also equations in 1.1.1.56–1.1.1.111 for fractional values of exponents.

∂x
+ g(x, y) ∂w
∂y
=0 11
◮ Coefficients of equations contain arbitrary powers of x and y.

∂w ∂w
56. + (ay + bxk ) = 0.
∂x ∂y
Z
−ax
Principal integral: Ξ = ye − b xk e−ax dx.
∂w ∂w
57. + (axk y + bxn ) = 0.
∂x ∂y
Z
a a
Principal integral: Ξ = y exp − xk+1 −b xn exp − xk+1 dx.
k+1 k+1
∂w ∂w
58. + (ay 2 + bxn ) = 0.
∂x ∂y
The principal integral Ξ(x, y) can be found as the general solution Ξ(x, y) = C of the spe-
cial Riccati equation yx′ = ay 2 + bxn , which is considered in the handbooks by G. M. Mur-
phy (1960), E. Kamke (1977), and A. D. Polyanin and V. F. Zaitsev (2003).
∂w ∂w
59. + (y 2 + anxn−1 − a2 x2n ) = 0.
∂x ∂y
1◦ . Principal integral for n 6= −1:
Z
E 2a n+1
Ξ= + E dx, E = exp x .
y − axn n+1
2◦ . Principal integral for n = −1 and a 6= − 12 :
xy + a + 1
Ξ= x2a+1 .
(2a + 1)(xy − a)
3◦ . Principal integral for n = −1 and a = − 12 :
2
Ξ= + ln |x|.
2xy + 1
⊙ Literature: V. F. Zaitsev and A. D. Polyanin (1996).
∂w ∂w
60. + (y 2 + axn y + axn−1 ) = 0.
∂x ∂y
Z
E a
Ξ= + x−2 E dx, E = exp x n+1
.
x(xy + 1) n+1
2◦ . Principal integral for n = −1 and a 6= 1:
xy + a
Ξ= xa−1 .
(a − 1)(xy + 1)
3◦ . Principal integral for n = −1 and a = 1:
1
Ξ= + ln |x|.
xy + 1
∂w ∂w
61. + (y 2 + axn y − abxn − b2 ) = 0.
∂x ∂y

Z
1 a n+1 a n+1
Ξ= exp 2bx + x + exp 2bx + x dx.
y−b n+1 n+1
2◦ . Principal integral for n = −1:

Z
xa e2bx
Ξ= + xa e2bx dx.
y−b
∂w ∂w
62. + (axn y 2 + bx−n−2 ) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ = ln |x| − , v = xn+1 y.
av 2 + (n + 1)v + b
∂w ∂w
63. + (axn y 2 + bmxm−1 − ab2 xn+2m ) = 0.
∂x ∂y
1◦ . Principal integral for n + m 6= −1:

Z
E n 2ab n+m+1
Ξ= +a x E dx, E = exp x .
y − bxm n+m+1
2◦ . Principal integral for n + m = −1 and m 6= 2ab:
x2ab a
Ξ= m
+ x2ab−m .
y − bx 2ab − m
3◦ . Principal integral for n + m = −1 and m = 2ab:
xm
Ξ= + a ln x.
y − bxm
⊙ Literature: A. D. Polyanin and V. F. Zaitsev (1996).
∂w ∂w
64. − (n + 1)xn y 2 − axn+m+1 y + axm = 0.
∂x ∂y
Principal integral:
Z
x−n−1 E −n−2 a n+m+2
Ξ = n+1 − (n + 1) x E dx, E = exp x .
x y−1 n+m+2
∂x
+ g(x, y) ∂w
∂y
=0 13
∂w ∂w
65. + (axn y 2 + bxm y + bcxm − ac2 xn ) = 0.
∂x ∂y
1◦ . Principal integral for m, n 6= −1:
Z
E b 2ac n+1
Ξ= + a xn E dx, E = exp x m+1
− x .
y+c m+1 n+1
2◦ . Principal integral for n = −1:
Z
x−2ac b m+1 −2ac−1 b m+1
Ξ= exp x +a x exp x dx.
y+c m+1 m+1
3◦ . Principal integral for m = −1:
Z
xb 2ac n+1 n+b 2ac n+1
Ξ= exp − x +a x exp − x dx.
y+c n+1 n+1
∂w ∂w
66. + axn y 2 − axn (bxm + c)y + bmxm−1 = 0.
∂x ∂y
1◦ . Principal integral for n 6= −1 and m + n 6= −1:
Z
E n abxn+m+1 acxn+1
Ξ= + a x E dx, E = exp + .
y − bxm − c n+m+1 n+1
2◦ . Principal integral for n = −1 and m 6= 0:
Z
xac ab m ac−1 ab m
y − bxm − c m m
3◦ . Principal integral for n 6= −1 and m = −1 − n:
Z
xab ac n+1 ab+n ac n+1
y − bx−n−1 − c n+1 n+1
∂w ∂w
67. − anxn−1 y 2 − cxm (axn + b) + cxm = 0.
∂x ∂y
1◦ . Principal integral for m 6= −1 and m + n 6= −1:
Z
E xn−1 E
Ξ= − an dx,
(ax + b) (axn + b)y − 1
n (axn + b)2

acxm+n+1 bcxm+1
E = exp + .
m+n+1 m+1
2◦ . Principal integral for m = −1 and n 6= 0:
Z bc+n−1
xbc ac n ac n x
Ξ= exp x − an exp x dx.
n n
(ax + b) (ax + b)y − 1 n n (axn + b)2
3◦ . Principal integral for n 6= −1 and m = −1 − n:
Z
xac bc xac+n−1 bc −n
Ξ= exp − x−n −an exp − x dx.
(axn +b) (axn +b)y−1 n (axn +b)2 n
∂w ∂w
68. + (axn y 2 + bxm y + ckxk−1 − bcxm+k − ac2 xn+2k ) = 0.
∂x ∂y
1◦ . Principal integral for m 6= −1 and n + k 6= −1:
Z
E n 2ac n+k+1 b m+1
Ξ= + a x E dx, E = exp x + x .
y − cxk n+k+1 m+1
2◦ . Principal integral for m = −1 and n + k =

6 −1:
Z
xb E b+n 2ac n+k+1
Ξ= + a x E dx, E = exp x .
y − cxk n+k+1
3◦ . Principal integral for m 6= −1 and n + k = −1:

Z
x2ac b m+1 2ac+n b m+1
y − cxk m+1 m+1
4◦ . Principal integral for m = −1, n + k = −1, and 2ac + b 6= k:
ay + (ac + b − k)xk 2ac+b−k

Ξ= x .
(2ac + b − k)(y − cxk )
5◦ . Principal integral for m = −1, n + k = −1, and 2ac + b = k:
xk
Ξ= + a ln x.
y − cxk
∂w ∂w
69. + (ax2n+1 y 3 + bx−n−2 ) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − ln |x|, v = xn+1 y.
av 3 + (n + 1)v + b
∂w ∂w
70. + (axn y 3 + 3abxn+m y 2 − bmxm−1 − 2ab3 xn+3m ) = 0.
∂x ∂y
1◦ . Principal integral for n + 2m 6= −1:
Z
E 6ab2
Ξ= + 2a xn E dx, E = exp − x n+2m+1
.
(y + bxm )2 n + 2m + 1
2◦ . Principal integral for n = −2m − 1:

2
x−6ab a 2
Ξ= + x−2(3ab +m) .
(y + bxm )2 3ab2 + m
3◦ . Principal integral for n = −2m − 1 and m = −3ab2 :
x2m
Ξ= + 2a ln |x|.
(y + bxm )2
∂x
+ g(x, y) ∂w
∂y
=0 15
∂w
71. + (axn y 3 + 3abxn+m y 2 + cxk y
∂x
∂w
− 2ab3 xn+3m + bcxm+k − bmxm−1 ) = 0.
∂y
1◦ . Principal integral for k 6= −1 and n + 2m 6= −1:
Z
E 2c k+1 6ab2
Ξ= + 2a xn E dx, E = exp x − xn+2m+1 .
(y + bxm )2 k+1 n + 2m + 1
2◦ . Principal integral for k = −1 and n + 2m 6= −1:

Z
x2c E2 n+2c 6ab2 n+2m+1
Ξ= + 2a x E2 dx, E2 = exp − x .
(y + bxm )2 n + 2m + 1
3◦ . Principal integral for k 6= −1 and n + 2m = −1:

2 Z
x−6ab E1 n−6ab2 2c k+1
Ξ= + 2a x E1 dx, E1 = exp x .
(y + bxm )2 k+1
4◦ . Principal integral for k = n + 2m = −1 and c 6= 3ab2 + m:

2
x2(c−3ab ) a 2
Ξ= m 2
+ 2
x2(c−3ab −m) .
(y + bx ) c − 3ab − m
5◦ . Principal integral for k = n + 2m = −1 and c = 3ab2 + m:
x2m
Ξ= + 2a ln |x|.
(y + bxm )2
n
∂w ∂w
72. + ay n + bx 1 − n = 0.
∂x ∂y
Principal integral:
Z 1
dv
Ξ= 1 − ln |x|, v = yx n−1 .
av n + n−1 v +b
∂w ∂w
73. + (axm−n−mn y n + bxm ) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ = ln |x| − , v = yx−m−1 .
av n − (m + 1)v + b
∂w ∂w
74. + (axn y k + bxm y) = 0.
∂x ∂y
This is a special case of equation 1.1.7.2 with f (x) = bxm and g(x) = axn .
∂w ∂w
75. x + (ay 2 + by + cx2b ) = 0.
∂x ∂y
1◦ . Principal integral for ac > 0:
r
b a −b
Ξ = √ arctan x y − xb .
ac c
2◦ . Principal integral for ac < 0:
√
b ax−b y − −ac
Ξ= √ ln √ − xb .
2 −ac ax−b y + −ac
∂w ∂w
76. x + ay 2 + (n + bxn )y + cx2n = 0.
∂x ∂y
1◦ . Principal integral for n 6= 0:
Z
dv 1
Ξ= 2
− xn , v = x−n y.
av + bv + c n
2◦ . Principal integral for n = 0:
Z
dy
Ξ= − ln |x|.
ay 2 + by + c
∂w ∂w
77. x + (axn y 2 + by + cx−n ) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − ln x, v = xn y.
av 2 + (b + n)v + c
∂w ∂w
78. x + (axn y 2 + my − ab2 xn+2m ) = 0.
∂x ∂y
1◦ . Principal integral for m + n 6= 0:
Z
xm E 2ab m+n
Ξ= + a xm+n−1 E dx, E = exp x .
y − bxm m+n
2◦ . Principal integral for m = −n:
x2ab (y + bxm )
Ξ= .
2b(y − bxm )
∂w ∂w
79. x + x2n y 2 + (m − n)y + x2m = 0.
∂x ∂y
xn+m
Principal integral: Ξ = arctan(xn−m y) − .
n+m
∂x
+ g(x, y) ∂w
∂y
=0 17
∂w ∂w
80. x + ax2n y 2 + (bxn − n)y + c = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ=n − xn , v = xn y.
av 2 + bv + c
∂w ∂w
81. x + ax2n+m y 2 + (bxn+m − n)y + cxm = 0.
∂x ∂y
1◦ . Principal integral for n + m 6= 0:
Z
dv xn+m
Ξ= − , v = xn y.
av 2 + bv + c n+m
2◦ . Principal integral for n + m = 0:

Z
dv
Ξ= 2
− ln x, v = xn y.
av + bv + c
∂w ∂w
82. x + (ay 3 + 3abxn y 2 − bnxn − 2ab3 x3n ) = 0.
∂x ∂y
Principal integral:
Z
E −1 3ab2 2n
Ξ= + 2a x E dx, E = exp − x .
(y + bxn )2 n
∂w ∂w
83. x + ax2n+1 y 3 + (bx − n)y + cx1−n = 0.
∂x ∂y
dv
Ξ= − x, v = xn y.
av 3 + bv + c
∂w ∂w
84. x + axn+2 y 3 + (bxn − 1)y + cxn−1 = 0.
∂x ∂y
dv 1
Ξ= − xn , v = xy.
av 3 + bv + c n
∂w ∂w
85. x + (y + axn−m y m + bxn−k y k ) = 0.
∂x ∂y
Z
dv xn−1 y
Ξ= − , v= .
av m + bv k n−1 x
2◦ . Principal integral for n = 1:

Z
dv y
Ξ= − ln |x|, v= .
av m + bv k x
∂w ∂w
86. y + xn−1 [(1 + 2n)x + an]y − nx2n (x + a) = 0.
∂x ∂y
Principal integral:
Z
n+1 n −1/n dv
Ξ = (x + ax − y) + , v = x(xn+1 + axn − y)−1/n .
a − v −n
∂w ∂w
87. y + [a(2n + k)xk + b]xn−1 y − (a2 nx2k + abxk − c)x2n−1 = 0.
∂x ∂y
Principal integral:
Z
−k E dv
Ξ=x E − ak , v = x−n y − axk ,
nv 2 − bv − c
Z
v dv
where E = exp −k .
nv 2 − bv − c
∂w ∂w
88. x(2axy + b) − a(m + 3)xy 2 + b(m + 2)y − cxm = 0.
∂x ∂y
m+2

Principal integral: Ξ = cx cxm − 2(m + 1)y(axy + b) .
∂w ∂w
89. x2 (2axy + b) − (4ax2 y 2 + 3bxy − cx2 − k) = 0.
∂x ∂y
Principal integral: Ξ = (cx2 + k)2 − 4cx3 y(axy + b).
∂w ∂w
90. axm + by n = 0.
∂x ∂y
1◦ . Principal integral for m 6= 1 and n 6= 1:
Ξ = b(n − 1)x1−m − a(m − 1)y 1−n .
2◦ . Principal integral for m = 1 and n 6= 1:

a
Ξ = b ln |x| + y 1−n .
n−1
3◦ . Principal integral for m 6= 1 and n = 1:
b
Ξ= x1−m + a ln |y|.
m−1
4◦ . Principal integral for m = n = 1:
Ξ = b ln |x| − a ln |y|.

∂x
+ g(x, y) ∂w
∂y
=0 19
∂w ∂w
91. axn + (by + cxm ) = 0.
∂x ∂y
Z
−F c b
Ξ=e y− e−F xm−n dx, F = x1−n .
a a(1 − n)
2◦ . Principal integral for n = 1 and am 6= b:
c am−b
Ξ = x−b/a y − x a .
am − b
3◦ . Principal integral for n = 1 and am = b:

c
Ξ = x−b/a y − ln |x|.
a
∂w ∂w
92. axk + (y n + bxm y) = 0, n 6= 1.
∂x ∂y
1◦ . Principal integral for m 6= k − 1:

Z
n−1 (1 − n)b
Ξ = e−F y 1−n + e−F x−k dx, F = xm−k+1 .
a a(m + k − 1)
2◦ . Principal integral for m = k − 1 and (n − 1)b 6= ma:
(n−1)b n−1 (n−1)b−ma

Ξ=x a y 1−n + x a .
(n − 1)b − ma
3◦ . Principal integral for m = k − 1 and (n − 1)b = ma:
(n−1)b n−1
Ξ=x a y 1−n + ln |x|.
a
∂w ∂w
93. x(axk + b) + αxn y 2 + (β − anxk )y + γx−n = 0.
∂x ∂y
Principal integral:
Z
E dv
Ξ = x−k E + ka , v = xn y,
αv 2 + (β + bn)v + γ
Z
dv
where E = exp kb .
αv 2 + (β + bn)v + γ
∂w ∂w
94. (y + Axn + a) − (nAxn−1 y + kxm + b) = 0.
∂x ∂y
2k
Principal integral: Ξ = y 2 + xm+1 + 2(Axn y + ay + bx).
m+1
∂w ∂w
95. (y + axn+1 + bxn ) + (anxn + cxn−1 )y = 0.
∂x ∂y
Principal integral:
Z
−1 E dv
Ξ=x E−a , v = x−n y + b,
nv 2
− (bn + c)v + bc
Z
v dv
where E = exp − .
nv 2 − (bn + c)v + bc
∂w ∂w
96. x(2axn y+b) − a(3n+m)xn y 2 +b(2n+m)y−Axm −Cx−n = 0.
∂x ∂y
Principal integral: Ξ = (Axn+m + C)2 − 2A(n + m)x2n+m y(axn y + b).
∂w ∂w
97. (axn + bx2 + xy) + (cxn + bxy + y 2 ) = 0.
∂x ∂y
Principal integral:
Z
1 y
Ξ= (ay − cx)n−2 + (v + b)(av − c)n−3 dv, v= .
n−2 x
∂w ∂w
98. (ay n + bx2 + cxy) + (ky n + bxy + cy 2 ) = 0.
∂x ∂y
Principal integral:
Z
1 (k − av)n−3 (b + cv) y
Ξ= (kx − ay)n−2 − dv, v= .
n−2 vn x
∂w ∂w
99. (axn + bxm + c) + (cy 2 − bxm−1 y + axn−2 ) = 0.
∂x ∂y
Principal integral:
Z Z
E dx xE (bxm + 2c) dx
Ξ=c n m
+ , E = exp − .
ax + bx + c xy + 1 x(axn + bxm + c)
∂w ∂w
100. (axn + bxm + c) + (axn−2 y 2 + bxm−1 y + c) = 0.
∂x ∂y
Principal integral:
Z Z
xn−2 E dx E (2axn + bxm ) dx
Ξ=a n m
+ , E = exp .
ax + bx + c y−x x(axn + bxm + c)
∂w ∂w
101. (axn + bxm + c) + (αxk y 2 + βxs y − αλ2 xk + βλxs ) = 0.
∂x ∂y
Principal integral:
Z Z
xk E dx E βxs − 2αλxk
Ξ=α n m
+ , E = exp dx .
ax + bx + c y+λ axn + bxm + c
∂x
+ g(x, y) ∂w
∂y
=0 21
∂w ∂w
102. x(axn + bxm + c) − sxk y 2 − (axn + bxm + c)y − sλxk+2 = 0.
∂x ∂y
√
y−x λ √ Z xk dx
Principal integral: Ξ = √ exp 2s λ .
y+x λ axn + bxm + c
∂w
103. (axn + bxm + c)
∂x
∂w
+ (ax + bx + c)y 2 − an(n − 1)xn−2 − bm(m − 1)xm−2
n m
= 0.
∂y
Principal integral:
Z
1 dx
Ξ= + .
(axn + bxm + c) (axn + bxm + c)y + anxn−1 + bmxm−1 (axn + bxm + c)2
∂w ∂w
104. (axn + by n + x) + (αxk y n−k + βxm y n−m + y) = 0.
∂x ∂y
Principal integral:
Z
1 n−1 E dv y
Ξ= x E− n−k n−m
, v= ,
n−1 αv + βv − bv n+1 − av x
Z
(bv n + a) dv
where E = exp (1 − n) .
αv n−k + βv n−m − bv n+1 − av
∂w
105. (axn + by n + Ax2 + Bxy)
∂x
∂w
+ (αxk y n−k + βxm y n−m + Axy + By 2 ) = 0.
∂y
Principal integral:
Z
1 n−2 (Bv + A)E dv y
Ξ= x E− n−k
, v= ,
n−2 αv + βv n−m − bv n+1 − av x
Z
(bv n + a) dv
where E = exp (2 − n) .
αv n−k + βv n−m − bv n+1 − av
∂w ∂w
106. (ay m + bxn + s) − (αxk + bnxn−1 y + β) = 0.
∂x ∂y
Principal integral:
Ξ = aϕ(y) + αψ(x) + bxn y + sy + βx,
where
 
 m+1  k+1
y if m 6= −1, x if k 6= −1,
ϕ(y) = m + 1 ψ(x) = k + 1

ln |y| 
ln |x|
if m = −1, if k = −1.
∂w ∂w
107. (axn y m + x) + (bxk y n+m−k + y) = 0.
∂x ∂y
Principal integral:
Z
1 n+m−1 E dv y
Ξ= x E−a , v= ,
n+m−1 v m (bv n−k − av) x
Z
dv
where E = exp a(1 − n − m) .
bv n−k − av
∂w ∂w
108. x(axn y m + α) − y(bxn y m + β) = 0.
∂x ∂y
Principal integral:
(y a xb )A (y α xβ )B mβ − nα mb − na
Ξ= + , where A= , B= .
A B aβ − bα aβ − bα
∂w ∂w
109. x(anxk y n+k + s) − y(bmxm+k y k + s) = 0.
∂x ∂y
Principal integral: Ξ = aky n + bkxm − s(xy)−k .

∂w ∂w
110. (axn y m + Ax2 + Bxy) + (bxk y n+m−k + Axy + By 2 ) = 0.
∂x ∂y
Principal integral:
Z
1 n+m−2 (Bv + A)E dv y
Ξ= x E−a , v= ,
n+m−2 v m (bv n−k − av) x
Z
dv
where E = exp a(2 − n − m) .
bv n−k − av
∂w ∂w
111. (axn y m + bxy k ) + (αy s + β) = 0.
∂x ∂y
Principal integral:
Z Z
1 ym E y k dy
Ξ= x1−n E − a dy, E = exp b(n − 1) .
1−n αy s + β αy s + β
1.1.2 Equations Containing Exponential Functions

◮ Coefficients of equations contain exponential functions.
∂w ∂w
1. + aeλx = 0.
∂x ∂y
Principal integral: Ξ = λy − aeλx .

∂w ∂w
2. + aeλx + b = 0.
∂x ∂y
Principal integral: Ξ = λ(bx − y) + aeλx .

∂x
+ g(x, y) ∂w
∂y
=0 23
∂w ∂w
3. + aeλy + b = 0.
∂x ∂y

Principal integral: Ξ = λ(bx − y) + lnb + aeλy .
∂w ∂w
4. + aeλy+βx + b = 0.
∂x ∂y
aλ (β+bλ)x
Principal integral: Ξ = ebλx−λy + e .
β + bλ
∂w ∂w
5. + aeλy+βx + beγx = 0.
∂x ∂y
Z
−λy βx bλ γx
Principal integral: Ξ = e E + aλ e E dx, where E = exp e .
γ
∂w ∂w
6. aeλx + beβy = 0.
∂x ∂y
1 −βy 1 −λx
Principal integral: Ξ = e − e .
βb λa
∂w ∂w
7. aeλx + b + ceβx + d = 0.
∂x ∂y
Z
ceβx + d
Principal integral: Ξ = y − dx.
aeλx + b
∂w ∂w
8. aeλx + b + ceβy + d = 0.
∂x ∂y

Principal integral: Ξ = λβ(dx − by) − dβ lnaeλx + b + bλ lnceβy + d.
∂w ∂w
9. aeλy + b + ceβx + d = 0.
∂x ∂y
Principal integral: Ξ = βλ(dx − by) + cλeβx − aβeλy .
∂w ∂w
10. aeλx + beβy + aλeλx = 0.
∂x ∂y
b
Principal integral: Ξ = aeλx−y − e(β−1)y .
β −1
∂w ∂w
11. aeλx+βy + cµ − beγx+µy + cλ = 0.
∂x ∂y
a b
Principal integral: Ξ = e(β−µ)y + e(γ−λ)x − ce−λx−µy .
β−µ γ−λ
◮ Coefficients of equations contain exponential and power-law functions.

∂w ∂w
12. + (y 2 + aλeλx − a2 e2λx ) = 0.
∂x ∂y
Principal integral:
Z
E 2a λx
Ξ= + E dx, E = exp e .
y − aeλx λ
∂w ∂w
13. + y 2 + by + a(λ − b)eλx − a2 e2λx = 0.
∂x ∂y
Principal integral:
Z
E 2a λx
Ξ= + E dx, E = exp e + bx .
y − aeλx λ
∂w ∂w
14. + (y 2 + aeλx y − abeλx − b2 ) = 0.
∂x ∂y
Principal integral:
Z
E a λx
Ξ= + E dx, E = exp 2bx + e .
y−b λ
∂w ∂w
15. − (y 2 − axeλx y + aeλx ) = 0.
∂x ∂y
Principal integral:
Z
E E a λx
Ξ= − dx, E = exp 2 (λx − 1)e .
x(xy − 1) x2 λ
∂w ∂w
16. + (aeλx y 2 + be−λx ) = 0.
∂x ∂y
dv
Ξ= − x, v = eλx y.
av 2 + λv + b
∂w ∂w
17. + aeλx y 2 + bµeµx − ab2 e(λ+2µ)x = 0.
∂x ∂y
Principal integral:
Z
E λx 2ab (λ+µ)x
Ξ= +a e E dx, E = exp e .
y − beµx λ+µ
∂w ∂w
18. + (aeλx y 2 + by + ce−λx ) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − x, v = eλx y.
av 2 + (b + λ)v + c
∂x
+ g(x, y) ∂w
∂y
=0 25
∂w ∂w
19. + aeλx y 2 + µy − ab2 e(λ+2µ)x = 0.
∂x ∂y
Principal integral:
Z
E λx 2ab (λ+µ)x
Ξ= +a e E dx, E = exp e + µx .
y − beµx λ+µ
∂w ∂w
20. + eλx y 2 + aeµx y + aλe(µ−λ)x = 0.
∂x ∂y
Principal integral:
Z
E λx a µx
Ξ= + e E dx, E = exp e − 2λx .
y + λe−λx µ
∂w ∂w
21. − λeλx y 2 − aeµx y + ae(µ−λ)x = 0.
∂x ∂y
Principal integral:
Z
E a µx
Ξ= −λ eλx E dx, E = exp e − 2λx .
y − e−λx µ
∂w ∂w
22. + aeλx y 2 + abe(λ+µ)x y − bµeµx = 0.
∂x ∂y
Principal integral:
Z
E λx ab (λ+µ)x
Ξ= +a e E dx, E = exp − e .
y + beµx λ+µ
∂w ∂w
23. + ae(2λ+µ)x y 2 + (be(λ+µ)x − λ)y + ceµx = 0.
∂x ∂y
Principal integral:
Z
dv 1
Ξ= − e(µ+λ)x , v = eλx y.
av 2 + bv + c µ+λ
∂w ∂w
24. + eλx (y − beµx )2 + bµeµx = 0.
∂x ∂y
1 1
Principal integral: Ξ = µx
+ eλx .
y − be λ
∂w ∂w
25. + (aeλx y 2 + bnxn−1 − ab2 eλx x2n ) = 0.
∂x ∂y
Principal integral:
Z Z
E λx n λx
Ξ= +a e E dx, E = exp 2ab x e dx .
y − bxn
∂w ∂w
26. + (eλx y 2 + axn y + aλxn e−λx ) = 0.
∂x ∂y
Principal integral:
Z
E λx a n+1
Ξ= + e E dx, E = exp x − 2λx .
y + λe−λx n+1
∂w ∂w
27. + (λeλx y 2 + axn eλx y − axn e2λx ) = 0.
∂x ∂y
Principal integral:
Z Z
e2λx E λx n −λx
Ξ= +λ e E dx, E = exp a x e dx .
y − eλx
∂w ∂w
28. + (aeλx y 2 − abxn eλx y + bnxn−1 ) = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= +a eλx E dx, E = exp ab xn eλx dx .
y − bxn
∂w ∂w
29. + (axn y 2 + bλeλx − ab2 xn e2λx ) = 0.
∂x ∂y
Principal integral:
Z Z
E n n λx
Ξ= +a x E dx, E = exp 2ab x e dx .
y − beλx
∂w ∂w
30. + (axn y 2 + λy − ab2 xn e2λx ) = 0.
∂x ∂y
Principal integral:
Z Z
E n n λx
Ξ= +a x E dx, E = exp λx + 2ab x e dx .
y − beλx
∂w ∂w
31. + (axn y 2 − abxn eλx y + bλeλx ) = 0.
∂x ∂y
Principal integral:
Z Z
E n n λx
Ξ= +a x E dx, E = exp ab x e dx .
y − beλx
∂w ∂w
32. + axn y 2 − axn (beλx + c)y + bλeλx = 0.
∂x ∂y
E
Ξ= + a xn E dx,
y − beλx − c
 Z

 ac n+1 n λx
exp x + ab x e dx if n 6= −1,
n
+ 1Z
where E =

 eλx
xac exp ab dx if n = −1.
x
∂x
+ g(x, y) ∂w
∂y
=0 27
∂w ∂w
33. + axn e2λx y 2 + (bxn eλx − λ)y + cxn = 0.
∂x ∂y
Principal integral:
Z Z
dv
Ξ= 2
− xn eλx dx, v = eλx y.
av + bv + c
∂w ∂w
34. + aeλx (y − bxn − c)2 + bnxn−1 = 0.
∂x ∂y
1 a
Principal integral: Ξ = n
+ eλx .
y − bx − c λ
∂w 2 2 ∂w
35. + y 2 + 2aλxeλx − a2 e2λx = 0.
∂x ∂y
Principal integral:
Z Z
E λx2
Ξ= 2 + E dx, E = exp 2a e dx .
y − aeλx
∂w 2 ∂w
36. + ae−λx y 2 + λxy + ab2 = 0.
∂x ∂y
Z
1 2

1
Principal integral: Ξ = arctan y exp − 2 λx − ab exp − 12 λx2 dx.
b
∂w 2 ∂w
37. + axn y 2 + λxy + ab2 xn eλx = 0.
∂x ∂y
Z
y 2

Principal integral: Ξ = arctan 1
exp − 2 λx − ab xn exp 1
2 λx
2
dx.
b
∂w ∂w
38. + (ae2λx y 3 + beλx y 2 + cy + de−λx ) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − x, v = eλx y.
av 3 + bv 2 + (c + λ)v + d
∂w ∂w
39. + (aeλx y 3 + 3abeλx y 2 + cy − 2ab3 eλx + bc) = 0.
∂x ∂y
Principal integral:
Z
e2cx E (λ+2c)x 6ab2 λx
Ξ= + 2a e E dx, E = exp − e .
(y + b)2 λ
∂w ∂w
40. x + (aeλx y 2 + ky + ab2 x2k eλx ) = 0.
∂x ∂y
Z
y
Principal integral: Ξ = arctan k − ab xk−1 eλx dx.
bx
∂w ∂w
41. x + ax2n eλx y 2 + (bxn eλx − n)y + ceλx = 0.
∂x ∂y
Principal integral:
Z Z
dv
Ξ= 2
− xn−1 eλx dx, v = xn y.
av + bv + c
∂w ∂w
42. y + eλx (2aλx + a + b)y − eλx (a2 λx2 + abx − c) = 0.
∂x ∂y
Principal integral:
Z
vE dv
Ξ = xE + , v = e−λx y − ax,
λv 2 − bv − c
Z
dv
where E = exp a .
λv 2 − bv − c
∂w ∂w
43. aeλx + by m = 0.
∂x ∂y
1 1 −λx
1◦ . Principal integral for m 6= 1: Ξ = y 1−m + e .
b(1 − m) λa
1 1 −λx
2◦ . Principal integral for m = 1: Ξ = ln y + e .
b λa
∂w ∂w
44. (aey + bx) + = 0.
∂x ∂y
a (1−b)y
1◦ . Principal integral for b 6= 1: Ξ = xe−by − e .
1−b
2◦ . Principal integral for b = 1: Ξ = xe−y − ay.
∂w ∂w
45. (axn eλy + bxy m ) + eµy = 0.
∂x ∂y
Principal integral:
Z Z
1
Ξ= x1−n E − a e (λ−µ)y
E dy, E = exp b(n − 1) y e m −µy
dy .
1−n
∂w ∂w
46. (axn y m + bxeλy ) + yk = 0.
∂x ∂y
Principal integral:
Z Z
1
Ξ= xn−1 E + a y m−k
E dy, E = exp b(n − 1) y −k λy
e dy .
n−1
∂w ∂w
47. (axn y m + bxy k ) + eλy = 0.
∂x ∂y
Principal integral:
Z Z
1
Ξ= x1−n E − a y e m −λy
E dy, E = exp b(n − 1) y ek −λy
dy .
1−n
∂x
+ g(x, y) ∂w
∂y
=0 29
1.1.3 Equations Containing Hyperbolic Functions

◮ Coefficients of equations contain hyperbolic sine.
∂w ∂w
1. + a sinh(λx) = 0.
∂x ∂y
Principal integral: Ξ = λy − a cosh(λx).
∂w ∂w
2. + a sinh(µy) = 0.
∂x ∂y

Principal integral: Ξ = aµx − lntanh( 12 µy).
∂w ∂w
3. + y 2 − a2 + aλ sinh(λx) − a2 sinh2 (λx) = 0.
∂x ∂y
Principal integral:
Z
E 2a
Ξ= + E dx, E = exp sinh(λx) .
y − a cosh(λx) λ
∂w ∂w
4. + λ sinh(λx)y 2 − sinh3 (λx) = 0.
∂x ∂y
Principal integral:
Z
exp 12 cosh(2λx)
Ξ= + λ sinh(λx) exp 12 cosh(2λx) dx.
y − cosh(λx)
∂w ∂w
5. + [a sinh2 (λx) − λ]y 2 − a sinh2 (λx) + λ − a = 0.
∂x ∂y
Principal integral:
Z
E λ
Ξ= + a− E dx,
sinh(λx) sinh(λx)y − cosh(λx) sinh2 (λx)

a
E = exp cosh(2λx) .
2λ
∂w ∂w
6. sinh(λx) + a sinh(µy) = 0.
∂x ∂y

Principal integral: Ξ = aµ lntanh( 12 λx) − λ lntanh( 12 µy).
∂w ∂w
7. sinh(µy) + a sinh(λx) = 0.
∂x ∂y
Principal integral: Ξ = λ cosh(µy) − aµ cosh(λx).
◮ Coefficients of equations contain hyperbolic cosine.

∂w ∂w
8. + a cosh(λx) = 0.
∂x ∂y
Principal integral: Ξ = a sinh(λx) − λy.
∂w ∂w
9. + a cosh(λy) = 0.
∂x ∂y
Principal integral: Ξ = aλx − 2 arctan(eλy ).

∂w ∂w
10. + [a cosh2 (λx) − λ]y 2 − a cosh2 (λx) + λ + a = 0.
∂x ∂y
Principal integral:
Z
E λ
Ξ= + a− E dx,
cosh(λx) cosh(λx)y − sinh(λx) cosh2 (λx)

a
E = exp cosh(2λx) .
2λ
∂w ∂w
11. 2 + [a − λ + a cosh(λx)]y 2 + a + λ − a cosh(λx) = 0.
∂x ∂y
Principal integral:
Z
E 1
Ξ= 1
+ a − λ + a cosh(λx) E dx,
y − tanh 2 λx
2
where Z
2(a−λ)
E = cosh( 12 λx) λ exp a cosh(λx) tanh 1
2 λx dx .
∂w ∂w
12. (axn + bx coshm y) + yk = 0.
∂x ∂y
Principal integral:
Z Z
1−n −k −k m
Ξ=x E + (n − 1)a y E dy, E = exp b(n − 1) y cosh y dy .
∂w ∂w
13. (axn + bx coshm y) + coshk (λy) = 0.
∂x ∂y
Principal integral:
Z Z
1−n E dy coshm y dy
Ξ=x E + (n − 1)a , E = exp b(n − 1) .
coshk (λy) coshk (λy)
∂w ∂w
14. (axn y m + bx) + coshk (λy) = 0.
∂x ∂y
Principal integral:
Z Z
y m E dy dy
Ξ = x1−n E + (n − 1)a , E = exp b(n − 1) .
coshk (λy) coshk (λy)
∂w ∂w
15. cosh(µy) + a cosh(λx) = 0.
∂x ∂y
Principal integral: Ξ = µa sinh(λx) − λ sinh(µy).
∂x
+ g(x, y) ∂w
∂y
=0 31
◮ Coefficients of equations contain hyperbolic tangent.

∂w ∂w
16. + a tanh(λx) = 0.
∂x ∂y

Principal integral: Ξ = λy − a ln cosh(λx) .
∂w ∂w
17. + a tanh(λy) = 0.
∂x ∂y

Principal integral: Ξ = aλx − lnsinh(λy).
∂w ∂w
18. + y 2 + aλ − a(a + λ) tanh2 (λx) = 0.
∂x ∂y
2a/λ Z
cosh(λx) 2a/λ
Principal integral: Ξ = + cosh(λx) dx.
y − a tanh(λx)
∂w ∂w
19. + y 2 + 3aλ − λ2 − a(a + λ) tanh2 (λx) = 0.
∂x ∂y
Principal integral:
2a/λ Z 2a/λ
cosh(λx) cosh(λx)
Ξ= + dx.
sinh2 (λx) y − a tanh(λx) + λ coth(λx) sinh2 (λx)
∂w ∂w
20. (axn + bx tanhm y) + yk = 0.
∂x ∂y
Principal integral:
Z Z
1−n −k −k m
Ξ=x E + (n − 1)a y E dy, E = exp b(n − 1) y tanh y dy .
∂w ∂w
21. (axn + bx tanhm y) + tanhk (λy) = 0.
∂x ∂y
Principal integral:
Z Z
1−n E dy tanhm y dy
Ξ=x E + (n − 1)a , E = exp b(n − 1) .
tanhk (λy) tanhk (λy)
∂w ∂w
22. (axn y m + bx) + tanhk (λy) = 0.
∂x ∂y
Principal integral:
Z Z
1−n y m E dy dy
Ξ=x E + (n − 1)a , E = exp b(n − 1) .
tanhk (λy) tanhk (λy)
∂w ∂w
23. (axn tanhm y + bx) + yk = 0.
∂x ∂y
1◦ . Principal integral for k 6= 1:
Z
1−n b(n − 1) 1−k
Ξ=x E + (n − 1)a y −k E tanhm y dy, E = exp y .
1−k
2◦ . Principal integral for k = 1:
Z
−b 1−n
Ξ = (xy ) + (n − 1)a y (n−1)b−1 tanhm y dy.
◮ Coefficients of equations contain hyperbolic cotangent.

∂w ∂w
24. + a coth(λx) = 0.
∂x ∂y

Principal integral: Ξ = λy − a lnsinh(λx).
∂w ∂w
25. + a coth(λy) = 0.
∂x ∂y

Principal integral: Ξ = aλx − ln cosh(λy) .
∂w ∂w
26. + y 2 + aλ − a(a + λ) coth2 (λx) = 0.
∂x ∂y
2a/λ Z
sinh(λx) 2a/λ
Principal integral: Ξ = + sinh(λx) dx.
y − a coth(λx)
∂w ∂w
27. + y 2 + 3aλ − λ2 − a(a + λ) coth2 (λx) = 0.
∂x ∂y
Principal integral:
2a/λ Z 2a/λ
sinh(λx) sinh(λx)
Ξ= + dx.
cosh2 (λx) y − a coth(λx) + λ tanh(λx) cosh2 (λx)
◮ Coefficients of equations contain different hyperbolic functions.

∂w ∂w
28. + a sinh(λx) cosh(µy) = 0.
∂x ∂y

Principal integral: Ξ = 2λ arctan eµy − aµ cosh(λx).
∂w ∂w
29. + a cosh(λx) sinh(µy) = 0.
∂x ∂y

Principal integral: Ξ = λ ln tanh 12 µy − aµ sinh(λx).
∂w ∂w
30. + y 2 − 2λ2 tanh2 (λx) − 2λ2 coth2 (λx) = 0.
∂x ∂y
Principal integral:
Z
sinh2 (λx) cosh2 (λx)
Ξ= + sinh2 (λx) cosh2 (λx) dx.
y − λ tanh(λx) − λ coth(λx)
∂w
31. + y 2 + λ(a + b) − 2ab − a(a + λ) tanh2 (λx)
∂x
∂w
− b(b + λ) coth2 (λx) = 0.
∂y
Principal integral:
2b 2a Z
sinh(λx) λ cosh(λx) λ 2b 2a
Ξ= + sinh(λx) λ cosh(λx) λ dx.
y − a tanh(λx) − b coth(λx)
∂x
+ g(x, y) ∂w
∂y
=0 33
∂w ∂w
32. sinh(λy) + a cosh(βx) = 0.
∂x ∂y
Principal integral: Ξ = β cosh(λy) − aλ sinh(βx).
∂w ∂w
33. axn coshm (λy) + bx + sinhk (βy) = 0.
∂x ∂y
Principal integral:
Z Z
1−n coshm (λy)E dy dy
Ξ=x E + (n − 1)a , E = exp b(n − 1) .
sinhk (βy) sinhk (βy)
1.1.4 Equations Containing Logarithmic Functions

◮ Coefficients of equations contain logarithmic functions.
∂w ∂w
1. + a lnk (λx) + b = 0.
∂x ∂y
Z
Principal integral: Ξ = y − bx − a lnk (λx) dx.
∂w ∂w
2. + a lnk (λy) + b = 0.
∂x ∂y
Z
dy
Principal integral: Ξ = x − k
.
a ln (λy) + b
∂w ∂w
3. + a lnk (λx) lnn (µy) = 0.
∂x ∂y
Z Z
k dy
Principal integral: Ξ = a ln (λx) dx − n .
ln (µy)
∂w ∂w
4. + a lnk (x + λy) = 0.
∂x ∂y
dz
Ξ=x− , z = x + λy.
1 + aλ lnk z
◮ Coefficients of equations contain logarithmic and power-law functions.
∂w ∂w
5. + axn lnk (λy) = 0.
∂x ∂y
Z
a dy
Principal integral: Ξ = xn+1 − k
.
n+1 ln (λy)
∂w ∂w
6. + ay n lnk (λx) = 0.
∂x ∂y
Z
1
Principal integral: Ξ = y 1−n − a lnk (λx) dx.
1−n
∂w ∂w
7. + y 2 + a ln(βx)y − ab ln(βx) − b2 = 0.
∂x ∂y
Principal integral:
Z
e(2b−a)x E
Ξ= + e(2b−a)x E dx, E = exp ax ln(βx) .
y−b
∂w ∂w
8. + y 2 + ax lnm (bx)y + a lnm (bx) = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= + x−2 E dx, E = exp a x lnm (bx) dx .
x(xy + 1)
∂w ∂w
9. + (axn y 2 − abxn+1 y ln x + b ln x + b) = 0.
∂x ∂y
Principal integral:
Z
E n ab n+2 1
Ξ= +a x E dx, E = exp x ln x − .
y − bx ln x n+2 n+2
∂w ∂w
10. − (n + 1)xn y 2 − axn+1 (ln x)m y + a(ln x)m = 0.
∂x ∂y
Principal integral:
Z Z
x−2(n+1) E
Ξ= − (n + 1) x−n−2 E dx, E = exp a xn+1 (ln x)m dx .
y − x−n−1
∂w ∂w
11. + a(ln x)n y 2 + bmxm−1 − ab2 x2m (ln x)n = 0.
∂x ∂y
Principal integral:
Z Z
E n m n
Ξ= +a (ln x) E dx, E = exp 2ab x (ln x) dx .
y − bxm
∂w ∂w
12. + a(ln x)n y 2 − abx(ln x)n+1 y + b ln x + b = 0.
∂x ∂y
Principal integral:
Z Z
E n n+1
Ξ= +a (ln x) E dx, E = exp ab x(ln x) dx .
y − bx ln x
∂w ∂w
13. + a(ln x)k (y − bxn − c)2 + bnxn−1 = 0.
∂x ∂y
Z
1
Principal integral: Ξ = +a (ln x)k dx.
y − bxn − c
∂x
+ g(x, y) ∂w
∂y
=0 35
∂w ∂w
14. + a(ln x)n y 2 + b(ln x)m y + bc(ln x)m − ac2 (ln x)n = 0.
∂x ∂y
Principal integral:
Z Z
E n m n
Ξ= +a (ln x) E dx, E = exp [b(ln x) − 2ac(ln x) ] dx .
y+c
∂w ∂w
15. x + (ay + b ln x)2 = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ = ln x − , v = ay + b ln x.
av 2 + b
∂w ∂w
16. x + xy 2 − A2 x ln2 (βx) + A = 0.
∂x ∂y
Principal integral:
Z
e−2Ax E
Ξ= + e−2Ax E dx, E = exp 2Ax ln(βx) .
y − A ln(βx)
∂w ∂w
17. x + xy 2 − A2 x ln2k (βx) + kA lnk−1 (βx) = 0.
∂x ∂y
Principal integral:
Z Z
E k
Ξ= + E dx, E = exp 2A ln (βx) dx .
y − A lnk (βx)
∂w ∂w
18. x + (axn y 2 + b − ab2 xn ln2 x) = 0.
∂x ∂y
Principal integral:
Z
E n−1 2abxn
Ξ= + ax E dx, E = exp (n ln x − 1) .
y − b ln x n2
∂w ∂w
19. x + a lnm (λx)y 2 + ky + ab2 x2k lnm (λx) = 0.
∂x ∂y
Z
y
Principal integral: Ξ = arctan − ab xk−1 lnm (λx) dx.
bxk
∂w ∂w
20. x + axn (y + b ln x)2 − b = 0.
∂x ∂y
1 a
Principal integral: Ξ = + xn .
y + b ln x n
∂w ∂w
21. x + ax2n (ln x)y 2 + (bxn ln x − n)y + c ln x = 0.
∂x ∂y
Principal integral:
Z Z
dv
Ξ= 2
− xn−1 ln x dx, v = xn y.
av + bv + c
∂w ∂w
22. xk + (ay n lnm x + by lns x) = 0.
∂x ∂y
Principal integral:
Z Z
1−n −k m −k s
Ξ=y E + (n − 1)a x E ln x dx, E = exp b(n − 1) x ln x dx .
∂w ∂w
23. (a ln x + b) + y 2 + c(ln x)n y − λ2 + λc(ln x)n = 0.
∂x ∂y
Principal integral:
Z Z
E E dx c(ln x)n − 2λ
Ξ= + , E = exp dx.
y+λ a ln x + b a ln x + b
∂w ∂w
24. (a ln x + b) + (ln x)n y 2 + cy − λ2 (ln x)n + cλ = 0.
∂x ∂y
Principal integral:
Z Z
E (ln x)n E dx c − 2λ(ln x)n
Ξ= + , E = exp dx .
y+λ a ln x + b a ln x + b

∂w ∂w
25. x2 ln(ax) − x2 y 2 ln(ax) + 1 = 0.
∂x ∂y
Z
x dx
Principal integral: Ξ = − 2 .
ln(ax) xy ln(ax) − 1 ln (ax)
∂w ∂w
26. lnk (λx) + (ay n + by lnm x) = 0.
∂x ∂y
Principal integral:
Z Z
1−n E dx lnm x dx
Ξ=y E + (n − 1)a , E = exp b(n − 1) .
lnk (λx) lnk (λx)
∂w ∂w
27. lnk (λx) + (ay n lnm x + by) = 0.
∂x ∂y
Principal integral:
Z Z
1−n E lnm x dx dx
Ξ=y E + (n − 1)a , E = exp b(n − 1) .
lnk (λx) lnk (λx)
1.1.5 Equations Containing Trigonometric Functions

◮ Coefficients of equations contain sine.
∂w ∂w
1. + a sink (λx) + b = 0.
∂x ∂y
Z
Principal integral: Ξ = y − bx − a sink (λx) dx.
∂x
+ g(x, y) ∂w
∂y
=0 37
∂w ∂w
2. + a sink (λy) + b = 0.
∂x ∂y
Z
dy
.
a sin (λy) + b
∂w ∂w
3. + a sink (λx) sinn (µy) = 0.
∂x ∂y
Z Z
dy
Principal integral: Ξ = a sink (λx) dx − .
sinn (µy)
∂w ∂w
4. + a sink (x + λy) = 0.
∂x ∂y
Principal integral:
Z
dz
Ξ=x− , z = x + λy.
1 + aλ sink z
∂w ∂w
5. + y 2 − a2 + aλ sin(λx) + a2 sin2 (λx) = 0.
∂x ∂y
Principal integral:
Z
E 2a
Ξ= +
E dx, E = exp − sin(λx) .
y + a cos(λx) λ
∂w ∂w
6. + y 2 + a sin(βx)y + ab sin(βx) − b2 = 0.
∂x ∂y
Principal integral:
Z
E a
Ξ= + E dx, E = exp −2bx − cos(βx) .
y+b β
∂w ∂w
7. + y 2 + ax sinm (bx)y + a sinm (bx) = 0.
∂x ∂y
Principal integral:
Z Z
E −2 m
Ξ= + x E dx, E = exp a x sin (bx) dx .
x(xy + 1)
∂w ∂w
8. + λ sin(λx)y 2 + λ sin3 (λx) = 0.
∂x ∂y
Principal integral:
Z
E
Ξ= +λ E sin(λx) dx, E = exp 12 cos(2λx) .
y + cos(λx)
∂w ∂w
9. 2 + [λ + a − a sin(λx)]y 2 + λ − a − a sin(λx) = 0.
∂x ∂y
Principal integral:
Z
E 1
Ξ= 1 1
+ λ + a − a sin(λx) E dx,
y − tan 2 λx + 4 π 2

1 a
E= exp sin(λx) .
1 − sin(λx) λ
∂w ∂w
10. + [λ + a sin2 (λx)]y 2 + λ − a + a sin2 (λx) = 0.
∂x ∂y
Principal integral:
Z
E 2
1 a
Ξ= + λ + a sin (λx) E dx, E= exp cos(2λx) .
y + cot(λx) sin2 (λx) 2λ
∂w ∂w
11. − (k + 1)xk y 2 − axk+1 (sin x)m y + a(sin x)m = 0.
∂x ∂y
Principal integral:
Z Z
E E dx k+1 m
Ξ= − (k + 1) , E = exp a x (sin x) dx .
xk+1 (xk+1 y − 1) xk+2
∂w ∂w
12. + a sink (λx + µ)(y − bxn − c)2 + y − bxn + bnxn−1 − c = 0.
∂x ∂y
Z
ex
Principal integral: Ξ = + a ex sink (λx + µ) dx.
y − bxn − c
∂w ∂w
13. x + a sinm (λx)y 2 + ky + ab2 x2k sinm (λx) = 0.
∂x ∂y
Z
y
Principal integral: Ξ = arctan − ab xk−1 sinm (λx) dx.
bxk
∂w ∂w
14. a sin(λx) + b + y 2 + c sin(µx)y − k2 + ck sin(µx) = 0.
∂x ∂y
Principal integral:
Z Z
E E dx c sin(µx) − 2k
Ξ= + , E = exp dx .
y+k a sin(λx) + b a sin(λx) + b
◮ Coefficients of equations contain cosine.
∂w ∂w
15. + a cosk (λx) + b = 0.
∂x ∂y
Z
Principal integral: Ξ = y − bx − a cosk (λx) dx.
∂w ∂w
16. + a cosk (λy) + b = 0.
∂x ∂y
Z
dy
Principal integral: Ξ = x − .
a cosk (λy) + b
∂w ∂w
17. + a cosk (λx) cosn (µy) = 0.
∂x ∂y
Z Z
k dy
Principal integral: Ξ = a cos (λx) dx − .
cosn (µy)
∂x
+ g(x, y) ∂w
∂y
=0 39
∂w ∂w
18. + a cosk (x + λy) = 0.
∂x ∂y
Principal integral:
Z
dz
Ξ=x− , z = x + λy.
1 + aλ cosk z
∂w ∂w
19. + y 2 − a2 + aλ cos(λx) + a2 cos2 (λx) = 0.
∂x ∂y
Principal integral:
Z
E 2a
Ξ= + E dx, E = exp − cos(λx) .
y − a sin(λx) λ
∂w ∂w
20. + λ cos(λx)y 2 + λ cos3 (λx) = 0.
∂x ∂y
Principal integral:
Z
E
Ξ= +λ E cos(λx) dx, E = exp − 12 cos(2λx) .
y − sin(λx)
∂w ∂w
21. 2 + [λ + a + a cos(λx)]y 2 + λ − a + a cos(λx) = 0.
∂x ∂y
Principal integral:
Z
E 1
Ξ= 1 + λ + a + a cos(λx) E dx,
y − tan( 2 λx) 2

1 a
E= exp − cos(λx) .
1 + cos(λx) λ
∂w ∂w
22. + [λ + a cos2 (λx)]y 2 + λ − a + a cos2 (λx) = 0.
∂x ∂y
Principal integral:
Z
E 1 a
Ξ= + λ + a cos2 (λx) E dx, E= exp − cos(2λx) .
y − tan(λx) cos2 (λx) 2λ
∂w ∂w
23. (axn y m + bx) + cosk (λy) = 0.
∂x ∂y
Principal integral:
Z Z
1−n y m E dy dy
Ξ=x E + a(n − 1) , E = exp b(n − 1) .
cosk (λy) cosk (λy)
∂w ∂w
24. (axn + bx cosm y) + yk = 0.
∂x ∂y
Principal integral:
Z Z
1−n −k cosm y dy
Ξ=x E + a(n − 1) y E dy, E = exp b(n − 1) .
yk
∂w ∂w
25. (axn + bx cosm y) + cosk (λy) = 0.
∂x ∂y
Principal integral:
Z Z
1−n E dy cosm y dy
Ξ=x E + a(n − 1) , E = exp b(n − 1) .
∂w ∂w
26. (axn cosm y + bx) + cosk (λy) = 0.
∂x ∂y
Principal integral:
Z Z
1−n cosm y E dy dy
Ξ=x E + a(n − 1) , E = exp b(n − 1) .
◮ Coefficients of equations contain tangent.

∂w ∂w
27. + a tank (λx) + b = 0.
∂x ∂y
Z
Principal integral: Ξ = y − bx − a tank (λx) dx.
∂w ∂w
28. + a tank (λy) + b = 0.
∂x ∂y
Z
dy
.
a tan (λy) + b
∂w ∂w
29. + a tank (λx) tann (µy) = 0.
∂x ∂y
Z Z
Principal integral: Ξ = a tan (λx) dx − cotn (µy) dy.
k
∂w ∂w
30. + y 2 + aλ + a(λ − a) tan2 (λx) = 0.
∂x ∂y
−2a/λ Z
cos(λx) −2a/λ
Principal integral: Ξ = + cos(λx) dx.
y − a tan(λx)
∂w ∂w
31. + y 2 + λ2 + 3aλ + a(λ − a) tan2 (λx) = 0.
∂x ∂y
Principal integral:
−2a/λ Z −2a/λ
cos(λx) cos(λx)
Ξ= + dx.
sin2 (λx) y − a tan(λx) + λ cot(λx) sin2 (λx)
∂w ∂w
32. + y 2 + ax tank (bx)y + a tank (bx) = 0.
∂x ∂y
Principal integral:
Z Z
E −2 k
Ξ= + x E dx, E = exp a x tan (bx) dx .
x(xy + 1)
∂x
+ g(x, y) ∂w
∂y
=0 41
∂w ∂w
33. − (k + 1)xk y 2 − axk+1 (tan x)m y + a(tan x)m = 0.
∂x ∂y
Principal integral:
Z Z
E E dx k+1 m
Ξ= − (k + 1) , E = exp a x (tan x) dx .
xk+1 (xk+1 y − 1) xk+2
∂w ∂w
34. + a tann (λx)y 2 − ab2 tann+2 (λx) + bλ tan2 (λx) + bλ = 0.
∂x ∂y
Principal integral:
Z Z
E n n+1
Ξ= +a E tan (λx) dx, E = exp 2ab tan (λx) dx .
y − b tan(λx)
∂w ∂w
35. + a tank (λx + µ)(y − bxn − c)2 + y − bxn + bnxn−1 − c = 0.
∂x ∂y
Z
ex
Principal integral: Ξ = + a ex tank (λx + µ) dx.
y − bxn − c
∂w ∂w
36. x + a tanm (λx)y 2 + ky + ab2 x2k tanm (λx) = 0.
∂x ∂y
Z
1 −k
Principal integral: Ξ = arctan x y − ab xk−1 tanm (λx) dx.
b
∂w ∂w
37. a tan(λx) + b + y 2 + c tan(µx)y − k2 + ck tan(µx) = 0.
∂x ∂y
Principal integral:
Z Z
E E dx c tan(µx) − 2k
Ξ= + , E = exp dx .
y+k a tan(λx) + b a tan(λx) + b
∂w ∂w
38. (axn y m + bx) + tank (λy) = 0.
∂x ∂y
Principal integral:
Z Z
y m E dy dy
Ξ = x1−n E + (n − 1)a , E = exp b(n − 1) .
tank (λy) tank (λy)
∂w ∂w
39. (axn + bx tanm y) + yk = 0.
∂x ∂y
Principal integral:
Z Z
1−n −k −k m
Ξ=x E + (n − 1)a y E dy, E = exp b(n − 1) y tan y dy .
∂w ∂w
40. (axn + bx tanm y) + tank (λy) = 0.
∂x ∂y
Principal integral:
Z Z
1−n E dy tanm y dy
Ξ=x E + (n − 1)a , E = exp b(n − 1) .
∂w ∂w
41. (axn tanm y + bx) + tank (λy) = 0.
∂x ∂y
Principal integral:
Z Z
1−n tanm yE dy dy
Ξ=x E + (n − 1)a , E = exp b(n − 1) .
◮ Coefficients of equations contain cotangent.
∂w ∂w
42. + a cotk (λx) + b = 0.
∂x ∂y
Z
Principal integral: Ξ = y − bx − a cotk (λx) dx.
∂w ∂w
43. + a cotk (λy) + b = 0.
∂x ∂y
Z
dy
.
a cot (λy) + b
∂w ∂w
44. + a cotk (x + λy) = 0.
∂x ∂y
Principal integral:
Z
dz
Ξ=x− , z = x + λy.
1 + aλ cotk z
∂w ∂w
45. + y 2 + aλ + a(λ − a) cot2 (λx) = 0.
∂x ∂y
−2a/λ Z
sin(λx) −2a/λ
Principal integral: Ξ = + sin(λx) dx.
y + a cot(λx)
∂w ∂w
46. + y 2 + λ2 + 3aλ + a(λ − a) cot2 (λx) = 0.
∂x ∂y
Principal integral:
−2a/λ Z −2a/λ
sin(λx) sin(λx)
Ξ= + dx.
cos2 (λx) y − λ tan(λx) + a cot(λx) cos2 (λx)
∂w ∂w
47. + y 2 − 2a cot(ax)y + b2 − a2 = 0.
∂x ∂y
−2
sin (bx) 1
Principal integral: Ξ = − cot(bx).
y − a cot(ax) + b cot(bx) b
∂w ∂w
48. cot(λx) + a cot(µy) = 0.
∂x ∂y

Principal integral: Ξ = aµ lncos(λx) − λ lncos(µy).
∂x
+ g(x, y) ∂w
∂y
=0 43
∂w ∂w
49. cot(µy) + a cot(λx) = 0.
∂x ∂y

Principal integral: Ξ = aµ lnsin(λx) − λ lnsin(µy).
∂w ∂w
50. cot(µy) + a cot2 (λx) = 0.
∂x ∂y

Principal integral: Ξ = λ lnsin(µy) + aµ cot(λx) + aλµx.
∂w ∂w
51. cot(y + a) + c cot(x + b) = 0.
∂x ∂y

Principal integral: Ξ = c lnsin(x + b) − lnsin(y + a).
∂w ∂w
52. cot(λx) cot(µy) +a = 0.
∂x ∂y

Principal integral: Ξ = λ lnsin(µy) + aµ lncos(λx).
∂w ∂w
53. cot(λx) cot(µy) + a cot(νx) = 0.
∂x ∂y
Z
cot(νx)
Principal integral: Ξ = aµ dx − lnsin(µy).
cot(λx)
◮ Coefficients of equations contain different trigonometric functions.
∂w ∂w
54. + a sink (λx) cosn (µy) = 0.
∂x ∂y
Z Z
k dy
Principal integral: Ξ = a sin (λx) dx − . In the special case a = 1, k = 1,
cosn (µy)
and n = −1 we have Ξ = µ cos(λx) + λ sin(µy).
∂w ∂w
55. + y 2 − y tan x + a(1 − a) cot2 x = 0.
∂x ∂y
1◦ . Principal integral for a 6= 12 :
(sin x)−2a cos x 1

Ξ= + (sin x)1−2a .
y + a cot x 1 − 2a
2◦ . Principal integral for a = 12 :

cos x
Ξ= + ln |sin x|.
y sin x + 12 cos x
∂w ∂w
56. + (y 2 − my tan x + b2 cos2m x) = 0.
∂x ∂y
Z
1 −m
Principal integral: Ξ = arctan y cos x − b cosm x dx.
b
∂w ∂w
57. + (y 2 + my cot x + b2 sinm x) = 0.
∂x ∂y
Z
1
Principal integral: Ξ = arctan y sin−m x − b sinm x dx.
b
∂w ∂w
58. + y 2 − 2λ2 tan2 (λx) − 2λ2 cot2 (λx) = 0.
∂x ∂y
Principal integral:
sin2 (λx) cos2 (λx) 1 1
Ξ= + x− sin(λx) cos(λx) cos(2λx).
y − λ cot(λx) + λ tan(λx) 8 8λ
∂w 2 ∂w
59. + y +λ(a+b)+2ab+a(λ−a) tan2 (λx)+b(λ−b) cot2 (λx) = 0.
∂x ∂y
Principal integral:
Z
E − 2a 2b
Ξ= + E dx, E = cos(λx) λ sin(λx) − λ .
y − a tan(λx) + b cot(λx)
∂w ∂w
60. + λ sin(λx)y 2 + a cosn (λx)y − a cosn−1 (λx) = 0.
∂x ∂y
Principal integral:
Z Z
E E sin(λx) n
Ξ= +λ dx, E = exp a cos (λx) dx .
cos(λx) y cos(λx) − 1 cos2 (λx)
∂w ∂w
61. + λ sin(λx)y 2 + a sin(λx)y − a tan(λx) = 0.
∂x ∂y
Principal integral:
Z
E E sin(λx) a
Ξ= +λ dx, E = exp − cos(λx) .
cos(λx) y cos(λx) − 1 cos2 (λx) λ
∂w ∂w
62. + λ sin(λx)y 2 + axn cos(λx)y − axn = 0.
∂x ∂y
Principal integral:
Z Z
E E sin(λx) n
Ξ= +λ dx, E = exp a x cos(λx) dx .
cos(λx) y cos(λx) − 1 cos2 (λx)
∂w ∂w
63. + Aeλx cos(ay) + Beµx sin(ay) + Aeλx = 0.
∂x ∂y
Z
ay aB µx aB µx
Principal integral: Ξ = tan exp − e − aA exp λx − e dx.
2 µ µ
∂w ∂w
64. sinn+1 (2x) + (ay 2 sin2n x + b cos2n x) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − 2−n−1 ln tan x, v = y tann x.
av 2 + n2n+1 v + b
∂x
+ g(x, y) ∂w
∂y
=0 45
1.1.6 Equations Containing Inverse Trigonometric Functions

◮ Coefficients of equations contain arcsine.
∂w ∂w
1. + a arcsink (λx) + b = 0.
∂x ∂y
Z
Principal integral: Ξ = y − bx − a arcsink (λx) dx.
∂w ∂w
2. + a arcsink (λy) + b = 0.
∂x ∂y
Z
dy
a arcsink (λy) + b
∂w ∂w
3. + k arcsinn (ax + by + c) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − x, v = ax + by + c.
a + bk arcsinn v
∂w ∂w
4. + a arcsink (λx) arcsinn (µy) = 0.
∂x ∂y
Z Z
k dy
Principal integral: Ξ = a arcsin (λx) dx − .
arcsinn (µy)
∂w ∂w
5. + y 2 + λ(arcsin x)n y − a2 + aλ(arcsin x)n = 0.
∂x ∂y
Principal integral:
Z Z
e−2ax E
Ξ= + e−2ax E dx, E = exp λ (arcsin x)n dx .
y+a
∂w ∂w
6. + y 2 + λx(arcsin x)n y + λ(arcsin x)n = 0.
∂x ∂y
Principal integral:
Z Z
E −2 n
Ξ= + x E dx, E = exp λ x(arcsin x) dx .
x(xy + 1)
∂w ∂w
7. − (k + 1)xk y 2 − λ(arcsin x)n (xk+1 y − 1) = 0.
∂x ∂y
Principal integral:
Z Z
E −k−2 k+1 n
Ξ= − (k + 1) x E dx, E = exp λ x (arcsin x) dx .
xk+1 (xk+1 y − 1)
∂w ∂w
8. + λ(arcsin x)n y 2 + ay + ab − b2 λ(arcsin x)n = 0.
∂x ∂y
Principal integral:
Z Z
eax E ax n n
Ξ= +λ e (arcsin x) E dx, E = exp −2bλ (arcsin x) dx .
y+b
∂w ∂w
9. + λ(arcsin x)n y 2 − bλxm (arcsin x)n y + bmxm−1 = 0.
∂x ∂y
Principal integral:
Z Z
E n m n
Ξ= +λ (arcsin x) E dx, E = exp bλ x (arcsin x) dx .
y − bxm
∂w ∂w
10. + λ(arcsin x)n y 2 + bmxm−1 − λb2 x2m (arcsin x)n = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= +λ (arcsin x)n E dx, E = exp 2bλ xm (arcsin x)n dx .
y − bxm
∂w ∂w
11. + λ(arcsin x)n (y − axm − b)2 + amxm−1 = 0.
∂x ∂y
Z
1
Principal integral: Ξ = + λ (arcsin x)n dx.
y − axm − b
∂w ∂w
12. x + λ(arcsin x)n y 2 + ky + λb2 x2k (arcsin x)n = 0.
∂x ∂y
Z
y
Principal integral: Ξ = arctan − λb xk−1 (arcsin x)n dx.
bxk
◮ Coefficients of equations contain arccosine.

∂w ∂w
13. + a arccosk (λx) + b = 0.
∂x ∂y
Z
Principal integral: Ξ = y − bx − a arccosk (λx) dx.
∂w ∂w
14. + a arccosk (λy) + b = 0.
∂x ∂y
Z
dy
a arccosk (λy) + b
∂w ∂w
15. + k arccosn (ax + by + c) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − x, v = ax + by + c.
a + bk arccosn v
∂x
+ g(x, y) ∂w
∂y
=0 47
∂w ∂w
16. + a arccosk (λx) arccosn (µy) = 0.
∂x ∂y
Z Z
dy
Principal integral: Ξ = a arccosk (λx) dx − .
arccosn (µy)
∂w ∂w
17. + y 2 + λ(arccos x)n y − a2 + aλ(arccos x)n = 0.
∂x ∂y
Principal integral:
Z Z
e−2ax E −2ax n
Ξ= + e E dx, E = exp λ (arccos x) dx .
y+a
∂w ∂w
18. + y 2 + λx(arccos x)n y + λ(arccos x)n = 0.
∂x ∂y
Principal integral:
Z Z
E −2 n
Ξ= + x E dx, E = exp λ x(arccos x) dx .
x(xy + 1)
∂w ∂w
19. − (k + 1)xk y 2 − λ(arccos x)n (xk+1 y − 1) = 0.
∂x ∂y
Principal integral:
Z Z
E −k−2 k+1 n
Ξ= −(k +1) x E dx, E = exp λ x (arccos x) dx .
xk+1 (xk+1 y − 1)
∂w ∂w
20. + λ(arccos x)n y 2 + ay + ab − b2 λ(arccos x)n = 0.
∂x ∂y
Principal integral:
Z Z
eax E ax n n
Ξ= +λ e (arccos x) E dx, E = exp −2bλ (arccos x) dx .
y+b
∂w ∂w
21. + λ(arccos x)n y 2 − bλxm (arccos x)n y + bmxm−1 = 0.
∂x ∂y
Principal integral:
Z Z
E n m n
Ξ= +λ (arccos x) E dx, E = exp bλ x (arccos x) dx .
y − bxm
∂w ∂w
22. + λ(arccos x)n y 2 + bmxm−1 − λb2 x2m (arccos x)n = 0.
∂x ∂y
Principal integral:
Z Z
E n m n
Ξ= +λ (arccos x) E dx, E = exp 2bλ x (arccos x) dx .
y − bxm
∂w ∂w
23. + λ(arccos x)n (y − axm − b)2 + amxm−1 = 0.
∂x ∂y
Z
1
Principal integral: Ξ = + λ (arccos x)n dx.
y − axm − b
∂w ∂w
24. x + λ(arccos x)n y 2 + ky + λb2 x2k (arccos x)n = 0.
∂x ∂y
Z
y
Principal integral: Ξ = arctan − λb xk−1 (arccos x)n dx.
bxk
◮ Coefficients of equations contain arctangent.

∂w ∂w
25. + a arctank (λx) + b = 0.
∂x ∂y
Z
Principal integral: Ξ = y − bx − a arctank (λx) dx.
∂w ∂w
26. + a arctank (λy) + b = 0.
∂x ∂y
Z
dy
a arctank (λy) + b
∂w ∂w
27. + k arctann (ax + by + c) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − x, v = ax + by + c.
a + bk arctann v
∂w ∂w
28. + a arctank (λx) arctann (µy) = 0.
∂x ∂y
Z Z
k dy
Principal integral: Ξ = a arctan (λx) dx − .
arctann (µy)
∂w ∂w
29. + y 2 + λ(arctan x)n y − a2 + aλ(arctan x)n = 0.
∂x ∂y
Principal integral:
Z Z
e−2ax E −2ax n
Ξ= + e E dx, E = exp λ (arctan x) dx .
y+a
∂w ∂w
30. + y 2 + λx(arctan x)n y + λ(arctan x)n = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= + x−2 E dx, E = exp λ x(arctan x)n dx .
x(xy + 1)
∂w ∂w
31. − (k + 1)xk y 2 − λ(arctan x)n (xk+1 y − 1) = 0.
∂x ∂y
Principal integral:
Z Z
E −k−2 k+1 n
Ξ= − (k + 1) x E dx, E = exp λ x (arctan x) dx .
xk+1 (xk+1 y − 1)
∂x
+ g(x, y) ∂w
∂y
=0 49
∂w ∂w
32. + λ(arctan x)n y 2 + ay + ab − b2 λ(arctan x)n = 0.
∂x ∂y
Principal integral:
Z Z
eax E ax n n
Ξ= +λ e (arctan x) E dx, E = exp −2bλ (arctan x) dx .
y+b
∂w ∂w
33. + λ(arctan x)n y 2 − bλxm (arctan x)n y + bmxm−1 = 0.
∂x ∂y
Principal integral:
Z Z
E n m n
Ξ= +λ (arctan x) E dx, E = exp bλ x (arctan x) dx .
y − bxm
∂w ∂w
34. + λ(arctan x)n y 2 + bmxm−1 − λb2 x2m (arctan x)n = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= +λ (arctan x)n E dx, E = exp 2bλ xm (arctan x)n dx .
y − bxm
∂w ∂w
35. + λ(arctan x)n (y − axm − b)2 + amxm−1 = 0.
∂x ∂y
Z
1
Principal integral: Ξ = + λ (arctan x)n dx.
y − axm − b
∂w ∂w
36. x + λ(arctan x)n y 2 + ky + λb2 x2k (arctan x)n = 0.
∂x ∂y
Z
y
Principal integral: Ξ = arctan − λb xk−1 (arctan x)n dx.
bxk
◮ Coefficients of equations contain arccotangent.

∂w ∂w
37. + a arccotk (λx) + b = 0.
∂x ∂y
Z
Principal integral: Ξ = y − bx − a arccotk (λx) dx.
∂w ∂w
38. + a arccotk (λy) + b = 0.
∂x ∂y
Z
dy
a arccot k (λy) + b
∂w ∂w
39. + k arccotn (ax + by + c) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − x, v = ax + by + c.
a + bk arccotn v
∂w ∂w
40. + a arccotk (λx) arccot n (µy) = 0.
∂x ∂y
Z Z
dy
Principal integral: Ξ = a arccotk (λx) dx − .
arccotn (µy)
∂w ∂w
41. + y 2 + λ(arccot x)n y − a2 + aλ(arccot x)n = 0.
∂x ∂y
Principal integral:
Z Z
e−2ax E −2ax n
Ξ= + e E dx, E = exp λ (arccot x) dx .
y+a
∂w ∂w
42. + y 2 + λx(arccot x)n y + λ(arccot x)n = 0.
∂x ∂y
Principal integral:
Z Z
E −2 n
Ξ= + x E dx, E = exp λ x(arccot x) dx .
x(xy + 1)
∂w ∂w
43. − (k + 1)xk y 2 − λ(arccot x)n (xk+1 y − 1) = 0.
∂x ∂y
Principal integral:
Z Z
E −k−2 k+1 n
Ξ= − (k + 1) x E dx, E = exp λ x (arccot x) dx .
xk+1 (xk+1 y − 1)
∂w ∂w
44. + λ(arccot x)n y 2 + ay + ab − b2 λ(arccot x)n = 0.
∂x ∂y
Principal integral:
Z Z
eax E ax n n
Ξ= +λ e (arccot x) E dx, E = exp −2bλ (arccot x) dx .
y+b
∂w ∂w
45. + λ(arccot x)n y 2 − bλxm (arccot x)n y + bmxm−1 = 0.
∂x ∂y
Principal integral:
Z Z
E n m n
Ξ= +λ (arccot x) E dx, E = exp bλ x (arccot x) dx .
y − bxm
∂w ∂w
46. + λ(arccot x)n y 2 + bmxm−1 − λb2 x2m (arccot x)n = 0.
∂x ∂y
Principal integral:
Z Z
E n m n
Ξ= +λ (arccot x) E dx, E = exp 2bλ x (arccot x) dx .
y − bxm
∂w ∂w
47. + λ(arccot x)n (y − axm − b)2 + amxm−1 = 0.
∂x ∂y
Z
1
Principal integral: Ξ = + λ (arccot x)n dx.
y − axm − b
∂x
+ g(x, y) ∂w
∂y
=0 51
∂w ∂w
48. x + λ(arccot x)n y 2 + ky + λb2 x2k (arccot x)n = 0.
∂x ∂y
Z
y
Principal integral: Ξ = arccot − λb xk−1 (arccot x)n dx.
bxk
1.1.7 Equations Containing Arbitrary Functions of x

◆ Notation: f = f (x), g = g(x), and h = h(x) are arbitrary functions, and a, b, k, n,
and λ are arbitrary parameters.
◮ Equations contain arbitrary and power-law functions.
∂w ∂w
1. + f (x)y + g(x) = 0.
∂x ∂y
Principal integral:
Z Z
−F −F
Ξ=e y− e g(x) dx, F = f (x) dx.
∂w ∂w
2. + f (x)y + g(x)y k = 0.
∂x ∂y
Principal integral:
Z Z
Ξ = e−F y 1−k − (1 − k) e−F g(x) dx, F = (1 − k) f (x) dx.
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂w ∂w
3. + (y 2 + f y − a2 − af ) = 0.
∂x ∂y
Principal integral:
Z Z
e2ax E 2ax
Ξ= + e E dx, E = exp f dx .
y−a
∂w ∂w
4. + (y 2 + xf y + f ) = 0.
∂x ∂y
Principal integral:
Z Z
E −2
Ξ= + x E dx, E = exp xf dx .
x(xy + 1)

∂w ∂w
5. − (k + 1)xk y 2 − xk+1 f y + f = 0.
∂x ∂y
Principal integral:
Z Z
E −k−2 k+1
Ξ= − (k + 1) x E dx, E = exp x f dx .
xk+1 (xk+1 y − 1)
∂w ∂w
6. + (f y 2 + ay − ab − b2 f ) = 0.
∂x ∂y
Principal integral:
Z Z
eax E ax
Ξ= + e f E dx, E = exp 2b f dx .
y−b
∂w ∂w
7. + (f y 2 − axn f y + anxn−1 ) = 0.
∂x ∂y
Principal integral:
Z Z
E n
Ξ= + f E dx, E = exp a x f dx .
y − axn
∂w ∂w
8. + (f y 2 + anxn−1 − a2 x2n f ) = 0.
∂x ∂y
Principal integral:
Z Z
E n
Ξ= + f E dx, E = exp 2a x f dx .
y − axn
∂w ∂w
9. + (f y 2 + gy − a2 f − ag) = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= + f E dx, E = exp (2af + g) dx.
y−a
∂w ∂w
10. + (f y 2 + gy + anxn−1 − axn g − a2 x2n f ) = 0.
∂x ∂y
Principal integral:
Z Z
E n
Ξ= + f E dx, E = exp (2ax f + g) dx .
y − axn
∂x
+ g(x, y) ∂w
∂y
=0 53
∂w ∂w
11. + f y 2 − axn gy + anxn−1 + a2 x2n (g − f ) = 0.
∂x ∂y
Principal integral:
Z Z
E n
Ξ= + f E dx, E = exp a x (2f − g) dx .
y − axn
∂w ∂w
12. x + (f y 2 + ny + ax2n f ) = 0.
∂x ∂y
1◦ . Principal integral for a > 0:
Z
y √
Ξ = arctan √ n − a xn−1 f dx.
ax
2◦ . Principal integral for a < 0:

Z
y p
Ξ = arctanh p + |a| xn−1 f dx.
|a|xn
∂w ∂w
13. x + x2n f y 2 + (axn f − n)y + bf = 0.
∂x ∂y
Principal integral:
Z Z
dv
Ξ= − xn−1 f dx, v = xn y.
v 2 + av + b
◮ Equations contain arbitrary and exponential functions.
∂w ∂w
14. + (aeλx y 2 + aeλx f y + λf ) = 0.
∂x ∂y
Principal integral:
Z Z
e−2λx E −λx λx
Ξ= + e E dx, E = exp a e f dx .
ay + λe−λx
∂w ∂w
15. + (f y 2 − aeλx f y + aλeλx ) = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= + f E dx, E = exp a eλx f dx .
y − aeλx

∂w ∂w
16. + (f y 2 + aλeλx − a2 e2λx f ) = 0.
∂x ∂y
Principal integral:
Z Z
E λx
Ξ= + f E dx, E = exp 2a e f dx .
y − aeλx
∂w ∂w
17. + (f y 2 + λy + ae2λx f ) = 0.
∂x ∂y
Z
e−λx y √
Ξ = arctan √ − a eλx f dx.
a

Z
e−λx y p
Ξ = arctanh p + |a| eλx f dx.
|a|
∂w ∂w
18. + f y 2 − (aeλx + b)f y + aλeλx = 0.
∂x ∂y
Principal integral:
Z Z
E λx
Ξ= + f E dx, E = exp (ae + b)f dx .
y − aeλx − b
∂w ∂w
19. + eλx f y 2 + (af − λ)y + be−λx f = 0.
∂x ∂y
Principal integral:
Z Z
dv
Ξ= − f (x) dx, v = eλx y.
v 2 + av + b
∂w ∂w
20. + (f y 2 + gy + aλeλx − aeλx g − a2 e2λx f ) = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= + f E dx, E = exp (2aeλx f + g) dx .
y − aeλx
∂w ∂w
21. + f y 2 − aeλx gy + aλeλx + a2 e2λx (g − f ) = 0.
∂x ∂y
Principal integral:
Z Z
E λx
Ξ= + f E dx, E = exp a e (2f − g) dx .
y − aeλx
∂x
+ g(x, y) ∂w
∂y
=0 55
∂w 2 2 ∂w
22. + f y 2 + 2aλxeλx − a2 f e2λx = 0.
∂x ∂y
Principal integral:
Z Z
E λx2
Ξ= + f E dx, E = exp 2a e f dx .
y − aeλx2
∂w 2 ∂w
23. + f y 2 + 2λxy + af e2λx = 0.
∂x ∂y
2 Z
e−λx y √ 2
Ξ = arctan √ − a eλx f dx.
a
2 Z
e−λx y p 2
Ξ = arctanh p + |a| eλx f dx.
|a|
∂w ∂w
24. + f (x)eλy + g(x) = 0.
∂x ∂y
Principal integral:
Z Z
−λy
Ξ=e E+λ f (x)E dx, E = exp λ g(x) dx .
◮ Equations contain arbitrary and hyperbolic functions.

∂w ∂w
25. + f y 2 − a2 f + aλ sinh(λx) − a2 f sinh2 (λx) = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= + f E dx, E = exp 2a f cosh(λx) dx .
y − a cosh(λx)
∂w ∂w
26. + f y 2 − a(af + λ) tanh2 (λx) + aλ = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= + f E dx, E = exp 2a f tanh(λx) dx .
y − a tanh(λx)
∂w ∂w
27. + f y 2 − a(af + λ) coth2 (λx) + aλ = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= + f E dx, E = exp 2a f coth(λx) dx .
y − a coth(λx)
◮ Equations contain arbitrary and logarithmic functions.

∂w ∂w
28. − ay 2 ln x − axy(ln x − 1)f + f = 0.
∂x ∂y
Principal integral:
Z
E E ln x dx
Ξ= − ,
x(ln x − 1)[axy(ln x − 1) − 1] x2 (ln x− 1)2
Z
E = exp a xf (ln x − 1) dx .
∂w ∂w
29. + f y 2 − ax(ln x)f y + a ln x + a = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= + f E dx, E = exp a xf ln x dx .
y − ax ln x
∂w ∂w
30. x + f y 2 + a − a2 (ln x)2 f = 0.
∂x ∂y
Principal integral:
Z Z
E −1 −1
Ξ= + x f E dx, E = exp 2a x f ln x dx .
y − a ln x
∂w ∂w
31. x + (y + a ln x)2 f − a = 0.
∂x ∂y
Z
1 f (x)
Principal integral: Ξ = + dx.
y + a ln x x
◮ Equations contain arbitrary and trigonometric functions.

∂w ∂w
32. + λ sin(λx)y 2 + f cos(λx)y − f = 0.
∂x ∂y
Principal integral:
Z Z
E sin(λx)
Ξ= +λ E dx, E = exp f cos(λx) dx .
cos(λx)[cos(λx)y − 1] cos2 (λx)
∂w ∂w
33. + f y 2 − a2 f + aλ sin(λx) + a2 f sin2 (λx) = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= + f E dx, E = exp −2a f cos(λx) dx .
y + a cos(λx)
∂x
+ g(x, y) ∂w
∂y
=0 57
∂w ∂w
34. + f y 2 − a2 f + aλ cos(λx) + a2 f cos2 (λx) = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= + f E dx, E = exp 2a f sin(λx) dx .
y − a sin(λx)
∂w ∂w
35. + f y 2 − a(af − λ) tan2 (λx) + aλ = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= + f E dx, E = exp 2a f tan(λx) dx .
y − a tan(λx)
∂w ∂w
36. + f y 2 − a(af − λ) cot2 (λx) + aλ = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= + f E dx, E = exp −2a f cot(λx) dx .
y + a cot(λx)
◮ Equations contain arbitrary functions and their derivatives.

∂w ∂w
37. + (f y 2 − f gy + gx′ ) = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= + f E dx, E = exp f g dx .
y−g
∂w ∂w
38. − (fx′ y 2 − f gy + g) = 0.
∂x ∂y
Principal integral:
Z Z
E fx′ E
Ξ= − dx, E = exp f g dx .
f (f y − 1) f2
∂w ∂w
39. + g(y − f )2 + fx′ = 0.
∂x ∂y
Z
1
Principal integral: Ξ = + g dx.
y−f

∂w fx′ 2 gx′ ∂w
40. + y − = 0.
∂x g f ∂y
Principal integral:
Z
1 fx′ dx
Ξ= + .
f (f y + g) f 2g
∂w ∂w
41. f2 + fx′ y 2 − g(y − f ) = 0.
∂x ∂y
Principal integral:
Z Z
f 2E g dx
Ξ= + fx′ E dx, E = exp − .
y−f f2

∂w f ′′ ∂w
42. + y 2 − xx = 0.
∂x f ∂y
Z
1 dx
Principal integral: Ξ = + .
f (f y + fx′ ) f2
∂w ∂w
43. g + af gy 3 + (bf g 3 + gx′ )y + cf g 4 = 0.
∂x ∂y
Principal integral:
Z Z
dv y
Ξ= 3
− f g2 dx, v= .
av + bv + c g
∂w ∂w
44. + f y 3 + 3f hy 2 + (g + 3f h2 )y + f h3 + gh − h′x = 0.
∂x ∂y
Principal integral:
Z Z
E
Ξ= +2 f E dx, E = exp 2 g dx .
(y + h)2
h i
∂w gx′ fx′ ∂w
45. + y3 + y + f gx′ = 0.
∂x f 2 (ag + b)3 f ∂y
Principal integral:
Z
dv 1 y
Ξ= − ln |ag + b|, v= .
v3 − av + 1 a f (ag + b)
h i
∂w af + bg y−g ′ y − f ′ ∂w
46. + (y − f )(y − g) y − h+ fx + gx = 0.
∂x a+b f −g g−f ∂y
Principal integral:
Z
a
af + bg −a−b
b ab 2
Ξ = E|y − f | |y − g| y − , E = exp − (f − g) h dx .
a+b a+b
∂x
+ g(x, y) ∂w
∂y
=0 59
∂w ∂w
47. + (f y 2 + gx′ y + af e2g ) = 0.
∂x ∂y
Z
e−g y √
Ξ = arctan √ − a f eg dx.
a

Z
e−g y p
Ξ = arctanh p + |a| f eg dx.
|a|
∂w ∂w
48. + (fx′ y 2 + aeλx f y + aeλx ) = 0.
∂x ∂y
Principal integral:
Z Z
E fx′ E λx
Ξ= + dx, E = exp a e f dx .
f (f y + 1) f2
1.1.8 Equations Containing Arbitrary Functions of Different

Arguments
◮ Equations contain arbitrary functions of x and arbitrary functions of y.
∂w ∂w
1. f (x) + g(y) = 0.
∂x ∂y
Z Z
dx dy
Principal integral: Ξ = − .
f (x) g(y)
∂w ∂w
2. f (x) + g(y) + fx′ (x) = 0.
∂x ∂y
Z
−y
Principal integral: Ξ = f (x)e − e−y g(y) dy.
∂w ∂w
3. xn f (y) + xg(y) + h(y) = 0.
∂x ∂y
Principal integral:
Z Z
1−n f (y)E g(y)
Ξ=x E + (n − 1) dy, E = exp (n − 1) dy .
h(y) h(y)

∂w ∂w
4. f (y) + amxn y m−1 − g(x) + anxn−1 y m = 0.
∂x ∂y
Z Z
Principal integral: Ξ = f (y) dy + g(x) dx + axn y m .

∂w ∂w
5. eαx f (y) + cβ − eβy g(x) + cα = 0.
∂x ∂y
Z Z
Principal integral: Ξ = e−βy f (y) dy + e−αx g(x) dx − ce−αx−βy .
◮ Equations contain one arbitrary function of complicated argument.

∂w ∂w
6. + f (ax + by + c) = 0, b 6= 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − x, v = ax + by + c.
a + bf (v)

∂w y ∂w
7. +f = 0.
∂x x ∂y
dv y
Ξ= − ln |x|, v= .
f (v) − v x
∂w ∂w
8. + f (y + axn + b) − anxn−1 = 0.
∂x ∂y
dv
Ξ= − x, v = y + axn + b.
f (v)
∂w ∂w
9. x + yf (xn y m ) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − ln |x|, v = xn y m .
v mf (v) + n
∂w ∂w
10. y m−1 + xn−1 f (axn + by m ) = 0.
∂x ∂y
Principal integral:
Z
dv 1
Ξ= − xn , v = axn + by m .
an + bmf (v) n
∂x
+ g(x, y) ∂w
∂y
=0 61
∂w ∂w
11. + e−λx f (eλx y) = 0.
∂x ∂y
dv
Ξ= − x, v = eλx y.
f (v) + λv
∂w ∂w
12. + eλy f (eλy x) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − ln |x|, v = eλy x.
v λvf (v) + 1
∂w ∂w
13. + yf (eαx y m ) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − x, v = eαx y m .
v α + mf (v)
∂w ∂w
14. x + f (xn eαy ) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − ln |x|, v = xn eαy .
v n + αf (v)
∂w ∂w
15. + eλx−βy f (aeλx + beβy ) = 0.
∂x ∂y
Principal integral:
Z
dv 1
Ξ= − eλx , v = aeλx + beβy .
aλ + bβf (v) λ
∂w ∂w
16. + f (y + aeλx + b) − aλeλx = 0.
∂x ∂y
dv
Ξ= − x, v = y + aeλx + b.
f (v)
∂w ∂w
17. αxy + αf (xn eαy ) − ny = 0.
∂x ∂y
Principal integral:
Z Z
1 −1 n αy n dv
Ξ = yE − v E dv, v=x e , E = exp 2 .
α α vf (v)
∂w ∂w
18. mx(ln y) + yf (xn y m ) − ny ln y = 0.
∂x ∂y
Principal integral:
Z Z
1 −1 n m n dv
Ξ = E ln y − v E dv, v = x y , E = exp .
m m vf (v)
∂w ∂w
19. + f (y + a tan x) − a tan2 x = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − x, v = y + a tan x.
a + f (v)
∂w ∂w
20. eλx + f (λx + ln y) = 0.
∂x ∂y
Principal integral:
Z
ev dv
Ξ= − x, v = λx + ln y.
f (v) + λev
∂w ∂w
21. + eλy f (λy + ln x) = 0.
∂x ∂y
Principal integral:
Z
dv
Ξ= − ln x, v = λy + ln x.
λev f (v) + 1
◮ Equations contain several arbitrary functions.

∂w ∂w
22. mx − ny − xy k f (x)g(xny m ) = 0.
∂x ∂y
Principal integral:
Z Z n(1−k)
1−k−m dv
Ξ= v m − x m f (x) dx, v = xn y m .
g(v)
∂w ∂w
23. yn − axn + g(x)f (y n+1 + axn+1 ) = 0.
∂x ∂y
Principal integral:
Z Z
dv
Ξ= + (n + 1) g(x) dx, v = y n+1 + axn+1 .
f (v)
∂x
+ g(x, y) ∂w
∂y
=0 63
h i h i
y y ∂w y y ∂w
24. f + xa h + g + yxa−1 h = 0.
x x ∂x x x ∂y
Principal integral:
Z
−a h(v)E dv y
Ξ=x E+a , v= ,
g(v) − vf (v) x
Z
f (v) dv
where E = exp a .
g(v) − vf (v)
∂w ∂w
25. f (ax + by) + bxg(ax + by) + h(ax + by) − axg(ax + by) = 0.
∂x ∂y
Principal integral:
Z
f (v)E dv
Ξ = xE − , v = ax + by,
af (v) + bh(v)
Z
g(v) dv
where E = exp −b .
af (v) + bh(v)
∂w ∂w
26. f (ax + by) + byg(ax + by) + h(ax + by) − ayg(ax + by) = 0.
∂x ∂y
Principal integral:
Z
h(v)E dv
Ξ = yE − , v = ax + by,
af (v) + bh(v)
Z
g(v) dv
where E = exp a .
af (v) + bh(v)
∂w ∂w
27. x f (xn y m ) + mxk g(xn y m ) + y h(xn y m ) − nxk g(xn y m ) = 0.
∂x ∂y
Principal integral:
Z
g(v)E dv
Ξ = x−k E + km
, v = xn y m ,
v nf (v) + mh(v)
Z
f (v) dv
where E = exp k .
v nf (v) + mh(v)
∂w ∂w
28. x f (xn y m ) + my k g(xn y m ) + y h(xn y m ) − ny k g(xn y m ) = 0.
∂x ∂y
Principal integral:
Z
g(v)E dv
Ξ = y −k E − kn , v = xn y m ,
v nf (v) + mh(v)
Z
h(v) dv
where E = exp k .
v nf (v) + mh(v)
∂w ∂w
29. x sf (xn y m ) − mg(xk y s ) + y ng(xk y s ) − kf (xn y m ) = 0.
∂x ∂y
Principal integral:
Z Z
dv dz
Ξ= − , v = xk y s , z = xn y m .
vg(v) zf (z)
∂w ∂w
30. fy − fx = 0.
∂x ∂y
Here fx and fy are the partial derivatives of the function f = f (x, y) with respect to x
and y.
General solution: w = Φ(f (x, y)), where Φ = Φ(ξ) is an arbitrary function.
∂w ∂w ∂f ∂g
31. f (x, y) − g(x, y) = 0, where = .
∂x ∂y ∂x ∂y
Principal integral: Z Z
y x
Ξ= f (x0 , t) dt + g(t, y) dt,
y0 x0
where x0 and y0 are arbitrary constants.

∂w ∂w
32. x + xf (x)g(xney ) − n = 0.
∂x ∂y
Principal integral:
Z Z
dv
Ξ= − f (x) dx, v = xn ey .
vg(v)
∂w ∂w
33. m + my k f (x)g(eαx y m ) − αy = 0.
∂x ∂y
Principal integral:
Z Z
1−k−m dv α(1 − k)
Ξ= v m − m f (x) exp x dx, v = eαx y m .
g(v) m
∂w ∂w
34. f (ax+by)+beλy g(ax+by) + h(ax+by)−aeλy g(ax+by) = 0.
∂x ∂y
Principal integral:
Z
−λy g(v)E dv
Ξ=e E − λa , v = ax + by,
af (v) + bh(v)
Z
h(v) dv
where E = exp λ .
af (v) + bh(v)
∂x
+ g(x, y) ∂w
∂y
=0 65
∂w ∂w
35. f (ax+by)+beαx g(ax+by) + h(ax+by)−aeαx g(ax+by) = 0.
∂x ∂y
Principal integral:
Z
−αx g(v)E dv
Ξ=e E + αb , v = ax + by,
af (v) + bh(v)
Z
f (v) dv
where E = exp α .
af (v) + bh(v)
∂w ∂w
36. x f (xn eαy ) + αyg(xn eαy ) + h(xn eαy ) − nyg(xn eαy ) = 0.
∂x ∂y
Principal integral:
Z
h(v)E dv
Ξ = yE − , v = xn eαy ,
v nf (v) + αh(v)
Z
g(v) dv
where E = exp n .
v nf (v) + αh(v)
∂w ∂w
37. f (eαx y m ) + mxg(eαx y m ) + y h(eαx y m ) − αxg(eαx y m ) = 0.
∂x ∂y
Principal integral:
Z
f (v)E dv
Ξ = xE − , v = eαx y m ,
v αf (v) + mh(v)
Z
g(v) dv
where E = exp −m .
v αf (v) + mh(v)
∂w ∂w
38. x + xyf (x)g(xn ln y) − ny ln y = 0.
∂x ∂y
Principal integral:
Z Z
dv
Ξ= − xn f (x) dx, v = xn ln y.
g(v)
∂w ∂w
39. x f (xn y m )+mg(xn y m ) ln y +y h(xn y m )−ng(xny m ) ln y = 0.
∂x ∂y
Principal integral:
Z
h(v)E dv
Ξ = E ln y − , v = xn y m ,
v nf (v) + mh(v)
Z
g(v) dv
where E = exp n .
v nf (v) + mh(v)
∂w ∂w
40. x f (xn y m )+mg(xn y m ) ln x +y h(xn y m )−ng(xn y m ) ln x = 0.
∂x ∂y
Principal integral:
Z
f (v)E dv
Ξ = E ln x − , v = xn y m ,
v nf (v) + mh(v)
Z
g(v) dv
where E = exp −m .
v nf (v) + mh(v)
∂w ∂w
41. cos y + f (x)g(sin x sin y) − cot x sin y = 0.
∂x ∂y
Principal integral:
Z Z
dv
Ξ= − f (x) sin x dx, v = sin x sin y.
g(v)
∂w ∂w
42. sin 2x + sin 2x cos2 yf (x)g(tan x tan y) − sin 2y = 0.
∂x ∂y
Principal integral:
Z Z
dv
Ξ= − f (x) tan x dx, v = tan x tan y.
g(v)
∂w ∂w
43. x + x cos2 yf (x)g(x2n tan y) − n sin 2y = 0.
∂x ∂y
Principal integral:
Z Z
dv
Ξ= − x2n f (x) dx, v = x2n tan y.
g(v)
∂w ∂w
44. + cos2 yf (x)g(e2x tan y) − sin 2y = 0.
∂x ∂y
Principal integral:
Z Z
dv
Ξ= − e2x f (x) dx, v = e2x tan y.
g(v)
1.2 Equations of the Form

∂w ∂w
f (x, y) + g(x, y) = h(x, y)
∂x ∂y
◆ The solutions given below contain an arbitrary function Φ = Φ(z).

∂x
+ g(x, y) ∂w
∂y
= h(x, y) 67

∂w ∂w
1. a +b = c.
∂x ∂y
The equation of a cylindrical surface. Two forms of the general solution:
c c
w= x + Φ(bx − ay), w= y + Φ(bx − ay).
a b
∂w ∂w
2. a +b = αx + βy + γ.
∂x ∂y
α 2 γ β
General solution: w = x + x + y 2 + Φ(bx − ay).
2a a 2b
∂w ∂w
3. ax +b = αx + βy + γ.
∂x ∂y
α γ β
General solution: w = x + ln |x| + y 2 + Φ ay − b ln |x| .
a a 2b
∂w ∂w
4. ax + bx = c.
∂x ∂y
c
General solution: w = ln |x| + Φ(bx − ay).
a
∂w ∂w
5. (ax + b) + (cy + d) = αx + βy + γ.
∂x ∂y
General solution:
α aγ − bα β dβ
w= x+ 2
ln |ax + b| + y − 2 ln |cy + d| + Φ |ax + b|c |cy + d|−a .
a a c c
∂w ∂w
6. ay +b = αx + βy + γ.
∂x ∂y
β αx + γ aα
General solution: w = x+ y − 2 y 3 + Φ 2bx − ay 2 .
a b 3b
∂w ∂w
7. ay + bx = c.
∂x ∂y
c √
General solution: w = √ ln ab x + ay + Φ ay 2 − bx2 .
ab
∂w ∂w
8. ay + bx = cx + ky.
∂x ∂y
bkx + acy
General solution: w = + Φ ay 2 − bx2 .
ab
∂w ∂w
9. a +b = cx2 + dy 2 + kxy + n.
∂x ∂y
General solution:
1
w= 2
b(2ac − bk)x3 + 2a2 dy 3 + 3abx(kxy + 2n) + Φ(bx − ay).
6a b
∂w ∂w
10. ax + by = cx2 + dy 2 + kxy + n.
∂x ∂y
General solution:


 1 (bcx2 + ady 2 ) + n ln |x| + k xy + Φ |x|−b/a y if a + b 6= 0,

w = 2ab a a+b
 1 (cx2 − dy 2 ) + 1 (kxy + n) ln |x| + Φ(xy)
 if a + b = 0.
2a a
∂w ∂w
11. ay + bx = cxy + d.
∂x ∂y
c 2 d √
General solution: w = x + √ ln ab x + ay + Φ(ay 2 − bx2 ).
2a ab
∂w ∂w
12. ax2 + by 2 = cx2 + dy 2 + kxy + nx + my + s.
∂x ∂y
General solution:

c s dy 2 kxy ax n m ax − by
w= x− − + ln + ln |x| + ln |y| + Φ .
a ax ax − by ax − by y a b xy
∂w ∂w
13. x2 + axy = by 2 .
∂x ∂y
General solution: 
b y2
 + Φ(|x|−a y) if a 6= 12 ,
w = 2a2− 1 x
b y ln |x| + Φ(|x|−1/2 y) if a = 1 .

x 2
∂w ∂w
14. ay 2 + bx2 = cx2 + d.
∂x ∂y
This is a special case of equation 1.2.7.14 with k = 2, f (x) = a, g(x) = bx2 , and h(x) =
cx2 + d.
∂w ∂w
15. ay 2 + bxy = cx2 + dy 2 .
∂x ∂y
General solution:
r r
ac + bd c ay 2 − bx2 b
w= x− arctan x + Φ(ay 2 − bx2 ).
ab b b ay 2 − bx2
∂x
+ g(x, y) ∂w
∂y
= h(x, y) 69
◮ Coefficients of equations contain other power-law functions.
∂w ∂w p
16. x +y =a x2 + y 2 .
∂x ∂y
p
2 2
y
General solution: w = a x + y + Φ .
x
∂w ∂w
17. ax + by = cxy 2 + dx2 y + k.
∂x ∂y
cxy 2 dx2 y k
General solution: w = + + ln |x| + Φ |x|−b/a y .
a + 2b 2a + b a
∂w ∂w
18. ay + bx = cx2 y + d.
∂x ∂y
c 3 d √
General solution: w = x + √ ln ab x + ay + Φ(ay 2 − bx2 ).
3a ab
∂w ∂w
19. (ax + b) + (cy + d) = kx3 + ny 3 .
∂x ∂y
General solution:

k 1 3 b 2 b2 b3
w= x − x + 2 x − 3 ln |ax + b|
a 3 2a a a

n 1 3 d d2 d3
+ y − y 2 + 2 y − 3 ln |cy + d| + Φ |ax + b|c |cy + d|−a .
c 3 2c c c
∂w ∂w
20. x2 + xy = y 2 (ax + by).
∂x ∂y

(ax + by)y 2 y
General solution: w = +Φ .
2x x
∂w ∂w
21. ax3 + by 3 = cx + d.
∂x ∂y
2
2cx + d ax − by 2
General solution: w = − +Φ .
2ax2 x2 y 2
∂w ∂w
22. a +b = cxn + dy m .
∂x ∂y
c d
General solution: w = Φ(bx − ay) + xn+1 + y m+1 .
a(n + 1) b(m + 1)
∂w ∂w
23. a +b = cxn y.
∂x ∂y
General solution:


 c[a(n + 2)y − bx]xn+1

 + Φ(bx − ay) if n 6= −1, −2;

 a2 (n + 1)(n + 2)

w = bc x 1 − ln |x| + c y ln |x| + Φ(bx − ay) if n = −1;
 a2
 a


 bc 1 + ln |x| − cy + Φ(bx − ay)
 if n = −2.
a2 ax
∂w ∂w
24. x +y = a(x2 + y 2 )k .
∂x ∂y

a 2 2 k y
General solution: w = (x + y ) + Φ .
2k x
∂w ∂w
25. ax + by = cxn y m .
∂x ∂y
General solution:
 c
 xn y m + Φ |y|a |x|−b if an + bm 6= 0,
w = an + bm
 c xn y m ln |x| + Φ |y|a |x|−b if an + bm = 0.
a
∂w ∂w
26. ax + by = cxn + dy m .
∂x ∂y
c n d m
General solution: w = x + y + Φ y a x−b .
an bm
∂w ∂w
27. mx + ny = (axn + by m )k .
∂x ∂y
1
General solution: w = (axn + by m )k + Φ y m x−n .
mnk
∂w ∂w
28. axn + by m = cxk + dy s .
∂x ∂y
This is a special case of equation 1.2.7.20. General solution:
c d
w= xk−n+1 + y s−m+1 + Φ(u),
a(k − n + 1) b(s − m + 1)
1 1
u= x1−n − y 1−m .
a(1 − n) b(1 − m)
∂w ∂w
29. axn + bxm y = cxk y s + d.
∂x ∂y
This is a special case of equation 1.2.7.34 with f (x) = axn , g(x) = bxm , and h(x, y) =
cxk y s + d.
∂x
+ g(x, y) ∂w
∂y
= h(x, y) 71
∂w ∂w
30. axn + bxm y + cxk = sxp y q + d.
∂x ∂y
This is a special case of equation 1.2.7.35 with f (x) = axn , g1 (x) = bxm , g0 (x) = cxk ,
and h(x, y) = sxp y q + d.
∂w ∂w
31. axn + bxm y k + cxl y = sxp y q + d.
∂x ∂y
This is a special case of equation 1.2.7.36 with f (x) = axn , g1 (x) = cxl , g0 (x) = bxm ,
∂w ∂w
32. ay k + bxn = cxm + d.
∂x ∂y
This is a special case of equation 1.2.7.14 with f (x) = a, g(x) = bxn , and h(x) = cxm + d.

∂w ∂w
1. a +b = ceλx + deµy .
∂x ∂y
c λx d µy
General solution: w = e + e + Φ(bx − ay).
aλ bµ
∂w ∂w
2. a +b = ceαx+βy .
∂x ∂y
General solution:
c
 eαx+βy + Φ(bx − ay) if aα + bβ 6= 0,
aα
w= c + bβ
 xeαx+βy + Φ(bx − ay) if aα + bβ = 0.
a
∂w ∂w
3. aeλx + beβy = c.
∂x ∂y
c −λx
General solution: w = − e + Φ(u), where u = bβe−λx − aλe−βy .
aλ
∂w ∂w
4. aeλy + beβx = c.
∂x ∂y
c(βx − λy)
General solution: w = + Φ(u), where u = aβeλy − bλeβx .
u
∂w ∂w
5. aeαx + beβy = ceγx−βy .
∂x ∂y
1 βy 1 αx
Introduce the notation u = e − e .
βb αa
General solution:


 c (γ−α)x −βy bβe−αx

 e e + + Φ(u) if γ 6= α, 2α,

 a(γ − α) a(γ −
2α)
c bβ
w= xe−βy − 2
(αx + 1)e−αx + Φ(u) if γ = α,

 a aα

 c bβ

 αx−βy
 e + (αx − 1) + Φ(u) if γ = 2α.
aα aα
∂w ∂w
6. aeαx + beβy = ceγx−2βy .
∂x ∂y
1 βy 1 αx
Introduce the notation u = e − e .
βb αa
1◦ . General solution for γ 6= α, γ 6= 2α, and γ 6= 3α:

c −2βy 2bβ −αx−βy 2b2β 2 −2αx (γ−α)x
w= e + e + 2 e e + Φ(u).
a(γ − α) a(γ − 2α) a (γ − 2α)(γ − 3α)
2◦ . General solution for γ = α:

c −2βy 2bβ −αx−βy b2 β 2 3 −2αx
w= xe − (αx + 1)e + 2 3 αx + e + Φ(u).
a aα2 a α 2
3◦ . General solution for γ = 2α:

c 2bβ
w= eαx−βy + (αx − 1) e−βy + Φ(u).
aα aα
4◦ . General solution for γ = 3α:

c 1 2(αx−βy) bβ αx−βy b2 β 2 3
w= e + e + 2 2 αx − + Φ(u).
aα 2 aα a α 2
∂w ∂w
7. aeαx + beβy = ceγx + seµy .
∂x ∂y
This is a special case of equation 1.2.7.20 with f (x) = aeαx , g(y) = beβy , h1 (x) = ceγx ,
and h2 (y) = seµy .
∂w ∂w
8. aeβx + beγx + ceλy = seµx + keδy + p.
∂x ∂y
This is a special case of equation 1.2.7.37 with f (x) = aeβx , g1 (x) = beγx , g0 (x) = c, and
h(x, y) = seµx + keδy + p.
∂w ∂w
9. aeβx + beγx + ceλy = seµx+δy + k.
∂x ∂y
h(x, y) = seµx+δy + k.
∂x
+ g(x, y) ∂w
∂y
= h(x, y) 73
∂w ∂w
10. aeβx + beγx+λy = ceµx+δy + k.
∂x ∂y
This is a special case of equation 1.2.7.37 with f (x) = aeβx , g1 (x) ≡ 0, g0 (x) = beγx , and
h(x, y) = ceµx+δy + k.
∂w ∂w
11. aeλy + beβx = ceγx + d.
∂x ∂y
This is a special case of equation 1.2.7.16 with f (x) = a, g(x) = beβx , and h(x) = ceγx +d.

∂w ∂w
12. a +b = cyeλx + kxeµy .
∂x ∂y

c λx b k µy a
General solution: w = e y− + e x− + Φ(bx − ay).
aλ aλ bµ bµ
∂w ∂w
13. +a = axk eλy .
∂x ∂y
This is a special case of equation 1.2.7.5 with f (x) = axk .
∂w ∂w
14. + (ay + beλx ) = ceβx .
∂x ∂y
This is a special case of equation 1.2.7.6 with f (x) = beλx and g(x) = ceβx .
∂w ∂w
15. + (aeλx y + beβx y k ) = ceµx .
∂x ∂y
This is a special case of equation 1.2.7.12 with f (x) = 1, g1 (x) = aeλx , g2 (x) = beβx , and
h(x) = ceµx .
∂w ∂w
16. + (axk + bxn eλy ) = ceβx .
∂x ∂y
This is a special case of equation 1.2.7.13 with f (x) = 1, g1 (x) = axk , g2 (x) = bxn , and
h(x) = ceβx .
∂w ∂w
17. x +y = axeλx+µy .
∂x ∂y

ax λx+µy y
General solution: w = e +Φ .
λx + µy x
∂w ∂w
18. x +y = ayeλx + bxeµy .
∂x ∂y

ay λx bx µy y
General solution: w = e + e +Φ .
λx µy x
∂w ∂w
19. axk + beλy = cxn + s.
∂x ∂y
This is a special case of equation 1.2.7.13 with f (x) = axk , g1 (x) = 0, g2 (x) = b, and
h(x) = cxn + s.
∂w ∂w
20. ay k + beλx = ceµx + s.
∂x ∂y
This is a special case of equation 1.2.7.14 with f (x) = a, g(x) = beλx , and h(x) = ceµx +s.
∂w ∂w
21. aeλx + by k = cxn + s.
∂x ∂y
This is a special case of equation 1.2.7.12 with f (x) = aeλx , g1 (x) = 0, g2 (x) = b, and
h(x) = cxn + s.
∂w ∂w
22. aeλy + bxk = ceµx + s.
∂x ∂y
This is a special case of equation 1.2.7.16 with f (x) = a, g(x) = bxk , and h(x) = ceµx + s.

∂w ∂w
1. a +b = c sinh(λx) + k sinh(µy).
∂x ∂y
c k
General solution: w = cosh(λx) + cosh(µy) + Φ(bx − ay).
aλ bµ
∂w ∂w
2. a +b = c sinh(λx + µy).
∂x ∂y
General solution:
 c
 cosh(λx + µy) + Φ(bx − ay) if aλ + bµ 6= 0,
w = aλ + bµ
 c x sinh(λx + µy) + Φ(bx − ay) if aλ + bµ = 0.
a
∂w ∂w
3. x +y = ax sinh(λx + µy).
∂x ∂y

ax y
General solution: w = cosh(λx + µy) + Φ .
λx + µy x
∂w ∂w
4. a + b sinhn (λx) = c sinhm (µx) + s sinhk (βy).
∂x ∂y
This is a special case of equation 1.2.7.35 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b sinhn (λx),
and h(x, y) = c sinhm (µx) + s sinhk (βy).
∂w ∂w
5. a + b sinhn (λy) = c sinhm (µx) + s sinhk (βy).
∂x ∂y
This is a special case of equation 1.2.7.20 with f (x) = a, g(y) = b sinhn (λy), h1 (x) =
c sinhm (µx), and h2 (y) = s sinhk (βy).
∂x
+ g(x, y) ∂w
∂y
= h(x, y) 75

∂w ∂w
6. a +b = c cosh(λx) + k cosh(µy).
∂x ∂y
c k
General solution: w = sinh(λx) + sinh(µy) + Φ(bx − ay).
aλ bµ
∂w ∂w
7. a +b = c cosh(λx + µy).
∂x ∂y
General solution:
 c
 sinh(λx + µy) + Φ(bx − ay) if aλ + bµ 6= 0,
w = aλ
c
+ bµ
 x cosh(λx + µy) + Φ(bx − ay) if aλ + bµ = 0.
a
∂w ∂w
8. x +y = ax cosh(λx + µy).
∂x ∂y

ax y
General solution: w = sinh(λx + µy) + Φ .
λx + µy x
∂w ∂w
9. a + b coshn (λx) = c coshm (µx) + s coshk (βy).
∂x ∂y
This is a special case of equation 1.2.7.35 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b coshn (λx),
and h(x, y) = c coshm (µx) + s coshk (βy).
∂w ∂w
10. a + b coshn (λy) = c coshm (µx) + s coshk (βy).
∂x ∂y
This is a special case of equation 1.2.7.20 with f (x) = a, g(y) = b coshn (λy), h1 (x) =
c coshm (µx), and h2 (y) = s coshk (βy).

∂w ∂w
11. a +b = c tanh(λx) + k tanh(µy).
∂x ∂y
c k
General solution: w = ln cosh(λx) + ln cosh(µy) + Φ(bx − ay).
aλ bµ
∂w ∂w
12. a +b = c tanh(λx + µy).
∂x ∂y
General solution:
 c
 ln cosh(λx + µy) + Φ(bx − ay) if aλ + bµ 6= 0,
w = aλ + bµ
 c x tanh(λx + µy) + Φ(bx − ay) if aλ + bµ = 0.
a
∂w ∂w
13. x +y = ax tanh(λx + µy).
∂x ∂y

ax y
General solution: w = ln cosh(λx + µy) + Φ .
λx + µy x
∂w ∂w
14. a + b tanhn (λx) = c tanhm (µx) + s tanhk (βy).
∂x ∂y
This is a special case of equation 1.2.7.35 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b tanhn (λx),
and h(x, y) = c tanhm (µx) + s tanhk (βy).
∂w ∂w
15. a + b tanhn (λy) = c tanhm (µx) + s tanhk (βy).
∂x ∂y
This is a special case of equation 1.2.7.20 with f (x) = a, g(y) = b tanhn (λy), h1 (x) =
c tanhm (µx), and h2 (y) = s tanhk (βy).
∂w ∂w
16. a +b = c coth(λx) + k coth(µy).
∂x ∂y
c k
General solution: w = lnsinh(λx) + ln sinh(µy) + Φ(bx − ay).
aλ bµ
∂w ∂w
17. a +b = c coth(λx + µy).
∂x ∂y
General solution:
 c
 lnsinh(λx + µy) + Φ(bx − ay) if aλ + bµ 6= 0,
w = aλ + bµ
 c x coth(λx + µy) + Φ(bx − ay) if aλ + bµ = 0.
a
∂w ∂w
18. x +y = ax coth(λx + µy).
∂x ∂y

ax y
General solution: w =
ln sinh(λx + µy) + Φ .
λx + µy x
∂w ∂w
19. a + b cothn (λx) = c cothm (µx) + s cothk (βy).
∂x ∂y
This is a special case of equation 1.2.7.35 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b cothn (λx),
and h(x, y) = c cothm (µx) + s cothk (βy).
∂w ∂w
20. a + b cothn (λy) = c cothm (µx) + s cothk (βy).
∂x ∂y
This is a special case of equation 1.2.7.20 with f (x) = a, g(y) = b cothn (λy), h1 (x) =
c cothm (µx), and h2 (y) = s cothk (βy).
∂w ∂w
21. a +b = c sinh(λx) + k cosh(µy).
∂x ∂y
c k
General solution: w = cosh(λx) + sinh(µy) + Φ(bx − ay).
aλ bµ
∂x
+ g(x, y) ∂w
∂y
= h(x, y) 77
∂w ∂w
22. a +b = tanh(λx) + k coth(µy).
∂x ∂y
1 k
General solution: w = ln cosh(λx) + ln sinh(µy) + Φ(bx − ay).
aλ bµ
∂w ∂w
23. a +b = sinh(λx) + k tanh(µy).
∂x ∂y
1 k
General solution: w = cosh(λx) + ln cosh(µy) + Φ(bx − ay).
aλ bµ
∂w ∂w
24. a + b cosh(µy) = sinh(λx).
∂x ∂y
1 µx
General solution: w = cosh(λx) + Φ(u), where u = bµx − 2a arctan tanh .
aλ 2
∂w ∂w
25. a + b sinh(µy) = cosh(λx).
∂x ∂y
1 µx

General solution: w = sinh(λx) + Φ(u), where u = bµx − a lntanh .
aλ 2

∂w ∂w
1. a +b = c ln(λx + βy).
∂x ∂y
General solution:


 c(λx + βy) ln(λx + βy) − 1 + Φ(bx − ay) if aλ 6= −bβ,

w = c aλ + bβ

 x ln(λx + βy) + Φ(bx − ay) if aλ = −bβ.
a
∂w ∂w
2. a +b = c ln(λx) + k ln(βy).
∂x ∂y
c k
General solution: w = x ln(λx) − 1 + y ln(βy) − 1 + Φ(bx − ay).
a b
∂w ∂w
3. a + b ln(λx) ln(βy) = c ln(γx).
∂x ∂y
General solution:
Z
c dy
w = x ln(γx) − 1 + Φ(u), where u = bx ln(λx) − 1 − a .
a ln(βy)
∂w ∂w
4. a + b lnn (λx) = c lnm (µx) + s lnk (βy).
∂x ∂y
This is a special case of equation 1.2.7.35 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b lnn (λx),
and h(x, y) = c lnm (µx) + s lnk (βy).
∂w ∂w
5. a + b lnn (λy) = c lnm (µx) + s lnk (βy).
∂x ∂y
This is a special case of equation 1.2.7.20 with f (x) = a, g(y) = b lnn (λy), h1 (x) =
c lnm (µx), and h2 (y) = s lnk (βy).
∂w ∂w
6. a lnn (λx) + b lnk (βy) = c lnm (γx).
∂x ∂y
General solution:
Z Z Z
c lnm (γx) dx dy
w= dx + Φ(u), where u = b −a .
a lnn (λx) lnn (λx) k
ln (βy)

∂w ∂w
7. a +b = cxn + s lnk (λy).
∂x ∂y
Z
c s
General solution: w = xn+1 + lnk (λy) dy + Φ(bx − ay).
a(n + 1) b
∂w ∂w
8. +a = by 2 + cxn y + s lnk (λx).
∂x ∂y
This is a special case of equation 1.2.7.3 with f (x) = b, g(x) = cxn , and h(x) = s lnk (λx).
∂w ∂w
9. +a = b lnk (λx) lnn (βy).
∂x ∂y
This is a special case of equation 1.2.7.18 with f (x) = b lnk (λx) and g(y) = lnn (βy).
∂w ∂w
10. + (ay + bxn ) = c lnk (λx).
∂x ∂y
This is a special case of equation 1.2.7.6 with f (x) = bxn and g(x) = c lnk (λx).
∂w ∂w
11. ax + by = xk (n ln x + m ln y).
∂x ∂y
This is a special case of equation 1.2.7.28 with f (u) = ln u.
∂w ∂w
12. axk + by n = c lnm (λx) + s lnl (βy).
∂x ∂y
General solution:
Z Z
c −k m s b 1−k a 1−n
w= x ln (λx) dx+ y −n lnl(βy) dy+Φ(u), u= x − y .
a b 1−k 1−n

∂w ∂w
1. a +b = c sin(λx) + k sin(µy).
∂x ∂y
c k
General solution: w = − cos(λx) − cos(µy) + Φ(bx − ay).
aλ bµ
∂x
+ g(x, y) ∂w
∂y
= h(x, y) 79
∂w ∂w
2. a +b = c sin(λx + µy).
∂x ∂y
General solution:
 c
− cos(λx + µy) + Φ(bx − ay) if aλ + bµ 6= 0,
w = c aλ + bµ
 x sin(λx + µy) + Φ(bx − ay) if aλ + bµ = 0.
a
∂w ∂w
3. x +y = ax sin(λx + µy).
∂x ∂y

ax y
General solution: w = − cos(λx + µy) + Φ .
λx + µy x
∂w ∂w
4. a + b sinn (λx) = c sinm (µx) + s sink (βy).
∂x ∂y
This is a special case of equation 1.2.7.35 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b sinn (λx),
and h(x, y) = c sinm (µx) + s sink (βy).
∂w ∂w
5. a + b sinn (λy) = c sinm (µx) + s sink (βy).
∂x ∂y
This is a special case of equation 1.2.7.20 with f (x) = a, g(y) = b sinn (λy), h1 (x) =
c sinm (µx), and h2 (y) = s sink (βy).

∂w ∂w
6. a +b = c cos(λx) + k cos(µy).
∂x ∂y
c k
General solution: w = sin(λx) + sin(µy) + Φ(bx − ay).
aλ bµ
∂w ∂w
7. a +b = c cos(λx + µy).
∂x ∂y
General solution:
 c
 sin(λx + µy) + Φ(bx − ay) if aλ + bµ 6= 0,
w = aλ + bµ
 c x cos(λx + µy) + Φ(bx − ay) if aλ + bµ = 0.
a
∂w ∂w
8. x +y = ax cos(λx + µy).
∂x ∂y

ax y
General solution: w = sin(λx + µy) + Φ .
λx + µy x
∂w ∂w
9. a + b cosn (λx) = c cosm (µx) + s cosk (βy).
∂x ∂y
This is a special case of equation 1.2.7.35 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b cosn (λx),
and h(x, y) = c cosm (µx) + s cosk (βy).
∂w ∂w
10. a + b cosn (λy) = c cosm (µx) + s cosk (βy).
∂x ∂y
This is a special case of equation 1.2.7.20 with f (x) = a, g(y) = b cosn (λy), h1 (x) =
c cosm (µx), and h2 (y) = s cosk (βy).

∂w ∂w
11. a +b = c tan(λx) + k tan(µy).
∂x ∂y
c k
General solution: w = − lncos(λx) − ln cos(µy) + Φ(bx − ay).
aλ bµ
∂w ∂w
12. a +b = c tan(λx + µy).
∂x ∂y
General solution:
 c
− lncos(λx + µy) + Φ(bx − ay) if aλ + bµ 6= 0,
w = c aλ + bµ
 x tan(λx + µy) + Φ(bx − ay) if aλ + bµ = 0.
a
∂w ∂w
13. x +y = ax tan(λx + µy).
∂x ∂y

ax y
General solution: w = −
ln cos(λx + µy) + Φ .
λx + µy x
∂w ∂w
14. a + b tann (λx) = c tanm (µx) + s tank (βy).
∂x ∂y
This is a special case of equation 1.2.7.35 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b tann (λx),
and h(x, y) = c tanm (µx) + s tank (βy).
∂w ∂w
15. a + b tann (λy) = c tanm (µx) + s tank (βy).
∂x ∂y
This is a special case of equation 1.2.7.20 with f (x) = a, g(y) = b tann (λy), h1 (x) =
c tanm (µx), and h2 (y) = s tank (βy).

∂w ∂w
16. a +b = c cot(λx) + k cot(µy).
∂x ∂y
c k
General solution: w = lnsin(λx) + ln sin(µy) + Φ(bx − ay).
aλ bµ
∂w ∂w
17. a +b = c cot(λx + µy).
∂x ∂y
General solution:
 c
 lnsin(λx + µy) + Φ(bx − ay) if aλ + bµ 6= 0,
w = aλ + bµ
 c x cot(λx + µy) + Φ(bx − ay) if aλ + bµ = 0.
a
∂w ∂w
18. x +y = ax cot(λx + µy).
∂x ∂y

ax y
General solution: w =
ln sin(λx + µy) + Φ .
λx + µy x
∂x
+ g(x, y) ∂w
∂y
= h(x, y) 81
∂w ∂w
19. a + b cotn (λx) = c cotm (µx) + s cotk (βy).
∂x ∂y
This is a special case of equation 1.2.7.35 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b cotn (λx),
and h(x, y) = c cotm (µx) + s cotk (βy).
∂w ∂w
20. a + b cotn (λy) = c cotm (µx) + s cotk (βy).
∂x ∂y
This is a special case of equation 1.2.7.20 with f (x) = a, g(y) = b cotn (λy), h1 (x) =
c cotm (µx), and h2 (y) = s cotk (βy).
∂w ∂w
21. a +b = sin(λx) + c cos(µy) + k.
∂x ∂y
k 1 c
General solution: w = x− cos(λx) + sin(µy) + Φ(bx − ay).
a aλ bµ
∂w ∂w
22. a +b = tan(λx) + c sin(µy) + k.
∂x ∂y
k 1 c
General solution: w = x− ln cos(λx) − cos(µy) + Φ(bx − ay).
a aλ bµ
∂w ∂w
23. a +b = sin(λx) cos(µy) + c.
∂x ∂y
General solution:


 c cos(λx − µy) cos(λx + µy)

 x− − + Φ(bx − ay) if aλ ± bµ 6= 0,

 a 2(aλ − bµ) 2(aλ + bµ)
c x hµ i cos(λx + µy)
w= x+ sin (bx − ay) − + Φ(bx − ay) if aλ − bµ = 0,

 a 2a a 2(aλ + bµ)

 c x h µ i cos(λx − µy)


 x− sin (bx − ay) − + Φ(bx − ay) if aλ + bµ = 0.
a 2a a 2(aλ − bµ)
∂w ∂w
24. a + b sin(µy) = cos(λx) + c.
∂x ∂y
c 1
General solution: w = x+ sin(λx) + Φ(u), where u = bµx − a lntan 1
2 µy
.
a aλ
∂w ∂w
25. a + b tan(µy) = sin(λx) + c.
∂x ∂y
c 1
General solution: w = x− cos(λx) + Φ(u), where u = bµx − a lnsin(µy).
a aλ
∂w ∂w
26. a + b tan(µy) = cot(λx) + c.
∂x ∂y
c 1
General solution: w = x− ln sin(λx) + Φ(u), where u = bµx − a lnsin(µy).
a aλ

∂w ∂w x y
1. a +b = c arcsin + k arcsin .
∂x ∂y λ β
General solution:
c x p k y p
w= x arcsin + λ2 − x2 + y arcsin + β 2 − y 2 + Φ(bx − ay).
a λ b β
∂w ∂w
2. a +b = c arcsin(λx + βy).
∂x ∂y
1◦ . General solution for aλ + bβ 6= 0:

c h p i
w= (λx + βy) arcsin(λx + βy) + 1 − (λx + βy)2 + Φ(bx − ay).
aλ + bβ
2◦ . General solution for aλ + bβ = 0:

c
w= x arcsin(λx + βy) + Φ(bx − ay).
a
∂w ∂w
3. x +y = ax arcsin(λx + βy).
∂x ∂y
p
1 − (λx + βy)2 y
General solution: w = ax arcsin(λx + βy) + +Φ .
λx + βy x
∂w ∂w
4. a + b arcsinn (λx) = c arcsinm (µx) + s arcsink (βy).
∂x ∂y
This is a special case of equation 1.2.7.35 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b arcsinn (λx),
and h(x, y) = c arcsinm (µx) + s arcsink (βy).
∂w ∂w
5. a + b arcsinn (λy) = c arcsinm (µx) + s arcsink (βy).
∂x ∂y
This is a special case of equation 1.2.7.20 with f (x) = a, g(y) = b arcsinn (λy), h1 (x) =
c arcsinm (µx), and h2 (y) = s arcsink (βy).
∂w ∂w x y
6. a +b = c arccos + k arccos .
∂x ∂y λ β
General solution:
c x p k y p
w= x arccos − λ2 − x2 + y arccos − β 2 − y 2 + Φ(bx − ay).
a λ b β
∂x
+ g(x, y) ∂w
∂y
= h(x, y) 83
∂w ∂w
7. a +b = c arccos(λx + βy).
∂x ∂y
c h p i
w= (λx + βy) arccos(λx + βy) − 1 − (λx + βy)2 + Φ(bx − ay).
aλ + bβ

c
w= x arccos(λx + βy) + Φ(bx − ay).
a
∂w ∂w
8. x +y = ax arccos(λx + βy).
∂x ∂y
p
1 − (λx + βy)2 y
General solution: w = ax arccos(λx + βy) − +Φ .
λx + βy x
∂w ∂w
9. a + b arccosn (λx) = c arccosm (µx) + s arccosk (βy).
∂x ∂y
This is a special case of equation 1.2.7.35 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b arccosn (λx),
and h(x, y) = c arccosm (µx) + s arccosk (βy).
∂w ∂w
10. a + b arccosn (λy) = c arccosm (µx) + s arccosk (βy).
∂x ∂y
This is a special case of equation 1.2.7.20 with f (x) = a, g(y) = b arccosn (λy), h1 (x) =
c arccosm (µx), and h2 (y) = s arccosk (βy).
∂w ∂w x y
11. a +b = c arctan + k arctan .
∂x ∂y λ β
General solution:

c x λ k y β
w= x arctan − ln(λ2 + x2 ) + y arctan − ln(β 2 + y 2 ) + Φ(bx − ay).
a λ 2 b β 2
∂w ∂w
12. a +b = c arctan(λx + βy).
∂x ∂y

c 1 2

w= (λx + βy) arctan(λx + βy) − ln 1 + (λx + βy) + Φ(bx − ay).
aλ + bβ 2

c
w= x arctan(λx + βy) + Φ(bx − ay).
a
∂w ∂w
13. x +y = ax arctan(λx + βy).
∂x ∂y
General solution:

1 2 x2 y
w = ax arctan(λx + βy) − ln x + 2
+Φ .
2(λx + βy) (λx + βy) x
∂w ∂w
14. a + b arctann (λx) = c arctanm (µx) + s arctank (βy).
∂x ∂y
This is a special case of equation 1.2.7.35 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b arctann (λx),
and h(x, y) = c arctanm (µx) + s arctank (βy).
∂w ∂w
15. a + b arctann (λy) = c arctanm (µx) + s arctank (βy).
∂x ∂y
This is a special case of equation 1.2.7.20 with f (x) = a, g(y) = b arctann (λy), h1 (x) =
c arctanm (µx), and h2 (y) = s arctank (βy).

∂w ∂w x y
16. a +b = c arccot + k arccot .
∂x ∂y λ β
General solution:

c x λ k y β
w= x arccot + ln(λ2 + x2 ) + y arccot + ln(β 2 + y 2 ) + Φ(bx − ay).
a λ 2 b β 2
∂w ∂w
17. a +b = c arccot(λx + βy).
∂x ∂y

c 1 2

w= (λx + βy) arccot(λx + βy) + ln 1 + (λx + βy) + Φ(bx − ay).
aλ + bβ 2
c
w = x arccot(λx + βy) + Φ(bx − ay).
a
∂w ∂w
18. x +y = ax arccot(λx + βy).
∂x ∂y
General solution:

1 x2 y
w = ax arccot(λx + βy) + ln x2 + + Φ .
2(λx + βy) (λx + βy)2 x
∂w ∂w
19. a + b arccotn (λx) = c arccot m (µx) + s arccot k (βy).
∂x ∂y
This is a special case of equation 1.2.7.35 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b arccotn (λx),
and h(x, y) = c arccotm (µx) + s arccotk (βy).
∂w ∂w
20. a + b arccotn (λy) = c arccot m (µx) + s arccot k (βy).
∂x ∂y
This is a special case of equation 1.2.7.20 with f (x) = a, g(y) = b arccotn (λy), h1 (x) =
c arccotm (µx), and h2 (y) = s arccotk (βy).
∂x
+ g(x, y) ∂w
∂y
= h(x, y) 85
1.2.7 Equations Containing Arbitrary Functions

◮ Coefficients of equations contain arbitrary functions of x.
∂w ∂w
1. a +b= f (x).
∂x ∂y
Z
1
General solution: w = f (x) dx + Φ(bx − ay).
a
∂w ∂w
2. +a = f (x)y.
∂x ∂y
Z x
General solution: w = (y − ax + at)f (t) dt + Φ(y − ax), where x0 may be taken as
x0
arbitrary.
∂w ∂w
3. +a = f (x)y 2 + g(x)y + h(x).
∂x ∂y
General solution:
w = ϕ(x)y 2 + ψ(x)y + χ(x) + Φ(y − ax),
where
Z Z Z

ϕ(x) = f (x) dx, ψ(x) = g(x) − 2aϕ(x) dx, χ(x) = h(x) − aψ(x) dx.
∂w ∂w
4. +a = f (x)y k .
∂x ∂y
Z x
General solution: w = (y − ax + at)k f (t) dt + Φ(y − ax), where x0 may be taken as
x0
arbitrary.
∂w ∂w
5. +a = f (x)eλy .
∂x ∂y
Z
λ(y−ax)
General solution: w = e f (x)eaλx dx + Φ(y − ax).
∂w ∂w
6. + [ay + f (x)] = g(x).
∂x ∂y
Z Z
General solution: w = g(x) dx + Φ(u), where u = e y − f (x)e−ax dx.
−ax
∂w ∂w
7. + ay + f (x) = g(x)y k .
∂x ∂y
This is a special case of equation 1.2.7.19 with h(y) = y k .

∂w ∂w
8. f (x) + yk = g(x).
∂x ∂y
General solution:
 Z

 1 1−k dx
Z  y + if k 6= 1,
g(x) k − 1 Z f (x)

w= dx + Φ(u), where u= dx
f (x) 
y exp − if k = 1.

f (x)
∂w ∂w
9. f (x) + (y + a) = by + c.
∂x ∂y
Z
dx
General solution: w = by +(c−ab) ln |y +a|+Φ(u), where u = (y +a) exp − .
f (x)
∂w ∂w
10. f (x) + (y + ax) = g(x).
∂x ∂y
General solution:
Z Z Z
g(x) −z xe−z dx
w= dx + Φ e y − a dx , where z = .
f (x) f (x) f (x)
∂w ∂w
11. f (x) + g1 (x)y + g0 (x) = h2 (x)y 2 + h1 (x)y + h0 (x).
∂x ∂y
General solution:
Z
g0
w = ϕ(x)y 2 + ψ(x)y + χ(x) + Φ(u), u = e−G y − e−G dx,
f
where Z Z
−2G h2 2G g1
ϕ(x) = e e dx, G = G(x) = dx,
f f
Z Z
h1 − 2g0 ϕ h0 − g0 ψ
ψ(x) = e−G eG dx, χ(x) = dx.
f f
∂w ∂w
12. f (x) + g1 (x)y + g2 (x)y k = h(x).
∂x ∂y
Z
h(x)
General solution: w = dx + Φ(u), where
f (x)
Z Z
g2 (x) g1 (x)
u = e−G y 1−k − (1 − k) e−G dx, G = (1 − k) dx.
f (x) f (x)
∂w ∂w
13. f (x) + g1 (x) + g2 (x)eλy = h(x).
∂x ∂y
Z
h(x)
General solution: w = dx + Φ(u), where
f (x)
Z Z
−λy g2 (x) g1 (x)
u=e E(x) + λ E(x) dx, E(x) = exp λ dx .
f (x) f (x)
∂x
+ g(x, y) ∂w
∂y
= h(x, y) 87
∂w ∂w
14. f (x)y k + g(x) = h(x).
∂x ∂y
Z x
h(t) − k
General solution: w = Φ(u) + u + E(t) k+1 dt, where
x0 f (t)
Z
k+1 g(x)
u=y − E(x), E(x) = (k + 1) dx, x0 may be taken as arbitrary.
f (x)
∂w ∂w
15. f (x)y k + g1 (x)y k+1 + g0 (x)
∂x ∂y
= h2 (x)y 3k+2 + h1 (x)y 2k+1 + h0 (x)y k .
The substitution z = y k+1 leads to an equation of the form 1.2.7.11:
∂w ∂w
f (x) + (k + 1) g1 (x)z + g0 (x) = h2 (x)z 2 + h1 (x)z + h0 (x).
∂x ∂z
∂w ∂w
16. f (x)eλy + g(x) = h(x).
∂x ∂y
General solution:
Z x Z
h(t) dt λy g(x)
w = Φ(u) + , u=e − E(x), E(x) = λ dx,
x0 f (t) u + E(t) f (x)
where x0 may be taken as arbitrary.
∂w ∂w
17. a +b = f (x) + g(y).
∂x ∂y
Z Z
1 1
General solution: w = f (x) dx + g(y) dy + Φ(bx − ay).
a b
∂w ∂w
18. +a = f (x)g(y).
∂x ∂y
Z x
General solution: w = f (t)g(y − ax + at) dt + Φ(y − ax), where x0 may be taken
x0
as arbitrary.
∂w ∂w
19. + ay + f (x) = g(x)h(y).
∂x ∂y
General solution:
Z Z Z
ax ax −ax −ax
w = g(x) h e u+e f (x)e dx dx+Φ(u), u=e y− f (x)e−ax dx.
In the integration, u is considered a parameter.

∂w ∂w
20. f (x) + g(y) = h1 (x) + h2 (y).
∂x ∂y
General solution:
Z Z Z Z
h1 (x) h2 (y) dx dy
w= dx + dy + Φ − .
f (x) g(y) f (x) g(y)
∂w ∂w
21. f1 (x) + f2 (x)y + f3 (x)y k = g(x)h(y).
∂x ∂y
Z
f2 (x)
The transformation ξ = dx, η = y 1−k leads to an equation of the form 1.2.7.19:
f1 (x)
∂w ∂w
+ (1 − k)η + F (ξ) = G(ξ)H(η),
∂ξ ∂η
f3 (x) g(x)
where F (ξ) = (1 − k) , G(ξ) = , and H(η) = h(y).
f2 (x) f2 (x)
∂w ∂w
22. f1 (x)g1(y) + f2 (x)g2 (y) = h1 (x)h2 (y).
∂x ∂y
Z Z
f2 (x) g1 (y)
The transformation ξ = dx, η = dy leads to an equation of the form
f1 (x) g2 (y)
1.2.7.18:
∂w ∂w h1 (x) h2 (y)
+ = F (ξ)G(η), where F (ξ) = , G(η) = .
∂ξ ∂η f2 (x) g1 (y)
∂w ∂w
23. f1 (x)g1(y) + f2 (x)g2 (y) = h1 (x) + h2 (y).
∂x ∂y
This is a special case of equation 1.2.7.38 with h(x, y) = h1 (x) + h2 (y).
◮ Equations contain arbitrary functions of complicated arguments.
∂w ∂w
24. a +b = f (αx + βy).
∂x ∂y
General solution:
 Z
 1
 f (z) dz + Φ(bx − ay) if aα + bβ =
6 0,
w = aα + bβ
 1 xf (αx + βy) + Φ(bx − ay)

if aα + bβ = 0,
a
where z = αx + βy.

∂w ∂w y y
25. x +y = xf+ yg .
∂x ∂y x x
y y
y
General solution: w = xf + yg +Φ .
x x x
∂x
+ g(x, y) ∂w
∂y
= h(x, y) 89
∂w ∂w
26. x +y = f (x2 + y 2 ).
∂x ∂y
y 1 Z dξ
General solution: w = Φ + f (ξ) , where ξ = x2 + y 2 .
x 2 ξ

∂w ∂w y
27. x +y + g(x2 + y 2 ).
= xf
∂x ∂y x
y y 1 Z dξ
General solution: w = Φ + xf + g(ξ) , where ξ = x2 + y 2 .
x x 2 ξ
∂w ∂w
28. ax + by = xk f (xn y m ).
∂x ∂y
General solution:
 Z an+bm m
 1

 xk−1 f x a u a dx + Φ(u) if an 6= −bm,

a
1 k
w=
 x f (xn y m ) + Φ(u) if an = −bm, k 6= 0,

 ak
 1 f (xn y m ) ln |x| + Φ(u)

if an = −bm, k = 0,
a
where u = y a x−b . In the integration, u is considered a parameter.
∂w ∂w
29. mx + ny= f (axn + by m ).
∂x ∂y
Z
1 dξ
General solution: w = Φ y m x−n + f (ξ) , where ξ = axn + by m .
nm ξ
∂w ∂w
30. x2 = y k f (αx + βy).
+ xy
∂x ∂y
Z y
yk k−2
General solution: w = z f (z) dz + Φ , where z = αx + βy.
x(αx + βy)k−1 x
f (x) ∂w g(y) ∂w
31. + = h f (x) + g(y) .
f ′ (x) ∂x g ′ (y) ∂y
General solution:
Z
dξ g(y)
w = Φ(u) + h(ξ) , where u= , ξ = f (x) + g(y).
ξ f (x)
◮ Equations contain arbitrary functions of two variables.

∂w ∂w
32. +a = f (x, y).
∂x ∂y
Z x
General solution: w = f (t, y − ax + at) dt + Φ(y − ax), where x0 may be taken as
x0
arbitrary.
∂w ∂w
33. ax + by = f (x, y).
∂x ∂y
General solution:
Z
1 1
w= f x, u1/a xb/a dx + Φ(u), where u = y a x−b .
a x
∂w ∂w
34. f (x) + g(x)y = h(x, y).
∂x ∂y
General solution:
Z Z
h(x, uG) y g
w = Φ(u) + dx, where u= , G = exp dx .
f G f
∂w ∂w
35. f (x) + g1 (x)y + g0 (x) = h(x, y).
∂x ∂y
General solution:
Z
h(x, uG + Q) y−Q
w = Φ(u) + dx, u= ,
f G
Z Z
g1 g0 dx
where G = exp dx and Q = G . In the integration, u is considered a
f fG
parameter.
∂w ∂w
36. f (x) + g1 (x)y + g0 (x)y k = h(x, y).
∂x ∂y
For k = 1, see equation 1.2.7.34. For k 6= 1, the substitution ξ = y 1−k leads to an equation
of the form 1.2.7.35:
∂w ∂w 1
f (x) + (1 − k) g1 (x)ξ + g0 (x) = h x, ξ 1−k .
∂x ∂ξ
∂w ∂w
37. f (x) + g1 (x) + g0 (x)eλy = h(x, y).
∂x ∂y
The substitution z = e−λy
leads to an equation of the form 1.2.7.35:

∂w ∂w 1
f (x) − λ g1 (x)z + g0 (x) = h x, − ln z .
∂x ∂z λ
∂w ∂w
38. f1 (x)g1(y) + f2 (x)g2 (y) = h(x, y).
∂x ∂y
Z Z
f2 (x) g1 (y)
f1 (x) g2 (y)
1.2.7.32:
∂w ∂w h(x, y)
+ = F (ξ, η), where F (ξ, η) = .
∂ξ ∂η f2 (x)g1 (y)
∂x
+ g(x, y) ∂w
∂y
= h(x, y)w 91

∂w ∂w
f (x, y) + g(x, y) = h(x, y)w
∂x ∂y

∂w ∂w
1. a +b = cw.
∂x ∂y
Two forms of the representation of the general solution:
c c
w = exp x Φ(bx − ay), w = exp y Φ(bx − ay).
a b
∂w ∂w
2. a +y = bw.
∂x ∂y

General solution: w = |y|b Φ |y|a e−x .
∂w ∂w
3. x +y = aw.
∂x ∂y
Differential equation for homogeneous functions of order a with two independent variables.
General solution: w = xa Φ(y/x).

∂w ∂w
4. x a −b = cyw.
∂x ∂y
n c o
General solution: w = exp 2 (bx + ay) ln x − bx Φ(bx + ay).
a
∂w ∂w
5. x +y = axw.
∂x ∂y
y
General solution: w = eax Φ .
x
∂w ∂w
6. (x − a) + (y − b) = w.
∂x ∂y
Differential equation of a conic surface with the vertex at the point (a, b, 0).
y−b
General solution: w = (x − a)Φ .
x−a
∂w ∂w
7. (y + ax) + (y − ax) = bw.
∂x ∂y
General solution:
Z
b p a+1 dv
w= ξ a+1 Φ ln ξ+ ,
2 v 2 + (a − 1)v + a
where ξ = y 2 + (a − 1)xy + ax2 and v = y/x.

∂w ∂w
8. a +b = (x2 − y 2 )w.
∂x ∂y

1 3 3
General solution: w = exp (bx − ay ) Φ(bx − ay).
3ab
∂w ∂w
9. x2 + axy = by 2 w.
∂x ∂y
General solution:


 b y2
exp Φ(x−a y) if a 6= 12 ,
2a
2 − 1 x

w=

 y
exp b ln x Φ(x−1/2 y) if a = 12 .
x
∂w ∂w
10. ax2 + by 2
= (x + cy)w.
∂x ∂y

1/a c/b b a
General solution: w = x y Φ − .
x y
∂w ∂w
11. x2 + ay 2 = (bx2 + cxy + dy 2 )w.
∂x ∂y
2
dy + abxy − bx2 cxy x x − ay
General solution: w = exp − ln Φ .
ay − x ay − x y xy
∂w ∂w
12. y2 + ax2 = (bx2 + cy 2 )w.
∂x ∂y

b
General solution: w = exp cx + y Φ ax3 − y 3 .
a
∂w ∂w
13. xy + ay 2 = (bx + cy + d)w.
∂x ∂y
General solution:

 (1 − a)d − abx
x exp
 c
Φ x−a y if a 6= 1,
a(a− 1)y
w=

 bx d y
exp + c ln |x| − Φ if a = 1.
y y x
∂w ∂w
14. x(ay + b) + (ay 2 − bx) = ayw.
∂x ∂y

ax − b x + y
General solution: w = (x + y)Φ
+ a ln .
x+y x
∂w ∂w
15. x(ky − x + a) − y(kx − y + a) = b(y − x)w.
∂x ∂y

b (x + y − a)k
General solution: w = (x + y − a) Φ .
xy
∂x
+ g(x, y) ∂w
∂y
= h(x, y)w 93
◮ Coefficients of equations contain other power-law functions.
∂w ∂w
16. a +b = (cx3 + dy 3 )w.
∂x ∂y

bcx4 + ady 4
General solution: w = exp Φ(bx − ay).
4ab
∂w ∂w p
17. x +y =a x2 + y 2 w.
∂x ∂y
p y
General solution: w = exp a x2 + y 2 Φ .
x
∂w ∂w
18. x2 + xy = y 2 (ax + by)w.
∂x ∂y

(ax + by)y 2 y
General solution: w = exp Φ .
2x x
∂w ∂w
19. x2 y + axy 2 = (bxy + cx + dy + k)w.
∂x ∂y
General solution:

 k d c

x b
exp − − − Φ x−a y if a 6= −1,
w= (a +1)xy x ay

 k c d
exp + b ln |x| + − Φ(xy) if a = −1.
xy y y
∂w ∂w
20. axy 2 + bx2 y = (any 2 + bmx2 )w.
∂x ∂y
General solution: w = xn y m Φ(ay 2 − bx2 ).
∂w ∂w
21. x3 + ay 3 = x2 (bx + cy)w.
∂x ∂y
General solution:
r r 2 2
x2 y 2 x x x − ay 2
w = exp c ln − a + + bx Φ .
x2 − ay 2 y 2 y x2 y 2
∂w ∂w
22. a +b = (cxn + dy m )w.
∂x ∂y

c n+1 d m+1
General solution: w = Φ(bx − ay) exp x + y .
a(n + 1) b(m + 1)
∂w ∂w
23. a +b = cxn yw.
∂x ∂y
General solution:


 c[a(n + 2)y − bx]xn+1

 exp Φ(bx − ay) if n 6= −1, −2;

 a2 (n + 1)(n + 2)

bc c
w = exp 2 x(1 − ln x) + y ln x Φ(bx − ay) if n = −1;

 a a

 bc cy


exp 2 (1 + ln x) − Φ(bx − ay) if n = −2.
a ax
∂w ∂w
24. x +y = a(x2 + y 2 )k w.
∂x ∂y
h a i y
General solution: w = exp (x2 + y 2 )k Φ .
2k x
∂w ∂w
25. ax + by = cxn y m w.
∂x ∂y
General solution:

 c

exp n
x y m
Φ y a x−b if an + bm 6= 0,
w= an + bm
 c n m

exp x y ln x Φ y a x−b if an + bm = 0.
a
∂w ∂w
26. ax + by = (cxn + ky m )w.
∂x ∂y

c n k m
General solution: w = exp x + y Φ y a x−b .
an bm
∂w ∂w
27. mx + ny = (axn + by m )k w.
∂x ∂y

1
General solution: w = exp (ax + by ) Φ y m x−n .
n m k
mnk
∂w ∂w
28. axn + by m = (cxk + dy s )w.
∂x ∂y
This is a special case of equation 1.3.7.19. General solution:

cxk−n+1 dy s−m+1 x1−n y 1−m
w = exp + Φ(u), u= − .
a(k − n + 1) b(s − m + 1) a(1 − n) b(1 − m)
∂w ∂w
29. axn + bxm y = cxk y s + d w.
∂x ∂y
This is a special case of equation 1.3.7.32 with f (x) = axn , g(x) = bxm , and h(x, y) =
cxk y s + d.
∂x
+ g(x, y) ∂w
∂y
= h(x, y)w 95
∂w ∂w
30. axn + bxm y + cxk = sxp y q + d w.
∂x ∂y
∂w ∂w
31. axn + bxm y k = cxp y q + s w.
∂x ∂y
This is a special case of equation 1.3.7.34 with f (x) = axn , g1 (x) ≡ 0, g0 (x) = bxm , and
h(x, y) = cxp y q + s.
∂w ∂w
32. ay k + bxn = cxm + s w.
∂x ∂y
This is a special case of equation 1.3.7.14 with f (x) = a, g(x) = bxn , and h(x) = cxm + s.
∂w ∂w
33. x xn + (2n − 1)y n + y y n + (2n − 1)xn = kn(xn + y n )w.
∂x ∂y

(xn − y n )2
General solution: w = (xn − y n )k Φ .
xy
∂w ∂w
34. x (n − 2)y n − 2xn + y 2y n − (n − 2)xn
∂x ∂y

= [a(n − 2) + 2b]y − [2a + b(n − 2)]xn w.
n
n
a b x + yn
General solution: w = x y Φ .
x2 y 2

∂w ∂w
1. a +b = ceαx+βy w.
∂x ∂y
General solution:


 c αx+βy
exp e Φ(bx − ay) if aα + bβ 6= 0,
w= aα + bβ

 c αx+βy
exp xe Φ(bx − ay) if aα + bβ = 0.
a
∂w ∂w
2. a +b = ceλx + keµy w.
∂x ∂y

c λx k µy
General solution: w = exp e + e Φ(bx − ay).
aλ bµ
∂w ∂w
3. aeλx + beβy
= cw.
∂x ∂y

c −λx
General solution: w = exp − e Φ bβe−λx − aλe−βy .
aλ
∂w ∂w
4. aeλy + beβx = cw.
∂x ∂y

c(βx − λy)
General solution: w = exp λy βx
Φ aβeλy − bλeβx .
aβe − bλe
∂w ∂w
5. aeλx + beβx = ceγy w.
∂x ∂y
This is a special case of equation 1.3.7.33 with f (x) = aeλx , g1 (x) ≡ 0, g0 (x) = beβx , and
h(x, y) = ceγy .
∂w ∂w
6. aeλx + beβy = ceγx + seδy w.
∂x ∂y
This is a special case of equation 1.3.7.19 with f (x) = aeλx , g(y) = beβy , h1 (x) = ceγx ,
and h2 (y) = seδy .
∂w ∂w
7. aeβx + beγx + ceλy = seµx + keδy + p w.
∂x ∂y
h(x, y) = seµx + keδy + p.
∂w ∂w
8. aeβx + beγx + ceλy = seµx+δy + k w.
∂x ∂y
h(x, y) = seµx+δy + k.
∂w ∂w
9. aeβx + beγx+λy = ceµx+δy + k w.
∂x ∂y
This is a special case of equation 1.3.7.35 with f (x) = aeβx , g1 (x) ≡ 0, g0 (x) = beγx , and
h(x, y) = ceµx+δy + k.
∂w ∂w
10. aeλy + beβx = ceµx + k w.
∂x ∂y
This is a special case of equation 1.3.7.15 with f (x) = a, g(x) = beβx , and h(x) = ceµx +k.
∂w ∂w
11. a +b = cyeλx + kxeµy w.
∂x ∂y

c λx b k µy a
General solution: w = exp e y− + e x− Φ(bx − ay).
aλ aλ bµ bµ
∂w ∂w
12. x +y = axeλx+µy w.
∂x ∂y

ax λx+µy y
General solution: w = exp e Φ .
λx + µy x
∂x
+ g(x, y) ∂w
∂y
= h(x, y)w 97
∂w ∂w
13. x +y = ayeλx + bxeµy w.
∂x ∂y

ay λx bx µy y
General solution: w = exp e + e Φ .
λx µy x
∂w ∂w
14. axk + beλy = cxn + s w.
∂x ∂y
This is a special case of equation 1.3.7.13 with f (x) = axk , g1 (x) = 0, g2 (x) = b, and
h(x) = cxn + s.
∂w ∂w
15. ay k + beλx = ceµx + s w.
∂x ∂y
This is a special case of equation 1.3.7.14 with f (x) = a, g(x) = beλx , and h(x) = ceµx +s.
∂w ∂w
16. aeλx + by k = cxn + s w.
∂x ∂y
This is a special case of equation 1.3.7.12 with f (x) = aeλx , g1 (x) = 0, g2 (x) = b, and
h(x) = cxn + s.
∂w ∂w
17. aeλy + bxk = ceµx + s w.
∂x ∂y
This is a special case of equation 1.3.7.15 with f (x) = a, g(x) = bxk , and h(x) = ceµx + s.

∂w ∂w
1. a +b = c sinh(λx) + k sinh(µy) w.
∂x ∂y

c k
General solution: w = exp cosh(λx) + cosh(µy) Φ(bx − ay).
aλ bµ
∂w ∂w
2. a +b = c sinh(λx + µy)w.
∂x ∂y
General solution:


 c
exp cosh(λx + µy) Φ(bx − ay) if aλ + bµ 6= 0,
w= aλ + bµ

 c
exp x sinh(λx + µy) Φ(bx − ay) if aλ + bµ = 0.
a
∂w ∂w
3. x +y = ax sinh(λx + µy)w.
∂x ∂y

ax y
General solution: w = exp cosh(λx + µy) Φ .
λx + µy x
∂w ∂w
4. a + b sinhn (λx) = c sinhm (µx) + s sinhk (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.33 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b sinhn (λx),
and h(x, y) = c sinhm (µx) + s sinhk (βy).
∂w ∂w
5. a + b sinhn (λy) = c sinhm (µx) + s sinhk (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.19 with f (x) = a, g(y) = b sinhn (λy), h1 (x) =
c sinhm (µx), and h2 (y) = s sinhk (βy).

∂w
∂w
6. a +b = c cosh(λx) + k cosh(µy) w.
∂x ∂y

c k
General solution: w = exp sinh(λx) + sinh(µy) Φ(bx − ay).
aλ bµ
∂w ∂w
7. a +b = c cosh(λx + µy)w.
∂x ∂y
General solution:


 c
exp sinh(λx + µy) Φ(bx − ay) if aλ + bµ 6= 0,
w= aλ + bµ

 c
exp x cosh(λx + µy) Φ(bx − ay) if aλ + bµ = 0.
a
∂w ∂w
8. x +y = ax cosh(λx + µy)w.
∂x ∂y

ax y
General solution: w = exp sinh(λx + µy) Φ .
λx + µy x
∂w ∂w
9. a + b coshn (λx) = c coshm (µx) + s coshk (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.33 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b coshn (λx),
and h(x, y) = c coshm (µx) + s coshk (βy).
∂w ∂w
10. a + b coshn (λy) = c coshm (µx) + s coshk (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.19 with f (x) = a, g(y) = b coshn (λy), h1 (x) =
c coshm (µx), and h2 (y) = s coshk (βy).

∂w
∂w
11. a +b = c tanh(λx) + k tanh(µy) w.
∂x ∂y

c k
General solution: w = exp ln cosh(λx) + ln cosh(µy) Φ(bx − ay).
aλ bµ
∂x
+ g(x, y) ∂w
∂y
= h(x, y)w 99
∂w ∂w
12. a +b = c tanh(λx + µy)w.
∂x ∂y
General solution:


 c
exp ln cosh(λx + µy) Φ(bx − ay) if aλ + bµ 6= 0,
w= aλ + bµ

 c
 exp x tanh(λx + µy) Φ(bx − ay) if aλ + bµ = 0.
a
∂w ∂w
13. x +y = ax tanh(λx + µy)w.
∂x ∂y

ax y
General solution: w = exp ln cosh(λx + µy) Φ .
λx + µy x
∂w ∂w
14. a + b tanhn (λx) = c tanhm (µx) + s tanhk (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.33 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b tanhn (λx),
and h(x, y) = c tanhm (µx) + s tanhk (βy).
∂w ∂w
15. a + b tanhn (λy) = c tanhm (µx) + s tanhk (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.19 with f (x) = a, g(y) = b tanhn (λy), h1 (x) =
c tanhm (µx), and h2 (y) = s tanhk (βy).

∂w ∂w
16. a +b = c coth(λx) + k coth(µy) w.
∂x ∂y

c k
General solution: w = exp
ln sinh(λx) +
ln sinh(µy) Φ(bx − ay).
aλ bµ
∂w ∂w
17. a +b = c coth(λx + µy)w.
∂x ∂y
General solution:

 c
exp
 lnsinh(λx + µy) Φ(bx − ay) if aλ + bµ 6= 0,
w= aλ + bµ

 c
exp x coth(λx + µy) Φ(bx − ay) if aλ + bµ = 0.
a
∂w ∂w
18. x +y = ax coth(λx + µy)w.
∂x ∂y

ax y
ln sinh(λx + µy) Φ .
λx + µy x
∂w ∂w
19. a + b cothn (λx) = c cothm (µx) + s cothk (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.33 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b cothn (λx),
and h(x, y) = c cothm (µx) + s cothk (βy).
∂w ∂w
20. a + b cothn (λy) = c cothm (µx) + s cothk (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.19 with f (x) = a, g(y) = b cothn (λy), h1 (x) =
c cothm (µx), and h2 (y) = s cothk (βy).

∂w
∂w
21. a +b = c sinh(λx) + k cosh(µy) w.
∂x ∂y

c k
General solution: w = exp cosh(λx) + sinh(µy) Φ(bx − ay).
aλ bµ
∂w ∂w
22. a +b = tanh(λx) + k coth(µy) w.
∂x ∂y
General solution: w = cosh1/aλ (λx) sinhk/bµ (µy)Φ(bx − ay).

∂w ∂w
23. + a sinh(µy) = b cosh(λx)w.
∂x ∂y

b µy
General solution: w = exp sinh(λx) Φ aµx − lntanh .
λ 2
∂w ∂w
24. + a sinh(µy) = b tanh(λx)w.
∂x ∂y

b/λ µy
General solution: w = cosh (λx)Φ aµx − lntanh .
2
∂w ∂w
25. a sinh(λx) + b cosh(µy) = w.
∂x ∂y

1/aλ λx
µy bµ λx
General solution: w = tanh Φ 2a arctan tanh + lncoth .
2 2 λ 2
∂w ∂w
26. a tanh(λx) + b coth(µy) = w.
∂x ∂y

General solution: w = sinh1/aλ (λx)Φ coshaλ (µy) sinh−bµ (λx) .

∂w ∂w
1. a +b = c ln(λx + βy)w.
∂x ∂y
General solution:

 c(λx + βy)
exp
 ln(λx + βy) − 1 Φ(bx − ay) if aλ 6= −bβ,
w= aλ + bβ

 c
exp x ln(λx + βy) Φ(bx − ay) if aλ = −bβ.
a
∂x
+ g(x, y) ∂w
∂y
= h(x, y)w 101
∂w
∂w
2. a +b = c ln(λx) + k ln(βy) w.
∂x ∂y

c k
General solution: w = exp x ln(λx) − 1 + y ln(βy) − 1 Φ(bx − ay).
a b
∂w ∂w
3. a + b lnn (λx) = c lnm (µx) + s lnk (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.33 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b lnn (λx),
and h(x, y) = c lnm (µx) + s lnk (βy).
∂w ∂w
4. a + b lnn (λy) = c lnm (µx) + s lnk (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.19 with f (x) = a, g(y) = b lnn (λy), h1 (x) =
c lnm (µx), and h2 (y) = s lnk (βy).
∂w ∂w
5. ln(βy) + a ln(λx)
= bw ln(βy).
∂x ∂y

General solution: w = ebx Φ(u), where u = ax 1 − ln(λx) + y ln(βy) − 1 .
∂w ∂w
6. a lnn (λx) + b lnk (βy) = c lnm (γx)w.
∂x ∂y
General solution:
Z Z Z
c lnm (γx) dx dy
w = Φ(u) exp dx , where u = b −a .
a lnn (λx) n
ln (λx) k
ln (βy)

∂w
∂w
7. a +b = cxn + s lnk (λy) w.
∂x ∂y
Z
c n+1 s k
General solution: w = Φ(bx − ay) exp x + ln (λy) dy .
a(n + 1) b
∂w ∂w
8. +a = by 2 + cxn y + s lnk (λx) w.
∂x ∂y
This is a special case of equation 1.3.7.3 with f (x) = b, g(x) = cxn , and h(x) = s lnk (λx).
∂w ∂w
9. +a = b lnk (λx) lnn (βy)w.
∂x ∂y
This is a special case of equation 1.3.7.17 with f (x) = b lnk (λx) and g(y) = lnn (βy).
∂w ∂w
10. + (ay + bxn ) = c lnk (λx)w.
∂x ∂y
This is a special case of equation 1.3.7.6 with f (x) = bxn and g(x) = c lnk (λx).
∂w ∂w
11. ax + by = xk (n ln x + m ln y)w.
∂x ∂y
This is a special case of equation 1.3.7.26 with f (u) = ln u.
∂w ∂w
12. axk + by n = c lnm (λx) + s lnl (βy) w.
∂x ∂y
General solution:
Z Z
c −k m s −n l
w = Φ(u) exp x ln (λx) dx + y ln (βy) dy ,
a b
b a
u= x1−k − y 1−n .
1−k 1−n

∂w ∂w
1. a +b = c sin(λx + µy)w.
∂x ∂y
General solution:


 c
exp − cos(λx + µy) Φ(bx − ay) if aλ + bµ 6= 0,
w= aλ + bµ

 c
exp x sin(λx + µy) Φ(bx − ay) if aλ + bµ = 0.
a
∂w ∂w
2. a +b = c sin(λx) + k sin(µy) w.
∂x ∂y

c k
General solution: w = exp − cos(λx) − cos(µy) Φ(bx − ay).
aλ bµ
∂w ∂w
3. x +y = ax sin(λx + µy)w.
∂x ∂y

ax y
General solution: w = exp − cos(λx + µy) Φ .
λx + µy x
∂w ∂w
4. a + b sinn (λx) = c sinm (µx) + s sink (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.33 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b sinn (λx),
and h(x, y) = c sinm (µx) + s sink (βy).
∂w ∂w
5. a + b sinn (λy) = c sinm (µx) + s sink (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.19 with f (x) = a, g(y) = b sinn (λy), h1 (x) =
c sinm (µx), and h2 (y) = s sink (βy).

∂w ∂w
6. a +b = c cos(λx + µy)w.
∂x ∂y
General solution:


 c
exp sin(λx + µy) Φ(bx − ay) if aλ + bµ 6= 0,
w= aλ + bµ

 c
exp x cos(λx + µy) Φ(bx − ay) if aλ + bµ = 0.
a
∂x
+ g(x, y) ∂w
∂y
= h(x, y)w 103
∂w
∂w
7. a +b = c cos(λx) + k cos(µy) w.
∂x ∂y

c k
General solution: w = exp sin(λx) + sin(µy) Φ(bx − ay).
aλ bµ
∂w ∂w
8. x +y = ax cos(λx + µy)w.
∂x ∂y

ax y
General solution: w = exp sin(λx + µy) Φ .
λx + µy x
∂w ∂w
9. a + b cosn (λx) = c cosm (µx) + s cosk (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.33 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b cosn (λx),
and h(x, y) = c cosm (µx) + s cosk (βy).
∂w ∂w
10. a + b cosn (λy) = c cosm (µx) + s cosk (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.19 with f (x) = a, g(y) = b cosn (λy), h1 (x) =
c cosm (µx), and h2 (y) = s cosk (βy).

∂w ∂w
11. a +b = c tan(λx + µy)w.
∂x ∂y
General solution:

 c

exp −
ln cos(λx + µy) Φ(bx − ay) if aλ 6= −bµ,
w= aλ + bµ

 c
exp x tan(λx + µy) Φ(bx − ay) if aλ = −bµ.
a
∂w ∂w
12. a +b = c tan(λx) + k tan(µy) w.
∂x ∂y

c k
General solution: w = exp −
ln cos(λx) −
ln cos(µy) Φ(bx − ay).
aλ bµ
∂w ∂w
13. x +y = ax tan(λx + µy)w.
∂x ∂y

ax y
General solution: w = exp − lncos(λx + µy) Φ .
λx + µy x
∂w ∂w
14. a + b tann (λx) = c tanm (µx) + s tank (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.33 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b tann (λx),
and h(x, y) = c tanm (µx) + s tank (βy).
∂w ∂w
15. a + b tann (λy) = c tanm (µx) + s tank (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.19 with f (x) = a, g(y) = b tann (λy), h1 (x) =
c tanm (µx), and h2 (y) = s tank (βy).

∂w ∂w
16. a +b = c cot(λx + µy)w.
∂x ∂y
General solution:

 c

exp
ln sin(λx + µy) Φ(bx − ay) if aλ 6= −bµ,
w= aλ + bµ

 c
exp x cot(λx + µy) Φ(bx − ay) if aλ = −bµ.
a
∂w ∂w
17. a +b = c cot(λx) + k cot(µy) w.
∂x ∂y

c k
General solution: w = exp lnsin(λx) + lnsin(µy) Φ(bx − ay).
aλ bµ
∂w ∂w
18. x +y = ax cot(λx + µy)w.
∂x ∂y

ax y
ln sin(λx + µy) Φ .
λx + µy x
∂w ∂w
19. a + b cotn (λx) = c cotm (µx) + s cotk (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.33 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b cotn (λx),
and h(x, y) = c cotm (µx) + s cotk (βy).
∂w ∂w
20. a + b cotn (λy) = c cotm (µx) + s cotk (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.19 with f (x) = a, g(y) = b cotn (λy), h1 (x) =
c cotm (µx), and h2 (y) = s cotk (βy).

∂w ∂w
21. +a = b sin(λx) + k cos(µy) w.
∂x ∂y

k b
General solution: w = exp sin(µy) − cos(λx) Φ(ax − y).
aµ λ
∂w ∂w
22. +a = b sin(λx) + k tan(µy) w.
∂x ∂y

b
General solution: w = exp − cos(λx) cosk/aµ (µy)Φ(ax − y).
λ
∂w ∂w
23. + a sin(µy) = bw tan(λx).
∂x ∂y

−b/λ µy
General solution: w = cos (λx)Φ aµx − lntan .
2
∂x
+ g(x, y) ∂w
∂y
= h(x, y)w 105
∂w ∂w
24. + a tan(µy)= bw sin(λx).
∂x ∂y

b
General solution: w = exp − cos(λx) Φ aµx − lnsin(µy) .
λ
∂w ∂w
25. sin(λx) +a
= bw cos(µy).
∂x ∂y

b λx
General solution: w = exp sin(µy) Φ λy + b lncot .
aµ 2
∂w ∂w
26. cot(λx) +a
= bw tan(µy).
∂x ∂y

General solution: w = cos−b/aµ (µy)Φ λy + b lncos(λx) .


∂w ∂w x y
1. a +b = c arcsin + k arcsin w.
∂x ∂y λ β
General solution:
k
c x p 2 2
y p 2 2
w = exp x arcsin + λ − x + y arcsin + β − y Φ(bx − ay).
a λ b β
∂w ∂w
2. a +b = c arcsin(λx + βy)w.
∂x ∂y
p
c(λx + βy) 1 − (λx + βy)2
w = exp arcsin(λx + βy) + Φ(bx − ay).
aλ + bβ aλ + bβ

c
w = exp x arcsin(λx + βy) Φ(bx − ay).
a
∂w ∂w
3. x +y = ax arcsin(λx + βy)w.
∂x ∂y
p
1 − (λx + βy)2 y
General solution: w = exp ax arcsin(λx + βy) + ax Φ .
λx + βy x
∂w ∂w
4. a + b arcsinn (λx) = c arcsinm (µx) + s arcsink (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.33 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b arcsinn (λx),
and h(x, y) = c arcsinm (µx) + s arcsink (βy).
∂w ∂w
5. a + b arcsinn (λy) = c arcsinm (µx) + s arcsink (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.19 with f (x) = a, g(y) = b arcsinn (λy), h1 (x) =
c arcsinm (µx), and h2 (y) = s arcsink (βy).

∂w ∂w x y
6. a +b = c arccos + k arccos w.
∂x ∂y λ β
General solution:
k
c x p 2 2
y p 2 2
w = exp x arccos − λ − x + y arccos − β − y Φ(bx − ay).
a λ b β
∂w ∂w
7. a +b = c arccos(λx + βy)w.
∂x ∂y

p
c(λx + βy) 1 − (λx + βy)2
w = exp arccos(λx + βy) − Φ(bx − ay).
aλ + bβ aλ + bβ

c
w = exp x arccos(λx + βy) Φ(bx − ay).
a
∂w ∂w
8. x +y = ax arccos(λx + βy)w.
∂x ∂y
p
1 − (λx + βy)2 y
General solution: w = exp ax arccos(λx + βy) − ax Φ .
λx + βy x
∂w ∂w
9. a + b arccosn (λx) = c arccosm (µx) + s arccosk (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.33 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b arccosn (λx),
and h(x, y) = c arccosm (µx) + s arccosk (βy).
∂w ∂w
10. a + b arccosn (λy) = c arccosm (µx) + s arccosk (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.19 with f (x) = a, g(y) = b arccosn (λy), h1 (x) =
c arccosm (µx), and h2 (y) = s arccosk (βy).

∂w ∂w x y
11. a +b = c arctan + k arctan w.
∂x ∂y λ β
General solution:
h
c x λ 2 2
i k h y β 2 2
i
w = exp x arctan − ln λ +x + y arctan − ln β +y Φ(bx−ay).
a λ 2 b β 2
∂x
+ g(x, y) ∂w
∂y
= h(x, y)w 107
∂w ∂w
12. a +b = c arctan(λx + βy)w.
∂x ∂y

c(λx + βy) ln[1 + (λx + βy)2 ]
w = exp arctan(λx + βy) − Φ(bx − ay).
aλ + bβ 2(aλ + bβ)

hc i
w = exp x arctan(λx + βy) Φ(bx − ay).
a
∂w ∂w
13. x +y = ax arctan(λx + βy)w.
∂x ∂y
General solution:

ax 2 x2 y
w = exp ax arctan(λx + βy) − ln x + 2
Φ .
∂w ∂w
14. a + b arctann (λx) = c arctanm (µx) + s arctank (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.33 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b arctann (λx),
and h(x, y) = c arctanm (µx) + s arctank (βy).
∂w ∂w
15. a + b arctann (λy) = c arctanm (µx) + s arctank (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.19 with f (x) = a, g(y) = b arctann (λy), h1 (x) =
c arctanm (µx), and h2 (y) = s arctank (βy).

∂w ∂w x y
16. a +b = c arccot + k arccot w.
∂x ∂y λ β
General solution:
h
c x λ 2 2
i k h y β 2 2
i
w = exp x arccot + ln λ +x + y arccot + ln β +y Φ(bx−ay).
a λ 2 b β 2
∂w ∂w
17. a +b = c arccot(λx + βy)w.
∂x ∂y

c(λx + βy) ln[1 + (λx + βy)2 ]
w = exp arccot(λx + βy) + Φ(bx − ay).
aλ + bβ 2(aλ + bβ)

hc i
w = exp x arccot(λx + βy) Φ(bx − ay).
a
∂w ∂w
18. x +y = ax arccot(λx + βy)w.
∂x ∂y
General solution:

ax 2 x2 y
w = exp ax arccot(λx + βy) + ln x + 2
Φ .
∂w ∂w
19. a + b arccotn (λx) = c arccotm (µx) + s arccot k (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.33 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b arccotn (λx),
and h(x, y) = c arccotm (µx) + s arccotk (βy).
∂w ∂w
20. a + b arccotn (λy) = c arccotm (µx) + s arccot k (βy) w.
∂x ∂y
This is a special case of equation 1.3.7.19 with f (x) = a, g(y) = b arccotn (λy), h1 (x) =
c arccotm (µx), and h2 (y) = s arccotk (βy).

∂w ∂w
1. a +b = f (x)w.
∂x ∂y
Z
1
General solution: w = exp f (x) dx Φ(bx − ay).
a
∂w ∂w
2. +a = f (x)yw.
∂x ∂y
Z x
General solution: w = exp (y − ax + at)f (t) dt Φ(y − ax), where x0 may be chosen
x0
arbitrarily.
∂w ∂w
3. +a = f (x)y 2 + g(x)y + h(x) w.
∂x ∂y
General solution:

w = exp ϕ(x)y 2 + ψ(x)y + χ(x) Φ(y − ax),
where
Z Z Z

ϕ(x) = f (x) dx, ψ(x) = g(x) − 2aϕ(x) dx, χ(x) = h(x) − aψ(x) dx.
∂w ∂w
4. +a = f (x)y k w.
∂x ∂y
Z x
General solution: w = exp (y − ax + at)k f (t) dt Φ(y − ax), where x0 can be chosen
x0
arbitrarily.
∂x
+ g(x, y) ∂w
∂y
= h(x, y)w 109
∂w ∂w
5. +a = f (x)eλy w.
∂x ∂y
Z
λ(y−ax) aλx
General solution: w = exp e f (x)e dx Φ(y − ax).
∂w ∂w
6. + ay + f (x) = g(x)w.
∂x ∂y
Z Z
−ax
General solution: w = exp g(x) dx Φ(u), where u = e y − f (x)e−ax dx.
∂w ∂w
7. + ay + f (x) = g(x)y k w.
∂x ∂y
This is a special case of equation 1.3.7.18 with h(y) = y k .

∂w ∂w
8. f (x) + yk = g(x)w.
∂x ∂y
General solution:
 Z
 1 dx
Z 
 y 1−k + if k 6= 1,
g(x) k − 1 Z f (x)

w = exp dx Φ(u), where u= dx
f (x) 

y exp − if k = 1.
f (x)
∂w ∂w
9. f (x) + (y + a) = (by + c)w.
∂x ∂y
Z
dx
General solution: w = (y + a)c−ab eby Φ(u), where u = (y + a) exp − .
f (x)
∂w ∂w
10. f (x) + (y + ax) = g(x)w.
∂x ∂y
General solution:
Z Z Z
g(x) −z xe−z dx
w = exp dx Φ e y − a dx , where z= .
f (x) f (x) f (x)
∂w ∂w
11. f (x) + g1 (x)y + g0 (x) = h2 (x)y 2 + h1 (x)y + h0 (x) w.
∂x ∂y
General solution:
Z
g0
w = exp ϕ(x)y 2 + ψ(x)y + χ(x) Φ(u), u=e −G
y− e−G dx,
f
where Z Z
−2G h2 2G g1
ϕ(x) = e e dx, G = G(x) = dx,
f f
Z Z
h1 − 2g0 ϕ h0 − g0 ψ
ψ(x) = e−G eG dx, χ(x) = dx.
f f

∂w ∂w
12. f (x) + g1 (x)y + g2 (x)y k = h(x)w.
∂x ∂y
Z
h(x)
General solution: w = exp dx Φ(u), where
f (x)
Z Z
−G 1−k −G g2 (x) g1 (x)
u=e y − (1 − k) e dx, G = (1 − k) dx.
f (x) f (x)
∂w ∂w
13. f (x) + g1 (x) + g2 (x)eλy = h(x)w.
∂x ∂y
Z
h(x)
General solution: w = exp dx Φ(u), where
f (x)
Z Z
−λy g2 (x) g1 (x)
u=e E(x) + λ E(x) dx, E(x) = exp λ dx .
f (x) f (x)
∂w ∂w
14. f (x)y k + g(x)
= h(x)w.
∂x ∂y
Z x
h(t) − k
General solution: w = Φ(u) exp u + E(t) k+1 dt , where
x0 f (t)
Z
k+1 g(x)
u=y − E(x), E(x) = (k + 1) dx, where x0 may be chosen arbitrarily.
f (x)
∂w ∂w
15. f (x)eλy + g(x) = h(x)w.
∂x ∂y
General solution:
Z x Z
h(t) dt g(x)
w = Φ(u) exp , u = eλy − E(x), E(x) = λ dx,
x0 f (t) u + E(t) f (x)
where x0 may be chosen arbitrarily.

∂w
∂w
16. a +b = f (x) + g(y) w.
∂x ∂y
Z Z
1 1
General solution: w = exp f (x) dx + g(y) dy Φ(bx − ay).
a b
∂w ∂w
17. +a = f (x)g(y)w.
∂x ∂y
Z x
General solution: w = exp f (t)g(y − ax + at) dt Φ(y − ax), where x0 may be taken
x0
as arbitrary.
∂x
+ g(x, y) ∂w
∂y
= h(x, y)w 111
∂w ∂w
18. + [ay + f (x)] = g(x)h(y)w.
∂x ∂y
The substitutions w = ±eu lead to an equation of the form 1.2.7.19:
∂u ∂u
+ [ay + f (x)] = g(x)h(y).
∂x ∂y
∂w ∂w
19. f (x) + g(y) = h1 (x) + h2 (y) w.
∂x ∂y
General solution:
Z Z Z Z
h1 (x) h2 (y) dx dy
w = exp dx + dy Φ dx − dy .
f (x) g(y) f (x) g(y)
∂w ∂w
20. f1 (x) + f2 (x)y + f3 (x)y k = g(x)h(y)w.
∂x ∂y
Z
f2 (x)
The transformation ξ = dx, η = y 1−k leads to an equation of the form 1.3.7.18:
f1 (x)
∂w ∂w
+ (1 − k)η + F (ξ) = G(ξ)H(η)w,
∂ξ ∂η
f3 (x) g(x)
where F (ξ) = (1 − k) , G(ξ) = , and H(η) = h(y).
f2 (x) f2 (x)
∂w ∂w
21. f1 (x)g1(y) + f2 (x)g2 (y) = h1 (x)h2 (y)w.
∂x ∂y
Z Z
f2 (x) g1 (y)
f1 (x) g2 (y)
1.3.7.17:
∂w ∂w h1 (x) h2 (y)
+ = F (ξ)G(η)w, where F (ξ) = , G(η) = .
∂ξ ∂η f2 (x) g1 (y)
∂w ∂w
22. f1 (x)g1(y) + f2 (x)g2 (y) = h1 (x) + h2 (y) w.
∂x ∂y
This is a special case of equation 1.3.7.36 with h(x, y) = h1 (x) + h2 (y).
◮ Equations contain arbitrary functions of complicated arguments.

∂w ∂w
23. a +b = f (αx + βy)w.
∂x ∂y
General solution:
 Z

 1
exp f (u) du Φ(bx − ay) if aα + bβ =
6 0,
w= aα + bβ

 1
exp xf (αx + βy) Φ(bx − ay) if aα + bβ = 0,
a
where u = αx + βy.

∂w ∂w y
24. x +y = xf
w.
∂x ∂y x

y y
General solution: w = exp xf Φ .
x x
∂w ∂w
25. x +y = f (x2 + y 2 )w.
∂x ∂y
y Z
1 dξ
General solution: w = Φ exp f (ξ) , where ξ = x2 + y 2 .
x 2 ξ
∂w ∂w
26. ax + by = xk f (xn y m )w.
∂x ∂y
General solution:
 Z an+bm m

 1 k−1

 exp x f x a u a dx Φ(u) if an 6= −bm;

 a
 1 k n m
w = exp x f (x y ) Φ(u) if an = −bm, k 6= 0;

 ak

 1


exp f (xn y m ) ln x Φ(u) if an = −bm, k = 0,
a
∂w ∂w
27. mx + ny= f (axn + by m )w.
∂x ∂y
Z
m −n
1 dξ
General solution: w = Φ y x exp f (ξ) , where ξ = axn + by m .
nm ξ
∂w ∂w
28. x2 + xy = y k f (αx + βy)w.
∂x ∂y
General solution:
Z
yk k−2 y
w = exp k−1
z f (z) dz Φ , where z = αx + βy.
x(αx + βy) x
f (x) ∂w g(y) ∂w
29. + = h f (x) + g(y) w.
f ′ (x) ∂x g ′ (y) ∂y
General solution:
Z
dξ g(y)
w = Φ(u) exp h(ξ) , where u = , ξ = f (x) + g(y).
ξ f (x)

∂w ∂w
30. +a = f (x, y)w.
∂x ∂y
Z x
General solution: w = exp f (t, y − ax + at) dt Φ(y − ax), where x0 may be taken
x0
as arbitrary.
∂x
+ g(x, y) ∂w
∂y
= h(x, y)w 113
∂w ∂w
31. ax + by = f (x, y)w.
∂x ∂y
General solution:
Z
1 1 1/a b/a

w = exp f x, u x dx Φ(u), where u = y a x−b .
a x
∂w ∂w
32. f (x) + g(x)y = h(x, y)w.
∂x ∂y
General solution:
Z Z
h(x, uG) y g
w = Φ(u) exp dx , where u= , G = exp dx .
f (x) G f
∂w ∂w
33. f (x) + g1 (x)y + g0 (x) = h(x, y)w.
∂x ∂y
General solution:
Z
h(x, uG + Q) y−Q
w = Φ(u) exp dx , u = ,
f (x) G
Z Z
g1 g0 dx
where G = exp dx and Q = G . In the integration, u is considered a
f fG
parameter.
∂w ∂w
34. f (x) + g1 (x)y + g0 (x)y k = h(x, y)w.
∂x ∂y
∂w ∂w 1
f (x) + (1 − k) g1 (x)ξ + g0 (x) = h x, ξ 1−k w.
∂x ∂ξ
∂w ∂w
35. f (x) + g1 (x) + g0 (x)eλy = h(x, y)w.
∂x ∂y
The substitution z = e−λy
∂w ∂w 1
f (x) − λ g1 (x)z + g0 (x) = h x, − ln z w.
∂x ∂z λ
∂w ∂w
36. f1 (x)g1(y) + f2 (x)g2 (y) = h(x, y)w.
∂x ∂y
Z Z
f2 (x) g1 (y)
f1 (x) g2 (y)
1.3.7.30:
∂w ∂w h(x, y)
+ = F (ξ, η)w, where F (ξ, η) = .
∂ξ ∂η f2 (x)g1 (y)

∂w ∂w
f (x, y) + g(x, y) = h1 (x, y)w + h0(x, y)
∂x ∂y

∂w ∂w
1. a +b = cw + d.
∂x ∂y
d
General solution: w = − + ecx/a Φ(bx − ay).
c
∂w ∂w
2. (x − a) + (y − b) = w − c.
∂x ∂y
Differential equation of a conic surface with the vertex at the point (a, b, c).
y−b
General solution: w = c + (x − a)Φ .
x−a
∂w ∂w
3. (ax + b) + (cx + d) = αw + β.
∂x ∂y
General solution:


− β + (ax + b)α/a Φ a(cx − ay) + (ad − bc) ln |ax + b|

if a 6= 0,
w= α
 β
− + eαx/b Φ x(cx + 2d) − 2by if a = 0.
α
∂w ∂w
4. (ax + b) + (cy + d) = αw + β.
∂x ∂y
General solution:


− β + (ax + b)α/a Φ (ax + b)−c/a (cy + d)

if a 6= 0,
w= α
 β
− + eαx/b Φ (cy + d)e−cx/b if a = 0.
α
∂w ∂w
5. (ax + b) + (cy + d) = αw + βy + γx.
∂x ∂y
1◦ . General solution for a 6= 0, a 6= α, and c 6= α:
γ(αx + b) β(αy + d)
w= − + (ax + b)α/a Φ (ax + b)−c/a (cy + d) .
α(a − α) α(α − c)
2◦ . General solution for a 6= 0, a = α, and c 6= α:

γ b + (ax + b) ln |ax + b| β(ay + d)
w= 2
− + (ax + b)Φ (ax + b)−c/a (cy + d) .
a a(a − c)
∂x
+ g(x, y) ∂w
∂y
= h1 (x, y)w + h0 (x, y) 115
3◦ . General solution for a 6= 0 and a = c = α:

bγ + dβ[γ(ax + b) + β(ay + d)] ln |ax + b| ay + d
w= + (ax + b)Φ .
a2 ax + b
4◦ . General solution for a = 0 and c 6= α:
γ(αx + b) β(αy + d)
w=− 2
− + eαx/b Φ (cy + d)e−cx/b .
α α(α − c)
5◦ . General solution for a = 0 and c = α:

(dβ − bγ)(cx + b) β
w= 2
+ xy + ecx/b Φ (cy + d)e−cx/b .
bc b
∂w ∂w
6. (ax + b) + (cx + dy) = αw + β.
∂x ∂y
1◦ . General solution for a 6= 0 and a 6= d:
β
w=− + (ax + b)α/a Φ c(dx + b) + d(d − a)y (ax + b)−d/a .
α
2◦ . General solution for a 6= 0 and a = d:

β α/a bc − a2 y
w = − + (ax + b) Φ + c ln |ax + b| .
α ax + b
3◦ . General solution for a = 0:

β
w=− + eαx/b Φ bc + d(cx + dy) e−dx/b .
α
∂w ∂w
7. (a1 x+a0 ) +(b2 y +b1 x+b0 ) = (c2 y +c1 x+c0 )w +k2 y +k1 x+k0 .
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a1 x + a0 , g1 (x) = b2 , g0 (x) =

b1 x + b0 , h(x, y) = c2 y + c1 x + c0 , and F (x, y) = k2 y + k1 x + k0 .
∂w ∂w
8. ay + (b1 x + b0 ) = (c1 x + c0 )w + s1 x + s0 .
∂x ∂y
This is a special case of equation 1.4.7.11 with k = 1, f1 (x) = a, f2 (x) = b1 x + b0 ,

g(x) = c1 x + c0 , and h(x) = s1 x + s0 .
∂w ∂w
9. a +b = cw + βxy + γ.
∂x ∂y
γ β
General solution: w = − − 3 (cx + a)(cy + b) + ab + ecx/a Φ(bx − ay).
c c
∂w ∂w
10. a +b = cw + x(βx + γy) + δ.
∂x ∂y
General solution:
δ 1
w = − − 3 β(cx + a)2 + γ(cx + a)(cy + b) + a(aβ + bγ) + ecx/a Φ(bx − ay).
c c
∂w ∂w
11. x +y = w + ax2 + by 2 + c.
∂x ∂y
General solution: w = ax2 + by 2 − c + xΦ(y/x).
∂w ∂w
12. ax + by = cw + x(βx + γy) + δ.
∂x ∂y
1◦ . General solution for c 6= 2a and c 6= a + b:
δ β γ
w=− + x2 + xy + xc/a Φ y|x|−b/a .
c 2a − c a+b−c
2◦ . General solution for c = 2a and a 6= b:
δ β γ
w=− + x2 ln |x| − xy + x2 Φ y|x|−b/a .
c a a−b
3◦ . General solution for c = a + b and a 6= b:
δ β γ
w=− + x2 + xy ln |x| + xΦ y|x|−b/a .
c a−b a
4◦ . General solution for c = 2a and a = b:
δ 1 y
w = − + x(βx + γy) ln |x| + Φ .
c a x
∂w ∂w
13. ay + (b2 x2 + b1 x + b0 ) = (c2 x2 + c1 x + c0 )w + s2 x2 + s1 x + s0 .
∂x ∂y
This is a special case of equation 1.4.7.11 with k = 1, f1 (x) = a, f2 (x) = b2 x2 + b1 x + b0 ,
g(x) = c2 x2 + c1 x + c0 , and h(x) = s2 x2 + s1 x + s0 .
∂w ∂w
14. ay 2 + (b1 x2 + b0 ) = (c1 x2 + c0 )w + s1 x2 + s0 .
∂x ∂y
This is a special case of equation 1.4.7.11 with k = 2, f1 (x) = a, f2 (x) = b1 x2 + b0 ,
g(x) = c1 x2 + c0 , and h(x) = s1 x2 + s0 .
∂w ∂w
15. (a1 x2 + a0 ) + (y + b2 x2 + b1 x + b0 )
∂x ∂y
= (c2 y + c1 x + c0 )w + k22 y 2 + k12 xy + k11 x2 + k0 .
This is a special case of equation 1.4.7.22 with f (x) = a1 x2 + a0 , g1 (x) = 1, g0 (x) =
b2 x2 + b1 x + b0 , h(x, y) = c2 y + c1 x + c0 , and F (x, y) = k22 y 2 + k12 xy + k11 x2 + k0 .
∂w ∂w
16. (a1 x2 + a0 ) + (b2 y 2 + b1 xy)
∂x ∂y
= (c2 y 2 + c1 x2 )w + s22 y 2 + s12 xy + s11 x2 + s0 .
This is a special case of equation 1.4.7.23 with k = 2, f (x) = a1 x2 + a0 , g1 (x) = b1 x,
g0 (x) = b2 , h(x, y) = c2 y 2 + c1 x2 , and F (x, y) = s22 y 2 + s12 xy + s11 x2 + s0 .
∂x
+ g(x, y) ∂w
∂y
= h1 (x, y)w + h0 (x, y) 117
◮ Coefficients of equations contain square roots.

∂w ∂w √
17. ax + by = αw + β xy + γ.
∂x ∂y
1◦ . General solution for 2α 6= a + b:
2β √ γ
w= xy − + xα/a Φ y|x|−b/a .
a + b − 2α α
◦
2 . General solution for 2α = a + b:
β√ 2γ √
w= xy ln |x| − + xy Φ y|x|−b/a .
a a+b
3◦ . General solution for α = a = −b:
1 √
w= β xy + γ + xΦ(xy).
a
∂w ∂w √
18. ax + by = λ xy w + βxy + γ.
∂x ∂y
1◦ . General solution for b 6= −a:

β√ β(a + b) 2λ √
w=− xy − 2
+ exp xy Φ x−b/a y .
λ 2λ a+b
2◦ . General solution for b = −a:

β√ λ√
w=− xy + exp xy ln |x| Φ(xy).
λ a
∂w ∂w √
19. ay + bx = αw + β x + γ.
∂x ∂y
This is a special case of equation 1.4.7.11 with k = 1, f1 (x) = a, f2 (x) = bx, g(x) = α,
√
and h(x) = β x + γ.
∂w √ ∂w √
20. ay +b x = αw + β x + γ.
∂x ∂y
√
This is a special case of equation 1.4.7.11 with k = 1, f1 (x) = a, f2 (x) = b x, g(x) = α,
√
√ ∂w √ ∂w
21. a x +b y = αw + βx + γy + δ.
∂x ∂y
General solution:
√ √
aβ x + bγ y βx + γy + δ a2 β + b2 γ 2α √ √ √
w=− − − + exp x Φ b x − a y .
α2 α 2α3 a
√ ∂w √ ∂w √
22. a x +b y = αw + β x + γ.
∂x ∂y
√
β x+γ aβ 2α √ √ √
General solution: w = − − + exp x Φ b x − a y .
α 2α2 a
√ ∂w √ ∂w √
23. a y +b x = αw + β x + γ.
∂x ∂y
√
This is a special case of equation 1.4.7.11 with k = 1/2, f1 (x) = a, f2 (x) = b x, g(x) = α,
√

∂w ∂w
24. a +b = cw + kxn y m .
∂x ∂y
Two forms of the representation of the general solution:
c k
Z c
n m
w = exp x Φ(bx − ay) + m+1 x (bx − u) exp − x dx ,
a a a
c k
Z c

w = exp y Φ(bx − ay) + n+1 y m (ay + u)n exp − y dy ,
b b b
where u = bx − ay. In the integration, u is considered a parameter.
∂w ∂w
25. a +y = bw + cxn y m .
∂x ∂y
General solution:
Z
b a −x m−b−1 n
w = y Φ(y e ) + c y (a ln y − ln u) dy , where u = y a e−x .

∂w ∂w
26. x +y = axw + bxn y m .
∂x ∂y
Z
y
General solution: w = eax Φ + bx−m y m xm+n−1 e−ax dx .
x
∂w ∂w p
27. x +y = a x2 + y 2 w + bxn y m .
∂x ∂y
General solution:
p y Z p y
2 2 −m m m+n−1 2
w = exp a x +y Φ +bx y x exp −ax 1+u dx , u= .
x x
∂w ∂w
28. ax + by = cxn y m w + pxk y s .
∂x ∂y
1◦ . General solution for an + bm 6= 0:
h i
c n m

w = exp x y Φ y a x−b + ψ(x, y) ,
an + bm
Z
− bs
s
bs+ak−a c m an+bm
ψ(x, y) = px a y x a exp − uax a dx,
an + bm
2◦ . General solution for an + bm = 0:
h i
c n m
w = exp x y ln x Φ y a x−b + ψ(x, y) ,
a


 −2
ak−bs
s c − bm
m
pk x a y exp − x a y (k ln x − 1) if k 6= 0,
a
ψ(x, y) =

 − bs c − bm
 12 px a y s exp − x a y m (ln x)2 if k = 0.
a
∂x
+ g(x, y) ∂w
∂y
= h1 (x, y)w + h0 (x, y) 119
∂w ∂w
29. ax + by = (cxn + py m )w + qxk y s .
∂x ∂y
General solution:
n
cx py m
w = exp + Φ y a x−b
an bm
Z
bs ak−a+bs cxn p m bm
+ qx− a y s x a exp − − uax a dx ,
an bm
∂w ∂w
30. x2 + axy = by 2 w + cxn y m .
∂x ∂y
1◦ . General solution for a 6= 1/2:
Z
b y2 −a −am m am+n−2 b 2 2a−1
w = exp Φ(x y) + cx y x exp − u x dx ,
2a − 1 x 2a − 1
where u = x−a y. In the integration, u is considered a parameter.
2◦ . General solution for a = 1/2:
2
y 2cxn y m
w = exp b ln x Φ(x−1/2 y) + .
x (m + 2n − 2)x − by 2
∂w ∂w
31. x2 + xy = y 2 (ax + by)w + cxn y m .
∂x ∂y
General solution:
Z
(ax + by)y 2 y −m m m+n−2 (a + bu)u2 x2
w = exp Φ + cx y x exp − dx ,
2x x 2
where u = y/x. In the integration, u is considered a parameter.
∂w ∂w
32. axn + bxm y = cxp y q w + sxγ y δ + d.
∂x ∂y
This is a special case of equation 1.4.7.21 with f (x) = axn , g(x) = bxm , h(x, y) = cxp y q ,
and F (x, y) = sxγ y δ + d.
∂w ∂w
33. axn + bxm y + cxk = sxp y q w + d.
∂x ∂y
h(x, y) = sxp y q , and F (x, y) = d.
∂w ∂w
34. axn + bxm y k = cw + sxp y q + d.
∂x ∂y
This is a special case of equation 1.4.7.23 with f (x) = axn , g1 (x) ≡ 0, g0 (x) = bxm ,
h(x, y) = c, and F (x, y) = sxp y q + d.
∂w ∂w
35. ay k + bxn = cw + sxm .
∂x ∂y
This is a special case of equation 1.4.7.11 with f1 (x) = a, f2 (x) = bxn , g(x) = c, and
h(x) = sxm .

∂w ∂w
1. a +b = ceλx + seµy w + keνx .
∂x ∂y
General solution:
c Z
λx s µy k c λx s µbx−µu
w = exp e + e Φ(bx − ay) + exp νx − e − e a dx ,
aλ bµ a aλ bµ
∂w ∂w
2. a +b = ceαx+βy w + keγx .
∂x ∂y
1◦ . General solution for aα + bβ 6= 0:
Z (aα+bβ)x−βu

c αx+βy k c
w = exp e Φ(bx−ay)+ exp γx− e a dx ,
aα + bβ a aα + bβ
2◦ . General solution for aα + bβ = 0:

c αx+βy keγx
w = exp xe Φ(bx − ay) + .
a aγ − ceαx+βy
∂w ∂w
3. aeλx + beβx = ceγy w + seµx+δy .
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = aeλx , g1 (x) ≡ 0, g0 (x) = beβx ,
h(x, y) = ceγy , and F (x, y) = seµx+δy .
∂w ∂w
4. aeβx + beγx + ceλy = sw + keµx+δy .
∂x ∂y
This is a special case of equation 1.4.7.24 with f (x) = aeβx , g1 (x) = beγx , g0 (x) = c,
h(x, y) = s, and F (x, y) = keµx+δy .
∂w ∂w
5. aeβx + beγx + ceλy = seµx+δy w + k.
∂x ∂y
This is a special case of equation 1.4.7.24 with f (x) = aeβx , g1 (x) = beγx , g0 (x) = c,
h(x, y) = seµx+δy , and F (x, y) = k.
∂w ∂w
6. aeβx + beγx+λy = ceσy w + keµx+δy + d.
∂x ∂y
This is a special case of equation 1.4.7.24 with f (x) = aeβx , g1 (x) ≡ 0, g0 (x) = beγx ,
h(x, y) = ceσy , and F (x, y) = keµx+δy + d.
∂w ∂w
7. aeλy + bxβx = cw + seγx .
∂x ∂y
This is a special case of equation 1.4.7.12 with f1 (x) = a, f2 (x) = bxβx , g(x) = c, and
h(x) = seγx .
∂x
+ g(x, y) ∂w
∂y
= h1 (x, y)w + h0 (x, y) 121
∂w ∂w
8. aeλy + bxβx = ceγx w + s.
∂x ∂y
This is a special case of equation 1.4.7.12 with f1 (x) = a, f2 (x) = bxβx , g(x) = ceγx , and
h(x) = s.

∂w ∂w
9. + (aeλx y + bxn ) = cw + keγx .
∂x ∂y
This is a special case of equation 1.4.7.7 with f (x) = 1, g1 (x) = aeλx , g0 (x) = bxn ,
h1 (x) = c, and h0 (x) = keγx .
∂w ∂w
10. + (aeλx y + beβx ) = cw + keγx.
∂x ∂y
This is a special case of equation 1.4.7.7 with f (x) = 1, g1 (x) = aeλx , g0 (x) = beβx ,
h1 (x) = c, and h0 (x) = keγx .
∂w ∂w
11. + (aeλx y + beβx ) = cw + kxn .
∂x ∂y
This is a special case of equation 1.4.7.7 with f (x) = 1, g1 (x) = aeλx , g0 (x) = beβx ,
h1 (x) = c, and h0 (x) = kxn .
∂w ∂w
12. + (aeλy + bxk ) = cw + keγx .
∂x ∂y
This is a special case of equation 1.4.7.10 with f (x) = 1, g1 (x) = bxk , g0 (x) = a, h2 (x) = c,
h1 (x) = 0, and h0 (x) = keγx .
∂w ∂w
13. x +y = axeλx+µy w + beνx .
∂x ∂y
General solution:
Z
ax λx+µy y a (λ+µu)x dx
w = exp e Φ + b exp νx − e ,
λx + µy x λ + µu x
∂w ∂w
14. x +y = ayeλx + bxeµy w + ceνx .
∂x ∂y
General solution:
Z
ay λx bx µy y au λx b µux dx
w = exp e + e Φ + c exp νx − e − e ,
λx µy x λ µu x
∂w ∂w
15. ay k + beλx = w + ceβx .
∂x ∂y
This is a special case of equation 1.4.7.11 with f1 (x) = a, f2 (x) = beλx , g(x) = 1, and
h(x) = ceβx .
∂w ∂w
16. aeλx + by = w + ceλx .
∂x ∂y
This is a special case of equation 1.4.7.7 with f (x) = aeλx , g1 (x) = b, g0 (x) = 0, h1 (x) = 1,
and h0 (x) = ceλx .
∂w ∂w
17. aeλy + bxk = w + ceβx .
∂x ∂y
This is a special case of equation 1.4.7.12 with f1 (x) = a, f2 (x) = bxk , g(x) = 1, and
h(x) = ceβx .
∂w ∂w
18. aeλy + beβx = w + cxk .
∂x ∂y
This is a special case of equation 1.4.7.12 with f1 (x) = a, f2 (x) = beβx , g(x) = 1, and
h(x) = cxk .

∂w ∂w
1. a +b = cw + sinhk (λx) sinhn (βy).
∂x ∂y
This is a special case of equation 1.4.7.13 with f (x) = sinhk (λx) and g(y) = sinhn (βy).
∂w ∂w
2. a +b = c sinhk (λx)w + s sinhn (βx).
∂x ∂y
This is a special case of equation 1.4.7.1 with f (x) = sinhk (λx) and g(y) = sinhn (βx).
∂w ∂w
3. a +b = c1 sinhn1 (λ1 x) + c2 sinhn2 (λ2 y) w
∂x ∂y
+ s1 sinhk1 (β1 x) + s2 sinhk2 (β2 y).
This is a special case of equation 1.4.7.16 with f (x)=c1 sinhn1(λ1x), g(y)=c2 sinhn2(λ2y),
p(x) = s1 sinhk1 (β1 x), and q(y) = s2 sinhk2 (β2 y).
∂w ∂w
4. a sinhn (λx) + b sinhm (µx) = c sinhk (νx)w + p sinhs (βy).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a sinhn (λx), g1 (x) ≡ 0, g0 (x) =
b sinhm (µx), h(x, y) = c sinhk (νx), and F (x, y) = p sinhs (βy).
∂w ∂w
5. a sinhn (λx) + b sinhm (µx) = c sinhk (νy)w + p sinhs (βx).
∂x ∂y
b sinhm (µx), h(x, y) = c sinhk (νy), and F (x, y) = p sinhs (βx).

∂w ∂w
6. a +b = cw + coshk (λx) coshn (βy).
∂x ∂y
This is a special case of equation 1.4.7.13 with f (x) = coshk (λx) and g(y) = coshn (βy).
∂x
+ g(x, y) ∂w
∂y
= h1 (x, y)w + h0 (x, y) 123
∂w ∂w
7. a +b = c coshk (λx)w + s coshn (βx).
∂x ∂y
This is a special case of equation 1.4.7.1 with f (x) = coshk (λx) and g(y) = coshn (βx).
∂w ∂w
8. a +b = c1 coshn1 (λ1 x) + c2 coshn2 (λ2 y) w
∂x ∂y
+ s1 coshk1 (β1 x) + s2 coshk2 (β2 y).
This is a special case of equation 1.4.7.16 in which f (x) = c1 coshn1 (λ1 x), g(y) =
c2 coshn2 (λ2 y), p(x) = s1 coshk1 (β1 x), and q(y) = s2 coshk2 (β2 y).
∂w ∂w
9. x +y = ax cosh(λx + µy)w + b cosh(νx).
∂x ∂y
General solution:
Z
ax sinh(λx + µy) y a sinh[(λ + µu)x] dx
w = exp Φ + b cosh(νx) exp − ,
λx + µy x λ + µu x
∂w ∂w
10. a coshn (λx) + b coshm (µx) = c coshk (νx)w + p coshs (βy).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a coshn (λx), g1 (x) ≡ 0, g0 (x) =
b coshm (µx), h(x, y) = c coshk (νx), and F (x, y) = p coshs (βy).
∂w ∂w
11. a coshn (λx) + b coshm (µx) = c coshk (νy)w + p coshs (βx).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a coshn (λx), g1 (x) ≡ 0, g0 (x) =
b coshm (µx), h(x, y) = c coshk (νy), and F (x, y) = p coshs (βx).

∂w ∂w
12. a +b = cw + tanhk (λx) tanhn (βy).
∂x ∂y
This is a special case of equation 1.4.7.13 with f (x) = tanhk (λx) and g(y) = tanhn (βy).
∂w ∂w
13. a +b = c tanhk (λx)w + s tanhn (βx).
∂x ∂y
This is a special case of equation 1.4.7.1 with f (x) = tanhk (λx) and g(y) = tanhn (βx).
∂w ∂w
14. a +b = c1 tanhn1 (λ1 x) + c2 tanhn2 (λ2 y) w
∂x ∂y
+ s1 tanhk1 (β1 x) + s2 tanhk2 (β2 y).
This is a special case of equation 1.4.7.16 in which f (x) = c1 tanhn1 (λ1 x), g(y) =
c2 tanhn2 (λ2 y), p(x) = s1 tanhk1 (β1 x), and q(y) = s2 tanhk2 (β2 y).
∂w ∂w
15. a tanhn (λx) + b tanhm (µx) = c tanhk (νx)w + p tanhs (βy).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a tanhn (λx), g1 (x) ≡ 0, g0 (x) =
b tanhm (µx), h(x, y) = c tanhk (νx), and F (x, y) = p tanhs (βy).
∂w ∂w
16. a tanhn (λx) + b tanhm (µx) = c tanhk (νy)w + p tanhs (βx).
∂x ∂y
b tanhm (µx), h(x, y) = c tanhk (νy), and F (x, y) = p tanhs (βx).

∂w ∂w
17. a +b = cw + cothk (λx) cothn (βy).
∂x ∂y
This is a special case of equation 1.4.7.13 with f (x) = cothk (λx) and g(y) = cothn (βy).
∂w ∂w
18. a +b = c cothk (λx)w + s cothn (βx).
∂x ∂y
This is a special case of equation 1.4.7.1 with f (x) = cothk (λx) and g(y) = cothn (βx).
∂w ∂w
19. a +b = c1 cothn1 (λ1 x) + c2 cothn2 (λ2 y) w
∂x ∂y
+ s1 cothk1 (β1 x) + s2 cothk2 (β2 y).
This is a special case of equation 1.4.7.16 in which f (x) = c1 cothn1 (λ1 x), g(y) =
c2 cothn2 (λ2 y), p(x) = s1 cothk1 (β1 x), and q(y) = s2 cothk2 (β2 y).
∂w ∂w
20. a cothn (λx) + b cothm (µx) = c cothk (νx)w + p coths (βy).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a cothn (λx), g1 (x) ≡ 0, g0 (x) =
b cothm (µx), h(x, y) = c cothk (νx), and F (x, y) = p coths (βy).
∂w ∂w
21. a cothn (λx) + b cothm (µx) = c cothk (νy)w + p coths (βx).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a cothn (λx), g1 (x) ≡ 0, g0 (x) =
b cothm (µx), h(x, y) = c cothk (νy), and F (x, y) = p coths (βx).

∂w ∂w
22. a +b = w + c1 sinhk (λx) + c2 coshn (βy).
∂x ∂y
This is a special case of equation 1.4.7.16 with f (x) = 0, g(y) = 1, p(x) = c1 sinhk (λx),
and q(y) = c2 coshn (βy).
∂w ∂w
23. a +b = cw + sinhk (λx) coshn (βy).
∂x ∂y
This is a special case of equation 1.4.7.13 with f (x) = sinhk (λx) and g(y) = coshn (βy).
∂w ∂w
24. a +b = cw + k tanh(λx) + s coth(µy).
∂x ∂y
General solution:
Z
cx/a 1 x bµ −ct/a
w=e Φ(bx − ay) − s coth (x − t) − µy − k tanh(λt) e dt .
a 0 a
∂x
+ g(x, y) ∂w
∂y
= h1 (x, y)w + h0 (x, y) 125
∂w ∂w
25. a + b sinh(λx) = cw + k cosh(µy).
∂x ∂y
General solution:
Z x
cx/a bµ −ct/a

w=e cosh µy + cosh(λt)−cosh(λx) e dt+Φ aλy −b cosh(λx) .
0 aλ
∂w ∂w
26. a sinhn (λx) + b coshm (µx) = c coshk (νx)w + p sinhs (βy).
∂x ∂y
b coshm (µx), h(x, y) = c coshk (νx), and F (x, y) = p sinhs (βy).
∂w ∂w
27. a tanhn (λx) + b cothm (µx) = c tanhk (νy)w + p coths (βx).
∂x ∂y
b cothm (µx), h(x, y) = c tanhk (νy), and F (x, y) = p coths (βx).

∂w ∂w
1. a +b = cw + lnk (λx) lnn (βy).
∂x ∂y
This is a special case of equation 1.4.7.13 with f (x) = lnk (λx) and g(y) = lnn (βy).
∂w ∂w
2. a +b = c lnk (λx)w + s lnn (βx).
∂x ∂y
This is a special case of equation 1.4.7.1 with f (x) = lnk (λx) and g(y) = lnn (βx).
∂w ∂w
3. a +b = c1 lnn1(λ1x)+c2 lnn2(λ2y) w+s1 lnk1(β1x)+s2 lnk2(β2y).
∂x ∂y
This is a special case of equation 1.4.7.16 with f (x) = c1 lnn1 (λ1 x), g(y) = c2 lnn2 (λ2 y),
p(x) = s1 lnk1 (β1 x), and q(y) = s2 lnk2 (β2 y).
∂w ∂w
4. a ln(λx) + b ln(µy) = cw + k.
∂x ∂y
General solution:
Z Z Z
k c dx dx dy
w=− + Φ(u) exp , u=b −a .
c a ln(λx) ln(λx) ln(µy)
∂w ∂w
5. a lnn (λx) + b lnm (µx) = c lnk (νx)w + p lns (βy) + q.
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a lnn (λx), g1 (x) ≡ 0, g0 (x) =
b lnm (µx), h(x, y) = c lnk (νx), and F (x, y) = p lns (βy) + q.
∂w ∂w
6. a lnn (λx) + b lnm (µx) = c lnk (νy)w + p lns (βx) + q.
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a lnn (λx), g1 (x) ≡ 0, g0 (x) =
b lnm (µx), h(x, y) = c lnk (νy), and F (x, y) = p lns (βx) + q.
∂w ∂w
7. a +b = w + c1 xk + c2 lnn (βy).
∂x ∂y
This is a special case of equation 1.4.7.16 with f (x) = 0, g(y) = 1, p(x) = c1 xk , and
q(y) = c2 lnn (βy).
∂w ∂w
8. a +b = cw + xk lnn (βy).
∂x ∂y
This is a special case of equation 1.4.7.13 with f (x) = xk and g(y) = lnn (βy).
∂w ∂w
9. axk + bxn = cw + s lnm (βx).
∂x ∂y
This is a special case of equation 1.4.7.7 with f (x) = axk , g1 (x) = 0, g0 (x) = bxn ,
h1 (x) = c, and h0 (x) = s lnm (βx).
∂w ∂w
10. axn + by k = cw + s lnm (βx).
∂x ∂y
This is a special case of equation 1.4.7.23 with f (x) = axn , g1 (x) = 0, g0 (x) = b, h(x, y) =
c, and F (x, y) = s lnm (βx).
∂w ∂w
11. axk + b lnn (λx) = cw + sxm .
∂x ∂y
This is a special case of equation 1.4.7.7 with f (x) = axk , g1 (x) = 0, g0 (x) = b lnn (λx),
h1 (x) = c, and h0 (x) = sxm .
∂w ∂w
12. ay k + bxn = cw + s lnm (βx).
∂x ∂y
This is a special case of equation 1.4.7.11 with f1 (x) = a, f2 (x) = bxn , g(x) = c, and
h(x) = s lnm (βx).
∂w ∂w
13. ay k + b lnn (λx) = cw + sxm .
∂x ∂y
This is a special case of equation 1.4.7.11 with f1 (x) = a, f2 (x) = b lnn (λx), g(x) = c,
and h(x) = sxm .

∂w ∂w
1. a +b = cw + k sin(λx + µy).
∂x ∂y
General solution:
k
w = ecx/a Φ(bx − ay) − (aλ + bµ) cos(λx + µy) + c sin(λx + µy) .
c2 + (aλ + bµ)2
∂x
+ g(x, y) ∂w
∂y
= h1 (x, y)w + h0 (x, y) 127
∂w ∂w
2. a +b = w + c1 sink (λx) + c2 sinn (βy).
∂x ∂y
This is a special case of equation 1.4.7.16 with f (x) = 0, g(y) = 1, p(x) = c1 sink (λx),
and q(y) = c2 sinn (βy).
∂w ∂w
3. a +b = cw + sink (λx) sinn (βy).
∂x ∂y
This is a special case of equation 1.4.7.13 with f (x) = sink (λx) and g(y) = sinn (βy).
∂w ∂w
4. ax + by = cw + k sin(λx + µy).
∂x ∂y
General solution:
Z x
c/a k −(a+c)/a b/a −b/a
−b/a

w=x t sin λt + µt x y dt + Φ x y .
a 0
∂w ∂w
5. x +y = ax sin(λx + µy)w + b sin(νx).
∂x ∂y
General solution:

ax y
w = exp − cos(λx + µy) Φ
λx + µy x
Z
a
+ b sin(νx) exp cos (λ + µu)x dx ,
λ + µu
∂w ∂w
6. a sinn (λx) + b sinm (µx) = c sink (νx)w + p sins (βy).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a sinn (λx), g1 (x) ≡ 0, g0 (x) =
b sinm (µx), h(x, y) = c sink (νx), and F (x, y) = p sins (βy).
∂w ∂w
7. a sinn (λx) + b sinm (µx) = c sink (νy)w + p sins (βx).
∂x ∂y
b sinm (µx), h(x, y) = c sink (νy), and F (x, y) = p sins (βx).

∂w ∂w
8. a +b = cw + k cos(λx + µy).
∂x ∂y
General solution:
k
w = ecx/a Φ(bx − ay) + (aλ + bµ) sin(λx + µy) − c cos(λx + µy) .
c2 + (aλ + bµ)2
∂w ∂w
9. a +b = w + c1 cosk (λx) + c2 cosn (βy).
∂x ∂y
This is a special case of equation 1.4.7.16 with f (x) = 0, g(y) = 1, p(x) = c1 cosk (λx),
and q(y) = c2 cosn (βy).
∂w ∂w
10. a +b = cw + cosk (λx) cosn (βy).
∂x ∂y
This is a special case of equation 1.4.7.13 with f (x) = cosk (λx) and g(y) = cosn (βy).
∂w ∂w
11. ax + by = cw + k cos(λx + µy).
∂x ∂y
General solution:
Z x
−b/a

w=x t cos λt + µt x y dt + Φ x y .
a 0
∂w ∂w
12. x +y = ax cos(λx + µy)w + b cos(νx).
∂x ∂y
General solution:

ax y
w = exp sin(λx + µy) Φ
λx + µy x
Z
a
+ b cos(νx) exp − sin (λ + µu)x dx ,
λ + µu
∂w ∂w
13. a cosn (λx) + b cosm (µx) = c cosk (νx)w + p coss (βy).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a cosn (λx), g1 (x) ≡ 0, g0 (x) =
b cosm (µx), h(x, y) = c cosk (νx), and F (x, y) = p coss (βy).

∂w ∂w
14. a +b = cw + k tan(λx + µy).
∂x ∂y
General solution:
Z
cx/a k x bµ bx −ct/a
w=e Φ(bx − ay) + tan λ + t+µ y− e dt .
a 0 a a
∂w ∂w
15. a +b = w + c1 tank (λx) + c2 tann (βy).
∂x ∂y
This is a special case of equation 1.4.7.16 with f (x) = 0, g(y) = 1, p(x) = c1 tank (λx),
and q(y) = c2 tann (βy).
∂w ∂w
16. a +b = cw + tank (λx) tann (βy).
∂x ∂y
This is a special case of equation 1.4.7.13 with f (x) = tank (λx) and g(y) = tann (βy).
∂w ∂w
17. a + b tan(µy) = c tan(λx)w + k tan(νx).
∂x ∂y
General solution:
Z
−c/aλ k c/aλ

w = cos(λx) cos(λx) tan(νx) dx + Φ bµx − a ln |sin(µy)| .
a
∂x
+ g(x, y) ∂w
∂y
= h1 (x, y)w + h0 (x, y) 129
∂w ∂w
18. ax + by = cw + k tan(λx + µy).
∂x ∂y
Z x
−b/a

General solution: w = x t tan λt + µt x y dt + Φ x y .
a 0
∂w ∂w
19. a tann (λx) + b tanm (µx) = c tank (νx)w + p tans (βy).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a tann (λx), g1 (x) ≡ 0, g0 (x) =
b tanm (µx), h(x, y) = c tank (νx), and F (x, y) = p tans (βy).
∂w ∂w
20. a tann (λx) + b tanm (µx) = c tank (νy)w + p tans (βx).
∂x ∂y
b tanm (µx), h(x, y) = c tank (νy), and F (x, y) = p tans (βx).

∂w ∂w
21. a +b = cw + k cot(λx + µy).
∂x ∂y
General solution:
Z
cx/a k x bµ bx −ct/a
w=e Φ(bx − ay) + cot λ + t+µ y− e dt .
a 0 a a
∂w ∂w
22. a +b = w + c1 cotk (λx) + c2 cotn (βy).
∂x ∂y
This is a special case of equation 1.4.7.16 with f (x) = 0, g(y) = 1, p(x) = c1 cotk (λx),
and q(y) = c2 cotn (βy).
∂w ∂w
23. a +b = cw + cotk (λx) cotn (βy).
∂x ∂y
This is a special case of equation 1.4.7.13 with f (x) = cotk (λx) and g(y) = cotn (βy).
∂w ∂w
24. a + b cot(µy) = c cot(λx)w + k cot(νx).
∂x ∂y
General solution:
Z
c/aλ k −c/aλ

w = sin(λx) sin(λx) cot(νx) dx + Φ bµx + a ln |cos(µy)| .
a
∂w ∂w
25. ax + by = cw + k cot(λx + µy).
∂x ∂y
Z x
−b/a

General solution: w = x t cot λt + µt x y dt + Φ x y .
a 0
∂w ∂w
26. a cotn (λx) + b cotm (µx) = c cotk (νx)w + p cots (βy).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a cotn (λx), g1 (x) ≡ 0, g0 (x) =
b cotm (µx), h(x, y) = c cotk (νx), and F (x, y) = p cots (βy).
∂w ∂w
27. a cotn (λx) + b cotm (µx) = c cotk (νy)w + p cots (βx).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a cotn (λx), g1 (x) ≡ 0, g0 (x) =
b cotm (µx), h(x, y) = c cotk (νy), and F (x, y) = p cots (βx).

∂w ∂w
28. a +b = w + c1 sink (λx) + c2 cosn (βy).
∂x ∂y
This is a special case of equation 1.4.7.16 with f (x) = 0, g(y) = 1, p(x) = c1 sink (λx),
and q(y) = c2 cosn (βy).
∂w ∂w
29. a +b = cw + sink (λx) cosn (βy).
∂x ∂y
This is a special case of equation 1.4.7.13 with f (x) = sink (λx) and g(y) = cosn (βy).
∂w ∂w
30. a + b sin(µy) = c sin(λx)w + k cos(νx) + s.
∂x ∂y
General solution:
Z
c 1 c
w = exp − cos(λx) s + k cos(νx) exp cos(λx) dx
aλ a aλ

µ

+ Φ bµx − a lntan y .
2
∂w ∂w
31. a + b sin(µy) = c sin(λx)w + k tan(νx) + s.
∂x ∂y
General solution:
Z
c 1 c
w = exp − cos(λx) s + k tan(νx) exp cos(λx) dx
aλ a aλ

µ
+ Φ bµx − a lntan y .
2
∂w ∂w
32. a + b tan(µy) = c tan(λx)w + k cot(νx) + s.
∂x ∂y
General solution:
Z
−c/aλ 1 c/aλ

w = cos(λx)
s + k cot(νx) cos(λx)
dx + Φ bµx − a ln sin(µy) .
a
∂w ∂w
33. a sinn (λx) + b cosm (µx) = c cosk (νx)w + p sins (βy).
∂x ∂y
b cosm (µx), h(x, y) = c cosk (νx), and F (x, y) = p sins (βy).
∂w ∂w
34. a tann (λx) + b cotm (µx) = c tank (νy)w + p cots (βx).
∂x ∂y
b cotm (µx), h(x, y) = c tank (νy), and F (x, y) = p cots (βx).
∂x
+ g(x, y) ∂w
∂y
= h1 (x, y)w + h0 (x, y) 131

∂w ∂w
1. a +b = w + c1 arcsink (λx) + c2 arcsinn (βy).
∂x ∂y
This is a special case of equation 1.4.7.16 with f (x) = 0, g(y) = 1, p(x) = c1 arcsink (λx),
and q(y) = c2 arcsinn (βy).
∂w ∂w
2. a +b = cw + arcsink (λx) arcsinn (βy).
∂x ∂y
This is a special case of equation 1.4.7.13 with f (x) = arcsink (λx) and g(y) = arcsinn (βy).
∂w ∂w
3. a +b = c1 arcsin(λ1 x) + c2 arcsin(λ2 y) w
∂x ∂y
+ s1 arcsinn (β1 x) + s2 arcsink (β2 y).
This is a special case of equation 1.4.7.16 in which f (x) = c1 arcsin(λ1 x), g(y) =
c2 arcsin(λ2 y), p(x) = s1 arcsinn (β1 x), and q(y) = s2 arcsink (β2 y).
∂w ∂w
4. a + b arcsinm (µx) = c arcsink (νx)w + p arcsinn (βy).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b arcsinm (µx),
h(x, y) = c arcsink (νx), and F (x, y) = p arcsinn (βy).
∂w ∂w
5. a + b arcsinm (µx) = c arcsink (νy)w + p arcsinn (βx).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b arcsinm (µx),
h(x, y) = c arcsink (νy), and F (x, y) = p arcsinn (βx).

∂w ∂w
6. a +b = w + c1 arccosk (λx) + c2 arccosn (βy).
∂x ∂y
This is a special case of equation 1.4.7.16 with f (x) = 0, g(y) = 1, p(x) = c1 arccosk (λx),
and q(y) = c2 arccosn (βy).
∂w ∂w
7. a +b = cw + arccosk (λx) arccosn (βy).
∂x ∂y
This is a special case of equation 1.4.7.13 in which f (x) = arccosk (λx) and g(y) =
arccosn (βy).
∂w ∂w
8. a +b = c1 arccos(λ1 x) + c2 arccos(λ2 y) w
∂x ∂y
+ s1 arccosn (β1 x) + s2 arccosk (β2 y).
This is a special case of equation 1.4.7.16 in which f (x) = c1 arccos(λ1 x), g(y) =
c2 arccos(λ2 y), p(x) = s1 arccosn (β1 x), and q(y) = s2 arccosk (β2 y).
∂w ∂w
9. a + b arccosm (µx) = c arccosk (νx)w + p arccosn (βy).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b arccosm (µx),
h(x, y) = c arccosk (νx), and F (x, y) = p arccosn (βy).
∂w ∂w
10. a + b arccosm (µx) = c arccosk (νy)w + p arccosn (βx).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b arccosm (µx),
h(x, y) = c arccosk (νy), and F (x, y) = p arccosn (βx).

∂w ∂w
11. a +b = w + c1 arctank (λx) + c2 arctann (βy).
∂x ∂y
This is a special case of equation 1.4.7.16 with f (x) = 0, g(y) = 1, p(x) = c1 arctank (λx),
and q(y) = c2 arctann (βy).
∂w ∂w
12. a +b = cw + arctank (λx) arctann (βy).
∂x ∂y
This is a special case of equation 1.4.7.13 with f (x) = arctank (λx) and g(y) = arctan n (βy).
∂w ∂w
13. a +b = c1 arctan(λ1 x) + c2 arctan(λ2 y) w
∂x ∂y
+ s1 arctann (β1 x) + s2 arctank (β2 y).
This is a special case of equation 1.4.7.16 in which f (x) = c1 arctan(λ1 x), g(y) =
c2 arctan(λ2 y), p(x) = s1 arctann (β1 x), and q(y) = s2 arctank (β2 y).
∂w ∂w
14. a + b arctanm (µx) = c arctank (νx)w + s arctann (βy).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b arctanm (µx),
h(x, y) = c arctank (νx), and F (x, y) = s arctann (βy).
∂w ∂w
15. a + b arctanm (µx) = c arctank (νy)w + s arctann (βx).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a, g1 (x) ≡ 0, g0 (x) = b arctanm (µx),
h(x, y) = c arctank (νy), and F (x, y) = s arctann (βx).

∂w ∂w
16. a +b = w + c1 arccot k (λx) + c2 arccot n (βy).
∂x ∂y
This is a special case of equation 1.4.7.16 with f (x) = 0, g(y) = 1, p(x) = c1 arccotk (λx),
and q(y) = c2 arccotn (βy).
∂w ∂w
17. a +b = cw + arccot k (λx) arccot n (βy).
∂x ∂y
This is a special case of equation 1.4.7.13 with f (x) = arccot k (λx) and g(y) = arccot n (βy).
∂x
+ g(x, y) ∂w
∂y
= h1 (x, y)w + h0 (x, y) 133
∂w ∂w
18. a +b = c1 arccot(λ1 x) + c2 arccot(λ2 y) w
∂x ∂y
+ s1 arccot n (β1 x) + s2 arccot k (β2 y).
This is a special case of equation 1.4.7.16 in which f (x) = c1 arccot(λ1 x), g(y) =
c2 arccot(λ2 y), p(x) = s1 arccotn (β1 x), and q(y) = s2 arccotk (β2 y).
∂w ∂w
19. a arccotn (λx) + b arccot m (µx) = c arccot k (νx)w + p arccot s (βy).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a arccotn (λx), g1 (x) ≡ 0, g0 (x) =
b arccotm (µx), h(x, y) = c arccot k (νx), and F (x, y) = p arccots (βy).
∂w ∂w
20. a arccotn (λx) + b arccot m (µx) = c arccot k (νy)w + p arccots (βx).
∂x ∂y
This is a special case of equation 1.4.7.22 with f (x) = a arccotn (λx), g1 (x) ≡ 0, g0 (x) =
b arccotm (µx), h(x, y) = c arccot k (νy), and F (x, y) = p arccots (βx).

∂w ∂w
1. a +b = f (x)w + g(x).
∂x ∂y
General solution:
Z Z Z
1 1 1
w = exp f (x) dx Φ(bx − ay) + g(x) exp − f (x) dx dx .
a a a
∂w ∂w
2. a +b = (cy + k)w + f (x).
∂x ∂y
General solution:

x
w = exp 2a(cy + k) − bcx Φ(bx − ay)
2a2
Z
1 x t
+ exp − 2 2a(cy + k) + bc(t − 2x) g(t) dt ,
a x0 2a

∂w ∂w
3. a +b = f (x)yw + g(x).
∂x ∂y
General solution:
Z
1 x g(t) dt
w = F (x, u) Φ(u) + , u = bx − ay,
a x0 F (t, u)
Z x
1
where F (x, u) = exp 2 (bτ − u)f (τ ) dτ .
a x0
∂w ∂w
4. ax + by = f (x)w + g(x).
∂x ∂y
General solution:
Z Z Z
1 f (x) dx −b/a
1 g(x) 1 f (x) dx
w = exp Φ x y + exp − dx .
a x a x a x
∂w ∂w
5. f (x) + (ay + b) = cw + g(x).
∂x ∂y
General solution:
Z Z Z Z
dx g(x) dx dx
w = exp c exp −c dx + Φ a − ln |ay + b| .
f (x) f (x) f (x) f (x)
∂w ∂w
6. f (x) + g(x) = h(x)w + p(x).
∂x ∂y
General solution:
Z Z Z Z
h(x) p(x) h(x) g(x)
w = exp dx exp − dx dx + Φ dx − y .
f (x) f (x) f (x) f (x)
∂w ∂w
7. f (x) + g1 (x)y + g0 (x) = h1 (x)w + h0 (x).
∂x ∂y
This is a special case of equation 1.4.7.22 with h(x, y) = h1 (x) and F (x, y) = h0 (x).
General solution:
Z
h0 (x) dx
w = H(x) Φ(u) + , u = yG(x) − S(x),
f (x)H(x)
where
Z Z Z
h1 (x) g1 (x) g0 (x)
H(x) = exp dx , G(x) = exp − dx , S(x) = G(x) dx.
f (x) f (x) f (x)
∂w ∂w
8. f (x) + g1 (x)y + g0 (x) = h2 (x)w + h1 (x)y + h0 (x).
∂x ∂y
This is a special case of equation 1.4.7.22 with h(x, y) = h2 (x) and F (x, y) = h1 (x)y +
h0 (x).
∂w ∂w
9. f (x) + g1 (x)y + g0 (x)y k = h2 (x)w + h1 (x)y n + h0 (x).
∂x ∂y
This is a special case of equation 1.4.7.23 with h(x, y) = h2 (x) and F (x, y) = h1 (x)y n +
h0 (x).
∂x
+ g(x, y) ∂w
∂y
= h1 (x, y)w + h0 (x, y) 135
∂w ∂w
10. f (x) + g1 (x) + g0 (x)eλy = h2 (x)w + h1 (x)eβy + h0 (x).
∂x ∂y
This is a special case of equation 1.4.7.24 with h(x, y) = h2 (x) and F (x, y) = h1 (x)eβy +
h0 (x).
∂w ∂w
11. f1 (x)y k + f2 (x) = g(x)w + h(x).
∂x ∂y
General solution:
Z x
h(t) − k dt
w = Φ(u)G(x, u) + G(x, u) u + F (t) k+1 ,
x0 f1 (t) G(t, u)
where Z
k+1 f2 (x)
u=y − F (x), F (x) = (k + 1) dx,
f1 (x)
Z x
g(t) − k
G(x, u) = exp u + F (t) k+1 dt ,
x0 f1 (t)
and x0 may be taken as arbitrary.
∂w ∂w
12. f1 (x)eλy + f2 (x) = g(x)w + h(x).
∂x ∂y
General solution:
Z x
h(t) dt
w = Φ(u)G(x, u) + G(x, u) ,
x0 f1 (t) u + F (t) G(t, u)
where
Z Z x
λy f2 (x) g(t) dt
u=e − F (x), F (x) = λ dx, G(x, u) = exp ,
f1 (x) x0 f1 (t) u + F (t)

∂w ∂w
13. a +b = cw + f (x)g(y).
∂x ∂y
General solution:
Z
1 x b(t − x) + ay −ct/a
w = ecx/a Φ(bx − ay) + f (t)g e dt ,
a x0 a
∂w ∂w
14. a +b = cw + xf (y) + yg(x).
∂x ∂y
General solution:
Z x
cx/a 1 b(t − x) + ay −ct/a
w=e Φ(bx−ay)+ 2 atf + b(t−x)+ay g(t) e dt .
a x0 a
∂w ∂w
15. a +b = f (x)w + g(x)h(y).
∂x ∂y
General solution:
Z
1 x g(t) bt − u
w = F (x) Φ(u) + g dt , u = bx − ay,
a x0 F (t) a
Z
1
where F (x) = exp f (x) dx .
a
∂w ∂w
16. a +b = f (x) + g(y) w + p(x) + q(y).
∂x ∂y
General solution:
Z Z
1 1
w = exp f (x) dx + g(y) dy Φ(bx − ay)
a b
Z Z
1 bx − u 1 bx − u
+ p(x) + q exp − f (x) + g dx dx ,
a a a a
∂w ∂w
17. ax + by = cw + f (x)g(y).
∂x ∂y
General solution:
Z x
c/a −b/a
1 −(a+c)/a b/a −b/a

w=x Φ x y + t f (t)g t x y dt .
a x0
∂w ∂w
18. f1 (x) + f2 (y) = aw + g1 (x) + g2 (y).
∂x ∂y
General solution:
Z Z
g1 (x) dx g2 (y) dy
w = E1 (x)Φ(u) + E1 (x) + E2 (y) ,
f1 (x)E1 (x) f2 (y)E2 (y)
where
Z Z Z Z
dx dy dx dy
E1 (x) = exp a , E2 (y) = exp a , u= − .
f1 (x) f2 (y) f1 (x) f2 (y)

∂w ∂w y
19. x +y = xf w + g(x, y).
∂x ∂y x
General solution:
Z x
y y dt y
w = exp xf Φ + g(t, ut) exp −tf (u) , u= ,
x x x0 t x
∂x
+ g(x, y) ∂w
∂y
= h1 (x, y)w + h0 (x, y) 137
∂w ∂w
20. ax + by = f (x, y)w + g(x, y).
∂x ∂y
General solution:
Z
1 1 1/a b/a

w = exp f x, u x dx Φ(u)
a x
Z Z
1 1 1/a b/a
1 1 1/a b/a

+ g x, u x exp − f x, u x dx dx ,
a x a x
∂w ∂w
21. f (x) + g(x)y = h(x, y)w + F (x, y).
∂x ∂y
General solution:
Z
F (x, uG) y
w = H(x, u) Φ(u) + dx , u= ,
f (x)H(x, u) G
Z Z
g(x) h(x, uG)
where G = G(x) = exp dx and H(x, u) = exp dx . In the
f (x) f (x)
integration, u is considered a parameter.
∂w ∂w
22. f (x) + g1 (x)y + g0 (x) = h(x, y)w + F (x, y).
∂x ∂y
General solution:
Z
F (x, uG + Q)
w = H(x, u) Φ(u) + dx ,
f (x)H(x, u)
where
Z
y−Q h(x, uG + Q)
u= , H(x, u) = exp dx ,
G f (x)
Z Z
g1 (x) g0 (x) dx
G = G(x) = exp dx , Q = Q(x) = G(x) .
f (x) f (x)G(x)
∂w ∂w
23. f (x) + g1 (x)y + g0 (x)y k = h(x, y)w + F (x, y).
∂x ∂y
∂w ∂w 1 1
f (x) + (1 − k) g1 (x)ξ + g0 (x) = h x, ξ 1−k w + F x, ξ 1−k .
∂x ∂ξ
∂w ∂w
24. f (x) + g1 (x) + g0 (x)eλy = h(x, y)w + F (x, y).
∂x ∂y
The substitution z = e−λy leads to an equation of the form 1.4.7.22:

∂w ∂w 1 1
f (x) − λ g1 (x)z + g0 (x) = h x, − ln z w + F x, − ln z .
∂x ∂z λ λ
Chapter 2
First-Order Equations with Three

or More Independent Variables

∂w ∂w ∂w
f (x, y, z) + g(x, y, z) + h(x, y, z) =0
∂x ∂y ∂z
◆ For brevity, only an integral basis
u1 = u1 (x, y), u2 = u2 (x, y)
of an equation will often be presented in Section 2.1. The general solution of the equation
is given by
w = Φ(u1 , u2 ),
where Φ = Φ(u1 , u2 ) is an arbitrary function of two variables.

◮ Coefficients of equations are linear in x, y, and z.
∂w ∂w ∂w
1. a +b +c = 0.
∂x ∂y ∂z
Integral basis: u1 = bx − ay, u2 = cx − az.
∂w ∂w ∂w
2. + ax + by = 0.
∂x ∂y ∂z
Integral basis: u1 = ax2 − 2y, u2 = 3z + bx(ax2 − 3y).
∂w ∂w ∂w
3. a + by + cz = 0.
∂x ∂y ∂z
Integral basis: u1 = |y|a e−bx , u2 = |z|a e−cx .

139
140 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂w ∂w ∂w
4. + az + by = 0.
∂x ∂y ∂z
Integral basis:
( √ √
2 2 by + ab z exp − ab x if ab > 0,
u1 = by − az , u2 = p p p
by cos |ab| x + |ab| z sin |ab| x if ab < 0.
∂w ∂w ∂w
5. x + ay + bz = 0.
∂x ∂y ∂z
Integral basis: u1 = |x|a/y, u2 = |x|b/z.

∂w ∂w ∂w
6. x + az + by = 0.
∂x ∂y ∂z
Integral basis:
 √ √

|x| ab by − ab z if ab > 0,
√
u1 = by 2 − az 2 , u2 = √
−ab −ab z

|x| exp − arctan if ab < 0.
by
∂w ∂w ∂w
7. x + (ax + by) + (αx + βy + γz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.1.21 with s1 = 1, s2 = b, and s3 = γ.

∂w ∂w ∂w
8. abx + (ay + bz) b −a = 0.
∂x ∂y ∂z

ay
Integral basis: u1 = ay + bz, u2 = x exp − .
ay + bz
∂w ∂w ∂w
9. abx + b(ay + bz) + a(ay − bz) = 0.
∂x ∂y ∂z
√
This is a special case of equation 2.1.1.21 with s1 = 1 and s2,3 = ± 2.
Integral basis:
√ √ √ √
u1 = ay + ( 2 − 1)bz |x|− 2 , u2 = ay − ( 2 + 1)bz |x| 2 .
Particular solution: w = a2 y 2 − 2abyz − b2 z 2 .
∂w ∂w ∂w
10. b2 cy + a2 cx − ab(ax + by) = 0.
∂x ∂y ∂z
Integral basis: u1 = ax + by + cz, u2 = a2 x2 − b2 y 2 .

2.1. Equations of the Form f (x, y, z) ∂w
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 141
∂w ∂w ∂w
11. cz + (ax + by) + (ax + by + cz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.1.21.
1◦ . Integral basis for a 6= b:
u1 = z − x − y,
2acz + (b − c − ρ)(ax + by) 2 ρ
u2 = acz + (b − c)(ax + by)z − (ax + by)2 b+c ,
2acz + (b − c + ρ)(ax + by)
where ρ2 = 4ac + (b − c)2 6= 0.

2◦ . Integral basis for a = b:

1 1 cy − ax
u1 = z − x − y, u2 = [a(x + y) + cz] exp + .
a c z−x−y
∂w ∂w ∂w
12. b2 cz − a2 cx + ab2 y = 0.
∂x ∂y ∂z
1
√
This is a special case of equation 2.1.1.21 with s1 = −1, s2 = 1 + i 3 , and s3 =
1
√ 2
2 1−i 3 .
∂w ∂w ∂w
13. (x + a) + (y + b) + (z + c) = 0.
∂x ∂y ∂z
x+a y+b
Integral basis: u1 = , u2 = .
z+c z+c
∂w ∂w ∂w
14. 2bc(ax − by) − ac(ax − by − cz) − ab(ax − by − 3cz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.1.21 with s1 = 0, s2 = 2, and s3 = 4.
∂w ∂w ∂w
15. bc(y − z) + ac(z − x) + ab(x − y) = 0.
∂x ∂y ∂z
Integral basis: u1 = ax + by + cz, u2 = ax2 + by 2 + cz 2 .
∂w ∂w ∂w
16. bc(by − 2cz) + ac(3cz − ax) + ab(2ax − 3by) = 0.
∂x ∂y ∂z
Integral basis: u1 = 3ax + 2by + cz, u2 = a2 x2 + b2 y 2 + c2 z 2 .
∂w ∂w ∂w
17. 2bc(by − cz) − ac(4ax − 3by − cz) + 3ab(4ax − by − 3cz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.1.21. Integral basis:
(8ax − 5by − 3cz)2

u1 = 3ax − 3by − cz, u2 = .
2ax − by − cz
∂w ∂w ∂w
18. (ax + y − z) − (x + ay − z) + (a − 1)(y − x) = 0.
∂x ∂y ∂z
p
This is a special case of equation 2.1.1.21 with s1 = 0 and s2,3 = ± (a + 3)(a − 1).
One of the integrals: u1 = x + y + z.
∂w ∂w
19. 2bc(3ax − 2by + cz) − 2ac(2ax − 5by + 3cz)
∂x ∂y
∂w
+ ab(2ax − 6by + 11cz) = 0.
∂z
This is a special case of equation 2.1.1.21 with s1 = 3abc, s2 = 6abc, and s3 = 18abc.
Integral basis:
(2ax + 2by + cz)2 (2ax − by − 2cz)3

u1 = , u2 = .
2ax − by − 2cz ax − 2by + 2cz
∂w ∂w ∂w
20. (Ax + cy + bz) + (cx + By + az) + (bx + ay + Cz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.1.21, where s is the root of the cubic equation
(A − s)(B − s)(C − s) − [a2 (A − s) + b2 (B − s) + c2 (C − s)] + 2abc = 0.
∂w ∂w
21. (a1 x + b1 y + c1 z + d1 ) + (a2 x + b2 y + c2 z + d2 )
∂x ∂y
∂w
+ (a3 x + b3 y + c3 z + d3 ) = 0.
∂z
An integral basis u1 , u2 of this equation is determined by the solution of the cubic equation

a1 − s b1 c1

a2 b2 − s c2 = 0, (1)

a3 b3 c3 − s
the solution of the linear algebraic system



αa1 + βa2 + γa3 = αs,
αb1 + βb2 + γb3 = βs, (2)


αc1 + βc2 + γc3 = γs,
and the value of the coefficient
D = αd1 + βd2 + γd3 . (3)
First, one finds the roots s of the cubic equation (1) and determines the solutions α, β, γ
of the linear system (2). Then one calculates the coefficient D of (3). Three cases are
possible.
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 143
1. If all s = D = 0, then one of the integrals has the form
u1 = αx + βy + γz.
Another integral, u2 , can be obtained by using the transformation specified in Section

13.1.3 (see paragraph The method of reducing the number of independent variables).
2. If equation (1) has two different nonzero roots, s1 and s2 , then two sets of numbers
αk , βk , γk (k = 1, 2), not all equal to zero within each set, can be found from system (2).
Then one of the integrals is
(α1 x + β1 y + γ1 z + D1 /s1 )s2

u1 = .
(α2 x + β2 y + γ2 z + D2 /s2 )s1
3. If equation (1) has three different roots, s1 , s2 , and s3 , other than zero, then the
integrals
(α1 x + β1 y + γ1 z + D1 /s1 )s2 (α1 x + β1 y + γ1 z + D1 /s1 )s3

u1 = , u2 =
(α2 x + β2 y + γ2 z + D2 /s2 )s1 (α3 x + β3 y + γ3 z + D3 /s3 )s1
form an integral basis. If there are multiple or zero roots, then we can take the uk which is
not constant to be one of the integrals and then use the transformation specified in Section
13.1.3 (for u = uk ).
◮ Coefficients of equations are quadratic in x, y, and z.
∂w ∂w ∂w
22. + a1 xy + b1 x2 + c1 x + (a2 xy + b2 x2 + c2 x = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.4 with f1 (x) = a1 x, f2 (x) = b1 x2 +c1 x, g1 (x) = a2 x,
and g2 (x) = b2 x2 + c2 x.
∂w ∂w ∂w
23. + a1 xy + b1 x2 + c1 x + (a2 xz + b2 x2 + c2 x = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.5 with f1 (x) = a1 x, f2 (x) = b1 x2 +c1 x, g1 (x) = a2 x,
and g2 (x) = b2 x2 + c2 x.
∂w ∂w ∂w
24. + a1 xy + b1 x2 + c1 x + (a2 yz + b2 y 2 + c2 y = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.12 with f1 (x) = a1 x, f2 (x) = b1 x2 + c1 x, g1 (y) =

a2 y, and g2 (y) = b2 y 2 + c2 y.
∂w ∂w ∂w
25. + a1 xy + b1 y 2 + (a2 xz + b2 z 2 = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.8 with k = m = 2, f1 (x) = a1 x, f2 (x) = b1 ,
g1 (x) = a2 x, and g2 (x) = b2 .
∂w ∂w ∂w
26. a + xz − xy = 0.
∂x ∂y ∂z
Integral basis: 2 2
2 2 x x
u1 = y + z , u2 = y sin + z cos .
2a 2a
The function u2 = x2 + 2a arctan(z/y) can also be taken to be the second integral.
∂w ∂w ∂w
27. cx + cy + (ax2 + by 2 ) = 0.
∂x ∂y ∂z
Integral basis: u1 = ax2 + by 2 − 2cz, u2 = y/x.
∂w ∂w ∂w
28. cz − a(2ax − b)y + a(2ax − b)z = 0.
∂x ∂y ∂z
Integral basis: u1 = yz, u2 = ax(ax − b) − cz.
∂w ∂w ∂w
29. acx2 − acxy − b2 y 2 = 0.
∂x ∂y ∂z
Integral basis: u1 = xy, u2 = 3acxyz − b2 y 3 .
∂w ∂w ∂w
30. ax2 + by 2 + cz 2 = 0.
∂x ∂y ∂z
Any two of the functions
1 1 1 1 1 1
u1 = − , u2 = − , u3 = −
by ax cz by ax cz
form an integral basis.
∂w ∂w ∂w
31. abx2 + cz 2 + 2abxz = 0.
∂x ∂y ∂z
x2 cz 2
Integral basis: u1 = , u2 = by − .
z 3ax
∂w ∂w ∂w
32. bcxy + a2 cx2 − by(2ax + cz) = 0.
∂x ∂y ∂z
Integral basis: u1 = a2 x2 − by 2 , u2 = x(ax + cz).
∂w ∂w ∂w
33. bcxy + c2 yz + b2 y 2 = 0.
∂x ∂y ∂z
by + cz
Integral basis: u1 = b2 y 2 − c2 z 2 , u2 = .
x
∂w ∂w ∂w
34. xy + y(y − a) + z(y − a) = 0.
∂x ∂y ∂z
y y−a
z x
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 145
∂w ∂w ∂w
35. by 2 − axy + cxz = 0.
∂x ∂y ∂z
Integral basis: u1 = ax2 + by 2 , u2 = y c z a .
∂w ∂w ∂w
36. cxz + 2axy − (2ax + cz)z = 0.
∂x ∂y ∂z
Integral basis: u1 = x(ax + cz), u2 = xyz.
∂w ∂w ∂w
37. cxz + cyz + abxy = 0.
∂x ∂y ∂z
y
Integral basis: u1 = , u2 = cz 2 − abxy.
x
∂w ∂w ∂w
38. cxz − cyz + (by 2 − ax) = 0.
∂x ∂y ∂z
Integral basis: u1 = xy, u2 = 2ax + by 2 + cz 2 .
∂w ∂w ∂w
39. cxz − cyz + (ax2 + by 2 ) = 0.
∂x ∂y ∂z
Integral basis: u1 = xy, u2 = ax2 − by 2 − cz 2 .
∂w ∂w ∂w
40. xz + yz + (ax2 + ay 2 + bz 2 ) = 0.
∂x ∂y ∂z
y a(x2 + y 2 ) + (b − 1)z 2
x (x2 + y 2 )b
∂w ∂w ∂w
41. 2cxz + 2cyz + (cz 2 − ax2 − by 2 ) = 0.
∂x ∂y ∂z
ax2 + by 2 + cz 2 ax2 + by 2 + cz 2
x y
∂w ∂w ∂w
42. bcyz + acxz + abxy = 0.
∂x ∂y ∂z
Integral basis: u1 = ax2 − by 2 , u2 = cz 2 − by 2 .
∂w ∂w ∂w
43. bc(x2 − a2 ) + c(bxy + acz) + b(cxz + aby) = 0.
∂x ∂y ∂z
by + cz by − cz
x−a x+a
∂w ∂w ∂w
44. bx(by + c) + (b2 y 2 − acx)
+ b2 yz = 0.
∂x ∂y
∂z

z ax − c ax + by
Integral basis: u1 = , u2 =
+ ln .
ax + by ax + by x
∂w ∂w ∂w
45. x(by − cz) + y(cz − ax) + z(ax − by) = 0.
∂x ∂y ∂z
Integral basis: u1 = ax + by + cz, u2 = xyz.
∂w ∂w ∂w
46. a(y + β)(z + γ) − b(x + α)(z + γ) − c(x + α)(y + β) = 0.
∂x ∂y ∂z
Integral basis: u1 = b(x + α)2 + a(y + β)2 , u2 = c(x + α)2 + a(z + γ)2 .
∂w ∂w ∂w
47. bc(acxz + b2 y 2 ) + ac(bcyz − 2a2 x2 ) − ab(2abxy + c2 z 2 ) = 0.
∂x ∂y ∂z
Integral basis: u1 = 2acxz − b2 y 2 , u2 = a2 x2 + bcyz.
∂w ∂w ∂w
48. a(y 2 + z 2 ) + x(bz − ay) − x(by + az) = 0.
∂x ∂y ∂z

Integral basis: u1 = x2 + y 2 + z 2 , u2 = 2a arctan(y/z) + b ln y 2 + z 2 .
∂w ∂w ∂w
49. b(by + cz)2 − ax(by + 2cz) + abxz = 0.
∂x ∂y ∂z
Integral basis: u1 = z(by + cz), u2 = ax2 + b2 y 2 − c2 z 2 .
∂w ∂w ∂w
50. (f0 x − f1 ) + (f0 y − f2 ) + (f0 z − f3 ) = 0,
∂x ∂y ∂z
fn = an + bn x + cn y + dn z.
Hesse’s equation (n = 0, 1, 2, 3).
By introducing the homogeneous coordinates x = ξ/τ , y = η/τ , and z = ζ/τ , we
arrive at an equation with linear coefficients for w = w(τ, ξ, η, ζ) having four independent
variables:
∂v ∂v ∂v ∂v
g0 + g1 + g2 + g3 = 0,
∂τ ∂ξ ∂η ∂ζ
where gn = an τ + bn ξ + cn η + dn ζ. For the solution of this equation, see 2.4.9.12 with
n = 4.
◮ Coefficients of equations contain other powers of x, y, and z.

∂w ∂w ∂w
51. 2b2 xz + by(b2 z 2 + 1) + axy(bz + 1)2 = 0.
∂x ∂y ∂z
Integral basis:

bz − axy axy 1 1
u1 = bz−axy, u2 =
ln − − ln |x|.
(bz − axy + 1)2 bz + 1 (bz + 1)(bz − axy + 1) 2
∂w ∂w ∂w
52. bcxy 2 + 2bcy 3+ 2(cyz − ax2 )2 = 0.
∂x ∂y ∂z

x2 by
Integral basis: u1 = , u2 = y exp .
y cyz − ax2
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 147
∂w ∂w ∂w
53. bc2 y 2 z + ac2 xz 2 − abxy 2 = 0.
∂x ∂y ∂z
Integral basis: u1 = ax2 + c2 z 2 , u2 = by 3 + c2 z 3 .
∂w ∂w ∂w
54. x(by 2 − cz 2 ) + y(cz 2 − ax2 ) + z(ax2 − by 2 ) = 0.
∂x ∂y ∂z
Integral basis: u1 = ax2 + by 2 + cz 2 , u2 = xyz.

∂w ∂w ∂w
55. by(3ax2 + by 2 + cz 2 ) − 2ax(ax2 + cz 2 ) + 2abxyz = 0.
∂x ∂y ∂z
ax2 + by 2 + cz 2 2ax2 + by 2
z z2
∂w
56. b[a(a2 x2 + b2 y 2 − 1)x + by]
∂x
∂w ∂w
+ a[b(a2x2 + b2 y 2 − 1)y − ax] + 2abz = 0.
∂y ∂z
The transformation ax = ξ, by = η leads to an equation of the same form with a = b = 1.
Integral basis:

a2 x2 + b2 y 2 − 1 by by
u1 = exp 2 arctan , u2 = z exp 2 arctan .
a2 x2 + b2 y 2 ax ax
∂w ∂w ∂w
57. x(b3 y 3 − 2a3 x3 ) + y(2b3 y 3 − a3 x3 ) + 9z(a3 x3 − b3 y 3 ) = 0.
∂x ∂y ∂z
ax by
Integral basis: u1 = x3 y 3 z, u2 = 2 2
+ 2 2.
b y a x
∂w ∂w ∂w
58. ax2 (abxy−c2z 2 ) + axy(abxy−c2z 2 ) + byz(bcyz+2a2 x2 ) = 0.
∂x ∂y ∂z
y a2 bx2 y + b2 cy 2 z + ac2 xz 2
x yz
∂w ∂w ∂w
59. x(cz 4 − by 4 ) + y(ax4 − 2cz 4 ) + z(2by 4 − ax4 ) = 0.
∂x ∂y ∂z
Integral basis: u1 = ax4 + by 4 + cz 4 , u2 = x2 yz.
∂w ∂w p ∂w
60. x +y + a x2 + y 2 = 0.
∂x ∂y ∂z
p
Integral basis: u1 = y/x, u2 = a x2 + y 2 − z.
∂w ∂w p ∂w
61. x +y + z−ax2 + y 2 + z 2 = 0.
∂x ∂y ∂z
p
Integral basis: u1 = y/x, u2 = xa−1 z + x2 + y 2 + z 2 .
p ∂w
p ∂w p p ∂w
62. z y2 + z2 +az x2 + z 2 − x y 2 + z 2 +ay x2 + z 2 = 0.
∂x ∂y ∂z
Integral basis:
u1 = r 2 , u2 = a arcsin(x/r) − arcsin(y/r), where r 2 = x2 + y 2 + z 2 .
p ∂w p ∂w p ∂w
63. (y − z) f (x) + (z − x) f (y) + (x − y) f (z) = 0,
∂x ∂y ∂z
where f (t) = a6 t + a5 t + a4 t + a3 t + a2 t2 + a1 t + a0 .
6 5 4 3
Integral basis:
p p p
(y−z) f (x)+(z−x) f (y)+(x−y) f (z) 2
u1 = −a6(x+y+z)2 −a5(x+y+z),
(y−z)(z−x)(x−y)
2 2 p p p
y z (y−z) f (x)+z 2x2(z−x) f (y)+x2y 2(x−y) f (z) 2
u2 =
xyz(y−z)(z−x)(x−y)

1 1 1 2 1 1 1
−a0 + + −a1 + + .
x y z x y z
◮ Coefficients of equations contain arbitrary powers of x, y, and z.
∂w ∂w ∂w
64. + axn y m + bxν y µ z λ = 0.
∂x ∂y ∂z
1◦ . Integral basis for m 6= 1 and n 6= −1:
1 k a
u1 = y − xn+1 , k = 1 − m;
k
 n + 1
Z


 z 1−λ ν ak n+1 µ/k
 − b x ku1 + x dx if λ 6= 1,
1−λ n+1
u2 = Z


 ν ak n+1 µ/k
 ln |z| − b x ku1 + x dx if λ = 1.
n+1
In the integration, u1 is considered a parameter.

2◦ . Integral basis for m = 1 and n 6= −1:
a
u1 = ln |y| − xs , s = n + 1;
 1−λ s Z

 z µ aµ s ν aµ s
 − by exp − x x exp x dx if λ 6= 1,
u2 = 1 − λ s Z s

 aµ aµ s
ln |z| − by µ exp − xs xν exp x dx if λ = 1.
s s
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 149
3◦ . Integral basis for m 6= 1 and n = −1:

1
u1 = y k − a ln |x|, k = 1 − m;
k
 1−λ Z
 z µ/k
 − b xν ku1 + ak ln |x| dx if λ 6= 1,
u2 = 1 − λ Z

ln |z| − b xν ku1 + ak ln |x| µ/k dx if λ = 1.

4◦ . Integral basis for m = 1 and n = −1:
 1−λ

 z bxν+1 y µ

 − if λ 6= 1, aµ + ν 6= −1;

 1−λ aµ + ν + 1

 z 1−λ

− by µ x−aµ ln |x| if λ 6= 1, aµ + ν = −1;
u1 = x−a y, u2 = 1 − λ

 bxν+1 y µ

 ln |z| − if λ = 1, aµ + ν 6= −1;



 aµ + ν + 1

ln |z| − by µ x−aµ ln |x| if λ = 1, aµ + ν = −1.
∂w ∂w ∂w
65. + a1 xn1 y + b1 xm1 + (a2 xn2 y + b2 xm2 = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.4 with f1 (x) = a1 xn1 , f2 (x) = b1 xm1 , g1 (x) =
a2 xn2 , and g2 (x) = b2 xm2 .
∂w ∂w ∂w
66. + a1 xn1 y + b1 xm1 + (a2 xn2 z + b2 xm2 = 0.
∂x ∂y ∂z
a2 xn2 , and g2 (x) = b2 xm2 .
∂w ∂w ∂w
67. + a1 xn1 y + b1 xm1 + (a2 y n2 z + b2 y m2 = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.12 with f1 (x) = a1 xn 1 ,
f2 (x) = b1 xm1 , g1 (y) =
a2 y n2 , and g2 (y) = b2 y m2 .
∂w ∂w ∂w
68. + a1 xn1 y + b1 xm1 y k1 + (a2 xn2 z + b2 xm2 z k2 = 0.
∂x ∂y ∂z
∂w ∂w ∂w
69. axn + by m + cz l = 0.
∂x ∂y ∂z
1◦ . Integral basis for n 6= 1 and m 6= 1:
 c a 1−l
b a  x1−n − z if l 6= 1,
u1 = x1−n − y 1−m , u2 = 1 − n 1 − l
1−n 1−m  c x1−n − a ln |z| if l = 1.
1−n
2◦ . Integral basis for m = l = 1:

c 
−c/a
x1−n − a ln |y| if n 6= 1,
u1 = x z, u2 = 1 − n
x−b/a y if n = 1.
∂w ∂w ∂w
70. ay m + bxn + cz l = 0.
∂x ∂y ∂z
1◦ . Integral basis for n 6= −1 and m 6= −1:
b a
u1 = xn+1 − y m+1 ,
n+1 m+1
 Z − m

 b(m + 1) n+1 m + 1 m+1 a 1−l

c x − u1 dx − z if l 6= 1;
a(n + 1) a 1−l
u2 = Z − m

 b(m + 1) n+1 m + 1 m+1

c x − u1 dx − a ln |z| if l = 1.
a(n + 1) a
2◦ . Integral basis for n 6= −1 and m = −1:
b
u1 = xn+1 − a ln |y|,
n
 + 1 Z

 b n+1 b n+1 a 1−l
 cy exp − x exp x dx − z if l 6= 1;
a(n + 1) Z a(n + 1) 1 − l
u2 =

 b b
cy exp − xn+1 exp xn+1 dx − a ln |z| if l = 1.
a(n + 1) a(n + 1)
3◦ . Integral basis for n = m = −1:


 a2 1−l

 (a + b)cxy − z if l 6= 1, a + b 6= 0;

 1−l
 a 1−l
u1 = x−b/a y, u2 = cxy ln |x| − 1 − l z if l 6= 1, a + b = 0;




 (a + b)cxy − a2 ln |z| if l = 1, a + b 6= 0;

cxy ln |x| − a ln |z| if l = 1, a + b = 0.
∂w ∂w ∂w
71. x(y n − z n ) + y(z n − xn ) + z(xn − y n ) = 0.
∂x ∂y ∂z
Integral basis: u1 = xyz, u2 = xn + y n + z n .

∂w ∂w ∂w
1. a + beαx + ceβy = 0.
∂x ∂y ∂z
Z
b αx eβy dy
Integral basis: u1 = e − ay, u2 = αz − c . In the integration, u1 is
α ay + u1
considered a parameter.
∂w ∂w ∂w
2. a + beαx + ceγz = 0.
∂x ∂y ∂z
b αx a
Integral basis: u1 = e − ay, u2 = cx + e−γz .
α γ
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 151
∂w ∂w ∂w
3. a + beβy + ceγz = 0.
∂x ∂y ∂z
a −βy c b
Integral basis: u1 = bx + e , u2 = − e−βy + e−γz .
β β γ
∂w ∂w ∂w
4. + A1 eα1 x + B1 eν1 x+λy + A2 eα2 x + B2 eν2 x+βz = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.10 with f1 (x) = A1 eα1 x , f2 (x) = B1 eν1 x , g1 (x) =
A2 eα2 x , and g2 (x) = B2 eν2 x .
∂w ∂w ∂w
5. aeαx + beβy + ceγz = 0.
∂x ∂y ∂z
1 −αx 1 −βy 1 1 −γz
Integral basis: u1 = − e + e , u2 = − e−βy + e .
aα bβ bβ cγ
∂w ∂w ∂w
6. aeβy + beαx + ceγz = 0.
∂x ∂y ∂z
1 αx 1 βy βy − αx 1 −γz
Integral basis: u1 = − e + e , u2 = αx βy
+ e .
aα bβ bβe − aαe cγ
∂w ∂w ∂w
7. (a1 + a2 eαx ) + (b1 + b2 eβy ) + (c1 + c2 eγz ) = 0.
∂x ∂y ∂z
Integral basis:
1 1
u1 = αx − ln(a1 + a2 eαx ) − βy − ln(b1 + b2 eβy ) ,
a1 α b1 β
1 1
u2 = αx − ln(a1 + a2 eαx ) − γz − ln(c1 + c2 eγz ) .
a1 α c1 γ
∂w ∂w ∂w
8. eβy (a1 + a2 eαx ) + eαx (b1 + b2 eβy ) + ceβy+γz = 0.
∂x ∂y ∂z
Integral basis:
1 1 1 1
u1 = ln(a1+a2eαx)− ln(b1+b2eβy ), u2 = αx−ln(a1+a2eαx) + e−γz .
a2α b1β a1α cγ

∂w ∂w ∂w
9. ayeαx + beβy + ceγz = 0.
∂x ∂y ∂z
1 −αx βy + 1 −βy 1 1 −γz
Integral basis: u1 = − e + e , u2 = − e−βy + e .
aα bβ 2 bβ cγ
∂w ∂w ∂w
10. axeαx + byeβy + czeγz = 0.
∂x ∂y ∂z
Integral basis:
Z Z
1 −1 −αx −2 −αx 1 −1 −βy −2 −βy
u1 = − x e + x e dx + y e + y e dy ,
aα bβ
Z Z
1 −1 −αx −2 −αx 1 −1 −γz −2 −γz
u2 = − x e + x e dx + z e + z e dz .
aα cγ
∂w ∂w 2 ∂w
11. + y 2 + aeαx (α − aeαx ) + z + bz + ceβx (β − b − ceβx ) = 0.
∂x ∂y ∂z
Integral basis:
Z
E1 2a αx
u1 = + E1 dx, E1 = exp e ,
y − aeαx α
Z
E2 2c βx
u2 = + E2 dx, E2 = exp e + bx .
z − ceβx β
∂w ∂w 2 ∂w
12. + y 2 +by +aeαx (α−b−aeαx ) + z +ceβx (z −k)−k2 = 0.
∂x ∂y ∂z
Integral basis:
Z
E1 2a αx
u1 = + E1 dx, E1 = exp e + bx ,
y − aeαx α
Z
E2 c βx
u2 = + E2 dx, E2 = exp e + 2kx .
z−k β
∂w ∂w ∂w
13. + ay 2 eαx + be−αx + deβx z 2 + ceγx (γ − cde(β+γ)x ) = 0.
∂x ∂y ∂z
Integral basis:
Z
dv
u1 = − x, v = eαx y;
av 2
+ αv + b
Z
E 2cd (β+γ)x
u2 = + d eβx E dx, E = exp e .
z − ceγx β+γ
∂w αx 2 ∂w ∂w
14. + be y +aeβx (β−abe(α+β)x ) + cz 2 eγx +dz+ke−γx = 0.
∂x ∂y ∂z
Integral basis:
Z
dv
u1 = − x, v = eγx z,
cv 2
+ (d + γ)v + k
Z
E 2ab (α+β)x
u2 = + b eαx E dx, E = exp e .
y − aeβx α+β
∂w ∂w ∂w
15. + aeαx y 2 + by + ce−αx + eβx z 2 + deγx (z + βe−βx ) = 0.
∂x ∂y ∂z
Integral basis:
Z
dv
u1 = − x, v = eαx y,
av 2
+ (b + α)v + c
Z
E d γx
u2 = + Eeβx dx, E = exp e − 2βx .
z + βe−βx γ
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 153
∂w ∂w γx 2 ∂w
16. + eαx y 2 +ayeβx +aαe(β−α)x + γe z +beδx (z +e−γx ) =
∂x ∂y ∂z
0.
Integral basis:
Z
E1 a βx
u1 = + eαx E1 dx, E1 = exp e − 2αx ,
y + αe−αx β
Z
E2 b δx
u2 = + γ eγx E2 dx, E2 = exp e − 2γx .
z + e−γx δ
∂w ∂w γx ∂w
17. + αeαx y 2 + aeβx (y + e−αx ) + e (z − beδx )2 + bδeδx = 0.
∂x ∂y ∂z
Integral basis:
Z
E 1 1
u1 = +α eαx E dx, u2 = + eγx ,
y + e−αx z − be−δx γ

a βx
where E = exp e − 2αx .
β
∂w ∂w
18. x + (a1 eαx y 2 + βy + a1 b21 x2β eαx )
∂x ∂y
∂w
+ a2 x z e + (b2 xn eλx − n)z + ceλx
2n 2 λx
= 0.
∂z
Integral basis:
Z
y
u1 = arctan − a b
1 1 xβ−1 eαx dx,
b1 xβ
Z Z
dv
u2 = − xn−1 eλx dx, v = xn z.
a2 v 2 + b2 v + c
∂w ∂w ∂w
19. + a1 eλ1 x y + b1 eβ1 x y k + a2 eλ2 x z + b2 eβ2 x z m = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.8 with f1 (x) = a1 eλ1 x , f2 (x) = b1 eβ1 x , g1 (x) =
a2 eλ2 x , and g2 (x) = b2 eβ2 x .
∂w ∂w ∂w
20. + a1 eβ1 x y + b1 eγ1 x y k + a2 eβ2 x + b2 eγ2 x+λz = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.9 with f1 (x) = a1 eβ1 x , f2 (x) = b1 eγ1 x , g1 (x) =
a2 eβ2 x , and g2 (x) = b2 eγ2 x .
∂w ∂w ∂w
21. + a1 xn + b1 xm eλy + a2 xk + b2 xs eβz = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.10 with f1 (x) = a1 xn , f2 (x) = b1 xm , g1 (x) = a2 xk ,

and g2 (x) = b2 xs .
∂w ∂w ∂w
22. (axn eλy + bxy m ) + eµy + (cy l z k + dy p z) = 0.
∂x ∂y ∂z
Integral basis:
Z Z
1−n (λ−µ)y m −µy
u1 = x E1 + a(n − 1) e E1 dy, E1 = exp b(n − 1) y e dy ,
Z Z
1−k
u2 = z E2 + c(k − 1) y l e−µy E2 dy, p −µy
E2 = exp d(k − 1) y e dy .
∂w 2 2 ∂w 2 ∂w
23. + y 2 + 2aαeαx − a2 e2αx + ce−2βx z 2 + 2βxz + b2 c = 0.
∂x ∂y ∂z
Integral basis:
Z Z
E αx2
u1 = + E dx, E = exp 2a e dx ,
y − aeαx2
Z
1 −βx2 2
u2 = arctan e z − bc e−βx dx.
b
∂w 2 ∂w
24. + ae−2αx y 2 + 2αxy + ab2
∂x ∂y
2 ∂w
+ cxβ z 2 + 2γxz + cd2 xβ e2γx = 0.
∂z
Integral basis:
Z Z
1 −αx2 2 1 −γx2 2
u1 = arctan e y − ab e−αx dx, u2 = arctan e z − cd xβ eγx dx.
b d

∂w ∂w ∂w
1. a +b + c sinh(λx) = 0.
∂x ∂y ∂z
Integral basis: u1 = bx − ay, u2 = aλz − c cosh(λx).
∂w ∂w ∂w
2. a + b sinh(βy) + c sinh(λx) = 0.
∂x ∂y ∂z
βy

Integral basis: u1 = bβx − a lntanh , u2 = aλz − c cosh(λx).
2
∂w ∂w ∂w
3. a + b sinh(βy) + c sinh(γz) = 0.
∂x ∂y ∂z
βy γz

Integral basis: u1 = bβx − a lntanh , u2 = cγx − a lntanh .
2 2
∂w ∂w ∂w
4. a sinh(λx) + b sinh(βy) + c sinh(γz) = 0.
∂x ∂y ∂z
Integral basis:
1 λx 1 βy 1 λx 1 γz
u1 = lntanh − lntanh , u2 = lntanh − lntanh .
aλ 2 bβ 2 aλ 2 cγ 2
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 155
∂w ∂w ∂w
5. a sinh(βy) + b sinh(λx) + c sinh(γz) = 0.
∂x ∂y ∂z
Integral basis:
u1 = bβ cosh(λx) − aλ cosh(βy),
Z
dx
u2 = c|λ|γ p + 2 sign(a) arctanh(eγz ).
2 2
[bβ cosh(λx) − u1 ] − a λ2

∂w ∂w ∂w
6. a sinh(βy) + b sinh(λx) + c sinh(λx) sinh(βy) sinh(γz) = 0.
∂x ∂y ∂z
1 1 1 1 γz
Integral basis: u1 = cosh(λx) − cosh(βy), u2 = cosh(λx) − lntanh .
aλ bβ aλ cγ 2

∂w ∂w ∂w
7. a +b + c cosh(βx) = 0.
∂x ∂y ∂z
Integral basis: u1 = bx − ay, u2 = aβz − c sinh(βx).
∂w ∂w ∂w
8. a + b cosh(βx) + c cosh(λx) = 0.
∂x ∂y ∂z
Integral basis: u1 = aβy − b sinh(βx), u2 = aλz − c sinh(λx).
∂w ∂w ∂w
9. a + b cosh(βy) + c cosh(λx) = 0.
∂x ∂y ∂z
βy

Integral basis: u1 = bβx − 2a arctantanh , u2 = aλz − c sinh(λx).
2
∂w ∂w ∂w
10. a + b cosh(βy) + c cosh(γz) = 0.
∂x ∂y ∂z
βy γz

Integral basis: u1 = bβx − 2a arctantanh , u2 = cγx − 2a arctantanh .
2 2
∂w ∂w ∂w
11. a cosh(λx) + b cosh(βy) + c cosh(γz) = 0.
∂x ∂y ∂z
Integral basis:
2 2 2 2
u1 = arctan eλx − arctan eβy , u2 = arctan eλx − arctan eγz .
aλ bβ aλ cγ
∂w ∂w ∂w
12. a cosh(βy) + b cosh(λx) + c cosh(γz) = 0.
∂x ∂y ∂z
Integral basis:
u1 = bβ sinh(λx) − aλ sinh(βy),
Z
dx
u2 = c|λ|γ p − 2 sign(a) arctan(eγz ).
[bβ sinh(λx) − u1 ]2 + a2 λ2
∂w ∂w ∂w
13. a +b + c tanh(γz) = 0.
∂x ∂y ∂z

Integral basis: u1 = bx − ay, u2 = cγx − a lnsinh(γz).
∂w ∂w ∂w
14. a + b tanh(βx) + c tanh(λx) = 0.
∂x ∂y ∂z

Integral basis: u1 = aβy − b lnsinh(βx), u2 = aλz − c lnsinh(λx).
∂w ∂w ∂w
15. a + b tanh(βy) + c tanh(λx) = 0.
∂x ∂y ∂z

Integral basis: u1 = bβx − a lnsinh(βy), u2 = aλz − c lncosh(λx).
∂w ∂w ∂w
16. a + b tanh(βy) + c tanh(γz) = 0.
∂x ∂y ∂z

Integral basis: u1 = bβx − a lnsinh(βy), u2 = cγx − a lnsinh(γz).
∂w ∂w ∂w
17. a tanh(λx) + b tanh(βy) + c tanh(γz) = 0.
∂x ∂y ∂z
sinhaλ (βy) sinhaλ (γz)

Integral basis: u1 = , u 2 = .
sinhbβ (λx) sinhcγ (λx)
∂w ∂w ∂w
18. a tanh(βy) + b tanh(λx) + c tanh(γz) = 0.
∂x ∂y ∂z
Integral basis:
Z n
coshaλ (βy) o
u1 = , u2 = cγ coth arccosh1/aλ u1 coshbβ (λx) dx − a lnsinh(γz).
coshbβ (λx)
∂w ∂w ∂w
19. a +b + c coth(λz) = 0.
∂x ∂y ∂z

Integral basis: u1 = bx − ay, u2 = cλx − a ln cosh(λz) .
∂w ∂w ∂w
20. a + b coth(βx) + c coth(λx) = 0.
∂x ∂y ∂z

Integral basis: u1 = aβy − b ln cosh(βx) , u2 = aλz − c ln cosh(λx) .
∂w ∂w ∂w
21. a + b coth(βy) + c coth(λx) = 0.
∂x ∂y ∂z

Integral basis: u1 = bβx − a ln cosh(βy) , u2 = aλz − c lnsinh(λx).
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 157
∂w ∂w ∂w
22. a + b coth(βy) + c coth(λz) = 0.
∂x ∂y ∂z

Integral basis: u1 = bβx − a ln cosh(βy) , u2 = cλx − a ln cosh(λz) .
∂w ∂w ∂w
23. a coth(λx) + b coth(βy) + c coth(γz) = 0.
∂x ∂y ∂z
coshaλ (βy) coshaλ (γz)

Integral basis: u1 = , u 2 = .
coshbβ (λx) coshcγ (λx)
∂w ∂w ∂w
24. a coth(βy) + b coth(λx) + c coth(γz) = 0.
∂x ∂y ∂z
Integral basis:
Z n
sinhaλ (βy) o
u1 = , u2 = cγ tanh arcsinh1/aλ u1 sinhbβ (λx) dx − a ln cosh(γz) .
sinhbβ (λx)

∂w ∂w ∂w
25. a sinh(λx) + b sinh(βy) + c cosh(γz) = 0.
∂x ∂y ∂z
Integral basis:
1 λx 1 βy 1 λx 2
u1 = lntanh − lntanh , u2 = lntanh − arctan eγz .
aλ 2 bβ 2 aλ 2 cγ
∂w ∂w ∂w
26. a sinh(λx) + b cosh(βy) + c cosh(γz) = 0.
∂x ∂y ∂z
Integral basis:
1 λx 2 1 λx 2
u1 = lntanh − arctan eβy , u2 = lntanh − arctan eγz .
aλ 2 bβ aλ 2 cγ
∂w ∂w ∂w
27. a sinh(βy) + b sinh(λx) + c sinh(λx) sinh(βy) cosh(γz) = 0.
∂x ∂y ∂z
1 1 1 2
Integral basis: u1 = cosh(λx) − cosh(βy), u2 = cosh(λx) − arctan eγz .
aλ bβ aλ cγ
∂w ∂w ∂w
28. a cosh(βy) + b tanh(λx) + c cosh(γz) = 0.
∂x ∂y ∂z
Integral basis:

u1 = bβ lncosh(λx) − aλ sinh(βy),
Z
dx
u2 = c|λ|γ p − 2 sign(a) arctan(eγz ).
2 2
[bβ ln |cosh(λx)| − u1 ] + a λ2

∂w ∂w ∂w
29. a coth(βy) + b tanh(λx) + c tanh(γz) = 0.
∂x ∂y ∂z
Integral basis:
Z n
sinhaλ (βy) o
u1 = , u2 = cγ coth arcsinh1/aλ u1 coshbβ (λx) dx − a lnsinh(γz).
coshbβ (λx)
∂w ∂w ∂w
30. a coth(βy) + b tanh(λx) + c coth(γz) = 0.
∂x ∂y ∂z
Integral basis:
Z n
sinhaλ (βy) o
u1 = , u2 = cγ coth arcsinh1/aλ u1 coshbβ (λx) dx − a lncosh(γz).
coshbβ (λx)

∂w ∂w ∂w
1. a +b + c ln(βy) ln(λz) = 0.
∂x ∂y ∂z
Z
dz
Integral basis: u1 = bx − ay, u2 = cy 1 − ln(βy) + b .
ln(λz)
∂w ∂w ∂w
2. a + b ln(βx) + c ln(λx) = 0.
∂x ∂y ∂z

Integral basis: u1 = bx 1 − ln(βx) + ay, u2 = cx 1 − ln(λx) + az.
∂w ∂w ∂w
3. a + b ln(βx) ln(λy) + c ln(µx) ln(γz) = 0.
∂x ∂y ∂z
Integral basis:
Z Z
dy dz
u1 = bx 1 − ln(βx) + a , u2 = cx 1 − ln(µx) + a .
ln(λy) ln(γz)
∂w ∂w ∂w
4. a ln(βx) + b ln(λy) + c ln(γz) = 0.
∂x ∂y ∂z
Integral basis:
Z Z Z Z
dx dy dx dz
u1 = b −a , u2 = c −a .
ln(βx) ln(λy) ln(βx) ln(γz)
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 159

∂w ∂w ∂w
5. + axn + b lnk (λx) = 0.
∂x ∂y ∂z
Z
a
Integral basis: u1 = y − xn+1 , u2 = z − b lnk (λx) dx.
n+1
∂w ∂w ∂w
6. + ay + c lnk (λx) + bz + s lnn (βx) = 0.
∂x ∂y ∂z
Integral basis:
Z Z
−ax k −ax −bx
u1 = ye −c ln (λx)e dx, u2 = ze −s lnn (βx)e−bx dx.
∂w ∂w ∂w
7. ax + by + c lnn (λx) + s lnk (βy) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.21 with f (x, y) = c lnn (λx) + s lnk (βy).
∂w ∂w ∂w
8. ax ln(λx) + by ln(βy) + cz ln(γz) = 0.
∂x ∂y ∂z

Integral basis: u1 = b lnln(λx) − a lnln(βy), u2 = c lnln(λx) − a lnln(γz).
∂w ∂w ∂w
9. ax ln(λx) + by ln(βy) + cz ln(λx) = 0.
∂x ∂y ∂z

Integral basis: u1 = b lnln(λx) − a lnln(βy), u2 = |x|c |z|−a .
∂w ∂w ∂w
10. ax(ln x)n + by(ln y)m + cz(ln z)k = 0.
∂x ∂y ∂z
Integral basis:
b a c a
u1 = (ln x)1−n − (ln y)1−m , u2 = (ln x)1−n − (ln z)1−k .
1−n 1−m 1−n 1−k

∂w ∂w ∂w
1. a +b + c sin(γz) = 0.
∂x ∂y ∂z
γz

Integral basis: u1 = bx − ay, u2 = cγx − a lntan .
2
∂w ∂w ∂w
2. a + b sin(βy) + c sin(λx) = 0.
∂x ∂y ∂z
βy

Integral basis: u1 = bβx − a lntan , u2 = aλz + c cos(λx).
2
∂w ∂w ∂w
3. a + b sin(βy) + c sin(γz) = 0.
∂x ∂y ∂z
βy γz

Integral basis: u1 = bβx − a lntan , u2 = cγx − a lntan .
2 2
∂w ∂w ∂w
4. a + b sin(λx) sin(βy) +c = 0.
∂x ∂y ∂z
aλ βy
Integral basis: u1 = cx − az, u2 = cos(λx) + lntan .
bβ 2
∂w ∂w ∂w
5. a + b sinn (λx) sinm (βy) + c sink (µx) sinl (γz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.16 with f1 (x) = a, f2 (x) = b sinn (λx), g(y) =
sinm (βy), f3 (x) = c sink (µx), and h(z) = sinl (γz).

∂w ∂w ∂w
6. a +b + c cos(λz) = 0.
∂x ∂y ∂z

Integral basis: u1 = bx − ay, u2 = cλx − a lntan 1
2 λz + π
4
.
∂w ∂w ∂w
7. a + b cos(βx) + c cos(λx) = 0.
∂x ∂y ∂z
Integral basis: u1 = aβy − b sin(βx), u2 = aλz − c sin(λx).
∂w ∂w ∂w
8. a + b cos(βy) + c cos(λx) = 0.
∂x ∂y ∂z

Integral basis: u1 = bβx − a lntan 12 βy + π4 , u2 = aλz − c sin(λx).
∂w ∂w ∂w
9. a + b cos(βy) + c cos(λz) = 0.
∂x ∂y ∂z

Integral basis: u1 = bβx − a lntan 12 βy + π4 , u2 = cλx − a lntan 1
2 λz + π
4
.
∂w ∂w ∂w
10. a + b cosn (λx) cosm (βy) + c cosk (µx) cosl (γz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.16 with f1 (x) = a, f2 (x) = b cosn (λx), g(y) =
cosm (βy), f3 (x) = c cosk (µx), and h(z) = cosl (γz).

∂w ∂w ∂w
11. a +b + c tan(λz) = 0.
∂x ∂y ∂z

Integral basis: u1 = bx − ay, u2 = cλx − a lnsin(λz).
∂w ∂w ∂w
12. a + b tan(βx) + c tan(λx) = 0.
∂x ∂y ∂z

Integral basis: u1 = aβy + b lncos(βx), u2 = aλz + c lncos(λx).
∂w ∂w ∂w
13. a + b tan(βy) + c tan(λx) = 0.
∂x ∂y ∂z

Integral basis: u1 = bβx − a lnsin(βy), u2 = aλz + c lncos(λx).
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 161
∂w ∂w ∂w
14. a + b tan(βy) + c tan(λz) = 0.
∂x ∂y ∂z

Integral basis: u1 = bβx − a lnsin(βy), u2 = cλx − a lnsin(λz).
∂w ∂w ∂w
15. µν tan(λx) + λν tan(µy) + λµ tan(νz) = 0.
∂x ∂y ∂z
sin(λx) sin(µy)
sin(µy) sin(νz)

∂w ∂w ∂w
16. a +b + c cot(λz) = 0.
∂x ∂y ∂z

Integral basis: u1 = bx − ay, u2 = cλx + a lncos(λz).
∂w ∂w ∂w
17. a + b cot(βx) + c cot(λx) = 0.
∂x ∂y ∂z

Integral basis: u1 = aβy − b lnsin(βx), u2 = aλz − c lnsin(λx).
∂w ∂w ∂w
18. a + b cot(βy) + c cot(λx) = 0.
∂x ∂y ∂z

Integral basis: u1 = bβx + a lncos(βy), u2 = aλz − c lnsin(λx).
∂w ∂w ∂w
19. a + b cot(βy) + c cot(λz) = 0.
∂x ∂y ∂z

Integral basis: u1 = bβx + a lncos(βy), u2 = cλx + a lncos(λz).
∂w ∂w ∂w
20. µν cot(λx) + λν cot(µy) + λµ cot(νz) = 0.
∂x ∂y ∂z
cos(λx) cos(µy)
cos(µy) cos(νz)

∂w ∂w ∂w
21. a +b + c sinn (λx) + s cosk (βy) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.18 with f (x, y) = c sinn (λx) + s cosk (βy).
∂w ∂w ∂w
22. a + b sin(βy) + c cos(λx) = 0.
∂x ∂y ∂z
βy

Integral basis: u1 = bβx − a lntan , u2 = aλz − c sin(λx).
2
∂w ∂w ∂w
23. + a sinn (λx) + b cosk (βx) = 0.
∂x ∂y ∂z
Z Z
Integral basis: u1 = y − a sinn (λx) dx, u2 = z − b cosk (βx) dx.
∂w ∂w ∂w
24. + a cosn (λx) + b sink (βy) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.12 with f1 (x) = 0, f2 (x) = a cosn (λx), g1 (y) = 0,
and g2 (y) = b sink (βy).
∂w ∂w ∂w
25. a + b tan(βy) + c cot(λx) = 0.
∂x ∂y ∂z

Integral basis: u1 = bβx − a lnsin(βy), u2 = aλz − c lnsin(λx).
∂w ∂w ∂w
26. + a cotn (λx) + b tank (βy) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.12 with f1 (x) = 0, f2 (x) = a cotn (λx), g1 (y) = 0,
and g2 (y) = b tank (βy).

∂w ∂w ∂w
1. a +b + c arcsinn (λx) arcsink (βz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.16 with f1 (x) = a, f2 (x) = b, f3 (x) = c arcsinn (λx),
g(y) = 1, and h(z) = arcsink (βz).
∂w ∂w ∂w
2. a +b + c arcsinn (λx) arcsinm (βy) arcsink (γz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.19 with f (x, y) = c arcsinn (λx) arcsinm (βy) and
g(z) = arcsink (γz).
∂w ∂w ∂w
3. a + b arcsinn (λx) + c arcsink (βx) = 0.
∂x ∂y ∂z
Integral basis:
Z Z
n
u1 = ay − b arcsin (λx) dx, u2 = az − c arcsink (βx) dx.
∂w ∂w ∂w
4. a + b arcsinn (λx) + c arcsink (βz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.16 with f1 (x) = a, f2 (x) = b arcsinn (λx), f3 (x) = 1,
g(y) = 1, and h(z) = c arcsink (βz).
∂w ∂w ∂w
5. a + b arcsinn (λy) + c arcsink (βz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.15 with f (x) = a, g(y) = b arcsinn (λy), and
h(z) = c arcsink (βz).
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 163
∂w ∂w ∂w
6. a +b + c arccosn (λx) arccosk (βz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.16 with f1 (x) = a, f2 (x) = b, f3 (x) = c arccosn (λx),
g(y) = 1, and h(z) = arccosk (βz).
∂w ∂w ∂w
7. a +b + c arccosn (λx) arccosm (βy) arccosk (γz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.19 with f (x, y) = c arccosn (λx) arccosm (βy), and
g(z) = arccosk (γz).
∂w ∂w ∂w
8. a + b arccosn (λx) + c arccosk (βx) = 0.
∂x ∂y ∂z
Integral basis:
Z Z
u1 = ay − b arccosn (λx) dx, u2 = az − c arccosk (βx) dx.
∂w ∂w ∂w
9. a + b arccosn (λx) + c arccosk (βz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.16 with f1 (x) = a, f2 (x) = b arccosn (λx), f3 (x) = 1,
g(y) = 1, and h(z) = c arccosk (βz).
∂w ∂w ∂w
10. a + b arccosn (λy) + c arccosk (βz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.15 with f (x) = a, g(y) = b arccosn (λy), and
h(z) = c arccosk (βz).
∂w ∂w ∂w
11. a +b + c arctann (λx) arctank (βz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.16 with f1 (x) = a, f2 (x) = b, f3 (x) = c arctann (λx),
g(y) = 1, and h(z) = arctank (βz).
∂w ∂w ∂w
12. a +b + c arctann (λx) arctanm (βy) arctank (γz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.19 with f (x, y) = c arctann (λx) arctanm (βy) and
g(z) = arctank (γz).
∂w ∂w ∂w
13. a + b arctann (λx) + c arctank (βx) = 0.
∂x ∂y ∂z
Integral basis:
Z Z
n
u1 = ay − b arctan (λx) dx, u2 = az − c arctank (βx) dx.
∂w ∂w ∂w
14. a + b arctann (λx) + c arctank (βz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.16 with f1 (x) = a, f2 (x) = b arctann (λx), f3 (x) = 1,
g(y) = 1, and h(z) = c arctank (βz).
∂w ∂w ∂w
15. a + b arctann (λy) + c arctank (βz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.15 with f (x) = a, g(y) = b arctann (λy), and
h(z) = c arctank (βz).

∂w ∂w ∂w
16. a +b + c arccotn (λx) arccot k (βz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.16 with f1 (x) = a, f2 (x) = b, f3 (x) = c arccot n (λx),
g(y) = 1, and h(z) = arccotk (βz).
∂w ∂w ∂w
17. a +b + c arccotn (λx) arccot m (βy) arccot k (γz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.19 with f (x, y) = c arccotn (λx) arccot m (βy) and
g(z) = arccotk (γz).
∂w ∂w ∂w
18. a + b arccotn (λx) + c arccot k (βx) = 0.
∂x ∂y ∂z
Integral basis:
Z Z
n
u1 = ay − b arccot (λx) dx, u2 = az − c arccotk (βx) dx.
∂w ∂w ∂w
19. a + b arccotn (λx) + c arccot k (βz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.16 with f1 (x) = a, f2 (x) = b arccotn (λx), f3 (x) = 1,
g(y) = 1, and h(z) = c arccotk (βz).
∂w ∂w ∂w
20. a + b arccotn (λy) + c arccot k (βz) = 0.
∂x ∂y ∂z
This is a special case of equation 2.1.7.15 with f (x) = a, g(y) = b arccotn (λy), and
h(z) = c arccotk (βz).

∂w ∂w ∂w
1. + f (x) + g(x) = 0.
∂x ∂y ∂z
Integral basis: Z Z
u1 = y − f (x) dx, u2 = z − g(x) dx.
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 165
∂w ∂w ∂w
2. + f (x)(y + a) + g(x)(z + b) = 0.
∂x ∂y ∂z
Integral basis:
Z Z
u1 = ln |y + a| − f (x) dx, u2 = ln |z + b| − g(x) dx.
∂w ∂w ∂w
3. + ay + f (x) + bz + g(x) = 0.
∂x ∂y ∂z
Integral basis:
Z Z
u1 = ye−ax − f (x)e−ax dx, u2 = ze−bx − g(x)e−bx dx.
∂w ∂w ∂w
4. + f1 (x)y + f2 (x) + g1 (x)y + g2 (x) = 0.
∂x ∂y ∂z
Integral basis:
Z Z
u1 = yF (x) − f2 (x)F (x) dx, F (x) = exp − f1 (x) dx ,
Z Z
g1 (x)
u2 = z − ϕ(x)y + f2 (x)ϕ(x) − g2 (x) dx, ϕ(x) = F (x) dx.
F (x)
∂w ∂w ∂w
5. + f1 (x)y + f2 (x) + g1 (x)z + g2 (x) = 0.
∂x ∂y ∂z
Integral basis:
Z Z
Z Z
u2 = zG(x) − g2 (x)G(x) dx, G(x) = exp − g1 (x) dx .
∂w ∂w ∂w
6. + f2 (x)y + f1 (x)z + f0 (x) + g2 (x)y + g1 (x)z + g0 (x) = 0.
∂x ∂y ∂z
One of the integrals has the form
u1 = ϕ(x)y + ψ(x)z + χ(x),
where the functions ϕ(x), ψ(x), and χ(x) are determined by solving the following system
of first-order ordinary differential equations:
ϕ′x + f2 ϕ + g2 ψ = 0,
ψx′ + f1 ϕ + g1 ψ = 0,
χ′x + f0 ϕ + g0 ψ = 0.
In some cases, this system can be integrated in quadrature. For example, this can be done
for g2 ≡ 0 (or f1 ≡ 0); in this case, one should begin the integration with the first (resp.,
second) equation. In the general case, the integration of the system can be reduced to the
solution of a second-order linear ordinary differential equation which follows from the first
two equations.
The general solution of the equation in question can be obtained using the technique de-
scribed in Section 13.1.3 (see paragraph The method of reducing the number of independent
variables).
∂w ∂w ∂w
7. + y 2 − a2 + aλ sinh(λx) − a2 sinh2 (λx) + f (x) sinh(γz) = 0.
∂x ∂y ∂z
Integral basis:
Z
1 γz
u1 = f (x) dx − lntanh ,
γ 2
Z
E 2a
u2 = + E dx, E = exp sinh(λx) .
∂w ∂w ∂w
8. + f1 (x)y + f2 (x)y k + g1 (x)z + g2 (x)z m = 0.
∂x ∂y ∂z
1◦ . For k 6= 1 and m 6= 1, the transformation ξ = y 1−k , η= z 1−m leads to an equation of

the form 2.1.7.5:
∂w ∂w ∂w
+ (1 − k) f1 (x)ξ + f2 (x) + (1 − m) g1 (x)η + g2 (x) = 0.
∂x ∂ξ ∂η
2◦ . For k 6= 1 and m = 1, the substitution ξ = y 1−k leads to an equation of the form 2.1.7.5:
∂w ∂w ∂w
+ (1 − k) f1 (x)ξ + f2 (x) + z g1 (x) + g2 (x) = 0.
∂x ∂ξ ∂z
3◦ . For k = m = 1, see equation 2.1.7.5.
∂w ∂w ∂w
9. + f1 (x)y + f2 (x)y k + g1 (x) + g2 (x)eλz = 0.
∂x ∂y ∂z
The transformation ξ = y 1−k , η= e−λz leads to an equation of the form 2.1.7.5:
∂w ∂w ∂w
+ (1 − k) f1 (x)ξ + f2 (x) − λ g1 (x)η + g2 (x) = 0.
∂x ∂ξ ∂η
∂w ∂w ∂w
10. + f1 (x) + f2 (x)eλy + g1 (x) + g2 (x)eβz = 0.
∂x ∂y ∂z
The transformation ξ = e−λy , η = e−βz leads to an equation of the form 2.1.7.5:
∂w ∂w ∂w
− λ f1 (x)ξ + f2 (x) − β g1 (x)η + g2 (x) = 0.
∂x ∂ξ ∂η
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 167
◮ Coefficients of equations contain arbitrary functions of different variables.
∂w ∂w ∂w
11. x +y + [z + f (x)g(y)] = 0.
∂x ∂y ∂z
Integral basis: Z
y z
u1 = , u2 = − x−2 f (x)g(u1 x) dx.
x x
∂w ∂w ∂w
12. + f1 (x)y + f2 (x) + g1 (y)z + g2 (y) = 0.
∂x ∂y ∂z
Integral basis:
Z Z
Z x Z x
u2 = zG(x, u1 ) − ḡ2 (t, u1 )G(t, u1 ) dt, G(x, u1 ) = exp − ḡ1 (t, u1 ) dt .
x0 x0
Here ḡ1 (x, u1 ) ≡ g1 (y), ḡ2 (x, u1 ) ≡ g2 (y) (y is expressed via x and u1 from the first
integral), and x0 is an arbitrary number.
∂w ∂w ∂w
13. + y 2 − a2 + aλ sinh(λx) − a2 sinh2 (λx) + f (x)g(z) = 0.
∂x ∂y ∂z
Integral basis:
Z Z
dz
u1 = f (x) dx − ,
g(z)
Z
E 2a
∂w ∂w ∂w
14. f (x) + zk + g(y) = 0.
∂x ∂y ∂z
Integral basis:
Z Z Z Z − k
1 k+1 dx k+1
u1 = g(y) dy− z , u2 = − (k+1) g(y) dy−(k+1)u1 dy.
k+1 f (x)
∂w ∂w ∂w
15. f (x) + g(y) + h(z) = 0.
∂x ∂y ∂z
Integral basis:
Z Z Z Z
dx dy dx dz
u1 = − , u2 = − .
f (x) g(y) f (x) h(z)
∂w ∂w ∂w
16. f1 (x) + f2 (x)g(y) + f3 (x)h(z) = 0.
∂x ∂y ∂z
Integral basis:
Z Z Z Z
f2 (x) dy f3 (x) dz
u1 = dx − , u2 = dx − .
f1 (x) g(y) f1 (x) h(z)
∂w ∂w ∂w
17. a sinh(βy) + b sinh(γx) + f1 (x)f2(z) sinh(βy) = 0.
∂x ∂y ∂z
Integral basis:
Z Z
1 1 1 dz
u1 = cosh(γx) − cosh(βy), u2 = f1 (x) dx − .
aγ bβ a f2 (z)
◮ Coefficients of equations contain arbitrary functions of two variables.
∂w ∂w ∂w
18. a +b + f (x, y) = 0.
∂x ∂y ∂z
Integral basis:
Z y
a(t − y)
u1 = bx − ay, u2 = bz − f x+ , t dt,
y0 b
where y0 may be taken as arbitrary.
∂w ∂w ∂w
19. a +b + f (x, y)g(z) = 0.
∂x ∂y ∂z
Integral basis:
Z Z y
dz a(t − y)
u1 = bx − ay, u2 = b − f x+ , t dt.
g(z) y0 b
∂w ∂w ∂w
20. x +y + [z + f (x, y)] = 0.
∂x ∂y ∂z
Integral basis: Z
x
y z y
u1 = , u2 = − f t, t t−2 dt,
x x x0 x
∂w ∂w ∂w
21. ax + by + f (x, y) = 0.
∂x ∂y ∂z
Integral basis:
Z y
u1 = xb y −a , u2 = bz − t−1 f xy −a/b ta/b , t dt.
y0
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 169
∂w ∂w ∂w
22. ax + by + f (x, y)g(z) = 0.
∂x ∂y ∂z
Integral basis:
Z Z y
b −a dz
u1 = x y , u2 = b − t−1 f xy −a/b ta/b , t dt.
g(z) y0
∂w ∂w ∂w
23. + f1 (x)y + f2 (x) + g(x, y)z + h(x, y) = 0.
∂x ∂y ∂z
Integral basis:
Z Z
Z x Z x
u2 = zG(x, u1 ) − h̄(t, u1 )G(t, u1 ) dt, G(x, u1 ) = exp − ḡ(t, u1 ) dt .
x0 x0
Here ḡ(x, u1 ) ≡ g(x, y) and h̄(x, u1 ) ≡ h(x, y) (y is expressed via x and u1 from the first
integral), and x0 is an arbitrary number.
∂w ∂w ∂w
24. + f1 (x)y + f2 (x)y k + g(x, y)z + h(x, y)z m = 0.
∂x ∂y ∂z
1◦ . For k 6= 1 and m 6= 1, the transformation ξ = y 1−k , η = z 1−m leads to an equation of

the form 2.1.7.23:
∂w ∂w ∂w
+ (1 − k) f1 (x)ξ + f2 (x) + (1 − m) ḡ(x, ξ)η + h̄(x, ξ) = 0,
∂x ∂ξ ∂η
1 1
where ḡ(x, ξ) ≡ g x, ξ 1−k and h̄(x, ξ) ≡ h x, ξ 1−k .
2◦ . For k 6= 1 and m = 1, the substitution ξ = y 1−k leads to an equation of the form
2.1.7.23:
∂w ∂w ∂w
+ (1 − k) f1 (x)ξ + f2 (x) + z ḡ(x, ξ) + h̄(x, ξ) = 0.
∂x ∂ξ ∂z
∂w ∂w ∂w
25. + f1 (x)y + f2 (x)y k + g(x, y) + h(x, y)eλz = 0.
∂x ∂y ∂z
The transformation ξ = y 1−k , η = e−λz leads to an equation of the form 2.1.7.23:
∂w ∂w ∂w
+ (1 − k) f1 (x)ξ + f2 (x) − λ ḡ(x, ξ)η + h̄(x, ξ) = 0,
∂x ∂ξ ∂η
1 1
where ḡ(x, ξ) ≡ g x, ξ 1−k and h̄(x, ξ) ≡ h x, ξ 1−k .
∂w ∂w ∂w
26. + f1 (x) + f2 (x)eλy + g(x, y)z + h(x, y)z k = 0.
∂x ∂y ∂z
The transformation ξ = e−λy , η= z 1−k leads to an equation of the form 2.1.7.23:

∂w ∂w ∂w
− λ f1 (x)ξ + f2 (x) + (1 − k) ḡ(x, ξ)η + h̄(x, ξ) = 0,
∂x ∂ξ ∂η

where ḡ(x, ξ) ≡ g x, − λ1 ln ξ and h̄(x, ξ) ≡ h x, − λ1 ln ξ .
∂w ∂w ∂w
27. + f1 (x) + f2 (x)eλy + g(x, y) + h(x, y)eβz = 0.
∂x ∂y ∂z

∂w ∂w ∂w
− λ f1 (x)ξ + f2 (x) − β ḡ(x, ξ)η + h̄(x, ξ) = 0,
∂x ∂ξ ∂η

where ḡ(x, ξ) ≡ g x, − λ1 ln ξ and h̄(x, ξ) ≡ h x, − λ1 ln ξ .
∂w ∂w ∂w
28. f1 (x)g1(y) + f2 (x)g2 (y) + h1 (x, y)z + h2 (x, y)z m = 0.
∂x ∂y ∂z
Z Z
f2 (x) g1 (y)
f1 (x) g2 (y)
2.1.7.24 with f1 ≡ 0, f2 ≡ 1, and k = 0:
∂w ∂w ∂w
+ + h̄1 (ξ, η)z + h̄2 (ξ, η)z m = 0,
∂ξ ∂η ∂z
h1 (x, y) h2 (x, y)
where h̄1 (ξ, η) ≡ and h̄2 (ξ, η) ≡ .
f2 (x)g1 (y) f2 (x)g1 (y)
∂w ∂w ∂w
29. f1 (x)g1(y) + f2 (x)g2 (y) + h1 (x, y) + h2 (x, y)eλz = 0.
∂x ∂y ∂z
Z Z
f2 (x) g1 (y)
f1 (x) g2 (y)
2.1.7.25 with f1 ≡ 0, f2 ≡ 1, and k = 0:
∂w ∂w ∂w
+ + h̄1 (ξ, η) + h̄2 (ξ, η)eλz = 0,
∂ξ ∂η ∂z
h1 (x, y) h2 (x, y)
where h̄1 (ξ, η) ≡ and h̄2 (ξ, η) ≡ .
f2 (x)g1 (y) f2 (x)g1 (y)

∂w ∂w ∂w
f1 + f2 + f3 = f4 , fn = fn (x, y, z)
∂x ∂y ∂z
◆ The solutions given below contain arbitrary functions of two variables Φ = Φ(u1 , u2 ),
where u1 = u1 (x, y, z) and u2 = u2 (x, y, z) are some functions.
2.2. Equations of the Form f1 ∂w
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 , fn = fn (x, y, z) 171

∂w ∂w ∂w
1. a +b +c = αx + βy + γz + δ.
∂x ∂y ∂z
α 2 β γ δ
General solution: w = x + y 2 + z 2 + x + Φ(bx − ay, cy − bz).
2a 2b 2c a
∂w ∂w ∂w
2. + az + by = cx + s.
∂x ∂y ∂z
General solution: w = 12 cx2 + sx + Φ(u1 , u2 ), where

( √ √
u1 = by − az , u2 = p p p
∂w ∂w ∂w
3. + (a1 x + a0 ) + (b1 x + b0 ) = αx + βy + γz + δ.
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = a1 x+a0 , g(x) = b1 x+b0 , h2 (x) = β,
h1 (x) = γ, and h0 (x) = αx + δ.
∂w ∂w ∂w
4. + (a2 y + a1 x + a0 ) + (b2 y + b1 x + b0 ) = c2 y + c1 z + c0 x + s.
∂x ∂y ∂z
This is a special case of equation 2.2.7.4 with f1 (x) = a2 , f2 (x) = a1 x + a0 , g1 (x) = b2 ,

g2 (x) = b1 x + b0 , h2 (x) = c2 , h1 (x) = c1 , and h0 (x) = c0 x + s.
∂w ∂w ∂w
5. + (ay + k1 x + k0 ) + (bz + s1 x + s0 ) = c1 x + c0 .
∂x ∂y ∂z
This is a special case of equation 2.2.7.3 with f (x) = k1 x + k0 , g(x) = s1 x + s0 , and

h(x) = c1 x + c0 .
∂w ∂w ∂w
6. ax + by + cz = αx + βy + γz + δ.
∂x ∂y ∂z
a
α β γ δ |y| |z|a
General solution: w = x + y + z + ln |x| + Φ , .
a b c a |x|b |x|c
∂w ∂w ∂w
7. x + az + by = c.
∂x ∂y ∂z
General solution: w = c ln |x| + Φ(u1 , u2 ), where

 √ √

√
u1 = by 2 − az 2 , u2 = √
−ab −ab z

by
∂w ∂w ∂w
8. abx + b(ay + bz) + a(ay − bz) = c.
∂x ∂y ∂z
c
General solution: w = ln |x| + Φ(u1 , u2 ), where
ab
√ √ √ √
u1 = ay + ( 2 − 1)bz |x|− 2 , u2 = ay − ( 2 + 1)bz |x| 2 .
c
Particular solution: w = ln |x| + Φ a2 y 2 − 2abyz − b2 z 2 .
ab
∂w ∂w ∂w
9. (a1 x + a0 ) + (b1 y + b0 ) + (c1 z + c0 ) = αx + βy + γz + δ.
∂x ∂y ∂z
1◦ . General solution for a1 b1 c1 6= 0:

α β γ 1 αa0 βb0 γc0
w= x+ y+ z+ δ− − − lna1 x + a0
a1 b1 c1 a1 a1 b1 c1

|b1 y + b0 |a1 |b1 y + b0 |c1
+Φ , .
|a1 x + a0 |b1 |c1 z + c0 |b1
2◦ . General solution for a1 b1 =

6 0 and c1 = 0:

α β γ 2 1 αa0 βb0 |b1 y + b0 |a1 c0 −b1 z
w = x+ y + z + δ− − z +Φ , |b1 y +b0 | e .
a1 b1 2c0 c0 a1 b1 |a1 x + a0 |b1
3◦ . General solution for a1 6= 0 and b1 = c1 = 0:

α β 2 γ 2 1 αa0
w= x+ y + z + δ− z + Φ |a1 x + a0 |b0 e−a1 y , c0 y − b0 z .
a1 2b0 2c0 c0 a1
4◦ . For a1 = b1 = c1 = 0, see equation 2.2.1.1.
∂w ∂w ∂w
10. a +b +c = αx2 + βy 2 + γz 2 + δ.
∂x ∂y ∂z
α 3 β γ δ
General solution: w = x + y 3 + z 3 + x + Φ(bx − ay, cy − bz).
3a 3b 3c a
∂w ∂w ∂w
11. + (a1 x2 + a0 ) + (b1 x2 + b0 ) = αx + βy + γz + δ.
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = a1 x2 + a0 , g(x) = b1 x2 + b0 ,

h2 (x) = β, h1 (x) = γ, and h0 (x) = αx + δ.
∂w ∂w ∂w
12. + (ay + k1 x2 + k0 ) + (bz + s1 x2 + s0 ) = c 1 x2 + c 0 .
∂x ∂y ∂z
This is a special case of equation 2.2.7.3 with f (x) = k1 x2 + k0 , g(x) = s1 x2 + s0 , and

h(x) = c1 x2 + c0 .
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 , fn = fn (x, y, z) 173
∂w ∂w ∂w
13. +(a2 xy+a1 x2 +a0 ) +(b2 xy+b1 x2 +b0 ) = c2 y+c1 z+c0 x+s.
∂x ∂y ∂z
This is a special case of equation 2.2.7.4 with f1 (x) = a2 x, f2 (x) = a1 x2 +a0 , g1 (x) = b2 x,
g2 (x) = b1 x2 + b0 , h2 (x) = c2 , h1 (x) = c1 , and h0 (x) = c0 x + s.
∂w ∂w ∂w
14. ax + by + cz = x(αx + βy + γz).
∂x ∂y ∂z
1◦ . General solution for b 6= −a and c 6= −a:
α 2 β γ
w= x + xy + xz + Φ x|y|−a/b , x|z|−a/c .
2a a+b a+c
2◦ . General solution for b = −a and c 6= −a:
1 γ
w= x αx + 2βy ln |x| + xz + Φ xy, x|z|−a/c .
2a a+c
3◦ . General solution for b = c = −a:
1
w= x αx + 2(βy + γz) ln |x| + Φ(xy, xz).
2a
∂w ∂w ∂w
15. ax2 + bxy + cxz = αx + βy + γz.
∂x ∂y ∂z
1◦ . General solution for b 6= a and c 6= a:

α 1 βy γz
w = ln |x| + + + Φ x|y|−a/b , x|z|−a/c .
a x b−a c−a
2◦ . General solution for b = a and c 6= a:

α 1 βy γz x −a/c
w = ln |x| + ln |x| + +Φ , x|z| .
a x a c−a y
3◦ . General solution for a = b = c:

ln |x| x x
w= (αx + βy + γz) + Φ , .
ax y z
∂w ∂w ∂w
16. ax2 + bxy + cz 2 = ky 2 .
∂x ∂y ∂z
1◦ . General solution for a 6= 2b:

ky 2 −a/b c a
w= + Φ xy , − .
(2b − a)x x z
2◦ . General solution for a = 2b:

ky 2 ln |x| x c a
w= +Φ 2, − .
ax y x z
∂w ∂w ∂w
17. ax2 + by 2 + cz 2 = kxy.
∂x ∂y ∂z

kxy ax b a c a
General solution: w = ln +Φ − , − .
ax − by y x y x z
∂w ∂w ∂w
18. ax2 + by 2 + cz 2 = αx2 + βy 2 + γz 2 .
∂x ∂y ∂z

α β γ b a c a
General solution: w = x + y + z + Φ − , − .
a b c x y x z
◮ Coefficients of equations contain other powers in x, y, and z.
∂w ∂w ∂w
19. +a +b = xyz.
∂x ∂y ∂z
General solution: w = 12 x2 yz − 16 x3 (az + by) + 1

12 abx
4
+ Φ(y − ax, z − bx).
∂w ∂w ∂w
20. a +b +c = kx3 + sy 2 .
∂x ∂y ∂z
k 4 s 3
General solution: w = x + y + Φ(bx − ay, cx − az).
4a 3a
∂w ∂w ∂w √
21. a + by + cz = kx + s x.
∂x ∂y ∂z
k 2 2s 3/2
General solution: w = x + x + Φ |y|a e−bx , |z|a e−cx .
2a 3a
∂w ∂w ∂w √
22. + az + by = c x + s.
∂x ∂y ∂z
General solution: w = 23 cx3/2 + sx + Φ(u1 , u2 ), where

( √ √
by + ab z exp − ab x if ab > 0,
u1 = by 2 − az 2 , u2 = p p p
∂w ∂w ∂w
23. ax2 + by 2 + cz 2 = kxyz.
∂x ∂y ∂z
General solution:

1 1 1 1
w = w0 (x, y, z) + Φ − , − ,
ax by ax cz
where w0 = w0 (x, y, z) is a particular solution,

ax ln(ax) by ln(by) cz ln(cz)
w0 = kxyz + + .
(ax − by)(ax − cz) (by − ax)(by − cz) (cz − ax)(cz − by)
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 , fn = fn (x, y, z) 175
∂w ∂w ∂w
24. a +b +c = αxn + βy m + γz k .
∂x ∂y ∂z
General solution:
α β γ
w= xn+1 + y m+1 + z k+1 + Φ(bx − ay, cx − az).
a(n + 1) b(m + 1) c(k + 1)
∂w ∂w ∂w
25. a + by + cz = αxn + βy m + γz k .
∂x ∂y ∂z
α β m γ k
General solution: w = xn+1 + y + z + Φ |y|a e−bx , |z|a e−cx .
a(n + 1) bm ck
∂w ∂w ∂w
26. + az + by = cxn .
∂x ∂y ∂z
c
General solution: w = xn+1 + Φ(u1 , u2 ), where
n+1
( √ √
u1 = by 2 − az 2 , u2 = p p p
∂w ∂w ∂w
27. ax + by + cz = αxn + βy m + γz k .
∂x ∂y ∂z
a
α n β m γ k |y| |z|a
General solution: w = x + y + z +Φ , .
an bm ck |x|b |x|c
∂w ∂w ∂w
28. x + az + by = cxn .
∂x ∂y ∂z
c n
General solution: w = x + Φ(u1 , u2 ), where
n
 √ √

√
u1 = by 2 − az 2 , u2 = √
−ab −ab z

by
∂w ∂w ∂w
29. abx + b(ay + bz) + a(ay − bz) = cxn .
∂x ∂y ∂z
c n
General solution: w = x + Φ(u1 , u2 ), where
abn
√ √ √ √
u1 = ay + ( 2 − 1)bz |x|− 2 , u2 = ay − ( 2 + 1)bz |x| 2 .
c n
Particular solution: w = x + Φ a2 y 2 − 2abyz − b2 z 2 .
abn
∂w ∂w ∂w
30. + axn y m + bxν y µ z λ = cxk .
∂x ∂y ∂z
General solution:

c  xk+1 if k 6= −1,
w = Φ(u1 , u2 ) + k + 1
c ln |x| if k = −1,
where u1 , u2 is the integral basis of the homogeneous equation 2.1.1.64.
∂w ∂w ∂w
31. + a1 xn1 y + b1 xm1 + (a2 xn2 y + b2 xm2 = c2 xk2 y + c1 xk1 z.
∂x ∂y ∂z
a2 xn2 , g2 (x) = b2 xm2 , h2 (x) = c2 xk2 , h1 (x) = c1 xk1 , and h0 (x) = 0.
∂w ∂w ∂w
32. + a1 xn1 y + b1 xm1 + (a2 xn2 z + b2 xm2 = c2 xk2 y + c1 xk1 z.
∂x ∂y ∂z
∂w ∂w ∂w
33. + a1 xn1 y + b1 y k + (a2 xn2 z + b2 z m = cxs .
∂x ∂y ∂z
This is a special case of equation 2.2.7.7 with f1 (x) = a1 xn1 , f2 (x) = b1 , g1 (x) = a2 xn2 ,
g2 (x) = b2 , and h(x) = cxs .
∂w ∂w ∂w
34. + a1 xn1 y + b1 y k + (a2 y n2 z + b2 z m = c1 xs1 + c2 y s2 .
∂x ∂y ∂z
This is a special case of equation 2.2.7.23 with f1 (x) = a1 xn1 , f2 (x) = b1 , g1 (x, y) = a2 y n2 ,
g2 (x, y) = b2 , and h(x, y, z) = c1 xs1 + c2 y s2 .
∂w ∂w p ∂w
35. x +y + a x2 + y 2 = bxn .
∂x ∂y ∂z

General solution: w = Φ u1 , u2 + w0 (x), where
(
y p (b/n)xn if n 6= 0,
u1 = , u2 = a x2 + y 2 − z, w0 (x) =
x b ln |x| if n = 0.
∂w ∂w p ∂w
36. x +y + z−a x2 + y 2 + z 2 = bxn .
∂x ∂y ∂z

General solution: w = Φ u1 , u2 + w0 (x), where
(
y p (b/n)xn if n 6= 0,
u1 = , u2 = |x|a−1 z + x2 + y 2 + z 2 , w0 (x) =
x b ln |x| if n = 0.
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 , fn = fn (x, y, z) 177

∂w ∂w ∂w
1. + aeλx + beβx = ceγx .
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = aeλx , g(x) = beβx , h2 (x) = h1 (x) = 0,
and h0 (x) = ceγx .
∂w ∂w ∂w
2. + aeλx + beβy = ceγy + seµz .
∂x ∂y ∂z
This is a special case of equation 2.2.7.22 with f1 (x) = 0, f2 (x) = aeλx , g1 (x, y) = 0,
g2 (x, y) = beβy , and h(x, y, z) = ceγy + seµz .
∂w ∂w ∂w
3. + aeλy + beβz = ceγx + seµz .
∂x ∂y ∂z
This is a special case of equation 2.2.7.26 with f1 (x) = 0, f2 (x) = a, g1 (x, y) = beβy ,
g2 (x, y) = 0, and h(x, y, z) = ceγx + seµz .
∂w ∂w ∂w
4. + A1 eα1 x + B1 eν1 x+λy + A2 eα2 x + B2 eν2 x+βz = keγz .
∂x ∂y ∂z
A2 eα2 x , g2 (x) = B2 eν2 x , and h(x) = keγz .
∂w ∂w ∂w
5. aeαx + beβy + ceγz = keλx .
∂x ∂y ∂z
General solution:  (λ−α)x
Φ(u1 , u2 ) + ke
 if λ 6= α,
w= a(λ − α)

Φ(u , u ) + k x
1 2 if λ = α,
a
1 1 −βy 1 1 −γz
where u1 = − e−αx + e and u2 = − e−βy + e .
aα bβ bβ cγ
∂w ∂w ∂w
6. aeβy + beαx + ceγz = keλx .
∂x ∂y ∂z
General solution:
 Z

 kα eλx dx

 Φ(u1 , u2 ) + if λ 6= α, λ 6= 0;

 bβ eαx + aαu1
k
w = Φ(u1 , u2 ) + y if λ = α 6= 0;

 b

 k −γz

Φ(u1 , u2 ) − e if λ = 0,
cγ
where u1 = − e + e and u2 = αx βy
+ e . In the integration,
u1 is considered a parameter.
∂w ∂w ∂w
7. (a1 + a2 eαx ) + (b1 + b2 eβy ) + (c1 + c2 eγz ) = k1 + k2 eαx .
∂x ∂y ∂z

k1 1 k2 k1
General solution: w = Φ(u1 , u2 ) + x+ − ln(a1 + a2 eαx ), where
a1 α a2 a1
1 1
a1 α b1 β
1 1
a1 α c1 γ
∂w ∂w ∂w
8. eβy (a1 + a2 eαx ) + eαx (b1 + b2 eβy ) + ceβy+γz
∂x ∂y ∂z
= k3 eβy (k1 + k2 eαx ).

k1 k3 k3 k2 k1
General solution: w = Φ(u1 , u2 ) + x+ − ln(a1 + a2 eαx ), where
a1 α a2 a1
1 1 1 1
u1 = ln(a1+a2eαx)− ln(b1+b2eβy ), u2 = αx−ln(a1+a2eαx) + e−γz .
a2α b1β a1α cγ
∂w ∂w ∂w
9. + axn + bxm = ceλx y + keβx z + seγx .
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = axn , g(x) = bxm , h2 (x) = ceλx ,
h1 (x) = keβx , and h0 (x) = seγx .
∂w ∂w ∂w
10. + aeλx + bxm = cxn y + keβx z + seγx .
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = aeλx , g(x) = bxm , h2 (x) = cxn ,
∂w ∂w ∂w
11. + aeλx + by = keβx z + seγx .
∂x ∂y ∂z
This is a special case of equation 2.2.7.4 with f1 (x) = 0, f2 (x) = aeλx , g1 (x) = b, g2 (x) =
h2 (x) = 0, h1 (x) = keβx , and h0 (x) = seγx .
∂w ∂w ∂w
12. + ay n + bz m = ceλx + keβy + seγz .
∂x ∂y ∂z
This is a special case of equation 2.2.7.10 with f (x) = 1, g(y) = ay n , h(z) = bz m ,

ϕ(x) = ceλx , ψ(y) = keβy , and χ(z) = seγz .
∂w ∂w ∂w
13. + aeβy + bz m = ceλx + ky n + seγz .
∂x ∂y ∂z
This is a special case of equation 2.2.7.10 with f (x) = 1, g(y) = aeβy , h(z) = bz m ,
ϕ(x) = ceλx , ψ(y) = ky n , and χ(z) = seγz .
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 , fn = fn (x, y, z) 179
∂w ∂w ∂w
14. + aeαx y 2 +be−αx + deβx z 2 +ceγx (γ −cde(β+γ)x ) = keλx .
∂x ∂y ∂z
k
General solution: w = eλx + Φ(u1 , u2 ), where u1 , u2 is an integral basis of the homo-
λ
geneous equation 2.1.2.15.
∂w ∂w ∂w
15. + a1 eλ1 x y + b1 eβ1 x y k + a2 eλ2 x z + b2 eβ2 x z m = cxs .
∂x ∂y ∂z
a2 eλ2 x , g2 (x) = b2 eβ2 x , and h(x) = cxs .
∂w ∂w ∂w
16. + a1 eβ1 x y + b1 eγ1 x y k + a2 eβ2 x + b2 eγ2 x+λz = cxs .
∂x ∂y ∂z
a2 eβ2 x , g2 (x) = b2 eγ2 x , and h(x) = cxs .
∂w ∂w ∂w
17. + a1 xn + b1 xm eλy + a2 xk + b2 xl eβz = cxs .
∂x ∂y ∂z
g2 (x) = b2 xl , and h(x) = cxs .

∂w ∂w ∂w
1. +a +b= c sinhk (λx) + s.
∂x ∂y ∂z
Z
General solution: w = Φ(y − ax, z − bx) + c sinhk (λx) dx + sx.
∂w ∂w ∂w
2. a +b + c sinh(λx) = k sinh(βy) + s sinh(γz).
∂x ∂y ∂z
General solution:
Z x
k s cγ
w = Φ(u1 , u2 ) + cosh(βy) − sinh cosh(λx) − cosh(λt) − γz dt,
bβ a 0 aλ
where u1 = bx − ay and u2 = aλz − c cosh(λx).
∂w ∂w ∂w
3. + a sinhn (βx) + b sinhk (λx) = c sinhm (γx) + s.
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = a sinhn (βx), g(x) = b sinhk (λx),
h2 (x) = h1 (x) = 0, and h0 (x) = c sinhm (γx) + s.
∂w ∂w ∂w
4. a + b sinh(βy)+ c sinh(λx) = k sinh(γz).
∂x ∂y ∂z
Z
k x cγ
General solution: w = Φ(u1 , u2 ) + sinh γz + cosh(λx) − cosh(λt) dt,
a 0 aλ
βy

where u1 = bβx − a lntanh and u2 = aλz − c cosh(λx).
2
∂w ∂w ∂w
5. a1 sinhn1 (λ1 x) + b1 sinhm1 (β1 y) + c1 sinhk1 (γ1 z)
∂x ∂y ∂z
= a2 sinhn2 (λ2 x) + b2 sinhm2 (β2 y) + c2 sinhk2 (γ2 z).
This is a special case of equation 2.2.7.10 in which f (x) = a1 sinhn1 (λ1 x), g(y) =
b1 sinhm1 (β1 y), h(z) = c1 sinhk1 (γ1 z), ϕ(x) = a2 sinhn2 (λ2 x), ψ(y) = b2 sinhm2 (β2 y),
and χ(z) = c2 sinhk2 (γ2 z).

∂w ∂w ∂w
6. +a +b= c coshk (λx) + s.
∂x ∂y ∂z
Z
General solution: w = Φ(y − ax, z − bx) + c coshk (λx) dx + sx.
∂w ∂w ∂w
7. a +b + c cosh(λx) = k cosh(βy) + s cosh(γz).
∂x ∂y ∂z
General solution:
Z x
k s cγ
w = Φ(u1 , u2 ) + sinh(βy) + cosh cosh(λt) − cosh(λx) + γz dt,
bβ a 0 aλ
where u1 = bx − ay and u2 = aλz − c sinh(λx).

∂w ∂w ∂w
8. + a coshn (βx) + b coshk (λx) = c coshm (γx) + s.
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = a coshn (βx), g(x) = b coshk (λx),
h2 (x) = h1 (x) = 0, and h0 (x) = c coshm (γx) + s.
∂w ∂w ∂w
9. a + b cosh(βy) + c cosh(λx) = k cosh(γz).
∂x ∂y ∂z
Z
k x cγ
General solution: w = Φ(u1 , u2 ) + cosh γz + sinh(λt) − sinh(λx) dt,
a 0 aλ
βy

where u1 = bβx − 2a arctantanh and u2 = aλz − c sinh(λx).
2
∂w ∂w ∂w
10. a + b cosh(βy) + c cosh(γz) = p cosh(λx) + q.
∂x ∂y ∂z
q p βy

General solution: w=Φ(u1, u2)+ x+ sinh(λx), where u1 =bβx−2a arctantanh
a aλ 2
γz
and u2 = cγx − 2a arctantanh .
2
∂w ∂w ∂w
11. a1 coshn1 (λ1 x) + b1 coshm1 (β1 y) + c1 coshk1 (γ1 z)
∂x ∂y ∂z
= a2 coshn2 (λ2 x) + b2 coshm2 (β2 y) + c2 coshk2 (γ2 z).
This is a special case of equation 2.2.7.10 in which f (x) = a1 coshn1 (λ1 x), g(y) =
b1 coshm1 (β1 y), h(z) = c1 coshk1 (γ1 z), ϕ(x) = a2 coshn2 (λ2 x), ψ(y) = b2 coshm2 (β2 y),
and χ(z) = c2 coshk2 (γ2 z).
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 , fn = fn (x, y, z) 181
∂w ∂w ∂w
12. +a = c tanhk (λx) + s.
+b
∂x ∂y ∂z
Z
General solution: w = Φ(y − ax, z − bx) + c tanhk (λx) dx + sx.
∂w ∂w ∂w
13. a +b + c tanh(γz) = k tanh(λx) + s tanh(βy).
∂x ∂y ∂z
k s
General solution: w = Φ(u1 , u2 ) + ln cosh(λx) + ln cosh(βy) , where u1 =
aλ bβ
bx − ay and u2 = cγx − a lnsinh(γz).
∂w ∂w ∂w
14. + a tanhn (βx) + b tanhk (λx) = c tanhm (γx) + s.
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = a tanhn (βx), g(x) = b tanhk (λx),
h2 (x) = h1 (x) = 0, and h0 (x) = c tanhm (γx) + s.
∂w ∂w ∂w
15. a + b tanh(βy) + c tanh(λx) = k tanh(γz).
∂x ∂y ∂z
General solution:
Z
k x cγ
w = Φ(u1 , u2 ) + tanh γz + ln |cosh(λx)| − ln |cosh(λt)| dt,
a 0 aλ

where u1 = bβx − a lnsinh(βy) and u2 = aλz − c ln cosh(λx) .
∂w ∂w ∂w
16. a + b tanh(βy) + c tanh(γz) = k tanh(λx).
∂x ∂y ∂z
k
General solution: w = Φ(u1 , u2 ) + ln cosh(λx) , where u1 = bβx − a lnsinh(βy)
aλ
and u2 = cγx − a lnsinh(γz).
∂w ∂w ∂w
17. a tanh(λx) + b tanh(βy) + c tanh(γz) = k.
∂x ∂y ∂z
General solution:
k |sinh(βy)|aλ |sinh(γz)|aλ
w = Φ(u1 , u2 ) + ln sinh(λx), u1 = , u2 = .
aλ |sinh(λx)|bβ |sinh(λx)|cγ
∂w ∂w ∂w
18. a1 tanhn1 (λ1 x) + b1 tanhm1 (β1 y) + c1 tanhk1 (γ1 z)
∂x ∂y ∂z
= a2 tanhn2 (λ2 x) + b2 tanhm2 (β2 y) + c2 tanhk2 (γ2 z).
This is a special case of equation 2.2.7.10 in which f (x) = a1 tanhn1 (λ1 x), g(y) =
b1 tanhm1 (β1 y), h(z) = c1 tanhk1 (γ1 z), ϕ(x) = a2 tanhn2 (λ2 x), ψ(y) = b2 tanhm2 (β2 y),
and χ(z) = c2 tanhk2 (γ2 z).
∂w ∂w ∂w
19. +a +b = c cothk (λx) + s.
∂x ∂y ∂z
Z
General solution: w = Φ(y − ax, z − bx) + c cothk (λx) dx + sx.
∂w ∂w ∂w
20. a +b + c coth(γz) = k coth(λx) + s coth(βy).
∂x ∂y ∂z
k s
General solution: w = Φ(u1 , u2 ) + ln sinh(λx) + ln sinh(βy), where u1 =
aλ bβ
bx − ay and u2 = cγx − a ln cosh(γz) .
∂w ∂w ∂w
21. + a cothn (βx) + b cothk (λx) = c cothm (γx) + s.
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = a cothn (βx), g(x) = b cothk (λx),
h2 (x) = h1 (x) = 0, and h0 (x) = c cothm (γx) + s.
∂w ∂w ∂w
22. a + b coth(βy) + c coth(λx) = k coth(γz).
∂x ∂y ∂z
General solution:
Z
k x cγ
w = Φ(u1 , u2 ) + coth γz + ln |sinh(λt)| − ln |sinh(λx)| dt,
a 0 aλ

where u1 = bβx − a ln cosh(βy) and u2 = aλz − c lnsinh(λx).
∂w ∂w ∂w
23. a + b coth(βy) + c coth(γz) = k coth(λx).
∂x ∂y ∂z
k
General solution: w = Φ(u1 , u2 ) + ln sinh(λx), where u1 = bβx − a ln cosh(βy)
aλ
and u2 = cγx − a ln cosh(γz) .
∂w ∂w ∂w
24. a1 cothn1 (λ1 x) + b1 cothm1 (β1 y) + c1 cothk1 (γ1 z)
∂x ∂y ∂z
= a2 cothn2 (λ2 x) + b2 cothm2 (β2 y) + c2 cothk2 (γ2 z).
This is a special case of equation 2.2.7.10 in which f (x) = a1 cothn1 (λ1 x), g(y) =
b1 cothm1 (β1 y), h(z) = c1 cothk1 (γ1 z), ϕ(x) = a2 cothn2 (λ2 x), ψ(y) = b2 cothm2 (β2 y),
and χ(z) = c2 cothk2 (γ2 z).
∂w ∂w ∂w
25. a +b + c sinhn (λy) = s coshm (βx) + k sinhl (γy).
∂x ∂y ∂z
This is a special case of equation 2.2.7.18 with f (x, y) = c sinhn (λy) and g(x, y) =
s coshm (βx) + k sinhl (γy).
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 , fn = fn (x, y, z) 183
∂w ∂w ∂w
26. + a sinhn (λx) + b coshm (βx) = s coshk (γx).
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = a sinhn (λx), g(x) = b coshm (βx),
h2 (x) = h1 (x) = 0, and h0 (x) = s coshk (γx).
∂w ∂w ∂w
27. + a coshn (λx) + b sinhm (βy) = s sinhk (γz).
∂x ∂y ∂z
This is a special case of equation 2.2.7.22 with f1 (x) = 0, f2 (x) = a coshn (λx), g1 (x, y) =
0, g2 (x, y) = b sinhm (βy), and h(x, y, z) = s sinhk (γz).
∂w ∂w ∂w
28. + a tanhn (λx) + b cothm (βx) = s cothk (γx).
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = a tanhn (λx), g(x) = b cothm (βx),
h2 (x) = h1 (x) = 0, and h0 (x) = s cothk (γx).
∂w ∂w ∂w
29. a sinh(λx) + b sinh(βy) + c sinh(γz) = k cosh(λx).
∂x ∂y ∂z
k
General solution: w = Φ(u1 , u2 ) + ln |sinh(λx)|, where
aλ

∂w ∂w ∂w
1. +a +b = c lnk (λx) + s.
∂x ∂y ∂z
Z
General solution: w = Φ(y − ax, z − bx) + c lnk (λx) dx + sx.
∂w ∂w ∂w
2. a +b + c ln(βy) ln(γz) = k ln(αx).
∂x ∂y ∂z
k
General solution: w = Φ(u1 , u2 ) + x ln(λx) − 1 , where
a
Z
dz
u1 = bx − ay, u2 = cy 1 − ln(βy) + b .
ln(γz)
∂w ∂w ∂w
3. + a lnn (βx) + b lnk (λx) = c lnm (γx) + s.
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = a lnn (βx), g(x) = b lnk (λx), h2 (x) =
h1 (x) = 0, and h0 (x) = c lnm (γx) + s.
∂w ∂w ∂w
4. + a lnn (λx) + b lnm (βy) = c lnk (γy) + s lnl (µz).
∂x ∂y ∂z
This is a special case of equation 2.2.7.22 with f1 (x) = 0, f2 (x) = a lnn (λx), g1 (x, y) = 0,
g2 (x, y) = b lnm (βy), and h(x, y, z) = c lnk (γy) + s lnl (µz).
∂w ∂w ∂w
5. a1 lnn1 (λ1 x) + b1 lnm1 (β1 y) + c1 lnk1 (γ1 z)
∂x ∂y ∂z
= a2 lnn2 (λ2 x) + b2 lnm2 (β2 y) + c2 lnk2 (γ2 z).
This is a special case of equation 2.2.7.10 with f (x) = a1 lnn1 (λ1 x), g(y) = b1 lnm1 (β1 y),
h(z) = c1 lnk1 (γ1 z), ϕ(x) = a2 lnn2 (λ2 x), ψ(y) = b2 lnm2 (β2 y), and χ(z) = c2 lnk2 (γ2 z).

∂w ∂w ∂w
6. a +b + cxn lnk (λy) = sy m lnl (βx).
∂x ∂y ∂z
This is a special case of equation 2.2.7.18 with f (x, y) = cxn lnk (λy), and g(x, y) =
sy m lnl (βx).
∂w ∂w ∂w
7. + axn + bxm = cy lnk (λx) + sz lnl (βx).
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = axn , g(x) = bxm , h2 (x) = c lnk (λx),
h1 (x) = s lnl (βx), and h0 (x) = 0.
∂w ∂w ∂w
8. + a lnn (λx) + by m = c lnk (βx) + s lnl (γz).
∂x ∂y ∂z
g2 (x, y) = by m , and h(x, y, z) = c lnk (βx) + s lnl (γz).
∂w ∂w ∂w
9. a lnn (λx) +z + b lnk (βy) = cxm + s ln(γy).
∂x ∂y ∂z
This is a special case of equation 2.2.7.11 with f (x) = a lnn (λx), g(y) = b lnk (βy), h2 (x) =
cxm , and h1 (y) = s ln(γy).
∂w ∂w ∂w
10. ax(ln x)n + by(ln y)m + cz(ln z)l = k(ln x)s .
∂x ∂y ∂z
General solution:

 k
 (ln x)s−n+1 if s + 1 6= n,
w = Φ(u1 , u2 ) + a(s − n + 1)
 k ln |ln x|

if s + 1 = n,
a
where
(ln x)1−n (ln y)1−m (ln x)1−n (ln z)1−l
u1 = − , u2 = − .
a(n − 1) b(m − 1) a(n − 1) c(l − 1)

∂w ∂w ∂w
1. +a +b = c sink (λx) + s.
∂x ∂y ∂z
Z
General solution: w = Φ(y − ax, z − bx) + c sink (λx) dx + sx.
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 , fn = fn (x, y, z) 185
∂w ∂w ∂w
2. a +b + c sin(γz) = k sin(αx) + s sin(βy).
∂x ∂y ∂z
k s
General solution: w = Φ(u1 , u2 ) − cos(αx) − cos(βy), where
aα bβ
γz

u1 = bx − ay, u2 = cγx − a lntan .
2
∂w ∂w ∂w
3. + a sinn (λx) + b sinm (βx) = c sink (γx).
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = a sinn (λx), g(x) = b sinm (βx),
h2 (x) = h1 (x) = 0, and h0 (x) = c sink (γx).
∂w ∂w ∂w
4. + a sinn (λx) + b sinm (βy) = c sink (γy) + s sinl (µz).
∂x ∂y ∂z
This is a special case of equation 2.2.7.22 with f1 (x) = 0, f2 (x) = a sinn (λx), g1 (x, y) = 0,
g2 (x, y) = b sinm (βy), and h(x, y, z) = c sink (γy) + s sinl (µz).
∂w ∂w ∂w
5. a + b sin(βy) + c sin(λx) = k sin(γz).
∂x ∂y ∂z
General solution:
Z x
k cγ
w = Φ(u1 , u2 ) + sin γz + cos(λx) − cos(λt) dt,
a 0 aλ
βy

where u1 = bβx − a lntan , u2 = aλz + c cos(λx).
2
∂w ∂w ∂w
6. a1 sinn1 (λ1 x) + b1 sinm1 (β1 y) + c1 sink1 (γ1 z)
∂x ∂y ∂z
= a2 sinn2 (λ2 x) + b2 sinm2 (β2 y) + c2 sink2 (γ2 z).
This is a special case of equation 2.2.7.10 in which f (x) = a1 sinn1 (λ1 x), g(y) =
b1 sinm1 (β1 y), h(z) = c1 sink1 (γ1 z), ϕ(x) = a2 sinn2 (λ2 x), ψ(y) = b2 sinm2 (β2 y), and
χ(z) = c2 sink2 (γ2 z).

∂w ∂w ∂w
7. +a +b = c cosk (λx) + s.
∂x ∂y ∂z
Z
General solution: w = Φ(y − ax, z − bx) + c cosk (λx) dx + sx.
∂w ∂w ∂w
8. a +b + c cos(βz) = k cos(λx) + s cos(γy).
∂x ∂y ∂z
k s
General solution: w = Φ(u1 , u2 ) + sin(λx) + sin(γy), where
aλ bγ

u1 = bx − ay, u2 = cβx − a lnsec(βz) + tan(βz).
∂w ∂w ∂w
9. + a cosn (βx) + b cosk (λx) = c cosm (γx) + s.
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = a cosn (βx), g(x) = b cosk (λx),
h2 (x) = h1 (x) = 0, and h0 (x) = c cosm (γx) + s.
∂w ∂w ∂w
10. + a cosn (λx) + b cosm (βy) = c cosk (γy) + s cosl (µz).
∂x ∂y ∂z
This is a special case of equation 2.2.7.22 with f1 (x) = 0, f2 (x) = a cosn (λx), g1 (x, y) = 0,
g2 (x, y) = b cosm (βy), and h(x, y, z) = c cosk (γy) + s cosl (µz).
∂w ∂w ∂w
11. a + b cos(βy) + c cos(λx) = k cos(γz).
∂x ∂y ∂z
General solution:
Z
k x cγ
w = Φ(u1 , u2 ) + cos γz + sin(λt) − sin(λx) dt,
a 0 aλ

where u1 = bβx − a lnsec(βy) + tan(βy), u2 = aλz − c sin(λx).
∂w ∂w ∂w
12. a1 cosn1 (λ1 x) + b1 cosm1 (β1 y) + c1 cosk1 (γ1 z)
∂x ∂y ∂z
= a2 cosn2 (λ2 x) + b2 cosm2 (β2 y) + c2 cosk2 (γ2 z).
This is a special case of equation 2.2.7.10 in which f (x) = a1 cosn1 (λ1 x), g(y) =
b1 cosm1 (β1 y), h(z) = c1 cosk1 (γ1 z), ϕ(x) = a2 cosn2 (λ2 x), ψ(y) = b2 cosm2 (β2 y), and
χ(z) = c2 cosk2 (γ2 z).
∂w ∂w ∂w
13. +a +b= c tank (λx) + s.
∂x ∂y ∂z
Z
General solution: w = Φ(y − ax, z − bx) + c tank (λx) dx + sx.
∂w ∂w ∂w
14. a +b + c tan(βz) = k tan(λx) + s tan(γy).
∂x ∂y ∂z
General solution:
k s
ln cos(λx) −
w = Φ(u1 , u2 ) − ln cos(γy),
aλ bγ

where u1 = bx − ay and u2 = cβx − a lnsin(βz).
∂w ∂w ∂w
15. + a tann (βx) + b tank (λx) = c tanm (γx) + s.
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = a tann (βx), g(x) = b tank (λx),
h2 (x) = h1 (x) = 0, and h0 (x) = c tanm (γx) + s.
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 , fn = fn (x, y, z) 187
∂w ∂w ∂w
16. + a tann (λx) + b tanm (βy) = c tank (γy) + s tanl (µz).
∂x ∂y ∂z
This is a special case of equation 2.2.7.22 with f1 (x) = 0, f2 (x) = a tann (λx), g1 (x, y) = 0,
g2 (x, y) = b tanm (βy), and h(x, y, z) = c tank (γy) + s tanl (µz).
∂w ∂w ∂w
17. a1 tann1 (λ1 x) + b1 tanm1 (β1 y) + c1 tank1 (γ1 z)
∂x ∂y ∂z
= a2 tann2 (λ2 x) + b2 tanm2 (β2 y) + c2 tank2 (γ2 z).
This is a special case of equation 2.2.7.10 in which f (x) = a1 tann1 (λ1 x), g(y) =
b1 tanm1 (β1 y), h(z) = c1 tank1 (γ1 z), ϕ(x) = a2 tann2 (λ2 x), ψ(y) = b2 tanm2 (β2 y),
and χ(z) = c2 tank2 (γ2 z).
∂w ∂w ∂w
18. +a +b = c cotk (λx) + s.
∂x ∂y ∂z
Z
General solution: w = Φ(y − ax, z − bx) + c cotk (λx) dx + sx.
∂w ∂w ∂w
19. a +b + c cot(γz) = k cot(λx) + s cot(βy).
∂x ∂y ∂z
General solution:
k s
ln sin(λx) +
w = Φ(u1 , u2 ) + ln sin(βy),
aλ bβ

where u1 = bx − ay, u2 = cγx + a lncos(γz).
∂w ∂w ∂w
20. + a cotn (βx) + b cotk (λx) = c cotm (γx) + s.
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = a cotn (βx), g(x) = b cotk (λx),
h2 (x) = h1 (x) = 0, and h0 (x) = c cotm (γx) + s.
∂w ∂w ∂w
21. + a cotn (λx) + b cotm (βy) = c cotk (γy) + s cotl (µz).
∂x ∂y ∂z
This is a special case of equation 2.2.7.22 with f1 (x) = 0, f2 (x) = a cotn (λx), g1 (x, y) = 0,
g2 (x, y) = b cotm (βy), and h(x, y, z) = c cotk (γy) + s cotl (µz).
∂w ∂w ∂w
22. a + b cot(βy) + c cot(λx) = k cot(γz).
∂x ∂y ∂z
General solution:
Z
k x cγ
w = Φ(u1 , u2 ) + cot γz + ln |sin(λt)| − ln |sin(λx)| dt,
a 0 aλ

where u1 = bβx + a lncos(βy) and u2 = aλz − c lnsin(λx).
∂w ∂w ∂w
23. a1 cotn1 (λ1 x) + b1 cotm1 (β1 y) + c1 cotk1 (γ1 z)
∂x ∂y ∂z
= a2 cotn2 (λ2 x) + b2 cotm2 (β2 y) + c2 cotk2 (γ2 z).
This is a special case of equation 2.2.7.10 in which f (x) = a1 cotn1 (λ1 x), g(y) =
b1 cotm1 (β1 y), h(z) = c1 cotk1 (γ1 z), ϕ(x) = a2 cotn2 (λ2 x), ψ(y) = b2 cotm2 (β2 y), and
χ(z) = c2 cotk2 (γ2 z).
∂w ∂w ∂w
24. + a sinn (λx) + b cosm (βx) = c sink (γx) + s.
∂x ∂y ∂z
This is a special case of equation 2.2.7.1 with f (x) = a sinn (λx), g(x) = b cosm (βx),
h2 (x) = h1 (x) = 0, and h0 (x) = c sink (γx) + s.
∂w ∂w ∂w
25. + a cosn (λx) + b sinm (βy) = c cosk (γy) + s sinl (µz).
∂x ∂y ∂z
g2 (x, y) = b sinm (βy), and h(x, y, z) = c cosk (γy) + s sinl (µz).
∂w ∂w ∂w
26. + a cosn (λx) + b tanm (βy) = c cosk (γy) + s tanl (µz).
∂x ∂y ∂z
g2 (x, y) = b tanm (βy), and h(x, y, z) = c cosk (γy) + s tanl (µz).
∂w ∂w ∂w
27. a1 sinn1 (λ1 x) + b1 cosm1 (β1 y) + c1 cosk1 (γ1 z)
∂x ∂y ∂z
= a2 cosn2 (λ2 x) + b2 sinm2 (β2 y) + c2 cosk2 (γ2 z).
This is a special case of equation 2.2.7.10 with f (x) = a1 sinn1 (λ1 x), g(y) = b1 cosm1 (β1 y),
h(z) = c1 cosk1 (γ1 z), ϕ(x) = a2 cosn2 (λ2 x), ψ(y) = b2 sinm2 (β2 y), χ(z) = c2 cosk2 (γ2 z).
∂w ∂w ∂w
28. a1 tann1 (λ1 x) + b1 cotm1 (β1 y) + c1 cotk1 (γ1 z)
∂x ∂y ∂z
= a2 cotn2 (λ2 x) + b2 tanm2 (β2 y) + c2 cotk2 (γ2 z).
b1 cotm1 (β1 y), h(z) = c1 cotk1 (γ1 z), ϕ(x) = a2 cotn2 (λ2 x), ψ(y) = b2 tanm2 (β2 y), and
χ(z) = c2 cotk2 (γ2 z).

∂w ∂w ∂w
1. +a +b = c arcsink (λx) + s.
∂x ∂y ∂z
Z
General solution: w = Φ(y − ax, z − bx) + c arcsink (λx) dx + sx.
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 , fn = fn (x, y, z) 189
∂w ∂w ∂w
2. a1 +a2 +a3 = b1 arcsin(λ1 x)+b2 arcsin(λ2 y)+b3 arcsin(λ3 z).
∂x ∂y ∂z
This is a special case of equation 2.2.7.10 with f (x) = a1 , g(y) = a2 , h(z) = a3 , ϕ(x) =
b1 arcsin(λ1 x), ψ(y) = b2 arcsin(λ2 y), and χ(z) = b3 arcsin(λ3 z).
∂w ∂w ∂w
3. a +b + c arcsinn (λx) arcsink (βz) = s arcsinm (γx).
∂x ∂y ∂z
f4 (x) = s arcsinm (γx), g(y) = 1, and h(z) = arcsink (βz).
∂w ∂w ∂w
4. a +b + c arcsinn (λx) arcsinm (βy) arcsink (γz) = s.
∂x ∂y ∂z
This is a special case of equation 2.2.7.19 with f (x, y) = c arcsinn (λx) arcsinm (βy),
g(z) = arcsink (γz), and h(x, y) = s.
∂w ∂w ∂w
5. a + b arcsinn (λx) + c arcsink (βz) = s arcsinm (γx).
∂x ∂y ∂z
f4 (x) = s arcsinm (γx), g(y) = 1, and h(z) = c arcsink (βz).
∂w ∂w ∂w
6. a + b arcsinn (λy) + c arcsink (βz) = s.
∂x ∂y ∂z
This is a special case of equation 2.2.7.10 with f (x) = a, g(y) = b arcsinn (λy), h(z) =
c arcsink (βz), ϕ(x) = s, and ψ(y) = χ(z) = 0.

∂w ∂w ∂w
7. +a +b = c arccosk (λx) + s.
∂x ∂y ∂z
Z
General solution: w = Φ(y − ax, z − bx) + c arccosk (λx) dx + sx.
∂w ∂w ∂w
8. a1 +a2 +a3 = b1 arccos(λ1 x)+b2 arccos(λ2 y)+b3 arccos(λ3 z).
∂x ∂y ∂z
b1 arccos(λ1 x), ψ(y) = b2 arccos(λ2 y), and χ(z) = b3 arccos(λ3 z).
∂w ∂w ∂w
9. a +b + c arccosn (λx) arccosk (βz) = s arccosm (γx).
∂x ∂y ∂z
f4 (x) = s arccosm (γx), g(y) = 1, and h(z) = arccosk (βz).
∂w ∂w ∂w
10. a + b arccosn (λx) + c arccosk (βz) = s arccosm (γx).
∂x ∂y ∂z
f4 (x) = s arccosm (γx), g(y) = 1, and h(z) = c arccosk (βz).
∂w ∂w ∂w
11. a + b arccosn (λy) + c arccosk (βz) = s.
∂x ∂y ∂z
This is a special case of equation 2.2.7.10 with f (x) = a, g(y) = b arccosn (λy), h(z) =
c arccosk (βz), ϕ(x) = s, and ψ(y) = χ(z) = 0.

∂w ∂w ∂w
12. +a +b = c arctank (λx) + s.
∂x ∂y ∂z
Z
General solution: w = Φ(y − ax, z − bx) + c arctank (λx) dx + sx.
∂w ∂w ∂w
13. a1 +a2 +a3 = b1 arctan(λ1 x)+b2 arctan(λ2 y)+b3 arctan(λ3 z).
∂x ∂y ∂z
b1 arctan(λ1 x), ψ(y) = b2 arctan(λ2 y), and χ(z) = b3 arctan(λ3 z).
∂w ∂w ∂w
14. a +b + c arctann (λx) arctank (βz) = s arctanm (γx).
∂x ∂y ∂z
f4 (x) = s arctanm (γx), g(y) = 1, and h(z) = arctank (βz).
∂w ∂w ∂w
15. a +b + c arctann (λx) arctanm (βy) arctank (γz) = s.
∂x ∂y ∂z
This is a special case of equation 2.2.7.19 with f (x, y) = c arctann (λx) arctanm (βy),
g(z) = arctank (γz), and h(x, y) = s.
∂w ∂w ∂w
16. a + b arctann (λx) + c arctank (βz) = s arctanm (γx).
∂x ∂y ∂z
f4 (x) = s arctanm (γx), g(y) = 1, and h(z) = c arctank (βz).

∂w ∂w ∂w
17. +a +b = c arccot k (λx) + s.
∂x ∂y ∂z
Z
General solution: w = Φ(y − ax, z − bx) + c arccotk (λx) dx + sx.
∂w ∂w ∂w
18. a1 + a2 + a3 = b1 arccot(λ1 x) + b2 arccot(λ2 y) + b3 arccot(λ3 z).
∂x ∂y ∂z
b1 arccot(λ1 x), ψ(y) = b2 arccot(λ2 y), and χ(z) = b3 arccot(λ3 z).
∂w ∂w ∂w
19. a +b + c arccotn (λx) arccot k (βz) = s arccot m (γx).
∂x ∂y ∂z
f4 (x) = s arccotm (γx), g(y) = 1, and h(z) = arccotk (βz).
∂w ∂w ∂w
20. a +b + c arccotn (λx) arccot m (βy) arccot k (γz) = s.
∂x ∂y ∂z
This is a special case of equation 2.2.7.19 with f (x, y) = c arccot n (λx) arccotm (βy),
g(z) = arccotk (γz), and h(x, y) = s.
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 , fn = fn (x, y, z) 191
∂w ∂w ∂w
21. a + b arccotn (λx) + c arccot k (βz) = s arccot m (γx).
∂x ∂y ∂z
f4 (x) = s arccotm (γx), g(y) = 1, and h(z) = c arccot k (βz).

◆ Throughout Section 2.2.7, sometimes only a particular solution w e of the nonhomoge-
neous equation and a basis u1 , u2 of the corresponding homogeneous equation are pre-
sented. The general solution can be obtained as w = w
e + Φ(u1 , u2 ), where Φ(u1 , u2 ) is an
arbitrary function of two variables.
∂w ∂w ∂w
1. + f (x) + g(x) = h2 (x)y + h1 (x)z + h0 (x).
∂x ∂y ∂z
General solution:
Z Z
w = H2 (x)y + H1 (x)z + H0 (x) − f (x)H2 (x) dx − g(x)H1 (x) dx + Φ(u1 , u2 ),
where
Z Z Z
Hk (x) = hk (x) dx (k = 0, 1, 2), u1 = y − f (x) dx, u2 = z − g(x) dx.
∂w ∂w ∂w
2. + f (x)(y + a) + g(x)(z + b) = h(x).
∂x ∂y ∂z
General solution:
Z Z Z
w = h(x) dx+Φ(u1 , u2 ), u1 = ln |y +a|− f (x) dx, u2 = ln |z +b|− g(x) dx.
∂w ∂w ∂w
3. + [ay + f (x)] + [bz + g(x)] = h(x).
∂x ∂y ∂z
General solution:
Z Z Z
−ax −ax
w= h(x) dx+Φ(u1, u2), u1 =ye − f (x)e dx, u2 =ze − g(x)e−bx dx.
−bx
∂w ∂w ∂w
4. + f1 (x)y + f2 (x) + g1 (x)y + g2 (x) = h2 (x)y + h1 (x)z +
∂x ∂y ∂z
h0 (x).
Particular solution:
Z

w=ϕ(x)y+ψ(x)z+
e h0(x)−f2(x)ϕ(x)−g2(x)ψ(x) dx,
Z Z Z
h2(x)−g1(x)ψ(x)
ϕ(x)=F (x) dx, ψ(x)= h1(x) dx, F (x)=exp − f1(x) dx .
F (x)
For an integral basis u1 , u2 of the homogeneous equation, see 2.1.7.4.
∂w ∂w ∂w
5. + f1 (x)y + f2 (x) + g1 (x)z + g2 (x) = h2 (x)y + h1 (x)z +
∂x ∂y ∂z
h0 (x).
Z

w
e = ϕ(x)y + ψ(x)z + h0 (x) − f2 (x)ϕ(x) − g2 (x)ψ(x) dx,
Z Z
h2 (x)
ϕ(x) = F (x) dx, F (x) = exp − f1 (x) dx ,
F (x)
Z Z
h1 (x)
ψ(x) = G(x) dx, G(x) = exp − g1 (x) dx .
G(x)
For an integral basis u1 , u2 of the homogeneous equation, see 2.1.7.5.
∂w ∂w
6. + y 2 − a2 + aλ sinh(λx) − a2 sinh2 (λx)
∂x ∂y
∂w
+ f (x) sinh(γz) = g(x).
∂z
Z
General solution: w = g(x) dx + Φ(u1 , u2 ), where
Z Z
1 γz E 2a
u1 = f (x) dx− lntanh , u2 = + E dx, E =exp sinh(λx) .
γ 2 y−a cosh(λx) λ
∂w ∂w ∂w
7. + f1 (x)y + f2 (x)y k + g1 (x)z + g2 (x)z m = h(x).
∂x ∂y ∂z
1◦ . For k 6= 1 and m 6= 1, the transformation
Z
1−k 1−m
ξ=y , η=z , W =w− h(x) dx

∂W ∂W ∂W
+ (1 − k) f1 (x)ξ + f2 (x) + (1 − m) g1 (x)η + g2 (x) = 0.
∂x ∂ξ ∂η
Z
◦ 1−k
2 . For k 6= 1 and m = 1, the transformation ξ = y , W = w − h(x) dx also leads
to an equation of the form 2.1.7.5.
3◦ . For k = m = 1 see equation 2.2.7.5 with h1 (x) = h2 (x) = 0.
∂w ∂w ∂w
8. + f1 (x)y + f2 (x)y k + g1 (x) + g2 (x)eλz = h(x).
∂x ∂y ∂z
The transformation
Z
ξ = y 1−k , η = e−λz , W =w− h(x) dx

∂W ∂W ∂W
+ (1 − k) f1 (x)ξ + f2 (x) − λ g1 (x)η + g2 (x) = 0.
∂x ∂ξ ∂η
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 , fn = fn (x, y, z) 193
∂w ∂w ∂w
9. + f1 (x) + f2 (x)eλy + g1 (x) + g2 (x)eβz = h(x).
∂x ∂y ∂z
The transformation
Z
−λy −βz
ξ=e , η=e , W =w− h(x) dx
∂W ∂W ∂W
− λ f1 (x)ξ + f2 (x) − β g1 (x)η + g2 (x) = 0.
∂x ∂ξ ∂η
∂w ∂w ∂w
10. f (x) + g(y) + h(z) = ϕ(x) + ψ(y) + χ(z).
∂x ∂y ∂z
General solution:
Z Z Z
ϕ(x) ψ(y) χ(z)
w= dx + dy + dz + Φ(u1 , u2 ),
f (x) g(y) h(z)
where Z Z Z Z
dx dy dx dz
u1 = − , u2 = − .
f (x) g(y) f (x) h(z)
∂w ∂w ∂w
11. f (x) +z + g(y) = h2 (x) + h1 (y).
∂x ∂y ∂z
General solution:
Z Z y
h2 (x) h1 (t) dt
w= dx + p + Φ(u1 , u2 ),
f (x) y0 2G(t) − 2G(y) + z 2
where
Z Z Z y
z2 dx dt
G(y) = g(y) dy, u1 = G(y) − , u2 = − p .
2 f (x) y0 2G(t) − 2G(y) + z 2
∂w ∂w ∂w
12. f1 (x) + f2 (x)g(y) + f3 (x)h(z) = f4 (x).
∂x ∂y ∂z
General solution:
Z Z Z Z Z
f4 (x) f2 (x) dy f3 (x) dz
w= dx+Φ(u1 , u2 ), u1 = dx− , u2 = dx− .
f1 (x) f1 (x) g(y) f1 (x) h(z)
∂w ∂w ∂w
13. + f1 (x)y + f2 (x) + g1 (x)z + g2 (y) = h1 (x) + h2 (y).
∂x ∂y ∂z
This is a special case of equation 2.2.7.22 with g1 (x, y) = g1 (x), g2 (x, y) = g2 (y), and
h(x, y, z) = h1 (x) + h2 (y).
∂w ∂w ∂w
14. + f1 (x)y + f2 (x)y k + g1 (y)z + g2 (x)z m = h1 (x) + h2(y).
∂x ∂y ∂z
This is a special case of equation 2.2.7.23 with g1 (x, y) = g1 (y), g2 (x, y) = g2 (x), and
h(x, y, z) = h1 (x) + h2 (y).
∂w ∂w ∂w
15. + f1 (x)y + f2 (x)y k + g1 (x) + g2 (y)eλz = h1 (x) + h2 (y).
∂x ∂y ∂z
h(x, y, z) = h1 (x) + h2 (y).
∂w ∂w ∂w
16. + f1 (x) + f2 (x)eλy + g1 (y)z + g2 (x)z k = h1 (x) + h2 (y).
∂x ∂y ∂z
h(x, y, z) = h1 (x) + h2 (y).
∂w ∂w ∂w
17. + f1 (x) + f2 (x)eλy + g1 (x) + g2 (y)eβz = h1 (x) + h2 (y).
∂x ∂y ∂z
h(x, y, z) = h1 (x) + h2 (y).

∂w ∂w ∂w
18. a +b + f (x, y) = g(x, y).
∂x ∂y ∂z
Z
1 y a(t − y)
General solution: w = Φ(u1 , u2 ) + g x+ , t dt, where
b y0 b
Z y
a(t − y)
u1 = bx − ay, u2 = bz − f x+ , t dt,
y0 b
and y0 may be taken as arbitrary.

∂w ∂w ∂w
19. a +b + f (x, y)g(z) = h(x, y).
∂x ∂y ∂z
Z
1 y a(t − y)
General solution: w = Φ(u1 , u2 ) + h x+ , t dt, where
b y0 b
Z Z y
dz a(t − y)
u1 = bx − ay, u2 = b − f x+ , t dt.
g(z) y0 b
∂w ∂w ∂w
20. x +y + [z + f (x, y)] = g(x, y).
∂x ∂y ∂z
Z y
xt dt
General solution: w = Φ(u1 , u2 ) + g ,t , where
y0 y t
Z y
y z xt dt
u1 = , u2 = − f ,t 2 .
x y y0 y t
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 , fn = fn (x, y, z) 195
∂w ∂w ∂w
21. ax + by
+ f (x, y)g(z) = h(x, y).
∂x ∂y ∂z
Z y
1
General solution: w = Φ(u1 , u2 ) + t−1 h xy −a/b ta/b , t dt, where
b y0
Z Z y
b −a dz
g(z) y0
∂w ∂w ∂w
22. + f1 (x)y + f2 (x) + g1 (x, y)z + g2 (x, y) = h(x, y, z).
∂x ∂y ∂z
General solution: Z x
w = Φ(u1 , u2 ) + h̄(t, u1 , u2 ) dt,
x0
where
Z Z
u1 = yF (x) − f2 (x)F (x) dx, F (x) = exp − f1 (x) dx , (1)
Z x Z x
u2 = zG(x, u1 ) − ḡ2 (t, u1 )G(t, u1 ) dt, G(x, u1 ) = exp − ḡ1 (t, u1 ) dt . (2)
x0 x0
Here ḡ1 (x, u1 ) ≡ g1 (x, y), ḡ2 (x, u1 ) ≡ g2 (x, y), h̄(x, u1 , u2 ) ≡ h(x, y, z) [in these func-
tions, y must be expressed via x and u1 from relation (1), and z must be expressed via
x, u1 , and u2 from relation (2)], and x0 is an arbitrary number.
∂w ∂w ∂w
23. + f1 (x)y + f2 (x)y k + g1 (x, y)z + g2 (x, y)z m = h(x, y, z).
∂x ∂y ∂z
1◦ . For k 6= 1 and m 6= 1, the transformation ξ = η= y 1−k , z 1−m
leads to an equation of
the form 2.2.7.22:
∂w ∂w ∂w
+ (1 − k) f1 (x)ξ + f2 (x) + (1 − m) ḡ1 (x, ξ)η + ḡ2 (x, ξ) = h̄(x, ξ, η),
∂x ∂ξ ∂η
where
1 1 1 1
ḡ1 (x, ξ) ≡ g1 x, ξ 1−k , ḡ2 (x, ξ) ≡ g2 x, ξ 1−k , h(x, ξ, η) ≡ h x, ξ 1−k , η 1−m .
2.2.7.22.
∂w ∂w ∂w
24. + f1 (x)y + f2 (x)y k + g1 (x, y) + g2 (x, y)eλz = h(x, y, z).
∂x ∂y ∂z
The transformation ξ = η= y 1−k , e−λz
∂w ∂w ∂w
+ (1 − k) f1 (x)ξ + f2 (x) − λ ḡ1 (x, ξ)η + ḡ2 (x, ξ) = h̄(x, ξ, η),
∂x ∂ξ ∂η
where
1 1 1 1
ḡ1 (x, ξ) ≡ g1 x, ξ 1−k , ḡ2 (x, ξ) ≡ g2 x, ξ 1−k , h̄(x, ξ, η) ≡ h x, ξ 1−k , − ln η .
λ
∂w ∂w ∂w
25. + f1 (x) + f2 (x)eλy + g1 (x, y)z + g2 (x, y)z k = h(x, y, z).
∂x ∂y ∂z
∂w ∂w ∂w
− λ f1 (x)ξ + f2 (x) + (1 − k) ḡ1 (x, ξ)η + ḡ2 (x, ξ) = h̄(x, ξ, η),
∂x ∂ξ ∂η
1
where ḡ1,2 (x, ξ) ≡ g1,2 x, − λ1 ln ξ and h̄(x, ξ, η) ≡ h x, − λ1 ln ξ, η 1−k .
∂w ∂w ∂w
26. + f1 (x) + f2 (x)eλy + g1 (x, y) + g2 (x, y)eβz = h(x, y, z).
∂x ∂y ∂z
∂w ∂w ∂w
− λ f1 (x)ξ + f2 (x) − β ḡ1 (x, ξ)η + ḡ2 (x, ξ) = h̄(x, ξ, η),
∂x ∂ξ ∂η

where ḡ1,2 (x, ξ) ≡ g1,2 x, − λ1 ln ξ and h̄(x, ξ, η) ≡ h x, − λ1 ln ξ, − β1 ln η .
∂w ∂w ∂w
27. f1(x)g1(y) +f2(x)g2(y) + h1(x, y)z+h2(x, y)z m =h3(x, y, z).
∂x ∂y ∂z
Z Z
f2 (x) g1 (y)
f1 (x) g2 (y)
2.2.7.23 with f1 ≡ 0, f2 ≡ 1, and k = 0:
∂w ∂w ∂w
+ + h̄1 (ξ, η)z + h̄2 (ξ, η)z m = h̄3 (ξ, η, z),
∂ξ ∂η ∂z
h1 (x, y) h2 (x, y) h3 (x, y, z)
where h̄1 (ξ, η) ≡ , h̄2 (ξ, η) ≡ , and h̄3 (ξ, η, z) ≡ .
f2 (x)g1 (y) f2 (x)g1 (y) f2 (x)g1 (y)
∂w ∂w ∂w
28. f1(x)g1(y) +f2(x)g2(y) + h1(x, y)+h2(x, y)eλz =h3(x, y, z).
∂x ∂y ∂z
Z Z
f2 (x) g1 (y)
f1 (x) g2 (y)
2.2.7.24 with f1 ≡ 0, f2 ≡ 1, and k = 0:
∂w ∂w ∂w
+ + h̄1 (ξ, η) + h̄2 (ξ, η)eλz = h̄3 (x, y, z),
∂ξ ∂η ∂z
h1 (x, y) h2 (x, y) h3 (x, y, z)
f2 (x)g1 (y) f2 (x)g1 (y) f2 (x)g1 (y)

∂w ∂w ∂w
f1 + f2 + f3 = f4 w, fn = fn (x, y, z)
∂x ∂y ∂z
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w, fn = fn (x, y, z) 197

∂w ∂w ∂w
1. a +b +c = (αx + βy + γz + δ)w.
∂x ∂y ∂z

α 2 β 2 γ 2 δ
General solution: w = exp x + y + z + x Φ(bx − ay, cy − bz).
2a 2b 2c a
∂w ∂w ∂w
2. + az + by = (cx + s)w.
∂x ∂y ∂z
1 2

General solution: w = exp 2 cx + sx Φ(u1 , u2 ), where
( √ √
u1 = by 2 − az 2 , u2 = p p p
∂w ∂w ∂w
3. + (a1 x + a0 ) + (b1 x + b0 ) = (αx + βy + γz + δ)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.1 with f (x) = a1 x+a0 , g(x) = b1 x+b0 , h2 (x) = β,
h1 (x) = γ, and h0 (x) = αx + δ.
∂w ∂w ∂w
4. + (a2 y + a1 x + a0 ) + (b2 y + b1 x + b0 ) = (c2 y + c1 z + c0 x + s)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.4 with f1 (x) = a2 , f2 (x) = a1 x + a0 , g1 (x) = b2 ,

g2 (x) = b1 x + b0 , h2 (x) = c2 , h1 (x) = c1 , and h0 (x) = c0 x + s.
∂w ∂w ∂w
5. + (ay + k1 x + k0 ) + (bz + s1 x + s0 ) = (c1 x + c0 )w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.3 with f (x) = k1 x + k0 , g(x) = s1 x + s0 , and

h(x) = c1 x + c0 .
∂w ∂w ∂w
6. ax + by + cz = (αx + βy + γz + δ)w.
∂x ∂y ∂z
a
δ/a α β γ |y| |z|a
General solution: w = |x| exp x+ y+ z Φ , .
a b c |x|b |x|c
∂w ∂w ∂w
7. x + az + by = cw.
∂x ∂y ∂z
General solution: w = |x|c Φ(u1 , u2 ), where

 √ √

√
u1 = by 2 − az 2 , u2 = √
−ab −ab z

by
∂w ∂w ∂w
8. abx + b(ay + bz) + a(ay − bz) = cw.
∂x ∂y ∂z
General solution: w = |x|c/(ab) Φ(u1 , u2 ), where

√ √ √ √
u1 = ay + ( 2 − 1)bz |x|− 2 , u2 = ay − ( 2 + 1)bz |x| 2 .

Particular solution: w = |x|c/(ab) Φ a2 y 2 − 2abyz − b2 z 2 .
∂w ∂w ∂w
9. (a1 x + a0 ) + (b1 y + b0 ) + (c1 z + c0 ) = (αx + βy + γz + δ)w.
∂x ∂y ∂z
1◦ . General solution for a1 b1 c1 6= 0:

α β γ 1 αa0 βb0 γc0
w = exp x+ y+ z+ δ− − − lna1 x + a0
a1 b1 c1 a1 a1 b1 c1

|b1 y + b0 |a1 |b1 y + b0 |c1
×Φ , .
|a1 x + a0 |b1 |c1 z + c0 |b1
2◦ . General solution for a1 b1 6= 0 and c1 = 0:

α β γ 2 1 αa0 βb0 |b1y+b0|a1 c0 −b1z
w=exp x+ y+ z + δ− − z Φ , |b1y+b0| e .
a1 b1 2c0 c0 a1 b1 |a1x+a0|b1
3◦ . General solution for a1 =

6 0 and b1 = c1 = 0:

α β 2 γ 2 1 αa0
w = exp x+ y + z + δ− z Φ |a1 x + a0 |b0 e−a1 y , c0 y − b0 z .
a1 2b0 2c0 c0 a1
4◦ . For a1 = b1 = c1 = 0, see equation 2.3.1.1.
∂w ∂w ∂w
10. a +b +c = (λx2 + βy 2 + γz 2 + δ)w.
∂x ∂y ∂z

λ 3 β 3 γ 3 δ
General solution: w = exp x + y + z + x Φ(bx − ay, cy − bz).
3a 3b 3c a
∂w ∂w ∂w
11. + (a1 x2 + a0 ) + (b1 x2 + b0 ) = (λx + βy + γz + δ)w.
∂x ∂y ∂z

h2 (x) = β, h1 (x) = γ, and h0 (x) = λx + δ.
∂w ∂w ∂w
12. + (ay + k1 x2 + k0 ) + (bz + s1 x2 + s0 ) = (c1 x2 + c0 )w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.3 with f (x) = k1 x2 + k0 , g(x) = s1 x2 + s0 , and

h(x) = c1 x2 + c0 .
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w, fn = fn (x, y, z) 199
∂w ∂w ∂w
13. + (a2 xy + a1 x2 + a0 ) + (b2 xy + b1 x2 + b0 )
∂x ∂y ∂z
= (c2 y + c1 z + c0 x + s)w.
This is a special case of equation 2.3.7.4 with f1 (x) = a2 x, f2 (x) = a1 x2 +a0 , g1 (x) = b2 x,
g2 (x) = b1 x2 + b0 , h2 (x) = c2 , h1 (x) = c1 , and h0 (x) = c0 x + s.
∂w ∂w ∂w
14. ax + by + cz = x(λx + βy + γz)w.
∂x ∂y ∂z
1◦ . General solution for b 6= −a and c 6= −a:

λ 2 β γ
w = exp x + xy + xz Φ x|y|−a/b , x|z|−a/c .
2a a+b a+c
2◦ . General solution for b = −a and c 6= −a:

1 γ
w = exp x λx + 2βy ln |x| + xz Φ xy, x|z|−a/c .
2a a+c
3◦ . General solution for b = c = −a:

1
w = exp x λx + 2(βy + γz) ln |x| Φ(xy, xz).
2a
∂w ∂w ∂w
15. ax2 + bxy + cxz = (λx + βy + γz)w.
∂x ∂y ∂z
1◦ . General solution for b 6= a and c 6= a:

λ/a 1 βy γz
w = |x| exp + Φ x|y|−a/b , x|z|−a/c .
x b−a c−a
2◦ . General solution for b = a and c 6= a:

λ/a 1 βy γz x −a/c
w = |x| exp ln |x| + Φ , x|z| .
x a c−a y
3◦ . General solution for a = b = c:

ln |x| x x
w = exp (λx + βy + γz) Φ , .
ax y z
∂w ∂w ∂w
16. ax2 + bxy + cz 2 = ky 2 w.
∂x ∂y ∂z
1◦ . General solution for a 6= 2b:

ky 2 −a/b c a
w = exp Φ x|y| , − .
(2b − a)x x z
2◦ . General solution for a = 2b:

ky 2 ln |x| x c a
w = exp Φ 2, − .
ax y x z
∂w ∂w ∂w
17. ax2 + by 2 + cz 2 = kxyw.
∂x ∂y ∂z

kxy ax b a c a
General solution: w = exp ln Φ − , − .
ax − by y x y x z
∂w ∂w ∂w
18. ax2 + by 2 + cz 2 = (λx2 + βy 2 + γz 2 )w.
∂x ∂y ∂z

λ β γ b a c a
General solution: w = exp x+ y+ z Φ − , − .
a b c x y x z
∂w ∂w ∂w
19. +a +b = xyzw.
∂x ∂y ∂z
1 2

General solution: w = exp 2x yz − 16 x3 (az + by) + 1
12 abx
4
Φ(y − ax, z − bx).
∂w ∂w ∂w
20. a +b +c = kx3 + sy 2 w.
∂x ∂y ∂z

k 4 s 3
General solution: w = exp x + y Φ(bx − ay, cx − az).
4a 3a
∂w ∂w ∂w √
21. a + by + cz = kx + s x w.
∂x ∂y ∂z

k 2 2s 3/2
General solution: w = exp x + x Φ |y|a e−bx , |z|a e−cx .
2a 3a
∂w ∂w ∂w √
22. + az + by = (c x + s)w.
∂x ∂y ∂z
2 3/2

General solution: w = exp 3 cx + sx Φ(u1 , u2 ), where
( √ √
u1 = by 2 − az 2 , u2 = p p p
∂w ∂w ∂w
23. ax2 + by 2 + cz 2 = kxyzw.
∂x ∂y ∂z
General solution:
1 1 1 1
w = w0 (x, y, z)Φ − , − ,
ax by ax cz
where

ax ln(ax) by ln(by) cz ln(cz)
w0(x, y, z)=exp kxyz + + .
(ax−by)(ax−cz) (by−ax)(by−cz) (cz−ax)(cz−by)
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w, fn = fn (x, y, z) 201
∂w ∂w ∂w
24. a +b +c = λxn + βy m + γz k w.
∂x ∂y ∂z
General solution:

λ n+1 β m+1 γ k+1
w = exp x + y + z Φ(bx − ay, cx − az).
a(n + 1) b(m + 1) c(k + 1)
∂w ∂w ∂w
25. a + by + cz = λxn + βy m + γz k w.
∂x ∂y ∂z

λ n+1 β m γ k
General solution: w = exp x + y + z Φ |y|a e−bx , |z|a e−cx .
a(n + 1) bm ck
∂w ∂w ∂w
26. + az + by = cxn w.
∂x ∂y ∂z

c n+1
General solution: w = exp x Φ(u1 , u2 ), where
n+1
( √ √
u1 = by − az , u2 = p p p
∂w ∂w ∂w
27. ax + by + cz = λxn + βy m + γz k w.
∂x ∂y ∂z
a
λ n β m γ k |y| |z|a
General solution: w = exp x + y + z Φ , .
an bm ck |x|b |x|c
∂w ∂w ∂w
28. x + az + by = cxn w.
∂x ∂y ∂z

c n
n
 √ √

2 2 √
u1 = by − az , u2 = √
−ab −ab z

by
∂w ∂w ∂w
29. abx + b(ay + bz) + a(ay − bz) = cxn w.
∂x ∂y ∂z

c n
abn
√ √ √ √
u1 = ay + ( 2 − 1)bz |x|− 2 , u2 = ay − ( 2 + 1)bz |x| 2 .

c n
Particular solution: w = exp x Φ a2 y 2 − 2abyz − b2 z 2 .
abn
∂w ∂w ∂w
30. + axn y m + bxν y µ z λ = cxk w.
∂x ∂y ∂z
General solution:

exp c k+1
x Φ(u1 , u2 ) if k =
6 −1,
w= k+1
 c
|x| Φ(u1 , u2 ) if k = −1,
where u1 , u2 are the integral basis of equation 2.1.1.64.
∂w ∂w ∂w
31. + a1 xn1 y+b1 xm1 +(a2 xn2 y+b2 xm2 = (c2 xk2 y+c1 xk1 z)w.
∂x ∂y ∂z
∂w ∂w ∂w
32. + a1 xn1 y+b1 xm1 +(a2 xn2 z+b2 xm2 = (c2 xk2 y+c1 xk1 z)w.
∂x ∂y ∂z
∂w ∂w ∂w
33. + a1 xn1 y + b1 y k + (a2 xn2 z + b2 z m = cxs w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.7 with f1 (x) = a1 xn1 , f2 (x) = b1 , g1 (x) = a2 xn2 ,
g2 (x) = b2 , and h(x) = cxs .
∂w ∂w ∂w
34. + a1 xn1 y + b1 y k + (a2 y n2 z + b2 z m
∂x ∂y ∂z
= (c1 xs1 + c2 y s2 + c3 z s3 )w.
This is a special case of equation 2.3.7.23 with f1 (x) = a1 xn1 , f2 (x) = b1 , g1 (x, y) = a2 y n2 ,
g2 (x, y) = b2 , and h(x, y, z) = c1 xs1 + c2 y s2 + c3 z s3 .
∂w ∂w p ∂w
35. x +y + a x2 + y 2 = bxn w.
∂x ∂y ∂z

General solution: w = w0 (x)Φ u1 , u2 , where
(
y p exp bxn/n if n 6= 0,
u1 = , u2 = a x2 + y 2 − z, w0 (x) =
x |x|b if n = 0.
∂w ∂w p ∂w
36. x +y + z−a x2 + y 2 + z 2 = bxn w.
∂x ∂y ∂z

General solution: w = w0 (x)Φ u1 , u2 , where
(
y p exp bxn/n if n 6= 0,
a−1
u1 = , u2 = |x| z+ x2 + y2 + z2 , w0 (x) =
x |x|b if n = 0.
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w, fn = fn (x, y, z) 203

∂w ∂w ∂w
1. + aeλx + beβx = ceγx w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.1 with f (x) = aeλx , g(x) = beβx , h2 (x) = h1 (x) = 0,
and h0 (x) = ceγx .
∂w ∂w ∂w
2. + aeλx + beβy = ceγy + seµz w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.22 with f1 (x) = 0, f2 (x) = aeλx , g1 (x, y) = 0,
g2 (x, y) = beβy , and h(x, y, z) = ceγy + seµz .
∂w ∂w ∂w
3. + aeλy + beβy = ceγx + seµz w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.26 with f1 (x) = 0, f2 (x) = a, g1 (x, y) = beβy ,
g2 (x, y) = 0, and h(x, y, z) = ceγx + seµz .
∂w ∂w ∂w
4. + A1 eα1 x + B1 eν1 x+λy + A2 eα2 x + B2 eν2 x+βz = keγz w.
∂x ∂y ∂z
A2 eα2 x , g2 (x) = B2 eν2 x , and h(x) = keγz .
∂w ∂w ∂w
5. aeαx + beβy + ceγz = keλx w.
∂x ∂y ∂z
General solution:


 k (λ−α)x
exp e Φ(u1 , u2 ) if λ 6= α,
w= a(λ− α)

 k
exp x Φ(u1 , u2 ) if λ = α,
a
1 −αx 1 −βy 1 1 −γz
where u1 = − e + e and u2 = − e−βy + e .
aα bβ bβ cγ
∂w ∂w ∂w
6. aeβy + beαx + ceγz = keλx w.
∂x ∂y ∂z
General solution:
 Z

 kα eλx dx

exp αx
Φ(u1 , u2 ) if λ 6= α, λ 6= 0;

 bβ e + aαu1

k
w = exp y Φ(u1 , u2 ) if λ = α 6= 0;

 b



 k
exp − e−γz Φ(u1 , u2 ) if λ = 0,
cγ
where u1 = − e + e and u2 = αx βy
+ e . In the integration,
u1 is considered a parameter.
∂w ∂w ∂w
7. (a1 + a2 eαx ) + (b1 + b2 eβy ) + (c1 + c2 eγz ) = k1 + k2 eαx w.
∂x ∂y ∂z

k1 1 k2 k1 αx
General solution: w = Φ(u1 , u2 ) exp x+ − ln(a1 + a2 e ) , where
a1 α a2 a1
1 1
a1 α b1 β
1 1
a1 α c1 γ
∂w ∂w ∂w
8. eβy (a1 + a2 eαx ) + eαx (b1 + b2 eβy ) + ceβy+γz
∂x ∂y ∂z
= k3 eβy (k1 + k2 eαx )w.

k1 k3 k3 k2 k1
General solution: w = Φ(u1 , u2 ) exp x+ − ln(a1 + a2 eαx ) , where
a1 α a2 a1
1 1 1 1
u1 = ln(a1 +a2 eαx )− ln(b1 +b2 eβy ), u2 = αx−ln(a1 +a2 eαx ) + e−γz .
a2 α b1 β a1 α cγ
∂w ∂w ∂w
9. + axn + bxm = ceλx y + keβx z + seγx w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.1 with f (x) = axn , g(x) = bxm , h2 (x) = ceλx ,
∂w ∂w ∂w
10. + aeλx + bxm = cxn y + keβx z + seγx w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.1 with f (x) = aeλx , g(x) = bxm , h2 (x) = cxn ,
∂w ∂w ∂w
11. + aeλx + by = keβx z + seγx w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.4 with f1 (x) = 0, f2 (x) = aeλx , g1 (x) = b, g2 (x) =
h2 (x) = 0, h1 (x) = keβx , and h0 (x) = seγx .
∂w ∂w ∂w
12. + ay n + bz m = ceλx + keβy + seγz w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.10 with f (x) = 1, g(y) = ay n , h(z) = bz m ,

ϕ(x) = ceλx , ψ(y) = keβy , and χ(z) = seγz .
∂w ∂w ∂w
13. + aeβy + bz m = ceλx + ky n + seγz w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.10 with f (x) = 1, g(y) = aeβy , h(z) = bz m ,
ϕ(x) = ceλx , ψ(y) = ky n , and χ(z) = seγz .
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w, fn = fn (x, y, z) 205
∂w ∂w ∂w
14. + a1 eλ1 x y + b1 eβ1 x y k + a2 eλ2 x z + b2 eβ2 x z m = cxs w.
∂x ∂y ∂z
a2 eλ2 x , g2 (x) = b2 eβ2 x , and h(x) = cxs .
∂w ∂w ∂w
15. + a1 eβ1 x y + b1 eγ1 x y k + a2 eβ2 x + b2 eγ2 x+λz = cxs w.
∂x ∂y ∂z
a2 eβ2 x , g2 (x) = b2 eγ2 x , and h(x) = cxs .
∂w ∂w ∂w
16. + a1 xn + b1 xm eλy + a2 xk + b2 xl eβz = cxs w.
∂x ∂y ∂z

g2 (x) = b2 xl , and h(x) = cxs .

∂w ∂w ∂w
1. +a +b = c sinhn (βx)w.
∂x ∂y ∂z
Z
n
General solution: w = Φ(y − ax, z − bx) exp c sinh (βx) dx .
∂w ∂w ∂w
2. a +b + c sinh(λx) = k sinh(βy) + s sinh(γz) w.
∂x ∂y ∂z
General solution:
Z x
k s cγ
w = Φ(u1 , u2 ) exp cosh(βy) − sinh cosh(λx) − cosh(λt) − γz dt ,
bβ a 0 aλ
where u1 = bx − ay and u2 = aλz − c cosh(λx).
∂w ∂w ∂w
3. + a sinhn (βx) + b sinhk (λx) = c sinhm (γx)w.
∂x ∂y ∂z
h2 (x) = h1 (x) = 0, and h0 (x) = c sinhm (γx).
∂w ∂w ∂w
4. a + b sinh(βy) + c sinh(λx) = k sinh(γz)w.
∂x ∂y ∂z
General solution:
Z x
k cγ
w = Φ(u1 , u2 ) exp sinh γz + cosh(λx) − cosh(λt) dt ,
a 0 aλ
βy

where u1 = bβx − a lntanh and u2 = aλz − c cosh(λx).
2
∂w ∂w ∂w
5. a1 sinhn1 (λ1 x) + b1 sinhm1 (β1 y) + c1 sinhk1 (γ1 z)
∂x ∂y ∂z

= a2 sinhn2 (λ2 x) + b2 sinhm2 (β2 y) + c2 sinhk2 (γ2 z) w.
This is a special case of equation 2.3.7.10 in which f (x) = a1 sinhn1 (λ1 x), g(y) =
b1 sinhm1 (β1 y), h(z) = c1 sinhk1 (γ1 z), ϕ(x) = a2 sinhn2 (λ2 x), ψ(y) = b2 sinhm2 (β2 y),
and χ(z) = c2 sinhk2 (γ2 z).
∂w ∂w ∂w
6. +a +b = c coshn (βx)w.
∂x ∂y ∂z
Z
n
General solution: w = Φ(y − ax, z − bx) exp c cosh (βx) dx .
∂w ∂w ∂w
7. a +b + c cosh(λx) = k cosh(βy) + s cosh(γz) w.
∂x ∂y ∂z
General solution:
Z x
k s cγ
w = Φ(u1 , u2 ) exp sinh(βy) + cosh cosh(λt) − cosh(λx) + γz dt ,
bβ a 0 aλ
where u1 = bx − ay and u2 = aλz − c sinh(λx).
∂w ∂w ∂w
8. + a coshn (βx) + b coshk (λx) = c coshm (γx)w.
∂x ∂y ∂z
h2 (x) = h1 (x) = 0, and h0 (x) = c coshm (γx).
∂w ∂w ∂w
9. a + b cosh(βy) + c cosh(λx) = k cosh(γz)w.
∂x ∂y ∂z
General solution:
Z x
k cγ
w = Φ(u1 , u2 ) exp cosh γz + sinh(λt) − sinh(λx) dt ,
a 0 aλ
βy

where u1 = bβx − 2a arctantanh and u2 = aλz − c sinh(λx).
2
∂w ∂w ∂w
10. a1 coshn1 (λ1 x) + b1 coshm1 (β1 y) + c1 coshk1 (γ1 z)
∂x ∂y ∂z

= a2 cosh (λ2 x) + b2 cosh (β2 y) + c2 coshk2 (γ2 z) w.
n2 m2
This is a special case of equation 2.3.7.10 in which f (x) = a1 coshn1 (λ1 x), g(y) =
b1 coshm1 (β1 y), h(z) = c1 coshk1 (γ1 z), ϕ(x) = a2 coshn2 (λ2 x), ψ(y) = b2 coshm2 (β2 y),
χ(z) = c2 coshk2 (γ2 z).
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w, fn = fn (x, y, z) 207
∂w ∂w ∂w
11. +a +b = c tanhn (βx)w.
∂x ∂y ∂z
Z
General solution: w = Φ(y − ax, z − bx) exp c tanhn (βx) dx .
∂w ∂w ∂w
12. a +b + c tanh(βz) = k tanh(λx) + s tanh(γy) w.
∂x ∂y ∂z
General solution: w = coshk/aλ (λx) coshs/bγ (γy)Φ(u1 , u2 ), where

u1 = bx − ay, u2 = cβx − a lnsinh(βz).
∂w ∂w ∂w
13. + a tanhn (βx) + b tanhk (λx) = c tanhm (γx)w.
∂x ∂y ∂z
h2 (x) = h1 (x) = 0, and h0 (x) = c tanhm (γx).
∂w ∂w ∂w
14. a + b tanh(βy) + c tanh(λx) = k tanh(γz)w.
∂x ∂y ∂z
General solution:
Z
k x cγ
w = Φ(u1 , u2 ) exp tanh γz + ln |cosh(λx)| − ln |cosh(λt)| dt ,
a 0 aλ

where u1 = bβx − a lnsinh(βy) and u2 = aλz − c lncosh(λx).
∂w ∂w ∂w
15. a + b tanh(βy) + c tanh(γz) = k tanh(λx)w.
∂x ∂y ∂z
General solution: w = coshk/aλ (λx)Φ(u1 , u2 ), where

u1 = bβx − a lnsinh(βy), u2 = cγx − a lnsinh(γz).
∂w ∂w ∂w
16. a1 tanhn1 (λ1 x) + b1 tanhm1 (β1 y) + c1 tanhk1 (γ1 z)
∂x ∂y ∂z

= a2 tanhn2 (λ2 x) + b2 tanhm2 (β2 y) + c2 tanhk2 (γ2 z) w.
This is a special case of equation 2.3.7.10 in which f (x) = a1 tanhn1 (λ1 x), g(y) =
b1 tanhm1 (β1 y), h(z) = c1 tanhk1 (γ1 z), ϕ(x) = a2 tanhn2 (λ2 x), ψ(y) = b2 tanhm2 (β2 y),
and χ(z) = c2 tanhk2 (γ2 z).
∂w ∂w ∂w
17. +a +b = c cothn (βx)w.
∂x ∂y ∂z
Z
n
General solution: w = Φ(y − ax, z − bx) exp c coth (βx) dx .
∂w ∂w ∂w
18. a +b + c coth(βz) = [k coth(λx) + s coth(γy)]w.
∂x ∂y ∂z
General solution: w = |sinh(λx)|k/aλ |sinh(γy)|s/bγ Φ(u1 , u2 ), where

u1 = bx − ay, u2 = cβx − a ln cosh(βz) .
∂w ∂w ∂w
19. + a cothn (βx) + b cothk (λx) = c cothm (γx)w.
∂x ∂y ∂z
h2 (x) = h1 (x) = 0, and h0 (x) = c cothm (γx).
∂w ∂w ∂w
20. a + b coth(βy) + c coth(λx) = k coth(γz)w.
∂x ∂y ∂z
General solution:

Z
k x cγ
w = Φ(u1 , u2 ) exp coth γz + ln |sinh(λt)| − ln |sinh(λx)| dt ,
a 0 aλ

where u1 = bβx − a ln cosh(βy) and u2 = aλz − c lnsinh(λx).
∂w ∂w ∂w
21. a + b coth(βy) + c coth(γz) = k coth(λx)w.
∂x ∂y ∂z
k/aλ
General solution: w = sinh(λx) Φ(u1 , u2 ), where

u1 = bβx − a ln cosh(βy) , u2 = cγx − a ln cosh(γz) .
∂w ∂w ∂w
22. a1 cothn1 (λ1 x) + b1 cothm1 (β1 y) + c1 cothk1 (γ1 z)
∂x ∂y ∂z

= a2 cothn2 (λ2 x) + b2 cothm2 (β2 y) + c2 cothk2 (γ2 z) w.
This is a special case of equation 2.3.7.10 in which f (x) = a1 cothn1 (λ1 x), g(y) =
b1 cothm1 (β1 y), h(z) = c1 cothk1 (γ1 z), ϕ(x) = a2 cothn2 (λ2 x), ψ(y) = b2 cothm2 (β2 y),
and χ(z) = c2 cothk2 (γ2 z).

∂w ∂w ∂w
23. a +b + c sinhn (λy) = s coshm (βx) + k sinhl (γy) w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.18 with f (x, y) = c sinhn (λy) and g(x, y) =
s coshm (βx) + k sinhl (γy).
∂w ∂w ∂w
24. + a sinhn (λx) + b coshm (βx) = s coshk (γx)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.1 with f (x) = a sinhn (λx), g(x) = b coshm (βx),
h2 (x) = h1 (x) = 0, and h0 (x) = s coshk (γx).
∂w ∂w ∂w
25. + a coshn (λx) + b sinhm (βy) = s sinhk (γz)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.22 with f1 (x) = 0, f2 (x) = a coshn (λx), g1 (x, y) =
0, g2 (x, y) = b sinhm (βy), and h(x, y, z) = s sinhk (γz).
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w, fn = fn (x, y, z) 209
∂w ∂w ∂w
26. + a tanhn (λx) + b cothm (βx) = s cothk (γx)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.1 with f (x) = a tanhn (λx), g(x) = b cothm (βx),
h2 (x) = h1 (x) = 0, and h0 (x) = s cothk (γx).
∂w ∂w ∂w
27. a sinh(λx) + b sinh(βy) + c sinh(γz) = k cosh(λx)w.
∂x ∂y ∂z
k/aλ
General solution: w = sinh(λx) Φ(u1 , u2 ), where

∂w ∂w ∂w
1. +a +b = c lnn (βx)w.
∂x ∂y ∂z
Z
n
General solution: w = Φ(y − ax, z − bx) exp c ln (βx) dx .
∂w ∂w ∂w
2. a +b + c lnn (βx) = s lnm (λy)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.18 with f (x, y) = c lnn (βx) and g(x, y) = s lnm (λy).
∂w ∂w ∂w
3. + a lnn (βx) + b lnk (λx) = c lnm (γx)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.1 with f (x) = a lnn (βx), g(x) = b lnk (λx), h2 (x) =
h1 (x) = 0, and h0 (x) = c lnm (γx).
∂w ∂w ∂w
4. + a lnn (βx) + b lnk (λy) = c lnm (γx)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.22 with f1 (x) = 0, f2 (x) = a lnn (βx), g1 (x, y) = 0,
g2 (x, y) = b lnk (λy), and h(x, y) = c lnm (γx).
∂w ∂w ∂w
5. a1 lnn1 (λ1 x) + b1 lnm1 (β1 y) + c1 lnk1 (γ1 z)
∂x ∂y ∂z

= a2 ln (λ2 x) + b2 ln (β2 y) + c2 lnk2 (γ2 z) w.
n2 m2
This is a special case of equation 2.3.7.10 with f (x) = a1 lnn1 (λ1 x), g(y) = b1 lnm1 (β1 y),
h(z) = c1 lnk1 (γ1 z), ϕ(x) = a2 lnn2 (λ2 x), ψ(y) = b2 lnm2 (β2 y), and χ(z) = c2 lnk2 (γ2 z).

∂w ∂w ∂w
6. a +b + cxn lnk (λy) = sy m lnl (βx)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.18 with f (x, y) = cxn lnk (λy) and g(x, y) =
sy m lnl (βx).
∂w ∂w ∂w
7. + axn + bxm = cy lnk (λx) + sz lnl (βx) w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.1 with f (x) = axn , g(x) = bxm , h2 (x) = c lnk (λx),
h1 (x) = s lnl (βx), and h0 (x) = 0.
∂w ∂w ∂w
8. + a lnn (λx) + by m = c lnk (βx) + s lnl (γz) w.
∂x ∂y ∂z
g2 (x, y) = by m , and h(x, y, z) = c lnk (βx) + s lnl (γz).
∂w ∂w ∂w
9. a lnn (λx) +z + b lnk (βy) = cxm + s ln(γy) w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.11 with f (x) = a lnn (λx), g(y) = b lnk (βy), h2 (x) =
cxm , and h1 (y) = s ln(γy).
∂w ∂w ∂w
10. ax(ln x)n + by(ln y)m + cz(ln z)l = k(ln x)s w.
∂x ∂y ∂z
General solution:

exp k s−n+1
(ln x) Φ(u1 , u2 ) if s + 1 6= n,
w= a(s − n + 1)

(ln x)k/a Φ(u1 , u2 ) if s + 1 = n,
where
u1 = − , u2 = − .
a(n − 1) b(m − 1) a(n − 1) c(l − 1)

∂w ∂w ∂w
1. +a +b = c sinn (λx)w.
∂x ∂y ∂z
Z
n
General solution: w = Φ(y − ax, z − bx) exp c sin (λx) dx .
∂w ∂w ∂w
2. a +b + c sin(λz) = k sin(γx) + s sin(βy) w.
∂x ∂y ∂z

k s
General solution: w = Φ(u1 , u2 ) exp − cos(γx) − cos(βy) , where
aγ bβ
λz

u1 = bx − ay, u2 = cλx − a lntan .
2
∂w ∂w ∂w
3. + a sinn (λx) + b sinm (βx) = c sink (γx)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.1 with f (x) = a sinn (λx), g(x) = b sinm (βx),
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w, fn = fn (x, y, z) 211
∂w ∂w ∂w
4. + a sinn (λx) + b sinm (βy) = c sink (γy) + s sinl (µz) w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.22 with f1 (x) = 0, f2 (x) = a sinn (λx), g1 (x, y) = 0,
g2 (x, y) = b sinm (βy), and h(x, y, z) = c sink (γy) + s sinl (µz).
∂w ∂w ∂w
5. a + b sin(βy) + c sin(λx) = k sin(γz).
∂x ∂y ∂z
General solution:
Z x
k cγ
w = Φ(u1 , u2 ) exp sin γz + cos(λx) − cos(λt) dt ,
a 0 aλ
βy

where u1 = bβx − a lntan and u2 = aλz + c cos(λx).
2
∂w ∂w ∂w
6. a1 sinn1 (λ1 x) + b1 sinm1 (β1 y) + c1 sink1 (γ1 z)
∂x ∂y ∂z

= a2 sin (λ2 x) + b2 sin (β2 y) + c2 sink2 (γ2 z) w.
n2 m2
This is a special case of equation 2.3.7.10 with f (x) = a1 sinn1 (λ1 x), g(y) = b1 sinm1 (β1 y),
h(z) = c1 sink1 (γ1 z), ϕ(x) = a2 sinn2 (λ2 x), ψ(y) = b2 sinm2 (β2 y), χ(z) = c2 sink2 (γ2 z).
∂w ∂w ∂w
7. +a +b = c cosn (βx)w.
∂x ∂y ∂z
Z
n
General solution: w = Φ(y − ax, z − bx) exp c cos (βx) dx .
∂w ∂w ∂w
8. a +b + c cos(βz) = k cos(λx) + s cos(γy) w.
∂x ∂y ∂z

k s
General solution: w = Φ(u1 , u2 ) exp sin(λx) + sin(γy) , where
aλ bγ

u1 = bx − ay, u2 = cβx − a lnsec(βz) + tan(βz).
∂w ∂w ∂w
9. + a cosn (βx) + b cosk (λx) = c cosm (γx)w.
∂x ∂y ∂z
h2 (x) = h1 (x) = 0, and h0 (x) = c cosm (γx).
∂w ∂w ∂w
10. + a cosn (βx) + b cosm (λy) = c cosk (γy) + s cosl (µz) w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.22 with f1 (x) = 0, f2 (x) = a cosn (βx), g1 (x, y) = 0,
g2 (x, y) = b cosm (λy), and h(x, y, z) = c cosk (γy) + s cosl (µz).
∂w ∂w ∂w
11. a + b cos(βy) + c cos(λx) = k cos(γz)w.
∂x ∂y ∂z
General solution:
Z
k x cγ
w = Φ(u1 , u2 ) exp cos γz + sin(λt) − sin(λx) dt ,
a 0 aλ

where u1 = bβx − a lnsec(βy) + tan(βy) and u2 = aλz − c sin(λx).
∂w ∂w ∂w
12. a1 cosn1 (λ1 x) + b1 cosm1 (β1 y) + c1 cosk1 (γ1 z)
∂x ∂y ∂z

= a2 cos (λ2 x) + b2 cos (β2 y) + c2 cosk2 (γ2 z) w.
n2 m2
This is a special case of equation 2.3.7.10 in which f (x) = a1 cosn1 (λ1 x), g(y) =
b1 cosm1 (β1 y), h(z) = c1 cosk1 (γ1 z), ϕ(x) = a2 cosn2 (λ2 x), ψ(y) = b2 cosm2 (β2 y), and
χ(z) = c2 cosk2 (γ2 z).

∂w ∂w ∂w
13. +a +b = c tann (βx)w.
∂x ∂y ∂z
Z
n
General solution: w = Φ(y − ax, z − bx) exp c tan (βx) dx .
∂w ∂w ∂w
14. a +b + c tan(βz) = k tan(λx) + s tan(γy) w.
∂x ∂y ∂z
General solution:
−k/aλ
w = cos(λx) cos(γy)−s/bγ Φ(u1 , u2 ),

where u1 = bx − ay and u2 = cβx − a lnsin(βz).
∂w ∂w ∂w
15. + a tann (βx) + b tank (λx) = c tanm (γx)w.
∂x ∂y ∂z
h2 (x) = h1 (x) = 0, and h0 (x) = c tanm (γx).
∂w ∂w ∂w
16. a + b tan(βy) + c tan(λx) = k tan(γz)w.
∂x ∂y ∂z
General solution:
Z
k x cγ
w = Φ(u1 , u2 ) exp tan γz + ln |cos(λx)| − ln |cos(λt)| dt ,
a 0 aλ

where u1 = bβx − a lnsin(βy) and u2 = aλz + c lncos(λx).
∂w ∂w ∂w
17. a1 tann1 (λ1 x) + b1 tanm1 (β1 y) + c1 tank1 (γ1 z)
∂x ∂y ∂z

= a2 tan (λ2 x) + b2 tan (β2 y) + c2 tank2 (γ2 z) w.
n2 m2
b1 tanm1 (β1 y), h(z) = c1 tank1 (γ1 z), ϕ(x) = a2 tann2 (λ2 x), ψ(y) = b2 tanm2 (β2 y),
and χ(z) = c2 tank2 (γ2 z).
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w, fn = fn (x, y, z) 213
∂w ∂w ∂w
18. +a +b = c cotn (βx)w.
∂x ∂y ∂z
Z
n
General solution: w = Φ(y − ax, z − bx) exp c cot (βx) dx .
∂w ∂w ∂w
19. a +b + c cot(βz) = k cot(λx) + s cot(γy) w.
∂x ∂y ∂z
k/aλ
General solution: w = sin(λx) sin(γy)s/bγ Φ(u1 , u2 ), where

u1 = bx − ay, u2 = cβx + a lncos(βz).
∂w ∂w ∂w
20. + a cotn (βx) + b cotk (λx) = c cotm (γx)w.
∂x ∂y ∂z
h2 (x) = h1 (x) = 0, and h0 (x) = c cotm (γx).
∂w ∂w ∂w
21. a + b cot(βy) + c cot(λx) = k cot(γz)w.
∂x ∂y ∂z
General solution:
Z
k x cγ
w = Φ(u1 , u2 ) exp cot γz + ln |sin(λt)| − ln |sin(λx)| dt ,
a 0 aλ

where u1 = bβx + a lncos(βy) and u2 = aλz − c lnsin(λx).
∂w ∂w ∂w
22. a1 cotn1 (λ1 x) + b1 cotm1 (β1 y) + c1 cotk1 (γ1 z)
∂x ∂y ∂z

= a2 cot (λ2 x) + b2 cot (β2 y) + c2 cotk2 (γ2 z) w.
n2 m2
This is a special case of equation 2.3.7.10 in which f (x) = a1 cotn1 (λ1 x), g(y) =
b1 cotm1 (β1 y), h(z) = c1 cotk1 (γ1 z), ϕ(x) = a2 cotn2 (λ2 x), ψ(y) = b2 cotm2 (β2 y), and
χ(z) = c2 cotk2 (γ2 z).
∂w ∂w ∂w
23. + a sinn (λx) + b cosm (βx) = c sink (γx)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.1 with f (x) = a sinn (λx), g(x) = b cosm (βx),
∂w ∂w ∂w
24. + a cosn (λx) + b sinm (βy) = c cosk (γy) + s sinl (µz) w.
∂x ∂y ∂z
g2 (x, y) = b sinm (βy), and h(x, y, z) = c cosk (γy) + s sinl (µz).
∂w ∂w ∂w
25. + a cosn (λx) + b tanm (βy) = c cosk (γy) + s tanl (µz) w.
∂x ∂y ∂z
g2 (x, y) = b tanm (βy), and h(x, y, z) = c cosk (γy) + s tanl (µz).
∂w ∂w ∂w
26. a1 sinn1 (λ1 x) + b1 cosm1 (β1 y) + c1 cosk1 (γ1 z)
∂x ∂y ∂z

= a2 cos (λ2 x) + b2 sin (β2 y) + c2 cosk2 (γ2 z) w.
n2 m2
This is a special case of equation 2.3.7.10 with f (x) = a1 sinn1 (λ1 x), g(y) = b1 cosm1 (β1 y),
h(z) = c1 cosk1 (γ1 z), ϕ(x) = a2 cosn2 (λ2 x), ψ(y) = b2 sinm2 (β2 y), χ(z) = c2 cosk2 (γ2 z).
∂w ∂w ∂w
27. a1 tann1 (λ1 x) + b1 cotm1 (β1 y) + c1 cotk1 (γ1 z)
∂x ∂y ∂z

= a2 cotn2 (λ2 x) + b2 tanm2 (β2 y) + c2 cotk2 (γ2 z) w.
b1 cotm1 (β1 y), h(z) = c1 cotk1 (γ1 z), ϕ(x) = a2 cotn2 (λ2 x), ψ(y) = b2 tanm2 (β2 y), and
χ(z) = c2 cotk2 (γ2 z).

∂w ∂w ∂w
1. +a +b= c arcsinn (βx)w.
∂x ∂y ∂z
Z
n
General solution: w = Φ(y − ax, z − bx) exp c arcsin (βx) dx .
∂w ∂w ∂w
2. a1 + a2 + a3
∂x ∂y
∂z
= b1 arcsin(λ1 x) + b2 arcsin(λ2 y) + b3 arcsin(λ3 z) w.
b1 arcsin(λ1 x), ψ(y) = b2 arcsin(λ2 y), and χ(z) = b3 arcsin(λ3 z).
∂w ∂w ∂w
3. a +b + c arcsinn (λx) arcsink (βz) = s arcsinm (γx)w.
∂x ∂y ∂z
f4 (x) = s arcsinm (γx), g(y) = 1, and h(z) = arcsink (βz).
∂w ∂w ∂w
4. a +b + c arcsinn (λx) arcsinm (βy) arcsink (γz) = sw.
∂x ∂y ∂z
This is a special case of equation 2.3.7.19 with f (x, y) = c arcsinn (λx) arcsinm (βy),
g(z) = arcsink (γz), and h(x, y) = s.
∂w ∂w ∂w
5. a + b arcsinn (λx) + c arcsink (βz) = s arcsinm (γx)w.
∂x ∂y ∂z
f4 (x) = s arcsinm (γx), g(y) = 1, and h(z) = c arcsink (βz).
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w, fn = fn (x, y, z) 215
∂w ∂w ∂w
6. a + b arcsinn (λy) + c arcsink (βz) = sw.
∂x ∂y ∂z
This is a special case of equation 2.3.7.10 with f (x) = a, g(y) = b arcsinn (λy), h(z) =
c arcsink (βz), ϕ(x) = s, and ψ(y) = χ(z) = 0.
∂w ∂w ∂w
7. +a +b = c arccosn (βx)w.
∂x ∂y ∂z
Z
n
General solution: w = Φ(y − ax, z − bx) exp c arccos (βx) dx .
∂w ∂w ∂w
8. a1 + a2 + a3
∂x ∂y
∂z
= b1 arccos(λ1 x) + b2 arccos(λ2 y) + b3 arccos(λ3 z) w.
b1 arccos(λ1 x), ψ(y) = b2 arccos(λ2 y), and χ(z) = b3 arccos(λ3 z).
∂w ∂w ∂w
9. a +b + c arccosn (λx) arccosk (βz) = s arccosm (γx)w.
∂x ∂y ∂z
f4 (x) = s arccosm (γx), g(y) = 1, and h(z) = arccosk (βz).
∂w ∂w ∂w
10. a +b + c arccosn (λx) arccosm (βy) arccosk (γz) = sw.
∂x ∂y ∂z
This is a special case of equation 2.3.7.19 with f (x, y) = c arccosn (λx) arccosm (βy),
g(z) = arccosk (γz), and h(x, y) = s.
∂w ∂w ∂w
11. a + b arccosn (λx) + c arccosk (βz) = s arccosm (γx)w.
∂x ∂y ∂z
f4 (x) = s arccosm (γx), g(y) = 1, and h(z) = c arccosk (βz).
∂w ∂w ∂w
12. a + b arccosn (λy) + c arccosk (βz) = sw.
∂x ∂y ∂z
This is a special case of equation 2.3.7.10 with f (x) = a, g(y) = b arccosn (λy), h(z) =
c arccosk (βz), ϕ(x) = s, and ψ(y) = χ(z) = 0.
∂w ∂w ∂w
13. +a +b = c arctann (βx)w.
∂x ∂y ∂z
Z
n
General solution: w = Φ(y − ax, z − bx) exp c arctan (βx) dx .
∂w ∂w ∂w
14. a1 + a2 + a3
∂x ∂y
∂z
= b1 arctan(λ1 x) + b2 arctan(λ2 y) + b3 arctan(λ3 z) w.
b1 arctan(λ1 x), ψ(y) = b2 arctan(λ2 y), and χ(z) = b3 arctan(λ3 z).
∂w ∂w ∂w
15. a +b + c arctann (λx) arctank (βz) = s arctanm (γx)w.
∂x ∂y ∂z
f4 (x) = s arctanm (γx), g(y) = 1, and h(z) = arctank (βz).
∂w ∂w ∂w
16. a +b + c arctann (λx) arctanm (βy) arctank (γz) = sw.
∂x ∂y ∂z
This is a special case of equation 2.3.7.19 with f (x, y) = c arctann (λx) arctanm (βy),
g(z) = arctank (γz), and h(x, y) = s.
∂w ∂w ∂w
17. a + b arctann (λx) + c arctank (βz) = s arctanm (γx)w.
∂x ∂y ∂z
f4 (x) = s arctanm (γx), g(y) = 1, and h(z) = c arctank (βz).

∂w ∂w ∂w
18. +a +b = c arccot n (βx)w.
∂x ∂y ∂z
Z
n
General solution: w = Φ(y − ax, z − bx) exp c arccot (βx) dx .
∂w ∂w ∂w
19. a1 + a2 + a3
∂x ∂y ∂z

= b1 arccot(λ1 x) + b2 arccot(λ2 y) + b3 arccot(λ3 z) w.
b1 arccot(λ1 x), ψ(y) = b2 arccot(λ2 y), and χ(z) = b3 arccot(λ3 z).
∂w ∂w ∂w
20. a +b + c arccotn (λx) arccot k (βz) = s arccot m (γx)w.
∂x ∂y ∂z
f4 (x) = s arccotm (γx), g(y) = 1, and h(z) = arccotk (βz).
∂w ∂w ∂w
21. a +b + c arccotn (λx) arccot m (βy) arccot k (γz) = sw.
∂x ∂y ∂z
This is a special case of equation 2.3.7.19 with f (x, y) = c arccot n (λx) arccotm (βy),
g(z) = arccotk (γz), and h(x, y) = s.
∂w ∂w ∂w
22. a + b arccotn (λx) + c arccot k (βz) = s arccot m (γx)w.
∂x ∂y ∂z
f4 (x) = s arccotm (γx), g(y) = 1, and h(z) = c arccot k (βz).
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w, fn = fn (x, y, z) 217

◆ Throughout Section 2.3.7, sometimes only a particular solution w e of the equation and
a basis u1 , u2 of the corresponding “truncated” equation with a zero right-hand side are
presented. The general solution can be obtained as w = w e Φ(u1 , u2 ), where Φ(u1 , u2 ) is
an arbitrary function of two variables.
∂w ∂w ∂w
1. + f (x) + g(x) = h2 (x)y + h1 (x)z + h0 (x) w.
∂x ∂y ∂z
General solution:
Z Z
w = exp H2 (x)y + H1 (x)z + H0 (x) − f (x)H2 (x) dx − g(x)H1 (x) dx Φ(u1 , u2 ),
where
Z Z Z
Hk (x) = hk (x) dx (k = 0, 1, 2), u1 = y − f (x) dx, u2 = z − g(x) dx.
∂w ∂w ∂w
2. + f (x)(y + a) + g(x)(z + b) = h(x)w.
∂x ∂y ∂z
General solution:
Z
w = exp h(x) dx Φ(u1 , u2 ),
Z Z
u1 = ln |y + a| − f (x) dx, u2 = ln |z + b| − g(x) dx.
∂w ∂w ∂w
3. + [ay + f (x)] + [bz + g(x)] = h(x)w.
∂x ∂y ∂z
General solution:
Z
w = exp h(x) dx Φ(u1 , u2 ),
Z Z
u1 = ye−ax − f (x)e−ax dx, u2 = ze−bx − g(x)e−bx dx.
∂w ∂w ∂w
4. + f1 (x)y + f2 (x) + g1 (x)y + g2 (x)
∂x ∂y ∂z

= h2 (x)y + h1 (x)z + h0 (x) w.
Z

we = exp ϕ(x)y+ψ(x)z+ h0(x)−f2(x)ϕ(x)−g2 (x)ψ(x) dx ,
Z Z Z
h2(x)−g1(x)ψ(x)
ϕ(x) = F (x) dx, ψ(x) = h1(x) dx, F (x) = exp − f1(x) dx .
F (x)
For an integral basis u1 , u2 of the corresponding “truncated” equation, see 2.1.7.4.
∂w ∂w ∂w
5. + f1 (x)y + f2 (x) + g1 (x)z + g2 (x)
∂x ∂y ∂z

= h2 (x)y + h1 (x)z + h0 (x) w.
Z

w̄ = exp ϕ(x)y + ψ(x)z + h0 (x) − f2 (x)ϕ(x) − g2 (x)ψ(x) dx ,
Z Z
h2 (x)
ϕ(x) = F (x) dx, F (x) = exp − f1 (x) dx ,
F (x)
Z Z
h1 (x)
ψ(x) = G(x) dx, G(x) = exp − g1 (x) dx .
G(x)
For an integral basis u1 , u2 of the corresponding “truncated” equation, see 2.1.7.5.
∂w ∂w
6. + y 2 − a2 + aλ sinh(λx) − a2 sinh2 (λx)
∂x ∂y
∂w
+ f (x) sinh(γz) = g(x)w.
∂z
Z
General solution: w = exp g(x) dx Φ(u1 , u2 ), where
Z Z
1 γz E 2a
u1 = f (x) dx− lntanh , u2 = + E dx, E =exp sinh(λx) .
γ 2 y−a cosh(λx) λ
∂w ∂w ∂w
7. + f1 (x)y + f2 (x)y k + g1 (x)z + g2 (x)z m = h(x)w.
∂x ∂y ∂z
1◦ . For k 6= 1 and m 6= 1, the transformation
Z
1−k 1−m
ξ=y , η=z , W = w exp − h(x) dx

∂W ∂W ∂W
+ (1 − k) f1 (x)ξ + f2 (x) + (1 − m) g1 (x)η + g2 (x) = 0.
∂x ∂ξ ∂η
Z
◦ 1−k
2 . For k 6= 1 and m = 1, the transformation ξ = y , W = w exp − h(x) dx also
leads to an equation of the form 2.1.7.5.
3◦ . For k = m = 1, see equation 2.3.7.5 with h1 (x) = h2 (x) = 0.
∂w ∂w ∂w
8. + f1 (x)y + f2 (x)y k + g1 (x) + g2 (x)eλz = h(x)w.
∂x ∂y ∂z
The transformation
Z
1−k −λz
ξ=y , η=e , W = w exp − h(x) dx

∂W ∂W ∂W
+ (1 − k) f1 (x)ξ + f2 (x) − λ g1 (x)η + g2 (x) = 0.
∂x ∂ξ ∂η
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w, fn = fn (x, y, z) 219
∂w ∂w ∂w
9. + f1 (x) + f2 (x)eλy + g1 (x) + g2 (x)eβz = h(x)w.
∂x ∂y ∂z
The transformation
Z
ξ = e−λy , η = e−βz , W = w exp − h(x) dx

∂W ∂W ∂W
− λ f1 (x)ξ + f2 (x) − β g1 (x)η + g2 (x) = 0.
∂x ∂ξ ∂η

∂w ∂w ∂w
10. f (x) + g(y) + h(z) = ϕ(x) + ψ(y) + χ(z) w.
∂x ∂y ∂z
General solution:
Z Z Z
ϕ(x) ψ(y) χ(z)
w = exp dx + dy + dz Φ(u1 , u2 ),
f (x) g(y) h(z)
where Z Z Z Z
dx dy dx dz
u1 = − , u2 = − .
f (x) g(y) f (x) h(z)
∂w ∂w ∂w
11. f (x) +z + g(y) = h2 (x) + h1 (y) w.
∂x ∂y ∂z
General solution:
Z Z y
h2 (x) h1 (t) dt
w = exp dx + p Φ(u1 , u2 ),
f (x) y0 2G(t) − 2G(y) + z 2
where
Z Z Z y
z2 dx dt
G(y) = g(y) dy, u1 = G(y) − , u2 = − p .
2 f (x) y0 2G(t) − 2G(y) + z 2
∂w ∂w ∂w
12. f1 (x) + f2 (x)g(y) + f3 (x)h(z) = f4 (x)w.
∂x ∂y ∂z
General solution:
Z
f4 (x)
w = exp dx Φ(u1 , u2 ),
f1 (x)
Z Z Z Z
f2 (x) dy f3 (x) dz
u1 = dx − , u2 = dx − .
f1 (x) g(y) f1 (x) h(z)
∂w ∂w ∂w
13. + f1 (x)y + f2 (x) + g1 (x)z + g2 (y) = h1 (x) + h2 (y) w.
∂x ∂y ∂z
h(x, y, z) = h1 (x) + h2 (y).
∂w ∂w ∂w
14. + f1 (x)y + f2 (x)y k + g1 (y)z + g2 (x)z m = h1 (x) +
∂x ∂y ∂z
h2 (y) w.
h(x, y, z) = h1 (x) + h2 (y).
∂w ∂w ∂w
15. + f1 (x)y+f2 (x)y k + g1 (x)+g2 (y)eλz = h1 (x)+h2 (y) w.
∂x ∂y ∂z
h(x, y, z) = h1 (x) + h2 (y).
∂w ∂w ∂w
16. + f1 (x)+f2 (x)eλy + g1 (y)z+g2 (x)z k = h1 (x)+h2 (y) w.
∂x ∂y ∂z
h(x, y, z) = h1 (x) + h2 (y).
∂w ∂w ∂w
17. + f1 (x)+f2 (x)eλy + g1 (x)+g2(y)eβz = h1 (x)+h2 (y) w.
∂x ∂y ∂z
h(x, y, z) = h1 (x) + h2 (y).

∂w ∂w ∂w
18. a +b + f (x, y) = g(x, y)w.
∂x ∂y ∂z
Z y
1 a(t − y)
General solution: w = Φ(u1 , u2 ) exp g x+ , t dt , where
b y0 b
Z y
a(t − y)
u1 = bx − ay, u2 = bz − f x+ , t dt,
y0 b
and y0 may be taken as arbitrary.
∂w ∂w ∂w
19. a +b + f (x, y)g(z) = h(x, y)w.
∂x ∂y ∂z
Z y
1 a(t − y)
General solution: w = Φ(u1 , u2 ) exp h x+ , t dt , where
b y0 b
Z Z y
dz a(t − y)
u1 = bx − ay, u2 = b − f x+ , t dt.
g(z) y0 b
∂w ∂w ∂w
20. x +y + [z + f (x, y)] = g(x, y)w.
∂x ∂y ∂z
Z y
xt dt
General solution: w = Φ(u1 , u2 ) exp g ,t , where
y0 y t
Z y
y z xt dt
u1 = , u2 = − f ,t 2 .
x y y0 y t
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w, fn = fn (x, y, z) 221
∂w ∂w ∂w
21. ax + by
+ f (x, y)g(z) = h(x, y)w.
∂x ∂y ∂z
Z y
1 −1 −a/b a/b

General solution: w = Φ(u1 , u2 ) exp t h xy t , t dt , where
b y0
Z Z y
b −a dz
g(z) y0
∂w ∂w ∂w
22. + f1 (x)y + f2 (x) + g1 (x, y)z + g2 (x, y) = h(x, y, z)w.
∂x ∂y ∂z
General solution: Z x
w = Φ(u1 , u2 ) exp h̄(t, u1 , u2 ) dt ,
x0
where
Z Z
u1 = yF (x) − f2 (x)F (x) dx, F (x) = exp − f1 (x) dx , (1)
Z x Z x
u2 = zG(x, u1 ) − ḡ2 (t, u1 )G(t, u1 ) dt, G(x, u1 ) = exp − ḡ1 (t, u1 ) dt . (2)
x0 x0
Here ḡ1 (x, u1 ) ≡ g1 (x, y), ḡ2 (x, u1 ) ≡ g2 (x, y), h̄(x, u1 , u2 ) ≡ h(x, y, z) [in these func-
tions, y must be expressed via x and u1 from relation (1), and z must be expressed via
x, u1 , and u2 from relation (2)], and x0 is an arbitrary number.
∂w ∂w ∂w
23. + f1 (x)y+f2 (x)y k + g1 (x, y)z+g2 (x, y)z m = h(x, y, z)w.
∂x ∂y ∂z
1◦ . For k 6= 1 and m 6= 1, the transformation ξ = y 1−k , η = z 1−m leads to an equation of
the form 2.3.7.22:
∂w ∂w ∂w
+ (1 − k) f1 (x)ξ + f2 (x) + (1 − m) ḡ1 (x, ξ)η + ḡ2 (x, ξ) = h̄(x, ξ, η)w,
∂x ∂ξ ∂η
where
1 1 1 1
ḡ1 (x, ξ) ≡ g1 x, ξ 1−k , ḡ2 (x, ξ) ≡ g2 x, ξ 1−k , h̄(x, ξ, η) ≡ h x, ξ 1−k , η 1−m .
2.3.7.22.
∂w ∂w ∂w
24. + f1 (x)y +f2 (x)y k + g1 (x, y)+g2 (x, y)eλz = h(x, y, z)w.
∂x ∂y ∂z
The transformation ξ = η= y 1−k , e−λz
∂w ∂w ∂w
+ (1 − k) f1 (x)ξ + f2 (x) − λ ḡ1 (x, ξ)η + ḡ2 (x, ξ) = h̄(x, ξ, η)w,
∂x ∂ξ ∂η
where
1 1 1 1
ḡ1 (x, ξ) ≡ g1 x, ξ 1−k , ḡ2 (x, ξ) ≡ g2 x, ξ 1−k , h̄(x, ξ, η) ≡ h x, ξ 1−k , − ln η .
λ
∂w ∂w ∂w
25. + f1 (x)+f2 (x)eλy + g1 (x, y)z +g2 (x, y)z k = h(x, y, z)w.
∂x ∂y ∂z
∂w ∂w ∂w
− λ f1 (x)ξ + f2 (x) + (1 − k) ḡ1 (x, ξ)η + ḡ2 (x, ξ) = h̄(x, ξ, η)w,
∂x ∂ξ ∂η
1
where ḡ1,2 (x, ξ) ≡ g1,2 x, − λ1 ln ξ and h̄(x, ξ, η) ≡ h x, − λ1 ln ξ, η 1−k .
∂w ∂w ∂w
26. + f1 (x) + f2 (x)eλy + g1 (x, y) + g2(x, y)eβz = h(x, y, z)w.
∂x ∂y ∂z
∂w ∂w ∂w
− λ f1 (x)ξ + f2 (x) − β ḡ1 (x, ξ)η + ḡ2 (x, ξ) = h̄(x, ξ, η)w,
∂x ∂ξ ∂η

where ḡ1,2 (x, ξ) ≡ g1,2 x, − λ1 ln ξ and h̄(x, ξ, η) ≡ h x, − λ1 ln ξ, − β1 ln η .
∂w ∂w ∂w
27. f1(x)g1(y) +f2(x)g2(y) + h1(x, y)z+h2(x, y)z m =s(x, y, z)w.
∂x ∂y ∂z
Z Z
f2 (x) g1 (y)
f1 (x) g2 (y)
2.3.7.23 with f1 ≡ 0, f2 ≡ 1, k = 0:
∂w ∂w ∂w
+ + h̄1 (ξ, η)z + h̄2 (ξ, η)z m = h̄3 (ξ, η, z)w,
∂ξ ∂η ∂z
h1 (x, y) h2 (x, y) s(x, y, z)
f2 (x)g1 (y) f2 (x)g1 (y) f2 (x)g1 (y)
∂w ∂w ∂w
28. f1(x)g1(y) +f2(x)g2(y) + h1(x, y)+h2(x, y)eλz =s(x, y, z)w.
∂x ∂y ∂z
Z Z
f2 (x) g1 (y)
f1 (x) g2 (y)
2.3.7.24 with f1 ≡ 0, f2 ≡ 1, and k = 0:
∂w ∂w ∂w
+ + h̄1 (ξ, η) + h̄2 (ξ, η)eλz = h̄3 (x, y, z)w,
∂ξ ∂η ∂z
h1 (x, y) h2 (x, y) s(x, y, z)
f2 (x)g1 (y) f2 (x)g1 (y) f2 (x)g1 (y)

∂w ∂w ∂w
f1 + f2 + f3 = f4 w + f5 , fn = fn (x, y, z)
∂x ∂y ∂z
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 223

∂w ∂w ∂w
1. a +b +c = (αx + β)w + px + q.
∂x ∂y ∂z
General solution:
Z
1 x x
w = exp (αx+ 2β) Φ(bx− ay, cx− az)+ (px+ q) exp − (αx+ 2β) dx .
a 2a 2a
∂w ∂w ∂w
2. + az + by = (cx + k)w + px + q.
∂x ∂y ∂z
Z
2

General solution: w = exp 1
2 cx + kx Φ(u1 , u2 ) + (px + q) exp − 12 cx2 − kx dx ,
where
( √ √
2 2 (by + ab z) exp − ab x if ab > 0,
u1 = by − az , u2 = √ √ √
by cos( −ab x) + −ab z sin( −ab x) if ab < 0.
∂w ∂w ∂w
3. + (a1 x + a0 ) + (b1 x + b0 ) = (c1 x + c0 )w + s1 x + s0 .
∂x ∂y ∂z
This is a special case of equation 2.4.7.1 with f (x) = a1 x + a0 , g(x) = b1 x + b0 , h(x) =

c1 x + c0 , and p(x) = s1 x + s0 .
∂w ∂w ∂w
4. + (b1 x + b0 ) + (c1 y + c0 ) = aw + s1 x + s0 .
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b1 x + b0 , g(y) = c1 y + c0 , and

h(x) = s1 x + s0 .
∂w ∂w ∂w
5. + (ay + k1 x + k0 ) + (bz + n1 x + n0 ) = (c1 x + c0 )w + s1 x + s0 .
∂x ∂y ∂z
This is a special case of equation 2.4.7.3 with f (x) = k1 x + k0 , g(x) = n1 x + n0 , h(x) =

c1 x + c0 , and p(x) = s1 x + s0 .
∂w ∂w ∂w
6. + (a2 y + a1 x + a0 ) + (b3 z + b2 y + b1 x + b0 )
∂x ∂y ∂z
= (c3 z + c2 y + c1 x + c0 )w + s3 z + s2 y + s1 x + s0 .
This is a special case of equation 2.4.7.20 with f1 (x) = a2 , f2 (x) = a1 x+a0 , g1 (x, y) = b3 ,
g2 (x, y) = b2 y + b1 x + b0 , h1 (x, y, z) = c3 z + c2 y + c1 x + c0 , and h2 (x, y, z) = s3 z +
s2 y + s1 x + s0 .
∂w ∂w ∂w
7. ax + bx + cz = (αx + β)w + px + q.
∂x ∂y ∂z
Z
1 β/a αx/a −c a
−(a+β)/a −αx/a
General solution: w = x e Φ bx−ay, x z + (px+q)x e dx .
a
∂w ∂w ∂w
8. ax + by + cz
= (αx + β)w + px + q.
∂x ∂y ∂z
Z
1 β/a αx/a b −a c −a
−(a+β)/a −αx/a
General solution: w = x e Φ xy ,xz + (px+q)x e dx .
a
∂w ∂w ∂w
9. x + az + by = (cx + k)w + px + q.
∂x ∂y ∂z
Z
General solution: w = xk ecx Φ(u1 , u2 ) + (px + q)x −k−1 −cx
e dx , where
 √ √

x ab (by − ab z) √ if ab > 0,

u1 = by 2 − az 2 , u2 = √
−ab −ab z

x exp − arctan if ab < 0.
by

∂w ∂w ∂w
10. + (a1 x2 + a0 ) + (b1 x2 + b0 ) = (c1 x + c0 )w + s1 x2 + s0 .
∂x ∂y ∂z

h(x) = c1 x + c0 , and p(x) = s1 x2 + s0 .
∂w ∂w ∂w
11. + (b1 x2 + b0 ) + (c1 y 2 + c0 ) = aw + s1 x2 + s0 .
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b1 x2 + b0 , g(y) = c1 y 2 + c0 , and

h(x) = s1 x2 + s0 .
∂w ∂w ∂w
12. +(ay +k1 x2 +k0 ) +(bz +n1 x2 +n0 ) = (c1 x+c0 )w +s1 x+s0 .
∂x ∂y ∂z
This is a special case of equation 2.4.7.3 with f (x) = k1 x2 + k0 , g(x) = n1 x2 + n0 ,
h(x) = c1 x + c0 , and p(x) = s1 x + s0 .
∂w ∂w ∂w
13. + (a2 xy + a1 x + a0 ) + (b3 yz + b2 y 2 + b1 x2 + b0 )
∂x ∂y ∂z
= (c3 z + c2 y + c1 x + c0 )w + s1 xy + s2 xz.
This is a special case of equation 2.4.7.20 with f1 (x) = a2 x, f2 (x) = a1 x + a0 , g1 (x, y) =
b3 y, g2 (x, y) = b2 y 2 + b1 x2 + b0 , h1 (x, y, z) = c3 z + c2 y + c1 x + c0 , and h2 (x, y, z) =
s1 xy + s2 xz.
∂w ∂w ∂w
14. ax + bx + cz = kxw + sx2 .
∂x ∂y ∂z
s
General solution: w = ekx/a Φ bx − ay, xc z −a − 2 (kx + a).
k
∂w ∂w ∂w
15. ax + by + cz = kxw + sx2 .
∂x ∂y ∂z
s
General solution: w = ekx/a Φ xb y −a , xc z −a − 2 (kx + a).
k
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 225
∂w ∂w ∂w
16. ax2 + by 2 + cz 2 = (kx + s)w + px + q.
∂x ∂y ∂z
General solution:
Z
k/a −sx/a b a c a 1 −(2a+k)/a sx/a
w=x e Φ − , − + (px + q)x e dx .
x y x z a

∂w √ ∂w √ ∂w √
17. + (a1 x + a0 ) + (b1 x + b0 ) = cw + s1 x + s0 .
∂x ∂y ∂z
√ √
This is a special case of equation 2.4.7.1 with f (x) = a1 x + a0 , g(x) = b1 x + b0 ,
√
h(x) = c, and p(x) = s1 x + s0 .
∂w ∂w ∂w
18. + (b1 x2 + b0 ) + (c1 y 3 + c0 ) = aw + s1 x3 + s0 .
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b1 x2 + b0 , g(y) = c1 y 3 + c0 , and
h(x) = s1 x3 + s0 .
∂w ∂w ∂w
19. + (ay + kx3 ) + (bz + nx3 ) = cw + sx2 .
∂x ∂y ∂z
This is a special case of equation 2.4.7.3 with f (x) = kx3 , g(x) = nx3 , h(x) = c, and
p(x) = sx2 .
∂w ∂w ∂w
20. + (a1 xy + a2 x3 ) + (b1 yz + b2 y 3 ) = (c1 z + c2 y)w + s1 x2 y +
∂x ∂y ∂z
s2 xz 2 .
This is a special case of equation 2.4.7.20 with f1 (x) = a1 x, f2 (x) = a2 x3 , g1 (x, y) = b1 y,
g2 (x, y) = b2 y 3 , h1 (x, y, z) = c1 z + c2 y, and h2 (x, y, z) = s1 x2 y + s2 xz 2 .
∂w ∂w ∂w
21. ax3 + by 3+ cz 3 = xw + kx + s.
∂x ∂y ∂z

1 b a c a s
General solution: w = exp − Φ 2 − 2 , 2 − 2 − + as − k.
ax x y x z x

∂w ∂w ∂w
22. a +b +c = kxn w + sxm .
∂x ∂y ∂z
1◦ . General solution for n 6= −1:
Z
k n+1 s m k n+1
w = exp x Φ(bx − ay, cx − az) + x exp − x dx .
a(n + 1) a a(n + 1)
2◦ . General solution for n = −1:


xk/a Φ(bx − ay, cx − az) + sxm+1
if a(m + 1) 6= k,
w= a(m + 1) − k
xk/a Φ(bx − ay, cx − az) + s xk/a ln x

if a(m + 1) = k.
a
∂w ∂w ∂w
23. a + by + cz = kxn w + sxm .
∂x ∂y ∂z
Z
a −bx a −cx
s xm k n+1
w = E(x)Φ y e ,z e + E(x) dx, E(x) = exp x .
a E(x) a(n + 1)



xk/a Φ y a e−bx , z a e−cx + sxm+1
if a(m + 1) 6= k,
w= a(m + 1) − k

xk/a Φ y a e−bx , z a e−cx + s xk/a ln x

if a(m + 1) = k.
a
∂w ∂w ∂w
24. + az + by = cxn w + sxm .
∂x ∂y ∂z
Z
c n+1 m c n+1
w = exp x Φ(u1 , u2 ) + s x exp − x dx ,
n+1 n+1
where
( √ √
u1 = by − az , u2 = p p p


xc Φ(u1 , u2 ) + s
xm+1 if m + 1 6= c,
w= m+1−c
xc Φ(u1 , u2 ) + sxc ln x if m + 1 = c,
where u1 and u2 are defined in Item 1◦ .
∂w ∂w ∂w
25. + axn + bxm = cxk w + sxl .
∂x ∂y ∂z
This is a special case of equation 2.4.7.1 with f (x) = axn , g(x) = bxm , h(x) = cxk , and
p(x) = sxl .
∂w ∂w ∂w
26. + bxn + cy m = aw + sxk .
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = bxn , g(y) = cy m , and h(x) = sxk .
∂w ∂w ∂w
27. + (ay + βxn ) + (bz + γxm ) = cxk w + sxl .
∂x ∂y ∂z
This is a special case of equation 2.4.7.3 with f (x) = βxn , g(x) = γxm , h(x) = cxk , and
p(x) = sxl .
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 227
∂w ∂w ∂w
28. + (a1 xn1 y + a2 xn2 ) + (b1 y m1 z + b2 y m2 ) = cw + s1 xy k1 +
∂x ∂y ∂z
s2 xk2 z.
This is a special case of equation 2.4.7.20 with f1 (x) = a1 xn1 , f2 (x) = a2 xn2 , g1 (x, y) =
b1 y m1 , g2 (x, y) = b2 y m2 , h1 (x, y, z) = c, and h2 (x, y, z) = s1 xy k1 + s2 xk2 z.
∂w ∂w ∂w
29. + a1 xλ1 y+a2 xλ2 y k + b1 xβ1 z+b2 xβ2 z m = c1 xγ1 w+c2 y γ2 .
∂x ∂y ∂z
This is a special case of equation 2.4.7.21 with f1 (x) = a1 xλ1 , f2 (x) = a2 xλ2 , g1 (x, y) =
b1 xβ1 , g2 (x, y) = b2 xβ2 , h1 (x, y, z) = c1 xγ1 , and h2 (x, y, z) = c2 y γ2 .
∂w ∂w ∂w
30. + a1 xλ1 y+a2 xλ2 y k + b1 y β1 z+b2 y β2 z m = c1 xγ1 w+c2 z γ2 .
∂x ∂y ∂z
This is a special case of equation 2.4.7.21 with f1 (x) = a1 xλ1 , f2 (x) = a2 xλ2 , g1 (x, y) =
b1 y β1 , g2 (x, y) = b2 y β2 , h1 (x, y, z) = c1 xγ1 , and h2 (x, y, z) = c2 z γ2 .
∂w ∂w ∂w
31. x + ay + bz = cxn w + kxm .
∂x ∂y ∂z
a Z
c n x xb m−1 c n
General solution: w = exp x Φ , +k x exp − x dx .
n y z n
∂w ∂w ∂w
32. x + az + by = cxn w + kxm .
∂x ∂y ∂z
1◦ . General solution for n 6= 0:
Z
c n m−1 c n
w = exp x Φ(u1 , u2 ) + k x exp − x dx ,
n n
where
 √ √

√
u1 = by 2 − az 2 , u2 = √
−ab −ab z

by
2◦ . General solution for n = 0:


xc Φ(u , u ) + k xm if m 6= c,
1 2
w= m−c
xc Φ(u , u ) + kxc ln x if m = c,
1 2

∂w ∂w ∂w
33. bcx + c(by + cz) + b(by − cz) = kxn w + sxm .
∂x ∂y ∂z
1◦ . General solution for n 6= 0:
Z
k n s m−1 k n
w = exp x Φ(u1 , u2 ) + x exp − x dx ,
bcn bc bcn
where
√ √ √ √
u1 = by + ( 2 − 1)cz |x|− 2 , u2 = by − ( 2 + 1)cz |x| 2 .
2◦ . General solution for n = 0:

 k
x bc Φ(u1 , u2 ) + s
xm if bcm 6= k,
w= bcm − k
xm Φ(u , u ) + s xm ln x if bcm = k,
1 2
bc
∂w ∂w ∂w
34. b1 xn1 + b2 y n2 + b3 z n3 = aw + c1 xk1 + c2 y k2 + c3 xk3 .
∂x ∂y ∂z
This is a special case of equation 2.4.7.16 with f1 (x) = b1 xn1 , f2 (y) = b2 y n2 , f3 (z) = b3 z n3 ,
g1 (x) = c1 xk1 , g2 (y) = c2 y k2 , and g3 (z) = c3 z k3 .
∂w ∂w ∂w
35. a 1 x n1 + a 2 y n2 + a 3 z n3 = bxk w + cxm .
∂x ∂y ∂z
This is a special case of equation 2.4.7.10 with f1 (x) = a1 xn1 , f2 (x) = f3 (x) = 1, f4 (x) =
bxk , f5 (x) = cxm , g(y) = a2 y n2 , and h(z) = a3 z n3 .

∂w ∂w ∂w
1. +a +b = ceβx w + keλx .
∂x ∂y ∂z
1◦ . General solution for β 6= 0:

Z
c βx c βx
w = exp e Φ(y − ax, z − bx) + k exp λx − e dx .
β β
2◦ . General solution for β = 0:
k
w = ecx Φ(y − ax, z − bx) + eλx .
λ−c
∂w ∂w ∂w
2. + aeβx + beλx = ceγx w + seµx .
∂x ∂y ∂z
This is a special case of equation 2.4.7.1 with f (x) = aeβx , g(x) = beλx , h(x) = ceγx , and
p(x) = seµx .
∂w ∂w ∂w
3. + beβx + ceλy = aw + seγx .
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = beβx , g(y) = ceλy , and h(x) = seγx .
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 229
∂w ∂w ∂w
4. + aeβx + beλz = cw + keγx .
∂x ∂y ∂z
General solution: 
ecx Φ(u , u ) + k eγx if c 6= γ,
1 2
w= γ −c
 γx
e Φ(u1 , u2 ) + kx if c = γ,
where u1 = aeβx − βy and u2 = bλx + e−λz .
∂w ∂w ∂w
5. + a1 eσx + a2 eλy + b1 eµy + b2 eβz = c1 w + c2 eνx .
∂x ∂y ∂z
This is a special case of equation 2.4.7.24 with f1 (x) = a1 eσx , f2 (x) = a2 , g1 (x, y) = b1 eµy ,
g2 (x, y) = b2 , h1 (x, y, z) = c1 , and h2 (x, y, z) = c2 eνx .
∂w ∂w ∂w
6. b1 eλ1 x + b2 eλ2 y + b3 eλ3 z = aw + c1 eβ1 x + c2 eβ2 y + c3 eβ3 z .
∂x ∂y ∂z
This is a special case of equation 2.4.7.16 with f1 (x) = b1 eλ1 x , f2 (y) = b2 eλ2 y , f3 (z) =
b3 eλ3 z , g1 (x) = c1 eβ1 x , g2 (y) = c2 eβ2 y , and g3 (z) = c3 eβ3 z .
∂w ∂w
7. a1 eσ1 x+β1 y + a2 eσ2 x+β2 y
∂x ∂y
∂w
+ b1 eν1 x+µ1 y + b2 eν2 x+µ2 y+λz = c1 w + c2 .
∂z
This is a special case of equation 2.4.7.27 with f1 (x) = a1 eσ1 x , g1 (y) = eβ1 y , f2 (x) =
a2 eσ2 x , g2 (y) = eβ2 y , h1 (x, y) = b1 eν1 x+µ1 y , h2 (x, y) = b2 eν2 x+µ2 y , ϕ1 (x, y, z) = c1 , and
ϕ2 (x, y, z) = c2 .
∂w ∂w ∂w
8. +a +b= ceβx w + kxn .
∂x ∂y ∂z
Z
c βx n c βx
General solution: w = exp e Φ(y − ax, z − bx) + k x exp − e dx .
β β
∂w ∂w ∂w
9. + axn + beλx = ceγx w + sxk .
∂x ∂y ∂z
This is a special case of equation 2.4.7.1 with f (x) = axn , g(x) = beλx , h(x) = ceγx , and
p(x) = sxk .
∂w ∂w ∂w
10. + beβx + cy n = aw + seγx .
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = beβx , g(y) = cy n , and h(x) = seγx .
∂w ∂w ∂w
11. + a1 y + a2 xy k + b1 x + b2 eβy+λz = c1 w + c2 eγx .
∂x ∂y ∂z
This is a special case of equation 2.4.7.22 with f1 (x) = a1 , f2 (x) = a2 x, g1 (x, y) = b1 x,
g2 (x, y) = b2 eβy , h1 (x, y, z) = c1 , and h2 (x, y, z) = c2 eγx .
∂w ∂w ∂w
12. + a1 x + a2 eλy + b1 z + b2 eβy z k = c1 w + c2 .
∂x ∂y ∂z
This is a special case of equation 2.4.7.23 with f1 (x) = a1 x, f2 (x) = a2 , g1 (x, y) = b1 ,
g2 (x, y) = b2 eβy , h1 (x, y, z) = c1 , and h2 (x, y, z) = c2 .
∂w ∂w ∂w
13. + a1 eµx + a2 eλy + b1 eνy + b2 eβz = c1 w + c2 .
∂x ∂y ∂z
This is a special case of equation 2.4.7.24 with f1 (x) = a1 eµx , f2 (x) = a2 , g1 (x, y) = b1 eνy ,
g2 (x, y) = b2 , h1 (x, y, z) = c1 , and h2 (x, y, z) = c2 .
∂w ∂w ∂w
14. + a1 eλ1 x y+a2 eλ2 x + b1 eβ1 x z+b2 eβ2 x = c1 eγ1 x w+c2 eγ2 x .
∂x ∂y ∂z
This is a special case of equation 2.4.7.20 with f1 (x) = a1 eλ1 x , f2 (x) = a2 eλ2 x , g1 (x, y) =
b1 eβ1 x , g2 (x, y) = b2 eβ2 x , h1 (x, y, z) = c1 eγ1 x , and h2 (x, y, z) = c2 eγ2 x .
∂w ∂w ∂w
15. + a1 eλ1 x y + a2 eλ2 x y k + b1 eβ1 x z + b2 eβ2 x z m
∂x ∂y ∂z
= c1 eγ1 x w + c2 eγ2 y .
b1 eβ1 x , g2 (x, y) = b2 eβ2 x , h1 (x, y, z) = c1 eγ1 x , and h2 (x, y, z) = c2 eγ2 y .
∂w ∂w ∂w
16. + a1 eλ1 x y + a2 eλ2 x y k + b1 eβ1 y z + b2 eβ2 y z m
∂x ∂y ∂z
= c1 eγ1 x w + c2 eγ2 z .
b1 eβ1 y , g2 (x, y) = b2 eβ2 y , h1 (x, y, z) = c1 eγ1 x , and h2 (x, y, z) = c2 eγ2 z .
∂w ∂w ∂w
17. a1 eβy + a2 eσx + b1 xn eµy + b2 y m eνx+λz = c1 w + c2 .
∂x ∂y ∂z
This is a special case of equation 2.4.7.27 with f1 (x) = 1, g1 (y) = a1 eβy , f2 (x) = a2 eσx ,
g2 (y) = 1, h1 (x, y) = b1 xn eµy , h2 (x, y) = b2 y m eνx , ϕ1 (x, y, z) = c1 , and ϕ2 (x, y, z) = c2 .

∂w ∂w ∂w
1. +a +b = c sinhn (βx)w + k sinhm (λx).
∂x ∂y ∂z
General solution:
Z Z
dx m n
w=E(x)Φ(y−ax, z−bx)+kE(x) sinh (λx) , E(x)=exp c sinh (βx) dx .
E(x)
∂w ∂w ∂w
2. a +b + c sinh(βz) = p sinh(λx) + q w + k sinh(γx).
∂x ∂y ∂z
General solution:
Z
k p cosh(λx)+qλx p cosh(λx)+qλx
w= exp Φ(u1, u2)+ sinh(γx) exp − dx ,
a aλ aλ
βz

where u1 = bx − ay and u2 = cβx − a lntanh .
2
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 231
∂w ∂w ∂w
3. + a sinhn (βx) + b sinhk (λx) = cw + s sinhm (µx).
∂x ∂y ∂z
h(x) = c, and p(x) = s sinhm (µx).
∂w ∂w ∂w
4. + b sinhn (βx) + c sinhk (λy) = aw + s sinhm (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b sinhn (βx), g(y) = c sinhk (λy), and
h(x) = s sinhm (µx).
∂w ∂w ∂w
5. b1 sinhn1 (λ1 x) + b2 sinhn2 (λ2 y) + b3 sinhn3 (λ3 z)
∂x ∂y ∂z
= aw + c1 sinhk1 (β1 x) + c2 sinhk2 (β2 y) + c3 sinhk3 (β3 z).
This is a special case of equation 2.4.7.16 in which f1 (x) = b1 sinhn1 (λ1 x), f2 (y) =
b2 sinhn2 (λ2 y), f3 (z) = b3 sinhn3 (λ3 z), g1 (x) = c1 sinhk1 (β1 x), g2 (y) = c2 sinhk2 (β2 y),
and g3 (z) = c3 sinhk3 (β3 z).
∂w ∂w ∂w
6. +a +b = c coshn (βx)w + k coshm (λx).
∂x ∂y ∂z
General solution:
Z Z
m dx n
w=E(x)Φ(y−ax, z−bx)+kE(x) cosh (λx) , E(x)=exp c cosh (βx) dx .
E(x)
∂w ∂w ∂w
7. a +b + c cosh(βz) = p cosh(λx) + q w + k cosh(γx).
∂x ∂y ∂z
General solution:
Z
k p sinh(λx)+qλx p sinh(λx)+qλx
w= exp Φ(u1, u2)+ cosh(γx) exp − dx ,
a aλ aλ
βz

where u1 = bx − ay and u2 = cβx − 2a arctantanh .
2
∂w ∂w ∂w
8. + a coshn (βx) + b coshk (λx) = cw + s coshm (µx).
∂x ∂y ∂z
h(x) = c, and p(x) = s coshm (µx).
∂w ∂w ∂w
9. + b coshn (βx) + c coshk (λy) = aw + s coshm (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b coshn (βx), g(y) = c coshk (λy),
and h(x) = s coshm (µx).
∂w ∂w ∂w
10. b1 coshn1 (λ1 x) + b2 coshn2 (λ2 y) + b3 coshn3 (λ3 z)
∂x ∂y ∂z
= aw + c1 coshk1 (β1 x) + c2 coshk2 (β2 y) + c3 coshk3 (β3 z).
This is a special case of equation 2.4.7.16 in which f1 (x) = b1 coshn1 (λ1 x), f2 (y) =
b2 coshn2 (λ2 y), f3 (z) = b3 coshn3 (λ3 z), g1 (x) = c1 coshk1 (β1 x), g2 (y) = c2 coshk2 (β2 y),
and g3 (z) = c3 coshk3 (β3 z).
∂w ∂w ∂w
11. +a +b = c tanhn (βx)w + k tanhm (λx).
∂x ∂y ∂z
General solution:
Z
dx
w = E(x)Φ(y − ax, z − bx) + kE(x) tanhm (λx) ,
E(x)
Z
n
E(x) = exp c tanh (βx) dx .
∂w ∂w ∂w
12. a +b + c tanh(βz) = p tanh(λx) + q w + k tanh(γx).
∂x ∂y ∂z
General solution:
Z
k qx/a p/aλ −qx/a −p/aλ
w= e cosh (λx) Φ(u1 , u2 ) + e cosh (λx) tanh(γx) dx ,
a

where u1 = bx − ay and u2 = cβx − a lnsinh(βz).
∂w ∂w ∂w
13. + a tanhn (βx) + b tanhk (λx) = cw + s tanhm (µx).
∂x ∂y ∂z
h(x) = c, and p(x) = s tanhm (µx).
∂w ∂w ∂w
14. + b tanhn (βx) + c tanhk (λy) = aw + s tanhm (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b tanhn (βx), g(y) = c tanhk (λy), and
h(x) = s tanhm (µx).
∂w ∂w ∂w
15. b1 tanhn1 (λ1 x) + b2 tanhn2 (λ2 y) + b3 tanhn3 (λ3 z)
∂x ∂y ∂z
= aw + c1 tanhk1 (β1 x) + c2 tanhk2 (β2 y) + c3 tanhk3 (β3 z).
This is a special case of equation 2.4.7.16 in which f1 (x) = b1 tanhn1 (λ1 x), f2 (y) =
b2 tanhn2 (λ2 y), f3 (z) = b3 tanhn3 (λ3 z), g1 (x) = c1 tanhk1 (β1 x), g2 (y) = c2 tanhk2 (β2 y),
and g3 (z) = c3 tanhk3 (β3 z).
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 233

∂w ∂w ∂w
16. +a +b = c cothn (βx)w + k cothm (λx).
∂x ∂y ∂z
General solution:
Z
dx
w = E(x)Φ(y − ax, z − bx) + kE(x) cothm (λx) ,
E(x)
Z
n
E(x) = exp c coth (βx) dx .
∂w ∂w ∂w
17. a +b + c coth(βz) = p coth(λx) + q w + k coth(γx).
∂x ∂y ∂z
General solution:
Z
k qx/a p/aλ −qx/a −p/aλ
w= e sinh (λx) Φ(u1 , u2 ) + e sinh (λx) coth(γx) dx ,
a

where u1 = bx − ay and u2 = cβx − a ln cosh(βz) .
∂w ∂w ∂w
18. + a cothn (βx) + b cothk (λx) = cw + s cothm (µx).
∂x ∂y ∂z
h(x) = c, and p(x) = s cothm (µx).
∂w ∂w ∂w
19. + b cothn (βx) + c cothk (λy) = aw + s cothm (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b cothn (βx), g(y) = c cothk (λy), and
h(x) = s cothm (µx).
∂w ∂w ∂w
20. b1 cothn1 (λ1 x) + b2 cothn2 (λ2 y) + b3 cothn3 (λ3 z)
∂x ∂y ∂z
= aw + c1 cothk1 (β1 x) + c2 cothk2 (β2 y) + c3 cothk3 (β3 z).
This is a special case of equation 2.4.7.16 in which f1 (x) = b1 cothn1 (λ1 x), f2 (y) =
b2 cothn2 (λ2 y), f3 (z) = b3 cothn3 (λ3 z), g1 (x) = c1 cothk1 (β1 x), g2 (y) = c2 cothk2 (β2 y),
and g3 (z) = c3 cothk3 (β3 z).

∂w ∂w ∂w
21. +a +b = c sinhn (βx)w + k coshm (λx).
∂x ∂y ∂z
General solution:
Z
dx
w = E(x)Φ(y − ax, z − bx) + kE(x) coshm (λx) ,
E(x)
Z
n
E(x) = exp c sinh (βx) dx .
∂w ∂w ∂w
22. +a +b = c tanhn (βx)w + k cothm (λx).
∂x ∂y ∂z
General solution:
Z
dx
w = E(x)Φ(y − ax, z − bx) + kE(x) cothm (λx) ,
E(x)
Z
n
E(x) = exp c tanh (βx) dx .
∂w ∂w ∂w
23. + b coshn (βx) + c sinhk (λy) = aw + s coshm (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b coshn (βx), g(y) = c sinhk (λy),
and h(x) = s coshm (µx).
∂w ∂w ∂w
24. + a tanhn (βx) + b cothk (λx) = cw + s tanhm (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.1 with f (x) = a tanhn (βx), g(x) = b cothk (λx),
h(x) = c, and p(x) = s tanhm (µx).
∂w ∂w ∂w
25. b1 sinhn1 (λ1 x) + b2 coshn2 (λ2 y) + b3 sinhn3 (λ3 z)
∂x ∂y ∂z
= aw + c1 coshk1 (β1 x) + c2 sinhk2 (β2 y) + c3 sinhk3 (β3 z).
This is a special case of equation 2.4.7.16 in which f1 (x) = b1 sinhn1 (λ1 x), f2 (y) =
b2 coshn2 (λ2 y), f3 (z) = b3 sinhn3 (λ3 z), g1 (x) = c1 coshk1 (β1 x), g2 (y) = c2 sinhk2 (β2 y),
and g3 (z) = c3 sinhk3 (β3 z).

∂w ∂w ∂w
1. +a +b = c lnn (βx)w + k lnm (λx).
∂x ∂y ∂z
General solution:
Z
dx
w = E(x)Φ(y − ax, z − bx) + kE(x) lnm (λx) ,
E(x)
Z
n
E(x) = exp c ln (βx) dx .
∂w ∂w ∂w
2. + a lnn (βx) + b lnk (λx) = cw + s lnm (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.1 with f (x) = a lnn (βx), g(x) = b lnk (λx), h(x) = c,
and p(x) = s lnm (µx).
∂w ∂w ∂w
3. + b lnn (βx) + c lnk (λy) = aw + s lnm (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b lnn (βx), g(y) = c lnk (λy), and
h(x) = s lnm (µx).
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 235
∂w ∂w ∂w
4. a ln(αx) + b ln(βy) + c ln(γz) = pw + q ln(λx).
∂x ∂y ∂z
General solution:
Z Z Z
q p dx ln(λx) p dx
w = exp Φ(u1 , u2 ) + exp − dx ,
a a ln(αx) ln(αx) a ln(αx)
where
Z Z Z Z
dx dy dx dz
u1 = b −a , u2 = c −a .
ln(αx) ln(βy) ln(αx) ln(γz)
∂w ∂w ∂w
5. b1 lnn1 (λ1 x) + b2 lnn2 (λ2 y) + b3 lnn3 (λ3 z)
∂x ∂y ∂z
= aw + c1 lnk1 (β1 x) + c2 lnk2 (β2 y) + c3 lnk3 (β3 z).
This is a special case of equation 2.4.7.16 with f1 (x) = b1 lnn1 (λ1 x), f2 (y) = b2 lnn2 (λ2 y),
f3 (z) = b3 lnn3 (λ3 z), g1 (x) = c1 lnk1 (β1 x), g2 (y) = c2 lnk2 (β2 y), and g3 (z) = c3 lnk3 (β3 z).

∂w ∂w ∂w
6. +a +b = c lnn (βx)w + kxm .
∂x ∂y ∂z
General solution:
Z Z
xm n
w = E(x)Φ(y − ax, z − bx) + kE(x) dx, E(x) = exp c ln (βx) dx .
E(x)
∂w ∂w ∂w
7. + axn + b lnk (λx) = cw + sxm .
∂x ∂y ∂z
This is a special case of equation 2.4.7.1 with f (x) = axn , g(x) = b lnk (λx), h(x) = c,
and p(x) = sxm .
∂w ∂w ∂w
8. + b lnn (βx) + cy k = aw + s lnm (λx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b lnn (βx), g(y) = cy k , and h(x) =
s lnm (λx).
∂w ∂w ∂w
9. b1 lnn1 (λ1 x) + b2 lnn2 (λ2 y) + b3 lnn3 (λ3 z)
∂x ∂y ∂z
= aw + c1 xk1 + c2 y k2 + c3 z k3 .
This is a special case of equation 2.4.7.16 with f1 (x) = b1 lnn1 (λ1 x), f2 (y) = b2 lnn2 (λ2 y),
f3 (z) = b3 lnn3 (λ3 z), g1 (x) = c1 xk1 , g2 (y) = c2 y k2 , and g3 (z) = c3 z k3 .
∂w ∂w ∂w
10. ax(ln x)n + by(ln y)m + cz(ln z)l = k(ln x)s w + p ln(νx).
∂x ∂y ∂z
Introduce the notation
u1 = − , u2 = − .
a(n − 1) b(m − 1) a(n − 1) c(l − 1)
1◦ . General solution for s + 1 6= n:

Z
k(ln x)s−n+1 p ln(νx) k(ln x)s−n+1
w = exp Φ(u1 , u2 ) + exp − dx .
a(s − n + 1) a x(ln x)n a(s − n + 1)
2◦ . General solution for s + 1 = n:

Z
k p ln(νx) k−an
w = (ln x) a Φ(u1 , u2 ) + (ln x) a dx .
a x

∂w ∂w ∂w
1. +a +b = c sinn (βx)w + k sinm (λx).
∂x ∂y ∂z
General solution:
Z Z
dx
w = E(x)Φ(y−ax, z−bx)+kE(x) sinm (λx) , E(x) = exp c sinn (βx) dx .
E(x)
∂w ∂w ∂w
2. + a sinn (βx) + b sink (λx) = cw + s sinm (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.1 with f (x) = a sinn (βx), g(x) = b sink (λx),
h(x) = c, and p(x) = s sinm (µx).
∂w ∂w ∂w
3. + b sinn (βx) + c sink (λy) = aw + s sinm (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b sinn (βx), g(y) = c sink (λy), and
h(x) = s sinm (µx).
∂w ∂w ∂w
4. a + b sin(βy) + c sin(γx) = p sin(µx)w + q sin(λx).
∂x ∂y ∂z
General solution:
Z
q p p
w = exp − cos(µx) Φ(u1 , u2 ) + sin(λx) exp cos(µx) dx ,
a aµ aµ
βy

where u1 = bβx − a lntan and u2 = aγz + c cos(γx).
2
∂w ∂w ∂w
5. a + b sin(βy) + c sin(γz) = p sin(µx)w + q sin(λx).
∂x ∂y ∂z
General solution:
Z
q p p
w = exp − cos(µx) Φ(u1 , u2 ) + sin(λx) exp cos(µx) dx ,
a aµ aµ
βy γz

where u1 = bβx − a lntan and u2 = cγx − a lntan .
2 2
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 237
∂w ∂w ∂w
6. b1 sinn1 (λ1 x) + b2 sinn2 (λ2 y) + b3 sinn3 (λ3 z)
∂x ∂y ∂z
= aw + c1 sink1 (β1 x) + c2 sink2 (β2 y) + c3 sink3 (β3 z).
This is a special case of equation 2.4.7.16 in which f1 (x) = b1 sinn1 (λ1 x), f2 (y) =
b2 sinn2 (λ2 y), f3 (z) = b3 sinn3 (λ3 z), g1 (x) = c1 sink1 (β1 x), g2 (y) = c2 sink2 (β2 y), and
g3 (z) = c3 sink3 (β3 z).

∂w ∂w ∂w
7. +a +b = c cosn (βx)w + k cosm (λx).
∂x ∂y ∂z
General solution:
Z Z
m dx n
w = E(x)Φ(y−ax, z−bx)+kE(x) cos (λx) , E(x) = exp c cos (βx) dx .
E(x)
∂w ∂w ∂w
8. + a cosn (βx) + b cosk (λx) = cw + s cosm (µx).
∂x ∂y ∂z
h(x) = c, and p(x) = s cosm (µx).
∂w ∂w ∂w
9. + b cosn (βx) + c cosk (λy) = aw + s cosm (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b cosn (βx), g(y) = c cosk (λy), and
h(x) = s cosm (µx).
∂w ∂w ∂w
10. a + b cos(βy) + c cos(γz) = p cos(µx)w + q cos(λx).
∂x ∂y ∂z
General solution:
Z
q p p
w = exp sin(µx) Φ(u1 , u2 ) + cos(λx) exp − sin(µx) dx ,
a aµ aµ

where u1 = bβx − a lnsec(βy) + tan(βy) and u2 = cγx − a lnsec(γz) + tan(γz).
∂w ∂w ∂w
11. a + b cos(βy) + c cos(γx) = p cos(µx)w + q cos(λx).
∂x ∂y ∂z
General solution:
Z
q p p
w = exp sin(µx) Φ(u1 , u2 ) + cos(λx) exp − sin(µx) dx ,
a aµ aµ

where u1 = bβx − a lnsec(βy) + tan(βy) and u2 = aγz − c sin(γx).
∂w ∂w ∂w
12. b1 cosn1 (λ1 x) + b2 cosn2 (λ2 y) + b3 cosn3 (λ3 z)
∂x ∂y ∂z
= aw + c1 cosk1 (β1 x) + c2 cosk2 (β2 y) + c3 cosk3 (β3 z).
This is a special case of equation 2.4.7.16 in which f1 (x) = b1 cosn1 (λ1 x), f2 (y) =
b2 cosn2 (λ2 y), f3 (z) = b3 cosn3 (λ3 z), g1 (x) = c1 cosk1 (β1 x), g2 (y) = c2 cosk2 (β2 y), and
g3 (z) = c3 cosk3 (β3 z).
∂w ∂w ∂w
13. +a +b = c tann (βx)w + k tanm (λx).
∂x ∂y ∂z
General solution:
Z Z
dx
w = E(x)Φ(y−ax, z−bx)+kE(x) tanm (λx) , E(x) = exp c tann (βx) dx .
E(x)
∂w ∂w ∂w
14. + a tann (βx) + b tank (λx) = cw + s tanm (µx).
∂x ∂y ∂z
h(x) = c, and p(x) = s tanm (µx).
∂w ∂w ∂w
15. + b tann (βx) + c tank (λy) = aw + s tanm (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b tann (βx), g(y) = c tank (λy), and
h(x) = s tanm (µx).
∂w ∂w ∂w
16. a + b tan(βy) + c tan(γz) = p tan(µx)w + q tan(λx).
∂x ∂y ∂z
Z
−p/aµ q p/aµ

General solution: w = cos(µx) Φ(u1 , u2 ) + cos(µx) tan(λx) dx ,
a
where u1 = bβx − a lnsin(βy) and u2 = cγx − a lnsin(γz).
∂w ∂w ∂w
17. a + b tan(βy) + c tan(γx) = p tan(µx)w + q tan(λx).
∂x ∂y ∂z
Z
−p/aµ q p/aµ

General solution: w = cos(µx) Φ(u1 , u2 ) + cos(µx) tan(λx) dx ,
a
where u1 = bβx − a lnsin(βy) and u2 = aγz + c lncos(γx).
∂w ∂w ∂w
18. b1 tann1 (λ1 x) + b2 tann2 (λ2 y) + b3 tann3 (λ3 z)
∂x ∂y ∂z
= aw + c1 tank1 (β1 x) + c2 tank2 (β2 y) + c3 tank3 (β3 z).
This is a special case of equation 2.4.7.16 in which f1 (x) = b1 tann1 (λ1 x), f2 (y) =
b2 tann2 (λ2 y), f3 (z) = b3 tann3 (λ3 z), g1 (x) = c1 tank1 (β1 x), g2 (y) = c2 tank2 (β2 y),
and g3 (z) = c3 tank3 (β3 z).
∂w ∂w ∂w
19. +a +b = c cotn (βx)w + k cotm (λx).
∂x ∂y ∂z
General solution:
Z Z
m dx n
w = E(x)Φ(y−ax, z−bx)+kE(x) cot (λx) , E(x) = exp c cot (βx) dx .
E(x)
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 239
∂w ∂w ∂w
20. + a cotn (βx) + b cotk (λx) = cw + s cotm (µx).
∂x ∂y ∂z
h(x) = c, and p(x) = s cotm (µx).
∂w ∂w ∂w
21. + b cotn (βx) + c cotk (λy) = aw + s cotm (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b cotn (βx), g(y) = c cotk (λy), and
h(x) = s cotm (µx).
∂w ∂w ∂w
22. a + b cot(βy) + c cot(γz) = p cot(µx)w + q cot(λx).
∂x ∂y ∂z
Z
p/aµ q −p/aµ

General solution: w= sin(µx) Φ(u1, u2)+ sin(µx) cot(λx) dx , where
a
u1 = bβx + a lncos(βy) and u2 = cγx + a lncos(γz).
∂w ∂w ∂w
23. a + b cot(βy) + c cot(γx) = p cot(µx)w + q cot(λx).
∂x ∂y ∂z
Z
p/aµ q −p/aµ

General solution: w= sin(µx) Φ(u1, u2)+ sin(µx) cot(λx) dx , where
a
u1 = bβx + a lncos(βy) and u2 = aγz − c lnsin(γx).
∂w ∂w ∂w
24. b1 cotn1 (λ1 x) + b2 cotn2 (λ2 y) + b3 cotn3 (λ3 z)
∂x ∂y ∂z
= aw + c1 cotk1 (β1 x) + c2 cotk2 (β2 y) + c3 cotk3 (β3 z).
This is a special case of equation 2.4.7.16 in which f1 (x) = b1 cotn1 (λ1 x), f2 (y) =
b2 cotn2 (λ2 y), f3 (z) = b3 cotn3 (λ3 z), g1 (x) = c1 cotk1 (β1 x), g2 (y) = c2 cotk2 (β2 y), and
g3 (z) = c3 cotk3 (β3 z).

∂w ∂w ∂w
25. +a +b = c sinn (βx)w + k cosm (λx).
∂x ∂y ∂z
General solution:
Z Z
m dx n
w = E(x)Φ(y−ax, z−bx)+kE(x) cos (λx) , E(x) = exp c sin (βx) dx .
E(x)
∂w ∂w ∂w
26. +a +b = c tann (βx)w + k cotm (λx).
∂x ∂y ∂z
General solution:
Z Z
m dx n
w = E(x)Φ(y−ax, z−bx)+kE(x) cot (λx) , E(x) = exp c tan (βx) dx .
E(x)
∂w ∂w ∂w
27. + b cosn (βx) + c sink (λy) = aw + s cosm (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b cosn (βx), g(y) = c sink (λy), and
h(x) = s cosm (µx).
∂w ∂w ∂w
28. + a tann (βx) + b cotk (λx) = cw + s tanm (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.1 with f (x) = a tann (βx), g(x) = b cotk (λx),
h(x) = c, and p(x) = s tanm (µx).
∂w ∂w ∂w
29. b1 sinn1 (λ1 x) + b2 cosn2 (λ2 y) + b3 sinn3 (λ3 z)
∂x ∂y ∂z
= aw + c1 cosk1 (β1 x) + c2 sink2 (β2 y) + c3 sink3 (β3 z).
This is a special case of equation 2.4.7.16 in which f1 (x) = b1 sinn1 (λ1 x), f2 (y) =
b2 cosn2 (λ2 y), f3 (z) = b3 sinn3 (λ3 z), g1 (x) = c1 cosk1 (β1 x), g2 (y) = c2 sink2 (β2 y), and
g3 (z) = c3 sink3 (β3 z).

∂w ∂w ∂w
1. +a +b = c arcsinn (βx)w + k arcsinm (λx).
∂x ∂y ∂z
General solution:
Z
dx
w = E(x)Φ(y − ax, z − bx) + kE(x) arcsinm (λx) ,
E(x)
Z
n
E(x) = exp c arcsin (βx) dx .
∂w ∂w ∂w
2. + a arcsink (λx) + b arcsinm (βx) = cw + s arcsinn (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.1 with f (x) = a arcsink (λx), g(x) = b arcsinm (βx),
h(x) = c, and p(x) = s arcsinn (µx).
∂w ∂w ∂w
3. + b arcsink (λx) + c arcsinm (βy) = aw + s arcsinn (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b arcsink (λx), g(y) = c arcsinm (βy),
and h(x) = s arcsinn (µx).
∂w ∂w ∂w
4. + a arcsink (λx) + b arcsinm (βz) = cw + s arcsinn (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.10 with f1 (x) = 1, f2 (x) = a arcsink (λx), f3 (x) = 1,
f4 (x) = c, f5 (x) = s arcsinn (µx), g(y) = 1, and h(z) = b arcsinm (βz).
∂w ∂w ∂w
5. + a arcsink (λy) + b arcsinm (βx) = cw + s arcsinn (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.10 with f1 (x) = f2 (x) = 1, f3 (x) = b arcsinm (βx),
f4 (x) = c, f5 (x) = s arcsinn (µx), g(y) = a arcsink (λy), and h(z) = 1.
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 241
∂w ∂w ∂w
6. +a +b = c arccosn (βx)w + k arccosm (λx).
∂x ∂y ∂z
General solution:
Z
dx
w = E(x)Φ(y − ax, z − bx) + kE(x) arccosm (λx) ,
E(x)
Z
n
E(x) = exp c arccos (βx) dx .
∂w ∂w ∂w
7. + a arccosk (λx) + b arccosm (βx) = cw + s arccosn (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.1 with f (x) = a arccosk (λx), g(x) = b arccosm (βx),
h(x) = c, and p(x) = s arccosn (µx).
∂w ∂w ∂w
8. + b arccosk (λx) + c arccosm (βy) = aw + s arccosn (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b arccosk (λx), g(y) = c arccos m (βy),
and h(x) = s arccosn (µx).
∂w ∂w ∂w
9. + a arccosk (λx) + b arccosm (βz) = cw + s arccosn (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.10 with f1 (x) = 1, f2 (x) = a arccosk (λx), f3 (x) = 1,
f4 (x) = c, f5 (x) = s arccosn (µx), g(y) = 1, and h(z) = b arccosm (βz).
∂w ∂w ∂w
10. + a arccosk (λy) + b arccosm (βx) = cw + s arccosn (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.10 with f1 (x) = f2 (x) = 1, f3 (x) = b arccosm (βx),
f4 (x) = c, f5 (x) = s arccosn (µx), g(y) = a arccosk (λy), and h(z) = 1.
∂w ∂w ∂w
11. +a +b = c arctann (βx)w + k arctanm (λx).
∂x ∂y ∂z
General solution:
Z
dx
w = E(x)Φ(y − ax, z − bx) + kE(x) arctanm (λx) ,
E(x)
Z
n
E(x) = exp c arctan (βx) dx .
∂w ∂w ∂w
12. + a arctank (λx) + b arctanm (βx) = cw + s arctann (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.1 with f (x) = a arctank (λx), g(x) = b arctanm (βx),
h(x) = c, and p(x) = s arctann (µx).
∂w ∂w ∂w
13. + b arctank (λx) + c arctanm (βy) = aw + s arctann (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b arctank (λx), g(y) = c arctanm (βy),
and h(x) = s arctann (µx).
∂w ∂w ∂w
14. + a arctank (λx) + b arctanm (βz) = cw + s arctann (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.10 with f1 (x) = 1, f2 (x) = a arctank (λx), f3 (x) = 1,
f4 (x) = c, f5 (x) = s arctann (µx), g(y) = 1, and h(z) = b arctanm (βz).
∂w ∂w ∂w
15. + a arctank (λy) + b arctanm (βx) = cw + s arctann (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.10 with f1 (x) = f2 (x) = 1, f3 (x) = b arctanm (βx),
f4 (x) = c, f5 (x) = s arctann (µx), g(y) = a arctank (λy), and h(z) = 1.
∂w ∂w ∂w
16. +a +b = c arccot n (βx)w + k arccot m (λx).
∂x ∂y ∂z
General solution:
Z
dx
w = E(x)Φ(y − ax, z − bx) + kE(x) arccotm (λx) ,
E(x)
Z
n
E(x) = exp c arccot (βx) dx .
∂w ∂w ∂w
17. + a arccotk (λx) + b arccot m (βx) = cw + s arccot n (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.1 with f (x) = a arccotk (λx), g(x) = b arccotm (βx),
h(x) = c, and p(x) = s arccotn (µx).
∂w ∂w ∂w
18. + b arccotk (λx) + c arccot m (βy) = aw + s arccotn (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.8 with f (x) = b arccotk (λx), g(y) = c arccotm (βy),
and h(x) = s arccotn (µx).
∂w ∂w ∂w
19. + a arccotk (λx) + b arccot m (βz) = cw + s arccotn (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.10 with f1 (x) = 1, f2 (x) = a arccot k (λx), f3 (x) = 1,
f4 (x) = c, f5 (x) = s arccotn (µx), g(y) = 1, and h(z) = b arccotm (βz).
∂w ∂w ∂w
20. + a arccotk (λy) + b arccot m (βx) = cw + s arccotn (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.10 with f1 (x) = f2 (x) = 1, f3 (x) = b arccotm (βx),
f4 (x) = c, f5 (x) = s arccotn (µx), g(y) = a arccot k (λy), and h(z) = 1.
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 243

∂w ∂w ∂w
1. + f (x) + g(x) = h(x)w + p(x).
∂x ∂y ∂z
Z Z Z
General solution: w = exp h(x) dx Φ(u1 , u2 ) + p(x) exp − h(x) dx dx ,
Z Z
where u1 = y − f (x) dx and u2 = z − g(x) dx.
∂w ∂w ∂w
2. + (y + a)f (x) + (z + b)g(x) = h(x)w + p(x).
∂x ∂y ∂z
Z Z Z
Z Z
where u1 = ln |y + a| − f (x) dx and u2 = ln |z + b| − g(x) dx.
∂w ∂w ∂w
3. + [ay + f (x)] + [bz + g(x)] = h(x)w + p(x).
∂x ∂y ∂z
Z Z
Z
Z Z
−ax −ax −bx
where u1 = ye − f (x)e dx and u2 = ze − g(x)e−bx dx.
∂w ∂w ∂w
4. + f (x)y k + g(x)z m = h(x)w + p(x).
∂x ∂y ∂z
Z Z
Z
Z Z
1 1−k 1 1−m
where u1 = y − f (x) dx and u2 = z − g(x) dx.
1−k 1−m
∂w ∂w ∂w
5. + f1 (x)y + f2 (x)y k + g1 (x)z + g2 (x)z m = h1 (x)w + h2 (x).
∂x ∂y ∂z
This is a special case of equation 2.4.7.21 in which g1 (x, y) = g1 (x), g2 (x, y) = g2 (x),
h1 (x, y, z) = h1 (x), and h2 (x, y, z) = h2 (x).
∂w ∂w ∂w
6. + f1 (x) + f2 (x)eλy + g1 (x) + g2 (x)eβz = h1 (x)w + h2 (x).
∂x ∂y ∂z
This is a special case of equation 2.4.7.24 in which g1 (x, y) = g1 (x), g2 (x, y) = g2 (x),
h1 (x, y, z) = h1 (x), and h2 (x, y, z) = h2 (x).
∂w ∂w ∂w
7. + y 2 − a2 + aλ sinh(λx) − a2 sinh2 (λx) + f (x) sinh(γz)
∂x ∂y ∂z
= g(x)w + h(x).
Z Z Z
General solution: w = exp g(x) dx Φ(u1 , u2 ) + h(x) exp − g(x) dx dx ,
where
Z
1 γz
u1 = f (x) dx − lntanh ,
γ 2
Z
E 2a

∂w ∂w ∂w
8. + f (x) + g(y) = aw + h(x).
∂x ∂y ∂z
Z
ax −ax
General solution: w = e Φ(u, v) + e h(x) dx , where
Z x Z

u = y − F (x), v=z− g u + F (t) dt, F (x) = f (x) dx,
x0

∂w ∂w ∂w
9. + f (x)y k + g(y)z m = aw + h(x).
∂x ∂y ∂z
The transformation
( (
1 1−k 1 1−m
Y (y) = 1−k y if k 6= 1,
Z(z) = 1−m z if m 6= 1,
ln |y| if k = 1, ln |z| if m = 1,

∂w ∂w ∂w
+ f (x) + ḡ(Y ) = aw + h(x), ḡ(Y ) ≡ g(y).
∂x ∂Y ∂Z
∂w ∂w ∂w
10. f1 (x) + f2 (x)g(y) + f3 (x)h(z) = f4 (x)w + f5 (x).
∂x ∂y ∂z
General solution:
Z Z
f5 (x) dx f4 (x)
w = F (x) Φ(u1 , u2 ) + , F (x) = exp dx ,
f1 (x) F (x) f1 (x)
where Z Z Z Z
f2 (x) dy f3 (x) dz
u1 = dx − , u2 = dx − .
f1 (x) g(y) f1 (x) h(z)
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 245
∂w ∂w ∂w
11. + f1 (x)y + f2 (x) + g1 (x)z + g2 (y) = h1 (x)w + h2 (y).
∂x ∂y ∂z
This is a special case of equation 2.4.7.20 in which g1 (x, y) = g1 (x), g2 (x, y) = g2 (y),
h1 (x, y, z) = h1 (x), and h2 (x, y, z) = h2 (y).
∂w ∂w ∂w
12. + f1 (x)y +f2(x)y k + g1 (y)z +g2 (x)z m = h1 (y)w +h2 (z).
∂x ∂y ∂z
This is a special case of equation 2.4.7.21 in which g1 (x, y) = g1 (y), g2 (x, y) = g2 (x),
h1 (x, y, z) = h1 (y), and h2 (x, y, z) = h2 (z).
∂w ∂w ∂w
13. + f1 (x)y + f2(x)y k + g1 (x) + g2(y)eλz = h1 (x)w + h2(y).
∂x ∂y ∂z
h1 (x, y, z) = h1 (x), h2 (x, y, z) = h2 (y).
∂w ∂w ∂w
14. + f1 (x) + f2 (x)eλy + g1 (y)z + g2 (x)z k = h1 (z)w + h2 (y).
∂x ∂y ∂z
This is a special case of equation 2.4.7.23 in which g1 (x, y) = g1 (y), g2 (x, y) = g2 (x),
h1 (x, y, z) = h1 (z), and h2 (x, y, z) = h2 (y).
∂w ∂w ∂w
15. + f1 (x) + f2(x)eλy + g1 (x) + g2(y)eβz = h1 (x)w + h2(y).
∂x ∂y ∂z
h1 (x, y, z) = h1 (x), and h2 (x, y, z) = h2 (y).
∂w ∂w ∂w
16. f1 (x) + f2 (y) + f3 (z) = aw + g1 (x) + g2 (y) + g3 (z).
∂x ∂y ∂z
General solution:
Z Z Z
g1(x) dx g2(y) dy g3(z) dz
w=E1(x)Φ(u1, u2)+E1(x) +E2(y) +E3(z) ,
f1(x)E1(x) f2(y)E2(y) f3(z)E3(z)
where
Z Z Z
dx dy dz
E1 (x) = exp a , E2 (y) = exp a , E3 (z) = exp a ,
f1 (x) f2 (y) f3 (z)
Z Z Z Z
dx dy dx dz
u1 = − , u2 = − .
f1 (x) f2 (y) f1 (x) f3 (z)
∂w ∂w ∂w
17. +a +b = F (x, y, z)w + G(x, y, z).
∂x ∂y ∂z
This is a special case of equation 2.4.7.20 with f1 (x) ≡ 0, f2 (x) = a, g1 (x, y) ≡ 0, and
g2 (x, y) = b.
∂w ∂w ∂w
18. x +y +z = aw + f (x, y, z).
∂x ∂y ∂z
General solution:
Z x
y z dt y z
w = xa Φ , + f (t, u1 t, u2 t) a+1 , u1 = , u2 = ,
x x x0 t x x
∂w ∂w ∂w
19. ax + by + cz = F (x, y, z)w + G(x, y, z).
∂x ∂y ∂z
The substitution x = eaξ leads to an equation of the form 2.4.7.20:
∂w ∂w ∂w
+ by + cz = F (eaξ , y, z)w + G(eaξ , y, z).
∂ξ ∂y ∂z
∂w ∂w ∂w
20. + f1 (x)y + f2 (x) + g1 (x, y)z + g2 (x, y)
∂x ∂y ∂z
= h1 (x, y, z)w + h2 (x, y, z).
General solution:
Z x Z x
h̄2 (t, u, v)
w = H(x, u, v) Φ(u, v) + dt , H(x, u, v) = exp h̄1 (t, u, v) dt ,
x0 H(t, u, v) x0
where
Z Z
u = yF (x) − f2 (x)F (x) dx, F (x) = exp − f1 (x) dx , (1)
Z x Z x
v = zG(x, u) − ḡ2 (t, u)G(t, u) dt, G(x, u) = exp − ḡ1 (t, u) dt . (2)
x0 x0
Here ḡn (x, u) ≡ gn (x, y), h̄n (x, u, v) ≡ hn (x, y, z) [in these functions, y must be expressed
via x and u from relation (1), and z must be expressed via x, u, and v from relation (2)],
and x0 is an arbitrary number.
∂w ∂w ∂w
21. + f1 (x)y + f2 (x)y k + g1 (x, y)z + g2 (x, y)z m
∂x ∂y ∂z
= h1 (x, y, z)w + h2 (x, y, z).
1◦ . For k 6= 1 and m 6= 1, the transformation ξ = y 1−k ,
η = z 1−m leads to an equation of
the form 2.4.7.20:
∂w ∂w ∂w
+ (1 − k) f1 (x)ξ + f2 (x) + (1 − m) ḡ1 (x, ξ)η + ḡ2 (x, ξ)
∂x ∂ξ ∂η
= h̄1 (x, ξ, η)w + h̄2 (x, ξ, η),
1 1 1
where ḡ1,2 (x, ξ) ≡ g1,2 x, ξ 1−k and h̄1,2 (x, ξ, η) ≡ h1,2 x, ξ 1−k , η 1−m .
2.4.7.20.
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 247
∂w ∂w ∂w
22. + f1 (x)y + f2 (x)y k + g1 (x, y) + g2 (x, y)eλz
∂x ∂y ∂z
= h1 (x, y, z)w + h2 (x, y, z).
The transformation ξ = y 1−k , η = e−λz leads to an equation of the form 2.4.7.20:
∂w ∂w ∂w
+(1−k) f1(x)ξ+f2(x) −λ ḡ1(x, ξ)η+ḡ2(x, ξ) = h̄1(x, ξ, η)w+h̄2 (x, ξ, η),
∂x ∂ξ ∂η
1 1
where ḡ1,2 (x, ξ) ≡ g1,2 x, ξ 1−k and h̄1,2 (x, ξ, η) ≡ h1,2 x, ξ 1−k , − λ1 ln η .
∂w ∂w ∂w
23. + f1 (x) + f2 (x)eλy + g1 (x, y)z + g2 (x, y)z k
∂x ∂y ∂z
= h1 (x, y, z)w + h2 (x, y, z).
The transformation ξ = e−λy , η = z 1−k leads to an equation of the form 2.4.7.20:
∂w ∂w ∂w
−λ f1(x)ξ+f2(x) +(1−k) ḡ1(x, ξ)η+ḡ2(x, ξ) = h̄1(x, ξ, η)w+h̄2 (x, ξ, η),
∂x ∂ξ ∂η
1
where ḡ1,2 (x, ξ) ≡ g1,2 x, − λ1 ln ξ and h̄1,2 (x, ξ, η) ≡ h1,2 x, − λ1 ln ξ, η 1−k .
∂w ∂w ∂w
24. + f1 (x) + f2 (x)eλy + g1 (x, y) + g2 (x, y)eβz
∂x ∂y ∂z
= h1 (x, y, z)w + h2 (x, y, z).
The transformation ξ = e−λy , η= e−βz leads to an equation of the form 2.4.7.20:
∂w ∂w ∂w
− λ f1 (x)ξ + f2 (x) − β ḡ1 (x, ξ)η + ḡ2 (x, ξ) = h̄1 (x, ξ, η)w + h̄2 (x, ξ, η),
∂x ∂ξ ∂η

where ḡ1,2 (x, ξ) ≡ g1,2 x, − λ1 ln ξ and h̄1,2 (x, ξ, η) ≡ h1,2 x, − λ1 ln ξ, − β1 ln η .
∂w ∂w ∂w
25. f0 (x) + f1 (x)y + f2 (x)y k + g1 (x, y)z + g2 (x, y)z m
∂x ∂y ∂z
= h1 (x, y, z)w + h2 (x, y, z).
On dividing the equation by f0 (x), we obtain an equation of the form 2.4.7.21.
∂w ∂w ∂w
26. f1 (x)g1(y) + f2 (x)g2 (y) + h1 (x, y)z + h2 (x, y)z m
∂x ∂y ∂z
= ϕ1 (x, y, z)w + ϕ2 (x, y, z).
Z Z
f2 (x) g1 (y)
f1 (x) g2 (y)
2.4.7.21 with f1 ≡ 0, f2 ≡ 1, and k = 0:
∂w ∂w ∂w
+ + h̄1 (ξ, η)z + h̄2 (ξ, η)z m = ϕ̄1 (ξ, η, z)w + ϕ̄2 (ξ, η, z),
∂ξ ∂η ∂z
hn (x, y) ϕn (x, y, z)
where h̄n (ξ, η) ≡ and ϕ̄n (ξ, η, z) ≡ ; n = 1, 2.
f2 (x)g1 (y) f2 (x)g1 (y)
∂w ∂w ∂w
27. f1 (x)g1(y) + f2 (x)g2 (y) + h1 (x, y) + h2 (x, y)eλz
∂x ∂y ∂z
= ϕ1 (x, y, z)w + ϕ2 (x, y, z).
Z Z
f2 (x) g1 (y)
f1 (x) g2 (y)
2.4.7.22 with f1 ≡ 0, f2 ≡ 1, and k = 0:
∂w ∂w ∂w
+ + h̄1 (ξ, η) + h̄2 (ξ, η)eλz = ϕ̄1 (ξ, η, z)w + ϕ̄2 (ξ, η, z),
∂ξ ∂η ∂z
hn (x, y) ϕn (x, y, z)
where h̄n (ξ, η) ≡ and ϕ̄n (ξ, η, z) ≡ ; n = 1, 2.
f2 (x)g1 (y) f2 (x)g1 (y)
∂w ∂w ∂w
28. + f1 (x, y, z) + f2 (x, y, z) = g(x)w + h(x).
∂x ∂y ∂z
General solution:
Z Z
h(x)
w = G(x)Φ(u1 , u2 ) + G(x) dx, G(x) = exp g(x) dx ,
G(x)
where u1 , u2 is an integral basis of the corresponding “truncated” homogeneous equation

∂u ∂u ∂u
with g(x) = h(x) = 0: + f1 (x, y, z) + f2 (x, y, z) = 0.
∂x ∂y ∂z
2.4.8 Underdetermined Equations Containing Operator div
1. div u = 0.
Let u = (u, v, w) be a vector field, where u, v, and w are its components depending on
rectangular Cartesian coordinates x, y, z. Then the homogeneous equation in question has
the form
∂u ∂v ∂w
div u ≡ + + = 0. (1)
∂x ∂y ∂z
This is an underdetermined equation, which contains three unknown functions u =
u(x, y, z), v = v(x, y, z), and w = w(x, y, z). A similar equation often occurs in various
systems of coupled equations and is called the continuity equation. Here we study this
equation separately.
Equation (1) is invariant with respect to the transformation
u = ū + ξ1 (y, z), v = v̄ + ξ2 (x, z), w = w̄ + ξ3 (x, y),
where ξ1 (y, z), ξ2 (x, z), and ξ3 (x, y) are arbitrary functions, which will be called the cali-
bration functions.
Consider various representations of the general solution of Eq. (1).
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 249
1◦ . Any two out of three functions u, v, and w can be assumed to be arbitrary, and the
remaining function is determined by a simple single integration of Eq. (1).
This method permits one to obtain the general solution in the form
u = u(x, y, z) is an arbitrary function,

v = v(x, y, z) is an arbitrary function,
Z
w = − (ux + vy ) dz + ξ3 (x, y).
2◦ . Suppose that ψ (1) = ψ (1) (x, y, z) and ψ (1) = ψ (2) (x, y, z) are two arbitrary twice
continuously differentiable functions.
2.1. The general solution of Eq. (1) can be represented, for example, in the simple form
u = (u, v, w), u = ψy(1) , v = −ψx(1) + ψz(2) , w = −ψy(2) . (2)
Remark 2.1. In hydrodynamics, ψ (1) and ψ (2) can be treated as two stream functions, which
allow one to reduce the original three-dimensional equations of motion of an incompressible fluid
with velocity components u, v, and w to equations for ψ (1) and ψ (2) (Polyanin & Zaitsev, 2012,
p. 1248). In the special case of ψ (2) ≡ 0 in (2), we have a usual representation of the fluid veloc-
ity components for two-dimensional flows in the plane (x, y) with w = 0 in terms of one stream
function.
Remark 2.2. Formulas (2) are invariant under the following transformation of the stream func-
tions:
ψ (1) = ψ̄ (1) + θz (x, z), ψ (2) = ψ̄ (2) + θx (x, z),
where θ(x, z) is an arbitrary (calibration) function.
2.2. The general solution of Eq. (1) can also be represented in the form
(1) (1) (2) (2)
u = a1 ψy − a3 ψz + b1 ψy − b3 ψz ,
(1) (1) (2) (2)
v = a2 ψz − a1 ψx + b2 ψz − b1 ψx , (3)
(1) (1) (2) (2)
w = a3 ψx − a2 ψy + b3 ψx − b2 ψy ,
where ai and bi are arbitrary constants such that a21 + a22 + a23 6= 0 and b21 + b22 + b23 6= 0.
Formulas (3) contain excessively many arbitrary constants. (Some of them can be set to 1,
0, or −1.) This gives us some freedom in representing the results. For example, by setting
a2 = a3 = b1 = b3 = 0 and a1 = b2 = 1 in (3), we obtain the simple representation (2).
3◦ . The general solution of Eq. (1) can be represented in the vector form
u = curl Ψ = (Ψ3y − Ψ2z , Ψ1z − Ψ3x , Ψ2x − Ψ1y ), (4)
where Ψ = (Ψ1 , Ψ2 , Ψ3 ) is an arbitrary vector function.

Remark 2.3. The representation of the solution in the form (2) corresponds to the special choice
of the components of the vector Ψ in the form
Ψ1 = ψ (2) , Ψ2 = 0, Ψ3 = ψ (1) ,
and the representation of the solution in the form (3) corresponds to the special case in which
Ψ1 = a2 ψ (1) + b2 ψ (2) , Ψ2 = a3 ψ (1) + b3 ψ (2) , Ψ3 = a1 ψ (1) + b1 ψ (2) .

Remark 2.4. The representation of the solution in the form (4) corresponds to the choice of the
stream functions ψ (1) and ψ (2) in (2) in the form
Z y Z y
∂ ∂
ψ (1) = Ψ3 − Ψ2 (x, ȳ, z) dȳ, ψ (2) = Ψ1 − Ψ2 (x, ȳ, z) dȳ, (5)
∂z y0 ∂x y0
where y0 is an arbitrary constant. The representation (5) is unique under the conditions ψ (1) |y=y0 =
Ψ3 |y=y0 and ψ (2) |y=y0 = Ψ1 |y=y0 .
2. div u = f (x).
The nonhomogeneous equation in the Cartesian coordinates has the form
∂u ∂v ∂w
+ + = f (x), (1)
∂x ∂y ∂z
where u = (u, v, w) and x = (x, y, z).
Particular solutions of equation (1) can be found by various methods.
1◦ . The simplest particular solutions are obtained if one sets any two out of the three func-
tions u, v, and w to be zero. For example, this method gives a particular solution of the
form Z
up = 0, vp = 0, wp = f (x) dz, (2)
R Rz
where f (x) dz ≡ z0 f (x, y, z̄) dz̄. From now on, the subscript p indicates that particular
solutions are considered.
2◦ . One often seeks a particular solution in the form of the gradient of a scalar function,
up = ∇ϕ; i.e., up = ϕx , vp = ϕy , wp = ϕz .
As a result, for ϕ one obtains the Poisson equation
∆ϕ = f (x), (3)
where ∆ is the Laplace operator. For solutions of Eq. (3), see Section 10.2.
The general solution of the nonhomogeneous equation (1) can be represented as the
sum of the general solution of the homogeneous equation (see Eq. 2.4.8.1) and a partic-
ular solution of the nonhomogeneous equation. For example, the general solution of the
nonhomogeneous equation (1) can be represented in the form
Z
(1) (1) (2) (2)
u = ψy , v = −ψx + ψz , w = −ψy + f (x) dz.
3. div u + f(x) · u + g(x) = 0.

In the Cartesian coordinates x, y, z, the equation becomes
∂u ∂v ∂w
+ + + f1 (x)u + f2 (x)v + f3 (x)w + g(x) = 0,
∂x ∂y ∂z
where u = (u, v, w), f = (f1 , f2 , f3 ), and x = (x, y, z).
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 251
This underdetermined equation is easy to integrate. (One arbitrarily specifies any two
out of three components of the vector u, and for the remaining component we have a linear
first-order ODE in which two spatial variables occur as parameters.) For example, if we ar-
bitrarily specify the second and third components of the vector u, then the general solution
can be represented in the form
Z
u = E A − F E −1 dx , v = v(x), w = w(x),
Z
F = vy + wz + f2 v + f3 w + g, E = exp − f1 dx ,
where v(x), w(x), and A = A(y, z) are arbitrary functions.
2.4.9 Equations with Four or More Independent Variables

◮ Equations containing power-law functions.
∂w ∂w ∂w ∂w
1. +a +b +c = 0.
∂x1 ∂x2 ∂x3 ∂x4
Integral basis: u1 = ax1 − x2 , u2 = bx1 − x3 , u3 = cx1 − x4 .
∂w ∂w ∂w ∂w
2. + ax1 + bx1 + cx1 = 0.
∂x1 ∂x2 ∂x3 ∂x4
Integral basis: u1 = x2 − 12 ax21 , u2 = x3 − 12 bx21 , u3 = x4 − 12 cx21 .
∂w ∂w ∂w ∂w
3. + ax2 + bx3 + cx4 = 0.
∂x1 ∂x2 ∂x3 ∂x4
Integral basis: u1 = ax1 − ln |x2 |, u2 = bx1 − ln |x3 |, u3 = cx1 − ln |x4 |.
∂w ∂w ∂w
4. βγδ + αγδ(γx3 − δx4 ) + αβδ(αx1 + βx2 + γx3 )
∂x1 ∂x2 ∂x3
∂w
+ αβγ(αx1 + βx2 + δx4 ) = 0.
∂x4
Integral basis:
u1 = βx2 − γx3 − δx4 , u2 = (γx3 − δx4 )e−αx1 ,

u3 = (αγx1 x3 − αδx1 x4 − αx1 − βx2 − δx4 − 1)e−αx1 .
∂w ∂w ∂w
5. αβγx1 + βγ(βx3 + γx4 ) + αγ(αx2 + γx4 )
∂x1 ∂x2 ∂x3
∂w
+ αβ(αx2 + βx3 ) = 0.
∂x4
Integral basis:
u1 = x1 (αx2 − βx3 ), u2 = x1 (αx2 − γx4 ), u3 = (αx2 + βx3 + γx4 )x−2

1 .
∂w ∂w
6. βγδ(βx2 + γx3 + δx4 ) + αγδ(αx1 + γx3 + δx4 )
∂x1 ∂x2
∂w ∂w
+ αβδ(αx1 + βx2 + δx4 ) + αβγ(αx1 + βx2 + γx3 ) = 0.
∂x3 ∂x4
Integral basis:
δx4 − βx2 δx4 − γx3
u1 = , u2 = , u3 = (δx4 − αx1 )3 (αx1 + βx2 + γx3 + δx4 ).
δx4 − αx1 δx4 − αx1
∂w ∂w ∂w ∂w
7. x1 x3 + x2 x3 + x23 + (x1 x2 + ax3 x4 ) = 0.
∂x1 ∂x2 ∂x3 ∂x4
Integral basis:
 x2
x2 x3 x1−a

1 + (a − 1)x4 x−a
1 if a 6= 1,
x3
u1 = , u2 = , u3 = x4 x ln x
x1 x1 
 −
2 1
if a = 1.
x1 x3
∂w ∂w ∂w ∂w
8. (γδx3 x4 − αβx1 x22 ) + αγx2 x3 + αγx23 + αγx3 x4 = 0.
∂x1∂x2 ∂x3 ∂x4
2
x3 x4 γδx3 x4 βx2
Integral basis: u1 = , u2 = , u3 = αx1 − 2 exp .
x2 x2 βx2 γx3
∂w ∂w ∂w ∂w
9. βγδx2 x3 x4 +αγδx1 x3 x4 +αβδx1 x2 x4 +αβγx1 x2 x3 = 0.
∂x1 ∂x2 ∂x3 ∂x4
Integral basis: u1 = αx21 − βx22 , u2 = βx22 − γx23 , u3 = γx23 − δx24 .
n
X ∂w
10. xk = aw.
∂xk
k=1
Equation for homogeneous functions of order a. General solution:

a x1 x2 xn−1
w = xn Φ , , ..., .
xn xn xn
n
X n
X
∂w
11. (X − xk ) = 0, X= xk .
∂xk
k=1 k=1
X − nxν
Integral basis: uν = ; ν = 1, 2, . . . , n − 1.
X − nxν+1
n
X n
X
∂w
12. ak0 + akl xl = 0.
∂xk
k=1 l=1
Let s1 , . . . , sn be the roots of the characteristic determinant

a11 − s a12 a13 ... a1n

a21 a22 − s a23 ... a2n

a31 a32 a33 − s . . . a3n .

.. .
.. .
.. .. ..
. . .

an1 an2 an3 . . . ann − s
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 253
For each si there exist n numbers bij not equal to zero simultaneously such that
n
X
ajm bij = bim si ; m = 1, 2, . . . , n,
j=1
and
n
X
di = am0 bim .
m=1
1◦ . If for some i we have si = di = 0, then one of the integrals is given by

n
X
ui = bij xj .
j=1
2◦ . If an si and an sj are distinct nonzero roots, then one of the integrals is

n
sj
di P
+ bim xm
si m=1
uj = n
si .
dj P
+ bjm xm
sj m=1
If all the roots si are different, this formula provides an integral basis.
3◦ . If there are multiple roots among si , then one can reduce the number of independent
variables by using the substitution specified in Section 13.1.3 (see paragraph The method
of reducing the number of independent variables).
n
X n
X
∂w
13. (A0 xk − Ak ) = 0, where Ak = ak0 + akl xl .
∂xk
k=1 l=1
Hesse’s equation. Introduce the homogeneous coordinates x1 = ξ1 /ξ0 , . . . , xn = ξn /ξ0 to

reduce Hesse’s equation to a constant coefficient equation for w = w(ξ0 , ξ1 , . . . , ξn ) with
n + 1 independent variables:
n
X n
X
∂w
Bk = 0, where Bk = akl ξl .
∂ξk
k=0 l=0
For the solution of this equation, see 2.4.9.12.

n
X
∂w ∂w
14. + a k xm
1
k
= bw + c.
∂x1 ∂xk
k=2
This is a special case of equation 2.4.9.32 with fk (x1 ) = ak xmk

1 , g(x1 ) = b, and h(x1 ) = c.
n
X
∂w ∂w
15. + ak xk + bk xm
1
k
= c1 w + c2 .
∂x1 ∂xk
k=2
This is a special case of equation 2.4.9.33 with fk (x1 ) = bk xmk
1 , g(x1 ) = c1 , and h(x1 ) = c2 .
n
X n
X
∂w
16. a k xm
k
k
= bw + ck xskk .
∂xk
k=1 k=1
This is a special case of equation 2.4.9.34 with fk (xk ) = ak xmk sk
k and gk (xk ) = ck xk .
n
X
∂w ∂w
17. + bk xm ak
k−1 xk + ck xk
k
= s1 w + s2 .
∂x1 ∂xk
k=2
This is a special case of equation 2.4.9.36 with fk (. . .) = bk xmk
k−1 , gk (. . .) = ck , h1 (. . .) = s1 ,
and h2 (. . .) = s2 .
◮ Other equations containing arbitrary parameters.

n
X
∂w ∂w
18. + ak eλk x1 = bw + c.
∂x1 ∂xk
k=2
This is a special case of equation 2.4.9.32 with fk (x1 ) = ak eλk x1 , g(x1 ) = b, and h(x1 ) = c.
n
X ∂w
19. ak eλk xk = bw + c.
∂xk
k=1
General solution:

 c b −λn xn
− + exp − e Φ(u1 , u2 , . . . , un−1 ) if b 6= 0,
w= b an λn
− c e−λn xn + Φ(u1 , u2 , . . . , un−1 )

if b = 0,
an λn
where um = am λm e−λn xn − an λn e−λm xm ; m = 1, . . . , n − 1.
n
X
∂w ∂w
20. + ak eβk xk−1 + bk eλk xk = c1 w + c2 .
∂x1 ∂xk
k=2
n
X
∂w ∂w
21. + ak xk + bk eλk x1 = c1 w + c2 .
∂x1 ∂xk
k=2
This is a special case of equation 2.4.9.33 in which fk (x1 ) = bk eλk x1 , g(x1 ) = c1 , and
h(x1 ) = c2 .
n
X
∂w β ∂w
22. + k
ak xk−1 + bk eλk xk = c1 w + c2 .
∂x1 ∂xk
k=2
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 255
n
X ∂w
23. ak sinh λk xk = bw + c.
∂xk
k=1
General solution:

b
− c + tanh 1 λn xn an λn Φ(u1 , u2 , . . . , un−1 ) if b 6= 0,

b 2
w= c

− lntanh 12 λn xn + Φ(u1 , u2 , . . . , un−1 ) if b = 0,
an λn

where um = cotham λm 12 λn xn tanhan λn 12 λm xm ; m = 1, . . . , n − 1.
n
X
∂w ∂w
24. + ak cosh(λk x1 ) = bw + c.
∂x1 ∂xk
k=2
This is a special case of equation 2.4.9.32 with fk (x1 ) = ak cosh(λk x1 ), g(x1 ) = b, and
h(x1 ) = c.
n
X ∂w
25. ak cosh λk xk = bw + c.
∂xk
k=1
General solution:

 c 2b λn x n

− b + exp a λ arctan tanh 2 Φ(u1 , u2 , . . . , un−1 ) if b 6= 0,
n n
w=

 2c λn xn
 arctan tanh + Φ(u1 , u2 , . . . , un−1 ) if b = 0.
an λn 2
1

The functions
um , m = 1, . .
. , n−1, are expressed as um = am λm arctan tanh 2 λn xn −
an λn arctan tanh 12 λm xm .
n
X ∂w
26. ak tanh λk xk = bw + c.
∂xk
k=1

This is a special case of equation 2.4.9.34 with fk (xk ) = ak tanh λk xk .
n
X ∂w
27. ak ln λk xk = bw + c.
∂xk
k=1

This is a special case of equation 2.4.9.34 with fk (xk ) = ak ln λk xk .
n
X
∂w ∂w
28. + ak sin(λk x1 ) = bw + c.
∂x1 ∂xk
k=2
This is a special case of equation 2.4.9.32 with fk (x1 ) = ak sin(λk x1 ), g(x1 ) = b, and
h(x1 ) = c.
n
X ∂w
29. ak sin λk xk = bw + c.
∂xk
k=1
General solution:

c b λn x n


− + tan an λn Φ(u1 , u2 , . . . , un−1 ) if b 6= 0,
b 2

w=

 c λn xn
 lntan + Φ(u1 , u2 , . . . , un−1 ) if b = 0,
an λn 2
1
1

where um = cotam λm 2 λn xn tanan λn 2 λm xm ; m = 1, . . . , n − 1.
n
X ∂w
30. ak cos λk xk = bw + c.
∂xk
k=1
General solution:

 c c
− + sec(λn xn ) + tan(λn xn ) an λn Φ(u1 , u2 , . . . , un−1 ) if b 6= 0,
w= b

 c lnsec(λ x ) + tan(λ x ) + Φ(u , u , . . . , u
 n n n n 1 2 n−1 ) if b = 0,
an λn
where

um = am λm lnsec(λn xn ) + tan(λn xn ) + an λn lnsec(λm xm ) + tan(λm xm );
m = 1, . . . , n − 1.
n
X ∂w
31. ak tan λk xk = bw + c.
∂xk
k=1

This is a special case of equation 2.4.9.34 with fk (xk ) = ak tan λk xk , g1 (x1 ) = c, and
g2 (x2 ) = gn (xn ) = 0.
◮ Equations containing arbitrary functions.
n
X
∂w ∂w
32. + fk (x1 ) = g(x1 )w + h(x1 ).
∂x1 ∂xk
k=2
General solution:
Z Z
h(x1 )
w = G(x1 ) Φ(u1 , u2 , . . . , un−1 ) + dx , G(x1 ) = exp g(x1 ) dx1 ,
G(x1 )
Z
where um = xm+1 − fm+1 (x1 ) dx1 ; m = 1, 2, . . . , n − 1.
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 257
n
X
∂w ∂w
33. + ak xk + fk (x1 ) = g(x1 )w + h(x1 ).
∂x1 ∂xk
k=2
General solution:
Z Z
h(x1 )
w = G(x1 ) Φ(u1 , u2 , . . . , un−1 ) + dx , G(x1 ) = exp g(x1 ) dx1 ,
G(x1 )
where
Z
um = xm+1 exp(−am+1 x1 ) − fm+1 (x1 ) exp(−am+1 x1 ) dx1 ; m = 1, . . . , n − 1.
n
X n
X
∂w
34. fk (xk ) = bw + gk (xk ).
∂xk
k=1 k=1
Introduce the notation
Z Z
dxm dxn
um = − ; m = 1, . . . , n − 1.
fm (xm ) fn (xn )
1◦ . General solution for b = 0:
n Z
X gk (xk )
w = Φ(u1 , u2 , . . . , un−1 ) + dxk .
fk (xk )
k=1
2◦ . General solution for b 6= 0:

n
X Z
gk (xk ) dxk
w = F1 (x1 )Φ(u1 , u2 , . . . , un−1 ) + Fk (xk ) ,
fk (xk )Fk (xk )
k=1
Z
dxk
Fk (xk ) = exp b .
fk (xk )
n
X
∂w ∂w
35. + xk fk (x1 , x2 , . . . , xk−1 ) + gk (x1 , x2 , . . . , xk−1 )
∂x1 ∂xk
k=2
= h1 (x1 , x2 , . . . , xk )w + h2 (x1 , x2 , . . . , xk ).
Change the variables x1 , x2 , x3 , . . . , xn for x1 , u2 , x3 , . . . , xn , where
Z Z
u2 = x2 F2 (x1 ) − g2 (x1 )F2 (x1 ) dx1 , F2 (x1 ) = exp − f2 (x1 ) dx1 , (1)
to obtain the equation

n
X
∂w ∂w
+ xk f¯k (x1 , u2 , x3 , . . . , xk−1 ) + ḡk (x1 , u2 , x3 , . . . , xk−1 )
∂x1 ∂xk
k=3
= h̄1 (x1 , u2 , x3 , . . . , xk )w + h̄2 (x1 , u2 , x3 , . . . , xk ), (2)
whose coefficients are defined as fk (x1 , x2 , x3 , . . . , xk−1 ) ≡ f¯k (x1 , u2 , x3 , . . . , xk−1 ), etc.
Equation (2) is similar to the original one, but contains fewer independent variables,
x1 , x3 , . . . , xn (there is no derivative of w with respect to u2 in the transformed equation,
and hence, u2 can be treated as a parameter). By applying transformations of the form (1)
successively, one can reduce the original partial differential equation to a first-order linear
ordinary differential equation for x1 , the coefficients of which depend on the parameters
u2 , . . . , un .
n
X
∂w ∂w
36. + xk fk (x1 , x2 , . . . , xk−1 ) + xakk gk (x1 , x2 , . . . , xk−1 )
∂x1 ∂xk
k=2
= h1 (x1 , x2 , . . . , xk )w + h2 (x1 , x2 , . . . , xk ).
1◦ . If a2 , . . . , an are not all equal to unity simultaneously, then the transformation zk =
x1−a
k
k
(k = 2, . . . , n) leads to an equation of the form 2.4.9.35:
n
X
∂w ∂w
+ (1 − ak ) zk fek (x1 , z2 , . . . , zk−1 ) + gek (x1 , z2 , . . . , zk−1 )
∂x1 ∂zk
k=2
=e
h1 (x1 , z2 , . . . , zk )w + e
h2 (x1 , z2 , . . . , zk ),
where fek (x1 , z2 , . . . , zk−1 ) ≡ fk (x1 , x2 , . . . , xn ), etc.

2◦ . If an am = 1 and the other ak 6= 1 (k 6= m), then the transformation zm = xm , zk = x1−a
k
k
(k 6= m) leads to an equation of the form 2.4.9.35.

n
X
∂w ∂w
37. + fk (x1 , x2 , . . . , xk−1 ) + exp(λk xk )gk (x1 , x2 , . . . , xk−1 )
∂x1 ∂xk
k=2
= h1 (x1 , x2 , . . . , xk )w + h2 (x1 , x2 , . . . , xk ).
The transformation zk = exp(−λk xk ) (k = 2, . . . , n) leads to an equation of the form
2.4.9.35:
n
X
∂w ∂w
− λk zk fek (x1 , z2 , . . . , zk−1 ) + gek (x1 , z2 , . . . , zk−1 )
∂x1 ∂zk
k=2
=e
h1 (x1 , z2 , . . . , zk )w + e
h2 (x1 , z2 , . . . , zk ),
where fek (x1 , z2 , . . . , zk−1 ) ≡ fk (x1 , x2 , . . . , xn ), etc.

n
X m
X n
X
∂w ∂w
38. (fk −f0 xk ) + ϕk (y1 , . . . , ym ) = 0, fk = ak0 + akl xl .
∂xk ∂yk
k=1 k=1 l=1
By applying the Hesse technique (see equation 2.4.9.13), one can make the first n + 1
coefficients linear. On introducing the homogeneous coordinates x1 = ξ1 /ξ0 , . . . , xn =
ξn /ξ0 , we arrive at the equation
n
X X m n
X
∂w ∂w
gk + ϕk (y1 , . . . , ym ) = 0, where gk = akl ξl .
∂ξk ∂yk
k=0 k=1 l=0
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 259
Z
dy1
In particular, if m = 1 and ϕ1 = ϕ(y1 ), then t = can be chosen to be the
ϕ(y1 )
independent variable in the characteristic equations, with the latter forming a linear system
ξk′ (t) = gk (k = 0, . . . , n).
Chapter 3
Second-Order
Parabolic Equations
with One Space Variable
3.1 Constant Coefficient Equations

∂w ∂2w
3.1.1 Heat Equation =a
∂t ∂x2
This equation is often encountered in the theory of heat and mass transfer. It describes one-
dimensional unsteady thermal processes in quiescent media or solids with constant thermal
diffusivity. A similar equation is used in studying corresponding one-dimensional unsteady
mass-exchange processes with constant diffusivity.
◮ Particular solutions (A, B, and µ are arbitrary constants).
w(x) = Ax + B,
w(x, t) = A(x2 + 2at) + B,
w(x, t) = A(x3 + 6atx) + B,
w(x, t) = A(x4 + 12atx2 + 12a2 t2 ) + B,
w(x, t) = A(x5 + 20atx3 + 60a2 t2 x) + B,
w(x, t) = A(x6 + 30atx4 + 180a2 t2 x2 + 120a3 t3 ) + B,
w(x, t) = A(x7 + 42atx5 + 420a2 t2 x3 + 840a3 t3 x) + B,
n
X (2n)(2n − 1) . . . (2n − 2k + 1)
w(x, t) = x2n + (at)k x2n−2k ,
k!
k=1
n
X
2n+1 (2n + 1)(2n) . . . (2n − 2k + 2)
w(x, t) = x + (at)k x2n−2k+1 ,
k!
k=1
2
w(x, t) = A exp(aµ t ± µx) + B,
w(x, t) = A exp(−aµ2 t) cos(µx) + B,
261
262 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
w(x, t) = A exp(−aµ2 t) sin(µx) + B,

1 x2
w(x, t) = A √ exp − + B,
t 4at

x x2
w(x, t) = A 3/2 exp − + B,
t 4at
w(x, t) = A exp(−µx) cos(µx − 2aµ2 t) + B,
w(x, t) = A exp(−µx) sin(µx − 2aµ2 t) + B,

x
w(x, t) = A erf √ + B,
2 at

x
w(x, t) = A erfc √ + B,
2 at
r
t x2 x x
w(x, t) = A exp − − √ erfc √ + B,
π 4at 2 a 2 at
Z z
2
where n is a positive integer, erf z ≡ √ exp(−ξ 2 ) dξ the error function (probability
π 0
integral), and erfc z = 1 − erf z the complementary error function (complementary proba-
bility integral).
Fundamental solution:

1 x2
E (x, t) = √ exp − .
2 πat 4at
⊙ Literature: H. S. Carslaw and J. C. Jaeger (1984), A. D. Polyanin, A. V. Vyazmin, A. I. Zhurov, and
D. A. Kazenin (1998).
◮ Formulas allowing the construction of particular solutions.

Suppose w = w(x, t) is a solution of the heat equation. Then the functions
w1 = Aw(±λx + C1 , λ2 t + C2 ),
w2 = A exp(λx + aλ2 t)w(x + 2aλt + C1 , t + C2 ),

A βx2 x γ + λt
w3 = p exp − w ± , , λδ − βγ = 1,
|δ + βt| 4a(δ + βt) δ + βt δ + βt
where A, C1 , C2 , β, δ, and λ are arbitrary constants, are also solutions of this equation. The
last formula with β = 1, γ = −1, δ = λ = 0 was obtained with the Appell transformation.
⊙ Literature: W. Miller, Jr. (1977), P. J. Olver (1986).
◮ Infinite series solutions.

A solution involving an arbitrary function of the space variable:
X∞
(at)n (2n) dm
w(x, t) = f (x) + f (x), fx(m) (x) = f (x),
n=1
n! x dxm
3.1. Constant Coefficient Equations 263
where f (x) is any infinitely differentiable function. This solution satisfies the initial con-
dition w(x, 0) = f (x). The sum is finite if f (x) is a polynomial.
Solutions involving arbitrary functions of time:
∞
X 1 (n)
w(x, t) = g(t) + x2n gt (t),
an (2n)!
n=1
∞
X 1 (n)
w(x, t) = xh(t) + x x2n ht (t),
an (2n + 1)!
n=1
where g(t) and h(t) are infinitely differentiable functions. The sums are finite if g(t)
and h(t) are polynomials. The first solution satisfies the boundary condition of the first
kind w(0, t) = g(t) and the second solution the boundary condition of the second kind
∂x w(0, t) = h(t).
⊙ Literature: H. S. Carslaw and J. C. Jaeger (1984).
◮ Transformations allowing separation of variables.

Table 3.1 presents transformations that reduce the heat equation to separable equations (the
identity transformation with ξ = x and g = 1 is omitted).
TABLE 3.1
Transformations of the form ξ = f (x, t), w = g(ξ, t) u(ξ, t) for which the equation
∂t w − ∂xx w = 0 admits multiplicatively separable particular solutions with u(ξ, t) = ϕ(t) ψ(ξ)
Function Factor Function ϕ = ϕ(t); Equation for

No ξ = f (x, t) g = g(ξ, t) ψ = ψ(ξ)
λ is arbitrary
x
1 ξ= √ g=1 ϕ = tλ ′′
ψξξ + 12 ξψξ′ − λψ = 0
t
1

2 ξ = x − 12 t2 g = exp − 12 ξt 1 3
ϕ = exp − 12 t + λt ′′
ψξξ + 2ξ −λ ψ =0
x exp(λ arctan t) 1 2

3 ξ= √ g = exp − 41 ξ 2 t ϕ= ′′
ψξξ + 4ξ −λ ψ =0
1 + t2 (1 + t2 )1/4
Remark 3.1. In general, the solution of the equation for ψ in the first row of Table 3.1 is
expressed in terms of degenerate hypergeometric functions. In the special case λ = 12 n (n =
0, 1, 2, . . . ), the equation admits solutions of the form ψ(ξ) = (i/2)n Hn (iξ/2), where Hn (z) is
the nth Hermite polynomial, i2 = −1. The solution of the equation for ψ in the second row of
Table 3.1 is expressed in terms of Bessel functions, and that in the third row, in terms of parabolic
cylinder functions.
⊙ Literature: E. Kalnins and W. Miller, Jr. (1974), W. Miller, Jr. (1977).
◮ Domain: −∞ < x < ∞. Cauchy problem.

An initial condition is prescribed:
w = f (x) at t = 0.
Solution:
Z ∞
1 (x − ξ)2
w(x, t) = √ exp − f (ξ) dξ.
2 πat −∞ 4at
Example 3.1. The initial temperatures in the domains |x| < x0 and |x| > x0 are constant and
equal to w1 and w2 , respectively, i.e.,
(
w1 for |x| < x0 ,
f (x) =
w2 for |x| > x0 .
Solution:

1 x0 − x x0 + x
w= (w1 − w2 ) erf √ + erf √ + w2 .
2 2 at 2 at
◮ Domain: 0 ≤ x < ∞. First boundary value problem.
The following conditions are prescribed:
w = f (x) at t = 0 (initial condition),

w = g(t) at x = 0 (boundary condition).
Solution:
Z ∞
1 (x − ξ)2 (x + ξ)2
w(x, t) = √ exp − − exp − f (ξ) dξ
2 πat 0 4at 4at
Z t
x x2 g(τ ) dτ
+ √ exp − .
2 πa 0 4a(t − τ ) (t − τ )3/2
Example 3.2. The initial temperature is linearly dependent on the space coordinate, f (x) =
w0 + bx. The temperature at the boundary is zero, g(t) = 0.
Solution:

x
w = w0 erf √ + bx.
2 at
The case of uniform initial temperature with f (x) = w0 corresponds to the value b = 0.
Example 3.3. The initial temperature is zero, f (x) = 0. The temperature at the boundary
increases linearly with time, g(t) = At.
Solution:

x2 x x x2
w = At 1 + erfc √ − √ exp − .
2at 2 at πat 4at
⊙ Literature: A. G. Butkovskiy (1979), H. S. Carslaw and J. C. Jaeger (1984).

◮ Domain: 0 ≤ x < ∞. Second boundary value problem.

∂x w = g(t) at x = 0 (boundary condition).
Solution:
Z ∞
1 (x − ξ)2 (x + ξ)2
w(x, t) = √ exp − + exp − f (ξ) dξ
2 πat 0 4at 4at
r Z t
a x2 g(τ )
− exp − √ dτ.
π 0 4a(t − τ ) t−τ
Example 3.4. The initial temperature is zero, f (x) = 0. A constant thermal flux is maintained
at the boundary all the time, g(t) = −Q.
Solution: r
at x2 x
w = 2Q exp − − Qx erfc √ .
π 4at 2 at
◮ Domain: 0 ≤ x < ∞. Third boundary value problem.

∂x w − kw = g(t) at x = 0 (boundary condition).
Solution:
Z ∞ Z t
w(x, t) = f (ξ)G(x, ξ, t) dξ − a g(τ )G(x, 0, t − τ ) dτ,
0 0
where

1 (x − ξ)2 (x + ξ)2
G(x, ξ, t) = √ exp − + exp −
2 πat 4at 4at
Z ∞ 2

(x + ξ + η)
− 2k exp − − kη dη .
0 4at
The improper integral may be calculated by the formula
Z ∞
(x + ξ + η)2 √ 2 x+ξ √
exp − − kη dη = πat exp ak t + k(x + ξ) erfc √ + k at .
0 4at 2 at
Example 3.5. The initial temperature is uniform, f (x) = w0 . The temperature of the contacting
medium is zero, g(t) = 0.
Solution:

x 2 x √
w = w0 erf √ + exp(kx + ak t) erfc √ + k at .
2 at 2 at
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980), H. S. Carslaw and J. C. Jaeger
(1984).
◮ Domain: 0 ≤ x ≤ l. First boundary value problem.

w = g1 (t) at x = 0 (boundary condition),
w = g2 (t) at x = l (boundary condition).
Solution:
∞
2X nπx an2 π 2 t
w(x, t) = sin exp − Mn (t),
l l l2
n=1
where
Z l Z t
nπξ anπ an2 π 2 τ
Mn (t) = f (ξ) sin dξ + exp g1 (τ ) − (−1)n g2 (τ ) dτ.
0 l l 0 l2
Remark 3.2. Using the relations [see Prudnikov, Brychkov, and Marichev (1986)]
X∞ ∞
X
sin nξ π−ξ sin nξ ξ
= (0 < ξ < 2π); (−1)n−1 = (−π < ξ < π),
n=1
n 2 n=1
n 2
one can transform the solution to
∞
x 2X nπ
w(x, t) = g1 (t) + g2 (t) − g1 (t) + sin(λn x) exp(−aλ2n t)Rn (t), λn = ,
l l n=1 l
where Z l
1
Rn (t) = f (ξ) sin(λn ξ) dξ − exp(aλ2n t) g1 (t) − (−1)n g2 (t)
0 λn
Z t

+ aλn exp(aλ2n τ ) g1 (τ ) − (−1)n g2 (τ ) dτ
0
Z l
1
=
f (ξ) sin(λn ξ) dξ − g1 (0) − (−1)n g2 (0)
0 λ n
Z t
1
− exp(aλ2n τ ) g1′ (τ ) − (−1)n g2′ (τ ) dτ.
λn 0
Note that another representation of the solution is given in Section 3.1.2 (see the first boundary
value problem for 0 ≤ x ≤ l).
Example 3.6. The initial temperature is uniform, f (x) = w0 . Both ends are maintained at zero
temperature, g1 (t) = g2 (t) = 0.
Solution:
∞
4w0 X 1 (2n + 1)πx a(2n + 1)2 π 2 t
w= sin exp − .
π n=0 (2n + 1) l l2
Example 3.7. The initial temperature is zero, f (x) = 0. The ends are maintained at uniform
temperatures, g1 (t) = w1 and g2 (t) = w2 .
Solution:
∞
x 2 X (−1)n w2 − w1 nπx an2 π 2 t
w = w1 + (w2 − w1 ) + sin exp − .
l π n=0 n l l2
◮ Domain: 0 ≤ x ≤ l. Second boundary value problem.

∂x w = g1 (t) at x = 0 (boundary condition),
∂x w = g2 (t) at x = l (boundary condition).
Solution:
Z l Z t Z t
w(x, t) = f (ξ)G(x, ξ, t) dξ −a g1 (τ )G(x, 0, t−τ ) dτ +a g2 (τ )G(x, l, t−τ ) dτ,
0 0 0
where
∞
1 2X nπx nπξ an2 π 2 t
G(x, ξ, t) = + cos cos exp − .
l l l l l2
n=1

(1984).
◮ Domain: 0 ≤ x ≤ l. Third boundary value problem (k1 > 0 and k2 > 0).

∂x w − k1 w = g1 (t) at x = 0 (boundary condition),
∂x w + k2 w = g2 (t) at x = l (boundary condition).
Solution:
Z l Z t Z t
w(x, t) = f (ξ)G(x, ξ, t) dξ −a g1 (τ )G(x, 0, t−τ ) dτ +a g2 (τ )G(x, l, t−τ ) dτ,
0 0 0
where
∞
X 1
G(x, ξ, t) = yn (x)yn (ξ) exp(−aµ2n t),
n=1
kyn k2

k1 2 k2 µ2n + k12 k1 l k12
yn (x) = cos(µn x) + sin(µn x), kyn k = + 2 + 1+ 2 .
µn 2µ2n µ2n + k22 2µn 2 µn
tan(µl) k1 + k2
Here, the µn are positive roots of the transcendental equation = 2 .
µ µ − k1 k2
◮ Domain: 0 ≤ x ≤ l. Mixed boundary value problems.

1◦ . The following conditions are prescribed:

Solution:
Z l Z t Z t
w(x, t) = f (ξ)G(x, ξ, t) dξ + a g1 (τ )Λ(x, t − τ ) dτ + a g2 (τ )G(x, l, t − τ ) dτ,
0 0 0
where
∞
2X π(2n + 1)x π(2n + 1)ξ aπ 2 (2n + 1)2 t
G(x, ξ, t) = sin sin exp − ,
l 2l 2l 4l2
n=0
∂

Λ(x, t) = G(x, ξ, t) .
∂ξ ξ=0

Solution:
Z l Z t Z t
w(x, t) = f (ξ)G(x, ξ, t) dξ − a g1 (τ )G(x, 0, t − τ ) dτ − a g2 (τ )H(x, t − τ ) dτ,
0 0 0
where
∞
2X π(2n + 1)x π(2n + 1)ξ aπ 2 (2n + 1)2 t
G(x, ξ, t) = cos cos exp − ,
l 2l 2l 4l2
n=0
∂

H(x, t) = G(x, ξ, t) .
∂ξ ξ=l
Note that Section 3.1.2 (see the mixed boundary value problems for 0 ≤ x ≤ l) also
gives other forms of representation of solutions to mixed boundary value problems.
Example 3.8. The initial temperature is zero, f (x) = 0. The left end is heat insulated, and the
right end is maintained at a constant temperature, g1 (t) = 0 and g2 (t) = A.
Solution:
∞
4A X (−1)n+1 π(2n + 1)x aπ 2 (2n + 1)2 t
w =A+ cos exp − .
π n=0 2n + 1 2l 4l2
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980), A. V. Bitsadze and D. F. Kalini-

chenko (1985).
◮ Problems without initial conditions.

In applications, problems are encountered in which the process is studied at a time instant
fairly remote from the initial instant and, in this case, the initial conditions do not practically
affect the distribution of the desired quantity at the observation instant. In such problems,
no initial condition is stated, and the boundary conditions are assumed to be prescribed for
all preceding time instants, −∞ < t. However, in addition, the boundedness condition in
the entire domain is imposed on the solution.
As an example, consider the first boundary value problem for the half-space 0 ≤ x < ∞
with the boundary conditions
w = g(t) at x = 0, w → 0 as x → ∞.
Solution:
Z t
x g(τ ) x2
w(x, t) = √ exp − dτ.
2 πa −∞ (t − τ )3/2 4a(t − τ )
Example 3.9. The temperature at the boundary is a harmonic function of time, i.e.,
g(t) = w0 cos(ωt + β).
Solution: r r
ω ω
w = w0 exp − x cos ωt − x+β .
2a 2a
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin, et al. (1964), A. N. Tikhonov and A. A. Samar-
skii (1990).
◮ Conjugate heat and mass transfer problems.

In such problems, one deals with two (or more) domains, V1 and V2 , with interface S. The
domains are filled by different media. Each of the media is characterized by its own thermal
conductivity, λ1 and λ2 , and thermal diffusivity, a1 and a2 . The processes in each of the
media are described by appropriate (different) equations of heat and mass transfer. The
thermal equilibrium conditions express the equality of the temperatures and of the thermal
fluxes at the interface. Below we consider a typical example of a conjugate problem (a
more detailed analysis of such problems is beyond the scope of this handbook).
Consider two semiinfinite solids (two semiinfinite quiescent media) with temperature
distributions, w1 = w1 (x, t) and w2 = w2 (x, t), governed by the equations
∂w1 ∂ 2 w1
= a1 (in the range 0 < x < ∞),
∂t ∂x2
∂w2 ∂ 2 w2
= a2 (in the range −∞ < x < 0).
∂t ∂x2
Either solid has its own temperature profile at the initial instant t = 0, and conjugate bound-
ary conditions are imposed at the interface x = 0; specifically,
w1 = f1 (x) at t = 0 (initial condition),
w2 = f2 (x) at t = 0 (initial condition),
w1 = w2 at x = 0 (boundary condition),
λ1 ∂x w1 = λ2 ∂x w2 at x = 0 (boundary condition).
Solution:
Z ∞
1 (x − ξ)2 (x + ξ)2
w1 (x, t) = √ f1 (ξ) exp − + exp − dξ
2 πa1 t 0 4a1 t 4a1 t
r Z t
a1 x2 g(τ ) dτ
− 2 exp − √ ,
πλ1 0 4a1 (t − τ ) t−τ
Z ∞
1 (x − ξ)2 (x + ξ)2
w2 (x, t) = √ f2 (−ξ) exp − + exp − dξ
2 πa2 t 0 4a2 t 4a2 t
r Z t
a2 x2 g(τ ) dτ
+ 2 exp − √ .
πλ2 0 4a2 (t − τ ) t−τ
The function g(t) is given by
Z t
λ1 λ2 d F (τ ) dτ
g(t) = √ √ p ,
π(λ1 a2 + λ2 a1 ) dt 0 τ (t − τ )
where
Z ∞ Z ∞
1 ξ2 1 ξ2
F (t) = √ f1 (ξ) exp − dξ − √ f2 (−ξ) exp − dξ.
a1 0 4a1 t a2 0 4a2 t
Example 3.10. The initial temperatures are uniform, f1 (x) = A and f2 (x) = B.
Solution:
w1 (x, t) − B K 1 x
= 1+ erf √ ,
A−B 1+K K 2 a1 t

w2 (x, t) − B K |x|
= erfc √ ,
A−B 1+K 2 a1 t
r
λ1 a2
where the quantity K = characterizes the thermal activity of the first medium with respect
λ2 a1
to the second medium.
⊙ Literature: A. V. Lykov (1967), H. S. Carslaw and J. C. Jaeger (1984).
∂w ∂2w
3.1.2 Equation of the Form =a + Φ(x, t)
∂t ∂x2
This sort of equation describes one-dimensional unsteady thermal processes in quiescent
media or solids with constant thermal diffusivity in the presence of a volume thermal source
dependent on the space coordinate and time.

w = f (x) at t = 0.
Solution:
Z ∞ Z tZ ∞
w(x, t) = f (ξ)G(x, ξ, t) dξ + Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ,
−∞ 0 −∞
where
1 (x − ξ)2
G(x, ξ, t) = √ exp − .
2 πat 4at


Solution:
Z ∞ Z t
w(x, t) = f (ξ)G(x, ξ, t) dξ + g(τ )H(x, t − τ ) dτ
0 0
Z tZ ∞
+ Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ,
0 0
where

1 (x − ξ)2 (x + ξ)2
G(x, ξ, t) = √ exp − − exp − ,
2 πat 4at 4at

x x2
H(x, t) = √ exp − .
2 πa t3/2 4at


Solution:
Z ∞ Z t
w(x, t) = G(x, ξ, t)f (ξ) dξ − a g(τ )G(x, 0, t − τ ) dτ
0 0
Z tZ ∞
+ Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ,
0 0
where

1 (x − ξ)2 (x + ξ)2
G(x, ξ, t) = √ exp − + exp − .
2 πat 4at 4at

Solution:
Z ∞ Z t
w(x, t) = f (ξ)G(x, ξ, t) dξ − a g(τ )G(x, 0, t − τ ) dτ
0 0
Z tZ ∞
+ Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ,
0 0
where

1 (x − ξ)2 (x + ξ)2
G(x, ξ, t) = √ exp − + exp −
2 πat 4at 4at
Z ∞ 2

(x + ξ + η)
0 4at
⊙ Literature: A. G. Butkovskiy (1979), B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).

Solution:
Z l Z tZ l
w(x, t) = f (ξ)G(x, ξ, t) dξ +Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ
0 0 0
Z t Z t
+a g1 (τ )H1 (x, t − τ ) dτ − a g2 (τ )H2 (x, t − τ ) dτ.
0 0
Two forms of representation of the Green’s function:
∞
2X nπx nπξ an2 π 2 t
G(x, ξ, t) = sin sin exp −
l n=1 l l l2
X∞
1 (x − ξ + 2nl)2 (x + ξ + 2nl)2
= √ exp − − exp − .
2 πat n=−∞ 4at 4at
The first series converges rapidly at large t and the second series at small t. The functions
H1 and H2 are expressed in terms of the Green’s function as
∂ ∂

H1 (x, t) = G(x, ξ, t) , H2 (x, t) = G(x, ξ, t) .
∂ξ ξ=0 ∂ξ ξ=l
⊙ Literature: A. G. Butkovskiy (1979), B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980), H. S.
Carslaw and J. C. Jaeger (1984).

Solution:
Z l Z tZ l
w(x, t) = f (ξ)G(x, ξ, t) dξ + Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ
0 0 0
Z t Z t
−a g1 (τ )G(x, 0, t − τ ) dτ + a g2 (τ )G(x, l, t − τ ) dτ.
0 0
∞
G(x, ξ, t) = + cos cos exp −
l l l l l2
n=1
∞
X
1 (x − ξ + 2nl)2 (x + ξ + 2nl)2
= √ exp − + exp − .
The first series converges rapidly at large t and the second series at small t.
◮ Domain: 0 ≤ x ≤ l. Third boundary value problem (k1 > 0, k2 > 0).

The solution is given by the formula presented in Section 3.1.1 (see the third boundary
value problem for 0 ≤ x ≤ l) with the additional term
Z tZ l
Φ(ξ, τ ) G(x, ξ, t − τ ) dξ dτ,
0 0
which takes into account the nonhomogeneity of the equation.

◮ Domain: 0 ≤ x ≤ l. Mixed boundary value problems.


Solution:
Z l Z tZ l
0 0 0
Z t Z t
∂
+a g1 (τ ) G(x, ξ, t − τ ) dτ + a g2 (τ )G(x, l, t − τ ) dτ.
0 ∂ξ ξ=0 0

∞
2X π(2n+1)x π(2n+1)ξ aπ 2(2n+1)2t
G(x, ξ, t) = sin sin exp −
l n=0 2l 2l 4l2
X ∞
1 n (x−ξ +2nl)2 (x+ξ +2nl)2
= √ (−1) exp − −exp − .

Solution:
Z l Z tZ l
w(x, t) = f (ξ)G(x, ξ, t) dξ +
Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ
0 0 0
t Z Z t
∂
−a g1 (τ )G(x, 0, t − τ ) dτ − a g2 (τ ) G(x, ξ, t − τ ) dτ.
0 0 ∂ξ ξ=l

∞
2X π(2n+1)x π(2n+1)ξ aπ 2(2n+1)2t
G(x, ξ, t) = cos cos exp −
l n=0 2l 2l 4l2
X∞
1 (x−ξ +2nl)2 (x+ξ +2nl)2
= √ (−1)n exp − +exp − .
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980), A. V. Bitsadze and D. F. Kalini-
chenko (1985).
∂w ∂2w
3.1.3 Equation of the Form =a + bw + Φ(x, t)
∂t ∂x2
Homogeneous equations of this form describe one-dimensional unsteady mass transfer in
a quiescent medium with a first-order volume chemical reaction; the cases b < 0 and b > 0
correspond to absorption and release of substance, respectively. A similar equation is used
to analyze appropriate one-dimensional thermal processes in which volume heat release
(b > 0) proportional to temperature occurs in the medium. Furthermore, this equation
governs heat transfer in a one-dimensional rod whose lateral surface exchanges heat with
the ambient medium having constant temperature; b > 0 if the temperature of the medium
is greater than that of the rod, and b < 0 otherwise.
◮ Homogeneous equation (Φ ≡ 0).
1◦ . Particular solutions:
p
w(x) = Aeλx + Be−λx , λ= −b/a,
bt
w(x, t) = (Ax + B)e ,

w(x, t) = A(x2 + 2at) + B ebt ,

w(x, t) = A(x3 + 6atx) + B ebt ,

w(x, t) = A(x4 + 12atx2 + 12a2 t2 ) + B ebt ,

w(x, t) = A(x5 + 20atx3 + 60a2 t2 x) + B ebt ,

w(x, t) = A(x6 + 30atx4 + 180a2 t2 x2 + 120a3 t3 ) + B ebt ,

w(x, t) = A exp (aµ2 + b)t ± µx + Bebt ,

1 x2
w(x, t) = A √ exp − + bt + Bebt ,
t 4at

x x2
w(x, t) = A 3/2 exp − + bt + Bebt ,
t 4at

w(x, t) = A exp (b − aµ2 )t cos(µx) + Bebt ,

w(x, t) = A exp (b − aµ2 )t sin(µx) + Bebt ,
w(x, t) = A exp(−µx + bt) cos(µx − 2aµ2 t) + Bebt ,
w(x, t) = A exp(−µx + bt) sin(µx − 2aµ2 t) + Bebt ,
p
w(x, t) = A exp(−µx) cos(βx − 2aβµt), β = µ2 + b/a,
p
w(x, t) = A exp(−µx) sin(βx − 2aβµt), β = µ2 + b/a,

x
w(x, t) = Aebt erf √ + Bebt ,
2 at

bt x
w(x, t) = Ae erfc √ + Bebt ,
2 at
where A, B, and µ are arbitrary constants.

2◦ . Fundamental solution:

1 x2
E (x, t) = √ exp − + bt .
2 πat 4at
◮ Reduction to the heat equation. Remarks on the Green’s functions.

The substitution w(x, t) = ebt u(x, t) leads to the nonhomogeneous heat equation
∂u ∂2u
= a 2 + e−bt Φ(x, t),
∂t ∂x
which is discussed in Section 3.1.2 in detail. The initial condition for the new variable u
remains the same, and the nonhomogeneous part in the boundary conditions is multiplied
by e−bt . Taking this into account, one can easily solve the original equation subject to the
initial and boundary conditions considered in Section 3.1.2.
In all the boundary value problems that are dealt with in the current subsection, the
Green’s function can be represented in the form
Gb (x, ξ, t) = ebt G0 (x, ξ, t),
where G0 (x, ξ, t) is the Green’s function for the heat equation that corresponds to b = 0.

w = f (x) at t = 0.
Solution:
Z ∞ Z tZ ∞
−∞ 0 −∞
where
1 (x − ξ)2
G(x, ξ, t) = √ exp − + bt .
2 πat 4at
⊙ Literature: A. G. Butkovskiy (1979).

Solution:
Z ∞ Z t
w(x, t) = f (ξ)G(x, ξ, t) dξ + g(τ )H(x, t − τ ) dτ
0 0
Z tZ ∞
+ Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ,
0 0
where

ebt (x − ξ)2 (x + ξ)2
G(x, ξ, t) = √ exp − − exp − ,
2 πat 4at 4at

xebt x2
H(x, t) = √ exp − .
2 πa t3/2 4at
⊙ Literature: A. D. Polyanin, A. V. Vyazmin, A. I. Zhurov, and D. A. Kazenin (1998).

Solution:
Z ∞ Z t
w(x, t) = G(x, ξ, t)f (ξ) dξ − a g(τ )G(x, 0, t − τ ) dτ
0 0
Z tZ ∞
+ Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ,
0 0
where
ebt (x − ξ)2 (x + ξ)2
G(x, ξ, t) = √ exp − + exp − .
2 πat 4at 4at

The solution w(x, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where

ebt (x − ξ)2 (x + ξ)2
G(x, ξ, t) = √ exp − + exp −
2 πat 4at 4at
Z ∞ 2

(x + ξ + η)
0 4at

Solution:
Z l Z tZ l
0 0 0
Z t Z t
+a g1 (τ )H1 (x, t − τ ) dτ − a g2 (τ )H2 (x, t − τ ) dτ,
0 0
where
∞
2 bt X nπx nπξ an2 π 2 t
G(x, ξ, t) = e sin sin exp − 2
,
l n=1
l l l
∂ ∂

H1 (x, t) = G(x, ξ, t) , H2 (x, t) = G(x, ξ, t) .
∂ξ ξ=0 ∂ξ ξ=l

Solution:
Z l Z tZ l
0 0 0
Z t Z t
−a g1 (τ )G(x, 0, t − τ ) dτ + a g2 (τ )G(x, l, t − τ ) dτ,
0 0
where
∞
bt 1 2X nπx nπξ an2 π 2 t
G(x, ξ, t) = e + cos cos exp − .
l l n=1 l l l2
◮ Domain: 0 ≤ x ≤ l. Third boundary value problem.

∞
X 1
G(x, ξ, t) = ebt yn (x)yn (ξ) exp(−aµ2n t),
kyn k2
n=1

k1 2 k2 µ2n + k12 k1 l k12
yn (x) = cos(µn x) + sin(µn x), kyn k = + 2 + 1+ 2 .
tan(µl) k1 + k2
µ µ − k1 k2
◮ Domain: 0 ≤ x ≤ l. Mixed boundary value problem.


Solution:
Z l Z tZ l
w(x, t) = f (ξ)G(x, ξ, t) dξ +Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ
0 0 0
Z t Z t
+a g1 (τ )Λ(x, t − τ ) dτ + a g2 (τ )G(x, l, t − τ ) dτ,
0 0
where
∞
2 bt X π(2n + 1)x π(2n + 1)ξ aπ 2 (2n + 1)2 t
G(x, ξ, t) = e sin sin exp − ,
l 2l 2l 4l2
n=0
∂

Λ(x, t) = G(x, ξ, t) .
∂ξ ξ=0
◮ Domain: 0 ≤ x < ∞. A problem with Φ = 0 and without an initial condition.

w = A cos(ωt + γ) at x = 0, w→0 as x → ∞.
Solution:
w = Ae−λx cos(ωt − βx + γ),
where √ 2 √ 1/2
ω + b2 − b 1/2 ω 2 + b2 + b
λ= , β= .
2a 2a
∂w ∂2w ∂w
3.1.4 Equation of the Form =a +b + Φ(x, t)
∂t ∂x2 ∂x
This equation is encountered in one-dimensional nonstationary problems of convective
mass transfer in a continuous medium that moves with a constant velocity; the case Φ ≡ 0
means that there is no absorption or release of substance.

w(x) = Ae−λx + B, λ = b/a,
w(x, t) = Ax + Abt + B,
w(x, t) = A(x + bt)2 + 2Aat + B,
w(x, t) = A(x + bt)3 + 6Aatx + B,

w(x, t) = A exp (aµ2 + bµ)t + µx + B,

1 (x + bt)2
w(x, t) = A √ exp − + B,
t 4at
w(x, t) = A exp(−aµ2 t) cos(µx + bµt) + B,
w(x, t) = A exp(−aµ2 t) sin(µx + bµt) + B,
p
w(x, t) = A exp(−µx) cos βx + β(b − 2aµ)t + B, β= µ2 − (b/a)µ,
p
w(x, t) = A exp(−µx) sin βx + β(b − 2aµ)t + B, β = µ2 − (b/a)µ,

x + bt
w(x, t) = A erf √ + B,
2 at

x + bt
w(x, t) = A erfc √ + B,
2 at

1 (x + bt)2
E (x, t) = √ exp − .
2 πat 4at
◮ Reduction to the heat equation. Remarks on the Green’s function.

1◦ . The substitution
b2 b
w(x, t) = exp(−βt − µx)u(x, t), β= , µ= (1)
4a 2a
leads to the nonhomogeneous heat equation
∂u ∂2u
= a 2 + exp(βt + µx)Φ(x, t), (2)
∂t ∂x
which is considered in Section 3.1.2 in detail.
2◦ . On passing from t, x to the new variables t, z = x + bt, we obtain the nonhomogeneous

heat equation
∂w ∂2w
= a 2 + Φ(z − bt, t),
∂t ∂z
which is treated in Section 3.1.2.
3◦ . For all first boundary value problems, the Green’s function can be represented as

b b2
Gb (x, ξ, t) = exp (ξ − x) − t G0 (x, ξ, t),
2a 4a
where G0 (x, ξ, t) is the Green’s function for the heat equation that corresponds to b = 0.

w = f (x) at t = 0.
Solution:
Z ∞ Z tZ ∞
−∞ 0 −∞
where
1 b(ξ − x) b2 t (x − ξ)2
G(x, ξ, t) = √ exp − − .
2 πat 2a 4a 4at


Solution:
Z ∞ Z t
w(x, t) = f (ξ)G(x, ξ, t) dξ + a g(τ )Λ(x, t − τ ) dτ
0 0
Z tZ ∞
+ Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ,
0 0
where

1 b(ξ − x) b2 t (x − ξ)2 (x + ξ)2
G(x, ξ, t) = √ exp − exp − − exp − ,
2 πat 2a 4a 4at 4at
∂

Λ(x, t) = G(x, ξ, t) .
∂ξ ξ=0


Substitution (1) reduces the considered equation to the nonhomogeneous heat equa-
tion (2) with the following initial and boundary conditions:
u = exp(µx)f (x) at t = 0,
(3)
∂x u − µu = exp(βt)g(t) at x = 0.
See Section 3.1.2 for the solution of the third boundary value problem (2)–(3) for
0 ≤ x < ∞.


(4)
∂x u − (k + µ)u = exp(βt)g(t) at x = 0.
See Section 3.1.2 for the solution of the third boundary value problem (2), (4) for
0 ≤ x < ∞.


Solution:
Z l Z tZ l
w(x, t) = f (ξ)G(x, ξ, t) dξ +
0 0 0
Z t Z t
0 0
where
∞
2 b b2 X πnx πnξ aπ 2n2
G(x, ξ, t) = exp (ξ − x) − t sin sin exp − 2 t ,
l 2a 4a n=1 l l l
∂ ∂

H1(x, t) = G(x, ξ, t) , H2(x, t) = G(x, ξ, t) .
∂ξ ξ=0 ∂ξ ξ=l

Solution:
Z l Z tZ l
0 0 0
Z t Z t
0 0
where

b bξ
G(x, ξ, t) = exp
a(ebl/a − 1) a
∞
2 b(ξ − x) b2 t X yn (x)yn (ξ) aπ 2 n2
+ exp − exp − 2 t ,
l 2a 4a 1 + µ2n l
n=1

πnx πnx bl
yn (x) = cos + µn sin , µn = .
l l 2aπn

∞
b(ξ − x) b2 t X 1
G(x, ξ, t) = exp − yn (x)yn (ξ) exp(−aµ2n t),
2a 4a Bn
n=1
2ak1 + b
yn (x) = cos(µn x) + sin(µn x),
2aµn
2ak2 − b 4a2 µ2n + (2ak1 + b)2 2ak1 + b l l(2ak1 + b)2
Bn = + + + ,
4aµ2n 4a2 µ2n + (2ak2 − b)2 4aµ2n 2 8a2 µ2n
and the µn are positive roots of the transcendental equation
tan(µl) 4a2 (k1 + k2 )
= 2 2 .
µ 4a µ − (2ak1 + b)(2ak2 − b)
∂w ∂2w ∂w
3.1.5 Equation of the Form =a +b + cw + Φ(x, t)
∂t ∂x2 ∂x
For Φ ≡ 0, this equation describes one-dimensional unsteady convective mass transfer
with a first-order volume chemical reaction in a continuous medium that moves with a
constant velocity. A similar equation is used for the analysis of the corresponding one-
dimensional thermal processes in a moving medium with volume heat release proportional
to temperature.

w(x, t) = ect (Ax + Abt + B),

w(x, t) = ect A(x + bt)2 + 2Aat + B ,

w(x, t) = ect A(x + bt)3 + 6Aatx + B ,
w(x, t) = Ae−λx+ct + Bect , λ = b/a,

w(x, t) = A exp (aµ2 + bµ + c)t + µx + Bect ,

1 (x + bt)2
w(x, t) = A √ exp − + ct + Bect ,
t 4at
w(x, t) = A exp(ct − aµ2 t) cos(µx + bµt) + Bect ,
w(x, t) = A exp(ct − aµ2 t) sin(µx + bµt) + Bect ,
p
w(x, t) = A exp(−µx) cos βx + β(b − 2aµ)t , β = µ2 − (b/a)µ + c/a,
p
w(x, t) = A exp(−µx) sin βx + β(b − 2aµ)t , β = µ2 − (b/a)µ + c/a,

x + bt
w(x, t) = Aect erf √ + Bect ,
2 at

ct x + bt
w(x, t) = Ae erfc √ + Bect ,
2 at

1 (x + bt)2
E (x, t) = √ exp − + ct .
2 πat 4at
◮ Reduction to the heat equation. Remarks on the Green’s functions.

b2 b
w(x, t) = exp(−βt − µx)u(x, t), β = −c + , µ= (1)
4a 2a
∂u ∂2u
= a 2 + exp(βt + µx)Φ(x, t), (2)
∂t ∂x
which is considered in Section 3.1.2 in detail.
2◦ . The transformation
w(x, t) = ect v(z, t), z = x + bt,
∂v ∂2v
= a 2 + e−ct Φ(z − bt, t),
∂t ∂z
which is treated in Section 3.1.2.
3◦ . For all first boundary value problems, the Green’s function can be represented as

b b2
Gb,c (x, ξ, t) = exp (ξ − x) + c − t G0,0 (x, ξ, t),
2a 4a
where G0,0 (x, ξ, t) is the Green’s function for the heat equation corresponding to b = c = 0.

w = f (x) at t = 0.
Solution:
Z ∞ Z tZ ∞
−∞ 0 −∞
where
1 b b2 (x − ξ)2
G(x, ξ, t) = √ exp (ξ − x) + c − t− .
2 πat 2a 4a 4at

Solution:
Z ∞ Z t
w(x, t) = f (ξ)G(x, ξ, t) dξ + a g(τ )Λ(x, t − τ ) dτ
0 0
Z tZ ∞
+ Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ,
0 0
where

1 b(ξ−x) b2 (x−ξ)2 (x+ξ)2
G(x, ξ, t)= √ exp + c− t exp − −exp − ,
2 πat 2a 4a 4at 4at
∂

Λ(x, t)= G(x, ξ, t) .
∂ξ ξ=0

(3)
∂x u − µu = exp(βt)g(t) at x = 0.
See Section 3.1.2 for the solution of the third boundary value problem (2)–(3) for
0 ≤ x < ∞.

(4)
∂x u − (k + µ)u = exp(βt)g(t) at x = 0.
See Section 3.1.2 for the solution of the third boundary value problem (2), (4) for
0 ≤ x < ∞.


Solution:
Z l Z tZ l
w(x, t) = f (ξ)G(x, ξ, t) dξ +
0 0 0
Z t Z t
0 0
where
X∞
2 b(ξ − x) b2 πnx πnξ aπ 2 n2
G(x, ξ, t) = exp + c− t sin sin exp − 2 t ,
l 2a 4a n=1
l l l
∂ ∂

H1 (x, t) = G(x, ξ, t) , H2 (x, t) = G(x, ξ, t) .
∂ξ ξ=0 ∂ξ ξ=l


Solution:
Z l Z tZ l
0 0 0
Z t Z t
0 0
where
X ∞
bξ 2 b(ξ − x) b2 yn (x)yn (ξ) aπ 2 n2
G(x, ξ, t) = A exp +ct + exp + c− t exp − t ,
a l 2a 4a n=1
1 + µ2n l2

b πnx πnx bl
A= bl/a
, y n (x) = cos + µ n sin , µn = .
a(e − 1) l l 2aπn


X∞
b(ξ − x) b2 1
G(x, ξ, t) = exp + c− t yn (x)yn (ξ) exp(−aµ2n t),
2a 4a n=1
B n
2ak1 + b
yn (x) = cos(µn x) + sin(µn x),
2aµn
2ak2 − b 4a2 µ2n + (2ak1 + b)2 2ak1 + b l l(2ak1 + b)2
Bn = + + + ,
4aµ2n 4a2 µ2n + (2ak2 − b)2 4aµ2n 2 8a2 µ2n
and the µn are positive roots of the transcendental equation
tan(µl) 4a2 (k1 + k2 )

= 2 2 .
µ 4a µ − (2ak1 + b)(2ak2 − b)
3.2 Heat Equation with Axial or Central Symmetry and

Related Equations
∂w
∂2w 1 ∂w

3.2.1 Equation of the Form =a +
∂t ∂r 2 r ∂r
This is a sourceless heat equation that describes one-dimensional unsteady thermal pro-
cesses having axial symmetry. It is often represented in the equivalent form

∂w a ∂ ∂w
= r .
∂t r ∂r ∂r
A similar equation is used for the analysis of the corresponding one-dimensional unsteady
diffusion processes.
w(r) = A + B ln r,
w(r, t) = A + B(r 2 + 4at),
w(r, t) = A + B(r 4 + 16atr 2 + 32a2 t2 ),
3.2. Heat Equation with Axial or Central Symmetry and Related Equations 289
Xn
4k [n(n − 1) . . . (n − k + 1)]2
w(r, t) = A + B r 2n + (at)k r 2n−2k ,
k!
k=1

w(r, t) = A + B 4at ln r + r 2 ln r − r 2 ,

B r2
w(r, t) = A + exp − ,
t 4at
Z ζ
dz r2
w(r, t) = A + B e−z , ζ = ,
1 z 4at
w(r, t) = A + B exp(−aµ2 t)J0 (µr),
w(r, t) = A + B exp(−aµ2 t)Y0 (µr),
2
B r + µ2 µr
w(r, t) = A + exp − I0 ,
t 4t 2t
2
B r + µ2 µr
w(r, t) = A + exp − K0 ,
t 4t 2t
where n is an arbitrary positive integer, J0 (z) and Y0 (z) are Bessel functions, and I0 (z)
and K0 (z) are modified Bessel functions.
Suppose w = w(r, t) is a solution of the original equation. Then the functions
w1 = Aw(±λr, λ2 t + C),

A βr 2 r γ + λt
w2 = exp − w ± , , λδ − βγ = 1,
δ + βt 4a(δ + βt) δ + βt δ + βt
where A, C, β, δ, and λ are arbitrary constants, are also solutions of this equation. The
second formula usually may be encountered with β = 1, γ = −1, and δ = λ = 0.
◮ Particular solutions in the form of an infinite series.

A solution containing an arbitrary function of the space variable:
∞
X (at)n n d2 1 d
w(r, t) = f (r) + L [f (r)], L≡ 2
+ ,
n! dr r dr
n=1
where f (r) is any infinitely differentiable function. This solution satisfies the initial con-
dition w(r, 0) = f (r). The sum is finite if f (r) is a polynomial that contains only even
powers.
A solution containing an arbitrary function of time:
∞
X 1 (n)
w(r, t) = g(t) + r 2n gt (t),
(4a)n (n!)2
n=1
where g(t) is any infinitely differentiable function. This solution is bounded at r = 0 and
possesses the properties
w(0, t) = g(t), ∂r w(0, t) = 0.
◮ Domain: 0 ≤ r ≤ R. First boundary value problem.

w = f (r) at t = 0 (initial condition),

w = g(t) at r = R (boundary condition),
|w| =
6 ∞ at r = 0 (boundedness condition).
Solution:
Z R Z t
w(r, t) = f (ξ)G(r, ξ, t) dξ − a g(τ )Λ(r, t − τ ) dτ.
0 0
Here,
∞
X
2ξ r ξ aµ2n t ∂
G(r, ξ, t) = 2 2 J 0 µ n J 0 µ n exp − 2
, Λ(r, t) = G(r, ξ, t) ,
n=1
R J1 (µn ) R R R ∂ξ ξ=R
where the µn are positive zeros of the Bessel function, J0 (µ) = 0. Below are the numerical
values of the first ten roots:
µ1 = 2.4048, µ2 = 5.5201, µ3 = 8.6537, µ4 = 11.7915, µ5 = 14.9309,
µ6 = 18.0711, µ7 = 21.2116, µ8 = 24.3525, µ9 = 27.4935, µ10 = 30.6346.
The zeroes of the Bessel function J0 (µ) may be approximated by the formula
µn = 2.4 + 3.13(n − 1) (n = 1, 2, 3, . . .),
which is accurate within 0.3%. As n → ∞, we have µn+1 − µn → π.

Example 3.11. The initial temperature of the cylinder is uniform, f (r) = w0 , and its lateral
surface is maintained all the time at a constant temperature, g(t) = wR .
Solution:
X∞
w(r, t) − wR 2 2 at r
= exp −µn 2 J0 µn .
w0 − wR µ J (µ )
n=1 n 1 n
R R
◮ Domain: 0 ≤ r ≤ R. Second boundary value problem.


∂r w = g(t) at r = R (boundary condition),
|w| =
Solution:
Z R Z t
w(r, t) = f (ξ)G(r, ξ, t) dξ + a g(τ )G(r, R, t − τ ) dτ.
0 0
Here,
∞
2 2 X ξ µn r µn ξ aµ2n t
G(r, ξ, t) = 2 ξ + 2 J0 J0 exp − 2 ,
R R n=1 J02 (µn ) R R R
where the µn are positive zeros of the first-order Bessel function, J1 (µ) = 0. Below are the
numerical values of the first ten roots:
µ1 = 3.8317, µ2 = 7.0156, µ3 = 10.1735, µ4 = 13.3237, µ5 = 16.4706,
µ6 = 19.6159, µ7 = 22.7601, µ8 = 25.9037, µ9 = 29.0468, µ10 = 32.1897.
As n → ∞, we have µn+1 − µn → π.
Example 3.12. The initial temperature of the cylinder is uniform, f (r) = w0 . The lateral surface
is maintained at constant thermal flux, g(t) = gR .
Solution:
X ∞
at 1 r2 2 2 at r
w(r, t) = w0 + gR R 2 2 − + − exp −µ n 2 J 0 µ n .
R 4 2R2 n=1 µ2n J0 (µn ) R R
◮ Domain: 0 ≤ r ≤ R. Third boundary value problem.

∂r w + kw = g(t) at r = R (boundary condition),
|w| =
Solution:
Z R Z t
0 0
Here,
∞
2 X µ2n ξ µn r µn ξ aµ2n t
G(r, ξ, t) = 2 J0 J0 exp − 2 ,
R
n=1
(k2 R2 + µ2n )J02 (µn ) R R R
where the µn are positive roots of the transcendental equation

µJ1 (µ) − kRJ0 (µ) = 0.
The numerical values of the first six roots µn can be found in Carslaw and Jaeger (1984).
Example 3.13. The initial temperature of the cylinder is uniform, f (r) = w0 . The temperature
of the environment is also uniform and is equal to wR , which corresponds to g(t) = kwR .
Solution:
X∞
w(r, t) − w0 aµ2n t µn r 2kR
=1− An exp − 2 J0 , An = 2 2 .
wR − w0 n=1
R R (k R + µ2n )J0 (µn )
⊙ Literature: A. V. Lykov (1967), B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980), H. S. Carslaw

and J. C. Jaeger (1984).
◮ Domain: R1 ≤ r ≤ R2 . First boundary value problem.


w = g1 (t) at r = R1 (boundary condition),
w = g2 (t) at r = R2 (boundary condition).
Solution:
Z R2 Z t Z t
w(r, t) = f (ξ)G(r, ξ, t) dξ + a g1 (τ )Λ1 (r, t − τ ) dτ − a g2 (τ )Λ2 (r, t − τ ) dτ.
R1 0 0
Here,
∞
π2 X µ2n J02 (sµn )ξ aµ2n t
G(r, ξ, t) = Ψn (r)Ψn (ξ) exp − 2 ,
2R12 n=1 J02 (µn ) − J02 (sµn ) R1

µn r µn r R2
Ψn (r) = Y0 (µn )J0 − J0 (µn )Y0 , s= ,
R1 R1 R1
∂ ∂

Λ1 (r, t) = G(r, ξ, t) , Λ2 (r, t) = G(r, ξ, t) ,
∂ξ ξ=R1 ∂ξ ξ=R2
where J0 (z) and Y0 (z) are Bessel functions; the µn are positive roots of the transcendental
equation
J0 (µ)Y0 (sµ) − J0 (sµ)Y0 (µ) = 0.
The numerical values of the first five roots µn = µn (s) range in the interval 1.4 ≤ s ≤ 4.0
and can be found in Carslaw and Jaeger (1984). See also Abramowitz and Stegun (1964).
Example 3.14. The initial temperature of the hollow cylinder is zero, and its interior and exterior
surfaces are held all the time at constant temperatures, g1 (t) = w1 and g2 (t) = w2 .
Solution:

1 R2 r
w(r, t) = w1 ln + w2 ln
ln s r R1
∞
X J0 (µn )[w2 J0 (µn ) − w1 J0 (sµn )]
aµ2n t
−π exp − Ψn (r).
n=1
J02 (µn ) − J02 (sµn ) R12
◮ Domain: R1 ≤ r ≤ R2 . Second boundary value problem.


∂r w = g1 (t) at r = R1 (boundary condition),
∂r w = g2 (t) at r = R2 (boundary condition).
Solution:
Z R2 Z t Z t
w(r, t) = f (ξ)G(r, ξ, t) dξ−a g1 (τ )G(r, R1 , t−τ ) dτ +a g2 (τ )G(r, R2 , t−τ ) dτ.
R1 0 0
Here,
∞
2ξ π2 X µ2n J12 (sµn )ξ aµ2n t
G(r, ξ, t) = 2 + Ψn (r)Ψn (ξ) exp − 2 ,
R2 − R12 2R12 n=1 J12 (µn ) − J12 (sµn ) R1

µn r µn r R2
Ψn (r) = Y1 (µn )J0 − J1 (µn )Y0 , s= ,
R1 R1 R1
where Jk (z) and Yk (z) are Bessel functions (k = 0, 1), and the µn are positive roots of the
transcendental equation
J1 (µ)Y1 (sµ) − J1 (sµ)Y1 (µ) = 0.
The numerical values of the first five roots µn = µn (s) can be found in Abramowitz and
Stegun (1964).
◮ Domain: R1 ≤ r ≤ R2 . Third boundary value problem.

∂r w − k1 w = g1 (t) at r = R1 (boundary condition),
∂r w + k2 w = g2 (t) at r = R2 (boundary condition).
Solution:
Z R2 Z t Z t
R1 0 0
Here,
∞
π 2 X λ2n
G(r, ξ, t) = [k2 J0 (λn R2 ) − λn J1 (λn R2 )]2 ξHn (r)Hn (ξ) exp(−aλ2n t),
2 Bn
n=1
where
Bn = (λ2n + k22)[k1 J0(λn R1) + λnJ1 (λnR1 )]2 − (λ2n + k12)[k2 J0(λn R2) − λnJ1 (λnR2 )]2,
Hn(r) = [k1 Y0 (λnR1) + λnY1 (λnR1)]J0 (λnr) − [k1 J0(λn R1) + λnJ1 (λnR1 )]Y0(λn r),
and the λn are positive roots of the transcendental equation
[k1 J0 (λR1 ) + λJ1 (λR1 )][k2 Y0 (λR2 ) − λY1 (λR2 )]

− [k2 J0 (λR2 ) − λJ1 (λR2 )][k1 Y0 (λR1 ) + λY1 (λR1 )] = 0.
◮ Domain: 0 ≤ r < ∞. Cauchy type problem.
This problem is encountered in the theory of diffusion wake behind a drop or a solid parti-
cle.
Given the initial condition
w = f (r) at t = 0,
the equation has the following bounded solution:

Z ∞ 2
1 ξ r + ξ2 rξ
w(r, t) = exp − I0 f (ξ) dξ,
2a 0 t 4at 2at
where I0 (ξ) is the modified Bessel function.

⊙ Literature: W. G. L. Sutton (1943), B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980), Yu. P. Gu-
palo, A. D. Polyanin, and Yu. S. Ryazantsev (1985).
◮ Domain: R ≤ r < ∞. Third boundary value problem.

∂r w − kw = g(t) at r = R (boundary condition),
|w| =
6 ∞ at r → ∞ (boundedness condition).
Solution:
Z ∞ Z t
w(r, t) = f (ξ)G(r, ξ, t) dξ − a g(τ )G(r, R, t − τ ) dτ,
R 0
where
Z ∞
G(r, ξ, t) = ξ exp −au2 t F (r, u)F (ξ, u)u du,
0
J0 (ur)[uY1 (uR) + kY0 (uR)] − Y0 (ur)[uJ1 (uR) + kJ0 (uR)]
F (r, u) = p .
[uJ1 (uR) + kJ0 (uR)]2 + [uY1 (uR) + kY0 (uR)]2
∂w
∂2w 1 ∂w

3.2.2 Equation of the Form =a + + Φ(r, t)
∂t ∂r 2 r ∂r
This equation is encountered in plane problems of heat conduction with heat release (the
function Φ is proportional to the amount of heat released per unit time in the volume under
consideration). The equation describes one-dimensional unsteady thermal processes having
axial symmetry.


|w| =
Solution:
Z R Z t Z tZ R
w(r, t) = f (ξ)G(r, ξ, t) dξ−a g(τ )Λ(r, t−τ ) dτ + Φ(ξ, τ )G(r, ξ, t−τ ) dξ dτ,
0 0 0 0
where
∞
X
2ξ r ξ aµ2n t
G(r, ξ, t) = J0 µn J0 µn exp − 2 ,
n=1
R2 J12 (µn ) R R R
∂

Λ(r, t) = G(r, ξ, t) .
∂ξ ξ=R
Here, the µn are positive zeros of the Bessel function, J0 (µn ) = 0. The numerical values
of the first ten roots µn are given in Section 3.2.1 (see the first boundary value problem for
0 ≤ r ≤ R).


|w| =
Solution:
Z R Z t
w(r, t) = f (ξ)G(r, ξ, t) dξ + a g(τ )G(r, R, t − τ ) dτ
0 0
Z tZ R
+ Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ.
0 0
Here,
∞
2 2 X ξ µn r µn ξ aµ2n t
G(r, ξ, t) = 2 ξ + 2 J0 J0 exp − 2 ,
R R J 2 (µn )
n=1 0
R R R
where the µn are positive zeros of the first-order Bessel function, J1 (µ) = 0. The numerical
values of the first ten roots µn can be found in Section 3.2.1 (see the second boundary value
problem for 0 ≤ r ≤ R).


|w| =
Solution:
Z R Z t
0 0
Z tZ R
+ Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ.
0 0
Here,
∞
2 X µ2n ξ µn r µn ξ aµ2n t
G(r, ξ, t) = 2 J0 J0 exp − 2 ,
R n=1 (k2 R2 + µ2n )J02 (µn ) R R R
µJ1 (µ) − kRJ0 (µ) = 0.


Solution:
Z R2 Z tZ R2
w(r, t) = f (ξ)G(r, ξ, t) dξ + Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ
R1 0 R1
Z t Z t
+a g1 (τ )Λ1 (r, t − τ ) dτ − a g2 (τ )Λ2 (r, t − τ ) dτ.
0 0
Here,
∞
X aµ2 t ∂G ∂G
G(r, ξ, t) = AnξΨn(r)Ψn(ξ) exp − n2 , Λ1(r, t) = , Λ (r, t) = ,
2
n=1
R1

π 2µ2nJ02(sµn) µn r µn r R2
An = 2
2 2
, Ψn(r) = Y0(µn)J0 −J0(µn)Y0 , s= ,
2R1 J0 (µn)−J0 (sµn) R1 R1 R1
where J0 (z) and Y0 (z) are Bessel functions, and the µn are positive roots of the transcen-
dental equation
J0 (µ)Y0 (sµ) − J0 (sµ)Y0 (µ) = 0.
The numerical values of the first five roots µn = µn (s) can be found in Carslaw and Jaeger
(1984).

Solution:
Z R2 Z tZ R2
R1 0 R1
Z t Z t
−a g1 (τ )G(r, R1 , t − τ ) dτ + a g2 (τ )G(r, R2 , t − τ ) dτ.
0 0
Here,
∞
2ξ π2 X µ2n J12 (sµn )ξ aµ2n t
G(r, ξ, t) = 2 + Ψn (r)Ψn (ξ) exp − 2 ,
R2 − R12 2R12 n=1 J12 (µn ) − J12 (sµn ) R1

µn r µn r R2
Ψn (r) = Y1 (µn )J0 − J1 (µn )Y0 , s= ,
R1 R1 R1
where Jk (z) and Yk (z) are Bessel functions of order k = 0, 1, and the µn are positive roots
of the transcendental equation
J1 (µ)Y1 (sµ) − J1 (sµ)Y1 (µ) = 0.
Stegun (1964).

The solution is given by the formula from Section 3.2.1 (see the third boundary value
problem for R1 ≤ r ≤ R2 ) with the additional term
Z t Z R2
Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ,
0 R1
The bounded solution of this equation subject to the initial condition
w = f (r) at t = 0
is given by the relations

Z ∞ Z tZ ∞
w(r, t) = G(r, ξ, t)f (ξ) dξ + G(r, ξ, t − τ )Φ(ξ, τ ) dξ dτ,
0 0 0
2
ξ r + ξ2 rξ
G(r, ξ, t) = exp − I0 ,
2at 4at 2at
where I0 (z) is the modified Bessel function.

⊙ Literature: W. G. L. Sutton (1943), B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
◮ Domain: R ≤ r < ∞. Third boundary value problem.

∂r w − kw = g(t) at r = R (boundary condition),
|w| =
6 ∞ at r → ∞ (boundedness condition).
The solution is given by the formula from Section 3.2.1 (see the third boundary value
problem for R ≤ r < ∞) with the additional term
Z tZ ∞
0 R
∂w
∂2w 2 ∂w

∂t ∂r 2 r ∂r
This is a sourceless heat equation that describes unsteady heat processes with central sym-
metry. It is often represented in the equivalent form

∂w a ∂ 2 ∂w
= 2 r .
∂t r ∂r ∂r
A similar equation is used for the analysis of the corresponding one-dimensional unsteady
diffusion processes.

1
w(r) = A + B ,
r
w(r, t) = A + B(r 2 + 6at),
w(r, t) = A + B(r 4 + 20atr 2 + 60a2 t2 ),
Xn
2n (2n + 1)(2n) . . . (2n − 2k + 2) k 2n−2k
w(r, t) = A + B r + (at) r ,
k!
k=1
t
w(r, t) = A + 2aB + Br,
r
w(r, t) = Ar −1 exp(aµ2 t ± µr) + B,

B r2
w(r, t) = A + 3/2 exp − ,
t 4at

B r2
w(r, t) = A + √ exp − ,
r t 4at
w(r, t) = Ar −1 exp(−aµ2 t) cos(µr) + B,
w(r, t) = Ar −1 exp(−aµ2 t) sin(µr) + B,
w(r, t) = Ar −1 exp(−µr) cos(µr − 2aµ2 t) + B,
w(r, t) = Ar −1 exp(−µr) sin(µr − 2aµ2 t) + B,

A r
w(r, t) = erf √ + B,
r 2 at

A r
w(r, t) = erfc √ + B,
r 2 at
where n is an arbitrary positive integer.
◮ Reduction to a constant coefficient equation. Some formulas.

1◦ . The substitution u(r, t) = rw(r, t) brings the original equation with variable coeffi-
cients to the constant coefficient equation
∂u ∂2u
=a 2,
∂t ∂r
which is discussed in Section 3.1.1 in detail.
2◦ . Suppose w = w(r, t) is a solution of the original equation. Then the functions
w1 = Aw(±λr, λ2 t + C),

A βr 2 r γ + λt
w2 = exp − w ± , , λδ − βγ = 1,
|δ + βt|3/2 4a(δ + βt) δ + βt δ + βt
where A, C, β, δ, and λ are arbitrary constants, are also solutions of this equation. The
second formula may usually be encountered with β = 1, γ = −1, and δ = λ = 0.
◮ Infinite series particular solutions.

∞
X (at)n n d2 2 d
w(r, t) = f (r) + L f (r) , L≡ 2
+ ,
n! dr r dr
n=1
where f (r) is any infinitely differentiable function. This solution satisfies the initial con-
dition w(r, 0) = f (r). The sum is finite if f (r) is a polynomial that contains only even
powers.
∞
X 1 (n)
w(r, t) = g(t) + r 2n gt (t),
n=1
an (2n + 1)!
where g(t) is any infinitely differentiable function. This solution is bounded at r = 0 and
w(0, t) = g(t), ∂r w(0, t) = 0.

|w| =
Solution:
∞
2 X1 πnr aπ 2 n2 t
w(r, t) = sin exp − Mn (t),
R n=1 r R R2
where
Z R Z t 2 2
πnξ n aπ n τ
Mn (t) = ξf (ξ) sin dξ − (−1) aπn g(τ ) exp dτ.
0 R 0 R2
Remark 3.3. Using the relation [see Prudnikov, Brychkov, and Marichev (1986)]
∞
X sin nz z
(−1)n−1 = (−π < z < π),
n=1
n 2
we rewrite the solution in the form
∞
2 X πn
w(r, t) = g(t) + sin(λn r) exp(−aλ2n t)Hn (t), λn = ,
Rr n=1 R
where
Z R Z t
R n 2 n
Hn(t)= ξf (ξ) sin(λnξ) dξ+(−1) g(t) exp(aλnt)−(−1) aπn g(τ ) exp(aλ2nτ ) dτ
0 λn 0
Z R Z
R R t ′
= ξf (ξ) sin(λnξ) dξ+(−1)n g(0)+(−1)n g (τ ) exp(aλ2nτ ) dτ.
0 λn λn 0 τ
Example 3.15. The initial temperature is uniform, f (r) = w0 , and the surface of the sphere is
maintained at constant temperature, g(t) = wR .
Solution:
∞
w(r, t) − wR 2R X (−1)n+1 πnr aπ 2 n2 t
= sin exp − .
w0 − wR πr n=1 n R R2
The average temperature w depends on time t as follows:

∞ Z
w − wR 6 X 1 aπ 2 n2 t 1
= 2 exp − , w= w dv,
w0 − wR π n=1 n2 R2 V v
where V is the volume of the sphere of radius R.


|w| =
Solution:
Z R Z t
w(r, t) = f (ξ)G(r, ξ, t) dξ + a g(τ )G(r, R, t − τ ) dτ,
0 0
where
∞
3ξ 2 2ξ X µ2n + 1 µn r µn ξ aµ2n t
G(r, ξ, t) = 3 + sin sin exp − 2 .
R Rr n=1 µ2n R R R
Here, the µn are positive roots of the transcendental equation tan µ − µ = 0. The first five
roots are
µ1 = 4.4934, µ2 = 7.7253, µ3 = 10.9041, µ4 = 14.0662, µ5 = 17.2208.
Example 3.16. The initial temperature of the sphere is uniform, f (r) = w0 . The thermal flux at
the sphere surface is a maintained constant, g(t) = gR .
Solution:
∞
3at 5r2 − 3R2 X 2R 1 µn r aµ2n t
w(r, t) = w0 + gR R + − sin exp − 2 ,
R2 10R2 n=1 n
µ3 cos(µn ) r R R
⊙ Literature: A. V. Lykov (1967), A. V. Bitsadze and D. F. Kalinichenko (1985).


|w| =
Solution:
Z R Z t
0 0
Here,
∞
2ξ X µ2n + (kR − 1)2 µn r µn ξ aµ2n t
G(r, ξ, t) = sin sin exp − 2 ,
Rr n=1 µ2n + kR(kR − 1) R R R

µ cot µ + kR − 1 = 0.
Example 3.17. The initial temperature of the sphere is uniform, f (r) = w0 . The temperature of
the ambient medium is zero, g(t) = 0.
Solution:
∞
2kR2 w0 X sin µn [µ2n + (kR − 1)2 ] µn r aµ2n t
w(r, t) = sin exp − ,
r n=1
µ2n [µ2n + kR(kR − 1)] R R2

Solution:
Z R2 Z t Z t
w(r, t) = f (ξ)G(r, ξ, t) dξ + a g1 (τ )Λ1 (r, t − τ ) dτ − a g2 (τ )Λ2 (r, t − τ ) dτ,
R1 0 0
where
X∞
2ξ πn(r − R1 ) πn(ξ − R1 ) π 2 n2 at
G(r, ξ, t) = sin sin exp − ,
(R2 − R1 )r R2 − R1 R2 − R1 (R2 − R1 )2
n=1
∂ ∂

Λ1 (r, t) = G(r, ξ, t) , Λ2 (r, t) = G(r, ξ, t) .
Example 3.18. The initial temperature is zero. The temperatures of the interior and exterior
surfaces of the spherical layer are maintained constants, g1 (t) = w1 and g2 (t) = w2 .
Solution:
R1 w1 (r − R1 )(R2 w2 − R1 w1 )
w(r, t) = +
r r(R2 − R1 )
X∞ n

2 (−1) R2 w2 − R1 w1 πn(r − R1 ) π 2 n2 at
+ sin exp − .
r n=1 πn R2 − R1 (R2 − R1 )2


Solution:
Z R2 Z t Z t
R1 0 0
Here,
X∞
3ξ 2 2ξ (1 + R22 λ2n )Ψn (r)Ψn (ξ) exp(−aλ2n t)
G(r, ξ, t) = + ,
R23 − R13 (R2 − R1 )r λ2 R12 + R22 + R1 R2 (1 + R1 R2 λ2n )
n=1 n
Ψn (r) = sin[λn (r − R1 )] + R1 λn cos[λn (r − R1 )],
where the λn are positive roots of the transcendental equation
(λ2 R1 R2 + 1) tan[λ(R2 − R1 )] − λ(R2 − R1 ) = 0.



Solution:
Z R2 Z t Z t
R1 0 0
Here,
∞
2ξ X (b22+R22λ2n)Ψn(r)Ψn(ξ) exp(−aλ2nt)
G(r, ξ, t)= ,
r
n=1
(R2−R1)(b21+R12λ2n)(b22+R22λ2n)+(b1R2+b2R1)(b1b2+R1R2λ2n)
Ψn(r)=b1 sin[λn(r−R1)]+R1λn cos[λn(r−R1)], b1 =k1R1+1, b2 =k2R2−1,
(b1 b2 − R1 R2 λ2 ) sin[λ(R2 − R1 )] + λ(R1 b2 + R2 b1 ) cos[λ(R2 − R1 )] = 0.

w = f (r) at t = 0
has the form

Z ∞
1 (r − ξ)2 (r + ξ)2
w(r, t) = √ ξ exp − − exp − f (ξ) dξ.
2r πat 0 4at 4at
◮ Domain: R ≤ r < ∞. First boundary value problem.


w = g(t) at r = R (boundary condition).
Solution:
Z ∞
1 (r − ξ)2 (r + ξ − 2R)2
w(r, t) = √ ξf (ξ) exp − − exp − dξ
2r πat R 4at 4at
Z ∞
2R (r − R)2 r−R
+ √ g t− 2
exp(−τ 2 ) dτ, z= √ .
r π z 4aτ 2 at
Example 3.19. The temperature of the ambient medium is uniform at the initial instant t = 0
and the boundary of the domain is held at constant temperature, that is, f (r) = w0 and g(t) = wR .
Solution:
w − w0 R r−R
= erfc √ ,
wR − w0 r 2 at
Z ∞
2
where erfc z = √ exp(−ξ 2 ) dξ is the error function.
π z
∂w
∂2w 2 ∂w

∂t ∂r 2 r ∂r
This equation is encountered in heat conduction problems with heat release; the function Φ
is proportional to the amount of heat released per unit time in the volume under consider-
ation. The equation describes one-dimensional unsteady thermal processes having central
symmetry.
The substitution u(r, t) = rw(r, t) brings the original nonhomogeneous equation with
variable coefficients to the nonhomogeneous constant coefficient equation
∂u ∂2u
= a 2 + rΦ(r, t),
∂t ∂r
which is considered in Section 3.1.2.

w 6= ∞ at r = 0 (boundedness condition).
Solution:
Z R Z t Z tZ R
w(r, t) = f (ξ)G(r, ξ, t) dξ −a g(τ )Λ(r, t−τ ) dτ + Φ(ξ, τ )G(r, ξ, t−τ ) dξ dτ,
0 0 0 0
where
∞
2ξ X nπr nπξ an2 π 2 t ∂
G(r, ξ, t) = sin sin exp − , Λ(r, t) = G(r, ξ, t) .
Rr n=1 R R R2 ∂ξ ξ=R

Solution:
Z R Z t
0 0
Z tZ R
+ Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ,
0 0
where
∞
3ξ 2 2ξ X µ2n + 1 µn r µn ξ aµ2n t
G(r, ξ, t) = 3 + sin sin exp − 2 .
R Rr n=1 µ2n R R R
Here, the µn are positive roots of the transcendental equation tan µ − µ = 0. The values of
the first five roots µn can be found in Section 3.2.3 (see the second boundary value problem
for 0 ≤ r ≤ R).
⊙ Literature: A. V. Lykov (1967), A. V. Bitsadze and D. F. Kalinichenko (1985).


Solution:
Z R Z t
0 0
Z tZ R
+ Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ.
0 0
Here,
∞
2ξ X µ2n + (kR − 1)2 µn r µn ξ aµ2n t
G(r, ξ, t) = sin sin exp − 2 ,
Rr µ2n + kR(kR − 1) R R R
n=1
where the µn are positive roots of the transcendental equation µ cot µ + kR − 1 = 0.



Solution:
Z R2 Z tZ R2
R1 0 R1
Z t Z t
+a g1 (τ )Λ1 (r, t − τ ) dτ − a g2 (τ )Λ2 (r, t − τ ) dτ,
0 0
where
X∞
2ξ πn(r−R1 ) πn(ξ −R1 ) π 2 n2 at
G(r,ξ,t)= sin sin exp − ,
(R2 −R1 )r n=1 R2 −R1 R2 −R1 (R2 −R1 )2
∂ ∂

Λ1 (r, t) = G(r, ξ, t) , Λ2 (r, t) = G(r, ξ, t) .

Solution:
Z R2 Z tZ R2
R1 0 R1
Z t Z t
−a g1 (τ )G(r, R1 , t − τ ) dτ + a g2 (τ )G(r, R2 , t − τ ) dτ,
0 0
where the function G(r, ξ, t) is the same as in Section 3.2.3 (see the second boundary value
problem for R1 ≤ r ≤ R2 ).

The solution is given by the formula of Section 3.2.3 (see the third boundary value
problem for R1 ≤ r ≤ R2 ) with the additional term
Z tZ R2
0 R1

w = f (r) at t = 0
is given by
Z ∞ Z tZ ∞
w(r, t) = f (ξ)G(r, ξ, t) dξ + Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ,
0 0 0

ξ (r − ξ)2 (r + ξ)2
G(r, ξ, t) = √ exp − − exp − .
2r πat 4at 4at

Solution:
Z ∞ Z tZ ∞ Z t
w(r,t)= f (ξ)G(r,ξ,t)dξ+ Φ(ξ,τ )G(r,ξ,t−τ )dξ dτ +a g(τ )Λ(r,t−τ )dτ,
R 0 R 0
where

ξ (r − ξ)2 (r + ξ − 2R)2
G(r, ξ, t) = √ exp − − exp − ,
2r πat 4at 4at

∂
Λ(r, t) = G(r, ξ, t) .
∂ξ ξ=R
∂w ∂2w 1 − 2β ∂w
3.2.5 Equation of the Form = +
∂t ∂x2 x ∂x
This dimensionless equation is encountered in problems of the diffusion boundary layer.
For β = 0, β = 12 , or β = − 12 , see the equations of Sections 3.2.1, 3.1.1, or 3.2.3, respec-
tively.
w(x) = A + Bx2β ,
w(x, t) = A + 4(1 − β)Bt + Bx2 ,
w(x, t) = A + 16(2 − β)(1 − β)Bt2 + 8(2 − β)Btx2 + Bx4 ,
n
X 4p
w(x, t) = x2n + sn,p sn−β,p tp x2(n−p) , sq,p = q(q − 1) . . . (q − p + 1),
p!
p=1
w(x, t) = A + 4(1 + β)Btx2β + Bx2β+2 ,

β−1 x2
w(x, t) = A + Bt exp − ,
4t

x2β x2
w(x, t) = A + B β+1 exp − ,
t 4t

x2
w(x, t) = A + Bγ β, ,
4t
w(x, t) = A + B exp(−µ2 t)xβ Jβ (µx),
w(x, t) = A + B exp(−µ2 t)xβ Yβ (µx),

xβ x2 + µ 2 µx
w(x, t) = A + B exp − Iβ ,
t 4t 2t

xβ x2 + µ 2 µx
w(x, t) = A + B exp − I−β ,
t 4t 2t

xβ x2 + µ 2 µx
w(x, t) = A + B exp − Kβ ,
t 4t 2t
where n is an arbitrary positive integer, γ(β, z) is the incomplete gamma function, Jβ (z)
and Yβ (z) are Bessel functions, and Iβ (z) and Kβ (z) are modified Bessel functions.
⊙ Literature: W. G. L. Sutton (1943), A. D. Polyanin (2001).
◮ Infinite series solutions.

X∞
1 n n d2 1 − 2β d
w(x, t) = f (x) + t L [f (x)], L≡ + ,
n=1
n! dx2 x dx
dition w(x, 0) = f (x). The sum is finite if f (x) is a polynomial that contains only even
powers.
∞
X 1 (n)
w(x, t) = g(t) + x2n gt (t),
n=1
4n n! (1 − β)(2 − β) . . . (n − β)
where g(t) is any infinitely differentiable function. This solution is bounded at x = 0 and
w(0, t) = g(t), ∂x w(0, t) = 0.
◮ Formulas and transformations for constructing particular solutions.

Suppose w = w(x, t) is a solution of the original equation. Then the functions
w1 = Aw(±λx, λ2 t + C),

β−1 bx2 x c + kt
w2 = A|a + bt| exp − w ± , , ak − bc = 1,
4(a + bt) a + bt a + bt
where A, C, a, b, and c are arbitrary constants, are also solutions of this equation. The
second formula usually may be encountered with a = k = 0, b = 1, and c = −1.
The substitution w = x2β u(x, t) brings the equation with parameter β to an equation
of the same type with parameter −β:
∂u ∂2u 1 + 2β ∂u
= 2
+ .
∂t ∂x x ∂x


Solution for 0 < β < 1:

Z
xβ ∞ 1−β x2 + ξ 2 ξx
w(x, t) = f (ξ)ξ exp − Iβ dξ
2t 0 4t 2t
Z t
x2β x2 dτ
+ 2β+1 g(τ ) exp − .
2 Γ(β + 1) 0 4(t − τ ) (t − τ )1+β
Example 3.20. For f (x) = w0 and g(t) = w1 , where f (x) = const and g(t) = const, we have
Z z
(w0 − w1 ) x2
w= γ β, + w1 , γ(β, z) = ξ β−1 e−ξ dξ.
Γ(β) 4t 0
Here, γ(β, z) is the incomplete gamma function and Γ(β) = γ(β, ∞) is the gamma function.
⊙ Literature: W. G. L. Sutton (1943).


x1−2β ∂x w = g(t) at x = 0 (boundary condition).

Z
xβ ∞ 1−β x2 + ξ 2 ξx
w(x, t) = f (ξ)ξ exp − I−β dξ
2t 0 4t 2t
Z t
22β−1 x2 dτ
− g(τ ) exp − .
Γ(1 − β) 0 4(t − τ ) (t − τ )1−β

w = 0 at t = 0 (initial condition),

x1−2β ∂x w + k(w0 − w) = 0 at x = 0 (boundary condition).
Z t
22β−1 k x2 dτ
w(x, t) = w0 ϕ(τ ) exp − ,
Γ(1 − β) 0 4(t − τ ) (t − τ )1−β
where the function ϕ(t) is given as the series
X∞
(−µtβ )n 22β−1 kΓ(β)
ϕ(t) = , µ= ,
n=0
Γ(nβ + 1) Γ(1 − β)
which is convergent for any t.

⊙ Literature: W. G. L. Sutton (1943).
∂w ∂2w 1 − 2β ∂w
3.2.6 Equation of the Form = + + Φ(x, t)
∂t ∂x2 x ∂x
This equation is encountered in problems of a diffusion boundary layer with sources/sinks
of substance. For β = 0, β = 12 , or β = − 12 , see the equations of Sections 3.2.2, 3.1.2,
or 3.2.4, respectively.

Z
xβ ∞ 1−β x2 + ξ 2 ξx
w(x, t) = f (ξ)ξ exp − Iβ dξ
2t 0 4t 2t
Z t
x2β x2 dτ
+ 2β+1 g(τ ) exp −
2 Γ(β + 1) 0 4(t − τ ) (t − τ )1+β
Z tZ ∞
1 xβ ξ 1−β x2 + ξ 2 ξx
+ Φ(ξ, τ ) exp − Iβ dξ dτ.
2 0 0 t−τ 4(t − τ ) 2(t − τ )


1−2β

x ∂x w = g(t) at x = 0 (boundary condition).

Z
xβ ∞ 1−β x2 + ξ 2 ξx
w(x, t) = f (ξ)ξ exp − I−β dξ
2t 0 4t 2t
Z t
22β−1 x2 dτ
− g(τ ) exp −
Γ(1 − β) 0 4(t − τ ) (t − τ )1−β
Z tZ ∞
1 xβ ξ 1−β x2 + ξ 2 ξx
+ Φ(ξ, τ ) exp − I−β dξ dτ.
2 0 0 t−τ 4(t − τ ) 2(t − τ )
⊙ Reference for Section 3.2.6: W. G. L. Sutton (1943).
3.3 Equations Containing Power Functions and Arbitrary

Parameters
∂w ∂2w
3.3.1 Equations of the Form =a + f (x, t)w
∂t ∂x2
◮ The function f depends on the space coordinate x alone.
Such equations are encountered in problems of heat and mass transfer with heat release (or
volume chemical reaction). The one-dimensional Schrödinger equation can be reduced to
this form by the change of variable t → −iht [the function −f (x) describes the potential
against the space coordinate; see Section 3.9.2].
∂w ∂2w
1. =a + (bx + c)w.
∂t ∂x2
This equation is a special case of equation 3.8.9 with s(x) = 1, p(x) = a, q(x) = −bx − c,
and Φ(x, t) = 0. Also, it is a special case of equation 3.8.1.6 with f (t) = b and g(t) = c.
1◦ . Particular solutions (A and µ are arbitrary constants):

w(x, t) = A exp btx + 13 ab2 t3 + ct ,

w(x, t) = A(x + abt2 ) exp btx + 13 ab2 t3 + ct ,

w(x, t) = A exp x(bt + µ) + 13 ab2 t3 + abµt2 + (aµ2 + c)t ,

p 2
w(x, t) = A exp(−µt) ξ J1/3 √ ξ 3/2 , ξ = bx + c + µ,
3b a

p 2 3/2
w(x, t) = A exp(−µt) ξ Y1/3 √ ξ , ξ = bx + c + µ,
3b a
where J1/3 (z) and Y1/3 (z) are Bessel functions of the first and second kind of order 1/3.
3.3. Equations Containing Power Functions and Arbitrary Parameters 313

w(x, t) = u(z, t) exp btx + 13 ab2 t3 + ct , z = x + abt2
leads to a constant coefficient equation, ∂t u = a∂zz u, which is considered in Section 3.1.1.

3◦ . Domain: −∞ < x < ∞. Cauchy problem.
w = f (x) at t = 0.
Solution:
Z ∞
1 (x + abt2 − ξ)2
w(x, t) = √ exp btx + 13 ab2 t3 + ct exp − f (ξ) dξ.
2 πat −∞ 4at
⊙ Literature: A. D. Polyanin, A. V. Vyazmin, A. I. Zhurov, and D. A. Kazenin (1998), see also W. Miller, Jr.
(1977).
∂w ∂2w
2. =a − (bx2 + c)w, b > 0.
∂t ∂x2
This equation is a special case of equation 3.8.9 with s(x) = 1, p(x) = a, q(x) = bx2 + c,
and Φ(x, t) = 0. In addition, it is a special case of equation 3.8.1.7 with f (t) = −c.
√
√ b 2
w(x, t) = A exp ab − c t + √ x ,
2 a
√
√ b 2
w(x, t) = A exp − ab + c t − √ x ,
2 a
√ r
b 2 c−µ 1 1 b 2
w(x, t) = A exp −µt − √ x Φ √ + , ; x ,
2 a 4 ab 4 2 a
√ r
b 2 c−µ 3 3 b 2
w(x, t) = A exp −µt − √ x xΦ √ + , ; x ,
2 a 4 ab 4 2 a
P∞ α(α + 1) . . . (α + m − 1) z m
where Φ(α, β; z) = 1 + is the degenerate hypergeo-
m=1 β(β + 1) . . . (β + m − 1) m!
metric function.
2◦ . In quantum mechanics the following particular solution is encountered:
√ 1 1 2
w(x, t) = e−[c+ ab (2n+1)]t
ψn (ξ), ψn (ξ) = √ n e− 2 ξ Hn (ξ),
π 1/4 2 n! x0
x a 1/4
ξ= , x0 = ,
x0 b
2 dn
−ξ 2 are the Hermite polynomials, n = 0, 1, 2, . . . These
where Hn (ξ) = (−1)n eξ dξ n e
solutions satisfy the normalization condition

Z ∞
|ψn (ξ)|2 dξ = 1.
−∞
3◦ . The transformation (A is any number)

√
b 2 √
w(x, t) = u(z, τ ) exp √ x + ( ab − c)t ,
2 a
√ √ √
a
z = x exp 2 ab t , τ = √ exp 4 ab t + A,
4 b
leads to the constant coefficient equation ∂τ u = ∂zz u, which is considered in Section 3.1.1.
⊙ Literature: A. D. Polyanin, A. V. Vyazmin, A. I. Zhurov, and D. A. Kazenin (1998), see also W. Miller, Jr.
(1977).
∂w ∂2w
3. =a + (bx2 − c)w, b > 0.
∂t ∂x2
This is a special case of equation 3.8.9 with s(x) = 1, p(x) = a, q(x) = c − bx2 , and
Φ(x, t) = 0. The transformation
√
1 b 2 √
w(x, t) = q √ exp 2√a x tan 2 ab t − ct u(z, τ ),
cos 2 ab t
√ √
x a
z= √ , τ = √ tan 2 ab t ,
cos 2 ab t 2 b
∂w ∂2w
4. =a + (bx2 + cx + k)w.
∂t ∂x2
The substitution z = x + c/(2b) leads to an equation of the form 3.3.1.2 (for b < 0) or
3.3.1.3 (for b > 0).
∂w ∂2w
5. =a + b + cx−2 w.
∂t ∂x2
1◦ . Particular solutions
2 )t √ 1 c
w(x, t) = e(b−aµ x AJν (µx) + BYν (µx) , ν2 = − ,
4 a
where Jν (z) and Yν (z) are Bessel functions; A, B, and µ are arbitrary constants.
2◦ . Domain: 0 ≤ x < ∞. First boundary value problem.

w=0 at x = 0 (boundary condition).
Solution:
Z ∞p
ebt x2 + ξ 2 ξx 1 c
w(x, t) = xξ exp − Iν f (ξ) dξ, ν2 = − ,
2at 0 4at 2at 4 a
where − 34 a < c < 14 a.

w(x, t) = ebt xk u(x, τ ), τ = at,
where k is a root of the quadratic equation ak2 − ak + c = 0 , leads to an equation of the

form 3.2.5:
∂u ∂2u 2k ∂u
= 2
+ .
∂τ ∂x x ∂x
See also W. Miller, Jr. (1977).
∂w ∂2w
6. =a + −bx2 + c + kx−2 w, b > 0.
∂t ∂x2
This is a special case of equation 3.8.1.2 with f (x) = −bx2 − c+ kx−2. The transformation
(A is any number)
√
b 2 √
w(x, t) = u(z, τ ) exp √ x + ( ab + c)t ,
2 a
√ 1 √
z = x exp 2 ab t , τ = √ exp 4 ab t + A,
4 ab
∂u ∂2u k
= + z −2 u.
∂τ ∂z 2 a
∂w ∂2w
7. =a + (bx2 − c + kx−2 )w, b > 0.
∂t ∂x2
This is a special case of equation 3.8.1.2 with f (x) = bx2 − c + kx−2 . The transformation
√
1 b 2 √
w(x, t) = q √ exp 2√a x tan 2 ab t − ct u(z, τ ),
cos 2 ab t
√ √
x a
z= √ , τ = √ tan 2 ab t ,
cos 2 ab t 2 b
∂u ∂2u k
= 2
+ z −2 u.
∂τ ∂z a
◮ The function f depends on time t alone.
∂w ∂2w
8. =a + bt + c w.
∂t ∂x2
This is a special case of equation 3.8.1.1 with f (t) = bt + c.
1◦ . Particular solutions (A, B, and µ are arbitrary constants):

w(x, t) = (Ax + B) exp 12 bt2 + ct ,

w(x, t) = A(x2 + 2at) exp 12 bt2 + ct ,

w(x, t) = A exp µx + 12 bt2 + (c + aµ2 )t ,

w(x, t) = A exp 12 bt2 + (c − aµ2 )t cos(µx),

w(x, t) = A exp 12 bt2 + (c − aµ2 )t sin(µx).

2◦ . The substitution w(x, t) = u(x, t) exp 12 bt2 + ct leads to a constant coefficient equa-
tion, ∂t u = a∂xx u, which is considered in Section 3.1.1.
∂w ∂2w
9. =a + btk w.
∂t ∂x2
This is a special case of equation 3.8.1.1 with f (t) = btk .
b
w(x, t) = (Ax + B) exp tk+1 ,
k+1
b
w(x, t) = A(x2 + 2at) exp tk+1 ,
k+1
b k+1
w(x, t) = A exp µx + aµ2 t + t ,
k+1
b
w(x, t) = A exp tk+1 − aµ2 t cos(µx),
k+1
b
w(x, t) = A exp tk+1 − aµ2 t sin(µx).
k+1
b k+1

2◦ . The substitution w(x, t) = u(x, t) exp k+1 t leads to a constant coefficient equa-
tion, ∂t u = a ∂xx u, which is considered in Section 3.1.1.
◮ The function f depends on both x and t.
∂w ∂ 2w
10. =a + (bx + ct + d)w.
∂t ∂x2
This is a special case of equation 3.8.1.6 with f (t) = b and g(t) = ct + d.

w(x, t) = A exp btx + 13 ab2 t3 + 12 ct2 + dt ,

w(x, t) = A(x + abt2 ) exp btx + 13 ab2 t3 + 12 ct2 + dt ,

p 2
1 2 √ 3/2
w(x, t) = A exp( 2 ct − µt) ξ J1/3 ξ , ξ = bx + d + µ,
3b a

p 2
w(x, t) = A exp( 12 ct2 − µt) ξ Y1/3 √ ξ 3/2 , ξ = bx + d + µ,
3b a
where J1/3 (ξ) and Y1/3 (ξ) are Bessel functions of the first and second kind.

w(x, t) = u(z, t) exp btx + 13 ab2 t3 + 12 ct2 + dt , z = x + abt2
∂w ∂ 2w
11. =a + x(bt + c)w.
∂t ∂x2

w(x, t) = A exp x 12 bt2 + ct + a 20 1 2 5
b t + 14 bct4 + 13 c2 t3 ,

w(x, t) = A x + a 13 bt3 + ct2 exp x 12 bt2 + ct + aφ(t) ,

w(x, t) = A exp x 12 bt2 + ct + µ + aµ 13 bt3 + ct2 + µt + aφ(t) ,
1 2 5
where φ(t) = 20 b t + 14 bct4 + 13 c2 t3 .
1 2
1 2 5 1 4 1 2 3
1 3 2

w(x, t) = u(z, t) exp x 2 bt +ct +a 20 b t + 4 bct + 3 c t , z = x+a 3 bt +ct
∂w ∂ 2w
12. =a + (bxt + cx + dt + e)w.
∂t ∂x2
This is a special case of equation 3.8.1.6 with f (t) = bt + c and g(t) = dt + e.
1◦ . Particular solution:
1 2
1 2 5

w(x, t) = exp x 2 bt + ct + a 20 b t + 14 bct4 + 13 c2 t3 + 12 dt2 + et .
1 2
1 2 5

w(x, t) = u(z, t) exp x 2 bt + ct + a 20 b t + 14 bct4 + 13 c2 t3 + 12 dt2 + et ,
z = x + a(bt2 + 2ct)
∂w ∂ 2w
13. =a + (−bx2 + ct + d)w.
∂t ∂x2
This is a special case of equation 3.8.1.7 with f (t) = ct + d.
1◦ . Particular solutions (A is an arbitrary constant):
r
1 b 2 1 2 √
w(x, t) = A exp x + ct + ab + d t ,
2 a 2
r
1 b 2 1 2 √
w(x, t) = Ax exp x + ct + 3 ab + d t .
2 a 2
2◦ . The transformation (A is any number)

r
1 b 2 1 2 √
w(x, t) = u(z, τ ) exp x + ct + ab + d t ,
2 a 2
r
√ 1 b √
z = x exp 2 ab t , τ = exp 4 ab t + A
4 a
∂w ∂ 2w
14. =a + x(−bx + ct + d)w.
∂t ∂x2
This is a special case of equation 3.8.1.8 with f (t) = ct + d.
∂w ∂ 2w
15. =a + bxtk w.
∂t ∂x2
This is a special case of equation 3.8.1.3 with f (t) = btk .

b k+1 ab2 2k+3
w(x, t) = A exp xt + t ,
k+1 (k+1)2 (2k+3)

2ab b ab2
w(x, t) = A x+ tk+2 exp xtk+1 + t 2k+3
,
(k+1)(k+2) k+1 (k+1)2 (2k+3)

b k+1 ab2 2k+3 2abµ k+2 2
w(x, t) = A exp xt +µx+ t + t +aµ t .
k+1 (k+1)2 (2k+3) (k+1)(k+2)

b k+1 ab2 2k+3 2ab
w(x, t) = u(z, t) exp xt + t , z = x+ tk+2
k+1 (k+1)2 (2k+3) (k+1)(k+2)
∂w ∂ 2w
16. =a + (bx2 tn + cxtm + dtk )w.
∂t ∂x2
This is a special case of equation 3.8.7.5 with n(t) = a, f (t) = g(t) = 0, h(t) = btn ,
s(t) = ctm , and p(t) = dtk .
∂w ∂2w ∂w
3.3.2 Equations of the Form =a + f (x, t)
∂t ∂x2 ∂x
2
∂w ∂ w ∂w
1. =a + (bt + c) .
∂t ∂x2 ∂x
w(x, t) = 2Ax + A(bt2 + 2ct) + B,
2
w(x, t) = A x + 12 bt2 + ct + 2aAt + B,

w(x, t) = A exp µx + 12 µbt2 + (aµ2 + µc)t + B.
2◦ . The substitution z = x + 12 bt2 + ct leads to a constant coefficient equation, ∂t w =
a∂zz w, which is considered in Section 3.1.1.
∂w ∂2w ∂w
2. =a + bx .
∂t ∂x2 ∂x
This equation is a special case of equation 3.8.2.2 with f (x) = bx and a special case of
equation 3.8.2.3 with f (t) = b.
Z b
w(x) = A exp − x2 dx + B,
2a
bt
w(x, t) = Axe + B,
w(x, t) = Abx2 e2bt + Aae2bt + B,

w(x, t) = A exp 2bµxebt + 2abµ2 e2bt + B.
2◦ . On passing from t, x to the new variables (A and B are any numbers)
A2 2bt
τ= e + B, z = Axebt ,
2b
for the function w(τ, z) we obtain a constant coefficient equation, ∂τ w = a∂zz w, which is
considered in Section 3.1.1.
w = f (x) at t = 0.
Solution:
−1/2 Z 2
2πa 2bt ∞ b xebt − ξ
w(x, t) = e −1 exp − f (ξ) dξ.
b −∞ 2a e2bt − 1
⊙ Literature: W. Miller, Jr. (1977), A. D. Polyanin, A. V. Vyazmin, A. I. Zhurov, and D. A. Kazenin (1998).
∂w ∂2w ∂w
3. =a + (bt2 + c) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.2.1 with f (t) = bt2 + c.


w(x, t) = A x + 13 bt3 + ct + B,
2
w(x, t) = A x + 13 bt3 + ct + 2aAt + B,

w(x, t) = A exp µx + 13 µbt3 + µ(aµ + c)t + B.
2◦ . On passing from t, x to the new variables t, z = x + 13 bt3 + ct, we obtain a constant

coefficient equation, ∂t w = a∂zz w, which is considered in Section 3.1.1.
∂w ∂2w ∂w
4. =a + x(bt + c) .
∂t ∂x2 ∂x
1 2

w(x, t) = Ax exp 2 bt
+ ct + B,
Z

w(x, t) = Ax exp bt + 2ct + 2Aa exp bt2 + 2ct dt + B,
2 2
h Z
1 2
2
i
w(x, t) = A exp µx exp 2 bt + ct + aµ exp bt2 + 2ct dt + B.
2◦ . On passing from t, x to the new variables (A is any number)

Z

τ= exp bt2 + 2ct dt + A, z = x exp 1 2
2 bt + ct ,
∂w ∂2w ∂w
5. = + (ax + bt + c) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.2.7 with f (t) = a and g(t) = bt + c. See also
equation 3.8.2.4.
∂w ∂2w x ∂w
6. =a +b .
∂t ∂x2 t ∂x
Ilkovič’s equation. It describes heat transfer to the surface of a growing drop that flows
out of a thin capillary into a fluid solution (the mass rate of flow of the fluid moving in
the capillary is assumed constant). This equation is a special case of equation 3.8.2.3 with
f (t) = b/t.
w(x, t) = Axtb + B,
w(x, t) = A(2b + 1)x2 t2b + 2Aat2b+1 + B,

b aµ2 2b+1
w(x, t) = A exp µxt + t + B.
2b + 1
2◦ . On passing from t, x to the new variables
1
τ= t2b+1 , z = xtb ,
2b + 1
3◦ . The solution of the original equation in the important special case where the drop sur-
face has a time-invariant temperature ws and the heat exchange occurs with an infinite
medium having an initial temperature w0 , namely,
w = w0 at t = 0 (initial condition),
w = ws at x = 0 (boundary condition),
w → w0 at x → ∞ (boundary condition),
is expressed in terms of the error function as follows:

√ Z ξ
w − ws 2b + 1 x 2
= erf √ √ , erf ξ = √ exp −ζ 2 dζ.
w0 − ws 2 a t π 0
⊙ Literature: Yu. P. Gupalo, A. D. Polyanin, and Yu. S. Ryazantsev (1985).
∂w ∂2w ∂w
7. =a + (btk x + ctm ) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.2.7 with f (t) = btk and g(t) = ctm .

∂w ∂2w b ∂w
8. =a + cx + .
∂t ∂x2 x ∂x
On passing from t, x to the new variables z, τ by the formulas
a 2ct
z = xect , τ= e + const,
2c
we obtain a simpler equation of the form 3.2.5:
∂w ∂2w µ ∂w b
= 2
+ , µ= .
∂τ ∂z z ∂z a
For µ = 1 and µ = 2, see also equations from Sections 3.2.1 and 3.2.3.

∂w ∂2w b ∂w
9. =a + ctn x + .
∂t ∂x2 x ∂x
This is a special case of equation 3.8.2.6 with f (t) = ctn .
3.3.3 Equations of the Form

∂w ∂2w ∂w
=a + f (x, t) + g(x, t)w + h(x, t)
∂t ∂x2 ∂x
∂w ∂2w ∂w
1. =a +b + (cx + d)w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.7.4 with n(t) = a, f (t) = 0, g(t) = b, h(t) = c, and
s(t) = d.

b 1 23 b2
w(x, t) = A exp ctx − x + ac t + d − t ,
2a 3 4a

2 b 1 23 b2
w(x, t) = A(x + act ) exp ctx − x + ac t + d − t ,
2a 3 4a

b 1 b2
w(x, t) = A exp x ct + µ − + ac2t3 + acµt2 + aµ2 + d − t ,
2a 3 4a

b p 2 3/2 b2
w(x, t) = A exp −µt − x ξ J1/3 √ ξ , ξ = cx + µ + d − ,
2a 3c a 4a

b p 2 3/2 b2
w(x, t) = A exp −µt − x ξ Y1/3 √ ξ , ξ = cx + µ + d − ,
2a 3c a 4a
where J1/3 (ξ) and Y1/3 (ξ) are the Bessel functions of the first and second kind of order 1/3,
respectively.

b 1 2 3 b2
w(x, t) = u(z, t) exp ctx − x + ac t + d − t , z = x + act2
2a 3 4a
∂w ∂2w ∂w
2. =a + bx + (cx + d)w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.7.4 with n(t) = a, f (t) = b, g(t) = 0, h(t) = c, and
s(t) = d.

c ac2
w(x, t) = A exp − x + d + 2 t ,
b b

2ac c ac2
w(x, t) = A x − 2 exp − x + b + d + 2 t ,
b b b
2
aµ 2bt bt 2ac c ac2
w(x, t) = A exp e + µe x − 2 − x + d + 2 t .
2b b b b
See 3.3.4.7 for more complicated solutions.

c ac2 a 2bt 2ac
bt
w(x, t) = u(z, τ ) exp − x + d + 2 t , τ= e , z =e x− 2 ,
b b 2b b
leads to a constant coefficient equation, ∂τ u = ∂zz u, which is considered in Section 3.1.1.
∂w ∂2w ∂w
3. =a + (bx + c) + (dx + e)w.
∂t ∂x2 ∂x
For b = 0, see equation 3.3.3.1. For b 6= 0, the substitution z = x + c/b leads to an equation
∂w ∂2w ∂w cd
= a 2 + bz + (dz + k)w, k =e− .
∂t ∂z ∂z b
∂w ∂2w ∂w
4. = + 2(ax + b) + (a2 x2 + 2abx + c)w.
∂t ∂x2 ∂x
This equation is a special case of equation 3.8.6.5 and
a special case of equation 3.8.7.5.
1 2
The substitution w(x, t) = u(x, t) exp − 2 ax − bx leads to a constant coefficient equa-
tion of the form 3.1.3 with Φ ≡ 0, namely, ∂t u = ∂xx u + (c − a − b2 )u.
∂w ∂2w ∂w
5. = + (ax + b) + (cx2 + dx + e)w.
∂t ∂x2 ∂x
This equation is a special case of equation 3.8.6.5 and a special case of equation 3.8.7.5.
1 2
w(x, t) = u(x, t) exp 2 Ax ,
where A is a root of the quadratic equation A2 + aA + c = 0, yields an equation of the form
3.8.7.3,
∂u ∂2u ∂u
= 2
+ (2A + a)x + b + (Ab + d)x + A + e u,
∂t ∂x ∂x
which is reduced to a constant coefficient equation.
1 2

w(x, t) = u(x, t) exp 2 Ax + Bx + Ct
leads to an equation of the analogous form:
∂u ∂2u ∂u
= 2
+ (2A + a)x + 2B + b
∂t ∂x ∂x
+ (A2 + Aa + c)x2 + (2AB + Ab + Ba + d)x + B 2 + Bb + A − C + e u.
By appropriately choosing the coefficients A, B, and C, one can simplify the original
equation in various ways.
∂w ∂2w ∂w
6. = + (ax + bt + c) + (sx2 + ptx + qt2 + kx + lt + m)w.
∂t ∂x2 ∂x
∂w ∂2w ∂w
7. =a + (btk x + ctm ) + stn w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.3.6 with f (t) = btk , g(t) = ctm , and h(t) = stn .

∂w ∂2w b ∂w k
8. =a + + c + 2 w.
∂t ∂x2 x ∂x x
w(x, t) = ect xλ u(x, τ ), τ = at,
where λ is a root of the quadratic equation aλ2 + (b − a)λ + k = 0, leads to an equation

of the form 3.2.5:
∂u ∂2u b 1 ∂u
= + 2λ + .
∂τ ∂x2 a x ∂x
2◦ . If w(x, t) is a solution of the original equation, then the functions

2 )τ
w1 = Aec(1−a w(±ax, a2 τ ), τ = t + B,

λ−1 x2 c x 1 1 b
w2 = Aτ exp − + cτ + 2 w ± , − 2 , λ= − ,
4aτ a τ aτ a τ 2 2a
where A and B are arbitrary constants, are also solutions of this equation.

∂w ∂2w b ∂w k
9. =a 2 + + c + 2 w + Φ(x, t).
∂t ∂x x ∂x x
The transformation
w(x, t) = ect xλ u(x, τ ), τ = at,
where λ is a root of the quadratic equation aλ2
+ (b − a)λ + k = 0, leads to an equation
of the form 3.2.6:

∂u ∂2u b 1 ∂u 1
= + 2λ + + Ψ(x, τ ), Ψ(x, τ ) = e−ct x−λ Φ(x, t).
∂τ ∂x2 a x ∂x a
∂w ∂2w ∂w
3.3.4 Equations of the Form = (ax+b) 2
+f (x, t) +g(x, t)w
∂t ∂x ∂x
∂w ∂ 2w
1. = ax .
∂t ∂x2
This is a special case of equation 3.8.6.1 with f (x) = ax. See also equation 3.3.6.6 with
n = 0.
1◦ . Particular solutions (A, B, C, and µ are arbitrary constants):
w(x) = Ax + B,
w(x, t) = 2Aatx + Ax2 + B,
w(x, t) = Aa2 t2 x + Aatx2 + 16 Ax3 + B,
w(x, t) = 2Aa3 t3 x + 3Aa2 t2 x2 + Aatx3 + 1
12 Ax
4
+ B,
n−1
X
[n(n − 1) . . . (n − k)]2
w(x, t) = xn + (at)k xn−k ,
n(n − k)k!
k=1
x
w(x, t) = A exp − + C,
at + B
Ax x
w(x, t) = exp − + C,
(at + B)2 at + B
w(x, t) = Aat + A(x ln x − x) + B,
w(x, t) = Aa2 t2 + 2Aat(x ln x − x) + A(x2 ln x − 52 x2 ) + B,

µt √ 2 √ 2 √
w(x, t) = e x AJ1 √ −µx + BY1 √ −µx for µ < 0,
a a

µt √ 2 √ 2 √
w(x, t) = e x AI1 √ µx + BK1 √ µx for µ > 0,
a a
where J1 (z) and Y1 (z) are Bessel functions, and I1 (z) and K1 (z) are modified Bessel
functions.
2◦ . A solution containing an arbitrary function of the space variable:
X∞
(at)n n d2
w(x, t) = f (x) + L [f (x)], L≡x ,
n=1
n! dx2
dition w(x, 0) = f (x). The sum is finite if f (x) is a polynomial.
3◦ . A solution containing an arbitrary function of time:

∞
X 1 (n−1)
w(x, t) = A + xg(t) + xn gt (t),
n[(n − 1)!]2 an−1
n=2
where g(t) is any infinitely differentiable function and A is an arbitrary number. This
solution possesses the properties
w(0, t) = A, ∂x w(0, t) = g(t).
4◦ . Suppose w = w(x, t) is a solution of the original equation. Then the functions
w1 = Aw(λx, λt + C),

βx x γ + λt
w2 = A exp − w , , λδ − βγ = 1,
a(δ + βt) (δ + βt)2 δ + βt
where A, C, β, δ, and λ are arbitrary constants, are also solutions of the equation.
∂w ∂ 2w
2. = ax + (−bx + c)w.
∂t ∂x2
The transformation
r r
b √ 1 a √
w(x, t) = u(z, τ ) exp x + ct , z = x exp 2 ab t , τ= exp 2 ab t
a 2 b
leads to a simpler equation of the form 3.3.4.1:
∂u ∂2u
=z 2.
∂τ ∂z
∂w ∂ 2w
3. = ax + bxtn w.
∂t ∂x2
This is a special case of equation 3.8.8.1 with f (t) = a, g(t) = 0, h(t) = btn , and s(t) = 0.
∂w ∂ 2w
4. = ax + (btn x + ctm )w.
∂t ∂x2
This is a special case of equation 3.8.8.1 with f (t) = a, g(t) = 0, h(t) = btn , and s(t) = ctm .
h i
∂w ∂2w ∂w
5. = a (x + b) 2 + .
∂t ∂x ∂x
This equation describes heat transfer in a quiescent medium (solid body) in the case of
thermal diffusivity as a linear function of the space coordinate.
1◦ . The original equation can be rewritten in a form more suitable for applications,

∂w ∂ ∂w
=a (x + b) .
∂t ∂x ∂x
2◦ . The substitution x = 14 z 2 − b leads to the equation

2
∂w ∂ w 1 ∂w
=a + ,
∂t ∂z 2 z ∂z
∂w ∂ 2w ∂w
6. = ax + bx + (cx + d)w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.8.1 with f (t) = a, g(t) = b, h(t) = c, and s(t) = d.
∂w ∂2w ∂w
7. = (a2 x + b2 ) + (a1 x + b1 ) + (a0 x + b0 )w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.6.5 with f (x) = a2 x + b2 , g(x) = a1 x + b1 , h(x) =
a0 x + b0 , and Φ ≡ 0.
Particular solutions of the original equation are presented in Table 3.2, where the func-
tion
J (a, b; x) = C1 Φ(a, b; x) + C2 Ψ(a, b; x), C1 , C2 are any numbers,
is an arbitrary solution of the degenerate hypergeometric equation

′′
xyxx + (b − x)yx′ − ay = 0,
and the function
Zν (x) = C1 Jν (x) + C2 Yν (x), C1 , C2 are any numbers,
is an arbitrary solution of the Bessel equation
x2 yxx
′′
+ xyx′ + (x2 − ν 2 )y = 0.
For the degenerate hypergeometric functions Φ(a, b; x) and Ψ(a, b; x), see Section 30.9
and books by Bateman and Erdélyi (1953, Vol. 1) and Abramowitz and Stegun (1964).
For the Bessel functions Jν (x) and Yν (x), see Section 30.6 and Bateman and Erdélyi
(1953, Vol. 2), Abramowitz and Stegun (1964), Nikiforov and Uvarov (1988), and Temme
(1996).
Remark 3.4. For a2 = 0 the original equation is a special case of equation 3.8.7.4 and can be
reduced to the constant coefficient heat equation that is considered in Section 3.1.1. In this case, a
number of solutions are not displayed in Table 3.2.
TABLE 3.2
Particular solutions of equation 3.3.4.7 for different values
of the determining parameters (µ is an arbitrary number)
Particular solution: w(x, t) = exp(hx − µt)F (ξ), where ξ = (x + γ)/p
Constraints h p γ F Parameters
a2 6= 0, D − a1 a2 b2 a = B(h)/A(h),
− J (a, b; ξ)
D 6= 0 2a2 A(h) a2 b = (a2 b1 − a1 b2 )a−2
2
a2 = 0, a0 2b2 h + b1 a = B(h)/(2a1 ),
− 1 J a, 12 ; kξ 2
a1 6= 0 a1 a1 k = −a1 /(2b2 )
1 2b2 h + b1
a2 6= 0, a1 b2 p α= − ,
− a2 α
ξ Z2α β ξ 2 2a
a21 = 4a0 a2 2a2 a2 p 2
β = 2 B(h)
a2 = a1 = 0, b1 4(b0 + µ)b2 − b21 2 a0 1/2
− 1 ξ 1/2 Z1/3 kξ 3/2 k=
a0 6= 0 2b2 4a0 b2 3 b2
Notation: D2 = a21 − 4a0 a2 , A(h) = 2a2 h + a1 , B(h) = b2 h2 + b1 h + b0 + µ

∂w ∂2w ∂w
= (ax2 + bx + c) + f (x, t) + g(x, t)w
∂t ∂x2 ∂x
∂w ∂2w ∂w
1. = ax2 + bx + cw.
∂t ∂x2 ∂x

w(x, t) = A ln |x| + B |x|n exp (c − an2 )t ,

w(x, t) = A 2at + ln2 |x| |x|n exp (c − an2 )t ,

w(x, t) = A|x|µ exp (c + aµ2 − 2anµ)t ,
where n = 12 (a − b)/a.
a−b
w(x, t) = |x|n exp (c − an2 )t u(z, t), z = ln |x|, n= ,
2a
∂w ∂2w
2. = ax2 + (btn + c)w.
∂t ∂x2
This is a special case of equation 3.8.4.2 with f (t) = btn + c.
The transformation

b n+1
w(x, t) = u(z, t) exp t + ct , z = ln |x|
n+1
leads to a constant coefficient equation of the form 3.1.4:
∂u ∂2u ∂u
=a 2 −a .
∂τ ∂z ∂z
∂w ∂2w ∂w
3. = ax2 + bx + (cxk + s)w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.6.5 with f (x) = ax2 , g(x) = bx, h(x) = cxk + s, and
Φ ≡ 0. For c = 0, see equation 3.3.5.1.
Particular solutions for c 6= 0:
r
−µt 2a
a−b 2 c k 1 p
w(x, t) = Ae x Jν x2 , ν = (a − b)2 − 4a(s + µ),
k a ak
r
−µt 2a
a−b 2 c k 1 p
w(x, t) = Ae x Yν x2 , ν = (a − b)2 − 4a(s + µ),
k a ak
where A and µ are arbitrary constants, and Jν (z) and Yν (z) are Bessel functions.
∂w ∂2w ∂w
4. = a 2 x2 + (a1 x2 + b1 x) + (a0 x2 + b0 x + c0 )w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.6.5 with f (x) = a2 x2 , g(x) = a1 x2 + b1 x, h(x) =
a0 x2 + b0 x + c0 , and Φ ≡ 0.
1◦ . Particular solutions for a21 6= 4a0 a2 :
b1
w(x, t) = A exp(−νt + µx)xk Φ α, 2k + ; −γx ,
a2
(1)
b1
w(x, t) = A exp(−νt + µx)xk Ψ α, 2k + ; −γx ,
a2
where A and ν are arbitrary constants,
p
a21 − 4a0 a2 − a1 (b1 + 2a2 k)µ + b0 + a1 k a1
µ= , α= , γ = 2µ + ,
2a2 2a2 µ + a1 a2
k = k(ν) is a root of the quadratic equation a2 k2 + (b1 − a2 )k + c0 + ν = 0, and Φ(α, β; z)
and Ψ(α, β; z) are the degenerate hypergeometric functions. [For the degenerate hyperge-
ometric functions, see Section 30.9 and the books by Abramowitz and Stegun (1964) and
Bateman and Erdélyi (1953, Vol. 1)].
2◦ . Particular solutions for a21 = 4a0 a2 :
a1 k m p x
w(x, t) = A exp −νt − x x ξ J2m 2 pξ , ξ= ,
2a2 a2
a1 k m p x (2)
w(x, t) = A exp −νt − x x ξ Y2m 2 pξ , ξ= ,
2a2 a2
where A and ν are arbitrary constants,
1 b1 a1
m= −k− , p=− (b1 + 2a2 k) + b0 + a1 k = 0,
2 2a2 2a2
k = k(ν) is a root of the quadratic equation a2 k2 + (b1 − a2 )k + c0 + ν = 0, and Jm (z)

and Ym (z) are Bessel functions. [For the Bessel functions, see Section 30.6 and the books
by Abramowitz and Stegun (1964) and Bateman and Erdélyi (1953, Vol. 2)].
Remark 3.5. In solutions (1) and (2), the parameter k can be regarded as arbitrary, and then
ν = −a2 k 2 − (b1 − a2 )k − c0 .
∂w ∂2w ∂w
5. = a 2 x2 + (a1 xk+1 + b1 x) + (a0 x2k + b0 xk + c0 )w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.6.5 with f (x) = a2 x2 , g(x) = a1 xk+1 + b1 x, h(x) =
a0 x2k + b0 xk + c0 , and Φ ≡ 0.
The substitution ξ = xk leads to an equation of the form 3.3.5.4:
∂w ∂2w ∂w
= a2 k2 ξ 2 2 + k a1 ξ 2 + βξ + (a0 ξ 2 + b0 ξ + c0 )w,
∂t ∂ξ ∂ξ
where β = b1 + a2 (k − 1).
∂w ∂2w ∂w
6. = (ax2 + b) 2 + ax + cw.
∂t ∂x ∂x
Z
dx
The substitution z = √ leads to the constant coefficient equation
ax2 + b
∂w ∂2w
= + cw,
∂t ∂z 2

∂w ∂ 2w ∂w
= f (x) + g(x, t) + h(x, t)w
∂t ∂x2 ∂x
∂w ∂2w ∂w
1. = ax3 + bxtk + ctm w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.8.7 with n = 3, f (t) = a, g(t) = btk , and h(t) = ctm .
∂w ∂2w
2. = ax4 + bw.
∂t ∂x2
This is a special case of equation 3.3.7.6 with n = m = 0.
w(x, t) = ebt (Ax + B),

A
w(x, t) = ebt 2Aatx + +B ,
h x
µi
w(x, t) = Ax exp (b + aµ2 )t + .
x
2◦ . The transformation w(x, t) = xebt u(ξ, t), ξ = 1/x leads to a constant coefficient
equation, ∂t u = a∂ξξ u, which is considered in Section 3.1.1.
∂w ∂2w
3. = (x2 + a2 )2 + Φ(x, t).
∂t ∂x2
Domain: 0 ≤ x ≤ l. First boundary value problem.
w = f (x) at t=0 (initial condition),

w = g(t) at x = 0 (boundary condition),
w = h(t) at x = l (boundary condition).
Solution:
Z tZ l Z l
w(x, t) = G(x, ξ, t − τ )Φ(ξ, τ ) dξ dτ +
G(x, ξ, t)f (ξ) dξ
0 0 0
Z t Z t
+ a4 g(τ )Λ1 (x, t − τ ) dτ − (a2 + l2 )2 h(τ )Λ2 (x, t − τ ) dτ.
0 0
Here, the Green’s function G is given by

X∞ 2
yn (x)yn (ξ) exp(−µ2n t) 2 πna
G(x, ξ, t) = 2 (ξ 2 + a2 )2
, µn = − a2 ,
n=1
ky n k arctan(l/a)
p
arctan(x/a) arctan(l/a)
yn (x) = x2 + a2 sin πn , kyn k2 = ,
arctan(l/a) 2a
and the functions Λ1 and Λ2 are expressed via the Green’s function as follows:
∂ ∂

Λ1 (x, t) = G(x, ξ, t) , Λ2 (x, t) = G(x, ξ, t) .
∂ξ ξ=0 ∂ξ ξ=l
∂w ∂ 2w
4. = (x − a1 )2 (x − a2 )2 − bw, a1 6= a2 .
∂t ∂x2
The transformation
x−a
1
w(x, t) = (x − a2 )e−bt u(ξ, τ ), ξ = ln , τ = (a1 − a2 )2 t
x − a2
leads to a constant coefficient equation,
∂u ∂2u ∂u
= 2
− ,
∂τ ∂ξ ∂ξ
∂w ∂ 2w
5. = (a2 x2 + a1 x + a0 )2 + bw.
∂t ∂x2
The transformation
Z
p dx
w(x, t) = exp (a2 a0 − 14 a21 +b)t |a2 x2 +a1 x+a0 | u(ξ, t), ξ=
a2 x2 +a1 x+a0
leads to a constant coefficient equation, ∂t u = ∂ξξ u, which is considered in Section 3.1.1.
∂w ∂2w
6. = ax1−n .
∂t ∂x2
This equation is encountered in diffusion boundary layer problems (see equation 3.9.1.3)
and is a special case of 3.8.6.1 with f (x) = ax1−n . In addition, it is a special case of
equation 3.3.5.1 with n = −1 and is an equation of the form 3.3.6.2 for n = −3 (in
both cases the equation is reduced to a constant coefficient equation). For n = 0, see
equation 3.3.4.1.
w(x) = Ax + B,
w(x, t) = Aan(n + 1)t + Axn+1 + B,
w(x, t) = Aa(n + 1)(n + 2)tx + Axn+2 + B,

x2n+2
w(x, t) = A an(n + 1)t2 + 2txn+1 + + B,
a(n + 1)(2n + 1)

2 n+2 x2n+3
w(x, t) = A a(n + 1)(n + 2)t x + 2tx + + B,
a(n + 1)(2n + 3)

n
− n+1 xn+1
w(x, t) = A + Bt exp − ,
a(n + 1)2 t

n+2
− n+1 xn+1
w(x, t) = A + Bxt exp − ,
a(n + 1)2 t
√ √
µt √ −µ q −µ q
w(x, t) = e x AJ 1 √ x + BY 1 √ x for µ < 0,
2q aq 2q aq
√ √
µt √ µ q µ q
w(x, t) = e x AI 1 √ x + BK 1 √ x for µ > 0,
2q aq 2q aq
where q = 12 (n + 1), Jν (z) and Yν (z) are Bessel functions, and Iν (z) and Kν (z) are
modified Bessel functions.
Suppose 2/(n + 1) = 2m + 1, where m is an integer. We have the following particular
solutions (A, B, and µ are arbitrary constants):
√ √
µt 1−2q m+1 µ q µ q
w(x, t) = e x(x D) A exp √ x + B exp − √ x for m ≥ 0,
aq aq
√ √
−m µ µ
w(x, t) = eµtx(x1−2q D) A exp √ xq + B exp − √ xq for m < 0,
aq aq
d n+1 1
where D = , q= = .
dx 2 2m + 1
2◦ . Suppose w = w(x, t) is a solution of the original equation. Then the functions
w1 =Aw(λx,λn+1 t+C),

A βk2 xn+1 x γ +λt 1
w2 = nk
exp − w 2k
, , k= , λδ −βγ=1,
|δ +βt| a(δ +βt) (δ +βt) δ +βt n+1
where A, C, β, δ, and λ are arbitrary constants, are also solutions of the equation.

w = w1 at x = 0 (boundary condition),
w → w0 at x → ∞ (boundary condition),
where w0 = const and w1 = const.

Solution:
w − w1 1 k 2 xn+1 1
= γ k, , k= ,
w0 − w1 Γ(k) at n+1
Z ζ
where Γ(k) = γ(k, ∞) is the gamma function and γ(k, ζ) = ζ k−1 e−ζ dζ is the incom-
0
plete gamma function.
n+1
τ = 14 a(n + 1)2 t, ξ=x 2
leads to the equation
∂w ∂2w 1 − 2k ∂w 1
= + , k= ,
∂τ ∂ξ 2 ξ ∂ξ n+1

5◦ . Two discrete transformations are worth mentioning. They preserve the form of the
original equation, but the parameter n is changed.
5.1. The point transformation
1 w
z= , u= (transformation F)
x x
leads to a similar equation,
∂u ∂2u
= az n+3 2 . (1)
∂t ∂z
The transformation F changes the equation parameter in accordance with the rule
F
n =⇒ −n − 2. The second application of the transformation F leads to the original
equation.
5.2. Using the Bäcklund transformation (see 3.8.6.1, Item 5.2)
∂v
ξ = xn , w= (transformation H)
∂ξ
and integrating the resulting equation with respect to ξ, we obtain
∂v n−1 ∂ 2 v
= an2 ξ n . (2)
∂t ∂ξ 2
The transformation H changes the equation parameter in accordance with the rule
H 1
n =⇒ . The second application of the transformation H leads to the original equa-
n
tion.
The composition of transformations G = H ◦ F changes the equation parameter in
G 1
accordance with the rule n =⇒ − .
n+2
The original equation reduces to a constant coefficient equation for n = −3 (see 3.3.6.2).
Substituting n = −3 into (2) yields the equation
∂v ∂2v
= Aξ 4/3 2 ,
∂t ∂ξ
which also can be reduced to a constant coefficient equation.
Likewise, using the transformations F, G, and H, one may find some other equations
of the given type that are reduced to a constant coefficient heat equation.
∂w ∂2w ∂w
7. = axn + bx .
∂t ∂x2 ∂x
This is a special case of equation 3.8.4.5 with f (t) = b. On passing from t, x to the new
variables
a
z = xebt , τ = eb(2−n)t + const,
b(2 − n)
we obtain an equation of the form 3.3.6.6:
∂w ∂2w
= zn 2 .
∂τ ∂z
2
∂w ∂ w ∂w
8. = a xn 2 + nxn−1 .
∂t ∂x ∂x
This equation describes heat transfer in a quiescent medium (solid body) in the case where
thermal diffusivity is a power-law function of the coordinate. The equation can be rewritten
in the form
∂w ∂ n ∂w
=a x ,
∂t ∂x ∂x
which is more customary for applications.
1◦ . For n = 2, see equation 3.3.5.1. For n 6= 2, by passing from t, x to the new variables
2−n
τ = 14 a(2 − n)2 t, z = x 2 , we obtain an equation of the form 3.2.5:
∂w ∂2w n 1 ∂w
= + .
∂τ ∂z 2 2 − n z ∂z
w(x, t) = x1−n u(ξ, t), ξ = x3−2n
leads to a similar equation

4−3n
∂u ∂ ∂u
=b ξ 3−2n , b = a(3 − 2n)2 .
∂t ∂ξ ∂ξ

∂w ∂2w ∂w
9. = a x2m 2 + mx2m−1 .
∂t ∂x ∂x
This is a special case of equation 3.8.4.7 with f (t) = g(t) = 0.

The substitution 
 1 x1−m if m 6= 1,
ξ = 1−m
ln |x| if m = 1,
leads to a constant coefficient equation, ∂t w = a∂ξξ w, which is considered in Section

3.1.1.
∂w ∂2w ∂w
10. = axn + x(btm + c) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.4.5 with f (t) = btm + c.
∂w ∂2w ∂w
3.3.7 Equations of the Form = f (x, t) + g(x, t) + h(x, t)w
∂t ∂x2 ∂x
∂w ∂ 2w
1. = axt + (bx + ctn )w.
∂t ∂x2
This is a special case of equation 3.8.8.1 with f (t) = at, g(t) = 0, h(t) = b, and s(t) = ctn .
∂w ∂2w ∂w
2. = axtk + bxtm + ctn w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.8.1 with f (t) = atk , g(t) = btm , h(t) = 0, and
s(t) = ctn .

∂w ∂2w ∂w
3. = x atk 2 + btm + ctn w .
∂t ∂x ∂x
This is a special case of equation 3.8.8.1 with f (t) = atk , g(t) = btm , h(t) = ctn , and
s(t) = 0.
∂w ∂2w ∂w
4. = ax2 tk + btm x + ctn w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.8.2 with f (t) = atk , g(t) = btm , and h(t) = ctn .
∂w ∂2w ∂w
5. = ax3 tm + bxtk + ctl w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.8.7 with n = 3, f (t) = atm , g(t) = btk , and h(t) = ctl .
∂w ∂2w
6. = ax4 tn + btm w.
∂t ∂x2
This is a special case of equation 3.8.8.4 with f (t) = atn and g(t) = btm .
1◦ . Particular solutions (A, B, and λ are arbitrary constants):

b
w(x, t) = (Ax + B) exp tm+1 ,
m+1

2a n+1 1 b m+1
w(x, t) = A t x+ exp t ,
n+1 x m+1

b m+1 aλ2 n+1 λ
w(x, t) = Ax exp t + t + .
m+1 n+1 x

b 1 a n+1
w(x, t) = x exp tm+1 u(ξ, τ ), ξ= , τ= t
m+1 x n+1
leads to a constant coefficient equation, ∂τ u = ∂ξξ u, which is considered in Section 3.1.1.
∂w ∂ 2w ∂w
7. = atn + (btm x + cti ) + (dtl x + etp )w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.7.4 (hence, it can be reduced to a constant coefficient
equation, which is considered in Section 3.1.1).
∂w ∂ 2w ∂w
8. = atn + (btm x + cti ) + (dtl x2 + etp x + stq )w.
∂t ∂x2 ∂x
∂w ∂2w ∂w
9. = axn tm + bxtk + ctp w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.8.7 with f (t) = atm , g(t) = btk , and h(t) = ctp .
∂w ∂2w
3.3.8 Liquid-Film Mass Transfer Equation (1 − y 2 ) =a
∂x ∂y 2
This equation describes steady-state heat and mass transfer in a fluid film with a parabolic
velocity profile. The variables have the following physical meanings: w is a dimensionless
temperature (concentration); x and y are dimensionless coordinates measured, respectively,
along and across the film (y = 0 corresponds to the free surface of the film and y = 1 to the
solid surface the film flows down), and Pe = 1/a is the Peclet number. Mixed boundary
conditions are usually encountered in practical applications.
◮ Particular solutions (A, B, k, and λ are arbitrary constants).
k 4 k 2
w(x, y) = kx − y + y + Ay + B,
12a 2a
w(x, y) = A exp −aλ2 x exp − 12 λy 2 Φ 14 − 14 λ, 12 ; λy 2 ,

w(x, y) = A exp −aλ2 x y exp − 12 λy 2 Φ 34 − 14 λ, 32 ; λy 2 ,
P∞ α(α + 1) . . . (α + m − 1) z m
where Φ(α, β; z) = 1 + is the degenerate hypergeo-
m=1 β(β + 1) . . . (β + m − 1) m!
metric function.
◮ Mass exchange between fluid film and gas.

The mass exchange between a fluid film and the gas above the free surface, provided that
the admixture concentration at the film surface is constant and there is no mass transfer
through the solid surface, meets the boundary conditions
w = 0 at x = 0 (0 < y < 1),

w = 1 at y = 0 (x > 0),
∂y w = 0 at y = 1 (x > 0).
The solution of the original equation subject to these boundary conditions is given by
∞
X
w(x, y) = 1 − Am exp −aλ2m x Fm (y),
m=1 (1)

Fm (y) = y exp − 12 λm y 2 Φ 34 − 14 λm , 32 ; λm y , 2
where the function Fm and the coefficients Am and λm are independent of the parameter a.
The eigenvalues λm are solutions of the transcendental equation

λm Φ 34 − 14 λm , 32 ; λm − Φ 34 − 14 λm , 12 ; λm = 0.
The series coefficients Am are calculated from

Z 1
(1 − y 2 )Fm (y) dy
Am = Z 10 , where m = 1, 2, . . .
2
2
(1 − y ) Fm (y) dy
0
Table 3.3 shows the first ten eigenvalues λm and coefficients Am .

TABLE 3.3
Eigenvalues λm and coefficients Am in solution (1)
m λm Am m λm Am
1 2.2631 1.3382 6 22.3181 −0.1873
2 6.2977 −0.5455 7 26.3197 0.1631
3 10.3077 0.3589 8 30.3209 −0.1449
4 14.3128 −0.2721 9 34.3219 0.1306
5 18.3159 0.2211 10 38.3227 −0.1191
The solution asymptotics as ax → 0 is given by

y
w = erfc √ ,
2 ax
Z ∞

where erfc z = exp −ξ 2 dξ is the complementary error function.
z
◮ Dissolution of a plate by a laminar fluid film.
The dissolution of a plate by a laminar fluid film, provided that the concentration at the
solid surface is constant and there is no mass flux from the film into the gas, satisfies the
boundary conditions
w = 0 at x = 0 (0 < y < 1),

∂y w = 0 at y = 0 (x > 0),
w = 1 at y = 1 (x > 0).
The solution of the original equation subject to these boundary conditions is given by
∞
X
w(x, y) = 1 − Am exp −aλ2m x Gm (y),
m=0 (2)

Gm (y) = exp − 12 λm y 2 Φ 14 − 14 λm , 1
2;
2
λm y ,
where the functions Gm and the constants Am and λm are independent of the parameter a.
The eigenvalues λm are solutions of the transcendental equation
1

Φ 4 − 14 λm , 1
2; λm = 0.
The following approximate relation is convenient to calculate λm :
λm = 4m + 1.68 (m = 0, 1, 2, . . . ). (3)
The maximum error of this formula is less than 0.2%.

The coefficients Am are approximated by
A0 = 1.2, Am = (−1)m 2.27 λ−7/6

m for m = 1, 2, 3, . . . ,
where the eigenvalues λm are defined by (3). The maximum error of the expressions for
Am is less than 0.1%.
The solution asymptotics as ax → 0 is given by
1 1 2
(1 − y)3
w=
1 Γ 3, 9ζ , ζ= ,
Γ 3
ax
Z ∞
where Γ(α, z) = e−ξ ξ α−1 dξ is the incomplete gamma function, Γ(α) = Γ(α, 0) is
z
the gamma function, and Γ 13 ≈ 2.679.
⊙ References for Section 3.3.8: Z. Rotem and J. E. Neilson (1966), E. J. Davis (1973), A. D. Polyanin,
A. M. Kutepov, A. V. Vyazmin, and D. A. Kazenin (2002).
∂w ∂w ∂2w
3.3.9 Equations of the Form f (x, y) + g(x, y) = + h(x, y)
∂x ∂y ∂y 2
∂w ∂w ∂2w
1. axn + bxk y = .
∂x ∂y ∂y 2
This is a special case of equation 3.9.1.1 with f (x) = axn and g(x) = bxk .
The transformation
Z
1 2b k−n+1 dx b k−n+1
t= exp − x , z = y exp − x
a a(k − n + 1) xn a(k − n + 1)
leads to a constant coefficient equation, ∂t w = ∂zz w, which is considered in Section 3.1.1.
∂w ∂w ∂2w
2. axn + bxk y = − cxm w.
∂x ∂y ∂y 2
This is a special case of equation 3.9.1.2 with f (x) = axn , g(x) = bxk , and h(x) = cxm .
∂w ∂w ∂2w
3. axm y n−1 + bxk y n = .
∂x ∂y ∂y 2
This is a special case of equation 3.9.1.3 with f (x) = axm and g(x) = bxk .
The transformation
Z
1 b(n + 1) k−m+1 dx b k−m+1
t= exp − x , z = y exp − x
a a(k − m + 1) xm a(k − m + 1)
∂w ∂2w
= z 1−n 2 .
∂t ∂z

y n ∂w b y k ∂w ∂2w
4. a √ + √ √ = .
x ∂x x x ∂y ∂y 2
This is a special case of equation 3.9.1.4 with f (z) = az n and g(z) = bz k .
◆ See Section 3.9.1 for other equations of this form.
3.4 Equations Containing Exponential Functions

and Arbitrary Parameters
∂w ∂2w
∂t ∂x2
∂w ∂2w
1. =a + beβt + c w.
∂t ∂x2
This is a special case of equation 3.8.1.1 with f (t) = beβt + c.
3.4. Equations Containing Exponential Functions and Arbitrary Parameters 339

b βt
w(x, t) = (Ax + B) exp e + ct ,
β

b βt
w(x, t) = A(x2 + 2at) exp e + ct ,
β

2 b βt
w(x, t) = A exp λx + aλ t + e + ct .
β
b
2◦ . The substitution w(x, t) = u(x, t) exp eβt + ct
leads to a constant coefficient
β
equation, ∂t u = a∂xx u, which is considered in Section 3.1.1.
∂w ∂2w
2. =a + (beβt + cx)w.
∂t ∂x2
This is a special case of equation 3.8.1.6 with f (t) = c and g(t) = beβt .
1◦ . Particular solutions (A and λ are arbitrary constants):

b 1
w(x, t) = A exp cxt + eβt + ac2 t3 ,
β 3

2
b βt 1 2 3
w(x, t) = A x + act exp cxt + e + ac t ,
β 3

b βt 1 2 3 2 2
w(x, t) = A exp x(ct + λ) + e + ac t + acλt + aλ t .
β 3

b βt 1 2 3
w(x, t) = u(z, t) exp cxt + e + ac t , z = x + act2
β 3
∂w ∂2w
3. =a + (beβx − c)w.
∂t ∂x2
This is a special case of equation 3.8.1.2 with f (x) = beβx − c.

Particular solutions (A and λ are arbitrary constants):
√ r
2 b βx/2 2 c−λ
w(x, t) = A exp(−λt)Jν √ e , ν= ,
β a β a
√ r
2 b βx/2 2 c−λ
w(x, t) = A exp(−λt)Yν √ e , ν= ,
β a β a
where Jν (ξ) and Yν (ξ) are the Bessel functions of the first and second kind, respectively.
∂w ∂2w
4. =a + (bxeβt + c)w.
∂t ∂x2
This is a special case of equation 3.8.1.6 with f (t) = beβt and g(t) = c.

b βt ab2 2βt
w(x, t) = A exp xe + 3 e + ct ,
β 2β

2ab βt b βt ab2 2βt
w(x, t) = A x + 2 e exp xe + 3 e + ct ,
β β 2β
2

b βt ab 2βt 2abλ βt 2
w(x, t) = A exp x e +λ + e + 2 e + (aλ + c)t .
β 2β 3 β

b βt ab2 2βt 2ab βt
w(x, t) = u(z, t) exp xe + 3 e + ct , z =x+ e
β 2β β2
∂w ∂2w
5. =a + x(beβt + c)w.
∂t ∂x2

b βt
w(x, t) = A exp x e + ct + aφ(t) ,
β

2b βt 2 b βt
w(x, t) = A x + a e + ct exp x e + ct + aφ(t) ,
β2 β

b βt 2b βt 2
w(x, t) = A exp x e + ct + λ + aλ e + ct + λt + aφ(t) ,
β β2
where φ(t) = 12 b2 β −3 e2βt + 2bcβ −3 (βt − 1)eβt + 13 c2 t3 .


b βt 2b βt 2
w(x, t) = u(z, t) exp x e + ct + aφ(t) , z =x+a e + ct
β β2
∂w ∂2w
6. =a + x2 (beβt + c)w.
∂t ∂x2
This is a special case of equation 3.8.7.5 with n(t) = a, f (t) = g(t) = s(t) = p(t) = 0,
and h(t) = beβt + c.
∂w ∂2w
7. =a + (beβt + ceλt )w.
∂t ∂x2
This is a special case of equation 3.8.1.1 with f (t) = beβt + ceλt .
1◦ . Particular solutions (A, B, and ν are arbitrary constants):

b βt c λt
w(x, t) = (Ax + B) exp e + e ,
β λ

2 b βt c λt
w(x, t) = A(x + 2at) exp e + e ,
β λ

2 b βt c λt
w(x, t) = A exp νx + aν t + e + e .
β λ

b βt c λt
2◦ .The substitution w(x, t) = u(x, t) exp e + e leads to a constant coefficient
β λ
equation, ∂t u = a∂xx u, which is considered in Section 3.1.1.
∂w ∂2w
8. =a + (beβx + ceλt + d)w.
∂t ∂x2
c
The substitution w(x, t) = u(x, t) exp eλt leads to an equation of the form 3.4.1.3:
λ
∂u ∂2u
= a 2 + (beβx + d)u.
∂t ∂x
∂w ∂2w
9. =a + (beβx+λt + c)w.
∂t ∂x2
For β = 0, see equation 3.4.1.1; for λ = 0, see equation 3.4.1.3.

For β 6= 0, the transformation
λ λ
w(x, t) = u(z, t)eµx , z =x+ t, where µ= ,
β 2aβ
∂u ∂2u
= a 2 + (beβz + c + aµ2 )u.
∂t ∂z
∂w ∂2w ∂w
∂t ∂x2 ∂x
∂w ∂2w ∂w
1. =a + (beβt + c) .
∂t ∂x2 ∂x


b βt
w(x, t) = Ax + A e + ct + B,
β
2
b βt
w(x, t) = A x + e + ct + 2aAt + B,
β

b
w(x, t) = A exp λx + λ eβt + (aλ2 + cλ)t + B.
β
b βt
2◦ . On passing from t, x to the new variables t, z = x + e + ct, we obtain a constant
β
∂w ∂2w ∂w
2. =a + (beβx + c) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.2.2 with f (x) = beβx + c.

c b βx
w(x, t) = A exp(−λt + kβx) Φ k, 2k + 1 + ;− e ,
aβ aβ
(1)
c b
w(x, t) = A exp(−λt + kβx) Ψ k, 2k + 1 + ; − eβx ,
aβ aβ
where k = k(λ) is a root of the quadratic equation aβ 2 k2 + cβk + λ = 0; Φ(α, ν; z) and

Ψ(α, ν; z) are degenerate hypergeometric functions. [Regarding the degenerate hyperge-
ometric functions, see Section 30.9 and the books by Abramowitz and Stegun (1964) and
Bateman and Erdélyi (1953, Vol. 1).]
Remark 3.6. In solutions (1), the parameter k can be considered arbitrary, and then we have
λ = −aβ 2 k 2 − cβk.
2◦ . Other particular solutions (A and B are arbitrary constants):

Z
b βx c
w(x) = A + B F (x) dx, F (x) = exp − e − x ,
aβ a
Z Z
dx
w(x, t) = Aat + A F (x) dx,
F (x)
Z Z Z
G(x) dx
w(x, t) = AatG(x) + A F (x) dx, G(x) = F (x) dx.
F (x)
3◦ . The substitution z = eβx leads to an equation of the form 3.3.5.4:
∂w ∂2w ∂w
= aβ 2 z 2 2 + βz(bz + c + aβ) .
∂t ∂z ∂z
∂w ∂2w ∂w
3. =a + x(beβt + c) .
∂t ∂x2 ∂x

b βt
w(x, t) = AxF (t) + B, F (t) = exp e + ct ,
β
Z
w(x, t) = Ax2 F 2 (t) + 2Aa F 2 (t) dt + B,
Z
w(x, t) = A exp λxF (t) + aλ2 F 2 (t) dt + B.

Z
2 b βt
τ = F (t) dt + A, z = xF (t), where F (t) = exp e + ct ,
β
∂w ∂2w ∂w
4. =a + (beβt + ceλt ) .
∂t ∂x2 ∂x

b βt c λt
w(x, t) = Ax + A e + e + B,
β λ

b βt c λt 2
w(x, t) = A x + e + e + 2aAt + B,
β λ

b βt c λt 2
w(x, t) = A exp µx + µ e + µ e + aµ t + B.
β λ
b βt c λt
2◦ . On passing from t, x to the new variables t, z = x+ e + e , we obtain a constant
β λ
∂w ∂2w ∂w
5. =a + (beβt+λx + c) .
∂t ∂x2 ∂x
For β = 0, see equation 3.4.2.2; for λ = 0, see equation 3.4.2.1.
For λ 6= 0, the substitution z = x + (β/λ)t leads to an equation of the form 3.4.2.2:

∂w ∂2w β ∂w
= a 2 + beλz + c − .
∂t ∂z λ ∂z
∂w ∂2w ∂w
6. =a + (beβt + cx) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.1.6 with f (t) = c and g(t) = beβt .
∂w ∂2w ∂w
7. =a + (bxeβt + c) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.1.6 with f (t) = beβt and g(t) = c.
∂w ∂2w ∂w
8. =a + (bxeβt + ceλt ) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.1.6 with f (t) = beβt and g(t) = ceλt .
∂w ∂2w ∂w
9. =a + x(beβt + ceλt ) .
∂t ∂x2 ∂x

b βt c λt
w(x, t) = AxF (t) + B, F (t) = exp e + e ,
β λ
Z
w(x, t) = Ax2 F 2 (t) + 2Aa F 2 (t) dt + B,
Z
2 2
w(x, t) = A exp µxF (t) + aµ F (t) dt + B.

Z
2 b βt c λt
τ = F (t) dt + A, z = xF (t), where F (t) = exp e + e ,
β λ

∂w ∂ 2w b ∂w
10. = a 2 + cxeβt + .
∂t ∂x x ∂x
This is a special case of equation 3.8.2.6 with f (t) = ceβt .
∂w ∂2w ∂w
3.4.3 Equations of the Form =a + f (x, t) + g(x, t)w
∂t ∂x2 ∂x
∂w ∂2w ∂w
1. =a +b + (ceβt + s)w.
∂t ∂x2 ∂x
The substitution

b c βt b2
w(x, t) = u(x, t) exp − x + e + s − t
2a β 4a
leads to a constant coefficient equation, ∂t u = a∂xx u, which is considered in Section 3.1.1.
∂w ∂2w ∂w
2. =a +b + (ceβx + s)w.
∂t ∂x2 ∂x
b

The substitution w(x, t) = u(x, t) exp − 2a x leads to an equation of the form 3.4.1.3:

∂u ∂2u βx b2
= a 2 + ce + s − u.
∂t ∂x 4a
∂w ∂2w ∂w
3. =a +b + (ceβt + seµt )w.
∂t ∂x2 ∂x
The substitution

b c s b2
w(x, t) = u(x, t) exp − x + eβt + eµt − t
2a β µ 4a
leads to a constant coefficient equation, ∂t u = a∂xx u, which is considered in Section 3.1.1.
∂w ∂2w ∂w
4. =a +b + (ceβx + seµt )w.
∂t ∂x2 ∂x

The substitution w(x, t) = u(x, t) exp µs eµt leads to an equation of the form 3.4.3.2:
∂u ∂2u ∂u
=a 2 +b + ceβx u.
∂t ∂x ∂x
∂w ∂2w ∂w
5. =a + βx + ce2βt w.
∂t ∂x2 ∂x
On passing from t, x to the new variables (A and B are any numbers)
A2 2βt
τ= e + B, z = Axeβt ,
2β
for the function w(τ, z) we obtain a constant coefficient equation of the form 3.1.3:
∂w ∂2w
= a 2 + cA−2 w.
∂τ ∂z
∂w ∂ 2w ∂w
6. = a2 + a1 eβx + b1 + a0 e2βx + b0 eβx + c0 w.
∂t ∂x2 ∂x
The substitution z = eβx leads to an equation of the form 3.3.5.4:
∂w ∂2w ∂w
= a2 β 2 z 2 2 + βz a1 z + b1 + a2 β + a0 z 2 + b0 z + c0 w.
∂t ∂z ∂z
∂w ∂2w ∂w
3.4.4 Equations of the Form = axn + f (x, t) + g(x, t)w
∂t ∂x2 ∂x
∂w ∂ 2w
1. = ax + beβt + cx w.
∂t ∂x2
This is a special case of equation 3.8.8.1 with f (t) = a, g(t) = 0, h(t) = c, and s(t) = beβt .
∂w ∂ 2w
2. = ax + bxeβt + c w.
∂t ∂x2
This is a special case of equation 3.8.8.1 with f (t) = a, g(t) = 0, h(t) = beβt , and s(t) = c.
∂w ∂ 2w ∂w
3. = ax + bx + ceβt + d w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.8.1 with f (t) = a, g(t) = b, h(t) = 0, and s(t) =
ceβt + d.
∂w ∂2w
4. = ax2 + beβt + c w.
∂t ∂x2
This is a special case of equation 3.8.8.2 with f (t) = a, g(t) = 0, and h(t) = beβt + c.
∂w ∂2w
5. = ax4 + beβt + c w.
∂t ∂x2
This is a special case of equation 3.8.8.4 with f (t) = a and g(t) = beβt + c.
∂w ∂2w ∂w
6. = axn + x beβt + c .
∂t ∂x2 ∂x
∂w ∂2w ∂w
7. = axn + bxeβt + ceµt w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.8.7 with f (t) = a, g(t) = beβt , and h(t) = ceµt .
∂w ∂2w ∂w
3.4.5 Equations of the Form = aeβx + f (x, t) + g(x, t)w
∂t ∂x2 ∂x
∂w ∂2w
1. = aeβx .
∂t ∂x2
This is a special case of equation 3.8.6.1 with f (x) = aeβx .
Particular solutions (A, B, and µ are arbitrary constants):
w(x) = Ax + B,

w(x, t) = A aβ 2 t + e−βx + B,

w(x, t) = A aβ 3 tx + βxe−βx + 2e−βx + B,

w(x, t) = A 2a2 β 4 t2 + 4aβ 2 te−βx + e−2βx + B,
r
2 µ 1

w(x, t) = A exp(−µt) J0 exp − 2 βx ,
β a
r
2 µ 1

w(x, t) = A exp(−µt) Y0 exp − 2 βx ,
β a
where J0 (ξ) and Y0 (ξ) are Bessel functions.
∂w ∂2w
2. = aeβx + bw.
∂t ∂x2
w(x, t) = ebt (Ax + B),

w(x, t) = Aebt aβ 2 t + e−βx + Bebt ,

w(x, t) = Aebt aβ 3 tx + βxe−βx + 2e−βx + Bebt ,

w(x, t) = Aebt 2a2 β 4 t2 + 4aβ 2 te−βx + e−2βx ,
r
2 µ+b 1

w(x, t) = A exp(−µt) J0 exp − 2 βx ,
β a
r
2 µ+b 1

w(x, t) = A exp(−µt) Y0 exp − 2 βx ,
β a
where J0 (ξ) and Y0 (ξ) are Bessel functions.

2◦ . The substitution w(x, t) = ebt u(x, t) leads to an equation of the form 3.4.5.1: ∂t u =
aeβx ∂xx u.
∂w ∂2w
3. = aeβx + (beµt + c)w.
∂t ∂x2
1◦ . Particular solutions (A, B, µ are arbitrary constants):

b µt
w(x) = (Ax + B) exp e + ct ,
µ

2 −βx
b µt
w(x, t) = A aβ t + e exp e + ct ,
µ

b µt
w(x, t) = A aβ 3 tx + βxe−βx + 2e−βx exp e + ct ,
µ

2 4 2 2 −βx −2βx
b µt
w(x, t) = A 2a β t + 4aβ te +e exp e + ct ,
µ
r
b µt 2 µ 1

w(x, t) = A exp e + (c − µ)t J0 exp − 2 βx ,
µ β a
r
b µt 2 µ 1

w(x, t) = A exp e + (c − µ)t Y0 exp − 2 βx .
µ β a
b µt

2◦ . The substitution w(x, t) = exp µe + ct u(x, t) leads to an equation of the form
3.4.5.1: ∂t u = aeβx ∂xx u.

∂w ∂2w ∂w
4. = a eβx 2 + βeβx .
∂t ∂x ∂x
This equation describes heat transfer in a quiescent medium (solid body) in the case where
thermal diffusivity is an exponential function of the coordinate. The equation can be rewrit-
ten in the divergence form
∂w ∂ βx ∂w
=a e ,
∂t ∂x ∂x
which is more customary for applications.
1◦ . Particular solutions (A, B, C, and µ are arbitrary constants):

e−βx
w(x, t) = A exp − 2 + B,
aβ t + C
w(x, t) = Aaβ 2 t − A(βx + 1)e−βx + B,
w(x, t) = 2Aaβ 2 te−βx + Ae−2βx + B,
w(x, t) = Aa2 β 4 t2 − 2Aaβ 2 t(βx + 1)e−βx − A(βx + 52 )e−2βx + B,
w(x, t) = Aa2 β 4 t2 e−βx + Aaβ 2 te−2βx + 16 Ae−3βx + B,
w(x, t) = 2Aa3 β 6 t3 e−βx + 3Aa2 β 4 t2 e−2βx + Aaβ 2 te−3βx + 1
12 Ae
−4βx
+ B,
n−1 2
X n(n − 1) . . . (n − k)
−nβx
w(x, t) = e + (aβ 2 t)k e(k−n)βx ,
n(n − k)k!
k=1
√ √
1 2 −µ − 1 βx 2 −µ − 1 βx
w(x, t) = eµt− 2 βx AJ1 √ e 2 + BY1 √ e 2 for µ < 0,
β a β a
√ √
µt− 12 βx 2 µ − 1 βx 2 µ − 1 βx
w(x, t) = e AI1 √ e 2 + BK1 √ e 2 for µ > 0,
β a β a
where J1 (z) and Y1 (z) are Bessel functions, and I1 (z) and K1 (z) are modified Bessel
functions.
2◦ . A solution containing an arbitrary function of the space variable:
∞
X
(at)n n d βx d
w(x, t) = f (x) + L f (x) , L≡ e ,
n! dx dx
n=1
dition w(x, 0) = f (x).
3◦ . A solution containing an arbitrary function of time:
∞
X
−βx 1 (n−1)
w(x, t) = A + e g(t) + 2 2 n−1
e−βnx gt (t),
n=2
n[(n − 1)!] (aβ )
where g(t) is any infinitely differentiable function. If g(t) is a polynomial, then the series
has finitely many terms.
4◦ . The transformation (C1 , C2 , and C3 are any numbers)

e−βx 1 C2 C2
w(x, t) = u(ξ, τ ) exp − 2 , ξ = x − ln , τ = C3 − ,
aβ (t + C1 ) β (t + C1 )2 t + C1
leads to the same equation, up to the notation,

∂u ∂ βξ ∂u
=a e .
∂τ ∂ξ ∂ξ
5◦ . The substitution z = e−βx leads to an equation of the form 3.3.4.1:
∂w ∂2w
= aβ 2 z 2 .
∂t ∂z
6◦ . A series solution of the original equation (under constant values of w at the boundary
and at the initial instant) can be found in Lykov (1967).
∂w ∂2w √ βx √ ∂w
5. = ae2βx + ae a βeβx + b .
∂t ∂x2 ∂x
This is a special case of equation 3.8.5.2 with f (t) = b and g(t) = 0. The substitution
1
ξ = √ (1 − e−βx ) leads to a constant coefficient equation of the form 3.1.4 with Φ ≡ 0:
aβ
∂w ∂2w ∂w
= 2
+b .
∂t ∂ξ ∂ξ
3.5. Equations Containing Hyperbolic Functions and Arbitrary Parameters 349
∂w ∂2w √ βx √ ∂w
6. = ae2βx + ae a βeβx + beµt + ceνt w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.5.2 with f (t) = beµt and g(t) = ceνt .
3.4.6 Other Equations

∂w ∂2w ∂w
1. = aeβt + beµt + ceνt w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.7.3 with f (t) = aeβt , g(t) = beµt , and h(t) = ceνt .
∂w ∂2w ∂w
2. = aeβt + bxeµt + cxeνt w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.7.4 with n(t) = aeβt , f (t) = beµt , g(t) = 0, h(t) = ceνt ,
and s(t) = 0.
∂w ∂2w
3. = aeβx+µt + beνt w.
∂t ∂x2
The transformation

b νt a µt
w(x, t) = exp e u(x, τ ), τ= e
ν µ
leads to an equation of the form 3.4.5.1: ∂τ u = eβx ∂xx u.
3.5 Equations Containing Hyperbolic Functions

3.5.1 Equations Containing a Hyperbolic Cosine
∂w ∂2w
1. =a + (b coshk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.1.1 with f (t) = b coshk ωt + c.
∂w ∂2w
2. =a + (b coshk ωt + cx)w.
∂t ∂x2
This is a special case of equation 3.8.1.6 with f (t) = c and g(t) = b coshk ωt.
∂w ∂2w
3. =a + x(b coshk ωt + c)w.
∂t ∂x2
∂w ∂2w
4. =a + (c coshk ωt − bx2 )w.
∂t ∂x2
This is a special case of equation 3.8.1.7 with f (t) = c coshk ωt.
∂w ∂2w ∂w
5. =a + (b coshk ωt + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
6. =a + x(b coshk ωt + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
7. =a +b + (c coshk ωt + s)w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.3.1 with f (t) = b and g(t) = c coshk ωt + s.

∂w ∂2w b ∂w
8. =a + cx coshk ωt + .
∂t ∂x2 x ∂x
This is a special case of equation 3.8.2.6 with f (t) = c coshk ωt.
∂w ∂ 2w
9. = ax + (bx coshk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.4.1 with f (t) = b coshk ωt and g(t) = c.
∂w ∂2w
10. = ax2 + (b coshk ωt + c)w.
∂t ∂x2
∂w ∂2w
11. = ax4 + (b coshk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.8.4 with f (t) = a and g(t) = b coshk ωt + c.
∂w ∂2w ∂w
12. = axn + x(b coshk ωt + c) .
∂t ∂x2 ∂x
3.5.2 Equations Containing a Hyperbolic Sine
∂w ∂2w
1. =a + (b sinhk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.1.1 with f (t) = b sinhk ωt + c.
∂w ∂2w
2. =a + (b sinhk ωt + cx)w.
∂t ∂x2
This is a special case of equation 3.8.1.6 with f (t) = c and g(t) = b sinhk ωt.
∂w ∂2w
3. =a + x(b sinhk ωt + c)w.
∂t ∂x2
∂w ∂2w
4. =a + (c sinhk ωt − bx2 )w.
∂t ∂x2
This is a special case of equation 3.8.1.7 with f (t) = c sinhk ωt.
∂w ∂2w ∂w
5. =a + (b sinhk ωt + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
6. =a + x(b sinhk ωt + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
7. =a +b + (c sinhk ωt + s)w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.3.1 with f (t) = b and g(t) = c sinhk ωt + s.

∂w ∂2w b ∂w
8. =a + cx sinhk ωt + .
∂t ∂x2 x ∂x
This is a special case of equation 3.8.2.6 with f (t) = c sinhk ωt.
∂w ∂ 2w
9. = ax + (bx sinhk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.4.1 with f (t) = b sinhk ωt and g(t) = c.
∂w ∂2w
10. = ax2 + (b sinhk ωt + c)w.
∂t ∂x2
∂w ∂2w
11. = ax4 + (b sinhk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.8.4 with f (t) = a and g(t) = b sinhk ωt + c.
∂w ∂2w ∂w
12. = axn + x(b sinhk ωt + c) .
∂t ∂x2 ∂x
3.5.3 Equations Containing a Hyperbolic Tangent
∂w ∂2w
1. =a + (b tanhk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.1.1 with f (t) = b tanhk ωt + c.
∂w ∂2w
2. =a + (b tanhk ωt + cx)w.
∂t ∂x2
This is a special case of equation 3.8.1.6 with f (t) = c and g(t) = b tanhk ωt.
∂w ∂2w
3. =a + x(b tanhk ωt + c)w.
∂t ∂x2
∂w ∂2w
4. =a + (c tanhk ωt − bx2 )w.
∂t ∂x2
This is a special case of equation 3.8.1.7 with f (t) = c tanhk ωt.
∂w ∂2w ∂w
5. =a + (b tanhk ωt + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
6. =a + x(b tanhk ωt + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
7. =a +b + (c tanhk ωt + s)w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.3.1 with f (t) = b and g(t) = c tanhk ωt + s.

∂w ∂2w b ∂w
8. =a + cx tanhk ωt + .
∂t ∂x2 x ∂x
This is a special case of equation 3.8.2.6 with f (t) = c tanhk ωt.
∂w ∂ 2w
9. = ax + (bx tanhk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.4.1 with f (t) = b tanhk ωt and g(t) = c.
∂w ∂2w
10. = ax2 + (b tanhk ωt + c)w.
∂t ∂x2
∂w ∂2w
11. = ax4 + (b tanhk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.8.4 with f (t) = a and g(t) = b tanhk ωt + c.
∂w ∂2w ∂w
12. = axn + x(b tanhk ωt + c) .
∂t ∂x2 ∂x
3.5.4 Equations Containing a Hyperbolic Cotangent
∂w ∂2w
1. =a + (b cothk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.1.1 with f (t) = b cothk ωt + c.
∂w ∂2w
2. =a + (b cothk ωt + cx)w.
∂t ∂x2
This is a special case of equation 3.8.1.6 with f (t) = c and g(t) = b cothk ωt.
∂w ∂2w
3. =a + x(b cothk ωt + c)w.
∂t ∂x2
∂w ∂2w
4. =a + (c cothk ωt − bx2 )w.
∂t ∂x2
This is a special case of equation 3.8.1.7 with f (t) = c cothk ωt.
∂w ∂2w ∂w
5. =a + (b cothk ωt + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
6. =a + x(b cothk ωt + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
7. =a +b + (c cothk ωt + s)w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.3.1 with f (t) = b and g(t) = c cothk ωt + s.

∂w ∂2w b ∂w
8. =a + cx cothk ωt + .
∂t ∂x2 x ∂x
This is a special case of equation 3.8.2.6 with f (t) = c cothk ωt.
∂w ∂ 2w
9. = ax + (bx cothk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.4.1 with f (t) = b cothk ωt and g(t) = c.
∂w ∂2w
10. = ax2 + (b cothk ωt + c)w.
∂t ∂x2
∂w ∂2w
11. = ax4 + (b cothk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.8.4 with f (t) = a and g(t) = b cothk ωt + c.
∂w ∂2w ∂w
12. = axn + x(b cothk ωt + c) .
∂t ∂x2 ∂x

3.6 Equations Containing Logarithmic Functions

∂w ∂2w ∂w
∂t ∂x2 ∂x
∂w ∂2w
1. =a + (b ln t + c)w.
∂t ∂x2
This is a special case of equation 3.8.1.1 with f (t) = b ln t + c.
The substitution w(x, t) = u(x, t) exp(bt ln t − bt + ct) leads to the constant coefficient
equation ∂t u = a∂xx2 u, which is considered in Section 3.1.1.
∂w ∂2w
2. =a + (bx + c ln t)w.
∂t ∂x2
This is a special case of equation 3.8.1.6 with f (t) = b and g(t) = c ln t.
∂w ∂2w
3. =a + x(b lnk t + c)w.
∂t ∂x2
This is a special case of equation 3.8.1.3 with f (t) = b lnk t + c.
∂w ∂2w
4. =a + (−bx2 + c lnk t)w.
∂t ∂x2
This is a special case of equation 3.8.1.7 with f (t) = c lnk t.
∂w ∂2w ∂w
5. =a + (b ln t + c) .
∂t ∂x2 ∂x
The change of variable z = x + bt ln t − bt + ct leads to the constant coefficient equation
2 w that is considered in Section 3.1.1.
∂t w = a∂zz
∂w ∂2w ∂w
6. =a + x(b ln t + c) .
∂t ∂x2 ∂x
On passing from t, x to the new variables (A and B are any numbers)
Z
τ = F 2 (t) dt + A, z = xF (t), where F (t) = B exp(bt ln t − bt + ct),
we arrive at the constant coefficient equation ∂τ w = a∂zz w for w(τ, z); this equation is
∂w ∂2w ∂w
3.6.2 Equations of the Form = axk + f (x, t) + g(x, t)w
∂t ∂x2 ∂x
∂w ∂ 2w
1. = ax + (bx + c ln t)w.
∂t ∂x2
This is a special case of equation 3.8.8.1 with f (t) = a, g(t) = 0, h(t) = b, and s(t) = c ln t.
3.6. Equations Containing Logarithmic Functions and Arbitrary Parameters 355
∂w ∂ 2w
2. = ax + (bx ln t + c)w.
∂t ∂x2
This is a special case of equation 3.8.8.1 with f (t) = a, g(t) = 0, h(t) = b ln t, and s(t) = c.
∂w ∂ 2w ∂w
3. = ax + bx + (c ln t + d)w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.8.1 with f (t) = a, g(t) = b, h(t) = 0, and s(t) =
c ln t + d.
∂w ∂2w
4. = ax2 + (b ln t + c)w.
∂t ∂x2
This is a special case of equation 3.8.8.2 with f (t) = a, g(t) = 0, and h(t) = b ln t + c.
∂w ∂2w
5. = ax2 + b ln x w.
∂t ∂x2
This is a special case of equation 3.8.8.3 with n(t) = a, f (t) = g(t) = h(t) = p(t) = 0,
and s(t) = b.
∂w ∂2w
6. = ax2 + btk ln x w.
∂t ∂x2
This is a special case of equation 3.8.8.3 with n(t) = a, f (t) = g(t) = h(t) = p(t) = 0,
and s(t) = btk .
∂w ∂2w
7. = ax2 + b ln2 x w.
∂t ∂x2
and h(t) = b. See also equation 3.8.6.5.
∂w ∂2w
8. = ax2 + btk ln2 x w.
∂t ∂x2
and h(t) = btk .
∂w ∂2w
9. = ax2 + b ln2 x + c ln x ln t + d ln2 t)w.
∂t ∂x2
This is a special case of equation 3.8.8.3 with n(t) = a, f (t) = g(t) = 0, h(t) = b,
s(t) = c ln t, and p(t) = d ln2 t.
∂w ∂2w
10. = ax4 + (b ln t + c)w.
∂t ∂x2
This is a special case of equation 3.8.8.4 with f (t) = a and g(t) = b ln t + c.
∂w ∂2w
11. = axn + (b ln t + c)w.
∂t ∂x2
This is a special case of equation 3.8.8.7 with f (t) = a, g(t) = 0, and h(t) = b ln t + c.
∂w ∂2w ∂w
12. = axn + x(b ln t + c) .
∂t ∂x2 ∂x
3.7 Equations Containing Trigonometric Functions and

Arbitrary Parameters
3.7.1 Equations Containing a Cosine
∂w ∂2w
1. =a + (b cosk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.1.1 with f (t) = b cosk ωt + c.
∂w ∂2w
2. =a + (b cosk ωt + cx)w.
∂t ∂x2
This is a special case of equation 3.8.1.6 with f (t) = c and g(t) = b cosk ωt.
∂w ∂2w
3. =a + x(b cosk ωt + c)w.
∂t ∂x2
∂w ∂2w
4. =a + (c cosk ωt − bx2 )w.
∂t ∂x2
This is a special case of equation 3.8.1.7 with f (t) = c cosk ωt.
∂w ∂2w ∂w
5. =a + (b cosk ωt + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
6. =a + x(b cosk ωt + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
7. =a +b + (c cosk ωt + s)w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.3.1 with f (t) = b and g(t) = c cosk ωt + s.

∂w ∂2w b ∂w
8. = a 2 + cx cosk ωt + .
∂t ∂x x ∂x
This is a special case of equation 3.8.2.6 with f (t) = c cosk ωt.
∂w ∂ 2w
9. = ax + (bx cosk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.4.1 with f (t) = b cosk ωt and g(t) = c.
∂w ∂2w
10. = ax2 + (b cosk ωt + c)w.
∂t ∂x2
∂w ∂2w
11. = ax4 + (b cosk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.8.4 with f (t) = a and g(t) = b cosk ωt + c.
3.7. Equations Containing Trigonometric Functions and Arbitrary Parameters 357
∂w ∂2w ∂w
12. = axn + x(b cosk ωt + c) .
∂t ∂x2 ∂x
3.7.2 Equations Containing a Sine
∂w ∂2w
1. =a + (b sink ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.1.1 with f (t) = b sink ωt + c.
∂w ∂2w
2. =a + (b sink ωt + cx)w.
∂t ∂x2
This is a special case of equation 3.8.1.6 with f (t) = c and g(t) = b sink ωt.
∂w ∂2w
3. =a + x(b sink ωt + c)w.
∂t ∂x2
∂w ∂2w
4. =a + (c sink ωt − bx2 )w.
∂t ∂x2
This is a special case of equation 3.8.1.7 with f (t) = c sink ωt.
∂w ∂2w ∂w
5. =a + (b sink ωt + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
6. =a + x(b sink ωt + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
7. =a +b + (c sink ωt + s)w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.3.1 with f (t) = b and g(t) = c sink ωt + s.

∂w ∂2w b ∂w
8. =a + cx sink ωt + .
∂t ∂x2 x ∂x
This is a special case of equation 3.8.2.6 with f (t) = c sink ωt.
∂w ∂ 2w
9. = ax + (bx sink ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.4.1 with f (t) = b sink ωt and g(t) = c.
∂w ∂2w
10. = ax2 + (b sink ωt + c)w.
∂t ∂x2
∂w ∂2w
11. = ax4 + (b sink ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.8.4 with f (t) = a and g(t) = b sink ωt + c.
∂w ∂2w ∂w
12. = axn + x(b sink ωt + c) .
∂t ∂x2 ∂x
3.7.3 Equations Containing a Tangent
∂w ∂2w
1. =a + (b tank ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.1.1 with f (t) = b tank ωt + c.
∂w ∂2w
2. =a + (b tank ωt + cx)w.
∂t ∂x2
This is a special case of equation 3.8.1.6 with f (t) = c and g(t) = b tank ωt.
∂w ∂2w
3. =a + x(b tank ωt + c)w.
∂t ∂x2
∂w ∂2w
4. =a + (c tank ωt − bx2 )w.
∂t ∂x2
This is a special case of equation 3.8.1.7 with f (t) = c tank ωt.
∂w ∂2w ∂w
5. =a + (b tank ωt + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
6. =a + x(b tank ωt + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
7. =a +b + (c tank ωt + s)w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.3.1 with f (t) = b and g(t) = c tank ωt + s.

∂w ∂2w b ∂w
8. =a + cx tank ωt + .
∂t ∂x2 x ∂x
This is a special case of equation 3.8.2.6 with f (t) = c tank ωt.
∂w ∂ 2w
9. = ax + (bx tank ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.4.1 with f (t) = b tank ωt and g(t) = c.
3.7. Equations Containing Trigonometric Functions and Arbitrary Parameters 359
∂w ∂2w
10. = ax2 + (b tank ωt + c)w.
∂t ∂x2
∂w ∂2w
11. = ax4 + (b tank ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.8.4 with f (t) = a and g(t) = b tank ωt + c.
∂w ∂2w ∂w
12. = axn + x(b tank ωt + c) .
∂t ∂x2 ∂x
3.7.4 Equations Containing a Cotangent
∂w ∂2w
1. =a + (b cotk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.1.1 with f (t) = b cotk ωt + c.
∂w ∂2w
2. =a + (b cotk ωt + cx)w.
∂t ∂x2
This is a special case of equation 3.8.1.6 with f (t) = c and g(t) = b cotk ωt.
∂w ∂2w
3. =a + x(b cotk ωt + c)w.
∂t ∂x2
∂w ∂2w
4. =a + (c cotk ωt − bx2 )w.
∂t ∂x2
This is a special case of equation 3.8.1.7 with f (t) = c cotk ωt.
∂w ∂2w ∂w
5. =a + (b cotk ωt + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
6. =a + x(b cotk ωt + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
7. =a +b + (c cotk ωt + s)w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.3.1 with f (t) = b and g(t) = c cotk ωt + s.

∂w ∂2w b ∂w
8. =a + cx cotk ωt + .
∂t ∂x2 x ∂x
This is a special case of equation 3.8.2.6 with f (t) = c cotk ωt.
∂w ∂ 2w
9. = ax + (bx cotk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.4.1 with f (t) = b cotk ωt and g(t) = c.
∂w ∂2w
10. = ax2 + (b cotk ωt + c)w.
∂t ∂x2
∂w ∂2w
11. = ax4 + (b cotk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.8.4 with f (t) = a and g(t) = b cotk ωt + c.
∂w ∂2w ∂w
12. = axn + x(b cotk ωt + c) .
∂t ∂x2 ∂x
3.8 Equations Containing Arbitrary Functions

∂w ∂2w
∂t ∂x2
∂w ∂2w
1. =a + f (t)w.
∂t ∂x2
hZ i
w(x, t) = (Ax + B) exp f (t) dt ,
hZ i
2
w(x, t) = A(x + 2at) exp f (t) dt ,
h Z i
2
w(x, t) = A exp λx + aλ t + f (t) dt ,
h Z i
2
w(x, t) = A cos(λx) exp −aλ t + f (t) dt ,
h Z i
2
w(x, t) = A sin(λx) exp −aλ t + f (t) dt .
hZ i
2◦ . The substitution w(x, t) = u(x, t) exp f (t) dt leads to a constant coefficient equa-
tion, ∂t u = a∂xx u, which is considered in Section 3.1.1.
∂w ∂2w
2. =a + f (x)w.
∂t ∂x2
This is a special case of equation 3.8.9 with s(x) ≡ 1, p(x) = a = const, q(x) = −f (x),
and Φ ≡ 0.
3.8. Equations Containing Arbitrary Functions 361
∂w ∂2w
3. =a + xf (t)w.
∂t ∂x2
h Z i Z
2
w(x, t) = A exp xF (t) + a F (t) dt , F (t) = f (t) dt,
h Z i h Z i
w(x, t) = A x + 2a F (t) dt exp xF (t) + a F 2 (t) dt ,
h Z Z i
2 2
w(x, t) = A exp xF (t) + λx + aλ t + a F (t) dt + 2aλ F (t) dt .
h Z i Z
2
w(x, t) = u(z, t) exp xF (t) + a F (t) dt , z = x + 2a F (t) dt,
Z
where F (t) = f (t) dt, leads to a constant coefficient equation, ∂t u = a∂zz u, which is
∂w ∂2w
4. =a + x2 f (t)w.
∂t ∂x2
∂w ∂2w
5. =a + f (x) + g(t) w.
∂t ∂x2
1◦ . There are particular solutions in the product form (λ is an arbitrary constant)
h Z i
w(x, t) = exp λt + g(t) dt ϕ(x),
where the function ϕ = ϕ(x) is determined by the ordinary differential equation

aϕ′′xx + f (x) − λ ϕ = 0.
hZ i
◦
2 . The substitution w(x, t) = u(x, t) exp g(t) dt leads to an equation of the form
3.8.1.2:
∂u ∂2u
= a 2 + f (x)u.
∂t ∂x
∂w ∂2w
6. =a + xf (t) + g(t) w.
∂t ∂x2
h Z Z i Z
w(x, t) = A exp xF (t) + a F 2 (t) dt + g(t) dt , F (t) = f (t) dt,
h Z i h Z Z i
w(x, t) = A x + 2a F (t) dt exp xF (t) + a F 2 (t) dt + g(t) dt ,
h Z Z Z i
w(x, t) = exp xF (t) + λx + aλ2 t + 2aλ F (t) dt + a F 2 (t) dt + g(t) dt .
h Z Z i Z
w(x, t) = u(z, t) exp xF (t) + a F 2 (t) dt + g(t) dt , z = x + 2a F (t) dt,
Z
where F (t) = f (t) dt, leads to a constant coefficient equation, ∂t u = a∂zz u, which is
∂w ∂2w
7. =a + −bx2 + f (t) w.
∂t ∂x2
1◦ . Particular solutions (A is an arbitrary constant):
r Z
1 b 2 √
w(x, t) = A exp x + ab t + f (t) dt ,
2 a
r Z
1 b 2 √
w(x, t) = Ax exp x + 3 ab t + f (t) dt .
2 a
2◦ . The transformation (C is any number)

r Z
1 b 2 √
w(x, t) = u(z, τ ) exp x + ab t + f (t) dt ,
2 a
r
√ 1 a √
z = x exp 2 ab t , τ = exp 4 ab t + C
4 b
∂w ∂2w
8. =a + x −bx + f (t) w.
∂t ∂x2
1◦ . Particular solution (A and B are arbitrary constants):
r Z t
1 b 2 √ 2
w(x, t) = exp x + xF (t) + ab t + a F (τ ) dτ ,
2 a A
√ Z t √
F (t) = exp 2 ab t f (τ ) exp −2 ab τ dτ.
B
r
1 b 2 √ 1 √
w(x, t) = exp x u(z, τ ), z = x exp 2 ab t , τ = √ exp 4 ab t
2 a 4 ab

∂u ∂2u 1
= a 2 + zΦ(τ ) + u,
∂τ ∂z 4τ

1 ln τ + ln n √
where Φ(τ ) = 3/2
f , n = 4 ab.
(nτ ) n
∂w ∂2w
9. =a + x2 f (t) + xg(t) + h(t) w.
∂t ∂x2
∂w ∂ 2w
10. =a + f (x − bt)w.
∂t ∂x2
On passing from t, x to the new variables t, ξ = x − bt, we obtain an equation of the form
3.8.6.5:
∂w ∂2w ∂w
=a 2 +b + f (ξ)w.
∂t ∂ξ ∂ξ
∂w ∂2w ∂w
∂t ∂x2 ∂x
∂w ∂2w ∂w
1. =a + f (t) .
∂t ∂x2 ∂x
This equation describes heat transfer in a moving medium where the velocity of motion is
an arbitrary function of time.

Z
w(x, t) = Ax + A f (t) dt + B,
Z 2
w(x, t) = A x + f (t) dt + 2aAt + B,
Z
2
w(x, t) = A exp λx + aλ t + λ f (t) dt + B.
Z
2◦ . On passing from t, x to the new variables t, z = x + f (t) dt, we obtain a constant
∂w ∂2w ∂w
2. =a + f (x) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.6.4. This equation describes heat transfer in a moving
medium where the velocity of motion is an arbitrary function of the coordinate.
1◦ . The equation has particular solutions of the form
w(x, t) = e−λt u(x),
where the function u = u(x) is determined by solving the following ordinary differential
equation with parameter λ:
au′′xx + f (x)u′x + λu = 0.
Other particular solutions (A and B are arbitrary constants):

Z Z
1
w(x) = A + B F (x) dx, F (x) = exp − f (x) dx ,
a
Z Z
dx
w(x, t) = Aat + A F (x) dx + B,
F (x)
Z Z Z
Φ(x) dx
w(x, t) = AatΦ(x) + A F (x) dx, Φ(x) = F (x) dx.
F (x)
More sophisticated solutions are specified below.

2◦ . The original equation admits particular solutions of the form
n
X
wn (x, t) = ti ϕn,i (x) (1)
i=0
for any f (x). Substituting the expression of (1) into the original equation and matching the
coefficients of like powers of t, we arrive at the following system of ordinary differential
equations for ϕn,i = ϕn,i (x):
aϕ′′n,n + f (x)ϕ′n,n = 0,
aϕ′′n,i + f (x)ϕ′n,i = (i + 1)ϕn,i+1 ; i = 0, 1, . . . , n − 1,
where the prime denotes the derivative with respect to x. Integrating these equations suc-
cessively in order of decreasing number i, we obtain the solution (A and B are any num-
bers):
Z Z
1
ϕn,n (x) = A + B F (x) dx, F (x) = exp − f (x) dx ,
a (2)
n−i

ϕn,i (x) = n(n − 1) . . . (i + 1)Lf ϕn,n (x) ; i = 0, 1, . . . , n − 1.
Here, the integral operator Lf is introduced as follows:

Z Z
1 y(x) dx
Lf [y(x)] ≡ F (x) dx. (3)
a F (x)

The powers of the operator are defined by Lif [y(x)] = Lf Li−1
f [y(x)] .
Formulas (1)–(3) give an exact analytical solution of the original equation for arbi-
trary f (x).
A linear combination of particular solutions
N
X
w(x, t) = Cn wn (x, t) (Cn are arbitrary constants)
n=0
is also a particular solution of the original equation.

∂w ∂2w ∂w
3. =a + xf (t) .
∂t ∂x2 ∂x
Generalized Ilkovič equation. The equation describes mass transfer to the surface of a
growing drop that flows out of a thin capillary into a fluid solution (the mass rate of flow of
the fluid moving through the capillary is an arbitrary function of time).
Z
w(x, t) = AxF (t) + B, F (t) = exp f (t) dt ,
Z
w(x, t) = Ax F (t) + 2Aa F 2 (t) dt + B,
2 2
Z
2 2
w(x, t) = A exp λxF (t) + aλ F (t) dt + B.

Z Z
τ = F 2 (t) dt + A, z = xF (t), where F (t) = exp f (t) dt ,
3◦ . Consider the special case where the heat exchange occurs with a semiinfinite medium;
the medium has a uniform temperature w0 at the initial instant t = 0 and the boundary x = 0
is maintained at a constant temperature w1 all the time. In this case, the original equation
subject to the initial and boundary conditions
w → w0 at x → ∞ (boundary condition)
has the solution
Z ξ
w − w1 z 2
= erf √ , erf ξ = √ exp −ζ 2 dζ,
w0 − w1 2 aτ π 0
Z t Z
2
τ= F (ζ) dζ, z = xF (t), F (t) = exp f (t) dt ,
0
where erf ξ is the error function.
∂w ∂2w ∂w
4. =a + f (x − bt) .
∂t ∂x2 ∂x
1◦ . Particular solutions (A and B are arbitrary constants):
Z Z
1 b
w(x, t) = A + B F (z) dz, F (z) = exp − f (z) dz − z ,
a a
Z Z
dz
w(x, t) = Aat + A F (z) dz,
F (z)
Z Z Z
Φ(z) dz
w(x, t) = AatΦ(z) + A F (z) dx, Φ(z) = F (z) dz,
F (z)
where z = x − bt.
2◦ . On passing from t, x to the new variables t, z = x − bt, we obtain a separable equation

∂w ∂2w ∂w
= a 2 + f (z) + b .
∂t ∂z ∂z

∂w ∂2w 1 x ∂w
5. =a 2 + √ f √ .
∂t ∂x t t ∂x
√
1◦ . On passing from t, x to the new variables τ = ln t, ξ = x/ t, we obtain a separable
equation of the form 3.8.2.2:
∂w ∂2w ∂w
= a 2 + f (ξ) + 12 ξ .
∂τ ∂ξ ∂ξ
2◦ . Consider the special case where the heat exchange occurs with a semiinfinite medium;
the medium has a uniform temperature w0 at the initial instant t = 0 and the boundary x = 0
is maintained at a constant temperature w1 all the time. In this case, the original equation
subject to the initial and boundary conditions
w → w0 at x → ∞ (boundary condition)
has the solution

R∞ Z
ξ exp −Φ(ξ) dξ
ξ
w − w0 1 2 1
= R∞ , Φ(ξ) = ξ + f (ξ) dξ,
w1 − w0 0 exp −Φ(ξ) dξ 4a a 0
√
where ξ = x/ t.
h i
∂w ∂2w b ∂w
6. =a + xf (t) + .
∂t ∂x2 x ∂x
Z
n b
w(x, t) = Ax exp n f (t) dt + B, n = 1 − ,
a
Z Z
2
w(x, t) = Ax F (t) + 2A(a + b) F (t) dt + B, F (t) = exp 2 f (t) dt .

Z Z
2
τ = F (t) dt + A, z = xF (t), where F (t) = exp f (t) dt ,
for the function w(τ, z) we obtain a simpler equation
∂w ∂2w b ∂w
=a 2 + ,
∂τ ∂z z ∂z
which is considered in Sections 3.2.1, 3.2.3, and 3.2.5.
∂w ∂2w ∂w
7. =a + xf (t) + g(t) .
∂t ∂x2 ∂x
The transformation (A, B, and C are any numbers)
Z Z Z
2
τ= F (t) dt + A, z = xF (t) + g(t)F (t) dt + C, F (t) = B exp f (t) dt ,
leads to a constant coefficient equation, ∂τ w = a∂zz w, which is considered in Section 3.1.1.
∂w ∂2w f (t) ∂w
8. =a + .
∂t ∂x2 x ∂x
Particular solutions:
Z
A x2 1 f (t)
w = √ exp − − dt + B,
t 4at 2a t
Z
w = x2 + 2at + 2 f (t) dt + A,
w = x4 + p(t)x2 + q(t),
where
Z Z

p(t) = 12at + 4 f (t) dt + A, q(t) = 2 a + f (t) p(t) dt + B,
with A and B being arbitrary constants. The second and third solutions are special cases of
a solution having the form
w = x2n + A2n−2 (t)x2n−2 + · · · + A2 (t)x2 + A0 (t),
which contains n arbitrary constants.

h i
∂w ∂2w g(t) ∂w
9. =a + xf (t) + .
∂t ∂x2 x ∂x
On passing from t, x to the new variables (A is any number)
Z Z
2
τ= F (t) dt + A, z = xF (t), where F (t) = exp f (t) dt ,
for the function w(τ, z) we obtain a simpler equation of the form 3.8.2.8:
∂w ∂2w ϕ(τ ) ∂w
=a 2 + .
∂τ ∂z z ∂z
The function ϕ = ϕ(τ ) is defined parametrically as

Z
g(t)
ϕ= , τ= F 2 (t) dt + A.
F (t)
∂w ∂2w ∂w
∂t ∂x2 ∂x
∂w ∂2w ∂w
1. =a + f (t) + g(t)w.
∂t ∂x2 ∂x
Z Z

w(x, t) = Ax + AF (t) + B exp g(t) dt , F (t) = f (t) dt,
Z
2
w(x, t) = A x + F (t) + 2at exp g(t) dt ,
Z
2
w(x, t) = A exp aλ t + g(t) dt ± λF (t) ± λx ,
Z
2

w(x, t) = A exp −aλ t + g(t) dt cos λx + λF (t) ,
Z
2

w(x, t) = A exp −aλ t + g(t) dt sin λx + λF (t) .
Z Z
w(x, t) = u(z, t) exp g(t) dt , z =x+ f (t) dt
∂w ∂2w ∂w
2. =a + f (x) + g(t)w.
∂t ∂x2 ∂x
Z
w(x, t) = A + B F (x) dx G(t),
Z Z
dx
w(x, t) = A at + F (x) dx G(t),
F (x)
Z Z
Ψ(x) dx
w(x, t) = A atΨ(x) + F (x) dx G(t).
F (x)
The following notation is used here:

Z Z Z
1
G(t) = exp g(t) dt , F (x) = exp − f (x) dx , Ψ(x) = F (x) dx.
a
Z
2◦ . The substitution w(x, t) = u(x, t) exp g(t) dt leads to an equation of the form
3.8.2.2:
∂u ∂2u ∂u
= a 2 + f (x) .
∂t ∂x ∂x
∂w ∂2w ∂w
3. =a + f (x) + g(x)w.
∂t ∂x2 ∂x
∂w ∂2w ∂w
4. =a + xf (t) + g(t)w.
∂t ∂x2 ∂x

Z Z
2
τ= F (t) dt + A, z = xF (t) + C, w(t, x) = u(τ, z) exp g(t) dt ,
Z
where F (t) = B exp f (t) dt , leads to a constant coefficient equation, ∂τ u = a∂zz u,
h i
∂w ∂2w b ∂w
5. =a + xf (t) + + g(t)w.
∂t ∂x2 x ∂x
Z
The substitution w(x, t) = u(x, t) exp g(t) dt leads to an equation of the form 3.8.2.6:

∂u ∂2u b ∂u
= a 2 + xf (t) + .
∂t ∂x x ∂x
For the special case b = 0, see equation 3.8.2.3.
∂w ∂2w ∂w
6. =a + [xf (t) + g(t)] + h(t)w.
∂t ∂x2 ∂x

Z Z Z
τ= F 2 (t) dt+A, z = xF (t)+ g(t)F (t) dt+C, w(t, x) = u(τ, z) exp h(t) dt ,
Z
where F (t) = B exp f (t) dt , leads to a constant coefficient equation, ∂τ u = a∂zz u,
∂w ∂2w ∂w
7. =a + [xf (t) + g(t)] + [xh(t) + s(t)]w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.7.4 with n(t) = a.
∂w ∂2w ∂w
8. =a + [xf (t) + g(t)] + [x2 h(t) + xs(t) + p(t)]w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.7.5 with n(t) = a.

∂w ∂2w f (t) ∂w
9. =a + + g(t)w.
∂t ∂x2 x ∂x
1◦ . Particular solutions (A, B, and C are arbitrary constants):
Z Z
A x2 1 f (t)
w = √ exp − − dt + g(t) dt ,
t 4at 2a t
Z Z
2
w = A exp g(t) dt x + 2at + 2 f (t) dt + B ,
Z

w = A exp g(t) dt x4 + p(t)x2 + q(t) ,
where
Z Z

p(t) = 12at + 4 f (t) dt + B, q(t) = 2 a + f (t) p(t) dt + C.
Z
2◦ . The substitution w = exp g(t) dt u(x, t) leads to an equation of the form 3.8.2.8
for u = u(x, t).
h i
∂w ∂ 2w g(t) ∂w
10. =a + xf (t) + + h(t)w.
∂t ∂x2 x ∂x
Z
The substitution w = exp h(t) dt u(x, t) leads to an equation of the form 3.8.2.9 for
u = u(x, t).
∂w ∂2w ∂w
3.8.4 Equations of the Form = axn + f (x, t) + g(x, t)w
∂t ∂x2 ∂x
∂w ∂ 2w
1. = ax + [xf (t) + g(t)]w.
∂t ∂x2
∂w ∂2w
2. = ax2 + f (t)w.
∂t ∂x2
This is a special case of equation 3.8.8.2. The transformation
Z
w(x, t) = u(z, t) exp f (t) dt , z = ln |x|
leads to a constant coefficient equation of the form 3.1.4:
∂u ∂2u ∂u
=a 2 −a .
∂τ ∂z ∂z
∂w ∂2w
3. = ax2 + ln x f (t)w.
∂t ∂x2
∂w ∂2w
4. = ax2 + ln2 x f (t)w.
∂t ∂x2
∂w ∂2w ∂w
5. = axn + xf (t) .
∂t ∂x2 ∂x
Z
w(x, t) = Ax exp f (t) dt + B,
Z
2−n
w(x, t) = Ax F (t) + Aa(n − 1)(n − 2) F (t) dt,
where Z
F (t) = exp (2 − n) f (t) dt .
2◦ . On passing from t, x to the new variables

Z
z = xF (t), τ =a F 2−n (t) dt,
where Z
F (t) = exp f (t) dt ,
we obtain an equation of the form 3.3.6.6:
∂w ∂2w
= zn 2 .
∂τ ∂z
∂w ∂2w ∂w
6. = axn + xf (t) + bw.
∂t ∂x2 ∂x
The substitution w(x, t) = ebt u(x, t) leads to an equation of the form 3.8.4.5:
∂u ∂2u ∂u
= axn 2 + xf (t) .
∂t ∂x ∂x
∂w ∂2w √ n√ ∂w
7. = ax2n + ax a nxn−1 + f (t) + g(t)w.
∂t ∂x2 ∂x
The substitution  1−n
x
1  if n 6= 1,
ξ= √ 1−n
a
ln |x| if n = 1
leads to a special case of equation 3.8.7.3, namely,
∂w ∂2w ∂w
= 2
+ f (t) + g(t)w.
∂t ∂ξ ∂ξ
∂w ∂2w ∂w
3.8.5 Equations of the Form = aeβx + f (x, t) + g(x, t)w
∂t ∂x2 ∂x
∂w ∂2w
1. = aeβx + f (t)w.
∂t ∂x2
Z
The substitution w(x, t) = u(x, t) exp f (t) dt leads to an equation of the form 3.4.5.1:
∂u ∂2u
= aeβx 2 .
∂t ∂x
∂w ∂2w √ βx √ ∂w
2. = ae2βx + ae a βeβx + f (t) + g(t)w.
∂t ∂x2 ∂x
1
The substitution ξ = √ 1 − e−βx leads to a special case of equation 3.8.7.3:
β a
∂w ∂2w ∂w
= 2
+ f (t) + g(t)w.
∂t ∂ξ ∂ξ
∂w ∂2w ∂w
3. = aeβx + f (x) .
∂t ∂x2 ∂x
w(x, t) = e−λt u(x), (1)
where the function u(x) is determined by solving the following linear ordinary differential
aeβx u′′xx + f (x)u′x + λu = 0. (2)

Z Z
1 −βx
w(x) = A + B F (x) dx, F (x) = exp − e f (x) dx ,
a
Z Z
dx
w(x, t) = Aat + A F (x) dx,
eβx F (x)
Z Z Z
Φ(x) dx
w(x, t) = AatΦ(x) + A F (x) dx, Φ(x) = F (x) dx.
eβx F (x)
∂w ∂2w ∂w
4. = aeβx + f (x) + g(t)w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.6.6. Z
The substitution w(x, t) = u(x, t) exp g(t) dt leads to an equation of the form
3.8.5.3:
∂u ∂2u ∂u
= aeβx 2 + f (x) .
∂t ∂x ∂x
∂w ∂ 2w ∂w
3.8.6 Equations of the Form = f (x) + g(x, t) + h(x, t)w
∂t ∂x2 ∂x
∂w ∂ 2w
1. = f (x) .
∂t ∂x2
This is an equation of the form 3.8.9 with s(x) = 1/f (x), p(x) ≡ 1, q(x) ≡ 0, and
Φ(x, t) ≡ 0.
w(x, t) = e−λt u(x), (1)
f (x)u′′xx + λu = 0. (2)
The procedure for constructing solutions to specific boundary value problems for the orig-
inal equations with the help of particular solutions of the form (1) is described in detail in
Sections 15.1.1 and 15.1.2.
The main problem here is to investigate the auxiliary equation (2), which is far from
always admitting a closed-form solution; therefore, recourse to numerical solution methods
is often necessary. Many specific solvable equations of the form (2) can be found in the
handbooks by Murphy (1960), Kamke (1977), and Polyanin and Zaitsev (2003).
2◦ . Particular solutions (A, B, and x0 are arbitrary constants):
w(x) = Ax + B,
Z x
x−ξ
w(x, t) = At + AF (x), F (x) = dξ,
x0 f (ξ)
Z x
x−ξ
w(x, t) = Atx + AG(x), G(x) = ξ dξ,
f (ξ)
Z x x0
x−ξ
w(x, t) = At2 + 2AtF (x) + 2A F (ξ) dξ,
x0 f (ξ)
Z x
2 x−ξ
w(x, t) = At x + 2AtG(x) + 2A G(ξ) dξ.
x0 f (ξ)
More sophisticated solutions are specified below in Item 3◦ .

3◦ . For any function f (x), the original equation admits exact analytical solutions of the
form
n−1
X
n
wn (x, t) = t + ti ϕn,i (x). (3)
i=0
Substituting expression (3) into the original equation and matching the coefficients of like
powers of t, we arrive at the following system of ordinary differential equations for ϕn,i =
ϕn,i (x):
f (x)ϕ′′n,i = (i + 1)ϕn,i+1 ,
i = 0, 1, . . . , n − 1; ϕn,n ≡ 1,
where the prime stands for the differentiation with respect to x. Integrating these equations
successively in order of decreasing number i, we obtain
ϕn,i (x) = n(n − 1) . . . (i + 1)Ln−i

f [1]. (4)

Z Z Z x
y(x) x−ξ
Lf [y(x)] ≡ dx dx = y(ξ) dξ + Ax + B, (5)
f (x) x0 f (ξ)
where x0 , A, and B are arbitrary constants.

i−1 The powers of the operator are defined by the
i
usual relation Lf [y(x)] = Lf Lf [y(x)] ; generally speaking, the constants A and B are
not the same in repeated actions of Lf in this formula.
Formulas (3) and (4) determine an exact analytical solution of the original equation for
arbitrary f (x).
A linear combination of particular solutions (3),
N
X
w(x, t) = Cn wn (x, t) (Cn are arbitrary constants)
n=0
is also a particular solution of the original equation.

The original equation also admits other exact analytical solutions, specifically,
n−1
X
wn (x, t) = tn x + ti φn,i (x), φn,i (x) = n(n − 1) . . . (i + 1)Ln−i
f [x],
i=0
where n is a positive integer and the operator Lf is given by relation (5). A linear combi-
nation of these solutions with a linear combination of solutions (3) is also a solution of the
original equation.
For the structure of other particular solutions, see equation 3.8.6.5 (Remark 3.7).
4◦ . The equation admits the following infinite-series solution that contains an arbitrary
function of the coordinate:
∞
X 1 n n d2
w(x, t) = Θ(x) + t L [Θ(x)], L ≡ f (x) ,
n! dx2
n=1
where Θ(x) is any infinitely differentiable function. This solution satisfies the initial con-
dition w(x, 0) = Θ(x).
5◦ . Below are two discrete transformations that preserve the form of the original equation;
the function f is subject to changes.
5.1. The transformation
1 w
z= , u= (point transformation)
x x
leads to a similar equation 2
∂u 1 ∂ u
= z4f .
∂t z ∂z 2
5.2. First, we perform the change of variable

Z
dx
ξ=
f (x)
to obtain the equation
∂w ∂ ∂w
= F (ξ) ,
∂t ∂ξ ∂ξ
where the function F = F (ξ) is defined parametrically as
Z
1 dx
F = , ξ= .
f (x) f (x)
Introducing the new dependent variable v = v(ξ, t) by the formula
∂v
w= (Bäcklund transformation)
∂ξ
and integrating the resulting equation with respect to ξ, we arrive at the desired equation
∂v ∂2v
= F (ξ) 2 .
∂t ∂ξ
(Here, the function v is defined up to an arbitrary additive term that depends on t.)
For power-law and exponential functions the above transformation acts as follows:
n
f (x) = bxn =⇒ F (ξ) = Aξ n−1 ,
f (x) = be−βx =⇒ F (ξ) = βξ,
1 n
where A = b(1 − n) n−1 .
b
∂w ∂ 2w
2. = f (x) + Φ(x, t).
∂t ∂x2
This is an equation of the form 3.8.9 with s(x) = 1/f (x), p(x) ≡ 1, and q(x) ≡ 0. For
Φ(x, t) ≡ 0, see equation 3.8.6.1.
1◦ . For
Φ(x, t) = gn (x)tn (n = 0, 1, 2, . . . )
and arbitrary functions f (x) and gn (x), the original equation has a particular solution of
the form
Xn
w̄n (x, t) = ti ψn,i (x). (1)
i=0
The functions ψn,i = ψn,i (x) are calculated by the formulas
(
−Lf gn (x) if i = n,
ψn,i (x) = n−i+1
(2)
−n(n − 1) . . . (i + 1)Lf gn (x) if i = 0, 1, . . . , n − 1
with the aid of the integral operator Lf that is defined by relation (5) in equation 3.8.6.1.
2◦ . If the nonhomogeneous part of the equation can be represented in the form
N
X
Φ(x, t) = gn (x)tn ,
n=1
then there is a particular solution that is the sum of particular solutions of the form (1):
N
X
w̄(x, t) = w̄n (x, t).
n=1
For example, if
Φ(x, t) = g(x)t + h(x),
where g(x) and h(x) are arbitrary functions, the original equation has a solution of the
form Z x
ψ(ξ) + h(ξ)
w̄(x, t) = −tψ(x) − (x − ξ) dξ,
x0 f (ξ)
Z x
g(ξ)
ψ(x) = (x − ξ) dξ, x0 is any.
x0 f (ξ)
For the structure of particular solutions for other Φ(x, t), see equation 3.8.6.5, Item 3◦ .
By summing different solutions of the homogeneous equation (see equation 3.8.6.1)
and any particular solution of the nonhomogeneous equation, one can obtain a wide class
of particular solutions of the nonhomogeneous equation.
h i
∂w ∂ ∂w
3. = f (x) .
∂t ∂x ∂x
This is a special case of equation 3.8.6.4 with g(x) = fx′ (x). The equation describes heat
transfer in a quiescent medium (solid body) in the case where the thermal diffusivity f (x)
is a coordinate dependent function.
Z
dx
w(x) = A + B ,
f (x)
Z
x dx
w(x, t) = At + A + B,
f (x)
Z Z Z
dx dx
w(x, t) = Atϕ(x) + A ϕ(x) dx + B, ϕ(x) = ,
f (x) f (x)
Z Z Z
2 dx x dx
w(x, t) = At + 2Atψ(x) + 2A ψ(x) dx + B, ψ(x) = ,
f (x) f (x)
Z Z
2 dx
w(x, t) = At ϕ(x) + 2AtI(x) + 2A I(x) dx + B,
f (x)
Z Z
dx
I(x) = ϕ(x) dx .
f (x)
2◦ . A solution in the form of an infinite series:

X∞
1 n n d d
w(x, t) = Θ(x) + t L Θ(x) , L≡ f (x) .
n! dx dx
n=1
It contains an arbitrary function of the space variable, Θ = Θ(x). This solution satisfies the
initial condition w(x, 0) = Θ(x).
Z Z
2 dx
w(x, t) = ϕ(x)u(ξ, t), ξ=− ϕ (x) dx, ϕ(x) = ,
f (x)
leads to the analogous equation

∂u ∂ ∂u
= F (ξ) ,
∂t ∂ξ ∂ξ
where the function F (ξ) is defined parametrically as
Z Z
dx
F (ξ) = f (x)ψ (x), ξ = − ϕ2 (x) dx,
4
ϕ(x) = .
f (x)
Z
dx
4◦ . The substitution z = leads to an equation of the form 3.8.6.1:
f (x)
∂w ∂2 w
= g(z) 2 ,
∂t ∂z
where the function g(z) is defined parametrically as
Z
1 dx
g(z) = , z= .
f (x) f (x)
∂w ∂ 2w ∂w
4. = f (x) + g(x) .
∂t ∂x2 ∂x
1◦ . This equation can be rewritten in the form

∂w ∂ ∂w
s(x) = p(x) , (1)
∂t ∂x ∂x
where Z Z
1 g(x) g(x)
s(x) = exp dx , p(x) = exp dx .
f (x) f (x) f (x)
For solutions of equation (1), see Section 3.8.9 with q(x) ≡ 0.
2◦ . There are particular solutions of the form
w(x, t) = e−λt u(x), (2)
where the function u(x) is identified by solving the following linear ordinary differential
f (x)u′′xx + g(x)u′x + λu = 0. (3)
A procedure for constructing solutions to specific boundary value problems for the original
equation with the aid of particular solutions (2) is described in detail in Sections 15.1.1 and
15.1.2. A good deal of specific solvable equations of the form (3) can be found in Murphy
(1960), Kamke (1977), and Polyanin and Zaitsev (2003).

Z Z
g(x)
w(x) = A + B F (x) dx, F (x) = exp − dx ,
f (x)
Z Z
dx
w(x, t) = At + A F (x) dx,
f (x)F (x)
Z Z Z
Φ(x) dx
w(x, t) = AtΦ(x) + A F (x) dx, Φ(x) = F (x) dx.
f (x)F (x)
More sophisticated solutions are presented below in Item 4◦ .

4◦ . For any f (x) and g(x), the original equation admits particular solutions of the form
n
X
wn (x, t) = ti ϕn,i (x). (4)
i=0
Substituting expression (4) into the original equation and matching the coefficients of like
powers of t, we arrive at the following system of ordinary differential equations for ϕn,i =
ϕn,i (x):
f (x)ϕ′′n,n + g(x)ϕ′n,n = 0,
f (x)ϕ′′n,i + g(x)ϕ′n,i = (i + 1)ϕn,i+1 , i = 0, 1, . . . , n − 1,
where the prime stands for the differentiation with respect to x. Integrating these equations
successively in order of decreasing number i, we obtain (A and B are any numbers)
Z Z
g(x)
ϕn,n (x) = A + B F (x) dx, F (x) = exp − dx ,
f (x) (5)
n−i

ϕn,i (x) = n(n − 1) . . . (i + 1)Lf ϕn,n (x) ; i = 0, 1, . . . , n − 1.

Z Z
y(x) dx
Lf [y(x)] ≡ F (x) dx. (6)
f (x)F (x)

The powers of the operator are defined as Lif [y(x)] = Lf Li−1
f [y(x)] .
Formulas (4)–(6) determine an exact analytical solution of the original equation for
arbitrary f (x).
A linear combination of particular solutions (4),
N
X
w(x, t) = Cn wn (x, t) (Cn are arbitrary numbers),
n=0
is also a particular solution of the homogeneous equation.

For the structure of other particular solutions, see equation 3.8.6.5 (Remark 3.7).
Z Z
g(x)
5◦ . The substitution ξ = ϕ(x) dx, ϕ(x) = exp − dx leads to an equation of
f (x)
the form 3.8.6.1:
∂w ∂2w
= F (ξ) 2 ,
∂t ∂ξ
where the function F = F (ξ) is determined by eliminating x from the relations
Z
2
F = f (x)ϕ (x), ξ = ϕ(x) dx.
6◦ . An infinite series solution containing an arbitrary function of the coordinate:
X∞
1 n n d2 d
w(x, t) = Θ(x) + t L [Θ(x)], L ≡ f (x) 2
+ g(x) ,
n=1
n! dx dx
where Θ(x) is any infinitely differentiable function. This solution satisfies the initial con-
dition w(x, 0) = Θ(x).
∂w ∂ 2w ∂w
5. = f (x) + g(x) + h(x)w + Φ(x, t).
∂t ∂x2 ∂x
1◦ . This equation can be rewritten in the form

∂w ∂ ∂w
s(x) = p(x) − q(x)w + s(x)Φ(x, t), (1)
∂t ∂x ∂x
where
Z Z
1 g(x) g(x)
s(x) = exp dx , p(x) = exp dx ,
f (x) f (x) f (x)
Z
h(x) g(x)
q(x) = − exp dx .
f (x) f (x)
For solutions of equation (1), see Section 3.8.9.

2◦ . Consider the homogeneous equation, i.e., the case Φ(x, t) ≡ 0.
2.1. There are particular solutions of the form
w(x, t) = e−λt u(x),

f (x)u′′xx + g(x)u′x + h(x) + λ u = 0.
2.2. Suppose we know a nontrivial particular solution w0 = w0 (x) of the ordinary

differential equation
f (x)w0′′ + g(x)w0′ + h(x)w0 = 0 (2)
that corresponds to the stationary case (∂t w ≡ 0). Then the functions
Z Z
F g
w(x) = Aw0 + Bw0 dx, F = exp − dx ,
w02 f
Z Z
F w02
w(x, t) = Atw0 + Aw0 dx dx,
w02 fF
Z Z Z
F w02 Ψ F
w(x, t) = Atw0 Ψ + Aw0 2 dx dx, Ψ = dx,
w0 fF w02
where A and B are arbitrary constants, are also particular solutions of the original equation.
By performing the change of variable w(x, t) = w0 (x)u(x, t), we arrive at the simpler
equation
∂u ∂2u w0′ (x) ∂u
= f (x) 2 + 2f (x) + g(x) .
∂t ∂x w0 (x) ∂x
It determines a wide class of more complicated analytical solutions of the original equation.
It follows, with reference to the results of Item 4◦ from 3.8.6.4, that any nontrivial par-
ticular solution of the auxiliary linear ordinary differential equation (2) generates infinitely
many particular solutions of the original partial differential equation.
For the structure of other particular solutions, see the remark at the end of Item 3◦ .
2.3. Let a particular nonstationary solution w0 = w0 (x, t) (∂t w0 6≡ 0) of the homoge-
neous equation be known. Then the functions
∂ n w0
wn (x, t) = (x, t),
∂tn
obtained by differentiating the solution w0 with respect to t, are also particular solutions of
the equation in question.
In addition, a new particular solution can be sought in the form
Z t
w̄(x, t) = w0 (x, τ ) dτ + φ(x), (3)
t0
where the unknown function φ(x) is determined on substituting expression (3) into the
original equation. On constructing solution (3), one can use the above approach to construct
another solution, and so on.
2.4. Case h(x) = h = const. Particular solutions (A and B are arbitrary constants):
Z Z
ht g(x)
w(x, t) = e A + B F (x) dx , F (x) = exp − dx ,
f (x)
Z Z
dx
w(x, t) = Aeht t + F (x) dx ,
f (x)F (x)
Z Z Z
ht Ψ(x) dx
w(x, t) = Ae tΨ(x) + F (x) dx , Ψ(x) = F (x) dx.
f (x)F (x)
The substitution w(x, t) = eht v(x, t) leads to an equation of the form 3.8.6.4:
∂v ∂2v ∂v
= f (x) 2 + g(x) .
∂t ∂x ∂x
3◦ . The structure of particular solutions w̄(x, t) of the nonhomogeneous equation 3.8.6.5

for some functions Φ(x, t) is presented in Table 3.4.
Remark 3.7. The homogeneous equation (with Φ ≡ 0) admits all the particular solutions spec-
ified in Table 3.4. In this case, n should be assumed an integer and β and λ arbitrary numbers.
TABLE 3.4
Structure of particular solutions of linear nonhomogeneous equations of special form
No Functions Φ(x, t) Form of particular solutions w̄(x, t) Remarks
P
n n is an integer; the equations
1 ϕ(x)tn ψm (x)tm for ψm (x) are solved consecutively,
m=0
starting with m = n
2 ϕ(x)eβt ψ(x)eβt ψ(x) is governed by a single equation
P
n n is an integer; the equations
3 ϕ(x)tn eβt eβt ψm (x)tm for ψm (x) are solved consecutively,
m=0
starting with m = n
the equations for ψ(x) and χ(x)
4 ϕ(x) sinh(βt) ψ(x)eβt + χ(x)e−βt
are independent
the equations for ψ(x) and χ(x)
5 ϕ(x) cosh(βt) ψ(x)eβt + χ(x)e−βt
are independent
ψ(x) and χ(x) are determined

6 ϕ(x) sin(βt) ψ(x) sin(βt) + χ(x) cos(βt)
by a system of equations

7 ϕ(x) cos(βt) ψ(x) sin(βt) + χ(x) cos(βt)

8 ϕ(x)eλt sin(βt) ψ(x)eλt sin(βt) + χ(x)eλt cos(βt)
9 ϕ(x)eλt cos(βt) ψ(x)eλt sin(βt) + χ(x)eλt cos(βt)
∂w ∂ 2w ∂w
6. = f (x) + g(x) + h(t)w.
∂t ∂x2 ∂x

Z
w(x, t) = A + B F (x) dx H(t),
Z Z
dx
w(x, t) = A t + F (x) dx H(t),
f (x)F (x)
Z Z
Ψ(x) dx
w(x, t) = A tΨ(x) + F (x) dx H(t),
f (x)F (x)
where
Z Z Z
g(x)
H(t) = exp h(t) dt , F (x) = exp − dx , Ψ(x) = F (x) dx.
f (x)
Z
◦
2 . The substitution w(x, t) = u(x, t) exp h(t) dt leads to an equation of the form
3.8.6.4:
∂u ∂2u ∂u
= f (x) 2 + g(x) .
∂t ∂x ∂x
∂w ∂ 2w ∂w
7. = f (x) + g(x)+ [h1 (x) + h2 (t)]w.
∂t ∂x2 ∂x
Z
The substitution w(x, t) = u(x, t) exp h2 (t) dt leads to an equation of the form
3.8.6.5:
∂u ∂2u ∂u
= f (x) 2 + g(x) + h1 (x)u.
∂t ∂x ∂x
∂w ∂ 2w ∂w
8. = f 2 (x) + f (x)[fx′ (x) + g(t)] + h(t)w.
∂t ∂x2 ∂x
Z
dx
The change of variable ξ = leads to a special case of equation 3.8.7.3:
f (x)
∂w ∂2w ∂w
= 2
+ g(t) + h(t)w.
∂t ∂ξ ∂ξ
∂w ∂2w ∂w
9. = f 2 2 + f fx′ + 2g + ϕ + (f gx′ + g 2 + gϕ + ψ)w,
∂t ∂x ∂x
where f = f (x), g = g(x), ϕ = ϕ(t), ψ = ψ(t).
The transformation
Z Z
g dx
w(x, t) = u(ξ, t) exp − dx , ξ=
f f (x)
leads to a special case of equation 3.8.7.3:
∂u ∂2u ∂u
= 2
+ ϕ(t) + ψ(t)u.
∂t ∂ξ ∂ξ
∂w ∂2w ∂w
3.8.7 Equations of the Form = f (t) + g(x, t) + h(x, t)w
∂t ∂x2 ∂x
∂w ∂ 2w ∂w
1. = f (t) + g(t) .
∂t ∂x2 ∂x
Z
w(x, t) = Ax + A g(t) dt + B,
Z 2 Z
w(x, t) = A x + g(t) dt + 2A f (t) dt + B,
Z Z
2
w(x, t) = A exp λx + λ f (t) dt + λ g(t) dt .
2◦ . On passing from t, x to the new variables (A and B are any numbers)

Z Z
τ = f (t) dt + A, z = x + g(t) dt + B,
for the function w(τ, z) we obtain a constant coefficient equation, ∂τ w = ∂zz w, which is
∂w ∂ 2w ∂w
2. = f (t) + xg(t) .
∂t ∂x2 ∂x
Z
w(x, t) = AxG(t) + B, G(t) = exp g(t) dt ,
Z
w(x, t) = Ax G (t) + 2A f (t)G2 (t) dt + B,
2 2
Z
2 2
w(x, t) = A exp λxG(t) + λ f (t)G (t) dt .

Z Z
2
τ = f (t)G (t) dt + A, z = xG(t), where G(t) = exp g(t) dt ,
for the function w(τ, z) we obtain a constant coefficient equation, ∂τ w = ∂zz w, which is
∂w ∂ 2w ∂w
3. = f (t) + g(t) + h(t)w.
∂t ∂x2 ∂x
The transformation
Z Z Z
w(x, t) = u(z, τ ) exp h(t) dt , z =x+ g(t) dt, τ= f (t) dt
∂w ∂ 2w ∂w
4. = n(t) + xf (t) + g(t) + xh(t) + s(t) w.
∂t ∂x2 ∂x
Let us perform the transformation

w(x, t) = exp xα(t) + β(t) u(z, τ ), τ = ϕ(t), z = xψ(t) + χ(t), (1)
where the unknown functions α(t), β(t), ϕ(t), ψ(t), and χ(t) are chosen so that the result-
ing equation is as simple as possible. For the new dependent variable u(z, τ ) we have
∂u ∂2u ∂u
ϕ′t = nψ 2 2 + x(f ψ − ψt′ ) + 2nψα + gψ − χ′t
∂τ ∂z ∂z
′ 2 ′
+ x(f α + h − αt ) + nα + gα + s − βt u.
Let the unknown functions satisfy the system of ordinary differential equations
ϕ′t = nψ 2 , (2)
ψt′ = f ψ, (3)
χ′t = 2nαψ + gψ, (4)
α′t = f α + h, (5)
βt′ 2
= nα + gα + s. (6)
Then the original equation can be reduced with the transformation (1)–(6) to the constant
coefficient equation
∂u ∂2u
= ,
∂τ ∂z 2
which is discussed in Section 3.1.1 in detail.
System (2)–(6) can be solved successively. To this end, we start with equation (3), for
example, in order (3) → (2) → (5) → (6) → (4). As a result, we obtain
Z
ψ = C1 exp f dt , C1 6= 0,
Z
ϕ = nψ 2 dt + C2 ,
Z
h
α=ψ dt + C3 ψ,
ψ
Z

β= nα2 + gα + s dt + C4 ,
Z
χ = (2nα + g)ψ dt + C5 ,
where C1 , C2 , C3 , C4 , and C5 are arbitrary constants.

Remark 3.8. Likewise, one can simplify the nonhomogeneous equation with an additional term
Φ(x, t) on the right-hand side.
∂w ∂ 2w ∂w
5. = n(t) + xf (t) + g(t) + x2 h(t) + xs(t) + p(t) w.
∂t ∂x2 ∂x
2

The substitution w(x, t) = exp ϕ(t)x u(x, t) leads to an equation
∂u ∂2u ∂u
= n 2 + x(4nϕ + f ) + g
∂t ∂x ∂x
+ [x2 (h + 2f ϕ + 4nϕ2 − ϕ′t ) + x(s + 2gϕ) + p + 2nϕ u. (1)
We choose the function ϕ = ϕ(t) so that it is a (particular) solution of the Riccati ordinary
differential equation
ϕ′t = 4nϕ2 + 2f ϕ + h. (2)
Then the transformed equation (1) becomes an equation of the form 3.8.7.4:
∂u ∂2u ∂u
= n 2 + x(4nϕ + f ) + g + [x(s + 2gϕ) + p + 2nϕ u.
∂t ∂x ∂x
A number of specific solvable Riccati equations (2) can be found in Murphy (1960),
Kamke (1977), and Polyanin and Zaitsev (2003).
In the special case where
n = af, h = bf, with a, b = const, f = f (t),
the roots of the quadratic equation 4aϕ2 + 2ϕ + b = 0 (ϕ = const) are particular solutions
of equation (2).
∂w ∂2w ∂w
3.8.8 Equations of the Form = f (x, t) + g(x, t) + h(x, t)w
∂t ∂x2 ∂x
∂w ∂ 2w ∂w
1. = xf (t) + xg(t) + xh(t) + s(t) w.
∂t ∂x2 ∂x
Z
τ = ϕ(t), z = xψ(t), w(x, t) = u(z, τ ) exp xα(t) + s(t) dt , (1)
where the unknown functions ϕ(t), ψ(t), and α(t) are chosen so that the resulting equation
is as simple as possible. For the new dependent variable u(z, τ ) we have
∂u ∂2u z ∂u z
ϕ′t = zf ψ 2 + 2f ψα + gψ − ψt′ + f α2 + gα + h − α′t u.
∂τ ∂z ψ ∂z ψ
Let the unknown functions satisfy the system of ordinary differential equations
ϕ′t = f ψ, (2)
ψt′ = 2f αψ + gψ, (3)
α′t 2
= f α + gα + h. (4)
Then the original equation can be reduced with the transformation (1)–(4) to a constant
coefficient equation of the form 3.3.4.1:
∂u ∂2u
=z 2.
∂τ ∂z
Let us solve system (2)–(4) successively, starting with equation (4) in the order (4) →
(3) → (2).
The Riccati equation (4) can be solved separately. A lot of specific solvable Riccati
equations can be found in Murphy (1960), Kamke (1977), and Polyanin and Zaitsev (2003).
Suppose a solution α = α(t) of equation (4) is known. Then the solutions of equations
(2) and (3) can be found in the form
Z Z
ψ(t) = C1 exp (2f α + g) dt , ϕ(t) = f ψ dt + C2 ,
where C1 and C2 are arbitrary constants.

Remark 3.9. The transformation (1)–(4) can also be used to simplify the nonhomogeneous
equation with an additional term Φ(x, t) on the right-hand side.
∂w ∂ 2w ∂w
2. = x2 f (t) + xg(t) + h(t)w.
∂t ∂x2 ∂x
The substitution x = ±eξ leads to an equation of the form 3.8.7.3:
∂w ∂2w ∂w
= f (t) 2 + g(t) − f (t) + h(t)w.
∂t ∂ξ ∂ξ
∂w ∂ 2w ∂w
3. = x2 n(t) 2
+x f (t) ln x+g(t) + h(t) ln2 x+s(t) ln x+p(t) w.
∂t ∂x ∂x
The substitution z = ln x leads to an equation of the form 3.8.7.5:
∂w ∂2w ∂w 2
= n(t) 2 + zf (t) + g(t) − n(t) + z h(t) + zs(t) + p(t) w.
∂t ∂z ∂z
∂w ∂ 2w
4. = x4 f (t) + g(t)w.
∂t ∂x2
Z
w(x, t) = (Ax + B) exp g(t) dt ,
Z Z
A
w(x, t) = 2Ax f (t) dt + Bx + exp g(t) dt ,
x
Z Z
λ
w(x, t) = Ax exp λ2 f (t) dt + g(t) dt + .
x
Z Z
1
w(x, t) = x exp g(t) dt u(ξ, τ ), ξ= , τ= f (t) dt
x
leads to a constant coefficient equation, ∂τ u = ∂ξξ u, which is considered in Section 3.1.1.
∂w ∂ 2w
5. = (x − a1 )2 (x − a2 )2 f (t) + g(t)w, a1 6= a2 .
∂t ∂x2
The transformation
Z Z
x − a1
w(x, t) = (x − a2 ) exp g(t) dt u(ξ, τ ),
ξ = ln , τ = (a1 − a2 ) 2
f (t) dt
x − a2
leads to a constant coefficient equation
∂u ∂2u ∂u
= − ,
∂τ ∂ξ 2 ∂ξ

∂w ∂ 2w
6. = (ax2 + bx + c)2 f (t) + g(t)w.
∂t ∂x2
The transformation
Z Z Z
2 1/2 dx
w(x, t) = |ax +bx+c| exp g(t) dt u(ξ, τ ), ξ= 2
, τ= f (t) dt
ax +bx+c
leads to a constant coefficient equation
∂u ∂2u
= + ac − 14 b2 )u,
∂τ ∂ξ 2
∂w ∂ 2w ∂w
7. = xn f (t) + xg(t) + h(t)w.
∂t ∂x2 ∂x
The transformation
Z Z
w(x, t) = u(z, τ ) exp h(t) dt , z = xG(t), τ= f (t)G2−n (t) dt,
Z
where G(t) = exp g(t) dt , leads to an equation of the form 3.3.6.6:
∂u ∂2u
= zn 2 .
∂τ ∂z
∂w ∂ 2w
8. = eβx f (t) + g(t)w.
∂t ∂x2
The transformation
Z Z
w(x, t) = exp g(t) dt u(x, τ ), τ= f (t) dt
∂u ∂2u
= eβx 2 .
∂τ ∂x
∂w ∂ 2w
9. = f (x)g(t) + h(t)w.
∂t ∂x2
1◦ . Particular solutions (A, B, and x0 are arbitrary constants):
w(x, t) = (Ax + B)H(t),

w(x, t) = A M (t) + F (x) H(t),

w(x, t) = A M (t)x + Ψ(x) H(t),
Z x
2 x−ξ
w(x, t) = A M (t) + 2M (t)F (x) + 2 F (ξ) dξ H(t),
x0 f (ξ)
Z x
2 x−ξ
w(x, t) = A M (t)x + 2M (t)Ψ(x) + 2 Ψ(ξ) dξ H(t).
x0 f (ξ)
Here we use the shorthand notation

Z Z Z x Z x
x−ξ x−ξ
H(t)=exp h(t)dt , M (t)= g(t)dt, F (x)= dξ, Ψ(x)= ξ dξ.
x0 f (ξ) x0 f (ξ)
Z Z
w(x, t) = exp h(t) dt u(x, τ ), τ= g(t) dt
leads to a simpler equation

∂u ∂2u
= f (x) 2 .
∂τ ∂x
A wide class of exact analytical solutions to this equation is specified in 3.8.6.1.

∂w ∂ ∂w
3.8.9 Equations of the Form s(x) = p(x) −q(x)w+Φ(x, t)
∂t ∂x ∂x
Equations of this form are often encountered in heat and mass transfer theory and chemical
engineering sciences. Throughout this subsection, we assume that the functions s, p, p′x ,
and q are continuous and s > 0, p > 0, and x1 ≤ x ≤ x2 .
◮ General formulas for solving linear nonhomogeneous boundary value problems.

The solution of the equation in question under the initial condition
w = f (x) at t=0 (1)
and the arbitrary linear nonhomogeneous boundary conditions
a1 ∂x w + b1 w = g1 (t) at x = x1 ,
(2)
a2 ∂x w + b2 w = g2 (t) at x = x2 ,
can be represented as the sum
Z t Z x2 Z x2
w(x, t) = Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ + s(ξ)f (ξ)G(x, ξ, t) dξ
0 x1 x1
Z t Z t
+ p(x1 ) g1 (τ )Λ1 (x, t − τ ) dτ + p(x2 ) g2 (τ )Λ2 (x, t − τ ) dτ. (3)
0 0
Here, the modified Green’s function is given by

X∞ Z x2
yn (x)yn (ξ) 2
G(x, ξ, t) = exp(−λn t), kyn k = s(x)yn2 (x) dx, (4)
n=1
kyn k2 x1
where the λn and yn (x) are the eigenvalues and corresponding eigenfunctions of the fol-
lowing Sturm–Liouville problem for a second-order linear ordinary differential equation:
[p(x)yx′ ]′x + [λs(x) − q(x)]y = 0,
a1 yx′ + b1 y = 0 at x = x1 , (5)
a2 yx′ + b2 y = 0 at x = x2 .
The functions Λ1 (x, t) and Λ2 (x, t) that occur in the integrands of the last two terms in
solution (3) are expressed via the Green’s function (4). Appropriate formulas will be given
below when considering specific boundary value problems.
◮ General properties of the Sturm–Liouville problem (5).

1◦ . There are infinitely many eigenvalues. All eigenvalues are real and different and can
be ordered so that λ1 < λ2 < λ3 < · · · , with λn → ∞ as n → ∞ (therefore, there can exist
only finitely many negative eigenvalues). Each eigenvalue is of multiplicity 1.
2◦ . The eigenfunctions are determined up to a constant multiplier. Each eigenfunction
yn (x) has exactly n − 1 zeros in the open interval (x1 , x2 ).
3◦ . Eigenfunctions yn (x) and ym (x), n 6= m, are orthogonal with weight s(x) on the
interval x1 ≤ x ≤ x2 :
Z x2
s(x)yn (x)ym (x) dx = 0 for n 6= m.
x1
4◦ . An arbitrary function F (x) that has a continuous derivative and satisfies the bound-
ary conditions of the Sturm–Liouville problem can be expanded into an absolutely and
uniformly convergent series in eigenfunctions:
∞
X Z x2
1
F (x) = Fn yn (x), Fn = s(x)F (x)yn (x) dx,
n=1
kyn k2 x1
where the norm kyn k2 is defined in (4).

5◦ . If the conditions
q(x) ≥ 0, a1 b1 ≤ 0, a2 b2 ≥ 0 (6)
are satisfied, there are no negative eigenvalues. If q ≡ 0 and b1 = b2 = 0, then λ1 = 0 is
the least eigenvalue, to which there corresponds the eigenfunction ϕ1 = const. Otherwise,
all eigenvalues are positive, provided that conditions (6) are satisfied.
6◦ . The following asymptotic relation holds for large eigenvalues as n → ∞:
Z x2 s
π 2 n2 s(x)
λn = 2
+ O(1), ∆ = dx. (7)
∆ x1 p(x)
Special, boundary value condition-dependent properties of the Sturm–Liouville prob-

lem are presented below.
Remark 3.10. Equation (5) can be reduced to one with p(x) ≡ 1 and s(x) ≡ 1 by the change of
variables
Z s
s(x) 1/4
ζ= dx, u(ζ) = p(x)s(x) y(x).
p(x)
The boundary conditions transform into boundary conditions of the same type.
◮ First boundary value problem: the case of a1 = a2 = 0 and b1 = b2 = 1.

The solution of the first boundary value problem with the initial condition (1) and the
boundary conditions
w = g1 (t) at x = x1 ,
w = g2 (t) at x = x2
is given by formulas (3)–(4) with
∂ ∂

Λ1 (x, t) = G(x, ξ, t) , Λ2 (x, t) = − G(x, ξ, t) .
∂ξ ξ=x1 ∂ξ ξ=x2
Some special properties of the Sturm–Liouville problem are worth mentioning.

1◦ . For n → ∞, the asymptotic relation (7) can be used to estimate eigenvalues λn . The
corresponding eigenfunctions yn (x) satisfy the asymptotic relation
1/4 Z r Z x2 r
yn (x) 4 πn x s(x) 1 s(x)
= sin dx + O , ∆= dx.
kyn k ∆2 p(x)s(x) ∆ x1 p(x) n x1 p(x)
2◦ . For q ≥ 0, the following upper estimate (Rayleigh principle) holds for the least eigen-
value: Z x2

p(x)(zx′ )2 + q(x)z 2 dx
λ1 ≤ x1 Z x2 , (8)
2
s(x)z dx
x1
where z = z(x) is any twice differentiable function that satisfies the conditions z(x1 ) =
z(x2 ) = 0. The equality in (8) is attained for z = y1 (x), where y1 (x) is the eigenfunction
of the Sturm–Liouville problem corresponding to the eigenvalue λ1 . To obtain particular
estimates, one may set z = (x − x1 )(x2 − x) or z = sin[π(x − x1 )/(x2 − x1 )] in (8).
3◦ . Suppose
0 < pmin ≤ p(x) ≤ pmax , 0 < qmin ≤ q(x) ≤ qmax , 0 < smin ≤ s(x) ≤ smax .
Then the following double-ended estimate holds for the eigenvalues:
pmin π 2 n2 qmin pmax π 2 n2 qmax

2
+ ≤ λn ≤ 2
+ .
smax (x2 − x1 ) smax smin (x2 − x1 ) smin
4◦ . In engineering calculations, the approximate formula
Z Z r
x2 x2
π 2 n2 1 q(x) s(x)
λn = 2
+ dx, where ∆= dx, (9)
∆ x2 − x1 x1 s(x) x1 p(x)
may be used to determine eigenvalues. This formula is exact if p(x)s(x) = const and
q(x)/s(x) = const (in particular, for constant p = p0 , q = q0 , and s = s0 ) and provides
correct asymptotics (7) for any p(x), q(x), and s(x). Furthermore, for p(x) = const and
s(x) = const, relation (9) gives two correct first terms as n → ∞; the same holds true if
p(x)s(x) = const.
5◦ . Suppose p(x) = s(x) = 1 and the function q = q(x) has a continuous derivative. Then
the following asymptotic relations hold for eigenvalues λn and eigenfunctions yn (x) as
n → ∞:
p
πn 1 1
λn = + Q(x1 , x2 ) + O ,
x2 − x1 πn n2
πn(x − x1 ) 1
yn (x) = sin − (x1 − x)Q(x, x2 )
x2 − x1 πn

πn(x − x1 ) 1
+ (x2 − x)Q(x1 , x) cos +O ,
x2 − x1 n2
where
Z v
1
Q(u, v) = q(x) dx. (10)
2 u
◮ Second boundary value problem: the case of a1 = a2 = 1 and b1 = b2 = 0.
The solution of the second boundary value problem with the initial condition (1) and the
boundary conditions
∂x w = g1 (t) at x = x1 ,
∂x w = g2 (t) at x = x2
is given by formulas (3)–(4) with
Λ1 (x, t) = −G(x, x1 , t), Λ2 (x, t) = G(x, x2 , t).
Some special properties of the Sturm–Liouville problem are worth mentioning.
1◦ . For q > 0, the upper estimate (8) holds for the least eigenvalue, with z = z(x) being any
twice differentiable function that satisfies the conditions zx′ (x1 ) = zx′ (x2 ) = 0. The equality
in (8) is attained for z = y1 (x), where y1 (x) is the eigenfunction of the Sturm–Liouville
problem corresponding to the eigenvalue λ1 .
n → ∞:
p
π(n − 1) 1 1
λn = + Q(x1 , x2 ) + O ,
x2 − x1 π(n − 1) n2
π(n − 1)(x − x1 ) 1 h
yn (x) = cos + (x1 − x)Q(x, x2 )
x2 − x1 π(n − 1)
i
π(n − 1)(x − x1 ) 1
+ (x2 − x)Q(x1 , x) sin +O ,
x2 − x1 n2
where the function Q(u, v) is defined by (10).

◮ Third boundary value problem: the case of a1 = a2 = 1, b1 6= 0, and b2 6= 0.

The solution of the third boundary value problem with the initial condition (1) and boundary
conditions (2), with a1 = a2 = 1, is given by relations (3)–(4) in which
Λ1 (x, t) = −G(x, x1 , t), Λ2 (x, t) = G(x, x2 , t).

Suppose p(x) = s(x) = 1 and the function q = q(x) has a continuous derivative. Then
n → ∞:
p h i
π(n − 1) 1 1
λn = + Q(x1 , x2 ) − b1 + b2 + O ,
x2 − x1 π(n − 1) n2
π(n − 1)(x − x1 ) 1 n
yn (x) = cos + (x1 − x) Q(x, x2 ) + b2
x2 − x1 π(n − 1)

o π(n − 1)(x − x1 ) 1
+ (x2 − x) Q(x1 , x) − b1 sin +O ,
x2 − x1 n2
where Q(u, v) is defined by (10).
◮ Mixed boundary value problem: the case of a1 = b2 = 0 and a2 = b1 = 1.

The solution of the mixed boundary value problem with the initial condition (1) and the
boundary conditions
w = g1 (t) at x = x1 ,
∂x w = g2 (t) at x = x2
is given by relations (3)–(4) with
∂

Λ1 (x, t) = G(x, ξ, t) , Λ2 (x, t) = G(x, x2 , t).
∂ξ ξ=x1
Below are some special properties of the Sturm–Liouville problem.

1◦ . For q ≥ 0, the upper estimate (8) holds for the least eigenvalue, with z = z(x) being any
twice differentiable function that satisfies the conditions z(x1 ) = 0 and zx′ (x2 ) = 0. The
equality in (8) is attained for z = y1 (x), where y1 (x) is the eigenfunction corresponding to
the eigenvalue λ1 .
n → ∞:
p
π(2n − 1) 1 1
λn = + Q(x1 , x2 ) + O ,
2(x2 − x1 ) π(2n − 1) n2
π(2n − 1)(x − x1 ) 2
yn (x) = sin − (x1 − x)Q(x, x2 )
2(x2 − x1 ) π(2n − 1)

π(2n − 1)(x − x1 ) 1
+ (x2 − x)Q(x1 , x) cos +O ,
2(x2 − x1 ) n2
where Q(u, v) is defined by (10).
3.9. Equations of Special Form 393
◮ Mixed boundary value problem: the case of a1 = b2 = 1 and a2 = b1 = 0.

The solution of the mixed boundary value problem with the initial condition (1) and the
boundary conditions
∂x w = g1 (t) at x = x1 ,
w = g2 (t) at x = x2
is given by formulas (3)–(4) in which
∂

Λ1 (x, t) = −G(x, x1 , t), Λ2 (x, t) = − G(x, ξ, t) .
∂ξ ξ=x2
Below are some special properties of the Sturm–Liouville problem.

1◦ . For q ≥ 0, the upper estimate (8) holds for the least eigenvalue, with z = z(x) being any
twice differentiable function that satisfies the conditions zx′ (x1 ) = 0 and z(x2 ) = 0. The
equality in (8) is attained for z = y1 (x), where y1 (x) is the eigenfunction corresponding to
the eigenvalue λ1 .
n → ∞:
p
π(2n − 1) 2 1
λn = + Q(x1 , x2 ) + O ,
2(x2 − x1 ) π(2n − 1) n2
π(2n − 1)(x − x1 ) 2
yn (x) = cos + (x1 − x)Q(x, x2 )
2(x2 − x1 ) π(2n − 1)

π(2n − 1)(x − x1 ) 1
+ (x2 − x)Q(x1 , x) sin +O ,
2(x2 − x1 ) n2
where the function Q(u, v) is defined by relation (10).
⊙ References for Section 3.8.9: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), E. Kamke
(1977), V. A. Marchenko (1986), V. S. Vladimirov (1988), B. M. Levitan and I. S. Sargsyan (1988), L. D.
Akulenko and S. V. Nesterov (1997), A. D. Polyanin (2001).
3.9 Equations of Special Form

3.9.1 Equations of the Diffusion (Thermal) Boundary Layer
∂w ∂w ∂2w
1. f (x) + g(x)y = .
∂x ∂y ∂y 2
This equation is encountered in diffusion boundary layer problems (mass exchange of drops
and bubbles with flow).
The transformation (A and B are any numbers)
Z 2 Z
h (x) g(x)
t= dx + A, z = yh(x), where h(x) = B exp − dx ,
f (x) f (x)
leads to a constant coefficient equation, ∂t w = ∂zz w, which is considered in Section 3.1.1.
⊙ Literature: V. G. Levich (1962), A. D. Polyanin and V. V. Dilman (1994), A. D. Polyanin, A. M. Kutepov,
A. V. Vyazmin, and D. A. Kazenin (2002).
∂w ∂w ∂2w
2. f (x) + g(x)y = − h(x)w.
∂x ∂y ∂y 2
This equation is encountered in diffusion boundary layer problems with a first-order volume
chemical reaction (usually h ≡ const).
Z Z 2
h(x) ϕ (x)
w(x, y) = u(t, z) exp − dx , t = dx + A, z = yϕ(x),
f (x) f (x)
Z
g(x)
where ϕ(x) = B exp − dx , leads to a constant coefficient equation, ∂t u = ∂zz u,
f (x)
⊙ Literature: Yu. P. Gupalo, A. D. Polyanin, and Yu. S. Ryazantsev (1985).
∂w ∂w ∂2w
3. f (x)y n−1 + g(x)y n = .
∂x ∂y ∂y 2
This equation is encountered in diffusion boundary layer problems (mass exchange of solid
particles, drops, and bubbles with flow).
Z n+1 Z
h (x) g(x)
f (x) f (x)
∂w ∂2w
= z 1−n 2 .
∂t ∂z
⊙ Literature: V. G. Levich (1962), A. D. Polyanin and V. V. Dilman (1994), A. D. Polyanin, A. V. Vyazmin,
A. I. Zhurov, and D. A. Kazenin (1998), A. D. Polyanin, A. M. Kutepov, A. V. Vyazmin, and D. A. Kazenin
(2002).

y ∂w 1 y ∂w ∂ 2w
4. f √ + √ g √ = .
x ∂x x x ∂y ∂y 2
This is a generalization of the problem of thermal boundary layer on a flat plate.

√
1◦ . By passing from x, y to the new variables t = ln x, ξ = y/ x, we arrive at the separable
equation
∂w ∂w ∂2w
f (ξ) + g(ξ) − 12 ξf (ξ) = .
∂t ∂ξ ∂ξ 2
There are particular solutions of the form
w(t, ξ) = Aeβt φ(ξ),
where the function φ(ξ) satisfies the ordinary differential equation

φ′′ξξ = g(ξ) − 12 ξf (ξ) φ′ξ + βf (ξ)φ.
2◦ . The solution of the original equation with the boundary conditions
x = 0, w = w0 ; y = 0, w = w1 ; y → ∞, w → w0
(w0 and w1 are some constants) is given by

R∞ Z
ξ exp −Ψ(ξ) dξ
ξ
w − w0 1
= R∞ , Ψ(ξ) = 2 ξf (ξ) − g(ξ) dξ,
w1 − w0 0 exp −Ψ(ξ) dξ 0
√
where ξ = y/ x. It is assumed that the inequality ξf (ξ) > 2g(ξ) holds for ξ > 0.
3◦ . The equation of a thermal boundary layer on a flat plate corresponds to

f (ξ) = Pr Fξ′ (ξ), g(ξ) = 12 Pr ξFξ′ (ξ) − F (ξ) ,
where F (ξ) is the Blasius solution in the problem of translational flow past a flat plate
and Pr is the Prandtl number (x is the coordinate measured along the plate and y is the
transverse coordinate to the plate surface). In this case the formulas in Item 2◦ transform
into Polhausen’s solution. See Schlichting (1981) for details.
h i
∂w b ∂w ∂2w
5. f (x) + g(x)y − = .
∂x y 2 ∂y ∂y
For b = 1, equations of this sort govern the concentration distribution in the internal region
of the diffusion wake behind a moving particle or drop.
Z 2 Z
h (x) g(x)
f (x) f (x)
leads to an equation of the form 3.2.5:
∂w ∂2w b ∂w
= 2
+ .
∂t ∂z z ∂z
⊙ Literature: Yu. P. Gupalo, A. D. Polyanin, and Yu. S. Ryazantsev (1985), A. D. Polyanin, A. V. Vyazmin,
A. I. Zhurov, and D. A. Kazenin (1998).
h i
∂w b ∂w ∂2w
6. f (x) + g(x)y − = + h(x)w.
∂x y 2 ∂y ∂y
Z

h(x)
The substitution w(x, y) = u(x, y) exp dx leads to an equation of the form
f (x)
3.9.1.5:
∂u b ∂u ∂2u
f (x) + g(x)y − = .
∂x y ∂y ∂y 2
h i
∂w b ∂w ∂2w
7. f (x)y n−1 + g(x)y n − = .
∂x y2 ∂y ∂y
Z n+1
h (x) 2 n+1
t= dx + A, ξ = yh(x) 2 ,
f (x) n+1
Z
g(x)
where h(x) = B exp − dx , leads to the equation
f (x)
∂w ∂2w 1 − 2β ∂w 1−b
= + , β= ,
∂t ∂ξ 2 ξ ∂ξ n+1
which is considered in Section 3.2.5 (see also equations in Sections 3.2.1 and 3.2.3).
3.9.2 One-Dimensional Schrödinger Equation

∂w ~2 ∂ 2 w
i~ =− + U (x)w
∂t 2m ∂x2
◮ Eigenvalue problem. Cauchy problem.
Schrödinger’s equation is the basic equation of quantum mechanics; w is the wave function,
i2 = −1, ~ is Planck’s constant, m is the mass of the particle, and U (x) is the potential
energy of the particle in the force field.
1◦ . In discrete spectrum problems, the particular solutions are sought in the form

iEn
w(x, t) = exp − t ψn (x),
~
where the eigenfunctions ψn and the respective energies En have to be determined by
solving the eigenvalue problem
d2 ψn 2m
2
+ 2 En − U (x) ψn = 0,
dx ~ Z ∞ (1)
ψn → 0 at x → ±∞, |ψn |2 dx = 1.
−∞
The last relation is the normalizing condition for ψn .

2◦ . In the cases where the eigenfunctions ψn (x) form an orthonormal basis in L2 (R), the
solution of the Cauchy problem for Schrödinger’s equation with the initial condition
w = f (x) at t=0 (2)
is given by
Z ∞ ∞
X
iEn
w(x, t) = G(x, ξ, t)f (ξ) dξ, G(x, ξ, t) = ψn (x)ψn (ξ) exp − t .
−∞ n=0
~
Various potentials U (x) are considered below and particular solutions of the bound-
ary value problem (1) or the Cauchy problem for Schrödinger’s equation are presented. In
some cases, nonnormalized eigenfunctions Ψn (x) are given instead of normalized eigen-
functions ψn (x); the former differ from the latter by a constant multiplier.
◮ Free particle: U (x) = 0.

The solution of the Cauchy problem with the initial condition (2) is given by
Z ∞
1 (x − ξ)2
w(x, t) = √ exp − f (ξ) dξ,
2 iπτ −∞ 4iτ
( p
~t √ eπi/4 |a| if a > 0,
τ= , ia = −πi/4
p
2m e |a| if a < 0.
⊙ Literature: W. Miller, Jr. (1977).
◮ Linear potential (motion in a uniform external field): U (x) = ax.

Solution of the Cauchy problem with the initial condition (2):
Z
1 1 2 3
∞ (x + bτ 2 − ξ)2
w(x, t) = √ exp −ibτ x − 3 ib τ exp − f (ξ) dξ,
2 iπτ −∞ 4iτ
~t 2am
τ= , b= 2 .
2m ~
1
◮ Linear harmonic oscillator: U (x) = 2
mω 2 x2 .
Eigenvalues:
1
En = ~ω n + 2 , n = 0, 1, . . .
Normalized eigenfunctions:
r
1 x ~
ψn (x) = 1/4 √ n exp − 12 ξ 2 Hn (ξ), ξ= , x0 = ,
π 2 n! x0 x0 mω
where Hn (ξ) are the Hermite polynomials.
The functions ψn (x) form an orthonormal basis in L2 (R).
⊙ Literature: S. G. Krein (1972), W. Miller, Jr. (1977), A. N. Tikhonov and A. A. Samarskii (1990).
◮ Isotropic free particle: U (x) = a/x2 .

Here, the variable x ≥ 0 plays the role of the radial coordinate, and a > 0. The equation
with U (x) = a/x2 results from Schrödinger’s equation for a free particle with n space
coordinates if one passes to spherical (cylindrical) coordinates and separates the angular
variables.
The solution of Schrödinger’s equation satisfying the initial condition (2) has the form
Z ∞ 2
exp − 12 iπ(µ + 1) sign t √ x + y2 xy
w(x, t) = xy exp i Jµ f (y) dy,
2|τ | 0 4τ 2|τ |
r
~t 2am 1
τ= , µ= 2
+ ≥ 1,
2m ~ 4
where Jµ (ξ) is the Bessel function.
1
◮ Isotropic harmonic oscillator: U (x) = 2
mω 2 x2 + ax−2 .
Here, the variable x ≥ 0 plays the role of the radial coordinate, and a > 0. The equation
with this U (x) results from Schrödinger’s equation for a harmonic oscillator with n space
coordinates if one passes to spherical (cylindrical) coordinates and separates the angular
variables.
Eigenvalues:
r
2am 1
En = −~ω(2n + µ + 1), µ= + ≥ 1, n = 0, 1, . . .
~2 4
Normalized eigenfunctions:
s r
2n! 2µ+1 x ~
ψn (x) = ξ 2 exp − 12 ξ 2 Lµn ξ 2 , ξ= , x0 = ,
Γ(n + 1 + µ)x0 x0 mω
where Lµn (z) is the nth generalized Laguerre polynomial with parameter µ. The norm
|ψn (x)|2 refers to the semiaxis x ≥ 0.
The functions ψn (x) form an orthonormal basis in L2 (R+ ).
◮ Morse potential: U (x) = U0 (e−2x/a − 2e−x/a ).
Eigenvalues:
2 √
1 1 a 2mU0
En = −U0 1 − (n + 2 ) , β= , 0 ≤ n < β − 2.
β ~
Eigenfunctions:
√
a −2mEn
ψn (x) = ξ s e−ξ/2 Φ(−n, 2s + 1, ξ), ξ = 2βe−x/a , s= ,
~
where Φ(a, b, ξ) is the degenerate hypergeometric function.

In this case the number of eigenvalues (energy levels) En and eigenfunctions ψn is
finite: n = 0, 1, . . . , nmax .
⊙ Literature: S. G. Krein (1964), L. D. Landau and E. M. Lifshitz (1974).
◮ Potential with a hyperbolic function: U (x) = −U0 cosh−2 (x/a).
Eigenvalues:
r !
~2 1 8mU0 a2
En = − (s − n)2 , s= −1 + 1+ , 0 ≤ n < s.
2ma2 2 ~2
Eigenfunctions:


 x −2s β −s β +s 1 2 x

 cosh a F ,− , , − sinh for even n,
2 2 2 a
ψn(x) = −2s

 x x 1+β −s 1−β −s 3 2 x

 sinh cosh F , , , − sinh for odd n,
a a 2 2 2 a
ap
where F (a, b, c, ξ) is the hypergeometric function and β = −2mEn .
~
The number of eigenvalues (energy levels) En and eigenfunctions ψn is finite in this
case: n = 0, 1, . . . , nmax .
⊙ Literature: S. G. Krein (1964).
◮ Potential with a trigonometric function: U (x) = U0 cot2 (πx/a).

Eigenvalues:
r !
π 2 ~2 2 1 8mU0 a2
En = (n + 2ns − s), s= −1 + 1+ , n = 0, 1, . . .
2ma2 2 π 2 ~2
Eigenfunctions:


 πx πx −2s 1−n−s n+1 3 2 πx

cos sin F , , , cos for even n,
a a 2 2 2 a
ψn (x) = −2s

 πx n+s n 1 πx

 sin F − , , , cos2 for odd n,
a 2 2 2 a
where F (a, b, c, ξ) is the hypergeometric

√ function.
In particular, if a = π~/ 2m, U0 = 2, and n = k − 1, we have
kπx kπx πx
Ek = k2 − 2, ψk (x) = k cos − sin cot , k = 1, 2, . . .
a a a
⊙ Literature: S. G. Krein (1964), L. D. Landau and E. M. Lifshitz (1974).
Chapter 4
Second-Order
Parabolic Equations
with Two Space Variables
∂w
4.1 Heat Equation = a∆2 w
∂t
4.1.1 Boundary Value Problems in Cartesian Coordinates

In rectangular Cartesian coordinates, the two-dimensional sourceless heat equation has the
form 2
∂w ∂ w ∂2w
=a + .
∂t ∂x2 ∂y 2
It governs two-dimensional unsteady heat transfer processes in quiescent media or solid
bodies with constant thermal diffusivity a. A similar equation is used to study analogous
two-dimensional unsteady mass transfer phenomena with constant diffusivity; in this case
the equation is called a diffusion equation.
◮ Particular solutions:
w(x, y) = Axy + C1 x + C2 y + C3 ,
w(x, y, t) = Ax2 + By 2 + 2a(A + B)t,
w(x, y, t) = A(x2 + 2at)(y 2 + 2at) + B,

w(x, y, t) = A exp k1 x + k2 y + (k12 + k22 )at + B,

w(x, y, t) = A cos(k1 x + C1 ) cos(k2 y + C2 ) exp −(k12 + k22 )at ,

w(x, y, t) = A cos(k1 x + C1 ) sinh(k2 y + C2 ) exp −(k12 − k22 )at ,

w(x, y, t) = A cos(k1 x + C1 ) cosh(k2 y + C2 ) exp −(k12 − k22 )at ,
w(x, y, t) = A exp(−µx − λy) cos(µx − 2aµ2 t + C1 ) cos(λy − 2aλ2 t + C2 ),

A (x − x0 )2 + (y − y0 )2
w(x, y, t) = exp − ,
t − t0 4a(t − t0 )
401
402 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES

x − x0 y − y0
w(x, y, t) = A erf √ erf √ + B,
2 at 2 at
where A, B, C1 , C2 , C3 , k1 , k2 , x0 , y0 , and t0 are arbitrary constants.

1 x2 + y 2
E (x, y, t) = exp − .
4πat 4at
◮ Formulas to construct particular solutions. Remarks on the Green’s functions.

1◦ . Apart from usual separable solutions w(x, y, t) = f1 (x)f2 (y)f3 (t), the equation in
question has more sophisticated solutions in the product form
w(x, y, t) = u(x, t)v(y, t),
where u = u(x, t) and v = v(y, t) are solutions of the one-dimensional heat equations
∂u ∂2u ∂v ∂2v
=a 2, =a 2,
∂t ∂x ∂t ∂y
2◦ . Suppose w = w(x, y, t) is a solution of the heat equation. Then the functions
w1 = Aw(±λx + C1 , ±λy + C2 , λ2 t + C3 ),
w2 = Aw(x cos β − y sin β + C1 , x sin β + y cos β + C2 , t + C3 ),

w3 = A exp λ1 x + λ2 y + a(λ21 + λ22 )t w(x + 2aλ1 t + C1 , y + 2aλ2 t + C2 , t + C3 ),

A β(x2 + y 2 ) x y γ + λt
w4 = exp − w , , , λδ − βγ = 1,
δ + βt 4a(δ + βt) δ + βt δ + βt δ + βt
where A, C1 , C2 , C3 , β, δ, λ, λ1 and λ2 are arbitrary constants, are also solutions of this
equation. The signs at λ’s in the formula for w1 are taken arbitrarily, independently of each
other.
3◦ . In all two-dimensional boundary value problems discussed in Section 4.1.1, the Green’s
function can be represented in the product form
G(x, y, ξ, η, t) = G1 (x, ξ, t) G2 (y, η, t),
where G1 (x, ξ, t) and G2 (y, η, t) are the Green’s functions of the corresponding one-di-
mensional boundary value problems (these functions are specified in Sections 3.1.1 and
3.1.2).
Example 4.1. The Green’s function of the first boundary value problem for a semiinfinite strip
(0 ≤ x ≤ l, 0 ≤ y < ∞), considered in Section 4.1.1, is the product of two one-dimensional Green’s
functions. The first Green’s function is that of the first boundary value problem on a closed interval
(0 ≤ x ≤ l) presented in Section 3.1.2. The second Green’s function is that of the first boundary
value problem on a semiinfinite interval (0 ≤ y < ∞) presented in Section 3.1.2, where x and ξ
must be renamed y and η, respectively.
∂w
4.1. Heat Equation ∂t
= a∆2 w 403
◮ Transformations that allow separation of variables.

Table 4.1 lists possible transformations that allow reduction of the two-dimensional heat
equation to a separable equation. All transformations of the independent variables have
the form (x, y, t) 7−→ (ξ, η, t). The transformations that can be obtained by interchange of
independent variables, x ⇋ y, are omitted.
The anharmonic oscillator functions are solutions of the second-order ordinary differ-
ential equation Fzz ′′ + (az 4 + bz 2 + c)F = 0. The Ince polynomials are the 2π-periodic
solutions of the Whittaker–Hill equation Fzz ′′ + k sin 2z F ′ + (a − bk cos 2z)F = 0; see

z
Arscott (1964, 1967).
◮ Domain: −∞ < x < ∞, −∞ < y < ∞. Cauchy problem.

w = f (x, y) at t = 0.
Solution:
Z ∞Z ∞
1 (x − ξ)2 + (y − η)2
w(x, y, t) = f (ξ, η) exp − dξ dη.
4πat −∞ −∞ 4at
Example 4.2. The initial temperature is piecewise-constant and equal to w1 in the domain
|x| < x0 , |y| < y0 and w2 in the domain |x| > x0 , |y| > y0 , specifically,
(
w1 for |x| < x0 , |y| < y0 ,
f (x, y) =
w2 for |x| > x0 , |y| > y0 .
Solution:

1 x0 − x x0 + x y0 − y y0 + y
w = (w1 − w2 ) erf √ + erf √ erf √ + erf √ + w2 .
4 2 at 2 at 2 at 2 at
If the initial temperature distribution f (x, y) is an infinitely differentiable function in

both arguments, then the solution can be represented in the series form
X∞
(at)n n ∂2 ∂2
w(x, y, t) = f (x, y) + L f (x, y) , L≡ + .
n=1
n! ∂x2 ∂y 2
Such a representation is useful for small t.

◮ Domain: 0 ≤ x < ∞, −∞ < y < ∞. First boundary value problem.

A half-plane is considered. The following conditions are prescribed:
w = f (x, y) at t = 0 (initial condition),

w = g(y, t) at x = 0 (boundary condition).
TABLE 4.1
Transformations (x, y, t) 7−→ (ξ, η, t) that allow solutions with R-separated
variables, w = exp[R(ξ, η, t)] f (ξ) g(η) h(t), for the two-dimensional heat
equation ∂t w = ∂xx w + ∂yy w. Everywhere, the function h(t) is exponential
No Transformations Factor exp R Function f (ξ) Function g(η)
x = ξ, Exponential Exponential
1 R=0
y=η function function
x =pξ, Exponential
2 y = η |t| R=0 Hermite function
function
p
x = ξ p|t|,
3 R=0 Hermite function Hermite function
y = η |t|
x = 12 (ξ 2 − η 2 ), Parabolic cylinder Parabolic cylinder

4 R=0
y = ξη function function
x = ξ cos η, Exponential
5 R=0 Bessel function
y = ξ sin η function
x = cosh ξ cos η, Modified Mathieu Mathieu
6 R=0
y = sinh ξ sin η function function
p
x = p|t| ξ cos η, Exponential
7 R=0 Laguerre function
y = |t| ξ sin η function
p
x = p|t| cosh ξ cos η,
8 R=0 Ince polynomial Ince polynomial
y = |t| sinh ξ sin η
x = ξ, Exponential
9 R = −aηt Airy function
y = η + at2 function
x = ξ, Exponential
10 R = − 14 η 2 t + 12 bη/t Airy function
y = ηt + b/t function
x√= ξ, Exponential Parabolic cylinder
11 R = − 14 η 2 t
y = η 1 + t2 function function
xp
= ξ, R = − 14 εη 2 t, Exponential
12 y = η |1 − t2 | Hermite function
ε = sign(1 − t2 ) function
x = ξt, Exponential Exponential
13 R = − 14 (ξ 2 + η 2 )t
y = ηt function function
x = ξ + at2 ,
14 R = −(aξ + bη)t Airy function Airy function
y = η + bt2
x = ξt + a/t, R = − 14 (ξ 2 + η 2 )t
15 Airy function Airy function
y = ηt + b/t + 12 (aξ + bη)/t
∂w
= a∆2 w 405
TABLE 4.1
(continued)
Function Function
No Transformations Factor exp R
f (ξ) g(η)
Parabolic Parabolic
x = 12 (ξ 2 − η 2 )t, 1
16 R = − 16 (ξ 2 + η 2 )2 t cylinder cylinder
y = ξηt
function function
√ Parabolic Parabolic
x = ξ √1 + t2 ,
17 R = − 14 (ξ 2 + η 2 )t cylinder cylinder
y = η 1 + t2
function function
p
x = ξ p|1 − t2 |, R = − 14 ε(ξ 2 + η 2 )t, Hermite Hermite
18
y = η |1 − t2 | ε = sign(1 − t2 ) function function
Anharmonic Anharmonic
x = 12 (ξ 2 − η 2 ) + at2 ,
19 R = − 12 a(ξ 2 − η 2 )t oscillator oscillator
y = ξη
function function
1 Anharmonic Anharmonic
x = 12 (ξ 2 − η 2 )t + a/t, R = − 16 (ξ 2 + η 2 )2 t
20 1 oscillator oscillator
y = ξηt + 4 a(ξ 2 − η 2 )/t
function function
x = ξt cos η, Bessel Exponential
21 R = − 14 ξ 2 t
y = ξt sin η function function
Modified
x = t cosh ξ cos η, Mathieu
22 R = − 14 (sinh2 ξ + cos2 η)t Mathieu
y = t sinh ξ sin η function
function
√
x = √1 + t2 ξ cos η, Whittaker Exponential
23 R = − 14 ξ 2 t
y = 1 + t2 ξ sin η function function
p
x = p|1 − t2 | ξ cos η, R = − 14 εξ 2 t, Laguerre Exponential
24
y = |1 − t2 | ξ sin η ε = sign(1 − t2 ) function function
√
x = √1 + t2 cosh ξ cos η, Ince Ince
25 2
R = − 14 (sinh2 ξ + cos2 η)t
y = 1 + t sinh ξ sin η polynomial polynomial
p
x = p|1 − t2 | cosh ξ cos η, R = − 14 ε(sinh2 ξ + cos2 η)t, Ince Ince
26 2 2
y = |1 − t | sinh ξ sin η ε = sign(1 − t ) polynomial polynomial
Solution:
Z ∞Z ∞
w(x, y, t) = f (ξ, η) G(x, y, ξ, η, t) dη dξ
0 −∞
Z tZ ∞
∂
+a g(η, τ ) G(x, y, ξ, η, t − τ ) dη dτ,
0 −∞ ∂ξ ξ=0
where

1 (x − ξ)2 + (y − η)2 (x + ξ)2 + (y − η)2
G(x, y, ξ, η, t) = exp − − exp − .
4πat 4at 4at
◮ Domain: 0 ≤ x < ∞, −∞ < y < ∞. Second boundary value problem.


∂x w = g(y, t) at x = 0 (boundary condition).
Solution:
Z ∞Z ∞
0 −∞
Z tZ ∞
−a g(η, τ ) G(x, y, 0, η, t − τ ) dη dτ,
0 −∞
where

1 (x − ξ)2 + (y − η)2 (x + ξ)2 + (y − η)2
G(x, y, ξ, η, t) = exp − + exp − .
4πat 4at 4at
◮ Domain: 0 ≤ x < ∞, −∞ < y < ∞. Third boundary value problem.


∂x w − kw = g(y, t) at x = 0 (boundary condition).
The solution w(x, y, t) is determined by the formula in the previous paragraph (for the

1 (y − η)2 (x − ξ)2 (x + ξ)2
G(x, y, ξ, η, t) = exp − exp − + exp −
4πat 4at 4at 4at
Z ∞ 2
(x + ξ + s)
− 2k exp − − ks ds .
0 4at
◮ Domain: 0 ≤ x < ∞, 0 ≤ y < ∞. First boundary value problem.

A quadrant of the plane is considered. The following conditions are prescribed:

w = g1 (y, t) at x = 0 (boundary condition),
w = g2 (x, t) at y = 0 (boundary condition).
∂w
= a∆2 w 407
Solution:
Z ∞Z ∞
w(x, y, t) = f (ξ, η) G(x, y, ξ, η, t) dξ dη
0 0
Z tZ
∞
∂
+a g1 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=0
Z tZ ∞
∂
+a g2 (ξ, τ ) G(x, y, ξ, η, t − τ ) dξ dτ,
0 0 ∂η η=0
where

1 (x − ξ)2 (x + ξ)2
G(x, y, ξ, η, t) = exp − − exp −
4πat 4at 4at
2
2

(y − η) (y + η)
× exp − − exp − .
4at 4at
Example 4.3. The initial temperature is uniform, f (x, y) = w0 . The boundary is maintained at
zero temperature, g1 (y, t) = g2 (x, t) = 0.
Solution:
x y
w = w0 erf √ erf √ .
2 at 2 at
◮ Domain: 0 ≤ x < ∞, 0 ≤ y < ∞. Second boundary value problem.


∂x w = g1 (y, t) at x = 0 (boundary condition),
∂y w = g2 (x, t) at y = 0 (boundary condition).
Solution:
Z ∞Z ∞
0 0
Z tZ ∞
−a g1 (η, τ ) G(x, y, 0, η, t − τ ) dη dτ
0 0
Z tZ ∞
−a g2 (ξ, τ ) G(x, y, ξ, 0, t − τ ) dξ dτ,
0 0
where

1 (x − ξ)2 (x + ξ)2
G(x, y, ξ, η, t) = exp − + exp −
4πat 4at 4at
2
2

(y − η) (y + η)
× exp − + exp − .
4at 4at
◮ Domain: 0 ≤ x < ∞, 0 ≤ y < ∞. Third boundary value problem.

w = f (x, y) at t=0 (initial condition),

∂x w − k1 w = g1 (y, t) at x = 0 (boundary condition),
∂y w − k2 w = g2 (x, t) at y = 0 (boundary condition).

1 (x − ξ)2 (x + ξ)2
G(x, y, ξ, η, t) = exp − + exp −
4πat 4at 4at

√ 2 x+ξ √
− 2k1 πat exp ak1 t + k1 (x + ξ) erfc √ + k1 at
2 at

(y − η)2 (y + η)2
× exp − + exp −
4at 4at

√ 2 y+η √
− 2k2 πat exp ak2 t + k2 (y + η) erfc √ + k2 at .
2 at
Example 4.4. The initial temperature is constant, f (x, y) = w0 . The temperature of the envi-
ronment is zero, g1 (y, t) = g2 (x, t) = 0.
Solution:

x 2 x √
w = w0 erf √ + exp(k1 x + ak1 t) erfc √ + k1 at
2 at 2 at

y 2 y √
× erf √ + exp(k2 y + ak2 t) erfc √ + k2 at .
2 at 2 at
◮ Domain: 0 ≤ x < ∞, 0 ≤ y < ∞. Mixed boundary value problems.

1◦ . A quadrant of the plane is considered. The following conditions are prescribed:

Solution:
Z ∞Z ∞
0 0
Z tZ ∞
∂
0 0 ∂ξ ξ=0
Z tZ ∞
−a g2 (ξ, τ ) G(x, y, ξ, 0, t − τ ) dξ dτ,
0 0
∂w
= a∆2 w 409
where

1 (x − ξ)2 (x + ξ)2
G(x, y, ξ, η, t) = exp − − exp −
4πat 4at 4at
2
2

(y − η) (y + η)
× exp − + exp − .
4at 4at
2◦ . A quadrant of the plane is considered. The following conditions are prescribed:

∂x w − kw = g1 (y, t) at x = 0 (boundary condition),
Solution:
Z ∞Z ∞
0 0
Z tZ ∞
−a g1 (η, τ ) G(x, y, 0, η, t − τ ) dη dτ
0 0
Z tZ ∞
∂
0 0 ∂η η=0
where

1 (y − η)2 (y + η)2
G(x, y, ξ, η, t) = exp − exp −
4πat 4at 4at
2

(x − ξ) (x + ξ)2
4at 4at

√ 2 x+ξ √
− 2k πat exp ak t + k(x + ξ) erfc √ + k at .
2 at
Example 4.5. The initial temperature is uniform, f (x, y) = w0 . Heat exchange with the envi-
ronment of zero temperature occurs at one side and the other side is maintained at zero temperature:
g1 (y, t) = g2 (x, t) = 0.
Solution:

x 2 x √ y
w = w0 erf √ + exp(kx + ak t) erfc √ + k at erf √ .
2 at 2 at 2 at
◮ Domain: 0 ≤ x ≤ l, 0 ≤ y < ∞. First boundary value problem.

A semiinfinite strip is considered. The following conditions are prescribed:

w = g2 (y, t) at x = l (boundary condition),
Solution:
Z ∞Z l
0 0
Z tZ ∞
∂
0 0 ∂ξ ξ=0
Z tZ ∞
∂
−a g2 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=l
Z tZ l
∂
0 0 ∂η η=0
where
G(x, y, ξ, η, t) = G1 (x, ξ, t) G2 (y, η, t),
∞
G1 (x, ξ, t) = sin sin exp − ,
l n=1 l l l2

1 (y − η)2 (y + η)2
G2 (y, η, t) = √ exp − − exp − .
2 πat 4at 4at
zero temperature, g1 (y, t) = g2 (y, t) = g3 (x, t) = 0.
Solution:
X ∞
4w0 y 1 (2n + 1)πx π 2 (2n + 1)2 at
w= erf √ sin exp − .
π 2 at n=0 2n + 1 l l2
◮ Domain: 0 ≤ x ≤ l, 0 ≤ y < ∞. Second boundary value problem.

∂x w = g2 (y, t) at x = l (boundary condition),
Solution:
Z ∞Z l
0 0
Z tZ ∞
−a g1 (η, τ ) G(x, y, 0, η, t − τ ) dη dτ
0 0
Z tZ ∞
+a g2 (η, τ ) G(x, y, l, η, t − τ ) dη dτ
0 0
Z tZ l
−a g3 (ξ, τ ) G(x, y, ξ, 0, t − τ ) dξ dτ,
0 0
∂w
= a∆2 w 411
where
G(x, y, ξ, η, t) = G1 (x, ξ, t) G2 (y, η, t),

∞
G1 (x, ξ, t) = + cos cos exp − ,
l l n=1 l l l2

1 (y − η)2 (y + η)2
G2 (y, η, t) = √ exp − + exp − .
2 πat 4at 4at
◮ Domain: 0 ≤ x ≤ l, 0 ≤ y < ∞. Third boundary value problem.


∂x w + k2 w = g2 (y, t) at x = l (boundary condition),
G(x, y, ξ, η, t) = G1 (x, ξ, t) G2 (y, η, t),

X∞
1
G1 (x, ξ, t) = yn (x)yn (ξ) exp(−aµ2n t),
kyn k2
n=1

1 (y − η)2 (y + η)2
G2 (y, η, t) = √ exp − + exp −
2 πat 4at 4at
Z ∞ 2

(y + η + ζ)
− 2k3 exp − − k3 ζ dζ .
0 4at
Here

k1 2 k2 µ2n + k12 k1 l k12
yn (x) = cos(µn x) + sin(µn x), kyn k = + 2 + 1+ 2 ,
tan(µl) k1 + k2
and the µn are positive roots of the transcendental equation = 2 .
µ µ − k1 k2
◮ Domain: 0 ≤ x ≤ l, 0 ≤ y < ∞. Mixed boundary value problems.

1◦ . A semiinfinite strip is considered. The following conditions are prescribed:

Solution:
Z ∞Z l
0 0
Z tZ
∞
∂
0 0 ∂ξ ξ=0
Z tZ ∞
∂
0 0 ∂ξ ξ=l
Z tZ l
−a g3 (ξ, τ ) G(x, y, ξ, 0, t − τ ) dξ dτ,
0 0
where
G(x, y, ξ, η, t) = G1 (x, ξ, t) G2 (y, η, t),

∞
l n=1 l l l2

1 (y − η)2 (y + η)2
G2 (y, η, t) = √ exp − + exp − .
2 πat 4at 4at
2◦ . A semiinfinite strip is considered. The following conditions are prescribed:

Solution:
Z ∞Z l
0 0
Z tZ ∞
−a g1 (η, τ ) G(x, y, 0, η, t − τ ) dη dτ
0 0
Z tZ ∞
+a g2 (η, τ ) G(x, y, l, η, t − τ ) dη dτ
0 0
Z tZ l
∂
0 0 ∂η η=0
where
G(x, y, ξ, η, t) = G1 (x, ξ, t) G2 (y, η, t),

∞
l l l l l2
n=1

1 (y − η)2 (y + η)2
G2 (y, η, t) = √ exp − − exp − .
2 πat 4at 4at
∂w
= a∆2 w 413
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 . First boundary value problem.

A rectangle is considered. The following conditions are prescribed:
w = g2 (y, t) at x = l1 (boundary condition),
w = g3 (x, t) at y = 0 (boundary condition),
w = g4 (x, t) at y = l2 (boundary condition).
Solution:
Z l1Z l2
0 0
Z t Z l2
∂
0 0 ∂ξ ξ=0
Z t Z l2
∂
0 0 ∂ξ ξ=l1
Z t Z l1
∂
+a g3 (ξ, τ ) G(x, y, ξ, η, t − τ ) dξ dτ
0 0 ∂η η=0
Z t Z l1
∂
−a g4 (ξ, τ ) G(x, y, ξ, η, t − τ ) dξ dτ,
0 0 ∂η η=l2
where
∞ ∞ 2
4 XX nπx nπξ mπy mπη n m2
G(x, y, ξ, η, t) = sin sin sin sin exp −π 2 2 + 2 at .
l1 l2 n=1 m=1 l1 l1 l2 l2 l1 l2
zero temperature, g1 (y, t) = g2 (y, t) = g3 (x, t) = g4 (x, t) = 0.
Solution:
∞
16w0 X 1 (2n + 1)πx π 2 (2n + 1)2 at
w= sin exp −
π2 n=0
2n + 1 l1 l12
X ∞
1 (2m + 1)πy π 2 (2m + 1)2 at
× sin exp − .
m=0
2m + 1 l2 l22
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 . Second boundary value problem.

w = f (x, y) at t=0 (initial condition),
∂x w = g2 (y, t) at x = l1 (boundary condition),
∂y w = g3 (x, t) at y = 0 (boundary condition),
∂y w = g4 (x, t) at y = l2 (boundary condition).
Solution:
Z l1Z l2
0 0
Z tZ l2
−a g1 (η, τ ) G(x, y, 0, η, t − τ ) dη dτ
0 0
Z tZ l2
+a g2 (η, τ ) G(x, y, l1 , η, t − τ ) dη dτ
0 0
Z tZ l1
−a g3 (ξ, τ ) G(x, y, ξ, 0, t − τ ) dξ dτ
0 0
Z tZ l1
+a g4 (ξ, τ ) G(x, y, ξ, l2 , t − τ ) dξ dτ,
0 0
where
X∞
1 π 2 n2 at nπx nπξ
G(x, y, ξ, η, t) = 1+2 exp − 2 cos cos
l1 l2 l1 l1 l1
n=1
X∞
π 2 m2 at mπy mπη
× 1+2 exp − 2 cos cos .
m=1
l2 l2 l2
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 . Third boundary value problem.

∂x w + k2 w = g2 (y, t) at x = l1 (boundary condition),
∂y w − k3 w = g3 (x, t) at y = 0 (boundary condition),
∂y w + k4 w = g4 (x, t) at y = l2 (boundary condition).
X∞ X ∞
ϕn(x)ϕn(ξ) 2 ψm(y)ψm(η) 2
G(x, y, ξ, η, t) = exp(−aµnt) exp(−aλmt) ,
n=1
kϕnk2 m=1
kψmk2

k1 2 k2 µ2n +k12 k1 l1 k12
ϕn(x) = cos(µnx)+ sin(µnx), kϕnk = 2 2 + + 1+ 2 ,
µn 2µn µn +k22 2µ2n 2 µn
2 2
k3 2 k4 λm +k3 k3 l2 k32
ψm(y) = cos(λmy)+ sin(λmy), kψmk = 2 2 + + 1+ 2 .
λm 2λm λm +k42 2λ2m 2 λm
Here, the µn and λm are positive roots of the transcendental equations
tan(µl1 ) k1 + k2 tan(λl2 ) k3 + k4
= 2 , = 2 .
µ µ − k1 k2 λ λ − k3 k4
∂w
= a∆2 w 415
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 . Mixed boundary value problems.
1◦ . A rectangle is considered. The following conditions are prescribed:

Solution:
Z l1Z l2
0 0
Z tZ
l2
∂
0 0 ∂ξ ξ=0
Z t Z l2
∂
0 0 ∂ξ ξ=l1
Z t Z l1
−a g3 (ξ, τ ) G(x, y, ξ, 0, t − τ ) dξ dτ
0 0
Z tZ l1
0 0
where
∞
4 X nπx nπξ π 2 n2 at
G(x, y, ξ, η, t) = sin sin exp −
l1 l2
n=1
l1 l1 l12
∞
X
1 mπy mπη π 2 m2 at
× + cos cos exp − .
2 l2
m=1
l2 l22

Solution:
Z l1Z l2
0 0
Z t Z l2
∂
0 0 ∂ξ ξ=0
Z t Z l2
+a g2 (η, τ ) G(x, y, l1 , η, t − τ ) dη dτ
0 0
Z t Z l1
∂
0 0 ∂η η=0
Z t Z l1
0 0
where
∞
4 X π(2n+1)x π(2n+1)ξ aπ 2(2n+1)2t
G(x, y, ξ, η, t)= sin sin exp −
l1l2
n=0
2l1 2l1 4l12
X ∞
π(2m+1)y π(2m+1)η aπ 2(2m+1)2t
× sin sin exp − .
m=0
2l2 2l2 4l22
4.1.2 Problems in Polar Coordinates

The sourceless heat equation with two space variables in the polar coordinate system r, ϕ
has the form
2 p
∂w ∂ w 1 ∂w 1 ∂2w
=a + + 2 , r= x2 + y 2 .
∂t ∂r 2 r ∂r r ∂ϕ2
One-dimensional problems with axial symmetry that have solutions of the form w =
w(r, t) are considered in Section 3.2.1.
◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ 2π. Cauchy problem.
w = f (r, ϕ) at t = 0.
Solution:
Z 2πZ ∞ 2
1 r + ξ 2 − 2rξ cos(ϕ − η)
w(r, ϕ, t) = ξ exp − f (ξ, η) dξ dη.
4πat 0 0 4at
∂w
= a∆2 w 417
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π. First boundary value problem.

A circle is considered. The following conditions are prescribed:
w = f (r, ϕ) at t = 0 (initial condition),

w = g(ϕ, t) at r = R (boundary condition).
Solution:
Z 2πZ R
w(r, ϕ, t) = f (ξ, η) G(r, ϕ, ξ, η, t)ξ dξ dη
0 0
Z t Z 2π
∂
− aR g(η, τ ) G(r, ϕ, ξ, η, t − τ ) dη dτ.
0 0 ∂ξ ξ=R
Here,
∞ ∞
1 XX An
G(r, ϕ, ξ, η, t) = Jn (µnm r)Jn (µnm ξ) cos[n(ϕ − η)] exp(−µ2nm at),
πR n=0 m=1 [Jn (µnm R)]2
2 ′
A0 = 1, An = 2 (n = 1, 2, . . .),
where Jn (ξ) are Bessel functions (the prime denotes a derivative with respect to the argu-
ment), and µnm are positive roots of the transcendental equation Jn (µR) = 0.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π. Second boundary value problem.


∂r w = g(ϕ, t) at r = R (boundary condition).
Solution:
Z 2πZ R
0 0
Z t Z 2π
+ aR g(η, τ ) G(r, ϕ, R, η, t − τ ) dη dτ.
0 0
Here,
∞ ∞
1 1 X X Anµ2nmJn(µnmr)Jn(µnmξ)
G(r, ϕ, ξ, η, t)= + cos[n(ϕ−η)] exp(−µ2nmat),
πR2 π (µ2nmR2−n2)[Jn(µnmR)]2
n=0 m=1
A0 =1, An =2 (n=1, 2, . . .),
where Jn (ξ) are Bessel functions, and µnm are positive roots of the transcendental equation
Jn′ (µR) = 0.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π. Third boundary value problem.

∂r w + kw = g(ϕ, t) at r = R (boundary condition).
The solution w(r, ϕ, t) is determined by the formula in the previous paragraph (for the
∞ ∞
1 X X Anµ2nmJn(µnmr)Jn(µnmξ)
G(r, ϕ, ξ, η, t)= cos[n(ϕ−η)] exp(−µ2nmat),
π (µ2nmR2+k 2R2−n2)[Jn(µnmR)]2
n=0 m=1
A0 =1, An =2 (n=1, 2, . . .).
Here, Jn (ξ) are Bessel functions, and µnm are positive roots of the transcendental equation
µJn′ (µR) + kJn (µR) = 0.
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π. First boundary value problem.

An annular domain is considered. The following conditions are prescribed:
w = g1 (ϕ, t) at r = R1 (boundary condition),
w = g2 (ϕ, t) at r = R2 (boundary condition).
Solution:
Z 2πZ R2
0 R1
Z t Z 2π
∂
+ aR1 g1 (η, τ ) G(r, ϕ, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R1
Z t Z 2π
∂
− aR2 g2 (η, τ ) G(r, ϕ, ξ, η, t − τ ) dη dτ.
0 0 ∂ξ ξ=R2
Here,
∞ ∞
πXX
G(r, ϕ, ξ, η, t) = An Bnm Zn (µnm r)Zn (µnm ξ) cos[n(ϕ − η)] exp(−µ2nm at),
2
n=0 m=1
(
1/2 if n = 0, µ2nm Jn2 (µnm R2 )
An = Bnm = 2 ,
1 if n 6= 0, Jn (µnm R1 ) − Jn2 (µnm R2 )
Zn (µnm r) = Jn (µnm R1 )Yn (µnm r) − Yn (µnm R1 )Jn (µnm r),
where Jn (r) and Yn (r) are Bessel functions, and µnm are positive roots of the transcen-
dental equation
Jn (µR1 )Yn (µR2 ) − Yn (µR1 )Jn (µR2 ) = 0.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
∂w
= a∆2 w 419
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π. Second boundary value problem.


∂r w = g1 (ϕ, t) at r = R1 (boundary condition),
∂r w = g2 (ϕ, t) at r = R2 (boundary condition).
Solution:
Z 2πZ R2
0 R1
Z t Z 2π
− aR1 g1 (η, τ ) G(r, ϕ, R1 , η, t − τ ) dη dτ
0 0
Z tZ 2π
+ aR2 g2 (η, τ ) G(r, ϕ, R2 , η, t − τ ) dη dτ.
0 0
Here,
1
G(r, ϕ, ξ, η, t) =
π(R22 − R12 )
∞ X
X ∞
1 An µ2nm Zn (µnm r)Zn (µnm ξ) cos[n(ϕ − η)] exp(−µ2nm at)
+ ,
π n=0 m=1
(µ2nm R22 − n2 )Zn2 (µnm R2 ) − (µ2nm R12 − n2 )Zn2 (µnm R1 )
′
Zn (µnm r) = Jn (µnm R1 )Yn (µnm r) − Yn′ (µnm R1 )Jn (µnm r),
where A0 = 1 and An = 2 for n = 1, 2, . . . ; Jn (r) and Yn (r) are Bessel functions, and
µnm are positive roots of the transcendental equation
Jn′ (µR1 )Yn′ (µR2 ) − Yn′ (µR1 )Jn′ (µR2 ) = 0.
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π. Third boundary value problem.


∂r w − k1 w = g1 (ϕ, t) at r = R1 (boundary condition),
∂r w + k2 w = g2 (ϕ, t) at r = R2 (boundary condition).
Solution:
Z 2πZ R2
0 R1
Z t Z 2π
− aR1 g1 (η, τ ) G(r, ϕ, R1 , η, t − τ ) dη dτ
0 0
Z tZ 2π
+ aR2 g2 (η, τ ) G(r, ϕ, R2 , η, t − τ ) dη dτ.
0 0
Here,
∞ ∞
1 X X An 2
G(r, ϕ, ξ, η, t) = µ Zn (µnm r)Zn (µnm ξ) cos[n(ϕ − η)] exp(−µ2nm at),
π Bnm nm
n=0 m=1
Bnm = (k22 R22 + µ2nm R22 −
n2 )Zn2 (µnm R2 ) − (k12 R12 + µ2nm R12 − n2 )Zn2 (µnm R1 ),

Zn (µnm r) = µnm Jn′ (µnm R1 ) − k1 Jn (µnm R1 ) Yn (µnm r)

− µnm Yn′ (µnm R1 ) − k1 Yn (µnm R1 ) Jn (µnm r),
where A0 = 1 and An = 2 for n = 1, 2, . . .; Jn (r) and Yn (r) are Bessel functions, and

µJn′ (µR1 ) − k1 Jn (µR1 ) µYn′ (µR2 ) + k2 Yn (µR2 )

= µYn′ (µR1 ) − k1 Yn (µR1 ) µJn′ (µR2 ) + k2 Jn (µR2 ) .
◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 . First boundary value problem.
A wedge domain is considered. The following conditions are prescribed:

w = g1 (r, t) at ϕ = 0 (boundary condition),
w = g2 (r, t) at ϕ = ϕ0 (boundary condition).
Solution:
Z ϕ0Z ∞
0 0
Z tZ ∞
1 ∂
+a g1 (ξ, τ ) G(r, ϕ, ξ, η, t − τ ) dξ dτ
0 0 ξ ∂η η=0
Z tZ ∞
1 ∂
−a g2 (ξ, τ ) G(r, ϕ, ξ, η, t − τ ) dξ dτ.
0 0 ξ ∂η η=ϕ0
Here,
2 ∞ rξ nπϕ nπη
1 r + ξ2 X
G(r, ϕ, ξ, η, t) = exp − Inπ/ϕ0 sin sin ,
aϕ0 t 4at n=1
2at ϕ0 ϕ0
where Iν (r) are modified Bessel functions.

∂w
= a∆2 w 421
◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 . Second boundary value problem.

A wedge domain is considered. The following conditions are prescribed:

−1
r ∂ϕ w = g1 (r, t) at ϕ = 0 (boundary condition),
−1
r ∂ϕ w = g2 (r, t) at ϕ = ϕ0 (boundary condition).
Solution:
Z ϕ0Z ∞
0 0
Z tZ ∞
−a g1 (ξ, τ ) G(r, ϕ, ξ, 0, t − τ ) dξ dτ
0 0
Z tZ ∞
+a g2 (ξ, τ ) G(r, ϕ, ξ, ϕ0 , t − τ ) dξ dτ.
0 0
Here, 2
1 r + ξ2 1 rξ
G(r, ϕ, ξ, η, t) = exp − I0
aϕ0 t 4at 2 2at
X∞ rξ nπϕ nπη
+ Inπ/ϕ0 cos cos ,
2at ϕ0 ϕ0
n=1
where Iν (r) are modified Bessel functions.

◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 . First boundary value problem.

A circular sector is considered. The following conditions are prescribed:

w = g1 (ϕ, t) at r = R (boundary condition),
Solution:
Z ϕ0Z R
0 0
Z t Z ϕ0
∂
− aR g1 (η, τ ) G(r, ϕ, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R
Z tZ R
1 ∂
0 0 ξ ∂η η=0
Z tZ R
1 ∂
0 0 ξ ∂η η=ϕ0
Here,
X∞ X ∞
Jnπ/ϕ0(µnmr)Jnπ/ϕ0(µnmξ) nπϕ nπη
G(r, ϕ, ξ, η, t)=4 2ϕ [J ′ 2
sin sin exp(−µ2nmat),
n=1 m=1
R 0 nπ/ϕ0(µ nm R)] ϕ0 ϕ0
where the Jnπ/ϕ0 (r) are Bessel functions, and the µnm are positive roots of the transcen-
dental equation Jnπ/ϕ0 (µR) = 0.
Example 4.8. The initial temperature is uniform, f (r, ϕ) = w0 . The boundary is maintained at
zero temperature, g1 (ϕ, t) = g2 (r, t) = g3 (r, t) = 0.
Solution:
∞ ∞ Z R
8w0 X sin(snϕ) X 2 Jsn (µnm r) (2n + 1)π
w= 2
exp(−µ nm at) ′ 2
Jsn (µnm ξ)ξ dξ, sn = ,
πR n=0 2n + 1 m=1 [Jsn (µnmR)] 0 ϕ0
where the µnm are positive roots of the transcendental equation Jsn (µR) = 0.
(1984).
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 . Second boundary value problem.

∂r w = g1 (ϕ, t) at r = R (boundary condition),
−1
−1
Solution:
Z ϕ0Z R
0 0
Z t Z ϕ0
+ aR g1 (η, τ ) G(r, ϕ, R, η, t − τ ) dη dτ
0 0
Z tZ R
−a g2 (ξ, τ ) G(r, ϕ, ξ, 0, t − τ ) dξ dτ
0 0
Z tZ R
+a g3 (ξ, τ ) G(r, ϕ, ξ, ϕ0 , t − τ ) dξ dτ.
0 0
Here,
X∞ X ∞
2 µ2nm Jnπ/ϕ0 (µnm r)Jnπ/ϕ0 (µnm ξ)
G(r, ϕ, ξ, η, t) = + 4ϕ0
R2 ϕ0
n=0 m=1
(R2 ϕ20 µ2nm − n2 π 2 )[Jnπ/ϕ0 (µnm R)]2
nπϕ nπη
× cos cos exp(−µ2nm at),
ϕ0 ϕ0
where the Jnπ/ϕ0 (r) are Bessel functions, and the µnm are positive roots of the transcen-
′
dental equation Jnπ/ϕ (µR) = 0.
0

∂w
= a∆2 w 423
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 . Mixed boundary value problem.

∂r w − kw = g(ϕ, t) at r = R (boundary condition),
∂ϕ w = 0 at ϕ = 0 (boundary condition),
∂ϕ w = 0 at ϕ = ϕ0 (boundary condition).
Solution:
Z ϕ0Z R
0 0
Z t Z ϕ0
+ aR g(η, τ ) G(r, ϕ, R, η, t − τ ) dη dτ.
0 0
Here,
∞ X
X ∞
G(r, ϕ, ξ, η, t) = AnmJsn(µnmr)Jsn(µnmξ) cos(snϕ) cos(snη) exp(−µ2nmat),
n=0 m=1
nπ 4µ2nm
sn = , Anm = 2 ,
ϕ0 ϕ0(µ2nmR2 +k2R2 −s2n) Jsn(µnmR)
where Jsn (r) are Bessel functions, and µnm are positive roots of the transcendental equa-
tion
µJs′ n (µR) + kJsn (µR) = 0.
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ ϕ0 . Different boundary value problems.
Some problems for this domain were studied in Budak, Samarskii, and Tikhonov (1980).
4.1.3 Axisymmetric Problems

In the case of angular symmetry, the two-dimensional sourceless heat equation in the cylin-
drical coordinate system has the form
2 p
∂w ∂ w 1 ∂w ∂2w
=a + + , r= x2 + y 2 .
∂t ∂r 2 r ∂r ∂z 2
This equation governs two-dimensional unsteady thermal processes in quiescent media or

solid bodies (bounded by coordinate surfaces of the cylindrical system) in the case where
the initial and boundary conditions are independent of the angular coordinate. A similar
equation is used to study analogous two-dimensional unsteady mass transfer phenomena.
◮ Particular solutions. Remarks on the Green’s functions.

1◦ . Apart from usual separable solutions w(r, z, t) = f1 (r)f2 (z)f3 (t), the equation in ques-
tion has more sophisticated solutions in the product form
w(r, z, t) = u(r, t)v(z, t),
where u = u(r, t) and v = v(z, t) are solutions of the simpler one-dimensional equations
2
∂u ∂ u 1 ∂u
=a + (see Section 3.2.1 for particular solutions of this equation),
∂t ∂r 2 r ∂r
∂v ∂ 2v
=a 2 (see Section 3.1.1 for particular solutions of this equation).
∂t ∂z
2◦ . For all two-dimensional boundary value problems considered in Section 4.1.3, the
Green’s function can be represented in the product form
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),
where G1 (r, ξ, t) and G2 (z, η, t) are the Green’s functions of appropriate one-dimensional
boundary value problems.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ z < ∞. First boundary value problem.

A semiinfinite circular cylinder is considered. The following conditions are prescribed:
w = f (r, z) at t = 0 (initial condition),
w = g1 (z, t) at r = R (boundary condition),
w = g2 (r, t) at z = 0 (boundary condition).
Solution:
Z ∞Z R
w(r, z, t) = 2π ξf (ξ, η) G(r, z, ξ, η, t) dξ dη
0 0
Z tZ ∞
∂
− 2πaR g1 (η, τ ) G(r, z, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R
Z tZ R
∂
+ 2πa ξg2 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ.
0 0 ∂η η=0
Here,
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),
∞
1 X 1 µn r µn ξ aµ2n t
G1 (r, ξ, t) = J0 J0 exp − 2 ,
πR2 n=1 J12 (µn ) R R R

1 (z − η)2 (z + η)2
G2 (z, η, t) = √ exp − − exp − ,
2 πat 4at 4at
where the µn are positive zeros of the Bessel function, J0 (µn ) = 0. The numerical values
of the first ten roots µn are specified in Section 3.2.1 (see the first boundary value problem
for 0 ≤ r ≤ R).
∂w
= a∆2 w 425
Example 4.9. The initial temperature is the same at every point of the cylinder, f (r, z) = w0 .
The lateral surface and the end face are maintained at zero temperature, g1 (r, t) = g2 (z, t) = 0.
Solution:
X ∞
2w0 z J0 (λn r) µn
w(r, z, t) = erf √ exp(−λ2n at), λn = .
R 2 at n=1 λn J1 (λn R) R
Example 4.10. The initial temperature of the cylinder is everywhere zero, f (r, z) = 0. The
lateral surface r = R is maintained at a constant temperature w0 , and the end face z = 0 at zero
temperature.
Solution:
∞
w0 X J0 (λn r) z
w(r, z, t) = w0 − 2 exp(−λ2n at) erf √
R n=1 λn J1 (λn R) 2 at

z √ z √
+ exp(λn z) erfc √ + λn at + exp(−λn z) erfc √ − λn at ,
2 at 2 at
where the λn are positive zeros of the Bessel function, J0 (λR) = 0.
(1984).
◮ Domain: 0 ≤ r ≤ R, 0 ≤ z < ∞. Second boundary value problem.

∂r w = g1 (z, t) at r = R (boundary condition),
∂z w = g2 (r, t) at z = 0 (boundary condition).
Solution:
Z ∞Z R
0 0
Z tZ ∞
+ 2πaR g1 (η, τ ) G(r, z, R, η, t − τ ) dη dτ
0 0
Z tZ R
− 2πa ξg2 (ξ, τ ) G(r, z, ξ, 0, t − τ ) dξ dτ.
0 0
Here,
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),
∞
1 1 X 1 µn r µn ξ aµ2n t
G1 (r, ξ, t) = + J0 J0 exp − 2 ,
πR2 πR2
n=1 0
J 2 (µn ) R R R

1 (z − η)2 (z + η)2
G2 (z, η, t) = √ exp − + exp − ,
2 πat 4at 4at
where the µn are positive zeros of the first-order Bessel function, J1 (µn ) = 0. The numer-
ical values of the first ten roots µn are specified in Section 3.2.1 (see the second boundary
value problem for 0 ≤ r ≤ R).
◮ Domain: 0 ≤ r ≤ R, 0 ≤ z < ∞. Third boundary value problem.

∂r w + k1 w = g1 (z, t) at r = R (boundary condition),
∂z w − k2 w = g2 (r, t) at z = 0 (boundary condition).
The solution w(r, z, t) is determined by the formula in the previous paragraph (for the
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),
∞
X
1 µ2n µn r µn ξ aµ2n t
G1 (r, ξ, t) = J0 J0 exp − 2 ,
πR2
n=1
(k12 R2 + µ2n )J02 (µn ) R R R

1 (z − η)2 (z + η)2
G2 (z, η, t) = √ exp − + exp −
2 πat 4at 4at
Z ∞ 2

(z + η + s)
− 2k2 exp − − k2 s ds .
0 4at
Here, J0 (µ) is the zeroth Bessel function, and the µn are positive roots of the transcendental
equation
µJ1 (µ) − k1 RJ0 (µ) = 0.
At the lateral surface and the end face, heat exchange of the cylinder with the zero temperature
environment occurs, g1 (z, t) = g2 (r, t) = 0.
Solution:
X ∞
2w0k1 z z √ J0(νnr) exp(−νn2 at)
w(r, z, t)= erf √ +exp(k2z+k22at) erfc √ +k2 at ,
R 2 at 2 at n=1
(k12+νn2 )J0(νnR)
where the νn are positive roots of the transcendental equation νJ1 (νR) − k1 J0 (νR) = 0.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ z < ∞. Mixed boundary value problems.

1◦ . A semiinfinite circular cylinder is considered. The following boundary conditions are
prescribed:
Solution:
Z ∞Z R
0 0
Z tZ ∞
∂
0 0 ∂ξ ξ=R
Z tZ R
− 2πa ξg2 (ξ, τ ) G(r, z, ξ, 0, t − τ ) dξ dτ.
0 0
∂w
= a∆2 w 427
Here,
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),

∞
G1 (r, ξ, t) = J0 J 0 exp − ,
πR2 n=1 J12 (µn ) R R R2

1 (z − η)2 (z + η)2
G2 (z, η, t) = √ exp − + exp − ,
2 πat 4at 4at
of the first ten µn are specified in Section 3.2.1 (see the first boundary value problem for
0 ≤ r ≤ R).
2◦ . A semiinfinite circular cylinder is considered. The following conditions are prescribed:

Solution:
Z ∞Z R
0 0
Z tZ ∞
0 0
Z tZ R
∂
+ 2πa ξg2 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ.
0 0 ∂η η=0
Here,
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),

∞
G1 (r, ξ, t) = + J0 J0 exp − 2 ,
πR2 πR2 n=1 J02 (µn ) R R R

1 (z − η)2 (z + η)2
G2 (z, η, t) = √ exp − − exp − ,
2 πat 4at 4at

∂r w + kw = g1 (z, t) at r = R (boundary condition),
The solution w(r, z, t) is determined by the formula in Item 2◦ (see above) where
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),
∞
X
G1 (r, ξ, t) = J0 J0 exp − 2 ,
πR2 n=1 (k2 R2 + µ2n )J02 (µn ) R R R

1 (z − η)2 (z + η)2
G2 (z, η, t) = √ exp − − exp − ,
2 πat 4at 4at
µJ1 (µ) − kRJ0 (µ) = 0.
Heat exchange of the cylinder with the zero temperature environment occurs at the lateral surface,
g1 (z, t) = 0. The end face is maintained at zero temperature, g2 (r, t) = 0.
Solution:
X ∞
2w0 k z J0 (λn r) µn
w(r, z, t) = erf √ 2 + λ2 )J (λ R)
exp(−λ2n at), λn = .
R 2 at n=1 (k n 0 n R
Example 4.13. The initial temperature of the cylinder is everywhere zero, f (r, z) = 0. Heat ex-
change of the cylinder with the zero temperature environment occurs at the lateral surface, g1 (z, t) =
0. The end face is maintained at a constant temperature, g2 (r, t) = w0 .
Solution:
∞
kw0 X J0(λnr)
w(r, z, t)= 2 exp(−λnz)
R n=1 (λ2n+k 2)J0(λnR)

√ z √ z µn
+exp(λnz) erfc λn at+ √ −exp(−λnz) erfc λn at− √ , λn = .
2 at 2 at R
◮ Domain: 0 ≤ r ≤ R, 0 ≤ z ≤ l. First boundary value problem.

A circular cylinder of finite length is considered. The following conditions are prescribed:
w = g2 (r, t) at z = 0 (boundary condition),
w = g3 (r, t) at z = l (boundary condition).
Solution:
Z lZ R
0 0
Z tZ l
∂
0 0 ∂ξ ξ=R
Z tZ R
∂
+ 2πa ξg2 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ
0 0 ∂η η=0
Z tZ R
∂
− 2πa ξg3 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ.
0 0 ∂η η=l
∂w
= a∆2 w 429
Here,
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),

∞
G1 (r, ξ, t) = J0 J0 exp − 2 ,
πR2
n=1 1
J 2 (µn ) R R R
∞
2X nπz nπη an2 π 2 t
G2 (z, η, t) = sin sin exp − ,
l n=1 l l l2
0 ≤ r ≤ R).
The lateral surface and the end faces are maintained at zero temperature, g1 (z, t) = g2 (r, t) =
g3 (r, t) = 0.
Solution:
∞
8w0 X 1 (2n + 1)πz a(2n + 1)2 π 2 t
w= sin exp −
π n=0
2n + 1 l l2
X∞
1 µn r µ2 at
× J0 exp − n 2 ,
µ J (µ )
n=1 n 1 n
R R
where µn are positive zeros of the Bessel function, J0 (µn ) = 0. The numerical values of the first
ten µn are specified in Section 3.2.1 (see the first boundary value problem for 0 ≤ r ≤ R).
◮ Domain: 0 ≤ r ≤ R, 0 ≤ z ≤ l. Second boundary value problem.


∂z w = g2 (r, t) at z = 0 (boundary condition),
∂z w = g3 (r, t) at z = l (boundary condition).
Solution:
Z lZ R
0 0
Z tZ l
0 0
Z tZ R
− 2πa ξg2 (ξ, τ ) G(r, z, ξ, 0, t − τ ) dξ dτ
0 0
Z tZ R
+ 2πa ξg3 (ξ, τ ) G(r, z, ξ, l, t − τ ) dξ dτ.
0 0
Here,
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),

∞
G1 (r, ξ, t) = + J 0 J0 exp − ,
πR2 πR2 n=1 J02 (µn ) R R R2
∞
1 2X nπz nπη an2 π 2 t
G2 (z, η, t) = + cos cos exp − ,
l l l l l2
n=1
◮ Domain: 0 ≤ r ≤ R, 0 ≤ z ≤ l. Third boundary value problem.

∂z w − k2 w = g2 (r, t) at z = 0 (boundary condition),
∂z w + k3 w = g3 (r, t) at z = l (boundary condition).
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),

∞
X
1 µ2n µn r µn ξ aµ2nt
G1 (r, ξ, t) = J0 J0 exp − 2 ,
πR2
n=1
(k12 R2 + µ2n)J02 (µn) R R R
∞
X
ϕm(z)ϕm (η) k2
G2 (z, η, t) = 2
exp(−aλ2mt), ϕm(z) = cos(λmz) + sin(λmz),
m=1
kϕm k λm

2 k3 λ2m + k22 k2 l k22
kϕmk = 2 2 + + 1+ 2 ,
2λm λm + k32 2λ2m 2 λm
and the µn and λm are positive roots of the transcendental equations
tan(λl) k2 + k3
µJ1 (µ) − k1 RJ0 (µ) = 0, = 2 .
λ λ − k2 k3
◮ Domain: 0 ≤ r ≤ R, 0 ≤ z ≤ l. Mixed boundary value problems.

1◦ . A circular cylinder of finite length is considered. The following conditions are pre-
scribed:
∂w
= a∆2 w 431
Solution:
Z lZ R
0 0
Z tZ l
∂
0 0 ∂ξ ξ=R
Z tZ R
0 0
Z tZ R
0 0
Here,
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),

∞
G1 (r, ξ, t) = 2 (µ ) J0 J0 exp − 2 ,
πR2 J 1 n R R R
n=1
∞
l l n=1 l l l2
0 ≤ r ≤ R).
scribed:
Solution:
Z lZ R
0 0
Z tZ l
0 0
Z tZ R
∂
0 0 ∂η η=0
Z tZ R
∂
0 0 ∂η η=l
Here,
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),
∞
G1 (r, ξ, t) = + J0 J0 exp − 2 ,
∞
l l l l2
n=1
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ z ≤ l. First boundary value problem.

A hollow circular cylinder of finite length is considered. The following conditions are
prescribed:
w = g1 (z, t) at r = R1 (boundary condition),
Solution:
Z lZ R2
0 R1
Z tZ l
∂
+ 2πaR1 g1 (η, τ ) G(r, z, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R1
Z tZ l
∂
− 2πaR2 g2 (η, τ ) G(r, z, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R2
Z t Z R2
∂
0 R1 ∂η η=0
Z t Z R2
∂
0 R1 ∂η η=l
Here,
G(r, z, ξ, η, t) = G1 (z, η, t) G2 (r, ξ, t),
∞
X
2 nπz nπη an2 π 2 t
l n=1 l l l2
∞
π X µ2n J02 (sµn ) aµ2n t
G2 (r, ξ, t) = Ψn (r)Ψn (ξ) exp − 2 ,
4R12 n=1 J02 (µn ) − J02 (sµn ) R1

µn r µn r R2
Ψn (r) = Y0 (µn )J0 − J0 (µn )Y0 , s= ,
R1 R1 R1
∂w
= a∆2 w 433
where J0 (µ) and Y0 (µ) are Bessel functions, the µn are positive roots of the transcendental
equation
J0 (µ)Y0 (sµ) − J0 (sµ)Y0 (µ) = 0.
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ z ≤ l. Second boundary value problem.
prescribed:
∂r w = g1 (z, t) at r = R1 (boundary condition),
Solution:
Z lZ R2
0 R1
Z tZ l
− 2πaR1 g1 (η, τ ) G(r, z, R1 , η, t − τ ) dη dτ
0 0
Z tZ l
+ 2πaR2 g2 (η, τ ) G(r, z, R2 , η, t − τ ) dη dτ
0 0
Z t Z R2
0 R1
Z tZ R2
0 R1
Here,
G(r, z, ξ, η, t) = G1 (z, η, t) G2 (r, ξ, t),

∞
X
1 2 nπz nπη an2π 2t
G1(z, η, t) = + cos cos exp − ,
l l l l l2
n=1
∞
1 π X µ2nJ12(sµn) aµ2nt
G2 (r, ξ, t) = + Ψn(r)Ψn(ξ) exp − 2 ,
π(R22 − R12) 4R12 n=1 J12 (µn) − J12(sµn) R1

µn r µn r R2
Ψn(r) = Y1(µn)J0 − J1(µn )Y0 , s= ,
R1 R1 R1
where Jk (µ) and Yk (µ) are Bessel functions of order k = 0, 1 and the µn are positive roots
of the transcendental equation
J1 (µ)Y1 (sµ) − J1 (sµ)Y1 (µ) = 0.

◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ z ≤ l. Third boundary value problem.
prescribed:

∂r w − k1 w = g1 (z, t) at r = R1 (boundary condition),
∂r w + k2 w = g2 (z, t) at r = R2 (boundary condition),
For the solution of this problem, see the last paragraph in Section 4.2.3 with Φ ≡ 0.
∂w
4.2 Heat Equation with a Source = a∆2w + Φ(x, y, t)
∂t
4.2.1 Problems in Cartesian Coordinates

In the rectangular Cartesian coordinate system, the heat equation has the form
2
∂w ∂ w ∂2w
=a + + Φ(x, y, t).
∂t ∂x2 ∂y 2
It governs two-dimensional unsteady thermal processes in quiescent media or solids with

constant thermal diffusivity in the cases where there are volume thermal sources or sinks.
w = f (x, y) at t = 0.
Solution:
Z ∞Z ∞
−∞ −∞
Z tZ ∞ Z ∞
+ Φ(ξ, η, τ ) G(x, y, ξ, η, t − τ ) dξ dη dτ,
0 −∞ −∞
where
1 (x − ξ)2 + (y − η)2
G(x, y, ξ, η, t) = exp − .
4πat 4at

∂w
4.2. Heat Equation with a Source ∂t
= a∆2 w + Φ(x, y, t) 435
◮ Domain: 0 ≤ x < ∞, −∞ < y < ∞. First boundary value problem.

w = g(y, t) at x = 0 (boundary condition).
Solution:
Z ∞Z ∞
0 −∞
Z tZ ∞ h ∂ i
+a g(η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
0 −∞ ∂ξ ξ=0
Z t Z ∞Z ∞
+ Φ(ξ, η, τ ) G(x, y, ξ, η, t − τ ) dη dξ dτ,
0 0 −∞
where

1 (x − ξ)2 + (y − η)2 (x + ξ)2 + (y − η)2
G(x, y, ξ, η, t) = exp − − exp − .
4πat 4at 4at
◮ Domain: 0 ≤ x < ∞, −∞ < y < ∞. Second boundary value problem.

∂x w = g(y, t) at x = 0 (boundary condition).
Solution:
Z ∞Z ∞
0 −∞
Z tZ ∞
−a g(η, τ ) G(x, y, 0, η, t − τ ) dη dτ
0 −∞
Z t Z ∞Z ∞
+ Φ(ξ, η, τ ) G(x, y, ξ, η, t − τ ) dη dξ dτ,
0 0 −∞
where

1 (x − ξ)2 + (y − η)2 (x + ξ)2 + (y − η)2
G(x, y, ξ, η, t) = exp − + exp − .
4πat 4at 4at
◮ Domain: 0 ≤ x < ∞, −∞ < y < ∞. Third boundary value problem.

∂x w − kw = g(y, t) at x = 0 (boundary condition).

1 (y − η)2 (x − ξ)2 (x + ξ)2
G(x, y, ξ, η, t) = exp − exp − + exp −
4πat 4at 4at 4at
Z ∞ 2

(x + ξ + s)
− 2k exp − − ks ds .
0 4at

Solution:
Z ∞Z ∞
0 0
Z tZ ∞ h ∂ i
0 0 ∂ξ ξ=0
Z tZ ∞ h ∂ i
0 0 ∂η η=0
Z t Z ∞Z ∞
0 0 0
where

1 (x − ξ)2 (x + ξ)2
G(x, y, ξ, η, t) = exp − − exp −
4πat 4at 4at
2
2

(y − η) (y + η)
× exp − − exp − .
4at 4at

∂w
= a∆2 w + Φ(x, y, t) 437


Solution:
Z ∞Z ∞
0 0
Z tZ ∞
−a g1 (η, τ ) G(x, y, 0, η, t − τ ) dη dτ
0 0
Z tZ ∞
−a g2 (ξ, τ ) G(x, y, ξ, 0, t − τ ) dξ dτ
0 0
Z tZ ∞Z ∞
0 0 0
where

1 (x − ξ)2 (x + ξ)2
G(x, y, ξ, η, t) = exp − + exp −
4πat 4at 4at
2 2
(y − η) (y + η)
× exp − + exp − .
4at 4at
◮ Domain: 0 ≤ x < ∞, 0 ≤ y < ∞. Third boundary value problem.



1 (x − ξ)2 (x + ξ)2
G(x, y, ξ, η, t) = exp − + exp −
4πat 4at 4at

√ 2 x+ξ √
− 2k1 πat exp ak1 t + k1 (x + ξ) erfc √ + k1 at
2 at

(y − η)2 (y + η)2
4at 4at

√ 2 y+η √
− 2k2 πat exp ak2 t + k2 (y + η) erfc √ + k2 at .
2 at
◮ Domain: 0 ≤ x < ∞, 0 ≤ y < ∞. Mixed boundary value problem.

Solution:
Z ∞Z ∞
0 0
Z tZ ∞ h ∂ i
0 0 ∂ξ ξ=0
Z tZ ∞
−a g2 (ξ, τ ) G(x, y, ξ, 0, t − τ ) dξ dτ
0 0
Z tZ ∞Z ∞
0 0 0
where

1 (x − ξ)2 (x + ξ)2
G(x, y, ξ, η, t) = exp − − exp −
4πat 4at 4at
2
2

(y − η) (y + η)
× exp − + exp − .
4at 4at
◮ Domain: 0 ≤ x ≤ l, 0 ≤ y < ∞. First boundary value problem.

The solution is given by the formula of Section 4.1.1 (see the first boundary value
problem for 0 ≤ x ≤ l, 0 ≤ y < ∞) with the additional term
Z tZ lZ ∞
Φ(ξ, η, τ ) G(x, y, ξ, η, t − τ ) dη dξ dτ,
0 0 0
which takes into account the equation’s nonhomogeneity.
◮ Domain: 0 ≤ x ≤ l, 0 ≤ y < ∞. Second boundary value problem.

∂w
= a∆2 w + Φ(x, y, t) 439
Solution:
Z ∞Z l Z tZ ∞Z l
w(x, y, t) = f (ξ, η) G(x, y, ξ, η, t) dξ dη+ Φ(ξ, η, τ ) G(x, y, ξ, η, t−τ ) dξ dη dτ
0 0 0 0 0
Z tZ ∞ Z tZ ∞
−a g1(η, τ ) G(x, y, 0, η, t−τ ) dη dτ +a g2(η, τ ) G(x, y, l, η, t−τ ) dη dτ
0 0 0 0
Z tZ l
−a g3(ξ, τ ) G(x, y, ξ, 0, t−τ ) dξ dτ,
0 0
where
G(x, y, ξ, η, t) = G1 (x, ξ, t) G2 (y, η, t),

∞ nπx nπξ
1 2X an2 π 2 t
l l l l l2
n=1

1 (y − η)2 (y + η)2
G2 (y, η, t) = √ exp − + exp − .
2 πat 4at 4at
◮ Domain: 0 ≤ x ≤ l, 0 ≤ y < ∞. Third boundary value problem.

∂x w + k2 w = g2 (y, t) at x = l (boundary condition),
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),

X∞
G1 (r, ξ, t) = J0 J0 exp − 2 ,
πR2
n=1
(k12 R2 + µ2n)J02 (µn) R R R
X∞
ϕm(z)ϕm (η) k2
G2 (z, η, t) = 2
exp(−aλ2mt), ϕm(z) = cos(λmz) + sin(λmz),
m=1
kϕm k λm

2 k3 λ2m + k22 k2 l k22
kϕmk = 2 2 + + 1+ 2 ,
tan(λl) k2 + k3
µJ1 (µ) − k1 RJ0 (µ) = 0, = 2 .
λ λ − k2 k3
◮ Domain: 0 ≤ x ≤ l, 0 ≤ y < ∞. Mixed boundary value problem.


The solution is given by the formula of Section 4.1.1 (see Item 1◦ of mixed boundary
value problems for 0 ≤ x ≤ l, 0 ≤ y < ∞) with the additional term
Z tZ lZ ∞
0 0 0


The solution is given by the formula of Section 4.1.1 (see the first boundary value
problem for 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 ) with the additional term
Z t Z l1Z l2
0 0 0



∂w
= a∆2 w + Φ(x, y, t) 441
Solution:
Z l1Z l2 Z tZ l1Z l2
w(x, y, t) = f (ξ, η) G(x, y, ξ, η, t) dη dξ+ Φ(ξ, η, τ ) G(x, y, ξ, η, t−τ ) dη dξ dτ
0 0 0 0 0
Z t Z l2 Z tZ l2
−a g1(η, τ ) G(x, y, 0, η, t−τ ) dη dτ +a g2(η, τ ) G(x, y, l1 , η, t−τ ) dη dτ
0 0 0 0
Z tZ l1 Z t Z l1
−a g3(ξ, τ ) G(x, y, ξ, 0, t−τ ) dξ dτ +a g4(ξ, τ ) G(x, y, ξ, l2 , t−τ ) dξ dτ,
0 0 0 0
where
X∞
1 π 2 n2 at nπx nπξ
G(x, y, ξ, η, t) = 1+2 exp − cos cos
l1 l2 n=1
l12 l1 l1
X∞
π 2 m2 at mπy mπη
× 1+2 exp − cos cos .
m=1
l22 l2 l2

X
∞ X ∞
ϕn(x)ϕn(ξ) 2 ψm(y)ψm(η) 2
G(x, y, ξ, η, t) = exp(−aµnt) exp(−aλmt) ,
n=1
kϕnk2 m=1
kψmk2

k1 k2 µ2 +k2 k1 l1 k2
ϕn(x) = cos(µnx)+ sin(µnx), kϕnk2 = 2 n2 12 + 2 + 1+ 12 ,
µn 2µn µn +k2 2µn 2 µn
2 2

k3 2 k4 λm +k3 k3 l2 k32
ψm(y) = cos(λmy)+ sin(λmy), kψmk = 2 2 + + 1+ 2 .
Here, the µn and λm are positive roots of the transcendental equations
= 2 , = 2 .
µ µ − k1 k2 λ λ − k3 k4
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 . Mixed boundary value problem.

The solution is given by the formula of Section 4.1.1 (see Item 1◦ of mixed boundary
value problems for 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 ) with the additional term
Z t Z l1Z l2
0 0 0

The heat equation with a volume source in the polar coordinate system r, ϕ is written as
2
∂w ∂ w 1 ∂w 1 ∂2w
=a + + 2 + Φ(r, ϕ, t).
∂t ∂r 2 r ∂r r ∂ϕ2
Solutions of the form w = w(r, t) that are independent of the angular coordinate ϕ and
govern plane thermal processes with central symmetry are presented in Section 3.2.2.
◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ 2π. Cauchy problem.
w = f (r, ϕ) at t = 0.
Solution:
Z 2πZ ∞
w(r, ϕ, t) = f (ξ, η) G(r, ϕ, ξ, η, t) ξ dξ dη
0 0
Z tZ 2πZ ∞
+ Φ(ξ, η, τ ) G(r, ϕ, ξ, η, t − τ ) ξ dξ dη dτ,
0 0 0
where 2
1 r + ξ 2 − 2rξ cos(ϕ − η)
G(r, ϕ, ξ, η, t) = exp − .
4πat 4at
∂w
= a∆2 w + Φ(x, y, t) 443

Solution:
Z 2πZ R
0 0
Z t Z 2π
∂
− aR g(η, τ ) G(r, ϕ, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R
Z t Z 2πZ R
+ Φ(ξ, η, τ ) G(r, ϕ, ξ, η, t − τ )ξ dξ dη dτ.
0 0 0
Here,
∞ ∞
1 XX An
G(r, ϕ, ξ, η, t) = Jn (µnm r)Jn (µnm ξ) cos[n(ϕ − η)] exp(−µ2nm at),
2 ′
A0 = 1, An = 2 (n = 1, 2, . . .),
where Jn (ξ) are Bessel functions (the prime denotes a derivative with respect to the argu-
ment), and µnm are positive roots of the transcendental equation Jn (µR) = 0.

Solution:
Z 2πZ R
0 0
Z t Z 2π
+ aR g(η, τ ) G(r, ϕ, R, η, t − τ ) dη dτ
0 0
Z t Z 2πZ R
+ Φ(ξ, η, τ ) G(r, ϕ, ξ, η, t − τ )ξ dξ dη dτ.
0 0 0
Here,
∞ ∞
G(r, ϕ, ξ, η, t)= + cos[n(ϕ−η)] exp(−µ2nmat),
πR2 π n=0 m=1 (µ2nmR2−n2)[Jn(µnmR)]2
A0 =1, An =2 (n=1, 2, . . .),
where Jn (ξ) are Bessel functions, and µnm are positive roots of the transcendental equation
Jn′ (µR) = 0.

∞ ∞
1 X X Anµ2nmJn(µnmr)Jn(µnmξ)
G(r, ϕ, ξ, η, t)= cos[n(ϕ−η)] exp(−µ2nmat),
π (µ2nmR2+k 2R2−n2)[Jn(µnmR)]2
n=0 m=1
A0 =1, An =2 (n=1, 2, . . .).
Here, Jn (ξ) are Bessel functions, and µnm are positive roots of the transcendental equation
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π. Different boundary value problems.

1◦ . The solution of the first boundary value problem for an annular domain is given by the
formula in Section 4.1.2 (see the first boundary value problem for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π)
with the additional term
Z t Z 2πZ R2
Φ(ξ, η, τ ) G(r, ϕ, ξ, η, t − τ )ξ dξ dη dτ, (1)
0 0 R1
which allows for the equation’s nonhomogeneity.

2◦ . The solution of the second boundary value problem for an annular domain is given by
the formula in Section 4.1.2 (see the second boundary value problem for R1 ≤ r ≤ R2 , 0 ≤
ϕ ≤ 2π) with the additional term (1); the Green’s function is also taken from Section 4.1.2.
3◦ . The solution of the third boundary value problem for an annular domain is given by the
formula in Section 4.1.2 (see the third boundary value problem for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π)
with the additional term (1); the Green’s function is also taken from Section 4.1.2.
◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 . Different boundary value problems.

1◦ . The solution of the first boundary value problem for a wedge domain is given by the
formula in Section 4.1.2 (see the first boundary value problem for 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 )
Z t Z ϕ0Z ∞
0 0 0
2◦ . The solution of the second boundary value problem for a wedge domain is given by the
formula in Section 4.1.2 (see the second boundary value problem for 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 )
∂w
= a∆2 w + Φ(x, y, t) 445
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 . Different boundary value problems.
1◦ . The solution of the first boundary value problem for a sector of a circle is given by the
formula of Section 4.1.2 (see the first boundary value problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 )
Z tZ ϕ0Z R
0 0 0
2◦ . The solution of the mixed boundary value problem for a sector of a circle is given by the
formula of Section 4.1.2 (see the mixed boundary value problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 )

In the case of axial symmetry, the heat equation in the cylindrical coordinate system is
written as
2
∂w ∂ w 1 ∂w ∂2w
=a + + + Φ(r, z, t),
∂t ∂r 2 r ∂r ∂z 2
provided there are heat sources or sinks.
One-dimensional axisymmetric problems that have solutions of the form w = w(r, t)
can be found in Section 3.2.2.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ z < ∞. First boundary value problem.

Solution:
Z ∞Z R
0 0
Z tZ ∞
∂
0 0 ∂ξ ξ=R
Z tZ R
∂
0 0 ∂η η=0
Z t Z ∞Z R
+ 2π ξΦ(ξ, η, τ ) G(r, z, ξ, η, t − τ ) dξ dη dτ.
0 0 0
Here,
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),

∞
G1 (r, ξ, t) = J0 J0 exp − 2 ,
πR2 n=1 J12 (µn ) R R R

1 (z − η)2 (z + η)2
G2 (z, η, t) = √ exp − − exp − ,
2 πat 4at 4at
0 ≤ r ≤ R).
◮ Domain: 0 ≤ r ≤ R, 0 ≤ z < ∞. Second boundary value problem.


Solution:
Z ∞Z R
0 0
Z tZ ∞
0 0
Z tZ R
0 0
Z t Z ∞Z R
0 0 0
Here,
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),

∞
G1 (r, ξ, t) = + J0 J0 exp − 2 ,
πR2 πR2
n=1 0
J 2 (µn ) R R R

1 (z − η)2 (z + η)2
G2 (z, η, t) = √ exp − + exp − ,
2 πat 4at 4at
∂w
= a∆2 w + Φ(x, y, t) 447
◮ Domain: 0 ≤ r ≤ R, 0 ≤ z < ∞. Third boundary value problem.

∂z w − k2 w = g2 (r, t) at z = 0 (boundary condition).
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),

∞
X
G1 (r, ξ, t) = J0 J0 exp − 2 ,
πR2 n=1 (k12 R2 + µ2n )J02 (µn ) R R R

1 (z − η)2 (z + η)2
G2 (z, η, t) = √ exp − + exp −
2 πat 4at 4at
Z ∞ 2
(z + η + s)
− 2k2 exp − − k2 s ds .
0 4at
Here, J0 (µ) is the zeroth Bessel function, and the µn are positive roots of the transcendental
equation
µJ1 (µ) − k1 RJ0 (µ) = 0.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ z < ∞. Mixed boundary value problems.

Solution:
Z ∞Z R
0 0
Z tZ ∞
∂
0 0 ∂ξ ξ=R
Z tZ R
0 0
Z tZ ∞Z R
0 0 0
Here,
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),

∞
G1 (r, ξ, t) = J0 J0 exp − 2 ,
πR2
n=1 1
J 2 (µn ) R R R
2
2

1 (z − η) (z + η)
G2 (z, η, t) = √ exp − + exp − ,
2 πat 4at 4at
0 ≤ r ≤ R).

Solution:
Z ∞Z R
0 0
Z tZ ∞
0 0
Z tZ R
∂
0 0 ∂η η=0
Z tZ ∞Z R
0 0 0
Here,
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),

∞
G1 (r, ξ, t) = + J0 J0 exp − 2 ,

1 (z − η)2 (z + η)2
G2 (z, η, t) = √ exp − − exp − ,
2 πat 4at 4at

∂r w + kw = g1 (z, t) at r = R (boundary condition),
∂w
= a∆2 w + Φ(x, y, t) 449
The solution w(r, z, t) is determined by the formula in Item 2◦ (see above) where
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),

∞
X
G1 (r, ξ, t) = J
2 R2 + µ2 )J 2 (µ ) 0
J0 exp − ,
πR2 (k n 0 n R R R2
n=1

1 (z − η)2 (z + η)2
G2 (z, η, t) = √ exp − − exp − ,
2 πat 4at 4at
µJ1 (µ) − kRJ0 (µ) = 0.

Solution:
Z lZ R
0 0
Z tZ l
∂
0 0 ∂ξ ξ=R
Z tZ R
∂
0 0 ∂η η=0
Z tZ R
∂
− 2πa ξg3 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ
0 0 ∂η η=l
Z tZ l Z R
0 0 0
Here,
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),

∞
G1 (r, ξ, t) = J0 J0 exp − 2 ,
πR2 n=1 J12 (µn ) R R R
∞
l l l l2
n=1
0 ≤ r ≤ R).

Solution:
Z lZ R
0 0
Z tZ l
0 0
Z tZ R
0 0
Z tZ R
+ 2πa ξg3 (ξ, τ ) G(r, z, ξ, l, t − τ ) dξ dτ
0 0
Z tZ l Z R
0 0 0
Here,
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),

∞
G1 (r, ξ, t) = + J0 J0 exp − 2 ,
πR2 πR2
n=1 0
J 2 (µn ) R R R
∞
l l l l l2
n=1

∂w
= a∆2 w + Φ(x, y, t) 451
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),
∞
X
G1 (r, ξ, t) = J0 J0 exp − 2 ,
πR2 n=1
(k12 R2 + µ2n)J02 (µn) R R R
X∞
ϕm(z)ϕm (η) k2
G2 (z, η, t) = exp(−aλ2mt), ϕm(z) = cos(λmz) + sin(λmz),
kϕm k2 λm
m=1

2 k3 λ2m + k22 k2 l k22
kϕmk = 2 2 + + 1+ 2 ,
tan(λl) k2 + k3
µJ1 (µ) − k1 RJ0 (µ) = 0, = 2 .
λ λ − k2 k3

scribed:
Solution:
Z lZ R
0 0
Z tZ l
∂
0 0 ∂ξ ξ=R
Z tZ R
0 0
Z tZ R
0 0
Z tZ l Z R
0 0 0
Here,
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),
∞
G1 (r, ξ, t) = J0 J0 exp − 2 ,
πR2
n=1 1
J 2 (µn ) R R R
∞
l l n=1 l l l2
0 ≤ r ≤ R).
scribed:
Solution:
Z lZ R
0 0
Z tZ l
0 0
Z tZ R
∂
0 0 ∂η η=0
Z tZ R
∂
0 0 ∂η η=l
Z tZ lZ R
0 0 0
Here,
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),
∞
G1 (r, ξ, t) = + J0 J0 exp − 2 ,
πR2 πR2
n=1 0
J 2 (µn ) R R R
∞
l n=1 l l l2

prescribed:
∂w
= a∆2 w + Φ(x, y, t) 453
The solution to the first boundary value problem for a hollow circular cylinder of interior
radius R1 , exterior radius R2 , and length l is given by the formula of Section 4.1.3 (see the
first boundary value problem for R1 ≤ r ≤ R2 , 0 ≤ z ≤ l) with the term
Z tZ lZ R2
2π ξΦ(ξ, η, τ ) G(r, z, ξ, η, t − τ ) dξ dη dτ (1)
0 0 R1
added; this term takes into account the equation’s nonhomogeneity.
prescribed:
The solution to the second boundary value problem for a finite hollow circular cylinder
is given by the formula of Section 4.1.3 (see the second boundary value problem for R1 ≤
r ≤ R2 , 0 ≤ z ≤ l) with the additional term (1). The Green’s function is the same as in
Section 4.1.3.
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ z ≤ l. Third boundary value problem.
prescribed:

Solution:
Z lZ R2
0 R1
Z tZ l
0 0
Z tZ l
0 0
Z tZ R2
0 R1
Z t Z R2
0 R1
Z t Z l Z R2
+ 2π ξΦ(ξ, η, τ ) G(r, ξ, z, η, t − τ ) dξ dη dτ.
0 0 R1
Here, the Green’s function is given by
G(r, z, ξ, η, t) = G1 (r, ξ, t) G2 (z, η, t),

X∞
π λ2n 2
G1 (r, ξ, t) = k2 J0 (λn R2 ) − λn J1 (λn R2 ) Hn (r)Hn (ξ) exp(−λ2n at),
4 n=1 Bn
∞
X ϕm (z)ϕm (η)
G2 (z, η, t) = 2
exp −µ2m at ,
kϕm k
m=1
with
2 2
Bn = (λ2n +k22) k1J0(λnR1)+λnJ1(λnR1) −(λ2n +k12) k2J0(λnR2)−λnJ1(λnR2) ,

Hn(r) = k1Y0(λnR1)+λnY1(λnR1) J0(λnr)− k1J0(λnR1)+λnJ1(λnR1) Y0(λnr),
k4 µ2m +k32 k3 l 2
ϕm(z) = µm cos(µmz)+k3 sin(µmz), kϕmk2 = 2 2 + + µm +k32 ,
2 µm +k4 2 2
where J0 (λ), J1 (λ), Y0 (λ), and Y1 (λ) are Bessel functions, the λn are positive roots of the

k1 J0 (λR1 ) + λJ1 (λR1 ) k2 Y0 (λR2 ) − λY1 (λR2 )

− k2 J0 (λR2 ) − λJ1 (λR2 ) k1 Y0 (λR1 ) + λY1 (λR1 ) = 0,
and the µm are positive roots of the transcendental equation

tan µl k3 + k4
= 2 .
µ µ − k3 k4
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ z ≤ l. Mixed boundary value problem.

prescribed:

4.3. Other Equations 455
Solution:
Z lZ R2
0 R1
Z tZ l
0 0
Z tZ l
0 0
Z t Z R2 h ∂ i
0 R1 ∂η η=0
Z t Z R2 h ∂ i
0 R1 ∂η η=l
Z t Z l Z R2
+ 2π ξΦ(ξ, η, τ ) G(r, ξ, z, η, t − τ ) dξ dη dτ.
0 0 R1
Here,
∞ πmz πmη
2 X π 2 m2 at
G(r, ξ, z, η, t) = G1 (r, ξ, t) sin sin exp − ,
l l l l2
m=1
where the expression of G1 (r, ξ, t) is specified in the previous paragraph (for the third
boundary value problem).
◆ Section 5.2.2 presents solutions of other boundary value problems; a more general,
three-dimensional equation is discussed there.
4.3 Other Equations

4.3.1 Equations Containing Arbitrary Parameters

∂w ∂2w ∂2w
1. =a + + (bx + cy + k)w.
∂t ∂x2 ∂y 2
The transformation

w(x, y, t) = u(ξ, η, t) exp (bx + cy + k)t + 13 a(b2 + c2 )t3 ,
ξ = x + abt2 , η = y + act2

leads to the two-dimensional heat equation ∂t u = a ∂ξξ u + ∂ηη u .
See also Niederer (1973) and Boyer (1974).
2
∂w ∂ w ∂2w
2. =a 2
+ 2
− bx2 + by 2 + k w, b > 0.
∂t ∂x ∂y
The transformation (C is an arbitrary constant)
r
1 b 2 2
√
w(x, y, t) = u(ξ, η, τ ) exp (x + y ) + 2 ab − k t ,
2 a
√ √ 1 √
ξ = x exp 2 ab t , η = y exp 2 ab t , τ = √ exp 4 ab t + C
4 ab

leads to the two-dimensional heat equation ∂τ u = a ∂ξξ u + ∂ηη u .
2
∂w ∂ w ∂2w
3. =a 2
+ 2
+ bx2 + by 2 − k w, b > 0.
∂t ∂x ∂y
The transformation
√
1 b √ 2 2

w(x, y, t) = √ exp √ tan 2 ab t x + y − kt u(ξ, η, τ ),
cos 2 ab t 2 a
√ √
x y a
ξ= √ , η = √ , τ = √ tan 2 ab t
cos 2 ab t cos 2 ab t 2 b
leads to the two-dimensional heat equation ∂τ u = ∂ξξ u + ∂ηη u.
∂w ∂2w ∂ 2w
4. = + + ax−2 + by −2 w.
∂t ∂x2 ∂y 2
This is a special case of equation 4.3.2.7. Boyer (1976) showed that this equation admits
the separation of variables into 25 systems of coordinates for ab = 0 and 15 systems of
coordinates for ab 6= 0.
2
∂w ∂ w ∂2w
5. =a 2
+ 2
+ (btn x + ctm y + stk )w.
∂t ∂x ∂y
This is a special case of equation 4.3.2.2 with f (t) = btn , g(t) = ctm , and h(t) = stk .
2
∂w ∂ w ∂2w
6. =a 2
+ 2
+ −b(x2 + y 2 ) + c1 tn1 x + c2 tn2 y + stk w.
∂t ∂x ∂y
This is a special case of equation 4.3.2.3 with f (t) = c1 tn1 , g(t) = c2 tn2 , and h(t) = stk .
2
∂w ∂ w ∂2w ∂w ∂w
7. =a 2
+ 2
+ b1 + b2 + cw.
∂t ∂x ∂y ∂x ∂y
This equation describes an unsteady temperature (concentration) field in a medium moving
with a constant velocity, provided there is volume release (absorption) of heat proportional
to temperature.
The substitution

w(x, y, t) = exp A1 x + A2 y + Bt U (x, y, t),
b1 b2 b2 + b22
A1 = − , A2 = − , B = c − 1 ,
2a 2a 4a
leads to the two-dimensional heat equation ∂t U = a∆2 U that is considered in Section 4.1.1.
2
∂w ∂ w ∂2w ∂w ∂w
8. =a 2
+ 2
+ b1 + b2 + (c1 x + c2 y + k)w.
∂t ∂x ∂y ∂x ∂y
The transformation

w(x, y, t) = exp (c1 x + c2 y)t + 13 a(c21 + c22 )t3 + 12 (b1 c1 + b2 c2 )t2 + kt U (ξ, η, t),
ξ = x + ac1 t2 + b1 t, η = y + ac2 t2 + b2 t
leads to the two-dimensional heat equation ∂t U = a(∂ξξ U + ∂ηη U ) that is considered in
Section 4.1.1.

∂w ∂2w ∂2w ∂w ∂w
9. =a + + b1 tn1 + b2 tn2 + (c1 tm1 x + c2 tm2 y + stk )w.
∂t ∂x2 ∂y 2 ∂x ∂y
This is a special case of equation 4.3.2.5. The equation can be reduced to the two-dimen-
sional heat equation treated in Section 4.1.1.

∂w ~2 ∂2w ∂2w
10. i~ + + = 0.
∂t 2m ∂x2 ∂y 2
Two-dimensional Schrödinger equation, i2 = −1.

im im 2 2 π
E (x, y, t) = − exp (x + y ) − i .
2π~2 t 2~t 2
⊙ Literature: V. S. Vladimirov, V. P. Mikhailov, A. A. Vasharin et al. (1974).

∂w ∂2w ∂2w
1. =a + + f (t)w.
∂t ∂x2 ∂y 2
This equation describes two-dimensional thermal phenomena in quiescent media or solids
with constant thermal diffusivities in the case of unsteady volume heat release proportional
to temperature. R
The substitution w(x, y, t) = exp f (t) dt U (x, y, t) leads to the two-dimensional
heat equation ∂t U = a(∂xx U + ∂yy U ) treated in Section 4.1.1.

∂w ∂2w ∂2w
2. =a + + [xf (t) + yg(t) + h(t)]w.
∂t ∂x2 ∂y 2
The transformation
Z Z
2 2
w(x, y, t) = u(ξ, η, t) exp xF (t) + yG(t) + H(t) + a F (t) dt + a G (t) dt ,
Z Z
ξ = x + 2a F (t) dt, η = y + 2a G(t) dt,
where Z Z Z
F (t) = f (t) dt, G(t) = g(t) dt, H(t) = h(t) dt,

leads to the two-dimensional heat equation ∂t u = a ∂ξξ u + ∂ηη u .

∂w ∂2w ∂2w
3. =a + + [−b(x2 + y 2 ) + f (t)x + g(t)y + h(t)]w.
∂t ∂x2 ∂y 2
1◦ . Case b > 0. The transformation
r
1 b 2 2
w(x, y, t) = u(ξ, η, τ ) exp (x + y ) ,
2 a
√ √ 1 √
ξ = x exp 2 ab t , η = y exp 2 ab t , τ = √ exp 4 ab t
4 ab

2
∂u ∂ u ∂2u
=a 2
+ 2
+ F (τ )ξ + G(τ )η + H(τ ) u,
∂t ∂ξ ∂η

1 ln(cτ ) 1 ln(cτ )
F (τ ) = f , G(τ ) = g ,
(cτ )3/2 c (cτ )3/2 c

1 ln(cτ ) 1 √
H(τ ) = h + , c = 4 ab.
cτ c 2τ
2◦ . Case b < 0. The transformation

√
−b √ 2 2
w(x, y, t) = v(ξ̄, η̄, τ̄ ) exp √ tan 2 −ab t (x + y ) ,
2 a
x y 1 √
ξ¯ = √ , η̄ = √ , τ̄ = √ tan 2 −ab t
cos 2 −ab t cos 2 −ab t 2 −ab
also leads to an equation of the form 2.3.2.2 (the transformed equation is not written out
here).
∂w ∂ 2w ∂ 2w
4. = a1 (t) + a2 (t) + Φ(x, y, t).
∂t ∂x2 ∂y 2
This is a special case of equation 4.3.2.8. Let 0 < a1 (t) < ∞ and 0 < a2 (t) < ∞.
For the first, second, third, and mixed boundary value problems treated in rectangular,
finite, or infinite domains (x1 ≤ x ≤ x2 , y1 ≤ y ≤ y2 ), the Green’s function admits
incomplete separation of variables and can be represented in the product form (see Section
17.5.2)
G(x, y, ξ, η, t, τ ) = G1 (x, ξ, T1 ) G2 (y, η, T2 ),

Z t Z t
T1 = a1 (η) dη, T2 = a2 (η) dη,
τ τ
where G1 = G1 (x, ξ, t) is the auxiliary Green’s function that corresponds to the one-
dimensional heat equation with a1 (t) = 1, a2 (t) = 0, and Φ(x, y, t) = 0 and with homoge-
neous boundary conditions at x = x1 and x = x2 (the functions G1 for various boundary
value problems can be found in Sections 3.1.1 and 3.1.2). Similarly, G2 = G2 (y, η, t) is
the auxiliary Green’s function corresponding to the one-dimensional heat equation with
a1 (t) = 0, a2 (t) = 1, and Φ(x, y, t) = 0 and with homogeneous boundary conditions at
y = y1 and y = y2 . Note that the Green’s functions G1 and G2 are introduced for τ = 0.
See Section 17.1.1 for the solution of various one-dimensional boundary value problems
with the help of the Green’s function.
Example 4.15. Domain: −∞ < x < ∞, −∞ < y < ∞. Cauchy problem.
w = f (x, y) at t = 0.
Solution:
Z tZ ∞ Z ∞
w(x, y, t) = Φ(ξ, η, τ ) G(x, y, ξ, η, t, τ ) dξ dη dτ
0 −∞ −∞
Z ∞Z ∞
+ f (ξ, η) G(x, y, ξ, η, t, 0) dξ dη,
−∞ −∞
where
Z t Z t
1 (x−ξ)2 (y −η)2
G(x, y, ξ, η, t, τ ) = √ exp − − , T1 = a1(η) dη, T2 = a2(η) dη.
4π T1T2 4T1 4T2 τ τ
Example 4.16. Domain: 0 ≤ x < ∞, 0 ≤ y < ∞. Second boundary value problem.


Solution:
Z tZ ∞Z ∞
w(x, y, t) = Φ(ξ, η, τ ) G(x, y, ξ, η, t, τ ) dξ dη dτ
0 0 0
Z ∞ Z ∞
+ f (ξ, η) G(x, y, ξ, η, t, 0) dξ dη
0 0
Z tZ ∞
− a1 (τ )g1 (η, τ ) G(x, y, 0, η, t, τ ) dη dτ
0 0
Z tZ ∞
− a2 (τ )g2 (ξ, τ ) G(x, y, ξ, 0, t, τ ) dξ dτ,
0 0
where

Z t
1 (x − ξ)2 (x + ξ)2
G1 (x, ξ, T1 ) = √ exp − + exp − , T1 = a1 (η) dη,
2 πT1 4T1 4T1 τ
Z t
1 (y − η)2 (y + η)2
G2 (y, η, T2 ) = √ exp − + exp − , T2 = a2 (η) dη.
2 πT2 4T2 4T2 τ
Example 4.17. Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 . First boundary value problem.


w = h1 (x, t) at y = 0 (boundary condition),
w = h2 (x, t) at y = l2 (boundary condition).
Solution:
Z t Z l1Z l2
w(x, y, t) = Φ(ξ, η, τ ) G(x, y, ξ, η, t, τ ) dη dξ dτ
0 0 0
Z l1Z l2
+ f (ξ, η) G(x, y, ξ, η, t, 0) dη dξ
0 0
Z tZ l2
∂
+ a1 (τ )g1 (η, τ ) G(x, y, ξ, η, t, τ ) dη dτ
0 0 ∂ξ ξ=0
Z t Z l2
∂
− a1 (τ )g2 (η, τ ) G(x, y, ξ, η, t, τ ) dη dτ
0 0 ∂ξ ξ=l1
Z t Z l1
∂
+ a2 (τ )h1 (ξ, τ ) G(x, y, ξ, η, t, τ ) dξ dτ
0 0 ∂η η=0
Z t Z l1
∂
− a2 (τ )h2 (ξ, τ ) G(x, y, ξ, η, t, τ ) dξ dτ,
0 0 ∂η η=l2
where

X∞ Z t
2 nπx nπξ n2 π 2 T 1
G1 (x, ξ, T1 ) = sin sin exp − , T 1 = a1 (η) dη,
l1 n=1 l1 l1 l1 2 τ
∞ Z t
2 X nπy nπη n2 π 2 T 2
G2 (y, η, T2 ) = sin sin exp − , T2 = a2 (η) dη.
l2 n=1 l2 l2 l2 2 τ
∂w ∂ 2w ∂ 2w ∂w
5. = a1 (t) + a2 (t) + [b1 (t)x + c1 (t)]
∂t ∂x2 ∂y 2 ∂x
∂w
+ [b2 (t)y + c2 (t)] + [s1 (t)x + s2 (t)y + p(t)]w.
∂y
The transformation

w(x, y, t) = exp f1 (t)x+f2 (t)y+g(t) u(ξ, η, t), ξ = h1 (t)x+r1 (t), η = h2 (t)y+r2 (t),
where
Z
hk (t) = Ak exp bk (t) dt ,
Z
sk (t)
fk (t) = hk (t) dt + Bk hk (t),
hk (t)
Z

rk (t) = 2ak (t)fk (t) + ck (t) hk (t) dt + Ck ,
Z

g(t) = a1 (t)f12 (t) + a2 (t)f22 (t) + c1 (t)f1 (t) + c2 (t)f2 (t) + p(t) dt + D,
(k = 1, 2; Ak , Bk , Ck , and D are arbitrary constants), leads to an equation of the form

4.3.2.4:
∂u ∂2u ∂2u
= a1 (t)h21 (t) 2 + a2 (t)h22 (t) 2 .
∂t ∂ξ ∂η
∂w ∂ 2w ∂ 2w ∂w ∂w
6. = a1 (t) + a2 (t) + [b1 (t)x + c1 (t)] + [b2 (t)y + c2 (t)]
∂t ∂x2 ∂y 2 ∂x ∂y
+ [s1 (t)x2 + s2 (t)y 2 + p1 (t)x + p2 (t)y + q(t)]w.
The substitution
w(x, y, t) = exp f1 (t)x2 + f2 (t)y 2 u(x, y, t),
where the functions f1 = f1 (t) and f2 = f2 (t) are solutions of the Riccati equations
f1′ = 4a1 (t)f12 + 2b1 (t)f1 + s1 (t),

f2′ = 4a2 (t)f22 + 2b2 (t)f2 + s2 (t),
leads to an equation of the form 4.3.2.5 for u = u(x, y, t).
∂w ∂ 2w ∂ 2w ∂w ∂w
7. = a1 (x) + a2 (y) + b1 (x) + b2 (y)
∂t ∂x2 ∂y 2 ∂x ∂y
+ [c1 (x) + c2 (y)]w + Φ(x, y, t).
Domain: x1 ≤ x ≤ x2 , y1 ≤ y ≤ y2 . Different boundary value problems:

s1 ∂x w − k1 w = g1 (y, t) at x = x1 (boundary condition),
s2 ∂x w + k2 w = g2 (y, t) at x = x2 (boundary condition),
s3 ∂y w − k3 w = g3 (x, t) at y = y1 (boundary condition),
s4 ∂y w + k4 w = g4 (x, t) at y = y2 (boundary condition).
By choosing appropriate parameters sn and kn (n = 1, 2, 3, 4), one obtains the first, sec-
ond, third, or mixed boundary value problem. If the domain is infinite, say, x2 = ∞, then
the corresponding boundary condition should be omitted; this is also true for x1 = −∞,
y1 = −∞, or y2 = ∞.
The Green’s function admits incomplete separation of variables; specifically, it can be
represented in the product form (see Section 17.5.2)
G(x, y, ξ, η, t) = G1 (x, ξ, t) G2 (y, η, t).
Here G1 = G1 (x, ξ, t) and G2 = G2 (y, η, t) are auxiliary Green’s functions determined

by solving the following simpler one-dimensional problems with homogeneous boundary
conditions:
∂G1 ∂ 2G1 ∂G1 ∂G2 ∂ 2G2 ∂G2
= a1(x) 2
+ b1(x) + c1(x) G1 , = a2(y) 2
+ b2(y) + c2(y) G2,
∂t ∂x ∂x ∂t ∂y ∂y
G1 = δ(x − ξ) at t = 0, G2 = δ(y − η) at t = 0,
s1∂xG1 − k1G1 = 0 at x = x1, s3∂y G2 − k3G2 = 0 at y = y1,
s2∂xG1 + k2G1 = 0 at x = x2, s4∂y G2 + k4G2 = 0 at y = y2,
where ξ and η are free parameters and δ(x) is the Dirac delta function.
The equation for G1 coincides with equation 3.8.6.5, which can be reduced to the equa-
tion in Section 3.8.9 (where the expression for the Green’s function can also be found). In
the general case, the equation for G2 differs from that for G1 only in notation.
∂w ∂ 2w ∂ 2w ∂w ∂w
8. = a1 (x, t) + a2 (y, t) + b1 (x, t) + b2 (y, t)
∂t ∂x2 ∂y 2 ∂x ∂y
+ [c1 (x, t) + c2 (y, t)]w + Φ(x, y, t).
Suppose that this equation is supplemented with the same initial and boundary conditions as
equation 4.3.2.7. Then the Green’s function for this problem admits incomplete separation
of variables and can be represented in the product form (see Section 17.5.2)
G(x, y, ξ, η, t, τ ) = G1 (x, ξ, t, τ ) G2 (y, η, t, τ ).
Here G1 = G1 (x, ξ, t, τ ) and G2 = G2 (y, η, t, τ ), t ≥ τ , are auxiliary Green’s functions

determined by solving the following simpler boundary value problems with homogeneous
boundary conditions:
∂G1 ∂ 2 G1 ∂G1
= a1 (x, t) + b1 (x, t) + c1 (x, t) G1 ,
∂t ∂x2 ∂x
G1 = δ(x − ξ) at t = τ,
s1 ∂x G1 − k1 G1 = 0 at x = x1 ,
s2 ∂x G1 + k2 G1 = 0 at x = x2 ,
∂G2 ∂ 2 G2 ∂G2
= a2 (y, t) 2
+ b2 (y, t) + c2 (y, t) G2 ,
∂t ∂y ∂y
G2 = δ(y − η) at t = τ,
s3 ∂y G2 − k3 G2 = 0 at y = y1 ,
s4 ∂y G2 + k4 G2 = 0 at y = y2 ,
where ξ, η, and τ are free parameters, and δ(x) is the Dirac delta function.
See Section 17.1.1 for the solution of various one-dimensional boundary value problems
with the help of the Green’s function.
Chapter 5
Second-Order
Parabolic Equations with
Three or More Space Variables
∂w
5.1 Heat Equation = a∆3 w
∂t
The three-dimensional sourceless heat equation in the rectangular Cartesian system of co-
ordinates has the form 2
∂w ∂ w ∂2w ∂2w
=a + + .
∂t ∂x2 ∂y 2 ∂z 2
It governs three-dimensional thermal phenomena in quiescent media or solids with constant
thermal diffusivity. A similar equation is used to study the corresponding three-dimensional
unsteady mass-exchange processes with constant diffusivity.
w(x, y, z, t) = Ax2 + By 2 + Cz 2 + 2a(A + B + C)t,
w(x, y, z, t) = A(x2 + 2at)(y 2 + 2at)(z 2 + 2at) + B,

w(x, y, z, t) = A exp k1x + k2 y + k3z + (k12 + k22 + k32)at + B,

w(x, y, z, t) = A cos(k1 x + C1) cos(k2 y + C2) cos(k3 z + C3) exp −(k12 + k22 + k32)at ,

w(x, y, z, t) = A cos(k1 x + C1) cos(k2 y + C2) sinh(k3 z + C3 ) exp −(k12 + k22 − k32 )at ,

w(x, y, z, t) = A cos(k1 x + C1) cos(k2 y + C2) cosh(k3 z + C3) exp −(k12 + k22 − k32 )at ,
w(x, y, z, t) = Ae−k1x−k2y−k3z cos(k1 x − 2ak12 t) cos(k2 y − 2ak22 t) cos(k3 z − 2ak32 t),

A (x − x0 )2 + (y − y0)2 + (z − z0 )2
w(x, y, z, t) = exp − ,
(t − t0)3/2 4a(t − t0 )

x − x0 y − y0 z − z0
w(x, y, z, t) = A erf √ erf √ erf √ + B,
2 at 2 at 2 at
where A, B, C, C1 , C2 , C3 , k1 , k2 , k3 , x0 , y0 , z0 , and t0 are arbitrary constants.
463
464 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES

1 x2 + y 2 + z 2
E (x, y, z, t) = exp − .
8(πat)3/2 4at
◮ Formulas to construct particular solutions. Remarks on the Green’s functions.

1◦ . Apart from usual solutions with separated variables,
w(x, y, z, t) = f1 (x)f2 (y)f3 (z)f4 (t),
the equation in question admits more sophisticated solutions in the product form
w(x, y, z, t) = u1 (x, t)u2 (y, t)u3 (z, t),
where the functions u1 = u1 (x, t), u2 = u2 (y, t), and u3 = u3 (y, t) are solutions of the
one-dimensional heat equations
∂u1 ∂ 2 u1 ∂u2 ∂ 2 u2 ∂u3 ∂ 2 u3
=a , =a , =a 2 ,
∂t ∂x2 ∂t ∂y 2 ∂t ∂z
treated in Section 3.1.1.
2◦ . Suppose w = w(x, y, z, t) is a solution of the three-dimensional heat equation. Then
the functions
w1 =Aw(±λx+C1, ±λy+C2, ±λz+C3, λ2t+C4),

w2 =A exp λ1x+λ2y+λ3z+(λ21+λ22+λ23)at w(x+2aλ1t, y+2aλ2t, z+2aλ3t, t),

A β(x2+y 2+z 2) x y z γ+λt
w3 = exp − w , , , , λδ−βγ =1,
|δ+βt|3/2 4a(δ+βt) δ+βt δ+βt δ+βt δ+βt
where A, C1 , C2 , C3 , C4 , λ, λ1 , λ2 , λ3 , β, and δ are arbitrary constants, are also solutions
of this equation. The signs at λ in the formula for w1 can be taken independently of one
another.
3◦ . For the three-dimensional boundary value problems considered in Section 5.1.1, the
Green’s function can be represented in the product form
G(x, y, z, ξ, η, ζ, t) = G1 (x, ξ, t) G2 (y, η, t) G3 (z, ζ, t),
where G1 (x, ξ, t), G2 (y, η, t), G3 (z, ζ, t) are the Green’s functions of the corresponding
one-dimensional boundary value problems; these functions can be found in Sections 3.1.1
and 3.1.2.
Example 5.1. The Green’s function of the mixed boundary value problem for a semiinfinite
layer (−∞ < x < ∞, 0 ≤ y < ∞, 0 ≤ z < l) with the initial and boundary conditions
w = f (x, y, z) at t = 0,
w = g1 (x, z, t) at y = 0, ∂z w = g2 (x, y, t) at z = 0, ∂z w = g3 (x, y, t) at z = l,
is the product of three one-dimensional Green’s functions from Section 3.1.2 (in which one needs
to carry out obvious renaming of variables): (i) for the Cauchy problem with −∞ < x < ∞, (ii) for
the first boundary value problem with 0 ≤ y < ∞, and (iii) for the second boundary value problem
with 0 ≤ z < l.
∂w
= a∆3 w 465
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, −∞ < z < ∞. Cauchy problem.
w = f (x, y, z) at t = 0.
Solution:
Z ∞ Z ∞ Z ∞
1 (x−ξ)2 +(y −η)2 +(z −ζ)2
w(x, y, z, t) = f (ξ, η, ζ) exp − dξ dη dζ.
8(πat)3/2 −∞ −∞ −∞ 4at
Example 5.2. The initial temperature is constant and is equal to w1 in the domain |x| < x0 ,
|y| < y0 , |z| < z0 and is equal to w2 in the domain |x| > x0 , |y| > y0 , |z| > z0 ; specifically,
(
w1 for |x| < x0 , |y| < y0 , |z| < z0 ,
f (x, y, z) =
w2 for |x| > x0 , |y| > y0 , |z| > z0 .
Solution:

1 x0 − x x0 + x
w= (w1 − w2 ) erf √ + erf √
8 2 at 2 at

y0 − y y0 + y z0 − z z0 + z
× erf √ + erf √ erf √ + erf √ + w2 .
2 at 2 at 2 at 2 at
◮ Domain: 0 ≤ x < ∞, −∞ < y < ∞, −∞ < z < ∞. First boundary value

problem.
A half-space is considered. The following conditions are prescribed:
w = f (x, y, z) at t = 0 (initial condition),

w = g(y, z, t) at x = 0 (boundary condition).
Solution:
Z ∞Z ∞Z ∞
w(x, y, z, t) = f (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
−∞ −∞ 0
Z tZ ∞ Z ∞
∂
+a g(η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dη dζ dτ,
0 −∞ −∞ ∂ξ ξ=0
where

1 (x−ξ)2 (x+ξ)2 (y−η)2+(z−ζ)2
G(x, y, z, ξ, η, ζ, t)= exp − −exp − exp − .
8(πat)3/2 4at 4at 4at

◮ Domain: 0 ≤ x < ∞, −∞ < y < ∞, −∞ < z < ∞. Second boundary value

problem.

∂x w = g(y, z, t) at x = 0 (boundary condition).
Solution:
Z ∞Z ∞Z ∞
−∞ −∞ 0
Z tZ ∞ Z ∞
−a g(η, ζ, τ ) G(x, y, z, 0, η, ζ, t − τ ) dη dζ dτ,
0 −∞ −∞
where

1 (x−ξ)2 (x+ξ)2 (y−η)2+(z−ζ)2
G(x, y, z, ξ, η, ζ, t)= exp − +exp − exp − .
◮ Domain: 0 ≤ x < ∞, −∞ < y < ∞, −∞ < z < ∞. Third boundary value

problem.

∂x w − kw = g(y, z, t) at x = 0 (boundary condition).
The solution w(x, y, z, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where

1 (y−η)2+(z−ζ)2 (x−ξ)2 (x+ξ)2
G(x, y, z, ξ, η, ζ, t)= exp − exp − +exp −

√ 2 x+ξ √
−2k πat exp k at+k(x+ξ) erfc √ +k at .
2 at
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, 0 ≤ z ≤ l. First boundary value

problem.
An infinite layer is considered. The following conditions are prescribed:

w = g1 (x, y, t) at z = 0 (boundary condition),
w = g2 (x, y, t) at z = l (boundary condition).
∂w
= a∆3 w 467
Solution:
Z lZ ∞Z ∞
0 −∞ −∞
Z tZ ∞ Z ∞
∂
+a g1 (ξ, η, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dτ
0 −∞ −∞ ∂ζ ζ=0
Z tZ ∞ Z ∞
∂
−a g2 (ξ, η, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dτ,
0 −∞ −∞ ∂ζ ζ=l
where
∞
1 (x − ξ)2 + (y − η)2 X nπz nπζ n2π 2at
G(x, y, z, ξ, η, ζ, t) = exp − sin sin exp − ,
2πalt 4at n=1
l l l2
or

1 (x − ξ)2 + (y − η)2
G(x, y, z, ξ, η, ζ, t) = exp −
8(πat)3/ 2 4at
X∞
(2nl + z − ζ)2 (2nl + z + ζ)2
× exp − − exp − .
n=−∞
4at 4at
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, 0 ≤ z ≤ l. Second boundary value

problem.

∂z w = g1 (x, y, t) at z = 0 (boundary condition),
∂z w = g2 (x, y, t) at z = l (boundary condition).
Solution:
Z lZ ∞Z ∞
0 −∞ −∞
Z tZ ∞ Z ∞
−a g1 (ξ, η, τ ) G(x, y, z, ξ, η, 0, t − τ ) dξ dη dτ
0 −∞ −∞
Z tZ ∞ Z ∞
+a g2 (ξ, η, τ ) G(x, y, z, ξ, η, l, t − τ ) dξ dη dτ,
0 −∞ −∞
where

1 (x − ξ)2 + (y − η)2
G(x, y, z, ξ, η, ζ, t) = exp −
4πalt 4at
X∞
nπz nπζ n2 π 2 at
× 1+2 cos cos exp − ,
n=1
l l l2
or

1 (x − ξ)2 + (y − η)2
G(x, y, z, ξ, η, ζ, t) = √ exp −
(2 πat )3 4at
∞
X
(z − ζ + 2nl)2 (z + ζ + 2nl)2
× exp − + exp − .
n=−∞
4at 4at
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, 0 ≤ z ≤ l. Third boundary value

problem.

∂z w − k1 w = g1 (x, y, t) at z = 0 (boundary condition),
∂z w + k2 w = g2 (x, y, t) at z = l (boundary condition).
∞
1 (x − ξ)2 + (y − η)2 X ϕn (z)ϕn (ζ)
G(x, y, z, ξ, η, ζ, t) = exp − 2
exp(−aµ2n t),
4πat 4at n=1
kϕ n k
2 2
k1 2 k2 µn + k1 k1 l k12
ϕn (z) = cos(µn z) + sin(µn z), kϕn k = + 2 + 1+ 2 .
tan(µl) k1 + k2
µ µ − k1 k2
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, 0 ≤ z ≤ l. Mixed boundary value

problem.

Solution:
Z lZ ∞Z ∞
0 −∞ −∞
Z tZ ∞ Z ∞
∂
0 −∞ −∞ ∂ζ ζ=0
Z tZ ∞ Z ∞
0 −∞ −∞
∂w
= a∆3 w 469
where

1 (x − ξ)2 + (y − η)2
G(x, y, z, ξ, η, ζ, t) = exp −
2πalt 4at
∞
X
(2n + 1)πz (2n + 1)πζ (2n + 1)2 π 2at
× sin sin exp − ,
2l 2l 4l2
n=0
or

1 (x−ξ)2 +(y−η)2
G(x, y, z, ξ, η, ζ, t) = √ exp −
(2 πat )3 4at
∞
X
n (z−ζ +2nl)2 (z+ζ +2nl)2
× (−1) exp − −exp − .
n=−∞
4at 4at
◮ Domain: −∞ < x < ∞, 0 ≤ y < ∞, 0 ≤ z ≤ l. First boundary value problem.
A semiinfinite layer is considered. The following conditions are prescribed:

w = g1 (x, z, t) at y = 0 (boundary condition),
Solution:
Z lZ ∞Z ∞
0 0 −∞
Z tZ lZ ∞
∂
+a g1 (ξ, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 −∞ ∂η η=0
Z t Z ∞Z ∞
∂
0 0 −∞ ∂ζ ζ=0
Z t Z ∞Z ∞
∂
0 0 −∞ ∂ζ ζ=l
where

1 (x − ξ)2 (y − η)2 (y + η)2
G(x, y, z, ξ, η, ζ, t) = exp − exp − − exp −
2πalt 4at 4at 4at
∞
X 2 2
nπz nπζ n π at
l l l2
n=1
◮ Domain: −∞ < x < ∞, 0 ≤ y < ∞, 0 ≤ z ≤ l. Second boundary value problem.

∂y w = g1 (x, z, t) at y = 0 (boundary condition),
Solution:
Z lZ ∞Z ∞
0 0 −∞
Z tZ l Z ∞
−a g1 (ξ, ζ, τ ) G(x, y, z, ξ, 0, ζ, t − τ ) dξ dζ dτ
0 0 −∞
Z tZ ∞Z ∞
0 0 −∞
Z tZ ∞Z ∞
0 0 −∞
where

1 (x − ξ)2 (y − η)2 (y + η)2
G(x, y, z, ξ, η, ζ, t) = exp − exp − + exp −
4πalt 4at 4at 4at
∞
X
nπz nπζ n2π 2at
× 1+2 cos cos exp − .
l l l2
n=1
◮ Domain: −∞ < x < ∞, 0 ≤ y < ∞, 0 ≤ z ≤ l. Third boundary value problem.

∂y w − k1 w = g1 (x, z, t) at y = 0 (boundary condition),
∂z w + k3 w = g3 (x, y, t) at z = l (boundary condition).
X∞
1 (x−ξ)2 ϕn(z)ϕn(ζ)
G(x, y, z, ξ, η, ζ, t)= exp − H(y, η, t) 2
exp(−aµ2nt),
4πat 4at n=1
kϕ n k
2
2
Z ∞
(y−η) (y+η) (y+η+s)2
H(y, η, t)=exp − +exp − −2k1 exp − −k1s ds.
4at 4at 0 4at
Here,

k2 2 k3 µ2n + k22 k2 l k22
ϕn (z) = cos(µn z) + sin(µn z), kϕn k = + 2 + 1+ 2 ,
∂w
= a∆3 w 471
tan(µl) k2 + k3
with the µn being positive roots of the transcendental equation = 2 .
µ µ − k2 k3
◮ Domain: −∞ < x < ∞, 0 ≤ y < ∞, 0 ≤ z ≤ l. Mixed boundary value problems.

1◦ . A semiinfinite layer is considered. The following conditions are prescribed:
Solution:
Z lZ ∞Z ∞
0 0 −∞
Z tZ l Z ∞
∂
0 0 −∞ ∂η η=0
Z t Z ∞Z ∞
0 0 −∞
Z tZ ∞Z ∞
0 0 −∞
where

1 (x − ξ)2 (y − η)2 (y + η)2
G(x, y, z, ξ, η, ζ, t) = exp − exp − − exp −
4πalt 4at 4at 4at
∞
X
nπz nπζ n2π 2at
× 1+2 cos cos exp − .
l l l2
n=1
2◦ . A semiinfinite layer is considered. The following conditions are prescribed:

Solution:
Z lZ ∞Z ∞
0 0 −∞
Z tZ lZ ∞
0 0 −∞
Z tZ ∞Z ∞
∂
0 0 −∞ ∂ζ ζ=0
Z t Z ∞Z ∞
∂
0 0 −∞ ∂ζ ζ=l
where

1 (x − ξ)2 (y − η)2 (y + η)2
G(x, y, z, ξ, η, ζ, t) = exp − exp − + exp −
2πalt 4at 4at 4at
X∞
nπz nπζ n2π 2 at
n=1
l l l2
◮ Domain: 0 ≤ x < ∞, 0 ≤ y < ∞, 0 ≤ z < ∞. First boundary value problem.

An octant is considered. The following conditions are prescribed:
w = g1 (y, z, t) at x = 0 (boundary condition),
w = g3 (x, y, t) at z = 0 (boundary condition).
Solution:
Z ∞Z ∞Z ∞
w(x, y, z, t) = G(x, y, z, ξ, η, ζ, t)f (ξ, η, ζ) dξ dη dζ
0 0 0
Z tZ ∞ ∞Z
∂
+a g1 (η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=0
Z t Z ∞Z ∞
∂
0 0 0 ∂η η=0
Z t Z ∞Z ∞
∂
+a g3 (ξ, η, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dτ,
0 0 0 ∂ζ ζ=0
where
1
G(x, y, z, ξ, η, ζ, t) = √ 3 H(x, ξ, t)H(y, η, t)H(z, ζ, t),
2 πat

(x − ξ)2 (x + ξ)2
H(x, ξ, t) = exp − − exp − .
4at 4at
Example 5.3. The initial temperature is uniform, f (x, y, z) = w0 . The faces are maintained at
zero temperature, g1 = g2 = g3 = 0.
Solution:
x y z
w = w0 erf √ erf √ erf √ .
2 at 2 at 2 at
◮ Domain: 0 ≤ x < ∞, 0 ≤ y < ∞, 0 ≤ z < ∞. Second boundary value problem.

∂x w = g1 (y, z, t) at x = 0 (boundary condition),
∂z w = g3 (x, y, t) at z = 0 (boundary condition).
∂w
= a∆3 w 473
Solution:
Z ∞Z ∞Z ∞
0 0 0
Z tZ ∞Z ∞
−a g1 (η, ζ, τ ) G(x, y, z, 0, η, ζ, t − τ ) dη dζ dτ
0 0 0
Z tZ ∞Z ∞
0 0 0
Z tZ ∞Z ∞
−a g3 (ξ, η, τ ) G(x, y, z, ξ, η, 0, t − τ ) dξ dη dτ,
0 0 0
where
1
G(x, y, z, ξ, η, ζ, t) = √ 3 H(x, ξ, t)H(y, η, t)H(z, ζ, t),
2 πat

(x − ξ)2 (x + ξ)2
H(x, ξ, t) = exp − + exp − .
4at 4at
◮ Domain: 0 ≤ x < ∞, 0 ≤ y < ∞, 0 ≤ z < ∞. Third boundary value problem.

∂x w − k1 w = g1 (y, z, t) at x = 0 (boundary condition),
∂z w − k3 w = g3 (x, y, t) at z = 0 (boundary condition).
1
G(x, y, z, ξ, η, ζ, t) = √ 3 H(x, ξ, t; k1 )H(y, η, t; k2 )H(z, ζ, t; k3 ),
2 πat

(x − ξ)2 (x + ξ)2
H(x, ξ, t; k) = exp − + exp −
4at 4at

√ 2 x+ξ √
− 2k πat exp ak t + k(x + ξ) erfc √ + k at .
2 at
Example 5.4. The initial temperature is uniform, f (x, y, z) = w0 . The temperature of the
contacting media is zero, g1 = g2 = g3 = 0.
Solution:

x x √
w = w0 erf √ + exp(k1 x + k12 at) erfc √ + k1 at
2 at 2 at

y y √
× erf √ + exp(k2 y + k22 at) erfc √ + k2 at
2 at 2 at

z z √
× erf √ + exp(k3 z + k32 at) erfc √ + k3 at .
2 at 2 at
◮ Domain: 0 ≤ x < ∞, 0 ≤ y < ∞, 0 ≤ z < ∞. Mixed boundary value problems.
1◦ . An octant is considered. The following conditions are prescribed:

Solution:
Z ∞Z ∞Z ∞
0 0 0
Z t Z ∞Z ∞
∂
0 0 0 ∂ξ ξ=0
Z t Z ∞Z ∞
0 0 0
Z t Z ∞Z ∞
0 0 0
where

1 (x − ξ)2 (x + ξ)2
G(x, y, z, ξ, η, ζ, t) = √ 3 exp − − exp − H(y, η, t)H(z, ζ, t),
2 πat 4at 4at

(y − η)2 (y + η)2
H(y, η, t) = exp − + exp − .
4at 4at
2◦ . An octant is considered. The following conditions are prescribed:

Solution:
Z ∞Z ∞Z ∞
0 0 0
Z t Z ∞Z ∞
∂
0 0 0 ∂ξ ξ=0
Z t Z ∞Z ∞
∂
0 0 0 ∂η η=0
Z t Z ∞Z ∞
0 0 0
∂w
= a∆3 w 475
where

1 (z − ζ)2 (z + ζ)2
G(x, y, z, ξ, η, ζ, t) = √ 3 H(x, ξ, t)H(y, η, t) exp − + exp − ,
2 πat 4at 4at

(x − ξ)2 (x + ξ)2
H(x, ξ, t) = exp − − exp − .
4at 4at
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , −∞ < z < ∞. First boundary value problem.
An infinite cylindrical domain of a rectangular cross-section is considered. The following

conditions are prescribed:

w = g2 (y, z, t) at x = l1 (boundary condition),
w = g4 (x, z, t) at y = l2 (boundary condition).
Solution:
Z ∞ Z l2Z l1
−∞ 0 0
Z t Z ∞ Z l2
∂
0 −∞ 0 ∂ξ ξ=0
Z t Z ∞ Z l2
∂
−a g2 (η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 −∞ 0 ∂ξ ξ=l1
Z t Z ∞ Z l1
∂
0 −∞ 0 ∂η η=0
Z t Z ∞ Z l1
∂
−a g4 (ξ, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dζ dτ,
0 −∞ 0 ∂η η=l2
where

1 (z − ζ)2
G(x, y, z, ξ, η, ζ, t) = √ exp − H1 (x, ξ, t)H2 (y, η, t),
2 πat 4at
∞
2 X πnx πnξ π 2 n2 at
H1 (x, ξ, t) = sin sin exp − ,
l1 n=1 l1 l1 l12
∞
2 X πny πnη π 2 n2 at
H2 (y, η, t) = sin sin exp − .
l2
n=1
l2 l2 l22
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , −∞ < z < ∞. Second boundary value

problem.

∂x w = g2 (y, z, t) at x = l1 (boundary condition),
∂y w = g4 (x, z, t) at y = l2 (boundary condition).
Solution:
Z ∞ Z l2Z l1
−∞ 0 0
Z t Z ∞ Z l2
0 −∞ 0
Z t Z ∞ Z l2
+a g2 (η, ζ, τ ) G(x, y, z, l1 , η, ζ, t − τ ) dη dζ dτ
0 −∞ 0
Z t Z ∞ Z l1
0 −∞ 0
Z t Z ∞ Z l1
+a g4 (ξ, ζ, τ ) G(x, y, z, ξ, l2 , ζ, t − τ ) dξ dζ dτ,
0 −∞ 0
where

1 (z − ζ)2
G(x, y, z, ξ, η, ζ, t) = √ exp − H1 (x, ξ, t)H2 (y, η, t),
2 πat 4at
X∞
1 πnx πnξ π 2 n2 at
H1 (x, ξ, t) = 1+2 cos cos exp − 2 ,
l1 n=1
l1 l1 l1
X∞
1 πny πnη π 2 n2 at
H2 (y, η, t) = 1+2 cos cos exp − .
l2 l2
n=1
l2 l22
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , −∞ < z < ∞. Third boundary value problem.


∂x w + k2 w = g2 (y, z, t) at x = l1 (boundary condition),
∂y w + k4 w = g4 (x, z, t) at y = l2 (boundary condition).
∂w
= a∆3 w 477

1 (z−ζ)2
G(x, y, z, ξ, η, ζ, t)= √ exp − H1(x, ξ, t)H2(y, η, t),
2 πat 4at
X∞ ∞
X
ϕn(x)ϕn(ξ) 2 ψm(y)ψm(η)
H1(x, ξ, t)= 2
exp(−aµnt), H2(y, η, t)= exp(−aλ2mt).
kϕnk kψmk2
n=1 m=1
Here,

k1 k2 µ2n +k12 k12 l1 k12
ϕn(x) = cos(µnx)+ sin(µnx),kϕnk = 2 2 + + 1+ 2 ,
2 2
k3 2 k4 λm +k3 k3 l2 k32
ψm(y) = cos(λmy)+ sin(λmy), kψmk = 2 2 + + 1+ 2 ;
the µn and λm are positive roots of the transcendental equations
= 2 , = 2 .
µ µ − k1 k2 λ λ − k3 k4
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , −∞ < z < ∞. Mixed boundary value problem.


∂y w = g4 (x, z, t) at y = l2 (boundary condition).
Solution:
Z ∞ Z l2Z l1
−∞ 0 0
Z tZ ∞ Z l2
∂
0 −∞ 0 ∂ξ ξ=0
Z t Z ∞ Z l2
∂
0 −∞ 0 ∂ξ ξ=l1
Z t Z ∞ Z l1
0 −∞ 0
Z tZ ∞ Z l1
+a g4 (ξ, ζ, τ ) G(x, y, z, ξ, l2 , ζ, t − τ ) dξ dζ dτ,
0 −∞ 0
where
∞
2 (z−ζ)2 X πnx πnξ π 2n2at
G(x, y, z, ξ, η, ζ, t)= √ exp − sin sin exp −
l1l2 πat 4at n=1
l1 l1 l12
∞
1 X πmx πmξ π 2m2at
× + cos cos exp − .
2 m=1 l2 l2 l22
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z < ∞. First boundary value problem.
A semiinfinite cylindrical domain of a rectangular cross-section is considered. The follow-

ing conditions are prescribed:

w = g4 (x, z, t) at y = l2 (boundary condition),
Solution:
Z ∞Z l2Z l1
0
Z t Z0 0
∞Z l2
∂
0 0 0 ∂ξ ξ=0
Z t Z ∞Z l2
∂
0 0 0 ∂ξ ξ=l1
Z t Z ∞Z l1
∂
0 0 0 ∂η η=0
Z t Z ∞Z l1
∂
−a g4 (ξ, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ∂η η=l2
Z t Z l2Z l1
∂
0 0 0 ∂ζ ζ=0
where
G(x, y, z, ξ, η, ζ, t) = G1 (x, ξ, t; l1 ) G1 (y, η, t; l2 ) G2 (z, ζ, t),

∞
2X πnx πnξ π 2 n2 at
G1 (x, ξ, t; l) = sin sin exp − ,
l n=1 l l l2

1 (z − ζ)2 (z + ζ)2
G2 (z, ζ, t) = √ exp − − exp − .
2 πat 4at 4at
∂w
= a∆3 w 479
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z < ∞. Second boundary value problem.


∂y w = g4 (x, z, t) at y = l2 (boundary condition),
Solution:
Z ∞Z l2Z l1
0 0 0
Z tZ ∞Z l2
0 0 0
Z tZ ∞Z l2
0 0 0
Z tZ ∞Z l1
0 0 0
Z tZ ∞Z l1
+a g4 (ξ, ζ, τ ) G(x, y, z, ξ, l2 , ζ, t − τ ) dξ dζ dτ
0 0 0
Z t Z l2Z l1
0 0 0
where

X∞
G1 (x, ξ, t) = 1+2 cos cos exp − ,
l1 n=1
l1 l1 l12
X∞
G2 (y, η, t) = 1+2 cos cos exp − 2 ,
l2 l2 l2 l 2
n=1

1 (z − ζ)2 (z + ζ)2
G3 (z, ζ, t) = √ exp − + exp − .
2 πat 4at 4at
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z < ∞. Third boundary value problem.

∂y w + k4 w = g4 (x, z, t) at y = l2 (boundary condition),
∂z w − k5 w = g5 (x, y, t) at z = 0 (boundary condition).
G(x, y, z, ξ, η, ζ, t) = H1 (x, ξ, t)H2 (y, η, t)H3 (z, ζ, t),
∞
X ϕn (x)ϕn (ξ) k1
H1(x, ξ, t) = 2
exp(−aµ2nt), ϕn(x) = cos(µnx) + sin(µnx),
kϕn k µn
n=1
X∞
ψm(y)ψm(η) k3
H2(y, η, t) = 2
exp(−aλ2mt), ψm(y) = cos(λmy) + sin(λmy),
kψmk λm
m=1

1 (z − ζ)2 (z + ζ)2
H3(z, ζ, t) = √ exp − + exp −
2 πat 4at 4at

2 z +ζ √
− k5 exp k5 at + k5 (z + ζ) erfc √ + k5 at .
2 at
Here,

k2 µ2n+k12 k1 l1 k12 k4 λ2m+k32 k3 l2 k32
kϕnk2 = + + 1+ , kψm k2
= + + 1+ ,
2µ2n µ2n+k22 2µ2n 2 µ2n 2λ2m λ2m+k42 2λ2m 2 λ2m
= 2 , = 2 .
µ µ − k1 k2 λ λ − k3 k4
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z < ∞. Mixed boundary value problems.

1◦ . A semiinfinite cylindrical domain of a rectangular cross-section is considered. The
following conditions are prescribed:
∂w
= a∆3 w 481
Solution:
Z ∞Z l2Z l1
0 0 0
Z tZ ∞Z l2
∂
0 0 0 ∂ξ ξ=0
Z t Z ∞Z l2
∂
0 0 0 ∂ξ ξ=l1
Z t Z ∞Z l1
∂
0 0 0 ∂η η=0
Z t Z ∞Z l1
∂
0 0 0 ∂η η=l2
Z t Z l2Z l1
0 0 0
where
G(x, y, z, ξ, η, ζ, t) = H1 (x, ξ, t; l1 )H2 (y, η, t; l2 )H3 (z, ζ, t),

∞
H1 (x, ξ, t; l1 ) = sin sin exp − ,
l1 l1
n=1
l1 l12
∞
H2 (y, η, t; l2 ) = sin sin exp − ,
l2
n=1
l2 l2 l22

1 (z − ζ)2 (z + ζ)2
H3 (z, ζ, t) = √ exp − + exp − .
2 πat 4at 4at
2◦ . A semiinfinite cylindrical domain of a rectangular cross-section is considered. The

following conditions are prescribed:

Solution:
Z ∞Z l2Z l1
0 0 0
Z tZ ∞Z l2
0 0 0
Z tZ ∞Z l2
0 0 0
Z tZ ∞Z l1
0 0 0
Z tZ ∞Z l1
0 0 0
Z tZ l2Z l1
∂
0 0 0 ∂ζ ζ=0
where
G(x, y, z, ξ, η, ζ, t) = H1 (x, ξ, t; l1 )H2 (y, η, t; l2 )H3 (z, ζ, t),

X∞
H1 (x, ξ, t; l1 ) = 1+2 cos cos exp − ,
l1 n=1
l1 l1 l12
X∞
H2 (y, η, t; l2 ) = 1+2 cos cos exp − ,
l2
n=1
l2 l2 l22

1 (z − ζ)2 (z + ζ)2
H3 (z, ζ, t) = √ exp − − exp − .
2 πat 4at 4at
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z ≤ l3 . First boundary value problem.
A rectangular parallelepiped is considered. The following conditions are prescribed:

w = g6 (x, y, t) at z = l3 (boundary condition).
∂w
= a∆3 w 483
Solution:
Z l3Z l2Z l1
0 0 0
Z t Z l3Z l2
∂
0 0 0 ∂ξ ξ=0
Z t Z l3Z l2
∂
0 0 0 ∂ξ ξ=l1
Z t Z l3Z l1
∂
0 0 0 ∂η η=0
Z t Z l3Z l1
∂
0 0 0 ∂η η=l2
Z t Z l2Z l1
∂
0 0 0 ∂ζ ζ=0
Z t Z l2Z l1
∂
0 0 0 ∂ζ ζ=l3
where

∞
l1
n=1
l1 l1 l12
∞
G2 (y, η, t) = sin sin exp − ,
l2
n=1
l2 l2 l22
∞
2 X πnz πnζ π 2 n2 at
G3 (z, ζ, t) = sin sin exp − .
l3
n=1
l3 l3 l32
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z ≤ l3 . Second boundary value problem.


∂z w = g6 (x, y, t) at z = l3 (boundary condition).
Solution:
Z l3Z l2Z l1
0 0 0
Z t Z l3Z l2
0 0 0
Z t Z l3Z l2
0 0 0
Z t Z l3Z l1
0 0 0
Z t Z l3Z l1
0 0 0
Z t Z l2Z l1
0 0 0
Z t Z l2Z l1
+a g6 (ξ, η, τ ) G(x, y, z, ξ, η, l3 , t − τ ) dξ dη dτ,
0 0 0
where
X∞
G1 (x, ξ, t) = 1+2 cos cos exp − ,
l1
n=1
l1 l1 l12
X∞
G2 (y, η, t) = 1+2 cos cos exp − ,
l2
n=1
l2 l2 l22
X∞
1 πnz πnζ π 2 n2 at
G3 (z, ζ, t) = 1+2 cos cos exp − .
l3 n=1
l3 l3 l32
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z ≤ l3 . Third boundary value problem.

∂y w + k4 w = g4 (x, z, t) at y = l2 (boundary condition),
∂z w + k6 w = g6 (x, y, t) at z = l3 (boundary condition).
∂w
= a∆3 w 485
G(x, y, z, ξ, η, ζ, t) = H1 (x, ξ, t)H2 (y, η, t)H3 (z, ζ, t),

X∞
ϕn (x)ϕn (ξ)
H1 (x, ξ, t) = 2
exp(−aµ2n t),
n=1
kϕn k
∞
X ψm (y)ψm (η)
H2 (y, η, t) = exp(−aλ2m t),
kψm k2
m=1
X∞
ρs (z)ρs (ζ)
H3 (z, ζ, t) = 2
exp(−aνs2 t).
s=1
kρ s k
Here,

k1 k2 µ2n +k12
2 k1 l1 k12
ϕn(x) = cos(µnx)+ sin(µnx), kϕnk = 2 2 + + 1+ 2 ,

k3 2 k4 λ2m +k32 k3 l2 k32
ψm(y) = cos(λmy)+ sin(λmy), kψmk = 2 2 + + 1+ 2 ,

k5 k6 ν 2 +k2 k5 l3 k2
ρs(x) = cos(νsx)+ sin(νsx), kρsk2 = 2 s2 52 + 2 + 1+ 52 .
νs 2νs νs +k6 2νs 2 νs
The µn , λm , and νs are positive roots of the transcendental equations
tan(µl1 ) k1 + k2 tan(λl2 ) k3 + k4 tan(νl3 ) k5 + k6

= 2 , = 2 , = 2 .
µ µ − k1 k2 λ λ − k3 k4 ν ν − k5 k6
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z ≤ l3 . Mixed boundary value problems.
1◦ . A rectangular parallelepiped is considered. The following conditions are prescribed:

Solution:
Z l3Z l2Z l1
0 0 0
Z t Z l3Z l2
∂
0 0 0 ∂ξ ξ=0
Z t Z l3Z l2
∂
0 0 0 ∂ξ ξ=l1
Z t Z l3Z l1
∂
0 0 0 ∂η η=0
Z t Z l3Z l1
∂
0 0 0 ∂η η=l2
Z t Z l2Z l1
0 0 0
Z t Z l2Z l1
0 0 0
where

∞
l1
n=1
l1 l1 l12
∞
2 X πky πkη π 2 k2 at
G2 (y, η, t) = sin sin exp − ,
l2 l2 l2 l22
k=1
∞
1 2 X πmz πmζ π 2 m2 at
G3 (z, ζ, t) = + cos cos exp − .
l3 l3
m=1
l3 l3 l32

∂w
= a∆3 w 487
Solution:
Z l3Z l2Z l1
0 0 0
Z t Z l3Z l2
∂
0 0 0 ∂ξ ξ=0
Z t Z l3Z l2
∂
0 0 0 ∂ξ ξ=l1
Z t Z l3Z l2
0 0 0
Z t Z l3Z l1
0 0 0
Z t Z l2Z l1
0 0 0
Z t Z l2Z l1
0 0 0
where

∞
l1 n=1 l1 l1 l12
∞
1 2 X πky πkη π 2 k2 at
G2 (y, η, t) = + cos cos exp − ,
l2 l2 l2 l2 l22
k=1
∞
1 2 X πmz πmζ π 2 m2 at
G3 (z, ζ, t) = + cos cos exp − .
l3 l3
m=1
l3 l3 l32
5.1.2 Problems in Cylindrical Coordinates

The three-dimensional sourceless heat equation in the cylindrical coordinate system has the
form
p
∂w 1 ∂ ∂w 1 ∂2w ∂2w
=a r + 2 + , r = x2 + y 2 .
∂t r ∂r ∂r r ∂ϕ2 ∂z 2
It is used to describe nonsymmetric unsteady processes in moving media or solids with

cylindrical or plane boundaries. A similar equation is used to study the corresponding
three-dimensional unsteady mass-exchange processes with constant diffusivity.
One-dimensional problems with axial symmetry that have solutions of the form w =
w(r, t) are discussed in Section 3.2.1. Two-dimensional problems whose solutions have
the form w = w(r, ϕ, t) or w = w(r, z, t) are considered in Sections 4.1.2 and 4.1.3.
◮ Remarks on the Green’s functions.

For the three-dimensional problems dealt with in Section 5.1.2, the Green’s function can
be represented in the product form
G(r, ϕ, z, ξ, η, ζ, t) = G1 (r, ϕ, ξ, η, t) G2 (z, ζ, t),
where G1 (r, ϕ, ξ, η, t) is the Green’s function of the two-dimensional boundary value prob-
lem (such functions are presented in Section 4.1.2), and G2 (z, ζ, t) is the Green’s func-
tion of the corresponding one-dimensional boundary value problem (such functions can be
found in Sections 3.1.1 and 3.1.2).
Example 5.5. The Green’s function of the first boundary value problem for a semiinfinite circu-
lar cylinder (0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞) with the initial and boundary conditions
w = f (r, ϕ, z) at t = 0,
w = g1 (ϕ, z, t) at r = R, w = g2 (r, ϕ, t) at z = 0,
is the product of the two-dimensional Green’s function of the first boundary value problem of Sec-
tion 4.1.2 (for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π) and the one-dimensional Green’s function of the first
boundary value problem of Section 3.1.2 (for 0 ≤ z < ∞), in which one should perform obvious
renaming of variables.
General formulas that enable one to obtain solutions of basic boundary value problems
with the help of the Green’s function can be found in Sections 17.4.1 and 17.5.2.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, −∞ < z < ∞. First boundary value problem.

An infinite circular cylinder is considered. The following conditions are prescribed:
w = f (r, ϕ, z) at t = 0 (initial condition),
w = g(ϕ, z, t) at r = R (boundary condition).
Solution:
Z ∞ Z 2πZ R
w(r, ϕ, z, t) = ξf (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t) dξ dη dζ
−∞ 0 0
Z t Z ∞ Z 2π
∂
− aR g(η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dη dζ dτ.
0 −∞ 0 ∂ξ ξ=R
Here,
G(r, ϕ, z, ξ, η, ζ, t)=G1(r, ϕ, ξ, η, t) G2(z, ζ, t),
∞ X
X ∞
1 An
G1(r, ϕ, ξ, η, t)= Jn(µnmr)Jn(µnmξ) cos[n(ϕ−η)] exp(−µ2nmat),
πR2 n=0 m=1 [Jn′ (µnmR)]2
(
1 (z−ζ)2 1 for n=0,
G2(z, ζ, t)= √ exp − , An =
2 πat 4at 2 for n=1, 2, . . . ,
where the Jn (ξ) are Bessel functions (the prime denotes a derivative with respect to the
argument) and the µnm are positive roots of the transcendental equation Jn (µR) = 0.
∂w
= a∆3 w 489
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, −∞ < z < ∞. Second boundary value

problem.
∂r w = g(ϕ, z, t) at r = R (boundary condition).
Solution:
Z ∞ Z 2πZ R
−∞ 0 0
Z t Z ∞ Z 2π
+ aR g(η, ζ, τ ) G(r, ϕ, z, R, η, ζ, t − τ ) dη dζ dτ.
0 −∞ 0
Here,
G(r, ϕ, z, ξ, η, ζ, t) = G1(r, ϕ, ξ, η, t) G2(z, ζ, t),
∞ ∞
1 1 X X Anµ2nmJn(µnmr)Jn (µnmξ)
G1(r, ϕ, ξ, η, t) = + cos[n(ϕ−η)] exp(−µ2nmat),
πR2 π n=0 m=1 (µ2nmR2 −n2)[Jn(µnmR)]2
(
1 (z −ζ)2 1 for n = 0,
G2(z, ζ, t) = √ exp − , An =
2 πat 4at 2 for n = 1, 2, . . . ,
where the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation Jn′ (µR) = 0.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, −∞ < z < ∞. Third boundary value

problem.
∂r w + kw = g(ϕ, z, t) at r = R (boundary condition).
The solution w(r, ϕ, z, t) is determined by the formula in the previous paragraph (for
X∞ X ∞
1 Anµ2nmJn(µnmr)Jn(µnmξ)
G1(r, ϕ, ξ, η, t) = cos[n(ϕ−η)] exp(−µ2nmat),
π n=0 m=1 (µ2nmR2 +k 2R2 −n2)[Jn(µnmR)]2
(
1 (z −ζ)2 1 for n = 0,
G2(z, ζ, t) = √ exp − , An =
2 πat 4at 2 for n = 1, 2, . . .
Here, the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞. First boundary value problem.


w = g1 (ϕ, z, t) at r = R (boundary condition),
w = g2 (r, ϕ, t) at z = 0 (boundary condition).
Solution:
Z ∞Z 2πZ R
0 0 0
Z t Z ∞Z 2π
∂
− aR g1 (η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=R
Z t Z 2πZ R
∂
+a ξg2 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dη dτ.
0 0 0 ∂ζ ζ=0
Here,

∞ ∞
1 XX An
G1 (r, ϕ, ξ, η, t) = Jn (µnm r)Jn (µnm ξ) cos[n(ϕ − η)] exp(−µ2nm at),
πR2 n=0 m=1 [Jn′ (µnm R)]2
(
1 (z − ζ)2 (z + ζ)2 1 for n = 0,
G2 (z, ζ, t) = √ exp − − exp − , An =
2 πat 4at 4at 2 for n = 1, 2, . . . ,
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞. Second boundary value problem.


∂r w = g1 (ϕ, z, t) at r = R (boundary condition),
∂z w = g2 (r, ϕ, t) at z = 0 (boundary condition).
Solution:
Z ∞Z 2πZ R
0 0 0
Z t Z ∞Z 2π
+ aR g1 (η, ζ, τ ) G(r, ϕ, z, R, η, ζ, t − τ ) dη dζ dτ
0 0 0
Z tZ 2πZ R
−a ξg2 (ξ, η, τ ) G(r, ϕ, z, ξ, η, 0, t − τ ) dξ dη dτ.
0 0 0
∂w
= a∆3 w 491
Here,

∞ ∞
1 1 X X An µ2nm Jn (µnm r)Jn (µnm ξ)
G1 (r, ϕ, ξ, η, t) = + cos[n(ϕ − η)] exp(−µ2nm at),
πR2 π n=0 m=1 (µ2nm R2 − n2 )[Jn (µnm R)]2
(
1 (z − ζ)2 (z + ζ)2 1 for n = 0,
G2 (z, ζ, t) = √ exp − + exp − , An =
2 πat 4at 4at 2 for n = 1, 2, . . . ,
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞. Third boundary value problem.

∂r w + k1 w = g(ϕ, z, t) at r = R (boundary condition),
∂z w − k2 w = g2 (r, ϕ, t) at z = 0 (boundary condition).

∞ X
X ∞
1 An µ2nm Jn (µnm r)Jn (µnm ξ)
G1 (r, ϕ, ξ, η, t) = 2 R2 + k 2 R2 − n2 )[J (µ 2
cos[n(ϕ − η)] exp(−µ2nm at),
π
n=0 m=1
(µ nm 1 n nm R)]

1 (z − ζ)2 (z + ζ)2
G2 (z, ζ, t) = √ exp − + exp −
2 πat 4at 4at
Z ∞ 2

(z + ζ + s)
− 2k2 exp − − k2 s ds .
0 4at
Here, A0 = 1 and An = 2 for n = 1, 2, . . . ; the Jn (ξ) are Bessel functions and the µnm
are positive roots of the transcendental equation
µJn′ (µR) + k1 Jn (µR) = 0.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞. Mixed boundary value problems.

Solution:
Z ∞Z 2πZ R
0 0 0
Z t Z ∞Z 2π
∂
0 0 0 ∂ξ ξ=R
Z tZ 2πZ R
−a ξg2 (ξ, η, τ ) G(r, ϕ, z, ξ, η, 0, t − τ ) dξ dη dτ.
0 0 0
Here,

∞ ∞
1 XX An
2 ′
(
1 (z − ζ)2 (z + ζ)2 1 for n = 0,
G2 (z, ζ, t) = √ exp − + exp − , An =
2 πat 4at 4at 2 for n = 1, 2, . . . ,

Solution:
Z ∞Z 2πZ R
0 0 0
Z t Z ∞Z 2π
0 0 0
Z tZ 2πZ R
∂
+a ξg2 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dη dτ.
0 0 0 ∂ζ ζ=0
Here,

∞ ∞
(
1 (z − ζ)2 (z + ζ)2 1 for n = 0,
G2 (z, ζ, t) = √ exp − − exp − , An =
2 πat 4at 4at 2 for n = 1, 2, . . . ,
∂w
= a∆3 w 493
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l. First boundary value problem.

w = g2 (r, ϕ, t) at z = 0 (boundary condition),
w = g3 (r, ϕ, t) at z = l (boundary condition).
Solution:
Z lZ 2πZ R
0 0 0
Z t Z l Z 2π
∂
0 0 0 ∂ξ ξ=R
Z t Z 2πZ R
∂
+a ξg2 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dη dτ
0 0 0 ∂ζ ζ=0
Z t Z 2πZ R
∂
−a ξg3 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dη dτ.
0 0 0 ∂ζ ζ=l
Here,

∞ ∞
1 XX An
πR2 n=0 m=1 [Jn′ (µnm R)]2
∞ (
2X nπz nπζ an2 π 2 t 1 for n = 0,
G2 (z, ζ, t) = sin sin exp − 2
, An =
l n=1 l l l 2 for n = 1, 2, . . . ,
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l. Second boundary value problem.

∂z w = g2 (r, ϕ, t) at z = 0 (boundary condition),
∂z w = g3 (r, ϕ, t) at z = l (boundary condition).
Solution:
Z lZ 2πZ R
0 0 0
Z t Z l Z 2π
0 0 0
Z t Z 2πZ R
−a ξg2 (ξ, η, τ ) G(r, ϕ, z, ξ, η, 0, t − τ ) dξ dη dτ
0 0 0
Z t Z 2πZ R
+a ξg3 (ξ, η, τ ) G(r, ϕ, z, ξ, η, l, t − τ ) dξ dη dτ.
0 0 0
Here,
∞ ∞
∞ (
1 2X nπx nπξ an2π 2t 1 for n = 0,
G2(z, ζ, t) = + cos cos exp − , An =
l l n=1 l l l2 2 for n = 1, 2, . . . ,
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l. Third boundary value problem.


∂z w − k2 w = g2 (r, ϕ, t) at z = 0 (boundary condition),
∂z w + k3 w = g3 (r, ϕ, t) at z = l (boundary condition).
X∞
hs (z)hs (ζ)
G(r, ϕ, z, ξ, η, ζ, t) = G1 (r, ϕ, ξ, η, t) exp(−aλ2s t),
s=1
khs k2
∞ ∞
1 XX An µ2nm Jn (µnm r)Jn (µnm ξ)
G1 (r, ϕ, ξ, η, t) = cos[n(ϕ − η)] exp(−µ2nm at),
π n=0 m=1 (µ2nm R2 + k12 R2 − n2 )[Jn (µnm R)]2

k2 2 k3 λ2s + k22 k2 l k22
hs (z) = cos(λs z) + sin(λs z), khs k = + 2 + 1+ 2 .
λs 2λ2s λ2s + k32 2λs 2 λs
Here, A0 = 1 and An = 2 for n = 1, 2, . . . ; the Jn (ξ) are Bessel functions; and the µnm
and λs are positive roots of the transcendental equations
tan(λl) k2 + k3
µJn′ (µR) + k1 Jn (µR) = 0, = 2 .
λ λ − k2 k3
∂w
= a∆3 w 495
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l. Mixed boundary value problems.
scribed:

Solution:
Z lZ 2πZ R
0 0 0
Z t Z l Z 2π
∂
0 0 0 ∂ξ ξ=R
Z tZ 2πZ R
−a ξg2 (ξ, η, τ ) G(r, ϕ, z, ξ, η, 0, t − τ ) dξ dη dτ
0 0 0
Z tZ 2πZ R
+a ξg3 (ξ, η, τ ) G(r, ϕ, z, ξ, η, l, t − τ ) dξ dη dτ.
0 0 0
Here,

X∞ X ∞
1 An
G1(r, ϕ, ξ, η, t)= 2 Jn(µnmr)Jn(µnmξ) cos[n(ϕ−η)] exp(−µ2nmat),
πR n=0 m=1 [Jn′ (µnmR)]2
∞ (
1 2X nπz nπζ an2π 2t 1 for n=0,
G2(z, ζ, t)= + cos cos exp − , An=
l l n=1 l l l2 2 for n=1, 2, . . . ,
scribed:

Solution:
Z lZ 2πZ R
0 0 0
Z t Z l Z 2π
0 0 0
Z tZ 2πZ R
∂
0 0 0 ∂ζ ζ=0
Z t Z 2πZ R
∂
0 0 0 ∂ζ ζ=l
Here,

∞ ∞
∞ (
2X nπz nπζ an2 π 2 t 1 for n = 0,
G2 (z, ζ, t) = sin sin exp − , An =
l n=1 l l l2 2 for n = 1, 2, . . . ,
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, −∞ < z < ∞. First boundary value

problem.
An infinite hollow circular cylinder is considered. The following conditions are prescribed:

w = g1 (ϕ, z, t) at r = R1 (boundary condition),
w = g2 (ϕ, z, t) at r = R2 (boundary condition).
Solution:
Z ∞ Z 2πZ R2
w(r, ϕ, z, t) = f (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
−∞ 0 R1
Z tZ ∞ Z 2π
∂
+ aR1 g1 (η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 −∞ 0 ∂ξ ξ=R1
Z t Z ∞ Z 2π
∂
− aR2 g2 (η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dη dζ dτ.
0 −∞ 0 ∂ξ ξ=R2
∂w
= a∆3 w 497
Here,

1 (z − ζ)2
G(r, ϕ, z, ξ, η, ζ, t) = √ exp − G1 (r, ϕ, ξ, η, t),
2 πat 4at
∞ ∞
πXX
G1 (r, ϕ, ξ, η, t) = An Bnm Zn (µnm r)Zn (µnm ξ) cos[n(ϕ − η)] exp(−µ2nm at),
2
n=0 m=1
(
1/2 for n = 0, µ2nm Jn2 (µnm R2 )
An = Bnm = 2 ,
1 for n 6= 0, Jn (µnm R1 ) − Jn2 (µnm R2 )
where the Jn (r) and Yn (r) are Bessel functions and the µnm are positive roots of the
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, −∞ < z < ∞. Second boundary value

problem.
∂r w = g1 (ϕ, z, t) at r = R1 (boundary condition),
∂r w = g2 (ϕ, z, t) at r = R2 (boundary condition).
Solution:
Z ∞ Z 2πZ R2
−∞ 0 R1
Z tZ ∞ Z 2π
− aR1 g1 (η, ζ, τ ) G(r, ϕ, z, R1 , ξ, η, ζ, t − τ ) dη dζ dτ
0 −∞ 0
Z t Z ∞ Z 2π
+ aR2 g2 (η, ζ, τ ) G(r, ϕ, z, R2 , η, ζ, t − τ ) dη dζ dτ.
0 −∞ 0
Here,

1 (z − ζ)2
2 πat 4at
1
G1 (r, ϕ, ξ, η, t) =
π(R2 − R12 )
2
∞ ∞
1 X X An µ2nm Zn (µnm r)Zn (µnm ξ) cos[n(ϕ − η)] exp(−µ2nm at)
+ ,
π n=0 m=1 (µ2nm R22 − n2 )Zn2 (µnm R2 ) − (µ2nm R12 − n2 )Zn2 (µnm R1 )
Zn (µnm r) = Jn′ (µnm R1 )Yn (µnm r) − Yn′ (µnm R1 )Jn (µnm r),
where A0 = 1 and An = 2 for n = 1, 2, . . . ; the Jn (r) and Yn (r) are Bessel functions (the
prime denotes a derivative with respect to the argument); and the µnm are positive roots of
the transcendental equation
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, −∞ < z < ∞. Third boundary value

problem.

∂r w − k1 w = g1 (ϕ, z, t) at r = R1 (boundary condition),
∂r w + k2 w = g2 (ϕ, z, t) at r = R2 (boundary condition).

1 (z −ζ)2
G(r, ϕ, z, ξ, η, ζ, t)= √ exp − G1(r, ϕ, ξ, η, t),
2 πat 4at
∞ ∞
1X X Anµ2nmZn(µnmr)Zn(µnmξ) cos[n(ϕ−η)] exp(−µ2nmat)
G1(r, ϕ, ξ, η, t)= ,
π n=0 m=1(k2 R2 +µ2nmR22 −n2)Zn2(µnmR2)−(k12R12 +µ2nmR12 −n2)Zn2(µnmR1)
2 2

Zn(µnmr)= µnmJn′ (µnmR1)−k1Jn(µnmR1) Yn(µnmr)

− µnmYn′ (µnmR1)−k1Yn(µnmR1) Jn(µnmr).
Here, A0 = 1 and An = 2 for n = 1, 2, . . . ; Jn (r) and Yn (r) are Bessel functions; and the


◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞. First boundary value problem.

A semiinfinite hollow circular cylinder is considered. The following conditions are pre-
scribed:
Solution:
Z ∞Z 2πZ R2
0 0 R1
Z t Z ∞Z 2π
∂
0 0 0 ∂ξ ξ=R1
Z t Z ∞Z 2π
∂
− aR2 g2 (η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=R2
Z t Z 2πZ R2
∂
+a g3 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) ξ dξ dη dτ.
0 0 R1 ∂ζ ζ=0
∂w
= a∆3 w 499
Here,

1 (z − ζ)2 (z + ζ)2
G(r, ϕ, z, ξ, η, ζ, t) = √ exp − − exp − G1 (r, ϕ, ξ, η, t),
2 πat 4at 4at
∞ ∞
πXX
2
n=0 m=1
(
1/2 for n = 0, µ2nm Jn2 (µnm R2 )
An = Bnm = 2 ,
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞. Second boundary value

problem.
scribed:
Solution:
Z ∞Z 2πZ R2
0 0 R1
Z t Z ∞Z 2π
− aR1 g1 (η, ζ, τ ) G(r, ϕ, z, R1 , η, ζ, t − τ ) dη dζ dτ
0 0 0
Z tZ ∞Z 2π
+ aR2 g2 (η, ζ, τ ) G(r, ϕ, z, R2 , η, ζ, t − τ ) dη dζ dτ
0 0 0
Z t Z 2πZ R2
−a g3 (ξ, η, τ ) G(r, ϕ, z, ξ, η, 0, t − τ )ξ dξ dη dτ.
0 0 R1
Here,

1 (z−ζ)2 (z+ζ)2
G(r, ϕ, z, ξ, η, ζ, t) = √ exp − +exp − G1(r, ϕ, ξ, η, t),
2 πat 4at 4at
1
G1(r, ϕ, ξ, η, t) =
π(R2 −R12)
2
∞ ∞
1 X X Anµ2nmZn(µnmr)Zn(µnmξ) cos[n(ϕ−η)] exp(−µ2nmat)
+ ,
π n=0 m=1 (µ2nmR22 −n2)Zn2(µnmR2)−(µ2nmR12 −n2)Zn2 (µnmR1)
Zn(µnmr) = Jn′ (µnmR1)Yn(µnmr)−Yn′ (µnmR1)Jn(µnmr),
where A0 = 1 and An = 2 for n = 1, 2, . . . , the Jn (r) and Yn (r) are Bessel functions, and
the µnm are positive roots of the transcendental equation
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞. Third boundary value problem.
scribed:

∂r w + k2 w = g2 (ϕ, z, t) at r = R2 (boundary condition),
∂z w − k3 w = g3 (r, ϕ, t) at z = 0 (boundary condition).
The solution w(r, ϕ, z, t) is determined by the formula from the previous paragraph (for
the second boundary value problem) in which
G(r, ϕ, z, ξ, η, ζ, t) = G1(z, ζ, t) G2(r, ϕ, ξ, η, t),

Z ∞
1 (z−ζ)2 (z+ζ)2 (z+ζ+s)2
G1(z, ζ, t)= √ exp − +exp − −2k3 exp − −k3s ds ,
2 πat 4at 4at 0 4at
X∞ X ∞
1 AnµnmZn(µnmr)Zn(µnmξ) cos[n(ϕ−η)] exp(−µ2nmat)
2
G2(r, ϕ, ξ, η, t)= ,
π n=0 m=1 (k22R22+µ2nmR22−n2)Zn2(µnmR2)−(k12R12+µ2nmR12−n2)Zn2(µnmR1)


− µnmYn′ (µnmR1)−k1Yn(µnmR1) Jn(µnmr),
where A0 = 1 and An = 2 for n = 1, 2, . . . , the Jn (r) and Yn (r) are Bessel functions, and


◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞. Mixed boundary value

problems.
1◦ . A semiinfinite hollow circular cylinder is considered. The following conditions are

prescribed:

∂w
= a∆3 w 501
Solution:
Z ∞Z 2πZ R2
0 0 R1
Z t Z ∞Z 2π
∂
0 0 0 ∂ξ ξ=R1
Z t Z ∞Z 2π
∂
0 0 0 ∂ξ ξ=R2
Z t Z 2πZ R2
0 0 R1
Here,

1 (z − ζ)2 (z + ζ)2
G(r, ϕ, z, ξ, η, ζ, t) = √ exp − + exp − G1 (r, ϕ, ξ, η, t),
2 πat 4at 4at
∞ ∞
πXX
2
n=0 m=1
(
1/2 for n = 0, µ2nm Jn2 (µnm R2 )
An = Bnm = 2 ,
2◦ . A semiinfinite hollow circular cylinder is considered. The following conditions are

prescribed:

Solution:
Z ∞Z 2πZ R2
0 0 R1
Z t Z ∞Z 2π
0 0 0
Z tZ ∞Z 2π
0 0 0
Z t Z 2πZ R2
∂
0 0 R1 ∂ζ ζ=0
Here,

1 (z−ζ)2 (z+ζ)2
G(r, ϕ, z, ξ, η, ζ, t) = √ exp − −exp − G1(r, ϕ, ξ, η, t),
2 πat 4at 4at
1
G1(r, ϕ, ξ, η, t) =
π(R2 −R12)
2
∞ ∞
+ ,
π
n=0 m=1
(µ2nmR22 −n2)Zn2(µnmR2)−(µ2nmR12 −n2)Zn2 (µnmR1)
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l. First boundary value problem.

Solution:
Z lZ 2πZ R2
0 0 R1
Z t Z l Z 2π
∂
0 0 0 ∂ξ ξ=R1
Z t Z l Z 2π
∂
0 0 0 ∂ξ ξ=R2
Z t Z 2πZ R2
∂
+a g3 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) ξ dξ dη dτ
0 0 R1 ∂ζ ζ=0
Z t Z 2πZ R2
∂
−a g4 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) ξ dξ dη dτ.
0 0 R1 ∂ζ ζ=l
∂w
= a∆3 w 503
Here,
X∞
2 nπz nπζ an2 π 2 t
G(r, ϕ, z, ξ, η, ζ, t) = G1 (r, ϕ, ξ, η, t) sin sin exp − ,
l l l l2
n=1
∞ ∞
πXX
2
( n=0 m=1
1/2 for n = 0, µ2nm Jn2 (µnm R2 )
An = Bnm = 2 ,
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l. Second boundary value problem.

Solution:
Z lZ 2πZ R2
0 0 R1
Z tZ lZ 2π
0 0 0
Z tZ lZ 2π
0 0 0
Z tZ 2πZ R2
−a g3 (ξ, η, τ ) G(r, ϕ, z, ξ, η, 0, t − τ )ξ dξ dη dτ
0 0 R1
Z tZ 2πZ R2
+a g4 (ξ, η, τ ) G(r, ϕ, z, ξ, η, l, t − τ )ξ dξ dη dτ.
0 0 R1
Here,
∞
1 2X nπz nπζ an2π 2t
G(r, ϕ, z, ξ, η, ζ, t) = G1(r, ϕ, ξ, η, t) + cos cos exp − ,
l l n=1 l l l2
1
G1(r, ϕ, ξ, η, t) =
π(R22 −R12)
∞ ∞
+ ,
π n=0 m=1 (µ2nmR22 −n2)Zn2 (µnmR2)−(µ2nmR12 −n2)Zn2 (µnmR1)
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l. Third boundary value problem.

G(r, ϕ, z, ξ, η, ζ, t) = G1 (r, ϕ, ξ, η, t) G2 (z, ζ, t).
Here, the first factor has the form
∞ ∞
1X X Anµ2nmZn(µnmr)Zn(µnmξ) cos[n(ϕ−η)] exp(−µ2nmat)
G1(r, ϕ, ξ, η, t)= ,
π n=0 m=1(k22R22 +µ2nmR22 −n2)Zn2(µnmR2)−(k12R12 +µ2nmR12 −n2)Zn2(µnmR1)


where A0 = 1 and An = 2 for n = 1, 2, . . . ; the Jn (r) and Yn (r) are Bessel functions; and
′
µJn (µR1 ) − k1 Jn (µR1 ) µYn′ (µR2 ) + k2 Yn (µR2 )

The second factor is given by
∞
X hs (z)hs (ζ)
G2 (z, ζ, t) = exp(−aλ2s t),
khs k2
s=1

k3 2 k4 λ2s + k32 k3 l k32
hs (z) = cos(λs z) + sin(λs z), khs k = + 2 + 1+ 2 ,
∂w
= a∆3 w 505
tan(λl) k3 + k4
where the λs are positive roots of the transcendental equation = 2 .
λ λ − k3 k4
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l. Mixed boundary value problems.
scribed:
Solution:
Z lZ 2πZ R2
0 0 R1
Z tZ l Z 2π
∂
0 0 0 ∂ξ ξ=R1
Z t Z l Z 2π
∂
0 0 0 ∂ξ ξ=R2
Z t Z 2πZ R2
0 0 R1
Z tZ 2πZ R2
0 0 R1
Here,
∞ nπz nπζ an2π 2t
1 2X
G(r, ϕ, z, ξ, η, ζ, t)=G1 (r, ϕ, ξ, η, t) + cos cos exp − 2 ,
l l l l l
n=1
∞ ∞
πXX
G1(r, ϕ, ξ, η, t)= AnBnmZn(µnmr)Zn(µnmξ) cos[n(ϕ−η)] exp(−µ2nmat),
2 n=0 m=1
(
1/2 for n=0, µ2 J 2(µnmR2)
An = Bnm = 2 nm n 2 ,
1 for n6= 0, Jn(µnmR1)−Jn(µnmR2)
Zn(µnmr)=Jn(µnmR1)Yn(µnmr)−Yn(µnmR1)Jn(µnmr),

scribed:
Solution:
Z lZ 2πZ R2
0 0 R1
Z t Z l Z 2π
0 0 0
Z tZ l Z 2π
0 0 0
Z t Z 2πZ R2
∂
0 0 R1 ∂ζ ζ=0
Z t Z 2πZ R2
∂
0 0 R1 ∂ζ ζ=l
Here,
X∞
2 nπz nπζ an2π 2t
G(r, ϕ, z, ξ, η, ζ, t)=G1 (r, ϕ, ξ, η, t) sin sin exp − 2 ,
l l l l
n=1
1
G1(r, ϕ, ξ, η, t)=
π(R22 −R12)
∞ ∞
+ ,
π
n=0 m=1
(µ2nmR22−n2)Zn2(µnmR2)−(µ2nmR12−n2)Zn2(µnmR1)
Zn(µnmr)=Jn′ (µnmR1)Yn(µnmr)−Yn′ (µnmR1)Jn(µnmr),
◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , −∞ < z < ∞. First boundary value problem.

A dihedral angle is considered. The following conditions are prescribed:

w = g1 (r, z, t) at ϕ = 0 (boundary condition),
w = g2 (r, z, t) at ϕ = ϕ0 (boundary condition).
∂w
= a∆3 w 507
Solution:
Z ∞ Z ϕ0Z ∞
−∞ 0 0
Z tZ ∞ Z ∞
1 ∂
+a g1 (ξ, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 −∞ 0 ξ ∂η η=0
Z tZ ∞ Z ∞
1 ∂
−a g2 (ξ, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dζ dτ.
0 −∞ 0 ξ ∂η η=ϕ0
Here,

1 (z − ζ)2
2 πat 4at
2 ∞
1 r + ξ2 X rξ nπϕ nπη
G1 (r, ϕ, ξ, η, t) = exp − Inπ/ϕ0 sin sin ,
aϕ0 t 4at n=1
2at ϕ0 ϕ0
where the Iν (r) are modified Bessel functions.
◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , −∞ < z < ∞. Second boundary value

problem.
A dihedral angle is considered. The following conditions are prescribed:

−1
r ∂ϕ w = g1 (r, z, t) at ϕ = 0 (boundary condition),
−1
r ∂ϕ w = g2 (r, z, t) at ϕ = ϕ0 (boundary condition).
Solution:
Z ∞ Z ϕ0Z ∞
−∞ 0 0
Z tZ ∞ Z ∞
−a g1 (ξ, ζ, τ ) G(r, ϕ, z, ξ, 0, ζ, t − τ ) dξ dζ dτ
0 −∞ 0
Z tZ ∞ Z ∞
+a g2 (ξ, ζ, τ ) G(r, ϕ, z, ξ, ϕ0 , ζ, t − τ ) dξ dζ dτ.
0 −∞ 0
Here,

1 (z − ζ)2
2 πat 4at
2
1 r + ξ2 1 rξ
G1 (r, ϕ, ξ, η, t) = exp − I0
aϕ0 t 4at 2 2at
X∞
rξ nπϕ nπη
n=1
2at ϕ0 ϕ0

◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z < ∞. First boundary value problem.

The upper half of a dihedral angle is considered. The following conditions are prescribed:
w = g2 (r, z, t) at ϕ = ϕ0 (boundary condition),
Solution:
Z ∞Z ϕ0Z ∞
0 0 0
Z t Z ∞Z ∞
1 ∂
0 0 0 ξ ∂η η=0
Z t Z ∞Z ∞
1 ∂
−a g2 (ξ, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ξ ∂η η=ϕ0
Z t Z ϕ0Z ∞
∂
0 0 0 ∂ζ ζ=0
Here,

1 (z − ζ)2 (z + ζ)2
G(r, ϕ, z, ξ, η, ζ, t) = √ exp − − exp − G1 (r, ϕ, ξ, η, t),
2 πat 4at 4at
2 ∞
aϕ0 t 4at n=1
2at ϕ0 ϕ0
◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z < ∞. Second boundary value problem.

The upper half of a dihedral angle is considered. The following conditions are prescribed:
−1
−1
r ∂ϕ w = g2 (r, z, t) at ϕ = ϕ0 (boundary condition),
Solution:
Z ∞Z ϕ0Z ∞
0 0 0
Z tZ ∞Z ∞
0 0 0
Z tZ ∞Z ∞
+a g2 (ξ, ζ, τ ) G(r, ϕ, z, ξ, ϕ0 , ζ, t − τ ) dξ dζ dτ
0 0 0
Z tZ ϕ0Z ∞
0 0 0
∂w
= a∆3 w 509
Here,

1 (z−ζ)2 (z+ζ)2
2 πat 4at 4at
2 2
1 r +ξ 1 rξ
G1(r, ϕ, ξ, η, t) = exp − I0
aϕ0t 4at 2 2at
X∞
rξ nπϕ nπη
n=1
2at ϕ 0 ϕ 0
◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z < ∞. Mixed boundary value problems.

1◦ . The upper half of a dihedral angle is considered. The following conditions are pre-
scribed:
Solution:
Z ∞Z ϕ0Z ∞
0 0 0
Z t Z ∞Z ∞
1 ∂
0 0 0 ξ ∂η η=0
Z t Z ∞Z ∞
1 ∂
0 0 0 ξ ∂η η=ϕ0
Z t Z ϕ0Z ∞
0 0 0
Here,

1 (z − ζ)2 (z + ζ)2
G(r, ϕ, z, ξ, η, ζ, t) = √ exp − + exp − G1 (r, ϕ, ξ, η, t),
2 πat 4at 4at
2 ∞
aϕ0 t 4at 2at ϕ0 ϕ0
n=1

2◦ . The upper half of a dihedral angle is considered. The following conditions are pre-
scribed:
r −1 ∂ϕ w = g1 (r, z, t) at ϕ = 0 (boundary condition),
−1
Solution:
Z ∞Z ϕ0Z ∞
0 0 0
Z tZ ∞Z ∞
0 0 0
Z tZ ∞Z ∞
0 0 0
Z t Z ϕ0Z ∞
∂
0 0 0 ∂ζ ζ=0
Here,

1 (z−ζ)2 (z+ζ)2
2 πat 4at 4at
2 2
1 r +ξ 1 rξ
G1(r, ϕ, ξ, η, t) = exp − I0
aϕ0t 4at 2 2at
X∞
rξ nπϕ nπη
n=1
2at ϕ 0 ϕ 0
◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ l. First boundary value problem.

A wedge domain of finite thickness is considered. The following conditions are prescribed:
Solution:
Z lZ ϕ0Z ∞
0 0 0
Z tZ lZ ∞
1 ∂
0 0 0 ξ ∂η η=0
Z tZ lZ ∞
1 ∂
0 0 0 ξ ∂η η=ϕ0
Z t Z ϕ0Z ∞
∂
0 0 0 ∂ζ ζ=0
Z t Z ϕ0Z ∞
∂
0 0 0 ∂ζ ζ=l
∂w
= a∆3 w 511
Here,
X ∞
2 nπz nπζ an2 π 2 t
G(r, ϕ, z, ξ, η, ζ, t) = G1 (r, ϕ, ξ, η, t) sin sin exp − ,
l n=1 l l l2
2 ∞
aϕ0 t 4at 2at ϕ0 ϕ0
n=1
◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ l. Second boundary value problem.

A wedge domain of finite thickness is considered. The following conditions are prescribed:

−1
r −1 ∂ϕ w = g2 (r, z, t) at ϕ = ϕ0 (boundary condition),
Solution:
Z lZ ϕ0Z ∞
0 0 0
Z tZ l Z ∞
0 0 0
Z tZ l Z ∞
0 0 0
Z tZ ϕ0Z ∞
0 0 0
Z tZ ϕ0Z ∞
0 0 0
Here,

2
1 r + ξ2 1 rξ
G1 (r, ϕ, ξ, η, t) = exp − I0
aϕ0 t 4at 2 2at
X∞
rξ nπϕ nπη
2at ϕ0 ϕ0
n=1
∞
1 2X nπz nπζ an2 π 2 t
G2 (z, ζ, t) = + cos cos exp − ,
l l n=1 l l l2

◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ l. Mixed boundary value problems.
1◦ . A wedge domain of finite thickness is considered. The following conditions are pre-
scribed:
Solution:
Z lZ ϕ0Z ∞
0 0 0
Z tZ lZ ∞
1 ∂
0 0 0 ξ ∂η η=0
Z tZ lZ ∞
1 ∂
0 0 0 ξ ∂η η=ϕ0
Z t Z ϕ0Z ∞
0 0 0
Z t Z ϕ0Z ∞
0 0 0
Here,

2 ∞
aϕ0 t 4at n=1
2at ϕ0 ϕ0
∞
1 2X nπz nπζ an2 π 2 t
G2 (z, ζ, t) = + cos cos exp − ,
l l l l l2
n=1
2◦ . A wedge domain of finite thickness is considered. The following conditions are pre-
scribed:
r −1 ∂ϕ w = g1 (r, z, t) at ϕ = 0 (boundary condition),
−1
∂w
= a∆3 w 513
Solution:
Z lZ ϕ0Z ∞
0 0 0
Z tZ l Z ∞
0 0 0
Z tZ l Z ∞
0 0 0
Z t Z ϕ0Z ∞
∂
0 0 0 ∂ζ ζ=0
Z t Z ϕ0Z ∞
∂
0 0 0 ∂ζ ζ=l
Here,
X ∞
2 nπz nπζ an2π 2t
G(r, ϕ, z, ξ, η, ζ, t) =G1(r, ϕ, ξ, η, t) sin sin exp − 2 ,
l n=1 l l l
2 2
1 r +ξ 1 rξ
G1(r, ϕ, ξ, η, t)= exp − I0
aϕ0t 4at 2 2at
∞
X
rξ nπϕ nπη
2at ϕ0 ϕ0
n=1
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , −∞ < z < ∞. First boundary value problem.

An infinite cylindrical sector is considered. The following conditions are prescribed:

w = g3 (r, z, t) at ϕ = ϕ0 (boundary condition).
Solution:
Z ∞ Z ϕ0Z R
−∞ 0 0
Z t Z ∞ Z ϕ0
∂
0 −∞ 0 ∂ξ ξ=R
Z tZ ∞ Z R
1 ∂
0 −∞ 0 ξ ∂η η=0
Z tZ ∞ Z R
1 ∂
−a g3 (ξ, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dζ dτ.
0 −∞ 0 ξ ∂η η=ϕ0
Here,

1 (z − ζ)2
2 πat 4at
∞ ∞
4 X X Jnπ/ϕ0 (µnm r)Jnπ/ϕ0 (µnm ξ)
G1 (r, ϕ, ξ, η, t) = 2 ′
R ϕ0 n=1m=1 [Jnπ/ϕ 0
(µnm R)]2

nπϕ nπη
× sin sin exp(−µ2nm at),
ϕ0 ϕ0
where the Jnπ/ϕ0 (r) are Bessel functions and the µnm are positive roots of the transcen-
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z < ∞. First boundary value problem.

A semiinfinite cylindrical sector is considered. The following conditions are prescribed:
Solution:
Z ∞Z ϕ0Z R
0 0 0
Z t Z ∞Z ϕ0
∂
0 0 0 ∂ξ ξ=R
Z t Z ∞Z R
1 ∂
0 0 0 ξ ∂η η=0
Z t Z ∞Z R
1 ∂
0 0 0 ξ ∂η η=ϕ0
Z t Z ϕ0Z R
∂
0 0 0 ∂ζ ζ=0
Here,

1 (z−ζ)2 (z+ζ)2
2 πat 4at 4at
∞ ∞
4 X X Jnπ/ϕ0(µnmr)Jnπ/ϕ0(µnmξ)
G1(r, ϕ, ξ, η, t) = 2 ′
(µnmR)]2

nπϕ nπη
×sin sin exp(−µ2nmat),
ϕ0 ϕ0
∂w
= a∆3 w 515
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z < ∞. Mixed boundary value problem.

A semiinfinite cylindrical sector is considered. The following conditions are prescribed:
Solution:
Z ∞Z ϕ0Z R
0 0 0
Z t Z ∞Z ϕ0
∂
0 0 0 ∂ξ ξ=R
Z t Z ∞Z R
1 ∂
0 0 0 ξ ∂η η=0
Z t Z ∞Z R
1 ∂
0 0 0 ξ ∂η η=ϕ0
Z t Z ϕ0Z R
0 0 0
Here,

1 (z−ζ)2 (z+ζ)2
2 πat 4at 4at
∞ ∞
4 X X Jnπ/ϕ0(µnmr)Jnπ/ϕ0(µnmξ)
G1(r, ϕ, ξ, η, t) = 2 ′
R ϕ0 [Jnπ/ϕ (µnmR)]2
n=1m=1 0

nπϕ nπη
ϕ0 ϕ0
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ l. First boundary value problem.

A cylindrical sector of finite thickness is considered. The following conditions are pre-
scribed:
Solution:
Z lZ ϕ0Z R
0 0 0
Z t Z l Z ϕ0
∂
0 0 0 ∂ξ ξ=R
Z tZ l Z R
1 ∂
0 0 0 ξ ∂η η=0
Z tZ l Z R
1 ∂
0 0 0 ξ ∂η η=ϕ0
Z t Z ϕ0Z R
∂
0 0 0 ∂ζ ζ=0
Z t Z ϕ0Z R
∂
0 0 0 ∂ζ ζ=l
Here,

∞ ∞
4 X X Jnπ/ϕ0 (µnm r)Jnπ/ϕ0 (µnm ξ)
G1 (r, ϕ, ξ, η, t) = 2 ′
R ϕ0 [Jnπ/ϕ (µnm R)]2
n=1m=1 0

nπϕ nπη
× sin sin exp(−µ2nm at),
ϕ0 ϕ0
∞
2X nπz nπζ an2 π 2 t
G2 (z, ζ, t) = sin sin exp − ,
l l l l2
n=1
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ l. Mixed boundary value problem.
scribed:
∂w
= a∆3 w 517
Solution:
Z lZ ϕ0Z R
0 0 0
Z t Z l Z ϕ0
∂
0 0 0 ∂ξ ξ=R
Z tZ l Z R
1 ∂
0 0 0 ξ ∂η η=0
Z tZ l Z R
1 ∂
0 0 0 ξ ∂η η=ϕ0
Z t Z ϕ0Z R
0 0 0
Z t Z ϕ0Z R
0 0 0
Here,
∞
1 2X nπz nπζ an2π 2t
G(r, ϕ, z, ξ, η, ζ, t) = G1(r, ϕ, ξ, η, t) + cos cos exp − ,
l l n=1 l l l2
∞ ∞
4 X X Jnπ/ϕ0(µnmr)Jnπ/ϕ0 (µnmξ)
G1(r, ϕ, ξ, η, t) = 2 ′
(µnmR)]2

nπϕ nπη
ϕ0 ϕ0
5.1.3 Problems in Spherical Coordinates

The heat equation in the spherical coordinate system has the form

∂w 1 ∂ 2 ∂w 1 ∂ ∂w 1 ∂2w
=a 2 r + 2 sin θ + 2 2 ,
∂t r ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂ϕ2
p
r = x2 + y 2 + z 2 .
This representation is convenient to describe three-dimensional heat and mass exchange
phenomena in domains bounded by coordinate surfaces of the spherical coordinate system.
One-dimensional problems with central symmetry that have solutions of the form w =
w(r, t) are discussed in Section 3.2.3.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π. First boundary value problem.

A spherical domain is considered. The following conditions are prescribed:
w = f (r, θ, ϕ) at t = 0 (initial condition),
w = g(θ, ϕ, t) at r = R (boundary condition).
Solution:
Z 2πZ π Z R
w(r, θ, ϕ, t)= f (ξ, η, ζ) G(r, θ, ϕ, ξ, η, ζ, t)ξ 2 sin η dξ dη dζ
0 0 0
Z tZ 2πZ π
2 ∂
−aR g(η, ζ, τ ) G(r, θ, ϕ, ξ, η, ζ, t−τ ) sin η dη dζ dτ,
0 0 0 ∂ξ ξ=R
where
X∞ X ∞ X n
1
G(r, θ, ϕ, ξ, η, ζ, t) = √ Ak Bnmk Jn+1/2 (λnm r)Jn+1/2 (λnm ξ)
2πR2 rξ n=0 m=1 k=0
× Pnk (cos θ)Pnk (cos η) cos[k(ϕ − ζ)] exp(−λ2nm at),
(
1 for k = 0, (2n + 1)(n − k)!
Ak = Bnmk = ′ 2 .
2 for k =
6 0, (n + k)! Jn+1/2 (λnm R)
Here, the Jn+1/2 (r) are Bessel functions, the Pnk (µ) are associated Legendre functions
expressed in terms of the Legendre polynomials Pn (µ) as follows:
dk 1 dn
Pnk (µ) = (1 − µ2 )k/2 Pn (µ), Pn (µ) = (µ2 − 1)n ;
dµk n! 2n dµn
and the λnm are positive roots of the transcendental equation Jn+1/2 (λR) = 0.
(1984).
◮ Domain: 0 ≤ r ≤ R, 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π. Second boundary value problem.


∂r w = g(θ, ϕ, t) at r = R (boundary condition).
Solution:
Z 2πZ π Z R
w(r, θ, ϕ, t) = f (ξ, η, ζ) G(r, θ, ϕ, ξ, η, ζ, t)ξ 2 sin η dξ dη dζ
0 0 0
Z t Z 2πZ π
+ aR2 g(η, ζ, τ ) G(r, θ, ϕ, R, η, ζ, t − τ ) sin η dη dζ dτ,
0 0 0
where
X∞ X ∞ Xn
3 1
G(r, θ, ϕ, ξ, η, ζ, t) = + √ Ak Bnmk Jn+1/2 (λnm r)Jn+1/2 (λnm ξ)
4πR3 2π rξ n=0 m=1
k=0
× Pn (cos θ)Pnk (cos η) cos[k(ϕ − ζ)] exp(−λ2nm at),
k
(
1 for k = 0, λ2nm (2n + 1)(n − k)!
Ak = Bnmk = 2 .
2 for k 6= 0, (n + k)! R2 λ2nm − n(n + 1) Jn+1/2 (λnm R)
∂w
= a∆3 w 519
Here, the Jn+1/2 (r) are Bessel functions, the Pnk (µ) are associated Legendre functions (see
the previous paragraph), and the λnm are positive roots of the transcendental equation
′
2λRJn+1/2 (λR) − Jn+1/2 (λR) = 0.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π. Third boundary value problem.


∂r w + kw = g(θ, ϕ, t) at r = R (boundary condition).
The solution w(r, θ, ϕ, t) is determined by the formula in the previous paragraph (for
∞ X
X ∞ X
n
1
G(r, θ, ϕ, ξ, η, ζ, t)= √ AsBnmsJn+1/2(λnmr)Jn+1/2(λnmξ)
2π rξ n=0 m=1 s=0
×Pns(cos θ)Pns(cos η) cos[s(ϕ−ζ)] exp(−λ2nmat),
(
1 if s=0, λ2nm(2n+1)(n−s)!
As = Bnms = 2 .
2 if s6= 0, (n+s)! R2λ2nm+(kR+n)(kR−n−1) Jn+1/2(λnmR)
Here, the Jn+1/2 (r) are Bessel functions, the Pns (µ) are associated Legendre functions (see
above), and the λnm are positive roots of the transcendental equation
′ 1

λRJn+1/2 (λR) + kR − 2 Jn+1/2 (λR) = 0.
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π. First boundary value problem.

A spherical layer is considered. The following conditions are prescribed:

w = g1 (θ, ϕ, t) at r = R1 (boundary condition),
w = g2 (θ, ϕ, t) at r = R2 (boundary condition).
Solution:
Z 2πZ π Z R2
w(r, θ, ϕ, t)= f (ξ, η, ζ) G(r, θ, ϕ, ξ, η, ζ, t)ξ 2 sin η dξ dη dζ
0 0 R1
Z tZ 2πZ π
∂
+aR12 g1(η, ζ, τ ) G(r, θ, ϕ, ξ, η, ζ, t−τ ) sin η dη dζ dτ
0 0 0 ∂ξ ξ=R1
Z tZ 2πZ π
2 ∂
−aR2 g2(η, ζ, τ ) G(r, θ, ϕ, ξ, η, ζ, t−τ ) sin η dη dζ dτ,
0 0 0 ∂ξ ξ=R2
where
∞ ∞ n
π XXX
G(r, θ, ϕ, ξ, η, ζ, t) = √ Ak Bnmk Zn+1/2 (λnmr)Zn+1/2 (λnmξ)
8 rξ n=0 m=1 k=0
× Pnk (cos θ)Pnk (cos η) cos[k(ϕ − ζ)] exp(−λ2nmat).
Here,
Zn+1/2 (λnm r) = Jn+1/2 (λnm R1 )Yn+1/2 (λnm r) − Yn+1/2 (λnm R1 )Jn+1/2 (λnm r),
(
1 for k = 0, λ2nm (2n + 1)(n − k)! Jn+1/2
2 (λnm R2 )
Ak = Bnmk = 2 2
,
2 for k 6= 0, (n + k)! Jn+1/2 (λnm R1 ) − Jn+1/2 (λnm R2 )
where the Jn+1/2 (r) are Bessel functions, the Pnk (µ) are associated Legendre functions
dk 1 dn
Pnk (µ) = (1 − µ2 )k/2 Pn (µ), Pn (µ) = (µ2 − 1)n ;
dµk n! 2n dµn
and the λnm are positive roots of the transcendental equation
Zn+1/2 (λR2 ) = 0.
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π. Second boundary value

problem.

∂r w = g1 (θ, ϕ, t) at r = R1 (boundary condition),
∂r w = g2 (θ, ϕ, t) at r = R2 (boundary condition).
Solution:
Z 2πZ π Z R2
0 0 R1
Z tZ 2πZ π
− aR12 g1 (η, ζ, τ ) G(r, θ, ϕ, R1 , η, ζ, t − τ ) sin η dη dζ dτ
0 0 0
Z tZ 2πZ π
+ aR22 g2 (η, ζ, τ ) G(r, θ, ϕ, R2 , η, ζ, t − τ ) sin η dη dζ dτ,
0 0 0
where
X∞ X ∞ Xn
3 1 Ak
G(r, θ, ϕ, ξ, η, ζ, t) = + √ Z (λnm r)
3 3
4π(R2 − R1 ) 4π rξ n=0 m=1 Bnmk n+1/2
k=0
× Zn+1/2 (λnm ξ)Pnk (cos θ)Pnk (cos η) cos[k(ϕ − ζ)] exp(−λ2nm at).
∂w
= a∆3 w 521
Here,
( Z R2
1 for k = 0, (n + k)! 2
Ak = Bnmk = rZn+1/2 (λnm r) dr,
2 for k 6= 0, (2n + 1)(n − k)! R1

′ 1
Zn+1/2 (λr) = λJn+1/2 (λR1 ) − J (λR1 ) Yn+1/2 (λr)
2R1 n+1/2

′ 1
− λYn+1/2 (λR1 ) − Y (λR1 ) Jn+1/2 (λr),
2R1 n+1/2
where the Jn+1/2 (r) and Yn+1/2 (r) are Bessel functions, the Pnk (µ) are associated Legen-
dre functions (see the previous paragraph), and the λnm are positive roots of the transcen-
dental equation
′ 1
λZn+1/2 (λR2 ) − Z (λR2 ) = 0.
2R2 n+1/2
The integrals that determine the coefficients Bnmk can be expressed in terms of the
Bessel functions and their derivatives; see Budak, Samarskii, and Tikhonov (1980).
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π. Third boundary value problem.


∂r w − k1 w = g1 (θ, ϕ, t) at r = R1 (boundary condition),
∂r w + k2 w = g2 (θ, ϕ, t) at r = R2 (boundary condition).
X∞ X ∞ X n
1 As
G(r, θ, ϕ, ξ, η, ζ, t) = √ Zn+1/2 (λnm r)Zn+1/2 (λnm ξ)
4π rξ n=0 m=1 s=0 Bnms
× Pns (cos θ)Pns (cos η) cos[s(ϕ − ζ)] exp(−λ2nm at).
Here,
( Z R2
1 for s = 0, (n + s)! 2
As = Bnms = rZn+1/2 (λnm r) dr,
2 for s =
6 0, (2n + 1)(n − s)! R1

′ 1
Zn+1/2 (λr) = λJn+1/2 (λR1 ) − k1 + Jn+1/2 (λR1 ) Yn+1/2 (λr)
2R1

′ 1
− λYn+1/2 (λR1 ) − k1 + Yn+1/2 (λR1 ) Jn+1/2 (λr),
2R1
where the Jn+1/2 (r) and Yn+1/2 (r) are Bessel functions, the Pns (µ) are associated Legen-
dre functions (see above), and the λnm are positive roots of the transcendental equation

′ 1
λZn+1/2 (λR2 ) + k2 − Zn+1/2 (λR2 ) = 0.
2R2
The integrals that determine the coefficients Bnms can be expressed in terms of the
Bessel functions and their derivatives.
◮ Domain: 0 ≤ r < ∞, 0 ≤ θ ≤ θ0 , 0 ≤ ϕ ≤ 2π. First boundary value problem.

A cone is considered. The following conditions are prescribed:
w = g(r, ϕ, t) at θ = θ0 (boundary condition).
Solution:
Z 2πZ θ0Z ∞
0 0 0
Z tZ 2πZ ∞
∂
−a g(ξ, ζ, τ ) sin η G(r, θ, ϕ, ξ, η, ζ, t−τ ) dξ dζ dτ,
0 0 0 ∂η η=θ0
where
X∞ X 2 2
1 Am(2ν +1) r +ξ rξ
G(r, θ, ϕ, ξ, η, ζ, t) = − √ exp − Iν+1/2
4πat rξ m=0 ν Bmν 4at 2at
×Pν−m(cos θ)Pν−m(cos η) cos[m(ϕ−ζ)],
(
1 for m = 0, 2 d −m d −m
Am = Bmν = (1−µ) P (µ) Pν (µ) .
2 for m 6= 0, dµ ν dν µ=cos θ0
Here, Pν−m (µ) is the modified Legendre function expressed as

−m 1 1 − µ m/2 1

Pν (µ) = F −ν, ν + 1, 1 + m; 2 − 12 µ ,
Γ(1 + m) 1 + µ
where F (a, b, c; µ) is the Gaussian hypergeometric function and Γ(z) is the gamma func-
tion. The summation with respect to ν is performed over all roots of the equation
Pν−m (cos θ0 ) = 0 that are greater than −1/2.
∂w
5.2 Heat Equation with Source = a∆3 w + Φ(x, y, z, t)
∂t
In the Cartesian coordinate system, the three-dimensional heat equation with a volume
source has the form
2
∂w ∂ w ∂2w ∂2w
=a + + + Φ(x, y, z, t).
∂t ∂x2 ∂y 2 ∂z 2
It describes three-dimensional unsteady thermal phenomena in quiescent media or solids
with constant thermal diffusivity. A similar equation is used to study the corresponding
three-dimensional mass transfer processes with constant diffusivity.
∂w
5.2. Heat Equation with Source ∂t
= a∆3 w + Φ(x, y, z, t) 523
w = f (x, y, z) at t = 0.
Solution:
Z ∞Z ∞Z ∞
−∞ −∞ −∞
Z tZ ∞ Z ∞ Z ∞
+ Φ(ξ, η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dζ dτ,
0 −∞ −∞ −∞
where

1 (x − ξ)2 + (y − η)2 + (z − ζ)2
G(x, y, z, ξ, η, ζ, t) = exp − .
8(πat)3/2 4at
◮ Domain: 0 ≤ x < ∞, −∞ < y < ∞, −∞ < z < ∞. First boundary value

problem.

w = g(y, z, t) at x = 0 (boundary condition).
Solution:
Z ∞Z ∞Z ∞
−∞ −∞ 0
Z tZ ∞ Z ∞
∂
+a g(η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 −∞ −∞ ∂ξ ξ=0
Z tZ ∞ Z ∞ Z ∞
0 −∞ −∞ 0
where

1 (x−ξ)2 (x+ξ)2 (y−η)2+(z−ζ)2
G(x, y, z, ξ, η, ζ, t)= exp − −exp − exp − .

◮ Domain: 0 ≤ x < ∞, −∞ < y < ∞, −∞ < z < ∞. Second boundary value

problem.
∂x w = g(y, z, t) at x = 0 (boundary condition).
Solution:
Z ∞Z ∞Z ∞
−∞ −∞ 0
Z tZ ∞ Z ∞
−a g(η, ζ, τ ) G(x, y, z, 0, η, ζ, t − τ ) dη dζ dτ
0 −∞ −∞
Z tZ ∞ Z ∞ Z ∞
0 −∞ −∞ 0
where

1 (x−ξ)2 (x+ξ)2 (y−η)2+(z−ζ)2
G(x, y, z, ξ, η, ζ, t)= exp − +exp − exp − .
◮ Domain: 0 ≤ x < ∞, −∞ < y < ∞, −∞ < z < ∞. Third boundary value

problem.
∂x w − kw = g(y, z, t) at x = 0 (boundary condition).

1 (y−η)2+(z−ζ)2 (x−ξ)2 (x+ξ)2
G(x, y, z, ξ, η, ζ, t)= exp − exp − +exp −

√ x+ξ √
−2k πat exp k 2at+k(x+ξ) erfc √ +k at .
2 at
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, 0 ≤ z ≤ l. Different boundary value

problems.
1◦ . The solution w(x, y, z, t) of the first boundary value problem for an infinite layer is
given by the formula in Section 5.1.1 (see the first boundary value problem for −∞ < x <
∞, −∞ < y < ∞, 0 ≤ z ≤ l) with the additional term
Z tZ lZ ∞ Z ∞
Φ(ξ, η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dζ dτ, (1)
0 0 −∞ −∞
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
∂w
= a∆3 w + Φ(x, y, z, t) 525
2◦ . The solution w(x, y, z, t) of the second boundary value problem for an infinite layer is
given by the formula in Section 5.1.1 (see the second boundary value problem for −∞ <
x < ∞, −∞ < y < ∞, 0 ≤ z ≤ l) with the additional term (1); the Green’s function is also
taken from Section 5.1.1.
3◦ . The solution w(x, y, z, t) of the third boundary value problem for an infinite layer is
given by the formula in Section 5.1.1 (see the third boundary value problem for −∞ < x <
∞, −∞ < y < ∞, 0 ≤ z ≤ l) with the additional term (1); the Green’s function is also
4◦ . The solution w(x, y, z, t) of a mixed boundary value problem for an infinite layer is
given by the formula in Section 5.1.1 (see the mixed boundary value problem for −∞ <
x < ∞, −∞ < y < ∞, 0 ≤ z ≤ l) with the additional term (1); the Green’s function is also
◮ Domain: −∞ < x < ∞, 0 ≤ y < ∞, 0 ≤ z ≤ l. Different boundary value

problems.
1◦ . The solution w(x, y, z, t) of the first boundary value problem for a semiinfinite layer is
given by the formula in Section 5.1.1 (see the first boundary value problem for −∞ < x <
∞, 0 ≤ y < ∞, 0 ≤ z ≤ l) with the additional term
Z t Z l Z ∞Z ∞
0 0 0 −∞
2◦ . The solution w(x, y, z, t) of the second boundary value problem for a semiinfinite layer
is given by the formula in Section 5.1.1 (see the second boundary value problem for −∞ <
x < ∞, 0 ≤ y < ∞, 0 ≤ z ≤ l) with the additional term (2); the Green’s function is also
3◦ . The solution w(x, y, z, t) of the third boundary value problem for a semiinfinite layer
is given by the formula in Section 5.1.1 (see the third boundary value problem for −∞ <
x < ∞, 0 ≤ y < ∞, 0 ≤ z ≤ l) with the additional term (2); the Green’s function is also
4◦ . The solutions w(x, y, z, t) of mixed boundary value problems for a semiinfinite layer
are given by the formulas in Section 5.1.1 (see the mixed boundary value problems for
−∞ < x < ∞, 0 ≤ y < ∞, 0 ≤ z ≤ l) with additional terms of the form (2); the Green’s
function is also taken from Section 5.1.1.
◮ Domain: 0 ≤ x < ∞, 0 ≤ y < ∞, 0 ≤ z < ∞. Different boundary value

problems.
1◦ . The solution w(x, y, z, t) of the first boundary value problem for the first octant is
given by the formula in Section 5.1.1 (see the first boundary value problem for 0 ≤ x < ∞,
0 ≤ y < ∞, 0 ≤ z < ∞) with the additional term

Z t Z ∞Z ∞Z ∞
0 0 0 0
2◦ . The solution w(x, y, z, t) of the second boundary value problem for the first octant is
given by the formula in Section 5.1.1 (see the second boundary value problem for 0 ≤ x <
∞, 0 ≤ y < ∞, 0 ≤ z < ∞) with the additional term (3); the Green’s function is also taken
from Section 5.1.1.
3◦ . The solution w(x, y, z, t) of the third boundary value problem for the first octant is
given by the formula in Section 5.1.1 (see the third boundary value problem for 0 ≤ x < ∞,
0 ≤ y < ∞, 0 ≤ z < ∞) with the additional term (3); the Green’s function is also taken
from Section 5.1.1.
4◦ . The solutions w(x, y, z, t) of mixed boundary value problems for the first octant are
given by the formulas in Section 5.1.1 (see the mixed boundary value problems for 0 ≤ x <
∞, 0 ≤ y < ∞, 0 ≤ z < ∞) with additional terms of the form (3); the Green’s function is
also taken from Section 5.1.1.
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , −∞ < z < ∞. Different boundary value

problems.
1◦ . The solution w(x, y, z, t) of the first boundary value problem in an infinite rectangular
domain is given by the formula in Section 5.1.1 (see the first boundary value problem for
0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , −∞ < z < ∞) with the additional term
Z t Z ∞ Z l2Z l1
0 −∞ 0 0
2◦ . The solution w(x, y, z, t) of the second boundary value problem in an infinite rectangu-
lar domain is given by the formula in Section 5.1.1 (see the second boundary value problem
for 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , −∞ < z < ∞) with the additional term (4); the Green’s function
is also taken from Section 5.1.1.
3◦ . The solution w(x, y, z, t) of the third boundary value problem in an infinite rectangular
domain is given by the formula in Section 5.1.1 (see the third boundary value problem for
0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , −∞ < z < ∞) with the additional term (4); the Green’s function
4◦ . The solution w(x, y, z, t) of a mixed boundary value problem in an infinite rectangular
domain is given by the formula in Section 5.1.1 (see the mixed boundary value problem for
0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , −∞ < z < ∞) with the additional term (4); the Green’s function
∂w
= a∆3 w + Φ(x, y, z, t) 527
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z < ∞. Different boundary value problems.
1◦ . The solution w(x, y, z, t) of the first boundary value problem in a semiinfinite rectan-
gular domain is given by the formula of Section 5.1.1 (see the first boundary value problem
for 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z < ∞) with the additional term
Z tZ ∞Z l2Z l1
0 0 0 0
2◦ . The solution w(x, y, z, t) of the second boundary value problem in a semiinfinite rect-
angular domain is given by the formula in Section 5.1.1 (see the second boundary value
problem for 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z < ∞) with the additional term (5); the Green’s
3◦ . The solution w(x, y, z, t) of the third boundary value problem in a semiinfinite rectan-
gular domain is given by the formula in Section 5.1.1 (see the third boundary value problem
for 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z < ∞) with the additional term (5); the Green’s function
4◦ . The solutions w(x, y, z, t) of mixed boundary value problems in a semiinfinite rect-
angular domain are given by the formulas in Section 5.1.1 (see the mixed boundary value
problems for 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z < ∞) with additional terms of the form (5);
the Green’s function is also taken from Section 5.1.1.
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z ≤ l3 . Different boundary value problems.
1◦ . The solution w(x, y, z, t) of the first boundary value problem for a rectangular paral-
lelepiped is given by the formula in Section 5.1.1 (see the first boundary value problem for
0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z ≤ l3 ) with the additional term
Z t Z l3Z l2Z l1
0 0 0 0
2◦ . The solution w(x, y, z, t) of the second boundary value problem for a rectangular par-
allelepiped is given by the formula in Section 5.1.1 (see the second boundary value problem
for 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z ≤ l3 ) with the additional term (6); the Green’s function
3◦ . The solution w(x, y, z, t) of the third boundary value problem for a rectangular paral-
lelepiped is given by the formula in Section 5.1.1 (see the third boundary value problem for
0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z ≤ l3 ) with the additional term (6); the Green’s function is
also taken from Section 5.1.1.
4◦ . The solutions w(x, y, z, t) of mixed boundary value problems for a rectangular paral-
lelepiped are given by the formulas in Section 5.1.1 (see the mixed boundary value prob-
lems for 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z ≤ l3 ) with additional terms of the form (6); the
Green’s function is also taken from Section 5.1.1.

In the cylindrical coordinate system, the heat equation with a volume source is written as

∂w 1 ∂ ∂w 1 ∂2w ∂2w
=a r + 2 + + Φ(r, ϕ, z, t).
∂t r ∂r ∂r r ∂ϕ2 ∂z 2
This representation is used to describe nonsymmetric unsteady thermal (diffusion) pro-

cesses in quiescent media or solids bounded by cylindrical surfaces and planes.

w = g(ϕ, z, t) at r = R (boundary condition).
Solution:
Z ∞ Z 2πZ R
−∞ 0 0
Z t Z ∞ Z 2π
∂
− aR g(η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 −∞ 0 ∂ξ ξ=R
Z t Z ∞ Z 2πZ R
+ Φ(ξ, η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ )ξ dξ dη dζ dτ.
0 −∞ 0 0
Here,
G(r, ϕ, z, ξ, η, ζ, t)=G1(r, ϕ, ξ, η, t) G2(z, ζ, t),

∞ X
X ∞
1 An
G1(r, ϕ, ξ, η, t)= Jn(µnmr)Jn(µnmξ) cos[n(ϕ−η)] exp(−µ2nmat),
πR n=0 m=1 [Jn(µnmR)]2
2 ′
(
1 (z−ζ)2 1 for n=0,
G2(z, ζ, t)= √ exp − , An =
2 πat 4at 2 for n=1, 2, . . . ,
∂w
= a∆3 w + Φ(x, y, z, t) 529

problem.

∂r w = g(ϕ, z, t) at r = R (boundary condition).
Solution:
Z ∞ Z 2πZ R
−∞ 0 0
Z t Z ∞ Z 2π
+ aR g(η, ζ, τ ) G(r, ϕ, z, R, η, ζ, t − τ ) dη dζ dτ
0 −∞ 0
Z t Z ∞ Z 2πZ R
+ Φ(ξ, η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ )ξ dξ dη dζ dτ.
0 −∞ 0 0
Here,

∞ ∞
1 1 X X Anµ2nmJn(µnmr)Jn (µnmξ)
(
1 (z −ζ)2 1 for n = 0,
G2(z, ζ, t) = √ exp − , An =
2 πat 4at 2 for n = 1, 2, . . . ,

problem.

∂r w + kw = g(ϕ, z, t) at r = R (boundary condition).

∞ X
X ∞
1 Anµ2nmJn(µnmr)Jn(µnmξ)
G1(r, ϕ, ξ, η, t) = cos[n(ϕ−η)] exp(−µ2nmat),
π n=0 m=1
(µnmR2 +k 2R2 −n2)[Jn(µnmR)]2
2
(
1 (z −ζ)2 1 for n = 0,
G2(z, ζ, t) = √ exp − , An =
2 πat 4at 2 for n = 1, 2, . . .
equation
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞. Different boundary value

problems.
1◦ . The solution of the first boundary value problem for a semiinfinite circular cylinder is
given by the formula in Section 5.1.2 (see the first boundary value problem for 0 ≤ r ≤ R,
0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞) with the additional term
Z t Z ∞Z 2πZ R
Φ(ξ, η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ )ξ dξ dη dζ dτ, (1)
0 0 0 0
2◦ . The solution of the second boundary value problem for a semiinfinite circular cylinder
is given by the formula in Section 5.1.2 (see the second boundary value problem for 0 ≤
r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞) with the additional term (1); the Green’s function is also
3◦ . The solution of the third boundary value problem for a semiinfinite circular cylinder is
the sum of the solution presented in Section 5.1.2 (see the third boundary value problem
for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞) and expression (1); the Green’s function is also
4◦ . The solutions of mixed boundary value problems for a semiinfinite circular cylinder
are given by the formulas in Section 5.1.2 (see the mixed boundary value problems for
0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞) with additional terms of the form (1); the Green’s
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l. Different boundary value problems.

1◦ . The solution w(r, ϕ, z, t) of the first boundary value problem for a circular cylinder of
finite length is given by the formula in Section 5.1.2 (see the first boundary value problem
for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l) with the additional term
Z t Z l Z 2πZ R
0 0 0 0
2◦ . The solution w(r, ϕ, z, t) of the second boundary value problem for a circular cylinder
of finite length is given by the formula in Section 5.1.2 (see the second boundary value
problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l) with the additional term (2); the Green’s
∂w
= a∆3 w + Φ(x, y, z, t) 531
3◦ . The solution w(r, ϕ, z, t) of the third boundary value problem for a circular cylinder
of finite length is the sum of the solution presented in Section 5.1.2 (see the third boundary
value problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l) and expression (2); the Green’s
4◦ . The solutions w(r, ϕ, z, t) of mixed boundary value problems for a circular cylinder
of finite length are given by the formulas in Section 5.1.2 (see the mixed boundary value
problems for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l) with additional terms of the form (2); the
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, −∞ < z < ∞. Different boundary value

problems.
1◦ . The solution w(r, ϕ, z, t) of the first boundary value problem for an infinite hollow
cylinder is given by the formula in Section 5.1.2 (see the first boundary value problem for
R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, −∞ < z < ∞) with the additional term
Z t Z ∞ Z 2πZ R2
0 −∞ 0 R1
2◦ . The solution w(r, ϕ, z, t) of the second boundary value problem for an infinite hollow
cylinder is given by the formula in Section 5.1.2 (see the second boundary value problem
for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, −∞ < z < ∞) with the additional term (3); the Green’s
3◦ . The solution w(r, ϕ, z, t) of the third boundary value problem for an infinite hollow
cylinder is the sum of the solution presented in Section 5.1.2 (see the third boundary value
problem for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, −∞ < z < ∞) and expression (3); the Green’s
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞. Different boundary value

problems.
1◦ . The solution w(r, ϕ, z, t) of the first boundary value problem for a semiinfinite hollow
cylinder is given by the formula in Section 5.1.2 (see the first boundary value problem for
R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞) with the additional term
Z t Z ∞Z 2πZ R2
0 0 0 R1
2◦ . The solution w(r, ϕ, z, t) of the second boundary value problem for a semiinfinite hol-
low cylinder is given by the formula in Section 5.1.2 (see the second boundary value prob-
lem for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞) with the additional term (4); the Green’s
3◦ . The solution w(r, ϕ, z, t) of the third boundary value problem for a semiinfinite hollow
cylinder is the sum of the solution presented in Section 5.1.2 (see the third boundary value
problem for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞) and expression (4); the Green’s
4◦ . The solutions w(r, ϕ, z, t) of mixed boundary value problems for a semiinfinite hollow
cylinder are given by the formulas in Section 5.1.2 (see the mixed boundary value problems
for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞) with additional terms of the form (4); the
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l. Different boundary value

problems.
1◦ . The solution w(r, ϕ, z, t) of the first boundary value problem for a hollow cylinder of
finite length is given by the formula in Section 5.1.2 (see the first boundary value problem
for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l) with the additional term
Z t Z l Z 2πZ R2
0 0 0 R1
2◦ . The solution w(r, ϕ, z, t) of the second boundary value problem for a hollow cylinder
of finite length is given by the formula in Section 5.1.2 (see the second boundary value
problem for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l) with the additional term (5); the Green’s
3◦ . The solution w(r, ϕ, z, t) of the third boundary value problem for a hollow cylinder of
finite length is the sum of the solution specified in Section 5.1.2 (see the third boundary
value problem for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l) and expression (5); the Green’s
4◦ . The solutions w(r, ϕ, z, t) of mixed boundary value problems for a hollow cylinder
of finite length are given by the formulas in Section 5.1.2 (see the mixed boundary value
problems for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l) with additional terms of the form (5);
◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , −∞ < z < ∞. Different boundary value

problems.
1◦ . The solution w(r, ϕ, z, t) of the first boundary value problem for an infinite wedge
0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , −∞ < z < ∞) with the additional term
Z t Z ∞ Z ϕ0Z ∞
0 −∞ 0 0
∂w
= a∆3 w + Φ(x, y, z, t) 533
2◦ . The solution w(r, ϕ, z, t) of the second boundary value problem for an infinite wedge
domain is given by the formula in Section 5.1.2 (see the second boundary value problem
for 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , −∞ < z < ∞) with the additional term (6); the Green’s
◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z < ∞. Different boundary value

problems.
1◦ . The solution w(r, ϕ, z, t) of the first boundary value problem for a semiinfinite wedge
0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z < ∞) with the additional term
Z t Z ∞Z ϕ0Z ∞
0 0 0 0
2◦ . The solution w(r, ϕ, z, t) of the second boundary value problem for a semiinfinite
wedge domain is given by the formula in Section 5.1.2 (see the second boundary value
problem for 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z < ∞) with the additional term (7); the Green’s
3◦ . The solutions w(r, ϕ, z, t) of mixed boundary value problems for a semiinfinite wedge
domain are given by the formulas in Section 5.1.2 (see the mixed boundary value problems
for 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z < ∞) with additional terms of the form (7); the Green’s
functions are taken from Section 5.1.2.
◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ l. Different boundary value problems.

1◦ . The solution w(r, ϕ, z, t) of the first boundary value problem for a wedge domain of
finite height is given by the formula in Section 5.1.2 (see the first boundary value problem
for 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ l) with the additional term
Z tZ lZ ϕ0Z ∞
0 0 0 0
2◦ . The solution w(r, ϕ, z, t) of the second boundary value problem for a wedge domain
of finite height is given by the formula in Section 5.1.2 (see the second boundary value
problem for 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ l) with the additional term (8); the Green’s
3◦ . The solutions w(r, ϕ, z, t) of mixed boundary value problems for a wedge domain of
finite height are given by the formulas in Section 5.1.2 (see the mixed boundary value
problems for 0 ≤ r < ∞, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ l) with additional terms of the form (8);
the Green’s functions are taken from Section 5.1.2.
◮ Different boundary value problems for a cylindrical sector.

1◦ . The solution w(r, ϕ, z, t) of the first boundary value problem for an unbounded cylin-
drical sector is given by the formula in Section 5.1.2 (see the first boundary value problem
for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , −∞ < z < ∞) with the additional term
Z t Z ∞ Z ϕ0Z R
Φ(ξ, η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ )ξ dξ dη dζ dτ,
0 −∞ 0 0
2◦ . The solution w(r, ϕ, z, t) of the first boundary value problem for a semibounded cylin-
drical sector is given by the formula in Section 5.1.2 (see the first boundary value problem
for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z < ∞) with the additional term
Z t Z ∞Z ϕ0 Z R
0 0 0 0
which allows for the equation’s nonhomogeneity; the Green’s function is also taken from
Section 5.1.2.
3◦ . The solution w(r, ϕ, z, t) of the mixed boundary value problem for a semibounded
cylindrical sector is given by the formula in Section 5.1.2 (see the mixed boundary value
problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z < ∞) with the additional term (9); the Green’s
4◦ . The solution w(r, ϕ, z, t) of the first boundary value problem for a cylindrical sector of
finite height is given by the formula in Section 5.1.2 (see the first boundary value problem
for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ l) with the additional term
Z t Z l Z ϕ0Z R
0 0 0 0
which allows for the equation’s nonhomogeneity; the Green’s function is also taken from
Section 5.1.2.
5◦ . The solution w(r, ϕ, z, t) of a mixed boundary value problem for a cylindrical sector
of finite height is given by the formula in Section 5.1.2 (see the mixed boundary value
problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ l) with the additional term (10); the Green’s

In the spherical coordinate system, the heat equation with a volume source has the form

∂w 1 ∂ 2 ∂w 1 ∂ ∂w 1 ∂2w
=a 2 r + 2 sin θ + 2 2 + Φ(r, θ, ϕ, t).
∂t r ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂ϕ2
w(r, t) are discussed in Section 3.2.4.
∂w
= a∆3 w + Φ(x, y, z, t) 535

Solution:
Z 2πZ π Z R
0 0 0
Z tZ 2πZ π
2 ∂
−aR g(η, ζ, τ ) G(r, θ, ϕ, ξ, η, ζ, t−τ ) sin η dη dζ dτ
0 0 0 ∂ξ ξ=R
Z tZ 2πZ π Z R
+ Φ(ξ, η, ζ, τ ) G(r, θ, ϕ, ξ, η, ζ, t−τ )ξ 2 sin η dξ dη dζ dτ,
0 0 0 0
where
X∞ X ∞ Xn
1
2πR2 rξ n=0 m=1
k=0
× Pn (cos θ)Pnk (cos η) cos[k(ϕ
k
− ζ)] exp(−λ2nm at),
(
1 for k = 0, (2n + 1)(n − k)!
Ak = Bnmk = ′ 2 .
2 for k =
6 0, (n + k)! Jn+1/2 (λnm R)
dk 1 dn
Pnk (µ) = (1 − µ2 )k/2 Pn (µ), Pn (µ) = (µ2 − 1)n ;
dµk n! 2n dµn
(1984).

Solution:
Z 2πZ π Z R
0 0 0
Z t Z 2πZ π
+ aR2 g(η, ζ, τ ) G(r, θ, ϕ, R, η, ζ, t − τ ) sin η dη dζ dτ
0 0 0
Z t Z 2πZ π Z R
+ Φ(ξ, η, ζ, τ ) G(r, θ, ϕ, ξ, η, ζ, t − τ )ξ 2 sin η dξ dη dζ dτ,
0 0 0 0
where
∞ X
X ∞ X
n
3 1
G(r, θ, ϕ, ξ, η, ζ, t) = + √ Ak Bnmk Jn+1/2 (λnm r)Jn+1/2 (λnm ξ)
4πR3 2π rξ
n=0 m=1 k=0
× Pnk (cos θ)Pnk (cos η) cos[k(ϕ − ζ)] exp(−λ2nm at),
(
1 for k = 0, λ2nm (2n + 1)(n − k)!
Ak = Bnmk = 2 .
the previous paragraph), and the λnm are positive roots of the transcendental equation
′
2λRJn+1/2 (λR) − Jn+1/2 (λR) = 0.

∞ X
X ∞ X
n
1
2π rξ n=0 m=1 s=0
×Pns(cos θ)Pns(cos η) cos[s(ϕ−ζ)] exp(−λ2nmat),
(
1 if s=0, λ2nm(2n+1)(n−s)!
As = Bnms = 2 .
2 if s6= 0, (n+s)! R2λ2nm+(kR+n)(kR−n−1) Jn+1/2(λnmR)
′

λRJn+1/2 (λR) + kR − 12 Jn+1/2 (λR) = 0.
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π. Different boundary value

problems.
1◦ . The solution w(r, θ, ϕ, t) of the first boundary value problem for a spherical layer is
given by the formula in Section 5.1.3 (see the first boundary value problem for R1 ≤ r ≤ R2 ,
0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π) with the additional term
Z t Z 2πZ π Z R2
Φ(ξ, η, ζ, τ ) G(r, θ, ϕ, ξ, η, ζ, t − τ )ξ 2 sin η dξ dη dζ dτ, (1)
0 0 0 R1
geneous equation with homogeneous initial and boundary conditions. The Green’s function
5.3. Other Equations with Three Space Variables 537
2◦ . The solution w(r, θ, ϕ, t) of the second boundary value problem for a spherical layer
is given by the formula in Section 5.1.3 (see the second boundary value problem for R1 ≤
r ≤ R2 , 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π) with the additional term (1); the Green’s function is also
3◦ . The solution w(r, θ, ϕ, t) of the third boundary value problem for a spherical layer is
the sum of the solution specified in Section 5.1.3 (see the third boundary value problem for
R1 ≤ r ≤ R2 , 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π) and expression (1); the Green’s function is also
◮ Domain: 0 ≤ r < ∞, 0 ≤ θ ≤ θ0 , 0 ≤ ϕ ≤ 2π. First boundary value problem.

The solution w(r, θ, ϕ, t) of the first boundary value problem for an infinite cone is given
by the formula in Section 5.1.3 (see the first boundary value problem for 0 ≤ r < ∞,
0 ≤ θ ≤ θ0 , 0 ≤ ϕ ≤ 2π) with the additional term
Z t Z 2πZ θ0Z ∞
Φ(ξ, η, ζ, τ ) G(r, θ, ϕ, ξ, η, ζ, t − τ )ξ 2 sin η dξ dη dζ dτ,
0 0 0 0
geneous equation with homogeneous initial and boundary conditions. The Green’s function
5.3 Other Equations with Three Space Variables


∂w ∂2w ∂2w ∂2w
1. =a + + + (b1 x + b2 y + b3 z + c)w.
∂t ∂x2 ∂y 2 ∂z 2
The transformation

w(x, y, z, t) = exp (b1 x + b2 y + b3 z)t + 13 a(b21 + b22 + b23 )t3 + ct u(ξ, η, ζ, t),
ξ = x + ab1 t2 , η = y + ab2 t2 , ζ = z + ab3 t2
leads to the three-dimensional heat equation ∂t u = a(∂ξξ u + ∂ηη u + ∂ζζ u) that is dealt
with in Section 5.1.1.
2
∂w ∂ w ∂2w ∂2w
2. =a 2
+ 2
+ 2
− b(x2 + y 2 + z 2 ) + c w, b > 0.
∂t ∂x ∂y ∂z
The transformation (A is any number)
√
ab 2 √
w(x, y, z, t) = exp (x + y 2 + z 2 ) + 3 ab − c t u(ξ, η, ζ, τ ),
2a
√ √
ξ = x exp 2 ab t , η = y exp 2 ab t ,
√ 1 √
ζ = z exp 2 ab t , τ = √ exp 4 ab t + A
4 ab
leads to the three-dimensional heat equation ∂τ u = a(∂ξξ u + ∂ηη u + ∂ζζ u) that is dealt

∂w ∂2w ∂2w ∂2w
3. =a + + + −b(x2 + y 2 + z 2 ) + c1 x + c2 y + c3 z + s w.
∂t ∂x2 ∂y 2 ∂z 2
This is a special case of equation 5.3.2.3 with fk (t) = ck , g(t) = s.
2
∂w ∂ w ∂2w ∂2w ∂w ∂w ∂w
4. =a 2
+ 2
+ 2
+ b1 + b2 + b3 + cw.
∂t ∂x ∂y ∂z ∂x ∂y ∂z
This equation governs the nonstationary temperature (concentration) field in a medium
moving with a constant velocity, provided there is volume release (absorption) of heat
proportional to temperature (concentration).
The substitution

w(x, y, z, t) = exp A1 x + A2 y + A3 z + Bt U (x, y, z, t),
where
b1 b2 b3 1 2
A1 = − , A2 = − , A3 = − , B =c− b1 + b22 + b23 ,
2a 2a 2a 4a
leads to the three-dimensional heat equation ∂t U = a∆3 U that is considered in Sec-
tion 5.1.1.
2
∂w ∂ w ∂2w ∂2w ∂w
5. =a 2
+ 2
+ 2
− by .
∂t ∂x ∂y ∂z ∂x
This equation is encountered in problems of convective heat and mass transfer in a simple
shear flow.
1
E (x, y, z, ξ, η, ζ, t) =
1 2 2 1/2
(4πat)3/2 1 + 12 b t
2
x − ξ − 12 bt(y + η) (y − η)2 + (z − ζ)2
× exp − 1 2 2
− .
4at 1 + 12 b t 4at
⊙ Literature: E. A. Novikov (1958).

∂w ∂ 2w ∂ 2w ∂ 2w ∂w ∂w ∂w
6. =a + + + b1 x + b2 y + b3 z + Φ(x, y, z, t).
∂t ∂x2 ∂y 2 ∂z 2 ∂x ∂x ∂z
This equation is encountered in problems of convective heat and mass transfer in a linear
shear flow.
Domain: −∞ < x < ∞, −∞ < y < ∞, −∞ < z < ∞. Cauchy problem.
w = f (x, y, z) at t = 0.
Solution:
Z ∞ Z ∞ Z ∞
w(x, y, z, t) = f (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dx dy dz
−∞ −∞ −∞
Z tZ ∞ Z ∞ Z ∞
+ Φ(ξ, η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dx dy dz dτ,
0 −∞ −∞ −∞
where
G(x, y, z, ξ, η, ζ, t) = H(x, ξ, t; b1 )H(y, η, t; b2 )H(z, ζ, t; b3 ),

−1/2 2
2πa 2bt b xebt − ξ
H(x, ξ, t; b) = (e − 1) exp − .
b 2a(e2bt − 1)

∂w ∂2w ∂2w ∂2w ∂w ∂w
7. =a + + + (b1 x + c1 ) + (b2 y + c2 )
∂t ∂x2 ∂y 2 ∂z 2 ∂x ∂y
∂w
+ (b3 z + c3 ) + (s1 x + s2 y + s3 z + p)w.
∂z
2
∂w ~2 ∂ w ∂2w ∂2w
8. i~ + + + = 0.
∂t 2m 2 ∂x 2 ∂y 2 ∂z
Three-dimensional Schrödinger equation, i2 = −1.

i m 3/2 m 2 2 2 3π
E (x, y, z, t) = − exp i (x + y + z ) − i .
~ 2π~t 2~t 4

∂w ∂ ∂w ∂ ∂w ∂ ∂w
9. = axn + by m + cz k .
∂t ∂x ∂x ∂y ∂y ∂z ∂z
This equation describes unsteady heat and mass transfer processes in inhomogeneous (anis-
otropic) media. It admits separable solutions, as well as solutions with incomplete sepa-
ration of variables (see Section 17.5.2). In addition, for n 6= 2, m 6= 2, k 6= 2 there are
particular solutions of the form

2 x2−n y 2−m z 2−k
w = w(ξ, t), ξ =4 + + ,
a(2 − n)2 b(2 − m)2 c(2 − k)2
where the function w(ξ, t) is determined by the one-dimensional nonstationary equation

∂w ∂2w A ∂w 1 1 1
= + , A = 2 + + − 1.
∂t ∂ξ 2 ξ ∂ξ 2−n 2−m 2−k
For solutions of this equation, see Sections 3.2.1, 3.2.3, and 3.2.5.

∂w ∂2w ∂2w ∂2w
1. =a + + + f (t)w.
∂t ∂x2 ∂y 2 ∂z 2
This equation describes three-dimensional unsteady thermal phenomena in quiescent media
or solids with constant thermal diffusivity, provided there is unsteady volume heat release
proportional to temperature. R
The substitution w(x, y, z, t) = exp f (t) dt U (x, y, z, t) leads to the usual heat equa-
tion ∂t U = a∆3 U that is dealt with in Section 5.1.1.

∂w ∂2w ∂2w ∂2w
2. =a + + + xf1 (t) + yf2 (t) + zf3 (t) + g(t) w.
∂t ∂x2 ∂y 2 ∂z 2
The transformation

w(x, y, z, t) = u(ξ, η, ζ, t) exp xF1 (t) + yF2 (t) + zF3 (t)
Z Z
2 2 2

+a F1 (t) + F2 (t) + F3 (t) dt + g(t) dt ,
Z Z
ξ = x + 2a F1 (t) dt, η = y + 2a F2 (t) dt,
Z Z
ζ = z + 2a F3 (t) dt, Fk (t) = fk (t) dt,

leads to the three-dimensional heat equation ∂t u = a ∂ξξ u + ∂ηη u + ∂ζζ u that is dealt
2
∂w ∂ w ∂2w ∂2w
3. =a + +
∂t ∂x2 ∂y 2 ∂z 2

+ −b(x + y 2 + z 2 ) + xf1 (t) + yf2 (t) + zf3 (t) + g(t) w.
2

√
ab 2 2 2
w(x, y, z, t)=u(ξ, η, ζ, τ ) exp x +y +z ,
2a
√ √
ξ =x exp 2 ab t , η=y exp 2 ab t ,
√ 1 √
ζ =z exp 2 ab t , τ = √ exp 4 ab t +A,
4 ab
where A is an arbitrary constant, leads to an equation of the form 5.3.2.2:

2
∂u ∂ u ∂2u ∂2u
=a 2
+ 2
+ 2
+ ξF1 (τ ) + ηF2 (τ ) + ζF3 (τ ) + G(τ ) u,
∂t ∂ξ ∂η ∂ζ

1 ln(cτ ) 1 ln(cτ ) 3 √
Fk (τ ) = f k , G(τ ) = g + , c = 4 ab, k = 1, 2, 3.
(cτ )3/2 c cτ c 4τ

√
−ab 2 2 2
√
w(x, y, z, t) = v(ξ1 , η1 , ζ1 , τ1 ) exp (x + y + z ) tan 2 −ab t ,
2a
x y
ξ1 = √ , η1 = √ ,
cos 2 −ab t cos 2 −ab t
z 1 √
ζ1 = √ , τ1 = √ tan 2 −ab t
cos 2 −ab t 2 −ab
also leads to an equation of the form 5.3.2.2 for v = v(ξ1 , η1 , ζ1 , τ1 ) (the equation for v is
not written out here).
∂w ∂ 2w ∂ 2w ∂ 2w
4. = a1 (t) + a2 (t) + a3 (t) + Φ(x, y, z, t).
∂t ∂x2 ∂y 2 ∂z 2
Here, 0 < ak (t) < ∞; k = 1, 2, 3.
Domain: −∞ < x < ∞, −∞ < y < ∞, −∞ < z < ∞. Cauchy problem.
w = f (x, y, z) at t = 0.
Solution:
Z tZ ∞Z ∞Z ∞
w(x, y, z, t) = Φ(ξ, η, ζ, τ ) G(x, y, z, ξ, η, ζ, t, τ ) dξ dη dζ dτ
0 −∞ −∞ −∞
Z ∞Z ∞Z ∞
+ f (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t, 0) dξ dη dζ,
−∞ −∞ −∞
where

1 (x − ξ)2 (y − η)2 (z − ζ)2
G(x, y, z, ξ, η, ζ, t, τ ) = √ exp − − − ,
8π 3/2 T1 T2 T3 4T1 4T2 4T3
Z t Z t Z t
T1 = a1 (σ) dσ, T2 = a2 (σ) dσ, T3 = a3 (σ) dσ.
τ τ τ
See also the more general equation 5.4.3.3, where other boundary value problems are
considered.
∂w ∂ 2w ∂ 2w ∂ 2w
5. = a1 (t) 2 + a2 (t) 2 + a3 (t) 2
∂t ∂x ∂y ∂z
∂w ∂w ∂w
+ b1 (t)x + c1 (t) + b2 (t)y + c2 (t) + b3 (t)z + c3 (t)
∂x ∂y
∂z
+ s1 (t)x + s2 (t)y + s3 (t)z + p(t) w.
The transformation

w(x, y, z, t) = exp f1 (t)x + f2 (t)y + f3 (t)z + g(t) u(ξ, η, ζ, t),
ξ = h1 (t)x + r1 (t), η = h2 (t)y + r2 (t), ζ = h3 (t)z + r3 (t),
where Z
Z
sk (t)
hk (t)
Z

Z X
3 Z

g(t) = ak (t)fk2 (t) + ck (t)fk (t) dt + p(t) dt + D,
k=1
(k = 1, 2, 3; Ak , Bk , Ck , D are arbitrary constants) leads to an equation of the form
5.3.2.4:
∂u ∂2u ∂2u ∂2u
= a1 (t)h21 (t) 2 + a2 (t)h22 (t) 2 + a3 (t)h23 (t) 2 .
∂t ∂ξ ∂η ∂ζ
∂w ∂ 2w ∂ 2w ∂ 2w
6. = a1(t) 2
+a2(t) 2 +a3(t) 2
∂t ∂x ∂y ∂z
∂w ∂w ∂w
+ b1(t)x+c1(t) + b2(t)y+c2(t) + b3(t)z+c3(t)
∂x ∂y ∂z

+ s1(t)x2 +s2(t)y 2 +s3(t)z 2 +p1(t)x+p2(t)y+p3(t)z+q(t) w.
The substitution

w(x, y, z, t) = exp f1 (t)x2 + f2 (t)y 2 + f3 (t)z 2 u(x, y, z, t),
where the functions fk = fk (t) are solutions of the respective Riccati equations
fk′ = 4ak (t)fk2 + 2bk (t)fk + sk (t) (k = 1, 2, 3),
leads to an equation of the form 5.3.2.5 for u = u(x, y, z, t).
2
X 2
X
∂w ∂w ∂w ∂ 2w ∂ 2w
7. + fn(t)+gn(t)x3 −a = Knm(t) +K33(t) .
∂t ∂xn ∂x3 ∂xn∂xm ∂x23
n=1 n,m=1
Equation of turbulent diffusion. It describes the diffusion of an admixture in a horizontal

stream whose velocity components are linear functions of the height.
3
1 1 X −1
E (x1 , x2 , x3 , ξ1 , ξ2 , ξ3 ) = p exp − Tij (t)yi yj ,
(4π)3/2 det |T | 4
i,j=1
Z t
Tij (t) = Sij (τ ) dτ.
0
Here, the following notation is used (n, m = 1, 2):

Z t
yn = xn − ξn − Fn (t) − x3 Gn (t) − a (t − τ )gn (τ ) dτ, y3 = x3 − ξ3 + at,
0
Snm (t) = Knm (t) + K33 (t) Gn (t) Gm (t), Sn3 (t) = S3n (t) = −K33 (t) Gn (t),
Z t Z t
S33 (t) = K33 (t), Fn (t) = fn (t) dt, Gn (t) = gn (t) dt,
0 0
where det |T | is the determinant of the matrix T with entries Tij (t) and Tij−1 (t) are the
2 (t) > 0, and
entries of the inverse of T. The inequalities T11 (t) > 0, T11 (t)T22 (t) − T12
det |T | > 0 are assumed to hold.
⊙ Literature: E. A. Novikov (1958).

∂w
ρ(x, y, z) = div[a(x, y, z)∇w] − q(x, y, z)w + Φ(x, y, z, t)
∂t
Equations of this form are often encountered in the theory of heat and mass transfer. For
brevity, the following notation is used:

∂ ∂w ∂ ∂w ∂ ∂w
div a(r)∇w = a(r) + a(r) + a(r) , r = {x, y, z}.
∂x ∂x ∂y ∂y ∂z ∂z
The problems presented in this subsection are assumed to refer to a simply connected
bounded domain V with smooth boundary S. It is also assumed that ρ(r) > 0, a(r) > 0,
and q(r) ≥ 0.
◮ First boundary value problem.
w = f (r) at t=0 (initial condition),

w = g(r, t) for r ∈ S (boundary condition).
Solution:
Z tZ Z
w(r, t) = Φ(ξ, τ )G(r, ξ, t−τ ) dVξ dτ + f (ξ)ρ(ξ)G(r, ξ, t) dVξ
0 V V
Z tZ
∂
− g(ξ, τ )a(ξ) G(r, ξ, t−τ ) dSξ dτ. (1)
0 S ∂Nξ
Here, the modified Green’s function is given by
X∞ Z
un (r)un (ξ) 2
G(r, ξ, t) = exp(−λn t), kun k = ρ(r)u2n (r) dV, ξ = {ξ1 , ξ2 , ξ3 },
n=1
kun k2 V
(2)
where the λn and un (r) are the eigenvalues and corresponding eigenfunctions of the Sturm–
Liouville problem for the following elliptic second-order equation with a homogeneous
boundary condition of the first kind:

div a(r)∇u − q(r)u + λρ(r)u = 0, (3)
u = 0 for r ∈ S. (4)
∂
The integration in solution (1) is carried out with respect to ξ1 , ξ2 , ξ3 ; ∂N ξ
denotes the
derivative along the outward normal to the surface S with respect to ξ1 , ξ2 , and ξ3 .
General properties of the Sturm–Liouville problem (3)–(4):
1◦ . There are countably many eigenvalues. All eigenvalues are real and can be ordered so
that λ1 ≤ λ2 ≤ λ3 ≤ · · · , with λn → ∞ as n → ∞; consequently, there can exist only
finitely many negative eigenvalues.
2◦ . For ρ(r) > 0, a(r) > 0, and q(r) ≥ 0, all eigenvalues are positive: λn > 0.
3◦ . The eigenfunctions are defined up to a constant multiplier. Any two eigenfunctions
un (r) and um (r) corresponding to different eigenvalues λn and λm are orthogonal with
weight ρ(r) in the domain V :
Z
ρ(r)un (r)um (r) dV = 0 for n 6= m.
V
4◦ . An arbitrary function F (r) that is twice continuously differentiable and satisfies the
boundary condition of the Sturm–Liouville problem (F = 0 for r ∈ S) can be expanded
into an absolutely and uniformly convergent series in the eigenvalues:
∞
X Z
1
F (r) = Fn un (r), Fn = F (r)ρ(r)un (r) dV,
kun k2 V
n=1
where the formula for kun k2 is given in (2).

Remark 5.1. In a three-dimensional problem, to each eigenvalue λn there generally correspond
(1) (2) (m)
finitely many linearly independent eigenfunctions un , un , . . . , un . These function can always
be replaced by their linear combinations
ū(k) (1) (k−1)

n = Ak,1 un + · · · + Ak,k−1 un + u(k)
n , k = 1, 2, . . . , m,
(1) (2) (m)
such that ūn , ūn , . . . , ūn are now pairwise orthogonal. Thus, without loss of generality, we
assume that all eigenfunctions are orthogonal.
◮ Second boundary value problem.


∂w
= g(r, t) for r ∈ S (boundary condition).
∂N
Solution:
Z tZ Z
w(r, t) = Φ(ξ, τ )G(r, ξ, t − τ ) dVξ dτ + f (ξ)ρ(ξ)G(r, ξ, t) dVξ
0 V V
Z tZ
+ g(ξ, τ )a(ξ)G(r, ξ, t − τ ) dSξ dτ. (5)
0 S
Here, the modified Green’s function is given by (2), where the λn and un (r) are the eigen-
values and corresponding eigenfunctions of the Sturm–Liouville problem for the elliptic
second-order equation (3) with a homogeneous boundary condition of the second kind,
∂u
= 0 for r ∈ S. (6)
∂N
For q(r) > 0 the general properties of the eigenvalue problem (3), (6) are the same as
for the first boundary value problem (with all λn > 0).
◮ Third boundary value problem.

w = f (r) at t=0 (initial condition),

∂w
+ k(r)w = g(r, t) for r ∈ S (boundary condition).
∂N
5.4. Equations with n Space Variables 545
The solution of the third boundary value problem is given by formulas (5) and (2),
Liouville problem for the second-order elliptic equation (3) with a homogeneous boundary
condition of the third kind,
∂u
+ k(r)u = 0 for r ∈ S. (7)
∂N
For q(r) ≥ 0 and k(r) > 0 the general properties of the eigenvalue problem (3), (7) are
the same as for the first boundary value problem.
Let k(r) = k = const. Denote the Green’s functions of the second and third boundary
value problems by G2 (r, ξ, t) and G3 (r, ξ, t, k), respectively. If q(r) > 0, then the following
limiting relation holds: G2 (r, ξ, t) = lim G3 (r, ξ, t, k).
k→0
⊙ Literature for Section 5.3.3: V. S. Vladimirov (1988), A. D. Polyanin (2000a, 2000c).
5.4 Equations with n Space Variables

∂w
5.4.1 Equations of the Form = a∆n w + Φ(x1 , . . . , xn , t)
∂t
This is an n-dimensional nonhomogeneous heat equation. In the Cartesian system of coor-
dinates, it is represented as
X ∂2w n
∂w
=a + Φ(x, t), x = {x1 , . . . , xn }.
∂t ∂x2k
k=1
The solutions of various problems for this equation can be constructed on the basis of
incomplete separation of variables (see Sections 16.2.2 and 17.5.2) taking into account the
results of Sections 3.1.1 and 3.1.2. Some examples of solving such problems can be found
below.

X
n n
X
2
w(x, t) = A exp km xm + at km ,
m=1 m=1
n
X n
Y
2
w(x, t) = A exp −at km cos(km xm + Cm ),
m=1 m=1
n
X Yn
2
w(x, t) = A exp − km xm cos(km xm − 2akm t + Cm ),
m=1 m=1
Xn
A 1 2
w(x, t) = exp − (xm − Cm ) ,
(t − t0 )n/2 4a(t − t0 )
m=1
n
Y
xm − Cm
w(x, t) = A erf √ ,
m=1
2 at
where x = {x1 , . . . , xn }; A, km , Cm , and t0 are arbitrary constants.

n
X
1 |x|2
E (x, t) = √ n exp − , |x|2 = x2k .
2 πat 4at
k=1
3◦ . Suppose w = w(x1 , . . . , xn , t) is a solution of the homogeneous equation. Then the

functions
w1 =Aw(±λx1+C1, . . . , ±λxn+Cn, λ2t+Cn+1),

X
n Xn
2
w2 =A exp λkxk+at λk w(x1+2aλ1t+C1, . . . , xn+2aλnt+Cn, t+Cn+1),
k=1 k=1
Xn
A β 2 x1 xn γ+λt
w3 = exp − xk w ,..., , , λδ−βγ =1,
|δ+βt|n/2 4a(δ+βt)
k=1
δ+βt δ+βt δ+βt
where A, C1 , . . . , Cn+1 , λ, λ1 , . . . , λn β, and δ are arbitrary constants, are also solutions

of the equation. The signs at λ in the formula for w1 can be taken independently of one
another.
◮ Domain: Rn = {−∞ < xk < ∞; k = 1, . . . , n}. Cauchy problem.

w = f (x) at t = 0.
Solution:
Z
1 |x − y|2
w(x, t) = √ n f (y) exp − dVy
2 πat Rn 4at
Z tZ
Φ(y, τ ) |x − y|2
+ p n exp − dVy dτ,
0 Rn 2 πa(t − τ ) 4a(t − τ )
where the following notation is used:

p
y = {y1 , . . . , yn }, |x − y| = (x1 − y1 )2 + · · · + (xn − yn )2 , dVy = dy1 dy2 . . . dyn .
⊙ Literature: V. S. Vladimirov (1988).
◮ Domain: V = {0 ≤ xk ≤ lk ; k = 1, . . . , n}. First boundary value problem.


w = gk (x, t) at xk = 0 (boundary conditions),
w = hk (x, t) at xk = lk (boundary conditions).
Solution:
Z tZ Z
w(x, t) = Φ(y, τ ) G(x, y, t − τ ) dVy dτ + f (y) G(x, y, t) dVy
0 V V
n Z tZ
X
∂
+a gk (y, τ ) G(x, y, t − τ ) dSy(k) dτ
0 S (k) ∂yk y =0
k=1 k
Xn Z tZ
∂
−a hk (y, τ ) G(x, y, t − τ ) dSy(k) dτ,
0 S (k) ∂yk yk =lk
k=1
dSy(k) = dy1 . . . dyk−1 dyk+1 . . . dyn ,
S (k) = {0 ≤ ym ≤ lm for m = 1, . . . , k − 1, k + 1, . . . , n}.
The Green’s function can be represented in the product form
n
Y
G(x, y, t) = Gk (xk , yk , t), (1)
k=1
where the Gk (xk , yk , t) are the Green’s functions of the respective one-dimensional bound-
ary value problems (see the first boundary value problem for 0 ≤ x ≤ l in Section 3.1.2):
∞
2 X mπx mπξ am2 π 2 t
Gk (xk , yk , t) = sin sin exp − 2 .
lk lk lk lk
m=1
◮ Domain: V = {0 ≤ xk ≤ lk ; k = 1, . . . , n}. Second boundary value problem.

∂xk w = gk (x, t) at xk = 0 (boundary conditions),
∂xk w = hk (x, t) at xk = lk (boundary conditions).
Solution:
Z tZ Z
0 V V
n Z tZ
X
−a gk (y, τ ) G(x, y, t − τ ) y dSy(k) dτ
k =0
k=1 0 S (k)
n Z tZ
X
+a hk (y, τ ) G(x, y, t − τ ) y dSy(k) dτ. (2)
k =lk
k=1 0 S (k)
The Green’s function can be represented as the product (1) of the corresponding one-
dimensional Green’s functions of the form (see the second boundary value problem for
0 ≤ x ≤ l in Section 3.1.2):
∞
1 2 X mπx mπξ am2 π 2 t
Gk (xk , yk , t) = + cos cos exp − .
lk lk lk
m=1
lk lk2
5.4.2 Other Equations Containing Arbitrary Parameters

n
X n
X
∂w ∂2w
1. =a + c+ bk xk w.
∂t ∂x2k
k=1 k=1
This is a special case of equation 5.4.3.1. The transformation

Xn n
X
1 3 2
w(x1 , . . . , xn , t) = exp t bk xk + 3 at bk +ct u(ξ1 , . . . , ξn , t), ξk = xk +abk t2
k=1 k=1
n
P
leads to the n-dimensional heat equation ∂t u = a ∂ξk ξk u that is dealt with in Sec-
k=1
tion 5.4.1.
n
X n
X
∂w ∂2w
2. =a − c+b x2k w, b > 0.
∂t ∂x2k
k=1 k=1
The transformation (A is any number)

r X n
1 b 2
√
w(x1 , . . . , xn , t) = u(ξ1 , . . . , ξn , τ ) exp xk + n ab − c t ,
2 a
k=1
√ √ 1 √
ξ1 = x1 exp 2 ab t , . . . , ξn = xn exp 2 ab t , τ = √ exp 4 ab t + A
4 ab
n
P
leads to the n-dimensional heat equation ∂τ u = a ∂ξk ξk u that is dealt with in Sec-
k=1
tion 5.4.1.
n
X n
X n
X
∂w ∂2w
3. =a + −b x2k + ck xk + s w.
∂t ∂x2k
k=1 k=1 k=1
This is a special case of equation 5.4.3.2 with fk (t) = ck and g(t) = s.

n
X n
X
∂w ∂2w ∂w
4. =a + bk + cw.
∂t ∂x2k ∂xk
k=1 k=1
The substitution
n n
1 X 1 X 2
w(x1 , . . . , xn , t) = exp At − bk xk U (x1 , . . . , xn , t), A=c− bk ,
2a 4a
k=1 k=1
leads to the n-dimensional heat equation ∂t U = a∆n U that is dealt with in Section 5.4.1.
n
X n
X ∂w X
n
∂w ∂2w
5. =a + bk xk + ck + sk xk + p w.
∂t ∂x2k ∂xk
k=1 k=1 k=1

n
X
∂w ~2 ∂2w
6. i~ + = 0.
∂t 2m ∂x2k
k=1
This is the n-dimensional Schrödinger equation, i2 = −1.

i m n/2 m 2 πn
E (x, t) = − exp i |x| − i , |x|2 = x21 + · · · + x2n .
~ 2π~t 2~t 4

n
X hX
n i
∂w ∂2w
1. =a + xk fk (t) + g(t) w.
∂t ∂x2k
k=1 k=1
The transformation
X
n n Z
X
w(x1 , . . . , xn , t) = exp xk Fk (t) + a Fk2 (t) dt + G(t) u(ξ1 , . . . , ξn , t),
k=1 k=1
Z Z Z
ξk = xk + 2a Fk (t) dt, Fk (t) = fk (t) dt, G(t) = g(t) dt,
n
P
leads to the n-dimensional heat equation ∂t u = a ∂ξk ξk u that is discussed in Sec-
k=1
tion 5.4.1.
n
X h n
X n
X i
∂w ∂2w
2. =a + −b x2k + xk fk (t) + g(t) w.
∂t ∂x2k
k=1 k=1 k=1
r X n
1 b 2
w(x1 , . . . , xn , t) = u(ξ1 , . . . , ξn , τ ) exp xk ,
2 a
k=1
√ √ 1 √
ξ1 = x1 exp 2 ab t , . . . , ξn = xn exp 2 ab t , τ = √ exp 4 ab t + C,
4 ab
where C is an arbitrary constant, leads to an equation of the form 5.4.3.1:
X n X n
∂u ∂2u
=a + ξk Fk (τ ) + G(τ ) u,
∂t ∂ξk2
k=1 k=1

1 ln(sτ ) 1 ln(sτ ) n √
Fk (τ ) = f k , G(τ ) = g + , s = 4 ab.
(sτ )3/2 s sτ s 4τ
√ n
−b √ X 2
w(x1 , . . . , xn , t) = v(z1 , . . . , zn , τ ) exp √ tan 2 −ab t xk ,
2 a
k=1
x x 1 √
z1 = √1 , . . . , zn = √n , τ = √ tan 2 −ab t
cos 2 −ab t cos 2 −ab t 2 −ab
also leads to an equation of the form 5.4.3.1 (this equation is not specified here).
n
X
∂w ∂ 2w
3. = ak (t) + Φ(x1 , . . . , xn , t).
∂t ∂x2k
k=1
The solutions of various problems for this equation can be constructed on the basis of
incomplete separation of variables (see Sections 16.2.2 and 17.5.2) taking into account the
results of Sections 3.1.1 and 3.1.2. Some examples of solving such problems are given
below. It is assumed that 0 < ak (t) < ∞, k = 1, . . . , n.
1◦ . Domain: Rn = {−∞ < xk < ∞; k = 1, . . . , n}. Cauchy problem.
w = f (x) at t = 0.
Solution:
Z tZ Z
w(x, t) = Φ(y, τ ) G(x, y, t, τ ) dVy dτ + f (y) G(x, y, t, 0) dVy ,
0 Rn Rn
where
X n Z t
1 (xk − yk )2
G(x, y, t, τ ) = n n/2 √ exp − , Tk = ak (η) dη,
2 π T1 T2 . . . Tn 4Tk τ
k=1
x = {x1 , . . . , xn }, y = {y1 , . . . , yn }, dVy = dy1 dy2 . . . dyn .
2◦ . Domain: V = {0 ≤ xk ≤ lk ; k = 1, . . . , n}. First boundary value problem.

Solution:
Z tZ Z
w(x, t) = Φ(y, τ ) G(x, y, t, τ ) dVy dτ + f (y) G(x, y, t) dVy
0 V V
n Z tZ
X
∂
+ ak (τ ) gk (y, τ ) G(x, y, t, τ ) dSy(k) dτ
0 S (k) ∂yk yk =0
k=1
Xn Z tZ
∂
− ak (τ ) hk (y, τ ) G(x, y, t, τ ) dSy(k) dτ,
0 S (k) ∂y k yk =lk
k=1

S (k) = {0 ≤ ym ≤ lm for m = 1, . . . , k − 1, k + 1, . . . , n}.

n
Y
G(x, y, t, τ ) = Gk (xk , yk , t, τ ), (1)
k=1
where the Gk (xk , yk , t, τ ) are the Green’s functions of the respective boundary value prob-
lems,
∞ mπx mπy m2π 2T Z t
2 X k k k
Gk(xk, yk, t, τ )= sin sin exp − , Tk = ak(σ) dσ.
lk
m=1
lk lk lk2 τ
(2)
3◦ . Domain: V = {0 ≤ xk ≤ lk ; k = 1, . . . , n}. Second boundary value problem.


Solution:
Z tZ Z
w(x, t) = Φ(y, τ ) G(x, y, t, τ ) dVy dτ + f (y) G(x, y, t) dVy
0 V V
n Z tZ
X
− ak (τ ) gk (y, τ ) G(x, y, t, τ ) y dSy(k) dτ
k =0
k=1 0 S (k)
Xn Z t Z

+ ak (τ ) hk (y, τ ) G(x, y, t, τ ) y dSy(k) dτ.
k =lk
k=1 0 S (k)
dimensional Green’s functions
∞
1 2 X mπxk mπyk m2 π 2 Tk
Gk (xk , yk , t, τ ) = + cos cos exp − 2 ,
lk lk lk lk lk
m=1
Z t
Tk = ak (σ) dσ.
τ
⊙ Literature: A. D. Polyanin (2000a, 2000b).
n
X n
X ∂w hX
n i
∂w ∂ 2w
4. = ak (t) 2 + bk (t)xk + ck (t) + sk (t)xk + p(t) w.
∂t ∂xk ∂xk
k=1 k=1 k=1

X
n
w(x1 , . . . , xn , t) = exp fk (t)xk + g(t) u(z1 , . . . , zn , t), zk = hk (t)xk + rk (t),
k=1
where the functions fk (t), g(t), hk (t), and rk (t) are given by the following expressions
(Ak , Bk , Ck , and D are arbitrary constants):

Z
Z
sk (t)
hk (t)
Z

Z X n Xn
2
g(t) = p(t) + ak (t)fk (t) + ck (t)fk (t) dt + D.
k=1 k=1
As a result, we arrive at an equation of the form 5.4.3.3 for the new dependent variable
u = u(z1 , . . . , zn , t):
n
∂u X ∂2u
= ak (t)h2k (t) 2 .
∂t ∂zk
k=1
n
X n
X
∂w ∂ 2w ∂w
5. = ak (t) + bk (t)xk + ck (t)
∂t ∂x2k ∂xk
k=1 k=1
hX
n n
X i
+ sk (t)x2k + pk (t)xk + q(t) w.
k=1 k=1
The substitution
X
n
w(x1 , . . . , xn , t) = exp fk (t)x2k u(x1 , . . . , xn , t),
k=1
where the functions fk = fk (t) are solutions of the Riccati equation
fk′ = 4ak (t)fk2 + 2bk (t)fk + sk (t) (k = 1, . . . , n),
leads to an equation of the form 5.4.3.4 for u = u(x1 , . . . , xn , t).
n h
X i
∂w ∂ 2w ∂w
6. − ak (xk , t) + bk (xk , t) + ck (xk , t)w = Φ(x1 , . . . , xn , t).
∂t ∂x2k ∂xk
k=1
Here, 0 < ak (xk , t) < ∞ for all k. We introduce the notation x = {x1 , . . . , xn }, y =
{y1 , . . . , yn } and consider the domain V = {αk ≤ xk ≤ βk , k = 1, . . . , n}, which is an
n-dimensional parallelepiped.
1◦ . First boundary value problem. The following conditions are prescribed:

w = gk (x, t) at xk = αk (boundary conditions),
w = hk (x, t) at xk = βk (boundary conditions).
Solution:
Z tZ Z
w(x, t) = Φ(y, τ ) G(x, y, t, τ ) dVy dτ + f (y) G(x, y, t, 0) dVy
0 V V
n Z tZ
X
∂
+ ak (αk , τ ) gk (y, τ ) G(x, y, t, τ ) dSy(k) dτ
(k) ∂y k
k=1 0 S yk =αk
n
X tZ Z
∂
− ak (βk , τ ) hk (y, τ ) G(x, y, t, τ ) dSy(k) dτ,
0 S (k) ∂y k yk =βk
k=1
where
dVy = dy1 dy2 . . . dyn , dSy(k) = dy1 . . . dyk−1 dyk+1 . . . dyn ,
S (k) = {αm ≤ ym ≤ βm for m = 1, . . . , k−1, k+1, . . . , n}.

n
Y
G(x, y, t, τ ) = Gk (xk , yk , t, τ ). (1)
k=1
Here, the Gk = Gk (xk , yk , t, τ ) are auxiliary Green’s functions that, for t > τ ≥ 0, satisfy
the one-dimensional linear homogeneous equations
∂Gk ∂ 2 Gk ∂Gk
− ak (xk , t) 2 − bk (xk , t) − ck (xk , t) Gk = 0 (k = 1, . . . , n) (2)
∂t ∂xk ∂xk
with nonhomogeneous initial conditions of a special form,
Gk = δ(xk − yk ) at t = τ, (3)
and homogeneous boundary conditions of the first kind,
Gk = 0 at xk = αk ,
Gk = 0 at xk = βk .
In determining the function Gk , the quantities yk and τ play the role of parameters; δ(x) is
the Dirac delta function.
2◦ . The second and third boundary value problems. The following conditions are pre-
scribed:

∂xk w − sk w = gk (x, t) at xk = αk (boundary conditions),
∂xk w + pk w = hk (x, t) at xk = βk (boundary conditions).
The second boundary value problem corresponds to sk = pk = 0.

Solution:
Z tZ Z
w(x, t) = Φ(y, τ ) G(x, y, t, τ ) dVy dτ + f (y) G(x, y, t, 0) dVy
0 V V
n Z tZ
X
− ak (αk , τ ) gk (y, τ ) G(x, y, t, τ ) y dSy(k) dτ
k =αk
k=1 0 S (k)
Xn Z tZ

+ ak (βk , τ ) hk (y, τ ) G(x, y, t, τ ) y dSy(k) dτ.
k =βk
k=1 0 S (k)
dimensional Green’s functions satisfying the linear equations (2) with the initial condi-
tions (3) and the homogeneous boundary conditions
∂xk Gk − sk Gk = 0 at xk = αk ,
∂xk Gk + pk Gk = 0 at xk = βk .
⊙ Literature: A. D. Polyanin (2000a, 2000b).
n
X h i
∂w ∂ ∂w
7. = aij (x1 , . . . , xn ) −q(x1 , . . . , xn )w +Φ(x1, . . . , xn , t).
∂t ∂xi ∂xj
i,j=1
The problems considered below are assumed to refer to a bounded domain V with smooth
surface S. We introduce the brief notation x = {x1 , . . . , xn } and assume that the condition
n
X n
X
aij (x)λi λj ≥ c λ2i , c > 0,
i,j=1 i=1
is satisfied; this condition imposes the requirement that the differential operator on the
right-hand side of the equation is elliptic.
1◦ . First boundary value problem. The following conditions are prescribed:
w = g(x, t) for x ∈ S (boundary condition).
Solution:
Z tZ Z
0 V V
Z tZ
∂
− g(y, τ ) G(x, y, t − τ ) dSy dτ. (1)
0 S ∂My
The Green’s function is given by
∞
X un (x)un (y)
G(x, y, t) = exp(−λn t),
kun k2
n=1 (2)
Z
2
kun k = u2n (x) dV, y = {y1 , . . . , yn },
V
where the λn and un (x) are the eigenvalues and corresponding eigenfunctions of the Sturm–
Liouville problem for the elliptic second-order equation with a homogeneous boundary
condition of the first kind
X n
∂ ∂u
aij (x) − q(x)u + λu = 0, (3)
∂xi ∂xj
i,j=1
u = 0 for x ∈ S. (4)
∂
The integration in solution (1) is carried out with respect to y1 , . . . , yn ; ∂My is the differ-
ential operator defined as
X n
∂G ∂G
≡ aij (y)Nj , (5)
∂My ∂yi
i,j=1
where N = {N1 , . . . , Nn } is the unit outward normal to the surface S. In the special case
where aii (x) = 1 and aij (x) = 0 for i 6= j, the operator of (5) coincides with the usual
operator of differentiation along the direction of the outward normal to the surface S.
1. There are countably many eigenvalues. All eigenvalues are real and can be ordered
so that λ1 ≤ λ2 ≤ λ3 ≤ · · · , with λn → ∞ as n → ∞; consequently, there can exist only
finitely many negative eigenvalues.
2. For q(x) ≥ 0, all eigenvalues are positive: λn > 0.
3. The eigenfunctions are defined up to a constant multiplier. Any two eigenfunctions
un (x) and um (x) corresponding to different eigenvalues λn and λm are orthogonal in the
domain V : Z
un (x)um (x) dV = 0 for n 6= m.
V
Remark 5.2. To each eigenvalue λn there generally correspond finitely many linearly indepen-
(1) (2) (m)
dent eigenfunctions un , un , . . . , un . These functions can always be replaced by their linear
combinations
ū(k) (1) (k−1)
n = Ak,1 un + · · · + Ak,k−1 un + u(k)
n , k = 1, 2, . . . , m,
(1) (2) (m)
such that ūn , ūn , . . . , ūn
are now pairwise orthogonal. Thus, without loss of generality, we
assume that all eigenfunctions are orthogonal.
2◦ . Second boundary value problem. The following conditions are prescribed:
∂w
= g(x, t) for x ∈ S (boundary condition).
∂Mx
Here, the left-hand side of the boundary condition is determined with the help of (5), where
G, y, y, and yk must be replaced by w, x, x, and xk , respectively.
Solution:
Z tZ Z
0 V V
Z tZ
+ g(y, τ ) G(x, y, t − τ ) dSy dτ. (6)
0 S
Here, the Green’s function is defined by (2), where the λn and un (x) are the eigenvalues and
corresponding eigenfunctions of the Sturm–Liouville problem for the elliptic second-order
equation (3) with a homogeneous boundary condition of the second kind:
∂u
= 0 for x ∈ S. (7)
∂Mx
For q(x) > 0, the general properties of the eigenvalue problem (3), (7) are the same as
for the first boundary value problem (see Item 1◦ ). For q(x) ≡ 0 the zero eigenvalue λ0 = 0
arises which corresponds to the eigenfunction u0 = const.
It should be noted that the Green’s function of the second boundary value problem can
be expressed in terms of the Green’s function of the third boundary value problem (see
Item 3◦ ).
3◦ . Third boundary value problem. The following conditions are prescribed:

∂w
+ k(x)w = g(x, t) for x ∈ S (boundary condition).
∂Mx
The solution of the third boundary value problem is given by relations (6) and (2), where
the λn and un (x) are the eigenvalues and corresponding eigenfunctions of the Sturm–
Liouville problem for the second-order elliptic equation (3) with a homogeneous boundary
condition of the third kind:
∂u
+ k(x)u = 0 for x ∈ S. (8)
∂Mx
For q(x) ≥ 0 and k(x) > 0, the general properties of the eigenvalue problem (3), (8) are
the same as for the first boundary value problem (see Item 1◦ ).
Let k(x) = k = const. Denote the Green’s functions of the second and third boundary
value problems by G2 (x, y, t) and G3 (x, y, t, k), respectively. Then the following relations
hold: 
 lim G3 (x, y, t, k)
k→0 if q(x) > 0;
G2 (x, y, t) = 1

 + lim G3 (x, y, t, k) if q(x) ≡ 0;
V0 k→0
Z
where V0 = dV is the volume of the domain in question.
V
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), A. D. Polyanin (2000a, 2000b).
Chapter 6
Second-Order
Hyperbolic Equations
with One Space Variable
6.1 Constant Coefficient Equations

∂2w ∂2w
6.1.1 Wave Equation = a2
∂t2 ∂x2
This equation is also known as the equation of vibration of a string. It is often encountered
in elasticity, aerodynamics, acoustics, and electrodynamics.
◮ General solution. Some formulas.

1◦ . General solution:
w(x, t) = ϕ(x + at) + ψ(x − at),
where ϕ(x) and ψ(x) are arbitrary functions.
Physical interpretation: The solution represents two traveling waves that propagate,
respectively, to the left and right along the x-axis at a constant speed a.
(
1 0 for z ≤ 0,
E (x, t) = ϑ at − |x| , ϑ(z) =
2a 1 for z > 0.
3◦ . Infinite series solutions containing arbitrary functions of the space variable:
X∞
(at)2n (2n) dm
w(x, t) = f (x) + fx (x), fx(m) (x) = f (x),
(2n)! dxm
n=1
∞
X (at)2n (2n)
w(x, t) = tg(x) + t g (x),
(2n + 1)! x
n=1
where f (x) and g(x) are any infinitely differentiable functions. The first solution satisfies
the initial conditions w(x, 0) = f (x) and ∂t w(x, 0) = 0, and the second w(x, 0) = 0 and
∂t w(x, 0) = g(x). The sums are finite if f (x) and g(x) are polynomials.
557
558 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
4◦ . Infinite series solutions containing arbitrary functions of time:

∞
X 1 (2n) (m) dm
w(x, t) = f (t) + 2n
x2n ft (t), ft (t) = f (t),
a (2n)! dtm
n=1
∞
X 1 (2n)
w(x, t) = xg(t) + x x2n gt (t),
n=1
a2n (2n + 1)!
where f (t) and g(t) are any infinitely differentiable functions. The sums are finite if f (t)
and g(t) are polynomials. The first solution satisfies the boundary condition of the first
kind w(0, t) = f (t), and the second solution satisfies the boundary condition of the second
kind ∂x w(0, t) = g(t).
5◦ . If w(x, t) is a solution of the wave equation, then the functions
w1 = Aw(±λx + C1 , ±λt + C2 ),

x − vt t − va−2 x
w2 = Aw p ,p ,
1 − (v/a)2 1 − (v/a)2

x t
w3 = Aw 2 , ,
x − a2 t2 x2 − a2 t2
are also solutions of the equation everywhere these functions are defined (A, C1 , C2 , v,
and λ are arbitrary constants). The signs at λ’s in the formula for w1 are taken arbitrarily,
independently of each other. The function w2 results from the invariance of the wave
equation under the Lorentz transformations.
⊙ Literature: G. N. Polozhii (1964), A. V. Bitsadze and D. F. Kalinichenko (1985).

Initial conditions are prescribed:
w = f (x) at t = 0,
∂t w = g(x) at t = 0.
Solution (D’Alembert’s formula):

Z x+at
1 1
w(x, t) = [f (x + at) + f (x − at)] + g(ξ) dξ.
2 2a x−at

1◦ . Problem with a homogeneous boundary condition:

∂t w = g(x) at t = 0 (initial condition),
w=0 at x = 0 (boundary condition).
Solution:
 Z x+at

 1 1 x
 [f (x + at) + f (x − at)] +
 g(ξ) dξ if t < ,
2 2a x−at a
w(x, t) = Z x+at

 1 1 x

 [f (x + at) − f (at − x)] + g(ξ) dξ if t > .
2 2a at−x a
2◦ . Problem with a nonhomogeneous boundary condition:

w = h(t) at x = 0 (boundary condition).
Solution:
 Z x+at

 1 1 x

 2 [f (x + at) + f (x − at)] + g(ξ) dξ if t < ,
2a x−at a
w(x, t) = Z x+at

 1 1 x x

 [f (x + at) − f (at − x)] + g(ξ) dξ + h t − if t > .
2 2a at−x a a
In the domain t < x/a the boundary conditions have no effect on the solution and the
expression of w(x, t) coincides with D’Alembert’s solution for an infinite line (see the
Cauchy problem above).
⊙ Literature: A. N. Tikhonov and A. A. Samarskii (1990).

1◦ . Problem with a homogeneous boundary condition:

∂x w = 0 at x = 0 (boundary condition).
Solution:

 1 1 x
 [f (x + at) + f (x − at)] + [G(x + at) − G(x − at)] if t < ,
w(x, t) = 2 2a a
 1 [f (x + at) + f (at − x)] +
 1
[G(x + at) + G(at − x)]
x
if t > ,
2 2a a
Z z
where G(z) = g(ξ) dξ.
0
2◦ . Problem with a nonhomogeneous boundary condition:

∂x w = h(t) at x = 0 (boundary condition).
Solution:

 1 1 x
 [f (x+at)+f (x−at)]+ [G(x+at)−G(x−at)] if t < ,
w(x, t) = 2 2a x a
 1 1
 [f (x+at)+f (at−x)]+ [G(x+at)+G(at−x)]−aH t− x
if t > ,
2 2a a a
Z z Z z
where G(z) = g(ξ) dξ and H(z) = h(ξ) dξ. In the domain t < x/a the boundary
0 0
conditions have no effect on the solution, and the expression of w(x, t) coincides with
D’Alembert’s solution for an infinite line (see the Cauchy problem above).
⊙ Literature: B. M. Budak, A. N. Tikhonov, and A. A. Samarskii (1980).
◮ Domain: 0 ≤ x < ∞. A problem without a initial condition for Φ = 0.

w = A cos(ωt + γ) at x = 0, w → 0 as x → 0.
Solution:
where √ 1/2 √ 2
ω 2 + b2 − b ω + b2 + b 1/2
λ= , β= .
2a 2a

1◦ . Vibration of a string with rigidly fixed ends. The following conditions are prescribed:

w=0 at x = 0 (boundary condition),
w=0 at x = l (boundary condition).
Solution:
∞
X πn
w(x, t) = An cos(λn at) + Bn sin(λn at) sin(λn x), λn = ,
l
n=1
Z l Z l
2 2
An = f (x) sin(λn x) dx, Bn = g(x) sin(λn x) dx.
l 0 aπn 0
Example 6.1. The initial shape of the string is a triangle with base 0 ≤ x ≤ l and height h at
x = c, i.e., 
 hx
 if 0 ≤ x ≤ c,
c
f (x) = h(l
 − x)
 if c ≤ x ≤ l.
l−c
The initial velocities of the string points are zero, g(x) = 0.

Solution:
X∞
2hl2 1 nπc nπx nπat
w(x, t) = sin sin cos .
π 2 c(l − c) n=1 n2 l l l
Example 6.2. Initially, the string has the shape of a parabola symmetric about the center of the
string with elevation h, so that
4h
f (x) = 2 x(l − x).
l
Solution:
∞
32h X 1 (2n + 1)πx (2n + 1)πat
w(x, t) = 3 sin cos .
π n=0 (2n + 1)3 l l
2◦ . For the solution of the first boundary value problem with a nonhomogeneous boundary
condition, see Section 6.1.2 with Φ(x, t) ≡ 0.
⊙ Literature: B. M. Budak, A. N. Tikhonov, and A. A. Samarskii (1980), A. V. Bitsadze and D. F. Kalini-
chenko (1985).
1◦ . Longitudinal vibration of an elastic rod with free ends. The following conditions are
prescribed:
∂x w = 0 at x = 0 (boundary condition),
∂x w = 0 at x = l (boundary condition).
Solution:
∞
X
w(x, t) = A0 + B0 t + An cos(λn at) + Bn sin(λn at) cos(λn x),
n=1
Z Z
πn 1 l 1 l
λn = , A0 = f (x) dx, B0 = g(x) dx,
l l 0 l 0
Z Z l
2 l 2
An = f (x) cos(λn x) dx, Bn = g(x) cos(λn x) dx.
l 0 aπn 0
2◦ . For the solution of the second boundary value problem with a nonhomogeneous bound-
ary condition, see Section 6.1.2 with Φ(x, t) ≡ 0.
⊙ Literature: A. V. Bitsadze and D. F. Kalinichenko (1985).

1◦ . Longitudinal vibration of an elastic rod with clamped ends in the case of equal stiffness
coefficients. The following conditions are prescribed:
∂x w − kw = 0 at x = 0 (boundary condition),
∂x w + kw = 0 at x = l (boundary condition).
Solution:
∞
X
w(x, t) = An cos(λn at) + Bn sin(λn at) sin(λn x + ϕn ),
n=1
where
Z l Z l
1 1
An = sin(λn x + ϕn )f (x) dx, Bn = sin(λn x + ϕn )g(x) dx,
kXn k2 0 aλn kXn k2 0
Z l
λn 2 l k
ϕn = arctan , kXn k = sin2 (λn x + ϕn ) dx =
+ 2 ;
k 0 2 k + λ2n

1 λ k
the λn are positive roots of the transcendental equation cot(λl) = − .
2 k λ
2◦ . Longitudinal vibration of an elastic rod with clamped ends in the case of different
stiffness coefficients. The following conditions are prescribed:
∂x w − k1 w = 0 at x = 0 (boundary condition),
∂x w + k2 w = 0 at x = l (boundary condition).
Solution:
∞
X
w(x, t) = An cos(λn at) + Bn sin(λn at) sin(λn x + ϕn ),
n=1
where
Z l Z l
1 1
An = sin(λn x + ϕn )f (x) dx, Bn = sin(λn x + ϕn )g(x) dx,
kXn k2 0 aλn kXn k2 0
Z l
λn l (λ2 + k1 k2 )(k1 + k2 )
ϕn = arctan , kXn k = 2
sin2 (λn x + ϕn ) dx = + n2 ;
k1 0 2 2(λn + k12 )(λ2n + k22 )
λ2 − k1 k2
the λn are positive roots of the transcendental equation cot(λl) = .
λ(k1 + k2 )
3◦ . For the solution of the third boundary value problem with nonhomogeneous bound-
ary conditions, see Section 6.1.2 (the third boundary value problem for 0 ≤ x ≤ l with
Φ(x, t) ≡ 0).

1◦ . Longitudinal vibration of an elastic rod with one end rigidly fixed and the other free.

Solution:
∞
X π(2n + 1)
w(x, t) = An cos(λn at) + Bn sin(λn at) sin(λn x), λn = ,
2l
n=0
Z l Z l
2 2
An = f (x) sin(λn x) dx, Bn = g(x) sin(λn x) dx.
l 0 alλn 0
2◦ . For the solution of the mixed boundary value problem with nonhomogeneous bound-
ary conditions, see Section 6.1.2 (the mixed boundary value problem for 0 ≤ x ≤ l with
Φ(x, t) ≡ 0).
⊙ Literature: M. M. Smirnov (1975), A. V. Bitsadze and D. F. Kalinichenko (1985).
◮ Goursat problem.
The boundary conditions are prescribed to the equation characteristics:
w = f (x) for x − at = 0 (0 ≤ x ≤ b),

w = g(x) for x + at = 0 (0 ≤ x ≤ c),
where f (0) = g(0).

Solution:
x + at x − at
w(x, t) = f +g − f (0).
2 2
The solution propagation domain is bounded by four lines:
x − at = 0, x + at = 0, x − at = 2c, x + at = 2b.
∂2w ∂2w
6.1.2 Equations of the Form = a2 + Φ(x, t)
∂t2 ∂x2
w = f (x) at t = 0,
∂t w = g(x) at t = 0.
Solution:
x+at
Z Zt x+a(t−τ
Z )
1 1 1
w(x, t) = [f (x − at) + f (x + at)] + g(ξ) dξ + Φ(ξ, τ ) dξ dτ.
2 2a 2a
x−at 0 x−a(t−τ )


w = h(t) at x = 0 (boundary condition).
Solution:
1
w(x, t) = w1 (x, t) + w2 (x, t),
2a
where
 x+at
 Z

 1 1 x

 [f (x + at) + f (x − at)] + g(ξ) dξ if t < ,

2 2a a
x−at
w1 (x, t) = x+at
 Z


 1 1 x x

 [f (x + at) − f (at − x)] + g(ξ) dξ + h t − if t > ,
2 2a a a
at−x

 Zt x+a(t−τ
Z )

 x



 Φ(ξ, τ ) dξ dτ if t < ,

 a
0 x−a(t−τ )
w2 (x, t) = t−x/a x+a(t−τ ) x+a(t−τ )

 Z Z Zt Z

 x

 Φ(ξ, τ ) dξ dτ + Φ(ξ, τ ) dξ dτ if t > .

 a
 0
a(t−τ )−x t−x/a x−a(t−τ )


∂x w = h(t) at x = 0 (boundary condition).
Solution:
1
w(x, t) = w1 (x, t) + w2 (x, t),
2a
where

 x+at
Z

 1 1 x

 [f (x + at) + f (x − at)] + g(ξ) dξ if t < ,

 2 2a a



 x−at

 x+at
Z
1 1
w1 (x, t) = [f (x + at) + f (at − x)] + g(ξ) dξ

 2 2a

 0

 t−x/a

 at−x
Z Z

 1 x

 + g(ξ) dξ − a h(ξ) dξ if t > ,

 2a a
0 0


 Zt x+a(t−τ
Z )

 x

 Φ(ξ, τ ) dξ dτ if t < ,

 a

 0 x−a(t−τ )




 t−x/a
Z x+a(t−τ
Z ) t−x/a
Z a(t−τ
Z )−x
w2 (x, t) = Φ(ξ, τ ) dξ dτ + Φ(ξ, τ ) dξ dτ



 0 0 0 0



 Z t x+a(t−τ
Z )

 x

 + Φ(ξ, τ ) dξ dτ if t > .

 a

t−x/a x−a(t−τ )
w = f0 (x) at t = 0 (initial condition),

∂t w = f1 (x) at t = 0 (initial condition),
Solution:
Z l Z l Z tZ l
∂
w(x, t) = f0(ξ)G(x, ξ, t) dξ + f1(ξ)G(x, ξ, t) dξ + Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ
∂t 0 0 0 0
Z t h ∂ i Z t h ∂ i
+ a2 g1(τ ) G(x, ξ, t − τ ) dτ − a2 g2(τ ) G(x, ξ, t − τ ) dτ,
0 ∂ξ ξ=0 0 ∂ξ ξ=l
where
∞
2 X1 nπx nπξ nπat
G(x, ξ, t) = sin sin sin .
aπ n l l l
n=1

Solution:
Z Z l Z tZ l
∂ l
w(x, t)= f0(ξ)G(x, ξ, t) dξ+ f1(ξ)G(x, ξ, t) dξ+ Φ(ξ, τ )G(x, ξ, t−τ ) dξ dτ
∂t 0 0 0 0
Z t Z t
−a2 g1(τ )G(x, 0, t−τ ) dτ +a2 g2(τ )G(x, l, t−τ ) dτ,
0 0
where
t
∞
2 X1 nπx nπξ nπat
G(x, ξ, t) = + cos cos sin .
l aπ n l l l
n=1

∞
1X 1
G(x, ξ, t) = sin(λn x + ϕn ) sin(λn ξ + ϕn ) sin(λn at),
a n=1 λn kun k2
λn l (λ2 + k1 k2 )(k1 + k2 )
ϕn = arctan , kun k2 = + n2 ;
k1 2 2(λn + k12 )(λ2n + k22 )
λ2 − k1 k2
the λn are positive roots of the transcendental equation cot(λl) = .
λ(k1 + k2 )

Solution:
Z Z l Z tZ l
∂ l
∂t 0 0 0 0
Z t h∂ i Z t
+a2 g1(τ ) G(x, ξ, t−τ ) dτ +a2 g2(τ )G(x, l, t−τ ) dτ,
0 ∂ξ ξ=0 0
where
∞
2 X 1 π(2n + 1)
G(x, ξ, t) = sin(λn x) sin(λn ξ) sin(λn at), λn = .
al λn 2l
n=0
∂2w ∂2w
6.1.3 Equation of the Form = a2 − bw + Φ(x, t)
∂t2 ∂x2
This equation with Φ(x, t) ≡ 0 and b > 0 is encountered in quantum field theory and a
number of applications and is referred to as the Klein–Gordon equation.
◮ Solutions of the homogeneous equation (Φ ≡ 0).

w(x, t) = exp(±µt)(Ax + B), b = −µ2 ,
w(x, t) = exp(±λx)(At + B), b = a2 λ2 ,
w(x, t) = cos(λx)[A cos(µt) + B sin(µt)], b = −a2 λ2 + µ2 ,
w(x, t) = sin(λx)[A cos(µt) + B sin(µt)], b = −a2 λ2 + µ2 ,
w(x, t) = exp(±µt)[A cos(λx) + B sin(λx)], b = −a2 λ2 − µ2 ,
w(x, t) = exp(±λx)[A cos(µt) + B sin(µt)], b = a2 λ2 + µ2 ,
w(x, t) = exp(±λx)[A exp(µt) + B exp(−µt)], b = a2 λ2 − µ2 ,
√
bp 2
w(x, t) = AJ0 (ξ) + BY0 (ξ), ξ = a (t + C1 )2 − (x + C2 )2 , b > 0,
a
√
−b p 2
w(x, t) = AI0 (ξ) + BK0 (ξ), ξ = a (t + C1 )2 − (x + C2 )2 , b < 0,
a
where A, B, C1 , and C2 are arbitrary constants, J0 (ξ) and Y0 (ξ) are Bessel functions, and
I0 (ξ) and K0 (ξ) are modified Bessel functions.
2◦ . Fundamental solutions:
ϑ(at − |x|) c p 2 2
E (x, t) = J0 a t − x2 for b = c2 > 0,
2a a
ϑ(at − |x|) c p 2 2
E (x, t) = I0 a t − x2 for b = −c2 < 0,
2a a
where ϑ(z) is the Heaviside unit step function (ϑ = 0 for z < 0 and ϑ = 1 for z ≥ 0), J0 (z)
is the Bessel function, and I0 (z) is the modified Bessel function.
◮ Some formulas and transformations of the homogeneous equation (Φ ≡ 0).

1◦ . Suppose w = w(x, t) is a solution of the Klein–Gordon equation. Then the functions
w1 = Aw(x + C1 , ±t + C2 ),
w2 = Aw(−x + C1 , ±t + C2 ),

w3 = Aw p ,p ,
1 − (v/a)2 1 − (v/a)2
where A, C1 , C2 , and v are arbitrary constants, are also solutions of this equation.
2◦ . Table 6.1 lists transformations of the independent variables that allow separation of
variables in the Klein–Gordon equation.
TABLE 6.1
Orthogonal coordinates u = u(x, t), v = v(x, t) admitting separable solutions w = F (u)G(v)
of the Klein–Gordon equation (a = 1; A1 , A2 , B1 , B2 , and λ are arbitrary constants)
Relation between Function F = F (u) Function G = G(v)

No
x, t and u, v (differential equation) (differential equation)
√ √ √ √
1 x = u, t = v F = A1 eu λ+b +A2 e−u λ+b G = B1 ev λ
+B2 e−v λ
√ √ √
x = u sinh v, F = u A1Jσ u b +A2Yσ u b ,
2 √ G = B1 eλv +B2 e−λv
t = u cosh v σ = 12 1+λ2
x = uv, F = A1Dλ(βu) + A2Dλ(−βu), G = B1 Dλ (βv)+B2 Dλ (−βv),
3 t = 12 (u2 +v 2 ) β = (−4b)1/4 β = (−4b)1/4
x = 12 (u2 +v 2 ), F = A1 Dλ (βu)+A2 Dλ (−βu), G = B1Dλ(βv) + B2Dλ(−βv),
4
t = uv β = (4b)1/4 β = (4b)1/4
√ √
x = − 12 (u − v)2 + u + v, F = U A1 J 1 (ξ)+A2 Y 1 (ξ) , G = V B1J 1 (η) + B2Y 1 (η) ,
5 3 √ 3 3 √ 3
t = 12 (u − v)2 + u + v U = u+λ, ξ = 23 b U 3/2 V = v +λ, η = 23 b V 3/2

t+x = cosh 12 (u−v) ,
6 F ′′ +(λ+b sinh u)F = 0 G′′ +(λ+b sinh v)G = 0
t−x = sinh 12 (u+v)
x = sinh(u − v) − 12 eu+v, F = A1 Jλ (βeu )+A u

√ 2 Yλ (βe ), G = B1 Iλ (βev )+B v
√ 1 Kλ (βe ),
7
t = sinh(u − v) + 12 eu+v β= b β= b
x = cosh(u − v) − 12 eu+v, F = A1 Jλ (βeu )+A u
√ 2 Yλ (βe ), G = B1 Jλ (βev )+B v
√ 1 Yλ (βe ),
8
t = cosh(u − v) + 12 eu+v β= b β= b
x = cosh u sinh v, F ′′ +(λ+ 12 b cosh 2u)F = 0, G′′ +(λ− 12 b cosh 2v)G = 0,
9
t = sinh u cosh v modified Mathieu equation modified Mathieu equation
x = sinh u sinh v, F ′′ +(λ+ 12 b cosh 2u)F = 0, G′′ +(λ+ 12 b cosh 2v)G = 0,
10
t = cosh u cosh v modified Mathieu equation modified Mathieu equation
x = sin u sin v, F ′′ +(λ− 12 b cos 2u)F = 0, G′′ +(λ− 12 b cos 2v)G = 0,
11
t = cos u cos v Mathieu equation Mathieu equation
Notation: Jσ (z) and Yσ (z) are Bessel functions, Iσ (z) and Kσ (z) are modified Bessel
functions, and Dλ (z) is the parabolic cylinder function.
⊙ Literature: E. Kalnins (1975), W. Miller, Jr. (1977).

w = f (x) at t = 0,
∂t w = g(x) at t = 0.
Solution for b = −c2 < 0:

Z p
1 ct x+at I1 c t2 − (x − ξ)2/a2
w(x, t) = [f (x + at) + f (x − at)] + p f (ξ) dξ
2 2a x−at t2 − (x − ξ)2/a2
Z x+at p
1
+ I0 c t2 − (x − ξ)2/a2 g(ξ) dξ
2a x−at
Z t Z x+a(t−τ ) p
1
+ I0 c (t − τ )2 − (x − ξ)2/a2 Φ(ξ, τ ) dξ dτ,
2a 0 x−a(t−τ )
where I0 (z) and I1 (z) are modified Bessel functions of the first kind.
Solution for b = c2 > 0:
Z p
1 ct x+at J1 c t2 − (x − ξ)2/a2
w(x, t) = [f (x + at) + f (x − at)] − p f (ξ) dξ
2 2a x−at t2 − (x − ξ)2/a2
Z x+at p
1
+ J0 c t2 − (x − ξ)2/a2 g(ξ) dξ
2a x−at
Z t Z x+a(t−τ ) p
1
+ J0 c (t − τ )2 − (x − ξ)2/a2 Φ(ξ, τ ) dξ dτ,
2a 0 x−a(t−τ )
where J0 (z) and J1 (z) are Bessel functions of the first kind.


Solution:
Z Z l Z tZ l
∂ l
∂t 0 0 0 0
Z t Z t
2 ∂ 2 ∂
+a g1(τ ) G(x, ξ, t−τ ) dτ −a g2(τ ) G(x, ξ, t−τ ) dτ,
0 ∂ξ ξ=0 0 ∂ξ ξ=l
where
∞ p
2X sin t a2 λ2n + b πn
G(x, ξ, t) = sin(λn x) sin(λn ξ) p , λn = .
l a2 λ2n + b l
n=1
Remark 6.1. Let b < 0 and a2 λ2n + b < 0 for n = 1, . . . , m and a2 λ2n + b > 0 for n =
m + 1, m + 2, . . . In this case the Green’s function is modified and acquires the form
m p
2X sinh t |a2 λ2n + b|
G(x, ξ, t) = sin(λn x) sin(λn ξ) p
l n=1 |a2 λ2n + b|
∞ p
2 X sin t a2 λ2n + b πn
+ sin(λn x) sin(λn ξ) p , λn = .
l n=m+1 a2 λ2n + b l
Analogously, the Green’s functions for the second, third, and mixed boundary value problems are
modified in similar cases.


Solution:
Z Z l Z tZ l
∂ l
∂t 0 0 0 0
Z t Z t
2 2
−a g1(τ )G(x, 0, t−τ ) dτ +a g2(τ )G(x, l, t−τ ) dτ,
0 0
where
∞ p
1 √ 2X sin t a2 λ2n + b πn
G(x, ξ, t) = √ sin t b + cos(λn x) cos(λn ξ) p , λn = .
l b l a2 λ2n + b l
n=1


∞ p
X yn (x)yn (ξ) sin t a2 λ2n + b
G(x, ξ, t) = p ,
n=1 kyn k2 a2 λ2n + b

k1 2 k2 λ2n + k12 k1 l k12
yn (x) = cos(λn x) + sin(λn x), kyn k = + 2 + 1+ 2 .
λn 2λ2n λ2n + k22 2λn 2 λn
tan(λl) k1 + k2
Here, the λn are positive roots of the transcendental equation = 2 .
λ λ − k1 k2

Solution:
Z Z l Z tZ l
∂ l
∂t 0 0 0 0
Z t Z t
2 ∂ 2
+a g1(τ ) G(x, ξ, t−τ ) dτ +a g2(τ )G(x, l, t−τ ) dτ,
0 ∂ξ ξ=0 0
where
∞ p
2X sin t a2 λ2n + b π(2n + 1)
l a2 λ2n + b 2l
n=0
∂2w ∂2w ∂w
6.1.4 Equation of the Form = a2 −b + Φ(x, t)
∂t2 ∂x2 ∂x
◮ Reduction to the nonhomogeneous Klein–Gordon equation.

The substitution w(x, t) = exp 12 bx/a2 u(x, t) leads the nonhomogeneous Klein–Gordon
equation

∂2u 2∂ u
2 b2 bx
=a − 2 u + exp − 2 Φ(x, t),
∂t2 ∂x2 4a 2a
which is discussed in Section 6.1.3.


w = f (x) at t = 0,
∂t w = g(x) at t = 0.
Solution:

1 bt 1 bt
w(x, t) = f (x+at) exp − + f (x−at) exp
2 2a 2 2a
Z x+at p
σt bx bξ J1 σ t2 −(x−ξ)2/a2
− exp exp − 2 p f (ξ) dξ
2a 2a2 x−at 2a t2 −(x−ξ)2/a2
Z x+at
1 bx bξ p
+ exp 2
exp − 2
J0 σ t2 −(x−ξ)2/a2 g(ξ) dξ
2a 2a x−at 2a
Z tZ x+a(t−τ )
1 b(x−ξ) p
+ exp J σ (t−τ )2 −(x−ξ)2/a2 Φ(ξ, τ ) dξ dτ,
0
2a 0 x−a(t−τ ) 2a2
where J0 (z) and J1 (z) are Bessel functions of the first kind, and σ = 12 |b|/a.


Solution:
Z Z l Z tZ l
∂ l
∂t 0 0 0 0
Z t Z t
2 ∂ 2 ∂
+a g1(τ ) G(x, ξ, t−τ ) dτ −a g2(τ ) G(x, ξ, t−τ ) dτ,
0 ∂ξ ξ=0 0 ∂ξ ξ=l
where
X∞
2 b πnx πnξ sin λn t
G(x, ξ, t) = exp (x − ξ) sin sin ,
l 2a2 l l λn
n=1
r
a2 π 2 n2 b2
λn = + .
l2 4a2


Solution:
Z Z l Z tZ l
∂ l
∂t 0 0 0 0
Z t Z t
2 2
−a g1(τ )G(x, 0, t−τ ) dτ +a g2(τ )G(x, l, t−τ ) dτ,
0 0
where
∞
bt bξ 2 b(x−ξ) X yn(x)yn(ξ) sin(λnt)
G(x, ξ, t)= 2 exp − 2 + exp ,
a 1−exp(−bl/a2) a l 2a2 n=1
λn(1+µ2n)
r
πnx bl πnx a2π 2n2 b2 bl
yn(x)=cos − 2 sin , λn = 2
+ 2 , µn = 2 .
l 2a πn l l 4a 2a πn


∞
b(x − ξ) X yn (x)yn (ξ) sin(aλn t)
G(x, ξ, t) = exp .
2a2 aλn Bn
n=1
Here,
r
2a2 k1 − b 2 +
b2
yn (x) = cos(µn x) + sin(µ n x), λn = µ n ,
2a2 µn 4a4
2a2 k2 + b 4a4 µ2n + (2a2 k1 − b)2 2a2 k1 − b l l(2a2 k1 − b)2
Bn = + + + ,
4a2 µ2n 4a4 µ2n + (2a2 k2 + b)2 4a2 µ2n 2 8a4 µ2n

tan(µl) 4a4 (k1 + k2 )
= 4 2 .
µ 4a µ − (2a2 k1 − b)(2a2 k2 + b)
∂2w ∂2w ∂w
6.1.5 Equation of the Form = a2 +b + cw + Φ(x, t)
∂t2 ∂x2 ∂x
◮ Reduction to the nonhomogeneous Klein–Gordon equation.

The substitution w(x, t) = exp − 12 a−2 bx u(x, t) leads to the equation
∂2u ∂2u
2
= a2 2 + c − 14 a−2 b2 u + exp 1 −2
2 a bx Φ(x, t),
∂t ∂x

w = f (x) at t = 0,
∂t w = g(x) at t = 0.
Solution for c − 14 a−2 b2 = σ 2 > 0:

1 bt 1 bt
w(x, t) = f (x+at) exp + f (x−at) exp −
2 2a 2 2a
Z x+at p
σt bx bξ I1 σ t2 −(x−ξ)2/a2
+ exp − 2 exp p f (ξ) dξ
2a 2a x−at 2a2 t2 −(x−ξ)2/a2
Z x+at p
1 bx bξ
+ exp − 2 exp 2
I 0 σ t2 −(x−ξ)2/a2 g(ξ) dξ
2a 2a x−at 2a
Z tZ x+a(t−τ ) p
1 b(ξ −x)
+ exp 2
I0 σ (t−τ )2 −(x−ξ)2/a2 Φ(ξ, τ ) dξ dτ,
2a 0 x−a(t−τ ) 2a
Solution for c − 14 a−2 b2 = −σ 2 < 0:

1 bt 1 bt
w(x, t) = f (x+at) exp + f (x−at) exp −
2 2a 2 2a
Z x+at p
σt bx bξ J1 σ t2 −(x−ξ)2/a2
− exp − 2 exp p f (ξ) dξ
2a 2a x−at 2a2 t2 −(x−ξ)2/a2
Z x+at
1 bx bξ p
+ exp − 2 exp J σ t 2 −(x−ξ)2/a2 g(ξ) dξ
0
2a 2a x−at 2a2
Z tZ x+a(t−τ )
1 b(ξ −x) p
+ exp 2
J0 σ (t−τ )2 −(x−ξ)2/a2 Φ(ξ, τ ) dξ dτ,
2a 0 x−a(t−τ ) 2a
where J0 (z) and J1 (z) are Bessel functions of the first kind.

Solution:
Z tZ l Z Z l
∂ l
w(x, t)= Φ(ξ, τ )G(x, ξ, t−τ ) dξ dτ + f0(ξ)G(x, ξ, t) dξ+ f1(ξ)G(x, ξ, t) dξ
0 0 ∂t 0 0
Z t Z t
∂ ∂
+a2 g1(τ ) G(x, ξ, t−τ ) dτ −a2 g2(τ ) G(x, ξ, t−τ ) dτ.
0 ∂ξ ξ=0 0 ∂ξ ξ=l
Let a2 π 2 + 14 a−2 b2 l2 − cl2 > 0. Then

∞ √
2 b(ξ − x) X πnx πnξ sin t λn
G(x, ξ, t) = exp sin sin √ ,
l 2a2 n=1
l l λn
a2 π 2 n2 b2
λn = + − c.
l2 4a2
Let
a2 π 2 n2 + 14 a−2 b2 l2 − cl2 ≤ 0 at n = 1, . . . , m;
a2 π 2 n2 + 14 a−2 b2 l2 − cl2 > 0 at n = m + 1, m + 2, . . .
Then
m √
2 b(ξ − x) X πnx πnξ sinh t βn
G(x, ξ, t) = exp sin sin √
l 2a2 n=1
l l βn
X∞ √
2 b(ξ − x) πnx πnξ sin t λn
+ exp sin sin √ ,
l 2a2 l l λn
n=m+1
a2 π 2 n2 b2 a2 π 2 n2 b2
βn = c − − , λn = + − c.
l2 4a2 l2 4a2
√ √
For βn = 0 the ratio sinh t βn / βn must be replaced by t.


Solution:
Z tZ l Z Z l
∂ l
w(x, t)= Φ(ξ, τ )G(x, ξ, t−τ ) dξ dτ + f0(ξ)G(x, ξ, t) dξ+ f1(ξ)G(x, ξ, t) dξ
0 0 ∂t 0 0
Z t Z t
−a2 g1(τ )G(x, 0, t−τ ) dτ +a2 g2(τ )G(x, l, t−τ ) dτ.
0 0
For c < 0,
p
b bξ sin t |c|
G(x, ξ, t) = 2 bl/a2 exp 2 p
a e −1 a |c|
X∞ √
2 b(ξ −x) yn(x)yn(ξ) sin t λn a2π 2n2 b2
+ exp √ , λn = + 2 −c,
l 2a2 n=1
1+µ 2
n λn l2 4a

πnx πnx bl
yn(x) = cos +µn sin , µn = 2 .
l l 2a πn
For c > 0,
√
b bξ sinh t c
G(x, ξ, t) = 2 bl/a2 exp 2 √
a e −1 a c
X
∞ √
2 b(ξ − x) yn (x)yn (ξ) sin t λn
+ exp √ ,
l 2a2 1 + µ2n λn
n=1
where the λn , yn (x), and µn were specified √ previously. If the inequality λn < 0 holds for
several first values n =p 1, . . . , m, then the λn in the corresponding terms of the series
should be replaced by |λn |, and the sines by the hyperbolic sines.


6.2. Wave Equations with Axial or Central Symmetry 577
∞ √
b(ξ − x) X yn (x)yn (ξ) sin t λn
G(x, ξ, t) = exp √ .
2a2 n=1
Bn λn
Here,
2a2 k1 + b b2
yn (x) = cos(µn x) + sin(µn x), λn = a2 µ2n + − c,
2a2 µn 4a2
2a2 k2 − b 4a4 µ2n + (2a2 k1 + b)2 2a2 k1 + b l l(2a2 k1 + b)2
Bn = + + + ,
4a2 µ2n 4a4 µ2n + (2a2 k2 − b)2 4a2 µ2n 2 8a4 µ2n
tan(µl) 4a4 (k1 + k2 )
= 4 2 .
µ 4a µ − (2a2 k1 + b)(2a2 k2 − b)
6.2 Wave Equations with Axial or Central Symmetry

∂2w 2 ∂2w 1 ∂w
∂t2 ∂r 2 r ∂r
p
This is the one-dimensional wave equation with axial symmetry, where r = x2 + y 2 is
the radial coordinate. In the problems considered for 0 ≤ r ≤ R the solutions bounded at
r = 0 are sought (this is not specifically stated below).

w = f0 (r) at t = 0 (initial condition),
∂t w = f1 (r) at t = 0 (initial condition),
Solution:
Z R Z R
∂
w(r, t) = f0 (ξ)G(r, ξ, t) dξ + f1 (ξ)G(r, ξ, t) dξ
∂t 0 0
Z t
∂
− a2 g(τ ) G(r, ξ, t − τ ) dτ,
0 ∂ξ ξ=R
where
∞
2ξ X 1 λn r λn ξ λn at
G(r, ξ, t) = J0 J0 sin .
aR n=1 λn J12 (λn ) R R R
Here, the λn are positive zeros of the Bessel function, J0 (λ) = 0. The numerical values
of the first ten λn are specified in Section 3.2.1 (see the first boundary value problem for
0 ≤ r ≤ R).

∂r w = g(t) at r = R (boundary condition).
Solution:
Z R Z R Z t
∂
w(r, t) = f0 (ξ)G(r, ξ, t) dξ + f1 (ξ)G(r, ξ, t) dξ +a2 g(τ )G(r, R, t−τ ) dτ,
∂t 0 0 0
where
∞
2tξ 2ξ X 1 λn r λn ξ λn at
G(r, ξ, t) = 2 + J0 J0 sin .
R aR n=1 λn J02 (λn ) R R R
Here, the λn are positive zeros of the first-order Bessel function, J1 (λ) = 0. The numerical
values of the first ten roots λn are specified in Section 3.2.1 (see the second boundary value
⊙ Literature: M. M. Smirnov (1975), B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).

∂r w + kw = g(t) at r = R (boundary condition).
The solution w(r, t) is determined by the formula in the previous paragraph (for the
∞
2ξ X λn λn r λn ξ λn at
G(r, ξ, t) = J0 J0 sin .
aR
n=1
(k2 R2 + λ2n )J02 (λn ) R R R
Here, the λn are positive roots of the transcendental equation
λJ1 (λ) − kRJ0 (λ) = 0.
The numerical values of the first six roots λn can be found in Carslaw and Jaeger (1984);
see also Abramowitz and Stegun (1964).


Solution:
Z R2 Z R2
∂
w(r, t) = f0(ξ)G(r, ξ, t) dξ + f1(ξ)G(r, ξ, t) dξ
∂t R1 R1
Z t Z t
2 ∂ 2 ∂
+a g1(τ ) G(r, ξ, t−τ ) dτ −a g2(τ ) G(r, ξ, t−τ ) dτ.
0 ∂ξ ξ=R1 0 ∂ξ ξ=R2
Here,
∞
X
λn at π 2 λ J 2 (sλn )
G(r, ξ, t) = An ξΨn (r)Ψn (ξ) sin , An = 2 n 0 ,
n=1
R1 2aR1 J0 (λn ) − J02 (sλn )

λn r λn r R2
Ψn (r) = Y0 (λn )J0 − J0 (λn )Y0 , s= ,
R1 R1 R1
where J0 (z) and Y0 (z) are Bessel functions, and the λn are positive roots of the transcen-
dental equation
J0 (λ)Y0 (sλ) − J0 (sλ)Y0 (λ) = 0.
The numerical values of the first five roots λn = λn (s) can be found in Abramowitz and
Stegun (1964) and Carslaw and Jaeger (1984).


Solution:
Z R2 Z R2
∂
∂t R1 R1
Z t Z t
− a2 g1 (τ )G(r, R1 , t − τ ) dτ + a2 g2 (τ )G(r, R2 , t − τ ) dτ.
0 0
Here,
∞
X
2tξ λnat π 2λnJ12(sλn)
G(r, ξ, t)= 2 + An ξΨ n(r)Ψ n (ξ) sin , A n = ,
R2 −R12 n=1 R1 2aR1 J12(λn)−J12(sλn)

λnr λnr R2
Ψn(r)=Y1(λn)J0 −J1(λn)Y0 , s= ,
R1 R1 R1
where Jk (z) and Yk (z) are Bessel functions (k = 0, 1); the λn are positive roots of the
J1 (λ)Y1 (sλ) − J1 (sλ)Y1 (λ) = 0.
The numerical values of the first five roots λn = λn (s) can be found in Abramowitz and
Stegun (1964).


∞
π 2 X µn 2
G(r, ξ, t) = k2 J0 (µn R2 ) − µn J1 (µn R2 ) ξHn (r)Hn (ξ) sin(µn at).
2a Bn
n=1
Here,
2 2
Bn = (µ2n +k22) k1J0(µnR1)+µnJ1(µnR1) −(µ2n +k12) k2J0(µnR2)−µnJ1(µnR2) ,

Hn(r) = k1Y0(µnR1)+µnY1(µnR1) J0(µnr)− k1J0(µnR1)+µnJ1(µnR1) Y0(µnr);
Jk (z) and Yk (z) are Bessel functions (k = 0, 1); and the µn are positive roots of the

k1 J0 (µR1 ) + µJ1 (µR1 ) k2 Y0 (µR2 ) − µY1 (µR2 )

− k2 J0 (µR2 ) − µJ1 (µR2 ) k1 Y0 (µR1 ) + µY1 (µR1 ) = 0.

∂2w 2 ∂2w 1 ∂w
∂t2 ∂r 2 r ∂r
◮ Domain: 0 ≤ r ≤ R. Different boundary value problems.
1◦ . The solution to the first boundary value problem for a circle of radius R is given by the
formula from Section 6.2.1 (see the first boundary value problem for 0 ≤ r ≤ R) with the
additional term
Z tZ R
Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ, (1)
0 0
2◦ . The solution to the second boundary value problem for a circle of radius R is given by
the formula from Section 6.2.1 (see the second boundary value problem for 0 ≤ r ≤ R)
3◦ . The solution to the third boundary value problem for a circle of radius R is the sum of
the solution presented in Section 6.2.1 (see the third boundary value problem for 0 ≤ r ≤ R)
and expression (1); the Green’s function is also taken from Section 6.2.1.
◮ Domain: R1 ≤ r ≤ R2 . Different boundary value problems.
1◦ . The solution to the first boundary value problem for an annular domain is given by the
formula from Section 6.2.1 (see the first boundary value problem for R1 ≤ r ≤ R2 ) with
the additional term
Z t Z R2
Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ, (2)
0 R1
2◦ . The solution to the second boundary value problem for an annular domain is given by
the formula from Section 6.2.1 (see the second boundary value problem for R1 ≤ r ≤ R2 )
3◦ . The solution to the third boundary value problem for an annular domain is the sum of
the solution presented in Section 6.2.1 (see the third boundary value problem for R1 ≤ r ≤
R2 ) and expression (2); the Green’s function is also taken from Section 6.2.1.

∂2w ∂2w 2 ∂w
6.2.3 Equation of the Form = a2 +
∂t2 ∂r 2 r ∂r
Thisp is the equation of one-dimensional vibration of a gas with central symmetry, where
r = x2 + y 2 + z 2 is the radial coordinate. In the problems considered for 0 ≤ r ≤ R the
solutions bounded at r = 0 are sought; this is not specifically stated below.
◮ General solution:
ϕ(r + at) + ψ(r − at)

w(t, r) = ,
r
where ϕ(r1 ) and ψ(r2 ) are arbitrary functions.
◮ Reduction to a constant coefficient equation.
The substitution u(r, t) = rw(r, t) leads to the constant coefficient equation
∂2u 2
2∂ u
= a ,
∂t2 ∂r 2
◮ Domain: 0 ≤ r < ∞. Cauchy problem.
w = f (r) at t = 0,
∂t w = g(r) at t = 0.
Solution:
Z r+at
1 1
w(r, t) = (r − at)f |r − at| + (r + at)f |r + at| + ξg |ξ| dξ.
2r 2ar r−at
Solution at the center r = 0:
w(0, t) = atf ′ (at) + f (at) + tg(at).

Solution:
Z R Z R
∂
∂t 0 0
Z t
2 ∂
−a g(τ ) G(r, ξ, t − τ ) dτ,
0 ∂ξ ξ=R
where
∞
2ξ X 1 nπr nπξ anπt
G(r, ξ, t) = sin sin sin .
πar n R R R
n=1

Solution:
Z R Z R Z t
∂
w(r, t) = f0(ξ)G(r, ξ, t) dξ + f1(ξ)G(r, ξ, t) dξ +a2 g(τ )G(r, R, t−τ ) dτ,
∂t 0 0 0
where
∞
3tξ 2 2ξ X µ2n + 1 µn r µn ξ µn at
G(r, ξ, t) = 3 + sin sin sin .
R ar µ3n R R R
n=1
Here, the µn are positive roots of the transcendental equation tan µ−µ = 0. The numerical
values of the first five roots µn are specified in Section 3.2.3 (see the second boundary value

∞
2ξ X µ2n + (kR − 1)2 µn r µn ξ µn at
ar µn µ2n + kR(kR − 1)
n=1
R R R
Here, the µn are positive roots of the transcendental equation
µ cot µ + kR − 1 = 0.

Solution:
Z R2 Z R2
∂
w(r, t)= f0(ξ)G(r, ξ, t) dξ+ f1(ξ)G(r, ξ, t) dξ
∂t R1 R1
Z t Z t
2 ∂ 2 ∂
+a g1(τ ) G(r, ξ, t−τ ) dτ −a g2(τ ) G(r, ξ, t−τ ) dτ,
0 ∂ξ ξ=R1 0 ∂ξ ξ=R2
where
∞
2ξ X 1 πn(r − R1 ) πn(ξ − R1 ) πnat
πar n R2 − R1 R2 − R1 R2 − R1
n=1


Solution:
Z R2 Z R2
∂
∂t R1 R1
Z t Z t
2 2
−a g1 (τ )G(r, R1 , t − τ ) dτ + a g2 (τ )G(r, R2 , t − τ ) dτ,
0 0
where
X∞
3tξ 2 2ξ (1 + R22 λ2n )Ψn (r)Ψn (ξ) sin(λn at)
G(r, ξ, t) = 3 + ,
R2 − R13 a(R2 − R1 )r λ3 R12 + R22 + R1 R2 (1 + R1 R2 λ2n )
n=1 n
Ψn (r) = sin[λn (r − R1 )] + R1 λn cos[λn (r − R1 )].
(λ2 R1 R2 + 1) tan[λ(R2 − R1 )] − λ(R2 − R1 ) = 0.


X∞
2ξ 1 (b22 + R22λ2n)Ψn(r)Ψn (ξ) sin(λnat)
G(r, ξ, t) = ,
a(R2 − R1)r n=1 λn (b1 + R1λn )(b22 + R22λ2n ) + (b1R2 + b2R1)(b1 b2 + R1R2λ2n)
2 2 2
Ψn(r) = b1 sin[λn (r − R1)] + R1λn cos[λn (r − R1)], b1 = k1R1 + 1, b2 = k2R2 − 1.

∂2w 2 ∂2w 2 ∂w
∂t2 ∂r 2 r ∂r
◮ Reduction to a nonhomogeneous constant coefficient equation.
The substitution u(r, t) = rw(r, t) leads to the nonhomogeneous constant coefficient equa-
tion
∂2u 2∂ u
2
= a + rΦ(r, t),
∂t2 ∂r 2
◮ Domain: 0 ≤ r < ∞. Cauchy problem.

w = f (r) at t = 0,
∂t w = g(r) at t = 0.
Solution:
Z r+at
1 1
w(r, t) = (r − at)f |r − at| + (r + at)f |r + at| + ξg |ξ| dξ
2r 2ar r−at
Z t Z r+a(t−τ )
1
+ dτ ξΦ |ξ|, τ dξ.
2ar 0 r−a(t−τ )
◮ Domain: 0 ≤ r ≤ R. Different boundary value problems.

1◦ . The solution to the first boundary value problem for a sphere of radius R is given by
the formula from Section 6.2.3 (see the first boundary value problem for 0 ≤ r ≤ R) with
the additional term Z Z t R
Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ, (1)
0 0
2◦ . The solution to the second boundary value problem for a sphere of radius R is given
by the formula from Section 6.2.3 (see the second boundary value problem for 0 ≤ r ≤ R)
3◦ . The solution to the third boundary value problem for a sphere of radius R is the sum of
the solution presented in Section 6.2.3 (see the third boundary value problem for 0 ≤ r ≤ R)
◮ Domain: R1 ≤ r ≤ R2 . Different boundary value problems.

1◦ . The solution to the first boundary value problem for a spherical layer is given by the
formula from Section 6.2.3 (see the first boundary value problem for R1 ≤ r ≤ R2 ) with
the additional term Z Z t R2
Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ, (2)
0 R1
2◦ . The solution to the second boundary value problem for a spherical layer is given by the
formula from Section 6.2.3 (see the second boundary value problem for R1 ≤ r ≤ R2 ) with
the additional term (2); the Green’s function is also taken from Section 6.2.3.
3◦ . The solution to the third boundary value problem for a spherical layer is the sum of the
solution presented in Section 6.2.3 (see the third boundary value problem for R1 ≤ r ≤ R2 )

∂2w 2 ∂2w 1 ∂w
6.2.5 Equation of the Form =a + − bw + Φ(r, t)
∂t2 ∂r 2 r ∂r
For b > 0 and Φ ≡ 0, this is the Klein–Gordon equation describing one-dimensional wave
phenomena with axial symmetry. In the problems considered for 0 ≤ r ≤ R the solutions
bounded at r = 0 are sought; this is not specifically stated below.


Solution:
Z R Z R
∂
∂t 0 0
Z t Z tZ R
2 ∂
−a g(τ ) G(r, ξ, t − τ ) dτ + Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ.
0 ∂ξ ξ=R 0 0
Here,
∞ √
2ξ X 1 µn r µn ξ sin t λn a2 µ2n
G(r, ξ, t) = 2 J0 J0 √ , λn = + b,
R n=1 J12 (µn ) R R λn R2
where the µn are positive zeros of the Bessel function, J0 (µ) = 0. The numerical values
0 ≤ r ≤ R).


Solution:
Z R Z R
∂
∂t 0 0
Z t Z tZ R
+ a2 g(τ )G(r, R, t − τ ) dτ + Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ.
0 0 0
Here,
√ ∞ √
2ξ sin t b 2ξ X 1 µn r µn ξ sin t λn
G(r, ξ, t) = √ + 2 2 (µ ) J0 J0 √ ,
R2 b R J0 n R R λn
n=1
a2 µ2n
λn = + b,
R2
values of the first ten µn are specified in Section 3.2.1 (see the second boundary value


∞ √
2 X µ2nξ µnr µnξ sin t λn a2µ2n
G(r, ξ, t)= 2 J0 J0 √ , λn = +b.
R n=1 (k2R2+µ2n)J02(µn) R R λn R2
µJ1 (µ) − kRJ0 (µ) = 0.
The numerical values of the first six roots µn can be found in Abramowitz and Stegun
(1964) and Carslaw and Jaeger (1984).


Solution:
Z tZ R2
w(r, t) = Φ(ξ, τ )G(r, ξ, t−τ ) dξ dτ
0 R1
Z R2 Z R2
∂
+ f0(ξ)G(r, ξ, t) dξ + f1(ξ)G(r, ξ, t) dξ
∂t R1 R1
Z t Z t
∂ ∂
+a2 g1(τ ) G(r, ξ, t−τ ) dτ −a2 g2(τ ) G(r, ξ, t−τ ) dτ.
0 ∂ξ ξ=R1 0 ∂ξ ξ=R2
Here,
∞ √
π2 X µ2n J02 (sµn )ξ sin t λn a2 µ2n
G(r, ξ, t) = 2 2 2 Ψn (r)Ψn (ξ) √ , λn = + b,
2R1 n=1 J0 (µn ) − J0 (sµn ) λn R12

µn r µn r R2
Ψn (r) = Y0 (µn )J0 − J0 (µn )Y0 , s= ,
R1 R1 R1
where J0 (z) and Y0 (z) are Bessel functions and the µn are positive roots of the transcen-
dental equation
J0 (µ)Y0 (sµ) − J0 (sµ)Y0 (µ) = 0.
Stegun (1964) and Carslaw and Jaeger (1984).


Solution:
Z tZ R2
w(r, t) = Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ
0 R1
Z R2 Z R2
∂
+ f0 (ξ)G(r, ξ, t) dξ + f1 (ξ)G(r, ξ, t) dξ
∂t R1 R1
Z t Z t
− a2 g1 (τ )G(r, R1 , t − τ ) dτ + a2 g2 (τ )G(r, R2 , t − τ ) dτ.
0 0
Here,
√ ∞ √
2ξ sin t b π2 X µ2n J12 (sµn )ξ sin t λn
G(r, ξ, t) = √ + Ψn (r)Ψn (ξ) √ ,
(R22 − R12 ) b 2R12 n=1 J12 (µn ) − J12 (sµn ) λn

µn r µn r a2 µ2n R2
Ψn (r) = Y1 (µn )J0 − J1 (µn )Y0 , λn = 2 + b, s = ,
R1 R1 R1 R1
where Jk (z) and Yk (z) are Bessel functions (k = 0, 1); the µn are positive roots of the
J1 (µ)Y1 (sµ) − J1 (sµ)Y1 (µ) = 0.
Stegun (1964).

∞ 2
π 2 X βn2 k2 J0 (βnR2 ) − βnJ1 (βn R2) p
G(r, ξ, t) = p ξHn(r)Hn(ξ) sin t a2βn2 + b .
2 n=1 Bn a2 βn2 + b
Here,
2 2
Bn = (βn2 +k22) k1J0(βnR1)+βnJ1(βnR1) −(βn2 +k12) k2J0(βnR2)−βnJ1(βnR2) ,

Hn(r) = k1Y0(βnR1)+βnY1(βnR1) J0(βnr)− k1J0(βnR1)+βnJ1(βnR1) Y0(βnr),
where the βn are positive roots of the transcendental equation

k1 J0 (βR1 ) + βJ1 (βR1 ) k2 Y0 (βR2 ) − βY1 (βR2 )

− k2 J0 (βR2 ) − βJ1 (βR2 ) k1 Y0 (βR1 ) + βY1 (βR1 ) = 0.

∂2w 2 ∂2w 2 ∂w
6.2.6 Equation of the Form =a + − bw + Φ(r, t)
∂t2 ∂r 2 r ∂r
For b > 0 and Φ ≡ 0, this is the Klein–Gordon equation describing one-dimensional wave
phenomena with central symmetry. In the problems considered for 0 ≤ r ≤ R the solutions
bounded at r = 0 are sought; this is not specifically stated below.

Solution:
Z R Z R
∂
∂t 0 0
Z t Z tZ R
2 ∂
−a g(τ ) G(r, ξ, t − τ ) dτ + Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ,
0 ∂ξ ξ=R 0 0
where
∞ √
2ξ X nπr nπξ sin t λn a2 π 2 n2
G(r, ξ, t) = sin sin √ , λn = + b.
Rr R R λn R2
n=1

Solution:
Z R Z R
∂
∂t 0 0
Z t Z tZ R
+ a2 g(τ )G(r, R, t − τ ) dτ + Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ,
0 0 0
where
√ ∞
3ξ 2 sin t b 2ξ X µ2n + 1 µn r µn ξ p
G(r, ξ, t) = √ + √ sin sin sin t λn ,
R3 b Rr µ2 λn
n=1 n
R R
a2 µ2n
λn = + b.
R2
Here, the µn are positive roots of the transcendental equation tan µ − µ = 0; for the
numerical values of the first five roots µn , see Section 3.2.3 (the second boundary value
problem with 0 ≤ r ≤ R).

∞ √
2ξ X µ2n+(kR−1)2 µnr µnξ sin t λn a2µ2n
G(r, ξ, t)= sin sin √ , λn = +b.
Rr µ2n+kR(kR−1) R R λn R2
n=1
Here, the µn are positive roots of the transcendental equation µ cot µ + kR − 1 = 0. The
numerical values of the first five roots µn can be found in Carslaw and Jaeger (1984).

Solution:
Z tZ R2
w(r, t) = Φ(ξ, τ )G(r, ξ, t−τ ) dξ dτ
0 R1
Z R2 Z R2
∂
+ f0(ξ)G(r, ξ, t) dξ + f1(ξ)G(r, ξ, t) dξ
∂t R1 R1
Z t Z t
∂ ∂
+a2 g1(τ ) G(r, ξ, t−τ ) dτ −a2 g2(τ ) G(r, ξ, t−τ ) dτ,
0 ∂ξ ξ=R1 0 ∂ξ ξ=R2
where
∞ √
2ξ X πn(r − R1 ) πn(ξ − R1 ) sin t λn
G(r, ξ, t) = sin sin √ ,
(R2 − R1 )r n=1 R2 − R1 R2 − R1 λn
a2 π 2 n2
λn = + b.
(R2 − R1 )2

Solution:
Z tZ R2
w(r, t) = Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ
0 R1
Z R2 Z R2
∂
+ f0 (ξ)G(r, ξ, t) dξ + f1 (ξ)G(r, ξ, t) dξ
∂t R1 R1
Z t Z t
2 2
−a g1 (τ )G(r, R1 , t − τ ) dτ + a g2 (τ )G(r, R2 , t − τ ) dτ.
0 0
Here,
√ ∞ p
3ξ 2 sin t b 2ξ X (1+R22λ2n)Ψn(r)Ψn(ξ) sin t a2λ2n+b
G(r, ξ, t)= √ + p ,
(R23−R13) b (R2−R1)r n=1 λ2n R12+R22+R1R2(1+R1R2λ2n) a2λ2n+b
Ψn(r)=sin[λn(r−R1)]+R1λn cos[λn(r−R1)],
(λ2 R1 R2 + 1) tan[λ(R2 − R1 )] − λ(R2 − R1 ) = 0.

∞
2ξ X (b22+R22λ2n)Ψn(r)Ψn(ξ) sin βnt
G(r, ξ, t)= ,
r β (R2−R1)(b21+R12λ2n)(b22+R22λ2n)+(b1R2+b2R1)(b1b2+R1R2λ2n)
n=1 n
Ψn(r)=b1 sin[λn(r−R1)]+R1λn cos[λn(r−R1)],
p
b1 =k1R1+1, b2 =k2R2−1, βn = a2λ2n+b.
6.3 Equations Containing Power Functions and Arbitrary

Parameters
∂2w ∂ 2w ∂w
6.3.1 Equations of the Form = (ax+b) +c +kw +Φ(x, t)
∂t2 ∂x2 ∂x

∂2w ∂ ∂w
1. = a2 x + Φ(x, t).
∂t2 ∂x ∂x
For Φ(x, t) ≡ 0, this equation governs small-amplitude free vibration of a hanging heavy
homogeneous thread (a2 is the acceleration due to gravity, w the deflection of the thread
from the vertical axis, and x the vertical coordinate).
1◦ . The substitution x = 14 r 2 leads to the equation
2
∂2w 2 ∂ w 1 ∂w 1 2

=a + +Φ 4r ,t ,
∂t2 ∂r 2 r ∂r
which is discussed in Sections 6.2.1–6.2.2.

2◦ . Domain: 0 ≤ x ≤ l. First boundary value problem.

w = g(t) at x = l (boundary condition),
w 6= ∞ at x = 0 (boundedness condition).
Solution:
Z l Z l
∂
w(x, t) = f0 (ξ)G(x, ξ, t) dξ + f1 (ξ)G(x, ξ, t) dξ
∂t 0 0
Z t Z tZ l
2 ∂
−a l g(τ ) G(x, ξ, t − τ ) dτ + Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ,
0 ∂ξ ξ=l 0 0
where
∞ r r
2 X 1 x ξ µn at
G(x, ξ, t) = √ J0 µn J0 µn sin √ .
a l n=1 µn J12 (µn ) l l 2 l
Here, the µn are positive zeros of the Bessel function, J0 (µ) = 0. The numerical values of
the first ten roots µn are specified in Section 3.2.1 (see the first boundary value problem for
0 ≤ r ≤ R).
⊙ Literature: M. M. Smirnov (1975).
3◦ . Domain: 0 ≤ x ≤ l. Second boundary value problem.


∂x w = g(t) at x = l (boundary condition),
Solution:
Z l Z l
∂
w(x, t) = f0 (ξ)G(x, ξ, t) dξ +
f1 (ξ)G(x, ξ, t) dξ
∂t 0 0
Z t Z tZ l
2
+a l g(τ )G(x, l, t − τ ) dτ + Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ,
0 0 0
where
∞ r r
t 2 X 1 x ξ µn at
G(x, ξ, t) = + √ J0 µn J0 µn sin √ .
l a l n=1 µn J02 (µn ) l l 2 l
Here, the µn are positive zeros of the first-order Bessel function, J1 (µ) = 0. The numerical
values of the first ten roots µn are specified in Section 3.2.1 (see the second boundary value
4◦ . Domain: 0 ≤ x ≤ l. Third boundary value problem.

∂x w + kw = g(t) at x = l (boundary condition),
The solution w(x, t) is given by the formula in Item 3◦ with

∞ r r
2 X µn x ξ µn at
G(x, ξ, t) = √ J0 µn J0 µn sin √ .
a l n=1 (4k2 l + µ2n )J02 (µn ) l l 2 l

√
µJ1 (µ) − 2k l J0 (µ) = 0.

∂2w ∂ ∂w
2. = a2 x − bw + Φ(x, t).
∂t2 ∂x ∂x
For b < 0 and Φ(x, t) ≡ 0, this equation describes small-amplitudepvibration of a heavy
homogeneous thread that rotates at a constant angular velocity ω = |b| about the vertical
axis (a2 is the acceleration due to gravity).
1◦ . The substitution x = 14 r 2 leads to the equation

2
∂2w 2 ∂ w 1 ∂w 1 2

=a + − bw + Φ 4r ,t ,
∂t2 ∂r 2 r ∂r


Solution:
Z l Z l
∂
∂t 0 0
Z t Z tZ l
∂
− a2 l g(τ ) G(x, ξ, t − τ ) dτ + Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ.
0 ∂ξ ξ=l 0 0
Here,
∞ r r √
1X 1 x ξ sin t λn a2 µ2n
G(x, ξ, t) = J0 µn J0 µn √ , λn = + b,
l J 2 (µn )
n=1 1
l l λn 4l
of the first ten roots µn are specified in Section 3.2.1 (see the first boundary value problem
for 0 ≤ r ≤ R).


Solution:
Z l Z l
∂
w(x, t) = f0 (ξ)G(x, ξ, t) dξ +
f1 (ξ)G(x, ξ, t) dξ
∂t 0 0
Z t Z tZ l
+ a2 l g(τ )G(x, l, t − τ ) dτ + Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ.
0 0 0
Here,
√ ∞ r r √
sin t b 1X 1 x ξ sin t λn
G(r, ξ, t) = √ + J0 µn J0 µn √ ,
l b l n=1 J02 (µn ) l l λn
a2 µ2n
λn = + b,
4l

∂x w + kw = g(t) at x = l (boundary condition).

∞
X r r √
µ2n x ξ sin t λn a2µ2n
G(r, ξ, t)= J 0 µ n J0 µ n √ , λn = +b.
l(4k2l+µ2n)J02(µn)
n=1
l l λn 4l

√
µJ1 (µ) − 2k l J0 (µ) = 0.
h i
∂2w ∂ ∂w
3. = a2 (l − x) .
∂t2 ∂x ∂x
This equation governs small-amplitude free vibration of a heavy homogeneous thread of
length l (a2 is the acceleration due to gravity, w the deflection of the thread from the vertical
axis, and x the vertical coordinate).
The change of variable z = l − x leads a special case of equation 6.3.1.1 with b = 0 and
Φ ≡ 0.

∂2w 2 2 ∂ 2w ∂w
4. = a x + , n = 1, 2, . . .
2
∂t 2n + 1 2∂x ∂x
General solution:
p p
∂ n−1 Φ( 2(2n + 1)x + at) + Ψ( 2(2n + 1)x − at)
w(x, t) = √ ,
∂xn−1 x
where Φ and Ψ are arbitrary functions.

∂2w ∂2w ∂w
5. = (ax + b) +a + cw + Φ(x, t).
∂t2 ∂x2 ∂x
The substitution z = ax + b leads to an equation of the form 6.3.1.2:

∂2w 2 ∂ ∂w z−b
=a z + cw + Φ ,t .
∂t2 ∂z ∂z a
∂2w ∂2w 1 ∂w
6. = (ax + b) + a + cw + Φ(x, t).
∂t2 ∂x2 2 ∂x
√
The substitution z = 2 ax + b leads to the equation
2
∂2w 2
2∂ w z − 4b
=a + cw + Φ ,t ,
∂t2 ∂z 2 4a
∂2w ∂ 2w ∂w
7. = (a2 x + b2 ) + (a1 x + b1 ) + (a0 x + b0 )w.
∂t2 ∂x2 ∂x
This is a special case of equation 6.5.3.4 with f (x) = a2 x + b2 , g(x) = a1 x + b1 , h(x) =

a0 x + b0 , and Φ ≡ 0.

x+q √ √
w(x, t) = exp(kx)F A sin t µ + B cos t µ for µ > 0,
p

x+q √ √
w(x, t) = exp(kx)F A sinh t −µ + B cosh t −µ for µ < 0,
p
where A, B, and µ are arbitrary constants; the coefficients k, p, q and the function F = F (ξ)
are listed in Table 6.2, in which
J (α, β; x) = C1 Φ(α, β; x) + C2 Ψ(α, β; x), C1 , C2 are arbitrary numbers,
′′ +(β−x)y ′ −αy=0,
is an arbitrary solution of the degenerate hypergeometric equation xyxx x
and
Zν (x) = C1 Jν (x) + C2 Yν (x), C1 , C2 are arbitrary numbers,
is an arbitrary solution of the Bessel equation x2 yxx

′′ + xy ′ + (x2 − ν 2 )y = 0.
x
For details about the degenerate hypergeometric functions Φ(a, b; x) and Ψ(a, b; x),
see Section 30.9 as well as the books by Abramowitz and Stegun (1964) and Bateman and
Erdélyi (1953, Vol. 1). For the Bessel functions Jν (x) and Yν (x), see Section 30.6 as well
as the books by Abramowitz and Stegun (1964) and Bateman and Erdélyi (1953, Vol. 2).
TABLE 6.2
The coefficients k, p, q and the function F = F (ξ) determining the form
of particular solutions to equation 6.3.1.7. Notation: E(k) = b2 k2 + b1 k + b0 + µ
Conditions k p q F = F (ξ) Parameters

√
a2 6= 0, D 6= 0 D − a1 a2 b2 α = E(k)/(2a2 k + a1),
2
− J (α, β; ξ)
D≡a1 − 4a0 a2 2a2 2a 2 k + a1 a2 β = (a2 b1 − a1 b2 )a−2
2
a2 = 0, a0 2b2 k + b1 α = E(k)/(2a1 ),
− 1 J α, 12 ; σξ 2
a1 6= 0 a1 a1 σ = −a1 /(2b2 )
1 2b2 k + b1
a2 6= 0, a1 b2 p α= − ,
2
− a2 ξ α Z2α σ ξ 2
p
2a2
a1 = 4a0 a2 2a2 a2
σ = 2 E(k)
a2 = a1 = 0, b1 4(b0 + µ)b2 − b21 1/2 2 a0 1/2
− 1 ξ Z1/3 σξ 3/2 σ=
a0 6= 0 2b2 4a0 b2 3 b2

∂2w ∂2w ∂w
= (ax2 + b) + cx + kw + Φ(x, t)
∂t2 ∂x2 ∂x
∂2w ∂ 2w
1. = x2 + Φ(x, t).
∂t2 ∂x2
This is a special case of equation 6.3.2.2 with a = 1 and b = c = 0.
1◦ . Domain: 1 ≤ x ≤ a. First boundary value problem.

w = g2 (t) at x = a (boundary condition).
Solution:
Z tZ a
w(x, t) = Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ
0 1
Z a Z a
∂
+ f0(ξ)G(x, ξ, t) dξ + f1(ξ)G(x, ξ, t) dξ
∂t 1 1
Z t Z t
∂ ∂
+ g1(τ ) G(x, ξ, t − τ ) dτ − a2 g2 (τ ) G(x, ξ, t − τ ) dτ,
0 ∂ξ ξ=1 0 ∂ξ ξ=a
where
√ ∞ q
2 x X 1 πn
G(x, ξ, t)= 3/2 sin(µn ln x) sin(µn ln ξ) sin(λnt), µn = , λn = µ2n+ 14 .
ξ ln a n=1 λn ln a
2◦ . Domain: 1 ≤ x ≤ a. Second boundary value problem.


∂x w = g2 (t) at x = a (boundary condition).
Solution:
Z tZ a
0
Z1 a Z a
∂
+ f0 (ξ)G(x, ξ, t) dξ + f1 (ξ)G(x, ξ, t) dξ
∂t 1 1
Z t Z t
− g1 (τ )G(x, 1, t − τ ) dτ + a2 g2 (τ )G(x, a, t − τ ) dτ,
0 0
where
√ ∞
at 8 x X µ2n
G(x, ξ, t) = 2
+ ϕn (x)ϕn (ξ) sin(λn t),
(a − 1)ξ ξ 3/2 ln a n=1 λn (1 + µ2n )
q
1 πn
ϕn (x) = cos(µn ln x) − sin(µn ln x), µn = , λn = µ2n + 14 .
2µn ln a
∂2w ∂2w ∂w
2. = ax2 + bx + cw + Φ(x, t).
∂t2 ∂x2 ∂x
The substitution x = kez (k 6= 0) leads to the constant coefficient equation ∂tt w = a∂zz w+
(b − a)∂z w + cw + Φ(kez , t), which is discussed in Section 6.1.5.
∂2w ∂ 2w ∂w
3. = (ax2 + b) 2 + ax + cw.
∂t2 ∂x ∂x
Z
dx
The substitution z = √ leads to the constant coefficient equation ∂tt w = ∂zz w+
ax2 + b
cw, which is discussed in Section 6.1.3.
h i
∂2w 2 ∂ 2 2 ∂w
4. 2
= a (l − x ) + Φ(x, t).
∂t ∂x ∂x
Domain: 0 ≤ x ≤ l. First boundary value problem.

w = g(t) at x = 0 (boundary condition),
w 6= ∞ at x = l (boundedness condition).
Solution:
Z l Z l
∂
w(x, t) = f0(ξ)G(x, ξ, t) dξ + f1(ξ)G(x, ξ, t) dξ
∂t 0 0
Z t Z tZ l
∂
+ a2l2 g(τ ) G(x, ξ, t − τ ) dτ + Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ.
0 ∂ξ ξ=0 0 0
Here,
∞
1 X 4n − 1 x ξ p
G(x, ξ, t) = P2n−1 P2n−1 sin(λn at), λn = 2n(2n − 1),
al n=1 λn l l
1 dk 2
where Pk (x) = k k
(x − 1)k are the Legendre polynomials.
2 k! dx

h i
∂2w ∂2w 2 ∂w n(n + 1)
1. = a2 + − w , n = 1, 2, 3, . . .
∂t2 ∂r2 r ∂r r2
General solution:
n
n 1 ∂ Φ(r + at) + Ψ(r − at)
w(r, t) = r ,
r ∂r r
where Φ(r1 ) and Ψ(r2 ) are arbitrary functions.

∂2w ∂2w α ∂w
2. = + .
∂t2 ∂x2 x ∂x
The hyperbolic Euler–Poisson–Darboux equation.
1◦ . For α = 1 and α = 2, see Sections 6.2.1–6.2.4. For α 6= 1, the substitution z = x1−α
∂2w 2α 2
2 α−1 ∂ w
= (1 − α) z .
∂t2 ∂z 2
2◦ . Suppose wα = wα (x, t) is a solution of the equation in question for a fixed value of the
parameter α. Then the functions w eα defined by the relations
∂wα
w
eα = ,
∂t
∂wα ∂wα
w
eα = x +t ,
∂x ∂t
∂wα ∂wα
w
eα = 2xt + (x2 + t2 ) + αtwα
∂x ∂t
are also solutions of this equation.
3◦ . Suppose wα = wα (x, t) is a solution of the equation in question for a fixed value of the
parameter α. Using this wα , one can construct solutions of the equation with other values
of the parameter by the formulas
w2−α = xα−1 wα ,
∂wα
wα−2 = x + (α − 1)wα ,
∂x
∂wα ∂wα
wα−2 = xt + x2 + (α − 1)twα ,
∂x ∂t
∂wα ∂wα 2
wα−2 = x(x2 + t2 ) + 2x2 t + x + (α − 1)t2 wα ,
∂x ∂t
1 ∂wα
wα+2 = ,
x ∂x
t ∂wα ∂wα
wα+2 = + ,
x ∂x ∂t
x2 + t2 ∂wα ∂wα
wα+2 = + 2t + αwα .
x ∂x ∂t
⊙ The results of Items 2◦ and 3◦ were obtained by A. V. Aksenov (2001).
∂2w ∂2w 2a ∂w
3. = + + b2 w, 0 < 2a < 1.
∂t2 ∂x2 x ∂x
General solution:
Z
1
Φ t + x(2ξ − 1) ¯ p
w(x, t) = J a−1 2bx ξ(1 − ξ) dξ
0 [ξ(1 − ξ)]1−a
Z 1
1−2a Ψ t + x(2ξ − 1) ¯ p
+x J−a 2bx ξ(1 − ξ) dξ,
0 [ξ(1 − ξ)]a
where Φ(ξ1 ) and Ψ(ξ2 ) are arbitrary functions; J¯−ν (z) = Γ(1 − ν)2−ν z ν J−ν (z); J−ν (z)
is the Bessel function.
∂2w ∂2w
4. = ax4 + Φ(x, t).
∂t2 ∂x2
The transformation z = 1/x, u = w/x leads to the equation

∂2u ∂2u 1
= a 2 + zΦ ,t ,
∂t2 ∂z z

∂2w ∂ 2w
5. = (ax + b)4 .
∂t2 ∂x2
The transformation
w 1 1
u= , z = at + , y = −at +
ax + b ax + b ax + b
leads to the equation ∂zy u = 0. Thus, the general solution of the original equation has the
form
w = (ax + b)[f (z) + g(y)],
where f = f (z) and g = g(y) are arbitrary functions.
⊙ Literature: N. H. Ibragimov (1994).
∂2w ∂2w
6. = (a2 − x2 )2 + Φ(x, t).
∂t2 ∂x2
Domain: −l ≤ x ≤ l. First boundary value problem.

w=0 at x = l (boundary condition),
w=0 at x = −l (boundary condition).
Solution for 0 < l < a:

Z Z l Z tZ l
∂ l
w(x, t)= f (ξ)G(x, ξ, t) dξ+ g(ξ)G(x, ξ, t) dξ+ Φ(ξ, τ )G(x, ξ, t−τ ) dξ dτ,
∂t −l −l 0 −l
where
∞
X
2a 1
G(x, ξ, t) =
2 2 2
ϕn (x)ϕn (ξ) sin(λn t),
k(ξ − a ) n=1 λn
p
πn πn a+x ap 2 2 a+l
ϕn (x) = a2 − x2 sin + ln , λn = π n + k2 , k = ln .
2 2k a−x k a−l
∂2w ∂2w
7. = (x − a1 )2 (x − a2 )2 , a1 6= a2 .
∂t2 ∂x2
The transformation

x − a1
w(x, t) = (x − a2 )u(ξ, τ ), ξ = ln , τ = |a1 − a2 |t
x − a2
leads to the constant coefficient equation ∂τ τ u = ∂ξξ u − ∂ξ u, which is discussed in Sec-
tion 6.1.4.
∂2w ∂2w
8. = (ax2 + bx + c)2 .
∂t2 ∂x2
The transformation
p Z
dx
w(x, t) = u(z, t) |ax2 + bx + c|, z=
ax2
+ bx + c

leads to the constant coefficient equation ∂tt u = ∂zz u + ac − 14 b2 u, which is discussed
in Section 6.1.3.

∂2w ∂ ∂w
9. = a2 xm + Φ(x, t).
∂t2 ∂x ∂x
1◦ .Domain: 0 ≤ x ≤ l. First boundary value problem.
1.1. Case 0 < m < 1:

w = g(t) at x = l (boundary condition).
Solution:
Z l Z l
∂
∂t
0 0
Z t Z tZ l
2 m ∂
−a l g(τ ) G(x, ξ, t − τ ) dτ + Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ. (1)
0 ∂ξ ξ=l 0 0
Here,
∞
X yn (x)yn (ξ) sin(λn at) µn m−2
G(x, ξ, t) = , λn = (2 − m)l 2 , (2)
akyn k2 λn 2
n=1
where
2−m Z l

1−m x 2 1− m
yn (x) = x 2 Jp µn , 2
kyn k = yn2 (x) dx,
p= ;
l 0 2− m
the µn are positive zeros of the Bessel function, Jp (µ) = 0.

1.2. Case 1 ≤ m < 2:

w 6= ∞ at x = 0 (boundedness condition),
The solution is given by the formulas presented in Item 1.1.

2◦ . Domain: 0 ≤ x ≤ l. Mixed boundary value problem.


m
(x ∂x w) = 0 at x = 0 (boundary condition),
The solution for 0 < m < 1 is given by relations (1) and (2) with
x 2−m Z l
1−m 2 2 1−m
yn (x) = x 2 J−p µn , kyn k = yn2 (x) dx, p= ;
l 0 2−m
the µn are positive zeros of the Bessel function, J−p (µ) = 0.

3◦ . For Φ ≡ 0, the change of variable z = x1−m leads to an equation of the form 6.3.3.10:
∂2w 2
m 2
2 m−1 ∂ w
= a (1 − m) z .
∂t2 ∂z 2
∂2w ∂2w
10. = a 2 xm .
∂t2 ∂x2
1◦ . Particular solutions (A1 , A2 , B1 , B2 , and µ are arbitrary constants):
√ h i
w(x, t) = x A1 J 1 µxq + A2 Y 1 µxq B1 sin(aqµt) + B2 cos(aqµt) ,
2q 2q
√ h i
w(x, t) = x A1 I 1 µxq + A2 K 1 µxq B1 sinh(aqµt) + B2 cosh(aqµt) ,
2q 2q
where q = 12 (2 − m); Jν (z) and Yν (z) are Bessel functions; Iν (z) and Kν (z) are modified
Bessel functions.
2◦ . Below are discrete transformations that preserve the form of the original equation; what
changes is the parameter n.
2.1. The point transformation
1 w
z= , u= (transformation P)
x x
∂2u 2
2 4−m ∂ u
= a z .
∂t2 ∂z 2
The transformation P changes the equation parameter in accordance with the rule
P
m =⇒ 4−m. The double application of the transformation P yields the original equation.
2.2. Suppose w = w(x, t) is a solution of the original equation. Then the function
v = v(ξ, τ ), which is related to the solution w = w(x, t) by the Bäcklund transformation
∂ 1
v(ξ, τ ) = w(x, t), x = ξ 1−m , τ = |1 − m|t (transformation B),
∂x
is a solution of a similar equation
∂2v m 2
2 m−1 ∂ v
= a ξ .
∂τ 2 ∂ξ 2
The transformation B changes the equation parameters in accordance with the rule
B m
m =⇒ . The double application of the transformation B yields the original equa-
m−1
tion.
2.3. The composition of transformations F = B ◦ P changes the equation parameter as

follows:
F 4 − m F 8 − 3m F 12 − 5m F 16 − 7m F
m =⇒ =⇒ =⇒ =⇒ =⇒ . . .
3−m 5 − 2m 7 − 3m 9 − 4m
The n-fold application of the transformation F yields the equation with parameter
Fn 4n − (2n − 1)m
m =⇒ . (1)
2n + 1 − nm
2.4. The composition of transformations G = P ◦ B changes the equation parameter as
follows:
G 4 − 3m G 8 − 5m G 12 − 7m G 16 − 9m G
m =⇒ =⇒ =⇒ =⇒ =⇒ . . .
1−m 3 − 2m 5 − 3m 7 − 4m
The n-fold application of the transformation G yields the equation with parameter
Gn 4n − (2n + 1)m
m =⇒ . (2)
2n − 1 − nm
2.5. Setting m = 0 in (1) and (2), we arrive at two families of equations
∂2w 4n
2 2n+1 ∂2w
= a x at n = 1, 2, . . . ;
∂t2 ∂x2
∂2w 4n
2 2n−1 ∂2w
= a x at n = 1, 2, . . . ;
∂t2 ∂x2
whose solutions can be obtained with the aid of the wave equation; for this constant coeffi-
cient wave equation, see Section 6.1.1.
3◦ . Below are some useful transformations that lead to other equations.
3.1. The substitution ξ = x1−m leads to an equation of the form 6.3.3.9:
m
∂2w 2 2 ∂ ∂w
= a (1 − m) ξ m−1 .
∂t2 ∂ξ ∂ξ
2−m
3.2. The transformation τ = 12 a|2 − m|t, ξ = x 2 leads to an equation of the form
6.3.3.2:
∂2w ∂2w m 1 ∂w
= + .
∂τ 2 ∂ξ 2 m − 2 ξ ∂ξ
∂2w ∂2w
11. = tm .
∂t2 ∂x2
w = f (x) at t = 0,
∂t w = g(x) at t = 0.
Solution for m > 0:

Z
Γ(2β) 1 2 m+2
w(x, t) = 2 f x+ t 2 (2ξ − 1) [ξ(1 − ξ)]β−1 dξ
Γ (β) 0 m+2
Z 1
Γ(2 − 2β) 2 m+2
+ 2 t g x+ t 2 (2ξ − 1) [ξ(1 − ξ)]−β dξ,
Γ (1 − β) 0 m+2
where Z ∞
m
β= , Γ(z) = e−s sz−1 ds.
2(m + 2) 0


Solution for m > −1:

∞ h 1 1 i
√ X
w(x, t) = t An J−p 2pλn t 2p + Bn Jp 2pλn t 2p sin(λn x),
n=1
Z l
p2 1
An = Γ(1 − p)(λn p) f (x) sin(λn x) dx, p= ,
l 0 m+2
Z l
−p 2 πn
Bn = Γ(1 + p)(λn p) g(x) sin(λn x) dx, λn = ,
l 0 l
where Γ(p) is the gamma function.
m−2
∂2w ∂2w ∂w
12. = tm + bt 2 , m ≥ 2.
∂t2 ∂x2 ∂x
Domain: −∞ < x < ∞. Cauchy problem.
w = f (x) at t = 0,
∂t w = g(x) at t = 0.
1◦ . Solution for |b| < 12 m:

Z
Γ(α + β) 1 2 m+2
w(x, t) = f x+ t 2 (2ξ − 1) ξ β−1 (1 − ξ)α−1 dξ
Γ(α)Γ(β) 0 m+2
Z 1
Γ(2 − α − β) 2 m+2
+ t g x+ t 2 (2ξ − 1) ξ −α (1 − ξ)−β dξ,
Γ(1 − α)Γ(1 − β) 0 m+2
6.4. Equations Containing the First Time Derivative 607
where Z ∞
m − 2b m + 2b
α= , β= , Γ(z) = e−s sz−1 ds.
2(m + 2) 2(m + 2) 0
2◦ . Solution for b = 12 m:
Z 1 m
2 m+2 2t 2 m+2 −
w(x, t) = f x+ t 2 + g x+ t 2 (2ξ −1) (1−ξ) m+2 dξ.
m+2 m+2 0 m+2
3◦ . Solution for b = − 12 m:
Z 1 m
2 m+2 2t 2 m+2 −
w(x, t) = f x− t 2 + g x+ t 2 (2ξ −1) (1−ξ) m+2 dξ.
m+2 m+2 0 m+2
h i
∂2w ∂ ∂w
13. (b + x)2 = a2 (b + x)2 .
∂t2 ∂x ∂x
General solution:
f (x + at) + g(x − at)
w(x, t) = ,
b+x
where f (y) and g(z) are arbitrary functions.
6.4 Equations Containing the First Time Derivative

∂2w ∂w ∂2w ∂w
6.4.1 Equations of the Form +k = a2 +b +cw+Φ(x, t)
∂t2 ∂t ∂x2 ∂x
∂2w ∂w ∂ 2w
1. +k = a2 + Φ(x, t).
∂t2 ∂t ∂x2
For Φ(x, t) ≡ 0, this equation governs free transverse vibration of a string, and also lon-
gitudinal vibration of a rod in a resisting medium with a velocity-proportional resistance
coefficient.

1◦ . The substitution w(x, t) = exp − 12 kt u(x, t) leads to the equation
∂2u 2
2∂ u

= a + 14 k2 u + exp 1
2 kt Φ(x, t),
∂t2 ∂x2
1 p
E (x, t) = ϑ at − |x| exp − 12 kt I0 1
2k t2 − x2/a2 ,
2a
where ϑ(z) is the Heaviside unit step function and I0 (z) is the modified Bessel function.

w = f (x) at t = 0,
∂t w = g(x) at t = 0.
Solution:

w(x, t)= 12 exp − 12 kt f (x+at)+f (x−at)
Z x+at 1
p
kt I k t 2−(x−ξ)2/a2
1
+ exp − 12 kt 2p
f (ξ) dξ
4a x−at t2−(x−ξ)2/a2
Z p
1 1
x+at
+ exp − 2 kt I0 12 k t2−(x−ξ)2/a2 g(ξ)+ 12 kf (ξ) dξ
2a x−at
Z tZ x+a(t−τ ) p
1
+ exp − 12 k(t−τ ) I0 12 k (t−τ )2−(x−ξ)2/a2 Φ(ξ, τ ) dξ dτ,
2a 0 x−a(t−τ )

Solution:
Z tZ l
0 0
Z l Z l
∂
+ f0(ξ)G(x, ξ, t) dξ + f1(ξ) + kf0(ξ) G(x, ξ, t) dξ
∂t 0 0
Z t Z t
2 ∂ 2 ∂
+a g1 (τ ) G(x, ξ, t − τ ) dτ − a g2(τ ) G(x, ξ, t − τ ) dτ,
0 ∂ξ ξ=0 0 ∂ξ ξ=l
where
∞ r
2 kt X πnx πnξ sin(λnt) a2π 2n2 k2
G(x, ξ, t)= exp − sin sin , λn = − .
l 2 l l λn l2 4
n=1
Example 6.3. Consider the homogeneous equation (Φ ≡ 0). The initial shape of the string is a
triangle with base 0 ≤ x ≤ l and height h at x = c, that is,

 hx
 if 0 ≤ x ≤ c,
f (x) = c
 h(l − x) if c ≤ x ≤ l.

l−c

Solution:
∞
2hl2 1
X 1 nπc nπx
w(x, t) = 2 exp − 2 kt sin sin Θn (t),
π c(l − c) n=1
n2 l l
where
 k 2πna

cos(λn t) + 2λn sin(λn t)


if k <
l
, s
 a2 n2 π 2 k 2
kt 2πna
Θn (t) = 1 + if k = , λn = − .

 2 l l2 4

cosh(λn t) + k sinh(λn t)

if k >
2πna
,
2λn l
⊙ Literature: M. M. Smirnov (1975), B. M. Budak, A. N. Tikhonov, and A. A. Samarskii (1980).
5◦ . For the second and third boundary value problems on the interval 0 ≤ x ≤ l, see
equation 6.4.1.2 (Items 5◦ and 6◦ with b = 0).
6◦ . Domain: 0 ≤ x < ∞. A problem without initial conditions for Φ = 0.
The following boundary conditions are prescribed:
Solution:
where √ 2 √ 2
ω k + ω 2 − ω 2 1/2 ω k + ω 2 + ω 2 1/2
λ= , β= .
2a2 2a2
∂2w ∂w ∂ 2w
2. +k = a2 + bw + Φ(x, t).
∂t2 ∂t ∂x2
Telegraph equation (with k > 0, b < 0, and Φ(x, t) ≡ 0).

1◦ . The substitution w(x, t) = exp − 12 kt u(x, t) leads to the equation
∂2u 2
2∂ u

= a + (b + 14 k2 )u + exp 1
2 kt Φ(x, t),
∂t2 ∂x2
1 p
E (x, t) = ϑ at − |x| exp − 12 kt I0 c t2 − x2/a2 for b + 14 k2 = c2 > 0,
2a
1 p
E (x, t) = ϑ at − |x| exp − 12 kt J0 c t2 − x2/a2 for b + 14 k2 = −c2 < 0,
2a
where ϑ(z) is the Heaviside unit step function, J0 (z) and J1 (z) are Bessel functions, and
I0 (z) and I1 (z) are modified Bessel functions.

w = f (x) at t = 0,
∂t w = g(x) at t = 0.
Solution for b + 14 k2 = c2 > 0:

Z p
ct 1
x+at I1 c t2−(x−ξ)2/a2
+ exp − 2 kt p f (ξ) dξ
Z
1 1
x+at p
+ exp − 2 kt I0 c t2−(x−ξ)2/a2 g(ξ)+ 12 kf (ξ) dξ
2a x−at
Z tZ x+a(t−τ ) p
1
+ exp − 12 k(t−τ ) I0 c (t−τ )2−(x−ξ)2/a2 Φ(ξ, τ ) dξ dτ.
2a 0 x−a(t−τ )
Solution for b + 14 k2 = −c2 < 0:

Z x+at p
ct J c t 2−(x−ξ)2/a2
1
− exp − 12 kt p f (ξ) dξ
Z
1 1
x+at p
+ exp − 2 kt J0 c t2−(x−ξ)2/a2 g(ξ)+ 12 kf (ξ) dξ
2a x−at
Z tZ x+a(t−τ )
1 p
+ exp − 12 k(t−τ ) J0 c (t−τ )2−(x−ξ)2/a2 Φ(ξ, τ ) dξ dτ.
2a 0 x−a(t−τ )


Solution:
Z tZ l
0 0
Z l Z l
∂
+ f0(ξ)G(x, ξ, t) dξ + f1(ξ) + kf0(ξ) G(x, ξ, t) dξ
∂t 0 0
Z t Z t
2 ∂ 2 ∂
+a g1 (τ ) G(x, ξ, t − τ ) dτ − a g2(τ ) G(x, ξ, t − τ ) dτ.
0 ∂ξ ξ=0 0 ∂ξ ξ=l
Let a2 π 2 − bl2 − 14 k2 l2 > 0. Then

∞ √
2 kt X πnx πnξ sin t λn a2π 2n2 k2
G(x, ξ, t)= exp − sin sin √ , λn = −b− .
l 2 l l λn l2 4
n=1
Let a2 π 2 n2 − bl2 − 14 k2 l2 ≤ 0 for n = 1, . . . , m and a2 π 2 n2 − bl2 − 14 k2 l2 > 0 for

n = m + 1, m + 2, . . . Then
m √
2 kt X πnx πnξ sinh t βn
G(x, ξ, t) = exp − sin sin √
l 2 l l βn
n=1
X ∞ √
2 kt πnx πnξ sin t λn
+ exp − sin sin √ ,
l 2 n=m+1 l l λn
k2 a2 π 2 n2 a2 π 2 n2 k2
βn = b + − 2
, λn = 2
−b− .
4 l l 4
Solution:
Z tZ l
0 0
Z l Z l
∂
+ f0 (ξ)G(x, ξ, t) dξ + f1 (ξ) + kf0 (ξ) G(x, ξ, t) dξ
∂t 0 0
Z t Z t
2 2
0 0
1 2
For p = b + 4k < 0,
p ∞ p
sin t |p| 2X sin t a2 µ2n − p
G(x, ξ, t) = exp − 12 kt p + cos(µn x) cos(µn ξ) p ,
l |p| l a2 µ2n − p
n=1
µn = πn/l.
For p = b + 14 k2 > 0,
√ ∞ p
sinh t p 2X sin t a2 µ2n − p
G(x, ξ, t) = exp − 12 kt √ + cos(µn x) cos(µn ξ) p ,
l p l a2 µ2n − p
n=1
µn = πn/l.
If the inequality
p a2 µ2n − p < 0 holds for several
p first values n = 1, . . . , m, then the ex-
pressions a µn − p should be replaced by |a2 µ2n − p| and the sines by the hyperbolic
2 2
sines in the corresponding terms of the series.


∂x w − s1 w = g1 (t) at x = 0 (boundary condition),
∂x w + s2 w = g2 (t) at x = l (boundary condition).
The solution w(x, t) is determined by the formula in Item 5◦ with
∞ p
X yn (x)yn (ξ) sin t a2 µ2n − p
1
G(x, ξ, t) = exp − 2 kt p , p = b + 14 k2 ,
2 2
Bn a µ n − p
n=1

s1 s2 µ2n + s21 s1 l s21
yn (x) = cos(µn x) + sin(µn x), Bn = + 2 + 1+ 2 .
µn 2µ2n µ2n + s22 2µn 2 µn
tan(µl) s1 + s2
µ µ − s1 s2
If the inequality 2 2
a µn − p < 0 holds for several first values n = 1, . . . , m, then the ex-
p p
pressions a µn − p should be replaced by |a µ2n − p| and the sines by the hyperbolic
2 2 2
sines in the corresponding terms of the series.

7◦ . Domain: 0 ≤ x < ∞. A problem without initial conditions for Φ = 0.
The following boundary conditions are prescribed:
Solution:
where
p 2 2 p 1/2
k ω + (ω 2 + b)2 − ω 2 − b 1/2 k2 ω 2 + (ω 2 + b)2 + ω 2 + b
λ= , β= .
2a2 2a2
∂2w ∂w ∂ 2w ∂w
3. +k = a2 +b + cw + Φ(x, t).
∂t2 ∂t ∂x2 ∂x

1◦ . The substitution w(x, t) = exp − 12 a−2 bx − 12 kt u(x, t) leads to the equation
∂2u 2∂ u
2
1 2 1 −2 2
1 −2

= a + c + 4 k − 4 a b u + exp 2 a bx + 12 kt Φ(x, t),
∂t2 ∂x2
r
1 bx kt x2 k2 b2
E (x, t)= ϑ at−|x| exp − 2 − 2
I0 σ t − 2 if c+ − 2 =σ 2 >0,
2a 2a 2 a 4 4a
r
1 bx kt x2 k2 b2
E (x, t)= ϑ at−|x| exp − 2 − J0 σ t2− 2 if c+ − 2 =−σ 2 <0,
2a 2a 2 a 4 4a
where ϑ(z) is the Heaviside unit step function, J0 (z) and J1 (z) are Bessel functions, and
I0 (z) and I1 (z) are modified Bessel functions.

w = f (x) at t = 0,
∂t w = g(x) at t = 0.
Solution for c + 14 k2 − 14 a−2 b2 = σ 2 > 0:

1 kt bt bt
w(x, t)= exp − f (x+at) exp +f (x−at) exp −
2 2 2a 2a
Z x+at p
σt bx kt bξ I1 σ t −(x−ξ)2/a2
2
+ exp − 2 − exp p f (ξ) dξ
2a 2a 2 x−at 2a2 t2−(x−ξ)2/a2
Z x+at
1 bx kt bξ p
+ exp − 2 − exp 2
I0 σ t2−(x−ξ)2/a2 g(ξ)+ 12 kf (ξ) dξ
2a 2a 2 x−at 2a
Z tZ x+a(t−τ )
1 b(ξ−x) k(t−τ ) p
+ exp 2
− I0 σ (t−τ )2−(x−ξ)2/a2 Φ(ξ, τ ) dξ dτ.
2a 0 x−a(t−τ ) 2a 2
Solution for c + 14 k2 − 14 a−2 b2 = −σ 2 < 0:

1 kt bt bt
w(x, t)= exp − f (x+at) exp +f (x−at) exp −
2 2 2a 2a
Z x+at p
σt bx kt bξ J1 σ t −(x−ξ)2/a2
2
− exp − 2 − exp p f (ξ) dξ
2a 2a 2 x−at 2a2 t2−(x−ξ)2/a2
Z x+at
1 bx kt bξ p
+ exp − 2 − exp 2
J0 σ t2−(x−ξ)2/a2 g(ξ)+ 12 kf (ξ) dξ
2a 2a 2 x−at 2a
Z tZ x+a(t−τ )
1 b(ξ−x) k(t−τ ) p
+ exp − J0 σ (t−τ )2−(x−ξ)2/a2 Φ(ξ, τ ) dξ dτ.
2a 0 x−a(t−τ ) 2a2 2

Solution:
Z tZ l
w(x, t) = Φ(ξ, τ )G(x, ξ, t−τ ) dξ dτ
0 0
Z l Z l
∂
+ f0(ξ)G(x, ξ, t) dξ + f1(ξ)+kf0(ξ) G(x, ξ, t) dξ
∂t 0 0
Z t Z t
2 ∂ 2 ∂
+a g1(τ ) G(x, ξ, t−τ ) dτ −a g2(τ ) G(x, ξ, t−τ ) dτ.
0 ∂ξ ξ=0 0 ∂ξ ξ=l
Let a2 π 2 + 14 a−2 b2 l2 − cl2 − 14 k2 l2 > 0. Then

∞ √
2 b(ξ − x) kt X πnx πnξ sin t λn
G(x, ξ, t) = exp − sin sin √ ,
l 2a2 2 n=1 l l λn
a2 π 2 n2 b2 k2
λn = + − c − .
l2 4a2 4
Let
a2 π 2 n2 + 14 a−2 b2 l2 − cl2 − 14 k2 l2 ≤ 0 for n = 1, . . . , m;
a2 π 2 n2 + 14 a−2 b2 l2 − cl2 − 14 k2 l2 > 0 for n = m + 1, m + 2, . . .
Then
m √
2 b(ξ − x) kt X πnx πnξ sinh t βn
G(x, ξ, t) = exp − sin sin √
l 2a2 2 n=1 l l βn
X∞ √
2 b(ξ − x) kt πnx πnξ sin t λn
+ exp − sin sin √ ,
l 2a2 2 l l λn
n=m+1
k2 a2 π 2 n2 b2 a2 π 2 n2 b2 k2
where βn = c + − 2
− 2 and λn = 2
+ 2 −c− .
4 l 4a l 4a 4


Solution:
Z tZ l
0 0
Z l Z l
∂
+ f0 (ξ)G(x, ξ, t) dξ + f1 (ξ) + kf0 (ξ) G(x, ξ, t) dξ
∂t 0 0
Z t Z t
2 2
0 0
For p = c + 14 k2 < 0,
p
bξ kt sin t |p|
G(x, ξ, t) = A exp 2 − p
a 2 |p|
X∞ √
2 b(ξ − x) kt yn (x)yn (ξ) sin t λn
+ exp − √ ,
l 2a2 2 1 + µ2n λn
n=1
where
b a2 π 2 n2 b2 k2
A= , λn = + − c − ,
a2 ebl/a2 − 1 l2 4a2 4

πnx πnx bl
yn (x) = cos + µn sin , µn = 2 .
l l 2a πn
For p = c + 14 k2 > 0,
√
bξ kt sinh t p
G(x, ξ, t) = A exp 2 − √
a 2 p
X∞ √
2 b(ξ − x) kt yn (x)yn (ξ) sin t λn
+ exp − √ ,
l 2a2 2 1 + µ2n λn
n=1
where the coefficient A, λn , µn and the functions yn (x) remain as before. If √

the inequality
λn < 0 holdsp for several first values n = 1, . . . , m, then the expressions λn must be
replaced by |λn | and the sines by the hyperbolic sines in the corresponding terms of the
series.


∂x w − s1 w = g1 (t) at x = 0 (boundary condition),
∂x w + s2 w = g2 (t) at x = l (boundary condition).
The solution w(x, t) is determined by the formula in Item 5◦ with

∞ √
b(ξ − x) kt X yn (x)yn (ξ) sin t λn
G(x, ξ, t) = exp − √ .
2a2 2 Bn λn
n=1
Here,
2a2 s1 + b 2 2 b2 k2
yn (x) = cos(µn x) + sin(µ n x), λn = a µ n + − c − ,
2a2 µn 4a2 4
2a2 s2 − b 4a4 µ2n + (2a2 s1 + b)2 2a2 s1 + b l l(2a2 s1 + b)2
Bn = + + + ,
4a2 µ2n 4a4 µ2n + (2a2 s2 − b)2 4a2 µ2n 2 8a4 µ2n
tan(µl) 4a4 (s1 + s2 )

= 4 2 .
µ 4a µ − (2a2 s1 + b)(2a2 s2 − b)

∂2w ∂w ∂ 2w ∂w
+k = f (x) + g(x) + h(x)w + Φ(x, t)
∂t2 ∂t ∂x2 ∂x

∂2w ∂w ∂2w 1 ∂w
1. +k = a2 + .
∂t2 ∂t ∂r2 r ∂r
This equation describes vibration of a circular membrane in a resisting medium character-
ized by a velocity-proportional resistance coefficient.
1◦ . Domain: 0 ≤ r ≤ R. First boundary value problem.

∂t w = g(r) at t = 0 (initial condition),
w=0 at x = R (boundary condition).
Solution:
∞ r
X µn r a2 µ2n k2
w(r, t) = exp − 12 kt An cos(λn t)+Bn sin(λn t) J0 , λn = 2
− .
n=1
R R 4
Here,
Z R Z R
2 µnr Ank 2 µnr
An = 2 2 f (r)J0 r dr, Bn = + 2 2 g(r)J0 r dr,
R J1 (µn) 0 R 2λn λnR J1 (µn) 0 R
where the µn are positive zeros of the Bessel function, J0 (µ) = 0.

2◦ . For the solution of the second and third boundary value problems, see equation 6.4.2.2
(Items 3◦ and 4◦ with b = 0).

∂2w ∂w ∂2w 1 ∂w
2. +k = a2 + − bw + Φ(r, t).
∂t2 ∂t ∂r2 r ∂r

1◦ . The substitution w(r, t) = exp − 12 kt u(r, t) leads to the equation
2
∂2u 2 ∂ u 1 ∂u
2
=a 2
+ − b − 14 k2 u + exp 1
2 kt Φ(r, t),
∂t ∂r r ∂r


Solution:
Z R Z R
∂
w(r, t) = f0 (ξ)G(r, ξ, t) dξ + f1 (ξ) + kf0 (ξ) G(r, ξ, t) dξ
∂t 0 0
Z t Z tZ R
2 ∂
−a g(τ ) G(r, ξ, t − τ ) dτ + Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ.
0 ∂ξ ξ=R 0 0
Here,
∞ √
X 2ξ µn r µn ξ sin t λn
G(r, ξ, t) = exp − 12 kt J0 J0 √ ,
n=1
R2 J12 (µn ) R R λn
a2 µ2n k2
λn = + b − ,
R2 4
0 ≤ r ≤ R).
3◦ . Domain: 0 ≤ r ≤ R. Second boundary value problem.

Solution:
Z R Z R
∂
w(r, t) = f0 (ξ)G(r, ξ, t) dξ + f1 (ξ) + kf0 (ξ) G(r, ξ, t) dξ
∂t 0 0
Z t Z tZ R
+ a2 g(τ )G(r, R, t − τ ) dτ + Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ.
0 0 0
Here,
√ ∞ √
2ξ sin t λ0 2 X ξ µnr µnξ sin t λn
G(r, ξ, t)=exp − 12 kt √ + 2 2(µ ) J0 R J0 R
√ ,
R2 λ0 R
n=1
J0 n λn
where λ0 = b − 14 k2 ; λn = a2 µ2n R−2 + b − 14 k2 ; the µn are positive zeros of the first-order

Bessel function, J1 (µ) = 0. The numerical values of the first ten roots µn are specified in
Section 3.2.1 (see the second boundary value problem for 0 ≤ r ≤ R).
4◦ . Domain: 0 ≤ r ≤ R. Third boundary value problem.

∂r w + sw = g(t) at r = R (boundary condition).
The solution w(r, t) is given by the formula in Item 3◦ with

∞ √
2 1
X µ2n ξ µn r µn ξ sin t λn
G(r, ξ, t) = 2 exp − 2 kt J0 J0 √ .
R
n=1
(s2 R2 + µ2n )J02 (µn ) R R λn
Here, λn = a2 µ2n R−2 +b− 14 k2 and the µn are positive roots of the transcendental equation
µJ1 (µ) − sRJ0 (µ) = 0.

∂2w ∂w ∂2w 2 ∂w
3. +k = a2 + − bw + Φ(r, t).
∂t2 ∂t ∂r2 r ∂r

1◦ . The substitution w(r, t) = exp − 12 kt u(r, t) leads to the equation
2
∂2u 2 ∂ u 2 ∂u
2
=a 2
+ − b − 14 k2 u + exp 1
2 kt Φ(r, t),
∂t ∂r r ∂r


Solution:
Z R Z R
∂
w(r, t) = f0 (ξ)G(r, ξ, t) dξ f1 (ξ) + kf0 (ξ) G(r, ξ, t) dξ
∂t 0 0
Z t Z tZ R
2 ∂
−a g(τ ) G(r, ξ, t − τ ) dτ + Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ,
0 ∂ξ ξ=R 0 0
where
∞ √
2ξ 1
X nπr nπξ sin t λn a2π 2n2 k2
G(r, ξ, t)= exp − 2 kt sin sin √ , λn = +b− .
Rr n=1
R R λn R2 4
3◦ . Domain: 0 ≤ r ≤ R. Second boundary value problem.


Solution:
Z R Z R
∂
w(r, t) = f0 (ξ)G(r, ξ, t) dξ f1 (ξ) + kf0 (ξ) G(r, ξ, t) dξ
∂t 0 0
Z t Z tZ R
+ a2 g(τ )G(r, R, t − τ ) dτ + Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ,
0 0 0
where
√
3ξ 2 sin t λ0
G(r, ξ, t) = exp − 12 kt √
R3 λ0
∞ p
2ξ X µ2n + 1 µn r µn ξ
+ √ sin sin sin t λn .
Rr µ2n λn
n=1
R R
Here, λ0 = b − 14 k2 ; λn = a2 µ2n R−2 + b − 14 k2 ; and the µn are positive roots of the

transcendental equation tan µ − µ = 0. The numerical values of the first five roots µn are
specified in Section 3.2.3 (see the second boundary value problem for 0 ≤ r ≤ R).
4◦ . Domain: 0 ≤ r ≤ R. Third boundary value problem.

∂r w + sw = g(t) at r = R (boundary condition).
The solution w(r, t) is given by the formula in Item 3◦ with

∞ √
2ξ 1
X µ2n + (sR − 1)2 µn r µn ξ sin t λn
G(r, ξ, t) = exp − 2 kt sin sin √ .
Rr µ2 + sR(sR − 1)
n=1 n
R R λn
Here, λn = a2 µ2n R−2 +b− 14 k2 and the µn are positive roots of the transcendental equation
µ cot µ+sR−1 = 0. The numerical values of the first six roots µn can be found in Carslaw
and Jaeger (1984).

∂2w ∂w 2 ∂ ∂w
4. + k = a x − bw + Φ(x, t).
∂t2 ∂t ∂x ∂x

1◦ . The substitution w(r, t) = exp − 12 kt u(r, t) leads to an equation of the form 6.3.1.2:

∂2u 2 ∂ ∂u 1 2
1

= a x − b − 4 k u + exp 2 kt Φ(x, t).
∂t2 ∂x ∂x
Solution:
Z l l Z
∂
w(x, t) = f0 (ξ)G(x, ξ, t) dξ + f1 (ξ) + kf0 (ξ) G(x, ξ, t) dξ
∂t 0 0
Z t Z tZ l
∂
− a2 l g(τ ) G(x, ξ, t − τ ) dτ + Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ,
0 ∂ξ ξ=l 0 0
where
∞ r r √
1 1
X 1 x ξ sin t λn
G(x, ξ, t) = exp − 2 kt J0 µn J0 µn √ .
l J 2 (µn )
n=1 1
l l λn
Here, λn = 14 a2 µ2n l−1 +b− 14 k2 ; the µn are positive zeros of the Bessel function, J0 (µ) = 0.
The numerical values of the first ten µn are specified in Section 3.2.1 (see the first boundary

Solution:
Z l Z l
∂
w(x, t) = f0 (ξ)G(x, ξ, t) dξ +
f1 (ξ) + kf0 (ξ) dξ
∂t 0 0
Z t Z tZ l
+ a2 l g(τ )G(x, l, t − τ ) dτ + Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ,
0 0 0
where
√
sin t λ0
G(r, ξ, t) = exp − 12 kt
√
l λ0
∞ r r √
1X 1 x ξ sin t λn
+ 2 (µ ) J0 µn J0 µn √ .
l J0 n l l λn
n=1
Here, λ0 = b − 14 k2 ; λn = 14 a2 µ2n l−1 + b − 14 k2 ; the µn are positive zeros of the first-order

Bessel function, J1 (µ) = 0. The numerical values of the first ten roots µn are specified in
Section 3.2.1 (see the second boundary value problem for 0 ≤ r ≤ R).

∂x w + kw = g(t) at x = l (boundary condition).

∞ r r √
1 1
X µ2n x ξ sin t λn
G(r, ξ, t) = exp − 2 kt J µn
2l+µ2 )J 2(µ ) 0
J0 µn √ .
l n=1
(4k n 0 n l l λn
1 2 2 −1 1 2
Here, λn = 4 a µn l +b− 4k , and the µn are positive roots of the transcendental
equation √
µJ1 (µ) − 2k l J0 (µ) = 0.
∂2w ∂w ∂2w 1 ∂w
5. +k = (axm + b) 2 + amxm−1 + cw.
∂t2 ∂t ∂x 2 ∂x
Z
dx
The substitution z = √ leads to a constant coefficient equation of the form
axm + b
6.4.1.2: ∂tt w + k∂t w = ∂zz w + cw.

∂2w k − 1 ∂w ∂2w
1. + = .
∂t2 t ∂t ∂x2
Darboux equation. Domain: −∞ < x < ∞. Cauchy problem.
w = f (x) at t = 0,
∂t w = 0 at t = 0.
Solution:
k
Z 1
Γ k−3
w(x, t) = √ 2
k 1
f (x + tξ)(1 − ξ 2 ) 2 dξ (k > 1).
πΓ 2 − 2 −1
⊙ Literature: R. Courant and D. Hilbert (1989).
∂2w 2a ∂w ∂2w
2. + = − b2 w.
∂t2 t ∂t ∂x2
w = f (x) at t = 0,
2a
t ∂t w = g(x) at t = 0.
Solution for 0 < 2a < 1:

Z p
Γ(2a) 1
w(x, t) = 2 f x + t(2ξ − 1) J¯a−1 2bt ξ(1 − ξ) ξ a−1 (1 − ξ)a−1 dξ
Γ (a) 0
Z 1 p
Γ(2 − 2a)t1−2a
+ 2
g x + t(2ξ − 1) J¯−a 2bt ξ(1 − ξ) ξ −a(1 − ξ)−a dξ,
(1 − 2a)Γ (1 − a) 0
where Z ∞
J¯ν (z) = 2ν Γ(1 + ν)z −ν Jν (z), Γ(ν) = e−s sν−1 ds.
0
∂2w 2a ∂w ∂2w
3. + = tm .
∂t2 t ∂t ∂x2
w = f (x) at t = 0,
t2a ∂t w = g(x) at t = 0.
Solution for 0 ≤ 2a < 1 and m > 0:

Z
Γ(2β) 1 2 2+m
w(x, t) = 2 f x+ t 2 (2ξ − 1) ξ β−1 (1 − ξ)β−1 dξ
Γ (β) 0 2+m
Z 1
Γ(2 − 2β)t1−2a 2 2+m
+ g x+ t 2 (2ξ − 1) ξ −β (1 − ξ)−β dξ,
(1 − 2a)Γ2 (1 − β) 0 2+m
where Z ∞
m + 4a
β= , Γ(z) = e−s sz−1 ds.
2(m + 2) 0
∂2w ∂w ∂2w ∂w
4. t2 + kt = a2 +b + cw.
∂t2 ∂t ∂x2 ∂x
The substitution t = Aeτ (A 6= 0) leads to a constant coefficient equation of the form

6.4.1.3:
∂2w ∂w 2
2∂ w ∂w
2
+ (k − 1) = a 2
+b + cw.
∂τ ∂τ ∂x ∂x
∂2w ∂w ∂ 2w ∂w
5. t2 + kt = a 2 x2 + bx + cw.
∂t2 ∂t ∂x2 ∂x
The transformation
t = Aeτ , x = Beξ (A 6= 0, B 6= 0)
leads to a constant coefficient equation of the form 6.4.1.3:
∂2w ∂w 2
2∂ w ∂w
2
+ (k − 1) = a 2
+ (b − a2 ) + cw.
∂τ ∂τ ∂ξ ∂ξ
∂2w ∂w ∂2w
6. tm + atm−1 = , 0 < m < 2.
∂t2 ∂t ∂x2

w = f (x) at t = 0,
a
t ∂t w = g(x) at t = 0.
1◦ . Solution for 12 m < a < 1:
Z
Γ(2β) 1 2 2−m
w(x, t) = 2 f x+ t 2 (2ξ − 1) ξ β−1 (1 − ξ)β−1 dξ
Γ (β) 0 2−m
Z 1
Γ(2 − 2β) 1−a 2 2−m
+ t g x+ t 2 (2ξ − 1) ξ −β (1 − ξ)−β dξ,
(1 − a)Γ2 (1 − β) 0 2 − m
where
Z ∞
2a − m
β= , Γ(z) = e−s sz−1 ds.
2(2 − m) 0
2◦ . Solution for a = 12 m:
Z y
f (y) + f (z) 1
w(x, t) = + g(ξ) dξ,
2 2 z
2 2−m 2 2−m
y =x− t 2 , z =x+ t 2 .
2−m 2−m
∂ 2w 1 ∂w ∂ 2w ∂w
7. (tm + k) + mtm−1 =a +b + cw.
∂t2 2 ∂t ∂x2 ∂x
Z
dt
The substitution τ = √ leads to the equation ∂τ τ w = a∂xx w + b∂x w + cw,
+k tm
6.5 Equations Containing Arbitrary Functions

6.5.1 Equations of the
Form
∂ 2w ∂ ∂w
s(x) 2 = p(x) − q(x)w + Φ(x, t)
∂t ∂x ∂x
It is assumed that the functions s, p, p′x , and q are continuous and the inequalities s > 0,
p > 0 hold for x1 ≤ x ≤ x2 .
◮ General relations to solve linear nonhomogeneous boundary value problems.

The solution of the equation in question under the general initial conditions
w = f0 (x) at t = 0,
(1)
∂t w = f1 (x) at t = 0
a1 ∂x w + b1 w = g1 (t) at x = x1 ,
(2)
a2 ∂x w + b2 w = g2 (t) at x = x2
Z t Z x2
0 x1
Z x2 Z x2
∂
+ s(ξ)f0 (ξ)G(x, ξ, t) dξ + s(ξ)f1 (ξ)G(x, ξ, t) dξ
∂t x1 x1
Z t Z t
+ p(x1 ) g1 (τ )Λ1 (x, t − τ ) dτ + p(x2 ) g2 (τ )Λ2 (x, t − τ ) dτ. (3)
0 0
Here, the modified Green’s function is determined by
∞ √ Z x2
X yn (x)yn (ξ) sin t λn 2
G(x, ξ, t) = √
2 λ
, ky n k = s(x)yn2 (x) dx, (4)
n=1
ky n k n x1
where the λn and yn (x) are the eigenvalues and corresponding eigenfunctions of the Sturm–
Liouville problem for the second-order linear ordinary differential equation
[p(x)yx′ ]′x + [λs(x) − q(x)]y = 0,
a1 yx′ + b1 y = 0 at x = x1 , (5)
a2 yx′ + b2 y = 0 at x = x2 .
The functions Λ1 (x, t) and Λ2 (x, t) that occur in the integrands of the last two terms in so-
lution (3) are expressed in terms of the Green’s function of (4). The corresponding formulas
will be specified below in studying specific boundary value problems.
General properties of the Sturm–Liouville problem (5):
1◦ . There are finitely many eigenvalues λ1 < λ2 < λ3 < · · · , with λn → ∞ as n → ∞;
hence the number of negative eigenvalues is finite.
2◦ . Any two eigenfunctions yn (x) and ym (x) for n 6= m are orthogonal to each other with
weight s(x) on the interval x1 ≤ x ≤ x2 ; specifically,
Z x2
s(x)yn (x)ym (x) dx = 0 at n 6= m.
x1
q(x) ≥ 0, a1 b1 ≤ 0, a2 b2 ≥ 0 (6)
are satisfied, then there are no negative eigenvalues. If q ≡ 0 and b1 = b2 = 0, the least
eigenvalue is λ1 = 0 and the corresponding eigenfunction is ϕ1 = const. In the other cases
where conditions (6) are satisfied, all eigenvalues are positive.
Remark 6.2. More detailed information about the properties of the Sturm–Liouville problem (5)
can be found in Section 3.8.9. Asymptotic and approximate formulas for eigenvalues and eigenfunc-
tions are also presented there.
◮ First boundary value problem (case a1 = a2 = 0, b1 = b2 = 1).

The solution of the first boundary value problem for the equation in question with the initial
conditions (1) and the boundary conditions
w = g1 (t) at x = x1 ,
w = g2 (t) at x = x2
is given by relations (3) and (4) in which

∂ ∂

Λ1 (x, t) = G(x, ξ, t) , Λ2 (x, t) = − G(x, ξ, t) .
◮ Second boundary value problem (case a1 = a2 = 1, b1 = b2 = 0).

The solution of the second boundary value problem for the equation in question with the
initial conditions (1) and the boundary conditions
∂x w = g1 (t) at x = x1 ,
∂x w = g2 (t) at x = x2
is given by relations (3) and (4) with
Λ1 (x, t) = −G(x, x1 , t), Λ2 (x, t) = G(x, x2 , t).
◮ Third boundary value problem (case a1 = a2 = 1, b1 6= 0, b2 6= 0).

The solution of the third boundary value problem for the equation in question with the
initial conditions (1) and the boundary conditions (2) with a1 = a2 = 1 is given by relations
(3) and (4) in which
Λ1 (x, t) = −G(x, x1 , t), Λ2 (x, t) = G(x, x2 , t).
◮ Mixed boundary value problem (case a1 = b2 = 0, a2 = b1 = 1).

The solution of the mixed boundary value problem for the equation in question with the
initial conditions (1) and the boundary conditions
w = g1 (t) at x = x1 ,
∂x w = g2 (t) at x = x2

∂

Λ1 (x, t) = G(x, ξ, t) , Λ2 (x, t) = G(x, x2 , t).
∂ξ ξ=x1
◮ Mixed boundary value problem (case a1 = b2 = 1, a2 = b1 = 0).
The solution of the mixed boundary value problem with the initial conditions (1) and the
boundary conditions
∂x w = g1 (t) at x = x1 ,
w = g2 (t) at x = x2
∂

Λ1 (x, t) = −G(x, x1 , t), Λ2 (x, t) = − G(x, ξ, t) .
∂ξ ξ=x2
⊙ Literature for Section 6.5.1: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), V. A. Marchenko
(1986), V. S. Vladimirov (1988), A. D. Polyanin (2000a).
6.5.2 Equations of the Formn h i o

∂2w ∂w ∂ ∂w
+ a(t) = b(t) p(x) − q(x)w + Φ(x, t)
∂t2 ∂t ∂x ∂x
It is assumed that the functions p, p′x , and q are continuous and p > 0 for x1 ≤ x ≤ x2 .
◮ General relations to solve linear nonhomogeneous boundary value problems.
The solution of the equation in question under the general initial conditions
w = f0 (x) at t = 0,
(1)
∂t w = f1 (x) at t = 0
s1 ∂x w + k1 w = g1 (t) at x = x1 ,
(2)
s2 ∂x w + k2 w = g2 (t) at x = x2

Z tZ x2
w(x, t) = Φ(ξ, τ )G(x, ξ, t, τ ) dξ dτ
0 x1
Z x2 Z x2
∂
− f0 (ξ) G(x, ξ, t, τ ) dξ + f1 (ξ) + a(0)f0 (ξ) G(x, ξ, t, 0) dξ
x1 ∂τ τ =0 x1
Z t Z t
+ p(x1 ) g1 (τ )b(τ )Λ1 (x, t, τ ) dτ + p(x2 ) g2 (τ )b(τ )Λ2 (x, t, τ ) dτ. (3)
0 0
Here, the modified Green’s function is determined by
X∞ Z x2
yn (x)yn (ξ) 2
G(x, ξ, t, τ ) = Un (t, τ ), kyn k = yn2 (x) dx, (4)
n=1
kyn k2 x1
where the λn and yn (x) are the eigenvalues and corresponding eigenfunctions of the Sturm–
Liouville problem for the following second-order linear ordinary differential equation with
homogeneous boundary conditions:
[p(x)yx′ ]′x + [λ − q(x)]y = 0,

s1 yx′ + k1 y = 0 at x = x1 , (5)
s2 yx′ + k2 y = 0 at x = x2 .
The functions Un = Un (t, τ ) are determined by solving the Cauchy problem for the linear
ordinary differential equation
Un′′ + a(t)Un′ + λn b(t)Un = 0,

(6)
Un = 0, Un′
t=τ
= 1.
t=τ
The prime denotes the derivative with respect to t, and τ is a free parameter occurring in
the initial conditions.
The functions Λ1 (x, t) and Λ2 (x, t) that occur in the integrands of the last two terms
in solution (3) are expressed in terms of the Green’s function of (4). The corresponding
formulas will be specified below when studying specific boundary value problems.
The properties of the Sturm–Liouville problem (5) are detailed in Section 3.8.9. Asymp-
totic and approximate formulas for eigenvalues and eigenfunctions are also presented there.
◮ First, second, third, and mixed boundary value problems.

1◦ . First boundary value problem. The solution of the equation in question with the initial
conditions (1) and boundary conditions (2) for s1 = s2 = 0 and k1 = k2 = 1 is given by
relations (3) and (4), where
∂ ∂

Λ1 (x, t, τ ) = G(x, ξ, t, τ ) , Λ2 (x, t, τ ) = − G(x, ξ, t, τ ) .
2◦ . Second boundary value problem. The solution of the equation with the initial condi-
tions (1) and boundary conditions (2) for s1 = s2 = 1 and k1 = k2 = 0 is given by relations
(3) and (4) with
Λ1 (x, t, τ ) = −G(x, x1 , t, τ ), Λ2 (x, t, τ ) = G(x, x2 , t, τ ).
3◦ . Third boundary value problem. The solution of the equation with the initial condi-
tions (1) and boundary conditions (2) for s1 = s2 = 1 and k1 k2 6= 0 is given by relations
(3) and (4) in which
4◦ . Mixed boundary value problem. The solution of the equation with the initial condi-
tions (1) and boundary conditions (2) for s1 = k2 = 0 and s2 = k1 = 1 is given by relations
(3) and (4) with
∂

Λ1 (x, t, τ ) = G(x, ξ, t, τ ) , Λ2 (x, t, τ ) = G(x, x2 , t, τ ).
∂ξ ξ=x1
5◦ . Mixed boundary value problem. The solution of the equation with the initial condi-
tions (1) and boundary conditions (2) for s1 = k2 = 1 and s2 = k1 = 0 is given by relations
(3) and (4) with
∂

Λ1 (x, t, τ ) = −G(x, x1 , t, τ ), Λ2 (x, t, τ ) = − G(x, ξ, t, τ ) .
∂ξ ξ=x2
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), A. V. Bitsadze and D. F. Kalini-
chenko (1985), A. D. Polyanin (2000a).

∂2w ∂ 2w
1. = f (x) .
∂t2 ∂x2
This is a special case of the equation of Section 6.5.1 with s(x) = 1/f (x), p(x) = 1, and
q = Φ = 0.
w = C1 xt + C2 t + C3 x + C4 ,
Z x
2 x−ξ
w = C1 t + C2 xt + C3 t + C4 x + 2C1 dξ + C5 ,
a f (ξ)
Z x
x−ξ
w = C1 t3 + C2 xt + C3 t + C4 x + 6C1 t dξ + C5 ,
a f (ξ)
Z x
2 (C1 ξ + C2 )
w = (C1 x + C2 )t + C3 xt + C4 t + C5 x + 2 (x − ξ) dξ + C6 ,
a f (ξ)
where C1 , . . . , C6 are arbitrary constants and a is an arbitrary real number.
2◦ . Separable particular solution:
w = (C1 eλt + C2 e−λt )H(x),
where C1 , C2 , and λ are arbitrary constants, and the function H = H(x) is determined by
the ordinary differential equation f (x)Hxx′′ − λ2 H = 0.

w = [C1 sin(λt) + C2 cos(λt)]Z(x),
where C1 , C2 , and λ are arbitrary constants, and the function Z = Z(x) is determined by
′′ + λ2 Z = 0.
the ordinary differential equation f (x)Zxx
4◦ . Particular solutions with even powers of t:
n
X
w= ϕk (x)t2k ,
k=0
where the functions ϕk = ϕk (x) are defined by the recurrence relations

ϕn (x) = An x + Bn ,
Z x
ϕk (ξ)
ϕk−1 (x) = Ak x + Bk + 2k(2k − 1) (x − ξ) dξ,
a f (ξ)
where Ak , Bk are arbitrary constants (k = n, . . . , 1).
5◦ . Particular solutions with odd powers of t:

n
X
w= ψk (x)t2k+1 ,
k=0
where the functions ψk = ψk (x) are defined by the recurrence relations
ψn (x) = An x + Bn ,
Z x
ψk (ξ)
ψk−1 (x) = Ak x + Bk + 2k(2k + 1) (x − ξ) dξ,
a f (ξ)

h i
∂2w ∂ ∂w
2. = f (x) .
∂t2 ∂x ∂x
This is a special case of the equation of Section 6.5.1 with s(x) = 1, p(x) = f (x), and
q = Φ = 0.
Z
2 C1 x + C3
w = C1 t + C2 t + 2 dx + C4 ,
f (x)
Z
C1 x + C3
w = C1 t3 + C2 t + 6t dx + C4 ,
f (x)
Z
dx
w = [C1 Φ(x) + C2 ]t + C3 Φ(x) + C4 ,Φ(x) = ,
f (x)
Z n Z o
2 1
w = [C1 Φ(x) + C2 ]t + C3 Φ(x) + C4 + 2 [C1 Φ(x) + C2 ] dx dx,
f (x)
where C1 , C2 , C3 , and C4 are arbitrary constants.

w = (C1 eλt + C2 e−λt )H(x),
the ordinary differential equation [f (x)Hx′ ]′x − λ2 H = 0.
w = [C1 sin(λt) + C2 cos(λt)]Z(x),
the ordinary differential equation [f (x)Zx′ ]′x + λ2 Z = 0.
4◦ . Particular solutions with even powers of t:
n
X
w= ζk (x)t2k ,
k=0
where the functions ζk = ζk (x) are defined by the recurrence relations

Z
dx
ζn (x) = An Φ(x) + Bn , Φ(x) = ,
f (x)
Z Z
1 n o
ζk−1 (x) = Ak Φ(x) + Bk + 2k(2k − 1) ζk (x) dx dx,
f (x)

5◦ . Particular solutions with odd powers of t:
n
X
w= ηk (x)t2k+1 ,
k=0
where the functions ηk = ηk (x) are defined by the recurrence relations

Z
dx
ηn (x) = An Φ(x) + Bn , Φ(x) = ,
f (x)
Z Z
1 n o
ηk−1 (x) = Ak Φ(x) + Bk + 2k(2k + 1) ηk (x) dx dx,
f (x)
∂2w ∂ 2w ∂w
3. = f (x) + g(x) + Φ(x, t), 0 < f (x) < ∞.
∂t2 ∂x2 ∂x
This equation can be rewritten in the form of the equation from Section 6.5.1 with q(x) ≡ 0:

∂2w ∂ ∂w
s(x) = p(x) + s(x)Φ(x, t),
∂t2 ∂x ∂x
where Z Z
1 g(x) g(x)
s(x) = exp dx , p(x) = exp dx .
f (x) f (x) f (x)
∂2w ∂ 2w ∂w
4. = f (x) + g(x) + h(x)w + Φ(x, t).
∂t2 ∂x2 ∂x
This equation can be rewritten in the form of the equation from Section 6.5.1:

∂2w ∂ ∂w
s(x) 2 = p(x) − q(x)w + s(x)Φ(x, t),
∂t ∂x ∂x
where Z Z
1 g(x) g(x)
s(x) = exp dx , p(x) = exp dx ,
f (x) f (x) f (x)
Z
h(x) g(x)
q(x) = − exp dx .
f (x) f (x)
∂2w ∂ 2w ∂w
5. = f (x) + g(x) + h1 (x) + h2 (t) w.
∂t2 ∂x2 ∂x
1◦ . There are separable solutions in the product form w(x, t) = ϕ(x)ψ(t), where the func-
tions ϕ = ϕ(x) and ψ = ψ(t) satisfy the ordinary differential equations (λ is an arbitrary
constant):

f (x)ϕ′′xx + g(x)ϕ′x + λ + h1 (x) ϕ = 0, ′′
ψtt + λ − h2 (t) ψ = 0.
2◦ . For the solution of various boundary value problems for the original equation, see
∂2w ∂ 2w 1 ′ ∂w
6. = f (x) + f (x) + bw.
∂t2 ∂x2 2 ∂x
Z
dx
The substitution z = p leads to the constant coefficient equation ∂tt w = ∂zz w +
f (x)
bw, which is discussed in Section 6.1.3.
∂2w ∂2w ∂w
7. = f2 + f (fx′ + 2g) + (f gx′ + g 2 )w, f = f (x), g = g(x).
∂t2 ∂x2 ∂x
The transformation
Z Z
g dx
w(x, t) = u(ξ, t) exp − dx , ξ=
f f (x)
leads to the wave equation ∂tt u = ∂ξξ u, which is discussed in Section 6.1.1.
∂2w ∂w ∂ 2w 1 ∂w
8. +a = f (x) 2 + f ′ (x) + bw.
∂t2 ∂t ∂x 2 ∂x
Z
dx
The substitution z = p leads to a constant coefficient equation of the form 6.4.1.2:
f (x)
∂tt w + a∂t w = ∂zz w + bw.
∂ 2w 1 ′ ∂w ∂2w ∂w
9. f (t) + f (t) =a +b + cw.
∂t2 2 ∂t ∂x2 ∂x
Z
dt
The substitution τ = p leads to the equation ∂τ τ w = a∂xx w + b∂x w + cw, which
f (t)
is discussed in Section 6.1.5.
∂ 2w 1 ′ ∂w ∂ 2w 1 ′ ∂w
10. f (t) + f (t) = g(x) + g (x) + cw.
∂t2 2 ∂t ∂x2 2 ∂x
Z Z
dt dx
The transformation τ = p , z= p leads to the constant coefficient equa-
f (t) g(x)
tion ∂τ τ w = ∂zz w + cw, which is discussed in Section 6.1.3.
Chapter 7
Second-Order
Hyperbolic Equations
∂ 2w
7.1 Wave Equation = a2∆2 w
∂t2
The wave equation with two space variables in the rectangular Cartesian system of coordi-
nates has the form
2
∂2w 2 ∂ w ∂2w
=a + .
∂t2 ∂x2 ∂y 2
◮ Particular solutions and some relations.
p
w(x, y, t) = A exp k1 x + k2 y ± at k12 + k22 ,
p
w(x, y, t) = A sin(k1 x + C1 ) sin(k2 y + C2 ) sin at k12 + k22 ,
p
w(x, y, t) = A sin(k1 x + C1 ) sin(k2 y + C2 ) cos at k12 + k22 ,
p
w(x, y, t) = A sinh(k1 x + C1 ) sinh(k2 y + C2 ) sinh at k12 + k22 ,
p
w(x, y, t) = A sinh(k1 x + C1 ) sinh(k2 y + C2 ) cosh at k12 + k22 ,
w(x, y, t) = ϕ(x sin β + y cos β + at) + ψ(x sin β + y cos β − at),
where A, C1 , C2 , k1 , k2 , and β are arbitrary constants, and ϕ(z) and ψ(z) are arbitrary
functions.
633
634 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T WO S PACE VARIABLES
2◦ . Particular solutions that are expressed in terms of solutions to simpler equations:

w(x, y, t)= A cos(ky)+B sin(ky) u(x, t) for ∂ttu=a2∂xxu−a2k2u, (1)
2 2 2
w(x, y, t)= A cosh(ky)+B sinh(ky) u(x, t) for ∂ttu=a ∂xxu+a k u, (2)
2
w(x, y, t)= A cos(kt)+B sin(kt) u(x, y) for ∂xxu+∂yy u=−(k/a) u, (3)

w(x, y, t)= A cosh(kt)+B sinh(kt) u(x, y) for ∂xxu+∂yy u=(k/a)2 u, (4)

at±y at∓y
w(x, y, t)=exp u(x, τ ), τ = for ∂τ u=b∂xxu. (5)
2b 2
For particular solutions of equations (1) and (2) for the function u(x, t), see the Klein–
Gordon equation 6.1.3. For particular solutions of equations (3) and (4) for the function
u(x, y), see Section 9.3.2. For particular solutions of the heat equation (5) for the function
u(x, τ ), see Section 3.1.1.
(
ϑ(at − r) 1 for z > 0,
E (x, y, t) = √ , ϑ(z) =
2πa a2 t2 − r 2 0 for z ≤ 0,
p
where r = x2 + y 2 .
4◦ . Infinite series solutions that contain arbitrary functions of the space variables:
X∞
(at)2n n ∂2 ∂2
w(x, y, t) = f (x, y) + ∆ f (x, y), ∆≡ + ,
n=1
(2n)! ∂x2 ∂y 2
∞
X (at)2n
w(x, y, t) = tg(x, y) + t ∆n g(x, y),
(2n + 1)!
n=1
where f (x, y) and g(x, y) are any infinitely differentiable functions. The first solution
satisfies the initial conditions w(x, y, 0) = f (x, y), ∂t w(x, y, 0) = 0 and the second solution
satisfies the initial conditions w(x, y, 0) = 0, ∂t w(x, y, 0) = g(x, y). The sums are finite if
f (x, y) and g(x, y) are bivariate polynomials.
5◦ . A wide class of solutions to the wave equation with two space variables is described
by the formulas
w(x, y, t) = Re F (θ) and w(x, y, t) = Im F (θ). (6)
Here, F (θ) is an arbitrary analytic function of the complex argument θ connected with the
variables (x, y, t) by the implicit relation
p
at − (x − x0 )θ + (y − y0 ) 1 − θ 2 = G(θ), (7)
where G(θ) is any analytic function and x0 , y0 are arbitrary constants. Solutions of the
forms (6), (7) find wide application in the theory of diffraction. If the argument θ ob-
tained by solving (7) with a prescribed G(θ) is real in some domain D, then one should set
Re F (θ) = F (θ) in relation (6) everywhere in D.
⊙ Literature: V. I. Smirnov (1974, Vol. 3, Pt. 2).
∂2w
7.1. Wave Equation ∂t2
= a 2 ∆2 w 635
6◦ . Suppose w = w(x, y, t) is a solution of the wave equation. Then the functions

w1 = Aw(±λx + C1 , ±λy + C2 , ±λt + C3 ),

w2 = Aw p , y, p ,
1 − (v/a)2 1 − (v/a)2

A x y t
w3 = p w 2 , , ,
|r 2 − a2 t2 | r − a2 t2 r 2 − a2 t2 r 2 − a2 t2

A x + k1 (a2 t2 − r 2 ) y + k2 (a2 t2 − r 2 ) at + k3 (a2 t2 − r 2 )
w4 = √ w , , ,
Ξ Ξ Ξ aΞ
r 2 = x2 + y 2 , Ξ = 1 − 2(k1 x + k2 y − ak3 t) + (k12 + k22 − k32 )(r 2 − a2 t2 ),
where A, C1 , C2 , C3 , k1 , k2 , k3 , v, and λ are arbitrary constants, are also solutions of
the equation. The signs at λ in the expression of w1 can be taken independently of one
another. The function w2 results from the invariance of the wave equation under the Lorentz
transformation.
More detailed information about particular solutions and transformations of the wave
equation with two space variables can be found in the references cited below.
⊙ Literature: E. Kalnins and W. Miller, Jr. (1975, 1976), W. Miller, Jr. (1977).

w = f (x, y) at t = 0,
∂t w = g(x, y) at t = 0.
Solution (Poisson’s formula):
ZZ ZZ
1 ∂ f (ξ, η) dξ dη 1 g(ξ, η) dξ dη
w(x, y, t) = p + p ,
2πa ∂t 2 2 2
a t − (ξ − x) − (η − y)2 2πa a t − (ξ − x)2 − (η − y)2
2 2
Cat Cat
where the integration is performed over the interior of the circle of radius at with center
at (x, y).
⊙ Literature: N. S. Koshlyakov, E. B. Gliner, and M. M. Smirnov (1970), A. N. Tikhonov and A. A. Samarskii
(1990).

w = f0 (x, y) at t = 0 (initial condition),
∂t w = f1 (x, y) at t = 0 (initial condition),
Solution:
Z l1Z l2
∂
w(x, y, t) = f0 (ξ, η)G(x, y, ξ, η, t) dη dξ
∂t 0 0
Z l1Z l2
+ f1 (ξ, η)G(x, y, ξ, η, t) dη dξ
0 0
Z tZ l2
2 ∂
0 0 ∂ξ ξ=0
Z t Z l2
2 ∂
0 0 ∂ξ ξ=l1
Z t Z l1
2 ∂
0 0 ∂η η=0
Z t Z l1
∂
− a2 g4 (ξ, τ ) G(x, y, ξ, η, t − τ ) dξ dτ,
0 0 ∂η η=l2
where
∞ ∞
4 XX 1
G(x, y, ξ, η, t) = sin(pn x) sin(qm y) sin(pn ξ) sin(qm η) sin(aλnm t),
al1 l2 λ
n=1 m=1 nm
nπ mπ p
pn = , qm = , λnm = p2n + qm 2 .
l1 l2
The problem of vibration of a rectangular membrane with sides l1 and l2 rigidly fixed in
its contour is characterized by homogeneous boundary conditions, gs ≡ 0 (s = 1, 2, 3, 4).
⊙ Literature: M. M. Smirnov (1964), B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).

∂2w
= a 2 ∆2 w 637
Solution:
Z l1Z l2
∂
∂t 0 0
Z l1Z l2
+ f1 (ξ, η)G(x, y, ξ, η, t) dη dξ
0 0
Z tZ l2
− a2 g1 (η, τ )G(x, y, 0, η, t − τ ) dη dτ
0 0
Z tZ l2
2
+a g2 (η, τ )G(x, y, l1 , η, t − τ ) dη dτ
0 0
Z tZ l1
2
−a g3 (ξ, τ )G(x, y, ξ, 0, t − τ ) dξ dτ
0 0
Z tZ l1
+ a2 g4 (ξ, τ )G(x, y, ξ, l2 , t − τ ) dξ dτ,
0 0
where
X∞ X ∞
t 2 A
G(x, y, ξ, η, t) = + p nm cos(pn x) cos(qm y)
l1 l2 al1 l2 n=0 m=0 p2n + qm 2

× cos(pn ξ) cos(qm η) sin(aλnm t) ,


0 for n = m = 0,
nπ mπ
pn = , qm = , Anm = 1 for nm = 0 (n 6= m),
l1 l2 

2 for nm 6= 0.
w = f0 (x, y) at t=0 (initial condition),

∂t w = f1 (x, y) at t=0 (initial condition),

∞ ∞
4XX 1
G(x, y, ξ, η, t) = p sin(µn x + εn ) sin(νm y + σm )
a n=1 m=1 Enm µ2n + νm 2
p
× sin(µn ξ + εn ) sin(νm η + σm ) sin at µ2n + νm 2 ,
µn νm
εn = arctan , σm = arctan ,
l1 l2
2 )(k + k )
(k1 k2 + µ2n )(k1 + k2 ) (k3 k4 + νm 3 4
Enm = l1 + l2 + ,
(k12 + µ2n )(k22 + µ2n ) (k32 + νm2 )(k 2 + ν 2 )
4 m
where the µn and νm are positive roots of the transcendental equations
µ2 − k1 k2 = (k1 + k2 )µ cot(l1 µ), ν 2 − k3 k4 = (k3 + k4 )ν cot(l2 ν).

Solution:
Z l1Z l2
∂
∂t 0 0
Z l1Z l2
+ f1 (ξ, η)G(x, y, ξ, η, t) dη dξ
0 0
Z t Z l2
2 ∂
0 0 ∂ξ ξ=0
Z t Z l2
2 ∂
0 0 ∂ξ ξ=l1
Z t Z l1
− a2 g3 (ξ, τ )G(x, y, ξ, 0, t − τ ) dξ dτ
0 0
Z t Z l1
0 0
∂2w
= a 2 ∆2 w 639
where
∞ ∞
2 X X Am
G(x, y, ξ, η, t) = sin(pn x) cos(qm y) sin(pn ξ) cos(qm η) sin(aλnm t),
al1 l2 λnm
n=1 m=0
(
nπ mπ p 1 for m = 0,
pn = , qm = , λnm = p2n + qm 2 , Am =
l1 l2 2 for m 6= 0.
Solution:
Z l1Z l2
∂
∂t 0 0
Z l1Z l2
+ f1 (ξ, η)G(x, y, ξ, η, t) dη dξ
0 0
Z tZ l2
2 ∂
0 0 ∂ξ ξ=0
Z tZ l2
+ a2 g2 (η, τ )G(x, y, l1 , η, t − τ ) dη dτ
0 0
Z tZ l1
2 ∂
0 0 ∂η η=0
Z tZ l1
0 0
where
∞ ∞
4 XX 1
G(x, y, ξ, η, t) = sin(pn x) sin(qm y) sin(pn ξ) sin(qm η) sin(aλnm t),
al1 l2 n=0 m=0 λnm
π(2n + 1) π(2m + 1) p
pn = , qm = , λnm = p2n + qm
2 .
2l1 2l2

The wave equation with two space variables in the polar coordinate system has the form
2 p
∂2w 2 ∂ w 1 ∂w 1 ∂2w
2
=a 2
+ + 2 2
, r = x2 + y 2 .
∂t ∂r r ∂r r ∂ϕ
One-dimensional solutions w = w(r, t) that are independent of the angular coordinate ϕ

are considered in Section 6.2.1.

w = f0 (r, ϕ) at t = 0 (initial condition),
∂t w = f1 (r, ϕ) at t = 0 (initial condition),
Solution:
Z 2πZ R
∂
w(r, ϕ, t) = f0 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
∂t 0 0
Z 2πZ R
+ f1 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
0 0
Z tZ 2π
2 ∂
−a R g(η, τ ) G(r, ϕ, ξ, η, t − τ ) dη dτ.
0 0 ∂ξ ξ=R
Here,
∞ ∞
1 XX An
G(r, ϕ, ξ, η, t) = Jn (µnm r)Jn (µnm ξ) cos[n(ϕ−η)] sin(µnm at),
πaR2 n=0 m=1 µnm [Jn′ (µnm R)]2
A0 = 1, An = 2 (n = 1, 2, . . .),
The problem of vibration of a circular membrane of radius R rigidly fixed in its contour
is characterized by the homogeneous boundary condition, g(ϕ, t) ≡ 0.
⊙ Literature: N. S. Koshlyakov, E. B. Gliner, and M. M. Smirnov (1970), A. G. Butkovskiy (1979), B. M. Bu-
dak, A. A. Samarskii, and A. N. Tikhonov (1980).

Solution:
Z 2πZ R
∂
∂t 0 0
Z 2πZ R
0 0
Z tZ 2π
+ a2 R g(η, τ )G(r, ϕ, R, η, t − τ ) dη dτ.
0 0
∂2w
= a 2 ∆2 w 641
Here,
∞ ∞
t 1 X X An µnm Jn (µnm r)Jn (µnm ξ)
G(r, ϕ, ξ, η, t) = + cos[n(ϕ − η)] sin(µnm at),
πR2 πa n=0 m=1 (µ2nm R2 − n2 )[Jn (µnm R)]2
A0 = 1, An = 2 (n = 1, 2, . . .),

∞ ∞
1 XX AnµnmJn(µnmr)Jn(µnmξ)
G(r, ϕ, ξ, η, t)= cos[n(ϕ−η)] sin(µnmat),
πa n=0 m=1 (µnmR2+k2R2−n2)[Jn(µnmR)]2
2
A0 =1, An =2 (n=1, 2, . . .).

equation

Solution:
Z 2πZ R2
∂
∂t 0 R1
Z 2πZ R2
0 R1
Z tZ 2π
2 ∂
+ a R1 g1 (η, τ ) G(r, ϕ, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R1
Z t Z 2π
2 ∂
− a R2 g2 (η, τ ) G(r, ϕ, ξ, η, t − τ ) dη dτ.
0 0 ∂ξ ξ=R2
Here,
∞ ∞
π XX
G(r, ϕ, ξ, η, t) = An Bnm Zn (µnm r)Zn (µnm ξ) cos[n(ϕ − η)] sin(µnm at),
2a n=0 m=1
(
1/2 for n = 0, µnm Jn2 (µnm R2 )
An = Bnm = 2 ,
where the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the


Solution:
Z 2πZ R2
∂
∂t 0 R1
Z 2πZ R2
0 R1
Z tZ 2π
− a2 R1 g1 (η, τ )G(r, ϕ, R1 , η, t − τ ) dη dτ
0 0
Z tZ 2π
2
+ a R2 g2 (η, τ )G(r, ϕ, R2 , η, t − τ ) dη dτ.
0 0
Here,
t
G(r, ϕ, ξ, η, t) =
π(R22 − R12)
∞ ∞
1 X X AnµnmZn(µnmr)Zn(µnmξ) cos[n(ϕ − η)] sin(µnmat)
+ ,
πa n=0 m=1 (µ2nmR22 − n2)Zn2 (µnmR2) − (µ2nmR12 − n2)Zn2 (µnmR1 )
Zn(µnmr) = Jn′ (µnmR1)Yn(µnmr) − Yn′ (µnmR1)Jn (µnmr),
where A0 = 1 and An = 2 for n = 1, 2, . . . ; the Jn (r) and Yn (r) are Bessel functions; and

∂2w
= a 2 ∆2 w 643

∞ ∞
1 X X An µnm
G(r, ϕ, ξ, η, t) = Zn (µnm r)Zn (µnm ξ) cos[n(ϕ − η)] sin(µnm at),
πa Bnm
n=0 m=1
Bnm = (k22 R22 + µnm R22 − n2 )Zn2 (µnm R2 ) − (k12 R12 + µ2nm R12 − n2 )Zn2 (µnm R1 ),
2

Zn (µnm r) = µnm Jn′ (µnm R1 ) − k1 Jn (µnm R1 ) Yn (µnm r)

− µnm Yn′ (µnm R1 ) − k1 Yn (µnm R1 ) Jn (µnm r).
Here, A0 = 1 and An = 2 for n = 1, 2, . . . ; the Jn (r) and Yn (r) are Bessel functions; and
′


Solution:
Z ϕ0Z R
∂
∂t 0 0
Z ϕ0Z R
0 0
Z t Z ϕ0
2 ∂
−a R g1 (η, τ ) G(r, ϕ, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R
Z tZ R
2 1 ∂
0 0 ξ ∂η η=0
Z tZ R
2 1 ∂
0 0 ξ ∂η η=ϕ0
Here,
X∞ X ∞
4 Jnπ/ϕ0 (µnm r)Jnπ/ϕ0 (µnm ξ)
G(r, ϕ, ξ, η, t) = 2 ′
aR ϕ0 n=1 m=1 µnm [Jnπ/ϕ 0
(µnm R)]2

nπϕ nπη
× sin sin sin(µnm at),
ϕ0 ϕ0

r −1 ∂ϕ w = g2 (r, t) at ϕ = 0 (boundary condition),
r −1 ∂ϕ w = g3 (r, t) at ϕ = ϕ0 (boundary condition).
Solution:
Z ϕ0Z R
∂
∂t 0 0
Z ϕ0Z R
0 0
Z tZ ϕ0
+ a2 R g1 (η, τ )G(r, ϕ, R, η, t − τ ) dη dτ
0 0
Z tZ R
2
−a g2 (ξ, τ )G(r, ϕ, ξ, 0, t − τ ) dξ dτ
0 0
Z tZ R
2
+a g3 (ξ, τ )G(r, ϕ, ξ, ϕ0 , t − τ ) dξ dτ.
0 0
Here,
∞ ∞
2t 4ϕ0 X X µnm Jnπ/ϕ0 (µnm r)Jnπ/ϕ0 (µnm ξ)
G(r, ϕ, ξ, η, t) = + 2
R2 ϕ0 a 2 2 2 2 2
n=0 m=1 (R ϕ0 µnm − n π ) Jnπ/ϕ0 (µnm R)

nπϕ nπη
× cos cos sin(µnm at),
ϕ0 ϕ0
′
0
∂2w
= a 2 ∆2 w 645

∂r w + kw = g(ϕ, t) at r = R (boundary condition),
Solution:
Z ϕ0Z R
∂
∂t 0 0
Z ϕ0Z R
0 0
Z tZ ϕ0
+ a2 R g(η, τ )G(r, ϕ, R, η, t − τ ) dη dτ.
0 0
Here,
∞ X
X ∞
G(r, ϕ, ξ, η, t) = Anm Jsn (µnm r)Jsn (µnm ξ) cos(sn ϕ) cos(sn η) sin(µnm at),
n=0 m=1
nπ 4µnm
sn = , Anm = 2 ,
ϕ0 aϕ0 (µ2nm R2 + k 2 R2 − s2n ) Jsn (µnm R)
where the Jsn (r) are Bessel functions and the µnm are positive roots of the transcendental
equation

In the axisymmetric case the wave equation in the cylindrical system of coordinates has the
form
2 p
∂2w 2 ∂ w 1 ∂w ∂2w
= a + + , r = x2 + y 2 .
∂t2 ∂r 2 r ∂r ∂z 2
One-dimensional problems with axial symmetry that have solutions w = w(r, t) are con-
sidered in Section 6.2.1.
In the solution of the problems considered below, the modified Green’s function
G(r, z, ξ, η, t) = 2πξG(r, z, ξ, η, t) is used for convenience.

w = f0 (r, z) at t = 0 (initial condition),
∂t w = f1 (r, z) at t = 0 (initial condition),
Solution:
Z lZ R
∂
w(r, z, t) = f0 (ξ, η)G(r, z, ξ, η, t) dξ dη
∂t 0 0
Z lZ R
+ f1 (ξ, η)G(r, z, ξ, η, t) dξ dη
0 0
Z tZ l
2 ∂
−a g1 (η, τ ) G(r, z, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R
Z tZ R
2 ∂
+a g2 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ
0 0 ∂η η=0
Z tZ R
2 ∂
−a g3 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ.
0 0 ∂η η=l
Here,
∞ ∞
4ξ X X 1 µn r µn ξ
G(r, z, ξ, η, t) = 2 J0 J0
R l n=1 m=1 J12 (µn ) R R
√
mπz mπη sin at λnm
× sin sin √ ,
l l a λnm
µ2n π 2 m2
λnm = + ,
R2 l2
0 ≤ r ≤ R).

∂2w
= a 2 ∆2 w 647
Solution:
Z lZ R Z lZ R
∂
w(r, z, t)= f0(ξ, η)G(r, z, ξ, η, t) dξ dη+ f1(ξ, η)G(r, z, ξ, η, t) dξ dη
∂t 0 0 0 0
Z tZ l
+a2 g1(η, τ )G(r, z, R, η, t−τ ) dη dτ
0 0
Z tZ R
2
−a g2(ξ, τ )G(r, z, ξ, 0, t−τ ) dξ dτ
0 0
Z tZ R
2
+a g3(ξ, τ )G(r, z, ξ, l, t−τ ) dξ dτ.
0 0
Here,
∞ ∞
2tξ 2ξ X X Anm µn r µn ξ
G(r, z, ξ, η, t) = 2 + 2 J0 J0
R l R l n=0 m=0 J02 (µn ) R R
√
× cos cos √ ,
l l a λnm

µ2n π 2 m2 0 for m = 0, n = 0,

λnm = 2 + , Anm = 1 for m = 0, n > 0,
R l2 

2 for m > 0,
where the µn are zeros of the first-order Bessel function, J1 (µ) = 0 (µ0 = 0). The numerical


∞ ∞ √
2ξ X X µn r µn ξ ϕm (z)ϕm (η) sin at λnm
G(r, z, ξ, η, t) = 2 An J0 J0 √ ,
R R R kϕm k2 a λnm
n=1 m=1
µ2n µ2n 2 k2
An = 2 2 2 2 , λnm = 2
+ βm , ϕm (z) = cos(βm z) + sin(βm z),
(k1 R + µn )J0 (µn ) R βm
2 + k2

2 k3 βm 2 k2 l k22
kϕm k = 2 β 2 + k2
+ 2 + 1+ 2 .
2βm m 3 2βm 2 βm
Here, the µn and βm are positive roots of the transcendental equations
tan(βl) k2 + k3
µJ1 (µ) − k1 RJ0 (µ) = 0, = 2 .
β β − k2 k3
scribed:
Solution:
Z lZ R Z lZ R
∂
∂t 0 0 0 0
Z tZ l
2 ∂
−a g1(η, τ ) G(r, z, ξ, η, t−τ ) dη dτ
0 0 ∂ξ ξ=R
Z tZ R
2
−a g2(ξ, τ )G(r, z, ξ, 0, t−τ ) dξ dτ
0 0
Z tZ R
+a2 g3(ξ, τ )G(r, z, ξ, l, t−τ ) dξ dτ.
0 0
Here,
∞ ∞
2ξ X X Am µn r µn ξ
G(r, z, ξ, η, t) = 2 J0 J0
R l
n=1 m=0 1
J 2 (µn ) R R
√
× cos cos √ ,
l l a λnm
(
µ2n π 2 m2 1 for m = 0,
λnm = 2 + 2
, Am =
R l 2 for m > 0,
0 ≤ r ≤ R).
∂2w
= a 2 ∆2 w 649
scribed:
Solution:
Z lZ R Z lZ R
∂
∂t 0 0 0 0
Z tZ l
2
+a g1(η, τ )G(r, z, R, η, t−τ ) dη dτ
0 0
Z tZ R
2 ∂
+a g2(ξ, τ ) G(r, z, ξ, η, t−τ ) dξ dτ
0 0 ∂η η=0
Z tZ R
2 ∂
−a g3(ξ, τ ) G(r, z, ξ, η, t−τ ) dξ dτ.
0 0 ∂η η=l
Here,
∞ ∞
G(r, z, ξ, η, t) = 2 2 (µ ) J0 J0
R l J0 n R R
n=0 m=1
√
× sin sin √ ,
l l a λnm
µ2n π 2 m2
λnm = + ,
R2 l2
prescribed:
Solution:
Z lZ R2 Z lZ R2
∂
∂t 0 R1 0 R1
Z tZ l
2 ∂
+a g1(η, τ ) G(r, z, ξ, η, t−τ ) dη dτ
0 0 ∂ξ ξ=R1
Z tZ l
2 ∂
−a g2(η, τ ) G(r, z, ξ, η, t−τ ) dη dτ
0 0 ∂ξ ξ=R2
Z tZ R2
∂
+a2 g3(ξ, τ ) G(r, z, ξ, η, t−τ ) dξ dτ
0 R1 ∂η η=0
Z tZ R2
2 ∂
−a g4(ξ, τ ) G(r, z, ξ, η, t−τ ) dξ dτ.
0 R1 ∂η η=l
Here,
∞ ∞ √
π2 ξ X X mπz mπη sin at λnm
G(r, z, ξ, η, t) = 2 An Ψn (r)Ψn (ξ) sin sin √ ,
R1 l n=1 m=1 l l a λnm

µ2 J 2 (sµn ) µn r µn r
An = 2 n 0 2 , Ψn (r) = Y0 (µn )J0 − J0 (µn )Y0 ,
J0 (µn ) − J0 (sµn ) R1 R1
R2 µ2 π 2 m2
s= , λnm = n2 + ,
R1 R1 l2
where J0 (µ) and Y0 (µ) are Bessel functions, and the µn are positive roots of the transcen-
dental equation
J0 (µ)Y0 (sµ) − J0 (sµ)Y0 (µ) = 0.

prescribed:
Solution:
Z l Z R2 Z lZ R2
∂
∂t 0 R1 0 R1
Z tZ l Z tZ l
−a2 g1(η, τ )G(r, z, R1, η, t−τ ) dη dτ +a2 g2(η, τ )G(r, z, R2, η, t−τ ) dη dτ
0 0 0 0
Z tZ R2 Z tZ R2
−a2 g3(ξ, τ )G(r, z, ξ, 0, t−τ ) dξ dτ +a2 g4(ξ, τ )G(r, z, ξ, l, t−τ ) dξ dτ.
0 R1 0 R1
∂2 w
7.2. Nonhomogeneous Wave Equation ∂t2
= a2 ∆2 w + Φ(x, y, t) 651
Here,
X∞
2tξ 4ξ 1 mπz mπη mπat
G(r, z, ξ, η, t) = + cos cos sin
(R22 − R12 )l πa(R22 − R12 ) m=1 m l l l
∞ ∞ √
π 2 ξ X X Am µ2n J12 (sµn ) mπz mπη sin at λnm
+ Ψn (r)Ψn (ξ) cos cos √ ,
2R12 l n=1 m=0 J12 (µn ) − J12 (sµn ) l l a λnm
where

µn r R2 µn r
Ψn (r) = Y1 (µn )J0 − J1 (µn )Y0 s= , ,
R1 R1 R1
(
1 for m = 0, µ2 π 2 m2
Am = λnm = n2 + ;
2 for m > 1, R1 l2
Jk (µ) and Yk (µ) are Bessel functions (k = 0, 1); and the µn are positive roots of the
J1 (µ)Y1 (sµ) − J1 (sµ)Y1 (µ) = 0.
7.2 Nonhomogeneous Wave Equation

∂ 2w
2
= a2 ∆2 w + Φ(x, y, t)
∂t
w = f (x, y) at t = 0,
∂t w = g(x, y) at t = 0.
Solution:
ZZ ZZ
1 ∂ f (ξ, η) dξ dη 1 g(ξ, η) dξ dη
w(x, y, t) = p + p
2πa ∂t a2 t2 − ρ2 2πa a2 t2 − ρ2
ρ≤at ρ≤at
Z t Z Z
1 Φ(ξ, η, τ ) dξ dη
+ p dτ, ρ2 = (ξ − x)2 + (η − y)2 .
2πa 0 a2 (t − τ )2 − ρ2
ρ≤a(t−τ )
⊙ Literature: N. S. Koshlyakov, E. B. Gliner, and M. M. Smirnov (1970).


The solution w(x, y, t) is given by the formula in Section 7.1.1 (see the first boundary
value problem for 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 ) with the additional term
Z t Z l1Z l2
Φ(ξ, η, τ )G(x, y, ξ, η, t − τ ) dη dξ dτ, (1)
0 0 0

The solution w(x, y, t) is given by the formula in Section 7.1.1 (see the second bound-
ary value problem for 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 ) with the additional term (1); the Green’s

∂2 w
= a2 ∆2 w + Φ(x, y, t) 653
The solution w(x, y, t) is given by the formula in Section 7.1.1 (see the third boundary
value problem for 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 ) with the additional term (1); the Green’s function

The solution w(x, y, t) is given by the formula in Section 7.1.1 (see Item 1◦ for the
mixed boundary value problems for 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 ) with the additional term (1).
The solution w(x, y, t) is given by the formula in Section 7.1.1 (see Item 2◦ for the
mixed boundary value problems for 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 ) with the additional term (1).

A nonhomogeneous wave equation in the polar coordinate system has the form
2 p
∂2w 2 ∂ w 1 ∂w 1 ∂2w
= a + + + Φ(r, ϕ, t), r = x2 + y 2 .
∂t2 ∂r 2 r ∂r r 2 ∂ϕ2
One-dimensional boundary value problems independent of the angular coordinate ϕ are

The solution w(r, ϕ, t) is given by the formula in Section 7.1.2 (see the first boundary
value problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π) with the additional term
Z tZ 2πZ R
Φ(ξ, η, τ )G(r, ϕ, ξ, η, t − τ )ξ dξ dη dτ, (1)
0 0 0
⊙ Literature: N. S. Koshlyakov, E. B. Gliner, and M. M. Smirnov (1970), B. M. Budak, A. A. Samarskii, and
A. N. Tikhonov (1980).


The solution w(r, ϕ, t) is given by the formula in Section 7.1.2 (see the second boundary
value problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π) with the additional term (1); the Green’s function


The solution w(r, ϕ, t) is given by the formula in Section 7.1.2 (see the third boundary
value problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π) with the additional term (1); the Green’s function


∂2 w
= a2 ∆2 w + Φ(x, y, t) 655
value problem for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π) with the additional term
Z t Z 2πZ R2
0 0 R1

value problem for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π) with the additional term (2); the Green’s

The solution w(r, ϕ, t) is given by the formula in Section 7.1.2 (see the third boundary
value problem for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π) with the additional term (2).

value problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 ) with the additional term
Z t Z ϕ0Z R
0 0 0

−1
value problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 ) with the additional term (3); the Green’s function

The solution w(r, ϕ, t) is given by the formula in Section 7.1.2 (see the mixed boundary
value problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 ) with the additional term (3); the Green’s function

In the axisymmetric case, a nonhomogeneous wave equation in the cylindrical system of
coordinates has the form
2 p
∂2w 2 ∂ w 1 ∂w ∂2w
= a + + + Φ(r, z, t), r = x2 + y 2 .
∂t2 ∂r 2 r ∂r ∂z 2

∂2 w
= a2 ∆2 w + Φ(x, y, t) 657
The solution w(r, z, t) is given by the formula in Section 7.1.3 (see the first boundary
value problem for 0 ≤ r ≤ R, 0 ≤ z ≤ l) with the additional term
Z tZ l Z R
Φ(ξ, η, τ )G(r, z, ξ, η, t − τ ) dξ dη dτ, (1)
0 0 0

The solution w(r, z, t) is given by the formula in Section 7.1.3 (see the second boundary
value problem for 0 ≤ r ≤ R, 0 ≤ z ≤ l) with the additional term (1); the Green’s function

The solution w(r, z, t) is given by the formula in Section 7.1.3 (see the third boundary
value problem for 0 ≤ r ≤ R, 0 ≤ z ≤ l) with the additional term (1).

scribed:
The solution w(r, z, t) is given by the formula in Section 7.1.3 (see Item 1◦ for the
mixed boundary value problems for 0 ≤ r ≤ R, 0 ≤ z ≤ l) with the additional term (1).
scribed:
The solution w(r, z, t) is given by the formula in Section 7.1.3 (see Item 2◦ for the
mixed boundary value problems for 0 ≤ r ≤ R, 0 ≤ z ≤ l) with the additional term (1);

prescribed:
The solution w(r, z, t) is given by the formula in Section 7.1.3 (see the first boundary
value problem for R1 ≤ r ≤ R2 , 0 ≤ z ≤ l) with the additional term
Z t Z l Z R2
Φ(ξ, η, τ )G(r, z, ξ, η, t − τ ) dξ dη dτ, (2)
0 0 R1

prescribed:
The solution w(r, z, t) is given by the formula in Section 7.1.3 (see the second boundary
value problem for R1 ≤ r ≤ R2 , 0 ≤ z ≤ l) with the additional term (2); the Green’s function
∂2w
7.3. Equations of the Form ∂t2
= a2 ∆2 w − bw + Φ(x, y, t) 659
∂ 2w
7.3 Equations of the Form = a2∆2 w − bw + Φ(x, y, t)
∂t2
The two-dimensional nonhomogeneous Klein–Gordon equation with two space variables
in the rectangular Cartesian coordinate system is written as
2
∂2w 2 ∂ w ∂2w
=a + − bw + Φ(x, y, t).
∂t2 ∂x2 ∂y 2
◮ Fundamental solutions.
1◦ . Case b = −σ 2 < 0:
p
ϑ(at − r) cosh σ t2 − r 2/a2 p
E (x, y, t) = p , r= x2 + y 2 ,
2πa2 t2 − r 2/a2
where ϑ(z) is the Heaviside unit step function.

2◦ . Case b = σ 2 > 0:
p
ϑ(at − r) cos σ t2 − r 2/a2 p
E (x, y, t) = p , r= x2 + y 2 .
2πa2 t2 − r 2/a2
⊙ Literature: V. S. Vladimirov, V. P. Mikhailov, A. A. Vasharin et al. (1974), B. M. Budak, A. A. Samarskii,

and A. N. Tikhonov (1980).

w = f (x, y) at t = 0,
∂t w = g(x, y) at t = 0.
1◦ . Solution for b = −a2 c2 < 0:

ZZ p
1 ∂ cosh c a2 t2 − ρ2
w(x, y, t) = f (ξ, η) p dξ dη
2πa ∂t a2 t2 − ρ2
ρ≤at
ZZ p
1 cosh c a2 t2 − ρ2
+ g(ξ, η) p dξ dη
2πa a2 t2 − ρ2
ρ≤at
Z t ZZ p
1 cosh c a2 (t − τ )2 − ρ2
+ dτ Φ(ξ, η, τ ) p dξ dη,
2πa 0 a2 (t − τ )2 − ρ2
ρ≤a(t−τ)
p
where ρ = (x − ξ)2 + (y − η)2 .
2◦ . Solution for b = a2 c2 > 0:

ZZ p
1 ∂ cos c a2 t2 − ρ2
w(x, y, t) = f (ξ, η) p dξ dη
2πa ∂t a2 t2 − ρ2
ρ≤at
ZZ p
1 cos c a2 t2 − ρ2
+ g(ξ, η) p dξ dη
2πa a2 t2 − ρ2
ρ≤at
Z t ZZ p
1 cos c a2 (t − τ )2 − ρ2
+ dτ Φ(ξ, η, τ ) p dξ dη,
2πa 0 a2 (t − τ )2 − ρ2
ρ≤a(t−τ)
p
where ρ = (x − ξ)2 + (y − η)2 .


Solution:
Z l1Z l2
∂
∂t 0 0
Z l1Z l2
+ f1 (ξ, η)G(x, y, ξ, η, t) dη dξ
0 0
Z tZ l2
2 ∂
0 0 ∂ξ ξ=0
Z t Z l2
2 ∂
0 0 ∂ξ ξ=l1
Z t Z l1
2 ∂
0 0 ∂η η=0
Z t Z l1
∂
− a2 g4 (ξ, τ ) G(x, y, ξ, η, t − τ ) dξ dτ
0 0 ∂η η=l2
Z t Z l1Z l2
+ Φ(ξ, η, τ )G(x, y, ξ, η, t − τ ) dη dξ dτ,
0 0 0
∂2w
= a2 ∆2 w − bw + Φ(x, y, t) 661
where
∞ ∞
4 XX 1
G(x, y, ξ, η, t) = sin(pn x) sin(qm y) sin(pn ξ) sin(qm η) sin(λnm t),
l1 l2 n=1 m=1 λnm
nπ mπ p
pn = , qm = , λnm = a2 p2n + a2 qm 2 + b.
l1 l2

Solution:
Z Z Z l1Z l2
∂ l1 l2
w(x, y, t)= f0(ξ, η)G(x, y, ξ, η, t) dη dξ+ f1(ξ, η)G(x, y, ξ, η, t) dη dξ
∂t 0 0 0 0
Z tZ l2
−a2 g1(η, τ )G(x, y, 0, η, t−τ ) dη dτ
0 0
Z tZ l2
2
+a g2(η, τ )G(x, y, l1 , η, t−τ ) dη dτ
0 0
Z tZ l1
2
−a g3(ξ, τ )G(x, y, ξ, 0, t−τ ) dξ dτ
0 0
Z tZ l1
2
+a g4(ξ, τ )G(x, y, ξ, l2 , t−τ ) dξ dτ
0 0
Z tZ l1Z l2
+ Φ(ξ, η, τ )G(x, y, ξ, η, t−τ ) dη dξ dτ,
0 0 0
where
√ ∞ ∞
sin t b 2 X X Anm
G(x, y, ξ, η, t)= √ + cos(pnx) cos(qmy) cos(pnξ) cos(qmη) sin(λnmt),
l1l2 b l1l2 n=0 m=0 λnm


0 for n=m=0,
nπ mπ p
pn = , qm = , λnm = a2p2n+a2qm 2 +b, Anm = 1 for nm=0 (n6= m),
l1 l2 

2 for nm6= 0.

∞ X
X ∞
1
G(x, y, ξ, η, t) = 4 p sin(µnx + εn) sin(νmy + σm)
n=1 m=1 Enm a2 µ2n + a2νm
2 +b
p
× sin(µnξ + εn) sin(νmη + σm) sin t a2µ2n + a2νm 2 +b ,

(k1 k2 + µ2n)(k1 + k2 ) (k3 k4 + νm2 )(k + k )
3 4
Enm = l1 + l2 + 2 ,
(k12 + µ2n)(k22 + µ2n) (k3 + νm2 )(k 2 + ν 2 )
4 m
µn νm
εn = arctan , σm = arctan .
l1 l2
Here, the µn and νm are positive roots of the transcendental equations
µ2 − k1 k2 = (k1 + k2 )µ cot(l1 µ),

ν 2 − k3 k4 = (k3 + k4 )ν cot(l2 ν).

∂2w
= a2 ∆2 w − bw + Φ(x, y, t) 663
Solution:
Z l1Z l2
∂
∂t 0 0
Z l1Z l2
+ f1 (ξ, η)G(x, y, ξ, η, t) dη dξ
0 0
Z t Z l2
2 ∂
0 0 ∂ξ ξ=0
Z t Z l2
2 ∂
0 0 ∂ξ ξ=l1
Z t Z l1
2
−a g3 (ξ, τ )G(x, y, ξ, 0, t − τ ) dξ dτ
0 0
Z t Z l1
2
+a g4 (ξ, τ )G(x, y, ξ, l2 , t − τ ) dξ dτ
0 0
Z t Z l1Z l2
0 0 0
where
∞ ∞
2 X X Am
G(x, y, ξ, η, t) = sin(pn x) cos(qm y) sin(pn ξ) cos(qm η) sin(λnm t),
l1 l2 λ
n=1 m=0 nm
(
pn = , qm = , λnm = a2 p2n + a2 qm2 + b, Am =
l1 l2 2 for m 6= 0.

Solution:
Z Z
∂ l1 l2
∂t 0 0
Z l1Z l2
+ f1 (ξ, η)G(x, y, ξ, η, t) dη dξ
0 0
Z t Z l2
2 ∂
0 0 ∂ξ ξ=0
Z t Z l2
0 0
Z tZ l1
2 ∂
0 0 ∂η η=0
Z tZ l1
+ a2 g4 (ξ, τ )G(x, y, ξ, l2 , t − τ ) dξ dτ
0 0
Z t Z l1Z l2
0 0 0
where
∞ ∞
4 XX 1
G(x, y, ξ, η, t) = sin(pn x) sin(qm y) sin(pn ξ) sin(qm η) sin(λnm t),
l1 l2 λnm
n=0 m=0
π(2n + 1) π(2m + 1) p
pn = , qm = , λnm = a2 p2n + a2 qm 2 + b.
2l1 2l2

A nonhomogeneous Klein–Gordon equation with two space variables in the polar coordi-
nate system has the form
2 p
∂2w 2 ∂ w 1 ∂w 1 ∂2w
=a + + 2 − bw + Φ(r, ϕ, t), r= x2 + y 2 .
∂t2 ∂r 2 r ∂r r ∂ϕ2
One-dimensional solutions w = w(r, t) independent of the angular coordinate ϕ are


Solution:
Z 2πZ R
∂
∂t 0 0
Z 2πZ R
0 0
Z t Z 2π
2 ∂
−a R g(η, τ ) G(r, ϕ, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R
Z t Z 2πZ R
+ Φ(ξ, η, τ )G(r, ϕ, ξ, η, t − τ )ξ dξ dη dτ.
0 0 0
∂2w
= a2 ∆2 w − bw + Φ(x, y, t) 665
Here,∗
∞ ∞
1 XX An sin λnmt
G(r, ϕ, ξ, η, t)= Jn(µnmr)Jn(µnmξ) cos[n(ϕ−η)] ,
πR2 [Jn′ (µnmR)]2 λnm
n=0 m=1
p
A0 =1, An =2 (n=1, 2, . . .), λnm = a2µ2nm+b,

Solution:
Z 2πZ R
∂
∂t 0 0
Z 2πZ R
0 0
Z t Z 2π
2
+a R g(η, τ )G(r, ϕ, R, η, t − τ ) dη dτ
0 0
Z t Z 2πZ R
0 0 0
Here,
√
sin t b
G(r, ϕ, ξ, η, t) = √
πR2 b
∞ ∞ p
1 X X An µ2nm Jn (µnm r)Jn (µnm ξ) sin t a2 µ2nm + b
+ cos[n(ϕ − η)] p ,
π (µ2nm R2 − n2 )[Jn (µnm R)]2 a2 µ2nm + b
n=0 m=1
where A0 = 1 and An = 2 for n = 1, 2, . . . ; the Jn (ξ) are Bessel functions; and the µm
are positive roots of the transcendental equation Jn′ (µR) = 0.

∗
p p
The ratios sin t a2 µ2nm + b a2 µ2nm + b in the expressions of the Green’s functions specified in
p p
Section 7.3.2 must be replaced with sinh t |a2 µ2nm + b| |a2 µ2nm + b| if a2 µ2nm + b < 0.
∞ ∞ p
1 X X Anµ2nm sin t a2µ2nm+b
G(r, ϕ, ξ, η, t)= Jn(µnmr)Jn(µnmξ) cos[n(ϕ−η)] p ,
π Bnm a2µ2nm+b
n=0 m=1
A0 =1, An =2 (n=1, 2, . . .), Bnm =(µ2nmR2+k2R2−n2)[Jn(µnmR)]2.
Here, the Jn (ξ) are Bessel functions and the µm are positive roots of the transcendental
equation

Solution:
Z 2πZ R2 Z 2πZ R2
∂
w(r, ϕ, t) = f0(ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη+ f1(ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
∂t 0 R1 0 R1
Z tZ 2π
∂
+a2R1 g1(η, τ ) G(r, ϕ, ξ, η, t−τ ) dη dτ
0 0 ∂ξ ξ=R1
Z tZ 2π
∂
−a2R2 g2(η, τ ) G(r, ϕ, ξ, η, t−τ ) dη dτ
0 0 ∂ξ ξ=R2
Z tZ 2πZ R2
+ Φ(ξ, η, τ )G(r, ϕ, ξ, η, t−τ )ξ dξ dη dτ.
0 0 R1
Here,
∞ ∞ p
πXX sin t a2µ2nm+b
G(r, ϕ, ξ, η, t)= AnBnmZn(µnmr)Zn(µnmξ) cos[n(ϕ−η)] p ,
2 a2µ2nm+b
n=0 m=1
(
1/2 for n=0, µ2 J 2(µnmR2)
An = Bnm = 2 nm n 2 ,
Zn(µnmr)=Jn(µnmR1)Yn(µnmr)−Yn(µnmR1)Jn(µnmr),

∂2w
= a2 ∆2 w − bw + Φ(x, y, t) 667

Solution:
Z 2πZ R2
∂
∂t 0 R1
Z 2πZ R2
0 R1
Z tZ 2π
2
− a R1 g1 (η, τ )G(r, ϕ, R1 , η, t − τ ) dη dτ
0 0
Z tZ 2π
+ a2 R2 g2 (η, τ )G(r, ϕ, R2 , η, t − τ ) dη dτ
0 0
Z t Z 2πZ R2
0 0 R1
Here,
√
sin t b
G(r, ϕ, ξ, η, t)= √
π(R22 −R12) b
∞ ∞ p
1 X X Anµ2nmZn(µnmr)Zn(µnmξ) cos[n(ϕ−η)] sin t a2µ2nm+b
+ p ,
π 2 2 2 2 2 2 2 2 a2µ2nm+b
n=0 m=1 (µnmR2−n )Zn(µnmR2)−(µnmR1−n )Zn(µnmR1)
where
(
1 for n = 0,
Zn (µnm r) = Jn′ (µnm R1 )Yn (µnm r) − Yn′ (µnm R1 )Jn (µnm r), An =
2 for n > 0,
the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the transcen-
dental equation

∞ ∞
1 X X An µ2nm
G(r, ϕ, ξ, η, t) = Znm (r)Znm (ξ) cos[n(ϕ − η)] sin(λnm t),
π n=0 m=1 Bnm λnm
Bnm = (k22 R22 + µ2nm R22 − n2 )Znm

2
(R2 ) − (k12 R12 + µ2nm R12 − n2 )Znm
2
(R1 ),

Znm (r) = µnm Jn′ (µnm R1 ) − k1 Jn (µnm R1 ) Yn (µnm r)

− µnm Yn′ (µnm R1 ) − k1 Yn (µnm R1 ) Jn (µnm r).
p
Here, A0 = 1 and An = 2 for n = 1, 2, . . . ; λnm = a2 µ2nm + b; the Jn (r) and Yn (r)
are Bessel functions; and the µnm are positive roots of the transcendental equation
′


Solution:
Z ϕ0Z R Z ϕ0Z R
∂
w(r, ϕ, t) = f0 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη + f1 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
∂t 0 0 0 0
Z t Z ϕ0
2 ∂
0 0 ∂ξ ξ=R
Z tZ R
1 ∂
+ a2 g2 (ξ, τ ) G(r, ϕ, ξ, η, t − τ ) dξ dτ
0 0 ξ ∂η η=0
Z tZ R
1 ∂
− a2 g3 (ξ, τ ) G(r, ϕ, ξ, η, t − τ ) dξ dτ
0 0 ξ ∂η η=ϕ0
Z t Z ϕ0Z R
0 0 0
Here,
∞ ∞
4 X X Jnπ/ϕ0(µnmr)Jnπ/ϕ0(µnmξ) nπϕ nπη sin λnmt
G(r, ϕ, ξ, η, t)= 2 ′ sin sin ,
R ϕ0 n=1 m=1 [Jnπ/ϕ 0
(µnmR)]2 ϕ0 ϕ0 λnm
p
dental equation Jnπ/ϕ0 (µR) = 0, and λnm = a2 µ2nm + b.
∂2w
= a2 ∆2 w − bw + Φ(x, y, t) 669

−1
Solution:
Z ϕ0Z R Z ϕ0Z R
∂
w(r, ϕ, t) = f0 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη + f1 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
∂t 0 0 0 0
Z tZ ϕ0
2
+a R g1 (η, τ )G(r, ϕ, R, η, t − τ ) dη dτ
0 0
Z tZ R
− a2 g2 (ξ, τ )G(r, ϕ, ξ, 0, t − τ ) dξ dτ
0 0
Z tZ R
+ a2 g3 (ξ, τ )G(r, ϕ, ξ, ϕ0 , t − τ ) dξ dτ
0 0
Z tZ ϕ0Z R
0 0 0
Here,
√ ∞ X ∞
2 sin t b X µ2nm Jnπ/ϕ0 (µnm r)Jnπ/ϕ0 (µnm ξ)
G(r, ϕ, ξ, η, t) = √ + 4ϕ0 2
R2 ϕ0 b 2 2 2 2 2
n=0 m=1 (R ϕ0 µnm − n π ) Jnπ/ϕ0 (µnm R)
p
nπϕ nπη sin t a2 µ2nm + b
× cos cos p ,
ϕ0 ϕ0 a2 µ2nm + b
′
0

Solution:
Z ϕ0Z R
∂
∂t 0 0
Z ϕ0Z R
0 0
Z tZ ϕ0
+ a2 R g(η, τ )G(r, ϕ, R, η, t − τ ) dη dτ
0 0
Z tZ ϕ0Z R
0 0 0
Here,
∞ X
X ∞
µ2
G(r, ϕ, ξ, η, t) = 4 p nm Jsn (µnm r)Jsn (µnm ξ)
2 2
n=0 m=1 Bnm a µnm + b
p
× cos(sn ϕ) cos(sn η) sin t a2 µ2nm + b ,
nπ 2
sn = , Bnm = ϕ0 (µ2nm R2 + k2 R2 − s2n ) Jsn (µnm R) ,
ϕ0
equation

In the axisymmetric case, a nonhomogeneous Klein–Gordon equation in the cylindrical
system of coordinates has the form
2 p
∂2w 2 ∂ w 1 ∂w ∂2w
= a + + − bw + Φ(r, z, t), r= x2 + y 2 .
∂t2 ∂r 2 r ∂r ∂z 2
In the solutions of the problems considered below, the modified Green’s function

∂2w
= a2 ∆2 w − bw + Φ(x, y, t) 671
Solution:
Z lZ R
∂
∂t 0 0
Z lZ R
+ f1 (ξ, η)G(r, z, ξ, η, t) dξ dη
0 0
Z tZ l
2 ∂
0 0 ∂ξ ξ=R
Z tZ R
2 ∂
0 0 ∂η η=0
Z tZ R
∂
− a2 g3 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ
0 0 ∂η η=l
Z tZ lZ R
+ Φ(ξ, η, τ )G(r, z, ξ, η, t − τ ) dξ dη dτ.
0 0 0
Here,
∞ ∞ √
4ξ X X 1 µnr µnξ mπz mπη sin t λnm
G(r, z, ξ, η, t)= 2 J0 J0 sin sin √ ,
R l n=1 m=1 J12(µn) R R l l λnm
a2µ2n a2π 2m2
λnm = + +b,
R2 l2
0 ≤ r ≤ R).


Solution:
Z lZ R
∂
∂t 0 0
Z lZ R
+ f1 (ξ, η)G(r, z, ξ, η, t) dξ dη
0 0
Z tZ l
+ a2 g1 (η, τ )G(r, z, R, η, t − τ ) dη dτ
0 0
Z tZ R
2
−a g2 (ξ, τ )G(r, z, ξ, 0, t − τ ) dξ dτ
0 0
Z tZ R
2
+a g3 (ξ, τ )G(r, z, ξ, l, t − τ ) dξ( dτ
0 0
Z tZ lZ R
0 0 0
Here,
√
2ξ sin t b
G(r, z, ξ, η, t)= √
R2l b
∞ ∞ √
2ξ X X Anm µnr µnξ mπz mπη sin t λnm
+ 2 J0 J0 cos cos √ ,
R l n=0 m=0 J02(µn) R R l l λnm

2 2 2 2 2

0 for m=0, n=0,
a µn a π m
λnm = 2 + +b, Anm = 1 for m=0, n>0,
R l2 

2 for m>0,

∞ ∞ √
2ξ X X µ2n µn r µn ξ ϕm (z)ϕm (η) sin t λnm
G(r, z, ξ, η, t) = 2 J0 J0 √ ,
R Bn R R kϕm k2 λnm
n=1 m=1
a2 µ2n k2
Bn = (k12 R2 +µ2n )J02 (µn ), λnm = 2
+a2 βm
2
+b, ϕm (z) = cos(βm z)+ sin(βm z),
R βm
2 + k2
k3 βm k2 l k22
kϕm k2 = 2 2 + k2
2
+ 2 + 1+ 2 .
2βm βm 3 2βm 2 βm
Here, the µn and βm are positive roots of the transcendental equations
tan(βl) k2 + k3
µJ1 (µ) − k1 RJ0 (µ) = 0, = 2 .
β β − k2 k3
∂2w
= a2 ∆2 w − bw + Φ(x, y, t) 673
scribed:
Solution:
Z lZ R
∂
∂t 0 0
Z lZ R
+ f1 (ξ, η)G(r, z, ξ, η, t) dξ dη
0 0
Z tZ l
2 ∂
0 0 ∂ξ ξ=R
Z tZ R
− a2 g2 (ξ, τ )G(r, z, ξ, 0, t − τ ) dξ dτ
0 0
Z tZ R
2
+a g3 (ξ, τ )G(r, z, ξ, l, t − τ ) dξ dτ
0 0
Z tZ lZ R
0 0 0
Here,
∞ ∞
2ξ X X Am µn r µn ξ
G(r, z, ξ, η, t) = 2 J0 J0
R l
n=1 m=0 1
J 2 (µn ) R R
√
mπz mπη sin t λnm
× cos cos √ ,
l l λnm
(
a2 µ2n a2 π 2 m2 1 for m = 0,
λnm = 2
+ 2
+ b, Am =
R l 2 for m > 0,
0 ≤ r ≤ R).
scribed:
Solution:
Z Z Z lZ R
∂ l R
w(r, z, t) = f0 (ξ, η)G(r, z, ξ, η, t) dξ dη + f1 (ξ, η)G(r, z, ξ, η, t) dξ dη
∂t 0 0 0 0
Z tZ l
2
+a g1 (η, τ )G(r, z, R, η, t − τ ) dη dτ
0 0
Z tZ R
2 ∂
0 0 ∂η η=0
Z tZ R
2 ∂
−a g3 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ
0 0 ∂η η=l
Z tZ l Z R
0 0 0
Here,
∞ ∞
G(r, z, ξ, η, t) = 2 J0 J0
R l n=0 m=1 J02 (µn ) R R
√
× sin sin √ ,
l l λnm
a2 µ2n a2 π 2 m2
λnm = 2
+ + b,
R l2

prescribed:
∂2w
= a2 ∆2 w − bw + Φ(x, y, t) 675
Solution:
Z l Z R2
∂
∂t 0 R1
Z l Z R2
+ f1 (ξ, η)G(r, z, ξ, η, t) dξ dη
0 R1
Z tZ l
2 ∂
+a g1 (η, τ ) G(r, z, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R1
Z tZ l
2 ∂
0 0 ∂ξ ξ=R2
Z t Z R2
2 ∂
0 R1 ∂η η=0
Z t Z R2
2 ∂
0 R1 ∂η η=l
Z t Z l Z R2
0 0 R1
Here,
∞ ∞
π2 ξ X X µ2n J02 (sµn )
G(r, z, ξ, η, t) = Ψn (r)Ψn (ξ)
R1 l n=1 m=1 J0 (µn ) − J02 (sµn )
2 2
√
× sin sin √ ,
l l λnm

µn r µn r R2 a2 µ2n a2 π 2 m2
Ψn (r) = Y0 (µn )J0 −J0 (µn )Y0 , s= , λnm = + +b,
R1 R1 R1 R12 l2
dental equation
J0 (µ)Y0 (sµ) − J0 (sµ)Y0 (µ) = 0.
prescribed:
Solution:
Z l Z R2
∂
∂t 0 R1
Z l Z R2
+ f1 (ξ, η)G(r, z, ξ, η, t) dξ dη
0 R1
Z tZ l
− a2 g1 (η, τ )G(r, z, R1 , η, t − τ ) dη dτ
0 0
Z tZ l
+ a2 g2 (η, τ )G(r, z, R2 , η, t − τ ) dη dτ
0 0
Z tZ R2
2
−a g3 (ξ, τ )G(r, z, ξ, 0, t − τ ) dξ dτ
0 R1
Z t Z R2
2
0 R1
Z tZ l Z R2
0 0 R1
Here,
√ ∞ √
2ξ sin t b 4ξ X mπz mπη sin t βm
G(r, z, ξ, η, t)= 2 √ + cos cos √
(R2 −R12)l b (R22−R12)l m=1 l l βm
∞ ∞ √
π 2ξ X X Amµ2nJ12(sµn) mπz mπη sin t λnm
+ 2 Ψn(r)Ψn(ξ) cos cos √ ,
2R1l n=1 m=0 J12(µn)−J12(sµn) l l λnm
where

µn r µn r R2
Ψn (r) = Y1 (µn )J0 − J1 (µn )Y0 , s= ,
R1 R1 R1
(
1 for m = 0, a2 π 2 m2 a2 µ2n a2 π 2 m2
Am = βm = + b, λnm = + + b;
2 for m > 1, l2 R12 l2
Jk (µ) and Yk (µ) are Bessel functions (k = 0, 1); and the µn are positive roots of the
J1 (µ)Y1 (sµ) − J1 (sµ)Y1 (µ) = 0.
7.4 Telegraph Equation

∂ 2w ∂w
2
+k = a2∆2 w − bw + Φ(x, y, t)
∂t ∂t
A two-dimensional nonhomogeneous telegraph equation in the rectangular Cartesian coor-
dinate system is written as
2
∂2w ∂w 2 ∂ w ∂2w
+k =a + − bw + Φ(x, y, t).
∂t2 ∂t ∂x2 ∂y 2
∂2 w
7.4. Telegraph Equation ∂t2
+ k ∂w
∂t
= a2 ∆2 w − bw + Φ(x, y, t) 677
◮ Reduction to the two-dimensional Klein–Gordon equation.

The substitution w(x, y, t) = exp − 12 kt u(x, y, t) leads to the equation
2
∂2u 2 ∂ u ∂2u
2
=a 2
+ 2
− b − 14 k2 u + exp 12 kt Φ(x, y, t),
∂t ∂x ∂y
1◦ . Case b − 14 k2 = σ 2 > 0:
p
cos σ t2 − r 2/a2
E (x, y, t) = ϑ(at − r) exp − 12 kt p ,
2πa2 t2 − r 2/a2
p
where r = x2 + y 2 and ϑ(z) is the Heaviside unit step function.
2◦ . Case b − 14 k2 = −σ 2 < 0:
p
cosh σ t 2 − r 2/a2
E (x, y, t) = ϑ(at − r) exp − 12 kt p .
2πa2 t2 − r 2/a2

w = f (x, y) at t = 0,
∂t w = g(x, y) at t = 0.
Solution:
ZZ
∂
w(x, y, t) = exp − 12 kt f (ξ, η)H(x, y, ξ, η, t) dξ dη
∂t
ρ≤at
ZZ

+ exp − 12 kt g(ξ, η) + 12 kf (ξ, η) H(x, y, ξ, η, t) dξ dη
ρ≤at
Z t ZZ

+ dτ exp − 12 k(t − τ ) Φ(ξ, η, τ )H(x, y, ξ, η, t − τ ) dξ dη.
0
ρ≤a(t−τ)
Here,
 p

 cos σ t2 − ρ2/a2

 p for b − 14 k2 = σ 2 > 0,
2πa2 pt2 − ρ2/a2
H(x, y, ξ, η, t) =
 cosh σ t2 − ρ2/a2


 p for b − 14 k2 = −σ 2 < 0,
2πa2 t2 − ρ2/a2
p
where ρ = (x − ξ)2 + (y − η)2 .

Solution:
Z l1Z l2
∂
∂t 0 0
Z l1Z l2

+ f1 (ξ, η) + kf0 (ξ, η) G(x, y, ξ, η, t) dη dξ
0 0
Z t Z l2
2 ∂
0 0 ∂ξ ξ=0
Z t Z l2
2 ∂
0 0 ∂ξ ξ=l1
Z t Z l1
2 ∂
0 0 ∂η η=0
Z t Z l1
2 ∂
−a g4 (ξ, τ ) G(x, y, ξ, η, t − τ ) dξ dτ
0 0 ∂η η=l2
Z t Z l1Z l2
0 0 0
where
∞ X ∞
4 X 1
G(x, y, ξ, η, t)= exp − 12 kt sin(pnx) sin(qmy) sin(pnξ) sin(qmη) sin(λnmt),
l1l2 n=1 m=1
λnm
q
nπ mπ 2 +b− 1 k 2.
pn = , qm = , λnm = a2p2n+a2qm 4
l1 l2

∂2 w
+ k ∂w
∂t
= a2 ∆2 w − bw + Φ(x, y, t) 679
Solution:
Z Z
∂ l1 l2
∂t 0 0
Z l1Z l2

0 0
Z tZ l2
− a2 g1 (η, τ )G(x, y, 0, η, t − τ ) dη dτ
0 0
Z t Z l2
0 0
Z tZ l1
− a2 g3 (ξ, τ )G(x, y, ξ, 0, t − τ ) dξ dτ
0 0
Z tZ l1
2
0 0
Z t Z l1Z l2
+ Φ(ξ, η, τ )G(x, y, ξ, η, t − τ ) dη dξ dτ.
0 0 0
Here,

sin λ00 t
G(x, y, ξ, η, t) = exp − 12 kt
l1 l2λ00
∞ ∞
2 X X Anm
+ cos(pnx) cos(qmy) cos(pnξ) cos(qmη) sin(λnmt) ,
l1l2 n=0 m=0 λnm
where

q 
0 for n=m=0,
nπ mπ 2 2 2 2 1 2
pn = , qm = , λnm = a pn+a qm+b− 4 k , Anm = 1 for nm=0 (n6= m),
l1 l2 

2 for nm6= 0.

∂x w − s1 w = g1 (y, t) at x = 0 (boundary condition),
∂x w + s2 w = g2 (y, t) at x = l1 (boundary condition),
∂y w − s3 w = g3 (x, t) at y = 0 (boundary condition),
∂y w + s4 w = g4 (x, t) at y = l2 (boundary condition).
∞ X
∞
X 1
G(x, y, ξ, η, t) = 4 exp − 12 kt sin(µn x + εn ) sin(νm y + σm )
Enm λmn
n=1m=1

× sin(µn ξ + εn ) sin(νm η + σm ) sin λmn t .
Here,
q
2 + b − 1 k2 ,
µn νm
λmn = a2 µ2n + a2 νm 4 εn = arctan , σm = arctan ,
l1 l2

(s1 s2 + µ2n )(s1 + s2 ) (s3 s4 + νm2 )(s + s )
3 4
Enm = l1 + l2 + ;
(s21 + µ2n )(s22 + µ2n ) (s23 + νm
2 )(s2 + ν 2 )
4 m
the µn and νm are positive roots of the transcendental equations
µ2 − s1 s2 = (s1 + s2 )µ cot(l1 µ), ν 2 − s3 s4 = (s3 + s4 )ν cot(l2 ν).

Solution:
Z l1Z l2
∂
∂t 0 0
Z l1Z l2

0 0
Z t Z l2
2 ∂
0 0 ∂ξ ξ=0
Z t Z l2
2 ∂
0 0 ∂ξ ξ=l1
Z t Z l1
2
−a g3 (ξ, τ )G(x, y, ξ, 0, t − τ ) dξ dτ
0 0
Z t Z l1
+ a2 g4 (ξ, τ )G(x, y, ξ, l2 , t − τ ) dξ dτ
0 0
Z t Z l1Z l2
0 0 0
∂2 w
+ k ∂w
∂t
= a2 ∆2 w − bw + Φ(x, y, t) 681
where
∞ X ∞
2 X Am
G(x, y, ξ, η, t) = exp − 12 kt sin(pnx) cos(qmy) sin(pnξ) cos(qm η) sin(λnm t),
l1 l2 λ
n=1 m=0 nm
(
q
nπ mπ 2 2 2 2 1 2 1 for m = 0,
pn = , qm = , λnm = a pn + a qm + b − 4 k , Am =
l1 l2 2 for m 6= 0.
w = g1 (y, t) at x=0 (boundary condition),
w = g3 (x, t) at y=0 (boundary condition),
Solution:
Z l1Z l2
∂
∂t 0 0
Z l1Z l2

0
Z t0Z l2
2 ∂
0 0 ∂ξ ξ=0
Z t Z l2
0 0
Z t Z l1
2 ∂
0 0 ∂η η=0
Z t Z l1
2
0 0
Z t Z l1Z l2
0 0 0
where
∞ X ∞
4 X 1
G(x, y, ξ, η, t) = exp − 12 kt sin(pn x) sin(qm y) sin(pn ξ) sin(qm η) sin(λnm t),
l1 l2 λ
n=0 m=0 nm
q
π(2n + 1) π(2m + 1) 2 + b − 1 k2 .
pn = , qm = , λnm = a2 p2n + a2 qm 4
2l1 2l2

A two-dimensional nonhomogeneous telegraph equation in the polar coordinate system has
the form
2 p
∂2w ∂w 2 ∂ w 1 ∂w 1 ∂2w
2
+k =a 2
+ + 2 2
− bw + Φ(r, ϕ, t), r = x2 + y 2 .
∂t ∂t ∂r r ∂r r ∂ϕ
For one-dimensional solutions w = w(r, t), see equation 6.4.2.2.


Solution:
Z 2πZ R
∂
∂t 0 0
Z 2πZ R

+ f1 (ξ, η) + kf0 (ξ, η) G(r, ϕ, ξ, η, t)ξ dξ dη
0 0
Z t Z 2π
2 ∂
−a R g(η, τ ) G(r, ϕ, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R
Z t Z 2πZ R
0 0 0
Here,
∞ X ∞ √
1 1
X An Jn (µnm r)Jn (µnm ξ) sin t λnm
G(r, ϕ, ξ, η, t) = exp − 2 kt cos[n(ϕ − η)] √ ,
πR2 n=0 m=1
[Jn′ (µnm R)]2 λnm
λnm = a2 µ2nm + b − 14 k 2 , A0 = 1, An = 2 (n = 1, 2, . . .),


Solution:
Z 2πZ R
∂
∂t 0 0
Z 2πZ R

0 0
Z t Z 2π
0 0
Z t Z 2πZ R
0 0 0
∂2 w
+ k ∂w
∂t
= a2 ∆2 w − bw + Φ(x, y, t) 683
Here,
" p
sin t b − k 2/4
G(r, ϕ, ξ, η, t) = exp − 12 kt p
πR2 b − k2/4
∞ ∞ √ #
1 X X Anµ2nmJn(µnmr)Jn(µnmξ) sin t λnm
+ cos[n(ϕ − η)] √ ,
π (µ2nmR2 − n2)[Jn(µnmR)]2 λnm
n=0 m=1
λnm = a2 µ2nm + b − 14 k2, A0 = 1, An = 2 (n = 1, 2, . . .),
where the Jn (ξ) are Bessel functions and the µm are positive roots of the transcendental

∂r w + sw = g(ϕ, t) at r = R (boundary condition).
∞ X ∞ √
1 1
X Anµ2nmJn(µnmr)Jn(µnmξ) cos[n(ϕ−η)] sin t λnm
G(r, ϕ, ξ, η, t)= exp − 2 kt √ ,
π n=0 m=1
(µ2nmR2+s2R2−n2)[Jn(µnmR)]2 λnm
λnm =a2µ2nm+b− 41 k 2, A0 =1, An =2 (n=1, 2, . . .).
Here, the Jn (ξ) are Bessel functions and the µm are positive roots of the transcendental
equation
µJn′ (µR) + sJn (µR) = 0.

Solution:
Z 2πZ R2
∂
∂t 0 R1
Z 2πZ R2

0 R1
Z tZ 2π
2 ∂
+ a R1 g1 (η, τ ) G(r, ϕ, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R1
Z t Z 2π
2 ∂
− a R2 g2 (η, τ ) G(r, ϕ, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R2
Z t Z 2πZ R2
0 0 R1
Here,
∞ X ∞ √
π 1
X sin t λnm
G(r, ϕ, ξ, η, t)= exp − 2 kt AnBnmZn(µnmr)Zn(µnmξ) cos[n(ϕ−η)] √ ,
2 n=0 m=1
λnm
(
1/2 for n=0, µ2 J 2(µnmR2)
An = Bnm = 2 nm n 2 ,
Zn(µnmr)=Jn(µnmR1)Yn(µnmr)−Yn(µnmR1)Jn(µnmr), λnm =a2µ2nm+b− 41 k 2,


Solution:
Z 2πZ R2
∂
∂t 0 R1
Z 2πZ R2

0 R1
Z tZ 2π
− a2 R1 g1 (η, τ )G(r, ϕ, R1 , η, t − τ ) dη dτ
0 0
∂2 w
+ k ∂w
∂t
= a2 ∆2 w − bw + Φ(x, y, t) 685
Z tZ 2π
2
+ a R2 g2 (η, τ )G(r, ϕ, R2 , η, t − τ ) dη dτ
0 0
Z t Z 2πZ R2
0 0 R1
Here,
" p
sin t b−k 2/4
G(r, ϕ, ξ, η, t)=exp − 21 kt p
π(R22−R12) b−k 2/4
∞ ∞ p #
1XX Anµ2nmZn(µnmr)Zn(µnmξ) cos[n(ϕ−η)] sin t a2µ2nm+b−k 2/4
+ p ,
π n=0 m=1 (µ2nmR22−n2)Zn2(µnmR2)−(µ2nmR12−n2)Zn2(µnmR1) a2µ2nm+b−k 2/4
where
(
1 for n = 0,
Zn (µnm r) = Jn′ (µnm R1 )Yn (µnm r) − Yn′ (µnm R1 )Jn (µnm r), An =
2 for n > 0,
dental equation

∂r w − s1 w = g1 (ϕ, t) at r = R1 (boundary condition),
∂r w + s2 w = g2 (ϕ, t) at r = R2 (boundary condition).
∞ X ∞
1 X An µ2nm
G(r, ϕ, ξ, η, t) = exp − 21 kt Zn (µnm r)Zn (µnm ξ) cos[n(ϕ−η)] sin(λnm t).
π n=0 m=1
Bnm λnm
Here,
( q
1 for n = 0,
An = λnm = a2 µ2nm + b − 14 k2 ,
2 for n > 0,
Bnm = (s22 R22 + µ2nm R22 − n2 )Zn2 (µnm R2 ) − (s21 R12 + µ2nm R12 − n2 )Zn2 (µnm R1 ),

Zn (µnm r) = µnm Jn′ (µnm R1 ) − s1 Jn (µnm R1 ) Yn (µnm r)

− µnm Yn′ (µnm R1 ) − s1 Yn (µnm R1 ) Jn (µnm r),
′
µJn (µR1 ) − s1 Jn (µR1 ) µYn′ (µR2 ) + s2 Yn (µR2 )

= µYn′ (µR1 ) − s1 Yn (µR1 ) µJn′ (µR2 ) + s2 Jn (µR2 ) .


Solution:
Z ϕ0Z R
∂
∂t 0 0
Z ϕ0Z R

0 0
Z t Z ϕ0
2 ∂
0 0 ∂ξ ξ=R
Z tZ R
2 1 ∂
0 0 ξ ∂η η=0
Z tZ R
1 ∂
− a2 g3 (ξ, τ ) G(r, ϕ, ξ, η, t − τ ) dξ dτ
0 0 ξ ∂η η=ϕ0
Z t Z ϕ0Z R
0 0 0
Here,
∞ X ∞
4 1
X Jnπ/ϕ0 (µnm r)Jnπ/ϕ0 (µnm ξ)
G(r, ϕ, ξ, η, t) = 2
exp − 2 kt ′
R ϕ0 [Jnπ/ϕ (µnm R)]2
n=1 m=1 0
p
nπϕ nπη sin t a2 µ2nm + b − k2/4
× sin sin p ,
ϕ0 ϕ0 a2 µ2nm + b − k2/4


−1
r −1 ∂ϕ w = g3 (r, t) at ϕ = ϕ0 (boundary condition).
∂2 w
+ k ∂w
∂t
= a2 ∆2 w − bw + Φ(x, y, t) 687
Solution:
Z ϕ0Z R
∂
∂t 0 0
Z ϕ0Z R

0 0
Z t Z ϕ0
+ a2 R g1 (η, τ )G(r, ϕ, R, η, t − τ ) dη dτ
0 0
Z tZ R
2
−a g2 (ξ, τ )G(r, ϕ, ξ, 0, t − τ ) dξ dτ
0 0
Z tZ R
+ a2 g3 (ξ, τ )G(r, ϕ, ξ, ϕ0 , t − τ ) dξ dτ
0 0
Z t Z ϕ0Z R
0 0 0
Here,
" p ∞ X∞
2 sin t b−k 2/4 X µ2nmJnπ/ϕ0(µnmr)Jnπ/ϕ0(µnmξ)
G(r, ϕ, ξ, η, t)=exp − 21 kt p +4ϕ0
R2ϕ0 b−k 2/4 n=0m=1
(R2ϕ20µ2nm−n2π 2)Jnπ/ϕ
2
0
(µnmR)
p #
nπϕ nπη sin t a2µ2nm+b−k 2/4
×cos cos p ,
ϕ0 ϕ0 a2µ2nm+b−k 2/4
′
0


∂r w + βw = g(ϕ, t) at r = R (boundary condition),
Solution:
Z ϕ0Z R
∂
∂t 0 0
Z ϕ0Z R

0 0
Z t Z ϕ0
0 0
Z t Z ϕ0Z R
0 0 0
Here,
∞ X
∞
X
G(r, ϕ, ξ, η, t) = exp − 12 kt Anm Jsn (µnm r)Jsn (µnm ξ) cos(sn ϕ) cos(sn η) sin λnm t ,
n=0 m=1
q
nπ 4µ2nm
sn = , Anm = 2 , λnm = a2 µ2nm + b − 14 k 2 ,
ϕ0 ϕ0 (µ2nm R2 + β 2 R2 − s2n ) Jsn (µnm R) λnm
equation
µJs′ n (µR) + βJsn (µR) = 0.

In the axisymmetric case, a nonhomogeneous telegraph equation in the cylindrical coordi-
nate system has the form
2 p
∂2w ∂w 2 ∂ w 1 ∂w ∂2w
+ k = a + + − bw + Φ(r, z, t), r = x2 + y 2 .
∂t2 ∂t ∂r 2 r ∂r ∂z 2
In the solutions of the problems considered below, the modified Green’s function

Solution:
Z lZ R
∂
∂t 0 0
Z lZ R

+ f1 (ξ, η) + kf0 (ξ, η) G(r, z, ξ, η, t) dξ dη
0 0
Z tZ l
2 ∂
0 0 ∂ξ ξ=R
Z tZ R
2 ∂
0 0 ∂η η=0
Z tZ R
2 ∂
0 0 ∂η η=l
Z tZ lZ R
0 0 0
∂2 w
+ k ∂w
∂t
= a2 ∆2 w − bw + Φ(x, y, t) 689
Here,
∞ ∞
4ξe−kt/2 X X 1 µn r µn ξ
G(r, z, ξ, η, t) = J0 J0
R2 l
n=1 m=1 1
J 2 (µn ) R R

mπz mπη sin(λnm t)
× sin sin ,
l l λnm
r
a2 µ2n a2 π 2 m 2 k2
λnm = + + b − ,
R2 l2 4
0 ≤ r ≤ R).


Solution:
Z lZ R
∂
∂t 0 0
Z lZ R

0 0
Z tZ l
2
+a g1 (η, τ )G(r, z, R, η, t − τ ) dη dτ
0 0
Z tZ R
2
−a g2 (ξ, τ )G(r, z, ξ, 0, t − τ ) dξ dτ
0 0
Z tZ R
2
0 0
Z tZ lZ R
0 0 0
Here,
" √ ∞ ∞
1
sin t c 1 X X Anm µn r µn ξ
G(r, z, ξ, η, t) = 2ξ exp − 2 kt √ + 2 J0 J0
R2 l c R l n=0 m=0 J02 (µn ) R R
√ #
× cos cos √ ,
l l λnm
where

k2 a2 µ2n a2 π 2 m2 k2 0 for m = 0, n = 0,

c=b− , λnm = + + b − , Anm = 1 for m = 0, n > 0,
4 R2 l2 4 

2 for m > 0,
and the µn are zeros of the first-order Bessel function, J1 (µ) = 0 (µ0 = 0). The numerical

∂r w + s1 w = g1 (z, t) at r = R (boundary condition),
∂z w − s2 w = g2 (r, t) at z = 0 (boundary condition),
∂z w + s3 w = g3 (r, t) at z = l (boundary condition).
∞ X ∞ √
2ξ 1
X µnr µnξ ϕm(z)ϕm(η) sin t λnm
G(r, z, ξ, η, t)= 2 exp − 2 kt AnJ0 J0 √ .
R R R kϕmk2 λ nm
n=1 m=1
Here,
µ2n a2 µ2n 2 2 k2
An = , λnm =
+ a βm + b − ,
(s21 R2 + µ2n )J02 (µn )
R2 4
2 + s2

s2 2 s3 βm 2 s2 l s22
ϕm (z) = cos(βm z) + sin(βm z), kϕm k = 2 β 2 + s2
+ 2 + 1+ 2 ;
βm 2βm m 3 2βm 2 βm
the µn and βm are positive roots of the transcendental equations
tan(βl) s2 + s3
µJ1 (µ) − s1 RJ0 (µ) = 0, = 2 .
β β − s2 s3

scribed:
∂2 w
+ k ∂w
∂t
= a2 ∆2 w − bw + Φ(x, y, t) 691
Solution:
Z lZ R
∂
∂t 0 0
Z lZ R

0 0
Z tZ l
2 ∂
0 0 ∂ξ ξ=R
Z tZ R
− a2 g2 (ξ, τ )G(r, z, ξ, 0, t − τ ) dξ dτ
0 0
Z tZ R
2
0 0
Z tZ lZ R
0 0 0
Here,
∞ ∞
2ξe−kt/2 X X Am µn r µn ξ
G(r, z, ξ, η, t) = J0 J0
R2 l
n=1 m=0 1
J 2 (µn ) R R

× cos cos ,
l l λnm
r (
a2 µ2n a2 π 2 m2 k2 1 for m = 0,
λnm = 2
+ 2
+b− , Am =
R l 4 2 for m > 0,
where the µn are zeros of the Bessel function, J0 (µ) = 0.
scribed:

Solution:
Z lZ R
∂
∂t 0 0
Z lZ R

0 0
Z tZ l
+ a2 g1 (η, τ )G(r, z, R, η, t − τ ) dη dτ
0 0
Z tZ R
2 ∂
0 0 ∂η η=0
Z tZ R
2 ∂
0 0 ∂η η=l
Z tZ lZ R
0 0 0
Here,
∞ ∞
4ξe−kt/2 X X 1 µn r µn ξ
G(r, z, ξ, η, t) = J0 J0
R2 l n=0 m=1 J02 (µn ) R R

× sin sin ,
l l λnm
r
a2 µ2n a2 π 2 m 2 k2
λnm = + + b − ,
R2 l2 4
prescribed:
7.5. Other Equations with Two Space Variables 693
Solution:
Z l Z R2
∂
∂t 0 R1
Z l Z R2

0 R1
Z tZ l
2 ∂
+a g1 (η, τ ) G(r, z, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R1
Z tZ l
2 ∂
0 0 ∂ξ ξ=R2
Z t Z R2
2 ∂
0 R1 ∂η η=0
Z t Z R2
2 ∂
0 R1 ∂η η=l
Z t Z l Z R2
0 0 R1
Here,
∞ ∞
π 2 ξ −kt/2 X X µ2n J02 (sµn )Ψn (r)Ψn (ξ)
G(r, z, ξ, η, t) = e
R12 l n=1 m=1
J02 (µn ) − J02 (sµn )
√
× sin sin √ ,
l l λnm

µn r µn r R2
Ψn (r) = Y0 (µn )J0 − J0 (µn )Y0 , s= ,
R1 R1 R1
a2 µ2n a2 π 2 m2 k2
λnm = 2 + 2
+b− ,
R1 l 4
dental equation
J0 (µ)Y0 (sµ) − J0 (sµ)Y0 (µ) = 0.
7.5 Other Equations with Two Space Variables

∂2w ∂w ∂2w ∂2w ∂w ∂w
1. +k = a2 + + b1 + b2 + cw.
∂t2 ∂t ∂x2 ∂y 2 ∂x ∂y
The transformation

1 b1 x + b2 y
w(x, y, t) = u(x, y, τ ) exp − kt − , τ = at
2 2a2
leads to the equation from Section 7.1.3:
∂2u ∂2u ∂2u c k2 1
= + + βu, β= + − 4 (b21 + b22 ).
∂τ 2 ∂x2 ∂y 2 a2 4a2 4a
∂2w m m−1 ∂w ∂2w ∂2w

2. tm + t = + .
∂t2 2 ∂t ∂x2 ∂y 2
Domain: −∞ < x < ∞, −∞ < y < ∞. Cauchy problem.
w = f (x, y) at t = 0,
tm/2 ∂t w = g(x, y) at t = 0.
Solution for 1 ≤ m < 2:

ZZ ZZ
1 m/2 ∂ f (ξ, η) dξ dη 1 g(ξ, η) dξ dη
w(x, y, t) = t p + p ,
2π ∂t km2 t2−m − ρ2 2π 2 t2−m − ρ2
km
Ct Ct
2 p
km = , ρ= (x − ξ)2 + (y − η)2 ,
2−m
where Ct = {ρ2 ≤ km
2 t2−m } is the circle with center at (x, y) and radius k t1/km .
m
Chapter 8
Second-Order
Hyperbolic Equations with
∂ 2w
8.1 Wave Equation = a2∆3 w
∂t2
The wave equation with three space variables in the rectangular Cartesian coordinate sys-
tem has the form 2
∂2w 2 ∂ w ∂2w ∂2w
=a + + .
∂t2 ∂x2 ∂y 2 ∂z 2
This equation is of fundamental importance in sound propagation theory, the propagation
of electromagnetic fields theory, and a number of other areas of physics and mechanics.
◮ Particular solutions and their properties.

p
w(x, y, z, t) = A exp k1x+k2y+k3z±at k12 +k22 +k32 ,
p
w(x, y, z, t) = A sin(k1x+C1) sin(k2y+C2) sin(k3z+C3) sin at k12 +k22 +k32 ,
p
w(x, y, z, t) = A sin(k1x+C1) sin(k2y+C2) sin(k3z+C3) cos at k12 +k22 +k32 ,
p
w(x, y, z, t) = A sinh(k1x+C1) sinh(k2y+C2) sinh(k3z+C3) sinh at k12 +k22 +k32 ,
p
w(x, y, z, t) = A sinh(k1x+C1) sinh(k2y+C2) sinh(k3z+C3) cosh at k12 +k22 +k32 ,
where A, C1 , C2 , C3 , k1 , k2 , and k3 are arbitrary constants.

1 p
E (x, y, z, t) = δ(a2 t2 − r 2 ), r= x2 + y 2 + z 2 ,
2πa
695
696 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
where δ(ξ) is the Dirac delta function.

3◦ . Infinite series solutions containing arbitrary functions of space variables:

X∞
(at)2n n ∂2 ∂2 ∂2
w(x, y, z, t) = f (x, y, z) + ∆ f (x, y, z), ∆≡ + + ,
n=1
(2n)! ∂x2 ∂y 2 ∂z 2
∞
X (at)2n
w(x, y, z, t) = tg(x, y, z) + t ∆n g(x, y, z),
(2n + 1)!
n=1
where f (x, y, z) and g(x, y, z) are any infinitely differentiable functions. The first solution
satisfies the initial conditions w(x, y, z, 0) = f (x, y, z), ∂t w(x, y, z, 0) = 0, while the
second solution satisfies the initial conditions w(x, y, z, 0) = 0, ∂t w(x, y, z, 0) = g(x, y, z).
The sums are finite if f (x, y, z) and g(x, y, z) are polynomials in x, y, z.
4◦ . Suppose w = w(x, y, z, t) is a solution of the wave equation. Then the functions
w1 = Aw(±λx + C1 , ±λy + C2 , ±λz + C3 , ±λt + C4 ),

w2 = Aw p , y, z, p ,
1 − (v/a)2 1 − (v/a)2

A x y z t
w3 = 2 w 2 , , , ,
r − a2 t2 r − a2 t2 r 2 − a2 t2 r 2 − a2 t2 r 2 − a2 t2
where A, Cn , v, and λ are arbitrary constants, are also solutions of the equation. The signs
at λ in the expression of w1 can be taken independently of one another. The function w2 is
a consequence of the invariance of the wave equation under the Lorentz transformation.
⊙ Literature: G. N. Polozhii (1964), W. Miller, Jr. (1977), A. V. Bitsadze and D. F. Kalinichenko (1985).

w = f (x, y, z) at t = 0,
∂t w = g(x, y, z) at t = 0.
Solution (Kirchhoff’s formula):
ZZ ZZ
1 ∂ f (ξ, η, ζ) 1 g(ξ, η, ζ)
w(x, y, z, t) = dS + dS,
4πa ∂t Sat r 4πa Sat r
p
r = (ξ − x)2 + (η − y)2 + (ζ − z)2 ,
where the integration is performed over the surface of the sphere of radius at with center at
(x, y, z).
⊙ Literature: N. S. Koshlyakov, E. B. Gliner, and M. M. Smirnov (1970), A. N. Tikhonov and A. A. Samarskii
(1990).
∂2w
= a 2 ∆3 w 697

w = f0 (x, y, z) at t = 0 (initial condition),
∂t w = f1 (x, y, z) at t = 0 (initial condition),
Solution:
Z l3Z l2Z l1
∂
w(x, y, z, t) = f0 (ξ, η, ζ)G(x, y, z, ξ, η, ζ, t) dξ dη dζ
∂t 0 0 0
Z l3Z l2Z l1
+ f1 (ξ, η, ζ)G(x, y, z, ξ, η, ζ, t) dξ dη dζ
0 0 0
Z t Z l3Z l2
2 ∂
0 0 0 ∂ξ ξ=0
Z t Z l3Z l2
2 ∂
0 0 0 ∂ξ ξ=l1
Z t Z l3Z l1
2 ∂
0 0 0 ∂η η=0
Z t Z l3Z l1
2 ∂
0 0 0 ∂η η=l2
Z t Z l2Z l1
2 ∂
0 0 0 ∂ζ ζ=0
Z t Z l2Z l1
2 ∂
−a g6 (ξ, η, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dτ.
0 0 0 ∂ζ ζ=l3
Here,
∞ ∞ ∞
8 XXX 1
G(x, y, z, ξ, η, ζ, t) = sin(αn x) sin(βm y) sin(γk z)
al1 l2 l3 λnmk
n=1 m=1 k=1
× sin(αn ξ) sin(βm η) sin(γk ζ) sin(aλnmk t),

where
nπ mπ kπ
αn = , βm = , γk = ,
l1 l2 l3
q
λnmk 2 + γ 2.
= α2n + βm k


Solution:
Z l3Z l2Z l1
∂
∂t 0 0 0
Z l3Z l2Z l1
0 0 0
Z t Z l3Z l2
− a2 g1 (η, ζ, τ )G(x, y, z, 0, η, ζ, t − τ ) dη dζ dτ
0 0 0
Z t Z l3Z l2
+ a2 g2 (η, ζ, τ )G(x, y, z, l1 , η, ζ, t − τ ) dη dζ dτ
0 0 0
Z t Z l3Z l1
2
−a g3 (ξ, ζ, τ )G(x, y, z, ξ, 0, ζ, t − τ ) dξ dζ dτ
0 0 0
Z t Z l3Z l1
2
+a g4 (ξ, ζ, τ )G(x, y, z, ξ, l2 , ζ, t − τ ) dξ dζ dτ
0 0 0
Z t Z l2Z l1
2
−a g5 (ξ, η, τ )G(x, y, z, ξ, η, 0, t − τ ) dξ dη dτ
0 0 0
Z t Z l2Z l1
+ a2 g6 (ξ, η, τ )G(x, y, z, ξ, η, l3 , t − τ ) dξ dη dτ,
0 0 0
where
X∞ X ∞ X∞
t 1 An Am Ak
G(x, y, z, ξ, η, ζ, t) = + cos(αn x) cos(βm y) cos(γk z)
l1 l2 l3 al1 l2 l3 n=0 m=0 λnmk
k=0
× cos(αn ξ) cos(βm η) cos(γk ζ) sin(aλnmk t),
(
q
nπ mπ kπ 2 2 2 1 for n = 0,
αn = , βm = , γk = , λnmk = αn + βm + γk , An =
l1 l2 l3 2 for n > 0.
The summation here is performed over the indices satisfying the condition n + m + k > 0;
the term corresponding to n = m = k = 0 is singled out.
∂2w
= a 2 ∆3 w 699


∂x w − s1 w = g1 (y, z, t) at x = 0 (boundary condition),
∂x w + s2 w = g2 (y, z, t) at x = l1 (boundary condition),
∂y w − s3 w = g3 (x, z, t) at y = 0 (boundary condition),
∂y w + s4 w = g4 (x, z, t) at y = l2 (boundary condition),
∂z w − s5 w = g5 (x, y, t) at z = 0 (boundary condition),
∂z w + s6 w = g6 (x, y, t) at z = l3 (boundary condition).
∞ ∞ ∞
8XXX 1
G(x, y, ξ, η, t) = q sin(αn x + εn ) sin(βm y + σm )
a n=1 m=1 E α2 + β2 + γ2
k=1 nmk n m k
q
× sin(γk z + νk ) sin(αn ξ + εn ) sin(βm η + σm ) sin(γk ζ + νk ) sin at α2n + βm 2 + γ2
k
with
αn βm γk
εn = arctan , σm = arctan , νk = arctan ,
l1 l2 l3

(s1 s2 + α2n )(s1 + s2 ) (s3 s4 + βm2
)(s3 + s4 ) (s5 s6 + γk2 )(s5 + s6 )
Enmk = l1 + 2 l2 + 2 l3 + .
(s1 + α2n )(s22 + α2n ) (s3 + βm2 )(s2 + β 2 )
4 m (s25 + γk2 )(s26 + γk2 )
Here, the αn , βm , and γk are positive roots of the transcendental equations
α2 − s1 s2 = (s1 + s2 )α cot(l1 α), β 2 − s3 s4 = (s3 + s4 )β cot(l2 β),

γ 2 − s5 s6 = (s5 + s6 )γ cot(l3 γ).


Solution:
Z l3Z l2Z l1
∂
∂t 0 0 0
Z l3Z l2Z l1
0 0 0
Z t Z l3Z l2
2 ∂
0 0 0 ∂ξ ξ=0
Z t Z l3Z l2
2 ∂
0 0 0 ∂ξ ξ=l1
Z t Z l3Z l1
2
0 0 0
Z t Z l3Z l1
+ a2 g4 (ξ, ζ, τ )G(x, y, z, ξ, l2 , ζ, t − τ ) dξ dζ dτ
0 0 0
Z t Z l2Z l1
− a2 g5 (ξ, η, τ )G(x, y, z, ξ, η, 0, t − τ ) dξ dη dτ
0 0 0
Z t Z l2Z l1
2
+a g6 (ξ, η, τ )G(x, y, z, ξ, η, l3 , t − τ ) dξ dη dτ.
0 0 0
Here,
∞ ∞ ∞
2 X X X Am Ak
G(x, y, z, ξ, η, ζ, t) = sin(αn x) cos(βm y) cos(γk z)
al1 l2 l3 n=1 m=0 λnmk
k=0
× sin(αn ξ) cos(βm η) cos(γk ζ) sin(aλnmk t),
where
nπ mπ kπ
αn = , βm = , γk = ,
l1 l2 l3
(
q 1 for m = 0,
λnmk = 2 + γ2,
α2n + βm Am =
k
2 for m > 0.

∂2w
= a 2 ∆3 w 701
Solution:
Z l3Z l2Z l1
∂
∂t 0 0 0
Z l3Z l2Z l1
0 0 0
Z t Z l3Z l2
2 ∂
0 0 0 ∂ξ ξ=0
Z t Z l3Z l2
0 0 0
Z t Z l3Z l1
2 ∂
0 0 0 ∂η η=0
Z t Z l3Z l1
2
0 0 0
Z t Z l2Z l1
∂
+ a2 g5 (ξ, η, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dτ
0 0 0 ∂ζ ζ=0
Z t Z l2Z l1
2
+a g6 (ξ, η, τ )G(x, y, z, ξ, η, l3 , t − τ ) dξ dη dτ.
0 0 0
Here,
∞ ∞ ∞
8 XXX 1
G(x, y, z, ξ, η, ζ, t) = sin(αn x) sin(βm y) sin(γk z)
al1 l2 l3 n=1 m=1 λnmk
k=1
× sin(αn ξ) sin(βm η) sin(γk ζ) sin(aλnmk t),
where
q
π(2n + 1) π(2m + 1) π(2k + 1) 2 + γ 2.
αn = , βm = , γk = , λnmk = α2n + βm k
2l1 2l2 2l3
The three-dimensional wave equation in the cylindrical coordinate system is written as

p
∂2w 2 1 ∂ ∂w 1 ∂2w ∂2w
=a r + 2 + , r= x2 + y 2 .
∂t2 r ∂r ∂r r ∂ϕ2 ∂z 2
One-dimensional problems with axial symmetry that have solutions w = w(r, t) are
considered in Section 6.2.1. Two-dimensional problems whose solutions have the form
w = w(r, ϕ, t) or w = w(r, z, t) are discussed in Sections 7.1.2 and 7.1.3.

w = f0 (r, ϕ, z) at t = 0 (initial condition),

∂t w = f1 (r, ϕ, z) at t = 0 (initial condition),
Solution:
Z l Z 2πZ R
∂
w(r, ϕ, z, t) = ξf0 (ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ, t) dξ dη dζ
∂t 0 0 0
Z l Z 2πZ R
+ ξf1 (ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ, t) dξ dη dζ
0 0 0
Z t Z l Z 2π
2 ∂
−a R g1 (η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=R
Z t Z 2πZ R
2 ∂
0 0 0 ∂ζ ζ=0
Z t Z 2πZ R
2 ∂
0 0 0 ∂ζ ζ=l
Here,
∞ ∞ ∞
2 XXX An
G(r, ϕ, z, ξ, η, ζ, t) = 2
√ Jn (µnm r)Jn (µnm ξ)
πaR l n=0 m=1
k=1
[Jn (µnm R)]2 λnmk
′
p
kπz kπζ
× cos[n(ϕ − η)] sin sin sin at λnmk ,
l l
(
k2 π2 1 for n = 0,
λnmk = µ2nm + 2 , An =
l 2 for n > 0,


∂2w
= a 2 ∆3 w 703
Solution:
Z l Z 2πZ R
∂
∂t 0 0 0
Z l Z 2πZ R
0 0 0
Z t Z l Z 2π
2
+a R g1 (η, ζ, τ )G(r, ϕ, z, R, η, ζ, t − τ ) dη dζ dτ
0 0 0
Z t Z 2πZ R
2
−a ξg2 (ξ, η, τ )G(r, ϕ, z, ξ, η, 0, t − τ ) dξ dη dτ
0 0 0
Z t Z 2πZ R
+ a2 ξg3 (ξ, η, τ )G(r, ϕ, z, ξ, η, l, t − τ ) dξ dη dτ.
0 0 0
Here,
∞
t 2 X1 kπx kπξ akπt
G(r, ϕ, z, ξ, η, ζ, t)= 2 + 2 2 cos cos sin
πR l π aR k l l l
k=1
∞ ∞ ∞
1 X X X AnAkµ2nmJn(µnmr)Jn(µnmξ) kπx kπξ sin(λnmkt)
+ 2 2 2 2
cos[n(ϕ−η)] cos cos ,
πl n=0 m=1 (µnmR −n )[Jn(µnmR)] l l λnmk
k=0
r (
k 2π 2 1 for n=0,
λnmk =a µ2nm+ 2 , An =
l 2 for n>0,

∞ ∞ ∞
1X XX An µ2nm
G(r, ϕ, z, ξ, η, ζ, t) =
π
n=0 m=1 s=1
(µ2nm R2 + k12 R2 − n2 )[Jn (µnm R)]2 khs k2 λnms
×Jn (µnm r)Jn (µnm ξ) cos[n(ϕ−η)]hs (z)hs (ζ) sin(λnms t),
p k2
λnms = a µ2nm + βs2 , hs (z) = cos(βs z) + sin(βs z),
βs

2 k3 βs2 + k22 k2 l k22
khs k = + 2 + 1+ 2 .
2βs2 βs2 + k32 2βs 2 βs
Here, A0 = 1 and An = 2 for n = 1, 2, . . . ; the Jn (ξ) are Bessel functions; and the µnm
and βs are positive roots of the transcendental equations
tan(βl) k2 + k3
µJn′ (µR) + k1 Jn (µR) = 0, = 2 .
β β − k2 k3
scribed:
w = f0 (r, ϕ, z) at t=0 (initial condition),
∂t w = f1 (r, ϕ, z) at t=0 (initial condition),
w = g1 (ϕ, z, t) at r=R (boundary condition),
∂z w = g2 (r, ϕ, t) at z=0 (boundary condition),
∂z w = g3 (r, ϕ, t) at z=l (boundary condition).
Solution:
Z l Z 2πZ R
∂
∂t 0 0 0
Z l Z 2πZ R
0 0 0
Z t Z l Z 2π
2 ∂
0 0 0 ∂ξ ξ=R
Z t Z 2πZ R
2
0 0 0
Z t Z 2πZ R
2
+a ξg3 (ξ, η, τ )G(r, ϕ, z, ξ, η, l, t − τ ) dξ dη dτ.
0 0 0
Here,
∞ ∞ ∞
1 XXX An Ak
G(r, ϕ, z, ξ, η, ζ, t) = 2
πaR l n=0 m=1
k=0
′
p
kπz kπζ
× cos[n(ϕ − η)] cos cos sin at λnmk ,
l l
(
k2 π2 1 for n = 0,
l 2 for n > 0,
∂2w
= a 2 ∆3 w 705
scribed:

Solution:
Z l Z 2πZ R
∂
∂t 0 0 0
Z l Z 2πZ R
0 0 0
Z tZ l Z 2π
+ a2 R g1 (η, ζ, τ )G(r, ϕ, z, R, η, ζ, t − τ ) dη dζ dτ
0 0 0
Z tZ 2πZ R
2 ∂
0 0 0 ∂ζ ζ=0
Z t Z 2πZ R
2 ∂
0 0 0 ∂ζ ζ=l
Here,
∞
2 X1 kπz kπζ kπat
G(r, ϕ, z, ξ, η, ζ, t) = 2 2 sin sin sin
π aR k l l l
k=1
∞ ∞ ∞
2 XXX An µ2nm Jn (µnm r)Jn (µnm ξ)
+ √
πal
n=0 m=1 k=1
(µ2nm R2 − n2 )[Jn (µnm R)]2 λnmk
p
kπz kπζ
× cos[n(ϕ − η)] sin sin sin at λnmk ,
l l
(
k2 π2 1 for n = 0,
l 2 for n > 0,
prescribed:

Solution:
Z l Z 2πZ R2
∂
w(r, ϕ, z, t) = f0 (ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
∂t 0 0 R1
Z l Z 2πZ R2
+ f1 (ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 R1
Z tZ l Z 2π
2 ∂
+ a R1 g1 (η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=R1
Z t Z l Z 2π
∂
− a2 R2 g2 (η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=R2
Z t Z 2πZ R2
2 ∂
0 0 R1 ∂ζ ζ=0
Z t Z 2πZ R2
2 ∂
0 0 R1 ∂ζ ζ=l
Here,
∞ ∞ ∞
π XXX An µ2nm Jn2 (µnm R2 )
G(r, ϕ, z, ξ, η, ζ, t) = Znm (r)Znm (ξ)
2l J 2 (µ R ) − Jn2 (µnm R2 )
n=0 m=1 k=1 n nm 1
√
kπz kπζ sin at λnmk
× cos[n(ϕ − η)] sin sin √ ,
l l a λnmk
(
1 for n = 0, k2 π2
An = λnmk = µ2nm + 2 ,
2 for n 6= 0, l
Znm (r) = Jn (µnm R1 )Yn (µnm r) − Yn (µnm R1 )Jn (µnm r),

∂2w
= a 2 ∆3 w 707

prescribed:
Solution:
Z l Z 2πZ R2
∂
∂t 0 0 R1
Z l Z 2πZ R2
0 0 R1
Z tZ l Z 2π
− a2 R1 g1 (η, ζ, τ )G(r, ϕ, z, R1 , η, ζ, t − τ ) dη dζ dτ
0 0 0
Z tZ l Z 2π
2
+ a R2 g2 (η, ζ, τ )G(r, ϕ, z, R2 , η, ζ, t − τ ) dη dζ dτ
0 0 0
Z t Z 2πZ R2
− a2 g3 (ξ, η, τ )G(r, ϕ, z, ξ, η, 0, t − τ )ξ dξ dη dτ
0 0 R1
Z tZ 2πZ R2
2
+a g4 (ξ, η, τ )G(r, ϕ, z, ξ, η, l, t − τ )ξ dξ dη dτ.
0 0 R1
Here,
∞
X
t 2 1 kπz kπζ kπat
G(r, ϕ, z, ξ, η, ζ, t) = + cos cos sin
π(R22 −R12)l π 2a(R22 −R12) k l l l
k=1
∞ ∞ ∞
1 XXX AnAk µ2nmZnm(r)Znm(ξ)
+
πl n=0 m=1 (µ2nmR22 −n2)Znm
2 (R )−(µ2 R2 −n2)Z 2 (R )
2 nm 1 nm 1
k=0
√
×cos[n(ϕ−η)] cos cos √ ,
l l a λnmk
where
(
1 for n = 0, k2 π2
An = λnmk = µ2nm + ,
2 for n 6= 0, l2
Znm (r) = Jn′ (µnm R1 )Yn (µnm r) − Yn′ (µnm R1 )Jn (µnm r);
dental equation

prescribed:
∞ ∞ ∞
1 XXX An µ2nm
G(r, ϕ, z, ξ, η, ζ, t) = p
πa khs k2 µ2 + λ2
nm s
n=0 m=1 s=1
p
Znm (r)Znm (ξ) cos[n(ϕ − η)]hs (z)hs (ζ) sin at µ2nm + λ2s
× 2 2 .
(k2 R2 + µ2nm R22 − n2 )Znm
2 (R ) − (k 2 R2 + µ2 R2 − n2 )Z 2 (R )
2 1 1 nm 1 nm 1
Here,
(
1 for n = 0, Znm (r) = µnm Jn′ (µnm R1 ) − k1 Jn (µnm R1 ) Yn (µnm r)
An =
2 for n =
6 0, − µnm Yn′ (µnm R1 ) − k1 Yn (µnm R1 ) Jn (µnm r),

k3 2 k4 λ2s + k32 k3 l k32
hs (z) = cos(λs z) + sin(λs z), khs k = + 2 + 1+ 2 ,
where the Jn (r) and Yn (r) are Bessel functions; the µnm are positive roots of the transcen-
dental equation
′

= µYn′ (µR1 ) − k1 Yn (µR1 ) µJn′ (µR2 ) + k2 Jn (µR2 ) ;
tan(λl) k3 + k4
and the λs are positive roots of the transcendental equation = 2 .
λ λ − k3 k4

1◦ . A hollow circular cylinder of finite length is considered. The following conditions are
prescribed:
∂2w
= a 2 ∆3 w 709
Solution:
Z l Z 2πZ R2
∂
∂t 0 0 R1
Z l Z 2πZ R2
0 0 R1
Z t Z l Z 2π
2 ∂
0 0 0 ∂ξ ξ=R1
Z t Z l Z 2π
2 ∂
− a R2 g2 (η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=R2
Z t Z 2πZ R2
− a2 g3 (ξ, η, τ )G(r, ϕ, z, ξ, η, 0, t − τ )ξ dξ dη dτ
0 0 R1
Z t Z 2πZ R2
2
0 0 R1
Here,
∞ ∞ ∞
π X X X AnAkµ2nmJn2(µnmR2)
G(r, ϕ, z, ξ, η, ζ, t)= Znm(r)Znm(ξ)
4l J 2(µ R )−Jn2(µnmR2)
n=0 m=1 k=0 n nm 1
√
×cos[n(ϕ−η)] cos cos √ ,
l l a λnmk
(
1 for n=0, k2π 2
An = λnmk =µ2nm+ 2 ,
2 for n6= 0, l
Znm(r)=Jn(µnmR1)Yn(µnmr)−Yn(µnmR1)Jn(µnmr),
prescribed:
w = f0 (r, ϕ, z) at t=0 (initial condition),

∂t w = f1 (r, ϕ, z) at t=0 (initial condition),
w = g3 (r, ϕ, t) at z=0 (boundary condition),
w = g4 (r, ϕ, t) at z=l (boundary condition).
Solution:
Z l Z 2πZ R2
∂
∂t 0 0 R1
Z l Z 2πZ R2
0 0 R1
Z t Z l Z 2π
2
− a R1 g1 (η, ζ, τ )G(r, ϕ, z, R1 , η, ζ, t − τ ) dη dζ dτ
0 0 0
Z t Z l Z 2π
2
Z t Z0 2πZ0 R02
2 ∂
0 0 R1 ∂ζ ζ=0
Z t Z 2πZ R2
2 ∂
0 0 R1 ∂ζ ζ=l
Here,
∞
X
2 1 kπz kπζ kπat
G(r, ϕ, z, ξ, η, ζ, t) = 2 sin sin sin
π a(R22 −R12) k l l l
k=1
∞ ∞ ∞
2 XXX Anµ2nmZnm(r)Znm(ξ)
+ 2 2 (R )−(µ2 R2 −n2)Z 2 (R )
πl
n=0 m=1 k=1
(µ2nmR2 −n2)Znm 2 nm 1 nm 1
√
×cos[n(ϕ−η)] sin sin √ ,
l l a λnmk
where
(
1 for n = 0, k2 π2
An = λnmk = µ2nm + ,
2 for n 6= 0, l2
dental equation

scribed:
∂2w
= a 2 ∆3 w 711
Solution:
Z l Z ϕ0Z R
∂
∂t 0 0 0
Z l Z ϕ0Z R
0 0 0
Z t Z l Z ϕ0
2 ∂
0 0 0 ∂ξ ξ=R
Z tZ lZ R
2 1 ∂
0 0 0 ξ ∂η η=0
Z tZ lZ R
2 1 ∂
0 0 0 ξ ∂η η=ϕ0
Z t Z ϕ0Z R
2 ∂
0 0 0 ∂ζ ζ=0
Z t Z ϕ0Z R
2 ∂
0 0 0 ∂ζ ζ=l
Here,
∞ ∞ ∞
8 X X X Jnπ/ϕ0 (µnm r)Jnπ/ϕ0 (µnm ξ) nπϕ
G(r, ϕ, z, ξ, η, ζ, t) = 2 ′ sin
R lϕ0 n=1 m=1 [Jnπ/ϕ (µnm R)]2 ϕ0
k=1 0
p
nπη kπz kπζ sin at µ2nm + k2 π 2/l2
× sin sin sin p ,
ϕ0 l l a µ2nm + k2 π 2/l2
scribed:
Solution:
Z l Z ϕ0Z R
∂
∂t 0 0 0
Z l Z ϕ0Z R
0 0 0
Z t Z l Z ϕ0
2 ∂
0 0 0 ∂ξ ξ=R
Z tZ lZ R
2 1 ∂
0 0 0 ξ ∂η η=0
Z tZ lZ R
2 1 ∂
0 0 0 ξ ∂η η=ϕ0
Z t Z ϕ0Z R
2
−a g4 (ξ, η, τ )G(r, ϕ, z, ξ, η, 0, t − τ )ξ dξ dη dτ
0 0 0
Z t Z ϕ0Z R
2
0 0 0
Here,
∞ ∞ ∞
4 X X X Ak Jnπ/ϕ0 (µnm r)Jnπ/ϕ0 (µnm ξ) nπϕ
G(r, ϕ, z, ξ, η, ζ, t) = 2 ′ sin
R lϕ0 [Jnπ/ϕ (µnm R)]2 ϕ0
n=1 m=1 k=0 0
p
nπη kπz kπζ sin at µ2nm + k2 π 2/l2
× sin cos cos p ,
ϕ0 l l a µ2nm + k2 π 2/l2
where A0 = 1 and Ak = 2 for k ≥ 1; the Jnπ/ϕ0 (r) are Bessel functions; and the µnm are
positive roots of the transcendental equation Jnπ/ϕ0 (µR) = 0.

The three-dimensional wave equation in the spherical coordinate system is represented as

∂2w 2 1 ∂ 2 ∂w 1 ∂ ∂w 1 ∂2w
= a r + sin θ + ,
∂t2 r 2 ∂r ∂r r 2 sin θ ∂θ ∂θ r 2 sin2 θ ∂ϕ2
p
r = x2 + y 2 + z 2 .
One-dimensional problems with central symmetry that have solutions w = w(r, t) are

w = f0 (r, θ, ϕ) at t = 0 (initial condition),
∂t w = f1 (r, θ, ϕ) at t = 0 (initial condition),
∂2w
= a 2 ∆3 w 713
Solution:
Z 2πZ πZ R
∂
w(r, θ, ϕ, t)= f0(ξ, η, ζ)G(r, θ, ϕ, ξ, η, ζ, t)ξ 2 sin η dξ dη dζ
∂t 0 0 0
Z 2πZ πZ R
+ f1(ξ, η, ζ)G(r, θ, ϕ, ξ, η, ζ, t)ξ 2 sin η dξ dη dζ
0 0 0
Z tZ 2πZ π
2 2 ∂
−a R g(η, ζ, τ ) G(r, θ, ϕ, ξ, η, ζ, t−τ ) sin η dη dζ dτ,
0 0 0 ∂ξ ξ=R
where
X∞ X ∞ Xn
1
2πaR2 rξ n=0 m=1
k=0
× Pnk (cos θ)Pnk (cos η) cos[k(ϕ − ζ)] sin(λnm at),
(
1 for k = 0, (2n + 1)(n − k)!
Ak = Bnmk = ′ 2 .
2 for k =
6 0, (n + k)! Jn+1/2 (λnm R) λnm
expressed in terms of the Legendre polynomials Pn (µ) as
dk 1 dn
Pnk (µ) = (1 − µ2 )k/2 Pn (µ), Pn (µ) = (µ2 − 1)n ,
dµk n! 2n dµn

Solution:
Z 2πZ π Z R
∂
w(r, θ, ϕ, t) = f0 (ξ, η, ζ)G(r, θ, ϕ, ξ, η, ζ, t)ξ 2 sin η dξ dη dζ
∂t 0 0 0
Z 2πZ π Z R
+ f1 (ξ, η, ζ)G(r, θ, ϕ, ξ, η, ζ, t)ξ 2 sin η dξ dη dζ
0 0 0
Z t Z 2πZ π
2 2
+a R g(η, ζ, τ )G(r, θ, ϕ, R, η, ζ, t − τ ) sin η dη dζ dτ,
0 0 0
where
X∞ X ∞ Xn
3t 1
G(r, θ, ϕ, ξ, η, ζ, t)= + √ AkBnmkJn+1/2(λnmr)Jn+1/2(λnmξ)
4πR3 2πa rξ n=0 m=1
k=0
×Pn (cos θ)Pnk(cos η) cos[k(ϕ−ζ)] sin(λnmat),
k
(
1 for k=0, λnm(2n+1)(n−k)!
Ak = Bnmk = 2 .
2 for k6= 0, (n+k)! R2λ2nm−n(n+1) Jn+1/2(λnmR)
the paragraph above), and the λnm are positive roots of the transcendental equation
′
2λRJn+1/2 (λR) − Jn+1/2 (λR) = 0.

X∞ X ∞ X n
1
2πa rξ n=0 m=1 s=0
×Pns(cos θ)Pns(cos η) cos[s(ϕ−ζ)] sin(λnmat),
(
1 for s=0, λnm(2n+1)(n−s)!
As = Bnms = 2 .
2 for s6= 0, (n+s)! R2λ2nm+(kR+n)(kR−n−1) Jn+1/2(λnmR)
′

λRJn+1/2 (λR) + kR − 12 Jn+1/2 (λR) = 0.

Solution:
Z Z Z
∂ 2π π R2
∂t 0 0 R1
Z 2πZ πZ R2
0 0 R1
Z tZ 2πZ π
2 ∂
+a R12 g1(η, ζ, τ ) G(r, θ, ϕ, ξ, η, ζ, t−τ ) sin η dη dζ dτ
0 0 0 ∂ξ ξ=R1
Z tZ 2πZ π
2 2 ∂
−a R2 g2(η, ζ, τ ) G(r, θ, ϕ, ξ, η, ζ, t−τ ) sin η dη dζ dτ,
0 0 0 ∂ξ ξ=R2
∂2w
= a 2 ∆3 w 715
where
∞ ∞ n
π XXX
G(r, θ, ϕ, ξ, η, ζ, t) = √ Ak Bnmk Zn+1/2 (λnm r)Zn+1/2 (λnm ξ)
8a rξ n=0 m=1 k=0
× Pnk (cos θ)Pnk (cos η) cos[k(ϕ − ζ)] sin(λnm at).
Here,
( 2
1 for k = 0, λnm (2n + 1)(n − k)! Jn+1/2 (λnm R2 )
Ak = Bnmk = 2 2
,
dk 1 dn
Pnk (µ) = (1 − µ2 )k/2 Pn (µ), Pn (µ) = (µ2 − 1)n ,
dµk n! 2n dµn
and the λnm are positive roots of the transcendental equation Zn+1/2 (λR2 ) = 0.

problem.

Solution:
Z 2πZ π Z R2
∂
∂t 0 0 R1
Z 2πZ π Z R2
0 0 R1
Z tZ 2πZ π
− a2 R12 g1 (η, ζ, τ )G(r, θ, ϕ, R1 , η, ζ, t − τ ) sin η dη dζ dτ
0 0 0
Z tZ 2πZ π
+ a2 R22 g2 (η, ζ, τ )G(r, θ, ϕ, R2 , η, ζ, t − τ ) sin η dη dζ dτ,
0 0 0
where
X∞ X ∞ Xn
3t 1 Ak
G(r, θ, ϕ, ξ, η, ζ, t) = + √ Zn+1/2 (λnm r)Zn+1/2 (λnm ξ)
4π(R23 − R13 ) 4πa rξ n=0 m=1 Bnmk
k=0
× Pn (cos θ)Pnk (cos η) cos[k(ϕ
k
− ζ)] sin(λnm at),
( Z
R2
1 for k = 0, λnm (n + k)! 2
2 for k 6= 0, (2n + 1)(n − k)! R1

′ 1
Zn+1/2 (λnm r) = λnm Jn+1/2 (λnm R1 ) − Jn+1/2 (λnm R1 ) Yn+1/2 (λnm r)
2R1

′ 1
− λnm Yn+1/2 (λnm R1 ) − Yn+1/2 (λnm R1 ) Jn+1/2 (λnm r).
2R1
Here, the Jn+1/2 (r) and Yn+1/2 (r) are Bessel functions, the Pnk (µ) are associated Legendre
functions (see the paragraph above), and the λnm are positive roots of the transcendental
equation
′ 1
λZn+1/2 (λR2 ) − Z (λR2 ) = 0.
2R2 n+1/2

X∞ X ∞ X n
1 As
4πa rξ n=0 m=1 s=0 Bnms
× Pns (cos θ)Pns (cos η) cos[s(ϕ − ζ)] sin(λnm at).
Here,
( Z R2
1 for s = 0, λnm (n + s)! 2
2 for s =
6 0, (2n + 1)(n − s)! R1

′ 1
2R1

′ 1
− λYn+1/2 (λR1 ) − k1 + Yn+1/2 (λR1 ) Jn+1/2 (λr),
2R1

′ 1
λZn+1/2 (λR2 ) + k2 − Zn+1/2 (λR2 ) = 0.
2R2
∂2w
= a2 ∆3 w + Φ(x, y, z, t) 717
8.2 Nonhomogeneous Wave Equation

∂ 2w
2
= a2 ∆3 w + Φ(x, y, z, t)
∂t
w = f (x, y, z) at t = 0,
∂t w = g(x, y, z) at t = 0.
Solution:
ZZ ZZ
1 ∂ f (ξ, η, ζ) 1 g(ξ, η, ζ)
w(x, y, z, t) = dS + dS
4πa ∂t r 4πa r
r=at r=at
ZZZ
1 1 r
+ Φ ξ, η, ζ, t − dξ dη dζ,
4πa2 r a
r≤at
p
r= (ξ − x)2 + (η − y)2 + (ζ − z)2 ,
where the integration is performed over the surface of the sphere (r = at) and the volume
of the sphere (r ≤ at) with center at (x, y, z).
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z ≤ l3 . Different boundary value problems.

1◦ . The solution w(x, y, z, t) of the first boundary value problem for a parallelepiped is
given by the formula in Section 8.1.1 (see the first boundary value problem for 0 ≤ x ≤ l1 ,
0 ≤ y ≤ l2 , 0 ≤ z ≤ l3 ) with the additional term
Z t Z l1Z l2Z l3
Φ(ξ, η, ζ, τ )G(x, y, z, ξ, η, ζ, t − τ ) dζ dη dξ dτ,
0 0 0 0
2◦ . The solution w(x, y, z, t) of the second boundary value problem for a parallelepiped
is given by the formula from Section 8.1.1 (see the second boundary value problem for
0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z ≤ l3 ) with the additional term specified in Item 1◦ above;
3◦ . The solution w(x, y, z, t) of the third boundary value problem for a parallelepiped is
given by the formula from Section 8.1.1 (see the third boundary value problem for 0 ≤ x ≤
l1 , 0 ≤ y ≤ l2 , 0 ≤ z ≤ l3 ) with the additional term specified in Item 1◦ above; the Green’s
4◦ . The solutions of mixed boundary value problems for a parallelepiped are given by the
formulas from Section 8.1.1 (see the mixed boundary value problems for 0 ≤ x ≤ l1 ,
0 ≤ y ≤ l2 , 0 ≤ z ≤ l3 ) to which one should add the term specified in Item 1◦ above; the

A three-dimensional nonhomogeneous wave equation in the cylindrical coordinate system
is written as

∂2w 2 1 ∂ ∂w 1 ∂2w ∂2w
=a r + 2 + + Φ(r, ϕ, z, t).
∂t2 r ∂r ∂r r ∂ϕ2 ∂z 2
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l. Different boundary value problems.

1◦ . The solution w(r, ϕ, z, t) of the first boundary value problem for a circular cylinder
of finite length is given by the formula from Section 8.1.2 (see the first boundary value
problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l) with the additional term
Z t Z l Z 2πZ R
Φ(ξ, η, ζ, τ )G(r, ϕ, z, ξ, η, ζ, t − τ )ξ dξ dη dζ dτ, (1)
0 0 0 0
2◦ . The solution w(r, ϕ, z, t) of the second boundary value problem for a circular cylinder
of finite length is given by the formula from Section 8.1.2 (see the second boundary value
problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l) with the additional term (1); the Green’s
3◦ . The solution w(r, ϕ, z, t) of the third boundary value problem for a circular cylinder
of finite length is the sum of the solution specified in Section 8.1.2 (see the third boundary
value problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l) and expression (1); the Green’s
4◦ . The solutions w(r, ϕ, z, t) of mixed boundary value problems for a circular cylinder of
finite length are given by the formulas from Section 8.1.2 (see the mixed boundary value
problems for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l) with additional terms of the form (1); the
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l. Different boundary value

problems.
1◦ . The solution w(r, ϕ, z, t) of the first boundary value problem for a hollow cylinder of
finite dimensions is given by the formula from Section 8.1.2 (see the first boundary value
problem for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l) with the additional term
Z t Z l Z 2πZ R2
0 0 0 R1
2◦ . The solution w(r, ϕ, z, t) of the second boundary value problem for a hollow cylinder
of finite dimensions is given by the formula from Section 8.1.2 (see the second boundary
value problem for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l) with the additional term (2); the
∂2w
= a2 ∆3 w + Φ(x, y, z, t) 719
3◦ . The solution w(r, ϕ, z, t) of the third boundary value problem for a hollow cylinder
of finite dimensions is the sum of the solution specified in Section 8.1.2 (see the third
boundary value problem for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l) and expression (2); the
4◦ . The solutions w(r, ϕ, z, t) of mixed boundary value problems for a hollow cylinder of
finite dimensions are given by the formulas from Section 8.1.2 (see the mixed boundary
value problems for R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ l) with additional terms of the
form (2); the Green’s function is also taken from Section 8.1.2.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ l. Different boundary value problems.

1◦ . The solution w(r, ϕ, z, t) of the first boundary value problem for a cylindrical sector
of finite thickness is given by the formula from Section 8.1.2 (see the first boundary value
problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ l) with the additional term
Z t Z l Z ϕ0Z R
0 0 0 0
2◦ . The solution w(r, ϕ, z, t) of a mixed boundary value problem for a cylindrical sector of
finite thickness is given by the formula from Section 8.1.2 (see the mixed boundary value
problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ l) with the additional term (3); the Green’s

A three-dimensional nonhomogeneous wave equation in the spherical coordinate system is
represented as

∂2w 2 1 ∂ 2 ∂w 1 ∂ ∂w 1 ∂2w
=a r + 2 sin θ + 2 2 + Φ(r, θ, ϕ, t).
∂t2 r 2 ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂ϕ2
◮ Domain: 0 ≤ r ≤ R, 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π. Boundary value problem.

1◦ . The solution w(r, θ, ϕ, t) of the first boundary value problem for a sphere is given
by the formula from Section 8.1.3 (see the first boundary value problem for 0 ≤ r ≤ R,
0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π) with the additional term
Z t Z 2πZ π Z R
Φ(ξ, η, ζ, τ )G(r, θ, ϕ, ξ, η, ζ, t − τ )ξ 2 sin η dξ dη dζ dτ, (1)
0 0 0 0
2◦ . The solution w(r, θ, ϕ, t) of the second boundary value problem for a sphere is given
by the formula from Section 8.1.3 (see the second boundary value problem for 0 ≤ r ≤ R,
0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π) with the additional term (1); the Green’s function is also taken
from Section 8.1.3.
3◦ . The solution w(r, θ, ϕ, t) of the third boundary value problem for a sphere is the sum of
the solution specified in Section 8.1.3 (see the third boundary value problem for 0 ≤ r ≤ R,
0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π) and expression (1); the Green’s function is also taken from
Section 8.1.3.
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π. Boundary value problems.

1◦ . The solution w(r, θ, ϕ, t) of the first boundary value problem for a spherical layer is
given by the formula from Section 8.1.3 (see the first boundary value problem for R1 ≤
r ≤ R2 , 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π) with the additional term
Z t Z 2πZ π Z R2
Φ(ξ, η, ζ, τ )G(r, θ, ϕ, ξ, η, ζ, t − τ )ξ 2 sin η dξ dη dζ dτ, (2)
0 0 0 R1
2◦ . The solution w(r, θ, ϕ, t) of the second boundary value problem for a spherical layer
is given by the formula from Section 8.1.3 (see the second boundary value problem for
R1 ≤ r ≤ R2 , 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π) with the additional term (2); the Green’s function
3◦ . The solution w(r, θ, ϕ, t) of the third boundary value problem for a spherical layer is
the sum of the solution specified in Section 8.1.3 (see the third boundary value problem for
R1 ≤ r ≤ R2 , 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π) and expression (2); the Green’s function is also

∂ 2w
2
= a2 ∆3 w − bw + Φ(x, y, z, t)
∂t
A three-dimensional nonhomogeneous Klein–Gordon equation in the rectangular Cartesian
2
∂2w 2 ∂ w ∂2w ∂2w
=a + + − bw + Φ(x, y, z, t).
∂t2 ∂x2 ∂y 2 ∂z 2
1◦ . For b = −c2 < 0,
p
1 δ(t − r/a) c I1 c t2 − r 2/a2
E (x, y, z, t) = − p ϑ(t − r/a) ,
4πa2 r a t2 − r 2/a2
p
where r = x2 + y 2 + z 2 , δ(ξ) is the Dirac delta function, ϑ(ξ) is the Heaviside unit step
function, and I1 (z) is the modified Bessel function.
∂2 w
= a2 ∆3 w − bw + Φ(x, y, z, t) 721
2◦ . For b = c2 > 0,
p
1 δ(t − r/a) c J1 c t2 − r 2/a2
E (x, y, z, t) = − p ϑ(t − r/a) ,
4πa2 r a t2 − r 2/a2
where J1 (z) is the Bessel function.
w = f (x, y, z) at t = 0,
∂t w = g(x, y, z) at t = 0.
Let a = 1 and Φ(x, y, z, t) ≡ 0.
1◦ . Solution for b = −c2 < 0:

Z t p
∂ 1 ∂ 2

w(x, y, z, t) = 2 2
r I0 c t − r Tr f (x, y, z) dr
∂t t ∂t 0
Z t p
1 ∂
+ r 2 I0 c t2 − r 2 Tr g(x, y, z) dr.
t ∂t 0

Here, I0 (z) is the modified Bessel function and Tr h(x, y, z) is the average of h(x, y, z)
over the spherical surface with center at (x, y, z) and radius r:
Z 2πZ π
1
Tr h(x, y, z) = h(x + r sin θ cos ϕ, y + r sin θ sin ϕ, z + r cos θ) sin θ dθ dϕ.
4π 0 0
2◦ . Solution for b = c2 > 0:

Z t p
∂ 1 ∂ 2

w(x, y, z, t) = 2 2
r J0 c t − r Tr f (x, y, z) dr
∂t t ∂t 0
Z t p
1 ∂
+ r 2 J0 c t2 − r 2 Tr g(x, y, z) dr,
t ∂t 0
where J0 (z) is the Bessel function.
⊙ Literature: V. I. Smirnov (1974, Vol. 2).


Solution:
Z l3Z l2Z l1
∂
∂t 0 0 0
Z l3Z l2Z l1
0 0 0
Z t Z l3Z l2
2 ∂
0 0 0 ∂ξ ξ=0
Z t Z l3Z l2
2 ∂
0 0 0 ∂ξ ξ=l1
Z t Z l3Z l1
2 ∂
0 0 0 ∂η η=0
Z t Z l3Z l1
2 ∂
0 0 0 ∂η η=l2
Z t Z l2Z l1
2 ∂
0 0 0 ∂ζ ζ=0
Z t Z l2Z l1
2 ∂
−a g6 (ξ, η, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dτ
0 0 0 ∂ζ ζ=l3
Z t Z l3Z l2Z l1
+ Φ(ξ, η, ζ, τ )G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dζ dτ.
0 0 0 0
Here,
X∞ X ∞ X ∞
8 1
G(x, y, z, ξ, η, ζ, t) = √ sin(αn x) sin(βm y) sin(γk z)
l1 l2 l3 λnmk
n=1 m=1 k=1
p

× sin(αn ξ) sin(βm η) sin(γk ζ) sin t λnmk ,
where
nπ mπ kπ
αn = , βm = , γk = , λnmk = a2 (α2n + βm
2
+ γk2 ) + b.
l1 l2 l3
∂2 w
= a2 ∆3 w − bw + Φ(x, y, z, t) 723


Solution:
Z l3Z l2Z l1
∂
∂t 0 0 0
Z l3Z l2Z l1
0 0 0
Z t Z l3Z l2
2
−a g1 (η, ζ, τ )G(x, y, z, 0, η, ζ, t − τ ) dη dζ dτ
0 0 0
Z t Z l3Z l2
2
+a g2 (η, ζ, τ )G(x, y, z, l1 , η, ζ, t − τ ) dη dζ dτ
0 0 0
Z t Z l3Z l1
2
0 0 0
Z t Z l3Z l1
0 0 0
Z t Z l2Z l1
2
0 0 0
Z t Z l2Z l1
2
+a g6 (ξ, η, τ )G(x, y, z, ξ, η, l3 , t − τ ) dξ dη dτ
0 0 0
Z t Z l3Z l2Z l1
+ Φ(ξ, η, ζ, τ )G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dζ dτ,
0 0 0 0
where
√ X∞ X ∞ X ∞
sin t b 1 An Am Ak
G(x, y, z, ξ, η, ζ, t) = √ + √ cos(αn x) cos(βm y)
l1 l2 l3 b l l l
1 2 3 n=0 m=0 λnmk
k=0
p

× cos(γk z) cos(αn ξ) cos(βm η) cos(γk ζ) sin t λnmk ,
(
nπ mπ kπ 2 2 2 2 1 for n = 0,
αn = , βm = , γk = , λnmk = a (αn +βm +γk )+b, An =
l1 l2 l3 2 for n > 0.
The summation is performed over the indices satisfying the condition n + m + k > 0; the
term corresponding to n = m = k = 0 is singled out.

∂x w − s1 w = g1 (y, z, t) at x = 0 (boundary condition),
∂y w − s3 w = g3 (x, z, t) at y = 0 (boundary condition),
∂z w − s5 w = g5 (x, y, t) at z = 0 (boundary condition),
∞ X
X ∞ X
∞
1
G(x, y, ξ, η, t) = 8 √ sin(αnx+εn) sin(βmy+σm) sin(γk z+νk )
n=1 m=1 k=1
Enmk λnmk
p
×sin(αnξ +εn) sin(βmη+σm) sin(γk ζ +νk) sin t λnmk ,
αn βm γk
εn = arctan , σm = arctan , νk = arctan , λnmk = a2(α2n +βm 2
+γk2)+b,
l1 l2 l3

(s1s2 +α2n)(s1 +s2) (s3s4 +βm2 )(s +s )
3 4 (s5s6 +γk2)(s5 +s6)
Enmk = l1 + 2 l2 + 2 l3 + 2 2 2 2 .
(s1 +α2n)(s22 +α2n) 2 )(s2 +β 2 )
(s3 +βm 4 m (s5 +γk )(s6 +γk )
Here, the αn , βm , and γk are positive roots of the transcendental equations
α2 − s1 s2 = (s1 + s2 )α cot(l1 α), β 2 − s3 s4 = (s3 + s4 )β cot(l2 β),
γ 2 − s5 s6 = (s5 + s6 )γ cot(l3 γ).

∂2 w
= a2 ∆3 w − bw + Φ(x, y, z, t) 725
Solution:
Z l3Z l2Z l1
∂
∂t 0 0 0
Z l3Z l2Z l1
0 0 0
Z t Z l3Z l2
2 ∂
0 0 0 ∂ξ ξ=0
Z t Z l3Z l2
2 ∂
0 0 0 ∂ξ ξ=l1
Z t Z l3Z l1
− a2 g3 (ξ, ζ, τ )G(x, y, z, ξ, 0, ζ, t − τ ) dξ dζ dτ
0 0 0
Z t Z l3Z l1
2
0 0 0
Z t Z l2Z l1
2
0 0 0
Z t Z l2Z l1
2
0 0 0
Z t Z l3Z l2Z l1
0 0 0 0
Here,
∞ ∞ ∞
2 X X X Am Ak
G(x, y, z, ξ, η, ζ, t) = √ sin(αn x) cos(βm y) cos(γk z)
l1 l2 l3 λnmk
n=1 m=0 k=0
p
× sin(αn ξ) cos(βm η) cos(γk ζ) sin t λnmk ,
where
( (
1 for m = 0, 1 for k = 0,
Am = Ak =
2 for m > 0, 2 for k > 0,
nπ mπ kπ
αn = , βm = , γk = , λnmk = a2 (α2n + βm
2
+ γk2 ) + b.
l1 l2 l3
Solution:
Z l3Z l2Z l1
∂
∂t 0 0 0
Z l3Z l2Z l1
0 0 0
Z t Z l3Z l2
∂
+ a2 g1 (η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=0
Z t Z l3Z l2
2
+a g2 (η, ζ, τ )G(x, y, z, l1 , η, ζ, t − τ ) dη dζ dτ
0 0 0
Z t Z l3Z l1
2 ∂
0 0 0 ∂η η=0
Z t Z l3Z l1
2
0 0 0
Z t Z l2Z l1
2 ∂
0 0 0 ∂ζ ζ=0
Z t Z l2Z l1
+ a2 g6 (ξ, η, τ )G(x, y, z, ξ, η, l3 , t − τ ) dξ dη dτ
0 0 0
Z t Z l3Z l2Z l1
0 0 0 0
Here,
∞ ∞ ∞
8 XXX 1
G(x, y, z, ξ, η, ζ, t) = √ sin(αn x) sin(βm y) sin(γk z)
l1 l2 l3 λnmk
n=1 m=1 k=1
p
× sin(αn ξ) sin(βm η) sin(γk ζ) sin t λnmk ,
where
π(2n + 1) π(2m + 1) π(2k + 1)
αn = , βm = , γk = ,
2l1 2l2 2l3
λnmk = a2 (α2n + βm2
+ γk2 ) + b.

A nonhomogeneous Klein–Gordon equation in the cylindrical coordinate system is written
as
p
∂2w 2 1 ∂ ∂w 1 ∂2w ∂2w
= a r + + − bw + Φ(r, ϕ, z, t), r = x2 + y 2 .
∂t2 r ∂r ∂r r 2 ∂ϕ2 ∂z 2
treated in Section 6.2.5. Two-dimensional problems whose solutions have the form w =
w(r, ϕ, t) or w = w(r, z, t) are considered in Sections 7.3.2 and 7.3.3.
∂2 w
= a2 ∆3 w − bw + Φ(x, y, z, t) 727

Solution:
Z l Z 2πZ R
∂
∂t 0 0 0
Z l Z 2πZ R
0 0 0
Z tZ lZ 2π
∂
− a2 R g1 (η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=R
Z t Z 2πZ R
2 ∂
0 0 0 ∂ζ ζ=0
Z t Z 2πZ R
2 ∂
−a ξg3 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dη dτ
0 0 0 ∂ζ ζ=l
Z t Z l Z 2πZ R
+ ξΦ(ξ, η, ζ, τ )G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dη dζ dτ.
0 0 0 0
Here,
∞ ∞ ∞
2 XXX An
G(r, ϕ, z, ξ, η, ζ, t) = 2
πR l
n=0 m=1 k=1
′
p
kπz kπζ
× cos[n(ϕ − η)] sin sin sin t λnmk ,
l l
where
(
a2 k 2 π 2 1 for n = 0,
λnmk = a2 µ2nm + + b, An =
l2 2 for n > 0,
the Jn (ξ) are Bessel functions (the prime denotes a derivative with respect to the argument),
and the µnm are positive roots of the transcendental equation Jn (µR) = 0.

Solution:
Z l Z 2πZ R
∂
∂t 0 0 0
Z l Z 2πZ R
0 0 0
Z t Z l Z 2π
2
+a R g1 (η, ζ, τ )G(r, ϕ, z, R, η, ζ, t − τ ) dη dζ dτ
0 0 0
Z t Z 2πZ R
2
0 0 0
Z t Z 2πZ R
+ a2 ξg3 (ξ, η, τ )G(r, ϕ, z, ξ, η, l, t − τ ) dξ dη dτ
0 0 0
Z t Z l Z 2πZ R
0 0 0 0
Here,
√ ∞
sin t b 2 X 1 kπx kπξ p
G(r, ϕ, z, ξ, η, ζ, t)= √ + 2 √ cos cos sin t βk
πR2l b πR l k=1 βk l l
∞ ∞ ∞
1 X X X AnAkµ2nmJn(µnmr)Jn(µnmξ) kπx kπξ sin(λnmkt)
+ cos[n(ϕ−η)] cos cos ,
πl n=0 m=1 (µ2nmR2−n2)[Jn(µnmR)]2 l l λnmk
k=0
r (
a2k 2π 2 a 2k 2π 2 1 for n=0,
βk = +b, λnmk = a2µ2nm+ 2 +b, An =
l2 l 2 for n>0,

∂2 w
= a2 ∆3 w − bw + Φ(x, y, z, t) 729
∞ ∞ ∞
1XXX An µ2nm
G(r, ϕ, z, ξ, η, ζ, t) = 2
π (µ2nm R2 + k1 R2 − n2 )[Jn (µnm R)]2 khs k2 λnms
n=0 m=1 s=1
× Jn (µnm r)Jn (µnm ξ) cos[n(ϕ − η)]hs (z)hs (ζ) sin(λnms t).
Here,
(
1 for n = 0, p
An = λnms = a2 µ2nm + a2 βs2 + b,
2 for n > 0,

k2 k3 βs2 + k22
2 k2 l k22
hs (z) = cos(βs z) + sin(βs z), khs k = + 2 + 1+ 2 ,
βs 2βs2 βs2 + k32 2βs 2 βs
the Jn (ξ) are Bessel functions, and the µnm and βs are positive roots of the transcendental
equations
tan(βl) k2 + k3
µJn′ (µR) + k1 Jn (µR) = 0, = 2 .
β β − k2 k3
scribed:
Solution:
Z l Z 2πZ R
∂
∂t 0 0 0
Z l Z 2πZ R
0 0 0
Z t Z l Z 2π
2 ∂
0 0 0 ∂ξ ξ=R
Z t Z 2πZ R
2
0 0 0
Z t Z 2πZ R
+ a2 ξg3 (ξ, η, τ )G(r, ϕ, z, ξ, η, l, t − τ ) dξ dη dτ
0 0 0
Z t Z l Z 2πZ R
0 0 0 0
Here,
∞ ∞ ∞
1 XXX An Ak
G(r, ϕ, z, ξ, η, ζ, t) = 2 ′
√
2 λ
Jn (µnm r)Jn (µnm ξ)
πR l n=0 m=1 [Jn (µ nm R)] nmk
k=0
p
kπz kπζ
× cos[n(ϕ − η)] cos cos sin t λnmk ,
l l
(
a2 k 2 π 2 1 for n = 0,
λnmk = a2 µ2nm + 2
+ b, An =
l 2 for n > 0,
scribed:

Solution:
Z l Z 2πZ R
∂
∂t 0 0 0
Z l Z 2πZ R
0 0 0
Z tZ lZ 2π
0 0 0
Z tZ 2πZ R
2 ∂
0 0 0 ∂ζ ζ=0
Z t Z 2πZ R
2 ∂
0 0 0 ∂ζ ζ=l
Z t Z l Z 2πZ R
0 0 0 0
∂2 w
= a2 ∆3 w − bw + Φ(x, y, z, t) 731
Here,
∞
2 X 1 kπz kπζ p
G(r, ϕ, z, ξ, η, ζ, t) = 2
√ sin sin sin t βk
πR l βk l l
k=1
∞
XX ∞ X
∞
2 An µ2nm
+ √ Jn (µnm r)
πl
n=0 m=1 k=1
(µ2nm R2 − n2 )[Jn (µnm R)]2 λnmk
p
kπz kπζ
× Jn (µnm ξ) cos[n(ϕ − η)] sin sin sin t λnmk ,
l l
(
2
a k π2 2 a2 k 2 π 2 1 for n = 0,
βk = 2
+ b, λnmk = a2 µ2nm + 2
+ b, An =
l l 2 for n > 0,

prescribed:
Solution:
Z l Z 2πZ R2
∂
∂t 0 0 R1
Z l Z 2πZ R2
0 0 R1
Z tZ lZ 2π
∂
+ a2 R1 g1 (η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=R1
Z t Z l Z 2π
2 ∂
0 0 0 ∂ξ ξ=R2
Z t Z 2πZ R2
2 ∂
0 0 R1 ∂ζ ζ=0
Z t Z 2πZ R2
2 ∂
−a g4 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) ξ dξ dη dτ
0 0 R1 ∂ζ ζ=l
Z t Z l Z 2πZ R2
+ Φ(ξ, η, ζ, τ )G(r, ϕ, z, ξ, η, ζ, t − τ )ξ dξ dη dζ dτ.
0 0 0 R1
Here,
∞ ∞ ∞
π XXX An µ2nm Jn2 (µnm R2 )
2l J 2 (µ R ) − Jn2 (µnm R2 )
n=0 m=1 k=1 n nm 1
√
kπz kπζ sin t λnmk
l l λnmk
where
(
1 for n = 0, a2 k2 π 2
An = λnmk = a2 µ2nm + + b,
2 for n =
6 0, l2
Znm (r) = Jn (µnm R1 )Yn (µnm r) − Yn (µnm R1 )Jn (µnm r);
dental equation

prescribed:
Solution:
Z l Z 2πZ R2
∂
∂t 0 0 R1
Z l Z 2πZ R2
0 0 R1
Z tZ lZ 2π
0 0 0
Z tZ lZ 2π
2
0 0 0
Z tZ 2πZ R2
2
0 0 R1
Z tZ 2πZ R2
+ a2 g4 (ξ, η, τ )G(r, ϕ, z, ξ, η, l, t − τ )ξ dξ dη dτ
0 0 R1
Z tZ l Z 2πZ R2
0 0 0 R1
∂2 w
= a2 ∆3 w − bw + Φ(x, y, z, t) 733
Here,
√ ∞ √
sin t b 2 X kπz kπζ sin t βk
G(r, ϕ, z, ξ, η, ζ, t) = √ + cos cos √
π(R22 − R12)l b π(R22 − R12)l k=1 l l βk
∞ ∞ ∞
1 XXX An Ak µ2nmZnm(r)Znm (ξ)
+ 2 2 (R ) − (µ2 R2 − n2)Z 2 (R )
πl n=0 m=1 (µnmR2 − n2)Znm
2
2 nm 1 nm 1
k=0
√
× cos[n(ϕ − η)] cos cos √ ,
l l λnmk
where
(
1 for n = 0, a2 k2 π 2 a2 k2 π 2
An = βk = + b, λnmk = a2 µ2nm + + b,
2 for n =
6 0, l2 l2
dental equation

prescribed:

∞ ∞ ∞
1 XXX A µ2
G(r, ϕ, z, ξ, η, ζ, t) = p n nm
π 2 a2 µ2 + a2 λ2 + b
n=0 m=1 s=1 khs k nm s
p
Znm (r)Znm (ξ) cos[n(ϕ − η)]hs (z)hs (ζ) sin t a2 µ2nm + a2 λ2s + b
× 2 2 .
(k2 R2 + µ2nm R22 − n2 )Znm
2 (R ) − (k 2 R2 + µ2 R2 − n2 )Z 2 (R )
2 1 1 nm 1 nm 1
Here,

Znm(r)= µnmJn′ (µnmR1)−k1Jn(µnmR1) Yn(µnmr)

(
1 for n=0, k3 2 k4 λ2s+k32 k3 l k32
An = hs(z)=cos(λsz)+ sin(λsz), khsk = 2 2 2 + 2 + 1+ 2 ,
2 for n6= 0, λs 2λs λs+k4 2λs 2 λs
′

= µYn′ (µR1 ) − k1 Yn (µR1 ) µJn′ (µR2 ) + k2 Jn (µR2 ) ,
tan(λl) k3 + k4
and the λs are positive roots of the transcendental equation = 2 .
λ λ − k3 k4

prescribed:

Solution:
Z l Z 2πZ R2
∂
∂t 0 0 R1
Z l Z 2πZ R2
0 0 R1
Z tZ lZ 2π
2 ∂
0 0 0 ∂ξ ξ=R1
Z t Z l Z 2π
2 ∂
0 0 0 ∂ξ ξ=R2
Z t Z 2πZ R2
2
0 0 R1
Z tZ 2πZ R2
0 0 R1
Z tZ lZ 2πZ R2
0 0 0 R1
Here,
∞ ∞ ∞
π X X X An Ak µ2nm Jn2 (µnm R2 )
4l J 2 (µ R ) − Jn2 (µnm R2 )
n=0 m=1 k=0 n nm 1
√
l l λnmk
∂2 w
= a2 ∆3 w − bw + Φ(x, y, z, t) 735
where
(
1 for n = 0, a2 k2 π 2
An = λnmk = a2 µ2nm + + b,
2 for n 6= 0, l2
Znm (r) = Jn (µnm R1 )Yn (µnm r) − Yn (µnm R1 )Jn (µnm r);
dental equation
prescribed:
Solution:
Z l Z 2πZ R2
∂
∂t 0 0 R1
Z l Z 2πZ R2
0 0 R1
Z tZ lZ 2π
0 0 0
Z t Z l Z 2π
+ a2 R2 g2 (η, ζ, τ )G(r, ϕ, z, R2 , η, ζ, t − τ ) dη dζ dτ
0 0 0
Z t Z 2πZ R2
2 ∂
0 0 R1 ∂ζ ζ=0
Z t Z 2πZ R2
2 ∂
0 0 R1 ∂ζ ζ=l
Z t Z l Z 2πZ R2
0 0 0 R1
Here,
∞ √
2 X kπz kπζ sin t βk
G(r, ϕ, z, ξ, η, ζ, t)= sin sin √
π(R22−R12)l l l βk
k=1
∞ ∞
XXX ∞
2 Anµ2nmZnm(r)Znm(ξ)
+
πl
n=0 m=1 k=1
(µ2nmR22−n2)Znm
2 (R )−(µ2 R2−n2)Z 2 (R )
2 nm 1 nm 1
√
×cos[n(ϕ−η)] sin sin √ ,
l l λnmk
where
(
1 for n = 0, a2 k2 π 2 a2 k2 π 2
An = βk = + b, λnmk = a2 µ2nm + + b,
2 for n =
6 0, l2 l2
dental equation
scribed:
Solution:
Z l Z ϕ0Z R
∂
∂t 0 0 0
Z l Z ϕ0Z R
0 0 0
Z t Z l Z ϕ0
2 ∂
0 0 0 ∂ξ ξ=R
Z tZ lZ R
2 1 ∂
0 0 0 ξ ∂η η=0
Z tZ lZ R
2 1 ∂
0 0 0 ξ ∂η η=ϕ0
Z t Z ϕ0Z R
∂
+ a2 g4 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) ξ dξ dη dτ
0 0 0 ∂ζ ζ=0
Z t Z ϕ0Z R
2 ∂
0 0 0 ∂ζ ζ=l
Z t Z l Z ϕ0Z R
0 0 0 0
∂2 w
= a2 ∆3 w − bw + Φ(x, y, z, t) 737
Here,
∞ ∞ ∞
8 X X X Jnπ/ϕ0(µnmr)Jnπ/ϕ0(µnmξ) nπϕ
G(r, ϕ, z, ξ, η, ζ, t)= 2 ′ sin
R lϕ0 n=1 m=1 [Jnπ/ϕ (µnmR)]2 ϕ0
k=1 0
p
nπη kπz kπζ sin t a2µ2nm+a2k2π 2l−2+b
×sin sin sin p ,
ϕ0 l l a2µ2nm+a2k2π 2l−2+b
scribed:
Solution:
Z l Z ϕ0Z R
∂
∂t 0 0 0
Z l Z ϕ0Z R
0 0 0
Z t Z l Z ϕ0
∂
− a2 R g1 (η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=R
Z tZ lZ R
2 1 ∂
0 0 0 ξ ∂η η=0
Z tZ lZ R
2 1 ∂
0 0 0 ξ ∂η η=ϕ0
Z t Z ϕ0Z R
2
0 0 0
Z t Z ϕ0Z R
0 0 0
Z t Z l Z ϕ0Z R
0 0 0 0
Here,
∞ ∞ ∞
4 X X X Ak Jnπ/ϕ0 (µnm r)Jnπ/ϕ0 (µnm ξ) nπϕ
G(r, ϕ, z, ξ, η, ζ, t) = 2 ′ sin
R lϕ0 n=1 m=1 [Jnπ/ϕ (µnm R)]2 ϕ0
k=0 0
p
nπη kπz kπζ sin t a2 µ2nm + a2 k2 π 2 l−2 + b
× sin cos cos p ,
ϕ0 l l a2 µ2nm + a2 k2 π 2 l−2 + b
where A0 = 1 and Ak = 2 for k ≥ 1; the Jnπ/ϕ0 (r) are Bessel functions; and the µnm are

A nonhomogeneous Klein–Gordon equation in the spherical coordinate system is written
as

∂ 2w 2 1 ∂ 2 ∂w 1 ∂ ∂w 1 ∂ 2w
=a r + 2 sin θ + 2 2 − bw + Φ(r, θ, ϕ, t),
∂t2 r2 ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂ϕ2
p
r = x2 + y 2 + z 2.
w(r, t) are treated in Section 6.2.6.


Solution:
Z 2πZ πZ R
∂
∂t 0 0 0
Z 2πZ πZ R
0 0 0
Z tZ 2πZ π
2 2 ∂
−a R g(η, ζ, τ ) G(r, θ, ϕ, ξ, η, ζ, t−τ ) sin η dη dζ dτ
0 0 0 ∂ξ ξ=R
Z tZ 2πZ πZ R
+ Φ(ξ, η, ζ, τ )G(r, θ, ϕ, ξ, η, ζ, t−τ )ξ 2 sin η dξ dη dζ dτ.
0 0 0 0
Here,
X∞ X ∞ X n
1
2πR2 rξ n=0 m=1
k=0
p
× Pnk (cos θ)Pnk (cos η) cos[k(ϕ − ζ)] sin t a2 λ2nm + b ,
∂2 w
= a2 ∆3 w − bw + Φ(x, y, z, t) 739
where
(
1 for k = 0, (2n + 1)(n − k)!
Ak = Bnmk = p ;
2 for k 6= 0, (n + ′
k)! [Jn+1/2 (λnm R)]2 a2 λ2nm + b
the Jn+1/2 (r) are Bessel functions, the Pnk (µ) are associated Legendre functions expressed
in terms of the Legendre polynomials Pn (µ) as
dk 1 dn
Pnk (µ) = (1 − µ2 )k/2 Pn (µ), Pn (µ) = (µ2 − 1)n ,
dµk n! 2n dµn


Solution:
Z 2πZ π Z R
∂
∂t 0 0 0
Z 2πZ π Z R
0 0 0
Z t Z 2πZ π
+ a2 R2 g(η, ζ, τ )G(r, θ, ϕ, R, η, ζ, t − τ ) sin η dη dζ dτ
0 0 0
Z t Z 2πZ π Z R
+ Φ(ξ, η, ζ, τ )G(r, θ, ϕ, ξ, η, ζ, t − τ )ξ 2 sin η dξ dη dζ dτ.
0 0 0 0
Here,
√ ∞ ∞ n
3 sin t b 1 X X X Ak Bnmk
G(r, θ, ϕ, ξ, η, ζ, t) = √ + √ p Jn+1/2 (λnm r)
4πR3 b 2π rξ n=0 m=1 k=0 a2 λ2nm + b
p
× Jn+1/2 (λnm ξ)Pnk (cos θ)Pnk (cos η) cos[k(ϕ − ζ)] sin t a2 λ2nm + b ,
where
(
1 for k = 0, λ2nm (2n + 1)(n − k)!
Ak = Bnmk = 2 ;
the Jn+1/2 (r) are Bessel functions, the Pnk (µ) are associated Legendre functions (see the
paragraph above), and the λnm are positive roots of the transcendental equation
′
2λRJn+1/2 (λR) − Jn+1/2 (λR) = 0.


∞ X
X ∞ X
n
1 A B
G(r, θ, ϕ, ξ, η, ζ, t) = √ p s nms Jn+1/2 (λnm r)Jn+1/2 (λnm ξ)
2π rξ n=0 m=1 s=0 a2 λ2nm + b
p
× Pns (cos θ)Pns (cos η) cos[s(ϕ − ζ)] sin t a2 λ2nm + b .
Here,
(
1 for s=0, λ2nm(2n+1)(n−s)!
As = Bnms = 2 ;
2 for s6= 0, (n+s)! R2λ2nm+(kR+n)(kR−n−1) Jn+1/2(λnmR)
the Jn+1/2 (r) are Bessel functions, the Pns (µ) are associated Legendre functions (see
′

λRJn+1/2 (λR) + kR − 12 Jn+1/2 (λR) = 0.

Solution:
Z 2πZ πZ R2
∂
∂t 0 0 R1
Z 2πZ πZ R2
0 0 R1
Z tZ 2πZ π
2 ∂
0 0 0 ∂ξ ξ=R1
Z tZ 2πZ π
2 2 ∂
−a R2 g2(η, ζ, τ ) G(r, θ, ϕ, ξ, η, ζ, t−τ ) sin η dη dζ dτ
0 0 0 ∂ξ ξ=R2
Z tZ 2πZ πZ R2
+ Φ(ξ, η, ζ, τ )G(r, θ, ϕ, ξ, η, ζ, t−τ )ξ 2 sin η dξ dη dζ dτ,
0 0 0 R1
∂2 w
= a2 ∆3 w − bw + Φ(x, y, z, t) 741
where
∞ ∞ n
π X X X Ak Bnmk
G(r, θ, ϕ, ξ, η, ζ, t) = √ p Zn+1/2 (λnm r)Zn+1/2 (λnm ξ)
8 rξ n=0 m=1 a2 λ2nm + b
k=0
p
× Pnk (cos θ)Pnk (cos η) cos[k(ϕ − ζ)] sin t a2 λ2nm + b .
Here,
( 2
1 for k = 0, λnm (2n + 1)(n − k)! Jn+1/2 (λnm R2 )
Ak = Bnmk = 2 2 ,
2 for k 6= 0, (n + k)! [Jn+1/2 (λnm R1 ) − Jn+1/2 (λnm R2 )]
dk 1 dn
Pnk (µ) = (1 − µ2 )k/2 Pn (µ), Pn (µ) = (µ2 − 1)n ,
dµk n! 2n dµn

problem.

Solution:
Z 2πZ π Z R2
∂
∂t 0 0 R1
Z 2πZ π Z R2
0 0 R1
Z tZ 2πZ π
− a2 R12 g1 (η, ζ, τ )G(r, θ, ϕ, R1 , η, ζ, t − τ ) sin η dη dζ dτ
0 0 0
Z t Z 2πZ π
+ a2 R22 g2 (η, ζ, τ )G(r, θ, ϕ, R2 , η, ζ, t − τ ) sin η dη dζ dτ
0 0 0
Z t Z 2πZ π Z R2
+ Φ(ξ, η, ζ, τ )G(r, θ, ϕ, ξ, η, ζ, t − τ )ξ 2 sin η dξ dη dζ dτ,
0 0 0 R1
where
√ ∞ X ∞ X n
3 sin t b 1 X Ak
G(r, θ, ϕ, ξ, η, ζ, t) = √ + √ Zn+1/2 (λnm r)
4π(R23 − R13 ) b 4π rξ n=0 m=1 k=0 Bnmk
p
k k sin t a2 λ2nm + b
× Zn+1/2 (λnm ξ)Pn (cos θ)Pn (cos η) cos[k(ϕ − ζ)] p .
a2 λ2nm + b
Here,
( Z R2
1 for k = 0, (n + k)! 2
2 for k 6= 0, (2n + 1)(n − k)! R1

′ 1
Zn+1/2 (λnm r) = λnm Jn+1/2 (λnm R1 ) − J (λnm R1 ) Yn+1/2 (λnm r)
2R1 n+1/2

′ 1
− λnm Yn+1/2 (λnm R1 ) − Y (λnm R1 ) Jn+1/2 (λnm r),
2R1 n+1/2
where the Jn+1/2 (r) and Yn+1/2 (r) are Bessel functions, the Pnk (µ) are associated Legen-
dre functions (see the paragraph above), and the λnm are positive roots of the transcendental
equation
′ 1
λZn+1/2 (λR2 ) − Z (λR2 ) = 0.
2R2 n+1/2

X∞ X ∞ X n
1 As
4π rξ n=0 m=1 s=0 Bnms p
s s sin t a2 λ2nm + b
× Pn (cos θ)Pn (cos η) cos[s(ϕ − ζ)] p .
a2 λ2nm + b
Here,
( Z R2
1 for s = 0, (n + s)! 2
2 for s =
6 0, (2n + 1)(n − s)! R1

′ 1
2R1

′ 1
− λYn+1/2 (λR1 ) − k1 + Yn+1/2 (λR1 ) Jn+1/2 (λr),
2R1
∂2w
+ k ∂w
∂t
= a2 ∆3 w − bw + Φ(x, y, z, t) 743

′ 1
λZn+1/2 (λR2 ) + k2 − Zn+1/2 (λR2 ) = 0.
2R2
8.4 Telegraph Equation

∂ 2w ∂w
2
+ k = a2∆3 w − bw + Φ(x, y, z, t)
∂t ∂t
A three-dimensional nonhomogeneous telegraph equation in the rectangular Cartesian sys-

tem of coordinates has the form
2
∂2w ∂w 2 ∂ w ∂2w ∂2w
+k =a + + − bw + Φ(x, y, z, t).
∂t2 ∂t ∂x2 ∂y 2 ∂z 2
◮ Reduction to the three-dimensional Klein–Gordon equation.

The substitution w(x, y, z, t) = exp − 12 kt u(x, y, z, t) leads to the equation
2
∂2u 2 ∂ u ∂2u ∂2u
2
=a 2
+ 2
+ 2
− b − 14 k2 u + exp 1
2 kt Φ(x, y, z, t),
∂t ∂x ∂y ∂z

Solution:
Z l3Z l2Z l1
∂
∂t 0 0 0
Z l3Z l2Z l1

+ f1 (ξ, η, ζ) + kf0 (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
0 0 0
Z t Z l3Z l2
∂
0 0 0 ∂ξ ξ=0
Z t Z l3Z l2
2 ∂
0 0 0 ∂ξ ξ=l1
Z t Z l3Z l1
2 ∂
0 0 0 ∂η η=0
Z t Z l3Z l1
2 ∂
0 0 0 ∂η η=l2
Z t Z l2Z l1
2 ∂
0 0 0 ∂ζ ζ=0
Z t Z l2Z l1
2 ∂
−a g6 (ξ, η, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dτ
0 0 0 ∂ζ ζ=l3
Z t Z l3Z l2Z l1
0 0 0 0
Here,
∞ ∞ ∞
8 X X X 1
G(x, y, z, ξ, η, ζ, t) = exp − 12 kt √ sin(αnx) sin(βmy) sin(γsz)
l1l2l3 λnms
n=1 m=1 s=1
p
×sin(αnξ) sin(βmη) sin(γsζ) sin t λnms ,
where
nπ mπ sπ
αn = , βm = , γs = , λnms = a2 (α2n + βm
2
+ γs2 ) + b − 14 k2 .
l1 l2 l3

∂2w
+ k ∂w
∂t
= a2 ∆3 w − bw + Φ(x, y, z, t) 745
Solution:
Z l3Z l2Z l1
∂
∂t
Z l3Z0 l2Z0 l1 0

0
Z t0Z l30Z l2
− a2 g1 (η, ζ, τ )G(x, y, z, 0, η, ζ, t − τ ) dη dζ dτ
0 0 0
Z t Z l3Z l2
0 0 0
Z t Z l3Z l1
2
Z0 t Z0 l3Z0 l1
0 0 0
Z t Z l2Z l1
− a2 g5 (ξ, η, τ )G(x, y, z, ξ, η, 0, t − τ ) dξ dη dτ
0 0 0
Z t Z l2Z l1
2
Z t Z0 l3Z0 0
Z
l2 l1
0 0 0 0
where
√ ∞ X ∞ X ∞
e−kt/2 sin t c X An Am As
G(x, y, z, ξ, η, ζ, t) = √ + √ cos(αn x) cos(βm y)
l1 l2 l3 c n=0 m=0 s=0
λnms
p

× cos(γs z) cos(αn ξ) cos(βm η) cos(γs ζ) sin t λnms ,
(
1 for n = 0, nπ mπ sπ
An = αn = , βm = , γs = ,
2 for n > 0, l1 l2 l3
c = b − 14 k2 , λnms = a2 (α2n + βm
2
+ γs2 ) + b − 14 k2 .
The summation is performed over the indices satisfying the condition n + m + s > 0; the
term corresponding to n = m = s = 0 is singled out.

w = f0 (x, y, z) at t=0 (initial condition),
∂t w = f1 (x, y, z) at t=0 (initial condition),
∂x w − s1 w = g1 (y, z, t) at x=0 (boundary condition),
∂y w − s3 w = g3 (x, z, t) at y=0 (boundary condition),
∂z w − s5 w = g5 (x, y, t) at z=0 (boundary condition),
∞ X ∞ X ∞
X 1
G(x, y, ξ, η, t) = 8 exp − 12 kt p sin(αn x + εn ) sin(βp y + σp )
n=1 p=1 q=1
E npq λnpq

p
× sin(γq z + νq ) sin(αn ξ + εn ) sin(βp η + σp ) sin(γq ζ + νq ) sin t λnpq .
Here,
αn βp γq
εn = arctan , σp = arctan , νq = arctan , λnpq = a2 (α2n + βp2 + γq2 ) + b − 14 k 2 ,
l1 l2 l3

2
(s1 s2 + αn )(s1 + s2 ) (s3 s4 + βp2 )(s3 + s4 ) (s5 s6 + γq2 )(s5 + s6 )
Enpq = l1 + l2 + l3 + ,
(s21 + α2n )(s22 + α2n ) (s23 + βp2 )(s24 + βp2 ) (s25 + γq2 )(s26 + γq2 )
where the αn , βp , and γq are positive roots of the transcendental equations
α2 − s1 s2 = (s1 + s2 )α cot(l1 α),
β 2 − s3 s4 = (s3 + s4 )β cot(l2 β),
γ 2 − s5 s6 = (s5 + s6 )γ cot(l3 γ).

w = f0 (x, y, z) at t=0 (initial condition),
∂t w = f1 (x, y, z) at t=0 (initial condition),
w = g1 (y, z, t) at x=0 (boundary condition),
∂y w = g3 (x, z, t) at y=0 (boundary condition),
∂z w = g5 (x, y, t) at z=0 (boundary condition),
Solution:
Z l3Z l2Z l1
∂
∂t 0 0 0
Z l3Z l2Z l1

0 0 0
Z t Z l3Z l2
∂
0 0 0 ∂ξ ξ=0
Z t Z l3Z l2
2 ∂
0 0 0 ∂ξ ξ=l1
Z t Z l3Z l1
− a2 g3 (ξ, ζ, τ )G(x, y, z, ξ, 0, ζ, t − τ ) dξ dζ dτ
0 0 0
∂2w
+ k ∂w
∂t
= a2 ∆3 w − bw + Φ(x, y, z, t) 747
Z t Z l3Z l1
0 0 0
Z t Z l2Z l1
2
0 0 0
Z t Z l2Z l1
2
0 0 0
Z t Z l3Z l2Z l1
0 0 0 0
where
∞ ∞ ∞
2 X X X AmAs
G(x, y, z, ξ, η, ζ, t) = exp − 12 kt √ sin(αnx) cos(βmy) cos(γsz)
l1l2l3 λnms
n=1 m=0 s=0
p
×sin(αnξ) cos(βmη) cos(γsζ) sin t λnms ,
(
1 for m = 0, nπ mπ sπ
Am = αn = , βm = , γs = ,
2 for m > 0, l1 l2 l3
λnms = a2(α2n +βm
2
+γs2)+b− 14 k2.

Solution:
Z l3Z l2Z l1
∂
∂t 0 0 0
Z l3Z l2Z l1

0 0 0
Z t Z l3Z l2
2 ∂
0 0 0 ∂ξ ξ=0
Z t Z l3Z l2
0 0 0
Z t Z l3Z l1
2 ∂
0 0 0 ∂η η=0
Z t Z l3Z l1
0 0 0
Z t Z l2Z l1
∂
+ a2 g5 (ξ, η, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dτ
0 0 0 ∂ζ ζ=0
Z t Z l2Z l1
2
0 0 0
Z t Z l3Z l2Z l1
0 0 0 0
where
∞ X ∞ X ∞
8 X 1
G(x, y, z, ξ, η, ζ, t) = exp − 21 kt √ sin(αn x) sin(βm y) sin(γs z)
l1 l2 l3 n=1 m=1 s=1
λnms
p
× sin(αn ξ) sin(βm η) sin(γs ζ) sin t λnms ,
π(2n + 1) π(2m + 1) π(2s + 1)
αn = , βm = , γs = , λnms = a2 (α2n + βm2
+ γs2 ) + b − 14 k 2 .
2l1 2l2 2l3
A three-dimensional nonhomogeneous telegraph equation in the cylindrical coordinate sys-

tem is written as
p
∂2w ∂w 2 1 ∂ ∂w 1 ∂2w ∂2w
+k = a r + + −bw+Φ(r, ϕ, z, t), r= x2 + y 2 .
∂t2 ∂t r ∂r ∂r r 2 ∂ϕ2 ∂z 2
treated in Section 6.4.2. Two-dimensional problems whose solutions have the form w =
w(r, ϕ, t) or w = w(r, z, t) are considered in Sections 7.4.2 and 7.4.3.

∂2w
+ k ∂w
∂t
= a2 ∆3 w − bw + Φ(x, y, z, t) 749
Solution:
Z l Z 2πZ R
∂
∂t 0 0 0
Z l Z 2πZ R

+ ξ f1 (ξ, η, ζ) + kf0 (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t) dξ dη dζ
0 0 0
Z t Z l Z 2π
2 ∂
0 0 0 ∂ξ ξ=R
Z t Z 2πZ R
2 ∂
0 0 0 ∂ζ ζ=0
Z t Z 2πZ R
2 ∂
0 0 0 ∂ζ ζ=l
Z t Z l Z 2πZ R
0 0 0 0
Here,
∞ ∞ ∞
2e−kt/2 X X X An
G(r, ϕ, z, ξ, η, ζ, t) = Jn (µnm r)Jn (µnm ξ)
πR l n=0 m=1 s=1 [Jn (µnm R)]2
2 ′
√
sπz sπζ sin t λnms
l l λnms
where
(
a2 s2 π 2 1 for n = 0,
λnms = a2 µ2nm + + b − 14 k2 , An =
l2 2 for n > 0,
the Jn (ξ) are Bessel functions (the prime denotes a derivative with respect to the argument),
and the µnm are positive roots of the transcendental equation Jn (µR) = 0.

Solution:
Z l Z 2πZ R
∂
∂t 0 0 0
Z l Z 2πZ R

0 0 0
Z t Z l Z 2π
0 0 0
Z t Z 2πZ R
2
0 0 0
Z t Z 2πZ R
2
+a ξg3 (ξ, η, τ )G(r, ϕ, z, ξ, η, l, t − τ ) dξ dη dτ
0 0 0
Z t Z l Z 2πZ R
0 0 0 0
Here,
√ ∞
sin t c 2 X 1 sπx sπξ p
G(r, ϕ, z, ξ, η, ζ, t)=exp − 12 kt 2
√ + 2
√ cos cos sin t βs
πR l c πR l s=1 βs l l
∞ ∞ ∞
1 X X X AnAsµ2nmJn(µnmr)Jn(µnmξ) sπx sπξ sin(λnmst)
+ cos[n(ϕ−η)] cos cos ,
πl n=0 m=1 s=0 (µ2nmR2−n2)[Jn(µnmR)]2 l l λnms
r (
k2 a2s2π 2 k2 2 2
a2s2π 2 k2 1 for n=0,
c=b− , βs = 2
+b− , λnms = a µnm+ 2 +b− , An =
4 l 4 l 4 2 for n>0,


∂r w + s1 w = g(ϕ, z, t) at r = R (boundary condition),
∂z w − s2 w = g2 (r, ϕ, t) at z = 0 (boundary condition),
∂z w + s3 w = g3 (r, ϕ, t) at z = l (boundary condition).
∞ ∞ ∞
1 X X X Anµ2nmJn(µnmr)Jn(µnmξ)
G(r, ϕ, z, ξ, η, ζ, t)= exp − 12 kt
π n=0 m=1 p=1
(µ2nmR2+s21R2−n2)[Jn(µnmR)]2
p
hp(z)hp(ζ) sin t λnmp
× cos[n(ϕ−η)] p .
khpk2 λnmp
∂2w
+ k ∂w
∂t
= a2 ∆3 w − bw + Φ(x, y, z, t) 751
Here, the Jn (ξ) are Bessel functions,

(
1 for n = 0,
An = λnmp = a2 µ2nm + a2 βp2 + b − 14 k2 ,
2 for n > 0,

s2 2s3 βp2 + s22 s2 l s22
hp (z) = cos(βp z) + sin(βp z), khp k = + 2 + 1+ 2 ;
βp 2βp2 βp2 + s23 2βp 2 βp
the µnm and βp are positive roots of the transcendental equations
tan(βl) s2 + s3
µJn′ (µR) + s1 Jn (µR) = 0, = 2 .
β β − s2 s3

scribed:
Solution:
Z l Z 2πZ R
∂
∂t 0 0 0
Z l Z 2πZ R

0 0 0
Z t Z l Z 2π
2 ∂
0 0 0 ∂ξ ξ=R
Z t Z 2πZ R
− a2 ξg2 (ξ, η, τ )G(r, ϕ, z, ξ, η, 0, t − τ ) dξ dη dτ
0 0 0
Z t Z 2πZ R
2
+a ξg3 (ξ, η, τ )G(r, ϕ, z, ξ, η, l, t − τ ) dξ dη dτ
0 0 0
Z t Z l Z 2πZ R
0 0 0 0
Here,
∞ X ∞ X ∞
1 1
X An As
G(r, ϕ, z, ξ, η, ζ, t) = exp − 2 kt √ Jn (µnm r)
πR2 l [J ′
n (µ nm R)]2 λ
nms
n=0 m=1 s=0
p
sπz sπζ
× Jn (µnm ξ) cos[n(ϕ − η)] cos cos sin t λnms ,
l l
(
2 2 a2 s2 π 2 1 2 1 for n = 0,
λnms = a µnm + 2
+ b − 4 k , An =
l 2 for n > 0,
scribed:
Solution:
Z l Z 2πZ R
∂
∂t 0 0 0
Z l Z 2πZ R

0 0 0
Z t Z l Z 2π
0 0 0
Z t Z 2πZ R
2 ∂
0 0 0 ∂ζ ζ=0
Z t Z 2πZ R
2 ∂
0 0 0 ∂ζ ζ=l
Z t Z l Z 2πZ R
0 0 0 0
Here,
∞
2 1
X 1 sπz sπζ p
G(r, ϕ, z, ξ, η, ζ, t) = 2
exp − 2 kt √ sin sin sin t βs
πR l s=1
βs l l
∞ ∞ ∞
2 X X X Anµ2nm
+ exp − 12 kt Jn(µnmr)
πl (µ2nmR2 −n2)[Jn(µnmR)]2
n=0 m=1 s=1
√
×Jn(µnmξ) cos[n(ϕ−η)] sin sin √ ,
l l λnms
(
a2s2π 2 1 for n = 0,
βs = 2
+b− 14 k2, An =
l 2 for n > 0,
a2s2π 2
λnms = a2µ2nm + +b− 14 k2,
l2
∂2w
+ k ∂w
∂t
= a2 ∆3 w − bw + Φ(x, y, z, t) 753

prescribed:
Solution:
Z l Z 2πZ R2
∂
∂t 0 0 R1
Z l Z 2πZ R2

+ f1 (ξ, η, ζ) + kf0 (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 R1
Z tZ lZ 2π
2 ∂
0 0 0 ∂ξ ξ=R1
Z t Z l Z 2π
2 ∂
0 0 0 ∂ξ ξ=R2
Z t Z 2πZ R2
2 ∂
0 0 R1 ∂ζ ζ=0
Z t Z 2πZ R2
2 ∂
0 0 R1 ∂ζ ζ=l
Z t Z l Z 2πZ R2
0 0 0 R1
Here,
∞ X ∞ X ∞
π X An µ2nm Jn2 (µnm R2 )
G(r, ϕ, z, ξ, η, ζ, t) = exp − 12 kt Znm (r)
2l Jn2 (µnm R1 ) − Jn2 (µnm R2 )
n=0 m=1 s=1
√
× Znm (ξ) cos[n(ϕ − η)] sin sin √ ,
l l λnms
(
1 for n = 0, a2 s2 π 2
An = λnms = a2 µ2nm + + b − 14 k2 ,
2 for n =
6 0, l2


prescribed:

Solution:
Z l Z 2πZ R2
∂
∂t 0 0 R1
Z l Z 2πZ R2

0 0 R1
Z tZ lZ 2π
0 0 0
Z tZ lZ 2π
2
0 0 0
Z tZ 2πZ R2
2
0 0 R1
Z tZ 2πZ R2
2
+a g4 (ξ, η, τ )G(r, ϕ, z, ξ, η, l, t − τ )ξ dξ dη dτ
0 0 R1
Z tZ lZ 2πZ R2
0 0 0 R1
Here,
√ ∞ √
e−kt/2 sin t c X sπz sπζ sin t βs
G(r, ϕ, z, ξ, η, ζ, t) = √ + 2 cos cos √
π(R22 − R12)l c s=1
l l βs
∞ ∞ ∞
e−kt/2 X X X An As µ2nmZnm(r)Znm (ξ)
+
πl n=0 m=1 s=0
(µ2nmR22 − n2)Znm2 (R ) − (µ2 R2 − n2 )Z 2 (R )
2 nm 1 nm 1
√
l l λnms
where
(
1 for n = 0, a2 s2 π 2 a2 s2 π 2
An = c = b− 14 k 2 , βs = +b− 1 2
4 k , λnms = a 2 2
µ nm + +b− 14 k 2 ,
2 for n =
6 0, l2 l2
∂2w
+ k ∂w
∂t
= a2 ∆3 w − bw + Φ(x, y, z, t) 755
dental equation
prescribed:

∂r w − s1 w = g1 (ϕ, z, t) at r = R1 (boundary condition),
∂r w + s2 w = g2 (ϕ, z, t) at r = R2 (boundary condition),
∂z w − s3 w = g3 (r, ϕ, t) at z = 0 (boundary condition),
∂z w + s4 w = g4 (r, ϕ, t) at z = l (boundary condition).
∞ ∞ ∞
1 X X X An µ2nm
G(r, ϕ, z, ξ, η, ζ, t) = exp − 12 kt p
π 2 a2 µ2 + a2 λ2 + b − k 2/4
n=0 m=1 p=1 khp k nm p
p
Znm (r)Znm (ξ) cos[n(ϕ − η)]hp (z)hp (ζ) sin t a2 µ2nm + a2 λ2p + b − k2/4
× .
(s22 R22 + µ2nm R22 − n2 )Znm
2 (R ) − (s2 R2 + µ2 R2 − n2 )Z 2 (R )
2 1 1 nm 1 nm 1
Here,

Znm (r) = µnm Jn′ (µnm R1 ) − s1 Jn (µnm R1 ) Yn (µnm r)

− µnm Yn′ (µnm R1 ) − s1 Yn (µnm R1 ) Jn (µnm r),
(
1 for n = 0, s3
An = hp (z) = cos(λp z) + sin(λp z),
2 for n 6= 0, λp

2 s4 λ2p + s23 s3 l s23
khp k = + 2 + 1+ 2 ,
2λ2p λ2p + s24 2λp 2 λp
where the Jn (r) and Yn (r) are Bessel functions, the µnm are positive roots of the transcen-
dental equation

µJn′ (µR1 ) − s1 Jn (µR1 ) µYn′ (µR2 ) + s2 Yn (µR2 )

= µYn′ (µR1 ) − s1 Yn (µR1 ) µJn′ (µR2 ) + s2 Jn (µR2 ) ,
tan(λl) s3 + s4
and the λp are positive roots of the transcendental equation = 2 .
λ λ − s3 s4

prescribed:

Solution:
Z l Z 2πZ R2
∂
∂t 0 0 R1
Z l Z 2πZ R2

0 0 R1
Z tZ lZ 2π
2 ∂
0 0 0 ∂ξ ξ=R1
Z t Z l Z 2π
2 ∂
0 0 0 ∂ξ ξ=R2
Z t Z 2πZ R2
2
0 0 R1
Z tZ 2πZ R2
0 0 R1
Z tZ lZ 2πZ R2
0 0 0 R1
Here,
∞ X ∞ X ∞
π 1
X An As µ2nm Jn2 (µnm R2 )
G(r, ϕ, z, ξ, η, ζ, t) = exp − 2 kt Znm (r)
4l J 2 (µ R ) − Jn2 (µnm R2 )
n=0 m=1 s=0 n nm 1
√
× Znm (ξ) cos[n(ϕ − η)] cos cos √ ,
l l λnms
(
1 for n = 0, a2 s2 π 2
An = λnms = a2 µ2nm + + b − 14 k2 ,
2 for n 6= 0, l2

∂2w
+ k ∂w
∂t
= a2 ∆3 w − bw + Φ(x, y, z, t) 757
prescribed:
Solution:
Z l Z 2πZ R2
∂
∂t 0 0 R1
Z l Z 2πZ R2

0 0 R1
Z tZ lZ 2π
0 0 0
Z tZ lZ 2π
2
0 0 0
Z t Z 2πZ R2
∂
+ a2 g3 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) ξ dξ dη dτ
0 0 R1 ∂ζ ζ=0
Z t Z 2πZ R2
2 ∂
0 0 R1 ∂ζ ζ=l
Z t Z l Z 2πZ R2
0 0 0 R1
Here,
∞ √
2e−kt/2 X sπz sπζ sin t βs
G(r, ϕ, z, ξ, η, ζ, t) = sin sin √
π(R22 − R12)l s=1 l l βs
∞ ∞ ∞
2e−kt/2 X X X An µ2nmZnm(r)Znm (ξ)
+ 2 2 2 2 (R ) − (µ2 R2 − n2 )Z 2 (R )
πl n=0 m=1 s=1 (µnmR2 − n )Znm 2 nm 1 nm 1
√
l l λnms
where
(
1 for n = 0, a2 s2 π 2 a2 s2 π 2
An = βs = + b − 14 k2 , λnms = a2 µ2nm + + b − 14 k2 ,
2 for n 6= 0, l2 l2
dental equation
scribed:
Solution:
Z l Z ϕ0Z R
∂
∂t 0 0 0
Z l Z ϕ0Z R

0 0 0
Z t Z l Z ϕ0
2 ∂
0 0 0 ∂ξ ξ=R
Z tZ l Z R
2 1 ∂
0 0 0 ξ ∂η η=0
Z tZ l Z R
2 1 ∂
0 0 0 ξ ∂η η=ϕ0
Z t Z ϕ0Z R
2 ∂
0 0 0 ∂ζ ζ=0
Z t Z ϕ0Z R
2 ∂
0 0 0 ∂ζ ζ=l
Z t Z l Z ϕ0Z R
0 0 0 0
Here,
∞ ∞ ∞
8e−kt/2 X X X Jnπ/ϕ0 (µnm r)Jnπ/ϕ0 (µnm ξ) nπϕ
G(r, ϕ, z, ξ, η, ζ, t) = 2 ′ sin
R lϕ0 n=1 m=1 s=1 [Jnπ/ϕ 0
(µnm R)]2 ϕ0
p
nπη sπz sπζ sin t a2 µ2nm + a2 s2 π 2 l−2 + b − k2/4
× sin sin sin p ,
ϕ0 l l a2 µ2nm + a2 s2 π 2 l−2 + b − k2/4
∂2w
+ k ∂w
∂t
= a2 ∆3 w − bw + Φ(x, y, z, t) 759
scribed:
Solution:
Z l Z ϕ0Z R
∂
∂t 0 0 0
Z l Z ϕ0Z R

0 0 0
Z t Z l Z ϕ0
2 ∂
0 0 0 ∂ξ ξ=R
Z tZ l Z R
2 1 ∂
0 0 0 ξ ∂η η=0
Z tZ l Z R
2 1 ∂
0 0 0 ξ ∂η η=ϕ0
Z t Z ϕ0Z R
2
0 0 0
Z t Z ϕ0Z R
2
+a g5 (ξ, η, τ )G(r, ϕ, z, ξ, η, l, t − τ )ξ dξ dη dτ
0 0 0
Z t Z l Z ϕ0Z R
0 0 0 0
Here,
∞ ∞ ∞
4e−kt/2 X X X As Jnπ/ϕ0 (µnm r)Jnπ/ϕ0 (µnm ξ) nπϕ
G(r, ϕ, z, ξ, η, ζ, t) = ′ sin
R2 lϕ0 n=1 m=1 s=0 [Jnπ/ϕ 0
(µnm R)]2 ϕ0
p
nπη sπz sπζ sin t a2 µ2nm + a2 s2 π 2 l−2 + b − k2/4
× sin cos cos p ,
ϕ0 l l a2 µ2nm + a2 s2 π 2 l−2 + b − k2/4
where A0 = 1 and As = 2 for s ≥ 1; the Jnπ/ϕ0 (r) are Bessel functions; and the µnm are

A three-dimensional nonhomogeneous telegraph equation in the spherical coordinate sys-
tem is written as

∂2w ∂w 2 1 ∂ 2 ∂w 1 ∂ ∂w 1 ∂2w
+k =a r + 2 sin θ + 2 2
∂t2 ∂t r 2 ∂r ∂r r sin θ ∂θ ∂θ r sin θ ∂ϕ2
− bw + Φ(r, θ, ϕ, t).


Solution:
Z 2πZ πZ R
∂
∂t 0 0 0
Z 2πZ πZ R

+ f1(ξ, η, ζ)+kf0(ξ, η, ζ) G(r, θ, ϕ, ξ, η, ζ, t)ξ 2 sin η dξ dη dζ
0 0 0
Z tZ 2πZ π
2 2 ∂
−a R g(η, ζ, τ ) G(r, θ, ϕ, ξ, η, ζ, t−τ ) sin η dη dζ dτ
0 0 0 ∂ξ ξ=R
Z tZ 2πZ πZ R
0 0 0 0
where
∞ X ∞ X n
1 1
X
G(r, θ, ϕ, ξ, η, ζ, t) = √ exp − 2 kt As Bnms Jn+1/2 (λnm r)
2πR2 rξ n=0 m=1 s=0
p
s s sin t a2 λ2nm + b − k2/4
× Jn+1/2 (λnm ξ)Pn (cos θ)Pn (cos η) cos[s(ϕ − ζ)] p ,
a2 λ2nm + b − k2/4
(
1 for s = 0, (2n + 1)(n − s)!
As = Bnms = ′ 2 .
2 for s 6= 0, (n + s)! Jn+1/2 (λnm R)
Here, the Jn+1/2 (r) are Bessel functions, the Pns (µ) are associated Legendre functions
ds 1 dn
Pns (µ) = (1 − µ2 )s/2 Pn (µ), Pn (µ) = (µ2 − 1)n ,
dµs n! 2n dµn
∂2w
+ k ∂w
∂t
= a2 ∆3 w − bw + Φ(x, y, z, t) 761

Solution:
Z 2πZ πZ R
∂
∂t 0 0 0
Z 2πZ πZ R

0 0 0
Z tZ 2πZ π
+a2R2 g(η, ζ, τ )G(r, θ, ϕ, R, η, ζ, t−τ ) sin η dη dζ dτ
0 0 0
Z tZ 2πZ πZ R
0 0 0 0
where
√ ∞ ∞ n
3e−kt/2 sin t c e−kt/2 X X X
G(r, θ, ϕ, ξ, η, ζ, t) = √ + √ AsBnmsJn+1/2(λnmr)
4πR3 c 2π rξ n=0 m=1 s=0
p
s s sin t a2λ2nm +c
×Jn+1/2(λnmξ)Pn (cos θ)Pn (cos η) cos[s(ϕ−ζ)] p ,
a2λ2nm +c
(
1 if s = 0, λ2nm(2n+1)(n−s)! 1 2
As = Bnms = 2 , c = b− 4 k .
2 if s 6= 0, (n+s)! R2λ2nm −n(n+1) Jn+1/2(λnmR)
the paragraph above), and the λnm are positive roots of the transcendental equation
′
2λRJn+1/2 (λR) − Jn+1/2 (λR) = 0.

∂r w + sw = g(θ, ϕ, t) at r = R (boundary condition).

∞ ∞ n
e−kt/2 X X X
G(r, θ, ϕ, ξ, η, ζ, t) = √ Al BnmlJn+1/2(λnmr)Jn+1/2(λnmξ)
2π rξ n=0 m=1
l=0
p
l l sin t a2λ2nm +b−k2/4
×Pn(cos θ)Pn(cos η) cos[l(ϕ−ζ)] p ,
a2λ2nm +b−k2/4
(
1 if l = 0, λ2nm(2n+1)(n−l)!
Al = Bnml = 2 .
2 if l 6= 0, (n+l)! R2λ2nm +(sR+n)(sR−n−1) Jn+1/2(λnmR)
Here, the Jn+1/2 (r) are Bessel functions, the Pnl (µ) are associated Legendre functions (see
′

λRJn+1/2 (λR) + sR − 12 Jn+1/2 (λR) = 0.

Solution:
Z 2πZ πZ R2
∂
∂t 0 0 R1
Z 2πZ πZ R2

0 0 R1
Z tZ 2πZ π
2 ∂
0 0 0 ∂ξ ξ=R1
Z tZ 2πZ π
∂
−a2R22 g2(η, ζ, τ ) G(r, θ, ϕ, ξ, η, ζ, t−τ ) sin η dη dζ dτ
0 0 0 ∂ξ ξ=R2
Z tZ 2πZ πZ R2
0 0 0 R1
where
∞ ∞ n
πe−kt/2 X X X
G(r, θ, ϕ, ξ, η, ζ, t) = √ As Bnms Zn+1/2 (λnm r)Zn+1/2 (λnm ξ)
8 rξ n=0 m=1 s=0
p
sin t a2 λ2 + b − k 2/4
nm
× Pns (cos θ)Pns (cos η) cos[s(ϕ − ζ)] p ,
a2 λ2nm + b − k2/4
( 2
1 for s = 0, λnm (2n + 1)(n − s)! Jn+1/2 (λnm R2 )
As = Bnms = 2 2
.
2 for s 6= 0, (n + s)! Jn+1/2 (λnm R1 ) − Jn+1/2 (λnm R2 )
∂2w
+ k ∂w
∂t
= a2 ∆3 w − bw + Φ(x, y, z, t) 763
Here, the Jn+1/2 (r) are Bessel functions, the Pns (µ) are associated Legendre functions
ds 1 dn
Pns (µ) = (1 − µ2 )s/2 Pn (µ), Pn (µ) = (µ2 − 1)n ,
dµs n! 2n dµn

problem.
Solution:
Z 2πZ πZ R2
∂
∂t 0 0 R1
Z 2πZ πZ R2

0 0 R1
Z tZ 2πZ π
−a2R12 g1(η, ζ, τ )G(r, θ, ϕ, R1 , η, ζ, t−τ ) sin η dη dζ dτ
0 0 0
Z tZ 2πZ π
+a2R22 g2(η, ζ, τ )G(r, θ, ϕ, R2 , η, ζ, t−τ ) sin η dη dζ dτ
0 0 0
Z tZ 2πZ π Z R2
0 0 0 R1
where
√ ∞ ∞ n
3e−kt/2 sin t c e−kt/2 X X X As
G(r, θ, ϕ, ξ, η, ζ, t) = √ + √ Zn+1/2 (λnm r)
4π(R23 − R13 ) c 4π rξ n=0 m=1 s=0 Bnms
p
s s sin t a2 λ2nm + c
× Zn+1/2 (λnm ξ)Pn (cos θ)Pn (cos η) cos[s(ϕ − ζ)] p .
a2 λ2nm + c
Here,
( Z R2
1 for s = 0, (n+s)! 2
As = Bnms = rZn+1/2 (λnmr) dr, c = b− 14 k2,
2 for s =
6 0, (2n+1)(n−s)! R1

′1
Zn+1/2(λnmr) = λnmJn+1/2
J (λnmR1)−
(λnmR1) Yn+1/2(λnmr)
2R1 n+1/2

′ 1
− λnmYn+1/2(λnmR1)− Y (λnmR1) Jn+1/2(λnmr),
2R1 n+1/2
dre functions (see the paragraph above), and the λnm are positive roots of the transcendental
equation
′ 1
λZn+1/2 (λR2 ) − Z (λR2 ) = 0.
2R2 n+1/2

∂r w − s1 w = g1 (θ, ϕ, t) at r = R1 (boundary condition),
∂r w + s2 w = g2 (θ, ϕ, t) at r = R2 (boundary condition).
∞ ∞ n
e−kt/2 X X X Al
4π rξ n=0 m=1 l=0 Bnml
p
l l sin t a2 λ2nm + c
× Pn (cos θ)Pn (cos η) cos[l(ϕ − ζ)] p .
a2 λ2nm + c
Here,
( Z R2
1 for l = 0, (n + l)! 2
Al = Bnml = rZn+1/2 (λnm r) dr, c = b− 14 k2 ,
2 for l 6= 0, (2n + 1)(n − l)! R1

′ 1
Zn+1/2 (λr) = λJn+1/2 (λR1 ) − s1 + Jn+1/2 (λR1 ) Yn+1/2 (λr)
2R1

′ 1
− λYn+1/2 (λR1 ) − s1 + Yn+1/2 (λR1 ) Jn+1/2 (λr),
2R1
where the Jn+1/2 (r) and Yn+1/2 (r) are Bessel functions, the Pnl (µ) are associated Legen-

′ 1
λZn+1/2 (λR2 ) + s2 − Zn+1/2 (λR2 ) = 0.
2R2


∂2w ∂ ∂w ∂ ∂w ∂ ∂w
1. = axn + by m + cz k .
∂t2 ∂x ∂x ∂y ∂y ∂z ∂z
This equation admits separable solutions. In addition, for n 6= 2, m 6= 2, and k 6= 2, there
are particular solutions of the form

2 x2−n y 2−m z 2−k
w = w(ξ, t), ξ =4 + + ,
a(2 − n)2 b(2 − m)2 c(2 − k)2
where w(ξ, t) is determined by the one-dimensional nonstationary equation

∂2w ∂2w A ∂w 1 1 1
= + , A = 2 + + − 1.
∂t2 ∂ξ 2 ξ ∂ξ 2−n 2−m 2−k

∂2w ∂w ∂2w ∂2w ∂ 2w ∂w ∂w ∂w
2. +k = a2 + + + b1 + b2 + b3 + cw.
∂t2 ∂t ∂x2 ∂y 2 ∂z 2 ∂x ∂y ∂z
The transformation

1 b1 x + b2 y + b3 z
w(x, y, z, t) = u(x, y, z, τ ) exp − kt − , τ = at
2 2a2
leads to the equation in Section 8.3.1:
∂2u ∂2u ∂2u ∂2u c k2 1 2 2 2

= + + + βu, β= + − b1 + b2 + b3 .
∂τ 2 ∂x2 ∂y 2 ∂z 2 a2 4a2 4a4
8.5.2 Equation of the Form

∂2w
ρ(x, y, z) 2 = div a(x, y, z)∇w − q(x, y, z)w + Φ(x, y, z, t)
∂t
Such equations are encountered when studying vibration of finite volumes. The equation is
written using the notation

∂ ∂w ∂ ∂w ∂ ∂w
div a(r)∇w = a(r) + a(r) + a(r) , r = {x, y, z}.
∂x ∂x ∂y ∂y ∂z ∂z
The problems for the equation in question are considered below for the interior of a
bounded domain V with smooth surface S. In what follows, it is assumed that ρ(r) > 0,
a(r) > 0, and q(r) ≥ 0.

The solution of the equation in question with the initial conditions
w = f0 (r) at t = 0,
(1)
∂t w = f1 (r) at t = 0
and the nonhomogeneous boundary conditions of the first kind

w = g(r, t) for r∈S (2)
can be written as the sum
Z Z
∂
w(r, t) = f0 (ξ)ρ(ξ)G(r, ξ, t) dVξ + f1 (ξ)ρ(ξ)G(r, ξ, t) dVξ
∂t V V
Z tZ
∂
− g(ξ, τ )a(ξ) G(r, ξ, t − τ ) dSξ dτ
0 S ∂Nξ
Z tZ
+ Φ(ξ, τ )G(r, ξ, t − τ ) dVξ dτ. (3)
0 V
Here, the modified Green’s function is expressed as
∞
X p
1
G(r, ξ, t) = √ un (r)un (ξ) sin( λn t),
n=1
λn kun k2
Z (4)
kun k2 = ρ(r)u2n (r) dV, ξ = {ξ1 , ξ2 , ξ3 },
V
Liouville problem for the second-order elliptic equation with homogeneous boundary con-
ditions of the first kind

div a(r)∇u − q(r)u + λρ(r)u = 0, (5)
u=0 for r ∈ S. (6)
∂
The integration in solution (3) is performed with respect to ξ1 , ξ2 , and ξ3 ; ∂N ξ
is the
derivative along the outward normal to the surface S with respect to ξ1 , ξ2 , and ξ3 .
1◦ . There are finitely many eigenvalues. All eigenvalues are real and can be ordered so that
λ1 ≤ λ2 ≤ λ3 ≤ · · · and λn → ∞ as n → ∞; therefore the number of negative eigenvalues
is finite.
2◦ . If ρ(r) > 0, a(r) > 0, and q(r) ≥ 0, then all eigenvalues are positive, λn > 0.
3◦ . An eigenfunction is determined up to a constant multiplier. Two eigenfunctions un (r)
and um (r) corresponding to different eigenvalues λn and λm are orthogonal with weight
ρ(r) in the domain V , that is,
Z
ρ(r)un (r)um (r) dV = 0 for n 6= m.
V
4◦ .An arbitrary function F (r) twice continuously differentiable and satisfying the bound-
ary condition of the Sturm–Liouville problem (F = 0 for r ∈ S) can be expanded into an
absolutely and uniformly convergent series in the eigenfunctions:
X∞ Z
1
F (r) = Fn un (r), Fn = F (r)ρ(r)un (r) dV,
n=1
kun k2 V
where kun k2 is defined in (4).

Remark 8.1. In a three-dimensional problem, finitely many linearly independent eigenfunctions

(1) (m)
un , . . . , un generally correspond to each eigenvalue λn . These functions can always be replaced
by their linear combinations
ū(k) (1) (k−1)

n = Ak,1 un + · · · + Ak,k−1 un + u(k)
n , k = 1, . . . , m,
(1) (m)
so that ūn , . . . , ūn are now orthogonal pairwise. For this reason, without loss of generality, all
eigenfunctions can be assumed orthogonal.

The solution of the equation with the initial conditions (1) and nonhomogeneous boundary
conditions of the second kind,
∂w
= g(r, t) for r ∈ S,
∂N
Z Z
∂
w(r, t) = f0 (ξ)ρ(ξ)G(r, ξ, t) dVξ + f1 (ξ)ρ(ξ)G(r, ξ, t) dVξ
∂t V V
Z tZ
+ g(ξ, τ )a(ξ)G(r, ξ, t − τ ) dSξ dτ (7)
0 S
Z tZ
+ Φ(ξ, τ )G(r, ξ, t − τ ) dVξ dτ.
0 V
Here, the modified Green’s function G is given by relation (4), the λn and un (r) are
the eigenvalues and corresponding eigenfunctions of the Sturm–Liouville problem for the
second-order elliptic equation (5) with homogeneous boundary conditions of the second
kind,
∂u
= 0 for r ∈ S. (8)
∂N
For q(r) > 0, the general properties of the eigenvalue problem (5), (8) are the same as
those of the first boundary value problem (all λn are positive).

The solution of the equation with the initial conditions (1) and nonhomogeneous boundary
conditions of the third kind,
∂w
+ k(r)w = g(r, t) for r ∈ S,
∂N
is determined by relations (7) and (4), where the λn and un (r) are the eigenvalues and
eigenfunctions of the Sturm–Liouville problem for the second-order elliptic equation (5)
with homogeneous boundary conditions of the third kind,
∂u
+ k(r)u = 0 for r ∈ S. (9)
∂N
If q(r) ≥ 0 and k(r) > 0, the general properties of the eigenvalue problem (5), (9)
are the same as those of the first boundary value problem (see the first paragraph of this
section).
Suppose k(r) = k = const. Denote the Green’s functions of the second and third
boundary value problems by G2 (r, ξ, t) and G3 (r, ξ, t, k), respectively. If q(r) > 0, the
limit relation G2 (r, ξ, t) = lim G3 (r, ξ, t, k) holds.
k→0
⊙ Literature for Section 8.5.2: V. S. Vladimirov (1988), A. D. Polyanin (2000a).

Throughout this section the following notation is used:
Pn ∂2w q
∆n w = 2 , x = {x1 , . . . , xn }, y = {y1 , . . . , yn }, |x| = x21 + · · · + x2n .
k=1 ∂xk
∂2w
8.6.1 Wave Equation = a2 ∆n w
∂t2
◮ Fundamental solution:
 n−2

 (−1) 2 n−1 ϑ(at − |x|)


 n+1 Γ 2 n−1 if n ≥ 2 is even;
E (x, t) = 2aπ 2 (a2 t2 − |x|2 ) 2

 n−3

 1 1 ∂ 2

2
δ(a2 t2 − |x|2 ) if n ≥ 3 is odd;
2πa 2πa t ∂t
where ϑ(z) is the Heaviside unit step function and δ(z) is the Dirac delta function.
◮ Properties of solutions.
Suppose w(x1 , . . . , xn , t) is a solution of the wave equation. Then the functions
w1 = Aw(±λx1 + C1 , . . . , ±λxn + Cn , ±λt + Cn+1 ),

x1 − vt t − va−2 x1
w2 = Aw p , x2 , . . . , xn , p ,
1 − (v/a)2 1 − (v/a)2

2
n−1
2 2 − 2 x1 xn t
w3 = A|r − a t | w 2 , ..., 2 , , r = |x|,
r − a2 t2 r − a2 t2 r 2 − a2 t2
are also solutions of this equation everywhere they are defined; A, C1 , . . . , Cn+1 , v, and λ
are arbitrary constants. The signs at λ in the expression of w1 can be taken independently
of one another.
◮ Domain: −∞ < xk < ∞; k = 1, . . . , n. Cauchy problem.
w = f (x) at t = 0,
∂t w = g(x) at t = 0.
Solution:
Z at
1 ∂ n−1 2 2 2
n−3
2 rT [f (x)] dr
w(x, t) = n−1 a t − r r
a (n − 2)! ∂tn−1 0
Z at
1 ∂ n−2 2 2 2
n−3
2 rTr [g(x)] dr.
+ n−1 a t − r
a (n − 2)! ∂tn−2 0
Here, Tr [f (x)] is the average of f over the surface of the sphere of radius r with center at x:
Z
1 2π n/2
Tr [f (x)] ≡ f (y) dSy , σn = ,
σn r n−1 Γ(n/2)
|x−y|=r
where σn r n−1 is the area of the surface of an n-dimensional sphere of radius r, dSy is the
area element of this surface, and |x − y|2 = (x1 − y1 )2 + · · · + (xn − yn )2 .
For odd n, the solution can be alternatively represented as
n−3
1 ∂ 1 ∂ 2
w(x, t) = tn−2 Tat [f (x)]
1 × 3 . . . (n − 2) ∂t t ∂t
n−3
1 1 ∂ 2
+ tn−2 Tat [g(x)] .
1 × 3 . . . (n − 2) t ∂t
For even n, the solution can be alternatively represented as
n−2 Z at
1 ∂ 1 ∂ 2 r n−1 dr
w(x, t) = Tr [f (x)] √
2 × 4 . . . (n − 2)an−1 ∂t t ∂t 0 a2 t2 − r 2
n−2 Z at
1 1 ∂ 2 r n−1 dr
+ T r [g(x)] √ .
2 × 4 . . . (n − 2)an−1 t ∂t 0 a2 t2 − r 2
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), R. Courant and D. Hilbert (1989),
D. Zwillinger (1998).
◮ Domain: 0 ≤ xk ≤ lk ; k = 1, . . . , n. Boundary value problems.
For solutions of the first, second, third, and mixed boundary value problems with nonho-
mogeneous conditions of general form, see Section 8.6.2 for Φ ≡ 0.
8.6.2 Nonhomogeneous Wave Equation

∂2w
= a2 ∆n w + Φ(x1 , . . . , xn , t)
∂t2
w = f (x) at t = 0,
∂t w = g(x) at t = 0.
Solution:
Z at
1 ∂ n−1 2 2 2
n−3
2 rT [f (x)] dr
w(x, t) = n−1 a t − r r
a (n − 2)! ∂tn−1 0
Z at
1 ∂ n−2 n−3
+ n−1 n−2
a2 t2 − r 2 2 rTr [g(x)] dr
a (n − 2)! ∂t 0
Z at Z aτ
1 ∂ n−2 2 2 2
n−3
2 rTr [Φ(x, t − τ )] dr.
+ n−1 dτ a τ − r
a (n − 2)! ∂tn−2 0 0
Here, Tr [f (x)] is the average of f over the spherical surface of radius r with center at x:
Z
1 2π n/2
Tr [f (x)] ≡ f (y) dS y , σ n = ,
σn r n−1 Γ(n/2)
|x−y|=r
where σn r n−1 is the area of the surface of an n-dimensional sphere of radius r and dSy is
the area element of this surface.
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), R. Courant and D. Hilbert (1989).

Solution:
Z tZ
w(x, t) = Φ(y, τ )G(x, y, t − τ ) dVy dτ
0
ZV Z
∂
+ f0 (y)G(x, y, t) dVy + f1 (y)G(x, y, t) dVy
∂t V V
n Z tZ
X
2 ∂
0 S (k) ∂yk yk =0
k=1
Xn Z t Z
∂
− a2 hk (y, τ ) G(x, y, t − τ ) dSy(k) dτ,
k=1
where
x={x1, . . . , xn}, y={y1, . . . , yn}, dVy =dy1 dy2 . . . dyn,
dSy(k) =dy1 . . . dyk−1 dyk+1 . . . dyn,
S (k) ={0≤ym ≤lm for m=1, . . . , k−1, k+1, . . . , n}.
Green’s function:
X∞ X ∞ X∞
2n
G(x, y, t) = ... sin(λs1 x1 ) sin(λs2 x2 ) . . . sin(λsn xn )
l1 l2 . . . ln s =1 s =1
1 2 s n =1 p
sin at λ2s1 + · · · + λ2sn
× sin(λs1 y1 ) sin(λs2 y2 ) . . . sin(λsn yn ) p ,
a λ2s1 + · · · + λ2sn
where
s1 π s2 π sn π
λs1 = , λs2 = , ..., λsn = .
l1 l2 ln

Solution:
Z tZ
Z0 V Z
V V
n Z tZ
X
− a2 gk (y, τ )G(x, y, t − τ ) y dSy(k) dτ
k =0
k=1 0 S (k)
n
X tZ Z

+ a2 hk (y, τ )G(x, y, t − τ ) y dSy(k) dτ.
k =lk
k=1 0 S (k)
Here,
X∞ X ∞ X∞ p
t 1 As As . . . Asn
G(x, y, t) = + ... p1 2 sin at λ2s1 +· · ·+λ2sn
l1l2 . . . ln l1l2 . . . ln s =0 s =0 s =0 a λs +· · ·+λs
2 2
1 2 n 1 n

×cos(λs1 x1) cos(λs2 x2) . . . cos(λsnxn) cos(λs1 y1) cos(λs2 y2) . . . cos(λsn yn) ,
where
(
s1 π s2 π sn π 1 if sm = 0,
λs1 = , λs2 = , . . . , λsn = ; Asm = m = 1, 2, . . . , n.
l1 l2 ln 2 if sm 6= 0,
The summation is performed over the indices satisfying the condition s1 + · · · + sn > 0;
the term corresponding to s1 = · · · = sn = 0 is singled out.
◮ Domain: V = {0 ≤ xk ≤ lk ; k = 1, . . . , n}. Third boundary value problem.

∂xk w − bk w = gk (x, t) at xk = 0 (boundary conditions),
∂xk w + ck w = hk (x, t) at xk = lk (boundary conditions).
∞ X ∞ ∞
p
X X sin at λ2s1 + λ2s2 + · · · + λ2sn
n
G(x, y, t) = 2 ... p
s1 =1 s2 =1 sn =1 aEs 1E s 2 . . . Es n λ2s1 + λ2s2 + · · · + λ2sn
× sin(λs1 x1 + ϕs1 ) sin(λs2 x2 + ϕs2 ) . . . sin(λsn xn + ϕsn )
× sin(λs1 y1 + ϕs1 ) sin(λs2 y2 + ϕs2 ) . . . sin(λsn yn + ϕsn ).
Here,
λsm (bm cm + λ2sm )(bm + cm )
ϕsm = arctan , Esm = lm + , m = 1, 2, . . . , n;
lm (b2m + λ2sm )(c2m + λ2sm )
the λsm are positive roots of the transcendental equations

1 bm cm
λ− = cot(lm λ), m = 1, 2, . . . , n.
bm + cm λ
◮ Domain: V = {0 ≤ xk ≤ lk ; k = 1, . . . , n}. Mixed boundary value problem.

Solution:
Z tZ
0
ZV Z
∂
∂t V V
n Z tZ
X
2 ∂
0 S (k) ∂yk yk =0
k=1
Xn Z t Z
+ a2 hk (y, τ )G(x, y, t − τ ) dSy(k) dτ,
k=1 0 S (k) yk =lk
where
X∞ X ∞ ∞
X
2n
l1 l2 . . . ln
s1 =1 s2 =1 sn =1
p
sin at λ2s1 + · · · + λ2sn
a λ2s1 + · · · + λ2sn
π(2s1 + 1) π(2s2 + 1) π(2sn + 1)
λs1 = , λs2 = , . . . , λsn = .
2l1 2l2 2ln
∂2w
8.6.3 Equations of the Form = a2 ∆n w − bw + Φ(x1 , . . . , xn , t)
∂t2
w = f (x) at t = 0,
∂t w = g(x) at t = 0,
where x = {x1 , . . . , xn }.
1◦ . Let b = −c2 < 0 and Φ ≡ 0. The solution is sought by the descent method in the form
1
w(x, t) = u(x, xn+1 , t), (1)
exp(cxn+1 )
where u is the solution of the Cauchy problem for the auxiliary (n + 1)-dimensional wave
equation
∂2u
= ∆n+1 u (2)
∂t2
with the initial conditions
u = exp(cxn+1 )f (x) at t = 0,
(3)
∂t u = exp(cxn+1 )g(x) at t = 0.
For the solution of the Cauchy problem (2), (3), see Section 8.6.1.
2◦ . Let b = c2 > 0 and Φ ≡ 0. In this case the function exp(cxn+1 ) in (1) and (3) must be
replaced by cos(cxn+1 ).

Solution:
Z tZ
0
ZV Z
∂
∂t V V
n Z tZ
X
2 ∂
0 S (k) ∂yk y =0
k=1 k
Xn Z tZ
∂
0 S (k) ∂yk yk =lk
k=1
where
x = {x1 , . . . , xn }, y = {y1 , . . . , yn }, dVy = dy1 dy2 . . . dyn ,
S (k) = {0 ≤ ym ≤ lm for m = 1, . . . , k − 1, k + 1, . . . , n}.
Green’s function:
X∞ X ∞ ∞
X
2n
l1 l2 . . . ln
s1 =1 s2 =1 sn =1
p
sin t a2 (λ2s1 + · · · + λ2sn ) + b
a2 (λ2s1 + · · · + λ2sn ) + b
where
s1 π s2 π sn π
λs1 = , λs2 = , ..., λsn = .
l1 l2 ln

Solution:
Z tZ
Z0 V
Z
V V
n Z tZ
X
k =0
k=1 0 S (k)
Xn Z t Z

k =lk
k=1 0 S (k)
Here,
X∞ X ∞ ∞
X
1
G(x, y, t) = ... As1 As2 . . . Asn cos(λs1 x1 ) cos(λs2 x2 ) . . . cos(λsn xn )
l1 l2 . . . ln s =0 s =0 sn =0
1 2
p
sin t a2 (λ2s1 + · · · + λ2sn ) + b
× cos(λs1 y1 ) cos(λs2 y2 ) . . . cos(λsn yn ) p ,
a2 (λ2s1 + · · · + λ2sn ) + b
where
(
s1 π s2 π sn π 1 for sm = 0,
λs1 = , λs2 = , ..., λsn = ; Asm = m = 1, 2, . . . , n.
l1 l2 ln 2 for sm 6= 0,
◮ Domain: V = {0 ≤ xk ≤ lk ; k = 1, . . . , n}. Third boundary value problem.


∞ X ∞ ∞
p
X X sin t a2 (λ2s1 + λ2s2 + · · · + λ2sn ) + b
n
G(x, y, t) = 2 ... p
s1 =1 s2 =1 sn =1 Es1Es2 . . . Esn a2 (λ2s1 + λ2s2 + · · · + λ2sn ) + b
× sin(λs1 x1 + ϕs1 ) sin(λs2 x2 + ϕs2 ) . . . sin(λsn xn + ϕsn )
× sin(λs1 y1 + ϕs1 ) sin(λs2 y2 + ϕs2 ) . . . sin(λsn yn + ϕsn ).
Here,
the λsm are positive roots of the transcendental equations

1 bm cm
λ− = cot(lm λ), m = 1, 2, . . . , n.
bm + cm λ
◮ Domain: V = {0 ≤ xk ≤ lk ; k = 1, . . . , n}. Mixed boundary value problem.


Solution:
Z tZ
0
ZV Z
∂
∂t V V
n Z tZ
X
2 ∂
0 S (k) ∂yk y =0
k=1 k
Xn Z tZ
+ a2 hk (y, τ )G(x, y, t − τ ) dSy(k) dτ.
k=1 0 S (k) yk =lk
Here,
X∞ X ∞ ∞
X
2n
l1 l2 . . . ln
s1 =1 s2 =1 sn =1
p
sin t a2 (λ2s1 + · · · + λ2sn ) + b
a2 (λ2s1 + · · · + λ2sn ) + b
where
π(2s1 + 1) π(2s2 + 1) π(2sn + 1)

λs1 = , λs2 = , ..., λsn = .
2l1 2l2 2ln
8.6.4 Equations Containing the First Time Derivative
∂2w ∂w
1. +β = a2 ∆n w − bw + Φ(x1, . . . , xn , t).
∂t2 ∂t
Nonhomogeneous telegraph equation with n space variables.

1◦ . The substitution w = exp − 12 βt u leads to the equation
∂2u
2
= a2 ∆n u − b − 14 β 2 u + exp 1
2 βt Φ(x1 , . . . , xn , t),
∂t
2◦ . Domain: V = {0 ≤ xk ≤ lk ; k = 1, . . . , n}. First boundary value problem.


Solution:
Z tZ
0 V
Z Z
∂
+ f0 (y)G(x, y, t) dVy + f1 (y) + βf0 (y) G(x, y, t) dVy
∂t V V
Xn Z tZ
2 ∂
(k) ∂y k
k=1 0 S yk =0
Xn Z t Z
∂
k=1
where
x = {x1 , . . . , xn }, y = {y1 , . . . , yn }, dVy = dy1 dy2 . . . dyn ,
S (k) = {0 ≤ ym ≤ lm for m = 1, . . . , k − 1, k + 1, . . . , n}.
Green’s function:
∞ ∞ ∞
2n e−βt/2 X X X
l1 l2 . . . ln
s1 =1 s2 =1 sn =1
p
sin t a2 (λ2s1 + · · · + λ2sn ) + b − β 2/4
a2 (λ2s1 + · · · + λ2sn ) + b − β 2/4
where
s1 π s2 π sn π
λs1 = , λs2 = , . . . , λsn = .
l1 l2 ln
3◦ . Domain: V = {0 ≤ xk ≤ lk ; k = 1, . . . , n}. Second boundary value problem.
Solution:
Z tZ
Z0 V
Z

V V
n Z tZ
X
k =0
k=1 0 S (k)
Xn Z tZ

k =lk
k=1 0 S (k)
Here,
X
∞ X
∞ ∞
X
e−βt/2
G(x, y, t) = ... As1 As2 . . . Asn cos(λs1 x1 ) cos(λs2 x2 ) . . . cos(λsn xn )
l1 l2 . . . ln s1 =0 s2 =0 sn =0
p
sin t a2 (λ2s1 + · · · + λ2sn ) + b − β 2/4
× cos(λs1 y1 ) cos(λs2 y2 ) . . . cos(λsn yn ) p ,
a2 (λ2s1 + · · · + λ2sn ) + b − β 2/4
where
(
s1 π s2 π sn π 1 for sm = 0,
λs1 = , λs2 = , ..., λsn = ; Asm = m = 1, 2, . . . , n.
l1 l2 ln 2 for sm 6= 0,
4◦ . Domain: V = {0 ≤ xk ≤ lk ; k = 1, . . . , n}. Third boundary value problem.


∞ X
∞ ∞
p
X X sin t a2(λ2s1+λ2s2+· · ·+λ2sn)+b−β 2/4
n −βt/2
G(x, y, t)=2 e ... p
s1=1 s2=1 sn=1 Es E
1 s 2 . . . Es n a2(λ2s1+λ2s2+· · ·+λ2sn)+b−β 2/4
×sin(λs1x1+ϕs1) sin(λs2x2+ϕs2) . . . sin(λsnxn+ϕsn)
×sin(λs1y1+ϕs1) sin(λs2y2+ϕs2) . . . sin(λsnyn+ϕsn).
Here,

the λsm are positive roots of the transcendental equation

1 bm cm
λ− = cot(lm λ), m = 1, 2, . . . , n.
bm + cm λ
5◦ . Domain: V = {0 ≤ xk ≤ lk ; k = 1, . . . , n}. Mixed boundary value problem.


Solution:
Z tZ
0 V
Z Z
∂
∂t V V
Xn Z tZ h i
∂
+ a2 gk (y, τ ) G(x, y, t − τ ) dSy(k) dτ
(k) ∂y k y =0
k=1 0 S k
n
X tZ Z h i
+ a2 hk (y, τ )G(x, y, t − τ ) dSy(k) dτ,
S (k) yk =lk
k=1 0
where
∞ ∞ ∞
2n e−βt/2 X X X
l1 l2 . . . ln s =1 s =1
1 2 sn =1 p
sin t a2 (λ2s1 + · · · + λ2sn ) + b − β 2/4
a2 (λ2s1 + · · · + λ2sn ) + b − β 2/4
π(2s1 + 1) π(2s2 + 1) π(2sn + 1)
λs1 = , λs2 = , . . . , λsn = .
2l1 2l2 2ln
n
X
∂2w ∂w ∂w
2. +β = a2 ∆n w + bk + cw.
∂t2 ∂t ∂xk
k=1
The transformation
n
1 1 X
w(x1 , . . . , xn , t) = u(x1 , . . . , xn , τ ) exp − βt − 2 bk xk , τ = at
2 2a
k=1
leads to the equation

n
∂2u c β2 1 X 2
= ∆n u + λu, λ= + − bk ,
∂τ 2 a2 4a2 4a4
k=1

∂2w n − 1 ∂w
3. + = ∆n w.
∂t2 t ∂t
Darboux equation. Cauchy problem.
w = f (x) at t = 0,
∂t w = 0 at t = 0.
Solution:
Z
1 2π n/2
w(x, t) = f (y) dSy , σn = ,
σn tn−1 Γ(n/2)
|x−y|=t
where dSy is the area element of the surface of an n-dimensional sphere of radius t (i.e.,
the solution w is the average of the function f over the sphere for radius t with center at x).
Chapter 9
Second-Order
Elliptic Equations
9.1 Laplace Equation ∆2 w = 0

The Laplace equation is often encountered in heat and mass transfer theory, fluid mechan-
ics, elasticity, electrostatics, and other areas of mechanics and physics. For example, in
heat and mass transfer theory, this equation describes steady-state temperature distribution
in the absence of heat sources and sinks in the domain under study.
A regular solution of the Laplace equation is called a harmonic function. The first
boundary value problem for the Laplace equation is often referred to as the Dirichlet prob-
lem, and the second boundary value problem as the Neumann problem.
Extremum principle: Given a domain D, a harmonic function w in D that is not iden-
tically constant in D cannot attain its maximum or minimum value at any interior point
of D.
9.1.1 Problems in Cartesian Coordinate System

The Laplace equation with two space variables in the rectangular Cartesian system of co-
ordinates is written as
∂2w ∂2w
+ = 0.
∂x2 ∂y 2
◮ Particular solutions and a method for their construction.

w(x, y) = Ax + By + C,
w(x, y) = A(x2 − y 2 ) + Bxy,
w(x, y) = A(x3 − 3xy 2 ) + B(3x2 y − y 3 ),
Ax + By
w(x, y) = 2 + C,
x + y2
781
782 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
w(x, y) = exp(±µx)(A cos µy + B sin µy),

w(x, y) = (A cos µx + B sin µx) exp(±µy),
w(x, y) = (A sinh µx + B cosh µx)(C cos µy + D sin µy),
w(x, y) = (A cos µx + B sin µx)(C sinh µy + D cosh µy),

w(x, y) = A ln (x − x0 )2 + (y − y0 )2 + B,
where A, B, C, D, x0 , y0 , and µ are arbitrary constants.

1 1 p
E (x, y) = ln , r= x2 + y 2 .
2π r
3◦ . If w(x, y) is a solution of the Laplace equation, then the functions
w1 = Aw(±λx + C1 , ±λy + C2 ),
w2 = Aw(x cos β + y sin β, −x sin β + y cos β),

x y
w3 = Aw , ,
x2 + y 2 x2 + y 2
are also solutions everywhere they are defined; A, C1 , C2 , β, and λ are arbitrary constants.
The signs at λ in w1 are taken independently of each other.
4◦ . A fairly general method for constructing particular solutions involves the following.
Let f (z) = u(x, y) + iv(x, y) be any analytic function of the complex variable z = x + iy
(u and v are real functions of the real variables x and y; i2 = −1). Then the real and
imaginary parts of f both satisfy the two-dimensional Laplace equation,
∆2 u = 0, ∆2 v = 0.
Recall that the Cauchy–Riemann conditions
∂u ∂v ∂u ∂v
= , =− (1)
∂x ∂y ∂y ∂x
are necessary and sufficient conditions for the function f to be analytic. Thus, by specify-
ing analytic functions f (z) and taking their real and imaginary parts, one obtains various
solutions of the two-dimensional Laplace equation.
⊙ Literature: M. A. Lavrent’ev and B. V. Shabat (1973), A. G. Sveshnikov and A. N. Tikhonov (1974),
A. V. Bitsadze and D. F. Kalinichenko (1985).
◮ Specific features of stating boundary value problems for the Laplace equation.
1◦ . For outer boundary value problems on the plane, it is (usually) required to set the
additional condition that the solution of the Laplace equation must be bounded at infinity.
2◦ . The solution of the second boundary value problem is determined up to an arbitrary
additive term.
9.1. Laplace Equation ∆2 w = 0 783
3◦ . Let the second boundary value problem in a closed bounded domain D with piecewise
smooth boundary Σ be characterized by the boundary condition∗
∂w
= f (r) for r ∈ Σ,
∂N
∂w
where ∂N is the derivative along the (outward) normal to Σ. The necessary and sufficient
condition of solvability of the problem has the form
Z
f (r) dΣ = 0. (2)
Σ
Remark 9.1. The same solvability condition occurs for the outer second boundary value prob-
lem if the domain is infinite but has a finite boundary.
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964).
◮ Domain: −∞ < x < ∞, 0 ≤ y < ∞. First boundary value problem.
A half-plane is considered. A boundary condition is prescribed:
w = f (x) at y = 0.
Solution:
Z ∞ Z π/2
1 yf (ξ) dξ 1
w(x, y) = 2 2
= f (x + y tan θ) dθ.
π −∞ (x − ξ) + y π −π/2
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), H. S. Carslaw and J. C. Jaeger

(1984).
◮ Domain: −∞ < x < ∞, 0 ≤ y < ∞. Second boundary value problem.
∂y w = f (x) at y = 0.
Solution: Z ∞ p
1
w(x, y) = f (ξ) ln (x − ξ)2 + y 2 dξ + C,
π −∞
where C is an arbitrary constant.

∗
More rigorously, Σ must satisfy the Lyapunov condition [see Babich, Kapilevich, Mikhlin et al. (1964)
and Tikhonov and Samarskii (1990)].
A quadrant of the plane is considered. Boundary conditions are prescribed:
w = f1 (y) at x = 0, w = f2 (x) at y = 0.
Solution:
Z ∞
4 f1 (η)η dη
w(x, y) = xy
π [x + (y − η)2 ][x2 + (y + η)2 ]
2
Z0 ∞
4 f2 (ξ)ξ dξ
+ xy 2 + y 2 ][(x + ξ)2 + y 2 ]
.
π 0 [(x − ξ)
◮ Domain: −∞ < x < ∞, 0 ≤ y ≤ a. First boundary value problem.
An infinite strip is considered. Boundary conditions are prescribed:
w = f1 (x) at y = 0, w = f2 (x) at y = a.
Solution:
Z ∞
1 πy f1 (ξ) dξ
w(x, y) = sin
2a a cosh[π(x − ξ)/a] − cos(πy/a)
Z−∞
∞
1 πy f2 (ξ) dξ
+ sin .
2a a −∞ cosh[π(x − ξ)/a] + cos(πy/a)
◮ Domain: −∞ < x < ∞, 0 ≤ y ≤ a. Second boundary value problem.
∂y w = f1 (x) at y = 0, ∂y w = f2 (x) at y = a.
Solution:
Z ∞
1
w(x, y) = f1 (ξ) ln cosh[π(x − ξ)/a] − cos(πy/a) dξ
2π −∞
Z ∞
1
− f2 (ξ) ln cosh[π(x − ξ)/a] + cos(πy/a) dξ + C,
2π −∞

◮ Domain: 0 ≤ x < ∞, 0 ≤ y ≤ a. First boundary value problem.

A semiinfinite strip is considered. Boundary conditions are prescribed:
w = f1 (y) at x = 0, w = f2 (x) at y = 0, w = f3 (x) at y = a.

Solution:
∞ Z a
2X nπx nπy nπη
w(x, y)= exp − sin f1(η) sin dη
a n=1 a a 0 a
Z ∞
1 πy 1 1
+ sin − f2(ξ) dξ
2a a 0 cosh[π(x−ξ)/a]−cos(πy/a) cosh[π(x+ξ)/a]−cos(πy/a)
Z ∞
1 πy 1 1
+ sin − f3(ξ) dξ.
2a a 0 cosh[π(x−ξ)/a]+cos(πy/a) cosh[π(x+ξ)/a]+cos(πy/a)
Example 9.1. Consider the first boundary value problem for the Laplace equation in a semiinfi-
nite strip with f1 (y) = 1 and f2 (x) = f3 (x) = 0.
Using the general formula and carrying out transformations, we obtain the solution

2 sin(πy/a)
w(x, y) = arctan .
π sinh(πx/a)
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b. First boundary value problem.

A rectangle is considered. Boundary conditions are prescribed:
w = f1 (y) at x = 0, w = f2 (y) at x = a,
w = f3 (x) at y = 0, w = f4 (x) at y = b.
Solution:
∞
X X ∞
nπ nπ nπ nπ
w(x, y) = An sinh (a − x) sin y + Bn sinh x sin y
n=1
b b n=1
b b
X∞ X ∞
nπ nπ nπ nπ
+ Cn sin x sinh (b − y) + Dn sin x sinh y ,
a a a a
n=1 n=1
where the coefficients An , Bn , Cn , and Dn are expressed as

Z b Z b
2 nπξ 2 nπξ
An = f1 (ξ) sin dξ, Bn = f2 (ξ) sin dξ,
λn 0 b λn 0 b
Z a Z a
2 nπξ 2 nπξ
Cn = f3 (ξ) sin dξ, Dn = f4 (ξ) sin dξ,
µn 0 a µn 0 a

nπa nπb
λn = b sinh , µn = a sinh .
b a
⊙ Literature: M. M. Smirnov (1975), H. S. Carslaw and J. C. Jaeger (1984).
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b. Second boundary value problem.

∂x w = f1 (y) at x = 0, ∂x w = f2 (y) at x = a,
∂y w = f3 (x) at y = 0, ∂y w = f4 (x) at y = b.
Solution:
A0 B0 C0 D0
w(x, y)=− (x−a)2+ x2− (x−b)2+ y 2+K
4a 4a 4b 4b
X∞ X ∞
An nπ nπ Bn nπ nπ
−b cosh (a−x) cos y +b cosh x cos y
λ
n=1 n
b b λ
n=1 n
b b
X∞ X ∞
Cn nπ nπ Dn nπ nπ
−a cos x cosh (b−y) +a cos x cosh y ,
µn a a µn a a
n=1 n=1
where K is an arbitrary constant, and the coefficients An , Bn , Cn , Dn , λn , and µn are

expressed as
Z Z
2 b nπξ 2 b nπξ
An = f1 (ξ) cos dξ, Bn = f2 (ξ) cos dξ,
b 0 b b 0 b
Z Z
2 a nπξ 2 a nπξ
Cn = f3 (ξ) cos dξ, Dn = f4 (ξ) cos dξ,
a 0 a a 0 a

nπa nπb
λn = nπ sinh , µn = nπ sinh .
b a
The solvability condition for the problem in question has the form (see Section 9.1.1,
condition (2)):
Z b Z b Z a Z a
f1 (y) dy + f2 (y) dy − f3 (x) dx − f4 (x) dx = 0.
0 0 0 0
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b. Third boundary value problem.

∂x w − k1 w = f1 (y) at x = 0, ∂x w + k2 w = f2 (y) at x = a,
∂y w − k3 w = f3 (x) at y = 0, ∂y w + k4 w = f4 (x) at y = b.
For the solution, see Section 9.2.2 (the third boundary value problem for 0 ≤ x ≤ a,
0 ≤ y ≤ b) with Φ ≡ 0.
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b. Mixed boundary value problems.

1◦ . A rectangle is considered. Boundary conditions are prescribed:
∂x w = f (y) at x = 0, ∂x w = g(y) at x = a,
w = h(x) at y = 0, w = s(x) at y = b.
Solution:
∞ ∞
b X fn πn πny b X gn πnx πny
w(x, y) = − cosh (a−x) sin + cosh sin
π n=1 nλn b b π n=1 nλn b b
∞
X hn ∞
X sn
πnx πn πnx πny
+ cos sinh (b−y) + cos sinh
µ
n=1 n
a a µ
n=1 n
a a
Z a Z a
b−y y
+ h(x) dx+ s(x) dx,
ab 0 ab 0
where
Z Z
2 b πnξ 2 b πnξ
fn = f (ξ) sin dξ, gn = g(ξ) sin dξ,
b 0 b b 0 b
Z Z
2 a πnξ 2 a πnξ
hn = h(ξ) cos dξ, sn = s(ξ) cos dξ,
a 0 a a 0 a

πna πnb
λn = sinh , µn = sinh .
b a

w = f (y) at x = 0, ∂x w = g(y) at x = a,
w = h(x) at y = 0, ∂y w = s(x) at y = b,
where f (0) = h(0).
Solution:
∞
X X∞
fn a−x y gn x y
w(x, y)= cosh λn sin λn +a sinh λn sin λn
n=0
cosh λn a a λ cosh λn
n=0 n
a a
X∞ X∞
hn x b−y sn x y
+ sin µn cosh µn +b sin µn sinh µn ,
n=0
cosh µn b b µ cosh µn
n=0 n
b b
where
Z Z
2 b π(2n + 1) 2 b π(2n + 1)
fn = f (ξ) sin ξ dξ, gn = g(ξ) sin ξ dξ,
b 0 2b b 0 2b
Z Z
2 a π(2n + 1) 2 a π(2n + 1)
hn = h(ξ) sin ξ dξ, sn = s(ξ) sin ξ dξ,
a 0 2a a 0 2a
π(2n + 1)a π(2n + 1)b
λn = , µn = .
2b 2a
9.1.2 Problems in Polar Coordinate System

The two-dimensional Laplace equation in the polar coordinate system is written as
p
1 ∂ ∂w 1 ∂2w
r + 2 = 0, r = x2 + y 2 .
r ∂r ∂r r ∂ϕ2
w(r) = A ln r + B,

m B
w(r, ϕ) = Ar + m (C cos mϕ + D sin mϕ),
r
where m = 1, 2, . . . ; A, B, C, and D are arbitrary constants.
◮ Domain: 0 ≤ r ≤ R or R ≤ r < ∞. First boundary value problem.
The condition
w = f (ϕ) at r=R
is set at the boundary of the circle; f (ϕ) is a given function.

1◦ . Solution of the inner problem (r ≤ R):
Z 2π
1 R2 − r 2
w(r, ϕ) = f (ψ) dψ.
2π 0 r 2 − 2Rr cos(ϕ − ψ) + R2
This formula is conventionally referred to as the Poisson integral.

Solution of the outer problem in series form:
∞
a0 X r n
w(r, ϕ) = + (an cos nϕ + bn sin nϕ),
2 R
n=1
Z
1 2π
an = f (ψ) cos(nψ) dψ, n = 0, 1, 2, . . . ,
π 0
Z
1 2π
bn = f (ψ) sin(nψ) dψ, n = 1, 2, 3, . . .
π 0
2◦ . Bounded solution of the outer problem (r ≥ R):
Z 2π
1 r 2 − R2
w(r, ϕ) = f (ψ) dψ.
2π 0 r 2 − 2Rr cos(ϕ − ψ) + R2
Bounded solution of the outer problem in series form:

∞
a0 X R n
2 n=1
r
where the coefficients a0 , an , and bn are defined by the same relations as in the inner
problem.
In hydrodynamics and other applications, outer problems are sometimes encountered
in which one has to consider unbounded solutions for r → ∞.
Example 9.2. The potential flow of an ideal (inviscid) incompressible fluid about a circular
cylinder of radius R with a constant incident velocity U at infinity is characterized by the following
boundary conditions for the stream function:
w=0 at r = R, w → U r sin ϕ as r → ∞.
Solution:
R2
w(r, ϕ) = U r − sin ϕ.
r
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), A. N. Tikhonov and A. A. Samar-
skii (1990).
◮ Domain: 0 ≤ r ≤ R or R ≤ r < ∞. Second boundary value problem.

The condition
∂r w = f (ϕ) at r=R
is set at the boundary of the circle. The function f (ϕ) must satisfy the solvability condition
Z 2π
f (ϕ) dϕ = 0.
0

Z 2π
R r 2 − 2Rr cos(ϕ − ψ) + R2
w(r, ϕ) = f (ψ) ln dψ + C,
2π 0 R2
where C is an arbitrary constant; this formula is known as the Dini integral.
Series solution of the inner problem:
∞
X R r n
w(r, ϕ) = (an cos nϕ + bn sin nϕ) + C,
n R
n=1
Z Z
1 2π 1 2π
an = f (ψ) cos(nψ) dψ, bn = f (ψ) sin(nψ) dψ,
π 0 π 0
2◦ . Solution of the outer problem (r ≥ R):
Z 2π
R r 2 − 2Rr cos(ϕ − ψ) + R2
w(r, ϕ) = − f (ψ) ln dψ + C,
2π 0 r2
Series solution of the outer problem:
∞
X R R n
w(r, ϕ) = − (an cos nϕ + bn sin nϕ) + C,
n r
n=1
where the coefficients an and bn are defined by the same relations as in the inner problem,
and C is an arbitrary constant.
◮ Domain: 0 ≤ r ≤ R or R ≤ r < ∞. Third boundary value problem.
The condition
∂r w + kw = f (ϕ) at r = R
is set at the circle boundary; f (ϕ) is a given function.

∞ n
a0 X R r
2k kR + n R
n=1
Z
1 2π
an = f (ψ) cos(nψ) dψ, n = 0, 1, 2, . . . ,
π 0
Z
1 2π
bn = f (ψ) sin(nψ) dψ, n = 1, 2, 3, . . .
π 0
2◦ . Solution of the outer problem (r ≥ R):

∞ n
a0 X R R
2k kR − n r
n=1
where the coefficients a0 , an , and bn are defined by the same relations as in the inner
problem.
An annular domain is considered. Boundary conditions are prescribed:
w = f1 (ϕ) at r = R1 , w = f2 (ϕ) at r = R2 .
Solution:
∞
X ∞
X 1
w(r, ϕ) = A0 +B0 ln r+ r n (An cos nϕ+Bn sin nϕ)+ (Cn cos nϕ+Dn sin nϕ),
rn
n=1 n=1
where the coefficients A0 , B0 , An , Bn , Cn , and Dn are expressed as
(1) (2) (2) (1)

1 a0 ln R2 − a0 ln R1 1 a0 − a0
A0 = , B0 = ,
2 ln R2 − ln R1 2 ln R2 − ln R1
(2) (1) (2) (1)
Rn an − R1n an Rn bn − R1n bn
An = 2 2n , Bn = 2 2n ,
R2 − R12n R2 − R12n
(1) (2) (1) (2)
R2n an − R1n an R2n bn − R1n bn
Cn = (R1 R2 )n , Dn = (R1 R2 )n .
R22n − R12n R22n − R12n
(i) (i)
Here, the an and bn (i = 1, 2) are the coefficients of the Fourier series expansions of the
functions f1 (ϕ) and f2 (ϕ):
Z
(i) 1 2π
an = fi (ψ) cos(nψ) dψ, n = 0, 1, 2, . . . ,
π 0
Z
1 2π
b(i)
n = fi (ψ) sin(nψ) dψ, n = 1, 2, 3, . . .
π 0

∂r w = f1 (ϕ) at r = R1 , ∂r w = f2 (ϕ) at r = R2 .
Solution:
∞
X ∞
X
n 1
w(r, ϕ) = B ln r + r (An cos nϕ+Bn sin nϕ)+ (Cn cos nϕ+Dn sin nϕ)+K.
rn
n=1 n=1
The coefficients B, An , Bn , Cn , and Dn are expressed as

(2) (1) (2) (1)
1 (1) Rn+1 an − R1n+1 an Rn+1 bn − R1n+1 bn
B = R1 a0 , An = 2 , Bn = 2 ,
2 n(R22n − R12n ) n(R22n − R12n )
(2) (1) (2) (1)
R1n−1 an − R2n−1 an R1n−1 bn − R2n−1 bn
Cn = (R1 R2 )n+1 , Dn = (R1 R2 )n+1 ,
n(R22n − R12n ) n(R22n − R12n )
(i) (i)
where the constants an and bn (i = 1, 2) are defined by the same relations as in the first
boundary value problem; K is an arbitrary constant.
(1) (2)
Remark 9.2. Note that the condition a0 R1 = a0 R2 must hold; this relation is a consequence
of the solvability condition for the problem,
Z Z
f1 dS − f2 dS = 0.
r=R1 r=R2
◮ Domain: R1 ≤ r ≤ R2 . Mixed boundary value problem.

∂r w = f1 (ϕ) at r = R1 , w = f2 (ϕ) at r = R2 .
Solution:
(2) (1) r
w(r, ϕ) = 12 a0 + 12 a0 R1 ln
R2
∞
X X∞
1
+ r n (An cos nϕ + Bn sin nϕ) + (Cn cos nϕ + Dn sin nϕ).
n=1 n=1
rn
Here, the coefficients An , Bn , Cn , and Dn are expressed as

(2) (1) (2) (1)
nR2n an + R1n+1 an nR2n bn + R1n+1 bn
An = , Bn = ,
n(R22n + R12n ) n(R22n + R12n )
n−1 (2) (1) n−1 (2) (1)
n+1 n nR1 an − R2n an n+1 n nR1 bn − R2n bn
Cn = R1 R2 , Dn = R1 R2 ,
n(R22n + R12n ) n(R22n + R12n )
(i) (i)
where the constants an and bn (i = 1, 2) are defined by the same formulas as in the first
boundary value problem.
9.1.3 Other Coordinate Systems. Conformal Mappings Method

◮ Parabolic, elliptic, and bipolar coordinate systems.
In a number of applications, it is convenient to solve the Laplace equation in other or-
thogonal systems of coordinates. Some of those commonly encountered are displayed in
Table 9.1. In all the coordinate systems presented, the Laplace equation ∆2 w = 0 is reduced
to the equation considered in Section 9.1.1 in detail (particular solutions and solutions to
boundary value problems are given there).
TABLE 9.1
Two-dimensional Laplace operator in some curvilinear orthogonal systems of coordinates
Coordinates Transformation (c > 0) Laplace operator, ∆2w

Parabolic coordinates x = cuv, y = 1
2
c(v 2 − u2) 1 ∂ 2w ∂ 2w
+
u, v −∞ < u < ∞, 0 ≤ v < ∞ c2(u2 + v 2) ∂u2 ∂v 2

Elliptic coordinates x = c cosh ξ cos η, y = c sinh ξ sin η 1 ∂ 2w ∂ 2w
+
ξ, η 0 ≤ ξ < ∞, 0 ≤ η < 2π c2(sinh2 ξ + sin2 η) ∂ξ 2 ∂η 2
c sinh τ c sin σ 2
Bipolar coordinates x= , y= 1 2 ∂ w ∂ 2w
cosh τ − cos σ cosh τ − cos σ (cosh τ − cos σ) +
σ, τ c2 ∂σ 2 ∂τ 2
0 ≤ σ < 2π, −∞ < τ < ∞
The orthogonal transformations presented in Table 9.1 can be written in the language
of complex variables as follows:
x + iy = − 12 ic(u + iv)2 (parabolic coordinates),

x + iy = c cosh(ξ + iη) (elliptic coordinates),

x + iy = ic cot 12 (σ + iτ ) (bipolar coordinates).
The real parts, as well as the imaginary parts, in both sides of these relations must be
equated to each other (i2 = −1).
Example 9.3. Plane hydrodynamic problems of potential flows of ideal (inviscid) incompress-
ible fluid are reduced to the Laplace equation for the stream function. In particular, the motion of an
elliptic cylinder with semiaxes a and b at a velocity U in the direction parallel to the major semiaxis
(a > b) in ideal fluid is described by the stream function
1/2
a+b
w(ξ, η) = −U b eξ sin η, c2 = a2 − b 2 ,
a−b
where ξ and η are the elliptic coordinates.
⊙ Literature: H. Lamb (1945), J. Happel and H. Brenner (1965), G. Korn and T. Korn (2000).
◮ Domain of arbitrary shape. Method of conformal mappings.

1◦ . Let ζ = ζ(z) be an analytic function that defines a conformal mapping from the com-
plex plane z = x + iy into a complex plane ζ = u + iv, where u = u(x, y) and v = v(x, y)
are new independent variables. With reference to the fact that the real and imaginary parts
of an analytic function satisfy the Cauchy–Riemann conditions, we have ∂x u = ∂y v and
∂y u = −∂x v, and hence
2
∂2w ∂2w ′ 2 ∂ w ∂2w
+ = |ζ (z)| + .
∂x2 ∂y 2 ∂u2 ∂v 2
Therefore, the Laplace equation in the xy-plane transforms under a conformal mapping
into the Laplace equation in the uv-plane.
2◦ . Any simply connected domain D in the xy-plane with a piecewise smooth boundary
can be mapped, with appropriate conformal mappings, onto the upper half-plane or into
a unit circle in the uv-plane. Consequently, a first and a second boundary value problem
for the Laplace equation in D can be reduced, respectively, to a first and a second bound-
ary value problem for the upper half-space or a circle; such problems are considered in
Section 9.2.4 presents conformal mappings of some domains onto the upper half-plane
or a unit circle. Moreover, examples of solving specific boundary value problems for the
Poisson equation by the conformal mappings method are given there; the Green’s functions
for a semicircle and a quadrant of a circle are obtained.
A large number of conformal mappings of various domains can be found, for example,
in the references cited below.
⊙ Literature: V. I. Lavrik and V. N. Savenkov (1970), M. A. Lavrent’ev and B. V. Shabat (1973), V. I. Ivanov
and M. K. Trubetskov (1994).
◮ Reduction of the two-dimensional Neumann problem to the Dirichlet problem.

Let the position of any point (x∗ , y∗ ) located on the boundary Σ of a domain D be spec-
ified by a parameter s, so that x∗ = x∗ (s) and y∗ = y∗ (s). Then a function of two vari-
ables, f (x, y), is determined on Σ by the parameter s as well, f (x, y)|Σ = f (x∗ (s), y∗ (s)) =
f∗ (s).
The solution of the two-dimensional Neumann problem for the Laplace equation
∆2 w = 0 in D with the boundary condition of the second kind
∂w
= f∗ (s) for r ∈ Σ
∂N
can be expressed in terms of the solution of the two-dimensional Dirichlet problem for the
Laplace equation ∆2 u = 0 in D with the boundary condition of the first kind
u = F∗ (s) for r ∈ Σ,
Z
where F∗ (s) = f∗ (s) ds, as follows:
Z x Z y
∂u ∂u
w(x, y) = (t, y0 ) dt − (x, t) dt + C.
x0 ∂y y0 ∂x
Here, (x0 , y0 ) are the coordinates of any point in D, and C is an arbitrary constant.
9.2 Poisson Equation ∆2 w = −Φ(x)

9.2.1 Preliminary Remarks. Solution Structure
Just as the Laplace equation, the Poisson equation is often encountered in heat and mass
transfer theory, fluid mechanics, elasticity, electrostatics, and other areas of mechanics and
physics. For example, it describes steady-state temperature distribution in the presence of
heat sources or sinks in the domain under study.
The Laplace equation is a special case of the Poisson equation with Φ ≡ 0.
In what follows, we consider a finite domain S with a sufficiently smooth boundary L.
Let r ∈ S and ρ ∈ S, where r = {x, y}, ρ = {ξ, η}, |r − ρ|2 = (x − ξ)2 + (y − η)2 .

The solution of the first boundary value problem for the Poisson equation
∆2 w = −Φ(r) (1)
in the domain S with the nonhomogeneous boundary condition
w = f (r) for r∈L
can be represented as
Z Z
∂G
w(r) = Φ(ρ)G(r, ρ) dSρ − f (ρ) dLρ . (2)
S L ∂Nρ
∂G
Here, G(r, ρ) is the Green’s function of the first boundary value problem, ∂N ρ
is the deriva-
tive of the Green’s function with respect to ξ, η along the outward normal N to the bound-
ary L. The integration is performed with respect to ξ, η, with dSρ = dξ dη.
The Green’s function G = G(r, ρ) of the first boundary value problem is determined
by the following conditions.
1◦ . The function G satisfies the Laplace equation in x, y in the domain S everywhere
1 1
except for the point (ξ, η), at which G has a singularity of the form 2π ln |r−ρ| .
9.2. Poisson Equation ∆2 w = −Φ(x) 795
2◦ . With respect to x, y, the function G satisfies the homogeneous boundary condition of

the first kind at the domain boundary, i.e., the condition G|L = 0.
The Green’s function can be represented in the form
1 1
G(r, ρ) = ln + u, (3)
2π |r − ρ|
where the auxiliary function u = u(r, ρ) is determined by solving the first boundary value
1 1
problem for the Laplace equation ∆2 u = 0 with the boundary condition u|L = − 2π ln |r−ρ| ;
in this problem, ρ is treated as a two-dimensional free parameter.
The Green’s function is symmetric with respect to its arguments: G(r, ρ) = G(ρ, r).
Remark 9.3. When using the polar coordinate system, one should set
r = {r, ϕ}, ρ = {ξ, η}, |r − ρ|2 = r2 + ξ 2 − 2rξ cos(ϕ − η), dSρ = ξ dξ dη

in relations (2) and (3).

The second boundary value problem for the Poisson equation (1) is characterized by the
boundary condition
∂w
= f (r) for r ∈ L.
∂N
The necessary solvability condition for this problem is
Z Z
Φ(r) dS + f (r) dL = 0. (4)
S L
The solution of the second boundary value problem, provided that condition (4) is sat-
isfied, can be represented as
Z Z
w(r) = Φ(ρ)G(r, ρ) dSρ + f (ρ)G(r, ρ) dLρ + C, (5)
S L
The Green’s function G = G(r, ρ) of the second boundary value problem is determined
by the following conditions:
1 1
2◦ . With respect to x, y, the function G satisfies the homogeneous boundary condition of
the second kind at the domain boundary:

∂G 1
= ,
∂N L L0
where L0 is the length of the boundary of S.
The Green’s function is unique up to an additive constant.
Remark 9.4. The Green’s
function cannot be determined by condition 1◦ and the homogeneous
∂G
boundary condition ∂N L
= 0. The point is that the problem is unsolvable for G in this case,
because, on representing G in the form (3), for u we obtain a problem with a nonhomogeneous
boundary condition of the second kind for which the solvability condition (4) now is not satisfied.

The solution of the third boundary value problem for the Poisson equation (1) in the do-
main S with the nonhomogeneous boundary condition
∂w
+ kw = f (r) for r ∈ L
∂N
is given by formula (5) with C = 0, where G = G(r, ρ) is the Green’s function of the third
boundary value problem and is determined by the following conditions:
1 1
2◦ . With respect to x, y, the function G satisfies the homogeneous
∂G boundary condition of
the third kind at the domain boundary, i.e., the condition ∂N + kG L = 0.
The Green’s function can be represented in the form (3); the auxiliary function u is iden-
tified by solving the corresponding third boundary value problem for the Laplace equation
∆2 u = 0.
The Green’s function is symmetric with respect to its arguments: G(r, ρ) = G(ρ, r).
⊙ Literature for Section 9.2.1: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), N. S. Koshlyakov,
E. B. Gliner, and M. M. Smirnov (1970).

The two-dimensional Poisson equation in the rectangular Cartesian coordinate system has
the form
∂2w ∂2w
+ + Φ(x, y) = 0.
∂x2 ∂y 2
◮ Particular solutions of the Poisson equation with a special right-hand side.

n P
P n
1◦ . If Φ(x, y) = aij exp(bi x + cj y), the equation has solutions of the form
i=1 j=1
n X
X n
aij
w(x, y) = − exp(bi x + cj y).
i=1 j=1
b2i + c2j
n P
P n
2◦ . If Φ(x, y) = aij sin(bi x + pi ) sin(cj y + qj ), the equation admits solutions of
i=1 j=1
the form
n X
X n
aij
w(x, y) = sin(bi x + pi ) sin(cj y + qj ).
b2
i=1 j=1 i
+ c2j
◮ Domain: −∞ < x < ∞, −∞ < y < ∞.

Solution:
Z ∞ Z ∞
1 1
w(x, y) = Φ(ξ, η) ln p dξ dη.
2π −∞ −∞ (x − ξ)2 + (y − η)2
w = f (x) at y = 0.
Solution:
Z ∞ Z ∞Z ∞
p
1 yf (ξ) dξ 1 (x − ξ)2 + (y + η)2
w(x, y) = 2 2
+ Φ(ξ, η) ln p dξ dη.
π −∞ (x − ξ) + y 2π 0 −∞ (x − ξ)2 + (y − η)2
∂y w = f (x) at y = 0.
Solution:
Z
1 ∞ p
w(x, y) = f (ξ) ln (x−ξ)2 +y 2 dξ
π −∞
Z ∞Z ∞
1 1 1
+ Φ(ξ, η) ln p +ln p dξ dη+C,
2π 0 −∞ (x−ξ)2 +(y−η)2 (x−ξ)2 +(y+η)2

w = f1 (x) at y = 0, w = f2 (x) at y = a.
Solution:
Z ∞
1 πy f1 (ξ) dξ
w(x, y) = sin
2a a cosh[π(x − ξ)/a] − cos(πy/a)
Z−∞∞
1 πy f2 (ξ) dξ
+ sin
2a a −∞ cosh[π(x − ξ)/a] + cos(πy/a)
Z aZ ∞
1 cosh[π(x − ξ)/a] − cos[π(y + η)/a]
+ Φ(ξ, η) ln dξ dη.
4π 0 −∞ cosh[π(x − ξ)/a] − cos[π(y − η)/a]


Solution:
Z ∞ Z ∞
w(x, y) = − f1 (ξ)G(x, y, ξ, 0) dξ + f2 (ξ)G(x, y, ξ, a) dξ
−∞ −∞
Z aZ ∞
+ Φ(ξ, η)G(x, y, ξ, η) dξ dη + C.
0 −∞
Here,
1 1
G(x, y, ξ, η) = ln
4π cosh[π(x − ξ)/a] − cos[π(y − η)/a]
1 1
+ ln ,
4π cosh[π(x − ξ)/a] − cos[π(y + η)/a]
◮ Domain: −∞ < x < ∞, 0 ≤ y ≤ a. Third boundary value problem.

∂y w − k1 w = f1 (x) at y = 0, ∂y w + k2 w = f2 (x) at y = a.
The solution w(x, y) is determined by the formula in the previous paragraph (for the
∞
1 X ϕn (y)ϕn (η)
G(x, y, ξ, η) = 2
exp −µn |x − ξ| ,
2 n=1 kϕn k µn

2 1 2 2 (k1 + k2 )(µ2n + k1 k2 )
ϕn (y) = µn cos(µn y)+k1 sin(µn y), kϕn k = (µn +k1 ) a+ .
2 (µ2n + k12 )(µ2n + k22 )
(k1 + k2 )µ
Here, the µn are positive roots of the transcendental equation tan(µa) = .
µ2 − k1 k2
◮ Domain: −∞ < x < ∞, 0 ≤ y ≤ a. Mixed boundary value problem.

w = f1 (x) at y = 0, ∂y w = f2 (x) at y = a.
Solution:
Z ∞ Z ∞
∂
w(x, y) = f1 (ξ) G(x, y, ξ, η) dξ + f2 (ξ)G(x, y, ξ, a) dξ
−∞ ∂η η=0 −∞
Z aZ ∞
+ Φ(ξ, η)G(x, y, ξ, η) dξ dη,
0 −∞
where
∞
1X 1 π(2n + 1)
G(x, y, ξ, η) = exp −µn |x − ξ| sin(µn y) sin(µn η), µn = .
a n=0 µn 2a

Solution:
Z a Z ∞
∂ ∂
w(x, y) = f1 (η) G(x, y, ξ, η) dη + f2 (ξ) G(x, y, ξ, η) dξ
0 ∂ξ ξ=0 0 ∂η η=0
Z ∞ Z aZ ∞
∂
− f3 (ξ) G(x, y, ξ, η) dξ + Φ(ξ, η)G(x, y, ξ, η) dξ dη,
0 ∂η η=a 0 0
where
G(x, y, ξ, η) = ln
1 cosh[π(x + ξ)/a] − cos[π(y + η)/a]
− ln .
4π cosh[π(x + ξ)/a] − cos[π(y − η)/a]
Alternatively, the Green’s function can be represented in the series form
∞
1X 1 πn
G(x, y, ξ, η)= exp −qn|x−ξ| −exp −qn|x+ξ| sin(qny) sin(qnη), qn = .
a qn a
n=1
⊙ Literature: N. N. Lebedev, I. P. Skal’skaya, and Ya. S. Uflyand (1955), A. G. Butkovskiy (1979).
◮ Domain: 0 ≤ x < ∞, 0 ≤ y ≤ a. Third boundary value problem.

∂xw−k1w=f1(y) at x=0, ∂y w−k2w=f2(x) at y=0, ∂yw+k3w=f3(x) at y=a.
Solution:
Z aZ ∞ Z a
w(x, y) = Φ(ξ, η)G(x, y, ξ, η) dξ dη − f1 (η)G(x, y, 0, η) dη
Z0 ∞ 0 Z ∞ 0
− f2 (ξ)G(x, y, ξ, 0) dξ + f3 (ξ)G(x, y, ξ, a) dξ,

0 0
where
∞
X
ϕn (y)ϕn (η)
G(x, y, ξ, η) = Hn (x, ξ),
kϕn k2 µn (µn + k1 )
n=1

2 1 2 2 (k2 + k3 )(µ2n + k2 k3 )
ϕn (y) = µn cos(µn y)+k2 sin(µn y), kϕn k = (µn +k2 ) a+ ,
2 (µ2n + k22 )(µ2n + k32 )
(
exp(−µn x) µn cosh(µn ξ) + k1 sinh(µn ξ) for x > ξ,
Hn (x, ξ) =
exp(−µn ξ) µn cosh(µn x) + k1 sinh(µn x) for ξ > x.
(k2 + k3 )µ
µ2 − k2 k3
◮ Domain: 0 ≤ x < ∞, 0 ≤ y ≤ a. Mixed boundary value problems.

1◦ . A semiinfinite strip is considered. Boundary conditions are prescribed:
w = f1 (y) at x = 0, ∂y w = f2 (x) at y = 0, ∂y w = f3 (x) at y = a.
Solution:
Z a Z ∞
∂
w(x, y) = f1 (η) G(x, y, ξ, η) dη − f2 (ξ)G(x, y, ξ, 0) dξ
0 ∂ξ ξ=0 0
Z ∞ Z aZ ∞
+ f3 (ξ)G(x, y, ξ, a) dξ + Φ(ξ, η)G(x, y, ξ, η) dξ dη,
0 0 0
where
∞
1 X εn
G(x, y, ξ, η) = exp −qn |x − ξ| − exp −qn |x + ξ| cos(qn y) cos(qn η),
2a q
n=0 n
(
πn 1 for n = 0,
qn = , ε=
a 2 for n 6= 0.
∂x w = f1 (y) at x = 0, w = f2 (x) at y = 0, w = f3 (x) at y = a.
Solution:
Z a Z ∞
∂
w(x, y) = − f1 (η)G(x, y, 0, η) dη + f2 (ξ) G(x, y, ξ, η) dξ
0 0 ∂η η=0
Z ∞ Z aZ ∞
∂
0 ∂η η=a 0 0
where
∞
1X 1 πn
G(x, y, ξ, η)= exp −qn|x−ξ| +exp −qn|x+ξ| sin(qny) sin(qnη), qn = .
a qn a
n=1

w = f1 (y) at x = 0, w = f2 (x) at y = 0.
Solution:
Z ∞ Z ∞
4 f1(η)η dη 4 f2(ξ)ξ dξ
w(x, y)= xy 2 2 2 2
+ xy
π 0 [x +(y−η) ][x +(y+η) ] π [(x−ξ) +y 2][(x+ξ)2 +y 2]
2
Z ∞Z ∞ p p0
1 (x−ξ)2 +(y+η)2 (x+ξ)2 +(y−η)2
+ Φ(ξ, η) ln p p dξ dη.
2π 0 0 (x−ξ)2 +(y−η)2 (x+ξ)2 +(y+η)2
⊙ Literature: V. S. Vladimirov, V. P. Mikhailov, A. A. Vasharin et al. (1974), A. G. Butkovskiy (1979).

w = f1 (y) at x = 0, w = f2 (y) at x = a,
w = f3 (x) at y = 0, w = f4 (x) at y = b.
Solution:
Z aZ b
w(x, y) = Φ(ξ, η)G(x, y, ξ, η) dη dξ
0 0
Z b Z b
∂ ∂
+ f1 (η) G(x, y, ξ, η) dη − f2 (η) G(x, y, ξ, η) dη
0 ∂ξ ξ=0 0 ∂ξ ξ=a
Z a Z a
∂ ∂
+ f3 (ξ) G(x, y, ξ, η) dξ − f4 (ξ) G(x, y, ξ, η) dξ.
0 ∂η η=0 0 ∂η η=b

∞ ∞
2 X sin(pn x) sin(pn ξ) 2 X sin(qm y) sin(qm η)
G(x, y, ξ, η) = Hn (y, η) = Qm (x, ξ),
a pn sinh(pn b) b qm sinh(qm a)
n=1 m=1
where
(
πn sinh(pn η) sinh[pn (b − y)] for b ≥ y > η ≥ 0,
pn = , Hn (y, η) =
a sinh(pn y) sinh[pn (b − η)] for b ≥ η > y ≥ 0,
(
πm sinh(qm ξ) sinh[qm (a − x)] for a ≥ x > ξ ≥ 0,
qm = , Qm (x, ξ) =
b sinh(qm x) sinh[qm (a − ξ)] for a ≥ ξ > x ≥ 0.
The Green’s function can be written in the form of a double series:
∞ ∞
4 X X sin(pnx) sin(qmy) sin(pnξ) sin(qmη) πn πm
G(x, y, ξ, η)= , pn = , qm = .
ab p2n+qm2 a b
n=1 m=1

∂x w − k1 w = f1 (y) at x = 0, ∂x w + k2 w = f2 (y) at x = a,
∂y w − k3 w = f3 (x) at y = 0, ∂y w + k4 w = f4 (x) at y = b.
Solution:
Z aZ b
0 0
Z b Z b
− f1 (η)G(x, y, 0, η) dη + f2 (η)G(x, y, a, η) dη
0 0
Z a Z a
− f3 (ξ)G(x, y, ξ, 0) dξ + f4 (ξ)G(x, y, ξ, b) dξ.
0 0
Here,
X∞ X ∞
ϕn(x)ϕn (ξ)ψm (y)ψm(η)
G(x, y, ξ, η) = ,
n=1 m=1
kϕnk2 kψmk2 (µ2n + λ2m)

k1 k2 µ2n + k12 k1 a k12
ϕn(x) = cos(µnx) + sin(µn x), kϕn k2 =
+ + 1 + ,
µn 2µ2n µ2n + k22 2µ2n 2 µ2n

k3 2 k4 λ2m + k32 k3 b k32
ψm(y) = cos(λmy) + sin(λmy), kψm k = 2 2 + + 1+ 2 ,
λm 2λm λm + k42 2λ2m 2 λm
where the µn and λm are positive roots of the transcendental equations
tan(µa) k1 + k2 tan(λb) k3 + k4
= 2 , = 2 .
µ µ − k1 k2 λ λ − k3 k4
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b. Mixed boundary value problem.

w = f1 (y) at x = 0, ∂x w = f2 (y) at x = a,
w = f3 (x) at y = 0, ∂y w = f4 (x) at y = b.
Solution:
Z aZ b
0 0
Z b Z b
∂
+ f1 (η) G(x, y, ξ, η) dη + f2 (η)G(x, y, a, η) dη
0 ∂ξ ξ=0 0
Z a Z a
∂
+ f3 (ξ) G(x, y, ξ, η) dξ + f4 (ξ)G(x, y, ξ, b) dξ.
0 ∂η η=0 0

∞ ∞
G(x, y, ξ, η) = Hn (y, η) = Qm (x, ξ),
a n=0 pn cosh(pn b) b m=0 qm cosh(qm a)
where
(
π(2n + 1) sinh(pn η) cosh[pn (b − y)] for b ≥ y > η ≥ 0,
pn = , Hn (y, η) =
a sinh(pn y) cosh[pn (b − η)] for b ≥ η > y ≥ 0,
(
π(2m + 1) sinh(qm ξ) cosh[qm (a − x)] for a ≥ x > ξ ≥ 0,
qm = , Qm (x, ξ) =
b sinh(qm x) cosh[qm (a − ξ)] for a ≥ ξ > x ≥ 0.
The Green’s function can be written in the form of a double series:
∞ ∞
4 X X sin(pn x) sin(qm y) sin(pn ξ) sin(qm η)
G(x, y, ξ, η) = ,
ab n=0 m=0 p2n + qm
2
π(2n + 1) π(2m + 1)
pn = , qm = .
2a 2b

The two-dimensional Poisson equation in the polar coordinate system is written as
p
1 ∂ ∂w 1 ∂2w
r + 2 2
+ Φ(r, ϕ) = 0, r = x2 + y 2 .
r ∂r ∂r r ∂ϕ

A circle is considered. A boundary condition is prescribed:
w = f (ϕ) at r = R.
Solution:
Z 2π Z 2πZ R
1 R2−r 2
w(r, ϕ)= f (η) 2 dη+ Φ(ξ, η)G(r, ϕ, ξ, η)ξ dξ dη,
2π 0 r −2Rr cos(ϕ−η)+R2 0 0
where
1 1 1 R
G(r, ϕ, ξ, η) = ln − ln ,
2π |r − r0 | 2π r0 |(R/r0 )2 r0 − r|
r = {x, y}, x = r cos ϕ, y = r sin ϕ,
r0 = {x0 , y0 }, x0 = ξ cos η, y0 = ξ sin η.
The magnitude of a vector difference is calculated as
|ar − br0 |2 = a2 r 2 − 2abrξ cos(ϕ − η) + b2 ξ 2
where a and b are any scalars. Thus, we obtain
1 r 2 ξ 2 − 2R2 rξ cos(ϕ − η) + R4
G(r, ϕ, ξ, η) = ln 2 2 .
4π R [r − 2rξ cos(ϕ − η) + ξ 2 ]
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), A. G. Butkovskiy (1979).

∂r w + kw = f (ϕ) at r = R.
Solution:
Z 2π Z 2πZ R
w(r, ϕ) = R f (η)G(r, ϕ, R, η) dη + Φ(ξ, η)G(r, ϕ, ξ, η)ξ dξ dη,
0 0 0
where
∞ ∞
1 XX An Jn (µnm r)Jn (µnm ξ)
G(r, ϕ, ξ, η) = cos[n(ϕ − η)],
π n=0 m=1 (µ2nm R2 + k2 R2 − n2 )[Jn (µnm R)]2
A0 = 1, An = 2 (n = 1, 2, . . .).
equation
◮ Domain: R ≤ r < ∞, 0 ≤ ϕ ≤ 2π. First boundary value problem.
The exterior of a circle is considered. A boundary condition is prescribed:
w = f (ϕ) at r = R.
Solution:
Z 2π Z 2πZ ∞
1 r 2−R2
w(r, ϕ)= f (η) 2 dη+ Φ(ξ, η)G(r, ϕ, ξ, η)ξ dξ dη,
2π 0 r −2Rr cos(ϕ−η)+R2 0 R
where the Green’s function G(r, ϕ, ξ, η) is defined by the formula presented in the first
paragraph for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π.
w = f1 (ϕ) at r = R1 , w = f2 (ϕ) at r = R2 .
Solution:
Z 2π Z 2π
∂ ∂
w(r, ϕ)=R1 f1(η) G(r, ϕ, ξ, η) dη−R2 f2(η) G(r, ϕ, ξ, η) dη
0 ∂ξ ξ=R1 0 ∂ξ ξ=R2
Z 2πZ R2
+ Φ(ξ, η)G(r, ϕ, ξ, η)ξ dξ dη.
0 R1
Here,
∞
1 X 1 R1
G(r, ϕ, ξ, η) = ln − ln ∗ ,
2π rn ξrn
n=0
where
rn2 = r 2 + ρ2n − 2rρn cos(ϕ − η), (rn∗ )2 = r 2 + (ρ∗n )2 − 2rρ∗n cos(ϕ − η),
(
(R1 /R2 )2k ξ for n = 2k, ∗ R12
ρn = ρn = .
(R2 /R1 )2k+2 ξ for n = 2k + 1, ρn
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ π. First boundary value problem.
A semicircle is considered. Boundary conditions are prescribed:
w = f1 (ϕ) at r = R, w = f2 (r) at ϕ = 0, w = f3 (r) at ϕ = π.

Solution:
Z π Z R
∂ 1 ∂
w(r, ϕ) = −R f1(η) G(r, ϕ, ξ, η) dη+ f2(ξ) G(r, ϕ, ξ, η) dξ
0 ∂ξ ξ=R 0 ξ ∂η η=0
Z R Z πZ R
1 ∂
− f3(ξ) G(r, ϕ, ξ, η) dξ + Φ(ξ, η)G(r, ϕ, ξ, η)ξ dξ dη,
0 ξ ∂η η=π 0 0
where
1 r 2ξ 2−2R2rξ cos(ϕ−η)+R4 1 r 2ξ 2−2R2rξ cos(ϕ+η)+R4

G(r, ϕ, ξ, η)= ln 2 2 − ln .
4π R [r −2rξ cos(ϕ−η)+ξ 2] 4π R2[r 2−2rξ cos(ϕ+η)+ξ 2]
See also Example 9.5 in Section 9.2.4.

◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ π/2. First boundary value problem.
A quadrant of a circle is considered. Boundary conditions are prescribed:
w = f1 (ϕ) at r = R, w = f2 (r) at ϕ = 0, w = f3 (r) at ϕ = π/2.
Solution:
Z
π/2 Z R
∂ 1 ∂
w(r, ϕ)=−R f1(η) G(r, ϕ, ξ, η) dη+ f2(ξ) G(r, ϕ, ξ, η) dξ
0 ∂ξ ξ=R 0 ξ ∂η η=0
Z R Z π/2Z R
1 ∂
− f3(ξ) G(r, ϕ, ξ, η) dξ+ Φ(ξ, η)G(r, ϕ, ξ, η)ξ dξ dη,
0 ξ ∂η η=π/2 0 0
where
G(r, ϕ, ξ, η) = G1 (r, ϕ, ξ, η)−G1 (r, ϕ, ξ, 2π −η)−G1 (r, ϕ, ξ, π −η)+G1 (r, ϕ, ξ, π +η),

1 r 2 ξ 2 − 2R2 rξ cos(ϕ − η) + R4
G1 (r, ϕ, ξ, η) = ln 2 2 .
4π R [r − 2rξ cos(ϕ − η) + ξ 2 ]

◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ β. First boundary value problem.
A circular sector is considered. Boundary conditions are prescribed:
w = f1 (ϕ) at r = R, w = f2 (r) at ϕ = 0, w = f3 (r) at ϕ = β.

Solution:
Z
β Z R
∂ 1 ∂
w(r, ϕ) = −R f1(η) G(r, ϕ, ξ, η) dη+ f2(ξ) G(r, ϕ, ξ, η) dξ
0 ∂ξ ξ=R 0 ξ ∂η η=0
Z R Z βZ R
1 ∂
− f3(ξ) G(r, ϕ, ξ, η) dξ + Φ(ξ, η)G(r, ϕ, ξ, η)ξ dξ dη.
0 ξ ∂η η=β 0 0
1◦ . For β = π/n, where n is a positive integer, the Green’s function is expressed as
n−1
X
G(r, ϕ, ξ, η) = G1 (r, ϕ, ξ, 2kβ + η) − G1 (r, ϕ, ξ, 2kβ − η) ,
k=0
1 r 2 ξ 2 − 2R2 rξ cos(ϕ − η) + R4
G1 (r, ϕ, ξ, η) = ln 2 2 .
4π R [r − 2rξ cos(ϕ − η) + ξ 2 ]
2◦ . For arbitrary β, the Green’s function is given by

π/β
1 z − ζ̄ π/β R2π/β − (ζ̄z)π/β
G(r, ϕ, ξ, η) = ln π/β ,
2π z − ζ π/β R2π/β − (ζz)π/β
where z = reiϕ , ζ = ξeiη , ζ̄ = ξe−iη , and i2 = −1.
◮ Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ β. First boundary value problem.
A wedge domain is considered. Boundary conditions are prescribed:
w = f1 (r) at ϕ = 0, w = f2 (r) at ϕ = β.
Solution:
Z ∞ Z ∞
1 ∂ 1 ∂
w(r, ϕ) = f1(ξ) G(r, ϕ, ξ, η) dξ − f2(ξ) G(r, ϕ, ξ, η) dξ
0 ξ ∂η η=0 0 ξ ∂η η=β
Z βZ ∞
+ Φ(ξ, η)G(r, ϕ, ξ, η)ξ dξ dη,
0 0
where
1 r 2π/β − 2(rξ)π/β cos[π(ϕ + η)/β] + ξ 2π/β
G(r, ϕ, ξ, η) = ln 2π/β .
4π r − 2(rξ)π/β cos[π(ϕ − η)/β] + ξ 2π/β
Alternatively, the Green’s function can be represented in the complex form
π/β
1 z − ζ̄ π/β
G(r, ϕ, ξ, η) = ln π/β , z = reiϕ , ζ = ξeiη , ζ̄ = ξe−iη , i2 = −1.
2π z − ζ π/β
9.2.4 Arbitrary Shape Domain. Conformal Mappings Method

◮ Description of the method. Tables of conformal mappings.
Any simply connected domain D in the xy-plane with a piecewise smooth boundary can be
mapped in a mutually unique way, with an appropriate conformal mapping, onto the upper
half-plane or into a unit circle in a uv-plane. Under a conformal mapping, a Poisson equa-
tion in the xy-plane transforms into a Poisson equation in the uv-plane; what is changed is
the function Φ, as well as the function f in the boundary condition. Consequently, a first
and a second boundary value problem for the plane domain D can be reduced, respectively,
to a first and a second boundary value problem for the upper half-plane or a unit circle. The
latter problems are considered above (see Sections 9.2.2 and 9.2.3).
A large number of conformal mappings (mappings defined by analytic functions) of
various domains onto the upper half-plane or a unit circle can be found, for example, in
Lavrik and Savenkov (1970), Lavrent’ev and Shabat (1973), and Ivanov and Trubetskov
(1994).
Table 9.2 presents conformal mappings of some domains D in the complex plane z onto
the upper half-plane Im√ ω ≥ 0pin the plane ω.
complex In the relations involving square

1 1
assumed that ζ = |ζ| cos 2 ϕ + i sin 2 ϕ , where ϕ = arg ζ (i.e., the first
roots, it is √
branch of ζ is taken).
Table 9.3 presents conformal mappings of some domains D in the complex plane z onto
the unit circle |ω| ≤ 1 in the complex plane ω.
◮ General formula for the Green’s function. Example boundary value problems.
Let a function ω = ω(z) define a conformal mapping of a domain D in the complex plane z
onto the upper half-plane in the complex plane ω. Then the Green’s function of the first
boundary value problem in D for the Poisson (Laplace) equation is expressed as

1 ω(z) − ω̄(ζ)
G(x, y, ξ, η) = ln , z = x + iy, ζ = ξ + iη, (1)
2π ω(z) − ω(ζ)
where ω(z) = u(x, y) + iv(x, y) and ω̄(z) = u(x, y) − iv(x, y).
The solution of the first boundary value problem for the Poisson equation is determined
by the above Green’s function in accordance with formula (2) specified in Section 9.2.1.
Example 9.4. Consider the first boundary value problem for the Poisson equation in the strip
−∞ < x < ∞, 0 ≤ y ≤ a. The function that maps this strip onto the upper half-plane has the form
ω(z) = exp(πz/a) (see the second row of Table 9.2). Substituting this expression into relation (1)
and performing elementary transformations, we obtain the Green’s function
G(x, y, ξ, η) = ln .
Example 9.5. Consider the first boundary value problem for the Poisson equation in a semicircle
of radius a such that D = {x2 + y 2 ≤ a2 , y ≥ 0}. The domain D is conformally mapped onto the
upper half-plane by the function ω(z) = −(z/a + a/z) (see the sixth row in Table 9.2). Substituting
this expression into (1), we arrive at the Green’s function
1 |z − ζ̄| |a2 − z ζ̄|
G(x, y, ξ, η) = ln , z = x + iy, ζ = ξ + iη.
2π |z − ζ| |a2 − zζ|
TABLE 9.2
Conformal mapping of some domains D in the z-plane onto the upper
half-plane Im ω ≥ 0 in the ω-plane. Notation: z = x + iy and ω = u + iv
No Domain D in the z-plane Transformation
First quadrant: ω = a2 z 2 + b,
1
0 ≤ x < ∞, 0 ≤ y < ∞ a, b are real numbers
Infinite strip of width a:
2 ω = exp(πz/a)
−∞ < x < ∞, 0 ≤ y ≤ a
Semiinfinite strip of width a:
3 ω = cosh(πz/a)
0 ≤ x < ∞, 0 ≤ y ≤ a
Plane with the cut √
4 ω= z
in the real axis
Interior of an infinite sector with angle β:
5 0 ≤ arg z ≤ β, 0 ≤ |z| < ∞ (0 < β ≤ 2π) ω = z π/β
Upper half of a circle of radius a: z a

6 ω=− −
x2 + y 2 ≤ a2 , y ≥ 0 a z
Quadrant of a circle of radius a: z2 a2
7 ω=− −
x2 + y 2 ≤ a2 , x ≥ 0, y ≥ 0 a2 z2
Sector of a circle of radius a with angle β: z π/β a π/β
8
x2 + y 2 ≤ a2 , 0 ≤ arg z ≤ β ω=− −
a z
Upper half-plane with a circular domain z a
9 ω= +
or radius a removed: y ≥ 0, x2 + y 2 ≥ a2 a z
Exterior of a parabola: q q
10 ω= z − 12 p − i 12 p
y 2 − 2px ≥ 0
Interior of a parabola: q
11 ω = i cosh π 12 z/p − 1
y 2 − 2px ≤ 0 4
Example 9.6. Consider the first boundary value problem for the Poisson equation in a quadrant
of a circle of radius a, so that D = {x2 + y 2 ≤ a2 , x ≥ 0, y ≥ 0}. The conformal mapping of the
domain D onto the upper half-plane is performed with the function ω(z) = −(z/a)2 − (a/z)2 (see
the seventh row of Table 9.2). Substituting this expression into (1) yields
1 |z 2 − ζ̄ 2 | |a4 − z 2 ζ̄ 2 |
G(x, y, ξ, η) = ln 2 , z = x + iy, ζ = ξ + iη.
2π |z − ζ 2 | |a4 − z 2 ζ 2 |
Let a function ω = ω(z) define a conformal mapping of a domain D in the complex
plane z onto the unit circle |ω| ≤ 1 in the complex plane ω. Then the Green’s function of
the first boundary value problem in D for the Laplace equation is given by

1 1 − ω̄(ζ)ω(z)
G(x, y, ξ, η) = ln , z = x + iy, ζ = ξ + iη. (2)
2π ω(z) − ω(ζ)
⊙ References for Section 9.2.4: N. N. Lebedev, I. P. Skal’skaya, and Ya. S. Uflyand (1955), A. G. Sveshnikov
9.3. Helmholtz Equation ∆2 w + λw = −Φ(x) 809
TABLE 9.3
Conformal mapping of some domains D in the z-plane onto the unit circle |ω| ≤ 1.
Notation: z = x + iy, ω = u + iv, z0 = x0 + iy0 , and z̄0 = x0 − iy0
No Domain D in z-plane Transformation
z − z0
Upper half-plane: ω = eiλ ,
1 −∞ < x < ∞, 0 ≤ y < ∞ z − z̄0
λ is a real number
z − z0
A circle of unit radius: ω = eiλ ,
2 x2 + y 2 ≤ 1 1 − z̄0 z
λ is a real number
Exterior of a circle of radius a: a
3 ω=
x2 + y 2 ≥ a2 z
Infinite strip of width a: exp(πz/a) − exp(πz0 /a)
4 ω=
−∞ < x < ∞, 0 ≤ y ≤ a exp(πz/a) − exp(πz̄0 /a)
Semicircle of radius a: z 2 + 2az − a2

5
x2 + y 2 ≤ a2 , x ≥ 0 ω=i
z 2 − 2az − a2
Sector of a unit circle with angle β: (1 + z π/β )2 − i(1 − z π/β )2
6 ω=
|z| ≤ 1, 0 ≤ arg z ≤ β (1 + z π/β )2 + i(1 − z π/β )2

Exterior of an ellipse with semiaxes a and b: 1 a+b
7 z= (a − b)ω +
(x/a)2 + (y/b)2 ≥ 1 2 ω
9.3 Helmholtz Equation ∆2 w + λw = −Φ(x)

Many problems related to steady-state oscillations (mechanical, acoustical, thermal, elec-
tromagnetic, etc.) lead to the two-dimensional Helmholtz equation. For λ < 0, this equa-
tion describes mass transfer processes with volume chemical reactions of the first order.
Moreover, any elliptic equation with constant coefficients can be reduced to the Helmholtz
equation.
9.3.1 General Remarks, Results, and Formulas

◮ Some definitions.
The Helmholtz equation is called homogeneous if Φ = 0 and nonhomogeneous if Φ 6= 0.

A homogeneous boundary value problem is a boundary value problem for the homoge-
neous Helmholtz equation with homogeneous boundary conditions; a particular solution of
a homogeneous boundary value problem is w = 0.
The values λn of the parameter λ for which there are nontrivial solutions (solutions
other than identical zero) of the homogeneous boundary value problem are called eigenval-
ues and the corresponding solutions, w = wn , are called eigenfunctions of the boundary
value problem.
In what follows, the first, second, and third boundary value problems for the two-
dimensional Helmholtz equation in a finite two-dimensional domain S with boundary L

are considered. For the third boundary value problem with the boundary condition
∂w
+ kw = 0 for r ∈ L,
∂N
∂w
it is assumed that k > 0. Here, ∂N is the derivative along the outward normal to the
contour L, and r = {x, y}.
◮ Properties of eigenvalues and eigenfunctions.

1◦ . There are infinitely many eigenvalues {λn }; the set of eigenvalues forms a discrete
spectrum for the given boundary value problem.
2◦ . All eigenvalues are positive, except for the eigenvalue λ0 = 0 existing in the second
boundary value problem (the corresponding eigenfunction is w0 = const). We number the
eigenvalues in order of increasing magnitude so that λ1 < λ2 < λ3 < · · · .
3◦ . The eigenvalues tend to infinity as the number n increases. The following asymptotic
estimate holds:
n S2
lim = ,
n→∞ λn 4π
where S2 is the area of the two-dimensional domain under study.
4◦ . The eigenfunctions wn = wn (x, y) are defined up to a constant multiplier. Any two
eigenfunctions corresponding to different eigenvalues, λn 6= λm , are orthogonal:
Z
wn wm dS = 0.
S
5◦ . Any twice continuously differentiable function f = f (r) that satisfies the boundary
conditions of a boundary value problem can be expanded into a uniformly convergent series
in the eigenfunctions of the boundary value problem:
∞
X Z Z
1 2
f= fn wn , where fn = f wn dS, kwn k = wn2 dS.
kwn k2 S S
n=1
If f is square summable, then the series converges in mean.

6◦ . The eigenvalues of the first boundary value problem do not increase if the domain is
extended.
Remark 9.5. In a two-dimensional problem, finitely many linearly independent eigenfunctions
(1) (2) (p)
wn , wn , . . . , wn generally correspond to each eigenvalue λn . These functions can always be
replaced by their linear combinations,
w̄n(j) = cj,1 wn(1) + · · · + cj,j−1 wn(j−1) + wn(j) , j = 1, 2, . . . , p,

(1) (2) (p)
so that the new eigenfunctions w̄n , w̄n , . . . , w̄n are pairwise orthogonal. Therefore, without
loss of generality, we assume that all the eigenfunctions are orthogonal.
◮ Nonhomogeneous Helmholtz equation with homogeneous boundary conditions.

Three cases are possible.
1◦ . If the equation parameter λ is not equal to any one of the eigenvalues, then there exists
the series solution
X∞ Z Z
An 1 2
w= wn , where An = 2
Φwn dS, kwn k = wn2 dS.
n=1
λn − λ kw n k S S
2◦ . If λ is equal to some eigenvalue, λ = λm , then the solution of the nonhomogeneous

problem exists only if the function Φ is orthogonal to wm , i.e.,
Z
Φwm dS = 0.
S
In this case the system is expressed as
m−1
X X∞ Z
An An 1
w= wn + wn + Cwm , An = Φwn dS,
λn − λm λn − λm kwn k2 S
n=1 n=m+1
Z
where kwn k2 = wn2 dS, and C is an arbitrary constant.
S
Z
◦
3 . If λ = λm and Φwm dS 6= 0, then the boundary value problem for the nonhomoge-
S
neous equation does not have solutions.
(j)
Remark 9.6. If pn mutually orthogonal eigenfunctions wn (j = 1, 2, . . . , pn ) correspond to
each eigenvalue λn , then, for λ 6= λn , the solution is written as
X∞ X pn (j) Z Z
An (j) (j) 1 2
w= wn , where An = (j) 2
Φwn(j) dS, kwn(j) k2 = wn(j) dS.
n=1 j=1 n
λ − λ kwn k S S
◮ Solution of nonhomogeneous boundary value problem of general form.

1◦ . The solution of the first boundary value problem for the Helmholtz equation with the
boundary condition
w = f (r) for r ∈ L
Z Z
∂
w(r) = Φ(ρ)G(r, ρ) dSρ − f (ρ) G(r, ρ) dLρ . (1)
S L ∂N ρ
∂
Here, r = {x, y} and ρ = {ξ, η} (r ∈ S, ρ ∈ S); ∂N ρ
denotes the derivative along the outward
normal to the contour L with respect to the variables ξ and η. The Green’s function is given
by the series
X∞
wn (r)wn (ρ)
G(r, ρ) = , λ 6= λn , (2)
kwn k2 (λn − λ)
n=1
where the wn and λn are the eigenfunctions and eigenvalues of the homogeneous first
2◦ . The solution of the second boundary value problem with the boundary condition
∂w
= f (r) for r∈L
∂N
can be written as
Z Z
w(r) = Φ(ρ)G(r, ρ) dSρ + f (ρ)G(r, ρ) dLρ . (3)
S L
Here, the Green’s function is given by the series

∞
X wn (r)wn (ρ)
1
G(r, ρ) = − + , λ 6= λn , (4)
S2 λ n=1 kwn k2 (λn − λ)
where S2 is the area of the two-dimensional domain under consideration, and the λn and wn
are the positive eigenvalues and the corresponding eigenfunctions of the homogeneous sec-
ond boundary value problem. For clarity, the term corresponding to the zero eigenvalue
λ0 = 0 (w0 = const) is singled out in (4).
3◦ . The solution of the third boundary value problem for the Helmholtz equation with the
boundary condition
∂w
+ kw = f (r) for r ∈ L
∂N
is given by formula (3), where the Green’s function is defined by series (2), which involves
the eigenfunctions wn and eigenvalues λn of the homogeneous third boundary value prob-
lem.
◮ Boundary conditions at infinity in the case of an infinite domain.
In what follows, the function Φ is assumed to be finite or sufficiently rapidly decaying as

r → ∞.
1◦ . For λ < 0, in the case of an infinite domain, the vanishing condition of the solution at
infinity is set,
w → 0 as r → ∞.
2◦ . For λ > 0, if the domain is unbounded, the radiation conditions (Sommerfeld condi-
tions) at infinity are used. In two-dimensional problems, these conditions are written as

√ √ ∂w √
lim r w = const, lim r + i λ w = 0, (5)
r→∞ r→∞ ∂r
where i2 = −1.
To identify a single solution, the principle of limit absorption and the principle of limit
amplitude are also used.

A two-dimensional nonhomogeneous Helmholtz equation in the rectangular Cartesian sys-
tem of coordinates has the form
∂2w ∂2w
+ + λw = −Φ(x, y).
∂x2 ∂y 2
1◦ . Particular solutions of the homogeneous equation (Φ ≡ 0):
w = (Ax + B)(C cos µy + D sin µy), λ = µ2 ,

w = (Ax + B)(C cosh µy + D sinh µy), λ = −µ2 ,
w = (A cos µx + B sin µx)(Cy + D), λ = µ2 ,
w = (A cosh µx + B sinh µx)(Cy + D), λ = −µ2 ,
w = (A cos µ1 x + B sin µ1 x)(C cos µ2 y + D sin µ2 y), λ = µ21 + µ22 ,
w = (A cos µ1 x + B sin µ1 x)(C cosh µ2 y + D sinh µ2 y), λ = µ21 − µ22 ,
w = (A cosh µ1 x + B sinh µ1 x)(C cos µ2 y + D sin µ2 y), λ = −µ21 + µ22 ,
w = (A cosh µ1 x + B sinh µ1 x)(C cosh µ2 y + D sinh µ2 y), λ = −µ21 − µ22 ,
where A, B, C, and D are arbitrary constants.

1
E (x, y) = K0 (sr) if λ = −s2 < 0,
2π
i (1)
E (x, y) = H0 (kr) if λ = k 2 > 0,
4
i (2)
E (x, y) = − H0 (kr) if λ = k2 > 0,
4
p (1)
where r = x2 + y 2 , K0 (z) is the modified Bessel function of the second kind, H0 (z)
(2)
and H0 (z) are the Hankel functions of the first and second kind of order 0, and i2 = −1.
The leading term of the asymptotic expansion of the fundamental solutions, as r → 0, is
1
given by 2π ln 1r .
3◦ . Suppose w = w(x, y) is a solution of the homogeneous Helmholtz equation. Then the
functions
w1 = w(x + C1 , ±y + C2 ),
w2 = w(−x + C1 , ±y + C2 ),
w3 = w(x cos θ + y sin θ + C1 , −x sin θ + y cos θ + C2 ),
where C1 , C2 , and θ are arbitrary constants, are also solutions of the equation.
◮ Domain: −∞ < x < ∞, −∞ < y < ∞.

1◦ . Solution for λ = −s2 < 0:
Z ∞Z ∞ p
1
w(x, y) = Φ(ξ, η)K0 (s̺) dξ dη, ̺= (x − ξ)2 + (y − η)2 .
2π −∞ −∞
2◦ . Solution for λ = k2 > 0:

Z Z p
i ∞ ∞ (2)
w(x, y) = − Φ(ξ, η)H0 (k̺) dξ dη, ̺= (x − ξ)2 + (y − η)2 .
4 −∞ −∞
The radiation conditions (Sommerfeld conditions) at infinity were used to obtain this solu-
tion (see Section 9.3.1, condition (5)).
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980), A. N. Tikhonov and A. A. Samarskii
(1990).

w = f (x) at y = 0.
Solution:
Z ∞ Z ∞Z ∞
∂
w(x, y) = f (ξ) G(x, y, ξ, η) dξ + Φ(ξ, η)G(x, y, ξ, η) dξ dη.
−∞ ∂η η=0 0 −∞
1◦ . The Green’s function for λ = −s2 < 0:

1
G(x, y, ξ, η) =
K0 (s̺1 ) − K0 (s̺2 ) ,
p 2π p
̺1 = (x − ξ) + (y − η)2 , ̺2 = (x − ξ)2 + (y + η)2 .
2
2◦ . The Green’s function for λ = k2 > 0:

i (2) (2)
G(x, y, ξ, η) = − H0 (k̺1 ) − H0 (k̺2 ) .
4
The radiation conditions at infinity were used to obtain this relation (see Section 9.3.1,
condition (5)).

∂y w = f (x) at y = 0.
Solution:
Z ∞ Z ∞Z ∞
w(x, y) = − f (ξ)G(x, y, ξ, 0) dξ + Φ(ξ, η)G(x, y, ξ, η) dξ dη.
−∞ 0 −∞
1
G(x, y, ξ, η) =
K0 (s̺1 ) + K0 (s̺2 ) ,
p 2π p
̺1 = (x − ξ) + (y − η)2 , ̺2 = (x − ξ)2 + (y + η)2 .
2
i (2) (2)
G(x, y, ξ, η) = − H0 (k̺1 ) + H0 (k̺2 ) .
4
The radiation conditions at infinity were used to obtain this relation (see Section 9.3.1,
condition (5)).
w = f1 (y) at x = 0, w = f2 (x) at y = 0.
Solution:
Z ∞ Z ∞
∂ ∂
w(x, y) = f1 (η) G(x, y, ξ, η) dη + f2 (ξ) G(x, y, ξ, η) dξ
0 ∂ξ ξ=0 0 ∂η η=0
Z ∞Z ∞
+ Φ(ξ, η)G(x, y, ξ, η) dξ dη.
0 0
1
G(x, y, ξ, η) = K0 (s̺1 ) − K0 (s̺2 ) − K0 (s̺3 ) + K0 (s̺4 ) ,
p 2π p
̺1 = (x − ξ)2 + (y − η)2 , ̺2 = (x − ξ)2 + (y + η)2 ,
p p
̺3 = (x + ξ)2 + (y − η)2 , ̺4 = (x + ξ)2 + (y + η)2 .
i (2) (2) (2) (2)

G(x, y, ξ, η) = − H0 (k̺1 ) − H0 (k̺2 ) − H0 (k̺3 ) + H0 (k̺4 ) .
4

∂x w = f1 (y) at x = 0, ∂y w = f2 (x) at y = 0.
Solution:
Z ∞ Z ∞
w(x, y) = − f1 (η)G(x, y, 0, η) dη − f2 (ξ)G(x, y, ξ, 0) dξ
0 0
Z ∞Z ∞
0 0

1
G(x, y, ξ, η) = K0 (s̺1 ) + K0 (s̺2 ) + K0 (s̺3 ) + K0 (s̺4 ) ,
p 2π p
̺1 = (x − ξ)2 + (y − η)2 , ̺2 = (x − ξ)2 + (y + η)2 ,
p p
̺3 = (x + ξ)2 + (y − η)2 , ̺4 = (x + ξ)2 + (y + η)2 .

i (2) (2) (2) (2)
G(x, y, ξ, η) = − H0 (k̺1 ) + H0 (k̺2 ) + H0 (k̺3 ) + H0 (k̺4 ) .
4

w = f1 (x) at y = 0, w = f2 (x) at y = a.
Solution:
Z ∞ Z ∞
∂ ∂
w(x, y) = f1 (ξ) G(x, y, ξ, η) dξ − f2 (ξ) G(x, y, ξ, η) dξ
−∞ ∂η η=0 −∞ ∂η η=a
Z aZ ∞
0 −∞
Green’s function:
∞
1X 1 πn p
G(x, y, ξ, η)= exp −βn|x−ξ| sin(qny) sin(qnη), qn = , βn = qn2 −λ.
a βn a
n=1
Alternatively, the Green’s function for λ = −s2 < 0 can be represented as

∞
1 X
G(x, y, ξ, η) = K0 (s̺1n ) − K0 (s̺2n ) ,
2π n=−∞
p p
̺n1 = (x − ξ)2 + (y − η − 2na)2 , ̺n2 = (x − ξ)2 + (y + η + 2na)2 .

Solution:
Z ∞ Z ∞
w(x, y) = − f1 (ξ)G(x, y, ξ, 0) dξ + f2 (ξ)G(x, y, ξ, a) dξ
Z a−∞
Z ∞ −∞

0 −∞
Green’s function:
∞
1 X εn
G(x, y, ξ, η) = exp −βn |x − ξ| cos(qn y) cos(qn η),
2a β
n=0 n
(
πn p 1 for n = 0,
qn = , βn = qn2 − λ, εn =
a 2 for n 6= 0.
Alternatively, the Green’s function for λ = −s2 < 0 can be represented as

∞
1 X
G(x, y, ξ, η) = K0 (s̺1n ) + K0 (s̺2n ) ,
2π n=−∞
p
̺n1 = (x − ξ)2 + (y − ηn1 )2 , ηn1 = 2na + η,
p
̺n2 = (x − ξ)2 + (y − ηn2 )2 , ηn2 = 2na − η.
◮ Domain: −∞ < x < ∞, 0 ≤ y ≤ a. Third boundary value problem.

∂y w − k1 w = f1 (x) at y = 0, ∂y w + k2 w = f2 (x) at y = a.
∞
1 X ϕn (y)ϕn (η) p
G(x, y, ξ, η) = exp −β n |x − ξ| , β n = µ2n − λ,
2 kϕn k2 βn
n=1

2 1 2 2 (k1 + k2 )(µ2n + k1 k2 )
ϕn (y) = µn cos(µn y)+k1 sin(µn y), kϕn k = (µn +k1 ) a+ .
2 (µ2n + k12 )(µ2n + k22 )
(k1 + k2 )µ
µ2 − k1 k2
◮ Domain: −∞ < x < ∞, 0 ≤ y ≤ a. Mixed boundary value problem.

w = f1 (x) at y = 0, ∂y w = f2 (x) at y = a.
Solution:
Z ∞ Z ∞
∂
w(x, y) = f1 (ξ) G(x, y, ξ, η) dξ + f2 (ξ)G(x, y, ξ, a) dξ
−∞ ∂η η=0 −∞
Z aZ ∞
+ Φ(ξ, η)G(x, y, ξ, η) dξ dη,
0 −∞
where
∞
1X 1
G(x, y, ξ, η) = exp −βn |x − ξ| sin(qn y) sin(qn η),
a β
n=0 n
π(2n + 1) p
qn = , βn = qn2 − λ.
2a

Solution:
Z aZ ∞ Z a
∂
w(x, y) = Φ(ξ, η)G(x, y, ξ, η) dξ dη + f1 (η) G(x, y, ξ, η) dη
0 0 0 ∂ξ ξ=0
Z ∞ Z ∞
∂ ∂
+ f2 (ξ) G(x, y, ξ, η) dξ − f3 (ξ) G(x, y, ξ, η) dξ,
0 ∂η η=0 0 ∂η η=a
where
∞
1X 1
G(x, y, ξ, η) = exp −βn |x − ξ| − exp −βn |x + ξ| sin(qn y) sin(qn η),
a β
n=1 n
πn p
qn = , βn = qn2 − λ.
a
◮ Domain: 0 ≤ x < ∞, 0 ≤ y ≤ a. Second boundary value problem.

∂x w = f1 (y) at x = 0, ∂y w = f2 (x) at y = 0, ∂y w = f3 (x) at y = a.
Solution:
Z aZ ∞ Z a
w(x, y) = Φ(ξ, η)G(x, y, ξ, η) dξ dη − f1 (η)G(x, y, 0, η) dη
0 0 0
Z ∞ Z ∞
− f2 (ξ)G(x, y, ξ, 0) dξ + f3 (ξ)G(x, y, ξ, a) dξ,
0 0
where
∞
1 X εn
G(x, y, ξ, η) = exp −βn |x − ξ| + exp −βn |x + ξ| cos(qn y) cos(qn η),
2a β
n=0 n
(
πn p 1 for n = 0,
qn = , βn = qn2 − λ, ε =
a 2 for n 6= 0.
◮ Domain: 0 ≤ x < ∞, 0 ≤ y ≤ a. Third boundary value problem.

∂xw−k1w=f1(y) at x=0, ∂y w−k2w=f2(x) at y=0, ∂yw+k3w=f3(x) at y=a.
∞
X p
ϕn (y)ϕn (η)
G(x, y, ξ, η) = Hn (x, ξ), βn = µ2n − λ,
kϕn k2 βn (βn + k1 )
n=1

1 2 2 2 (k2 + k3 )(µ2n + k2 k3 )
ϕn (y) = µn cos(µn y)+k2 sin(µn y), kϕn k = (µn +k2 ) a+ ,
2 (µ2n + k22 )(µ2n + k32 )
(
exp(−βn x) βn cosh(βn ξ) + k1 sinh(βn ξ) for x > ξ,
Hn (x, ξ) =
exp(−βn ξ) βn cosh(βn x) + k1 sinh(βn x) for ξ > x.
(k2 + k3 )µ
µ2 − k2 k3
◮ Domain: 0 ≤ x < ∞, 0 ≤ y ≤ a. Mixed boundary value problems.

w = f1 (y) at x = 0, ∂y w = f2 (x) at y = 0, ∂y w = f3 (x) at y = a.
Solution:
Z
a Z ∞
∂
w(x, y) = f1 (η) G(x, y, ξ, η) dη − f2 (ξ)G(x, y, ξ, 0) dξ
0 ∂ξ ξ=0 0
Z ∞ Z aZ ∞
+ f3 (ξ)G(x, y, ξ, a) dξ + Φ(ξ, η)G(x, y, ξ, η) dξ dη,
0 0 0
where
∞
1 X εn
G(x, y, ξ, η) = exp −βn |x − ξ| − exp −βn |x + ξ| cos(qn y) cos(qn η),
2a n=0 βn
(
πn p 1 for n = 0,
qn = , βn = qn2 − λ, ε =
a 2 for n 6= 0.
∂x w = f1 (y) at x = 0, w = f2 (x) at y = 0, w = f3 (x) at y = a.
Solution:
Z a Z ∞
∂
w(x, y) = − f1 (η)G(x, y, 0, η) dη + f2 (ξ) G(x, y, ξ, η) dξ
0 0 ∂η η=0
Z ∞ Z aZ ∞
∂
0 ∂η η=a 0 0
where
∞
1X 1
G(x, y, ξ, η) = exp −βn |x − ξ| + exp −βn |x + ξ| sin(qn y) sin(qn η),
a β
n=1 n
πn p
qn = , βn = qn2 − λ.
a

w = f1 (y) at x = 0, w = f2 (y) at x = a,
w = f3 (x) at y = 0, w = f4 (x) at y = b.
1◦ . Eigenvalues of the one-dimensional problem (it is convenient to label them with a

double subscript):
2
2 n m2
λnm = π + 2 ; n = 1, 2, . . . ; m = 1, 2, . . .
a2 b
Eigenfunctions and the norm squared:

nπx mπy ab
wnm = sin sin , kwnm k2 = .
a b 4
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), B. M. Budak, A. A. Samarskii,

2◦ . Solution for λ 6= λnm :

Z aZ b
0 0
Z b Z b
∂ ∂
+ f1 (η) G(x, y, ξ, η) dη − f2 (η) G(x, y, ξ, η) dη
0 ∂ξ ξ=0 0 ∂ξ ξ=a
Z a Z a
∂ ∂
+ f3 (ξ) G(x, y, ξ, η) dξ − f4 (ξ) G(x, y, ξ, η) dξ.
0 ∂η η=0 0 ∂η η=b

∞ ∞
G(x, y, ξ, η) = Hn (y, η) = Qm (x, ξ),
a n=1 βn sinh(βn b) b m=1 µm sinh(µm a)
where
πn p πm p
pn = , βn = p2n − λ, qm = , µm = q m 2 − λ,
a ( b
sinh(βn η) sinh[βn (b − y)] for b ≥ y > η ≥ 0,
Hn (y, η) =
sinh(βn y) sinh[βn (b − η)] for b ≥ η > y ≥ 0,
(
sinh(µm ξ) sinh[µm (a − x)] for a ≥ x > ξ ≥ 0,
Qm (x, ξ) =
sinh(µm x) sinh[µm (a − ξ)] for a ≥ ξ > x ≥ 0.
Alternatively, the Green’s function can be written as the double series

∞ ∞
4 X X sin(pnx) sin(qmy) sin(pnξ) sin(qmη) πn πm
G(x, y, ξ, η)= , pn = , qm = .
ab p2n+qm
2 −λ a b
n=1 m=1
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b. Second boundary value problem.

∂x w = f1 (y) at x = 0, ∂x w = f2 (y) at x = a,
∂y w = f3 (x) at y = 0, ∂y w = f4 (x) at y = b.
1◦ . Eigenvalues of the homogeneous problem:

2
n m2
λnm = π 2 + ; n = 0, 1, 2, . . . ; m = 0, 1, 2, . . .
a2 b2

(
nπx mπy ab 1 if n=0,
wnm =cos cos , kwnmk2 = (1+δn0)(1+δm0), δn0 =
a b 4 0 if n6= 0.

2◦ . Solution for λ 6= λnm :

Z aZ b
0 0
Z b Z b
− f1 (η)G(x, y, 0, η) dη + f2 (η)G(x, y, a, η) dη
0 0
Z a Z a
− f3 (ξ)G(x, y, ξ, 0) dξ + f4 (ξ)G(x, y, ξ, b) dξ.
0 0

∞
1 X εn cos(pn x) cos(pn ξ)
G(x, y, ξ, η) = Hn (y, η)
a n=0 βn sinh(βn b)
∞
1 X εm cos(qm y) cos(qm η)
= Qm (x, ξ),
b µm sinh(µm a)
m=0
where
(
πn cosh(βn η) cosh[βn (b − y)] for y > η,
pn = , Hn (y, η) =
a cosh(βn y) cosh[βn (b − η)] for η > y,
(
πm cosh(µm ξ) cosh[µm (a − x)] for x > ξ,
qm = , Qm (x, ξ) =
b cosh(µm x) cosh[µm (a − ξ)] for ξ > x,
(
p p 1 for n = 0,
βn = p2n − λ, µm = qm 2 − λ, εn =
2 for n 6= 0.
The Green’s function can also be written as the double series

∞ ∞
1 X X εn εm cos(pn x) cos(qm y) cos(pn ξ) cos(qm η)
G(x, y, ξ, η) = ,
ab n=0 m=0 p2n + qm
2 −λ
πn πm
pn = , qm = .
a b
◆ Only the eigenvalues and eigenfunctions of homogeneous boundary value problems for
the homogeneous Helmholtz equation (with Φ ≡ 0) are presented below. The solutions of
the corresponding nonhomogeneous boundary value problems (with Φ 6≡ 0) can be con-
structed by the relations specified in Section 9.3.1.

∂x w − k1 w = 0 at x = 0, ∂x w + k2 w = 0 at x = a,
∂y w − k3 w = 0 at y = 0, ∂y w + k4 w = 0 at y = b.
Eigenvalues:
λnm = µ2n + νm
2
,
(k1 + k2 )µ (k3 + k4 )ν
tan(µa) = , tan(νb) = .
µ2 − k1 k2 ν 2 − k3 k4
Eigenfunctions:
wnm = (µn cos µn x + k1 sin µn x)(νm cos νm y + k3 sin νm y).

The square of the norm of an eigenfunction:

2 1 2 2 2 2 (k1 + k2 )(µ2n + k1 k2 ) (k3 + k4 )(νm2 +k k )
3 4
kwnm k = (µn +k1 )(νm +k3 ) a+ b+ 2 .
4 (µ2n + k12 )(µ2n + k22 ) (νm + k32 )(νm
2 + k2 )
4
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b. Mixed boundary value problems.

w = 0 at x = 0, w = 0 at x = a,
∂y w = 0 at y = 0, ∂y w = 0 at y = b.
Eigenvalues:

2 n2 m2
λnm = π + 2 ; n = 1, 2, 3, . . . ; m = 0, 1, 2, . . .
a2 b

(
nπx mπy 2 ab 1 for m = 0,
wnm = sin cos , kwnm k = (1 + δm0 ), δm0 =
a b 4 0 for m 6= 0.
w = 0 at x = 0, ∂x w = 0 at x = a,
w = 0 at y = 0, ∂y w = 0 at y = b.
Eigenvalues:

π 2 (2n + 1)2 (2m + 1)2
λnm = + ; n = 0, 1, 2, . . . ; m = 0, 1, 2, . . .
4 a2 b2

π(2n + 1)x π(2m + 1)y ab
wnm = sin sin , kwnm k2 = .
2a 2b 4
◮ First boundary value problem for a triangular domain.

The sides of the triangle are defined by the equations
x = 0, y = 0, y = a − x.
The unknown quantity is zero for these sides.

Eigenvalues:
π2 2 2

λnm = (n + m) + m ; n = 1, 2, . . . ; m = 1, 2, . . .
a2
Eigenfunctions:

π π n π π
wnm = sin (n + m)x sin my − (−1) sin mx sin (n + m)y .
a a a a
◮ Second boundary value problem for a triangular domain.

The sides of the triangle are defined by the equations
x = 0, y = 0, y = a − x.
The normal derivative of the unknown quantity for these sides is zero.
Eigenvalues:
π2
λnm = 2
(n + m)2 + m2 ; n = 0, 1, . . . ; m = 0, 1, . . .
a
Eigenfunctions:

π π π π
wnm = cos (n + m)x cos my − (−1)n cos mx cos (n + m)y .
a a a a

A two-dimensional nonhomogeneous Helmholtz equation in the polar coordinate system is
written as
p
1 ∂ ∂w 1 ∂2w
r + 2 2
+ λw = −Φ(r, ϕ), r = x2 + y 2 .
r ∂r ∂r r ∂ϕ
◮ Particular solutions of the homogeneous equation (Φ ≡ 0):
w = [AJ0 (µr) + BY0 (µr)](Cϕ + D), λ = µ2 ,

w = [AI0 (µr) + BK0 (µr)](Cϕ + D), λ = −µ2 ,
w = [AJm (µr) + BYm (µr)](C cos mϕ + D sin mϕ), λ = µ2 ,
w = [AIm (µr) + BKm (µr)](C cos mϕ + D sin mϕ), λ = −µ2 ,
where m = 1, 2, . . . ; A, B, C, D are arbitrary constants; the Jm (µ) and Ym (µ) are Bessel
functions; and the Im (µ) and Km (µ) are modified Bessel functions.

w = 0 at r = R.
Eigenvalues:
µ2nm
λnm = ; n = 0, 1, 2, . . . ; m = 1, 2, 3, . . .
R2
Here, the µnm are positive zeros of the Bessel functions, Jn (µ) = 0.
Eigenfunctions:
p p
(1) (2)
wnm = Jn r λnm cos nϕ, wnm = Jn r λnm sin nϕ.
(1) √
Eigenfunctions possessing the axial symmetry property: w0m = J0 r λ0m .
The square of the norm of an eigenfunction is given by
(
(k) 2 1 2 ′ 2 1 for i = j,
kwnm k = 2 πR (1 + δn0 )[Jn (µnm )] , k = 1, 2; δij =
0 for i =
6 j.

∂r w = 0 at r = R.
Eigenvalues:
µ2nm
λnm = ,
R2
where the µnm are roots of the quadratic equation Jn′ (µ) = 0.
Eigenfunctions:
p p
(1) (2)
wnm = Jn (r λnm ) cos nϕ, wnm = Jn (r λnm ) sin nϕ.
Here, n = 0, 1, 2, . . . ; for n 6= 0, the parameter m assumes the values m = 1, 2, 3, . . . ;

for n = 0, a root µ00 = 0 (the corresponding eigenfunction is w00 = 1).
(1) √
Eigenfunctions possessing the axial symmetry property: w0m = J0 r λ0m .
(k) 2 π 2 R2 (1 + δn0 ) 2
kwnm k = (µnm − n2 )[Jn (µnm )]2 , kw00 k2 = πR2 ,
2µ2nm
(
1 for i = j,
where k = 1, 2; δij =
0 for i 6= j.

∂r w + kw = 0 at r = R.
Eigenvalues:
µ2nm
λnm = ; n = 0, 1, 2, . . . ; m = 1, 2, 3, . . .
R2
Here, the µnm is the mth root of the transcendental equation µJn′ (µ) + kRJn (µ) = 0.
Eigenfunctions:
p p
(1) (2)
wnm = Jn r λnm cos nϕ, wnm = Jn r λnm sin nϕ.

(
2 1 for i=j,
(1) 2 (2) 2 πR (1+δn0) 2 2
kwnm k =kwnm k = (k R +µ2nm−n2)[Jn(µnm)]2, δij =
2µ2nm 0 for i6= j.


w = 0 at r = R1 , w = 0 at r = R2 .
Eigenvalues:
λnm = µ2nm ; n = 0, 1, 2, . . . ; m = 1, 2, 3, . . .
Here, the µnm are positive roots of the transcendental equation
Jn (µR1 )Yn (µR2 ) − Jn (µR2 )Yn (µR1 ) = 0.
Eigenfunctions:
(1)
wnm = [Jn (µnm r)Yn (µnm R1 ) − Jn (µnm R1 )Yn (µnm r)] cos nϕ,
(2)
wnm = [Jn (µnm r)Yn (µnm R1 ) − Jn (µnm R1 )Yn (µnm r)] sin nϕ.

(
(1) 2 (2) 2 2(1 + δn0 ) Jn2 (µnm R1 ) − Jn2 (µnm R2 ) 1 for i = j,
kwnm k = kwnm k = , δij =
πµ2nm Jn2 (µnm R2 ) 0 for i =
6 j.


∂r w = 0 at r = R1 , ∂r w = 0 at r = R2 .
Eigenvalues:
λnm = µ2nm ; n = 0, 1, 2, . . . ; m = 0, 1, 2, . . .
Here, the µnm are roots of the quadratic equation
Jn′ (µR1 )Yn′ (µR2 ) − Jn′ (µR2 )Yn′ (µR1 ) = 0.

(1)
If n = 0, there is a root µ00 = 0 and the corresponding eigenfunction is w00 = 1.
Eigenfunctions:
(1)
wnm = [Jn (µnm r)Yn′ (µnm R1 ) − Jn′ (µnm R1 )Yn (µnm r)] cos nϕ,
(2)
wnm = [Jn (µnm r)Yn′ (µnm R1 ) − Jn′ (µnm R1 )Yn (µnm r)] sin nϕ.

′
(1) 2 (2) 2 2(1+δn0) n2 Jn(µnmR1) 2 n2
kwnmk = kwnmk = 1− 2 2 − 1− 2 2 ,
πµ2nm R2µnm Jn′ (µnmR2) R1µnm
(
(1) 2 2 2 1 for i = j,
kw00 k = π(R2 −R1); δij =
0 for i 6= j.

∂r w − kw = 0 at r = R1 , ∂r w + kw = 0 at r = R2 .
Eigenvalues:
λnm = µ2nm ; n = 0, 1, 2, . . . ; m = 1, 2, 3, . . . ;
where the µnm are positive roots of the transcendental equation
A1 (µR1 )B2 (µR2 ) − A2 (µR2 )B1 (µR1 ) = 0.
Here, we use the notation

k k
A1 (µR) = Jn′ (µR) − Jn (µR), B1 (µR) = Yn′ (µR) − Yn (µR),
µ µ
k k
A2 (µR) = Jn′ (µR) + Jn (µR), B2 (µR) = Yn′ (µR) + Yn (µR).
µ µ
Eigenfunctions:
(1)
wnm = [B1 (µnm R1 )Jn (µnm r) − A1 (µnm R1 )Yn (µnm r)] cos nϕ,
(2)
wnm = [B1 (µnm R1 )Jn (µnm r) − A1 (µnm R1 )Yn (µnm r)] sin nϕ.
The square of the norm of an eigenfunction is given by (s = 1, 2)

(s) 2 1 2
′ 2 n2 2
kwnm k = 2 πεn R2 Fnm (R2 ) + 1 − 2 2 Fnm (R2 )
R2 µnm

1 2
′ 2 n2 2
− 2 πεn R1 Fnm (R1 ) + 1 − 2 2 Fnm (R1 ) ,
R1 µnm
(
2 for i = j,
Fnm (r) = B1 (µnm R1 )Jn (µnm r) − A1 (µnm R1 )Yn (µnm r), εij =
1 for i 6= j.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ α. First boundary value problem.

w=0 at r = R, w = 0 at ϕ = 0, w = 0 at ϕ = α.
Eigenvalues:
µ2nm
λnm = ; n = 1, 2, 3, . . . ; m = 1, 2, 3, . . .
R2
Here, the µnm are positive zeros of the Bessel functions, J nπ (µ) = 0.
α
Eigenfunctions:
r nπ
wnm = J nπ µnm sin ϕ .
α R α
αR2 h ′ i2
kwnm k2 = J nπ (µnm ) .
4 α

◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ α. Second boundary value problem.

∂r w = 0 at r = R, ∂ϕ w = 0 at ϕ = 0, ∂ϕ w = 0 at ϕ = α.
Eigenvalues:
µ2nm
λnm = ; n = 0, 1, 2, . . . ; m = 0, 1, 2, . . .
R2
Here, the µnm are roots of the quadratic equation J ′nπ (µ) = 0.
α
Eigenfunctions:
r nπ
wnm = J nπ µnm cos ϕ , w00 = 1.
α R α

2 αR2 n2 h i2 αR2
kwnm k = (1 + δn0 ) 1 − 2 J nπ (µnm ) , kw00 k2 = .
4 µnm α 2
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ α. Third boundary value problem.

∂r w+k1 w = 0 at r = R, ∂ϕ w−k2 w = 0 at ϕ = 0, ∂ϕ w+k3 w = 0 at ϕ = α.
Eigenvalues:
µ2nm
λnm = ; n = 1, 2, 3, . . . ; m = 1, 2, 3, . . .
R2
Here, the µnm are positive roots of the transcendental equation µJν′ n (µ) + k1 RJνn (µ) = 0;
(k2 + k3 )ν
the νn are positive roots of the transcendental equation tan(αν) = 2 .
ν − k2 k3
Eigenfunctions:

r νn cos(νn ϕ) + k2 sin(νn ϕ)
wnm = Jνn µnm p .
R νn2 + k22

2 R2 (k2 + k3 )(νn2 + k2 k3 ) k12 R2 − νn2
kwnm k = α+ 1+ Jν2n (µnm ).
4 (νn2 + k22 )(νn2 + k32 ) µ2nm
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ α. First boundary value problem.

Boundary conditions are prescribed:
w=0 at r = R1 , w = 0 at r = R2 ,
w=0 at ϕ = 0, w = 0 at ϕ = α.
Eigenvalues:
λnm = µ2nm ,
where the µnm are positive roots of the transcendental equation

nπ
Jνn (µR1 )Yνn (µR2 ) − Jνn (µR2 )Yνn (µR1 ) = 0, νn = .
α
Eigenfunctions:

wnm = Jνn (µnm r)Yνn (µnm R1 ) − Jνn (µnm R1 )Yνn (µnm r) sin(νn ϕ).
2 2
2 α Jνn (µnm R1 ) − Jνn (µnm R2 )
kwnm k = 2 2 2 .
π µnm Jνn (µnm R2 )
9.3.4 Other Orthogonal Coordinate Systems. Elliptic Domain

In the first and second paragraphs below, two other orthogonal systems of coordinates are
described in which the homogeneous Helmholtz equation admits separation of variables.
◮ Parabolic coordinate system.

In the parabolic coordinates that are introduced by the relations
x = 12 (ξ 2 − η 2 ), y = ξη (0 ≤ ξ < ∞, −∞ < η < ∞),
the Helmholtz equation has the form
∂2w ∂2w
+ + λ(ξ 2 + η 2 )w = 0.
∂ξ 2 ∂η 2
Setting w = f (ξ)g(η), we arrive at the following linear ordinary differential equations
for f = f (ξ) and g = g(η):
f ′′ + (λξ 2 + k)f = 0, g′′ + (λη 2 − k)g = 0,
where k is the separation constant. The general solutions of these equations are given by
f (ξ)=A1Dµ−1/2(σξ)+A2Dµ−1/2(−σξ), g(η)=B1D−µ−1/2(ση)+B2D−µ−1/2(−ση),
µ= 12 k(−λ)−1/2, σ=(−4λ)1/4.
Here, A1 , B1 , A2 , and B2 are arbitrary constants, and Dν (z) is the parabolic cylinder
function,

1/2 1 2
Γ 12 ν 1 1 2
−1/2 Γ − 12 1 ν 3 1 2

Dν (z) = 2 exp − 4 z Φ − 2 , 2 ; 2 z +2 zΦ 2 − 2 , 2 ; 2 z .
Γ 12 − ν2 Γ − ν2
For ν = n = 0, 1, 2, . . . , we have
dn
Dn(z)=2−n/2 exp − 14 z 2 Hn 2−1/2z , where Hn(z)=(−1)n exp z 2 n exp −z 2 .
z
⊙ Literature: M. Abramowitz and I. Stegun (1964), W. Miller, Jr. (1977).
◮ Elliptic coordinate system.

In the elliptic coordinates that are introduced by the relations
x = a cosh u cos v, y = a sinh u sin v (0 ≤ u < ∞, 0 ≤ v < 2π, a > 0),
the Helmholtz equation is expressed as
∂2w ∂2w
+ + a2 λ(cosh2 u − cos2 v)w = 0.
∂u2 ∂v 2
Setting w = F (u)G(v), we arrive at the following linear ordinary differential equations
for F = F (u) and G = G(v):

F ′′ + 12 a2 λ cosh 2u − k F = 0, G′′ − 12 a2 λ cos 2v − k G = 0,
where k is the separation constant. The solutions of these equations periodic in v are given
by ( (
Cen (u, q), cen (v, q),
F (u) = G(v) = q = 14 a2 λ,
Sen (u, q), sen (v, q),
where Cen (u, q) and Sen (u, q) are the modified Mathieu functions, while cen (v, q) and
sen (v, q) are the Mathieu functions; to each value of q there is a corresponding k = kn (q).
⊙ Literature: M. Abramowitz and I. Stegun (1964), W. Miller, Jr. (1977).
◮ Domain: (x/a)2 + (y/b)2 ≤ 1. First boundary value problem.

The unknown quantity is zero at the boundary of the elliptic domain:
w = 0 if (x/a)2 + (y/b)2 = 1 (a ≥ b).
The first three eigenvalues and eigenfunctions are given by the approximate relations
2

γ10 1 1
λ1 = + 2 , w1 (R) = J0 (γ10 R),
2 a2 b
2 3
γ11 1

(c) (c)
λ2 = 2
+ 2 , w2 (R, ϕ) = J1 (γ11 R) cos ϕ,
4 a b
2
(s) γ11 1 3 (s)
λ2 = + 2 , w2 (R, ϕ) = J1 (γ11 R) sin ϕ,
4 a2 b
where γ10 = 2.4048 and γ11 = 3.8317 are pthe first roots of the Bessel functions J0 and J1 ,
i.e., J0 (γ10 ) = 0 and J1 (γ11 ) = 0; R = (x/a)2 + (y/b)2 .
The above relations were obtained using the generalized (nonorthogonal) polar coordi-
nates R, ϕ defined by
x = aR cos ϕ, y = bR sin ϕ (0 ≤ R ≤ 1, 0 ≤ ϕ ≤ 2π)
and the variational method.

p
For ε = 1 − (b/a)2 ≤ 0.9, the above formulas provide an accuracy of 1% for λ1
(c) (s) (c)
and 2% for λ2 and λ2 . For ε ≤ 0.5, the errors in calculating λ1 and λ2 do not exceed
(s)
0.01%, and the maximum error in determining λ2 is 0.12%. In the limit case ε = 0 that
corresponds to a circular domain, the above formulas are exact.
⊙ Literature: L. D. Akulenko and S. V. Nesterov (2000).
9.4 Other Equations

9.4.1 Stationary Schrödinger Equation ∆2 w = f (x, y)w
∂2w ∂2w
1. + = a(x2 + y 2 )w.
∂x2 ∂y 2
The transformation
z = 12 (x2 − y 2 ), ζ = xy
leads to the Helmholtz equation
∂2w ∂2w
+ − aw = 0,
∂z 2 ∂ζ 2
∂2w ∂2w
2. + = a(x2 + y 2 )2 w.
∂x2 ∂y 2
The transformation
z = 13 x3 − xy 2 , ζ = x2 y − 13 y 3
∂2w ∂2w
+ − aw = 0,
∂z 2 ∂ζ 2
∂2w ∂2w
3. + = a(x2 + y 2 )k w.
∂x2 ∂y 2
This is a special case of equation 9.4.1.7 for f (u) = auk . Table 9.4 presents transformations
that reduce this equation to the Helmholtz equation that is discussed in Section 9.3.2; the
sixth row involves the imaginary unit, i2 = −1.
∂2w ∂2w
4. + = aeβx w.
∂x2 ∂y 2
The transformation
1
1
1
1

u(x, y) = exp 2 βx cos 2 βy , v(x, y) = exp 2 βx sin 2 βy

∂2w ∂2w
+ = 4aβ −2 w,
∂u2 ∂v 2
TABLE 9.4
∂2w ∂2 w
Transformations reducing equation 9.4.1.3 to the Helmholtz equation ∂ξ2
+ ∂η 2
= bw
No Exponent k Transformation Factor b
1 2
1 k=1 ξ= 2 (x − y 2 ), η = xy b=a
1 3
2 k=2 ξ= 3x − xy 2 , η = x2 y − 13 y 3 b=a
y
3 k = −1 ξ= 1
2 ln(x2 + y 2 ), η = arctan b=a
x
x y
4 k = −2 ξ=− , η= 2 b=a
x2 + y 2 x + y2
5 k = − 12 x= 1 2
2 (ξ − η 2 ), y = ξη b = 2a
(x + iy)k+1 + (x − iy)k+1 (x + iy)k+1 − (x − iy)k+1

6 k = ±3, ±4, . . . ξ = , η= b=a
2(k + 1) 2(k + 1)i
k is any ρk+1 cos[(k + 1)ϕ] ρk+1 sin[(k + 1)ϕ]

7 ξ= , η= b=a
(k 6= −1) k+1 k+1
x = ρ cos ϕ, y = ρ sin ϕ
∂2w ∂2w
5. + = keax+by w.
∂x2 ∂y 2
The transformation
ξ = ax + by, η = bx − ay
∂2w ∂2w k
2
+ 2
= 2 eξ w.
∂ξ ∂η a + b2
∂2w ∂2w
6. + = f (ax + by)w.
∂x2 ∂y 2
This is a special case of equation 9.4.1.9 for g(u) = 0. Particular solutions:

w(x, y) = C1 cos[k(bx − ay)] + C2 sin[k(bx − ay)] ϕ(ax + by),
where C1 , C2 , and k are arbitrary constants, and the function ϕ = ϕ(ξ) is determined by
the ordinary differential equation

′′ 1 2
ϕξξ − 2 f (ξ) + k ϕ = 0.
a + b2
∂2w ∂2w
7. + = f (x2 + y 2 )w.
∂x2 ∂y 2
1◦ . This equation admits separation of variables in the polar coordinates ρ, ϕ (x = ρ cos ϕ,
y = ρ sin ϕ). Particular solution:

w(x, y) = C1 cos(kϕ) + C2 sin(kϕ) U (ρ),
where C1 , C2 , and k are arbitrary constants, and the function U = U (ρ) is determined by
the ordinary differential equation

ρ(ρUρ′ )′ρ − k2 + ρ2 f (ρ2 ) U = 0.
z = 12 (x2 − y 2 ), ζ = xy
√
∂2w ∂2w f (2 u )
+ = F (z 2 + ζ 2 )w, F (u) = √ .
∂z 2 ∂ζ 2 2 u
√
In the special case f (u) = 2a, we have F (u) = a/ u. For f (u) = bu3 , we obtain an
equation of the form 9.4.1.1 with F (u) = 4bu.
∂2w ∂2w
8. + = [f (x) + g(y)]w.
∂x2 ∂y 2
A particular separable solution:
w(x, y) = ϕ(x)ψ(y),
where the functions ϕ(x) and ψ(y) are determined by the second-order ordinary differential
equations
ϕ′′xx − [f (x) − C]ϕ = 0, ψyy′′
− [g(y) + C]ψ = 0,
∂2w ∂2w
9. + = [f (ax + by) + g(bx − ay)]w.
∂x2 ∂y 2
The transformation
ξ = ax + by, η = bx − ay

∂2w ∂2w f (ξ) g(η)
+ = 2 + 2 w.
∂ξ 2 ∂η 2 a + b2 a + b2
∂2w ∂2w
10. + = (x2 + y 2 )[f (x2 − y 2 ) + g(xy)]w.
∂x2 ∂y 2
The transformation
z = 12 (x2 − y 2 ), ζ = xy
∂2w ∂2w
+ = [f (2z) + g(ζ)]w.
∂z 2 ∂ζ 2
9.4.2 Convective Heat and Mass Transfer Equations

∂2w ∂2w ∂w
1. + =α .
∂x2 ∂y 2 ∂x
This is a convective heat and mass transfer equation. It describes a stationary temperature
(concentration) field in a continuous medium moving with a constant velocity along the
x-axis. In particular, it models convective-molecular heat transfer from a heated flat plate
in a flow of a thermal-transfer ideal fluid moving along the plate. This occurs, for example,
if a liquid-metal coolant flows past a flat plate or if a plate is in a seepage flow through a
granular medium.
In the sequel, it is assumed that the equation is written in dimensionless variables x, y
related to the characteristic length (for a flat plate of length 2h, the characteristic length is
taken to be h).

1◦ . The substitution w(x, y) = exp 12 αx U (x, y) brings the original equation to the Helm-
holtz equation
∂2U ∂2U 1
2
+ 2
= α2 U.
∂x ∂y 4
Particular solutions of this equation in Cartesian and polar coordinates can be found in
2◦ . In the elliptic coordinates
x = cosh ζ cos η, y = sinh ζ sin η
a wide class of particular solutions (vanishing as ζ → ∞) can be indicated; this class of

solutions of the original equation is represented in series form as
∞
1
X 1 2
w = exp 2 αx Am cem (η, −q) Fekm (ζ, −q), q = − 16 α ,
m=0
where the Am are arbitrary constants, the cem (η, −q) are the Mathieu functions, and the
Fekm (ζ, −q) are the modified Mathieu functions [e.g., see McLachlan (1947) and Bateman
and Erdélyi (1955)].
3◦ . Consider the first boundary value problem in the upper half-plane (−∞ < x < ∞,
0 ≤ y < ∞). We assume that the surface of a plate of finite length is maintained at a
constant temperature w0 and the medium has a temperature w∞ = const far away from the
plate:
w = w0 for y = 0, |x| < 1,
∂y w = 0 for y = 0, |x| > 1,
w → w∞ for x2 + y 2 → ∞.
The solution of this problem in the elliptic coordinates ζ, η (see Item 2◦ ) has the form
∞
1
X Fekm (ζ, −q)
w(η, ζ) = w∞ + (w0 − w∞ ) exp 2 α cos η cosh ζ Dm cem (η, −q) ,
m=0
Ferm (0, −q)
where
ce2n(0, −q) (2n) 1 ce2n+1(0, −q) (2n+1) 1

D2n = 2 A0 , D2n+1 = − αB1 , q = − α2.
ce2n(0, q) 2 ce2n+1(0, q) 16
(2n) (2n+1)
Here, the A0 and B1 are the coefficients in the series expansions of the Mathieu
functions; these can be found in McLachlan (1947).
4◦ . Consider the second boundary value problem in the upper half-plane (−∞ < x < ∞,
0 ≤ y < ∞). We assume that a thermal flux is prescribed on the surface of a plate of finite
length and the medium has a constant temperature far away from the plate:
∂y w = f (x) for y = 0, |x| < 1,

∂y w = 0 for y = 0, |x| > 1,
w → w∞ as x2 + y 2 → ∞.
The solution of this problem in the Cartesian coordinates has the form
Z
1 1 1 1 p
w(x, y) = w∞ − f (ξ) exp α(x − ξ) K α (x − ξ)2 + y 2 dξ,
2 0 2
π −1
where K0 (z) is the modified Bessel function of the second kind.

⊙ Literature: P. V. Cherpakov (1975), A. A. Borzykh and G. P. Cherepanov (1978).
∂2w ∂2w ∂w ∂w
2. + =α +β + γw.
∂x2 ∂y 2 ∂x ∂y
This equation describes a stationary temperature field in a medium moving with a constant
velocity, provided there is volume release heat (or absorption) proportional to temperature.
The substitution

w(x, y) = exp 12 (αx + βy) U (x, y)
brings the original equation to the Helmholtz equation
∂2U ∂2U
2
+ 2
= γ + 14 α2 + 14 β 2 U,
∂x ∂y
which is discussed in Sections 9.3.1 through 9.3.3.
∂2w ∂2w ∂w
3. + = Pe (1 − y 2 ) .
∂x2 ∂y 2 ∂x
The Graetz–Nusselt equation. It governs steady-state heat exchange in a laminar fluid flow
with a parabolic velocity profile in a plane channel. The equation is written in terms of the
dimensionless Cartesian coordinates x, y related to the channel half-width h; Pe = U h/a
is the Peclet number and U is the fluid velocity at the channel axis (y = 0). The walls of
the channel correspond to y = ±1.
w(y) = A + By,
w(x, y) = 12Ax + A Pe (6y 2 − y 4 ) + B,
Xm
λ2n
w(x, y) = An exp − x fn (y).
Pe
n=1
Here, A, B, An , and λn are arbitrary constants, and the functions fn are defined by

fn (y) = exp − 12 λn y 2 Φ αn , 12 ; λn y 2 , αn = 14 − 14 λn − 14 λ3n Pe−2 , (1)
∞
P α(α+1)...(α+k−1) ξ k
where Φ(α, β; ξ) = 1 + β(β+1)...(β+k−1) k! is the degenerate hypergeometric function.
k=1
2◦ . Let the walls of the channel be maintained at a constant temperature, w = 0 for x < 0
and w = w0 for x > 0. Due to the symmetry of the problem about the x-axis, it suffices to
consider only half of the domain, 0 ≤ y ≤ 1. The boundary conditions are written as
(
∂w 0 for x < 0,
y = 0, = 0; y = 1, w=
∂y w0 for x > 0;
x → −∞, w → 0; x → ∞, w → w0 .
The solution of the original equation under these boundary conditions is sought in the form
X∞ 2
µn
w(x, y) = w0 Bn exp x gn (y) for x < 0,
n=1
Pe
X∞
λ2n
w(x, y) = w0 1 − An exp − x fn (y) for x > 0.
Pe
n=1
The series coefficients must satisfy the matching conditions at the boundary:

w(x, y)x→0, x<0 − w(x, y)x→0, x>0 = 0,

∂x w(x, y)x→0, x<0 − ∂x w(x, y)x→0, x>0 = 0.
For x > 0, the function fn (y) is defined by relation (1), where the eigenvalues λn are
roots of the transcendental equation

Φ αn , 12 ; λn = 0, where αn = 14 − 14 λn − 14 λ3n Pe−2 .
For Pe → ∞, it is convenient to use the following approximate relation to identify

the λn :
λn = 4(n − 1) + 1.68 (n = 1, 2, 3, . . . ). (2)
The error of this formula does not exceed 0.2%. The corresponding numerical values of
the coefficients An are rather well approximated by the relations
A1 = 1.2, An = 2.27 (−1)n−1 λ−7/6

n for n = 2, 3, 4, . . . ,
whose maximum error is less than 0.1%, provided that the λn are calculated by (2).
For Pe → 0, the following asymptotic relations hold:
q 4(−1)n−1
λn = π n − 12 Pe, An = , fn (y) = cos π n− 12 y n = 1, 2, 3, . . .
π 2 (2n − 1)2
No results for x < 0 are given here, because they are of secondary importance in appli-
cations.
3◦ . Let a constant thermal flux be prescribed at the walls for x > 0 and let, for x < 0, the
walls be insulated from heat and the temperature vanishes as x → −∞. Then the boundary
conditions have the form
(
∂w ∂w 0 for x < 0,
y = 0, = 0; y = 1, = x → −∞, w → 0.
∂y ∂y q for x > 0;
In the domain of thermal stabilization, the asymptotic behavior of the solution (as x → ∞)
is as follows:

3 x 3 2 1 4 9 39
w(x, y) = q + y − y + − .
2 Pe 4 8 4 Pe2 280
⊙ Literature: L. Graetz (1883), W. Nusselt (1910), C. A. Deavours (1974), A. D. Polyanin, A. M. Kutepov,

A. V. Vyazmin, and D. A. Kazenin (2002).
∂2w 1 ∂w ∂2w ∂w
4. + + = Pe (1 − r 2 ) .
∂r2 r ∂r ∂z 2 ∂z
This equation governs steady-state heat exchange in a laminar fluid flow with parabolic
(Poiseuille’s) velocity profile in a circular tube. The equation is written in terms of the
dimensionless cylindrical coordinates x, y related to the tube radius R; Pe = U R/a is the
Peclet number and U is the fluid velocity at the tube axis (at r = 0). The walls of the tube
correspond to r = 1.
w(r) = A + B ln r,
w(r, z) = 16Az + A Pe (4r 2 − r 4 ) + B,
Xm
λ2n
w(r, z) = An exp − z fn (r).
n=1
Pe
Here, A, B, An , and λn are arbitrary constants, and the functions fn are defined by

fn (r) = exp − 12 λn r 2 Φ αn , 1; λn r 2 , αn = 1
2 − 14 λn − 14 λ3n Pe−2 , (1)
where Φ(α, β; ξ) is the degenerate hypergeometric function (see equation 9.4.2.3, Item 1◦ ).
2◦ . Let the tube wall be maintained at a constant temperature such that w = 0 for z < 0
and w = w0 for z > 0. The boundary conditions are written as
(
∂w 0 for z < 0,
r = 0, = 0; r = 1, w=
∂r w0 for z > 0;
z → −∞, w → 0; z → ∞, w → w0 .
The solution of the original equation under these boundary conditions is sought in the form
X∞ 2
µn
w(r, z) = w0 Bn exp z gn (r) for z < 0,
Pe
n=1
X∞
λ2n
w(r, z) = w0 1 − An exp − z fn (r) for z > 0.
n=1
Pe
The series coefficients must satisfy the matching conditions at the boundary,

w(r, z)z→0, z<0 − w(r, z)z→0, z>0 = 0,

∂z w(r, z)z→0, z<0 − ∂z w(r, z)z→0, z>0 = 0.
For z > 0, the functions fn (r) are defined by relations (1), where the eigenvalues λn
are roots of the transcendental equation

Φ αn , 1; λn = 0, where αn = 12 − 14 λn − 14 λ3n Pe−2 .
For Pe → ∞, it is convenient to use the following approximate relation to identify the
λn :
λn = 4 (n − 1) + 2.7 (n = 1, 2, 3, . . . ). (2)
The error of this formula does not exceed 0.3%. The corresponding numerical values of
the coefficients An are rather well approximated by the relations
An = 2.85 (−1)n−1 λ−2/3
n for n = 1, 2, 3, . . . ,
whose maximum error is 0.5%,
No results for z < 0 are given here, since they are of secondary importance in applica-
tions.
3◦ . Let a constant thermal flux be prescribed at the wall for z > 0 and let, for z < 0, the
tube surface be insulated from heat and the temperature vanishes as z → −∞. Then the
boundary conditions have the form
(
∂w ∂w 0 for z < 0,
r = 0, = 0; r = 1, = z → −∞, w → 0.
∂r ∂r q for z > 0;
In the domain of thermal stabilization, the asymptotic behavior of the solution (as z → ∞)
is as follows:
z 2 1 4 8 7
w(r, z) = q 4 +r − r + − .
Pe 4 Pe2 24
⊙ Literature: C. A. Deavours (1974), A. D. Polyanin, A. M. Kutepov, A. V. Vyazmin, and D. A. Kazenin
(2002).
∂2w ∂2w ∂w
5. + = f (y) .
∂x2 ∂y 2 ∂x
This equation describes steady-state heat exchange in a laminar fluid flow with an arbitrary
velocity profile f = f (y) in a plane channel.
Z y
w(x, y) = Ax + A (y − ξ)f (ξ) dξ + By + C, (1)
y0
m
X
w(x, y) = B + An exp(−βn x)un (y). (2)
n=1
Here, A, B, C, y0 , An , and βn are arbitrary constants, and the functions un = un (y) are
determined by the second-order linear ordinary differential equation
d2 un
2
+ βn f (y) + βn2 un = 0.
dy
2◦ . Solution (1) describes the temperature distribution far away from the inlet section of
the tube, in the domain of thermal stabilization, provided that a constant thermal flux is
prescribed at the channel walls.
2
∂ w ∂2w ∂w ∂w
6. a 2
+ 2
= v1 (x, y) + v2 (x, y) .
∂x ∂y ∂x ∂y
This is an equation of steady-state convective heat and mass transfer in the Cartesian co-
ordinate system. Here, v1 = v1 (x, y) and v2 = v2 (x, y) are the components of the fluid
velocity that are assumed to be known from the solution of the hydrodynamic problem.
1◦ . In plane problems of convective heat exchange in liquid metals modeled by an ideal
fluid, as well as in describing seepage (filtration) streams employing the model of potential
flows, the fluid velocity components v1 (x, y) and v2 (x, y) can be expressed in terms of the
potential ϕ = ϕ(x, y) and stream function ψ = ψ(x, y) as follows:
∂ϕ ∂ψ ∂ϕ ∂ψ
v1 = =− , v2 = = . (1)
∂x ∂y ∂y ∂x
The function ϕ is determined by solving the Laplace equation ∆ϕ = 0. In specific problems,
the potential ϕ and stream function ψ may be identified by invoking the complex variable
theory [e.g., see Lavrent’ev and Shabat (1973) and Sedov (1980)].
By passing in the convective heat exchange equation from x, y to the new variables
ϕ, ψ (Boussinesq transformation) and taking into account (1), we arrive a simpler equation
with constant coefficients of the form 9.4.2.1:
∂2w ∂2w 1 ∂w
2
+ = . (2)
∂ϕ ∂ψ 2 a ∂ϕ
The Boussinesq transformation brings any plane contour in a potential flow to a cut
in the ϕ-axis, simultaneously with the reduction of the original equation to the form (2).
Consequently, the heat transfer problem of a potential flow about this contour is reduced
to the heat exchange problem of a longitudinal flow of an ideal fluid past a flat plate (see
equation 9.4.2.1, Items 3◦ and 4◦ ).
2◦ . Asymptotic analyses of plane problems on heat/mass exchange of bodies of various

shape with laminar translational and shear flows of a viscous (and ideal) incompressible
fluid for large and small Peclet numbers were carried out in the references cited below. In
the thermal boundary layer approximation, the solution of the heat exchange problem for
a flat plate in a longitudinal translational flow of a viscous incompressible fluid at large
Reynolds numbers is presented in 3.9.1.4, Item 3◦ .
⊙ Literature: V. G. Levich (1962), P. V. Cherpakov (1975), A. A. Borzykh and G. P. Cherepanov (1978),
Yu. P. Gupalo, A. D. Polyanin, and Yu. S. Ryazantsev (1985), A. D. Polyanin, A. M. Kutepov, A. V. Vyazmin,
and D. A. Kazenin (2002).

1 ∂ ∂w 1 ∂ ∂w ∂w sin θ ∂w
7. r2 + sin θ = cos θ − .
r2 ∂r ∂r r2 sin θ ∂θ ∂θ ∂r r ∂θ
This is a special case of equation 9.4.1.8 with a = 1, vr = cos θ, and vθ = − sin θ. This
equation is obtained from the equation ∂xx w + ∂yy w + ∂zz w = ∂x w by the passage to the
spherical coordinate system in the axisymmetric case.
The general solution satisfying the decay condition (w → 0 as r → ∞) is expressed as
π 1/2 r cos θ X∞ r

w(r, θ) = exp An Kn+ 1 Pn (cos θ),
r 2 n=0
2 2
where the An are arbitrary constants. The Legendre polynomials Pn (ξ) and the modified
Bessel functions Kn+ 1 (z) are given by
2
1 dn 2 r π 1/2 r X
n
(n + m)!
Pn (ξ) = (ξ − 1)n , Kn+ 1 = exp − .
n! 2n dξ n 2 2 r 2 (n − m)! m! r m
m=0
⊙ Literature: P. L. Rimmer (1968).
h i
1 ∂ ∂w 1 ∂ ∂w ∂w v ∂w
8. a r2 + sin θ = vr + θ .
r2 ∂r ∂r r2 sin θ ∂θ ∂θ ∂r r ∂θ
This equation is often encountered in axisymmetric problems of convective heat and mass
exchange of solid particles, drops, and bubbles with a flow of a viscous incompressible
fluid. The fluid velocity components vr = vr (r, θ) and vθ = vθ (r, θ) can be expressed in
terms of the stream function ψ = ψ(r, θ) as
1 ∂ψ 1 ∂ψ
vr = , vθ = − . (1)
r 2 sin θ ∂θ r sin θ ∂r
Asymptotic analyses for a wide class of axisymmetric problems on heat/mass exchange
of solid particles, drops, and bubbles of various shape with a laminar translational or strain-
ing flow of a viscous incompressible fluid at large and small Peclet numbers Pe = U R/a
are performed in the books cited below. The Peclet number is written in terms of the charac-
teristic velocity U (e.g., the unperturbed fluid velocity far away from the particle in the case
of translation flow), the characteristic size of the particle R (e.g., the radius for a spherical
particle), and the thermal conductivity or diffusion coefficient a.
The following boundary conditions are usually specified:
w = w0 at r = R, w → w∞ as r → ∞, (2)
where R is the particle radius, w0 the temperature at the particle surface, and w∞ the
temperature far away from the particle (w0 and w∞ are constant).
Convective mass transfer problems are characterized by large Peclet numbers. To solve
such problems, the diffusion boundary layer approximation is often used; in this case, the
left-hand side of the equation takes into account only the diffusion mass transfer in the
normal direction to the particle surface (the tangential mass transfer is neglected). The con-
vective terms on the right-hand side are partially preserved—the fluid velocity components
are approximated by their leading terms of the asymptotic expansion near the phase sur-
face. Presented below are some important results obtained by solving the original equation
under the boundary conditions (2) in the diffusion boundary layer approximation.
Example 9.7. For the translational Stokes flow of a viscous incompressible fluid about a spher-
ical bubble, the stream function is expressed as
ψ(r, θ) = 1
2 U r(r − R) sin2 θ.
Here, U is the unperturbed fluid velocity in the incident flow, R the bubble radius (the value θ = π
corresponds to the front critical point at the bubble surface).
In this case, the solution of the convective heat/mass transfer equation with the boundary con-
ditions (2) for Pe = U R/a ≫ 1 in the diffusion boundary layer approximation is given by
q
3 r 1 − cos θ
w(r, θ) = w0 + (w∞ − w0 ) erf ξ, ξ= 8 Pe −1 √ ,
R 2 − cos θ
where erf ξ is the error function.
Example 9.8. For the translational Stokes flow of a viscous incompressible fluid about a solid
spherical particle, the stream function is expressed as

1 R
ψ(r, θ) = U (r − R)2 2 + sin2 θ.
4 r
Here, the notation is the same as in the case of a bubble above.

For a solid particle, the solution of the convective heat/mass transfer equation with the boundary
conditions (2) for Pe = U R/a ≫ 1 in the diffusion boundary layer approximation is given by
1
−1 1
Pe (r − R)3 sin3 θ
w(r, θ) = w0 + (w∞ − w0 ) Γ 3 γ 3,ξ , ξ= ,
3R3 π − θ + 12 sin 2θ
Z ξ
where Γ(β) is the gamma function and γ(β, ξ) = e−z z β−1 dz is the incomplete gamma function.
0
⊙ Literature: V. G. Levich (1962), Yu. P. Gupalo, A. D. Polyanin, and Yu. S. Ryazantsev (1985),
A. D. Polyanin, A. M. Kutepov, A. V. Vyazmin, and D. A. Kazenin (2002).
9.4.3 Equations of Heat and Mass Transfer in Anisotropic Media

∂ ∂w ∂ ∂w
1. axn + by m = 0.
∂x ∂x ∂y ∂y
This is a two-dimensional equation of the heat and mass transfer theory in a inhomoge-
neous anisotropic medium. Here, a1 (x) = axn and a2 (y) = by m are the principal thermal
diffusivities.
1◦ . Particular solutions (A, B, C are arbitrary constants):
w(x, y) = Ax1−n + By 1−m + C,

x2−n y 2−m
w(x, y) = A − + B,
a(2 − n) b(2 − m)
w(x, y) = Ax1−n y 1−m + B.
2◦ . For n 6= 2 and m 6= 2, there are particular solutions of the form

1/2
w = w(ξ), ξ = b(2 − m)2 x2−n + a(2 − n)2 y 2−m .
The function w = w(ξ) is determined by the ordinary differential equation
′′ A ′ 4 − nm
wξξ + w = 0, A= . (1)
ξ ξ (2 − n)(2 − m)
The general solution of equation (1) is given by

(
C1 ξ 1−A + C2 for A 6= 1,
w(ξ) =
C1 ln ξ + C2 for A = 1,

3◦ . There are multiplicative separable particular solutions in the form
w(x, y) = ϕ(x)ψ(y), (2)
where ϕ(x) and ψ(y) are determined by the following second-order linear ordinary differ-
ential equations (A1 is an arbitrary constant):
(axn ϕ′x )′x = −A1 ϕ, (3)

m
(by ψy′ )′y = A1 ψ. (4)
The solution of equation (3) is given by

 1−n h 2−n 2−n i

x 2 C1 Jν βx 2 + C2 Yν βx 2 for A1 > 0,
ϕ(x) = 1−n h 2−n 2−n i

x 2 C1 Iν βx 2 + C2 Kν βx 2 for A1 < 0,
r
|1 − n| 2 |A1 |
ν= , β= ,
2−n 2−n a
where C1 and C2 are arbitrary constants, Jν (z) and Yν (z) are Bessel functions, and Iν (z)
and Kν (z) are modified Bessel functions.
The solution of equation (4) is expressed as
 1−m h 2−m 2−m i

y 2 C1 Jσ µy 2 + C2 Yσ µy 2 for A1 < 0,
ψ(y) = 1−m h 2−m 2−m i

y 2 C1 Iσ µy 2 + C2 Kσ µy 2 for A1 > 0,
r
|1 − m| 2 |A1 |
σ= , µ= ,
2−m 2−m b
The sum of solutions of the form (2) corresponding to different values of the param-
eter A1 is also a solution of the original equation; the solutions of some boundary value
problems may be obtained by separation of variables.
4◦ . See equation 9.4.3.3, Item 4◦ , for c = 0.

∂ ∂w ∂ ∂w
2. axn + by m = c.
∂x ∂x ∂y ∂y
This is a two-dimensional equation of the heat and mass transfer theory with constant vol-
ume release of heat in an inhomogeneous anisotropic medium. Here, a1 (x) = axn and
a2 (y) = by m are the principal thermal diffusivities.
1/2
w = w(ξ), ξ = b(2 − m)2 x2−n + a(2 − n)2 y 2−m . (1)
′′ A ′
wξξ + w = B, (2)
ξ ξ
where
4 − nm 4c
A= , B= . (3)
(2 − n)(2 − m) ab(2 − n)2 (2 − m)2
 1−A + C + B 2
 C1 ξ
 2 2(A+1) ξ for A 6= ±1,
w(ξ) = C1 ln ξ + C2 + 14 Bξ 2 for A = 1,


C1 ξ 2 + C2 + 12 Bξ 2 ln ξ for A = −1,

c
w(x, y) = U (x, y) + x2−n
a(2 − n)
leads to a homogeneous equation of the form 9.4.3.1:

∂ n ∂U ∂ m ∂U
ax + by = 0.
∂x ∂x ∂y ∂y

∂ ∂w ∂ ∂w
3. axn + by m = cw.
∂x ∂x ∂y ∂y
This is a two-dimensional equation of the heat and mass transfer theory with a linear source
in an inhomogeneous anisotropic medium.
1/2
w = w(ξ), ξ = b(2 − m)2 x2−n + a(2 − n)2 y 2−m .
′′ A ′
wξξ + w = Bw, (1)
ξ ξ
where
4 − nm 4c
A= , B= .
(2 − n)(2 − m) ab(2 − n)2 (2 − m)2
1−A h
p p i
w(ξ) = ξ C1 Jν
ξ |B| + C2 Yν ξ |B|
2 for B < 0,
1−A h √ √ i
w(ξ) = ξ 2 C1 Iν ξ B + C2 Kν ξ B for B > 0,
where ν = 12 |1 − A|; C1 and C2 are arbitrary constants; Jν (z) and Yν (z) are Bessel
functions; and Iν (z) and Kν (z) are modified Bessel functions.
2◦ . There are multiplicative separable particular solutions of the form
w(x, y) = ϕ(x)ψ(y),
(axn ϕ′x )′x = A1 ϕ, (by m ψy′ )′y = (c − A1 )ψ. (2)
The solutions of equations (2) are expressed in terms of the Bessel functions (or modified
Bessel functions); see equation 9.4.3.1, Item 3◦ .
3◦ . There are additively separable particular solutions of the form
w(x, y) = f (x) + g(y),
where f (x) and g(y) are determined by the following second-order linear ordinary differ-
(axn fx′ )′x − cf = A2 , (by m gy′ )′y − cg = −A2 . (3)
The solutions of equations (3) are expressed in terms of the Bessel functions (or modified
Bessel functions).
4◦ . The transformation (specified by A. I. Zhurov, private communication, 2001)

2−n 2−m
x 2 = Ar cos θ, y 2 = Br sin θ,
where A2 = a(2 − n)2 and B 2 = b(2 − m)2 , leads to the equation
∂ 2w 4−nm 1 ∂w 1 ∂ 2w 2 (nm−n−m) cos 2θ+(n−m) ∂w

+ + − = 4cw,
∂r 2 (2−n)(2−m) r ∂r r 2 ∂θ 2 r 2 (2−n)(2−m) sin 2θ ∂θ
which admits separable solutions of the form w(r, θ) = F1 (r)F2 (θ).

h i h i
∂ ∂w ∂ ∂w
4. a(x + k)n + b(y + s)m = c.
∂x ∂x ∂y ∂y
The transformation ζ = x + k, η = y + s leads to an equation of the form 9.4.3.2:

∂ n ∂w ∂ m ∂w
aζ + bη = c.
∂ζ ∂ζ ∂η ∂η
h i h i
∂ ∂w ∂ ∂w
5. a(x + k)n + b(y + s)m = cw.
∂x ∂x ∂y ∂y
The transformation ζ = x + k, η = y + s leads to an equation of the form 9.4.3.3:

∂ n ∂w ∂ m ∂w
aζ + bη = cw.
∂ζ ∂ζ ∂η ∂η

∂ ∂w ∂ ∂w
6. aeβx + beµy = 0.
∂x ∂x ∂y ∂y
This is a two-dimensional equation of the heat and mass transfer theory in an inhomoge-
neous anisotropic medium. Here, a1 (x) = aeβx and a2 (y) = beµy are the principal thermal
diffusivities.
1◦ . Particular solutions (A, B, C are arbitrary constants):
w(x, y) = Ae−βx + Be−µy + C,

A A
w(x, y) = 2
(βx + 1)e−βx − 2 (µy + 1)e−µy + B,
aβ bµ
−βx−µy
w(x, y) = Ae + B.
w(x, y) = ϕ(x)ψ(y), (1)
(aeβx ϕ′x )′x = −A1 ϕ, (2)

µy
(be ψy′ )′y = A1 ψ. (3)

(
e−βx/2 C1 J1 ke−βx/2 + C2 Y1 ke−βx/2 for A1 > 0,
ϕ(x) = −βx/2
−βx/2
−βx/2

e C1 I1 ke + C2 K1 ke for A1 < 0,
p
where k = −(2/β) |A1 |/a; C1 and C2 are arbitrary constants; J1 (z) and Y1 (z) are Bessel
functions; and I1 (z) and K1 (z) are modified Bessel functions.
(
e−µy/2 C1 J1 se−µy/2 + C2 Y1 se−µy/2 for A1 < 0,
ψ(y) = −µy/2
−µy/2
−µy/2

e C1 I1 se + C2 K1 se for A1 > 0,
p
where s = −(2/µ) |A1 |/b; C1 and C2 are arbitrary constants.
The sum of solutions of the form (1) corresponding to different values of the parame-
ter A1 is also a solution of the original equation.
3◦ . See equation 9.4.3.8, Item 3◦ , for c = 0.

∂ ∂w ∂ ∂w
7. aeβx + beµy = c.
∂x ∂x ∂y ∂y
This is a two-dimensional equation of the heat and mass transfer theory with constant vol-
ume release of heat in an inhomogeneous anisotropic medium. Here, a1 (x) = aeβx and
a2 (y) = beµy are the principal thermal diffusivities.
The substitution
c
w(x, y) = U (x, y) − (βx + 1)e−βx
aβ 2
leads to a homogeneous equation of the form 9.4.3.6:

∂ ∂U ∂ ∂U
aeβx + beµy = 0.
∂x ∂x ∂y ∂y

∂ ∂w ∂ ∂w
8. aeβx + beµy = cw.
∂x ∂x ∂y ∂y
in an inhomogeneous anisotropic medium.
1◦ . For βµ 6= 0, there are particular solutions of the form
w = w(ξ), ξ = bµ2 e−βx + aβ 2 e−µy )1/2 .
′′ 1 ′ 4c
wξξ − w = Bw, B= .
ξ ξ abβ 2 µ2
For the solution of this equation, see 9.4.3.3 (Item 1◦ for A = −1).
2◦ . The original equation admits multiplicative (and additively) separable solutions. See
equation 7.4.3.12 with f (x) = aeβx and g(y) = beµy .
e−βx/2 = Ar cos θ, e−µy/2 = Br sin θ,
where A2 = aβ 2 and B 2 = bµ2 , leads to the equation

∂2w 1 ∂w 1 ∂2w 2 ∂w
2
− + 2 2
− 2 cot 2θ = 4cw,
∂r r ∂r r ∂θ r ∂θ

∂ ∂w ∂ ∂w
9. axn + beβy = cw.
∂x ∂x ∂y ∂x
1◦ . For n 6= 2 and β 6= 0, there are particular solutions of the form
x2−n e−βy
w = w(r), r2 = + .
a(2 − n)2 bβ 2
The function w = w(r) is determined by the ordinary differential equation
∂2w n 1 ∂w
+ = 4cw.
∂r 2 2 − n r ∂r
For the solution of this equation, see 9.4.3.3 (Item 1◦ ).
2◦ . The original equation admits multiplicative (and additively) separable solutions. See
equation 7.4.3.12 with f (x) = axn and g(y) = beβy .
1 1
x1− 2 n = Ar cos θ, e− 2 βy = Br sin θ,
where A2 = a(2 − n)2 and B 2 = bβ 2 , leads to the equation

∂2w n 1 ∂w 1 ∂2w 2 (1 − n) cos 2θ + 1 ∂w
2
+ + 2 2
− 2 = 4cw,
∂r 2 − n r ∂r r ∂θ r (2 − n) sin 2θ ∂θ
h i
∂ ∂w ∂2w
10. f (x) + 2
= 0.
∂x ∂x ∂y
Z
C1 x + C3
w = C1 y 2 + C2 y − 2 dx + C4 ,
f (x)
Z
3 C1 x + C3
w = C1 y + C2 y − 6y dx + C4 ,
f (x)
Z
dx
w = [C1 Φ(x) + C2 ]y + C3 Φ(x) + C4 ,Φ(x) = ,
f (x)
Z Z
2 1
w = [C1 Φ(x) + C2 ]y + C3 Φ(x) + C4 − 2 [C1 Φ(x) + C2 ] dx dx,
f (x)
where C1 , C2 , C3 , and C4 are arbitrary constants.
w = (C1 eλy + C2 e−λy )H(x),
the ordinary differential equation [f (x)Hx′ ]′x + λ2 H = 0.
w = [C1 sin(λy) + C2 cos(λy)]Z(x),
the ordinary differential equation [f (x)Zx′ ]′x − λ2 Z = 0.
4◦ . Particular solutions with even powers of y:
n
X
w= ζk (x)y 2k ,
k=0
where the functions ζk = ζk (x) are defined by the recurrence relations

Z
dx
ζn (x) = An Φ(x) + Bn , Φ(x) = ,
f (x)
Z Z
1
ζk−1 (x) = Ak Φ(x) + Bk − 2k(2k − 1) ζk (x) dx dx,
f (x)
where Ak and Bk are arbitrary constants (k = n, . . . , 1).

5◦ . Particular solutions with odd powers of y:
n
X
w= ηk (x)y 2k+1 ,
k=0
where the functions ηk = ηk (x) are defined by the recurrence relations

Z
dx
ηn (x) = An Φ(x) + Bn , Φ(x) = ,
f (x)
Z Z
1
ηk−1 (x) = Ak Φ(x) + Bk − 2k(2k + 1) ηk (x) dx dx,
f (x)

h i h i
∂ ∂w ∂ ∂w
11. f (x) + g(y) = 0.
∂x ∂x ∂y ∂y
This is a two-dimensional sourceless equation of the heat and mass transfer theory in an
inhomogeneous anisotropic medium. The functions f = f (x) and g = g(y) are the principal
thermal diffusivities.
Z Z
dx dy
w(x, y) = A1 + B1 + C1 ,
f (x) g(y)
Z Z
x dx y dy
w(x, y) = A2 − A2 + B2 ,
f (x) g(y)
Z Z
dx dy
w(x, y) = A3 + B3 ,
f (x) g(y)
where A1 , A2 , A3 , B1 , B2 , B3 , and C1 are arbitrary constants. A linear combination of

these solutions is also a solution of the original equation.
w(x, y) = ϕ(x)ψ(y), (1)
ential equations (A is an arbitrary constant):
(f ϕ′x )′x = Aϕ, f = f (x),

(2)
(gψy′ )′y = −Aψ, g = g(y).
The sum of solutions of the form (1) corresponding to different values of the parameter A
in (2) is also a solution of the original equation (the solutions of some boundary value
problems may be obtained by separation of variables).
h i h i
∂ ∂w ∂ ∂w
12. f (x) + g(y) = βw.
∂x ∂x ∂y ∂y
in an inhomogeneous anisotropic medium. The functions f = f (x) and g = g(y) are the
principal thermal diffusivities.
w(x, y) = ϕ(x)ψ(y), (1)
ential equations (A is an arbitrary constant):
(f ϕ′x )′x = Aϕ, f = f (x),

(2)
(gψy′ )′y = (β − A)ψ, g = g(y).
The sum of solutions of the form (1) corresponding to different values of the parameter A
in (2) is also a solution of the original equation; the solutions of some boundary value
problems may be obtained by separation of variables.
2◦ . There are additively separable particular solutions of the form
w(x, y) = Φ(x) + Ψ(y),

where Φ(x) and Ψ(y) are determined by the following second-order linear ordinary differ-
ential equations (C is an arbitrary constant):
(f Φ′x )′x − βΦ = C, f = f (x),

(gΨ′y )′y − βΨ = −C, g = g(y).
In the special case β = 0, the solutions of these equations can be represented as

Z Z
x dx dx
Φ(x) = C + A1 + B1 ,
f (x) f (x)
Z Z
y dy dy
Ψ(y) = −C + A2 + B2 ,
g(y) g(y)
where A1 , A2 , B1 , and B2 are arbitrary constants.
9.4.4 Other Equations Arising in Applications

∂2w ∂2w
1. y + = 0.
∂x2 ∂y 2
Tricomi equation. It is used to describe near-sonic flows of gas.
w = Axy + Bx + Cy + D,
w = A(3x2 − y 3 ) + B(x3 − xy 3 ) + C(6yx2 − y 4 ),

2◦ . Particular solutions with even powers of x:
n
X
w= ϕk (y)x2k ,
k=0
where the functions ϕk = ϕk (y) are defined by the recurrence relations

Z y
ϕn (y) = An y + Bn , ϕk−1 (y) = Ak y + Bk − 2k(2k − 1) (y − t)tϕk (t) dt,
0

3◦ . Particular solutions with odd powers of x:
n
X
w= ψk (y)x2k+1 ,
k=0
where the functions ψk = ψk (y) are defined by the recurrence relations

Z y
ψn (y) = An y + Bn , ψk−1 (y) = Ak y + Bk − 2k(2k + 1) (y − t)tψk (t) dt,
0

4◦ . Separable particular solutions:

√
w(x, y) = A sinh(3λx) + B cosh(3λx) y CJ1/3 (2λy 3/2 ) + DY1/3 (2λy 3/2 ) ,
√
w(x, y) = A sin(3λx) + B cos(3λx) y CI1/3 (2λy 3/2 ) + DK1/3 (2λy 3/2 ) ,
where A, B, C, D, and λ are arbitrary constants, J1/3 (z) and Y1/3 (z) are Bessel functions,
and I1/3 (z) and K1/3 (z) are modified Bessel functions.
5◦ . For y > 0, see also equation 7.4.4.2 with n = 1. For y < 0, the change of variable
y = −t leads to an equation of the form 6.3.3.11 with m = 1.
∂2w ∂2w
2. yn + = 0.
∂x2 ∂y 2
w = Axy + Bx + Cy + D,
2A
w = Ax2 − y n+2 ,
(n + 1)(n + 2)
6A
w = Ax3 − xy n+2 ,
(n + 1)(n + 2)
2A
w = Ayx2 − y n+3 ,
(n + 2)(n + 3)
2◦ . Particular solutions with even powers of x:
m
X
w= ϕk (y)x2k ,
k=0
where the functions ϕk = ϕk (y) are defined by the recurrence relations

Z y
ϕm (y) = Am y + Bm , ϕk−1 (y) = Ak y + Bk − 2k(2k − 1) (y − t)tn ϕk (t) dt,
a
where Ak and Bk are arbitrary constants (k = m, . . . , 1), and a is any number.

3◦ . Particular solutions with odd powers of x:
m
X
w= ψk (y)x2k+1 ,
k=0
where the functions ψk = ψk (y) are defined by the recurrence relations

Z y
ψm (y) = Am y + Bm , ψk−1 (y) = Ak y + Bk − 2k(2k + 1) (y − t)tn ψk (t) dt,
a
where Ak and Bk are arbitrary constants (k = m, . . . , 1), and a is any number.

4◦ . Separable particular solutions:

√
w(x, y) = A sinh(λqx)+B cosh(λqx) y CJ 1 (λy q )+DY 1 (λy q ) , q = 12 (n+2),
2q 2q
√
w(x, y) = A sin(λqx)+B cos(λqx) y CI 1 (λy q )+DK 1 (λy q ) ,
2q 2q
where A, B, C, D, and λ are arbitrary constants, Jν (z) and Yν (z) are Bessel functions,
and Iν (z) and Kν (z) are modified Bessel functions.
5◦ . Fundamental solutions (for y > 0):
n r22
w1 (x, y, x0 , y0 ) = k1 (r12 )−β F (β, β, 2β; 1 − ξ), β= , ξ= ,
2(n + 2) r12
w2 (x, y, x0 , y0 ) = k2 (r12 )−β (1 − ξ)1−2β F (1 − β, 1 − β, 2 − 2β; 1 − ξ).
Here, F (a, b, c; ξ) is the hypergeometric function and
n+2 2β 2
2 2 4 n+2
2 1 4 Γ (β)
r1 = (x − x0 ) + 2
y 2 + y0 , k1 = ,
(n + 2) 4π n + 2 Γ(2β)
n+2 2β 2
2 2 4 n+2
2 1 4 Γ (1 − β)
r2 = (x − x0 ) + 2
y 2 − y0 , k2 = ,
(n + 2) 4π n + 2 Γ(2 − 2β)
where Γ(β) is the gamma function; x0 and y0 are arbitrary constants.
The fundamental solutions satisfy the conditions

∂y w1 y=0 = 0, w2 y=0 = 0 (x and x0 are any, y0 > 0).
The solutions of some boundary value problems can be found in the first book cited
below.
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), A. D. Polyanin (2001).
∂2w α ∂w ∂2w
3. + + = 0.
∂r2 r ∂r ∂z 2
Elliptic analogue of the Euler–Poisson–Darboux equation.
1◦ . For α = 1, see Sections 10.1.2 and 10.2.3 with w = w(r, z). For α 6= 1, the transfor-
mation x = (1 − α)z, y = r 1−α leads to an equation of the form 9.4.4.2:
2α ∂2w ∂2w
y 1−α + = 0.
∂x2 ∂y 2
2◦ . Suppose wα = wα (r, z) is a solution of the equation in question for a fixed value of the
parameter α. Then the functions w eα defined by the relations
∂wα
w
eα = ,
∂z
∂wα ∂wα
weα = r +z ,
∂r ∂z
∂wα ∂wα
weα = 2rz + (z 2 − r 2 ) + αzwα
∂r ∂z
are also solutions of this equation.
3◦ . Suppose wα = wα (r, z) is a solution of the equation in question for a fixed value of the
parameter α. Using this wα , one can construct solutions of the equation with other values
of the parameter by the formulas
w2−α = r α−1 wα ,
∂wα
wα−2 = r + (α − 1)wα ,
∂r
∂wα ∂wα
wα−2 = rz − r2 + (α − 1)zwα ,
∂r ∂z
∂wα ∂wα 2
wα−2 = r(r 2 − z 2 ) + 2r 2 z + r − (α − 1)z 2 wα ,
∂r ∂z
1 ∂wα
wα+2 = ,
r ∂r
z ∂wα ∂wα
wα+2 = − ,
r ∂r ∂z
r 2 − z 2 ∂wα ∂wα
wα+2 = + 2z + αwα .
r ∂r ∂z
⊙ Literature: A. V. Aksenov (2001).
∂2w ∂ 2w
4. + f (x) = 0.
∂x2 ∂y 2
w = C1 xy + C2 y + C3 x + C4 ,
Z x
2
w = C1 y + C2 xy + C3 y + C4 x − 2C1 (x − t)f (t) dt + C5 ,
a
Z x
w = C1 y 3 + C2 xy + C3 y + C4 x − 6C1 y (x − t)f (t) dt + C5 ,
a
Z x
2
w = (C1 x + C2 )y + C3 xy + C4 y + C5 x − 2 (x − t)(C1 t + C2 )f (t) dt + C6 ,
a
where C1 , . . . , C6 are arbitrary constants and a is any number.

w = (C1 eλy + C2 e−λy )H(x),
the ordinary differential equation Hxx′′ + λ2 f (x)H = 0.
w = [C1 sin(λy) + C2 cos(λy)]Z(x),
′′ − λ2 f (x)Z = 0.
the ordinary differential equation Zxx
4◦ . Particular solutions with even powers of y:

n
X
w= ϕk (x)y 2k ,
k=0
where the functions ϕk = ϕk (x) are defined by the recurrence relations

Z x
ϕn (x) = An x + Bn , ϕk−1 (x) = Ak x + Bk − 2k(2k − 1) (x − t)f (t)ϕk (t) dt,
a
where Ak and Bk are arbitrary constants (k = n, . . . , 1), and a is any number.

5◦ . Particular solutions with odd powers of y:
n
X
w= ψk (x)y 2k+1 ,
k=0
where the functions ψk = ψk (x) are defined by the recurrence relations

Z x
ψn (x) = An x + Bn , ψk−1 (x) = Ak x + Bk − 2k(2k + 1) (x − t)f (t)ψk (t) dt,
a
where Ak and Bk are arbitrary constants (k = n, . . . , 1), and a is any number.

h i h i
∂ ∂w ∂ ∂w
5. f1 (x) + f2 (y) + λ g1 (x) + g2 (y) w = 0.
∂x ∂x ∂y ∂y
This equation is encountered in the theory of vibration of inhomogeneous membranes. Its
separable solutions are sought in the form w(x, y) = ϕ(x)ψ(y).
The article cited below presents an algorithm for accelerated convergence of solutions
to eigenvalue boundary value problems for this equation.
⊙ Literature: L. D. Akulenko and S. V. Nesterov (1999).

∂ 2w ∂2w ∂w
a(x) + + b(x) + c(x)w = −Φ(x, y)
∂x2 ∂y 2 ∂x
◮ Statements of boundary value problems. Relations for the Green’s function.
Consider two-dimensional boundary value problems for the equation
∂2w ∂2w ∂w
a(x) 2
+ 2
+ b(x) + c(x)w = −Φ(x, y) (1)
∂x ∂y ∂x
with general boundary conditions in x,
s1 ∂x w − k1 w = f1 (y) at x = x1 ,
(2)
s2 ∂x w + k2 w = f2 (y) at x = x2 ,
and different boundary conditions in y. We assume that the coefficients of equation (1) and
the boundary conditions (2) meet the requirement
a(x), b(x), c(x) are continuous functions; a(x) > 0, |s1 | + |k1 | > 0, |s2 | + |k2 | > 0,
where x1 ≤ x ≤ x2 .
In the general case, the Green’s function can be represented as
X∞
un (x)un (ξ)
G(x, y, ξ, η) = ρ(ξ) 2
Ψn (y, η; λn ). (3)
n=1
kun k
Here, Z Z x2
1 b(x) 2
ρ(x) = exp dx , kun k = ρ(x)u2n (x) dx, (4)
a(x) a(x) x1
and the λn and un (x) are the eigenvalues and eigenfunctions of the homogeneous boundary
value problem for the ordinary differential equation
a(x)u′′xx + b(x)u′x + [λ + c(x)]u = 0, (5)

s1 u′x − k1 u = 0 at x = x1 , (6)
s2 u′x + k2 u = 0 at x = x2 . (7)
The functions Ψn for various boundary conditions in y are specified in Table 9.5.
Equation (5) can be rewritten in self-adjoint form as
[p(x)u′x ]′x + [λρ(x) − q(x)]u = 0, (8)
where the functions p(x) and q(x) are given by

Z Z
b(x) c(x) b(x)
p(x) = exp dx , q(x) = − exp dx ,
a(x) a(x) a(x)
and ρ(x) is defined in (4).

The eigenvalue problem (8), (6), (7) possesses the following properties:
1◦ . All eigenvalues λ1 , λ2 , . . . are real and λn → ∞ as n → ∞.
2◦ . The system of eigenfunctions {u1 (x), u2 (x), . . . } is orthogonal on the interval x1 ≤
x ≤ x2 with weight ρ(x), that is,
Z x2
ρ(x)un (x)um (x) dx = 0 for n 6= m.
x1
q(x) ≥ 0, s1 k1 ≥ 0, s2 k2 ≥ 0 (9)
are satisfied, there are no negative eigenvalues. If q ≡ 0 and k1 = k2 = 0, then the least
eigenvalue is λ0 = 0 and the corresponding eigenfunction is u0 = const; in this case, the
summation in (3) must start with n = 0. In the other cases, if conditions (9) are satisfied,
all eigenvalues are positive; for example, the first inequality in (9) holds if c(x) ≤ 0.
TABLE 9.5 √
The functions Ψn in (3) for various boundary conditions. Notation: βn = λn
Domain Boundary conditions Function Ψn(y, η; λn)
1 −βn|y−η|
−∞ < y < ∞ |w| < ∞ as y → ±∞ e
2βn

1 e−βny sinh(βnη) if y > η,
0≤y<∞ w = 0 at y = 0
βn e−βnη sinh(βny) if η > y

1 e−βny cosh(βnη) if y > η,
0≤y<∞ ∂y w = 0 at y = 0
βn e−βnη cosh(βny) if η > y

1 e−βny [βn cosh(βnη)+k3 sinh(βnη)] if y > η,
0≤y<∞ ∂y w−k3w = 0 at y = 0
βn(βn +k3) e−βnη [βn cosh(βny)+k3 sinh(βny)] if η > y

w=0 at y = 0, 1 sinh(βnη) sinh[βn(h−y)] if y > η,
0≤y≤h
w=0 at y=h βn sinh(βnh) sinh(βny) sinh[βn(h−η)] if η > y

∂y w = 0 at y = 0, 1 cosh(βnη) cosh[βn(h−y)] if y > η,
0≤y≤h
∂y w = 0 at y=h βn sinh(βnh) cosh(βny) cosh[βn(h−η)] if η > y

w=0 at y = 0, 1 sinh(βnη) cosh[βn(h−y)] if y > η,
0≤y≤h
∂y w = 0 at y=h βn cosh(βnh) sinh(βny) cosh[βn(h−η)] if η > y
Section 3.8.9 presents some relations for estimating the eigenvalues λn and eigenfunc-
tions un (x).
The Green’s function of the two-dimensional third boundary value problem (1)–(2)
augmented by the boundary conditions
∂w
− k3 w = 0 at y = 0,
∂y
∂w
+ k4 w = 0 at y = h
∂y
is given by relation (3) with



 βn cosh(βnη)+k3 sinh(βnη) βn cosh[βn(h−y)]+k4 sinh[βn(h−y)]



 βn βn(k3+k4) cosh(βnh)+(βn2 +k3k4) sinh(βnh)



 if y>η,
Ψn(y, η; λn)=

 βn cosh(βny)+k3 sinh(βny) βn cosh[βn(h−η)]+k4 sinh[βn(h−η)]



 βn βn(k3+k4) cosh(βnh)+(βn2 +k3k4) sinh(βnh)



 if y<η.
◮ Representation of solutions to boundary value problems using the Green’s func-

tion.
1◦ . The solution∗ of the first boundary value problem for equation (1) with the boundary
conditions
w = f1 (y) at x = x1 , w = f2 (y) at x = x2 ,
w = f3 (x) at y = 0, w = f4 (x) at y = h
is expressed in terms of the Green’s function as
Z h
∂
w(x, y) = a(x1 ) f1 (η) G(x, y, ξ, η) dη
0 ∂ξ ξ=x1
Z h
∂
− a(x2 ) f2 (η) G(x, y, ξ, η) dη
0 ∂ξ ξ=x2
Z x2 Z x2
∂ ∂
+ f3 (ξ) G(x, y, ξ, η) dξ − f4 (ξ) G(x, y, ξ, η) dξ
x1 ∂η η=0 x1 ∂η η=h
Z x2Z h
+ Φ(ξ, η)G(x, y, ξ, η) dη dξ.
x1 0
2◦ . The solution of the second boundary value problem for equation (1) with boundary
conditions
∂x w = f1 (y) at x = x1 , ∂x w = f2 (y) at x = x2 ,
∂y w = f3 (x) at y = 0, ∂y w = f4 (x) at y = h
is expressed in terms of the Green’s function as

Z h Z h
w(x, y) = −a(x1 ) f1 (η)G(x, y, x1 , η) dη + a(x2 ) f2 (η)G(x, y, x2 , η) dη
0 0
Z x2 Z x2
− f3 (ξ)G(x, y, ξ, 0) dξ + f4 (ξ)G(x, y, ξ, h) dξ
x1 x1
Z x2Z h
x1 0
3◦ . The solution of the third boundary value problem for equation (1) in terms of the
Green’s function is represented in the same way as the solution of the second boundary
value problem (the Green’s function is now different).
∗
For unbounded domains, the condition of boundedness of the solution as y → ±∞ is set; in Table 9.5,
this condition is omitted.
Chapter 10
Second-Order
Elliptic Equations with
10.1 Laplace Equation ∆3 w = 0

The three-dimensional Laplace equation is often encountered in heat and mass transfer
theory, fluid mechanics, elasticity, electrostatics, and other areas of mechanics and physics.
For example, in heat and mass transfer theory, this equation describes stationary tempera-
ture distribution in the absence of heat sources and sinks in the domain under study.
A regular solution of the Laplace equation is called a harmonic function. The first
boundary value problem for the Laplace equation is often referred to as the Dirichlet prob-
lem, and the second boundary value problem as the Neumann problem.
Extremum principle: Given a domain D, a harmonic function w in D that is not iden-
tically constant in D cannot attain its maximum or minimum value at any interior point
of D.

The three-dimensional Laplace equation in the rectangular Cartesian system of coordinates
is written as
∂2w ∂2w ∂2w
2
+ 2
+ = 0.
∂x ∂y ∂z 2

w(x, y, z) = Ax + By + Cz + D,
w(x, y, z) = Ax2 + By 2 − (A + B)z 2 + Cxy + Dxz + Eyz,
w(x, y, z) = cos(µ1 x + µ2 y) exp(±µz),
w(x, y, z) = sin(µ1 x + µ2 y) exp(±µz),
859
860 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
w(x, y, z) = exp(µ1 x + µ2 y) cos(µz + A),

w(x, y, z) = exp(±µx) cos(µ1 y + A) cos(µ2 z + B),
w(x, y, z) = cosh(µ1 x) cosh(µ2 y) cos(µz + B),
w(x, y, z) = cosh(µ1 x) sinh(µ2 y) cos(µz + B),
w(x, y, z) = cosh(µx) cos(µ1 y + A) cos(µ2 z + B),
w(x, y, z) = sinh(µ1 x) sinh(µ2 y) sin(µz + B),
w(x, y, z) = sinh(µx) sin(µ1 y + A) sin(µ2 z + B),
p
where A, B, C, D, E, µ1 , and µ2 are arbitrary constants, and µ = µ21 + µ22 .
1
E (x, y, z) = p .
4π x2 + y 2 + z 2
3◦ . Suppose w = w(x, y, z) is a solution of the Laplace equation. Then the functions
w1 = Aw(±λx+C1, ±λy+C2, ±λx+C3),

p
A x y z
w2 = w 2 , 2 , 2 , r = x2 +y 2 +z 2,
r r r r

A x−ar 2 y−br 2 z−cr 2
w3 = √ w , , , Ξ = 1−2(ax+by+cz)+(a2 +b2 +c2)r 2,
Ξ Ξ Ξ Ξ
where A, Cn , a, b, c, and λ are arbitrary constants, are also solutions of this equation. The
signs at λ in the expression of w1 can be taken independently of one another.
⊙ Literature: W. Miller, Jr. (1977), R. Courant and D. Hilbert (1989).
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, 0 ≤ z < ∞. First boundary value

problem.
A half-space is considered. A boundary condition is prescribed:
w = f (x, y) at z = 0.
Solution:
Z ∞Z ∞
1 zf (ξ, η) dξ dη
w(x, y, z) = 3/2 .
2π −∞ −∞ (x − ξ)2 + (y − η)2 + z 2
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, 0 ≤ z < ∞. Second boundary value

problem.
∂z w = f (x, y) at z = 0.
Solution:
Z ∞Z ∞
1 f (ξ, η) dξ dη
w(x, y, z) = − p + C,
2π −∞ −∞ (x − ξ)2 + (y − η)2 + z 2
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ≤ z ≤ c. First boundary value problem.

A rectangular parallelepiped is considered. Boundary conditions are prescribed:
w = f1 (y, z) at x = 0, w = f2 (y, z) at x = a,
w = f3 (x, z) at y = 0, w = f4 (x, z) at y = b,
w = f5 (x, y) at z = 0, w = f6 (x, y) at z = c.
Solution:
∞ X
X ∞ 2 sinh(λ1 x) + f 1 sinh[λ1 (a − x)]
fnm πny πmz
nm nm nm
w(x, y, z) = sin sin
sinh(λ1nma) b c
n=1 m=1
X∞ X ∞ 4 sinh(λ2 y) + f 3 sinh[λ2 (b − y)]
fnm πnx πmz
nm nm nm
+ sin sin
n=1 m=1
sinh(λ2nmb) a c
X∞ X ∞ 6 sinh(λ3 z) + f 5 sinh[λ3 (c − z)]
fnm πnx πmy
nm nm nm
+ sin sin ,
sinh(λ3nmc) a b
n=1 m=1
where the constant coefficients are given by
r r r
1 n2 m2 2 n2 m2 3 n2 m2
λnm = π + , λ nm = π + , λnm = π + ,
b2 c2 a2 c2 a2 b2
 Z bZ c πny πmz

 4

 bc f i (y, z) sin sin dy dz for i = 1, 2;

 b c
 4 Z0 a Z0 c πnx πmz
i
fnm = fi (x, z) sin sin dx dz for i = 3, 4;

 ac Z0 Z0 a c

 a b πnx πmy

 4
 fi (x, y) sin sin dx dy for i = 5, 6.
ab 0 0 a b
Example 10.1. The planes x = 0 and x = a have constant temperatures w1 and w2 , respectively.
The other planes are maintained at zero temperature (f3 = f4 = f5 = f6 = 0).
Solution:
∞ ∞
16 X X w2 sinh(µnm x) + w1 sinh[µnm (a − x)]
w= 2 sin(pn y) sin(qm z),
π n=1 m=1 (2n − 1)(2m − 1) sinh(µnm a)
π(2n − 1) π(2m − 1) p
pn = , qm = , µnm = p2n + qm
2 .
b c
and A. N. Tikhonov (1980), H. S. Carslaw and J. C. Jaeger (1984).
◆ For the solution of other boundary value problems for the three-dimensional Laplace
equation in the Cartesian coordinate system, see Section 10.2.2 for Φ ≡ 0.

The three-dimensional Laplace equation in the cylindrical coordinate system is written as
p
1 ∂ ∂w 1 ∂2w ∂2w
∆3 w ≡ r + 2 + = 0, r = x2 + y 2 .
r ∂r ∂r r ∂ϕ2 ∂z 2

B m
w(r, ϕ, z) = Ar + m (C cos mϕ + D sin mϕ)(α + βz),
r
w(r, ϕ, z) = Jm (µr)(A cos mϕ + B sin mϕ)(C cosh µz + D sinh µz),
w(r, ϕ, z) = Ym (µr)(A cos mϕ + B sin mϕ)(C cosh µz + D sinh µz),
w(r, ϕ, z) = Im (µr)(A cos mϕ + B sin mϕ)(C cos µz + D sin µz),
w(r, ϕ, z) = Km (µr)(A cos mϕ + B sin mϕ)(C cos µz + D sin µz),
where m = 0, 1, 2, . . . ; A, B, C, D, α, β, and µ are arbitrary constants; the Jm (ξ)
and Ym (ξ) are Bessel functions; and the Im (ξ) and Km (ξ) are modified Bessel functions.
◮ Domain: 0 ≤ r ≤ a, 0 ≤ ϕ ≤ 2π, −∞ < z < ∞. First boundary value problem.

An infinite circular cylinder is considered. A boundary condition is prescribed:
w = f (ϕ, z) at r = a.
Solution:
∞ ∞ Z
1 X X Jn (λnm r) ∞
w(r, ϕ, z) = − ′
An (ξ) cos nϕ + Bn (ξ) sin nϕ exp −λnm |ξ − z| dξ,
a n=0 m=1 Jn (λnm a) −∞
where the Jn (r) are the Bessel functions and λmn are positive roots of the transcendental
equation Jn (aλ) = 0. The functions An (z) and Bn (z) are the coefficients of the Fourier
series expansion of f (ϕ, z),
Z 2π Z 2π
εn 1
An (z) = f (ϕ, z) cos(nϕ) dϕ, Bn (z) = f (ϕ, z) sin(nϕ) dϕ,
π 0 π 0
where ε0 = 1/2 and εn = 1 for n = 1, 2, . . .

If the surface temperature is independent of ϕ, i.e., f (z, ϕ) = f (z), then the solution
takes the form
∞ Z
1 X J0 (λm r) ∞
w(r, ϕ, z) = f (z + ζ) + f (z − ζ) exp −λm ζ dζ,
a J1 (λm a) 0
m=1
where the λm are positive roots of the transcendental equation J0 (aλ) = 0.

◮ Domain: 0 ≤ r ≤ a, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ b. First boundary value problem.

A circular cylinder of finite length is considered. Boundary conditions are prescribed:
w = f (ϕ, z) at r = a, w = g1 (r, ϕ) at z = 0, w = g2 (r, ϕ) at z = b.
Solution:
πmr
∞ In
∞ X
X
w(r, ϕ, z) = b A cos nϕ + B sin nϕ sin πmz
πma nm nm
n=0 m=1 In
b
b
µ (b − z)
mn
X∞ X ∞ µ r sinh
+ Jn
mn (1) (1)
Cnm cos nϕ + Dnm sin nϕ a
µ b
a mn
n=0 m=1 sinh
µ za
mn
X∞ X ∞ µ r sinh
+ Jn
mn (2) (2)
Cnm cos nϕ + Dnm sin nϕ a
µ b,
a mn
n=0 m=1 sinh
a
where the Jn (r) are Bessel functions, the In (r) are modified Bessel functions, and µmn is
(i) (i)
the mth root of the equation Jn (µ) = 0. The coefficients Anm , Bnm , Cnm , and Dnm are
defined by
Z Z πmz
εn 2π b
Anm = f (ϕ, z) cos(nϕ) sin dϕ dz,
πb 0 0 b
Z 2πZ b πmz
2
Bnm = f (ϕ, z) sin(nϕ) sin dϕ dz,
πb 0 0 b
Z 2πZ a µ r
(i) εn mn
Cnm = gi (r, ϕ) cos(nϕ)J n r dr dϕ,
πa2 [Jn′ (µmn )]2 0 0 a
Z 2πZ a µ r
(i) 2 mn
Dnm = 2 ′ 2
gi (r, ϕ) sin(nϕ)J n r dr dϕ,
πa [Jn (µmn )] 0 0 a
(
1 for n = 0,
εn = i = 1, 2.
2 for n 6= 0,

equation in the cylindrical coordinate system, see Section 10.2.3 for Φ ≡ 0.

The three-dimensional Laplace equation in the spherical coordinate system is written as
p
1 ∂ 2 ∂w 1 ∂ ∂w 1 ∂2w
r + sin θ + = 0, r = x2 + y 2 + z 2 .
r 2 ∂r ∂r r 2 sin θ ∂θ ∂θ r 2 sin2 θ ∂ϕ2
B
w(r) = A + ,
r
n B
w(r, θ) = Ar + n+1 Pn (cos θ),
r

B
w(r, θ, ϕ) = Ar + n+1 Pnm (cos θ)(C cos mϕ + D sin mϕ),
n
r
where n = 0, 1, 2, . . . ; m = 0, 1, 2, . . . , n; A, B, C, D are arbitrary constants; the Pn (ξ)
are the Legendre polynomials; and the Pnm (ξ) are associated Legendre functions, which are
expressed as
1 dn 2 dm
Pn (x) = (x − 1)n , Pnm (x) = (1 − x2 )m/2 Pn (x).
n! 2n dxn dxm
◮ Domain: 0 ≤ r ≤ R or R ≤ r < ∞. First boundary value problem.

A boundary condition at the sphere surface is prescribed:
w = f (θ, ϕ) at r = R.
1◦ . Solution of the inner problem (for r ≤ R):

Z 2πZ π
R R2 − r 2
w(r, θ, ϕ) = f (θ0 , ϕ0 ) 2 sin θ0 dθ0 dϕ0 ,
4π 0 0 (r − 2Rr cos γ + R2 )3/2
cos γ = cos θ cos θ0 + sin θ sin θ0 cos(ϕ − ϕ0 ).
This formula is conventionally called the Poisson integral for a sphere.

Series solution:
X∞ n X n
r
w(r, θ, ϕ)= Yn(θ, ϕ), Yn(θ, ϕ)= (Anm cos mϕ+Bnm sin mϕ)Pnm(cos θ),
n=0
R m=0
where
Z 2πZ π
1
A00 = f (θ, ϕ) sin θ dθ dϕ,
4π 0 0
Z Z
(2n + 1)(n − m)! 2π π
Anm = f (θ, ϕ)Pnm (cos θ) cos mϕ sin θ dθ dϕ,
2π(n + m)! 0 0
Z Z
(2n + 1)(n − m)! 2π π
Bnm = f (θ, ϕ)Pnm (cos θ) sin mϕ sin θ dθ dϕ.
2π(n + m)! 0 0
2◦ . Solution of the outer problem (for r ≥ R):

Z 2πZ π
R r 2 − R2
w(r, θ, ϕ) = f (θ0 , ϕ0 ) 2 sin θ0 dθ0 dϕ0 ,
4π 0 0 (r − 2Rr cos γ + R2 )3/2
where cos γ is expressed in the same way as in the inner problem.

Series solution:
∞
X R n+1
w(r, θ, ϕ) = Yn (θ, ϕ),
r
n=0
Xn
Yn (θ, ϕ) = (Anm cos mϕ + Bnm sin mϕ)Pnm (cos θ),
m=0
where the coefficients Anm and Bnm are defined by the same relations as in the inner
problem.
⊙ Literature: G. N. Polozhii (1964), V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), B. M. Bu-
dak, A. A. Samarskii, and A. N. Tikhonov (1980), A. N. Tikhonov and A. A. Samarskii (1990).
◮ Domain: 0 ≤ r ≤ R or R ≤ r < ∞. Second boundary value problem.

A boundary condition at the sphere surface is prescribed:
∂w
= f (θ, ϕ) at r = R.
∂r
The function f (θ, ϕ) must satisfy the solvability condition
Z 2πZ π
f (θ, ϕ) sin θ dθ dϕ = 0.
0 0
1◦ . Solution of the inner problem (for r ≤ R):

Z 2πZ π
R 1 2
w(r, θ, ϕ) = f (θ0 , ϕ0 ) ln(R + r1 − r cos γ) − sin θ0 dθ0 dϕ0 ,
4π 0 0 R r1
p
r1 = r 2 − 2Rr cos γ + R2 , cos γ = cos θ cos θ0 + sin θ sin θ0 cos(ϕ − ϕ0 ).
Series solution:
∞ X n
X R r n
w(r, θ, ϕ) = (Anm cos mϕ + Bnm sin mϕ)Pnm (cos θ) + C,
n=1 m=0
n R
where the coefficients Anm and Bnm are expressed in the same way as in the inner first
boundary value problem (see the paragraph above), and C is an arbitrary constant.
2◦ . Solution of the outer problem (for r ≥ R):
Z 2πZ π
R 1 R + r1 − r cos γ 2
w(r, θ, ϕ) = − f (θ0 , ϕ0 ) ln − sin θ0 dθ0 dϕ0 ,
4π 0 0 R r(1 − cos γ) r1
p
r1 = r 2 − 2Rr cos γ + R2 , cos γ = cos θ cos θ0 + sin θ sin θ0 cos(ϕ − ϕ0 ).
Series solution:
∞ X
X n n+1
R R
w(r, θ, ϕ) = − (Anm cos mϕ + Bnm sin mϕ)Pnm (cos θ) + C,
n+1 r
n=0 m=0
where the coefficients Anm and Bnm are expressed in the same way as in the inner first
boundary value problem, and C is an arbitrary constant.
3◦ . Outer boundary value problems where unbounded solutions as r → ∞ are sought are
also encountered in applications.
Example 10.2. A potential translational flow of an ideal incompressible fluid about a sphere of
radius R is governed by the Laplace equation with the boundary conditions:
∂r w = 0 at r = R, |w − U r cos θ| → 0 as r → ∞,
where w is the potential, U the unperturbed flow velocity at infinity; the fluid velocity is expressed
in terms of the potential as v = ∇ϕ.
Solution:
R3
w = U r 1 + 3 cos θ.
2r
This solution is a special case of the second formula from the first paragraph for n = 1.
⊙ Literature: G. N. Polozhii (1964), V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), B. M. Bu-
dak, A. A. Samarskii, and A. N. Tikhonov (1980), L. G. Loitsyanskii (1996).
equation in the spherical coordinate system, see Section 10.2.4 for Φ ≡ 0.
10.1.4 Other Orthogonal Curvilinear Systems of Coordinates

The three-dimensional Laplace equation admits separation of variables in the eleven or-
thogonal coordinate systems that are listed in Table 10.1.
For the general ellipsoidal and conical coordinate systems, the functions f , g, and h are
determined by Lamé equations that involve the Jacobian elliptic function sn z = sn(z, k).
The solutions of these equations under some conditions can be represented in the form of
finite series called Lamé polynomials. For details about the Lamé equation and its solutions,
see Whittaker and Watson (1963), Bateman and Erdélyi (1955), Arscott (1964), and Miller,
Jr. (1977).
There are also coordinate systems that allow the so-called R-separation of variables
of the three-dimensional Laplace equation. p Such solutions in the new coordinate system,
µ, ν, ρ, can be represented in the form w = R (µ, ν, ρ) f (µ)g(ν)h(ρ). Coordinates that
allow the R-separation of variables are listed in Table 10.2.
Only the bicylindrical and toroidal coordinate systems are fairly widely used in applica-
tions. In three subsequent coordinate systems, the functions f = f (µ) and g = g(ρ) are de-
termined by identical equations. With the change of variables µ = sn2 (α, k), ρ = sn2 (β, k),
where k = a−1/2 , these equations are reduced to Lamé equations (α and β are the new in-
dependent variables).
⊙ Literature for Section 10.1.4: M. Bôcher (1894), N. N. Lebedev, I. P. Skal’skaya, and Ya. S. Uflyand (1955),
F. M. Morse and H. Feshbach (1953, Vols. 1–2), P. Moon and D. Spencer (1988), A. Makarov, J. Smorodinsky,
K. Valiev, and P. Winternitz (1967), W. Miller, Jr. (1977).
10.2 Poisson Equation ∆3 w + Φ(x) = 0

10.2.1 Preliminary Remarks. Solution Structure
Like the three-dimensional Laplace equation, the three-dimensional Poisson equation is of-
ten encountered in heat and mass transfer theory, fluid mechanics, elasticity, electrostatics,
10.2. Poisson Equation ∆3 w + Φ(x) = 0 867
TABLE 10.1
Orthogonal coordinates x̄, ȳ, z̄ that allow separable solutions of the form
w = f (x̄)g(ȳ)h(z̄) for the three-dimensional Laplace equation ∆3 w = 0
Coordinates Transformations Particular solutions (or equations for f , g, h)
x = x, w = cos(k1 x + s1 ) cos(k2 y + s2 ) cosh(k3 z + s3 ),

Cartesian
y = y, where k12 + k22 = k32 ;
x, y, z
z=z see also Section 10.1.1 (particular solutions)
x = r cos ϕ, w = [AJn (kr) + BYn (kr)] cos(nϕ + c) exp(±kz),

Cylindrical
y = r sin ϕ, Jn (z) and Yn (z) are the Bessel functions;
r, ϕ, z
Parabolic x= 1
2
(ξ 2 − η 2 ), w = Dµ−1/2 (±σξ)D
√ −µ−1/2 (±ση) cos(kz + s),
cylindrical y = ξη, where σ = 2k, Dµ (z) is the parabolic
ξ, η, z z=z cylinder function

Cen (u, −q) cen (v, −q) cos(kz + s),
Elliptic x = a cosh u cos v, w=
cylindrical y = a sinh u sin v, Sen (u, −q) sen (v, −q) cos(kz + s),
Cen and Sen are the modified Mathieu functions,
u, v, z z=z
cen and sen are the Mathieu functions, q = 14 a2 k2
x = r sin θ cos ϕ, w = (Ar n + Br −n−1 )Pnk (cos θ) cos(kϕ + s),

Spherical
y = r sin θ sin ϕ, Pnk (ξ) are the associated Legendre functions,
r, θ, ϕ
z = r cos θ see also Section 10.1.3 (particular solutions)
Prolate x = a sinh u sin v cos ϕ,

w = Pnk (cosh u)Pnk (cos v) cos(kϕ + s),
spheroidal y = a sinh u sin v sin ϕ,
Pnk (ξ) are the associated Legendre functions
u, v, ϕ z = a cosh u cos v
Oblate x = a cosh u sin v cos ϕ,
w = Pnk (−i sinh u)Pnk (cos v) cos(kϕ + s),
spheroidal y = a cosh u sin v sin ϕ,
Pnk (ξ) are the associated Legendre functions
u, v, ϕ z = a sinh u cos v
x = aξη cos ϕ, w = I±k (βξ)J±k (βη) cos(kϕ + s),
Parabolic
y = aξη sin ϕ, Jk (z) are the Bessel functions,
ξ, η, ϕ
z = 12 a(ξ 2 − η 2 ) Ik (z) are the modified Bessel functions

x = 2a cosh u cos v sinh ϕ, Cen (u, −b) cen (v, −b) Cen (ϕ + iπ/2, −b),
y = 2a sinh u sin v cosh ϕ, w=
Paraboloidal Sen (u, −b) sen (v, −b) Sen (ϕ + iπ/2, −b),
u, v, ϕ z = 12 a(cosh 2u b = 12 aβ; cen and sen are the Mathieu functions,
+ cos 2v − cosh 2ϕ) Cen and Sen are the modified Mathieu functions
q ′′
x= (µ−a)(ν−a)(ρ−a)
, fξξ + (β2 + β1 sn2 ξ)f = 0 (Lamé equation),
General a(a−1)
q gηη + (β2 + β1 sn2 η)g = 0 (Lamé equation),
′′
ellipsoidal y= (µ−1)(ν−1)(ρ−1)
,
p 1−a h′′ζζ + (β2 + β1 sn2 ζ)g = 0 (Lamé equation),
µ, ν, ρ µνρ
z= a µ = sn2 (ξ, k), ν = sn2 (η, k), ρ = sn2 (ζ, k), k = a−1/2
q f (ρ) = Aρn + Bρ−n−1 , n = 0, 1, . . . ,

(aµ−1)(aν−1) ′′
x=ρ 1−a
, gξξ + [β − n(n + 1)k2 sn2 ξ]g = 0 (Lamé equation),
Conical q
y = ρ a(µ−1)(ν−1) , hηη + [β − n(n + 1)k2 sn2 η]h = 0 (Lamé equation),
′′
ρ, µ, ν
√ a−1 2 2
where µ = sn (ξ, k), ν = sn (η, k), k = a ,1/2
z = ρ aµν
g and h are expressed in terms of the Lamé polynomials
TABLE 10.2
p Coordinates x̄, ȳ, z̄ that allow R-separated solutions of the form
w = R(x̄, ȳ, z̄) f (x̄)g(ȳ)h(z̄) for the three-dimensional Laplace equation ∆3 w = 0
New coordinates, Transformations Functions f , g, h

function R of coordinates (equations for f , g, h)
x=cR−1 sin α cos ϕ, m m

f (α)=A1P n (cos α)+A
2Qn (cos
α),
Bicylindrical y=cR−1 sin α sin ϕ,
g(β)=B1 cosh (n+ 2 )β +B2 sinh (n+ 12 )β ,
1
coordinates α, β, ϕ, z=cR−1 sinh β;
h(ϕ)=C1 cos(mϕ)+C2 sin(mϕ),
R=cosh β−cos α 0≤α≤π, β is any,
n=0, 1, 2, . . . ; m=0, 1, 2, . . .
0≤ϕ<2π
m
x=cR−1 sinh α cos ϕ, f (α)=A1Pn−1/2 (cosh α)+A2Qm n−1/2(cosh α),
Toroidal
y=cR−1 sinh α sin ϕ, g(β)=B1 cos(nβ)+B2 sin(nβ),
coordinates α, β, ϕ,
z=cR−1 sin β; h(ϕ)=C1 cos(mϕ)+C2 sin(mϕ),
R=cosh α−cos β
α≥0, −π≤β≤π, 0≤ϕ<2π n=0, 1, 2, . . . ; m=0, 1, 2, . . .
p p
Coordinates
q µ, ρ, ϕ, x=R−1 cos ϕ, U (µ) [ U (µ) f ′]′+ ( 14 −n2)µ−λ f =0,
p p
R= (µ−a)(a−ρ)
a(a−1)
y=R−1 p sin ϕ, U (ρ) [ U (ρ) g ′]′+ ( 14 −n2)ρ−λ g=0,
q z=R−1 −µρ/a; h(ϕ)=C1 cos(nϕ)+C2 sin(nϕ),
(µ−1)(1−ρ)
− a−1 µ>a>1, ρ<0, 0≤ϕ<2π U (t)=4t(t−1)(t−a)
p p
x=R−1 cos ϕ, U (µ) [ U (µ) f ′]′+ ( 14 −n2)µ−λ f =0,
p µ, ρ, ϕ,
Coordinates
y=R−1 p p
R= µρ
q a q sin ϕ, U (ρ) [ U (ρ) g ′]′+ ( 14 −n2)ρ−λ g=0,
z=R−1 (µ−a)(a−ρ) ; h(ϕ)=C1 cos(nϕ)+C2 sin(nϕ),
+ (µ−1)(ρ−1)
a−1
a(a−1)
1<ρ<a<µ, 0≤ϕ<2π U (t)=4t(t−1)(t−a)
p p
q µ, ρ, ϕ,
Coordinates x=R−1 cos ϕ, p
U (µ) [ U (µ) f ′]′+ ( 14 −n2)µ−λ f =0,
p
R=2 Re i(µ−a)(ρ−a)
a(a−b)
, y=Rp −1
sin ϕ, U (ρ) [ U (ρ) g ′]′+ ( 14 −n2)ρ−λ g=0,
−1
a= b̄=α+iβ, z=R −µρ/(ab); h(ϕ)=C1 cos(nϕ)+C2 sin(nϕ),
α, β are real numbers µ>0, ρ<0, 0≤ϕ<2π U (t)=4t(t−a)(t−b)
q
x=R−1 (µ−a)(ν−a)(ρ−a)
, p p ′ ′ 2
pU (µ) [ pU (µ) f ′]′−(3µ2 +λ1µ+λ2)f =0,
(b−a)(a−1)a
q
pµ, ν, ρ,
Coordinates y=R−1 (µ−b)(ν−b)(ρ−b) , pU (ν) [p U (ν) g ′]′−(3ν2 +λ1ν+λ2)g=0,
R=1+ µνρ ab
q (a−b)(b−1)b U (ρ) [ U (ρ) h ] −(3ρ +λ1ρ+λ2)h=0,
z=R−1 (µ−1)(ν−1)(ρ−1) ;
(a−1)(b−1) U (t)=16t(t−1)(t−a)(t−b)
0<ρ<1<ν <b<µ<a
q p p
Coordinates µ, ν, ρ, x=R−1 (µ−1)(ν−1)(ρ−1)
′ ′ 2
q
p(a−1)(b−1)
, pU (µ) [ pU (µ) f ′]′−(3µ2 +λ1µ+λ2)f =0,
R=2 Re (µ−a)(ν−a)(ρ−a) ,
ia(a−1)(a−b) y=R−1 − µνρab
, pU (ν) [p U (ν) g ′]′−(3ν2 +λ1ν+λ2)g=0,
a= b̄=α+iβ, z=R−1; U (ρ) [ U (ρ) h ] −(3ρ +λ1ρ+λ2)h=0,
α, β are real numbers ρ<0<µ<1<ν U (t)=16t(t−1)(t−a)(t−b)
and other areas of mechanics and physics. In particular, the Poisson equation describes sta-
tionary temperature distribution in the presence of thermal sources or sinks in the domain
under consideration.
The Laplace equation is a special case of the Poisson equation with Φ ≡ 0.
Throughout this section, we consider a three-dimensional bounded domain V with a
sufficiently smooth boundary S. We assume that r ∈ V and ρ ∈ V , where r = {x, y, z}
and ρ = {ξ, η, ζ}.

The solution of the first boundary value problem for the Poisson equation
∆3 w + Φ(r) = 0 (1)
in a domain V with the nonhomogeneous boundary condition
w = f (r) for r ∈ S

Z Z
∂G
w(r) = Φ(ρ)G(r, ρ) dVρ − f (ρ) dSρ . (2)
V S ∂Nρ
∂G
Here, G(r, ρ) is the Green’s function of the first boundary value problem, ∂N ρ
is the deriva-
tive of the Green’s function with respect to ξ, η, ζ along the outward normal N to the bound-
ary S of the domain V . The integration is everywhere with respect to ξ, η, ζ.
The volume elements in solution (2) for basic coordinate systems are presented in Ta-
ble 10.3. In addition, the expressions of the gradients are given, which enable one to find
∂G
the derivative along the normal in accordance with the formula ∂N ρ
= (N · ∇ρ G).
TABLE 10.3
The volume elements and distances occurring in relations (2)
and (5) in some coordinate systems. In all cases, ρ = {ξ, η, ζ}
Coordinate Volume Gradient, ∇ρ u Distance,

system element, dVρ (|iξ | = |iη | = |iζ | = 1) d = |r − ρ|
Cartesian p
dξ dη dζ iξ ∂u
∂ξ
+ iη ∂u
∂η
+ iζ ∂u
∂ζ d= (x − ξ)2 + (y − η)2 + (z − ζ)2
r = {x, y, z}
Cylindrical p
ξ dξ dη dζ iξ ∂u
∂ξ
+ iη 1ξ ∂u
∂η
+ iζ ∂u
∂ζ d= r 2 + ξ 2 − 2rξ cos(ϕ − η) + (z − ζ)2
r = {r, ϕ, z}
p
Spherical d = r 2 + ξ 2 − 2rξ cos γ, where
2 iξ ∂u + iη 1ξ ∂u 1
+ iζ ξ sin ∂u
r = {r, θ, ϕ} ξ sin η dξ dη dζ ∂ξ ∂η η ∂ζ
cos γ = cos θ cos η + sin θ sin η cos(ϕ − ζ)
The Green’s function G = G(r, ρ) of the first boundary value problem is determined
1◦ . The function G satisfies the Laplace equation with respect to x, y, z in the domain V
everywhere except for the point (ξ, η, ζ), at which it can have a singularity of the form
1 1
4π |r−ρ| .
2◦ . The function G, with respect to x, y, z, satisfies the homogeneous boundary condition
of the first kind at the boundary, i.e., the condition G|S = 0.
The Green’s function can be represented as
1 1
G(r, ρ) = + u, (3)
4π |r − ρ|
where the auxiliary function u = u(r, ρ) is determined by solving the first boundary value
1 1
problem for the Laplace equation ∆3 u = 0 with the boundary condition u|S = − 4π |r−ρ| ;
the vector quantity ρ in this problem is treated as a three-dimensional free parameter.
The Green’s function possesses the symmetry property with respect to the arguments:
G(r, ρ) = G(ρ, r).
The construction of Green’s functions is discussed in Section 10.3.2 for λ = 0.
Remark 10.1. For outer first boundary value problems for the Laplace equation, the following
condition is usually set at infinity: |w| < A/|r| (|r| → ∞, A = const).

The second boundary value problem for the Poisson equation (1) is characterized by the
boundary condition
∂w
= f (r) for r ∈ S.
∂N
Necessary condition solvability of the inner problem:
Z Z
Φ(r) dr + f (r) dS = 0. (4)
V S
The solution of the second boundary value problem can be written as

Z Z
w(r) = Φ(ρ)G(r, ρ) dVρ + f (ρ)G(r, ρ) dSρ + C, (5)
V S
where C is an arbitrary constant, provided that the solvability condition is met.

The Green’s function G = G(r, ρ) of the second boundary value problem is determined
1◦ . The function G satisfies the Laplace equation with respect to x, y, z in the domain V
1 1
everywhere except for the point (ξ, η, ζ) at which it has a singularity of the form 4π |r−ρ| .
2◦ . The function G, with respect to x, y, z, satisfies the homogeneous condition of the

second kind at the boundary, i.e., the condition

∂G 1
= ,
∂N S S0
where S0 is the area of the surface S.

The Green’s function is unique up to an additive constant.
Remark 10.2. The Green’s function cannot be identified with condition 1◦ and the homoge-
∂G
neous boundary condition ∂N S
= 0; this problem for G has no solution, because, on represent-
ing G in the form (3), for u we obtain a problem with a nonhomogeneous boundary condition of the
second kind, for which the solvability condition (2) is not met.
Remark 10.3. Condition (4) is not extended to the outer second boundary value problem (for
infinite domain).

The solution of the third boundary value problem for the Poisson equation (1) in a bounded
domain V with the nonhomogeneous boundary condition
∂w
+ kw = f (r) for r ∈ S
∂N
is given by relation (5) with C = 0, where G = G(r, ρ) is the Green’s function of the third
boundary value problem; the Green’s function is determined by the following conditions:
1◦ . The function G satisfies the Laplace equation with respect to x, y, z in V everywhere
1 1
except for the point (ξ, η, ζ) at which it has a singularity of the form 4π |r−ρ| .
2◦ . The function G, with respect to x, y, z, satisfiesthe homogeneous
boundary condition
∂G
of the third kind at the boundary, i.e., the condition ∂N + kG S = 0.
The Green’s function can be represented in the form (3), where the auxiliary function u
is determined by solving the corresponding third boundary value problem for the Laplace
equation ∆3 u = 0.
The construction of Green’s functions is discussed in Section 10.3.2 for λ = 0.
⊙ Literature for Section 10.2.1: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), N. S. Koshlya-
kov, E. B. Gliner, and M. M. Smirnov (1970).

The three-dimensional Poisson equation in the rectangular Cartesian system of coordinates
has the form
∂2w ∂2w ∂2w
2
+ 2
+ + Φ(x, y, z) = 0.
∂x ∂y ∂z 2
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, −∞ < z < ∞.

Solution:
Z ∞Z ∞Z ∞
1 Φ(ξ, η, ζ) dξ dη dζ
w(x, y, z) = p .
4π −∞ −∞ −∞ (x − ξ)2 + (y − η)2 + (z − ζ)2
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, 0 ≤ z < ∞.

First boundary value problem.
w = f (x, y) at z = 0.
Solution:
Z ∞Z ∞
1 zf (ξ, η) dξ dη
w(x, y, z) = 3/2
2π (x − ξ)2 + (y − η)2 + z 2
−∞ −∞
Z ∞Z ∞ Z ∞
1 1 1
+ − Φ(ξ, η, ζ) dξ dη dζ,
4π 0 −∞ −∞ R− R+
where
p p
R− = (x − ξ)2 + (y − η)2 + (z − ζ)2 , R+ = (x − ξ)2 + (y − η)2 + (z + ζ)2 .
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, 0 ≤ z < ∞. Third boundary value

problem.
∂z w − kw = f (x, y) at z = 0.
Solution:
Z ∞Z ∞
w(x, y, z) = − f (ξ, η)G(x, y, z, ξ, η, 0) dξ dη
Z ∞−∞
Z ∞−∞
Z ∞
+ Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dξ dη dζ,
0 −∞ −∞
where

1 1 1
G(x, y, z, ξ, η, ζ)= p +p
4π (x−ξ)2+(y−η)2+(z−ζ)2 (x−ξ)2+(y−η)2+(z+ζ)2
Z ∞
exp(−ks) ds
−2k p .
0 (x−ξ)2+(y−η)2+(z+ζ+s)2
◮ Domain: −∞ < x < ∞, 0 ≤ y < ∞, 0 ≤ z < ∞. First boundary value problem.

A dihedral angle is considered. Boundary conditions are prescribed:
w = f1 (x, z) at y = 0, w = f2 (x, y) at z = 0.
Solution:
Z ∞Z ∞
∂
w(x, y, z) = f1 (ξ, ζ) G(x, y, z, ξ, η, ζ) dξ dζ
0 −∞ ∂η η=0
Z ∞Z ∞
∂
+ f2 (ξ, η) G(x, y, z, ξ, η, ζ) dξ dη
0 −∞ ∂ζ ζ=0
Z ∞Z ∞ Z ∞
0 0 −∞
where

1 1 1
G(x, y, z, ξ, η, ζ)= p −p
4π 2 2
(x−ξ) +(y−η) +(z−ζ)2 (x−ξ) +(y−η)2+(z+ζ)2
2

1 1
−p +p .
(x−ξ)2+(y+η)2+(z−ζ)2 (x−ξ)2+(y+η)2+(z+ζ)2
◮ Domain: 0 ≤ x < ∞, 0 ≤ y < ∞, 0 ≤ z < ∞. First boundary value problem.

An octant is considered. Boundary conditions are prescribed:
w = f1 (y, z) at x = 0, w = f2 (x, z) at y = 0, w = f3 (x, y) at z = 0.
Solution:
Z ∞Z ∞
∂
w(x, y, z) = f1 (η, ζ) G(x, y, z, ξ, η, ζ) dη dζ
0 0 ∂ξ ξ=0
Z ∞Z ∞
∂
+ f2 (ξ, ζ) G(x, y, z, ξ, η, ζ) dξ dζ
0 0 ∂η η=0
Z ∞Z ∞
∂
+ f3 (ξ, η) G(x, y, z, ξ, η, ζ) dξ dη
0 0 ∂ζ ζ=0
Z ∞Z ∞Z ∞
0 0 0
where

1 1 1
G(x, y, z, ξ, η, ζ)= p −p
4π 2 2
(x−ξ) +(y−η) +(z−ζ)2 (x−ξ) +(y−η)2+(z+ζ)2
2
1 1
−p +p
(x−ξ)2+(y+η)2+(z−ζ)2 (x−ξ)2+(y+η)2+(z+ζ)2
1 1
−p +p
2 2
(x+ξ) +(y−η) +(z−ζ) 2 (x+ξ) +(y−η)2+(z+ζ)2
2

1 1
+p −p .
(x+ξ)2+(y+η)2+(z−ζ)2 (x+ξ)2+(y+η)2+(z+ζ)2
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, 0 ≤ z ≤ a. First boundary value

problem.
An infinite layer is considered. Boundary conditions are prescribed:
w = f1 (x, y) at z = 0, w = f2 (x, y) at z = a.
Solution:
Z ∞Z ∞
∂
w(x, y, z) = f1 (ξ, η) G(x, y, z, ξ, η, ζ) dξ dη
−∞ −∞ ∂ζ ζ=0
Z ∞Z ∞
∂
− f2 (ξ, η) G(x, y, z, ξ, η, ζ) dξ dη
−∞ −∞ ∂ζ ζ=a
Z aZ ∞ Z ∞
+ Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dξ dη dζ.
0 −∞ −∞
Green’s function:
∞
1 X 1 1
G(x, y, z, ξ, η, ζ) = − ,
4π n=−∞ rn1 rn2
where p
rn1 = (x − ξ)2 + (y − η)2 + (z − ζ − 2na)2 ,
p
rn2 = (x − ξ)2 + (y − η)2 + (z + ζ − 2na)2 .
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, 0 ≤ z ≤ a. Mixed boundary value

problem.
w = f1 (x, y) at z = 0, ∂z w = f2 (x, y) at z = a.
Solution:
Z ∞Z ∞
∂
−∞ −∞ ∂ζ ζ=0
Z ∞Z ∞
+ f2 (ξ, η)G(x, y, z, ξ, η, a) dξ dη
Z−∞Z
a ∞
−∞
Z ∞
0 −∞ −∞
Green’s function:
∞
1 X 1 1
G(x, y, z, ξ, η, ζ) = − ,
4π n=−∞ rn1 rn2
where p
rn1 = (x − ξ)2 + (y − η)2 + [z − (−1)n ζ − 2na]2 ,
p
rn2 = (x − ξ)2 + (y − η)2 + [z + (−1)n ζ − 2na]2 .
◮ Domain: 0 ≤ x < ∞, −∞ < y < ∞, 0 ≤ z ≤ a. First boundary value problem.
A semiinfinite layer is considered. Boundary conditions are prescribed:
w = f1 (y, z) at x = 0, w = f2 (x, y) at z = 0, w = f3 (x, y) at z = a.

Solution:
Z aZ ∞
∂
w(x, y, z) = f1 (η, ζ) G(x, y, z, ξ, η, ζ) dη dζ
0 −∞ ∂ξ ξ=0
Z ∞Z ∞
∂
+ f2 (ξ, η) G(x, y, z, ξ, η, ζ) dξ dη
−∞ 0 ∂ζ ζ=0
Z ∞Z ∞
∂
− f3 (ξ, η) G(x, y, z, ξ, η, ζ) dξ dη
−∞ 0 ∂ζ ζ=a
Z aZ ∞ Z ∞
0 −∞ 0
Green’s function:
∞
1 X 1 1 1 1
G(x, y, z, ξ, η, ζ) = − − + ,
4π n=−∞ rn1 rn2 rn3 rn4
where p
rn1 = (x − ξ)2 + (y − η)2 + (z − ζ − 2na)2 ,
p
rn2 = (x − ξ)2 + (y − η)2 + (z + ζ − 2na)2 ,
p
rn3 = (x + ξ)2 + (y − η)2 + (z − ζ − 2na)2 ,
p
rn4 = (x + ξ)2 + (y − η)2 + (z + ζ − 2na)2 .
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, −∞ < z < ∞. First boundary value problem.

An infinite cylindrical domain of a rectangular cross-section is considered. Boundary con-
ditions are prescribed:
w = f3 (x, z) at y = 0, w = f4 (x, z) at y = b.
Solution:
Z bZ ∞
∂
w(x, y, z) = f1 (η, ζ) G(x, y, z, ξ, η, ζ) dζ dη
0 −∞ ∂ξ ξ=0
Z bZ ∞
∂
− f2 (η, ζ) G(x, y, z, ξ, η, ζ) dζ dη
0 −∞ ∂ξ ξ=a
Z aZ ∞
∂
+ f3 (ξ, ζ) G(x, y, z, ξ, η, ζ) dζ dξ
0 −∞ ∂η η=0
Z aZ ∞
∂
− f4 (ξ, ζ) G(x, y, z, ξ, η, ζ) dζ dξ
0 −∞ ∂η η=b
Z aZ bZ ∞
+ Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dζ dη dξ.
0 0 −∞
Green’s function:
∞ ∞
2 XX 1
G(x, y, z, ξ, η, ζ) = sin(pn x) sin(qm y)
ab n=1 m=1 βnm

× sin(pn ξ) sin(qm η) exp(−βnm |z − ζ|) ,
nπ mπ p
pn = , qm = , βnm = p2n + qm
2 .
a b
Alternatively, the Green’s function can be represented as
∞ ∞
1 X X 1 1 1 1
G(x, y, z, ξ, η, ζ) = (1)
− (2) − (3) + (4) ,
4π n=−∞ m=−∞ rnm rnm rnm rnm
where p
(1)
rnm = (x − ξ − 2na)2 + (y − η − 2mb)2 + (z − ζ)2 ,
(2)
p
rnm = (x + ξ − 2na)2 + (y − η − 2mb)2 + (z − ζ)2 ,
(3)
p
rnm = (x − ξ − 2na)2 + (y + η − 2mb)2 + (z − ζ)2 ,
(4)
p
rnm = (x + ξ − 2na)2 + (y + η − 2mb)2 + (z − ζ)2 .
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, −∞ < z < ∞. Third boundary value problem.

∂x w − k1 w = f1 (y, z) at x = 0, ∂x w + k2 w = f2 (y, z) at x = a,
∂y w − k3 w = f3 (x, z) at y = 0, ∂y w + k4 w = f4 (x, z) at y = b.
Solution:
Z bZ ∞
w(x, y, z) = − f1 (η, ζ)G(x, y, z, 0, η, ζ) dζ dη
0 −∞
Z bZ ∞
+ f2 (η, ζ)G(x, y, z, a, η, ζ) dζ dη
0 −∞
Z aZ ∞
− f3 (ξ, ζ)G(x, y, z, ξ, 0, ζ) dζ dξ
0 −∞
Z aZ ∞
+ f4 (ξ, ζ)G(x, y, z, ξ, b, ζ) dζ dξ
0 −∞
Z aZ bZ ∞
0 0 −∞
Green’s function:
∞ ∞
1 X X unm (x, y)unm (ξ, η)
G(x, y, z, ξ, η, ζ) = exp(−βnm |z − ζ|),
2 n=1 m=1 kunm k2 βnm
where
p
wnm(x, y) = (µn cos µnx+k1 sin µnx)(νm cos νmy+k3 sin νmy), βnm = µ2n +νm 2,
2 +k k )
2 1 2 2 2 2 (k1 +k2)(µ2n +k1k2) (k3 +k4)(νm 3 4
kwnmk = (µn +k1 )(νm +k3 ) a+ 2 2 b+ 2 .
4 2 2
(µn +k1 )(µn +k2 ) (νm +k3 )(νm +k42)
2 2
(k1 + k2 )µ (k3 + k4 )ν
µ2 − k1 k2 ν 2 − k3 k4
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, −∞ < z < ∞. Mixed boundary value problems.
1◦ . An infinite cylindrical domain of a rectangular cross-section is considered. Boundary

w = f1 (y, z) at x = 0, ∂x w = f2 (y, z) at x = a,
w = f3 (x, z) at y = 0, ∂y w = f4 (x, z) at y = b.
Solution:
Z bZ ∞
∂
0 −∞ ∂ξ ξ=0
Z bZ ∞
+ f2 (η, ζ)G(x, y, z, a, η, ζ) dζ dη
0 −∞
Z aZ ∞
∂
+ f3 (ξ, ζ) G(x, y, z, ξ, η, ζ) dζ dξ
0 −∞ ∂η η=0
Z aZ ∞
+ f4 (ξ, ζ)G(x, y, z, ξ, b, ζ) dζ dξ
0 −∞
Z aZ bZ ∞
0 0 −∞
Green’s function:
∞ ∞
2 XX 1
G(x, y, z, ξ, η, ζ) = sin(pn x) sin(qm y) sin(pn ξ) sin(qm η) exp(−βnm |z − ζ|),
ab n=0 m=0 βnm
where
(2n + 1)π (2m + 1)π p
pn = , qm = , βnm = p2n + qm
2 .
2a 2b
∂y w = f3 (x, z) at y = 0, ∂y w = f4 (x, z) at y = b.
Solution:
Z bZ ∞
∂
0 −∞ ∂ξ ξ=0
Z bZ ∞
∂
− f2 (η, ζ) G(x, y, z, ξ, η, ζ) dζ dη
0 −∞ ∂ξ ξ=a
Z aZ ∞
− f3 (ξ, ζ)G(x, y, z, ξ, 0, ζ) dζ dξ
0 −∞
Z aZ ∞
+ f4 (ξ, ζ)G(x, y, z, ξ, b, ζ) dζ dξ
0 −∞
Z aZ bZ ∞
0 0 −∞
Green’s function:
∞ ∞
1 X X Am
G(x, y, z, ξ, η, ζ) = sin(pn x) cos(qm y) sin(pn ξ) cos(qm η) exp(−βnm |z − ζ|),
ab n=1 m=0 βnm
where
(
pn = , qm = , βnm = p2n + qm
2 , Am =
a b 2 for m 6= 0.
◆ Below, in Section 10.2.2, we present only Green’s functions; the complete solution is
constructed with the formulas given in Section 10.2.1.
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ x, −∞ < z < ∞. First boundary value problem.

An infinite cylindrical domain of triangular cross-section is considered. Boundary condi-
tions are prescribed:
w = f1 (y, z) at x = 0, w = f2 (x, z) at y = 0, w = f3 (x, z) at y = x.
Green’s function:
G(x, y, z, ξ, η, ζ) = H(x, y, z, ξ, η, ζ) − H(x, y, z, η, ξ, ζ),
where
∞ ∞
2 XX 1
H(x, y, z, ξ, η, ζ) = sin(pn x) sin(pm y) sin(pn ξ) sin(pm η) exp(−βnm |z − ζ|),
a2 n=1 m=1 βnm
nπ mπ p
pn = , pm = , βnm = p2n + p2m .
a a
An alternative representation of the Green’s function can be obtained by setting
∞ ∞
1 X X 1 1 1 1
H(x, y, z, ξ, η, ζ) = (1)
− (2)
− (3)
+ (4)
,
4π n=−∞ m=−∞ rnm rnm rnm rnm
where p
(1)
rnm = (x − ξ − 2na)2 + (y − η − 2ma)2 + (z − ζ)2 ,
(2)
p
rnm = (x + ξ − 2na)2 + (y − η − 2ma)2 + (z − ζ)2 ,
(3)
p
rnm = (x − ξ − 2na)2 + (y + η − 2ma)2 + (z − ζ)2 ,
(4)
p
rnm = (x + ξ − 2na)2 + (y + η − 2ma)2 + (z − ζ)2 .
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ≤ z < ∞. First boundary value problem.
A semiinfinite cylindrical domain of a rectangular cross-section is considered. Boundary

w = f5 (x, y) at z = 0.
Green’s function:
∞ ∞
4 XX 1
G(x, y, z, ξ, η, ζ) = sin(pn x) sin(qm y) sin(pn ξ) sin(qm η)Hnm (z, ζ),
ab n=1 m=1 βnm
nπ mπ p
pn = , qm = , βnm = p2n + qm 2 ,
a( b
exp(−βnm z) sinh(βnm ζ) for z > ζ ≥ 0,
Hnm (z, ζ) =
exp(−βnm ζ) sinh(βnm z) for ζ > z ≥ 0.
An alternative representation of the Green’s function:

∞ ∞
1 X X 1 1 1 1 1 1 1 1
G(x, y, z, ξ, η, ζ) = (1)
− (2)
− (3)
+ (4)
− (5)
+ (6)
+ (7)
− (8)
,
4π n=−∞ m=−∞ rnm rnm rnm rnm rnm rnm rnm rnm
where p
(1)
rnm = (x − ξ − 2na)2 + (y − η − 2mb)2 + (z − ζ)2 ,
(2)
p
rnm = (x + ξ − 2na)2 + (y − η − 2mb)2 + (z − ζ)2 ,
(3)
p
rnm = (x − ξ − 2na)2 + (y + η − 2mb)2 + (z − ζ)2 ,
p
(4)
rnm = (x + ξ − 2na)2 + (y + η − 2mb)2 + (z − ζ)2 ,
p
(5)
rnm = (x − ξ − 2na)2 + (y − η − 2mb)2 + (z + ζ)2 ,
(6)
p
rnm = (x + ξ − 2na)2 + (y − η − 2mb)2 + (z + ζ)2 ,
(7)
p
rnm = (x − ξ − 2na)2 + (y + η − 2mb)2 + (z + ζ)2 ,
(8)
p
rnm = (x + ξ − 2na)2 + (y + η − 2mb)2 + (z + ζ)2 .
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ≤ z < ∞. Third boundary value problem.

∂y w − k3 w = f3 (x, z) at y = 0, ∂y w + k4 w = f4 (x, z) at y = b,
∂z w − k5 w = f5 (x, y) at z = 0.
Green’s function:
X∞ X ∞
unm (x, y)unm (ξ, η)
G(x, y, z, ξ, η, ζ) = Hnm (z, ζ),
n=1 m=1
kunm k2
where
wnm(x, y) = (µn cos µnx+k1 sin µnx)(νm cos νmy +k3 sin νmy),

2 1 2 2 2 2 (k1 +k2)(µ2n +k1k2) (k3 +k4)(νm2
+k3k4)
kwnmk = (µn +k1 )(νm +k3 ) a+ b+ 2 ,
4 (µ2n +k12)(µ2n +k22) (νm +k32)(νm2 +k 2)
4


 exp(−βnmz) βnm cosh(βnmζ)+k5 sinh(βnmζ)
 if z > ζ, p
Hnm(z, ζ) = βnm(βnm +k5) βnm = µ2n +νm 2.

 exp(−β nm ζ) β nm cosh(β nm z)+k 5 sinh(β nm z)
 if ζ > z,
βnm(βnm +k5)
(k1 + k2 )µ (k3 + k4 )ν
µ2 − k1 k2 ν 2 − k3 k4
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ≤ z < ∞. Mixed boundary value problems.

1◦ . A semiinfinite cylindrical domain of a rectangular cross-section is considered. Bound-
ary conditions are prescribed:
∂z w = f5 (x, y) at z = 0.
Green’s function:
∞ ∞
4 XX 1
ab n=1 m=1 βnm
nπ mπ p
pn = , qm = , βnm = p2n + qm 2 ,
a( b
exp(−βnm z) cosh(βnm ζ) for z > ζ ≥ 0,
Hnm (z, ζ) =
exp(−βnm ζ) cosh(βnm z) for ζ > z ≥ 0.

∂x w = f1 (y, z) at x = 0, ∂x w = f2 (y, z) at x = a,
∂y w = f3 (x, z) at y = 0, ∂y w = f4 (x, z) at y = b,
w = f5 (x, y) at z = 0.
Green’s function:
∞ ∞
1 X X An Am
G(x, y, z, ξ, η, ζ) = cos(pn x) cos(qm y) cos(pn ξ) cos(qm η)Hnm (z, ζ),
ab βnm
n=0 m=0
(
nπ mπ p 1 for n = 0,
pn = , qm = , βnm = p2n + qm 2 , An =
a b 2 for n 6= 0,
(
Hnm (z, ζ) =

1◦ . Representation of the Green’s function in the form of a double series:
∞ ∞
4 XX
G(x, y, z, ξ, η, ζ) = sin(pn x) sin(qm y) sin(pn ξ) sin(qm η)Fnm (z, ζ),
ab n=1 m=1


 sinh(βnm ζ) sinh[βnm (c − z)]
 for c ≥ z > ζ ≥ 0,
Fnm (z, ζ) = βnm sinh(βnm c)

 sinh(βnm z) sinh[βnm (c − ζ)]
 for c ≥ ζ > z ≥ 0,
βnm sinh(βnm c)
πn πm p
pn = , qm = , βnm = p2n + qm 2 .
a b
This relation can be used to obtain two other representations of the Green’s function by
means of the following cyclic permutations:
(x, ξ, a)
ր ց
(z, ζ, c) ←− (y, η, b)
2◦ . Representation of the Green’s function in the form of a triple series:
∞ ∞ ∞
8 X X X sin(pn x) sin(qm y) sin(sk z) sin(pn ξ) sin(qm η) sin(sk ζ)
G(x, y, z, ξ, η, ζ) = 2 + s2
,
abc n=1 m=1 p2n + qm k
k=1
πn πm πk
pn = , qm = , sk = .
a b c
3◦ . An alternative representation of the Green’s function in the form of a triple series:

∞ ∞ ∞
1 X X X 1 1 1 1
G(x, y, z, ξ, η, ζ) = (1)
− (2) − (3) + (4)
4π n=−∞ m=−∞
k=−∞ rnmk rnmk rnmk rnmk

1 1 1 1
− (5)
+ (6)
+ (7)
− (8)
,
rnmk rnmk rnmk rnmk
where p
(1)
rnmk = (x − ξ − 2na)2 + (y − η − 2mb)2 + (z − ζ − 2kc)2 ,
(2)
p
rnmk = (x + ξ − 2na)2 + (y − η − 2mb)2 + (z − ζ − 2kc)2 ,
(3)
p
rnmk = (x − ξ − 2na)2 + (y + η − 2mb)2 + (z − ζ − 2kc)2 ,
(4)
p
rnmk = (x + ξ − 2na)2 + (y + η − 2mb)2 + (z − ζ − 2kc)2 ,
(5)
p
rnmk = (x − ξ − 2na)2 + (y − η − 2mb)2 + (z + ζ − 2kc)2 ,
(6)
p
rnmk = (x + ξ − 2na)2 + (y − η − 2mb)2 + (z + ζ − 2kc)2 ,
(7)
p
rnmk = (x − ξ − 2na)2 + (y + η − 2mb)2 + (z + ζ − 2kc)2 ,
(8)
p
rnmk = (x + ξ − 2na)2 + (y + η − 2mb)2 + (z + ζ − 2kc)2 .
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ≤ z ≤ c. Third boundary value problem.
∂z w − k5 w = f5 (x, y) at z = 0, ∂z w + k6 w = f6 (x, y) at z = c.
Green’s function:
X∞ X ∞ X ∞
ϕn(x)ϕn(ξ)ψm(y)ψm(η)χs(z)χs(ζ)
G(x, y, z, ξ, η, ζ) = ,
kϕnk2kψmk2kχs(ζ)k2(µ2n +λ2m +νs2)
n=1 m=1 s=1

k1 k2 µ2 +k2 k1 a k2
ϕn(x) = cos(µnx)+ sin(µnx), kϕnk2 = 2 n2 12 + 2 + 1+ 12 ,
µn 2µn µn +k2 2µn 2 µn
2 2
k3 2 k4 λm +k3 k3 b k32
ψm(y) = cos(λmy)+ sin(λmy), kψmk = 2 2 + + 1+ 2 ,
2 2
2

k5 k6 ν +k k5 c k
χs(z) = cos(νsz)+ sin(νsz), kχsk2 = 2 s2 52 + 2 + 1+ 52 ,
νs 2νs νs +k6 2νs 2 νs
where the µn , λm , and νs are positive roots of the transcendental equations
tan(µa) k1 + k2 tan(λb) k3 + k4 tan(νc) k5 + k6

= 2 , = 2 , = 2 .
µ µ − k1 k2 λ λ − k3 k4 ν ν − k5 k6
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ≤ z ≤ c. Mixed boundary value problem.

w = f3 (x, z) at y = 0, ∂y w = f4 (x, z) at y = b,
w = f5 (x, y) at z = 0, ∂z w = f6 (x, y) at z = c.
Solution:
Z aZ bZ c
w(x, y, z)= Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dζ dη dξ
0 0 0
Z bZ c Z bZ c
∂
+ f1(η, ζ) G(x, y, z, ξ, η, ζ) dζ dη+ f2(η, ζ)G(x, y, z, a, η, ζ) dζ dη
0 0 ∂ξ ξ=0 0 0
Z aZ c Z aZ c
∂
+ f3(ξ, ζ) G(x, y, z, ξ, η, ζ) dζ dξ+ f4(ξ, ζ)G(x, y, z, ξ, b, ζ) dζ dξ
0 0 ∂η η=0 0 0
Z aZ b Z aZ b
∂
+ f5(ξ, η) G(x, y, z, ξ, η, ζ) dη dξ+ f6(ξ, η)G(x, y, z, ξ, η, c) dη dξ.
0 0 ∂ζ ζ=0 0 0
1◦ . A double-series representation of the Green’s function:

∞ ∞
4 XX
G(x, y, z, ξ, η, ζ) = sin(pn x) sin(qm y) sin(pn ξ) sin(qm η)Fnm (z, ζ),
ab n=0 m=0


 sinh(βnm ζ) cosh[βnm (c − z)]
 for c ≥ z > ζ ≥ 0,
Fnm (z, ζ) = βnm cosh(β nm c)

 sinh(βnm z) cosh[βnm (c − ζ)]
 for c ≥ ζ > z ≥ 0,
βnm cosh(βnm c)
π(2n + 1) π(2m + 1) p
pn = , qm = , βnm = p2n + qm 2 .
2a 2b
This relation can be used to obtain two other representations of the Green’s function by
means of the following cyclic permutations:
(x, ξ, a)
ր ց
(z, ζ, c) ←− (y, η, b)
2◦ . A triple series representation of the Green’s function:
∞ ∞ ∞
8 X X X sin(pn x) sin(qm y) sin(sk z) sin(pn ξ) sin(qm η) sin(sk ζ)
G(x, y, z, ξ, η, ζ) = 2 + s2
,
abc n=0 m=0 p2n + qm k
k=0
π(2n + 1) π(2m + 1) π(2k + 1)
pn = , qm = , sk = .
2a 2b 2c

The three-dimensional Poisson equation in the cylindrical coordinate system is written as
p
1 ∂ ∂w 1 ∂2w ∂2w
r + 2 + = −Φ(r, ϕ, z), r = x2 + y 2 .
r ∂r ∂r r ∂ϕ2 ∂z 2
w = f (ϕ, z) at r = R.
Solution:
Z 2πZ ∞
∂
w(r, ϕ, z) = −R f (η, ζ) G(r, ϕ, z, ξ, η, ζ) dζ dη
0 −∞ ∂ξ ξ=R
Z R Z 2πZ ∞
+ Φ(ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ)ξ dζ dη dξ.
0 0 −∞
Green’s function:
∞ ∞
1 X X An Jn (µnm r)Jn (µnm ξ)
G(r, ϕ, z, ξ, η, ζ) = 2 cos[n(ϕ − η)] exp −µnm |z − ζ| ,
2πR2 n=0 m=1 ′
Jn (µnm R) µnm
where A0 = 1 and An = 2 for n 6= 0; the Jn (ξ) are Bessel functions; and the µnm are
positive roots of the transcendental equation Jn (µR) = 0.

problem.
∂r w + kw = f (ϕ, z) at r = R.
Solution:
Z 2πZ ∞
w(r, ϕ, z) = R f (η, ζ)G(r, ϕ, z, R, η, ζ) dζ dη
0 −∞
Z R Z 2πZ ∞
0 0 −∞
Green’s function:
∞ ∞
1 X X An µnm Jn (µnm r)Jn (µnm ξ) cos[n(ϕ − η)]
G(r, ϕ, z, ξ, η, ζ) = exp −µnm |z − ζ| ,
2π n=0 m=1 (µ2nm R2 + k 2 R2 − n2 )Jn2 (µnm R)
where A0 = 1 and An = 2 for n 6= 0; the Jn (ξ) are Bessel functions; and the µnm are
positive roots of the transcendental equation


A semiinfinite circular cylinder is considered. Boundary conditions are prescribed:
w = f1 (ϕ, z) at r = R, w = f2 (r, ϕ) at z = 0.
Solution:
Z
2πZ ∞
∂
w(r, ϕ, z) = −R f1 (η, ζ) G(r, ϕ, z, ξ, η, ζ) dζ dη
0 0 ∂ξ ξ=R
Z 2πZ R
∂
+ f2 (ξ, η) G(r, ϕ, z, ξ, η, ζ) ξ dξ dη
0 0 ∂ζ ζ=0
Z R Z 2πZ ∞
0 0 0
Green’s function:
∞ ∞
G(r, ϕ, z, ξ, η, ζ) = 2 cos[n(ϕ − η)]Fnm (z, ζ),
2πR2 n=0 m=1 Jn′ (µnm R) µnm
(
1 for n = 0,
Fnm (z, ζ) = exp(−µnm |z − ζ|) − exp(−µnm |z + ζ|), An =
2 for n 6= 0,
where the µnm are positive roots of the transcendental equation Jn (µR) = 0.

∂r w + k1 w = f1 (ϕ, z) at r = R, ∂z w − k2 w = f2 (r, ϕ) at z = 0.
Solution:
Z 2πZ ∞
w(r, ϕ, z) = R f1 (η, ζ)G(r, ϕ, z, R, η, ζ) dζ dη
0 0
Z 2πZ R
− f2 (ξ, η)G(r, ϕ, z, R, η, 0)ξ dξ dη
0 0
Z R Z 2πZ ∞
0 0 0
Green’s function:
∞ ∞
1 X X Anµ2nmJn(µnmr)Jn(µnmξ) cos[n(ϕ−η)]
G(r, ϕ, z, ξ, η, ζ)= Fnm(z, ζ),
π n=0 m=1 (µ2nmR2+k12R2−n2)Jn2(µnmR)

( 
 exp(−µnmz) µnm cosh(µnmζ)+k2 sinh(µnmζ)
 for z >ζ,
1 for n=0, µnm(µnm+k2)
An = Fnm(z, ζ)=
2 for n6= 0, 
 exp(−µnmζ) µnm cosh(µnmz)+k2 sinh(µnmz)
 for ζ >z,
µnm(µnm+k2)
equation
µJn′ (µR) + k1 Jn (µR) = 0.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞. Mixed boundary value problem.

w = f1 (ϕ, z) at r = R, ∂z w = f2 (r, ϕ) at z = 0.
Solution:
Z 2πZ ∞
∂
0 0 ∂ξ ξ=R
Z 2πZ R
− f2 (ξ, η)G(r, ϕ, z, ξ, η, 0)ξ dξ dη
0 0
Z R Z 2πZ ∞
0 0 0
Green’s function:
∞ ∞
2πR2 Jn′ (µnm R) µnm
n=0 m=1
(
1 for n = 0,
Fnm (z, ζ) = exp(−µnm |z − ζ|) + exp(−µnm |z + ζ|), An =
2 for n 6= 0,
where the Jn (ξ) are Bessel functions and the µnm are roots of the transcendental equation
Jn (µR) = 0.
◆ Below, in Section 10.2.3, we present only Green’s functions; the complete solution is
constructed with the formulas given in Section 10.2.1. See also Section 10.3.2 for λ = 0.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ a. First boundary value problem.

w = f1 (ϕ, z) at r = R, w = f2 (r, ϕ) at z = 0, w = f3 (r, ϕ) at z = a.
A double series representation of the Green’s function:
∞ ∞
1 XX AnJn(µnmr)Jn(µnmξ)
G(r, ϕ, z, ξ, η, ζ)= 2 2 cos[n(ϕ−η)]Fnm(z, ζ),
πR ′
n=0 m=1 Jn(µnmR) µnm sinh(µnma)
( (
sinh(µnmζ) sinh[µnm(a−z)] for a≥z>ζ ≥0, 1 for n=0,
Fnm(z, ζ)= An =
sinh(µnmz) sinh[µnm(a−ζ)] for a≥ζ >z≥0, 2 for n6= 0,
A triple series representation of the Green’s function:
∞ ∞ ∞
2a X X X A
G(r, ϕ, z, ξ, η, ζ) = 2
n Jn (µnm r)
πR
n=0 m=1 k=1
[Jn (µnm R)] (aµnm )2 + (πk)2
′ 2

kπz kπζ
×Jn (µnm ξ) cos[n(ϕ − η)] sin sin .
a a
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ a. Third boundary value problem.

∂r w + k1 w = f1 (ϕ, z) at r = R,
∂z w − k2 w = f2 (r, ϕ) at z = 0,
∂z w + k3 w = f3 (r, ϕ) at z = a.
Green’s function:
∞ ∞ ∞
1 X X X Anµ2nmJn (µnmr)Jn(µnmξ) cos[n(ϕ − η)]hs (z)hs (ζ)
G(r, ϕ, z, ξ, η, ζ) = 2 R2 + k 2 R2 − n2)(µ2 + λ2)[J (µ 2 kh k2
,
π (µ nm 1 nm s n nm R)] s
n=0 m=1 s=1

k2 2 k3 λ2s + k22 k2 a k22
hs (z) = cos(λsz) + sin(λs z), khs k = 2 2 + + 1+ 2 .
λs 2λs λs + k32 2λ2s 2 λs
Here, A0 = 1 and An = 2 for n 6= 0; the Jn (ξ) are Bessel functions; and the µnm and λs
are positive roots of the transcendental equations
tan(λa) k2 + k3
µJn′ (µR) + k1 Jn (µR) = 0, = 2 .
λ λ − k2 k3
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ a. Mixed boundary value problem.

w = f1 (ϕ, z) at r = R, ∂z w = f2 (r, ϕ) at z = 0, ∂z w = f3 (r, ϕ) at z = a.
Green’s function:
∞ ∞ ∞
a XXX A A
G(r, ϕ, z, ξ, η, ζ) = 2
n k Jn (µnm r)
πR
n=0 m=1 k=1
[Jn (µnm R)] (aµnm )2 + (πk)2
′ 2

kπz kπζ
×Jn (µnm ξ) cos[n(ϕ − η)] cos cos ,
a a
where A0 = 1 and An = 2 for n 6= 0; the Jn (ξ) are Bessel functions (the prime denotes a
derivative with respect to the argument); and the µnm are positive roots of the transcenden-
tal equation Jn (µR) = 0.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ a. First boundary value problem.

A cylindrical sector of finite thickness is considered. Boundary conditions are prescribed:
w = f1(r, z) at ϕ = 0, w = f2(r, z) at ϕ = ϕ0, w = f3(ϕ, z) at r = R,
w = f4(r, ϕ) at z = 0, w = f5(r, ϕ) at z = a.
Green’s function:
∞ ∞ ∞
8a X X X Jnπ/ϕ0 (µnm r)Jnπ/ϕ0 (µnm ξ)
G(r, ϕ, z, ξ, η, ζ) = 2 ′

R ϕ0 [J (µ R)]2 (aµ )2 + (πk)2
n=1 m=1 k=1 nπ/ϕ0 nm nm

nπϕ nπη kπz kπζ
× sin sin sin sin ,
ϕ0 ϕ0 a a
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ a. Mixed boundary value problem.

w=f1(r, z) at ϕ=0, w=f2(r, z) at ϕ=ϕ0, w=f3(ϕ, z) at r=R,

∂z w=f4(r, ϕ) at z=0, ∂z w=f5(r, ϕ) at z=a.
Green’s function:
∞ ∞ ∞
4a X X X Ak Jnπ/ϕ0 (µnm r)Jnπ/ϕ0 (µnm ξ)
G(r, ϕ, z, ξ, η, ζ) =
R2 ϕ0 ′
[Jnπ/ϕ (µnm R)]2 (aµnm )2 + (πk)2
n=1 m=1 k=0 0

nπϕ nπη kπz kπζ
× sin sin cos cos ,
ϕ0 ϕ0 a a
where A0 = 1 and Ak = 2 for k 6= 0; the Jnπ/ϕ0 (r) are Bessel functions; and the µnm are

The three-dimensional Poisson equation in the spherical coordinate system is written as

1 ∂ 2 ∂w 1 ∂ ∂w 1 ∂2w
r + sin θ + = −Φ(r, θ, ϕ),
p
where r = x2 + y 2 + z 2 .
◆ Only Green’s functions are presented below; the complete solutions can be constructed
with the formulas given in Section 10.2.1.

A spherical domain is considered. A boundary condition is prescribed:
w = f (ϕ, θ) at r = R.
Green’s function:
1 1
G(r, θ, ϕ, ξ, η, ζ) = p − p ,
4π r2 − 2rξ cos γ + ξ2 4π r2ξ 2 − 2R2 rξ cos γ + R4
cos γ = cos θ cos η + sin θ sin η cos(ϕ − ζ).
An alternative representation of the Green’s function:

1 1 1 R
G(r, r0 ) = − , r0 = |r0 |,
4π |r − r0 | 4π r0 |(R/r0 )2 r0 − r|
where
r = {x, y, z}, x = r sin θ cos ϕ, y = r sin θ sin ϕ, z = r cos θ
r0 = {x0 , y0 , z0 }, x0 = ξ sin η cos ζ, y0 = ξ sin η sin ζ, z0 = ξ cos η.

∂r w = f (ϕ, θ) at r = R.
Green’s function:

1 1 R 1 2R2
G(r, θ, ϕ, ξ, η, ζ) = + + ln 2 ,
4π |r − r0 | |r0 | |r1 | R R + |r0 | |r1 | − (r · r0 )
where
p p
|r − r0 | = r 2 − 2rξ cos γ + ξ 2 , |r0 | |r1 | = r 2 ξ 2 − 2R2 rξ cos γ + R4 ,
|r0 | = ξ, (r · r0 ) = rξ cos γ, cos γ = cos θ cos η + sin θ sin η cos(ϕ − ζ).
For a solution of the second boundary value problem to exist the solvability condition
must be satisfied (see Section 10.2.1).

∂r w + kw = f (θ, ϕ) at r = R.
Green’s function:
∞ X
X ∞ X
n
1
G(r, θ, ϕ, ξ, η, ζ)= √ AsBnmsJn+1/2(λnmr)Jn+1/2(λnmξ)
2π rξ n=0 m=1 s=0
×Pns(cos θ)Pns(cos η) cos[s(ϕ−ζ)],
(
1 for s=0, (2n+1)(n−s)!
As = Bnms = 2 .
2 for s6= 0, 2 2
(n+s)! R λnm+(kR+n)(kR−n−1) Jn+1/2(λnmR)
Here, the Jn+1/2 (r) are Bessel functions, the Pns (µ) are associated Legendre functions,
which are expressed in terms of the Legendre polynomials Pn (µ) as
ds 1 dn
Pns (µ) = (1 − µ2 )s/2 Pn (µ), Pn (µ) = (µ2 − 1)n ,
dµs n! 2n dµn
and the λnm are positive roots of the transcendental equation
′

λRJn+1/2 (λR) + kR − 12 Jn+1/2 (λR) = 0.
◮ Domain: R ≤ r < ∞, 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π. First boundary value problem.

Three-dimensional space with a spherical cavity is considered. A boundary condition is
prescribed:
w = f (ϕ, θ) at r = R.
The Green’s function of the outer first boundary value problem is given by the same
relation as that for the inner first boundary value problem (see the first paragraph in this
section), except that r ≥ R and ξ ≥ R.
◮ Domain: R ≤ r < ∞, 0 ≤ θ ≤ π, 0 ≤ ϕ ≤ 2π. Second boundary value problem.

Three-dimensional space with a spherical cavity is considered. A boundary condition is
prescribed:
∂r w = f (ϕ, θ) at r = R.
Green’s function:

1 1 R 1 (1 − cos γ)|r| |r0 |
G(r, θ, ϕ, ξ, η, ζ) = + + ln 2 ,
4π |r − r0 | |r0 | |r1 | R R + |r0 | |r1 | − (r · r0 )
where
p
|r| = r, |r0 | = ξ, |r − r0 | = r 2 − 2rξ cos γ + ξ 2 ,
p
|r0 | |r1 | = r 2 ξ 2 − 2R2 rξ cos γ + R4 , (r · r0 ) = rξ cos γ,

A spherical layer is considered. Boundary conditions are prescribed:
w = f1 (θ, ϕ) at r = R1 , w = f2 (θ, ϕ) at r = R2 .
Green’s function:
∞ ∞ n
π XXX
G(r, θ, ϕ, ξ, η, ζ) = √ Ak Bnmk Zn+1/2 (λnm r)Zn+1/2 (λnm ξ)
8 rξ n=0 m=1
k=0
× Pnk (cos θ)Pnk (cos η) cos[k(ϕ − ζ)],
where
( 2
1 for k = 0, (2n + 1)(n − k)! Jn+1/2 (λnm R2 )
Ak = Bnmk = 2 2
;
the Jn+1/2 (r) are Bessel functions, the Pnk (µ) are associated Legendre functions (see
above), and the λnm are positive roots of the transcendental equation Zn+1/2 (λR2 ) = 0.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ θ ≤ π/2, 0 ≤ ϕ ≤ 2π. First boundary value problem.

A hemisphere is considered. Boundary conditions are prescribed:
w = f1 (ϕ, θ) at r = R, w = f2 (r, ϕ) at θ = π/2.
Green’s function in the spherical coordinate system:
G(r, θ, ϕ, ξ, η, ζ) = Gs (r, θ, ϕ, ξ, η, ζ) − Gs (r, θ, ϕ, ξ, π − η, ζ),

1 1
Gs (r, θ, ϕ, ξ, η, ζ) = p − p ,
2
4π r − 2rξ cos γ + ξ 2 4π r ξ − 2R2 rξ cos γ + R4
2 2
where Gs (r, θ, ϕ, ξ, η, ζ) is the Green’s functions for a sphere.

Green’s function in the Cartesian coordinate system:

1 1 R 1 1 R
G(x, y, z, x0 , y0 , z0 ) = − − − ,
4π |r − r0 | |r0 | |r − r∗0 | 4π |r − r1 | |r0 | |r − r∗1 |
r = {x, y, z}, r0 = {x0 , y0 , z0 }, r1 = {x0 , y0 , −z0 }, r∗k = (R/r0 )2 rk , k = 0, 1.

◮ Domain: 0 ≤ r ≤ R, 0 ≤ θ ≤ π/2, 0 ≤ ϕ ≤ π. First boundary value problem.

A quarter of a sphere is considered. Boundary conditions are prescribed:
w = f1 (ϕ, θ) at r = R, w = f2 (r, ϕ) at θ = π/2,

w = f3 (r, θ) at ϕ = 0, w = f4 (r, θ) at ϕ = π.
Green’s function in the spherical coordinate system:
G(r, θ, ϕ, ξ, η, ζ) = Gs (r, θ, ϕ, ξ, η, ζ) − Gs (r, θ, ϕ, ξ, π − η, ζ)

+ Gs (r, θ, ϕ, ξ, π − η, 2π − ζ) − Gs (r, θ, ϕ, ξ, η, 2π − ζ),
where Gs (r, θ, ϕ, ξ, η, ζ) is the Green’s function for a sphere; see the previous paragraph.
Green’s function in the Cartesian coordinate system:
1
1 X n+k 1 R
G(x, y, z, x0 , y0 , z0 ) = (−1) − ,
4π |r − rnk | |r0 | |r − r∗nk |
n,k=0
r = {x, y, z}, r0 = {x0 , y0 , z0 }, rnk = {x0 , (−1)n y0 , (−1)k z0 }, r∗nk = (R/r0 )2 rnk ,
where r0 = |r0 |; n = 0, 1; k = 0, 1.
10.3 Helmholtz Equation ∆3 w + λw = −Φ(x)

A variety of problems related to steady-state oscillations (mechanical, acoustic, thermal,
electromagnetic, etc.) lead to the three-dimensional Helmholtz equation with λ > 0. This
equation governs mass transfer phenomena with volume chemical reaction of the first order
for λ < 0. Any elliptic equation with constant coefficients can be reduced to the Helmholtz
equation.
The Helmholtz equation is called homogeneous if Φ = 0 and nonhomogeneous if Φ 6= 0.
10.3.1 Homogeneous Helmholtz Equation. Eigenvalue problems

◮ Some definitions. Eigenvalue problems
A homogeneous boundary value problem is a boundary value problem for a homogeneous
equation with homogeneous boundary conditions; w = 0 is a particular solution of a ho-
mogeneous boundary value problem.
The values λn of the parameter λ for which there are nontrivial solutions (i.e., not iden-
tically zero solutions) of a homogeneous boundary value problem are called eigenvalues.
The corresponding solutions, w = wn , are called eigenfunctions of this boundary value
problem.
In what follows, we consider simultaneously the first, second, and third boundary value
problems for the three-dimensional homogeneous Helmholtz equation
∆3 w + λw = 0 (1)
in a finite three-dimensional domain V with a sufficiently smooth surface S. Let the homo-
geneous boundary condition have the form
∂w
s + kw = 0 for r ∈ S, (2)
∂N
∂w
where ∂N is the derivative along the outward normal to the surface S, sk ≥ 0, and
r = {x, y, z}. By appropriately choosing the constants s and k, one can obtain bound-
ary conditions of the first (s = 0, k = 1), second (s = 1, k = 0), and third (s = 1, k > 0)
kind.
◮ General properties of eigenvalues and eigenfunctions.

1◦ . There are infinitely many eigenvalues {λn }; they form the discrete spectrum of the
boundary value problem (1)–(2).
2◦ . All eigenvalues are positive, except for one eigenvalue λ0 = 0 of the second bound-
ary value problem (the corresponding eigenfunction is w0 = const). The eigenvalues are
assumed to be ordered so that λ≤ < λ2 ≤ λ3 ≤ · · · .
3◦ . The eigenvalues tend to infinity as the number n increases. The following asymptotic
estimate holds:
n V3
lim 3/2 = ,
n→∞ λ
n
6π 2
where V3 is the volume of the domain under consideration.
4◦ . The eigenfunctions are defined up to a constant factor, and they can be chosen to be
real. Any two eigenfunctions, wn and wm , that correspond to distinct eigenvalues λn 6= λm
are orthogonal; that is, Z
wn wm dV = 0.
V
Distinct eigenfunctions corresponding to coinciding eigenvalues λn = λm can be chosen
to be orthogonal.
5◦ . Any twice continuously differentiable function f = f (r) that satisfies the boundary
conditions of a boundary value problem can be expanded into a uniformly convergent series
in the eigenfunctions of this boundary value problem; specifically,
∞
X Z Z
1 2
f= an wn , where an = f wn dV, kwn k = wn2 dV.
n=1
kwn k2 V V
If f is square integrable, then the series is convergent in mean.

6◦ . The eigenvalues of the first boundary value problem decrease if the domain is extended.
The nonzero eigenvalues of the second boundary value problem decrease if the domain is
extended.
7◦ . Let λn and wn = wn (x) be the eigenvalues and eigenfunctions of problem (1)–(2) in
a domain V with the first or second boundary conditions (s = 0, k = 1, or k = 0, s = 1).
Then the eigenvalues and eigenfunctions λ′n and wn′ of the similar problem in the domain
V ′ obtained from V by dilation with ratio k (the shape of the domain remains unchanged)
are determined by the formulas
λn x
λ′n = 2 , wn′ = wn .
k k
10.3.2 Nonhomogeneous Helmholtz Equation. General Remarks,

Results, and Formulas
◮ Nonhomogeneous Helmholtz equation with homogeneous boundary conditions.
Three cases are possible.
1◦ . If λ is not equal to any one of the eigenvalues, then the solution of the problem is given
by
∞
X Z Z
An 1 2
w= wn , where An = Φwn dV, kwn k = wn2 dV.
λn − λ kwn k2 V V
n=1
2◦ . If λ coincides with one of the eigenvalues, λ = λm , then the condition of the orthogo-
nality of the function Φ to the eigenfunction wm ,
Z
Φwm dV = 0,
V
is a necessary condition for a solution of the nonhomogeneous problem to exist. The solu-
tion is then given by
m−1
X X∞ Z
An An 1
w= wn + wn + Cwm , An = Φwn dV,
n=1
λn − λm λ − λm
n=m+1 n
kwn k2 V
R
where C is an arbitrary constant and kwn k2 = V wn2 dV .
R
3◦ . If λ = λm and V Φwm dV 6= 0, then the boundary value problem for the nonhomoge-
neous equation has no solution.
Remark 10.4. If to each eigenvalue λn there are corresponding pn mutually orthogonal eigen-
(s)
functions wn (s = 1, . . . , pn ), then the solution is written as
X∞ X pn (s) Z Z
An 1 2
w= wn(s) , where A(s)
n = (s)
Φwn(s) dV, kwn(s) k2 = wn(s) dV,
n=1 s=1
λn − λ kwn k2 V V
provided that λ 6= λn .
◮ Solution of nonhomogeneous boundary value problems of general form.

1◦ . The solution of the first boundary value problem for the Helmholtz equation with the
boundary condition
w = f (r) for r ∈ S
Z Z
∂
w(r) = Φ(ρ)G(r, ρ) dVρ − f (ρ) G(r, ρ) dSρ . (1)
V S ∂N ρ
∂
Here, r = {x, y, z}, ρ = {ξ, η, ζ} (r ∈ V , ρ ∈ V ); ∂N ρ
denotes the derivative along the
outward normal to the surface S with respect to ξ, η, ζ. The Green’s function is given by
the series
∞
X wn (r)wn (ρ)
G(r, ρ) = , λ 6= λn , (2)
kwn k2 (λn − λ)
n=1
where the wn and λn are the eigenfunctions and eigenvalues of the homogeneous first
2◦ . The solution of the second boundary value problem with the boundary condition
∂w
= f (r) for r∈S
∂N
Z Z
w(r) = Φ(ρ)G(r, ρ) dVρ + f (ρ)G(r, ρ) dSρ . (3)
V S
Here, the Green’s function is given by the series

∞
X wn (r)wn (ρ)
1
G(r, ρ) = − + , (4)
V3 λ kwn k2 (λn − λ)
n=1
where V3 is the volume of the three-dimensional domain under consideration, and the λn
and wn are the positive eigenvalues and corresponding eigenfunctions of the homogeneous
second boundary value problem. For clarity, the term corresponding to the zero eigenvalue
λ0 = 0 (w0 = const) is singled out in (4). It is assumed that λ 6= 0 and λ 6= λn .
3◦ . The solution of the third boundary value problem for the Helmholtz equation with the
boundary condition
∂w
+ kw = f (r) for r ∈ S
∂N
is given by relation (3) in which the Green’s function is defined by series (2) with the
eigenfunctions wn and eigenvalues λn of the homogeneous third boundary value problem.
4◦ . Let nonhomogeneous boundary conditions of various types be set on different por-
m
P
tions Si of the surface S = Si ,
i=1
Γi [w] = fi (r) for r ∈ Si .

Then the solution of the corresponding mixed boundary value problem can be written as
Z Xm Z
w(r) = Φ(ρ)G(r, ρ) dVρ + fi (ρ)Λi (r, ρ) dSρ(i) ,
V i=1 Si
where

− ∂ G(r, ρ) if a first-kind boundary condition is set on Si ,
Λi (r, ρ) = ∂Nρ

G(r, ρ) if a second- or third-kind boundary condition is set on Si .
The Green’s function is expressed by series (2) that involves the eigenfunctions wn and
eigenvalues λn of the homogeneous mixed boundary value problem.
◮ Boundary conditions at infinity in the case of an unbounded domain.

Below it is assumed that the function Φ is finite or sufficiently rapidly decaying as r → ∞.
1◦ . If λ < 0 and the domain is unbounded, the additional condition that the solution must
vanish at infinity is set:
w → 0 as r → ∞. (5)
2◦ . If λ > 0, the radiation conditions (Sommerfeld conditions) are often used at infinity. In
three-dimensional problems, these conditions are expressed as

∂w √
lim rw = const, lim r + i λ w = 0, (6)
r→∞ r→∞ ∂r
where i2 = −1.
The principle of limit absorption and the principle of limit amplitude are also employed
to separate a single solution.
◮ Green’s function for an infinite cylindrical domain of arbitrary cross-section.
Consider the three-dimensional Helmholtz equation
∆3 w + λw = −Φ(r) (7)
inside an infinite cylindrical domain V = {(x, y) ∈ D, −∞ < z < ∞} with arbitrary

cross-section D. On the surface of this domain, let S = {(x, y) ∈ L, −∞ < z < ∞},
where L is the boundary of D, the homogeneous boundary condition of general form
∂w
s + kw = 0 for r∈S (8)
∂N
be set, with sk ≥ 0. By appropriately choosing the constants s and k in (8), one can obtain
boundary conditions of the first (s = 0, k = 1), second (s = 1, k = 0), and third (sk 6= 0)
kind.
The Green’s function of the first or third boundary value problem can be represented in
the form∗
∞
1 X un (x, y)un (ξ, η) −√µn −λ |z−ζ|
G(x, y, z, ξ, η, ζ) = √ e ,
2 n=1 kun k2 µn − λ
Z (9)
2
kun k = u2n (x, y) dx dy,
D
where the µn and un are the eigenvalues and eigenfunctions of the corresponding two-
dimensional boundary value problem in D,
∆2 u + µu = 0 for (x, y) ∈ D,
∂u (10)
s + ku = 0 for (x, y) ∈ L.
∂N
Recall that all µn are positive.

In the second boundary value problem, the zero eigenvalue µ0 = 0 appears, and hence
the summation in (9) must start with n = 0. In this case, u0 = 1 and ku0 k2 = D2 , where
D2 is the area of the cross-section D.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980), A. N. Tikhonov and A. A. Samarskii
(1990).
∗
Here and later in this section, the cross-section D is assumed to have finite dimensions.
◮ Green’s function for a semiinfinite cylindrical domain.
1◦ . The Green’s function of the three-dimensional first boundary value problem for equa-
tion (7) in a semiinfinite cylindrical domain V = {(x, y) ∈ D, 0 ≤ z < ∞} with arbitrary
cross-section D is given by
∞
X un (x, y)un (ξ, η)
G(x, y, z, ξ, η, ζ) = Hn (z, ζ), (11)
kun k2
n=1
where
1
Hn (z, ζ) = exp(−βn |z − ζ| ) − exp(−βn |z + ζ| )
2β
n

 1
 exp(−βn z) sinh(βn ζ) for z > ζ ≥ 0, p (12)
βn
= βn = µn − λ.
 1

 exp(−βn ζ) sinh(βn z) for ζ > z ≥ 0,
βn
Relations (11) and (12) involve the eigenfunctions un and eigenvalues µn of the two-dimen-
sional first boundary value problem (10) with s = 0 and k = 1.
2◦ . The Green’s function of the three-dimensional second boundary value problem for
equation (7) in a semiinfinite cylindrical domain V = {(x, y) ∈ D, 0 ≤ z < ∞} with
arbitrary cross-section D is given by
∞
X un (x, y)un (ξ, η)
1
G(x, y, z, ξ, η, ζ) = H0 (z, ζ) + Hn (z, ζ), (13)
D2 kun k2
n=1
where
1
Hn (z, ζ) = exp(−βn |z − ζ| ) + exp(−βn |z + ζ| )
2β
n
 1 exp(−β z) cosh(β ζ) for z > ζ ≥ 0,
 (14)
 n n p
βn
= βn = µn − λ.

 1
 exp(−βn ζ) cosh(βn z) for ζ > z ≥ 0,
βn
Relations (13) and (14) involve the eigenfunctions un and eigenvalues µn of the two-
dimensional second boundary value problem (10) with s = 1 and k = 0. Note that in (13)
the term corresponding to the zero eigenvalue µ0 = 0 is specially singled out; D2 is the
area of the cross-section D.
3◦ . The Green’s function of the three-dimensional third boundary value problem for equa-
tion (7) with the boundary conditions
∂w ∂w
− k1 w = 0 for z = 0, + k2 w = 0 for r∈S
∂z ∂N
in a semiinfinite cylindrical domain V = {(x, y) ∈ D, 0 ≤ z < ∞} with arbitrary cross-

section D and lateral surface S is given by relation (11) with

 exp(−βn z) βn cosh(βn ζ) + k1 sinh(βn ζ)

 if z > ζ ≥ 0,

βn (βn + k1 )
Hn (z, ζ) =

 exp(−βn ζ) βn cosh(βn z) + k1 sinh(βn z) (15)

 if ζ > z ≥ 0,
βn (βn + k1 )
p
βn = µn − λ.
sional third boundary value problem (10) with s = 1 and k = k2 .
4◦ . The Green’s function of the three-dimensional mixed boundary value problem for
equation (7) with a second-kind boundary condition at the end face and a first-kind bound-
ary condition at the lateral surface is given by relations (11) and (14), where the µn and un
are the eigenvalues and eigenfunctions of the two-dimensional first boundary value prob-
lem (10) with s = 0 and k = 1.
The Green’s functions of other mixed boundary value problems can be constructed
likewise.
◮ Green’s function for a cylindrical domain of finite dimensions.

1◦ . The Green’s function of the three-dimensional first boundary value problem for equa-
tion (7) in a cylindrical domain of finite dimensions V = {(x, y) ∈ D, 0 ≤ z ≤ a} with
arbitrary cross-section D is given by relation (11) with

 sinh(βn ζ) sinh[βn (a − z)]

 if a ≥ z > ζ ≥ 0,
βn sinh(βn a) p
Hn (z, ζ) = βn = µn − λ. (16)

 sinh(βn z) sinh[βn (a − ζ)]
 if a ≥ ζ > z ≥ 0,
βn sinh(βn a)
Another representation of the Green’s function:
∞ ∞
2 X X un (x, y)un (ξ, η) sin(qm z) sin(qm ζ) πm
G(x, y, z, ξ, η, ζ) = , qm = .
a kun k2 (µn + qm2 − λ) a
n=1 m=1
It is a consequence of formula (2).

2◦ . The Green’s function of the three-dimensional second boundary value problem for
equation (7) in a cylindrical domain of finite dimensions V = {(x, y) ∈ D, 0 ≤ z ≤ a}
with arbitrary cross-section D is given by relation (13) with

 cosh(βn ζ) cosh[βn (a − z)]

 if a ≥ z > ζ ≥ 0,
βn sinh(βn a) p
Hn (z, ζ) = βn = µn − λ. (17)

 cosh(βn z) cosh[βn (a − ζ)]
 if a ≥ ζ > z ≥ 0,
βn sinh(βn a)
Relations (13) and (17) involve the eigenfunctions un and eigenvalues µn of the two-
dimensional second boundary value problem (10) with s = 1 and k = 0.
Another representation of the Green’s function:
∞ ∞
1 X X εm un (x, y)un (ξ, η) cos(qm z) cos(qm ζ)
G(x, y, z, ξ, η, ζ) = ,
a kun k2 (µn + qm
2 − λ)
n=0 m=0
(
πm 1 for m = 0,
qm = , εm = µ0 = 0, u0 = 1.
a 2 for m 6= 0,
It is a consequence of formula (4).
3◦ . The Green’s function of the three-dimensional third boundary value problem for equa-
∂w ∂w ∂w
− k1 w = 0 at z = 0, + k2 w = 0 at z = a, + k3 w = 0 for r ∈ S
∂z ∂z ∂N
in a cylindrical domain of finite dimensions V = {(x, y) ∈ D, 0 ≤ z ≤ a} with arbitrary
cross-section D and lateral surface S is given by relation (11) with


 [βn cosh(βnζ)+k1 sinh(βnζ)] βn cosh[βn(a−z)]+k2 sinh[βn(a−z)]

 if z >ζ,
βn[βn(k1+k2) cosh(βna)+(βn2 +k1k2) sinh(βna)]
Hn(z, ζ)= (18)

 [βn cosh(βnz)+k1 sinh(βnz)] βn cosh[βn(a−ζ)]+k2 sinh[βn(a−ζ)]

 if z <ζ,
βn[βn(k1+k2) cosh(βna)+(βn2 +k1k2) sinh(βna)]
p
βn = µn−λ (0≤z ≤a, 0≤ζ ≤a).
sional third boundary value problem (10) with s = 1 and k = k3 .
4◦ . The Green’s function of the three-dimensional mixed boundary value problem for
equation (7) with second-kind boundary conditions at the end faces and a first-kind bound-
ary condition at the lateral surface is given by relations (11) and (17), where the µn and un
are the eigenvalues and eigenfunctions of the two-dimensional first boundary value prob-
lem (10) with s = 0 and k = 1.
The Green’s function of the three-dimensional mixed boundary value problem for equa-
w = 0 for z = 0, ∂z w = 0 for z = a, w = 0 for r∈S
in a cylindrical domain of finite dimensions V = {(x, y) ∈ D, 0 ≤ z ≤ a} with arbitrary
cross-section D and lateral surface S is given by relation (11) with

 sinh(βn ζ) cosh[βn (a − z)]

 if a ≥ z > ζ ≥ 0,
βn cosh(βn a) p
Hn (z, ζ) = βn = µn − λ. (19)

 sinh(βn z) cosh[βn (a − ζ)]
 if a ≥ ζ > z ≥ 0,
βn cosh(βn a)
The Green’s functions of other mixed boundary value problems can be constructed
likewise.

The three-dimensional nonhomogeneous Helmholtz equation in the rectangular Cartesian
∂2w ∂2w ∂2w

+ + + λw = −Φ(x, y, z).
∂x2 ∂y 2 ∂z 2

w=(A1 cos kx+A2 sin kx)(B1 cos my+B2 sin my)(C1z+C2), λ=k2+m2;
w=(A1 cos kx+A2 sin kx)(B1 cosh my+B2 sinh my)(C1z+C2), λ=k2−m2;
w=(A1 cos kx+A2 sin kx)(B1 cos my+B2 sin my) cos nz, λ=k2+m2+n2;
w=(A1 cos kx+A2 sin kx)(B1 cos my+B2 sin my) sin nz, λ=k2+m2+n2;
w=(A1 cosh kx+A2 sinh kx)(B1 cos my+B2 sin my) cos nz, λ=−k2+m2+n2;
w=(A1 cosh kx+A2 sinh kx)(B1 cos my+B2 sin my) sin nz, λ=−k2+m2+n2;
w=(A1 cosh kx+A2 sinh kx)(B1 cosh my+B2 sinh my) cos nz, λ=−k2−m2+n2;
w=(A1 cosh kx+A2 sinh kx)(B1 cosh my+B2 sinh my) sin nz, λ=−k2−m2+n2;
w=(A1 cosh kx+A2 sinh kx)(B1 cosh my+B2 sinh my) cosh nz, λ=−k2−m2−n2,
w=(A1 cosh kx+A2 sinh kx)(B1 cosh my+B2 sinh my) sinh nz, λ=−k2−m2−n2,
where A1 , A2 , B1 , B2 , C1 , and C2 are arbitrary constants.
Fundamental solutions:
1
E (x, y, z) = exp(−kr), λ = −k 2 < 0,
4πr
1
E (x, y, z) = exp(∓ikr), λ = k 2 > 0,
4πr
p
where r = x2 + y 2 + z 2 , k > 0, i2 = −1.
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, −∞ < z < ∞.

1◦ . Solution for λ = −k2 < 0:
Z ∞Z ∞Z ∞ p
1 exp −k (x−ξ)2+(y−η)2+(z−ζ)2
w(x, y, z)= Φ(ξ, η, ζ) p dξ dη dζ.
4π −∞ −∞ −∞ (x−ξ)2+(y−η)2+(z−ζ)2
2◦ . Solution for λ = k2 > 0:

Z ∞Z ∞Z ∞ p
1 exp −ik (x−ξ)2+(y−η)2+(z−ζ)2
w(x, y, z)= Φ(ξ, η, ζ) p dξ dη dζ.
4π −∞ −∞ −∞ (x−ξ)2+(y−η)2+(z−ζ)2
This solution was obtained taking into account the radiation condition at infinity (see Sec-
tion 10.3.2, condition (6)).
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, 0 ≤ z < ∞. First boundary value

problem.
w = f (x, y) at z = 0.
Solution:
Z ∞Z ∞
∂
w(x, y, z) = f (ξ, η) G(x, y, z, ξ, η, ζ) dξ dη
−∞ −∞ ∂ζ ζ=0
Z ∞Z ∞ Z ∞
0 −∞ −∞
Green’s function for λ = −k2 < 0:
exp(−kR1 ) exp(−kR2 )
G(x, y, z, ξ, η, ζ) = − ,
4πR1 4πR2
p p
R1 = (x − ξ)2 + (y − η)2 + (z − ζ)2 , R2 = (x − ξ)2 + (y − η)2 + (z + ζ)2 .
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, 0 ≤ z < ∞. Second boundary value

problem.
∂z w = f (x, y) at z = 0.
Solution:
Z ∞Z ∞
w(x, y, z) = − f (ξ, η)G(x, y, z, ξ, η, 0) dξ dη
Z ∞−∞
Z ∞−∞
Z ∞
0 −∞ −∞
exp(−kR1 ) exp(−kR2 )
G(x, y, z, ξ, η, ζ) = + ,
4πR1 4πR2
p p
R1 = (x − ξ)2 + (y − η)2 + (z − ζ)2 , R2 = (x − ξ)2 + (y − η)2 + (z + ζ)2 .

◮ Domain: −∞ < x < ∞, 0 ≤ y < ∞, 0 ≤ z < ∞. First boundary value problem.
w = f1 (x, z) at y = 0, w = f2 (x, y) at z = 0.
Solution:
Z ∞Z ∞
∂
w(x, y, z) = f1 (ξ, ζ) G(x, y, z, ξ, η, ζ) dξ dζ
0 −∞ ∂η η=0
Z ∞Z ∞
∂
+ f2 (ξ, η) G(x, y, z, ξ, η, ζ) dξ dη
0 −∞ ∂ζ ζ=0
Z ∞Z ∞Z ∞
0 0 −∞
exp(−kR1 ) exp(−kR2 ) exp(−kR3 ) exp(−kR4 )

G(x, y, z, ξ, η, ζ) = − − + ,
4πR1 4πR2 4πR3 4πR4
p p
R1 = (x − ξ)2 + (y − η)2 + (z − ζ)2 , R2 = (x − ξ)2 + (y − η)2 + (z + ζ)2 ,
p p
R3 = (x − ξ)2 + (y + η)2 + (z − ζ)2 , R4 = (x − ξ)2 + (y + η)2 + (z + ζ)2 .
◮ Domain: −∞ < x < ∞, 0 ≤ y < ∞, 0 ≤ z < ∞. Second boundary value

problem.
∂y w = f1 (x, z) at y = 0, ∂z w = f2 (x, y) at z = 0.
Solution:
Z ∞Z ∞
w(x, y, z) = − f1 (ξ, ζ)G(x, y, z, ξ, 0, ζ) dξ dζ
0 −∞
Z ∞Z ∞
− f2 (ξ, η)G(x, y, z, ξ, η, 0) dξ dη
Z0 ∞Z−∞Z
∞ ∞
0 0 −∞
exp(−kR1 ) exp(−kR2 ) exp(−kR3 ) exp(−kR4 )

G(x, y, z, ξ, η, ζ) = + + + ,
4πR1 4πR2 4πR3 4πR4
p p
R1 = (x − ξ)2 + (y − η)2 + (z − ζ)2 , R2 = (x − ξ)2 + (y − η)2 + (z + ζ)2 ,
p p
R3 = (x − ξ)2 + (y + η)2 + (z − ζ)2 , R4 = (x − ξ)2 + (y + η)2 + (z + ζ)2 .
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, 0 ≤ z ≤ a. First boundary value

problem.
w = f1 (x, y) at z = 0, w = f2 (x, y) at z = a.
Solution:
Z ∞Z ∞
∂
−∞ −∞ ∂ζ ζ=0
Z ∞Z ∞
∂
− f2 (ξ, η) G(x, y, z, ξ, η, ζ) dξ dη
−∞ −∞ ∂ζ ζ=a
Z aZ ∞ Z ∞
0 −∞ −∞
Green’s function for λ = −k 2 < 0:

X∞
exp(−kRn1 ) exp(−kR2n )
G(x, y, z, ξ, η, ζ) = − ,
n=−∞
4πRn1 4πR2n
p
R1n = (x − ξ)2 + (y − η)2 + (z − ζ − 2na)2 ,
p
R2n = (x − ξ)2 + (y − η)2 + (z + ζ − 2na)2 .
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, 0 ≤ z ≤ a. Second boundary value

problem.
∂z w = f1 (x, y) at z = 0, ∂z w = f2 (x, y) at z = a.
Solution:
Z ∞Z ∞
w(x, y, z) = − f1 (ξ, η)G(x, y, z, ξ, η, 0) dξ dη
−∞ −∞
Z ∞Z ∞
+ f2 (ξ, η)G(x, y, z, ξ, η, a) dξ dη
Z−∞Z
a ∞
−∞
Z ∞
0 −∞ −∞

X∞
exp(−kRn1 ) exp(−kR2n )
G(x, y, z, ξ, η, ζ) = + ,
n=−∞
4πRn1 4πR2n
p
R1n = (x − ξ)2 + (y − η)2 + (z − ζ − 2na)2 ,
p
R2n = (x − ξ)2 + (y − η)2 + (z + ζ − 2na)2 .
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, −∞ < z < ∞. First boundary value problem.

w = f3 (x, z) at y = 0, w = f4 (x, z) at y = b.
Solution:
Z bZ ∞
∂
0 −∞ ∂ξ ξ=0
Z bZ ∞
∂
− f2 (η, ζ) G(x, y, z, ξ, η, ζ) dζ dη
0 −∞ ∂ξ ξ=a
Z aZ ∞
∂
+ f3 (ξ, ζ) G(x, y, z, ξ, η, ζ) dζ dξ
0 −∞ ∂η η=0
Z aZ ∞
∂
− f4 (ξ, ζ) G(x, y, z, ξ, η, ζ) dζ dξ
0 −∞ ∂η η=b
Z aZ bZ ∞
0 0 −∞
Green’s function:
∞ ∞
2 XX 1
G(x, y, z, ξ, η, ζ) = sin(pn x) sin(qm y)
ab n=1 m=1 βnm

× sin(pn ξ) sin(qm η) exp(−βnm |z − ζ|) ,
nπ mπ p
pn = , qm = , βnm = p2n + qm
2 − λ.
a b
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, −∞ < z < ∞. Second boundary value problem.

Solution:
Z bZ ∞ Z bZ ∞
w(x, y, z)=− f1(η, ζ)G(x, y, z, 0, η, ζ) dζ dη+ f2(η, ζ)G(x, y, z, a, η, ζ) dζ dη
0 −∞ 0 −∞
Z aZ ∞ Z aZ ∞
− f3(ξ, ζ)G(x, y, z, ξ, 0, ζ) dζ dξ+ f4(ξ, ζ)G(x, y, z, ξ, b, ζ) dζ dξ
0 −∞ 0 −∞
Z aZ bZ ∞
0 0 −∞
Green’s function:
∞ ∞
1 X X An Am
G(x, y, z, ξ, η, ζ) = cos(pn x) cos(qm y) cos(pn ξ) cos(qm η) exp(−βnm |z−ζ|),
2ab n=0 m=0 βnm
(
pn = , qm = 2 2
, βnm = pn + qm − λ, An =
a b 2 for n 6= 0.
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, −∞ < z < ∞. Third boundary value problem.

∂y w − k3 w = f3 (x, z) at y = 0, ∂y w + k4 w = f4 (x, z) at y = b.
The solution w(x, y, z) is determined by the formula in the previous paragraph (for the
∞ ∞
1 X X unm (x, y)unm (ξ, η)
G(x, y, z, ξ, η, ζ) = exp(−βnm |z − ζ|).
2 n=1 m=1 kunm k2 βnm
Here,
p
wnm(x, y)=(µn cos µnx+k1 sin µnx)(νm cos νmy+k3 sin νmy), βnm = µ2n+νm 2 −λ,

2 1 2 2 2 2 (k1+k2)(µ2n+k1k2) (k3+k4)(νm2 +k k )
3 4
kwnmk = (µn+k1 )(νm+k3 ) a+ 2 2 2 2 b+ 2 ,
4 (µn+k1 )(µn+k2 ) (νm+k32)(νm2 +k 2)
4
(k1 + k2 )µ (k3 + k4 )ν
µ2 − k1 k2 ν 2 − k3 k4
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, −∞ < z < ∞. Mixed boundary value problems.

w = f3 (x, z) at y = 0, ∂y w = f4 (x, z) at y = b.
Solution:
Z bZ ∞
∂
0 −∞ ∂ξ ξ=0
Z bZ ∞
+ f2 (η, ζ)G(x, y, z, a, η, ζ) dζ dη
0 −∞
Z aZ ∞
∂
+ f3 (ξ, ζ) G(x, y, z, ξ, η, ζ) dζ dξ
0 −∞ ∂η η=0
Z aZ ∞
+ f4 (ξ, ζ)G(x, y, z, ξ, b, ζ) dζ dξ
0 −∞
Z aZ bZ ∞
0 0 −∞
Green’s function:
∞ ∞
2 XX 1
G(x, y, z, ξ, η, ζ) = sin(pn x) sin(qm y) sin(pn ξ) sin(qm η) exp(−βnm |z − ζ|),
ab n=0 m=0 βnm
(2n + 1)π (2m + 1)π p
pn = , qm = , βnm = p2n + qm
2 − λ.
2a 2b
Solution:
Z bZ ∞
∂
0 −∞ ∂ξ ξ=0
Z bZ ∞
∂
− f2 (η, ζ) G(x, y, z, ξ, η, ζ) dζ dη
0 −∞ ∂ξ ξ=a
Z aZ ∞
− f3 (ξ, ζ)G(x, y, z, ξ, 0, ζ) dζ dξ
0 −∞
Z aZ ∞
+ f4 (ξ, ζ)G(x, y, z, ξ, b, ζ) dζ dξ
0 −∞
Z aZ bZ ∞
0 0 −∞
Green’s function:
∞ ∞
1 X X Am
G(x, y, z, ξ, η, ζ) = sin(pn x) cos(qm y) sin(pn ξ) cos(qm η) exp(−βnm |z − ζ|),
ab n=1 m=0 βnm
(
pn = , qm = , βnm = p2n + qm 2 − λ, Am =
a b 2 for m 6= 0.
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ≤ z < ∞. First boundary value problem.

w = f5 (x, y) at z = 0.
Solution:
Z bZ ∞
∂
0 0 ∂ξ ξ=0
Z bZ ∞
∂
− f2 (η, ζ) G(x, y, z, ξ, η, ζ) dζ dη
0 0 ∂ξ ξ=a
Z aZ ∞
∂
+ f3 (ξ, ζ) G(x, y, z, ξ, η, ζ) dζ dξ
0 0 ∂η η=0
Z aZ ∞
∂
− f4 (ξ, ζ) G(x, y, z, ξ, η, ζ) dζ dξ
0 0 ∂η η=b
Z aZ b
∂
+ f5 (ξ, η) G(x, y, z, ξ, η, ζ) dη dξ
0 0 ∂ζ ζ=0
Z aZ b Z ∞
0 0 0
Green’s function:
∞ ∞
4 XX 1
ab βnm
n=1 m=1
nπ mπ p
pn = , qm = , βnm = p2n + qm 2 − λ,
a ( b
Hnm (z, ζ) =
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ≤ z < ∞. Second boundary value problem.

∂z w = f5 (x, y) at z = 0.
Solution:
Z aZ bZ ∞
w(x, y, z) = Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dζ dη dξ
0 0 0
Z bZ ∞ Z bZ ∞
− f1 (η, ζ)G(x, y, z, 0, η, ζ) dζ dη + f2 (η, ζ)G(x, y, z, a, η, ζ) dζ dη
0 0
Z aZ ∞ Z 0a Z0 ∞
− f3 (ξ, ζ)G(x, y, z, ξ, 0, ζ) dζ dξ + f4 (ξ, ζ)G(x, y, z, ξ, b, ζ) dζ dξ
0 0 0 0
Z aZ b
− f5 (ξ, η)G(x, y, z, ξ, η, 0) dη dξ.
0 0
Green’s function:
∞ ∞
1 X X An Am
G(x, y, z, ξ, η, ζ) = cos(pn x) cos(qm y) cos(pn ξ) cos(qm η)Hnm (z, ζ),
ab n=0 m=0 βnm
(
pn = , qm = , βnm = p2n + qm 2 − λ, An =
a b 2 for n 6= 0,
(
Hnm (z, ζ) =
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ≤ z < ∞. Third boundary value problem.

∂z w − k5 w = f5 (x, y) at z = 0.
The solution w(x, y, z) is determined by the formula in the previous paragraph (for the
∞ X
X ∞
unm (x, y)unm (ξ, η)
G(x, y, z, ξ, η, ζ) = Hnm (z, ζ).
kunm k2
n=1 m=1
Here,
wnm(x, y) = (µn cos µnx+k1 sin µnx)(νm cos νmy+k3 sin νmy),

2 1 2 2 2 2 (k1 +k2)(µ2n +k1k2) (k3 +k4)(νm2 +k k )
3 4
kwnmk = (µn +k1 )(νm +k3 ) a+ b+ 2 ,
4 (µ2n +k12)(µ2n +k22) (νm +k32)(νm2 +k 2)
4

 exp(−βnmz)[βnm cosh(βnmζ)+k5 sinh(βnmζ)]

 for z > ζ,
βnm(βnm +k5)
Hnm(z, ζ) =
 exp(−βnmζ)[βnm cosh(βnmz)+k5 sinh(βnmz)]

 for ζ > z,
βnm(βnm +k5)
p
βnm = µ2n +νm 2 −λ,
(k1 + k2 )µ (k3 + k4 )ν
µ2 − k1 k2 ν 2 − k3 k4
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ≤ z < ∞. Mixed boundary value problems.

∂z w = f5 (x, y) at z = 0.
Solution:
Z bZ ∞
∂
0 0 ∂ξ ξ=0
Z bZ ∞
∂
− f2 (η, ζ) G(x, y, z, ξ, η, ζ) dζ dη
0 0 ∂ξ ξ=a
Z aZ ∞
∂
+ f3 (ξ, ζ) G(x, y, z, ξ, η, ζ) dζ dξ
0 0 ∂η η=0
Z aZ ∞
∂
− f4 (ξ, ζ) G(x, y, z, ξ, η, ζ) dζ dξ
0 0 ∂η η=b
Z aZ b
− f5 (ξ, η)G(x, y, z, ξ, η, 0) dη dξ
0 0
Z aZ b Z ∞
0 0 0
Green’s function:
∞ ∞
4 XX 1
ab βnm
n=1 m=1
nπ mπ p
pn = , qm = , βnm = p2n + qm 2 − λ,
a ( b
Hnm (z, ζ) =

w = f5 (x, y) at z = 0.
Solution:
Z aZ bZ ∞
w(x, y, z) = Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dζ dη dξ
0 0 0
Z bZ ∞ Z bZ ∞
− f1 (η, ζ)G(x, y, z, 0, η, ζ) dζ dη + f2 (η, ζ)G(x, y, z, a, η, ζ) dζ dη
Z0 a Z0 ∞ Z 0a Z0 ∞
− f3 (ξ, ζ)G(x, y, z, ξ, 0, ζ) dζ dξ + f4 (ξ, ζ)G(x, y, z, ξ, b, ζ) dζ dξ
0 0 0 0
Z aZ b
∂
+ f5 (ξ, η) G(x, y, z, ξ, η, ξ) dη dξ.
0 0 ∂ζ ζ=0
Green’s function:
∞ ∞
1 X X An Am
G(x, y, z, ξ, η, ζ) = cos(pn x) cos(qm y) cos(pn ξ) cos(qm η)Hnm (z, ζ).
ab n=0 m=0 βnm
Here,
(
pn = , qm = , βnm = p2n + qm 2 − λ, An =
a b 2 for n =
6 0,
(
Hnm (z, ζ) =


n2 m2 k2
λnmk = π 2 + + ; n, m, k = 1, 2, 3, . . .
a2 b2 c2

πnx πmy πkz abc
wnmk = sin sin sin , kwnmk k2 = .
a b c 8
2◦ . A double-series representation of the Green’s function:

∞ ∞
4 XX
ab n=1 m=1

 sinh(βnm ζ) sinh[βnm (c − z)]

 for c ≥ z > ζ ≥ 0,
βnm sinh(βnm c)
Hnm (z, ζ) =

 sinh(βnm z) sinh[βnm (c − ζ)]
 for c ≥ ζ > z ≥ 0,
βnm sinh(βnm c)
πn πm p
pn = , qm = , βnm = p2n + qm 2 − λ.
a b
This relation can be used to obtain two other representations of the Green’s function with
the aid of the cyclic permutations of triples:
(x, ξ, a)
ր ց
(z, ζ, c) ←− (y, η, b)
∞ ∞ ∞
8 X X X sin(pnx) sin(qmy) sin(sk z) sin(pnξ) sin(qmη) sin(sk ζ)
G(x, y, z, ξ, η, ζ) = 2 +s2 −λ
,
abc n=1 m=1 p2n +qm k
k=1
πn πm πk
pn = , qm = , sk = .
a b c
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ≤ z ≤ c. Second boundary value problem.

∂z w = f5 (x, y) at z = 0, ∂z w = f6 (x, y) at z = c.
2
2 n m2 k2
λnmk = π + 2 + 2 ; n, m, k = 0, 1, 2, . . .
a2 b c
Eigenfunctions:

πnx πmy πkz
wnmk = cos cos cos .
a b c
The square of the norm of an eigenfunction is defined as
(
abc
2 1 for n = 0,
kwnmk k = (1 + δn0 )(1 + δm0 )(1 + δk0 ), δn0 =
8 0 for n =
6 0.
2◦ . A double series representation of the Green’s function:

∞ ∞
1 XX
G(x, y, z, ξ, η, ζ) = εn εm cos(pn x) cos(qm y) cos(pn ξ) cos(qm η)Hnm (z, ζ),
ab
n=0 m=0

 cosh(βnm ζ) cosh[βnm (c − z)]

 for c ≥ z > ζ ≥ 0,
βnm sinh(βnm c)
Hnm (z, ζ) =

 cosh(βnm z) cosh[βnm (c − ζ)]
 for c ≥ ζ > z ≥ 0,
βnm sinh(βnm c)
(
πn πm p 1 for n = 0,
pn = , qm = , βnm = p2n + qm 2 − λ, εn =
a b 2 for n 6= 0.
This relation can be used to obtain two other representations of the Green’s function with
the aid of the cyclic permutations:
(x, ξ, a)
ր ց
(z, ζ, c) ←− (y, η, b)

∞ ∞ ∞
1 XXX εn εm εk
G(x, y, z, ξ, η, ζ) = 2 2 + s2 − λ
cos(pn x) cos(qm y)
abc n=0 m=0 p + qm
k=0 n k

× cos(sk z) cos(pn ξ) cos(qm η) cos(sk ζ) ,
πn πm πk
pn = , qm = , sk = .
a b c
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ≤ z ≤ c. Third boundary value problem.

∂z w − k5 w = f5 (x, y) at z = 0, ∂z w + k6 w = f6 (x, y) at z = c.
Eigenvalues of the homogeneous problem:
λnml = µ2n + νm
2
+ σl2 ; n, m, l = 1, 2, 3, . . .
Here, the µn , νm , and σl are positive roots of the transcendental equations
(k1 + k2 )µ (k3 + k4 )ν (k5 + k6 )σ

tan(µa) = , tan(νb) = , tan(σc) = .
µ2 − k1 k2 ν 2 − k3 k4 σ 2 − k5 k6
Eigenfunctions:
1
wnml = (µn cos µn x + k1 sin µn x)(νm cos νm y + k3 sin νm y)(σl cos σl z + k5 sin σl z),
An Bm Cl
q q q
An = µ2n + k12 , Bm = νm 2 + k2 ,
3 C l = σl2 + k52 .

1 (k1 +k2)(µ2n +k1k2) (k3 +k4)(νm2
+k3k4) (k5 +k6)(σl2 +k5k6)
kwnmlk2 = a+ b+ c+ .
8 (µ2n +k12)(µ2n +k22) 2 +k 2)(ν 2 +k 2)
(νm 3 m 4 (σl2 +k52)(σl2 +k62)
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ≤ z ≤ c. Mixed boundary value problems.

1◦ . A rectangular parallelepiped is considered. Boundary conditions are prescribed:
2
2 n m2 k2
λnmk = π + 2 + 2 ; n, m = 1, 2, 3, . . . ; k = 0, 1, 2, . . .
a2 b c
Eigenfunctions:

πnx πmy πkz
wnmk = sin sin cos .
a b c
(
abc 1 for k = 0,
kwnmk k2 = (1 + δk0 ), δk0 =
8 0 for k 6= 0.
2◦ . A rectangular parallelepiped is considered. Boundary conditions are prescribed:
2
2 n m2 k2
λnmk = π + 2 + 2 ; n = 1, 2, 3, . . . ; m, k = 0, 1, 2, . . .
a2 b c
Eigenfunctions:

πnx πmy πkz
wnmk = sin cos cos .
a b c
(
abc 1 for m = 0,
kwnmk k2 = (1 + δm0 )(1 + δk0 ), δm0 =
8 0 for m =
6 0.
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ x, 0 ≤ z ≤ c. First boundary value problem.

A right prism whose base is an isosceles right-angled triangle is considered. Boundary
w = f1 (y, z) at x = 0, w = f2 (x, z) at y = 0, w = f3 (x, z) at y = x,

π2 π2 k2
λnmk = 2
(n + m)2 + m2 + 2 ; n, m, k = 1, 2, 3, . . .
a c
Eigenfunctions:

π πmy n πmx π πkz
wnmk = sin (n+m)x sin −(−1) sin sin (n+m)y sin .
a a a a c
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ x, 0 ≤ z ≤ c. Second boundary value problem.

A right prism whose base is an isosceles right-angled triangle is considered. Boundary
∂xw = f1(y, z) at x = 0, ∂y w = f2(x, z) at y = 0, ∂N w = f3(x, z) at y = x,

∂z w = f4(x, y) at z = 0, ∂z w = f5(x, y) at z = c,
where ∂N w = N · ∇w = √12 (∂x w + ∂y w).

π2 2 2
π2 k2
λnmk = (n + m) + m + 2 ; n, m, k = 0, 1, 2, . . .
a2 c
Eigenfunctions:

wnmk = cos (n+m)x cos −(−1) cos cos (n+m)y cos .
a a a a c
◮ Domain: 0 ≤ x ≤ a, 0 ≤ y ≤ x, 0 ≤ z ≤ c. Mixed boundary value problems.

1◦ . A right prism whose base is an isosceles right-angled triangle is considered. Boundary
w = f1(y, z) at x = 0, w = f2(x, z) at y = 0, w = f3(x, z) at y = x,

∂z w = f4(x, y) at z = 0, ∂z w = f5(x, y) at z = c.
π2 2 2
π2 k2
λnmk = (n + m) + m + 2 ; n, m = 1, 2, 3, . . . ; k = 0, 1, 2, . . .
a2 c
Eigenfunctions:

wnmk = sin (n+m)x sin −(−1) sin sin (n+m)y cos .
a a a a c
2◦ . A right prism whose base is an isosceles right-angled triangle is considered. Boundary

∂xw=f1(y, z) at x=0, ∂y w=f2(x, z) at y=0, ∂N w=f3(x, z) at y=x,

w=f4(x, y) at z=0, w=f5(x, y) at z=c,
where ∂N w = N · ∇w = √12 (∂x w + ∂y w).

π2 2 2
π2 k2
λnmk = (n + m) + m + 2 ; n, m = 0, 1, 2, . . . ; k = 1, 2, 3, . . .
a2 c
Eigenfunctions:

π πmy πmx π πkz
wnmk = cos (n+m)x cos −(−1)n cos cos (n+m)y sin .
a a a a c

The three-dimensional nonhomogeneous Helmholtz equation in the cylindrical coordinate
system is written as
p
1 ∂ ∂w 1 ∂2w ∂2w
r + 2 + + λw = −Φ(r, ϕ, z), r= x2 + y 2 .
r ∂r ∂r r ∂ϕ2 ∂z 2
√ √
w = AJ0 r λ + BY0 r λ (C1 ϕ + D1 )(C2 z + D2 ),
p
w = Jm r λ − µ2 (A cos mϕ + B sin mϕ)(C cos µz + D sin µz), λ > µ2 ,
p
w = Ym r λ − µ2 (A cos mϕ + B sin mϕ)(C cos µz + D sin µz), λ > µ2 ,
p
w = Jm r λ + µ2 (A cos mϕ + B sin mϕ)(C cosh µz + D sinh µz), λ > −µ2 ,
p
w = Ym r λ + µ2 (A cos mϕ + B sin mϕ)(C cosh µz + D sinh µz), λ > −µ2 ,
p
w = Im r µ2 − λ (A cos mϕ + B sin mϕ)(C cos µz + D sin µz), λ < µ2 ,
p
w = Ym r µ2 − λ (A cos mϕ + B sin mϕ)(C cos µz + D sin µz), λ < µ2 ,
where m = 0, 1, 2, . . . ; A, B, C, D, C1 , C2 , D1 , D2 , and µ are arbitrary constants;
the Jm (ξ) and Ym (ξ) are Bessel functions; and the Im (ξ) and Km (ξ) are modified Bessel
functions.

w = f (ϕ, z) at r = R.
Solution:
Z 2πZ ∞
∂
w(r, ϕ, z) = −R f (η, ζ) G(r, ϕ, z, ξ, η, ζ) dζ dη
0 −∞ ∂ξ ξ=R
Z R Z 2πZ ∞
0 0 −∞
Here,
∞ ∞
1 X X AnJn(µnmr)Jn(µnmξ)
G(r, ϕ, z, ξ, η, ζ)= 2 2 cos[n(ϕ−η)] exp −βnm|z−ζ| ,
2πR n=0 m=1 Jn′ (µnmR) βnm
(
p 1 for n=0,
βnm = µ2nm−λ, An =
2 for n6= 0,
equation Jn (µR) = 0.

problem.
∂r w = f (ϕ, z) at r = R.
Solution:
Z 2πZ ∞
0 −∞
Z R Z 2πZ ∞
0 0 −∞
Here,
√
exp − −λ |z−ζ|
G(r, ϕ, z, ξ, η, ζ)= √
2πR2 −λ
∞ ∞
+ 2 2 2 2
exp −βnm|z−ζ| ,
2π (µnmR −n )Jn(µnmR)βnm
n=0 m=1
(
p 1 for n=0,
2
βnm = µnm−λ, An =
2 for n6= 0,

problem.
∂r w + kw = f (ϕ, z) at r = R.
Solution:
Z 2πZ ∞
0 −∞
Z R Z 2πZ ∞
0 0 −∞
Here,
∞ ∞
G(r, ϕ, z, ξ, η, ζ)= 2 2 2 2 2 2
exp −βnm|z−ζ| ,
2π (µnmR +k R −n )Jn(µnmR)βnm
n=0 m=1
(
p 1 for n=0,
2
βnm = µnm−λ, An =
2 for n6= 0,
equation

w = f1 (ϕ, z) at r = R, w = f2 (r, ϕ) at z = 0.
Solution:
Z 2πZ ∞
∂
0 0 ∂ξ ξ=R
Z 2πZ R
∂
+ f2 (ξ, η) G(r, ϕ, z, ξ, η, ζ) ξ dξ dη
0 0 ∂ζ ζ=0
Z R Z 2πZ ∞
0 0 0
Here,
∞ ∞
2πR2 Jn′ (µnm R) βnm
n=0 m=1
Fnm (z, ζ) = exp(−βnm |z − ζ|) − exp(−βnm |z + ζ|),
(
p 1 for n = 0,
βnm = µ2nm − λ, An =
2 for n 6= 0,
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞. Second boundary value problem.
∂r w = f1 (ϕ, z) at r = R, ∂z w = f2 (r, ϕ) at z = 0.
Solution:
Z 2πZ ∞
0 0
Z 2πZ R
− f2 (ξ, η)G(r, ϕ, z, R, η, 0)ξ dξ dη
0 0
Z R Z 2πZ ∞
0 0 0
Here,
√ √
exp − −λ |z − ζ| + exp − −λ |z + ζ|
G(r, ϕ, z, ξ, η, ζ) = √
2πR2 −λ
∞ ∞
1 X X An µ2nm Jn (µnm r)Jn (µnm ξ) cos[n(ϕ − η)]
+ Fnm (z, ζ),
2π (µ2nm R2 − n2 )Jn2 (µnm R)βnm
n=0 m=1
Fnm (z, ζ) = exp(−βnm |z − ζ|) + exp(−βnm |z + ζ|),
(
p 1 for n = 0,
2 for n 6= 0,
∂r w + k1 w = f (ϕ, z) at r = R, ∂z w − k2 w = f2 (r, ϕ) at z = 0.
Solution:
Z 2πZ ∞
0 0
Z 2πZ R
− f2 (ξ, η)G(r, ϕ, z, ξ, η, 0)ξ dξ dη
0 0
Z R Z 2πZ ∞
0 0 0
Here,
∞ ∞
1 X X An µ2nm Jn (µnm r)Jn (µnm ξ) cos[n(ϕ − η)]
G(r, ϕ, z, ξ, η, ζ) = 2 R2 + k 2 R2 − n2 )J 2 (µ
Fnm (z, ζ),
π (µ nm 1 n nm R)
n=0 m=1


 exp(−βnm z)[βnm cosh(βnm ζ) + k2 sinh(βnm ζ)]
 for z > ζ,
β (βnm + k2 )
Fnm (z, ζ) = exp(−β ζ)[β nm cosh(βnm z) + k2 sinh(βnm z)]

 nm nm
 for ζ > z,
βnm (βnm + k2 )
(
p 1 for n = 0,
2 for n 6= 0,
equation
µJn′ (µR) + k1 Jn (µR) = 0.
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z < ∞. Mixed boundary value problem.

w = f1 (ϕ, z) at r = R, ∂z w = f2 (r, ϕ) at z = 0.
Solution:
Z 2πZ ∞
∂
0 0 ∂ξ ξ=R
Z 2πZ R
− f2 (ξ, η)G(r, ϕ, z, ξ, η, 0)ξ dξ dη
0 0
Z R Z 2πZ ∞
0 0 0
Here,
∞ ∞
2πR2 Jn′ (µnm R) βnm
n=0 m=1
Fnm (z, ζ) = exp(−βnm |z − ζ|) + exp(−βnm |z + ζ|),
(
p 1 for n = 0,
2 for n 6= 0,
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ a. First boundary value problem.

w = 0 at r = R, w = 0 at z = 0, w = 0 at z = a.
Eigenvalues:
π2 k2 µ2
λnmk = 2
+ nm ; n = 0, 1, . . . ; m, k = 1, 2, . . .
a R2
Here, the µnm are positive zeros of the Bessel functions, Jn (µ) = 0.
Eigenfunctions:

(1) r πkz
wnmk = Jn µnm cos(nϕ) sin ,
R a

(2) r πkz
wnmk = Jn µnm sin(nϕ) sin .
R a
Eigenfunctions possessing the axial symmetry property:

(1) r πkz
w0mk = J0 µ0m sin .
R a
(
(1) (2) πR2 a 2 1 for n = m,
kwnmk k2 = kwnmk k2 = (1 + δn0 ) Jn′ (µnm ) , δnm =
4 0 for n 6= m.

◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ a. Second boundary value problem.

∂r w = 0 at r = R, ∂z w = 0 at z = 0, ∂z w = 0 at z = a.
Eigenvalues:
π2 k2 µ2nm
λ000 = 0, λnmk = + ; n = 0, 1, . . . ; k, m = 0, 1, . . .
a2 R2
Here, the µnm are roots of the quadratic equation Jn′ (µ) = 0.
Eigenfunctions:

(1) r πkz (1)
wnmk = Jn µnm cos(nϕ) cos , w000 = 1,
R a

(2) r πkz
wnmk = Jn µnm sin(nϕ) cos .
R a
(1) (2) πR2 a 2 (1)

kwnmk k2 = kwnmk k2 = 2
(1 + δn0 )(µ2nm − n2 ) Jn (µnm ) , kw000 k2 = πR2 a,
4µnm
where δn0 is the Kronecker delta.

◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ a. Third boundary value problem.

∂r w+k1 w = 0 at r = R, ∂z w−k2 w = 0 at z = 0, ∂z w+k3 w = 0 at z = a.
Eigenvalues:
µ2nm
λnml = νl2 + ,
R2
where the νl and µnm are positive roots of the transcendental equations
(k2 + k3 )ν
tan(νa) = , µJn′ (µ) + Rk1 Jn (µ) = 0.
ν 2 − k2 k3
Eigenfunctions:

(1) r νl cos νl z + k2 sin νl z
wnml = Jn µnm cos(nϕ) q ,
R νl2 + k22

(2) r νl cos νl z + k2 sin νl z
wnml = Jn µnm sin(nϕ) q .
R ν 2 + k2 l 2

(i) πR2 2 (k2 + k3 )(νl2 + k2 k3 )
kwnml k2 2 2 2 2
= 2 (1 + δn0 )(R k1 + µnm − n ) Jn (µnm ) a + ,
4µnm (νl2 + k22 )(νl2 + k32 )
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ a. First boundary value problem.

A hollow circular cylinder of finite length is considered. Boundary conditions are pre-
scribed:
w = 0 at r = R1 , w = 0 at r = R2 ,
w = 0 at z = 0, w = 0 at z = a.
Eigenvalues:
π2 k2
λnmk = + µ2nm ; n = 0, 1, 2, . . . ; m, k = 1, 2, 3, . . .
a2
Here, the µnm are positive roots of the transcendental equation
Eigenfunctions:

(1) πkz
wnmk = [Jn (µnm r)Yn (µnm R1 ) − Jn (µnm R1 )Yn (µnm r)] cos(nϕ) sin ,
a

(2) πkz
wnmk = [Jn (µnm r)Yn (µnm R1 ) − Jn (µnm R1 )Yn (µnm r)] sin(nϕ) sin .
a
2 2 (
(1) (2) a [Jn (µ nm R1 ) −[Jn (µ nm R2) 1 if i=j,
kwnmkk2 =kwnmkk2 = 2 (1+δn0) 2 , δij =
πµnm [Jn(µnmR2) 0 if i6= j.

◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ a. Second boundary value problem.

A hollow circular cylinder of finite length is considered. Boundary conditions are pre-
scribed:
∂r w = 0 at r = R1 , ∂r w = 0 at r = R2 ,
∂z w = 0 at z = 0, ∂z w = 0 at z = a.
Eigenvalues:
π2 k2
λnmk = + µ2nm ; n, m, k = 0, 1, 2, . . .
a2
Eigenfunctions:

(1) πkz
wnmk = [Jn (µnm r)Yn′ (µnm R1 ) − Jn′ (µnm R1 )Yn (µnm r)] cos(nϕ) cos ,
a

(2) πkz
wnmk = [Jn (µnm r)Yn′ (µnm R1 ) − Jn′ (µnm R1 )Yn (µnm r)] sin(nϕ) cos .
a
(1)
To the zero eigenvalue λ000 = 0 there is a corresponding eigenfunction w000 = 1.
2
(1) (2) a(1+δn0)(1+δk0) n2 Jn′ (µnmR1) n2
kwnmk k2 = kwnmkk2 = 1− − 1− ,
πµ2nm R22µ2nm Jn′ (µnmR2) R12µ2nm
◮ Domain: R1 ≤ r ≤ R2 , 0 ≤ ϕ ≤ 2π, 0 ≤ z ≤ a. Mixed boundary value problems.

1◦ . A hollow circular cylinder of finite length is considered. Boundary conditions are
prescribed:
w = 0 at r = R1 , w = 0 at r = R2 ,
∂z w = 0 at z = 0, ∂z w = 0 at z = a.
Eigenvalues:
π2 k2
λnmk = + µ2nm ; n, k = 0, 1, 2, . . . ; m = 1, 2, 3, . . .
a2
Eigenfunctions:

(1) πkz
wnmk = [Jn (µnm r)Yn (µnm R1 ) − Jn (µnm R1 )Yn (µnm r)] cos(nϕ) cos ,
a

(2) πkz
wnmk = [Jn (µnm r)Yn (µnm R1 ) − Jn (µnm R1 )Yn (µnm r)] sin(nϕ) cos .
a
(
(1) (2) aεn εk [Jn (µnm R1 )]2 − [Jn (µnm R2 )]2 2 if n = 0,
kwnmk k2 = kwnmk k2 = , εn =
πµ2nm [Jn (µnm R2 )]2 1 if n 6= 0.
2◦ . A hollow circular cylinder of finite length is considered. Boundary conditions are
prescribed:
∂r w = 0 at r = R1 , ∂r w = 0 at r = R2 ,
w = 0 at z = 0, w=0 at z = a.
Eigenvalues:
π2 k2
λnmk = + µ2nm ; n = 0, 1, 2, . . . ; m, k = 1, 2, 3, . . .
a2
Eigenfunctions:

(1) πkz
wnmk = [Jn (µnm r)Yn′ (µnm R1 )
− Jn′ (µnm R1 )Yn (µnm r)] cos(nϕ) sin ,
a

(2) πkz
wnmk = [Jn (µnm r)Yn′ (µnm R1 ) − Jn′ (µnm R1 )Yn (µnm r)] sin(nϕ) sin .
a
′
(1) 2 (2) 2 aεn n2 Jn (µnm R1 ) 2 n2
kwnmk k = kwnmk k = 1− 2 2 − 1− 2 2 ,
πµ2nm R2 µnm Jn′ (µnm R2 ) R1 µnm
where εn is defined in Item 1◦ .
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ a. First boundary value problem.

w = 0 at ϕ = 0, w = 0 at ϕ = ϕ0 , w = 0 at r = R,
w = 0 at z = 0, w = 0 at z = a.
Eigenvalues:
π2 k2 µ2nm
λnmk = + ; n, m, k = 1, 2, 3, . . .
a2 R2
Here, the µnm are positive roots of the transcendental equation Jnπ/ϕ0 (µ) = 0.
Eigenfunctions:

µnm r nπϕ kπz
wnmk = Jnπ/ϕ0 sin sin .
R ϕ0 a
′ 2
kwnmk k2 = 18 aR2 ϕ0 Jnπ/ϕ 0
(µnm ) .
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ ϕ0 , 0 ≤ z ≤ a. Mixed boundary value problem.

w = 0 at ϕ = 0, w = 0 at ϕ = ϕ0 , w = 0 at r = R,
∂z w = 0 at z = 0, ∂z w = 0 at z = a.
Eigenvalues:
π2 k2 µ2nm
λnmk = + ; n, m = 1, 2, 3, . . . ; k = 0, 1, 2, . . .
a2 R2
Here, the µnm are positive roots of the transcendental equation Jnπ/ϕ0 (µ) = 0.
Eigenfunctions:

µnm r nπϕ kπz
wnmk = Jnπ/ϕ0 sin cos .
R ϕ0 a
(
′ 2 1 for k = 0,
kwnmk k2 = 18 aR2 ϕ0 (1 + δk0 ) Jnπ/ϕ 0
(µnm ) , δk0 =
0 for k 6= 0.

The three-dimensional homogeneous Helmholtz equation in the spherical coordinate sys-
tem is written as
p
1 ∂ 2 ∂w 1 ∂ ∂w 1 ∂ 2w
r + sin θ + +λw=0, r= x2+y 2+z 2.
1
w= (A sin µr + B cos µr), λ = µ2 ,
r
1
w= (A sinh µr + B cosh µr), λ = −µ2 ,
r
1
w= √ Jn+1/2 (µr)Pnm (cos θ)(A cos mϕ + B sin mϕ), λ = µ2 ,
r
1
w= √ Yn+1/2 (µr)Pnm (cos θ)(A cos mϕ + B sin mϕ), λ = µ2 ,
r
1
w= √ In+1/2 (µr)Pnm (cos θ)(A cos mϕ + B sin mϕ), λ = −µ2 ,
r
1
w= √ Kn+1/2 (µr)Pnm (cos θ)(A cos mϕ + B sin mϕ), λ = −µ2 ,
r
where n, m = 0, 1, 2, . . . ; A and B are arbitrary constants; Jν (ξ) and Yν (ξ) are Bessel
functions; Iν (ξ) and Kν (ξ) are modified Bessel functions; and the Pnm (ξ) are associated
Legendre functions, which are expressed in terms of the Legendre polynomials Pn (ξ) as
dm 1 dn 2
Pnm (ξ) = (1 − ξ 2 )m/2 Pn (ξ), Pn (ξ) = (ξ − 1)n .
dξ m n! 2n dξ n

1◦ . A spherical domain is considered. A homogeneous boundary condition is prescribed,
w = 0 at r = R.
Eigenvalues:
µ2nk
λnk = ; n = 0, 1, 2, . . . ; k = 1, 2, 3, . . .
R2
Here, the µnk are positive zeros of the Bessel functions, Jn+1/2 (µ) = 0. Note that the
Jn+1/2 (µ) can be expressed in terms of elementary functions, see Bateman and Erdélyi
(1953, Vol. 2).
Eigenfunctions:

(1) 1 r
wnmk = √ J µnk Pnm (cos θ) cos mϕ, m = 0, 1, 2, . . . ;
r n+1/2 R

(2) 1 r
wnmk = √ Jn+1/2 µnk Pnm (cos θ) sin mϕ, m = 1, 2, 3, . . .
r R
Here, the Pnm (ξ) are associated Legendre functions.

Eigenfunctions possessing central symmetry (i.e., independent of θ and ϕ):

(1) r
w00k = J1/2 µ0k .
R
Eigenfunctions possessing axial symmetry (i.e., independent of ϕ):

(1) r
wn0k = Jn+1/2 µnk Pn (cos θ).
R

(
(1) πR2 (1 + δm0 )(n + m)! ′ 2 1 for m = 0,
kwnmk k2 = Jn+1/2 (µnk ) , δm0 =
(2n + 1)(n − m)! 0 for m 6= 0,
(1) (2)
kwnmk k2 = kwnmk k2 , m = 1, 2, 3, . . .
2◦ . A spherical domain is considered. A nonhomogeneous boundary condition is pre-

scribed,
w = f (θ, ϕ) at r = R.
Solution:
∞ X
n √
X Ψn (r λ ) m 1
w(r, θ, ϕ) = fnm √ Yn (θ, ϕ), Ψn (x) = √ Jn+1/2 (x),
n=0 m=−n Ψn (R λ ) x
where
Z 2πZ π
1 2πεm (n + m)!
fnm = f (θ, ϕ)Ynm (θ, ϕ) sin θ dθ dϕ, kYnm k = ,
kYnm k 0 0 2n + 1 (n − m)!

(
Pn (cos θ)
 for m = 0,
2 for m = 0,
Ynm (θ, ϕ) = Pnm (cos θ) sin mϕ for m = 1, 2, . . . , εm =

 |m| 1 for m 6= 0.
Pn (cos θ) cos mϕ for m = −1, −2, . . . ,
√
The solution was written out under the assumption that Jn+1/2 (R λ ) 6= 0, where n =
0, 1, 2, . . .
⊙ Literature: M. M. Smirnov (1975), A. N. Tikhonov and A. A. Samarskii (1990).

∂r w = 0 at r = R.
Eigenvalues:
µ2nk
λ00 = 0, λnk = ; n = 0, 1, 2, . . . ; k = 1, 2, 3, . . .
R2
Here, the µnk are roots of the quadratic equation

′
2µJn+1/2 (µ) − Jn+1/2 (µ) = 0.
Eigenfunctions:

(1) (1) 1 r
w000 = 1, wnmk = √ Jn+1/2 µnk Pnm (cos θ) cos mϕ, m = 0, 1, 2, . . . ;
r R

(2) 1 r
wnmk = √ J µnk Pnm (cos θ) sin mϕ, m = 1, 2, 3, . . .
r n+1/2 R

(1) (1) πR2 εm (n + m)! n(n + 1) 2
kw000 k2 = 43 πR3 , kwnmk k2 = 1− Jn+1/2 (µnk ),
(2n + 1)(n − m)! µ2nk
(1) (2)
kwnmk k2 = kwnmk k2 , m = 1, 2, 3, . . . ,
(
2 for m = 0,
where εm =
1 for m =
6 0.

∂r w + sw = 0 at r = R.
Eigenvalues:
µ2nk
λnk = ; n = 0, 1, 2, . . . ; k = 1, 2, 3, . . .
R2
Here, the µnk are positive roots of the transcendental equation
′
2µJn+1/2 (µ) − (1 − 2Rs)Jn+1/2 (µ) = 0.
Eigenfunctions:

(1) 1 r
wnmk = √ Jn+1/2 µnk Pnm (cos θ) cos mϕ, m = 0, 1, 2, . . . ;
r R

(2) 1 r
wnmk = √ Jn+1/2 µnk Pnm (cos θ) sin mϕ, m = 1, 2, 3, . . .
r R
Here, the Pnm (ξ) are associated Legendre functions.

(1) 2 πR2 εm (n + m)! (Rs + n)(Rs − n − 1) 2
kwnmk k = 1+ Jn+1/2 (µnk ),
(2n + 1)(n − m)! µ2nk
(
2 for m = 0, (1) (2)
εm = kwnmk k2 = kwnmk k2 , m = 1, 2, 3, . . .
1 for m 6= 0,

A spherical cavity is considered and the dependent variable is prescribed at its surface:
w = f (θ, ϕ) at r = R,
and the radiation conditions are prescribed at infinity (see Section 10.3.2, condition (6)).
Solution for λ = k 2 > 0:
X∞ X n
Ξn (kr) m 1 (2)
w(r, θ, ϕ) = fnm Yn (θ, ϕ), Ξn (ρ) = √ Hn+1/2 (ρ),
n=0 m=−n
Ξ n (kR) ρ
Z 2πZ π
1 2πεm (n + m)!
fnm = m
f (θ, ϕ)Ynm (θ, ϕ) sin θ dθ dϕ, kYnm k = ,
kYn k 0 0 2n + 1 (n − m)!

(
Pn (cos θ)
 for m = 0,
2 for m = 0,
m
Yn (θ, ϕ) = Pnm (cos θ) sin mϕ for m = 1, 2, . . . , εm =

 |m| 1 for m 6= 0,
Pn (cos θ) cos mϕ for m = −1, −2, . . . ,
(2)
where Hn+1/2 (ρ) is the Hankel function of the second kind.

A spherical layer is considered. Boundary conditions are prescribed:
w = 0 at r = R1 , w = 0 at r = R2 .
Eigenvalues:
λnk = µ2nk ; n = 0, 1, 2, . . . ; k = 1, 2, 3, . . .
Here, the µnk are positive roots of the transcendental equation
Jn+1/2 (µR1 )Yn+1/2 (µR2 ) − Jn+1/2 (µR2 )Yn+1/2 (µR1 ) = 0.
Eigenfunctions:
(1) 1
wnmk = √ Zn+1/2 (µnk r)Pnm (cos θ) cos mϕ, m = 0, 1, 2, . . . ;
r
(2) 1
wnmk = √ Zn+1/2 (µnk r)Pnm (cos θ) sin mϕ, m = 1, 2, 3, . . .
r
Here, the Pnm (ξ) are associated Legendre functions and
Zn+1/2 (µr) = Jn+1/2 (µR1 )Yn+1/2 (µr) − Yn+1/2 (µR1 )Jn+1/2 (µr).
2 2
(
(1) 2 4εm(n+m)! Jn+1/2(µnkR1)−Jn+1/2(µnkR2) 2 if m=0,
kwnmkk = , εm =
π(2n+1)(n−m)! µ2nkJn+1/2
2 (µnkR2) 1 if m6= 0,
(1) (2)
kwnmkk2 =kwnmkk2, m=1, 2, 3, . . .
10.3.6 Other Orthogonal Curvilinear Coordinates

The homogenous three-dimensional Helmholtz equation admits separation of variables in
the eleven orthogonal systems of coordinates listed in Table 10.4.
For the parabolic cylindrical system of coordinates, the multipliers f and g are ex-
pressed in terms of the parabolic cylinder functions as
f (ξ)=A1Dµ−1/2(σξ)+A2Dµ−1/2(−σξ), g(η)=B1D−µ−1/2(ση)+B2D−µ−1/2(−ση),
1/4
µ= 21 β(k2−λ)−1/2, σ= 4(k2−λ) ,
where A1 , B1 , A2 , and B2 are arbitrary constants.

For the elliptic cylindrical system of coordinates, the functions f and g are determined
by the modified Mathieu equation and Mathieu equation, respectively, so that
( (
Cen (u, q), cen (v, q),
f (u) = g(v) = q = 14 a2 (λ − k2 ),
Sen (u, q), sen (v, q),
where Cen (u, q) and Sen (u, q) are the modified Mathieu functions, and cen (v, q) and
sen (v, q) are the Mathieu functions; to each value of the parameter q there are certain
corresponding eigenvalues β = βn (q) [see Abramowitz and Stegun (1964)].
In the prolate and oblate spheroidal systems of coordinates, the equations for f and g
are different forms of the spheroidal wave equation, whose bounded solutions are given by
f (u) = Ps|k| 2
n (cosh u, a λ), g(u) = Ps|k| 2
n (cos v, a λ) for prolate spheroid,
f (u) = Ps|k| 2 |k|
n (−i sinh u, a λ), g(u) = Psn (cos v, −a λ)
2
for oblate spheroid,
k is an integer, n = 0, 1, 2, . . . , −n ≤ k ≤ n,
where Pskn (z, a) are the spheroidal wave functions; see Bateman and Erdélyi (1955, Vol. 3),
Arscott (1964), and Meixner and Schäfke (1965). The separation of variables for the
Helmholtz equation in modified prolate and oblate spheroidal systems of coordinates, as
well as the spheroidal wave functions, are discussed in Abramowitz and Stegun (1964).
In the parabolic coordinate system, the solutions of the equations for f and g are ex-
pressed in terms of the degenerate hypergeometric functions [see Miller, Jr. (1977)] as
follows:

k 1 2
β k+1 2
√
f (ξ) = ξ exp ± 2 ωξ Φ − + , k + 1; ∓ωξ , ω = −λ,
4ω 2

k 1 2
β k+1 2
g(η) = η exp ± 2 ωη Φ + , k + 1; ∓ωη .
4ω 2
In the case of the paraboloidal coordinate system, the equations for f , g, and h are
reduced to the Whittaker–Hill equation

G′′θθ + µ + 18 b2 + bc cos 2θ − 18 b2 cos 4θ G = 0.
Denote by gcn (θ; b, c) and gsn (θ; b, c), respectively, the even and odd 2π-periodic solutions
of the Whittaker–Hill equation, which is a generalization of the Mathieu equation. The
TABLE 10.4
Orthogonal coordinates x̄, ȳ, z̄ that allow separable solutions of the form
w = f (x̄)g(ȳ)h(z̄) for the three-dimensional Helmholtz equation ∆3 w + λw = 0
Coordinates Transformations Particular solutions (or equations for f , g, h)
x = x, w = cos(k1 x + s1 ) cos(k2 y + s2 ) cos(k3 z + s3 ),

Cartesian
y = y, where k12 + k22 + k32 = λ;
x, y, z
x = r cos ϕ, w = [AJn(βr)+BYn(βr)] cos(nϕ+c) cos(kz+s),

Cylindrical
y = r sin ϕ, where Jn and Yn are the Bessel functions,
r, ϕ, z
z=z k 2 + β 2 = λ; see also Section 10.3.4
1 2 w = f (ξ)g(η) cos(kz + s),
Parabolic x= 2 (ξ − η 2 ),
cylindrical y = ξη, f ′′ + [(λ − k 2 )ξ 2 + β]f = 0,
ξ, η, z z=z g ′′ + [(λ − k 2 )η 2 − β]g = 0
Elliptic x = a cosh u cos v, w = f (u)g(v) cos(kz + s),

cylindrical y = a sinh u sin v, f ′′ + [ 12 a2 (λ − k 2 ) cosh 2u − β]f = 0,
u, v, z z=z g ′′ − [ 12 a2 (λ − k 2 ) cos 2v − β]g = 0
x = r sin θ cos ϕ, w = r−1/2 Jn+1/2 (βr)Pnm (cos θ) cos(mϕ + s),

Spherical
r, θ, ϕ y = r sin θ sin ϕ, w = r−1/2 Yn+1/2 (βr)Pnm (cos θ) cos(mϕ + s),
z = r cos θ where λ = β 2 ; see also Section 10.3.5
Prolate x = a sinh u sin v cos ϕ, w = f (u)g(v) cos(kϕ + s),

spheroidal y = a sinh u sin v sin ϕ, f ′′ + f ′ coth u + (−β + a2λ sinh2 u − k 2/sinh2 u)f = 0,
u, v, ϕ z = a cosh u cos v g ′′ + g ′ cot v + (β + a2λ sin2 v − k 2/sin2 v)g = 0
Oblate x = a cosh u sin v cos ϕ, w = f (u)g(v) cos(kϕ + s),

2 2
spheroidal y = a cosh u sin v sin ϕ, f ′′ + f ′ tanh u + (−β + a2λ cosh u + k 2/cosh u)f = 0,
2 2
u, v, ϕ z = a sinh u cos v g ′′ + g ′ cot v + (β − a2λ sin v − k 2/sin v)g = 0
x = ξη cos ϕ, w = f (ξ)g(η) cos(kϕ + s),

Parabolic
y = ξη sin ϕ, ξ 2 f ′′ + ξf ′ + (λξ 4 − βξ 2 − k 2 )f = 0,
ξ, η, ϕ
z = 12 (ξ 2 − η 2 ) η 2 g ′′ + ηg ′ + (λη 4 + βη 2 − k 2 )g = 0
x = 2a cosh u cos v sinh ϕ,

f ′′ + (−k − aβ cosh 2u + 12 a2 λ cosh 4u)f = 0,
Paraboloidal y = 2a sinh u sin v cosh ϕ,
g ′′ + (k + aβ cos 2v − 12 a2 λ cos 4v)g = 0,
u, v, ϕ z = 12 a(cosh 2u
h + (−k + aβ cosh 2ϕ − 21 a2 λ cosh 4ϕ)h = 0
′′
+ cos 2v − cosh 2ϕ)
q p p
x = (µ−a)(ν−a)(ρ−a)
a(a−1) , 4 pϕ(µ) [ pϕ(µ) f ′ ]′ + (λµ2 + β1 µ + β2 )f = 0,
General q
ellipsoidal y = (µ−1)(ν−1)(ρ−1) 4pϕ(ν) [pϕ(ν) g ′ ]′ + (λν 2 + β1 ν + β2 )g = 0,
,
µ, ν, ρ q 1−a 4 ϕ(ρ) [ ϕ(ρ) h′ ]′ + (λρ2 + β1 ρ + β2 )h = 0,
µνρ
z= a
ϕ(t) = t(t − 1)(t − a)
q √
x=ρ (aµ−1)(aν−1)
, w = ρ−1/2 J±(n+1/2) ρ λ g(ξ)h(η),
1−a
Conical q g ′′ + [β − n(n + 1)k 2 sn2 ξ]g = 0,
ρ, µ, ν y = ρ a(µ−1)(ν−1) , h′′ + [β − n(n + 1)k 2 sn2 η]h = 0, √
√ a−1
z = ρ aµν where µ = sn2 (ξ, k), ν = sn2 (η, k), k = a
subscript n = 0, 1, 2, . . . labels the discrete eigenvalues µ = µn . Each of the solutions

gcn and gsn can be represented in the form of an infinite convergent trigonometric series in
cos nθ and sin nθ, respectively; see Urvin and Arscott (1970). The functions f , g, and h can
be expressed in terms of the periodic solutions of the Whittaker–Hill equation as follows
[Miller, Jr. (1977)]:
( (
gcn iu; 2aω, 12 β/ω , gcn v; 2aω, 12 β/ω ,
f (u) = g(v) =
gsn iu; 2aω, 12 β/ω , gsn v; 2aω, 12 β/ω ,
(
gcn iϕ + π2 ; 2aω, 12 β/ω ,
h(ϕ) =
gsn iϕ + π2 ; 2aω, 12 β/ω ,
√
where ω = λ and k = µn − 12 a2 λ.
For the general ellipsoidal coordinates, the functions f , g, and h are expressed in terms
of the ellipsoidal wave functions; for details, see Arscott (1964) and Miller, Jr. (1977).
For the conical coordinate system, the functions g and h are determined by the Lamé
equations that involve the Jacobian elliptic function sn z = sn(z, k).
The unambiguity conditions for the transformation yield n = 0, 1, 2, . . . It is known
that, for any positive integer n, there exist exactly 2n + 1 solutions corresponding to 2n + 1
different eigenvalues β. These solutions can be represented the form of finite series known
as Lamé polynomials. For more details about the Lamé equation and its solutions, see
Whittaker and Watson (1963), Arscott (1964), Bateman and Erdélyi (1955), and Miller, Jr.
(1977).
Unlike the Laplace equation, there are no nontrivial transformations for the three-
dimensional Helmholtz equation that allow the R-separation of variables.
⊙ Literature for Section 10.3.5: F. M. Morse and H. Feshbach (1953, Vols. 1–2), P. Moon and D. Spencer
(1988), A. Makarov, J. Smorodinsky, K. Valiev, and P. Winternitz (1967), W. Miller, Jr. (1977).


∂2w ∂2w ∂2w a
1. + + + λ+ w = 0, r 2 = x2 + y 2 + z 2 .
∂x2 ∂y 2 ∂z 2 r
Schrödinger’s equation. It governs the motion of an electron in the Coulomb field of a

nucleus (a > 0).
The desired solutions must satisfy the normalizing condition
Z ∞Z ∞Z ∞
|w(x, y, z)|2 dx dy dz = 1.
−∞ −∞ −∞
Eigenvalues:
a2
λn = − ; n = 1, 2, 3, . . .
4n2
Normalized eigenfunctions (in the spherical coordinate system r, θ, ϕ):

3/2s k
2 (2k+1)(k−m)! (n−k−1)! ar ar ar (m)
wnmk = exp − L2k+1
n−k−1 Yk (θ, ϕ),
n 4πεmn(n+k)! (m+k)! n 2n n
n = 1, 2, 3, . . . ; m = 0, ±1, ±2, . . . , ±k; k = 0, 1, 2 . . . , n−1;
where

(
2 for m = 0, Pk (cos θ)
 for m = 0,
(m) m
εm = Yk (θ, ϕ) = Pk (cos θ) sin mϕ for m = 1, 2, . . . ,
1 for m 6= 0, 
 |m|
Pk (cos θ) cos mϕ for m = −1, −2, . . . ,
1 −s x dk 2 m/2 d
m
Lsk (x) = x e x k+s −x
e , Pk
m
(x) = (1 − x ) Pk (x),
k! dxk dxm
1 dk 2
Pk (x) = (x − 1)k .
k! 2k dxk
These relations involve the generalized Laguerre polynomials Lsk (x) and the associated
Legendre functions Pnm (ξ); the Pn (ξ) are the Legendre polynomials.
⊙ Literature: G. Korn and T. Korn (2000), A. N. Tikhonov and A. A. Samarskii (1990).
∂2w ∂2w ∂2w ∂w

2. + + = ay .
∂x2 ∂y 2 ∂z 2 ∂x
This equation is encountered in problems of convective heat and mass transfer in a simple
shear flow.
Z ∞
1 1
E (x, y, z, ξ, η, ζ) = 3/2
p
(4π) 0 3 1 2 2
t (1 + 12 a t )

[x − ξ − 12 at(y + η)]2 (y − η)2 + (z − ζ)2
× exp − 1 2 2 − dt.
4t(1 + 12 a t ) 4t
⊙ Literature: E. A. Novikov (1958), D. E. Elrick (1962).
∂2w ∂2w ∂2w ∂w ∂w ∂w

3. + + + a1 x + a2 y + a3 z = 0.
∂x2 ∂y 2 ∂z 2 ∂x ∂y ∂z
This equation is encountered in problems of convective heat and mass transfer in a straining
flow.
Z ∞
E (x, y, z, ξ, η, ζ) = F (x, ξ, t; a1 )F (y, η, t; a2 )F (z, ζ, t; a3 ) dt,
0
−1/2
2π 2at a(xeat − ξ)2
F (x, ξ, t; a) = e −1 exp − .
a 2(e2at − 1)
3
X
∂2w ∂2w ∂2w ∂w
4. + + = ank xn .
∂x21 ∂x22 ∂x23 ∂xk
n,k=1
This equation is encountered in problems of convective heat and mass transfer in an arbi-
trary linear shear flow.
The solution that corresponds to a source of unit power at the origin of coordinates is
given by
Z ∞ 3
X
1 bnk (t)xn xk dt
w(x1 , x2 , x3 ) = exp − p .
(4π)3/2 0 4D(t)
n,k=1
D(t)
Here, D = D(t) is the determinant of the matrix B = {Bnk }; the bnk = bnk (t) are the
cofactors of the entries Bnk = Bnk (t); the Bnk are determined by solving the following
system of ordinary differential equations with constant coefficients:
3
X 3
X
dBnk
= δnk + anm Bkm + akm Bnm ,
dt
m=1 m=1
Bnk → δnk t as t → 0 (initial conditions),
where δnn = 1 (n = 1, 2, 3) and δnk = 0 if n 6= k.

⊙ Literature: G. K. Batchelor (1979).
h i h i h i
∂ ∂w ∂ ∂w ∂ ∂w
5. f1 (x) + f2 (y) + f3 (z) = βw.
∂x ∂x ∂y ∂y ∂z ∂z
This is a three-dimensional linear equation of heat and mass transfer theory with a source
in an inhomogeneous anisotropic medium. Here, f1 = f1 (x), f2 = f2 (y), and f3 = f3 (z)
are the principal thermal diffusivities.
1◦ . The equation admits multiplicatively separable solutions, w = ϕ1 (x)ϕ2 (y)ϕ3 (z).
2◦ . There are also additively separable solutions, w = ψ1 (x) + ψ2 (y) + ψ3 (z).
3◦ . If f1 = axn , f2 = by m , and f3 = cz k (n 6= 2, m 6= 2, k 6= 2), there are particular

solutions of the form

2 x2−n y 2−m z 2−k
w = w(ξ), ξ =4 + + ,
a(2 − n)2 b(2 − m)2 c(2 − k)2
where the function w(ξ) is determined by the ordinary differential equation

d2 w A dw 1 1 1
2
+ = βw, A=2 + + − 1,
dξ ξ dξ 2−n 2−m 2−k
whose solutions are expressed in terms of the Bessel functions.


div [a(x, y, z)∇w] − q(x, y, z)w = −Φ(x, y, z)
Equations of this sort are often encountered in heat and mass transfer theory. For brevity,
the equation is written using the notation

∂ ∂w ∂ ∂w ∂ ∂w
div[a(r)∇w] = a(r) + a(r) + a(r) , r = {x, y, z}.
∂x ∂x ∂y ∂y ∂z ∂z
In what follows, the problems for the equation in question will be considered in a
bounded domain V with a sufficiently smooth surface S. It is assumed that a(r) > 0
and q(r) ≥ 0.

The following boundary condition of the first kind is imposed:
w = f (r) for r ∈ S.
Solution:
Z Z
∂
w(r) = Φ(ρ)G(r, ρ) dVρ − f (ρ)a(ρ) G(r, ρ) dSρ . (1)
V S ∂Nρ
Here, the Green’s function is given by
∞
X Z
un (r)un (ρ) 2
G(r, ρ) = , kun k = u2n (r) dV, ρ = {ξ, η, ζ}, (2)
kun k2 λn V
n=1
where the λn and un (r) are the eigenvalues and eigenfunctions of the Sturm–Liouville
problem for the following second-order elliptic equation with a homogeneous boundary
condition of the first kind:

div a(r)∇u − q(r)u + λu = 0, (3)
u = 0 for r ∈ S. (4)
∂
The integration in (1) is performed with respect to ξ, η, ζ; ∂N ρ
denotes the derivative along
the outward normal to the surface S with respect to ξ, η, ζ.
1◦ . There are countably many eigenvalues. All eigenvalues are real and can be ordered so
that λ1 ≤ λ2 ≤ λ3 ≤ · · · , with λn → ∞ as n → ∞; therefore the number of negative
eigenvalues is finite.
2◦ . If a(r) > 0 and q(r) ≥ 0, all eigenvalues are positive, λn > 0.
3◦ . The eigenfunctions are defined up to a constant factor, and they can be chosen to be
real. Two arbitrary eigenfunctions, un (r) and um (r), corresponding to distinct eigenvalues
λn 6= λm are orthogonal to each other in V ,
Z
un (r)um (r) dV = 0.
V
Distinct eigenfunctions corresponding to coinciding eigenvalues λn = λm can be chosen
to be orthogonal.
4◦ . An arbitrary function F (r) that is twice continuously differentiable and satisfies the
boundary condition of the Sturm–Liouville problem (F = 0 for r ∈ S) can be expanded
into an absolutely and uniformly convergent series in the eigenfunctions; specifically,
∞
X Z
1
F (r) = Fn un (r), Fn = F (r)un (r) dV,
kun k2 V
n=1
where the squared norm kun k2 is defined in (2).

5◦ . An increase in a and/or a decay in q results in an increase in the eigenvalues. The
eigenvalues decrease if the domain is extended.
6◦ . Constraint-strengthening modifications of the eigenvalue problem (including extra con-
ditions on w) result in nondecreasing of the eigenvalues.

A boundary condition of the second kind is imposed,
∂w
= f (r) for r ∈ S.
∂N
It is assumed that q(r) > 0.
Solution:
Z Z
w(r) = Φ(ρ)G(r, ρ) dVρ + f (ρ)a(ρ)G(r, ρ) dSρ . (5)
V S
Here, the Green’s function is defined by relation (2), where the λn and un (r) are the eigen-
values and eigenfunctions of the Sturm–Liouville problem for the second-order elliptic
equation (3) with the following homogeneous boundary condition of the second kind:
∂u
= 0 for r ∈ S. (6)
∂N
If q(r) > 0, the general properties of the eigenvalue problem (3), (6) are the same as
those of the first boundary value problem (see the paragraph above).

The following boundary condition of the third kind is set:
∂w
+ k(r)w = f (r) for r ∈ S.
∂N
The solution of the third boundary value problem is given by relations (5) and (2), where
the λn and un (r) are the eigenvalues and eigenfunctions of the Sturm–Liouville problem for
the second-order elliptic equation (3) with the following homogeneous boundary condition
of the third kind:
∂u
+ k(r)u = 0 for r ∈ S. (7)
∂N
If q(r) ≥ 0 and k(r) > 0, the general properties of the eigenvalue problem (3), (7) are
the same as those of the first boundary value problem (see the first paragraph).
Let k(r) = k = const. Denote the Green’s functions of the second and third boundary
value problems by G2 (r, ρ) and G3 (r, ρ, k), respectively. For q(r) > 0, the following limit
relation holds:
G2 (r, ρ) = lim G3 (r, ρ, k).
k→0
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1984), G. A. Korn and T. M. Korn (2000).

10.5.1 Laplace Equation ∆n w = 0
The n-dimensional Laplace equation in the rectangular Cartesian system of coordinates
x1 , . . . , xn has the form
∂2w ∂2w ∂2w

2 + 2 + ··· + = 0.
∂x1 ∂x2 ∂x2n
For n = 2 and n = 3, see Sections 9.1.1 and 10.1.1.

A regular solution of the Laplace equation is called a harmonic function.
p
In what follows we use the notation: x = {x1 , . . . , xn } and |x| = x21 + · · · + x2n .
◮ Particular solutions.
1 2π n/2
E (x) = − , σn = (n ≥ 3).
(n − 2)σn |x|n−2 Γ(n/2)
2◦ . Solution containing arbitrary functions of n − 1 variables:

∞
X 2k
k xn k x2k+1
n k
w(x1 , . . . , xn ) = (−1) ∆ f (x1 , . . . , xn−1 ) + ∆ g(x1 , . . . , xn−1 ) ,
(2k)! (2k + 1)!
k=0
where f (x1 , . . . , xn−1 ) and g(x1 , . . . , xn−1 ) are arbitrary infinitely differentiable func-
tions.
3◦ . Let w(x1 , . . . , xn ) be a harmonic function. Then the functions
w1 = Aw(±λx1 + C1 , . . . , ±λxn + Cn ),

A x1 xn
w2 = n−2 w ,..., 2 ,
|x| |x|2 |x|
are also harmonic functions everywhere they are defined; A, C1 , . . . , Cn , and λ are arbitrary
constants. The signs at λ in the expression of w1 can be taken independently of one another.
⊙ Literature: A. V. Bitsadze and D. F. Kalinichenko (1985), R. Courant and D. Hilbert (1989).
◮ Domain: −∞ < x1 < ∞, . . . , −∞ < xn−1 < ∞, 0 ≤ xn < ∞.

The first boundary value problem for an n-dimensional half-space is considered. A bound-
ary condition is prescribed:
w = f (x1 , . . . , xn−1 ) at xn = 0.
Solution:
Z ∞ Z ∞ n−1 −n/2
Γ(n/2) P
w(x1, . . . , xn)= ... (yk−xk)2+x2n xnf (y1, . . . , yn−1) dy1 . . . dyn−1,
π n/2 −∞ −∞ k=1
where Γ(z) is the gamma function.

◮ Domain: |x| ≤ 1. First boundary value problem.

A sphere of unit radius in the n-dimensional space is considered. A boundary condition is
prescribed:
w = f (x) for |x| = 1.
Solution (Poisson integral):
Z
Γ(n/2) 1 − |x|2
w(x) = f (y) dSy .
2π n/2 |y − x|n
|y|=1
1. ∆n w = −Φ(x1 , . . . , xn ).
This is the Poisson equation in n independent variables. For n = 2 and n = 3, see Sec-
tions 9.2 and 10.2.
1◦ . Solution:
Z
Γ(n/2) Φ(y1 , . . . , yn ) dy1 . . . dyn
w(x1 , . . . , xn ) = .
2(n − 2)π n/2 Rn
n−2
(x1 − y1 )2 + · · · + (xn − yn )2 2
2◦ . Domain: 0 ≤ xk ≤ ak ; k = 1, . . . , n. First boundary value problem.

w = fk (x1 , . . . , xk−1 , xk+1 , . . . , xn ) at xk = 0,

w = gk (x1 , . . . , xk−1 , xk+1 , . . . , xn ) at xk = ak .
Green’s function:
∞
X ∞
X
2n sin(pk1x1) sin(pk1y1) . . . sin(pknxn) sin(pknyn)
G(x1, . . . , xn, y1, . . . , yn)= ... ,
a1 . . . an p2k1+· · ·+p2kn
k1=1 kn=1
πk1 πk2 πkn
pk1 = , pk2 = , ..., pkn = .
a1 a2 an
3◦ . Domain: 0 ≤ xk ≤ ak ; k = 1, . . . , n. Mixed boundary value problem.
w = fk (x1 , . . . , xk−1 , xk+1 , . . . , xn ) at xk = 0,

∂xk w = gk (x1 , . . . , xk−1 , xk+1 , . . . , xn ) at xk = ak .
Green’s function:
∞
X ∞
X
2n sin(pk1x1) sin(pk1y1) . . . sin(pknxn) sin(pknyn)
G(x1, . . . , xn, y1, . . . , yn)= ... ,
a1 . . . an p2k1+· · ·+p2kn
k1=0 kn=0
π(2k1+1) π(2k2+1) π(2kn+1)
pk1 = , pk2 = , ..., pkn = .
2a1 2a2 2an
2. ∆n w + λw = 0.
This is the Helmholtz equation in n independent variables. For n = 2 and n = 3, see
Sections 9.3 and 10.3.
1◦ . Fundamental solution for λ = k2 > 0:
n−2
k 2 n−2
E (x, y) = n−2 r− 2 Y n−2 (kr), r = |x − y| for even n,
4(2π) 2 2
n−2
k 2 − n−2
E (x, y) = r
2 J (kr) for odd n,
n−2
1 − n−2
2
4(2π) 2 sin 2 πn
where Jν (z) and Yν (z) are Bessel functions.

2◦ . Domain: 0 ≤ xk ≤ ak ; k = 1, . . . , n. First boundary value problem.
w = fk (x1 , . . . , xk−1 , xk+1 , . . . , xn ) at xk = 0,

w = gk (x1 , . . . , xk−1 , xk+1 , . . . , xn ) at xk = ak .
Green’s function:
∞X ∞
X sin(pk x1) sin(pk y1) . . . sin(pk xn) sin(pk yn)
2n 1 1 n n
G(x1, . . . , xn, y1, . . . , yn)= ... ,
a1 a2 . . . an p2k1+· · ·+p2kn−λ
k1=1 kn=1
πk1 πk2 πkn
pk1 = , pk2 = , ..., pkn = .
a1 a2 an
3◦ . Domain: 0 ≤ xk ≤ ak ; k = 1, . . . , n. Second boundary value problem.
∂xk w = fk (x1 , . . . , xk−1 , xk+1 , . . . , xn ) at xk = 0,

Green’s function:
∞
X ∞
X cos(pk1x1) cos(pk1y1) . . . cos(pknxn) cos(pknyn)
G(x1, . . . , xn, y1, . . . , yn)= ... Ak1k2...kn ,
p2k1+· · ·+p2kn−λ
k1=0 kn=0
(
εk1εk2 . . . εkn πk1 πk2 πkn 1 for m=0,
Ak1k2...kn = , pk1 = , pk2 = , . . . , pkn = , εm =
a1a2 . . . an a1 a2 an 2 for m6= 0.
4◦ . Domain: 0 ≤ xk ≤ ak ; k = 1, . . . , n. Mixed boundary value problem.

w = fk (x1 , . . . , xk−1 , xk+1 , . . . , xn ) at xk = 0,

Green’s function:
∞
X ∞
X sin(pk x1) sin(pk y1) . . . sin(pk xn) sin(pk yn)
2n 1 1 n n
G(x1, . . . , xn, y1, . . . , yn)= ... ,
a1a2 . . . an p2k1+· · ·+p2kn−λ
k1=0 kn=0
π(2k1+1) π(2k2+1) π(2kn+1)
pk1 = , pk2 = , ..., pkn = .
2a1 2a2 2an
n
X ∂ 2w
3. aij = 0.
∂xi ∂xj
i,j=1
P
n n
P
It is assumed that for any real numbers y1 , . . . , yn the relation aij yi yj ≥ k yi2
i,j=1 i=1
holds, where k is some positive constant.

 X n − n−2

 Γ(n/2) 2

 √ bij (x i −y i )(x j −y j ) if n≥3,
 2(n−2)π n/2 A
i,j=1
E (x1, . . . , xn, y1, . . . , yn)= X 2 −1/2
 1


 √ ln b (x −y )(x −y ) if n=2,

 2π A ij i i j j
i,j=1
where A is the determinant of the matrix A = {aij } and the bij are the entries of the inverse
of A.
n−1
X
∂2w ∂ ∂w
4. + xβ
n + λw = 0.
∂x2i ∂xn ∂xn
i=1
Domain: ai ≤ xi ≤ bi (i = 1, . . . , n − 1), 0 ≤ xn ≤ c.
1◦ . Case 0 < β < 1. First boundary value problem. The condition w = 0 is set on the
entire boundary of the domain.
Eigenvalues and eigenfunctions:
n−1
X ki2 π 2 (2 − β)2 γνm
λk1 ,...,kn−1 ,m = + ,
(bi − ai )2 4c2−β
i=1
1−β x 2−β n−1
Y
2 n 2 ki π(xi − ai )
wk1 ,...,kn−1 ,m = xn Jν γνm sin ,
a bi − ai
i=1
where γνm is the mth positive root of the equation Jν (γ) = 0,
1−β
k1 , . . . , kn−1 = 1, 2, . . . ; m = 1, 2, . . . ; ν= .
2−β
2◦ . Case 1 ≤ β < 2. Boundary conditions: the solution must be bounded at xn = 0, and
the condition w = 0 must hold on the rest of the boundary of the domain.
The eigenvalues and eigenfunctions of this problem are given by the relations of Item 1◦
with ν = (β − 1)/(2 − β).
Chapter 11
Higher-Order Partial Differential

Equations
11.1 Third-Order Partial Differential Equations
11.1.1 One-Dimensional Equations Containing the First Derivative

in t
∂w ∂ 3w
1. + = Φ(x, t).
∂t ∂x3
Linearized Korteweg–de Vries equation.
1◦ . Particular solutions of the homogeneous equation with Φ(x, t) = 0:
w(x, t) = a(x3 − 6t) + bx2 + cx + k,

w(x, t) = a(x5 − 60x2 t) + b(x4 − 24xt),
w(x, t) = a sin(λx + λ3 t) + b cos(λx + λ3 t) + c,
w(x, t) = a sinh(λx − λ3 t) + b cosh(λx − λ3 t) + c,
√
w(x, t) = exp −λ3 t a exp λx + b exp − 12 λx sin 3
2 λx +c ,
where a, b, c, k, and λ are arbitrary constants.
2◦ . Solution given by a formal series in powers of t (Φ(x, t) = 0):
∞
X tk d3k f (x)
w(x, t) = f (x) + (−1)k ,
k! dx3k
k=1
where f (x) is an arbitrary infinitely differentiable function. This solution satisfies the initial
condition w(x, 0) = f (x). If a polynomial of degree n is taken for the function f (x), then
the solution is a polynomial in x of degree n as well.
941
942 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
3◦ . Solution given by a formal series in powers of x (Φ(x, t) = 0):

∞
X 3k dk f (t) ∞
X 3k+1 dk g(t)
k x k x
w(x, t) = (−1) + (−1)
(3k)! dtk (3k + 1)! dtk
k=0 k=0
∞
X x3k+2 k
d h(t)
+ (−1)k ,
(3k + 2)! dtk
k=0
where f (t), g(t), and h(t) are arbitrary infinitely differentiable functions and d0f (t)/dt0 =
f (t). This solution satisfies the boundary conditions w(0, t) = f (t), ∂x w(0, t) = g(t), and
∂xx w(0, t) = h(t). If polynomials of degree ≤ n are taken for the functions f (t), g(t), and
h(t), then the solution is a polynomial in t of degree ≤ n as well.
Z
1 ∞ 1 x
E (x, t) = cos(tξ 3 + xξ) dξ = Ai(z), z= ,
π 0 (3t)1/3 (3t)1/3
R∞
where Ai(z) = π1 0 cos 13 ξ 3 + zξ dξ is the Airy function.
w = f (x) at t = 0.
Solution:
Z ∞ Z tZ ∞
w(x, t) = E (x − y, t)f (y) dy + E (x − y, t − τ )Φ(y, τ ) dy dτ.
−∞ 0 −∞

Initial and boundary conditions are prescribed:
w = 0 at t = 0, w = f (t) at x = 0, w → 0 as x → ∞.
Solution: Z
t
x f (τ )
w(x, t) = 3 Ai′′ dτ,
0 (t − τ )1/3 t−τ
′′
where Ai (z) is the second derivative of the Airy function.
7◦ . Domain: −∞ < x ≤ 0. First boundary value problem.
Initial and boundary conditions are prescribed:
w=0 at t = 0, w = f (t) at x = 0, w → 0 as x → −∞.
Solution: Z
t
3 ′′ x f (τ )
w(x, t) = − Ai dτ.
2 0 (t − τ )1/3 t−τ
⊙ Literature: A. V. Faminskii (1999), N. A. Kudryashov and D. I. Sinelshchikov (2014).
11.1. Third-Order Partial Differential Equations 943
∂w ∂ 3w ∂2w
2. +a −b = Φ(x, t).
∂t ∂x3 ∂x2
Linearized Burgers–Korteweg–de Vries equation.
Z ∞ Z
1 2 3 ixu 1 ∞
E = exp[(−bu + iau )t]e du = exp(−bu2 ) cos(au3 t + xu) du.
2π −∞ π 0
w = f (x) at t = 0.
Solution:
Z ∞ Z tZ ∞
−∞ 0 −∞
∂w ∂3w
3. = ax6 .
∂t ∂x3
The transformation
u(z, τ ) = wx−2 , z = 1/x, τ = at
leads to a constant coefficient equation of the form 11.1.1.1:
∂u ∂3u
=− 3.
∂τ ∂z
∂w ∂ 3w ∂w
4. = k(t) + xf (t) + g(t) + h(t)w.
∂t ∂x3 ∂x
The transformation
Z Z Z
w(x, t) = u(z, τ ) exp h(t) dt , z = xF (t) + g(t)F (t) dt, τ= k(t)F 3 (t) dt,
Z
where F (t) = exp f (t) dt , leads to the linearized Korteweg–de Vries equation of the
form 11.1.1.1:
∂u ∂3u
= .
∂τ ∂z 3
∂w ∂3w
5. = (ax2 + bx + c)3 .
∂t ∂x3
This is a special case of equation 11.7.4.5 with k = 1 and n = 3. The transformation
Z
dx
w(x, t) = u(z, t)(ax2 + bx + c), z=
ax2 + bx + c
leads to the constant coefficient equation
∂u ∂3u ∂u
= + (4ac − b2 ) .
∂t ∂z 3 ∂z
11.1.2 One-Dimensional Equations Containing the Second Derivative

in t
∂2w ∂ 3w
1. =a .
∂t2 ∂x3
w(x, t) = A(x3 + 3at2 ) + Bx2 + C1 x + C2 t + C3 ,
w(x, t) = A(x5 + 30ax2 t2 ) + B(x4 + 12axt2 ),
p p
w(x, t) = eβx A exp − aβ 3 t + B exp aβ 3 t ,
p p
w(x, t) = e−βx A cos aβ 3 t + B sin aβ 3 t ,
where A, B, Cn , and β are arbitrary constants.
2◦ . Solution given by a formal series in powers of t:
∞
X ∞
ak t2k d3k f (x) X ak t2k+1 d3k g(x)
w(x, t) = + ,
(2k)! dx3k (2k + 1)! dx3k
k=0 k=0
where f (x) and g(x) are arbitrary infinitely differentiable functions and d0f (x)/dx0 =
f (x). This solution satisfies the initial conditions w(x, 0) = f (x) and ∂t w(x, 0) = g(x). If
polynomials of degree ≤ n are taken for the functions f (x) and g(x), then the solution is a
polynomial in x of degree ≤ n as well.
3◦ . Solution given by a formal series in powers of x:
∞
X ∞
x3k d2k f (t) X x3k+1 d2k g(t)
w(x, t) = +
ak (3k)! dt2k ak (3k + 1)! dt2k
k=0 k=0
∞
X x3k+2 2k
d h(t)
+ ,
ak (3k + 2)! dt2k
k=0
where f (t), g(t), and h(t) are arbitrary infinitely differentiable functions and d0f (t)/dt0 =
f (t). This solution satisfies the boundary conditions w(0, t) = f (t), ∂x w(0, t) = g(t), and
∂xx w(0, t) = h(t). If polynomials of degree ≤ n are taken for the functions f (t), g(t), and
h(t), then the solution is a polynomial in t of degree ≤ n as well.
∂2w ∂3w
2. = ax6 .
∂t2 ∂x3
The transformation z = 1/x, u = wx−2 leads to the constant coefficient equation of the
form 11.1.2.1:
∂2u ∂3u
= −a .
∂t2 ∂z 3
∂2w ∂3w
3. = (ax2 + bx + c)3 .
∂t2 ∂x3
This is a special case of equation 11.7.4.5 with k = 2 and n = 3. The transformation
Z
dx
w(x, t) = u(z, t)(ax2 + bx + c), z=
ax2 + bx + c
leads to the constant coefficient equation
∂2u ∂3u ∂u
= + (4ac − b2 ) .
∂t2 ∂z 3 ∂z
11.1.3 One-Dimensional Equations Containing a Mixed Derivative

and the First Derivative in t
∂w ∂ 2w ∂3w
1. −a −b = 0.
∂t ∂x2 ∂t∂x2
Equation of filtration of a compressible fluid in a cracked porous medium (e.g., see Baren-
blatt, Zheltov, and Kochina (1960), Barenblatt (1963)). This equation also describes one-
dimensional unsteady motions of second-grade non-Newtonian fluids (e.g., see Puri (1984),
Christov (2010)).
w(x, t) = Ax4 + (12aAt + B)x2 + 12a2 At2 + 2a(12A + B)t + C,

h i
aβ 2 t
w(x, t) = exp A exp(−βx) + B exp(βx) ,
1 − bβ 2
h i
aβ 2 t
w(x, t) = exp − A cos(βx) + B sin(βx) ,
1 + bβ 2
where A, B, C, and β are arbitrary constants. The last solution is periodic in x.

2◦ . Solutions periodic in t:

w(x, t) = e−λx A cos(ωt − µx) + B sin(ωt − µx) ,
√ 2 √ 2
ω a + b2 + bω 2 1/2 ω a + b2 − bω 2 1/2
λ=± , µ=± .
2(a2 + b2 ω 2 ) 2(a2 + b2 ω 2 )
Here one simultaneously takes only the upper or only the lower signs; A and B are arbitrary
constants.
For λ > 0, these formulas give the solution of the Stokes second problem for a second-
grade fluid with the boundary conditions
w = A cos(ωt) + B sin(ωt) at x = 0, w → 0 as x → ∞.
This is a problem without initial conditions, −∞ < t < ∞; see also Item 12◦ .
Z ∞
1 1 aξ 2 t
Ee (x, t) = exp − + ixξ dξ
2π −∞ 1 + bξ 2 1 + bξ 2
Z (1)
1 ∞ aξ 2 t cos(xξ)
= exp − 2
dξ.
π 0 1 + bξ 1 + bξ 2

w = f (x) at t = 0. (2)
One also assumes that w → 0 and ∂x w → 0 as |x| → ∞.

Solution: Z ∞
w(x, t) = Ee (x − y, t)[f (y) − bf ′′ (y)] dy,
−∞
where the function Ee (x, t) is defined in (1).

5◦ . Domain: 0 ≤ x < ∞. First boundary value problem with initial condition (2).
A boundary condition is prescribed:
w = 0 at x = 0.
Solution:
Z ∞
w(x, t) = Ee (x − y, t) − Ee (x + y, t) [f (y) − bf ′′ (y)] dy,
0
where the function Ee (x, t) is defined in (1).

6◦ . Domain: 0 ≤ x < ∞. Second boundary value problem with initial condition (2).
A boundary condition is prescribed:
∂x w = 0 at x = 0.
Solution:
Z ∞
w(x, t) = Ee (x − y, t) + Ee (x + y, t) [f (y) − bf ′′ (y)] dy.
0
7◦ . Domain: 0 ≤ x ≤ l. First boundary value problem with initial condition (2).

w = 0 at x = 0, w=0 at x = l.
Solution:
∞
X
aβn2 t
w(x, t) = An sin(βn x) exp − ,
1 + bβn2
n=1
Zl
2 πn
An = f (x) sin(βn x) dx, βn = .
l 0 l
8◦ . Domain: 0 ≤ x ≤ l. Second boundary value problem with initial condition (2).
∂x w = 0 at x = 0, ∂x w = 0 at x = l.
Solution:
∞
X
aβn2 t
w(x, t) = A0 + An cos(βn x) exp − ,
n=1
1 + bβn2
Z Z
1 l 2 l πn
A0 = f (x) dx, An = f (x) cos(βn x) dx, βn = .
l 0 l 0 l
9◦ . Domain: 0 ≤ x ≤ l. Mixed boundary value problem with initial condition (2).
w = 0 at x = 0, ∂x w = 0 at x = l.
Solution:
∞
X aβn2 t
1 + bβn2
n=1
Z
2 l π(2n + 1)
An = f (x) sin(βn x) dx, βn = .
l 0 2l
10◦ . Domain: 0 ≤ x ≤ l. Mixed boundary value problem with initial condition (2).
∂x w = 0 at x = 0, w=0 at x = l.
Solution:
∞
X aβn2 t
w(x, t) = An cos(βn x) exp − ,
1 + bβn2
n=1
Z l
2 π(2n + 1)
An = f (x) cos(βn x) dx, βn = .
l 0 2l
11◦ . Domain: 0 ≤ x < ∞. First boundary value problem. The following conditions are
prescribed:
w=0 at t = 0 (initial condition),
w = f (t) at x=0 (boundary condition),
w→0 as x → ∞ (boundary condition).
It is assumed that the initial and boundary conditions are consistent; i.e., f (0) = 0.
Solution:
Z
2 ∞
w(x, t) = U (ξ, t) sin(xξ) dξ,
π 0
Z t
ξ aξ 2 (t − τ )
U (ξ, t) = ϕ(τ ) exp − dτ, ϕ(τ ) = af (τ ) + bf ′ (τ ).
1 + bξ 2 0 1 + bξ 2
An alternative representation of the solution:
Z t
w(x, t) = ϕ(τ )G(x, t − τ ) dτ, ϕ(τ ) = af (τ ) + bf ′ (τ ),
0
Z
2 ∞ ξ sin(xξ) aξ 2 t
G(x, t) = exp − dξ.
π 0 1 + bξ 2 1 + bξ 2
12◦ . Domain: 0 ≤ x < ∞. Problems without initial conditions (−∞ < t < ∞).
In applications, problems are encountered in which the process is studied at a time
instant fairly remote from the initial instant; in this case, the initial conditions essentially
do not affect the distribution of the desired variable at the observation instant. In such
problems, no initial condition is stated, and the boundary conditions are assumed to be
prescribed for all preceding time instants, −∞ < t. However, in addition, the boundedness
condition in the entire domain is imposed on the solution.
As an example, consider the first boundary value problem for the half-space 0 ≤ x < ∞
w = f (t) at x = 0, w→0 as x → ∞.
The estimate |f (t)| < C exp(−λ|t|) with C > 0 and λ > a/b is assumed to hold as t → −∞.
Solution:
Z t
w(x, t) = ϕ(τ )G(x, t − τ ) dτ, ϕ(τ ) = af (τ ) + bf ′ (τ ),
−∞
Z
2 ∞ ξ sin(xξ) aξ 2 t
G(x, t) = exp − dξ.
π 0 1 + bξ 2 1 + bξ 2
13◦ . The Stokes first problem (a special case of Item 12◦ ). The problem of unidirectional
plane flow of a second-grade fluid in a half-plane caused by an impulsive motion of a plate
is characterized by the boundary conditions
w = U0 ϑ(t) at x = 0, w → 0 as x → ∞,
where ϑ(t) is the Heaviside unit step function and −∞ < t < ∞.
Three representations of the solution:
Z
2 ∞ sin(xξ) aξ 2 t
w(x, t) = U0 ϑ(t) 1 − exp − dξ
π 0 ξ(1 + bξ 2 ) 1 + bξ 2
Z r
1 1/b η dη
= U0 ϑ(t) 1 − exp(−atη) sin x
π 0 1 − bη η
Z ∞ r
x atζ
= U0 ϑ(t)e−at/b e−ζ erfc √ I0 2 dζ,
0 2 bζ b
where erfc z is the complementary error function and I0 (y) is the modified Bessel function
of the first kind of order zero.
14◦ . A modified Stokes second problem (a special case of Item 12◦ ). The transient version
of the Stokes second problem for a second-grade fluid is characterized by the boundary
conditions
w = U0 ϑ(t) cos(ωt) at x = 0, w→0 as x → ∞,
Solution:
Z ∞
2 ξ sin(xξ) a2 ξ 2 aξ 2 t
w(x, t) = U0 ϑ(t) − exp −
π 0 a2 ξ 4 + ω 2 (1 + bξ 2 )2 1 + bξ 2 1 + bξ 2

+ aω sin(ωt) + [a2 ξ 2 + bω 2 (1 + bξ 2 )] cos(ωt) dξ.
15◦ . Domain: 0 ≤ x < ∞. Transient filtration of a fluid in an adit (w is the pressure).

w = w0 at t = 0, w = w1 + (w0 − w1 )e−at/b at x = 0,
where w1 is the pressure in the stratum to the left of the boundary x = 0.
Solution:

at
w(x, t) = w1 + (w0 − w1 ) exp −
b
Z ∞
2 sin(xξ) aξ 2 t at
+ (w0 − w1 ) exp − − exp − dξ.
π 0 ξ 1 + bξ 2 b
⊙ Literature for Items 13◦ –15◦ : G. I. Barenblatt, Yu. P. Zheltov, and I. N. Kochina (1960), P. Puri (1984),
R. Bandelli, K. R. Rajagopal, and G. P. Galdi (1995), C. Fetecâu and J. Zierep (2001), I. C. Christov (2010),
I. C. Christov and C. I. Christov (2010), I. C. Christov and P. M. Jordan (2012).
∂w ∂ 2w ∂3w
2. −a −b = Φ(x, t).
∂t ∂x2 ∂t∂x2
w = f (x) at t = 0.
Solution:
Z tZ ∞ Z ∞
w(x, t) = Ee (x − y, t − τ )Φ(y, τ ) dy dτ + Ee (x − y, t)[f (y) − bf ′′ (y)] dy,
0 −∞ −∞
where
Z ∞
1 1 aξ 2 t
Ee (x, t) = exp − + ixξ dξ
2π −∞ 1 + bξ 2 1 + bξ 2
Z
1 ∞ aξ 2 t cos(xξ)
= exp − dξ.
π 0 1 + bξ 2 1 + bξ 2
2◦ . In Items 3◦ –6◦ , we consider problems on an interval 0 ≤ x ≤ l with the general initial
condition
w = f (x) at t = 0
and various homogeneous boundary conditions. The solution can be represented in terms
of the Green’s function as
Z tZ l Z l
w(x, t) = G(x, ξ, t − τ )Φ(ξ, τ ) dξ dτ + G(x, ξ, t)[f (ξ) − bf ′′ (ξ)] dξ.
0 0 0
The function f (x) is assumed to be consistent with the homogeneous boundary conditions
for w at x = 0 and x = l.

w = 0 at x = 0, w=0 at x = l.
Green’s function:
∞
2 X sin(βn x) sin(βn ξ) aβn2 t πn
G(x, ξ, t) = exp − , βn = .
l 1 + bβn2 1 + bβn2 l
n=1

∂x w = 0 at x = 0, ∂x w = 0 at x = l.
Green’s function:
∞
1 2 X cos(βn x) cos(βn ξ) aβn2 t πn
G(x, ξ, t) = + exp − , βn = .
l l 1 + bβn2 1 + bβn2 l
n=1

w = 0 at x = 0, ∂x w = 0 at x = l.
Green’s function:
∞
2 X sin(βn x) sin(βn ξ) aβn2 t π(2n + 1)
G(x, ξ, t) = exp − , βn = .
l 1 + bβn2 1 + bβn2 2l
n=0

∂x w = 0 at x = 0, w=0 at x = l.
Green’s function:
∞
2 X cos(βn x) cos(βn ξ) aβn2 t π(2n + 1)
G(x, ξ, t) = exp − , βn = .
l n=0 1 + bβn2 1 + bβn2 2l
∂w ∂ 2w ∂3w
3. −a −b + cw = Φ(x, t).
∂t ∂x2 ∂t∂x2
The substitution w = e−ct u leads to an equation of the form 11.1.3.2:
∂u ∂2u ∂3u
− (a − bc) 2 − b = ect Φ(x, t).
∂t ∂x ∂t∂x2
11.1.4 One-Dimensional Equations Containing a Mixed Derivative

and the Second Derivative in t
∂2w ∂2w ∂3w
1. −a −b = 0.
∂t2 ∂x2 ∂t∂x2
This equation describes one-dimensional unsteady motions of viscous compressible baro-
tropic fluids.
w(x, t) = (A1 t + A2 )x2 + (B1 t + B2 )x + 13 aA1 t3 + (aA2 + bA1 )t2 + C1 t + C2 ;

βt βx βx
w(x, t) = e A1 exp − √ + A2 exp √ , a + bβ > 0;
a + bβ a + bβ

βt βx βx
w(x, t) = e A1 cos p + A2 sin p , a + bβ < 0,
|a + bβ| |a + bβ|
where An , Bn , Cn , and β are arbitrary constants. The last solution is periodic in x.


√ 2 √ 2
a + b2 ω 2 − a 1/2 a + b2 ω 2 + a 1/2
λ = ±ω , µ = ±ω .
2(a2 + b2 ω 2 ) 2(a2 + b2 ω 2 )
constants.

Solution:
∞
X πn
w(x, t) = An ψn1 (t) + Bn ψn2 (t) sin(βn x), βn = ,
l
n=1
where
λn1 exp(−λn2 t) − λn2 exp(−λn1 t) exp(−λn2 t) − exp(−λn1 t)
ψn1 (t) = , ψn2 (t) = ,
λn1 − λn2 λn1 − λn2
p p
bβ 2 + βn b2 βn2 − 4a bβ 2 − βn b2 βn2 − 4a
λn1 = n , λn2 = n ,
2 2
Z l Z l
2 2
An = f (x) sin(βn x) dx, Bn = g(x) sin(βn x) dx.
l 0 l 0
2 2
− 4a > 0 as well as for b2 βn2 − 4a < 0. In the
p holds for b βn p
Remark 11.1. This solution
latter case, one should write b2 βn2 − 4a = i 4a − b2 βn2 , where i2 = −1, and transform the
exponentials occurring in the functions ψn1 (t) and ψn2 (t) by using the Moivre formulas.
Solution:
∞
X πn
w(x, t) = A0 + B0 t + An ψn1 (t) + Bn ψn2 (t) cos(βn x), βn = ,
l
n=1
where
λn1 exp(−λn2 t) − λn2 exp(−λn1 t) exp(−λn2 t) − exp(−λn1 t)
ψn1 (t) = , ψn2 (t) = ,
λn1 − λn2 λn1 − λn2
p p
bβn2 + βn b2 βn2 − 4a bβn2 − βn b2 βn2 − 4a
λn1 = , λn2 = ,
2 2
Z l Z l
1 1
A0 = f (x) dx, B0 = g(x) dx,
l 0 l 0
Z Z
2 l 2 l
An = f (x) cos(βn x) dx, Bn = g(x) cos(βn x) dx, n = 1, 2, . . . .
l 0 l 0
∂ 2w ∂w ∂2w ∂3w
2. c + =a +b .
∂t2 ∂t ∂x2 ∂t∂x2
This equation describes one-dimensional unsteady motions of a viscoelastic incompressible
Oldroyd-B fluid (a > 0, b > 0, and c > 0).
1◦ . Solutions periodic in x:

w(x, t) = e−γt A cos(βx) + B sin(βx) ,
1h 2 p i
γ1,2 = bβ + 1 ± (bβ 2 + 1)2 − 4acβ 2 ,
2c
where A, B, and β are arbitrary constants.

p
ω (bcω 2 + a)2 + (ac − b)2 ω 2 − (ac − b)ω 2 1/2
λ=± ,
2(a2 + b2 ω 2 )
p
ω (bcω 2 + a)2 + (ac − b)2 ω 2 + (ac − b)ω 2 1/2
µ=± .
2(a2 + b2 ω 2 )
constants.
3◦ . The Stokes first problem. The problem of unidirectional plane flow of an Oldroyd-B
fluid in a half-plane due to the impulsive motion of the plate is characterized by the bound-
ary conditions
w = U0 ϑ(t) at x = 0, w → 0 as x → ∞,
Solution:
Z
1 1 ∞
w(x, t) = U0 ϑ(t) + exp −xg(η)[cos h(η) − sin h(η)]
2 π 0

tη dη
× sin − xg(η)[cos h(η) + sin h(η)] ,
c η
where
1/2 1/4
η 1 + η2 1 −1 b
g(η) = , h(η) = tan η − tan−1 (kη) , k= .
2ac 1 + k2 η2 2 ac
4◦ . For the special case of b = ac, the equation admits the factorization
(c∂t + 1)(∂t − a∂xx )[w] = 0.
Hence its general solution has the form
w = f (x) exp(−t/c) + u(x, t),
where f (x) is an arbitrary function and the function u = u(x, t) is an arbitrary solution of
the heat equation ∂t u − a∂xx u = 0.
The solution of the Stokes first problem (see Item 3◦ ) with b = ac can be expressed via
the complementary error function

x
w(x, t) = U0 ϑ(t) erfc √ .
2 at
⊙ Literature for Item 3◦ : R. I. Tanner (1962), I. C. Christov (2010).
∂2w ∂w ∂ 2w ∂ 3w
3. ε +σ =a +b + kw.
∂t2 ∂t ∂x2 ∂t∂x2
1◦ . Solutions periodic in x:

w(x, t) = e−γt A cos(βx) + B sin(βx) ,
1 h 2 p i
γ1,2 = bβ + σ ± (bβ 2 + σ)2 + 4ε(k − aβ 2 ) , ε 6= 0,
2ε

p
ω 2 (bεω 2 + aσ + bk)2 + [(aε − bσ)ω 2 + ak]2 − (aε − bσ)ω 2 − ak 1/2
λ=± ,
2(a2 + b2 ω 2 )
p 2
ω (bεω 2 + aσ + bk)2 + [(aε − bσ)ω 2 + ak]2 + (aε − bσ)ω 2 + ak 1/2
µ=± .
2(a2 + b2 ω 2 )
constants.
3◦ . The substitution w = e−λt u, where λ is a root of the quadratic equation ελ2 −σλ−k = 0,
∂2u ∂u ∂2u ∂3u

ε + (σ − 2λ) = (a − bλ) + b .
∂t2 ∂t ∂x2 ∂t∂x2
11.1.5 Two- and Three-Dimensional Equations
∂w ∂
1. − a∆w − b ∆w = 0.
∂t ∂t
Equation of filtration of a fluid in a cracked porous medium (e.g., see Barenblatt, Zheltov,
and Kochina (1960), Barenblatt (1963)).
1◦ . Fundamental solution in two-dimensional case:
Z
ϑ(t) ∞ ρJ0 (rρ) aρ2 t
Ee (x, t) = exp − dρ,
2π 0 1 + bρ2 1 + bρ2
p
where r = |x| = x2 + y 2 and J0 (z) is the Bessel function.
2◦ . Fundamental solution in three-dimensional case:
Z
ϑ(t) ∞ ρ sin(rρ) aρ2 t
Ee (x, t) = exp − dρ,
2π 2 r 0 1 + bρ2 1 + bρ2
p
where r = |x| = x2 + y 2 + z 2 .
3◦ . Domain: x ∈ Rn . Cauchy problem.
w = f (x) at t = 0.
Solution:
Z

w= Ee (x − y, t) f (y) − b∆y f (y) dVy , dVy = dy1 . . . dyn ,
Rn
where ∆y is the Laplace operator in the integration variables y1 , . . . , yn . See Items 1◦

and 2◦ for the function Ee (x, t) with n = 2 and n = 3.
4◦ . Consider an open bounded domain V with boundary S. The following conditions are
prescribed:
∂w
αw + β = 0 at x ∈ S (boundary condition),
∂n
where α = α(x) ≥ 0, β = β(x) ≥ 0, α + β > 0 (all functions are assumed to be continuous
in V ), x = (x, y, z), and ∂w/∂n is the outward normal derivative on S. The boundary
conditions of the first and second kind correspond to the special values β = 0 (α = 1) and
α = 0 (β = 1), respectively.
Solution:
∞
X
aλk t
w(x, t) = Ak uk (x) exp − ,
1 + bλk
k=1
Z Z
1 2
Ak = f (x)uk (x) dv, kuk k = u2k (x) dv,
kuk k2 V V
where λk and uk (x) are the eigenvalues and eigenfunctions of the auxiliary problem
∆u + λu = 0, x ∈ V,
∂u
αu + β = 0, x ∈ S.
∂n
Let us list the main properties of the eigenvalues λk and eigenfunctions uk (x) (e.g., see
Miranda (1970) and Vladimirov (1988)).
1. All eigenvalues are nonnegative. They can be numbered in ascending order,
0 ≤ λ1 ≤ λ2 ≤ · · · , λk → ∞ as k → ∞,
where each λk is repeated a number of times equal to its multiplicity.

2. The eigenfunctions uk = uk (x) can be chosen to be real and orthogonal,
Z
uk um dv = 0 if k 6= m.
V
3. Every function f (x) twice continuously differentiable in V , continuously differ-

entiable in V ∪ S, and satisfying the same boundary condition as w and the condition
∆f ∈ L2 (V ) can be expanded in a Fourier series in the orthonormal eigenfunction system,
∞
X Z
1
f (x) = Ak uk (x), Ak = f (x)uk (x) dv.
kuk k2 V
k=1
4. The system of eigenfunctions is complete in L2 (V ).

5. A necessary and sufficient condition that λ = 0 is an eigenvalue is that α = 0; then
λ1 = 0 is a simple eigenvalue, and the corresponding eigenfunction is u1 = const.
∂w ∂
2. − a∆w − b ∆w = Φ(x, t).
∂t ∂t
Equation of filtration of a fluid in a cracked porous medium.
Domain: x ∈ Rn . Cauchy problem.
w = f (x) at t = 0.
Solution:
Z tZ Z

w= Ee (x − y, t − τ )Φ(y, τ ) dVy dτ + Ee (x − y, t) f (y) − b∆y f (y) dVy ,
0 Rn Rn
where ∆y is the Laplace operator in the integration variables y1 , . . . , yn .

For the function Ee (x, t) with n = 2 and n = 3, see Items 1◦ and 2◦ in Eq. 11.1.5.1.
∂2w ∂
3. − a∆w − b ∆w = 0.
∂t2 ∂t
This equation arises when decomposing unsteady equations describing 3D motions of vis-
cous compressible barotropic fluids (see Section 12.12.2).
Consider a finite open domain V with boundary S. The following conditions are pre-
scribed:
∂w
αw + β = 0 at x ∈ S (boundary condition),
∂n
Solution:
∞
X
w(x, t) = Ak ψk1 (t) + Bk ψk2 (t) uk (x),
k=1
where
γk1 exp(−γk2 t) − γk2 exp(−γk1 t) exp(−γk2 t) − exp(−γk1 t)

ψk1 (t) = , ψk2 (t) = ,
γk1 − γk2 γk1 − γk2
q q
bλk + b2 λ2k − 4aλk bλk − b2 λ2k − 4aλk
γk1 = , γk2 = ,
Z 2 Z 2
1 1
Ak = 2
f (x)uk (x) dv, Bk = g(x)uk (x) dv,
kuk k V kuk k2 V
and the eigenvalues λk and eigenfunctions uk (x) are the same as in the solution of equation
11.1.5.1, Item 4◦ .
11.2. Fourth-Order One-Dimensional Nonstationary Equations 957
11.2 Fourth-Order One-Dimensional Nonstationary

Equations
∂w ∂4w
11.2.1 Equation of the Form + a2 = Φ(x, t)
∂t ∂x4
◮ Particular solutions and the formal series solution.
w(x) = Ax3 + Bx2 + Cx + D,

w(x, t) = A(x5 − 120a2 xt) + B(x4 − 24a2 t),

w(x, t) = A sin(λx) + B cos(λx) + C sinh(λx) + D cosh(λx) exp(−λ4 a2 t),
where A, B, C, D, and λ are arbitrary constants.

2◦ . Formal series solution with Φ(x, t) = 0:
∞
X a2k tk d 4k f (x)
w(x, t) = f (x) + (−1)k ,
k! dx4k
k=1
condition w(x, 0) = f (x). If the function f (x) is a polynomial of degree n, then the solution
is a polynomial in x of degree n as well.
◮ Fundamental solution and the Cauchy problem.
Z ∞
1
E (x, t) = exp(−a2 ξ 4 t) cos(xξ) dξ.
π 0

w = f (x) at t = 0.
Solution:
Z ∞ Z tZ ∞
−∞ 0 −∞
◮ Solutions of boundary value problems in terms of the Green’s function.
1◦ . Consider problems on an interval 0 ≤ x ≤ l with the general initial condition
w = f (x) at t=0
Z l Z tZ l
w(x, t) = f (ξ)G(x, ξ, t) dξ + Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ. (1)
0 0 0
2◦ . The Green’s functions can be evaluated from the formula

∞
X ϕn (x)ϕn (ξ)
G(x, ξ, t) = exp(−λ4n a2 t), (2)
kϕn k2
n=1
where the λn and ϕn (x) are determined by solving the self-adjoint eigenvalue problem for
the fourth-order ordinary differential equation
ϕ′′′′ − λ4 ϕ = 0
subject to appropriate boundary conditions; the prime denotes differentiation with respect
to x. The norms of eigenfunctions can be calculated by the formula
Z l
2 l 2 l 2 l
kϕn k = ϕ2n (x) dx = ϕn (l) + 4 ϕ′′n (l) − 4 ϕ′n (l)ϕ′′′
n (l). (3)
0 4 4λn 2λn
Relations (2) and (3) are written under the assumption that λ = 0 is not an eigenvalue.
◮ Green’s functions for various boundary value problems.

Items 1◦ through 8◦ below present the Green’s functions for various types of boundary
conditions. The solutions of these boundary value problems are represented via the Green’s
functions by formula (1).
1◦ . Domain: 0 ≤ x ≤ l. The function and its first derivative are prescribed at the boundaries:
w = ∂x w = 0 at x = 0, w = ∂x w = 0 at x = l.
Green’s function:
∞
4X λ4n 4 2
G(x, ξ, t) = ϕn (x)ϕn (ξ) exp(−λn a t),
l n=1 ϕ′′n (l) 2
where

ϕn (x) = sinh(λn l) − sin(λn l) cosh(λn x) − cos(λn x)

− cosh(λn l) − cos(λn l) sinh(λn x) − sin(λn x) ;
the λn are positive roots of the transcendental equation cosh(λl) cos(λl) = 1. The numer-
ical values of the roots can be calculated from the formulas
µn π
λn = , where µ1 = 4.730, µ2 = 7.859, µn = (2n + 1) for n ≥ 3. (4)
l 2
2◦ . Domain: 0 ≤ x ≤ l. The function and its second derivative are prescribed at the
boundaries:
w = ∂xx w = 0 at x = 0, w = ∂xx w = 0 at x = l.
Green’s function:
∞
2X πn
G(x, ξ, t) = sin(λn x) sin(λn ξ) exp(−λ4n a2 t), λn = .
l l
n=1
3◦ . Domain: 0 ≤ x ≤ l. The first and third derivatives are prescribed at the boundaries:
∂x w = ∂xxx w = 0 at x = 0, ∂x w = ∂xxx w = 0 at x = l.
Green’s function:
∞
1 2X πn
G(x, ξ, t) = + cos(λn x) cos(λn ξ) exp(−λ4n a2 t), λn = .
l l l
n=1
4◦ . Domain: 0 ≤ x ≤ l. The second and third derivatives are prescribed at the boundaries:
wxx = ∂xxx w = 0 at x = 0, wxx = ∂xxx w = 0 at x = l.
Green’s function:
∞
1 3 4 X ϕn (x)ϕn (ξ)
G(x, ξ, t) = + 3 (2x − l)(2ξ − l) + exp(−λ4n a2 t),
l l l ϕ2n (l)
n=1
where

ϕn (x) = sinh(λn l) − sin(λn l) cosh(λn x) + cos(λn x)

− cosh(λn l) − cos(λn l) sinh(λn x) + sin(λn x)
and the λn are positive roots of the transcendental equation cosh(λl) cos(λl) = 1. For the
numerical values of the roots, see the end of Item 1◦ .
5◦ . Domain: 0 ≤ x ≤ l. Mixed conditions are prescribed at the boundaries (case 1):
w = ∂x w = 0 at x = 0, w = ∂xx w = 0 at x = l.
Green’s function:
∞
2 X 4 ϕn (x)ϕn (ξ)
G(x, ξ, t) = λn ′ exp(−λ4n a2 t),
l |ϕn (l)ϕ′′′
n (l)|
n=1
where


− cosh(λn l) − cos(λn l) sinh(λn x) − sin(λn x) ,
and the λn are positive roots of the transcendental equation tan(λl) − tanh(λl) = 0.
w = ∂x w = 0 at x = 0, ∂xx w = ∂xxx w = 0 at x = l.
Green’s function:
∞
4 X ϕn (x)ϕn (ξ)
G(x, ξ, t) = exp(−λ4n a2 t),
l n=1 ϕ2n (l)
where

ϕn (x) = sinh(λn l) + sin(λn l) cosh(λn x) − cos(λn x)

− cosh(λn l) + cos(λn l) sinh(λn x) − sin(λn x)
and the λn are positive roots of the transcendental equation cosh(λl) cos(λl) = −1.
w = ∂xx w = 0 at x = 0, ∂x w = ∂xxx w = 0 at x = l.
Green’s function:
∞
2X π(2n + 1)
G(x, ξ, t) = sin(λn x) sin(λn ξ) exp(−λ4n a2 t), λn = .
l 2l
n=0
w = ∂xx w = 0 at x = 0, ∂xx w = ∂xxx w = 0 at x = l.
Green’s function:
∞
4 X ϕn (x)ϕn (ξ)
G(x, ξ, t) = exp(−λ4n a2 t),
l ϕ2n (l)
n=1
where
ϕn (x) = sin(λn l) sinh(λn x) + sinh(λn l) sin(λn x);
the λn are positive roots of the transcendental equation tan(λl) − tanh(λl) = 0.
∂2w ∂4w
11.2.2 Equation of the Form + a2 =0
∂t2 ∂x4
This equation is encountered when studying transverse vibrations of elastic bars.
◮ Particular solutions and the formal series solution.

w(x, t) = (Ax3 + Bx2 + Cx + D)t + A1 x3 + B1 x2 + C1 x + D1 ,

w(x, t) = A sin(λx) + B cos(λx) + C sinh(λx) + D cosh(λx) sin(λ2 at),

w(x, t) = A sin(λx) + B cos(λx) + C sinh(λx) + D cosh(λx) cos(λ2 at),

w(x, t) = exp(−λx) A sin(λx) + B cos(λx) C exp(−2λ2 at) + D exp(2λ2 at) ,

w(x, t) = exp(λx) A sin(λx) + B cos(λx) C exp(−2λ2 at) + D exp(2λ2 at) ,
where A, B, C, D, A1 , B1 , C1 , D1 , and λ are arbitrary constants.

2◦ . Formal series solution:
∞
X ∞
a2k t2k d 4k f (x) X
k
2k 2k+1 d 4k g(x)
k a t
w(x, t) = (−1) + (−1) ,
(2k)! dx4k (2k + 1)! dx4k
k=0 k=0
the functions f (x) and g(x) are polynomials of degree ≤ n, then the solution is a polyno-
mial in x of degree ≤ n as well.
◮ Fundamental solution and the Cauchy problem.

Z ∞ Z
1 sin(aξ 2 t) ixξ 1 ∞ sin(aξ 2 t)
E (x, t) = e dξ = cos(xξ) dξ
2π −∞ aξ 2 π 0 aξ 2
2 2 2
x x x √ x π
= S −C + πa sin + ,
2a 4at 4at 4at 4
where C(z) and S(z) are the Fresnel integrals.

w = f (x) at t = 0, ∂t w = ag′′ (x) at t = 0.
Boussinesq solution:
Z ∞ √
1
w(x, t) = √ f x − 2ξ at cos ξ 2 + sin ξ 2 dξ
2π −∞
Z ∞
1 √
+ √ g x − 2ξ at cos ξ 2 − sin ξ 2 dξ.
a 2π −∞
⊙ Literature: I. Sneddon (1951).

◮ Boundary value problems.

1◦ . Domain: 0 ≤ x < ∞. The problem on free vibrations of a semi-infinite bar.
w=0 at t = 0, ∂t w = 0 at t = 0 (initial conditions),
w = f (t) at x = 0, ∂xx w = 0 at x = 0 (boundary conditions).
Boussinesq solution:
Z ∞
1 x2 ξ2 ξ2
w(x, t) = √ √ f t− sin + cos dξ.
π x/ 2at 2aξ 2 2 2
2◦ . For solutions of various boundary value problems in the domain 0 ≤ x ≤ l, see Sec-
tion 11.2.3 for Φ ≡ 0.
∂2w ∂4w
11.2.3 Equation of the Form + a2 = Φ(x, t)
∂t2 ∂x4
This equation is encountered when studying forced (transverse) vibrations of elastic bars.

1◦ . In this section we consider boundary value problems on an interval 0 ≤ x ≤ l with the
general initial conditions
w = f (x) at t = 0, ∂t w = g(x) at t = 0
Z l Z l
∂
w(x, t) = f (ξ)G(x, ξ, t) dξ + g(ξ)G(x, ξ, t) dξ
∂t 0 0
Z tZ l (1)
+ Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ.
0 0
∞
1 X ϕn (x)ϕn (ξ)
G(x, ξ, t) = sin(λ2n at), (2)
a λ2n kϕn k2
n=1
where the λn and ϕn (x) are determined by solving the self-adjoint eigenvalue problem for
the fourth-order ordinary differential equation
ϕ′′′′ − λ4 ϕ = 0
subject to appropriate boundary conditions; the prime denotes differentiation with respect
to x. The norms of eigenfunctions can be calculated by Krylov’s formula [see Krylov
(1949)]:
Z l
l l 2 l
kϕn k2 = ϕ2n (x) dx = ϕ2n (l) + 4 ϕ′′n (l) − 4 ϕ′n (l)ϕ′′′ n (l). (3)
0 4 4λn 2λn
Relations (2) and (3) are written under the assumption that λ = 0 is not an eigenvalue.

1◦ . Domain: 0 ≤ x ≤ l. Both ends of the bar are clamped.
w = ∂x w = 0 at x = 0, w = ∂x w = 0 at x = l.
Green’s function:
∞
4 X λ2n 2
G(x, ξ, t) = ϕn (x)ϕn (ξ) sin(λn at),
al ϕ′′ (l) 2
n=1 n
where


− cosh(λn l) − cos(λn l) sinh(λn x) − sin(λn x)
and the λn are positive roots of the transcendental equation cosh(λl) cos(λl) = 1. The
numerical values of the roots can be calculated from the formulas
µn π
λn = , where µ1 = 4.730, µ2 = 7.859, µn = (2n + 1) for n ≥ 3.
l 2
2◦ . Domain: 0 ≤ x ≤ l. Both ends of the bar are hinged.

Green’s function:
∞
2l X 1 πn
G(x, ξ, t) = sin(λn x) sin(λn ξ) sin(λ2n at), λn = .
aπ 2 n2 l
n=1
⊙ Literature: A. N. Krylov (1949), B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
3◦ . Domain: 0 ≤ x ≤ l. Both ends of the bar are free.

Green’s function:
∞
t 3t 4 X ϕn (x)ϕn (ξ)
G(x, ξ, t) = + 3 (2x − l)(2ξ − l) + sin(λ2n at),
l l al n=1 λ2n ϕ2n (l)
where


numerical values of the roots, see the end of Item 1◦ .
The first two terms in the expression for the Green’s function correspond to the zero
(1)
eigenvalue λ0 = 0, to which there correspond two orthogonal eigenfunctions w0 = 1 and
(2) (1) (2)
w0 = 2x − l with kw0 k2 = l and kw0 k2 = 13 l3 .
⊙ Literature: A. N. Krylov (1949).
4◦ . Domain: 0 ≤ x ≤ l. One end of the bar is clamped and the other is hinged.
w = ∂x w = 0 at x = 0, w = ∂xx w = 0 at x = l.
Green’s function:
∞
2 X 2 ϕn (x)ϕn (ξ)
G(x, ξ, t) = λ sin(λ2n at),
al n=1 n |ϕ′n (l)ϕ′′′
n (l)|
where


5◦ . Domain: 0 ≤ x ≤ l. One end of the bar is clamped and the other is free.
Green’s function:
∞
4 X ϕn (x)ϕn (ξ)
G(x, ξ, t) = sin(λ2n at),
al λ2n ϕ2n (l)
n=1
where


6◦ . Domain: 0 ≤ x ≤ l. One end of the bar is hinged and the other is free.
Green’s function:
∞
4 X ϕn (x)ϕn (ξ)
G(x, ξ, t) = sin(λ2n at),
al n=1 λ2n ϕ2n (l)
where
ϕn (x) = sin(λn l) sinh(λn x) + sinh(λn l) sin(λn x)
7◦ . Domain: 0 ≤ x ≤ l. The first and third derivatives are prescribed at the ends:
Green’s function:
∞
t 2 X 1 πn
G(x, ξ, t) = + cos(λn x) cos(λn ξ) sin(λ2n at), λn = .
l al n=1 λ2n l
8◦ . Domain: 0 ≤ x ≤ l. Mixed boundary conditions are prescribed at the ends:
Green’s function:
∞
2 X 1 π(2n + 1)
G(x, ξ, t) = sin(λn x) sin(λn ξ) sin(λ2n at), λn = .
al n=0 λ2n 2l
∂2w ∂4w
11.2.4 Equation of the Form + a2 + kw = Φ(x, t)
∂t2 ∂x4
◮ Particular solutions.
Particular solutions of the homogeneous equation (Φ ≡ 0):

√
w(x, t) = (Ax3 + Bx2 + Cx + D) sin t k ,
√
w(x, t) = (Ax3 + Bx2 + Cx + D) cos t k ,
p
w(x, t) = A sin(λx) + B cos(λx) + C sinh(λx) + D cosh(λx) sin t a2 λ4 + k ,
p
w(x, t) = A sin(λx) + B cos(λx) + C sinh(λx) + D cosh(λx) cos t a2 λ4 + k ,
where A, B, C, D, and λ are arbitrary constants.

◮ Solution of boundary value problems in terms of the Green’s function.
1◦ . Consider boundary value problems on the interval 0 ≤ x ≤ l with the general initial
conditions
w = f (x) at t = 0, ∂t w = g(x) at t = 0
Z l Z l
∂
w(x, t) = f (ξ)G(x, ξ, t) dξ + g(ξ)G(x, ξ, t) dξ
∂t 0 0
Z tZ l (1)
+ Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ.
0 0

∞ p
X ϕn (x)ϕn (ξ) sin t a2 λ4n + k
G(x, ξ, t) = 2
p , (2)
n=1
kϕ n k a2 λ4 + k
n
where the λn and ϕn (x) are determined by solving the self-adjoint eigenvalue problem
for the fourth-order ordinary differential equation ϕ′′′′ − λ4 ϕ = 0 subject to appropriate
boundary conditions. The norms of eigenfunctions can be calculated by formula (3) in
Section 11.2.3.
1◦ . Domain: 0 ≤ x ≤ l. The function and its first derivative are prescribed at the ends:
w = ∂x w = 0 at x = 0, w = ∂x w = 0 at x = l.
Green’s function:
∞ p
4 X 4 ϕn (x)ϕn (ξ) sin t a2 λ4n + k d2 ϕn
G(x, ξ, t) = λn p , ϕ′′n (x) = ,
l ϕ′′ (l) 2 a2 λ4n + k dx2
n=1 n
where


and the λn are positive roots of the transcendental equation cosh(λl) cos(λl) = 1.
2◦ . Domain: 0 ≤ x ≤ l. The function and its second derivative are prescribed at the ends:
Green’s function:
∞ p
2X sin t a2 λ4n + k πn
l a2 λ4n + k l
n=1
3◦ . Domain: 0 ≤ x ≤ l. The first and third derivatives are prescribed at the ends:
Green’s function:
√ ∞ p
sin t k 2X sin t a2 λ4n + k πn
G(x, ξ, t) = √ + cos(λn x) cos(λn ξ) p , λn = .
l k l a2 λ4 + k
n
l
n=1
4◦ . Domain: 0 ≤ x ≤ l. The second and third derivatives are prescribed at the ends:
Green’s function:
√ ∞ p
3 sin t k 4 X ϕn (x)ϕn (ξ) sin t a2 λ4n + k
G(x, ξ, t) = 1+ 2 (2x−l)(2ξ −l) √ + p ,
l l k l n=1 ϕ2n (l) a2 λ4n + k
where


numerical values of the roots, see formulas (4) in Section 11.2.1.
5◦ . Domain: 0 ≤ x ≤ l. Mixed boundary conditions are prescribed at the ends (case 1):
w = ∂x w = 0 at x = 0, w = ∂xx w = 0 at x = l.
Green’s function:
∞ p
2 X 4 ϕn (x)ϕn (ξ) sin t a2 λ4n + k
G(x, ξ, t) = λ p ,
l n=1 n |ϕ′n (l)ϕ′′′
n (l)| a2 λ4n + k
where


Green’s function:
∞ p
4 X ϕn (x)ϕn (ξ) sin t a2 λ4n + k
G(x, ξ, t) = p ,
l n=1 ϕ2n (l) a2 λ4n + k
where


Green’s function:
∞ p
2X sin t a2 λ4n + k π(2n + 1)
l a2 λ4n + k 2l
n=0
Green’s function:
∞ p
4 X ϕn (x)ϕn (ξ) sin t a2 λ4n + k
G(x, ξ, t) = p ,
l ϕ2n (l) a2 λ4n + k
n=1
where
ϕn (x) = sin(λn l) sinh(λn x) + sinh(λn l) sin(λn x);
the λn are positive roots of the transcendental equation tan(λl) − tanh(λl) = 0.
11.2.5 Other Equations without Mixed Derivatives

◮ Equations containing the first derivative with respect to t.
∂w ∂4w
1. + a2 + kw = Φ(x, t).
∂t ∂x4
The change of variable w(x, t) = e−kt u(x, t) leads to the equation
∂u ∂4u
+ a2 4 = ekt Φ(x, t),
∂t ∂x
∂w ∂4w
2. = ax8 .
∂t ∂x4
This is a special case of equation 11.7.4.3 with k = 1 and n = 4.
∂w ∂ 4w ∂w
3. = k(t) + [xf (t) + g(t)] + h(t)w.
∂t ∂x4 ∂x
This is a special case of equation 11.7.4.1 with n = 4. The transformation
Z Z Z
w(x, t) = u(z, τ ) exp h(t) dt , z = xF (t) + g(t)F (t) dt, τ = k(t)F 4 (t) dt,
Z
where F (t) = exp f (t) dt , leads to the constant coefficient equation
∂u ∂4u
= ,
∂τ ∂z 4
∂w ∂4w
4. = (ax2 + bx + c)4 .
∂t ∂x4
◮ Equations containing the second derivative with respect to t.

∂2w ∂ 4w
5. − a2 = 0.
∂t2 ∂x4
w = w1 + w2 ,
where w1 and w2 are arbitrary solutions of two second-order equations
∂w1 ∂ 2 w1 ∂w2 ∂ 2 w2
−a = 0, + a = 0.
∂t ∂x2 ∂t ∂x2
2◦ . Formal series solution:
X∞ ∞
a2k t2k d 4k f (x) X a2k t2k+1 d 4k g(x)
w(x, t) = + ,
(2k)! dx4k (2k + 1)! dx4k
k=0 k=0
the functions f (x) and g(x) are polynomials of degree ≤ n, then the solution is a polyno-
mial in x of degree ≤ n as well.
∂2w ∂w ∂4w
6. +k + a2 = Φ(x, t).
∂t2 ∂t ∂x4
For Φ(x, t) ≡ 0, this equation describes transverse vibrations of an elastic bar in a resisting
medium with resistance coefficient proportional to the velocity.
The change of variable w(x, t) = exp − 12 kt u(x, t) leads to the equation
∂2u 4
2∂ u

+ a − 14 k2 u = exp 1
2 kt Φ(x, t),
∂t2 ∂x4
∂2w ∂4w
7. = ax8 .
∂t2 ∂x4
∂2w ∂4w
8. = (ax2 + bx + c)4 .
∂t2 ∂x4
◮ Equation containing the fourth derivative with respect to t.
∂4w ∂4w
9. − = 0.
∂t4 ∂x4
1◦ . Particular solution:
w(x, t) = f (x − t) + g(x + t),
where f (z1 ) and g(z2 ) are arbitrary functions.
w = w1 + w2 ,
where w1 and w2 are arbitrary solutions of two second-order equations
∂ 2 w1 ∂ 2 w1 ∂ 2 w2 ∂ 2 w2
− = 0, + = 0.
∂t2 ∂x2 ∂t2 ∂x2
p
1 x 1
E (x, t) = t ln x2 + t2 − x arctan − (t + x) ln |t + x|
2π t 2

1 1 1
− (t − x) ln |t − x| + |t + x| + |t − x| .
2 8 8

w = 0 at t = 0, ∂t w = 0 at t = 0, ∂tt w = f (x) at t = 0.
Solution:
Z ∞
w(x, t) = E (x − ξ, t)f (ξ) dξ.
−∞
⊙ Literature: G. E. Shilov (1965).

11.2.6 Equations Containing Second Derivative in x and Mixed

Derivatives
∂2w ∂2w ∂4w
1. − − = 0.
∂t2 ∂x2 ∂t2 ∂x2
One-dimensional wave equation with strong dispersion. This equation describes the dy-
namics of interior one-dimensional wave motions in an exponentially stratified fluid as
well as one-dimensional longitudinal vibrations of a rigid Rayleigh bar of constant cross-
section. This is a special case of equation 11.2.6.4 with a(x) = b(x) = c(x) = 1 and
Φ(x, t) = 0.

λx λx
w = C1 exp(−λt) + C2 exp(λt) C3 exp − √ + C4 exp √ ,
1 + λ2 1 + λ2

λt λt
w = C1 cos(λx) + C2 sin(λx) C3 cos √ + C4 sin √ ,
1 + λ2 1 + λ2

λx λx
w = C1 cos(λt) + C2 sin(λt) C3 exp − √ + C4 exp √ , |λ| > 1,
λ2 − 1 λ2 − 1
where C1 , C2 , C3 , C4 , and λ are arbitrary constants.
Z ∞
1 ei|x|ξ tξ
Ee (x, t) = p sin p dξ,
2π −∞ ξ 1 + ξ 2 1 + ξ2
Z
1 ∞ cos(|x|ξ) tξ
= p sin p dξ, i2 = −1.
π 0 ξ 1 + ξ2 1 + ξ2
Here the absolute value is only written to emphasize that the function Ee (x, t) is even in the
variable x.
The following asymptotic formulas hold:
(i) For x = 0 and t → ∞,
Z
1 t 1
Ee (0, t) = J0 (ξ) dξ = + O(t−1/2 ).
2 0 2
(ii) For |x|/t ∈ [δ, 1 − δ], where δ ∈ (0, 1/2), and t → ∞,
1 sin[(t2/3 − |x|2/3 )3/2 − π/4]

Ee (x, t) = + √ + O(t−1 ).
2 6π(t2/3 − |x|2/3 )3/4
(iii) For |x| = t and t → ∞,
1
Ee (x, t) = + O(t−1 ).
6
(iv) For |x|/t ∈ [1 + δ, ∞), where δ > 0, and |x| → ∞,
exp[−(|x|2/3 − t2/3 )3/2 ]

Ee (x, t) = √ 1 + O(|x|−1/2 ) .
2/3
2 6π(|x| − t ) 2/3 3/4
(v) For t → 0,
Ee (x, t) = 12 te−|x| 1 + O(t2 ) .
w = f0 (x) at t = 0,
∂t w = f1 (x) at t = 0.
Solution:
Z ∞
∂
w(x, t) = Ee (x − y, t)[f0 (y) − f0′′ (y)] dy
∂t −∞
Z ∞
+ Ee (x − y, t)[f1 (y) − f1′′ (y)] dy.
−∞

w = f0 (x) at t = 0,
∂t w = f1 (x) at t = 0,
w = g(t) at x = 0.
Solution:
Z ∞ Z ∞
∂
w(x, t) = G1 (x, y, t)[f0 (y) − f0′′ (y)] dy
+ G1 (x, y, t)[f1 (y) − f1′′ (y)] dy
∂t 0 0
2 Z t
∂
− 2g(t) Ee (x, t) −2 B[Ee ](x, t − τ )g(τ ) dτ,
∂t2 t=0 0
∂2
G1 (x, y, t) = Ee (x − y, t) − Ee (x + y, t), B[u](x, t) = u(x, t) + u(x, t).
∂t2
5◦ . Domain: 0 ≤ x < ∞. A boundary value problem.
w = f0 (x) at t = 0,
∂t w = f1 (x) at t = 0,
∂x w + ∂ttx w = g(t) at x = 0.
Solution:
Z ∞ Z ∞
∂ ′′
w(x, t) = G2 (x, y, t)[f0 (y) − f0 (y)] dy + G2 (x, y, t)[f1 (y) − f1′′ (y)] dy
∂t 0 0
Z t
− G2 (x, 0, t − τ )g(τ ) dτ, G2 (x, y, t) = Ee (x − y, t) + Ee (x + y, t).
0


Solution:
∞ h
X i
bn πn βn
w(x, t) = an cos(λn t) + sin(λn t) sin(βn x), βn = , λn = p ,
n=1
λ n l 1 + βn
2
Z Z
2 l 2 l
an = f0 (ξ) sin(βn ξ) dξ, bn = f1 (ξ) sin(βn ξ) dξ.
l 0 l 0
An alternative representation of the solution:
Z l Z l
∂
w(x, t) = G(x, ξ, t)f0 (ξ) dξ + G(x, ξ, t)f1 (ξ) dξ,
∂t 0 0
∞
2X 1
G(x, ξ, t) = sin(βn x) sin(βn ξ) sin(λn t).
l n=1 λn


Solution:
∞ h
X i
bn
w(x, t) = a0 + b0 t + an cos(λn t) +
sin(λn t) cos(βn x),
n=1
λn
Z l Z
πn βn 1 1 l
βn = , λn = p , a0 = f0 (ξ) dξ, bn = f1 (ξ) dξ,
l 1 + βn2 l 0 l 0
Z Z
2 l 2 l
l 0 l 0

Solution:
∞ h
X i
bn
w(x, t) = an cos(λn t) + sin(λn t) sin(βn x),
λn
n=0
π(2n + 1) βn
βn = , λn = p ,
2l 1 + βn2
Z l Z
2 2 l
l 0 l 0
⊙ Literature for equation 11.2.6.1: S. A. Gabov and B. B. Orazov (1986), S. A. Gabov and A. G. Sveshnikov
(1990), I. A. Fedotov, A. D. Polyanin, and M. Yu. Shatalov (2007), I. A. Fedotov, Y. Gai, A. D. Polyanin, and
M. Yu. Shatalov (2008).
∂2w ∂2w ∂4w

2. − − = Φ(x, t).
∂t2 ∂x2 ∂t2 ∂x2
This is a special case of equation 11.2.6.4 with a(x) = b(x) = c(x) = 1. For Φ(x, t) = 0,
see equation 11.2.6.1.
w = f0 (x) at t = 0,
∂t w = f1 (x) at t = 0.
The solution of the Cauchy problem is given by the formula in Item 3◦ of equation
11.2.6.1 with the additional term
Z tZ ∞
Ee (x − y, t − τ )Φ(y, τ ) dy dτ.
0 −∞

w = f0 (x) at t = 0,
∂t w = f1 (x) at t = 0,
w = g(t) at x = 0.
The solution of the first boundary value problem is given by the formula in Item 4◦ of
equation 11.2.6.1 with the additional term
Z tZ ∞
G1 (x, y, t − τ )Φ(y, τ ) dy dτ.
0 0
3◦ . Domain: 0 ≤ x < ∞. A boundary value problem.

w = f0 (x) at t = 0,
∂t w = f1 (x) at t = 0,
∂x w + ∂ttx w = g(t) at x = 0.
The solution of the boundary value problem is given by the formula in Item 5◦ of
Z tZ ∞
G2 (x, y, t − τ )Φ(y, τ ) dy dτ.
0 0


The solution of the boundary value problem is given by the formula in Item 6◦ of
X∞ Z t
sin(βn x) βn (t − τ )
p ϕn (τ ) sin p dτ,
2 1 + βn2
n=1 βn 1 + βn 0
where Z l
πn 2
βn = , ϕn (τ ) = Φ(ξ, τ ) sin(βn ξ) dξ.
l l 0
⊙ Literature: S. A. Gabov and A. G. Sveshnikov (1990).
∂2w ∂2w ∂4w

3. − − + σ2 w = Φ(x, t), σ2 6= 1.
∂t2 ∂x2 ∂t2 ∂x2
One-dimensional wave equation with strong dispersion taking into account the rotation of
an exponentially stratified fluid.
1◦ . Particular solutions of the homogeneous equation with Φ ≡ 0:
r 2 r 2
λ + σ2 λ + σ2
w = C1 exp(−λt) + C2 exp(λt) C3 exp −x + C 4 exp x ,
λ2 + 1 λ2 + 1
r 2 r 2
λ + σ2 λ + σ2
w = C1 cos(λx) + C2 sin(λx) C3 cos t + C 4 sin t ,
λ2 + 1 λ2 + 1
r 2 r 2
λ − σ2 λ − σ2
w = C1 cos(λt) + C2 sin(λt) C3 exp −x + C4 exp x ,
λ2 − 1 λ2 − 1
where C1 , C2 , C3 , C4 , and λ are arbitrary constants.

Z ∞ p 2
1 eiµx σ + µ2
Ee (x, t) = p p sin t p dµ, i2 = −1.
2π −∞ 1 + µ2 σ 2 + µ2 1 + µ2

w = f0 (x) at t = 0,
∂t w = f1 (x) at t = 0.
Solution:
Z tZ ∞
w(x, t) = E (x − y, t − τ )Φ(y, τ ) dy dτ
0 −∞
Z ∞ Z ∞
∂
+ E (x − y, t)[f0 (y) − f0′′ (y)] dy + E (x − y, t)[f1 (y) − f1′′ (y)] dy.
∂t −∞ −∞

h i
∂ 2w ∂ ∂w ∂3w
4. a(x) − b(x) + c(x) = Φ(x, t).
∂t2 ∂x ∂x ∂ 2 t∂x
This equation describes one-dimensional longitudinal vibrations of a rigid Rayleigh bar of
variable cross-section and takes into account the inertial effects of the transverse motion.
The coefficients of the equation are positive functions and can be expressed as follows via
geometric and physical variables:
a(x) = ρ(x)S(x), b(x) = S(x)E(x), c(x) = ρ(x)ν 2 (x)I(x),
where ρ(x) is the bar mass density, S(x) is the transverse cross-section area, E(x) is the
Young modulus, ν(x) is the Poisson ratio, and I(x) is the polar moment of inertia.
◮ Initial and boundary conditions.

This equation is considered with the general boundary conditions
∂w
w = f (x) at t = 0, = g(x) at t=0 (1)
∂t
and with two types of boundary conditions given below.
(i) Geometric boundary conditions (rigid clamping of the bar ends):
w = 0 at x = 0, w=0 at x = l. (2)
(ii) Dynamic boundary conditions (the bar ends are fixed with the use of springs and
lumped masses):
∂w ∂3w ∂2w
b(x) + c(x) 2 − K1 w − M1 2 = 0 at x = 0,
∂x ∂ t∂x ∂t (3)
∂w ∂3w ∂2w
b(x) + c(x) 2 + K2 w + M2 2 = 0 at x = l,
∂x ∂ t∂x ∂t
where Kn ≥ 0 and Mn ≥ 0. The boundary conditions (3) express the balance of elastic
deformation forces of the Rayleigh bar (the first two terms), elastic deformation forces of
the springs (the terms proportional to Kn ), and the inertia forces exerted on the bar by the
accelerated lumped masses (the terms proportional to Mn ).
◮ Free vibrations of the Rayleigh bar. The Sturm–Liouville problem.
Free vibrations are determined by the relations
w(x, t) = [A cos(ωt) + B sin(ωt)]ϕ(x), Φ(x, t) = 0.
By substituting these relations into the equation and the dynamic boundary conditions (3),
we arrive at the nonclassical Sturm–Liouville boundary value problem
[p(x, ω)ϕ′x ]′x + ω 2 a(x)ϕ = 0, (4)

p(0, ω)ϕ′x 2
− (K1 − ω M1 )ϕ = 0 at x = 0, (5)
p(l, ω)ϕ′x 2
+ (K2 − ω M2 )ϕ = 0 at x = l, (6)
where
p(x, ω) = b(x) − ω 2 c(x). (7)
We point out that the spectral parameter ω occurs in Eq. (4) as well as in the boundary
conditions (5)–(6) in a complicated way. Without loss of generality, we can assume that
ω > 0.
In the general case, the eigenvalues ωn of the Sturm–Liouville problem (4)–(7) are
bounded and sit on the interval
p
0 ≤ |ωn | < ω∞ = inf b(x)/c(x),
0≤x≤l
which is ensured by the positivity of the coefficient p(x, ωn ) in Eq. (4).
◮ Two types of orthogonality of eigenfunctions.
Let ωm and ωn be two distinct eigenvalues of the Sturm–Liouville problem (4)–(7), and let
ϕm (x) and ϕn (x) be the corresponding eigenfunctions.
1◦ . First type of orthogonality is the mass orthogonality of eigenfunctions for the case of
the dynamic boundary conditions (3),
Z l
a(x)ϕm (x)ϕn (x) + c(x)ϕ′m (x)ϕ′n (x) dx
0 (8)
+ M1 ϕm (0)ϕn (0) + M2 ϕm (l)ϕn (l) = 0, m 6= n.
The corresponding squared mass norm of the eigenfunction is given by

Z l
kϕn k21 = a(x)ϕ2n (x) + c(x)ϕ′n2 (x) dx + M1 ϕ2n (0) + M2 ϕ2n (l). (9)
0
For the geometric boundary conditions (2), one should set M1 = M2 = 0 in formulas
(8) and (9).
2◦ . Second type of orthogonality is the stiffness orthogonality of eigenfunctions for the

case of the dynamic boundary conditions (3),
Z l
b(x)ϕ′m (x)ϕ′n (x) dx + K1 ϕm (0)ϕn (0) + K2 ϕm (l)ϕn (l) = 0, m 6= n. (10)
0
The corresponding squared stiffness norm of the eigenfunction is given by

Z l
2
kϕn k2 = b(x)[ϕ′n (x)]2 dx + K1 ϕ2n (0) + K2 ϕ2n (l) = ωn2 kϕn k21 . (11)
0
For the geometric boundary conditions (2), one should set K1 = K2 = 0 in formulas
(10) and (11).
◮ Two representations of solutions of the general initial-boundary value problem via

Green’s functions.
1◦ . The solution of the problem on longitudinal vibrations of a Rayleigh bar with the gen-
eral initial conditions (1) and the dynamic boundary conditions (3) can be represented in
the form
Z l
∂ ′ ∂
w(x, t) = a(y)f (y)G1 (x, y, t) + c(y)f (y) G1 (x, y, t) dy
∂t 0 ∂y
Z l
′ ∂
+ a(y)g(y)G1 (x, y, t) + c(y)g (y) G1 (x, y, t) dy
0 ∂y
Z tZ l
+ Φ(y, τ )G1 (x, y, t − τ ) dy dτ (12)
0 0

∂
+ M1 f (0) G1 (x, 0, t) + g(0)G1 (x, 0, t)
∂t

∂
+ M2 f (l) G1 (x, l, t) + g(l)G1 (x, l, t) ,
∂t
where G1 (x, y, t) is the mass Green’s function

∞
X 1
G1 (x, y, t) = ϕn (x)ϕn (y) sin(ωn t). (13)
n=1
ωn kϕn k21
For the geometric boundary conditions (2), one should set M1 = M2 = 0 in the above
formula for w(x, t).
Remark 11.2. The solution (12) can be reduced to the form
Z l
∂
w(x, t) = G1 (x, y, t) a(y)f (y) − [c(y)f ′ (y)]′ dy
∂t 0
Z l

+ G1 (x, y, t) a(y)g(y) − [c(y)g ′ (y)]′ dy + · · · ,
0
where we should add the last three lines in (12).

2◦ . An alternative representation of the solution of the problem on longitudinal vibrations

of a Rayleigh bar:
Z l
′ ∂2 ′ ∂
w(x, t) = b(y) f (y) G2 (x, y, t) + g (y) G2 (x, y, t) dy
0 ∂t∂y ∂y
Z tZ l 2
∂
− Φ(y, τ ) 2 G2 (x, y, t − τ ) dy dτ
0 ∂t
0
∂
+ K1 f (0) G2 (x, 0, t) + g(0)G2 (x, 0, t)
∂t

∂
+ K2 f (l) G2 (x, l, t) + g(l)G2 (x, l, t) ,
∂t
where G2 (x, y, t) is the stiffness Green’s function
∞
X 1
G2 (x, y, t) = ϕn (x)ϕn (y) sin(ωn t).
n=1
ωn kϕn k22
For the geometric boundary conditions (2), one should set K1 = K2 = 0 in the above
formula for w(x, t).
◮ Homogeneous bar of constant cross-section.

Consider longitudinal vibrations of a homogeneous Rayleigh bar of constant cross-section.
In this case, the coefficients of the original equation are constants, and the general solution
of Eq. (4) is given by the formula
r
a
ϕ(x) = C1 cos(λx) + C2 sin(λx), λ=ω , (14)
b − ω2c
where Cp1 and C2 are arbitrary constants. The vibration frequencies are bounded, |ω| <
ω∞ = b/c. In the degenerate case (c = 0), we have the usual model of vibrations
p of a
thin bar described by a second-order wave equation with λ = ω/v, where v = b/a is the
phase speed of propagation of perturbations in the bar.
For the dynamic boundary conditions (3) in the absence of lumped masses, M1 =
M2 = 0, the dispersion equation for the eigenvalues ωn has the form
(p2 λ2 − K1 K2 ) sin(λl) − (K1 + K2 )pλ cos(λl) = 0,
where p = b − ω 2 c and λ is defined in (14).

For the geometric boundary conditions (2), one should set K1 = K2 = 0 in the dispersion
equation, which results in the following eigenvalues and eigenfunctions:
s
b(πn)2 πnx
ωn = , ϕn (x) = sin , n = 1, 2, . . . (15)
al2 + c(πn)2 l
Formulas (12), (13), and (15) permit one to obtain the solution of the problem on the
longitudinal vibrations of a cylindrical Rayleigh bar with the general initial conditions (1)
and the homogeneous boundary conditions (2). See also equation 11.2.6.1.
◮ Uniform conical bar.

Consider a homogeneous conical bar of circular cross-section with constant physical pa-
rameters ρ, E, and ν. The cross-sectional area of the cone and the moment of inertia with
respect to the axis are given by the formulas
S(x) = π(x − x0 )2 , I(x) = 12 π(x − x0 )4 ,
where x0 is the coordinate of the cone vertex (x0 < 0). The coefficients of the original
equation have the form
a(x) = πρ(x − x0 )2 , b(x) = πE(x − x0 )2 , c(x) = 12 πρν 2 (x − x0 )4 .
In this case, the general solution of Eq. (4) can be represented in the form
Pσ (µ(x − x0 )) Qσ (µ(x − x0 ))
ϕ(x) = C1 + C2 ,
x − x0 x − x0
r r
ρ 1 9 2
µ = νω , σ=− + + 2,
2E 2 4 ν
where C1 and C2 are arbitrary constants and Pσ (z) and Qσ (z) are the Legendre functions
of the first and second kind, respectively. One can show that 0 < µ(x−x0 ) ≤ µ(l −x0 ) < 1.
⊙ Literature for equation 11.2.6.4: I. A. Fedotov, A. D. Polyanin, and M. Yu. Shatalov (2007), I. A. Fedotov,
Y. Gai, A. D. Polyanin, and M. Yu. Shatalov (2008).
11.2.7 Equations Containing Fourth Derivative in x and Mixed

Derivatives
◮ Equations containing the second derivative with respect to t.

∂ ∂2 2
1. − 2
w = 0.
∂t ∂x
1◦ . General solution (two representations):
w(x, t) = tu1 (x, t) + u2 (x, t),
w(x, t) = xu1 (x, t) + u2 (x, t),
where uk = uk (x, t) is an arbitrary function satisfying the heat equation ∂t uk − ∂xx uk = 0;
k = 1, 2.
√
t x2
E (x, t) = √ exp − .
2 π 4t
w = 0 at t = 0, ∂t w = f (x) at t = 0.
Solution: √ Z ∞
t (x − ξ)2
w(x, t) = √ exp − f (ξ) dξ.
2 π −∞ 4t
∂2w ∂2w ∂4w ∂ 4w

2. −a −b +c = Φ(x, t).
∂t2 ∂x2 ∂ 2 t∂x2 ∂x4
Rayleigh–Bishop equation with a > 0, b > 0, and c > 0. This equation describes one-
dimensional longitudinal vibrations of a thick short circular bar of constant cross-section.
1◦ . Particular solutions for the homogeneous equation with Φ(x, t) = 0:
s
a − cβ 2
w = [A exp(−βx) + B exp(βx)][C cos(λt) + sin(λt)], λ=β ,
bβ 2 − 1
s
a − cβ 2
w = [A exp(−βx) + B exp(βx)][C exp(−λt) + D exp(λt)], λ=β ,
1 − bβ 2
s
a + cβ 2
w = [A cos(βx) + B sin(βx)][C cos(λt) + sin(λt)], λ=β ,
1 + bβ 2
where A, B, C, D, and β are arbitrary constants. (It is assumed in all cases that the
radicands are positive.)
Z ∞ s
1 exp(−i|x|ξ) a + cξ 2
Ee (x, t) = p sin tξ dξ
2π −∞ ξ (1 + bξ 2 )(a + cξ 2 ) 1 + bξ 2
Z s
1 ∞ cos(|x|ξ) a + cξ 2
= p sin tξ dξ.
π 0 ξ (1 + bξ 2 )(a + cξ 2 ) 1 + bξ 2
Here the absolute value is only written to emphasize that the function Ee (x, t) is even in the
variable x.
w = f (x) at t = 0,
∂t w = g(x) at t = 0.
Solution:
Z tZ ∞
w(x, t) = Ee (x − y, t − τ )Φ(y, τ ) dy dτ
0 −∞
Z ∞ Z ∞
∂
+ Ee (x − y, t)[f (y) − f ′′ (y)] dy + Ee (x − y, t)[g(y) − g′′ (y)] dy.
∂t −∞ −∞

w = f (x) at t = 0,
∂t w = g(x) at t = 0,
w=0 at x = 0.
Solution:
Z tZ ∞ Z ∞
∂
w(x, t) = G(x − y, t − τ )Φ(y, τ ) dy dτ + G(x − y, t)[f (y) − f ′′ (y)] dy
∂t
Z0 ∞ 0 0
+ G(x − y, t)[g(y) − g ′′ (y)] dy, G(x, y, t) = Ee (x − y, t) − Ee (x + y, t).

0
5◦ . Domain: 0 ≤ x ≤ l. The boundary value problem corresponding to the rigid clamping

of the bar ends.

w = ∂xx w = 0 at x = 0 (boundary conditions),
w = ∂xx w = 0 at x = l (boundary conditions).
It is assumed that f (0) = f (l) = g(0) = g(l) = 0.

Solution:
Z tZ l
w(x, t) = G(x − y, t − τ )Φ(y, τ ) dy dτ
0 0
Z l Z l (1)
∂ ′′ ′′
+ G(x − y, t)[f (y) − f (y)] dy + G(x − y, t)[g(y) − g (y)] dy,
∂t 0 0
where
∞
2X 1
G(x, ξ, t) = sin(βn x) sin(βn y) sin(λn t),
l (1 + bβn2 )λn
n=1
s
πn a + cβn2
βn = , λn = βn .
l 1 + bβn2
6◦ . Domain: 0 ≤ x ≤ l. A boundary value problem.

w = ∂xx w = 0 at x = 0 (boundary conditions),
wx = ∂xxx w = 0 at x = l (boundary conditions).
The solution is determined by formula (1), where

∞
2X 1
G(x, ξ, t) = sin(βn x) sin(βn y) sin(λn t),
l (1 + bβn2 )λn
n=1
s
π(2n − 1) a + cβn2
βn = , λn = βn .
2l 1 + bβn2
7◦ . Domain: 0 ≤ x ≤ l. A boundary value problem.


wx = ∂xxx w = 0 at x = 0 (boundary conditions),
w = ∂xx w = 0 at x = l (boundary conditions).
The solution is determined by formula (1), where

∞
2X 1
G(x, ξ, t) = cos(βn x) cos(βn y) sin(λn t),
l n=1 (1 + bβn2 )λn
s
π(2n − 1) a + cβn2
βn = , λn = βn .
2l 1 + bβn2
h i h i h i
∂ 2w ∂ ∂w ∂ ∂ 3w ∂2 ∂ 2w
3. a(x) − b(x) − c(x) + d(x) = Φ(x, t).
∂t2 ∂x ∂x ∂x ∂ 2 t∂x ∂x2 ∂x2
This equation describes one-dimensional longitudinal vibrations of a thick short bar that
is a body of revolution around the x-axis. The Rayleigh–Bishop model is used, which takes
into account the lateral displacements as well as the transverse shear stresses. The coeffi-
cients of the equation are positive functions and can be expressed as follows via geometric
and physical variables:
a(x) = ρS(x), b(x) = ES(x), c(x) = ρν 2 I(x), d(x) = µν 2 I(x),
where ρ is the bar mass density, E is the Young modulus, ν is the Poisson ratio, µ =
E
2(1+ν) is the shear modulus, S(x) is the transverse cross-section area, and I(x) is the polar
moment of inertia.
◮ Initial and boundary conditions.

This equation is considered with the general initial conditions
∂w
w = f (x) at t = 0, = g(x) at t=0 (1)
∂t
and the homogeneous boundary conditions
∂2w ∂2w
w= = 0 at x = 0, w= = 0 at x = l, (2)
∂x2 ∂x2
which correspond to the rigid clamping of the bar ends.
◮ Free vibrations of a Rayleigh bar. The Sturm–Liouville problem.

Free vibrations are determined by the relations
w(x, t) = [A cos(ωt) + B sin(ωt)]ϕ(x), Φ(x, t) = 0. (3)

By substituting these relations into the equation and the boundary conditions (2), we arrive
at the nonclassical Sturm–Liouville eigenvalue problem
[d(x)ϕ′′ ]′′ + {[ω 2 c(x) − b(x)]ϕ′ }′ = ω 2 a(x)ϕ, (4)
′′
ϕ=ϕ =0 at x = 0, (5)
′′
ϕ=ϕ =0 at x = l, (6)
where the primes stand for derivatives with respect to x. We point out that the spectral
parameter ω occurs in Eq. (4) in a complicated way.
◮ Two types of orthogonality of eigenfunctions.

Let ωm and ωn be two distinct eigenvalues of the Sturm–Liouville problem (4)–(6), and let
ϕm (x) and ϕn (x) be the corresponding eigenfunctions.
1◦ . First type of orthogonality of eigenfunctions:
Z l

a(x)ϕm (x)ϕn (x) + c(x)ϕ′m (x)ϕ′n (x) dx = 0, m 6= n.
0
The corresponding squared norm of the eigenfunction is given by
Z l
2

kϕn k1 = a(x)ϕ2n (x) + c(x)ϕ′n2 (x) dx.
0
2◦ . Second type of orthogonality of eigenfunctions:
Z l

b(x)ϕ′m (x)ϕ′n (x) + d(x)ϕ′′m (x)ϕ′′n (x)] dx = 0, m 6= n.
0
The corresponding squared norm of the eigenfunction is given by
Z l
2

kϕn k2 = b(x)[ϕ′n (x)]2 + d(x)[ϕ′′n (x)]2 dx = ωn2 kϕn k21 .
0
◮ Representation of the solution of a general initial-boundary value problem via the

Green’s function.
1◦ . The solution of the problem on the longitudinal vibrations of a Rayleigh–Bishop bar
with the general initial conditions (1) and the homogeneous boundary conditions (2) can be
represented in the form
Z l
∂ ′ ∂
w(x, t) = a(y)f (y)G(x, y, t) + c(y)f (y) G(x, y, t) dy
∂t 0 ∂y
Z l
′ ∂
+ a(y)g(y)G(x, y, t) + c(y)g (y) G(x, y, t) dy (7)
0 ∂y
Z tZ l
+ Φ(y, τ )G(x, y, t − τ ) dy dτ,
0 0
where G(x, y, t) is the Green’s function
∞
X 1
G(x, y, t) = ϕn (x)ϕn (y) sin(ωn t). (8)
n=1
ωn kϕn k21
◮ Homogeneous cylindrical bar of constant cross-section.

Let us look at longitudinal vibrations of a homogeneous cylindrical bar of constant cross-
section. In this case, the coefficients of the original equation are constants, and the general
solution of the corresponding Sturm–Liouville problem (4)–(6) results in the following
eigenvalues and eigenfunctions:
s
πn bl2 + d(πn)2 πnx
ωn = , ϕn (x) = sin , n = 1, 2, . . . (9)
l al2 + c(πn)2 l
Formulas (7)–(9) permit one to obtain the solution of the problem on the longitudinal vi-
brations of a cylindrical Rayleigh–Bishop bar with the general initial conditions (1) and
homogeneous boundary conditions (2).
⊙ Literature for equation 11.2.7.3: I. A. Fedotov, A. D. Polyanin, M. Yu. Shatalov, and H. M. Tenkam (2010).
◮ Equations containing the fourth derivative with respect to t.
∂4w ∂4w ∂4w

4. −2 + = 0.
∂t4 ∂t2 ∂x2 ∂x4
General solution (three representations):

w(x, t) = f1 (t − x) + f2 (t + x) + t g1 (t − x) + g2 (t + x) ,

w(x, t) = f1 (t − x) + f2 (t + x) + x g1 (t − x) + g2 (t + x) ,
w(x, t) = f1 (t − x) + f2 (t + x) + (t + x)g1 (t − x) + (t − x)g2 (t + x),
where f1 (y), f2 (z), g1 (y), and g2 (z) are arbitrary functions.


∂2 ∂2 ∂2 ∂2
5. − a2 − b2 w = 0.
∂t2 ∂x2 ∂t2 ∂x2
1◦ . General solution with a 6= b:
w = f1 (x − at) + f2 (x + at) + f3 (x − bt) + f4 (x + bt), (1)
where f1 (z1 ), f2 (z2 ), f3 (z3 ), and f4 (z4 ) are arbitrary functions.

2◦ . General solution with a = b:
w = f1 (x − at) + f2 (x + at) + tf3 (x − at) + tf4 (x + at) (2)
where f1 (z1 ), f2 (z2 ), f3 (z3 ), and f4 (z4 ) are arbitrary functions.

w(x, 0) = ϕ0 (x), wt (x, 0) = ϕ1 (x), wtt (x, 0) = ϕ2 (x), wttt (x, 0) = ϕ3 (x). (3)
For a 6= b, the solution of the problem is described by formula (1), where

1 2 b2 1
f1 (x) = −b ϕ0 (x) + Φ1 (x) + Φ2 (x) − Φ3 (x) ,
2(a2 − b2 ) a 2a

1 2 b2 1
f2 (x) = −b ϕ0 (x) − Φ1 (x) + Φ2 (x) + Φ3 (x) ,
2(a2 − b2 ) a 2a

1 2 a2 1
f3 (x) = −a ϕ0 (x) + Φ1 (x) + Φ2 (x) − Φ3 (x) ,
2(b2 − a2 ) b 2b
2
1 2 a 1
f4 (x) = −a ϕ0 (x) − Φ 1 (x) + Φ 2 (x) + Φ 3 (x) ,
2(b2 − a2 ) b 2b
Z x
Φn (x) = ϕn (ξ)(x − ξ)n−1 dξ, n = 1, 2, 3,
x0
and x0 is an arbitrary constant.

For a = b, the solution of the problem is described by formula (2), where
Z x Z x
1 3 1
f1 (x) = ϕ0 (x) − ϕ1 (ξ) dξ + 3 ϕ3 (ξ)(x − ξ)2 dξ,
2 4a x0 8a x0
Z x Z x
1 3 1
f2 (x) = ϕ0 (x) + ϕ1 (ξ) dξ − 3 ϕ3 (ξ)(x − ξ)2 dξ,
2 4a x0 8a x0
Z x Z x
1 ′ 1 1 1
f3 (x) = aϕ0 (x) − ϕ1 (x) − ϕ2 (ξ) dξ + 2 ϕ3 (ξ)(x − ξ) dξ,
4 4 4a x0 4a x0
Z x Z x
1 1 1 1
f4 (x) = − aϕ′0 (x) − ϕ1 (x) + ϕ2 (ξ) dξ + 2 ϕ3 (ξ)(x − ξ) dξ.
4 4 4a x0 4a x0
11.3 Two-Dimensional Nonstationary Fourth-Order

Equations

∂w 2 ∂4w ∂4w
11.3.1 Equation of the Form +a + = Φ(x, y, t)
∂t ∂x4 ∂y 4
In this section we consider boundary value problems in a rectangular domain 0 ≤ x ≤ l1 ,
0 ≤ y ≤ l2 with the general initial condition
w = f (x, y) at t=0 (1)
Z l1Z l2
w(x, y, t) = f (ξ, η)G(x, y, ξ, η, t) dη dξ
0 0
Z t Z l1Z l2 (2)
+ Φ(ξ, η, τ )G(x, y, ξ, η, t − τ ) dη dξ dτ.
0 0 0
11.3. Two-Dimensional Nonstationary Fourth-Order Equations 987
◮ Green’s functions for various boundary value problems (0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 ).

1◦ . The function and its first derivatives are prescribed on the sides of the rectangle:
w = ∂x w = 0 at x = 0, w = ∂x w = 0 at x = l1 ,
w = ∂y w = 0 at y = 0, w = ∂y w = 0 at y = l2 .
Green’s function:
∞ ∞
16 X X p4n qm
4 ϕ (x)ψ (y)ϕ (ξ)ψ (η) exp −(p4 + q 4 )a2 t
n m n m n m
G(x, y, ξ, η, t) = ,
l1 l2 ′′ (l ) 2
ϕ′′n (l1 )ψm
n=1 m=1 2
d2 ϕn d2 ψm
ϕ′′n (x) = , ′′
ψm (y) = .
dx2 dy 2
Here
ϕn (x) = sinh(pn l1 ) − sin(pn l1 ) cosh(pn x) − cos(pn x)

− cosh(pn l1 ) − cos(pn l1 ) sinh(pn x) − sin(pn x) ,

ψm (y) = sinh(qm l2 ) − sin(qm l2 ) cosh(qm y) − cos(qm y)

− cosh(qm l2 ) − cos(qm l2 ) sinh(qm y) − sin(qm y) ,
where the pn and qm are positive roots of the transcendental equations
cosh(pl1 ) cos(pl1 ) = 1, cosh(ql2 ) cos(ql2 ) = 1 (qm = pm l1 /l2 ).
2◦ . The function and its second derivatives are prescribed on the sides of the rectangle:
w = ∂xx w = 0 at x = 0, w = ∂xx w = 0 at x = l1 ,
w = ∂yy w = 0 at y = 0, w = ∂yy w = 0 at y = l2 .
Green’s function:
∞ ∞
4 XX
G(x, y, ξ, η, t) = sin(pn x) sin(qm y) sin(pn ξ) sin(qm η) exp −(p4n +qm
4
)a2 t ,
l1 l2
n=1 m=1
πn πm
pn = , qm = .
l1 l2
3◦ . The first and third derivatives are prescribed on the sides of the rectangle:
wx = ∂xxx w = 0 at x = 0, wx = ∂xxx w = 0 at x = l1 ,
wy = ∂yyy w = 0 at y = 0, wy = ∂yyy w = 0 at y = l2 .
Green’s function:
∞ ∞
1 XX
G(x, y, ξ, η, t) = εn εm cos(pn x) sin(qm y)
l1 l2 n=0 m=0

× cos(pn ξ) cos(qm η) exp −(p4n + qm 4
)a2 t ,
(
πn πm 1 for n = 0,
pn = , qm = , εn =
l1 l2 2 for n 6= 0.
4◦ . The second and third derivatives are prescribed on the sides of the rectangle:
wxx = ∂xxx w = 0 at x = 0, wxx = ∂xxx w = 0 at x = l1 ,

wyy = ∂yyy w = 0 at y = 0, wyy = ∂yyy w = 0 at y = l2 .
Green’s function:
G(x, y, ξ, η, t) = G1 (x, ξ, t)G2 (y, η, t),

∞
1 3 4 X ϕn (x)ϕn (ξ)
G1 (x, ξ, t) = + 3 (2x − l1 )(2ξ − l1 ) + exp(−p4n a2 t),
l1 l1 l1 n=1 ϕ2n (l1 )
∞
1 3 4 X ψm (y)ψm (η) 4 2
G2 (y, η, t) = + 3 (2y − l2 )(2η − l2 ) + 2 (l )
exp(−qm a t).
l2 l2 l2 ψm 2
m=1
Here
ϕn (x) = sinh(pn l1 ) − sin(pn l1 ) cosh(pn x) + cos(pn x)

− cosh(pn l1 ) − cos(pn l1 ) sinh(pn x) + sin(pn x) ,

ψm (y) = sinh(qm l2 ) − sin(qm l2 ) cosh(qm y) + cos(qm y)

− cosh(qm l2 ) − cos(qm l2 ) sinh(qm y) + sin(qm y) ,
cosh(pl1 ) cos(pl1 ) = 1, cosh(ql2 ) cos(ql2 ) = 1.
5◦ . Mixed boundary conditions are prescribed on the sides of the rectangle:
w = ∂xx w = 0 at x = 0, ∂x w = ∂xxx w = 0 at x = l1 ,
w = ∂yy w = 0 at y = 0, ∂y w = ∂yyy w = 0 at y = l2 .
Green’s function:
∞ ∞
4 XX
G(x, y, ξ, η, t) = sin(pn x) sin(qm y) sin(pn ξ) sin(qm η) exp −(p4n +qm
4
)a2 t ,
l1 l2
n=0 m=0
π(2n + 1) π(2m + 1)
pn = , qm = .
2l1 2l2
∂2w
11.3.2 Equation of the Form + a2 ∆∆w = 0
∂t2
◮ Preliminary remarks. Particular solutions.
1◦ . This equation describes two-dimensional free transverse vibrations of a thin elastic
plate; the unknown w is the deflection (transverse displacement) of the plate midplane
points relative to the original plane position. Here ∆∆ = ∆2 and ∆ is the Laplace operator
defined as
( 2
∂ ∂2
∂x2
+ ∂y 2 in the Cartesian coordinate system,
∆ = ∂2 1 ∂ 1 ∂2
∂r 2
+ r ∂r + r2 ∂ϕ2 in the polar coordinate system.

w(x, y, t) = A1 sin(k1 x) + B1 cos(k1 x) A2 sin(k2 y) + B2 cos(k2 y) sin (k12 + k22 )at ,

w(x, y, t) = A1 sin(k1 x) + B1 cos(k1 x) A2 sin(k2 y) + B2 cos(k2 y) cos (k12 + k22 )at ,

w(x, y, t) = A1 sinh(k1 x) + B1 cosh(k1 x) sinh(k2 y) sin (k12 + k22 )at ,

w(x, y, t) = A1 sinh(k1 x) + B1 cosh(k1 x) cosh(k2 y) sin (k12 + k22 )at ,

w(x, y, t) = A1 sinh(k1 x) + B1 cosh(k1 x) sinh(k2 y) cos (k12 + k22 )at ,

w(x, y, t) = A1 sinh(k1 x) + B1 cosh(k1 x) cosh(k2 y) cos (k12 + k22 )at ,

w(r, ϕ, t) = A1 Jn (kr) + A2 Yn (kr) + A3 In (kr) + A4 Kn (kr) cos(nϕ) sin(k 2 at),

w(r, ϕ, t) = A1 Jn (kr) + A2 Yn (kr) + A3 In (kr) + A4 Kn (kr) sin(nϕ) cos(k 2 at),
where A1 , A2 , A3 , A4 , B1 , B2 , k, k1 , k2 are arbitrary constants, Jn (ξ) and Yn (ξ) are Bessel

functions of the first and second p kind, In (ξ) and Kn (ξ) are modified Bessel functions of
the first and second kind, r = x2 + y 2 , and n = 0, 1, 2, . . .
◮ Cauchy problem.
1◦ . Domain: −∞ < x < ∞, −∞ < y < ∞. Cauchy problem.
w = f (x, y) at t = 0, ∂t w = g(x, y) at t = 0.
Poisson solution:
Z Z
1 ∞ ∞ √ √
w(x, y, t) = f x + 2ξ at, y + 2η at sin ξ 2 + η 2 dξ dη
π −∞ −∞
Z Z ∞Z ∞
1 t √ √
+ dτ g x + 2ξ aτ , y + 2η aτ sin ξ 2 + η 2 dξ dη.
π 0 −∞ −∞
Green’s function:
Z t
1 (x − ξ)2 + (y − η)2 dτ
G(x, y, ξ, η, t) = sin .
4πa 0 4aτ τ
⊙ Literature: A. N. Krylov (1949), I. Sneddon (1951), B. M. Budak, A. A. Samarskii, and A. N. Tikhonov
(1980).
2◦ . Domain: 0 ≤ r < ∞, 0 ≤ ϕ ≤ 2π. Cauchy problem.

Initial conditions for the symmetric case in the polar coordinate system:
w = f (r) at t = 0, ∂t w = 0 at t = 0.
Solution:
Z ∞
1 ξr ξ 2 + r2
w(r, t) = ξf (ξ)J0 sin dξ,
2at 0 2at 4at
where J0 (z) is the zeroth Bessel function.
⊙ Literature: I. Sneddon (1951), B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).

Consider boundary value problems in the rectangular domain 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 with
the general initial conditions
w = f (x, y) at t = 0, ∂t w = g(x, y) at t = 0 (1)
Z l1Z l2
∂
w(x, y, t) = f (ξ, η)G(x, y, ξ, η, t) dη dξ
∂t 0 0
Z l1Z l2 (2)
+ g(ξ, η)G(x, y, ξ, η, t) dη dξ.
0 0

conditions in Cartesian coordinates. The solutions of these boundary value problems are
represented via the Green’s functions by formula (2).
1◦ . Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 . All sides of the plate are hinged.
Green’s function:
∞ ∞
4 XX sin(λnm at)
G(x, y, ξ, η, t) = sin(pn x) sin(qm y) sin(pn ξ) sin(qm η) ,
al1 l2 n=1 m=1 λnm
πn πm
pn = , qm = , λnm = p2n + qm 2
.
l1 l2
2◦ . Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 . The 1st and 3rd derivatives are prescribed on the
sides:
∂x w = ∂xxx w = 0 at x = 0, ∂x w = ∂xxx w = 0 at x = l1 ,
∂y w = ∂yyy w = 0 at y = 0, ∂y w = ∂yyy w = 0 at y = l2 .
Green’s function:
∞ ∞
1 XX sin(λnm at)
G(x, y, ξ, η, t) = εn εm cos(pn x) cos(qm y) cos(pn ξ) cos(qm η) ,
al1 l2 n=0 m=0 λnm
(
πn πm 1 for n = 0,
pn = , qm = , λnm = p2n + qm 2
, εn =
l1 l2 2 for n 6= 0.
For n = m = 0, the ratio sin(λnm at)/λnm must be replaced by at.

3◦ . Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 . Mixed boundary conditions are set on the sides:
Green’s function:
∞ ∞
2 XX sin(λnm at)
G(x, y, ξ, η, t) = εm sin(pn x) cos(qm y) sin(pn ξ) cos(qm η) ,
al1 l2 λnm
n=1 m=0
(
πn πm 1 for m = 0,
pn = , qm = , λnm = p2n + qm 2
, εm =
l1 l2 2 for m 6= 0.
◮ Solution of the boundary value problem on transverse vibrations of a circular

plate.
Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π. Initial and boundary conditions for symmetric transverse
vibrations of a circular plate of radius R with clamped contour in the polar coordinate
system:
w = f (r) at t = 0, ∂t w = g(t) at t = 0;
w=0 at r = R, ∂r w = 0 at r = R.
Solution:
∞
X
w(r, t) = An cos(akn2 t) + Bn sin(akn2 t) Ψn (r),
n=1
Ψn (r) = I0 (kn R)J0 (kn r) − J0 (kn R)I0 (kn r),
where the kn are positive roots of the transcendental equation (the prime denotes the deriva-
tive)
J0 (kR)I0′ (kR) − I0 (kR)J0′ (kR) = 0,
and the coefficients An and Bn are given by
Z R Z R
1 1
An = f (r)Ψn (r)r dr, Bn = g(r)Ψn (r)r dr,
kΨn k2 0 akn2 kΨn k2 0
2
kΨn k2 = 14 R6 Ψ′′n (R) = R2 J02 (kn R)I02 (kn R).
∂2w
11.3.3 Equation of the Form + a2 ∆∆w + kw = Φ(x, y, t)
∂t2

Z l1 Z l2 Z l1 Z l2
∂
w= f (ξ, η)G(x, y, ξ, η, t) dη dξ + g(ξ, η)G(x, y, ξ, η, t) dη dξ
∂t 0 0 0 0
Z t Z l1 Z l2
+ Φ(ξ, η, τ )G(x, y, z, ξ, η, ζ, t − τ ) dη dξ dτ. (2)
0 0 0

Green’s function:
∞ ∞
4 XX sin(λnm t)
l1 l2 λnm
n=1 m=1
πn πm p
pn = , qm = , λnm = a2 (p2n + qm 2 )2 + k.
l1 l2
Green’s function:
∞ ∞
1 XX sin(λnm t)
l1 l2 λnm
n=0 m=0
(
pn = , qm = , λnm = a2 (p2n + qm 2 )2 + k, εn =
l1 l2 2 for n 6= 0.
3◦ . Mixed boundary conditions are prescribed on the sides of the rectangle:
Green’s function:
∞ ∞
2 XX sin(λnm t)
G(x, y, ξ, η, t) = εm sin(pn x) cos(qm y) sin(pn ξ) cos(qm η) ,
l1 l2 λnm
n=1 m=0
(
πn πm p 1 for m = 0,
pn = , qm = , λnm = a2 (p2n + qm 2 )2 + k, εm =
l1 l2 2 for m 6= 0.
∂2w
∂4w ∂4w

2
11.3.4 Equation of the Form +a + + kw = Φ(x, y, t)
∂t2 ∂x4 ∂y 4
Z l1 Z l2 Z l1 Z l2
∂
w= f (ξ, η)G(x, y, ξ, η, t) dη dξ + g(ξ, η)G(x, y, ξ, η, t) dη dξ
∂t 0 0 0 0
Z t Z l1 Z l2
+ Φ(ξ, η, τ )G(x, y, z, ξ, η, ζ, t − τ ) dη dξ dτ. (2)
0 0 0

w = ∂x w = 0 at x = 0, w = ∂x w = 0 at x = l1 ,
w = ∂y w = 0 at y = 0, w = ∂y w = 0 at y = l2 .
Green’s function:
∞ ∞
16 X X p4n qm
4 sin(λnm t)
G(x, y, ξ, η, t) = 2 ϕn (x)ψm (y)ϕn (ξ)ψm (η) ,
l1 l2 ϕ′′ (l )ψ ′′ (l )
n 1 m 2
λnm
n=1 m=1
p d2 ϕn d2 ψm
λnm = a2 (p4n + qm4 ) + k, ϕ′′n (x) = , ψm
′′
(y) = .
dx2 dy 2
Here



Green’s function:
∞ ∞
4 XX sin(λnm t)
l1 l2 λnm
n=1 m=1
πn πm p
pn = , qm = , λnm = a2 (p4n + qm 4 ) + k.
l1 l2
Green’s function:
∞ ∞
1 XX sin(λnm t)
l1 l2 λnm
n=0 m=0
(
pn = , qm = 2 4 4
, λnm = a (pn + qm ) + k, εn =
l1 l2 2 for n 6= 0.
11.3.5 Other Two-Dimensional Nonstationary Fourth-Order

Equations

∂2 ∂ 2w ∂2w ∂2w ∂ 2w
1. + + a2 + b2 = 0.
∂t2 ∂x2 ∂y 2 ∂x2 ∂y 2
Two-dimensional equation of gravitational–gyroscopic waves in the Boussinesq approxi-
mation. In what follows, we assume that a ≥ 0 and b ≥ 0.
s
a2 k12 + b2 k22
w = (c1 e−k1 x + c2 ek1 x )(c3 e−k2 y + c4 ek2 y ) cos(λt + c5 ), λ= ;
k12 + k22
s
−k1 x k1 x a2 k12 − b2 k22
w = (c1 e + c2 e ) cos(k2 y + c3 ) cos(λt + c4 ), λ= ;
k12 − k22
s
a2 k12 − b2 k22
w = cos(k1 x + c1 )(c2 e−k2 y + c3 ek2 y ) cos(λt + c4 ), λ= ;
k12 − k22
s
a2 k12 + b2 k22
w = cos(k1 x + c1 ) cos(k2 y + c2 ) cos(λt + c3 ), λ= ;
k12 + k22
s
b2 k22 − a2 k12
w = (c1 e−k1 x + c2 ek1 x ) cos(k2 y + c3 )(c4 e−λt + c5 eλt ), λ= ;
k12 − k22
s
−k2 y k2 y −λt λt a2 k12 − b2 k22
w = cos(k1 x + c1 )(c2 e + c3 e )(c4 e + c5 e ), λ= ,
k22 − k12
where c1 , c2 , c3 , k1 , and k2 are arbitrary constants and the radicands are assumed to be
positive.
2◦ . Singular particular solution:

Z t
1 1 |r|∗
w2 (r, t) = t ln |r| + ln t + C − 1 − Ci ξ dξ,
2π 2π 0 |r|
Z ∞ Z ξ (1)
cos z cos z − 1
Ci(ξ) = − dz = C + ln ξ + dz,
ξ z 0 z
where |r| = (x2 + y 2 )1/2 , |r|∗ = (b2 x2 + a2 y 2 )1/2 , Ci(ξ) is the cosine integral, and C =
0.5772 . . . is the Euler constant.
The singular solution (1) satisfies the initial conditions
∂w2 1
w2 = 0, = ln |r| at t = 0. (2)
∂t 2π
Ee (r, t) = (I − aJat ∗)(I − bJbt ∗)w2 (r, t),
Z t
(3)
(I − aJat ∗)f (t) ≡ f (t) − a J1 (a(t − τ ))f (τ ) dτ,
0
where I is the identity operator and J1 (τ ) is the Bessel function.
The fundamental solution satisfies the initial conditions
∂Ee 1
Ee = 0, =− ln |r| at t = 0.
∂t 2π
4◦ . Set λ = λ(r) = |r|∗ /|r|. For |x| ≥ δ > 0, |y| ≥ δ > 0, and a > b > 0, the following
asymptotic formula holds for the fundamental solution as t → ∞:
s
1 2 sin(at − π/4) sin(bt − π/4)
Ee (r, t) = √ + √ ln |r|
2π π|a2 − b2 | at bt
s
1 1 sin(at − π/4) λ2 (r) sin(bt − π/4) λ2 (r)
+ √ ln1 − + √ ln1 −
2π 2π|a2 − b2 | at a2 bt b2
r
1 π cos(bt − π/4) 1 sin(λ(r)t)
− 2 2
√ + − p p ,
2π 2|a − b | bt 2πabt 2πt a − λ2 (r) b2 − λ2 (r)
2
where terms of the order of t−3/2 are dropped. For b > a > 0, one should exchange a and
b in the asymptotic formula.
5◦ . Cauchy problem (t ≥ 0, r ∈ R2 ). Initial conditions are prescribed:
w = f (r) at t = 0,
∂t w = g(r) at t = 0.
Solution:
Z Z
∂ ′
w(r, t) = Ee (r − r , t)∆′2 f (r′ ) dV ′ + Ee (r − r′ , t)∆′2 g(r′ ) dV ′ ,
∂t R2 R2
where ∆′2 is the two-dimensional Laplace operator in the integration variables (x′ , y ′ ) and
dV ′ = dx′ dy ′ .

∂2 ∂ 2w ∂2w ∂2w ∂ 2w
2. + + a2 + b2 = Φ(x, y, t).
∂t2 ∂x2 ∂y 2 ∂x2 ∂y 2
For Φ(x, y, t) = 0, see equation 11.3.5.1.
Cauchy problem (t ≥ 0, r ∈ R2 ). Initial conditions:
w = f (r) at t = 0,
∂t w = g(r) at t = 0,
where r = (x, y).

Solution:
Z Z
∂ ′ ′ ′ ′
w(r, t) = Ee (r − r , t)∆2 f (r ) dV + Ee (r − r′ , t)∆′2 g(r′ ) dV ′
∂t R2 R2
Z tZ
+ Ee (r − r′ , t − τ )Φ(r′ , τ ) dV ′ dτ,
0 R2
where E (r, t) is the fundamental solution given in Item 3◦ of equation 11.3.5.1, ∆′2 is the
two-dimensional Laplace operator in the integration variables (x′ , y ′ ), and dV ′ = dx′ dy ′ .

∂2 ∂ 2w ∂2w ∂ 2w ∂2w
3. + − + = 0.
∂t2 ∂x2 ∂y 2 ∂t2 ∂x2
This equation describes interior waves propagating in an infinite flat channel bounded by
horizontal rigid walls and filled with an exponentially stratified fluid.
α
w = (A1 e−αx + A2 eαx )(B1 e−βy + B2 eβy )(C1 e−λt + C2 eλt ), λ= p ,
1 − α2 − β 2
α
w = (A1 e−αx + A2 eαx )(B1 e−βy + B2 eβy ) cos(λt + C1 ), λ = p ,
α + β2 − 1
2
α
w = (A1 e−αx + A2 eαx ) cos(βy + C1 ) cos(λt + C2 ), λ = p ,
α − β2 − 1
2
α
w = cos(αx + C1 )(B1 e−βy + B2 eβy ) cos(λt + C2 ), λ = p ,
1 + α2 − β 2
α
w = cos(αx + C1 ) cos(βy + C2 ) cos(λt + C3 ), λ = p ,
1 + α2 + β 2
where An , Bn , Cn , α, and β are arbitrary constants.

2◦ . Domain: S = {−∞ < x < ∞, 0 ≤ y ≤ π}. First boundary value problem.
w = f0 (x, y) at t = 0,
∂t w = f1 (x, y) at t = 0,
w=0 at y = 0,
w=0 at y = π.
11.4. Three- and n-Dimensional Nonstationary Fourth-Order Equations 997
Solution:
Z Z
∂ ′ ′ ′
w(r, t) = G(r, r , t)M[f0 (r )] dV + G(r, r′ , t)M[f1 (r′ )] dV ′ ,
∂t S′ S′
where
∂2f ∂2f
r = (x, y), r′ = (x′ , y ′ ), M[f (r′ )] ≡
+ − f, dV ′ = dx′ dy ′ ,
∂x′2 ∂y ′2
√
X∞
sin(ny) sin(ny ′ ) Z ∞ exp iµ|x − x′ | 1 + n2 tµ
′
G(r, r , t) = − √ p sin p dµ.
n=1
π 2 1 + n2 −∞ µ 1 + µ2 1 + µ2
11.4 Three- and n-Dimensional Nonstationary

Fourth-Order Equations
∂2w
11.4.1 Equation of the Form + a2 ∆∆w = 0
∂t2
◮ Three-dimensional case.
1◦ . Domain: −∞ < x < ∞, −∞ < y < ∞, −∞ < z < ∞. Cauchy problem.
w = f (x, y, z) at t = 0, ∂t w = 0 at t = 0.
Solution:
Z ∞Z ∞Z ∞
1
w(x, y, z, t) = √ 3 f (x + ξ, y + η, z + ζ)
2 πat −∞ −∞ −∞
2
ξ + η2 + ζ 2 3π
× cos − dξ dη dζ.
4at 4
2◦ . Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 , 0 ≤ z ≤ l3 . Boundary value problem.

Initial conditions:
w = f (x, y, z) at t = 0, ∂t w = g(x, y, z) at t = 0.
Boundary conditions:
w = ∂yy w = 0 at y = 0, w = ∂yy w = 0 at y = l2 ,
w = ∂zz w = 0 at z = 0, w = ∂zz w = 0 at z = l3 .
Solution:
Z l1Z l2Z l3
∂
w(x, y, z, t) = f (ξ, η, ζ)G(x, y, z, ξ, η, ζ, t) dζ dη dξ
∂t 0 0 0
Z l1Z l2Z l3
+ g(ξ, η, ζ)G(x, y, z, ξ, η, ζ, t) dζ dη dξ,
0 0 0
where
∞ ∞ ∞
8 XXX 1
G(x, y, z, ξ, η, ζ, t) = sin(pn x) sin(qm y) sin(sk z)
al1 l2 l3 λnmk
n=1 m=1 k=1
× sin(pn ξ) sin(qm η) sin(sk ζ) sin(λnmk at),
πn πm πk
pn = , qm = , sk = , λnmk = p2n + qm 2
+ s2k .
l1 l2 l3
◮ n-dimensional case.
1◦ . Domain: Rn = {−∞ < xk < ∞; k = 1, . . . , n}. Cauchy problem.
w = f (x) at t = 0, ∂t w = 0 at t = 0,
where x = {x1 , . . . , xn }.
Solution:
Z
1 |x − y| πn
w(x, t) = √ n f (y) cos − dVy ,
2 πat Rn 4at 4
where y = {y1 , . . . , yn } and dVy = dy1 . . . dyn .
2◦ . Domain: V = {0 ≤ xk ≤ lk ; k = 1, 2, . . . n}. Boundary value problem.

Initial conditions:
w = f (x) at t = 0, ∂t w = g(x) at t = 0.
Boundary conditions:
w = ∂xk xk w = 0 at xk = 0, w = ∂xk xk w = 0 at xk = lk .
Solution:
Z Z
∂
w(x, t) = f (y)G(x, y, t) dVy + g(y)G(x, y, t) dVy ,
∂t V V
where
∞ X
X ∞ ∞
X
2n 1
G(x, y, t) = ... sin(pk1 x1 ) sin(pk2 x2 ) . . . sin(pkn xn )
al1 l2 . . . ln λk
k1 =1 k2 =1 kn =1

× sin(pk1 y1 ) sin(pk2 y2 ) . . . sin(pkn yn ) sin λk at ,
πk1 πk2 πkn
pk1 = , pk2 = , . . . , pkn = , λk = p2k1 + p2k2 + · · · + p2kn .
l1 l2 ln
11.4.2 Equations Containing Mixed Derivatives

∂2w ∂2
1. − a∆w − b ∆w = 0.
∂t2 ∂t2
Let a > 0 and b > 0. Consider a bounded open domain V with boundary S. The following

∂w
αw + β =0 at x ∈ S (boundary condition),
∂n
Solution:
∞
X
w(x, t) = Ak ψk1 (t) + Bk ψk2 (t) uk (x),
k=1
where
r r r
1 + bλn aλn aλn
ψk1 (t) = sin t , ψk2 (t) = cos t ,
aλn 1 + bλn 1 + bλn
Z Z
1 1
Ak = f (x)uk (x) dv, Bk = g(x)uk (x) dv,
kuk k2 V kuk k2 V
and λk and uk (x) are the eigenvalues and eigenfunctions of the auxiliary problem
∆u + λu = 0, x ∈ V,
∂u
αu + β = 0, x ∈ S.
∂n
For the main properties of the eigenvalues λk and eigenfunctions uk (x), see Item 4◦ of
equation 11.1.5.1.

∂
2. ∆ − ν∆ w = Φ(x, t).
∂t
Here x = (x, y, z) and ∆ is the three-dimensional Laplace operator. Equations of this kind
occur in the hydrodynamics of a viscous incompressible fluid (see Section 12.7.3).
1◦ . The general solution of the homogeneous equation with Φ ≡ 0:
w = w1 + w2 ,
where w1 and w2 are arbitrary solutions of the Laplace and the heat equations
∂w2
∆w1 = 0, − ν∆w2 = 0.
∂t
2◦ . A particular solution of the nonhomogeneous equation for an arbitrary domain V∗ lying

inside the rectangular parallelepiped V = {0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ≤ z ≤ c} can be
represented in the form of the series
∞ X
X ∞ X
∞
wp (x, t) = ϕnmk (t)unmk (x),
n=1 m=1 k=1
where
Z t
1
ϕnmk (t) = − exp −νλnmk (t − τ ) Anmk (τ ) dτ,
λnmk 0

πnx πmy πkz
unmk (x) = sin sin sin ,
a b c
2 Z
2 n m2 k2 8
λnmk =π + 2 + 2 , Anmk (t) = f (x, t)unmk (x) dV.
a2 b c abc V

∂2 ∂
3. ∆ σ 2 + − ν∆ w = Φ(x, t).
∂t ∂t
Here x = (x, y, z) and ∆ is the three-dimensional Laplace operator. Equations of this kind
arise when decomposing the equations of a viscoelastic incompressible Maxwell fluid (see
Section 12.9.3).
w = w1 + w2 ,
where w1 and w2 are arbitrary solutions of the Laplace and the telegraph equations
∂ 2 w2 ∂w2
∆w1 = 0, σ 2
+ − ν∆w2 = 0.
∂t ∂t
2◦ . A particular solution of the nonhomogeneous equation for an arbitrary domain V∗ lying

inside the rectangular parallelepiped V = {0 ≤ x ≤ a, 0 ≤ y ≤ b, 0 ≤ z ≤ c} can be
represented in the form of the series
∞ X
X ∞ X
∞
n=1 m=1 k=1
where
 Z t

 2 1 snmk

 − Anmk (τ ) exp − (t − τ ) sin (t − τ ) dτ

 snmk λnmk 0 2σ 2σ



 if 4νσλnmk − 1 = s2nmk > 0,

 Z

 2 t
1 snmk
− Anmk (τ ) exp − (t − τ ) sinh (t − τ ) dτ
ϕnmk (t) = snmk λnmk 0 2σ 2σ



 if 4νσλnmk − 1 = −s2nmk < 0,

 Z t

 1 1

 − Anmk (τ ) exp − (t − τ ) (t − τ ) dτ

 σλnmk 0 2σ


 if 4νσλnmk − 1 = 0,

πnx πmy πkz
a b c
2 Z
2 n m2 k2 8
λnmk = π + 2 + 2 , Anmk (t) = f (x, t)unmk (x) dV.
a2 b c abc V
2
∂2 ∂
4. − c21 ∆ − c 2
2 ∆ w = Φ(x, t).
∂t2 2 ∂t
Here x = (x, y, z) and ∆ is the three-dimensional Laplace operator. Similar equations
occur in elasticity theory (see Section 12.6.3).
w = w1 + w2 ,
where w1 and w2 are arbitrary solutions of two independent wave equations
∂ 2 w1 ∂ 2 w2
− c21 ∆w1 = 0, − c22 ∆w2 = 0.
∂t2 ∂t2
2◦ . If c21 6= c22 , then a particular solution of the nonhomogeneous equation for an arbitrary
domain V∗ lying inside the rectangular parallelepiped V = {0 ≤ x ≤ a1 , 0 ≤ y ≤ a2 , 0 ≤
z ≤ a3 } can be represented in the form of the series
∞ X
X ∞ X
∞
n=1 m=1 k=1
where
Z t
1/2 1/2
c1 sin c2 λnmk (t − τ ) − c2 sin c1 λnmk (t − τ )
ϕnmk (t) = 3/2
Anmk (τ ) dτ,
0 c1 c2 λnmk (c21 − c22 )

πnx πmy πkz
a1 a2 a3
2 Z
2 n m2 k2 8
λnmk =π + 2 + 2 , Anmk (t) = f (x, t)unmk (x) dV.
a21 a2 a3 a1 a2 a3 V

∂2 ∂ 2w ∂2w ∂2w ∂ 2w ∂2w ∂2w
5. + + + a2 + + b2 = 0.
∂t2 ∂x2 ∂y 2 ∂z 2 ∂x2 ∂y 2 ∂z 2
Equation of gravitational–gyroscopic waves in the Boussinesq approximation. In the spe-
cial case of a = 0, it is known as the Sobolev equation and describes small vibrations of a
homogeneous inviscid rotating fluid. It is assumed in what follows that a ≥ 0 and b ≥ 0.
w = exp(k1 x + k2 y + k3 z)[c1 cos(λt) + c2 sin(λt)],
w = cos(k1 x + c1 ) cos(k2 y + c2 ) cos(k3 z + c3 ) cos(λt + c4 ),
where c1 , c2 , c3 , k1 , k2 , and k3 are arbitrary constants, and
s
a2 (k12 + k22 ) + b2 k32
λ= .
k12 + k22 + k32
2◦ . Singular particular solution:
Z t|r|∗ /|r|
1
w3 (r, t) = − J0 (ξ) dξ,
4π|r|∗ 0 (1)
2 2 2 1/2 2 2 2 2 2 1/2
|r| = (x + y + z ) , |r|∗ = [b (x + y ) + a z ] ,
where J0 (ξ) is the Bessel function. Note that min{a, b} ≤ |r|∗ /|r| ≤ max{a, b}.
The singular solution (1) satisfies the initial conditions
∂w3 1
w3 = 0, =− at t = 0. (2)
∂t 4π|r|
Z t
1 |r|∗
Ee (r, t) = − J0 (a(t − τ ))J0 τ dτ. (3)
4π|r| 0 |r|
The fundamental solution satisfies the initial conditions (2) with w3 replaced by Ee .
On the z-axis, which corresponds to x = y = 0 and |r|∗ /|r| = a, formula (3) implies
that
sin(at)
Ee (r, t)x=y=0 = − .
4πa|z|
4◦ . Let a 6= b. Set λ = λ(r) = |r|∗ /|r|. An alternative representation for the fundamental
solution with λ 6= a is given by
Z 1
1 sin(atµ) dµ
Ee (r, t) = − 2 p p .
2π a|r| λ/a 1 − µ2 µ2 − (λ/a)2
Let |λ − a| ≥ δ > 0, where δ is a fixed sufficiently small number. (Thus, a narrow
domain near the z-axis is eliminated.) Then the leading term of the asymptotic expansion
of the fundamental solution as at → ∞ has the form
r
1 πa sin(λt − π/4)
Ee (r, t) = − 2 2 2
√
2π 2λ|a − λ | |r| at
r
1 π sin(at − π/4) 1 1
− 2 √ + O .
2π 2|a2 − λ2 | |r| at |r| at
11.5. Fourth-Order Stationary Equations 1003
5◦ . The Cauchy problem (t ≥ 0, r ∈ R3 ). The initial conditions:
w = f (r) at t = 0,
∂t w = g(r) at t = 0.
Solution:
Z Z
∂
w(r, t) = Ee (r − r′ , t)∆′3 f (r′ ) dV ′ + Ee (r − r′ , t)∆′3 g(r′ ) dV ′ ,
∂t R3 R3
where ∆′3 is the three-dimensional Laplace operator in the integration variables (x′ , y ′ , z ′ )
and dV ′ = dx′ dy ′ dz ′ .
⊙ Literature: S. Ya. Sekerzh–Zenkovich (1979), S. A. Gabov and A. G. Sveshnikov (1990).

∂2 ∂ 2w ∂2w ∂2w ∂ 2w ∂2w ∂2w
6. + + + a2 + + b2 = Φ(x, y, z, t).
∂t2 ∂x2 ∂y 2 ∂z 2 ∂x2 ∂y 2 ∂z 2
The Cauchy problem (t ≥ 0, r ∈ R3 ). The initial conditions:
w = f (r) at t = 0,
∂t w = g(r) at t = 0,
where r = (x, y, z).

Solution:
Z Z
∂ ′ ′ ′ ′
w(r, t) = Ee (r − r , t)∆3 f (r ) dV + Ee (r − r′ , t)∆′3 g(r′ ) dV ′
∂t R3 R3
Z tZ
+ Ee (r − r′ , t − τ )Φ(r′ , τ ) dV ′ dτ,
0 R3
where Ee (r, t) is the fundamental solution given in Item 3◦ of the preceding equation,
∆′3 is the three-dimensional Laplace operator in the integration variables x′ , y ′ , z ′ , and
dV ′ = dx′ dy ′ dz ′ .
11.5 Fourth-Order Stationary Equations

11.5.1 Biharmonic Equation ∆∆w = 0
The biharmonic equation is encountered in plane problems of elasticity (w is the Airy stress
function). It is also used to describe slow flows of viscous incompressible fluids (in this
case w is the stream function).
All solutions of the Laplace equation ∆w = 0 (see Sections 9.1 and 10.1) are also
solutions of the biharmonic equation.
◮ Two-dimensional equation. Particular solutions. Fundamental solution.
1◦ . In the rectangular Cartesian system of coordinates, the biharmonic operator has the
form
∂4 ∂4 ∂4
∆∆ ≡ ∆2 = + 2 + .
∂x4 ∂x2 ∂y 2 ∂y 4
w(x, y) = Ax3 + Bx2 y + Cxy 2 + Dy 3 + ax2 + bxy + cy 2 + αx + βy + γ,

w(x, y) = (A cosh βx + B sinh βx + Cx cosh βx + Dx sinh βx)(a cos βy + b sin βy),
w(x, y) = (A cos βx + B sin βx + Cx cos βx + Dx sin βx)(a cosh βy + b sinh βy),
p
w(x, y) = Ar 2 ln r + Br 2 + C ln r + D, r = (x − a)2 + (y − b)2 ,
w(x, y) = (Ax + By + C)(D cosh βx + E sinh βx)(a cos βy + b sin βy),
w(x, y) = (Ax + By + C)(D cosh βy + E sinh βy)(a cos βx + b sin βx),
w(x, y) = (x2 + y 2 )(D cosh βx + E sinh βx)(a cos βy + b sin βy),
w(x, y) = (x2 + y 2 )(D cosh βy + E sinh βy)(a cos βx + b sin βx),
where A, B, C, D, E, a, b, c, α, β, and γ are arbitrary constants.

3◦ . Particular solutions of the biharmonic equation in some orthogonal curvilinear coordi-
nate systems are listed in Table 11.1.
TABLE 11.1
Particular solutions of the biharmonic equation in some orthogonal curvilinear
coordinate systems; A, B, C, D, a, b, and λ are arbitrary constants
Transformation Particular solutions
w = (Ar 2+λ + Br 2−λ + Cr λ + Dr −λ ) cos(λϕ),

Polar coordinates r, ϕ:
w = (Ar 2+λ + Br 2−λ + Cr λ + Dr −λ ) sin(λϕ),
x = r cos ϕ, y = r sin ϕ
w = Ar 2 ln r + Br 2 + C ln r + D (at λ = 0)
Bipolar coordinates ξ, η: a cos λη + b sin λη

w= A cosh(λ + 1)ξ
c sinh ξ c sin η cosh ξ − cos η
x= , y=
cosh ξ − cos η cosh ξ − cos η + B sinh(λ + 1)ξ + C cosh(λ − 1)ξ + D sinh(λ − 1)ξ
Degenerate bipolar a cos(λv) + b sin(λv)

w= A cosh(λu) + B sinh(λu)
coordinates u, v: (u2 + v 2 )2
u v
x= 2 , y=− 2 + Cu cosh(λu) + Du sinh(λu)
u + v2 u + v2
⊙ Literature: N. N. Lebedev, I. P. Skal’skaya, and Ya. S. Uflyand (1955).
1 2 p
Ee (x, y) = r ln r, r= x2 + y 2 . (1)
8π
◮ Two-dimensional equation. Various representations of the general solution.

1◦ . Various representations of the general solution of the biharmonic equation in terms of
harmonic functions:
w(x, y) = xu1 (x, y) + u2 (x, y),

w(x, y) = yu1 (x, y) + u2 (x, y),
w(x, y) = (x2 + y 2 )u1 (x, y) + u2 (x, y),
where u1 and u2 are arbitrary functions satisfying the Laplace equation ∆uk = 0 (k = 1, 2).
2◦ . Complex form of representation of the general solution:

w(x, y) = Re zf (z) + g(z) ,
where f (z) and g(z) are arbitrary analytic functions of the complex variable z = x + iy;
z = x − iy, i2 = −1. The symbol Re[A] stands for the real part of a complex number A.
◮ Two-dimensional boundary value problems for the upper half-plane.

1◦ . Domain: −∞ < x < ∞, 0 ≤ y < ∞. The desired function and its derivative along the
normal are prescribed on the boundary:
w = 0 at y = 0, ∂y w = f (x) at y = 0.
Solution:
Z ∞
1 y2
w(x, y) = f (ξ)G(x − ξ, y) dξ, G(x, y) = .
−∞ π x2 + y 2
2◦ . Domain: −∞ < x < ∞, 0 ≤ y < ∞. The derivatives of the desired function are
prescribed on the boundary:
∂x w = f (x) at y = 0, ∂y w = g(x) at y = 0.
Solution:
Z
1 ∞ x−ξ y(x − ξ)
w(x, y) = f (ξ) arctan + dξ
π −∞ y (x − ξ)2 + y 2
Z
y2 ∞ g(ξ) dξ
+ + C,
π −∞ (x − ξ)2 + y 2

Example 11.1. Consider the problem of a slow (Stokes) inflow of a viscous fluid into a half-
plane through a slit of width 2a at a constant velocity U that makes an angle β with the normal to
the boundary (the angle is measured off from the normal direction counterclockwise).
With the stream function w introduced by the relations vx = − ∂w ∂w
∂y and vy = ∂x (vx and vy are
the fluid velocity components), the problem is reduced to the special case of the previous problem
with
( (
U cos β for |x| < a, U sin β for |x| < a,
f (x) = g(x) =
0 for |x| > a, 0 for |x| > a.
Dean’s solution:

U y
w(x, y) = (x − a) cos β + y sin β arctan
π x−a

U y
− (x + a) cos β + y sin β arctan + C.
π x+a
◮ Two-dimensional boundary value problem for a disk.
Domain: 0 ≤ r ≤ a, 0 ≤ ϕ ≤ 2π. Boundary conditions in the polar coordinate system:
w = f (ϕ) at r = a, ∂r w = g(ϕ) at r = a.
Solution:
Z 2π
1 2 2 2 [a − r cos(η − ϕ)]f (η) dη
w(r, ϕ) = (r − a )
2πa 0 [r 2 + a2 − 2ar cos(η − ϕ)]2
Z 2π
1 g(η) dη
− .
2 0 r 2 + a2 − 2ar cos(η − ϕ)
◮ Three-dimensional equation. Particular solutions.
In the rectangular Cartesian coordinate system, the three-dimensional biharmonic operator

is expressed as
∂4 ∂4 ∂4 ∂4 ∂4 ∂4
∆∆ ≡ ∆2 = + + + 2 + 2 + 2 .
∂x4 ∂y 4 ∂z 4 ∂x2 ∂y 2 ∂x2 ∂z 2 ∂y 2 ∂z 2
1◦ . Particular solutions in the Cartesian coordinate system:
D p
w(x, y, z) = Ar 2 + Br + C + , r = (x − a)2 + (y − b)2 + (z − c)2 ,
r p

w(x, y, z) = Ax sin(βx + C1 ) + B sin(βx + C2 ) sin(µy) exp ±z β 2 + µ2 ,
p
w(x, y, z) = Ax sin(βx + C1 ) + B sin(βx + C2 ) cos(µy) exp ±z β 2 + µ2 ,
p
w(x, y, z) = Ax sin(βx + C1 ) + B sin(βx + C2 ) sinh(µy) exp ±z β 2 − µ2 ,
p
w(x, y, z) = Ax sin(βx + C1 ) + B sin(βx + C2 ) cosh(µy) exp ±z β 2 − µ2 ,
p
w(x, y, z) = Ax sinh(βx + C1 ) + B sinh(βx + C2 ) sinh(µy) sin z β 2 + µ2 ,
p
w(x, y, z) = Ax cosh(βx + C1 ) + B cosh(βx + C2 ) sinh(µy) sin z β 2 + µ2 ,
p
w(x, y, z) = Ax sinh(βx + C1 ) + B sinh(βx + C2 ) cosh(µy) cos z β 2 + µ2 ,
p
w(x, y, z) = Ax cosh(βx + C1 ) + B cosh(βx + C2 ) cosh(µy) cos z β 2 + µ2 ,
where A, B, C, C1 , C2 , D, β, and µ are arbitrary constants.

p
2◦ . Particular solutions in the cylindrical coordinate system r = x2 + y 2 :
w(r, ϕ, z) = Jn (µr)(Ar cos ϕ + Br sin ϕ + C)(a1 cos nϕ + b1 sin nϕ) exp(±µz),

w(r, ϕ, z) = Yn (µr)(Ar cos ϕ + Br sin ϕ + C)(a1 cos nϕ + b1 sin nϕ) exp(±µz),
w(r, ϕ, z) = In (µr)(Ar cos ϕ + Br sin ϕ + C)(a1 cos nϕ + b1 sin nϕ) sin(µz + β),
w(r, ϕ, z) = Kn (µr)(Ar cos ϕ + Br sin ϕ + C)(a1 cos nϕ + b1 sin nϕ) sin(µz + β),
w(r, ϕ, z) = Jn (µr) cos(nϕ + β)(a1 cosh µz + b1 sinh µz + a2 z cosh µz + b2 z sinh µz),
w(r, ϕ, z) = Yn (µr) cos(nϕ + β)(a1 cosh µz + b1 sinh µz + a2 z cosh µz + b2 z sinh µz),
w(r, ϕ, z) = In (µr) cos(nϕ + β)(a1 cos µz + b1 sin µz + a2 z cos µz + b2 z sin µz),
w(r, ϕ, z) = Kn (µr) cos(nϕ + β)(a1 cos µz + b1 sin µz + a2 z cos µz + b2 z sin µz),
where n = 0, 1, 2, . . . ; A, B, C, a1 , a2 , b1 , b2 , β, and µ are arbitrary constants; the

Jn (ξ) and Yn (ξ) are Bessel functions; and the In (ξ) and Kn (ξ) are modified Bessel func-
tions.
p
3◦ . Particular solutions in the spherical coordinate system r = x2 + y 2 + z 2 :
w(r) = Ar 2 + Br + C + Dr −1 ,

w(r, θ) = Ar n+2 + Br n + Cr 1−n + Dr −1−n Pn (cos θ),

w(r, θ, ϕ) = Ar n+2 + Br n + Cr 1−n + Dr −1−n Pnm (cos θ)(a cos mϕ + b sin mϕ),
where n = 0, 1, 2, . . . ; m = 0, 1, 2, . . . , n; A, B, C, D, a, and b are arbitrary constants;

the Pn (ξ) are the Legendre polynomials; and the Pnm (ξ) are the associated Legendre func-
tions defined by
1 dn 2 dm
Pn (x) = (x − 1)n , Pnm (x) = (1 − x2 )m/2 Pn (x).
n! 2n dxn dxm
◮ Three-dimensional equation. Fundamental solution. General solution.
1 p 2
Ee (x, y, z) = − x + y2 + z2 . (2)
8π
2◦ . Various representations of the general solution of the biharmonic equation in terms of

harmonic functions:
w(x, y, z) = xu1 (x, y, z) + u2 (x, y, z),

w(x, y, z) = (x2 + y 2 + z 2 )u1 (x, y, z) + u2 (x, y, z),
where u1 and u2 are arbitrary functions satisfying the three-dimensional Laplace equation
∆3 uk = 0 (k = 1, 2). The coefficient x of u1 in the first formula can be replaced by y or z.
◮ n-dimensional equation.
n
X n
X n
X
w(x) = Aijk xi xj xk + Bij xi xj + Ci xi + D,
i,j,k=1 i,j=1 i=1
Xn
w(x) = Ar 2 + B + Cr 4−n + Dr 2−n , r2 = (xk − αk )2 ,
k=1
X
n n
X
2−n
w(x) = (A + Br ) Ci xi + D , r2 = (xk − αk )2 ,
i=1 k=1
p X n n−1
Y n−1
X

w(x) = exp ±xn λn Ai xi + B sin(αk xk + βk ), λn = α2k ,
i=1 k=1 k=1
X
n m−1
Y Y
n m−1
X n
X
w(x) = Ai xi + B sin(αk xk + βk ) sinh(γk xk ) , α2k = γk2 ,
i=1 k=1 k=m k=1 k=m
where the Aijk , Bij , Ai , Ci , A, B, C, D, αk , βk , and γk are arbitrary constants.


 Γ(n/2)|x|4−n
 for n = 3, 5, 6, 7, . . . ;
n/2
Ee (x) = 4π (n − 2)(n − 4)
− 1 ln |x|

for n = 4.
8π 2
For n = 2, see formula (1) above.
3◦ . Various representations of solutions of the biharmonic equation in terms of harmonic

functions:
w(x) = xs u1 (x) + u2 (x), s = 1, 2, . . . , n;
Xn
2 2
w(x) = |x| u1 (x) + u2 (x), |x| = x2k ,
k=1
where u1 and u2 are arbitrary functions satisfying the n-dimensional Laplace equation
∆n um = 0 (m = 1, 2).
11.5.2 Equation of the Form ∆∆w = Φ

Nonhomogeneous biharmonic equation. It is encountered in problems of elasticity and
hydrodynamics.
◮ Domain: −∞ < x < ∞, −∞ < y < ∞. Particular solution.
Particular solution with Φ = Φ(x, y):

Z ∞Z ∞ p
1 2 2
w(x, y) = Φ(ξ, η)Ee (x−ξ, y−η) dξ dη, Ee (x, y) = (x +y ) ln x2 + y 2 .
−∞ −∞ 8π
◮ Boundary value problems in the Cartesian coordinates.
1◦ . The upper half-plane is considered (−∞ < x < ∞, 0 ≤ y < ∞). The derivatives are
prescribed on the boundary:
∂x w = f (x) at y = 0, ∂y w = g(x) at y = 0.
Solution:
Z Z
1 ∞ x−ξ y(x − ξ) y2 ∞ g(ξ) dξ
w(x, y) = f (ξ) arctan + 2 2
dξ +
π −∞ y (x − ξ) + y π −∞ (x − ξ)2 + y 2
Z ∞ Z ∞
1 1 2 2 2 R+
+ dξ (R+ − R− ) − R− ln Φ(ξ, η) dη + C,
8π −∞ 0 2 R−
where C is an arbitrary constant and
2
R+ = (x − ξ)2 + (y + η)2 , 2
R− = (x − ξ)2 + (y − η)2 .

2◦ . A rectangle is considered (0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 ). The sides of the plate are hinged.

Solution: Z l1Z l2
w(x, y) = Φ(ξ, η)G(x, y, ξ, η) dη dξ,
0 0
where
∞ ∞
4 XX 1
G(x, y, ξ, η) = 2 2 )2
sin(pn x) sin(qm y) sin(pn ξ) sin(qm η),
l1 l2 (pn + qm
n=1 m=1
πn πm
pn = , qm = .
l1 l2
◮ Domain: 0 ≤ r ≤ 1, 0 ≤ ϕ ≤ 2π. Various boundary value problems for the disk.

1◦ . Consider boundary value problems in the circle 0 ≤ r ≤ 1, 0 ≤ ϕ ≤ 2π with vari-
ous homogeneous boundary conditions in the polar coordinate system, Φ = Φ(r, ϕ). The
solution can be represented in terms of the Green’s function as
Z 1 Z 2π
w(r, ϕ) = Φ(ξ, η)G(r, ϕ, ξ, η) dη dξ.
0 0
2◦ . Boundary conditions are prescribed:
w = 0 at r = 1, ∂r w = 0 at r = 1.
These boundary conditions correspond to the rigid clamping of the circular plate. (The de-
flection w and the rotation angle ∂r w of the normal element vanish on the support contour.)
Green’s function:

1 2 2
1 + r 2 ξ 2 − 2ξr cos(ϕ − η)
G0 (r, ϕ, ξ, η) = ξ r + ξ − 2ξr cos(ϕ − η) −1
16π r 2 + ξ 2 − 2ξr cos(ϕ − η)
2
r + ξ 2 − 2ξr cos(ϕ − η)
+ ln .
1 + r 2 ξ 2 − 2ξr cos(ϕ − η)
Here the Green’s function is equipped with the zero subscript, which will prove useful in
the subsequent exposition.
⊙ Literature: J. H. Michell (1902).

w = 0 at r = 1, ∂rr w = 0 at r = 1.
Green’s function:
√
ξ (ξ 2 − 1)(r 2 − 1) i ϕ−η p ϕ−η
G(r, ϕ, ξ, η) = G0 (r, ϕ, ξ, η) + √ e 2 tanh−1 ξr e−i 2
8π r
ϕ−η p i ϕ−η
p
+ e−i 2 tanh−1 ξr e 2 − rξ ,
where G0 (r, ϕ, ξ, η) is the Green’s function in Item 2◦ .

ν
w=0 at r = 1, ∂rr w +
∂r w = 0 at r = 1.
r
These boundary conditions correspond to a hinged circular plate. (The deflection w and the
bending torque vanish on the support contour; ν is the Poisson ratio.)
Green’s function:

ξ(ξ 2 − 1)(r 2 − 1) ν+1 ν+3 −i(ϕ−η)
G(r, ϕ, ξ, η) = G0 (r, ϕ, ξ, η) + F 1, ; ; ξre
8π(ν + 1) 2 2

ν+1 ν+3
+ F 1, ; ; ξrei(ϕ−η) − 1 ,
2 2
where G0 (r, ϕ, ξ, η) is the Green’s function in Item 2◦ and F (a, b; c; z) is the hypergeo-
metric function.
ν
w = 0 at r = 1, ∂rr w + + γ ∂r w = 0
at r = 1.
r
These boundary conditions correspond to an elastically supported circular plate. (The de-
flection w on the support contour is zero, and the bending torque is proportional to the angle
of rotation of the normal element; γ characterizes the elasticity of the support contour.)
The Green’s function in this case can be obtained from the Green’s function given in
Item 4◦ by the replacement of the parameter ν with ν + γ.
6◦ . Singular solution with Φ(r, ϕ) = δ(r − ξ)δ(ϕ − η):
ξ(ξ 2 − 1)(r 2 − 1)
w = G0 (r, ϕ, ξ, η) − ln 1 + r 2 ξ 2 − 2ξr cos (ϕ − η) ,
16π
where δ(r) is the Dirac delta function and ξ and η are arbitrary constants. This solution
satisfies the boundary conditions
ξ(1 − ξ 2 )
w = 0 at r = 1, ∂rr w − ∂r w = at r = 1.
4π
7◦ . Solution with Φ(r, ϕ) = α + βr cos ϕ:

r2 − 1 2 12α 4rβ cos ϕ
w= (3α + βr cos ϕ)(r − 1) − − .
192 1+ν+γ 3+ν+γ
This solution satisfies the boundary conditions in Item 5◦ .
⊙ The results of Items 3◦ –7◦ were obtained by S. A. Lychev (private communication, 2014).
◮ Domain: −∞ < x < ∞, −∞ < y < ∞, −∞ < z < ∞. Particular solution.

Particular solution with Φ = Φ(x, y, z):
Z ∞Z ∞Z ∞
w(x, y) = Φ(ξ, η, ζ)Ee (x − ξ, y − η, z − ζ) dξ dη dζ,
−∞ −∞ −∞
1 p 2
Ee (x, y, z) = − x + y2 + z2 .
8π
⊙ Literature: O. A. Ladyzhenskaya (1969).
11.5.3 Equation of the Form ∆∆w − λw = Φ(x, y)

1◦ . This equation describes the shapes of two-dimensional free transverse vibrations of
a thin elastic plate; the function w defines the deflection (transverse displacement) of the
plate midplane points relative to the original plane position, and k = λ1/4 is the frequency
parameter. Here ∆∆ = ∆2 is the biharmonic operator and ∆ is the Laplace operator defined
as ( 2
∂ ∂2
∂x2 + ∂y 2 in the Cartesian coordinate system,
∆ = ∂2 1 ∂ 1 ∂2
∂r 2
+ r ∂r + r2 ∂ϕ2 in the polar coordinate system.
2◦ . Particular solutions (A, B, C, and D are arbitrary constants):

w(x, y) = A sin(k1 x) + B cos(k1 x) C sin(k2 y) + D cos(k2 y) , λ = (k12 + k22 )2 ,

w(x, y) = A sin(k1 x) + B cos(k1 x) C exp(−k2 y) + D exp(k2 y) , λ = (k12 − k22 )2 ,

w(x, y) = A exp(−k1 x) + B exp(k1 x) C sin(k2 y) + D cos(k2 y) , λ = (k12 − k22 )2 ,

w(x, y) = A exp(−k1 x) + B exp(k1 x) C exp(−k2 y) + exp(k2 y) , λ = (k12 + k22 )2 ,

w(r, ϕ) = AJn (kr) + BYn (kr) + CIn (kr) + DKn (kr) cos(nϕ), λ = k4 > 0,

w(r, ϕ) = AJn (kr) + BYn (kr) + CIn (kr) + DKn (kr) sin(nϕ), λ = k4 > 0,
where the Jn (ξ) and Yn (ξ) are Bessel functions of the first and second kind,
p the In (ξ)
and Kn (ξ) are modified Bessel functions of the first and second kind, r = x2 + y 2 , and
n = 0, 1, 2, . . .
3◦ . General solution with λ > 0:
w(x, y) = u1 (x, y) + u2 (x, y),
where u1 and u2 are arbitrary functions satisfying the Helmholtz equations
√ √
∆u1 + λ u1 = 0, ∆u2 − λ u2 = 0.
For the solutions of these equations, see Section 9.3.
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 . Boundary value problem.

Solution: Z l1Z l2
w(x, y) = Φ(ξ, η)G(x, y, ξ, η) dη dξ,
0 0
where
∞ ∞
4 X X sin(pn x) sin(qm y) sin(pn ξ) sin(qm η)
G(x, y, ξ, η, t) = ,
l1 l2 (p2n + qm
2 )2 − λ
n=1 m=1
πn πm
pn = , qm = .
l1 l2
◮ Domain: 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π. Eigenvalue problem with Φ ≡ 0.

The unknown and its normal derivative are zero on the boundary of a circular domain:
∂w
w= = 0 at r = R.
∂r
Eigenvalues:
4
βnm
λnm = , n = 0, 1, 2, . . . , m = 1, 2, 3, . . . ,
R4
where the βnm are positive roots of the transcendental equation
Jn (β)In′ (β) − In (β)Jn′ (β) = 0.
Numerical values of some roots:
β01 = 3.196, β02 = 6.306, β03 = 9.439, β04 = 12.58;

β11 = 4.611, β12 = 7.799, β13 = 10.96, β14 = 14.11;
β21 = 5.906, β22 = 9.197, β23 = 12.40, β24 = 15.58,
β31 = 7.144, β32 = 10.54, β33 = 13.79, β34 = 17.01.
Eigenvalues:
h r r i
(c)
wnm (r, ϕ) = In (βnm )Jn βnm − Jn (βnm )In βnm cos(nϕ),
h R R i
(s) r r
wnm (r, ϕ) = In (βnm )Jn βnm − Jn (βnm )In βnm sin(nϕ).
R R
⊙ Literature: V. V. Bolotin (1978).
◮ Domain: (x/a)2 + (y/b)2 ≤ 1. Eigenvalue problem with Φ ≡ 0.

The unknown and its normal derivative are zero on the boundary of an elliptic domain:
∂w
w= = 0 on (x/a)2 + (y/b)2 = 1 (a ≥ b).
∂N
Eigenvalues and eigenfunctions (approximate formulas):
β04 3 3 2
λ0 = + + , w0 (ρ) = I0 (β0 )J0 (β0 ρ) − J0 (β0 )I0 (β0 ρ),
8 a4 b4 a2 b2
(c) β4 5 1 2 (c)
λ11 = 11 + + , w11 (ρ, ϕ) = f (ρ) cos ϕ,
8 a4 b4 a2 b2
(s) β4 1 5 2 (s)
λ11 = 11 + + , w11 (ρ, ϕ) = f (ρ) sin ϕ,
8 a4 b4 a2 b2
where
p
ρ= (x/a)2 + (y/b)2 , f (ρ) = I1 (β11 )J1 (β11 ρ) − J1 (β11 )I1 (β11 ρ) ,
Jn (z) and In (z) are the Bessel and the modified Bessel functions, and β0 = 3.196 and
β11 = 4.611 are the least roots of the transcendental equations
J0 (β)I1 (β) + J1 (β)I0 (β) = 0,

J1 (β)I1′ (β) − J1′ (β)I1 (β) = 0.
The above formulas were obtained with the aid of generalized (nonorthogonal) polar
coordinates ρ, ϕ defined by
x = aρ cos ϕ, y = bρ sin ϕ (0 ≤ ρ ≤ 1, 0 ≤ ϕ ≤ 2π)
and the variational method. p

The maximum error in the eigenvalue λ1 for ε = 1 − (b/a)2 ≤ 0.86 is less than 1%.
(c) (s)
The errors in λ11 and λ11 for ε ≤ 0.6 do not exceed 2%. In the limit case ε = 0 that
corresponds to a circular domain, the formulas provide exact results.
⊙ Literature: L. D. Akulenko, S. V. Nesterov, and A. L. Popov (2001).
∂4w ∂4w
11.5.4 Equation of the Form + = Φ(x, y)
∂x4 ∂y 4
◮ Homogeneous equation. Particular solutions. General solution.
1◦ . Particular solutions with Φ ≡ 0:

w(x, y) = A sin(λx + β) + B exp(−λx) + C exp(λx) exp √12 λy sin √12 λy ,

w(x, y) = A sin(λx + β) + B exp(−λx) + C exp(λx) exp √12 λy cos √12 λy ,

w(x, y) = A sin(λx + β) + B exp(−λx) + C exp(λx) exp − √12 λy sin √12 λy ,

w(x, y) = A sin(λx + β) + B exp(−λx) + C exp(λx) exp − √12 λy cos √12 λy ,
where A, B, C, β, and λ are arbitrary constants.


w(x, y) = Re f (z1 ) + g(z2 ) .
Here f (z1 ) and g(z2 ) are arbitrary analytic functions of the complex variables z1 = x −
√1 (1 + i)y and z2 = x + √1 (1 + i)y. The symbol Re[A] stands for the real part of a
2 2
complex number A.
⊙ Literature: A. V. Bitsadze end D. F. Kalinichenko (1985).
◮ Homogeneous equation. Boundary value problem.

The upper half-space is considered (−∞ < x < ∞, 0 ≤ y < ∞).
w = 0 at y = 0, ∂y w = f (x) at y = 0.
Solution: Z ∞
w(x, y) = f (ξ)G(x − ξ, y) dξ,
−∞
where √ √
1 x 2 x 2
G(x, y) = √
arctan 1 − + arctan 1 + .
π 2 y y
◮ Nonhomogeneous equation. Boundary value problems in a rectangle.

Consider problems in the rectangular domain 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 with various homoge-
neous boundary conditions. The solution can be expressed in terms of the Green’s function
as Z Z l1 l2
w(x, y) = Φ(ξ, η)G(x, y, ξ, η) dη dξ.
0 0
Given below are the Green’s functions for two types of boundary conditions.
w = ∂x w = 0 at x = 0, w = ∂x w = 0 at x = l1 ,
w = ∂y w = 0 at y = 0, w = ∂y w = 0 at y = l2 .
Green’s function:
∞ ∞
16 X X p4n qm 4 ϕ (x)ψ (y)ϕ (ξ)ψ (η)
n m n m
G(x, y, ξ, η) = 2 .
l1 l2 4 4 ′′ ′′
n=1 m=1 (pn + qm ) ϕn (l1 )ψm (l2 )
Here



where pn and qm are positive roots of the transcendental equations
Green’s function:
∞ ∞
4 XX 1
G(x, y, ξ, η) = sin(pn x) sin(qm y) sin(pn ξ) sin(qm η),
l1 l2 p4n + qm4
n=1 m=1
πn πm
pn = , qm = .
l1 l2
∂4w ∂4w
11.5.5 Equation of the Form + + kw = Φ(x, y)
∂x4 ∂y 4

w(x, y) = A sin(λx) + B cos(λx) + C sinh(λx) + D cosh(λx) exp(βy) sin(βy),

w(x, y) = A sin(λx) + B cos(λx) + C sinh(λx) + D cosh(λx) exp(βy) cos(βy),

w(x, y) = A sin(λx) + B cos(λx) + C sinh(λx) + D cosh(λx) exp(−βy) sin(βy),

w(x, y) = A sin(λx) + B cos(λx) + C sinh(λx) + D cosh(λx) exp(−βy) cos(βy),
where β = √1 (λ4 + k)1/4 ; A, B, C, D, and λ are arbitrary constants.
2
◮ Domain: 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 . Boundary value problems.

1◦ . Consider problems in the rectangular domain with various homogeneous boundary
conditions. The solution can be expressed in terms of the Green’s function as
Z l1Z l2
w(x, y) = Φ(ξ, η)G(x, y, ξ, η) dη dξ.
0 0
Given below are the Green’s functions for two types of boundary conditions.
w = ∂x w = 0 at x = 0, w = ∂x w = 0 at x = l1 ,
w = ∂y w = 0 at y = 0, w = ∂y w = 0 at y = l2 .
Green’s function:
∞ ∞
16 X X p4n qm 4 ϕ (x)ψ (y)ϕ (ξ)ψ (η)
n m n m
G(x, y, ξ, η) = 2 .
l1 l2 4 4 ′′ ′′
n=1 m=1 (pn + qm + k) ϕn (l1 )ψm (l2 )
Here



where pn and qm are positive roots of the transcendental equations
Green’s function:
∞ ∞
4 XX 1
G(x, y, ξ, η) = sin(pn x) sin(qm y) sin(pn ξ) sin(qm η),
l1 l2 p4n + qm4 +k
n=1 m=1
πn πm
pn = , qm = .
l1 l2
11.5.6 Stokes Equation (Axisymmetric Flows of Viscous Fluids)

◮ Stokes equation for the stream function in the spherical coordinate system.
1◦ . The Stokes equation for the stream function in the axisymmetric case is written as

∂2 sin θ ∂ 1 ∂
E 2 (E 2 w) = 0, E2 ≡ + .
∂r 2 r 2 ∂θ sin θ ∂θ
It describes slow axisymmetric flows of viscous incompressible fluids; w is the stream
function, and (r, θ) are the spherical coordinates. The fluid velocity components are related
1 ∂w 1 ∂w
to the stream function by vr = 2 and vθ = − .
r sin θ ∂θ r sin θ ∂r
en , B
2◦ . General solution (An , Bn , Cn , Dn , A en , C
en , and De n are arbitrary constants):
∞
X
w(r, θ) = An r n + Bn r 1−n + Cn r n+2 + Dn r 3−n Jn (cos θ)
n=0
∞ (1)
X
+ en r + B
(A nen r 1−n en r
+C n+2 e nr
+D 3−n
Hn (cos θ),
n=2
where Jn (ζ) and Hn (ζ) are the Gegenbauer functions of the first and second kind, respec-
tively. These are linearly related to the Legendre functions Pn (ζ) and Qn (ζ) by
Pn−2 (ζ) − Pn (ζ) Qn−2 (ζ) − Qn (ζ)
Jn (ζ) = , Hn (ζ) = (n ≥ 2).
2n − 1 2n − 1
The Gegenbauer functions of the first kind are represented in the form of finite power
series as
n−2 2
1 d ζ − 1 n−1
Jn (ζ) = −
(n − 1)! dζ 2

1 · 3 . . . (2n − 3) n n(n − 1) n−2 n(n − 1)(n − 2)(n − 3) n−4
= ζ − ζ + ζ −· · · .
1·2 . . . n 2(2n − 3) 2 · 4(2n − 3)(2n − 5)
In particular,
J0 (ζ) = 1, J1 (ζ) = −ζ, J2 (ζ) = 12 (1 − ζ 2 ), J3 (ζ) = 12 ζ(1 − ζ 2 ),

J4 (ζ) = 18 (1 − ζ 2 )(5ζ 2 − 1), J5 (ζ) = 18 ζ(1 − ζ 2 )(7ζ 2 − 3).
The Gegenbauer functions of the second kind are defined as

1 1+ζ
H0 (ζ) = −ζ, H1 (ζ) = −1, Hn (ζ) = Jn (ζ) ln + Kn (ζ) at n ≥ 2,
2 1−ζ
where the functions Kn (ζ) are expressed in terms of the Gegenbauer functions of the first
kind as
1
n≤k≤ 12 n+ 12
2 X (2n − 4k + 1) (2k − 1)(n − k)
Kn (ζ) = − 1− Jn−2k+1 (ζ);
(2k − 1)(n − k) n(n − 1)
k
the series start with J0 or J1 depending on whether n is odd or even. In particular,
K2 (ζ) = 12 ζ, K3 (ζ) = 16 (3 ζ 2 − 2), K4 (ζ) = 1

24 ζ(15 ζ
2
− 13),
1 4 2
K5 (ζ) = 120 (105 ζ − 115 ζ + 16).
For n ≥ 2, the Gegenbauer functions of the second kind take infinite values at the
points ζ = ±1, which correspond to θ = 0 and θ = π. Therefore, if physically there are no
singularities in the problem, then the constants in (1) labeled with a tilde must be zero. In
an overwhelming majority of problems on the flow about particles, drops, or bubbles, the
stream function in the spherical coordinates is given by formula (1) with
en = B
A1 = A0 = B1 = B0 = C1 = C0 = D1 = D0 = 0; A en = C
en = D
e n = 0 for n = 2, 3, . . .
Example 11.2. In the problem on the translational Stokes flow about a solid spherical particle,
the following boundary conditions are imposed on the stream function w:
w=0 at r = R, ∂r w = 0 at r = R, w → 12 U r2 sin2 θ as r → ∞,
where R is the radius of the particle and U is the unperturbed fluid velocity at infinity.
Stokes solution:
1 R
w(r, θ) = U (r − R)2 2 + sin2 θ.
4 r
Here the only remaining terms are those in the first sum in (1) with n = 2.
Example 11.3. In the problem on the axisymmetric straining Stokes flow about a solid spherical
particle, the following boundary conditions are imposed on the stream function w:
3
w=0 at r = R, ∂r w = 0 at r = R, w→ 1
2 Er sin2 θ cos θ as r → ∞,
where R is the particle radius and E is the shear modulus.

Solution: 3
1 r 5 3 R2
w(r, θ) = ER3 − + sin2 θ cos θ.
2 R3 2 2 r2
Here the only remaining terms are those in the first sum in (1) with n = 3.
Example 11.4. Solving the problem of the translational Stokes flow about a spherical drop (or
bubble) is reduced to solving the Stokes equation outside and inside the drop. The boundary con-
dition at infinity is given in Example 11.2. Transmission boundary conditions are set on the drop
surface; these conditions can be found in the references cited below and are not written out here.
Hadamard–Rybczynski solution:

1 2 3β + 2 R β R3
w(r, θ) = U r 2 − + sin2 θ for r > R,
4 β+1 r β + 1 r3
2
U r
w(r, θ) = r 2
− 1 sin2 θ for r < R,
4(β + 1) R2
where R is the drop radius, U is the unperturbed fluid velocity at infinity, and β is the ratio of
dynamic viscosities of the fluids inside and outside the drop. (The value β = 0 corresponds to a gas
bubble; β = ∞, to a solid particle.)
Example 11.5. Solving the problem of the axisymmetric straining Stokes flow about a spherical
drop (or bubble) is reduced to solving the Stokes equation outside and inside the drop. The boundary
condition at infinity is specified in Example 11.3. Transmission boundary conditions are set at the
drop surface; these conditions can be found in the references cited below and are not written out
here.
Taylor solution:
3
1 r 1 5β + 2 3 β R2
w(r, θ) = ER 3
− + sin2 θ cos θ for r > R,
2 R3 2 β+1 2 β + 1 r2

3 ER3 r3 r2
w(r, θ) = − 1 sin2 θ cos θ for r < R,
4 β + 1 R3 R2
where R is the drop radius, E is the shear modulus, and β is the ratio of dynamic viscosities of the
fluids inside and outside the drop. (The value β = 0 corresponds to a gas bubble; β = ∞, to a solid
particle.)
⊙ Literature: G. I. Taylor (1932), V. G. Levich (1962), J. Happel and H. Brenner (1965), A. D. Polyanin,
A. M. Kutepov, A. V. Vyazmin, and D. A. Kazenin (2002).
◮ Stokes equation in the bipolar coordinate system.

1◦ . When studying axisymmetric problems of a flow about two spherical particles (drops,
bubbles), one uses the bipolar coordinates ξ, η; these are related to the cylindrical coordi-
nates ρ = r cos θ, z = r sin θ by
a sin ξ a sinh η
ρ= , z= .
cosh η − cos ξ cosh η − cos ξ
2◦ . The general solution of the equation E 2 (E 2 w) = 0 in the bipolar coordinate system

has the form
X∞ X∞
1
w(ξ, η) = Jn+1 (cos ξ)fn (η) + Hn+1 (cos ξ)gn (η) ,
(cosh η − cos ξ)3/2 n=0 n=0

fn (η) = An cosh (n− 12 )η +Bn sinh (n− 12 )η +Cn cosh (n+ 32 )η +Dn sinh (n+ 32 )η ,

en cosh (n− 1 )η + B
gn (η) = A en sinh (n− 1 )η + C en cosh (n+ 3 )η + De n sinh (n+ 3 )η ,
2 2 2 2
en , B
where An , Bn , Cn , Dn , A en , C
en , and D
e n are arbitrary constants and Jn (ζ) and Hn (ζ)
are the Gegenbauer functions.
⊙ Literature: J. Happel and H. Brenner (1965).
◮ Stokes equation in the oblate spheroidal coordinate system.

1◦ . When studying axisymmetric problems of flows about spheroidal particles, one uses
the oblate spheroidal coordinates ξ, η; these are related to the cylindrical coordinates ρ =
r cos θ, z = r sin θ by
ρ = c cosh ξ sin η, z = c sinh ξ cos η.

2◦ . The solution of the equation E 2 (E 2 w) = 0 that describes the flow of a fluid about a
oblate spheroid in the direction parallel to the spheroid axis is expressed as

1 2 2 2 [λ/(λ2 + 1)] − [(λ20 − 1)/(λ20 + 1)] arccot λ
w = U c cosh ξ sin η 1 − ,
2 [λ0/(λ20 + 1)] − [(λ20 − 1)/(λ20 + 1)] arccot λ0
λ = sinh ξ, λ0 = sinh ξ0 .
Here w is the stream function, U is the fluid velocity at infinity,
√ and c and λ0 are the
constants related to the spheroid semiaxes a and b (a > b) by c = a2 − b2 and λ0 = b/c.
⊙ Literature: J. Happel and H. Brenner (1965).
11.6 Higher-Order Linear Equations with Constant

Coefficients
◆ Throughout Section 11.6, the following notation is used :
x = {x1 , . . . , xn }, y = {y1 , . . . , yn }, ω = {ω1 , . . . , ωn }, ξ = {ξ1 , . . . , ξn },

p p
|x| = x21 + · · · + x2n , |ω| = ω12 + · · · + ωn2 , ω · x = ω1 x1 + · · · + ωn xn .
11.6.1 Fundamental Solutions. Cauchy Problem

◮ Domain: Rn = {−∞ < xk < ∞; k = 1, . . . , n}. Fundamental solution.
Let P be a constant coefficient linear differential operator such that

∂ X
M
∂ ∂s
P ,..., ≡ as1 ,...,sn s1 , s = s1 + · · · + sn ,
∂x1 ∂xn ∂x1 . . . ∂xsnn
s=0
where s1 , . . . , sn are nonnegative integers, as1 ,...,sn are some constants, and M is the order
of the operator. A generalized function (distribution) Ee (x) = Ee (x1 , . . . , xn ) that satisfies
the equation
∂ ∂
P ,..., Ee (x) = δ(x),
∂x1 ∂xn
where δ(x) = δ(x1 ) . . . δ(xn ) is the Dirac delta function in the n-dimensional Euclidian
space, is called a fundamental solution corresponding to the operator P .
Any constant coefficient linear differential operator has a fundamental solution Ee (x).
term w0 (x) that is an
The fundamental solution is not unique; it is defined up to an additive
arbitrary solution of the homogeneous equation P ∂x∂ 1 , . . . , ∂x∂ n w0 (x) = 0.
A solution of the nonhomogeneous equation

∂ ∂
P ,..., w = Φ(x)
∂x1 ∂xn
with an arbitrary right-hand side has the form
Z
w(x) = Ee (x) ∗ Φ(x), Ee (x) ∗ Φ(x) = Ee (x − y)Φ(y) dVy .
Rn
11.6. Higher-Order Linear Equations with Constant Coefficients 1021
Here dVy = dy1 . . . dyn and the convolution Ee ∗ Φ is assumed to make sense.
⊙ Literature: G. E. Shilov (1965), S. G. Krein (1972), L. Hörmander (1983), V. S. Vladimirov (1988).
◮ Domain: 0 ≤ t < ∞, −∞ < xk < ∞; k = 1, . . . , n. Cauchy problem.

∂

Now let P ∂t , ∂x∂ 1 , . . . , ∂x∂ n be a constant coefficient linear differential operator of or-
der m with respect to t. Then a distribution E (t, x) = E (t, x1 , . . . , xn ) that is a solution of
the homogeneous equation

∂ ∂ ∂
P , ,..., E (t, x) = 0 (1)
∂t ∂x1 ∂xn
and satisfies the initial conditions∗

∂E ∂ m−2 E ∂ m−1 E
E t=0 = 0, = 0, . . . , = 0, = δ(x) (2)
∂t t=0 ∂tm−2 t=0 ∂tm−1 t=0
is called the fundamental solution of the Cauchy problem corresponding to the operator P .
The solution of the Cauchy problem for the linear differential equation

∂ ∂ ∂
P , ,..., w=0 (3)
∂t ∂x1 ∂xn
with the special initial conditions

∂w ∂ m−2 w ∂ m−1 w
wt=0 = 0, = 0, . . . , = 0, = f (x) (4)
∂t t=0 ∂tm−2 t=0 ∂tm−1 t=0
is given by
Z
w(t, x) = E (t, x) ∗ f (x), E (t, x) ∗ f (x) ≡ E (t, x − y)f (y) dVy .
Rn
◮ Solution of the Cauchy problem for general initial conditions.

If the general initial conditions

∂w ∂ m−1 w
w = f0 (x), = f1 (x), ..., = fm−1 (x) (5)
t=0 ∂t t=0 ∂tm−1 t=0
are set, then the solution of Eq. (3) is sought in the form
∂E (t, x) ∂ m−1 E (t, x)

w(t, x) = E (t, x) ∗ ϕ0 (x) + ∗ ϕ1 (x) + · · · + ∗ ϕm−1 (x). (6)
∂t ∂tm−1
∗
The number of initial conditions can be less than m (see Section 11.6.4).
Each term in (6) satisfies Eq. (3), and the functions ϕm−1 , ϕm−2 , . . . , ϕ0 are determined
successively from the linear system
f0 (x) = ϕm−1 (x),

∂ m E (0, x)
f1 (x) = ϕm−2 (x) + ∗ ϕm−1 (x),
∂tm
.................................................
∂ m E (0, x) ∂ m+k−1 E (0, x)
fk (x) = ϕm−k−1 (x) + ∗ ϕm−k (x) + · · · + ∗ ϕm−1 (x),
∂tm ∂tm+k−1
k = 2, . . . , m − 1.
This system of equations is obtained by successively differentiating relation (6) followed

by substituting t = 0 and taking into account the initial conditions (2) and (5).
◮ Nonhomogeneous constant coefficient linear equations of special form.

Consider a nonhomogeneous constant coefficient linear equation of the special form
∂mw X ∂
m−1
∂
k
∂ w
m
+ Pk ,..., = Φ(t, x). (7)
∂t ∂x1 ∂xn ∂tk
k=0
Let E (t, x) = E (t, x1 , . . . , xn ) be the fundamental solution of the Cauchy problem for
the homogeneous equation (7) (Φ ≡ 0) with the initial conditions (2). Then the function
Ee (t, x) = ϑ(t)E (t, x),
where ϑ(t) is the Heaviside unit step function (ϑ = 0 for t ≤ 0 and ϑ = 1 for t > 0), is a
fundamental solution of Eq. (7) with Φ(x, t) = δ(t)δ(x).
The solution of the Cauchy problem for the linear differential equation (7) with the
special initial conditions (4) is given by
Z tZ Z
w(t, x) = E (t − τ, x − y)Φ(τ, y) dVy dτ + E (t, x − y)f (y) dVy .
0 Rn Rn
11.6.2 Elliptic Operators and Elliptic Equations

◮ Homogeneous elliptic differential operator.
A constant coefficient linear homogeneous differential operator of order k has the form
X n
∂ ∂ ∂ s1 ∂ sn X
Pk ,..., ≡ as1 ,...,sn ... , si = k,
∂x1 ∂xn ∂x1 ∂xn
i=1
where s1 , . . . , sn are nonnegative integers. From now on, we adopt the notation

∂ s1 ∂ sn ∂ s1 +···+sn
... ≡ .
∂x1 ∂xn ∂xs11 . . . ∂xsnn
A linear homogeneous differential operator of order k possesses the property

∂ ∂ k ∂ ∂
Pk b ,...,b = b Pk ,..., , b 6= 0 is an arbitrary constant.
∂x1 ∂xn ∂x1 ∂xn
A linear homogeneous differential operator Pk is said to be elliptic if the replacement

of the symbols ∂x∂ 1 , . . . , ∂x∂ n in Pk with the variables ω1 , . . . , ωn gives a polynomial
Pk (ω1 , . . . , ωn ) that does not vanish for real ω 6= 0, i.e., if
X
Pk (ω1 , . . . , ωn ) ≡ as1 ,...,sn ω1s1 . . . ωnsn 6= 0 if ω ∈ Rn , |ω| =
6 0.
A linear differential equation
X n
X
∂ ∂ ∂ s1 +···+sn w
Pk ,..., w≡ as1 ,...,sn s1 = 0, si = k, (1)
∂x1 ∂xn ∂x1 . . . ∂xsnn
i=1
is said to be elliptic if the linear homogeneous differential operator Pk is elliptic.
◮ Elliptic differential operator of the general form.
In general, a constant coefficient linear differential operator of order k has the form
k−1
X
∂ ∂ ∂ ∂ ∂ ∂
Pk ,..., = Pk ,..., + Pi ,..., ,
∂x1 ∂xn ∂x1 ∂xn ∂x1 ∂xn
i=0
where Pk is the principal part of the operator and Pj (j = 0, 1 . . . , k) is a linear homoge-

neous differential operator of order j. The operator Pk is said to be elliptic if its principal
part Pk is elliptic.
A linear differential equation

∂ ∂
Pk ,..., w=0 (2)
∂x1 ∂xn
is said to be elliptic if the linear differential operator Pk is elliptic.
Remark 11.3. A linear elliptic operator and a linear elliptic differential equation can only be of
even order k = 2m, where m is a positive integer.
◮ Fundamental solution of a homogeneous elliptic equation.

The fundamental solution of the homogeneous elliptic equation (1) with k = 2m is given
by
n−1 Z
(−1) 2 dΩn
Ee (x) = n−1
|ω · x|2m−n if n is odd and 2m ≥ n,
4(2π) (2m − n)! Ωn P2m (ω)
n−2 Z
(−1) 2 dΩn
Ee (x) = n
|ω · x|2m−n ln |ω · x| if n is even and 2m ≥ n,
(2π) (2m − n)! Ωn P2m (ω)
n−1 Z
(−1) 2 dΩn
Ee (x) = n−1 δ(n−2m−1) (ω · x) if n is odd and 2m < n,
Ω P2m (ω)
2(2π) 2 n
n Z
(−1) 2 (n − 2m − 1)! dΩn
Ee (x) = n
|ω · x|2m−n if n is even and 2m < n.
(2π) Ωn P2m (ω)
Here the integration is over the surface of the n-dimensional sphere Ωn of unit radius de-
fined by the equation |ω| = 1, ω ·x = ω1 x1 + · · ·+ ωn xn , and P2m (ω) = P2m (ω1 , . . . , ωn ).
The fundamental solution is an ordinary function analytic at any point x 6= 0; this
function is described in a neighborhood of the origin (as |x| → 0) by the relations
(
bn,m |x|2m−n if n is odd or n is even and n > 2m,
Ee (x) = 2m−n
cn,m |x| ln |x| if n is even and n ≤ 2m.
Here bn,m and cn,m are some nonzero constants. If 2m > n, then the fundamental solution
has continuous derivatives of order ≤ 2m − n − 1 at the origin.
◮ Fundamental solution of the general elliptic equation.

The fundamental solution of the general elliptic equation (2) with k = 2m is determined
by the relation Z
Ee (x) = Zω (ω · x, −n) dΩn , (3)
Ωn
where
Z ∞
1 2π n/2
Zω (ξ, λ) = n−1 G(ξ − η, ω)|η|λ dη, σn = .
2 Γ λ+1 −∞ Γ(n/2)
σn π 2
Here the function G(ξ, ω) is the fundamental solution of the constant coefficient linear

d d
P2m ω1 , . . . , ωn G(ξ, ω) = δ(ξ).
dξ dξ
If n is odd, then the fundamental solution (3) can be represented as
Z n−1 n−1
∂ (−1) 2
Ee (x) = An n−1
G(ξ, ω) dΩn , An = n−1 .
Ωn ∂ξ 1× 3×. . .×(n − 2) σn (2π) 2
⊙ Literature: I. M. Gelfand and G. E. Shilov (1959), S. G. Krein (1972).
11.6.3 Hyperbolic Operators and Hyperbolic Equations

∂

Let P ∂t , ∂x∂ 1 , . . . , ∂x∂ n be a constant coefficient linear homogeneous differential opera-
tor of order m with respect to t. The operator P is said to be hyperbolic if, for any real
n
P
numbers ω1 , . . . , ωn such that ωs2 = 1, the mth-order algebraic equation
s=1
P (λ, ω1 , . . . , ωn ) = 0
for λ has m distinct real roots.

Fundamental solution of the Cauchy problem for m ≥ n − 1:
n+1 Z
(−1) 2 [sign(ξ · x + t)]m−1
E (t, x) = n−1
(ξ · x + t)m−n−1 dσH
2(2π) (m − n − 1)! |∇H| sign(ξ · ∇H)
H=0
if n is odd,
n Z
2(−1) 2 (ξ · x + t)m−n−1 ξ · x + t
E (t, x) = ln dσH
n
(2π) (m − n − 1)! |∇H| sign(ξ · ∇H) ξ · x
H=0
if n is even.
Here and in the following, we use the notation
H = P (1, ξ1 , . . . , ξn ),
2
∂H ∂H 2 1/2 ∂H ∂H
|∇H| = + ··· + , ξ · ∇H = ξ1 + · · · + ξn ,
∂ξ1 ∂ξn ∂ξ1 ∂ξn
and dσH is the element of the surface H = 0.
Fundamental solution of the Cauchy problem for m < n − 1:
n+1 Z
(−1) 2 δ(n−m) (ξ · x + t)
E (t, x) = dσH if n is odd,
(2π)n−1 |∇H| sign(ξ · ∇H)
H=0
n Z
(−1) 2 (n − m)! (ξ · x + t)m−n−1
E (t, x) = dσH if n is even.
(2π)n |∇H| sign(ξ · ∇H)
H=0
11.6.4 Regular Equations. Number of Initial Conditions in the Cauchy

Problem
◮ Equations with two independent variables (0 ≤ t < ∞, −∞ < x < ∞).
1◦ . Consider the constant coefficient linear differential equation
∂mw X ∂ ∂k w
m−1
= pk i , (1)
∂tm ∂x ∂tk
k=0
where pk (z) is a polynomial and i2 = −1. Let r = r(σ) be the number of roots (taking into
account their multiplicities) of the characteristic equation
m−1
X
m
λ − pk (σ)λk = 0 (2)
k=0
whose real parts are nonpositive (or bounded above) for a given σ. If r is the same (up to a
set of measure zero) for all σ ∈ (−∞, ∞), then Eq. (1) is said to be regular with regularity
index r.
Classical equations such as the heat, wave, and Laplace equations are regular.
2◦ . In the Cauchy problem for a regular equation (1), one should set r initial conditions of
the form

∂w ∂ r−1 w
wt=0 = f0 (x), = f1 (x), . . . , = fr−1 (x). (3)
∂t t=0 ∂tr−1 t=0
We point out that the regularity index r can generally differ from the order m of
the equation with respect to t. In particular, for the two-dimensional Laplace equation
∂tt w = −∂xx w, we have r = 1 and m = 2; here y is replaced by t and the first boundary
value problem in the upper half-plane t ≥ 0 is considered. For the heat equation ∂t w = ∂xx w
and wave equation ∂tt w = ∂xx w, we have r = m = 1 and r = m = 2, respectively.
Example 11.6. The regularity indices for some fourth-order equations are as follows:
∂2w ∂4w ∂2w ∂4w

2
− a2 4 = 0 (r = 1, m = 2); 2
+ a2 4 = 0 (r = 2, m = 2);
∂t ∂x ∂t ∂x
2 2
∂ ∂2 ∂4w ∂ 4
w
+ w=0 (r = 2, m = 4); − a2 4 = 0 (r = 3, m = 4).
∂t2 ∂x2 ∂t 4 ∂x
3◦ . The special solution E = E (t, x) that satisfies the initial conditions

∂E ∂ r−2 E ∂ r−1 E
E t=0 = 0, = 0, . . . , = 0, = δ(x) (4)
∂t t=0 ∂tr−2 t=0 ∂tr−1 t=0
is called the fundamental solution.

The fundamental solution can be found by applying the Fourier transform in the space
variable to Eq. (1) (with w = E ) and the initial conditions (4).
Example 11.7. Consider the polyharmonic equation
n
∂2 ∂2
+ w = 0. (5)
∂t2 ∂x2
∂2 ∂
2
Taking into account the representation ∂x2 = − i ∂x , we rewrite the characteristic equation (2)
in the form
(λ2 − σ 2 )n = 0.
It has only one solution, λ = −|σ|, whose real part is nonpositive. Considering the multiplicity of
the root, we find that the regularity index r is equal to n.
In Eq. (5) with w = E and the initial conditions (4) with r = n, we make the Fourier transform
with respect to the space variable,
Z ∞
U (t, σ) = eiσx E (t, x) dx.
−∞
As a result, we arrive at the ordinary differential equation

2 n
d 2
− σ U =0 (6)
dt2
(n−2) (n−1)
U t=0 = 0, Ut′ t=0 = 0, ..., Ut t=0
= 0, Ut t=0
= 1. (7)
The bounded solution of problem (6), (7) is given by

tn−1 −|σ|t
U (t, σ) = e .
(n − 1)!
By applying the inverse Fourier transform, we obtain the fundamental solution of the polyharmonic
equation in the form
Z ∞ Z ∞
1 −iσx 1 tn−1
E (t, x) = e U (t, σ) dσ = e−iσx−|σ|t dσ
2π −∞ 2π (n − 1)! −∞
Z ∞ Z 0
1 tn−1 −iσx−σt −iσx+σt
= e dσ + e dσ
2π (n − 1)! 0 −∞

1 tn−1 1 1 1 tn−1 t
= + = .
2π (n − 1)! t + ix t − ix π (n − 1)! t + x2
2
4◦ . For general initial conditions of the form (3), the solution of Eq. (1) is determined on
the basis of the fundamental solution from the relation
∂E (t, x) ∂ r−1 E (t, x)
w(t, x) = E (t, x) ∗ ϕ0 (x) + ∗ ϕ1 (x) + · · · + ∗ ϕr−1 (x). (8)
∂t ∂tr−1
Each term in (8) satisfies Eq. (1), and the functions ϕr−1 , ϕr−2 , . . . , ϕ0 are calculated
successively by solving the linear system
f0 (x) = ϕr−1 (x),

∂ r E (0, x)
f1 (x) = ϕr−2 (x) + ∗ ϕr−1 (x),
∂tr
...............................................
∂ r E (0, x) ∂ r+k−1 E (0, x)
fk (x) = ϕr−k−1 (x) + r
∗ ϕr−k (x) + · · · + ∗ ϕr−1 (x),
∂t ∂tr+k−1
k = 2, . . . , r − 1.
This system of equations is obtained by successively differentiating relation (8) followed
by substituting t = 0 and by taking into account the initial conditions (3) and (4).
In the special case of f0 (x) = f1 (x) = · · · = fr−2 (x) = 0, one should set ϕ0 (x) =
fr−1 (x) and ϕ1 (x) = · · · = ϕr−1 (x) = 0 in (8).
◮ Equations with many independent variables (0 ≤ t < ∞, x ∈ Rn ).

Solving the Cauchy problem for the constant coefficient linear differential equation

∂ ∂ ∂
P , ,..., w=0 (9)
∂t ∂x1 ∂xn
with arbitrarily many space variables x1 , . . . , xn can be reduced to solving the Cauchy
problem for an equation with one space variable ξ. Take the auxiliary linear differential
operator
∂ ∂ ∂ ∂ ∂
Pω , ≡P , ω1 , . . . , ωn
∂t ∂ξ ∂t ∂ξ ∂ξ
that depends on two independent variables t and ξ so that the Cauchy problem for the
equation
∂ ∂
Pω , v=0 (10)
∂t ∂ξ
is well posed. Then the fundamental solution of the Cauchy problem for the original equa-
tion (9) is given by Z
E (t, x) = vω (t, ω · x, −n) dΩn .
Ωn
Here
Z ∞
1 2π n/2
vω (t, ξ, λ) = n−1 Gω (t, ξ − η)|η|λ dη, σn = ,
2 Γ λ+1 −∞ Γ(n/2)
σn π 2
where Gω (t, ξ) is the fundamental solution of the Cauchy problem for the auxiliary equa-
tion (10).
If the number of space variables is odd, one can use the simpler formula
n−1
n−1
Z
(−1) 2
2 ! dn−1
E (t, x) = n−1 Gω (t, ξ) dΩn , ξ = ω · x.
Ωn dξ n−1
σn π 2 (n − 1)!
Remark 11.4. The above relations hold for all equations for which the Cauchy problem is well
posed.
◮ Stationary homogeneous regular equations (x ∈ Rn ).

A linear differential operator Pk ∂x∂ 1 , . . . , ∂x∂ n is said to be regular if it is homogeneous
and if the gradient of the function Pk (ω1 , . . . , ωn ) on the set defined by the equation
Pk (ω1 , . . . , ωn ) = 0
is everywhere nonzero whenever |ω| =

6 0.

The fundamental solution of the linear regular PDE Pk ∂x∂ 1 , . . . , ∂
∂xn w = 0 generated
by the linear regular differential operator Pk is expressed as
Z
ϕnk (ω · x)
Ee (x) = dΩn , (11)
Ωn Pk (ω)
where the function ϕnk (z) is defined by

n−2
(−1) 2
ϕnk (z) = z k−n ln |z| if n is even and k ≥ n,
(2π)n (k − n)!
n+2k
(−1) 2 (n − k − 1)! k−n
ϕnk (z) = z if n is even and k < n,
(2π)n
n−1
(−1) 2
ϕnk (z) = z k−n sign z if n is odd and k ≥ n,
4(2π)n−1 (k − n)!
n−1
(−1) 2 (n−k−1)
ϕnk (z) = δ (z) if n is odd and k < n.
2(2π)n−1
The integral in (11) is understood in the sense of its regularized value; i.e.,
Z
ϕnk (ω · x)
Ee (x) = lim Iε (x), Iε (x) = dΩ(ε)
n ,
ε→0 Ωn
(ε) P k (ω)
(ε)
where Ωn is the set of points on the sphere of unit radius for which |Pk (ω)| > ε.
11.6.5 Some Equations with Two Independent Variables Containing

the First Derivative in t
∂w ∂ 2n w
1. +a = Φ(x, t), n = 1, 2, . . . .
∂t ∂x2n
2n 2n−1
X
x
w(x, t) = b − at + ck xk ,
(2n)!
k=0
2n+2
2n+1
2x x
w(x, t) = b − ax2 t + c − axt ,
(2n + 2)! (2n + 1)!
2n

w(x, t) = exp(−aλ ) b exp(−λx) + c exp(λx) ,

w(x, t) = exp a(−1)n+1 λ2n b cos(λx) + c sin(λx) ,

w(x, t) = exp −aλ2n t b exp −λx + c exp λx
n−1
X πk
+ exp(λx cos βk )[bk cos(λx sin βk ) + ck sin(λx sin βk )] , βk = ,
n
k=1
where b, bk , c, ck , and λ are arbitrary constants.

2◦ . Formal series solution for Φ ≡ 0:

∞
X ak tk d 2nk f (x)
w(x, t) = f (x) + (−1)k ,
k! dx2nk
k=1
condition w(x, 0) = f (x). If the function f (x) is a polynomial of degree m, then the
solution is a polynomial in x of degree m as well.
3◦ . Let a = (−1)n α2 , where α > 0.
Z ∞ Z
1 1 ∞
E (x, t) = exp −α2 tξ 2n + ixξ dξ = exp −α2 tξ 2n cos(xξ) dξ.
2π −∞ π 0
By using the self-similar variables
1 − 1
ζ = ξ 2nα2 t 2n , z = x 2nα2 t 2n ,
one can reduce the fundamental solution to the form

Z ∞ 1
1 1
2 − 2n
E (x, t) = 2nα t I2n (z), I2n (z) = exp − ζ 2n cos(zζ) dζ.
π 0 2n
The function I2n (z) is known as the hyper-Airy function of even index. These functions
are bounded, integrable, and infinitely differentiable.
The function I2n (z) can be represented in the form of the series
∞
X (2n)νk −1 2k + 1
I2n (z) = (−1)k Γ(νk )z 2k , νk = ,
(2k)! 2n
k=0
where Γ(ν) is the gamma function.

The leading term of the asymptotic expansion for I2n (z), n > 1, as z → ±∞ has the
form
1/2 n−1
2n

. 2 − 2n−1 2n − 1 n−1
I2n (z) = |z| exp − cos π |z| 2n−1
π(2n − 1) 2n 2n − 1
2n

2n − 1 n−1 π n−1
× cos sin π |z| 2n−1 − .
2n 2n − 1 2 2n − 1
w = f (x) at t = 0.
Solution with a = (−1)n α2 :
Z ∞ Z tZ ∞
w(x, t) = E (x − y, t)f (y) dy + E (x − y, t − τ )Φ(y, τ ) dy dτ,
−∞ 0 −∞
where the function E (x, t) is defined in Item 3◦ .

⊙ Literature: A. Winter (1957), R. B. Paris and A. D. Wood (1987), A. Karlsson and S. Ritke (1998), G. Kris-
tensson, A. Karlsson, and S. Rikte (2002), N. A. Kudryashov and D. I. Sinelshchikov (2014).
∂w ∂ 2n+1 w
2. +a = Φ(x, t), n = 1, 2, . . . .
∂t ∂x2n+1
2n+1 X 2n
x
w(x, t) = b − at + ck xk ,
(2n + 1)!
k=0
2n+3
2n+2
2x x
w(x, t) = b − ax2 t + c − axt ,
(2n + 3)! (2n + 2)!
w(x, t) = b exp(λx − aλ2n+1 t) + c exp(−λx + aλ2n+1 t),

w(x, t) = b sin λx + (−1)n+1 aλ2n+1 t + c cos λx + (−1)n+1 aλ2n+1 t ,

w(x, t) = exp −aλ2n+1 t b exp λx
X n
2πk
+ exp(λx cos βk )[bk cos(λx sin βk ) + ck sin(λx sin βk )] , βk = ,
2n + 1
k=1
where b, bk , c, ck , and λ are arbitrary constants.
2◦ . Formal series solution for Φ ≡ 0:

∞
X ak tk d 2nk+k f (x)
w(x, t) = f (x) + (−1)k ,
k! dx2nk+k
k=1
condition w(x, 0) = f (x). If the function f (x) is a polynomial of degree m, then solution
is a polynomial in x of degree m as well.
3◦ . Let a = (−1)n+1 α2 , where α > 0.

Z ∞ Z
1 2 2n+1 1 ∞
E (x, t) = exp i(α tξ + xξ) dξ = cos(α2 tξ 2n+1 + xξ) dξ.
2π −∞ π 0
1 − 1
ζ = ξ (2n + 1)α2 t 2n+1 , z = x (2n + 1)α2 t 2n+1 ,
1 − 1
E (x, t) = (2n + 1)α2 t 2n+1 I2n+1 (z),
Zπ ∞ 1
I2n+1 (z) = cos ζ 2n+1 + zζ dζ
0 2n + 1
Z ∞ 1
2nζ 2n−1 2n+1
= sin ζ + zζ dζ.
0 (z + ζ 2n )2 2n + 1
The family of integrals I2n+1 (z) is called the hyper-Airy function of odd index. The
leading term of the asymptotic expansion for I2n+1 (z) as z → ∞ has the form

. 1 − 2n−1 2n π n−1 2n+1
I2n+1 (z) = √ z 4n exp − cos z 2n
πn 2n + 1 2 n

2n π n−1 2n+1 π n−1
× cos sin z 2n − .
2n + 1 2 n 2 2n
For I2n+1 (z) as z → −∞ we have

. 1 2n−1 2n 2n+1 π
I2n+1 (z) = √ |z|− 4n cos |z| 2n − .
πn 2n + 1 4
w = f (x) at t = 0.
Solution with a = (−1)n+1 α2 :

Z ∞ Z tZ ∞
−∞ 0 −∞
where the function Ee (x, t) is defined in Item 3◦ .

⊙ Literature: A. Winter (1957), R. B. Paris and A. D. Wood (1987), A. Karlsson and S. Ritke (1998), G. Kris-
tensson, A. Karlsson, and S. Rikte (2002), N. A. Kudryashov and D. I. Sinelshchikov (2014).
n
X
∂w ∂ 2k w
3. + ak = Φ(x, t).
∂t ∂x2k
k=0
n
X
w = exp(−λt)[A exp(−βx) + B exp(βx)], λ= ak β 2k ,
k=0
n
X
w = exp(−λt)[A cos(βx) + B sin(βx)], λ= (−1)k ak β 2k ,
k=0

2◦ . The condition
n
X
P (ξ) = (−1)k ak ξ 2k ≥ 0 (1)
k=0
is assumed to be met for all real ξ. The condition is satisfied, say, if ak = (−1)k bk with
bk ≥ 0 for all k.
Z ∞ Z
1 1 ∞
E (x, t) = exp −tP (ξ) + ixξ dξ = exp −tP (ξ) cos(xξ) dξ. (2)
2π −∞ π 0

w = f (x) at t = 0.
Solution:
Z ∞ Z tZ ∞
−∞ 0 −∞

Remark 11.5. One can weaken condition (1) by replacing it with the condition (−1)n an > 0.
n
X
∂w ∂ 2k+1 w
4. + ak = Φ(x, t).
∂t ∂x2k+1
k=0

n
X
w = A exp(βx − λt) + B exp(−βx + λt), λ= ak β 2k+1 ,
k=0
n
X
w = A cos(βx + λt) + B sin(βx + λt), λ= (−1)k+1 ak β 2k+1 ,
k=0

n
X
P (ξ) = (−1)k+1 ak ξ 2k+1 ≥ 0
k=0
is assumed to be met for real ξ ≥ 0. The condition is satisfied, say, if ak = (−1)k+1 bk with
bk ≥ 0 for all k.
Z ∞ Z
1 1 ∞
E (x, t) = exp itP (ξ) + ixξ dξ = cos tP (ξ) + xξ dξ.
2π −∞ π 0

w = f (x) at t = 0.
Solution:
Z ∞ Z tZ ∞
−∞ 0 −∞

n
X
∂w ∂kw
5. + ak = Φ(x, t).
∂t ∂xk
k=0
For n = 2m and a1 = a3 = · · · = a2m−1 = 0, see equation 11.6.5.3; for n = 2m + 1 and

a0 = a2 = · · · = a2m = 0, see equation 11.6.5.4.
n
X
w = exp(βx − λt), λ= ak β k ,
k=0

w = exp −P (β)t cos βx + Q(β)t ,

w = exp −P (β)t sin βx + Q(β)t ,
where β is an arbitrary constant and
2k≤n
X 2k+1≤n
X
P (β) = (−1)k a2k β 2k , Q(β) = (−1)k+1 a2k+1 β 2k+1 .
k=0 k=0
2k≤n
X
P (ξ) = (−1)k a2k ξ 2k > 0
k=0
is assumed to be met for real ξ > 0.

Z
1 ∞
E (x, t) = exp −tP (ξ) cos tQ(ξ) + xξ dξ,
π 0
2k+1≤n
X
Q(ξ) = (−1)k+1 a2k+1 ξ 2k+1 .
k=0

w = f (x) at t = 0.
Solution:
Z ∞ Z tZ ∞
−∞ 0 −∞

⊙ Literature: S. G. Krein (1972), V. S. Vladimirov, V. P. Mikhailov, A. A. Vasharin et al. (1974).
n
X
∂w ∂3w ∂ 2k w
6. −b + ak = Φ(x, t).
∂t ∂t∂x2 ∂x2k
k=0

n
X
1
w = exp(−λt)[A exp(−βx) + B exp(βx)], λ= ak β 2k ,
1 − bβ 2
k=0
n
X
1
w = exp(−λt)[A cos(βx) + B sin(βx)], λ= (−1)k ak β 2k ,
1 + bβ 2
k=0

P
2◦ . Let b > 0, and let the condition P (ξ) = nk=0 (−1)k ak ξ 2k ≥ 0 be satisfied for all real
ξ. The condition is satisfied, say, if ak = (−1)k ck with ck ≥ 0 for all k.
Z ∞
1 1 P (ξ)
Ee (x, t) = exp − t + ixξ dξ
2π −∞ 1 + bξ 2 1 + bξ 2
Z
1 ∞ 1 P (ξ)
= exp − t cos(|x|ξ) dξ.
π 0 1 + bξ 2 1 + bξ 2

w = f (x) at t = 0.
Solution:
Z tZ ∞ Z ∞
w(x, t) = Ee (x − y, t − τ )Φ(y, τ ) dy dτ + Ee (x − y, t) f (y) − bf ′′ (y) dy.
0 −∞ −∞
11.6.6 Some Equations with Two Independent Variables Containing

the Second Derivative in t
∂2w ∂ 2n w
1. +a = Φ(x, t).
∂t2 ∂x2n
p
w = [A exp(−βx) + B exp(βx)][C cos(λt) + D sin(λt)], λ = aβ 2n ,
p
w = [A exp(−βx) + B exp(βx)][C exp(−λt) + D exp(λt)], λ = −aβ 2n ,
p
w = [A cos(βx) + B sin(βx)][C cos(λt) + D sin(λt)], λ = (−1)n aβ 2n ,
p
w = [A cos(βx) + B sin(βx)][C exp(−λt) + D exp(λt)], λ = (−1)n+1 aβ 2n ,
where A, B, C, D, and β are arbitrary constants. (In all cases, the sign of the coefficient a
is chosen to ensure that the radicands are positive).
2◦ . General solution with Φ(x, t) = 0 and a = −α2 < 0:
w = w1 + w2 ,
where w1 and w2 are arbitrary solutions of two simpler nth-order equations
∂w1 ∂ n w1 ∂w2 ∂ n w2
+α = 0, −α = 0.
∂t ∂xn ∂t ∂xn
See equation 11.6.5.1 or 11.6.5.2 for information on the solution of these equations de-
pending on whether n is odd or even.
3◦ . Formal series solution for Φ(x, t) = 0:
∞
X k 2k d 2nk f (x) ∞
X k 2k+1 d 2nk g(x)
ka t k a t
w(x, t) = (−1) + (−1) ,
(2k)! dx2nk (2k + 1)! dx2nk
k=0 k=0
f (x). This solution satisfies the initial conditions w(x, 0) = f (x) and ∂t w(x, 0) = g(x).
If the functions f (x) and g(x) are polynomials of degree ≤ m, then the solution is a
polynomial in x of degree ≤ m as well.
4◦ . Let a = (−1)n α2 , where α > 0.
Z ∞ Z
1 sin(αtξ n ) 1 ∞ sin(αtξ n )
E (x, t) = exp(ixξ) dξ = cos(xξ) dξ.
2π −∞ αξ n π 0 αξ n
ζ = ξ(αt)1/n , z = x(αt)−1/n ,

Z ∞
t(n−1)/n sin(ζ n )
E (x, t) = I(z), I(z) = cos(zζ) dζ.
πα1/n 0 ζn
5◦ . Domain: −∞ < x < ∞. Cauchy problem with a = (−1)n α2 , where α > 0.

w = f (x) at t = 0,
∂t w = g(x) at t = 0.
Solution:
Z tZ ∞
0
Z−∞
∞ Z ∞
∂
+ E (x − y, t)f (y) dy + E (x − y, t)g(y) dy.
∂t −∞ −∞
n
X
∂2w ∂ 2k w
2. + ak = Φ(x, t).
∂t2 ∂x2k
k=0
1◦ . Particular solutions of the homogeneous equation for Φ(x, t) = 0:
X
n 1/2
w = [A exp(−βx) + B exp(βx)] sin(λt + C), λ= ak β 2k ,
k=0
Xn 1/2
2k
w = [A exp(−βx) + B exp(βx)] exp(±λt), λ= − ak β ,
k=0
X
n 1/2
k 2k
w = sin(βx + A) sin(λt + B), λ= (−1) ak β ,
k=0
Xn 1/2
k 2k
w = sin(βx + A)[B exp(−λt) + C exp(λt)], λ= − (−1) ak β ,
k=0
where A, B, C, and β are arbitrary constants. (In all cases, the radicands are assumed to
be positive.)
P
2◦ . Let b > 0, and let the condition P (ξ) = nk=0 (−1)k ak ξ 2k ≥ 0 be satisfied for all real
ξ. The condition is satisfied if, say, ak = (−1)k ck with ck ≥ 0 for all k.
Z ∞
1 1 p
E (x, t) = p sin t P (ξ) exp(ixξ) dξ
2π −∞ P (ξ)
Z
1 ∞ 1 p
= p sin t P (ξ) cos(|x|ξ) dξ.
π 0 P (ξ)

w = f (x) at t = 0,
∂t w = g(x) at t = 0.
Solution:
Z tZ ∞
0
Z−∞
∞ Z ∞
∂
+ E (x − y, t)f (y) dy + E (x − y, t)g(y) dy.
∂t −∞ −∞
n
X
∂2w ∂4w ∂ 2k w
3. −b + ak = Φ(x, t).
∂t2 ∂t2 ∂x2 ∂x2k
k=0
1◦ . Particular solutions of the homogeneous equation for Φ(x, t) = 0:
X n 1/2
1 2k
w = [A exp(−βx) + B exp(βx)] sin(λt + C), λ = ak β ,
1 − bβ 2
k=0
X n 1/2
1 2k
w = [A exp(−βx) + B exp(βx)] exp(±λt), λ= ak β ,
bβ 2 − 1
k=0
X n 1/2
1 k 2k
w = sin(βx + A) sin(λt + B), λ= (−1) ak β ,
1 + bβ 2
k=0
Xn 1/2
1 k 2k
w = sin(βx + A)[B exp(−λt) + C exp(λt)], λ = − (−1) ak β ,
1 + bβ 2
k=0
where A, B, C, and β are arbitrary constants.

2◦ . Let b > 0, and let the condition
n
X
P (ξ) = (−1)k ak ξ 2k ≥ 0
k=0
be satisfied for all real ξ. The condition is satisfied if, say, ak = (−1)k ck with ck ≥ 0 for
all k.
Z ∞ s
1 1 P (ξ)
Ee (x, t) = p sin t exp(ixξ) dξ
2π −∞ (1 + bξ 2 )P (ξ) 1 + bξ 2
Z s
1 ∞ 1 P (ξ)
= p sin t cos(|x|ξ) dξ.
π 0 (1 + bξ 2 )P (ξ) 1 + bξ 2

w = f (x) at t = 0,
∂t w = g(x) at t = 0.
Solution:
Z tZ ∞
w(x, t) = Ee (x − y, t − τ )Φ(y, τ ) dy dτ
0 −∞
Z Z ∞
∂ ∞ ′′

+ Ee (x − y, t) f (y) − bf (y) dy + Ee (x − y, t) g(y) − bg′′ (y) dy.
∂t −∞ −∞
11.6.7 Other Equations with Two Independent Variables

∂ ∂ n
1. −a w = 0, n = 1, 2, . . .
∂t ∂x
n−1
X
w(x, t) = tk fk (x + at),
k=0
n−1
X
w(x, t) = xk fk (x + at),
k=0
where the fk = fk (z) are arbitrary functions.

tn−1
E (x, t) = δ(x + at).
(n − 1)!

∂ ∂2 n
2. − w = 0, n = 1, 2, . . .
∂t ∂x2
n−1
X
w(x, t) = tk uk (x, t),
k=0
n−1
X
w(x, t) = xk uk (x, t),
k=0
where the uk = uk (x, t) are arbitrary solutions of the heat equations ∂t uk − ∂xx uk = 0.

1 n−3/2 x2
E (x, t) = √ t exp − .
2 π (n − 1)! 4t


∂w ∂ n−2 w ∂ n−1 w
w = 0, = 0, ..., = 0, = f (x).
t=0 ∂t t=0 ∂tn−2 t=0 ∂tn−1 t=0
Solution: Z ∞
w(x, t) = f (ξ)E (x − ξ, t) dξ.
−∞


∂2 ∂2 n
3. − w = 0, n = 1, 2, . . .
∂t2 ∂x2
n−1
X
w(x, t) = tk fk (x + t) + gk (x − t) ,
k=0
n−1
X
w(x, t) = xk fk (x + t) + gk (x − t) ,
k=0
where fk = fk (y) and gk = gk (z) are arbitrary functions.
n−1
X (2t)k (x − t)2n−k−2
(−1)n−1
E (x, t) = n sign(x − t)
4 (n − 1)! k!(n − k − 1)!
k=0
n−1
X
n (−2t)k (x + t)2n−k−2
+ (−1) sign(x + t) .
k!(n − k − 1)!
k=0

∂2 ∂2 n
4. + w = 0, n = 1, 2, . . .
∂x2 ∂y 2
This is the polyharmonic equation of order n with two independent variables.
n−1
X p
w(x, y) = r 2k uk (x, y), r= x2 + y 2 ,
k=0
n−1
X
w(x, y) = xk uk (x, y),
k=0
where the uk (x, y) are arbitrary harmonic functions (∆uk = 0). In the second relation, xk
can be replaced by y k .
2◦ . Domain: −∞ < x < ∞, −∞ < y < ∞. Fundamental solution:
1 p
2n−2
Ee (x, y) = 2 r ln r, r = x2 + y 2 .
π22n−1 (n − 1)!
3◦ . Domain: −∞ < x < ∞, 0 ≤ y < ∞. Boundary value problem.

∂w ∂ n−2 w ∂ n−1 w

w y=0 = 0, = 0, . . . , = 0, = f (x).
∂y y=0 ∂y n−2 y=0 ∂y n−1 y=0
Solution:
Z ∞
1 yn
w(x, y) = f (ξ)G(x − ξ, y) dξ, G(x, y) = .
−∞ π(n − 1)! x2 + y 2
⊙ Literature: G. E. Shilov (1965), L. D. Faddeev (1998).

∂2 ∂2 n
5. + w = Φ(x, y), n = 1, 2, . . .
∂x2 ∂y 2
This is a nonhomogeneous polyharmonic equation of order n with two independent vari-

ables.
Z ∞Z ∞
1
w(x, y) = Φ(ξ, η)[(x − ξ)2 + (y − η)2 ]n−1 ln[(x − ξ)2 + (y − η)2 ] dξ dη,
kn −∞ −∞
kn = π22n [(n − 1)!]2 .
The general solution is given by the sum of any particular solution of the nonhomogeneous
equation and the general solution of the homogeneous equation (see equation 11.6.7.4,
Item 1◦ ).
m
X ∂2 ∂2
6. ak ∆k w = 0, ∆≡ + .
∂x2 ∂y 2
k=0
m
X
w(x, y) = un (x, y),
n=1
where the un are solutions of the Helmholtz equations ∆un − λn un = 0 and the λn are
m
P
roots of the characteristic equation ak λk = 0.
k=0
11.6.8 Equations with Three and More Independent Variables

n
X ∂2
1. ∆m
n w = 0, ∆n = .
∂x2k
k=1
This is the polyharmonic equation of order m with n independent variables. For m = 1,

see Sections 9.1 and 10.1. For m = 2, see Section 11.5.1. For n = 2, see equation 11.6.7.4.
m−1
X
w(x) = xsj us (x), j = 1, 2, . . . , n,
s=0
where the us (x) are arbitrary harmonic functions (∆n us = 0).

2◦ . Fundamental solution for m ≥ 1 and n ≥ 3:
(
bn,m |x|2m−n if n is odd or n is even and n > 2m,
Ee (x) = 2m−n
cn,m |x| ln |x| if n is even and n ≤ 2m.
Here
an,m Γ(n/2)
bn,m = , an,m = ,
(2 − n)(4 − n) . . . (2m − n) 2m (m − 1)! π n/2
an,m
cn,m = ,
(2 − n)(4 − n) . . . (2m0 − 2 − n)(2m0 + 2 − n)(2m0 + 4 − n) . . . (2m − n)
where m0 = n/2. The expression of the coefficient cn,m can be obtained formally from the
expression of bn,m by removing the zero factor (2m0 − n) from the denominator.
m
X
∂w
2. + as ∆sn w = Φ(x, t).
∂t
s=1
P ∂2
Here x = (x1 , . . . , xn ) and ∆n = ns=1 ∂x 2.
s
Let an initial-boundary value problem for this equation be stated in a bounded do-
main V . Particular solutions of the nonhomogeneous equation can be found as follows.
Consider the auxiliary problem
∆u + λu = 0 (1)
for the Helmholtz equation in a bounded domain V1 containing V (moreover, the bound-
aries of V and V1 do not have common points) with the homogeneous boundary condition
of the first kind
u|x∈S1 = 0. (2)
Let λk and uk (x) be the eigenvalues and eigenfunctions of problem (1)–(2). All eigen-
values are positive. We arrange them in the nondescending order λ1 ≤ λ2 ≤ . . . . It is well
known that the eigenfunctions are orthogonal,
Z
ui (x)uj (x) dv1 = 0 if λi 6= λj ,
V1
and can be chosen to be real. The eigenfunctions corresponding to coinciding eigenvalues

λi = λj can be chosen to be orthogonal as well. The system of eigenfunctions uk is
complete in L2 (V1 ).
Assume that the function Φ(x, t) is defined in V1 . Let us expand it in the eigenfunctions
uk ,
X∞
Φ(x, t) = Ak (t)uk (x),
k=1 (3)
Z Z
1 2 2
Ak (t) = Φ(x, t)uk (x) dv1 , kuk k = uk (x) dv1 .
kuk k2 V1 V1
We seek a particular solution of the original equation in the form

∞
X
w= fk (t)uk (x). (4)
k=1
Since the eigenfunctions uk satisfy Eq. (1), we have ∆s uk = (−λk )s uk . In view of the
representation (3), for the functions fk = fk (t) we obtain the first-order linear ODEs
m
X
fk′ + βk f = Ak (t), βk = as (−λk )s .
s=1
The general solutions of these equations have the form

Z t
fk (t) = Ck e−βk t + e−βk t eβk τ Ak (τ ) dτ, (5)
0
where the Ck are arbitrary constants. Formulas (4)–(5) give a particular solution of the
original inhomogeneous equation.
We point out that for V1 one can take a domain of simple geometric shape (for example,
an n-dimensional cube or sphere) for which the eigenvalues and eigenfunctions are known.
Remark 11.6. Instead of the boundary condition (2) of the first kind, one can use homogeneous
boundary conditions of the second or third kind.
m
X
∂2w
3. + as ∆sn w = Φ(x, t).
∂t2
s=1
P ∂2
Here x = (x1 , . . . , xn ) and ∆n = ns=1 ∂x 2.
s
Let λk and uk (x) be the eigenvalues and eigenfunctions of the auxiliary problem for the
homogeneous Helmholtz equation described in 11.6.8.2 (see Eqs. (1)–(2)).
Let us expand Φ(x, t) in a series in the eigenfunctions uk (x). We seek a particular
solution of the equation in question in the form
∞
X
w= fk (t)uk (x).
k=1
As result, for the functions fk = fk (t) we obtain the second-order linear ODEs
m
X Z
1
fk′′ + βk f = Ak (t), βk = s
as (−λk ) , Ak (t) = Φ(x, t)uk (x) dv1 .
s=1
kuk k2 V1
The general solution of these equations has the form

 Z

 1 t

 C 1 cos(b k t) + C 2 sin(bk t) + Ak (τ ) sin[bk (t − τ )] dτ if βk = b2k > 0,

 bk 0Z

1 t
fk = C1 cosh(bk t) + C2 sinh(bk t) + Ak (τ ) sinh[bk (t − τ )] dτ if βk = −b2k < 0,

 Z t bk 0




C1 t + C2 + Ak (τ )(t − τ ) dτ if βk = 0,
0

n
X ∂nw
4. ak = 0.
∂tk ∂xn−k
k=0
1◦ . Let the algebraic equation
n
X
ak λk = 0 (1)
k=0
have m distinct real roots λ1 , . . . , λm (which corresponds to the homogeneous hyperbolic
equation). Then the function
Xm
w= fk (x + λk t),
k=1
where the fk (zk ) are arbitrary functions, is a solution of the equation.

2◦ . Let λs be a real root of Eq. (1) of multiplicity r. Then the function
r
X
w= xk gk (x + λs t),
k=1
where gk (z) are arbitrary functions, is a solution of the equation.

3◦ . Let λp,1 = bp + icp and λp,2 = bp − icp be complex conjugate roots of Eq. (1) (p =
1, . . . , q). Then the original equation has solutions wp that satisfy second-order elliptic
equations of the form
∂ 2 wp ∂ 2 wp 2
2
2 ∂ wp
− 2b p + (b p + cp ) = 0. (2)
∂t2 ∂t∂x ∂x2
Each equation (2) can be reduced to the Laplace equation
∂ 2 Up ∂ 2 Up
+ =0
∂ξp2 ∂ηp2
by the transformation
wp = Up (ξp , ηp ), ξp = (b2p + c2p )t + bp x, ηp = cp x.

m
X
5. ak Lk [w] = 0.
k=0
Here L is any constant coefficient linear differential operator with arbitrarily many inde-
pendent variables x1 , . . . , xn .
m
X
w(x1 , . . . , xn ) = Cs us (x1 , . . . , xn ),
s=1
where the us are solutions of the equations L[us ] − λs us = 0, the λs are roots of the
m
P
characteristic equation ak λk = 0, and the Cs are arbitrary constants.
k=0
11.7. Higher-Order Linear Equations with Variable Coefficients 1045
m
X
6. ak Lk M m−k [w] = 0.
k=0
Here L and M are any constant coefficient linear differential operators with arbitrarily
many independent variables x1 , . . . , xn .
Solution:
m
X
w(x1 , . . . , xn ) = Cs us (x1 , . . . , xn ),
s=1
where the us are solutions of the equations L[us ] − λs M [us ] = 0, the λs are roots of the
m
P
characteristic equation ak λk = 0, and the Cs are arbitrary constants.
k=0
7. Lm . . . L2 L1 w = 0.
Here L1 , . . . Lm are any linear differential operator with constant coefficients and arbitrarily
many independent variables x1 , . . . , xn .
Solution:
Xm
w= wk ,
k=1
where Lk wk = 0 and k = 1, . . . , m. (If Li 6= Lj for all i and j, then this is the general
solution.)
11.7 Higher-Order Linear Equations with Variable

Coefficients
11.7.1 Equations Containing the First Time Derivative
◮ Statement of the problem for an equation with two independent variables.
Consider the linear nonhomogeneous partial differential equation
∂w
− Lx,t [w] = Φ(x, t), (1)
∂t
where Lx,t is a general linear differential operator of order n with respect to the space
variable x,
Xn
∂k w
Lx,t [w] ≡ ak (x, t) k , (2)
∂x
k=0
whose coefficients ak = ak (x, t) are sufficiently smooth functions of both arguments for
t ≥ 0 and x1 ≤ x ≤ x2 . The subscripts x and t indicate that the operator Lx,t depends on
the variables x and t.
We set the initial condition
w = f (x) at t=0 (3)

and the general nonhomogeneous boundary conditions

n−1
X (1) ∂k w
Γ(1)
m [w] ≡ bmk (t) (1)
= gm (t) at x = x1 (m = 1, . . . , s),
∂xk
k=0
(4)
n−1
X (2) ∂k w
Γ(2)
m [w] ≡
(2)
bmk (t) k = gm (t) at x = x2 (m = s + 1, . . . , n),
∂x
k=0
(1)
where s ≥ 1 and n ≥ s + 1. We assume that both sets of the boundary forms Γm [w]
(2)
(m = 1, . . . , s) and Γm [w] (m = s + 1, . . . , n) are linearly independent, which means that
for any nonzero ψm = ψm (t) the following relations hold:
s
X n
X
ψm (t)Γ(1)
m [w] 6≡ 0, ψm (t)Γ(2)
m [w] 6≡ 0.
m=1 m=s+1
In what follows, we deal with the nonstationary boundary value problem (1)–(4). It is
assumed that there exist solutions of the problems considered in what follows.
◮ Homogeneous boundary conditions. The Green’s function.

The solution of Eq. (1) with the initial condition (3) and the homogeneous boundary con-
ditions
Γ(1)
m [w] = 0 at x = x1 (m = 1, . . . , s),
(2)
(5)
Γm [w] = 0 at x = x2 (m = s + 1, . . . , n)
can be written as
Z x2 Z tZ x2
w(x, t) = f (y)G(x, y, t, 0) dy + Φ(y, τ )G(x, y, t, τ ) dy dτ. (6)
x1 0 x1
Here G(x, y, t, τ ) is the Green’s function, which satisfies, for t > τ ≥ 0, the homogeneous
equation
∂G
− Lx,t [G] = 0 (7)
∂t
with the special nonhomogeneous initial condition
G = δ(x − y) at t = τ (8)
Γ(1)
m [G] = 0 at x = x1 (m = 1, . . . , s),
(9)
Γ(2)
m [G] = 0 at x = x2 (m = s + 1, . . . , n).
The quantities y and τ appear in problem (7)–(9) as free parameters (x1 ≤ y ≤ x2 ), and
δ(x) is the Dirac delta function.
We point out that the Green’s function G is independent of the functions Φ(x, t), f (x),
(1) (2)
gm (t), and gm (t) that characterize various inhomogeneities of the boundary value prob-
(1) (2)
lem. If the coefficients ak , bmk , and bmk determining the differential operator (2) and
boundary conditions (4) are independent of time t, then the Green’s function depends only
on three arguments, G(x, y, t, τ ) = G(x, y, t − τ ).
⊙ Literature: Mathematical Encyclopedia (1977, Vol. 1).
◮ Nonhomogeneous boundary conditions. Preliminary transformations.

To solve the problem with nonhomogeneous boundary conditions (1), (3), (4), we choose a
sufficiently smooth “test function” ϕ = ϕ(x, t) that satisfies the same boundary conditions
as the unknown function; thus,
Γ(1) (1)
m [ϕ] = gm (t) at x = x1 (m = 1, . . . , s),
(10)
Γ(2) (2)
m [ϕ] = gm (t) at x = x2 (m = s + 1, . . . , n).
Otherwise the choice of the “test function” ϕ is arbitrary and is not linked to the solution
of the equation in question; there are infinitely many such functions.
Let us pass from w = w(x, t) to the new unknown u = u(x, t) by the relation
w(x, t) = u(x, t) + ϕ(x, t). (11)
By substituting (11) into (1), (3), and (4), we arrive at the problem for an equation with a modified
right-hand side,
∂u ∂ϕ
− Lx,t [u] = Φ(x, t), Φ(x, t) = Φ(x, t) − + Lx,t [ϕ], (12)
∂t ∂t
subject to the nonhomogeneous initial condition
u = f (x) − ϕ(x, 0) at t = 0 (13)
Γ(1)
m [u] = 0 at x = x1 (m = 1, . . . , s),
(14)
Γ(2)
m [u] = 0 at x = x2 (m = s + 1, . . . , n).
The solution of problem (12)–(14) can be found using the Green’s function by formula (6) in
which one should replace w by u, Φ(x, t) by Φ(x, t), and f (x) by f (x) = f (x) − ϕ(x, 0). Taking
into account relation (11), for w we obtain the representation
Z x2 Z t Z x2
w(x, t) = f (y)G(x, y, t, 0) dy + Φ(y, τ )G(x, y, t, τ ) dy dτ + ϕ(x, t)
x1 0 x1
Z x2 Z tZ x2
∂ϕ
− ϕ(y, 0)G(x, y, t, 0) dy − (y, τ )G(x, y, t, τ ) dy dτ (15)
x1 0 x1 ∂τ
Z t Z x2
+ G(x, y, t, τ )Ly,τ [ϕ(y, τ )] dy dτ.
0 x1
Changing the order of integration and integrating by parts with respect to τ , we find, with
reference to the initial condition (8) for the Green’s function,
Z t Z t
∂ϕ ∂G
G dτ = ϕ(y, t)δ(x − y) − ϕ(y, 0)G(x, y, t, 0) − ϕ(y, τ ) (x, y, t, τ ) dτ. (16)
0 ∂τ 0 ∂τ
We transform the inner integral of the last term in (15) using the Lagrange–Green’s formula [see
Kamke (1977)] to obtain
Z x2 Z x2
y=x2
GLy,τ [ϕ] dy = ϕL∗y,τ [G] dy + L [ϕ, G]y=x1 , (17)
x1 x1
n
X k n−1
X X
∂ ∂qϕ ∂p
L∗y,τ [G] ≡ (−1)k k ak (y, τ )G , L [ϕ, G] ≡ (−1)p q p
ar+1 (y, τ )G ,
∂y r=0 p+q=r
∂y ∂y
k=0
where L∗x,t [w]

is the differential form adjoint to Lx,t [w] in (2), ϕ = ϕ(y, τ ), and p and q are non-
negative integers.
Using relations (16) and (17), we rewrite solution (15) in the form
Z x2 Z t Z x2
w(x, t) = f (y)G(x, y, t, 0) dy + Φ(y, τ )G(x, y, t, τ ) dy dτ
x1 0 x1
Z t
(18)
y=x
+ L [ϕ, G]y=x21 dτ.
0
This formula was derived taking into account the fact that the Green’s function with respect
to y and τ satisfies the adjoint equation∗
∂G
+ L∗y,τ [G] = 0.
∂τ
For subsequent analysis, it is convenient to represent the bilinear differential form
L [ϕ, G] as
n−1
X n−k−1
X
∂k ϕ ∂s
L [ϕ, G] = Ψk [G], Ψk [G] = (−1)s as+k+1 (y, τ )G . (19)
∂y k ∂y s
k=0 s=0
Note that in the special case where the operator (2) is binomial,
∂nw
Lx,t [w] = an + a0 (x, t)w, an = const,
∂xn
the differential forms in (19) are written as
n−1
X ∂ k ϕ ∂ n−k−1 G ∂ n−k−1 G
L [ϕ, G] = an (−1)n−k−1 , Ψk [G] = an (−1)n−k−1 .
∂y k ∂y n−k−1 ∂y n−k−1
k=0
◮ Nonhomogeneous boundary conditions of the special form.

Consider the following nonhomogeneous boundary conditions of special form that are often
encountered in applications:
∂ km w (1)
= gkm (t) at x = x1 (m = 1, . . . , s),
∂xkm (20)
∂ km w (2)
= gkm (t) at x = x2 (m = s + 1, . . . , n).
∂xkm
∗
This equation can be derived by considering the case of homogeneous initial and boundary conditions
and using arbitrariness in the choice of the test function ϕ = ϕ(x, t); it should be taken into account that the
solution itself must be independent of the specific form of ϕ, because ϕ does not occur in the original statement
of the problem. By appropriately selecting the test function, one can also derive the boundary conditions (21).
Without loss of generality, we assume that the following inequalities hold:
n − 1 ≥ k1 > k2 > · · · > ks , n − 1 ≥ ks+1 > ks+2 > · · · > kn .
The Green’s function satisfies the corresponding homogeneous boundary conditions

(1) (2)
that can be obtained from (20) by replacing w by G and by setting gkm (t) = gkm (t) = 0.
The homogeneous boundary conditions adjoint to (20), which must be satisfied by the
Green’s function with respect to y and τ , have the form
Ψkβ [G] = 0 at x = x1 (kβ 6= km , β = s + 1, . . . , n; m = 1, . . . , s),

(21)
Ψkβ [G] = 0 at x = x2 (kβ 6= km , β = 1, . . . , s; m = s + 1, . . . , n).
These conditions involve the linear differential forms Ψk [G] defined in (19). For each
endpoint of the interval in question, the set {kβ } of indices in the boundary operators (21)
together with the set {km } of the orders of derivatives in the boundary conditions (20) make
up a complete set of nonnegative integers from 0 to n − 1.
Taking into account the fact that the test function ϕ must satisfy the boundary condi-
tions (20) and the Green’s function G must satisfy conditions (21), we rewrite the solu-
tion (18) to obtain
Z x2 Z tZ x2
x1 0 x1
s
X Z t n
X Z t (22)
(1) (2)
− gkm (τ )Ψkm [G]y=x1 dτ + gkm (τ )Ψkm [G]y=x2 dτ,
m=1 0 m=s+1 0
where the Ψkm [G] are differential operators with respect to y defined in (19).
If the Green’s function is known, then formula (22) can be used to obtain the solution
of the nonhomogeneous boundary value problem (1), (3), (20) readily for arbitrary Φ(x, t),
(1) (2)
f (x), gkm (t) (m = 1, . . . , s), and gkm (t) (m = s + 1, . . . , n).
◮ General nonhomogeneous boundary conditions.
On solving (4) for the highest derivatives, we reduce the boundary conditions (4) to the
canonical form
kX
m −1
∂ km w (1) ∂iw (1)
+ cmi (t) = hkm (t) at x = x1 (m = 1, . . . , s),
∂xkm ∂xi
i=0
(23)
kX
m −1
∂ km w (2) ∂iw (2)
+ cmi (t) i = hkm (t) at x = x2 (m = s + 1, . . . , n),
∂xkm ∂x
i=0
where the leading terms in distinct boundary conditions are distinct,
n − 1 ≥ k1 > k2 > · · · > ks , n − 1 ≥ ks+1 > ks+2 > · · · > kn .

The sums in (23) do not contain the derivatives of orders k1 , . . . , ks (for x = x1 ) and
ks+1 , . . . , kn (for x = x2 ); thus,
(1)
cmi (t) = 0 at i = kj (j = 1, . . . , s),
(2)
cmi (t) = 0 at i = kj (j = s + 1, . . . , n).
It can be shown that the solution of problem (1), (3), (23) is given by
Z x2 Z t Z x2
x1 0 x1
s Z
X t n
X Z t (24)
(1) (2)
− hkm (τ )Ψkm [G]y=x1 dτ + hkm (τ )Ψkm [G]y=x2 dτ,
m=1 0 m=s+1 0
where the Ψkm [G] are the differential operators with respect to y defined in (19). Rela-
tion (24) is similar to (22) but contains the Green’s function satisfying the more compli-
cated boundary conditions that can be obtained from (23) by substituting G for w and by
(1) (2)
setting hkm (t) = hkm (t) = 0.
⊙ Literature for Section 11.7.1: A. G. Butkovskiy (1979, 1982), A. D. Polyanin (2000a, 2002).
11.7.2 Equations Containing the Second Time Derivative

◮ Homogeneous initial and boundary conditions.
Consider the linear nonhomogeneous differential equation
n
∂2w ∂w X ∂k w
2
+ ψ(x, t) − ak (x, t) k = Φ(x, t). (1)
∂t ∂t ∂x
k=0
We set the homogeneous initial conditions
w = 0 at t = 0,
(2)
∂t w = 0 at t = 0
Γ(1)
m [w] = 0 at x = x1 (m = 1, . . . , s),
(3)
Γ(2)
m [w] = 0 at x = x2 (m = s + 1, . . . , n),
(1) (2)
where the boundary operators Γm [w] and Γm [w] are defined in Section 11.7.1, see formu-
las (4).
The solution of problem (1)–(3) can be represented in the form∗
Z t Z x2
w(x, t) = Φ(y, τ )G(x, y, t, τ ) dy dτ. (4)
0 x1
∗
Problem (1)–(3) is assumed to be well posed.
Here G = G(x, y, t, τ ) is the Green’s function; for t > τ ≥ 0, it satisfies the homogeneous
equation
n
∂2G ∂G X ∂k G
+ ψ(x, t) − a k (x, t) =0 (5)
∂t2 ∂t ∂xk
k=0
with the special semihomogeneous initial conditions
G=0 at t = τ,
(6)
∂t G = δ(x − y) at t = τ
and the corresponding homogeneous boundary conditions
Γ(1)
m [G] = 0 at x = x1 (m = 1, . . . , s),
(7)
Γ(2)
m [G] = 0 at x = x2 (m = s + 1, . . . , n).
The quantities y and τ appear in problem (5)–(7) as free parameters (x1 ≤ y ≤ x2 ), and
δ(x) is the Dirac delta function.
One can verify formula (4) by a straightforward substitution into the equation and the
initial and boundary conditions (1)–(3) and by taking into account properties (5)–(7) of the
Green’s function.
◮ Nonhomogeneous initial and boundary conditions.

Consider the linear nonhomogeneous differential equation (1) with the general nonhomo-
geneous initial conditions
w = f0 (x) at t = 0,
(8)
∂t w = f1 (x) at t = 0
and the nonhomogeneous boundary conditions reduced to the canonical form (the reduction
of boundary conditions to the canonical form is described at the end of Section 9.7.1):
kX
m −1
∂ km w (1) ∂iw (1)
k
+ cmi (t) = hkm (t) at x = x1 (m = 1, . . . , s),
∂x m ∂xi
i=0
(9)
kX
m −1
∂ km w (2) ∂iw (2)
+ cmi (t) i = hkm (t) at x = x2 (m = s + 1, . . . , n).
∂xkm ∂x
i=0
Introducing a test function ϕ = ϕ(x, t) that satisfies the nonhomogeneous initial and
boundary conditions (8), (9) and using the same line of reasoning as in Section 11.7.1 for a
simpler equation, we arrive at the solution of problem (1), (8), (9) in the form
Z t Z x2 Z x2
∂
w(x, t) = Φ(y, τ )G(x, y, t, τ ) dy dτ − f0 (y) G(x, y, t, τ ) dy
0 x1 x1 ∂τ τ =0
Z x2

+ f1 (y) + f0 (y)ψ(y, 0) G(x, y, t, 0) dy (10)
x1
s Z
X t n
X Z t
(1) (2)
− hkm (τ )Ψkm [G]y=x1 dτ + hkm (τ )Ψkm [G]y=x2 dτ,
m=1 0 m=s+1 0
where the Ψkm [G] are differential operators with respect to y defined in Section 11.7.1; see
relations (19).
Remark 11.7. If the coefficients of Eq. (1) and those of the boundary conditions (9) are time
independent, i.e., if
(1) (2)
ψ = ψ(x), ak = ak (x), cmi = const, cmi = const,
then in solution (10) one should set
∂ ∂ e
e y, t − τ ),
G(x, y, t, τ ) = G(x, G(x, y, t, τ ) = − G(x, y, t).
∂τ τ =0 ∂t
11.7.3 Nonstationary Problems with Many Space Variables

◮ Equations with the first partial derivative with respect to t.
Consider the following linear differential operator with respect to variables x1 , . . . , xn :
X ∂ k1 +···+kn w
Lx,t [w] ≡ Ak1 ,...,kn (x1 , . . . , xn , t) . (1)
∂xk11 . . . ∂xknn
The coefficients Ak1 ,...,kn of the operator are assumed to be sufficiently smooth functions of
x1 , . . . , xn and t (and also bounded if necessary). The coefficients of the highest derivatives
are assumed to be everywhere nonzero.
1◦ . Cauchy problem (t ≥ 0, x ∈ Rn ). The solution of the Cauchy problem for the linear
nonhomogeneous parabolic differential equation with variable coefficients
∂w
− Lx,t [w] = Φ(x, t) (2)
∂t
under the initial conditions
w = f (x) at t = 0 (3)
is given by
Z tZ Z
w(x, t) = Φ(y, τ )E (x, y, t, τ ) dVy dτ + f (y)E (x, y, t, 0) dVy . (4)
0 Rn Rn
Here dVy = dy1 . . . dyn and E = E (x, y, t, τ ) is the fundamental solution of the Cauchy
problem, which satisfies the equation
∂E
− Lx,t [E ] = 0 (5)
∂t
for t > τ ≥ 0 and the special initial condition

E t=τ = δ(x − y). (6)
The variables y and τ appear in problem (5), (6) as free parameters (y ∈ Rn ), and δ(x) is
the n-dimensional Dirac delta function.
If the coefficients Ak1 ,...,kn of operator (1) are independent of time t, then the funda-
mental solution only depends on three arguments, E (x, y, t, τ ) = E (x, y, t − τ ). If the
coefficients of operator (1) are constants, then E (x, y, t, τ ) = E (x − y, t − τ ).
2◦ . Boundary value problems (t ≥ 0, x ∈ D). The solutions of linear boundary value prob-
lems in a spatial domain D for Eq. (2) with initial condition (3) and homogeneous boundary
conditions for x ∈ ∂D (these conditions are not written out here) are given by formula (4)
in which the domain of integration Rn should be replaced by D. Here by E we mean
the Green’s function that must satisfy, apart from Eq. (5) and the boundary condition (6),
the same homogeneous boundary conditions for x ∈ ∂D as the original equation (2). For
boundary value problems, the parameter y belongs to the same domain as x; i.e., y ∈ D.
⊙ Literature: Mathematical Encyclopedia (1977, Vol. 1).
◮ Equations with the second partial derivative with respect to t.

1◦ . Cauchy problem (t ≥ 0, x ∈ Rn ). The solution of the Cauchy problem for the linear
nonhomogeneous differential equation with variable coefficients
∂2w
− Lx,t [w] = Φ(x, t) (7)
∂t2
under the initial conditions
w = f (x) at t = 0,
(8)
∂t w = g(x) at t = 0
is given by
Z tZ
w(x, t) = Φ(y, τ )E (x, y, t, τ ) dVy dτ
Rn
Z0 Z
∂
− f (y) E (x, y, t, τ ) dVy + g(y)E (x, y, t, 0) dVy .
Rn ∂τ τ =0 Rn
Here E = E (x, y, t, τ ) is the fundamental solution of the Cauchy problem
∂2E
− Lx,t [E ] = 0,
∂t2
∂E
E t=τ = 0, = δ(x − y),
∂t t=τ
where y and τ play the role of parameters.

If the coefficients Ak1 ,...,kn of operator (1) are independent of time t, then the funda-
mental solution only depends
on three arguments, E (x, y, t, τ ) = E (x, y, t − τ ), and the
∂
relation ∂τ E (x, y, t, τ )τ =0 = − ∂t
∂
E (x, y, t) holds. If the coefficients of operator (1) are
constant, then E (x, y, t, τ ) = E (x − y, t − τ ).
2◦ . The solution of the Cauchy problem for the more complicated linear nonhomogeneous
differential equation with variable coefficients
∂2w ∂w
2
+ ψ(x, t) − Lx,t [w] = Φ(x, t)
∂t ∂t
with initial conditions (8) is expressed as

Z tZ Z
∂
w(x, t) = Φ(y, τ )E (x, y, t, τ ) dVy dτ − f (y) E (x, y, t, τ ) dVy
0 Rn Rn ∂τ τ =0
Z (9)

+ g(y) + ψ(y, 0)f (y) E (x, y, t, 0) dVy .
Rn
Here E (x, y, t, τ ) is the corresponding fundamental solution of the Cauchy problem
∂2E ∂E
2
+ ψ(x, t) − Lx,t [E ] = 0,
∂t ∂t
∂E
E t=τ = 0, = δ(x − y).
∂t t=τ
3◦ . Boundary value problems (t ≥ 0, x ∈ D). The solutions of linear boundary value
problems in a spatial domain D for Eq. (7) with initial condition (8) and homogeneous
boundary conditions for x ∈ ∂D (these conditions are not written out here) are given by
formula (9) in which the domain of integration Rn should be replaced by D. Here by E we
mean the Green’s function that must satisfy, apart from Eq. (7) and the initial conditions (8),
the same homogeneous boundary conditions as the solutions of the original equation (7).
11.7.4 Some Special Equations with Variable Coefficients

∂w ∂ nw ∂w
1. = k(t) + xf (t) + g(t) + h(t)w.
∂t ∂xn ∂x
The transformation
Z Z Z
w(x, t) = u(z, τ ) exp h(t) dt , z = xF (t) + g(t)F (t) dt, τ= k(t)F n (t) dt,
Z
where F (t) = exp f (t) dt , leads to the simpler constant coefficient equation
∂u ∂nu
= .
∂τ ∂z n
Concerning the solutions of this equation, see equation 11.6.5.1 (for even n) and equa-
tion 11.6.5.2 (for odd n).
n
X
∂w ∂ kw
2. + ak (t) = Φ(x, t).
∂t ∂xk
k=0
Z t X
n
k
w = exp βx − λ(t) , λ(t) = ak (τ )β dτ,
t0 k=0
Z t Z t
w = exp − p(τ ) dτ cos βx + q(τ ) dτ ,
t0 t0
Z t Z t
w = exp − p(τ ) dτ sin βx + q(τ ) dτ ,
t0 t0
where β and t0 are arbitrary constants and

2k≤n
X 2k+1≤n
X
p(t) = (−1)k a2k (t)β 2k , q(t) = (−1)k+1 a2k+1 (t)β 2k+1 .
k=0 k=0
2◦ . Let the condition

2k≤n
X
P (ξ, t) = (−1)k a2k (t)ξ 2k > 0
k=0
be satisfied for ξ > 0 and t > 0.
Z Z t Z t
1 ∞
E (x, t, τ ) = exp − P (ξ, µ) dµ cos Q(ξ, µ) dµ + xξ dξ,
π 0 τ τ
2k+1≤n
X
Q(ξ, t) = (−1)k+1 a2k+1 (t)ξ 2k+1 .
k=0

w = f (x) at t = 0.
Solution:
Z ∞ Z tZ ∞
w(x, t) = E (x − y, t, 0)f (y) dy + E (x − y, t, τ )Φ(y, τ ) dy dτ,
−∞ 0 −∞
where the function E (x, t, τ ) is defined in Item 2◦ .
∂kw ∂nw
3. = ax2n .
∂tk ∂xn
The transformation z = 1/x, u = wx1−n leads to the constant coefficient equation
∂k u n
n∂ u
= a(−1) .
∂tk ∂z n
n
X
∂kw ∂ mw
4. = a m xm .
∂tk ∂xm
m=0
The change of variable z = ln |x| leads to a constant coefficient equation.
∂kw ∂nw
5. = (ax2 + bx + c)n .
∂tk ∂xn
The transformation
Z
n−1 dx
w(x, t) = u(z, t)|ax2 + bx + c| 2 , z=
ax2 + bx + c
leads to a constant coefficient equation.
n
∂
6. − Lx w = 0, n = 1, 2, . . .
∂t
Here Lx is a linear differential operator of any order with respect to the space variable x
whose coefficients can depend on x.
n−1
X
w(x, t) = tk uk (x, t),
k=0
where the uk = uk (x, t) are arbitrary functions that satisfy the original equation with n = 1:
(∂t − Lx )uk = 0.
tn−1
En (x, t) = E1 (x, t),
(n − 1)!
where E1 (x, t) is the fundamental solution of the equation with n = 1.

Remark 11.8. The linear differential operator Lx can involve arbitrarily many space variables.
n
∂2
7. − Lx w = 0, n = 1, 2, . . .
∂t2
Here Lx is a linear differential operator of any order with respect to the space variable x
whose coefficients can depend on x.
n−1
X
w(x, t) = tk uk (x, t),
k=0
where the uk = uk (x, t) are arbitrary functions that satisfy the original equation with n = 1:
(∂tt − Lx )uk = 0.
2◦ . Suppose that the Cauchy problem for the special case of the equation with n = 1 is well
posed if only one initial condition is set at t = 0; this means that the constant coefficient
differential operator Lx is such that the equation with n = 1 is regular with regularity index
r = 1. Then the fundamental solution of the original equation can be found by the formula
tn−1
En (x, t) = E1 (x, t),
(n − 1)!
where E1 (x, t) is the fundamental solution for n = 1.

Remark 11.9. The linear differential operator Lx can involve arbitrarily many space variables.
m
X
8. ak Lk [w] = 0.
k=0
Here L is any linear differential operator that depends on arbitrarily many independent
variables x1 , . . . , xn .
Solution:
m
X
w(x1 , . . . , xn ) = Cs us (x1 , . . . , xn ),
s=1
where the us are solutions of the simpler equations L[us ] − λs us = 0, the λs are roots of
m
P
the characteristic equation ak λk = 0, and the Cs are arbitrary constants.
k=0
Chapter 12
Systems of Linear
Partial Differential Equations
12.1 Preliminary Remarks. Some Notation and Helpful

Relations
Let f and u = u e1 + v e2 + w e3 be arbitrary sufficiently smooth scalar and vector func-
tions, and let e1 , e2 , and e3 be the unit coordinate vectors corresponding to the Cartesian
coordinates x, y, and z. Then, by definition, we have
∇f = fx e1 + fy e2 + fz e3 ,
div u = ∇ · u = ux + vy + wz ,
curl u = ∇ × u = (wy − vz )e1 + (uz − wx )e2 + (vx − uy )e3 ,
∆f = fxx + fyy + fzz ,
where the subscripts x, y, and z stand for the derivatives. The following differential rela-
tions hold:
curl ∇f = 0, div curl u = 0, div ∇f = ∆f,
curl curl u = ∇ div u − ∆u, ∆(xf ) = x∆f + 2∇f, x = (x, y, z),
curl[∆(xf )] = curl(x∆f ), ∆(x · u) = x · (∆u) + 2 div u,
which are often used in what follows in Sections 12.5–12.19.
12.2 Systems of Two First-Order Equations

∂u ∂u ∂w ∂w
1. =a + f1 (t)u + g1 (t)w, =a + f2 (t)u + g2 (t)w.
∂t ∂x ∂t ∂x
Variable coefficient first-order linear system.
General solution:
u = ϕ1 (t)U (x + at) + ϕ2 (t)W (x + at),
w = ψ1 (t)U (x + at) + ψ2 (t)W (x + at),
1059
1060 S YSTEMS OF L INEAR PARTIAL D IFFERENTIAL E QUATIONS
where U = U (z) and W = W (z) are arbitrary functions and the pairs of functions ϕ1 =
ϕ1 (t), ψ1 = ψ1 (t) and ϕ2 = ϕ2 (t), ψ2 = ψ2 (t) are linearly independent (fundamental)
solutions of the system of linear ordinary differential equations
ϕ′t = f1 (t)ϕ + g1 (t)ψ,
ψt′ = f2 (t)ϕ + g2 (t)ψ.
∂u ∂u
2. = a(t) + f1 (t)u + g1 (t)w + h1 (t),
∂t ∂x
∂w ∂w
= a(t) + f2 (t)u + g2 (t)w + h2 (t).
∂t ∂x
General solution:
Z
u = ϕ1 (t)U (z) + ϕ2 (t)W (z) + u0 (t),
z =x+ a(t) dt,
w = ψ1 (t)U (z) + ψ2 (t)W (z) + w0 (t),
where U = U (z) and W = W (z) are arbitrary functions, the pairs of functions ϕ1 = ϕ1 (t),
ψ1 = ψ1 (t) and ϕ2 = ϕ2 (t), ψ2 = ψ2 (t) are linearly independent (fundamental) solutions
of the system of linear ordinary differential equations
ϕ′t = f1 (t)ϕ + g1 (t)ψ,
ψt′ = f2 (t)ϕ + g2 (t)ψ,
and u0 = u0 (t), w0 = w0 (t) is a solution of the system of linear ordinary differential
equations
u′0 = f1 (t)u0 + g1 (t)w0 + h1 (t),
w0′ = f2 (t)u0 + g2 (t)w0 + h2 (t).
∂u ∂w ∂w ∂u
3. C + + Au = 0, L + + Rw = 0.
∂t ∂x ∂t ∂x
This system of equations occurs in the theory of propagation of quasistationary electric
oscillations in cables. (Here u is the voltage, w is the current, L is the self-induction, R is
the ohmic resistance, C is the capacitance, and A is the creepage.)
1◦ . Let us introduce an auxiliary function ψ by the formulas
∂ψ ∂ψ
u= , w = −C − Aψ.
∂x ∂t
Then the first equation in the system is satisfied identically, and the second equation is
reduced to the second-order linear equation
∂2ψ ∂ψ ∂2ψ
CL + (AL + CR) − + ARψ = 0. (1)
∂t2 ∂t ∂x2
The substitution
AL + CR
ψ = e−µt θ(x, t), µ= (2)
2CL
reduces Eq. (1) to the simpler form
∂2θ 1 ∂2θ AL − CR
2
= 2
+ β 2 θ, β = .
∂t CL ∂x 2CL
For the solutions of this equation, see Section 3.1.3.
12.2. Systems of Two First-Order Equations 1061
2◦ . With AL = CR, the general solution of the system is given by the formulas
u = e−µt [ϕ1 (x − at) + ϕ2 (x + at)], a = (CL)−1/2 ,

w = be−µt [ϕ1 (x − at) − ϕ2 (x + at)], b = (C/L)1/2 ,
where ϕ1 (z1 ) and ϕ2 (z2 ) are arbitrary functions and the coefficient µ is defined in (2).
⊙ Literature: V. I. Smirnov (1974, Vol. 2, pp. 571–577).
∂u ∂u ∂w ∂w
4. a1 + b1 + c1 + d1 + p1 u + q1 w = 0,
∂t ∂x ∂t ∂x
∂u ∂u ∂w ∂w
a2 + b2 + c2 + d2 + p2 u + q2 w = 0.
∂t ∂x ∂t ∂x
It is a system of constant coefficient first-order homogeneous linear PDEs.
1◦ . Set
∂ψ ∂ψ ∂ψ ∂ψ
u = c1 + d1 + q1 ψ, w = −a1 − b1 − p1 ψ. (1)
∂t ∂x ∂t ∂x
By substituting (1) into the second equation of the system, we arrive at a second-order
constant coefficient linear PDE for ψ,
∂2ψ ∂2ψ ∂2ψ

(a2 c1 − a1 c2 ) + (a d
2 1 − a d
1 2 + b c
2 1 − b c
1 2 ) + (b d
2 1 − b d
1 2 )
∂t2 ∂x∂t ∂x2
∂ψ ∂ψ
+ (a2 q1 − a1 q2 + c1 p2 − c2 p1 ) + (b2 q1 − b1 q2 + d1 p2 − d2 p1 ) (2)
∂t ∂x
+ (p2 q1 − q2 p1 )ψ = 0.
The substitution of the expressions (1) into the first equation of the system results in an
identity.
Remark 12.1. A similar technique of reducing two linear equations to one equation can be used
if, instead of the second equation, we deal with any other linear or nonlinear PDE of arbitrary order.
(The first equation remains the same.)
If we set
∂ψ ∂ψ ∂ψ ∂ψ
u = c2 + d2 + q2 ψ, w = −a2 − b2 − p2 ψ,
∂t ∂x ∂t ∂x
then the first equation of the system gives Eq. (2), and the second equation gives an identity.
2◦ . Consider the nondegenerate case ∆ = a1 d1 − b1 c1 6= 0. The transformation
1
U = exp(βx + λt)(a1 u + c1 w), u= exp(−βx − λt)(d1 U − c1 W ),
∆
1
W = exp(βx + λt)(b1 u + d1 w), w= exp(−βx − λt)(a1 W − b1 U ),
∆
where β and λ are a solution of the linear algebraic system
a1 λ + b1 β = p1 ,
c1 λ + d1 β = q1 ,
leads to the system

∂U ∂W
+ = 0, (3)
∂t ∂x
∂U ∂U ∂W ∂W
A +B +C +D + P U + QW = 0. (4)
∂t ∂x ∂t ∂x
Here Eq. (4) uses the notation
A = d1 a2 − b1 c2 , B = d1 b2 − b1 d2 , C = a1 c2 − c1 a2 , D = a1 d2 − c1 b2 ,
P = −Aλ − Bβ + d1 p2 − b1 q2 , Q = −Cλ − Dβ − c1 p2 + a1 q2 .
Let us differentiate Eq. (4) with respect to x and then eliminate the derivative Wx with
the help of the first equation. As a result, we obtain a second-order PDE for U ,
∂2U ∂2U ∂2U ∂U ∂U
C + (D − A) − B +Q −P = 0. (5)
∂t2 ∂x∂t ∂x2 ∂t ∂x
In a similar way, one can derive exactly the same second-order PDE for W .
We can satisfy Eq. (3) identically by setting U = ∂Ψ/∂x and W = −∂Ψ/∂t, where
Ψ is a new unknown function (an analog of the stream function in hydrodynamic prob-
lems). Then Eq. (4) leads to the second-order constant coefficient linear partial differential
equation considered in Section 14.1.1 (see Eq. 14.1.1.8).
3◦ . Consider the degenerate case ∆ = a1 d1 − b1 c1 = 0.
3.1. Let a1 6= 0 and c1 6= 0. Then one can set b1 /a1 = d1 /c1 = k. By passing from u
and w to the new unknown functions u and v = a1 u + c1 w, we transform the first equation
of the system to the form

∂v ∂v q1 a1
+k + v + p1 − q1 u = 0. (6)
∂t ∂x c1 c1
If c1 p1 − a1 q1 6= 0, then, by eliminating u from (6) and from the transformed second equa-
tion of the system, we obtain a second-order constant coefficient linear partial differential
equation for the function v. If c1 p1 − a1 g1 = 0, then Eq. (6) permits finding the function v
immediately (because the equation is independent of u).
3.2. Let a1 = c1 = 0. By passing from u and w to the new unknown functions u and
v = b1 u + d1 w, we reduce the first equation of the system to the form

∂v q1 b1
+ v + p1 − q1 u = 0. (7)
∂x d1 d1
If d1 p1 − b1 q1 6= 0, then, by eliminating u from (7) and from the transformed second equa-
tion of the system, we obtain a second-order constant coefficient linear partial differential
equation for the function v. If d1 p1 − b1 q1 = 0, then Eq. (7) permits finding the function v
immediately (because the equation is independent of u).
3.3. Let a1 = b1 = 0 and c1 6= 0. For p1 6= 0, the simple elimination of u from the
system leads to a second-order constant coefficient linear partial differential equation for
the function w. For p1 = 0, the first equation of the system permits finding the function w
immediately.
12.3. Systems of Two Second-Order Equations 1063
3.4. Let c1 = d1 = 0 and a1 6= 0. For q1 6= 0, the simple elimination of w from the

system leads to a second-order constant coefficient linear partial differential equation for
the function u. For q1 = 0, the first equation of the system permits finding the function u
immediately.
12.3 Systems of Two Second-Order Equations

12.3.1 Systems of Parabolic Equations
∂u ∂ 2u ∂w ∂2w
1. =a + b1 u + c1 w, =a + b2 u + c2 w.
∂t ∂x2 ∂t ∂x2
Constant coefficient second-order linear system of parabolic type.
Solution:
b1 − λ2 b1 − λ1
u= eλ1 t θ1 − eλ2 t θ2 ,
b2 (λ1 − λ2 ) b2 (λ1 − λ2 )
1
w= eλ1 t θ1 − eλ2 t θ2 ,
λ1 − λ2
where λ1 and λ2 are roots of the quadratic equation
λ2 − (b1 + c2 )λ + b1 c2 − b2 c1 = 0
and the functions θn = θn (x, t) satisfy the independent linear heat equations
∂θ1 ∂ 2 θ1 ∂θ2 ∂ 2 θ2
=a 2 , =a 2 .
∂t ∂x ∂t ∂x
∂u ∂ 2u ∂w ∂ 2w
2. =a + f1 (t)u + g1 (t)w, =a + f2 (t)u + g2 (t)w.
∂t ∂x2 ∂t ∂x2
Variable coefficient second-order linear system of parabolic type.
Solution:
u = ϕ1 (t)U (x, t) + ϕ2 (t)W (x, t),
w = ψ1 (t)U (x, t) + ψ2 (t)W (x, t),
where the pairs of functions ϕ1 = ϕ1 (t), ψ1 = ψ1 (t) and ϕ2 = ϕ2 (t), ψ2 = ψ2 (t) are
linearly independent (fundamental) solutions of the system of linear ordinary differential
equations
ϕ′t = f1 (t)ϕ + g1 (t)ψ,
ψt′ = f2 (t)ϕ + g2 (t)ψ,
and the functions U = U (x, t) and W = W (x, t) satisfy the independent linear heat equa-
tions
∂U ∂2U ∂W ∂2W
=a 2, =a .
∂t ∂x ∂t ∂x2
⊙ Literature for Section 12.3.1: A. D. Polyanin and A. V. Manzhirov (2007, pp. 1341–1342).
12.3.2 Systems of Hyperbolic or Elliptic Equations
∂2u ∂2u ∂2w ∂2w

1. =k + a1 u + b1 w, =k + a2 u + b2 w.
∂t2 ∂x2 ∂t2 ∂x2
Constant coefficient second-order linear system of hyperbolic type.
Solution:
a1 − λ2 a1 − λ1 1
u= θ1 − θ2 , w= θ1 − θ2 ,
a2 (λ1 − λ2 ) a2 (λ1 − λ2 ) λ1 − λ2
λ2 − (a1 + b2 )λ + a1 b2 − a2 b1 = 0
and the functions θn = θn (x, t) satisfy the independent linear Klein–Gordon equations
∂ 2 θ1 ∂ 2 θ1 ∂ 2 θ2 ∂ 2 θ2
= k + λ1 θ1 , = k + λ2 θ2 .
∂t2 ∂x2 ∂t2 ∂x2
∂2u ∂2u ∂2w ∂2w
2. + = a1 u + b1 w, + = a2 u + b2 w.
∂x2 ∂y 2 ∂x2 ∂y 2
Constant coefficient second-order linear system of elliptic type.
Solution:
a1 − λ2 a1 − λ1 1
u= θ1 − θ2 , w= θ1 − θ2 ,
a2 (λ1 − λ2 ) a2 (λ1 − λ2 ) λ1 − λ2
λ2 − (a1 + b2 )λ + a1 b2 − a2 b1 = 0
and the functions θn = θn (x, y) satisfy the independent linear Helmholtz equations
∂ 2 θ1 ∂ 2 θ1 ∂ 2 θ2 ∂ 2 θ2
+ = λ1 θ1 , + = λ2 θ2 .
∂x2 ∂y 2 ∂x2 ∂y 2
⊙ Literature for Section 12.3.2: A. D. Polyanin and A. V. Manzhirov (2007, p. 1342).
12.4 Systems of Two Higher-Order Equations

∂u ∂w
1. = L[u] + f1 (t)u + g1 (t)w, = L[w] + f2 (t)u + g2 (t)w.
∂t ∂t
Here L is an arbitrary linear differential operator with respect to the coordinates x1 , . . . , xn
(of any order in derivatives) whose coefficients may depend on x1 , . . . , xn , t. It is assumed
that L[const] = 0.
Solution:
u = ϕ1 (t)U (x1 , . . . , xn , t) + ϕ2 (t)W (x1 , . . . , xn , t),
w = ψ1 (t)U (x1 , . . . , xn , t) + ψ2 (t)W (x1 , . . . , xn , t),
12.4. Systems of Two Higher-Order Equations 1065
where the two pairs of functions ϕ1 = ϕ1 (t), ψ1 = ψ1 (t) and ϕ2 = ϕ2 (t), ψ2 = ψ2 (t) are
linearly independent (fundamental) solutions of the system of first-order linear ordinary
differential equations
ϕ′t = f1 (t)ϕ + g1 (t)ψ,
ψt′ = f2 (t)ϕ + g2 (t)ψ
and the functions U = U (x1 , . . . , xn , t) and W = W (x1 , . . . , xn , t) satisfy the independent
linear equations
∂U ∂W
= L[U ], = L[W ].
∂t ∂t
⊙ Literature: A. D. Polyanin and A. V. Manzhirov (2007, p. 1374).
∂2u ∂2w
2. = L[u] + a1 u + b1 w, = L[w] + a2 u + b2 w.
∂t2 ∂t2
Here L is an arbitrary linear differential operator with respect to the coordinates x1 , . . . , xn
(of any order in derivatives).
Solution:
a1 − λ2 a1 − λ1 1
u= θ1 − θ2 , w= θ1 − θ2 ,
a2 (λ1 − λ2 ) a2 (λ1 − λ2 ) λ1 − λ2
λ2 − (a1 + b2 )λ + a1 b2 − a2 b1 = 0
and the functions θn = θn (x1 , . . . , xn , t) satisfy the independent linear equations
∂ 2 θ1 ∂ 2 θ2
= L[θ1 ] + λ1 θ1 , = L[θ2 ] + λ2 θ2 .
∂t2 ∂t2
⊙ Literature: A. D. Polyanin and A. V. Manzhirov (2007, p. 1374).
3. L[u] = a1 u + b1 w, L[w] = a2 u + b2 w.
Here L is an arbitrary linear differential operator with respect to the variables x1 , . . . , xn , t
(of any order in derivatives).
Solution:
a1 − λ2 a1 − λ1 1
u= θ1 − θ2 , w= θ1 − θ2 ,
a2 (λ1 − λ2 ) a2 (λ1 − λ2 ) λ1 − λ2
λ2 − (a1 + b2 )λ + a1 b2 − a2 b1 = 0
and the functions θn = θn (x1 , . . . , xn , t) satisfy the independent linear equations
L[θ1 ] = λ1 θ1 , L[θ2 ] = λ2 θ2 .
4. L1 [u] + L2 [w] = 0, M [u, w] = 0.

Here L1 and L2 are arbitrary constant coefficient linear differential operators with respect
to the variables x and t (of any order in derivatives) and M is an arbitrary linear or nonlinear
differential operator with respect to the variables x and t whose coefficients may depend
on the independent variables.
One can identically satisfy the first equation by setting
u = L2 [ψ], w = −L1 [ψ], (1)
where ψ is the new unknown function (an analog of the stream function in hydrodynamic
problems). By substituting (1) into the second equation of the system, we obtain a equation
for ψ,
M L2 [ψ], −L1 [ψ] = 0.
12.5 Simplest Systems Containing Vector Functions and

Operators div and curl
12.5.1 Equation curl u = A(x)
◮ Homogeneous equation.
Let u = (u, v, w) be a vector field, where u, v, w are its components depending on the
rectangular Cartesian coordinates x, y, z. Then the homogeneous equation
curl u = 0 (1)
is equivalent to the system
wy − vz = 0, uz − wx = 0, vx − uy = 0 (2)
of three coupled first-order linear equations for the components of u.

The general solution of the vector equation (1) or system (2) is
u = ∇ϕ or u = ϕx , v = ϕy , w = ϕz ,
where ϕ = ϕ(x, y, z) is an arbitrary function.
◮ Nonhomogeneous equation.
Consider the nonhomogeneous vector equation
curl u = A(x), (3)

where A(x) = a(x), b(x), c(x) is a given vector function and x = (x, y, z).
The solvability condition for this equation has the form
div A(x) ≡ ax + by + cz = 0. (4)

12.5. Simplest Systems Containing Vector Functions and Operators div and curl 1067
The vector equation (3) is equivalent to the system
wy − vz = a(x), uz − wx = b(x), vx − uy = c(x) (5)
of three coupled first-order linear equations for the components of u.

Let condition (4) be satisfied. We seek a solution of system (5) in the form
u = ϕx , v = ϕy + η, w = ϕz + ζ. (6)
By substituting (6) into (5), we have
ηx = c(x), ζx = −b(x), ζy − ηz = a(x). (7)
The first two equations can be integrated in an elementary way. The resulting solutions
are substituted into the third equation, and then we use condition (4). Finally, after simple
transformations, we obtain
Z x Z x Z y
η= c(x̄, y, z) dx̄, ζ=− b(x̄, y, z) dx̄ + a(x0 , ȳ, z) dȳ, (8)
x0 x0 y0
where x0 and y0 are arbitrary constants lying in the domain of Eq. (3).
Formulas (6) and (8), where ϕ = ϕ(x, y, z) is an arbitrary function, give the general
solution of the nonhomogeneous vector equation (3) (or system (5)) under the solvability
condition (4).
12.5.2 Simple Systems of Equations Containing Operators div and

curl
◮ Homogeneous system.
Consider the homogeneous system of equations
div u = 0, curl u = 0.
It is an overdetermined system, because it consists of four scalar equations for the three
components of the vector u.
The general solution of the system has the form
u = ∇ϕ, (9)
where ϕ = ϕ(x) is an arbitrary harmonic function satisfying the Laplace equation
∆ϕ = 0.
◮ Nonhomogeneous system I.
Consider an overdetermined nonhomogeneous system of equations consisting of two cou-
pled equations,
div u = f (x), curl u = 0. (10)
The general solution of system (10) has the form (9), where the function ϕ = ϕ(x)
satisfies the Poisson equation
∆ϕ = f (x).
A particular solution of the equation in a bounded domain V has the form
Z
1 f (x′ )
ϕ(x) = − dV ′ ,
4π V |x − x′ |
where dV ′ = dx′ dy ′ dz ′ .
◮ Nonhomogeneous system II.

Consider an overdetermined nonhomogeneous system of equations consisting of two cou-
pled equations,
div u = 0, curl u = A(x). (11)
We assume that the vector function A(x) satisfies the solvability condition (4).
1◦ . We seek a solution of system (11) under condition (4) in the form (6), (8). As a result,
for the function ϕ we obtain the Poisson equation
∆ϕ + g(x) = 0,
where
Z x Z x Z y
g(x) = cy (x̄, y, z) dx̄ − bz (x̄, y, z) dx̄ + az (x0 , ȳ, z) dȳ. (12)
x0 x0 y0
2◦ . A particular solution of system (11) under condition (4) in a bounded domain V has
the form Z
1 A(x′ )
u = curl ψ, ψ(x) = dV ′ .
4π V |x − x′ |
◮ Nonhomogeneous system III (Helmholtz problem on the reconstruction of a vector

field from its divergence and curl).
Consider the overdetermined nonhomogeneous system of equations
div u = f (x), curl u = A(x), (13)
which generalizes systems (10) and (11). We assume that the solvability condition (4) is
satisfied.
12.5. Simplest Systems Containing Vector Functions and Operators div and curl 1069
1◦ . We seek a solution of problem (13) in the form (6), (8). For the function ϕ, we obtain
the Poisson equation
∆ϕ + g(x) = f (x), (14)
where the function g(x) is determined by formula (12).
2◦ . The solution of problem (13) can be represented in the form
u = ∇ϕ + curl ψ + ∇θ,
Z Z
1 f (x′ ) 1 A(x′ )
ϕ(x) = − dV ′ , ψ(x) = dV ′ ,
4π V |x − x′ | 4π V |x − x′ |
where the function θ satisfies the Laplace equation, ∆θ = 0. The surface V may be un-
bounded if both integrals converge and decay as |x| → ∞ at least at the rate of |x|−(1+ε) ,
where ε > 0.
3◦ . Equations (13) are often supplemented with the boundary condition
(n · u) = h(x) at x ∈ S,
where n is the normal vector on the boundary S of the domain V and h(x) is a given
function. For this problem to be solvable, the solvability condition
R (4) should be
R satisfied
and, in the case of a bounded domain V , the additional condition V f (x) dV = S h(x) dS
should hold.
12.5.3 Two Representations of Vector Functions

◮ Stokes–Helmholtz representation of a vector function.
Each sufficiently smooth vector function f = f(x) can be represented as the sum of a poten-
tial and a solenoidal vector (the Stokes–Helmholtz decomposition):
f = ∇γ + curl ω. (15)
In what follows, we assume that the vector f is given and we need to determine the
scalar and vector functions γ and ω.
1◦ . The scalar function γ satisfies the Poisson equation
∆γ = div f, (16)
which can be derived by applying the operator div to relation (15). It follows from (16) that
the function γ is not uniquely determined. Assume that we have constructed a solution γ
of Eq. (16). Then we have an equation of the form (3) for the vector ω,
curl ω = f − ∇γ, (17)
whose right-hand side is known and satisfies the solvability condition. The vector ω can
be found by formulas (6) and (8), where u, v, and w should be replaced by ω1 , ω2 , and ω3 ,
respectively, and a(x), b(x), and c(x) are the components of the vector f−∇γ. Formulas (6)
include the arbitrary function ϕ, and hence the vector ω is determined nonuniquely.
2◦ . The scalar and vector functions γ and ω in (15), for example, can be sought in the form
γ = div U, ω = − curl U. (18)
By substituting (18) into (15) and by taking into account the identity
curl curl U = ∇ div U − ∆U,
we obtain Poisson’s vector equation for the Lamé vector potential U,
∆U = f. (19)
The solutions of Eq. (19) permit using formulas (18) to find the functions γ and ω
determining the decomposition (15). A particular solution of Eq. (19) has the form
Z
1 f(x′ )
U=− dV ′ .
4π V |x − x′ |
The Stokes–Helmholtz representation (15) of a vector function contains four scalar

functions γ, ω1 , ω2 , and ω3 .
◮ Representation of a vector function via a potential and two stream functions.

The vector function f = (f1 , f2 , f3 ) can also be represented in the form
f = ∇γ + h, h = (h1 , h2 , h3 ),
(20)
h1 = θy(1) , h2 = −θx(1) + θz(2) , h3 = −θy(2) .
Here the potential γ is the same as in (15) and the stream functions θ (1) and θ (2) can be
expressed via the components of the vector ω = (ω1 , ω2 , ω3 ) occurring in the decomposition
(15) as follows:
Z y Z y
∂ ∂
θ (1) = ω3 − ω2 (x, ȳ, z) dȳ, θ (2) = ω1 − ω2 (x, ȳ, z) dȳ, (21)
∂z a ∂x a
where a is an arbitrary constant.

The representation (20)–(21) of the mass force contains three scalar functions γ, θ (1) ,
and θ (2) (one less than the Stokes–Helmholtz representation (15)), and the vector h satisfies
the incompressibility condition div h = 0.
Remark 12.2. The conditions under which the vector function f = f(x) in a finite or infinite do-
main V admits the Stokes–Helmholtz representation (15), as well as related questions (concerning
the determination and uniqueness of γ and ω = (ω1 , ω2 , ω3 ), the smoothness of the boundary of V ,
etc.) are discussed in the literature cited below.
⊙ Literature for Section 12.5: G. G. Stokes (1849), H. Helmholtz (1858), P. M. Morse and H. Feshbach
(1953, Section 1.5), M. E. Gurtin (1962), D. A. W. Pecknold (1971), A. C. Eringen and E. S. Suhubi (1975),
L. Morino (1986), G. A. Korn and T. M. Korn (2000), G. B. Arfken and H. J. Weber (2005).
12.6. Elasticity Equations 1071
12.6 Elasticity Equations

12.6.1 Elasticity Equations in Various Coordinate Systems
◮ Vector forms of the elasticity equations.
The linear elasticity equations (in displacements) have the form
ρutt = µ∆u + (λ + µ)∇ div u + ρf, (1)
where u is the displacement vector, t is time, ρ is the material density, µ is the shear
modulus (or the modulus of elasticity of the second kind), λ is the Lamé coefficient, ∆ is
the Laplace operator, and f is the mass force.
The vector equation (1) is a short representation of the elasticity equations, which de-
scribe the motion of a linearly elastic isotropic body (medium) in the case of small strains.
The vector elasticity equation (1) is sometimes represented as follows:
utt = c22 ∆u + (c21 − c22 )∇ div u + f, (2)
where c1 and c2 are the longitudinal and transverse wave velocities, respectively, which are
determined by the formulas
1/2 1/2
λ + 2µ µ
c1 = , c2 = . (3)
ρ ρ
Remark 12.3. The longitudinal waves, which propagate at the velocity c1 , are also known as
primary waves (or p-waves), because these waves are recorded first in geophysics, while the trans-
verse waves, which propagate at the velocity c2 , are also known as secondary waves (or s-waves).
In view of the identity ∇ div u ≡ ∆u + curl curl u and the formula curl u ≡ ∇ × u,
Eqs. (1) and (2) can be represented in the form
ρutt = (λ + 2µ)∇ div u − µ curl curl u + ρf,

utt = c21 ∇ div u − c22 curl curl u + f.
◮ Elasticity equations in rectangular Cartesian coordinates.

In the rectangular Cartesian coordinates (x, y, z) the elasticity equations (1) have the form
2
∂ 2 u1 ∂ u1 ∂ 2 u1 ∂ 2 u1 ∂
ρ 2 =µ + + + (λ + µ) div u + ρf1 ,
∂t ∂x2 ∂y 2 ∂z 2 ∂x

∂ 2 u2 ∂ 2 u2 ∂ 2 u2 ∂ 2 u2 ∂
ρ 2 =µ + + + (λ + µ) div u + ρf2 ,
∂t ∂x2 ∂y 2 ∂z 2 ∂y
2 (4)
∂ 2 u3 ∂ u3 ∂ 2 u3 ∂ 2 u3 ∂
ρ 2 =µ + + + (λ + µ) div u + ρf3 ,
∂t ∂x2 ∂y 2 ∂z 2 ∂z
∂u1 ∂u2 ∂u3
div u ≡ + + ,
∂x ∂y ∂z
where u1 , u2 , u3 are the displacement vector components and f1 = f1 (x, y, z, t), f2 =

f2 (x, y, z, t), f3 = f3 (x, y, z, t) are the mass force components.
Equations (4) can be represented in the form
X ∂3
∂ 2 ui
ρ 2 = Sij + ρfi , j = 1, 2, 3,
∂t ∂xj
j=1
where we have used the notation x1 = x, x2 = y, x3 = z and the stress tensor components
Sij are expressed via the displacement vector components as follows:
3
X
∂ui ∂ui
Sii = λ div u + 2µ , div u = ;
∂xi ∂xi
i=1

∂ui ∂uj
Sij = Sji = µ + for i 6= j.
∂xj ∂xi
◮ Elasticity equations in the cylindrical coordinates.
In the cylindrical coordinates (r, ϕ, z) the elasticity equations (1) have the form

∂ 2 ur ur 2 ∂uϕ ∂
ρ 2 = µ ∆ur − 2 − 2 + (λ + µ) div u + ρfr ,
∂t r r ∂ϕ ∂r

∂ 2 uϕ uϕ 2 ∂ur 1 ∂
ρ 2 = µ ∆uϕ − 2 + 2 + (λ + µ) div u + ρfϕ ,
∂t r r ∂ϕ r ∂ϕ
∂ 2 uz ∂
ρ 2 = µ∆uz + (λ + µ) div u + ρfz ,
∂t ∂z
∂ur 1 ∂uϕ ∂uz ur
div u ≡ + + + ,
∂r r ∂ϕ ∂z r
where ur , uϕ , and uz are the displacement vector components, fr = fr (r, ϕ, z, t), fϕ =

fϕ (r, ϕ, z, t), and fz = fz (r, ϕ, z, t) are the mass force components, and the Laplace oper-
ator ∆ is given by the formula
∂2 1 ∂ 1 ∂2 ∂2
∆≡ + + + .
∂r 2 r ∂r r 2 ∂ϕ2 ∂z 2
◮ Elasticity equations in the spherical coordinates.
In the spherical coordinates (R, θ, ϕ) the elasticity equations (1) have the form
∂ 2 uR ∂
ρ 2
= (λ + µ) div u + µ∆uR
∂t ∂R
2µ ∂uθ 1 ∂uϕ
− 2 + + uR + cot θ uθ + ρfR ,
R ∂θ sin θ ∂ϕ
∂ 2 uθ 1 ∂
ρ 2
= (λ + µ) div u + µ∆uθ
∂t R ∂θ
µ 2 ∂uR ∂uϕ
+ 2 2 2 sin θ − 2 cos θ − uθ + ρfθ ,
R sin θ ∂θ ∂ϕ
∂ 2 uϕ 1 ∂
ρ 2 = (λ + µ) div u + µ∆uϕ
∂t R sin θ ∂ϕ

µ ∂uR ∂uθ
+ 2 2 2 sin θ + 2 cos θ − uϕ + ρfϕ ,
R sin θ ∂ϕ ∂ϕ
1 ∂ 1 ∂ 1 ∂uϕ
div u ≡ 2 R2 uR + sin θuθ + ,
R ∂R R sin θ ∂θ R sin θ ∂ϕ
where uR , uθ , and uϕ are the displacement vector components and the Laplace operator ∆
is given by the formula

1 ∂ 2 ∂ 1 ∂ ∂ 1 ∂2
∆≡ 2 R + 2 sin θ + 2 2 .
R ∂R ∂R R sin θ ∂θ ∂θ R sin θ ∂ϕ2
◮ Invariant transformations.
The elasticity equations (1) are invariant under the following transformations:
u=e
u + a1 x + a2 y + a3 y + a4 t + b,
e + curl Ψ◦ ,
u=u
u=e
u + ∇Φ,
where an and b are arbitrary constant vectors and the vector function Ψ◦ and the scalar
function Φ are arbitrary solutions of the wave equations
Ψ◦tt − c22 ∆Ψ◦ = 0,
Φtt − c21 ∆Φ = 0.
12.6.2 Various Forms of Decompositions of Homogeneous Elasticity

Equations with f = 0
◮ Decomposition based on two stream functions.
Any solution of the homogeneous elasticity equations (4) with f = 0 can also be represented
as
u1 = ϕx + ψy(1) , u2 = ϕy − ψx(1) + ψz(2) , u3 = ϕz − ψy(2) , (5)
where the functions ϕ = ϕ(x, y, z), ψ (1) = ψ (1) (x, y, z), and ψ (2) = ψ (2) (x, y, z) are
solutions of the wave equations
ϕtt − c21 ∆ϕ = 0, (6)
(1) (2)
ψtt − c22 ∆ψ (1) = 0, ψtt − c22 ∆ψ (2) = 0. (7)
Remark 12.4. For ϕ = 0, formulas (5) satisfy the incompressibility condition
div u = (u1 )x + (u2 )y + (u3 )z = 0,
and hence the functions ψ (1) and ψ (2) can be interpreted as stream functions.
◮ Using the Stokes–Helmholtz representation of the displacement vector.

Every solution of Eqs. (1) with f = 0 can be represented by the formulas
u = ∇ϕ + curl Ψ, (8)
where Ψ = (Ψ1 , Ψ2 , Ψ3 ) is a solution of the vector wave equation
Ψtt − c22 ∆Ψ = 0 (9)
and the function ϕ is a solution of the scalar wave equation (6).
◮ Cauchy–Kovalevskaya solution.
Any solution of Eqs. (1) with f = 0 can be represented in the form
u = 1 [w] + (c21 − c22 )∇ div w, (10)
where the vector function w satisfies the equation
2 1 [w] = 0. (11)
Here and in the following, the d’Alembert operators 1 and 2 are given by
1 ≡ ∂t2 − c21 ∆, 2 ≡ ∂t2 − c22 ∆. (12)
If ut 6≡ 0, then the general solution of Eq. (10) can be represented as the sum
w = w1 + w2 ,
where w1 and w2 are arbitrary solutions of the two simpler equations
1 [w1 ] = 0, 2 [w2 ] = 0.
◮ Chadwick–Trowbridge solution (toroidal–poloidal decomposition).

Any solution of Eqs. (1) with f = 0 can be represented in the form
u = ∇ϕ + curl(xξ) + curl curl(xη), x = (x, y, z),
where ϕ, ξ, and η are scalar functions satisfying the wave equations
1 ϕ = 0, 2 ξ = 0, 2 η = 0. (13)
◮ Two other representations of solutions via three scalar functions.

The solutions Eqs. (1) with f = 0 can also be represented in the following two forms:
u = ∇ϕ + curl(aξ + bη), a · b 6= 0,
u = ∇ϕ + curl(aξ + xη), |a| =
6 0,
where ϕ, ξ, and η are scalar functions satisfying the wave equations (13).
◮ Steady-state elasticity equations. The biharmonic equation for the displacement

vector.
For the steady-state elasticity equations (1) with ut = 0, the displacement field u satisfies
the biharmonic equation
∆∆u = 0. (14)
Remark 12.5. Equation (13) remains valid for the steady-state nonhomogeneous elasticity equa-
tions (1) if the mass force satisfies the conditions div f = 0 and curl f = 0. (This was proved by
Cauchy in 1828.)
◮ Steady-state elasticity equations. Slobodyanskii’s solution.
The solution of the steady-state elasticity equations (1) with f = 0 can be represented in the
form
2(2µ + λ)
u= w + (x · ∇)w − x div w,
µ+λ
where the vector function w satisfies the Laplace equation
∆w = 0.
12.6.3 Various Forms of Decompositions for Nonhomogeneous

Elasticity Equations
◮ Lamé decomposition of the elasticity equations.
Assume that the mass force in the elasticity equations (1) is represented as the sum of
potential and solenoidal components (this is the Stokes–Helmholtz representation of the
mass force; see Section 12.5.3),
f = ∇γ + curl ω.
In this case, the displacement can be represented by formula (8), where the scalar function
ϕ and the vector function Ψ satisfy the nonhomogeneous wave equations
ϕtt − c21 ∆ϕ = γ,
Ψtt − c22 ∆Ψ = ω.
◮ Cauchy–Kovalevskaya decomposition not requiring the force to split into compo-

nents.
Any solution of Eqs. (1) can be represented in the form (10), where the vector function w
satisfies the equation
2 1 [w] = f.
Here the d’Alembert operators 1 and 2 are defined in (12).

◮ Integro-differential representation of solutions.

A fairly broad class of solutions of the elasticity vector equations (1) can be represented in
the form
Z
(c21 − c22 ) ′ r dv ′
u=ϑ+ ∇ div ϑ x , t − , r = |x′ − x|,
4πc21 r≤c1 t c1 r
where the integration is over the ball of radius c1 t centered at x and the vector function ϑ
satisfies the nonhomogeneous wave equation
ϑtt − c22 ∆ϑ = f.
◮ Papkovich–Neuber decomposition for the steady-state elasticity equations.

Let ut = 0. Then the solution of the elasticity equations (1) can be represented in the form
h λ+µ i
u = v+∇ Φ− x·v , (15)
2(λ + 2µ)
where the vector function v and the scalar function Φ satisfy the Poisson equations
ρ(λ + µ)
µ∆v = −ρf, ∆Φ = − x · f. (16)
2µ(λ + 2µ)
Remark 12.6. The Papkovich–Neuber solution (15)–(16) admits a generalization and can be
∆H + µ + λ
u = v + ∇ Φ − x · (Qv) , Q= ,
2(λ + 2µ)
where H is an arbitrary constant coefficient linear differential operator, the vector function v is a
solution of the first equation in (16), and the scalar function Φ satisfies the Poisson equation
ρ ρ
∆Φ + x · (Qf) + H[div f] = 0.
µ µ(λ + 2µ)
12.6.4 Cauchy Problem and Its Solution. Fundamental Solution

Matrix
The 3D Cauchy problem for the elasticity equations (1) in the infinite domain R3 is deter-
mined by the initial conditions
u|t=0 = u0 (x), ut |t=0 = u1 (x), (17)
where u0 (x) and u1 (x) are given vector functions.

The main role in the construction of the solution of the Cauchy problem is played by the
fundamental solutions of Eq. (1), which represent the displacements of an infinite elastic
space subjected to a lumped force applied at a point x0 ≡ (x01 , x02 , x03 ) with magnitude δ(t),
where δ(t) is the Dirac delta function.
12.7. Stokes Equations for Viscous Incompressible Fluids 1077
If the force acts in the direction of the xk -axis (k = 1, 2, 3), then the displacement
(k)
components uj (x, t) = Γkj (k, j = 1, 2, 3) are given by the formulas

1 ∂r ∂r δ(ζ1 ) ∂r ∂r δ(ζ2 )
Γkj (x − x0 , t) = + δkj −
4πρ ∂xk ∂xj c21 r ∂xk ∂xj c2 r
2
r r
+ θ(ζ1 ) − θ(ζ2 ) + ζ1 θ(ζ1 ) − ζ2 θ(ζ2 ) ,
c1 c2
where c1 and c2 are the characteristic velocities (3) and we have used the notation
r r
r = |x − x0 |, ζ1 = t − , ζ2 = t − ,
c1 c2
( (
1 if ζ ≥ 0, 1 if k = j,
θ(ζ) = δkj =
0 if ζ < 0, 0 if k = 6 j.
The solution of the Cauchy problem (1) with the initial conditions (17) can be repre-
sented via the symmetric fundamental solution matrix Γ(x − x0 , t) = kΓkj (x − x0 , t)k by
the Volterra formula
Z Z
0 ∂ 0
u(x , t) = Γ(x − x, t)u0 (x) dx + Γ(x0 − x, t)u1 (x) dx
∂t r≤c1 t r≤c1 t
Z tZ
+ Γ(x0 − x, t − τ )f(x, τ ) dx dτ,
0 r≤c1 (t−τ )
where we have used the notation dx = dx1 dx2 dx3 .

⊙ Literature for Section 12.6: C.-L.-M.-H. Navier (1821), A.-L. Cauchy (1828), S. Kovalevski (1885),
C. Somigliana (1889), B. Galerkin (1930), P. F. Papkovich (1932), U. Neuber (1934), M. G. Slobodianskii
(1959), P. Chadwick and E. A. Trowbridge (1967), M. E. Gurtin (1972), W. Nowacki (1975), A. C. Eringen
and E. S. Suhubi (1975), Mathematical Encyclopedia (1979, pp. 149–154), D. S. Chandrasechariah (1988),
A. D. Polyanin and S. A. Lychev (2014a).
12.7 Stokes Equations for Viscous Incompressible

Fluids
12.7.1 Stokes Equations in Various Coordinate Systems
◮ Vector form of the Stokes equations.
The closed system of equations for slow (creeping) flows of a viscous incompressible New-
tonian fluid has the form
ut = −∇p + ν∆u + f, (1)

div u = 0, (2)
where u = (u1 , u2 , u3 ) is the fluid velocity, t is time, ν is the kinematic viscosity of the
fluid, p is the ratio of the pressure to the fluid density, f = (f1 , f2 , f3 ) is the mass force
(e.g., the gravitational force), ∇ is the gradient operator, and ∆ is the Laplace operator.
The vector equation (1) is a concise representation of the Stokes equations, and Eq. (2)
is the continuity equation. In what follows, for brevity, system (1)–(2) will be referred to
as the Stokes equations (without distinguishing the continuity equation), and the ratio p of
the pressure to the fluid density will be referred to as the pressure.
To derive the Stokes equations, one linearizes the Navier–Stokes equations using the
order-of-magnitude relations
|u| ∼ ε, |p − p0 | ∼ ε, |f| ∼ ε,
where ε is a small parameter.

We apply the operator div to Eq. (1), take into account (2), and obtain an equation for
the pressure,
∆p = div f.
The Stokes equations (1)–(2) are invariant under the transformation
u + curl Ψ◦ ,
u=e p = pe,
where Ψ◦ is an arbitrary solution of the heat equation Ψ◦t − ν∆Ψ◦ = 0.

The Stokes equations (1)–(2) are also invariant under the transformation
u=e
u + ∇Φ, p = pe − Φt + p0 (t),
where p0 (t) is an arbitrary function and Φ = Φ(x, t) is an arbitrary solution of the Laplace
equation ∆Φ = 0.
◮ Problems for the Stokes equations.

Cauchy problem. One seeks a solution of Eqs. (1)–(2) in R3 for t > 0 with the initial
condition
u|t=0 = u0 (x), (3)
where u0 (x) is a given vector function.
Interior initial-boundary value problem. To describe the motion of a fluid in an im-
movable bounded domain V , one supplements the initial condition (3) with the no-slip
condition on the boundary S of the domain,
u|x∈S = 0. (4)
Exterior initial-boundary value problems. In the problem on the flow, homogeneous

at infinity with velocity a, past an immovable bounded body, one supplements the initial
condition (3) with the boundary conditions
u|x∈S = 0, u||x|→∞ → a, (5)
where S is the surface of the body.

In the problem on the motion of a bounded body at a constant velocity a in an immov-

able fluid, the initial condition (3) is supplemented with the boundary conditions
u|x∈S = a, u||x|→∞ → 0. (6)
We point out that no conditions on the pressure p are needed in the above-stated prob-
lems for the Stokes equations.
◮ Stokes equations in the rectangular Cartesian coordinates.
In the rectangular Cartesian coordinates (x, y, z), the Stokes equations (1)–(2) have the
form 2
∂u1 ∂p ∂ u1 ∂ 2 u1 ∂ 2 u1
=− +ν + + + f1 ,
∂t ∂x ∂x2 ∂y 2 ∂z 2
2
∂u2 ∂p ∂ u2 ∂ 2 u2 ∂ 2 u2
=− +ν + + + f2 ,
∂t ∂y ∂x2 ∂y 2 ∂z 2
2 (7)
∂u3 ∂p ∂ u3 ∂ 2 u3 ∂ 2 u3
=− +ν + + + f3 ,
∂t ∂z ∂x2 ∂y 2 ∂z 2
∂u1 ∂u2 ∂u3
+ + = 0,
∂x ∂y ∂z
where u1 , u2 , u3 are the fluid velocity components.
Remark 12.7. For f1 = f2 = f3 = 0, the scaling transformation
b2 aν
x = bx̄, y = bȳ, z = bz̄, t= t̄, u1 = aū1 , u2 = aū2 , u3 = aū3 , p= p̄,
ν b
where a and b are arbitrary (nonzero) constants, reduces system (7) to the same form with ν = 1.
◮ Stokes equations in the cylindrical coordinates.
The cylindrical coordinates (r, ϕ, z) are related to the rectangular Cartesian coordinates
(x, y, z) by
p
r= x2 + y 2 , tan ϕ = y/x, z=z (sin ϕ = y/r);
x = r cos ϕ, y = r sin ϕ, z = z,
where 0 ≤ ϕ ≤ 2π.
The fluid velocity components in the cylindrical coordinates are expressed in terms of
those in the rectangular Cartesian coordinates as follows:
ur = u1 cos ϕ + u2 sin ϕ, uϕ = u2 cos ϕ − u1 sin ϕ, uz = u3 .
Remark 12.8. For uz = 0, the cylindrical coordinates r and ϕ are also used as the polar coordi-
nates on the xy-plane.
The Stokes equations (1)–(2) in the cylindrical coordinates become

∂ur ∂p ur 2 ∂uϕ
=− + ν ∆ur − 2 − 2 + fr ,
∂t ∂r r r ∂ϕ

∂uϕ 1 ∂p uϕ 2 ∂ur
=− + ν ∆uϕ − 2 + 2 + fϕ ,
∂t r ∂ϕ r r ∂ϕ (8)
∂uz ∂p
=− + ν∆uz + fz ,
∂t ∂z
∂ur 1 ∂uϕ ∂uz ur
+ + + = 0,
∂r r ∂ϕ ∂z r
where ur , uϕ , uz are the fluid velocity components and the Laplace operator ∆ is given by
the formula
∂2 1 ∂ 1 ∂2 ∂2
∆≡ + + + .
∂r 2 r ∂r r 2 ∂ϕ2 ∂z 2
◮ Stokes equations in the spherical coordinates.
The spherical coordinates (R, θ, ϕ) are related to the rectangular Cartesian coordinates
(x, y, z) by
p
z y y
R= x2 + y2 + z2, θ = arccos , tan ϕ = sin ϕ = p ;
R x x2 + y 2
x = R sin θ cos ϕ, y = R sin θ sin ϕ, z = R cos θ,
where 0 ≤ θ ≤ π and 0 ≤ ϕ ≤ 2π.

The Stokes equations (1)–(2) in the spherical coordinates become

∂uR ∂p 2ν ∂uθ 1 ∂uϕ
=− + ν∆uR − 2 + + uR + cot θ uθ + fR ,
∂t ∂R R ∂θ sin θ ∂ϕ

∂uθ 1 ∂p ν ∂uR ∂uϕ
=− + ν∆uθ + 2 2 2 sin2 θ − 2 cos θ − uθ + fθ ,
∂t R ∂θ R sin θ ∂θ ∂ϕ

∂uϕ 1 ∂p ν ∂uR ∂uθ
=− + ν∆uϕ + 2 2 2 sin θ + 2 cos θ − uϕ + fϕ ,
∂t R sin θ ∂ϕ R sin θ ∂ϕ ∂ϕ
∂ ∂ ∂
R2 sin θuR + R sin θuθ + Ruϕ = 0,
∂R ∂θ ∂ϕ
where uR , uθ , uϕ are the fluid velocity components and the Laplace operator ∆ is given by
the formula

1 ∂ 2 ∂ 1 ∂ ∂ 1 ∂2
∆≡ 2 R + 2 sin θ + 2 2 .
R ∂R ∂R R sin θ ∂θ ∂θ R sin θ ∂ϕ2
12.7.2 Various Forms of Decompositions for the Stokes Equations

with f = 0
Each solution of Eqs. (1)–(2) with f = 0 can be represented in the form
u = ∇ϕ + v, p = p0 − ϕt ,
(9)
v = (v1 , v2 , v3 ), v1 = ψy(1) , v2 = −ψx(1) + ψz(2) , v3 = −ψy(2) ,
where p0 = p0 (t) is an arbitrary function, ψ (1) = ψ (1) (x, t) and ψ (2) = ψ (2) (x, t) are
arbitrary solutions of the heat equation
ψt − ν∆ψ = 0, (10)
and the function ϕ = ϕ(x, t) is a solution of the Laplace equation
∆ϕ = 0. (11)
Remark 12.9. The functions ψ (1) and ψ (2) in (9) can be interpreted as two stream functions
permitting one to eliminate the continuity equation from the 3D incompressible fluid equations
(where the fluid velocity components are denoted by v1 , v2 , v3 ). In the special case of ψ (2) = 0,
we obtain the usual representation of the fluid velocity components for 2D planar flows with v3 = 0
and with one stream function.
Remark 12.10. The Laplace equation (11) does not include time t explicitly, but the arbitrary
constants Cn occurring in the solution of this steady-state equation may depend on t arbitrarily.
◮ Using the Stokes–Helmholtz representation of the fluid velocity.

u = ∇ϕ + curl Ψ, p = p0 − ϕt , (12)
where p0 = p0 (t) is an arbitrary function, Ψ = (Ψ1 , Ψ2 , Ψ3 ) is an arbitrary solution of the

vector heat equation
Ψt − ν∆Ψ = 0, (13)
and the function ϕ is a solution of the Laplace equation (11).
◮ Special types of decomposition using representations of the vector u via three scalar
functions.
Table 12.1 describes various decompositions of the Stokes equations (1)–(2) with f = 0 on
the basis of a representation of solutions by formulas of the form (12),
u = ∇ϕ + curl Ψn , p = p0 − ϕt , (14)
where p0 = p0 (t) is an arbitrary function and the vector function Ψn (n = 1, . . . , 7) can be

expressed in a special way via two scalar functions ψ (1) and ψ (2) . To ensure completeness
Table 12.1: Various decompositions of the Stokes equations (1)–(2) with f = 0 based on
the representation (14) of the solutions
n Form of vector function Ψn Order of decomposition Remarks
Is equivalent to
1 e1 ψ (2) + e3 ψ (1) First
the representation (9)
2 aψ (1) + bψ (2) First a · b 6= 0
(1) (2)
3 aψ + curl(bψ ) Second a · b 6= 0
(1) (2)
4 aψ + xψ First |a| 6= 0
(1) (2)
5 aψ + curl(xψ ) Second |a| 6= 0
(1) (2)
6 xψ + curl(aψ ) Second |a| 6= 0
(1) (2)
7 xψ + curl(xψ ) Second No remark
and clarity, Table 12.1 also includes the decomposition determined by formulas (9) and
Eqs. (10)–(11). We use the following notation: e1 , e2 , and e3 are unit vectors in a Cartesian
coordinate system, and a and b are arbitrary constant vectors.
For all solutions of the Stokes equations (1)–(2) based on formula (14) and the seven
distinct representations of the vector function Ψn in Table 12.1, the functions ψ (1) =
ψ (1) (x, t) and ψ (2) = ψ (2) (x, t) are arbitrary solutions of the heat equation (10), and the
function ϕ satisfies the Laplace equation (11).
Note that all above-described special representations of solutions determined by for-
mula (14) and Table 12.1 contain one unknown function less than the Stokes–Helmholtz
representation (12).
◮ Steady-state Stokes equations. Solution of Slobodyanskii type.

The solution of the steady-state Stokes equations (1)–(2) with f = 0 can be represented in
the form
u = 2w + (x · ∇)w − x div w, p = p0 − 2νρ div w,
where p0 is an arbitrary constant and the vector function w satisfies the Laplace equation
∆w = 0.
12.7.3 Various Forms of Decompositions for the Stokes Equations

with f 6= 0
◮ Using the Stokes–Helmholtz representation of the mass force.
Let the mass force in the Stokes equations (1)–(2) be written as the sum of potential and
solenoidal components (see Section 12.5.3)
f = ∇γ + curl ω. (15)
Then the solution of system (1)–(2) can be represented in the form
u = ∇ϕ + curl Ψ, p = p0 − ϕt + γ, (16)
where p0 = p0 (t) is an arbitrary function and the functions ϕ and Ψ satisfy the independent
equations
Ψt − ν∆Ψ = ω,
(17)
∆ϕ = 0.
◮ Decomposition not requiring the mass force to split into components.

The solution of the Stokes equations (1)–(2) can be represented in the form
u = − curl curl U ≡ ∆U − ∇ div U, p = p0 (t) + div(Ut − ν∆U),
where p0 (t) is an arbitrary function and the vector function U satisfies the equation
∆(∂t − ν∆)[U] = f.
◮ Papkovich–Neuber decomposition for the steady-state Stokes equations.

Let ut = 0. Then the solution of the Stokes equations (1)–(2) can be represented in the
form
1
u=v+∇ Φ− x·v , p = p0 − νρ div v, (17a)
2
where p0 is an arbitrary constant and the vector function v and the scalar function Φ satisfy
the Poisson equations
1
ν∆v = −f, ν∆Φ = − x · f. (17b)
2
Remark 12.11. The solution (17a)–(17b) admits a generalization and can be represented in the
form
1
u = v + ∇ Φ + x · (Qv) , p = p0 − νρ div v, Q = ∆H − ,
2
where H is an arbitrary constant coefficient linear differential operator, the vector function v is a
solution of the first equation in (17b), and the scalar function Φ satisfies the Poisson equation
ν∆Φ = x · (Qf) + 2H[div f].
◮ Incomplete asymmetric decomposition of the Stokes equations.

Each solution of system (1)–(2) with f = (f1 , f2 , f3 ) can be represented in the form
u = ∇ϕ + e2 v2 + e3 v3 , p = −ϕt + ν∆ϕ + F,
Z x
(18)
F = F (x, t) = F (x, y, z, t) = f1 (x1 , y, z, t) dx1 ,
0
where e2 and e3 are the unit vectors of the Cartesian coordinate y- and z-axes, the scalar
functions v2 = v2 (x, t) and v3 = v3 (x, t) satisfy the two independent linear nonhomoge-
neous heat equations
(v2 )t − ν∆v2 = f2 − Fy , (v3 )t − ν∆v3 = f3 − Fz , (19)
and the function ϕ is related to v2 and v3 by the equation
∆ϕ + (v2 )y + (v3 )z = 0. (20)
◮ Incomplete symmetric decomposition of the Stokes equations.

Every solution of system (1)–(2) can also be represented in the symmetric form
u = ∇ϕ + v, p = −ϕt + ν∆ϕ + G, (21)
where the vector function v = (v1 , v2 , v3 ) satisfies the independent linear nonhomogeneous
equation
vt − ∆v = f − ∇G (22)
and the functions ϕ and v are related by the single equation
∆ϕ + div v = 0. (23)
Formulas (21) and Eqs. (22) include an arbitrary scalar function G = G(x, t).
By setting v1 = 0 and G = F in (21)–(23), we obtain an asymmetric representation of
the solutions (18)–(20).
12.7.4 General Solution of the Steady-State Homogeneous Stokes

Equations
The general solution of the steady-state homogeneous Stokes equations (1)–(2) with ut ≡ 0
and f = 0 is given by the formulas
X∞
u= ∇ϕn + ∇ × (xψn )
n=−∞
∞
X
1 (n + 3) 2 n
+ |x| ∇pn − xpn ,
ν 2(n + 1)(2n + 3) (n + 1)(2n + 3)
n=−∞,n6=−1
X p
p= pn , x = (x, y, z), |x| = R = x2 + y 2 + z 2 ,
where ϕn , ψn , and pn are arbitrary spherical harmonics of order n, which satisfy the
Laplace equations
∆ϕn = 0, ∆ψn = 0, ∆pn = 0
and have the form
n
X
ϕn = |x|n Pnm (cos θ) anm cos(mϕ) + e
anm sin(mϕ) ,
m=0
Xn

ψn = |x|n Pnm (cos θ) bnm cos(mϕ) + ebnm sin(mϕ) ,
m=0
Xn

pn = |x|n Pnm (cos θ) cnm cos(mϕ) + e
cnm sin(mϕ) ,
m=0
where anm , eanm , bnm , ebnm , cnm , and e

cnm are arbitrary constants and the Pnm are the
associated Legendre polynomials,
dm 1 dn 2
Pnm (ζ) = (1 − ζ 2 )m/2 P n (ζ), Pn (ζ) = (ζ − 1)n ,
dζ m n!2n dζ n
n = 1, 2, 3, . . . , m = 0, 1, 2, . . .
12.7.5 Solution of the Steady-State Nonhomogeneous Stokes

Equations
Consider the auxiliary problems
ν∆uk − ∇pk = δ(x − y), div uk = 0, k = 1, 2, 3,
(24)
|uk | → 0 as |x| → ∞, |pk | → 0 as |x| → ∞,
where δ(x) is the Dirac delta function, and y plays the role of a parameter.
The solution of problems (24) is given by the Lorentz formulas
3
X
1 δkj (xk − yk )(xj − yj )
uk (x, y) = ukj (x, y)ej , ukj (x, y) = − + ,
8πν |x − y| |x − y|3
j=1
xk − y k
pk (x, y) = − ,
4π|x − y|3
(25)
where ej is the unit vector of the jth coordinate axis and δkj is the Kronecker delta.
The solution of the steady-state nonhomogeneous Stokes equations (1) and (2) in a
domain V with sufficiently smooth boundary ∂V can be expressed via the functions (25)
and the mass force f = (f1 , f2 , f3 ) as follows:
Z X
3 Z X
3
u(x) = uk (x, y)fk (y) dVy , p(x) = pk (x, y)fk (y) dVy , (26)
V k=1 V k=1
where dVy = dy1 dy2 dy3 . Formulas (26) are the volume potentials for the steady-state
Stokes equations.
12.7.6 Solution of the Cauchy Problem

Consider the Cauchy problem for the Stokes equations (1)–(2) with the homogeneous initial
condition
u = 0 at t = 0. (27)
The solution of problem (1)–(2), (27) for x ∈ R3 and t > 0 has the form
Z tZ
u = (u1 , u2 , u3 ), uk = uk (x, t) = wk (x − y, t − τ ) · f(y, τ ) dy dτ, (28)
0 R3
Z tZ 3
X
p = p(x, t) = qk (x − y, t − τ )fk (y, τ ) dy dτ, dy = dy1 dy2 dy3 , (29)
0 R3 k=1
where the fk (x, t) are the components of the mass force f and the functions wk and qk are
given by the formulas
Z
1 ∂ Γ(x − y, t)
wk (x, t) = Γ(x, t)ek + ∇ dy, (30)
4π ∂xk R3 |y|

∂ 1 1 |x|2
qk (x, t) = − δ(t), Γ(x, t) = exp − . (31)
∂xk 4π|x| (4νπt)3/2 4νt
Here ek is the unit vector of the xk -axis and δ(t) is the Dirac delta function.
The vectors wk in (28) and (30) can be represented in the form
wk (x, t) = curl curl Uk ≡ −∆Uk + ∇ div Uk ,
where
Z Z z
1 Γ(x − y, t) 1 |x| 2
Uk (x, t) = ek dy = θ √ ek , θ(z) = e−ξ dξ.
4π R3 |y| 2π 3/2 |x| 2 νt 0
⊙ Literature for Section 12.7: H. Lamb (1945), R. Berker (1963), J. Happel and H. Brenner (1965), O. A. La-
dyzhenskaya (1969), G. K. Batchelor (1970), H. Schlichting (1981), L. G. Loitsyanskiy (1996), A. D. Polyanin
and A. V. Vyazmin (2012, 2013a), A. D. Polyanin and A. I. Zhurov (2013), A. D. Polyanin and S. A. Lychev
(2014a).
12.8 Oseen Equations for Viscous Incompressible Fluids

12.8.1 Vector Form of Oseen Equations. Some Remarks
The Oseen equations for a viscous incompressible Newtonian fluid have the form
ut + (a · ∇)u = −∇p + ν∆u + f,

(1)
div u = 0,
where (a · ∇)u = a1 ux + a2 uy + a3 uz , a1 , a2 , and a3 are some constants and the remaining

notation is the same as in the Stokes equations (see Eqs. (1)–(2) in Section 12.7).
The Oseen equations (1) describe the evolution of small perturbations of velocity com-
ponents for a viscous incompressible fluid near the unperturbed flow with constant velocity
vector a = (a1 , a2 , a3 ) and constant pressure p0 . To derive the Oseen equations, one lin-
earizes the Navier–Stokes equations on the basis of the order-of-magnitude relations
|u − a| ∼ ε, |p − p0 | ∼ ε, |f| ∼ ε,
where ε is a small parameter.

In the special case of a1 = a2 = a3 = 0, the Oseen equations become the Stokes
equations.
In the time-dependent case of ut 6≡ 0, the transition in the Oseen equations (1) from x,
y, z, and t to the new independent variables x̄ = x − a1 t, ȳ = y − a2 t, z̄ = z − a3 t, and
t̄ = t results in the Stokes equations (see Eqs. (1)–(2) in Section 12.7).
The Oseen equations (1) are invariant under the transformation
u + curl Ψ◦ ,
u=e p = pe,
where Ψ◦ is an arbitrary solution of the equation Ψ◦t + (a · ∇)Ψ◦ − ν∆Ψ◦ = 0.

From (1), we obtain an equation for the pressure,
∆p = div f.
12.8. Oseen Equations for Viscous Incompressible Fluids 1087
12.8.2 Various Forms of Decompositions for the Oseen Equations

with f = 0
Any solution of the Oseen equations (1) with f = 0 can be represented by the formulas
u = ∇ϕ + v, p = p0 − ϕt − (a · ∇)ϕ,
(2)
where p0 = p0 (t) is an arbitrary function, ψ (1) = ψ (1) (x, t) and ψ (2) = ψ (2) (x, t) are some
solutions of the convective heat equation
ψt + (a · ∇)ψ − ν∆ψ = 0, (3)
and the function ϕ = ϕ(x, t) is a solution of the Laplace equation ∆ϕ = 0.
Every solution of the Oseen equations (1) with f = 0 can be represented by the formulas
u = ∇ϕ + curl Ψ, p = p0 − ϕt − (a · ∇)ϕ, (4)
where p0 = p0 (t) is an arbitrary function, Ψ = (Ψ1 , Ψ2 , Ψ3 ) is a solution of the convective

heat vector equation
Ψt + (a · ∇)Ψ − ν∆Ψ = 0,
and the function ϕ is a solution of the Laplace equation ∆ϕ = 0.
◮ Decomposition of the steady-state Oseen equations.
Let the vector a be directed along the x-axis; i.e., a = ae1 . Then the solution of the steady-
state Oseen equations (1) can be represented in the form
1 a
u = ∇ϕ + ∇θ − e1 θ + e2 ψz − e3 ψy , p = p0 − aρϕx , k= , (5)
2k 2ν
where p0 is an arbitrary constant and the functions ϕ, θ, and ψ satisfy the equations
∆ϕ = 0, (∆ − 2k∂x )θ = 0, (∆ − 2k∂x )ψ = 0.
The transformations
e
θ = ekx θ, ψ = ekx ψe
reduce the last two equations to the Helmholtz equations
∆θe − k2 θe = 0, ∆ψe − k 2 ψe = 0.
12.8.3 Various Forms of Decompositions for the Oseen Equations

with f 6= 0
Let the mass force in the Oseen equations (1) be represented as the sum f = ∇γ + curl ω
(see Section 12.5.3). Then the solution of system (1) can be represented by the formulas
u = ∇ϕ + curl Ψ, p = p0 − ϕt − (a · ∇)ϕ + γ, (6)
where p0 = p0 (t) is an arbitrary function, the vector function Ψ satisfies the equation
Ψt + (a · ∇)Ψ − ν∆Ψ = ω,

The solution of the Oseen equations (1) can be represented in the form
u = − curl curl U ≡ ∆U − ∇ div U, p = p0 (t) + div[Ut + (a · ∇)U − ν∆U],
∆[Ut + (a · ∇)U − ν∆U] = f.
◮ Incomplete symmetric decomposition of the Oseen equations.

Any solution of the Oseen equations (1) can also be represented by the formulas
u = ∇ϕ + v, p = −ϕt − (a · ∇)ϕ + ν∆ϕ + G, (7)
equation
vt + (a · ∇)v − ∆v = f − ∇G (8)
and the function ϕ satisfies Eq. (23) in Section 12.7. Formulas (7) and Eqs. (8) contain an
arbitrary scalar function G = G(x, t).
By setting v1 = 0 and Gx = f1 in (7)–(8), we obtain an asymmetric representation of
solutions of the Oseen equations (1).
12.8.4 Oseen Equations with Variable Coefficients

The coefficients of the Oseen equations (1) may arbitrarily depend on time, a1 = a1 (t),
a2 = a2 (t), a3 = a3 (t). This corresponds to the linearization of the Navier–Stokes equations
in a neighborhood of an exact solution on the basis of the order-of-magnitude relations
|u − a| ∼ ε, |p − p0 + a′t · x| ∼ ε, |f| ∼ ε,
where ε is a small parameter, a = (a1 , a2 , a3 ), and x = (x, y, z).

12.9. Maxwell Equations for Viscoelastic Incompressible Fluids 1089
R
The
R passage from x,
R y, z, t to the new independent variables x̄ = x − a1 dt, ȳ =
y − a2 dt, z̄ = z − a3 dt, t̄ = t in the Oseen equations with variable coefficients (1)
gives the Stokes equations (see Eqs. (1)–(2) in Section 12.7).
The above-described decompositions of the Oseen equations with constant coefficients
remain valid for the Oseen equations with variable coefficients.
⊙ Literature for Section 12.8: C. W. Oseen (1927), H. Lamb (1945), G. K. Batchelor (1970), A. D. Polyanin
and A. V. Vyazmin (2012, 2013a), A. D. Polyanin and S. A. Lychev (2014a).
12.9 Maxwell Equations for Viscoelastic Incompressible

Fluids
12.9.1 Vector Form of the Maxwell Equations
Slow motions of incompressible Maxwell viscoelastic fluids are described by the coupled
linear equations
τ utt + ut = −∇p + ν∆u + f,
(1)
div u = 0,
where τ is the relaxation time and all remaining notation is as in the Stokes equations (see
Eqs. (1)–(2) in Section 12.7).
Some linearized hyperbolic modifications of the Navier–Stokes equations can be re-
duced to Eqs. (1) as well.
In the special case of τ = 0, Eqs. (1) become the Stokes equations.
The Maxwell equations (1) are invariant under the transformation
u + curl Ψ◦ ,
u=e p = pe,
where Ψ◦ is an arbitrary solution of the telegraph equation τ Ψ◦tt + Ψ◦t − ν∆Ψ◦ = 0.

The Maxwell equations (1) are also invariant under the transformation
u=e
u + ∇Φ, p = pe − τ Φtt − Φt + p0 (t),
where p0 (t) is an arbitrary function and Φ is an arbitrary solution of the Laplace equation
∆Φ = 0.
From (1), one obtains an equation for the pressure,
∆p = div f.
12.9.2 Various Forms of Decompositions for the Maxwell Equations

with f = 0
Each solution of the Maxwell equations (1) with f = 0 can be represented by the formulas
u = ∇ϕ + v, p = p0 − τ ϕtt − ϕt ,
(2)
solutions of the telegraph equation
τ ψtt + ψt − ν∆ψ = 0, (3)
Any solution of the Maxwell equations (1) with f = 0 can be represented by the formulas
u = ∇ϕ + curl Ψ, p = p0 − τ ϕtt − ϕt , (4)
where p0 = p0 (t) is an arbitrary function, Ψ = (Ψ1 , Ψ2 , Ψ3 ) is a solution of the vector

telegraph equation
τ Ψt + Ψt − ν∆Ψ = 0,
◮ Special types of decompositions using representations of the vector u via three

scalar functions.
Other representations of solutions of the Maxwell equations (1) with f = 0 can be repre-
sented with the use of formulas of the form (4)
u = ∇ϕ + curl Ψn , p = p0 − τ ϕtt − ϕt (5)
and Table 12.1, where the functions ψ (1) = ψ (1) (x, t) and ψ (2) = ψ (2) (x, t) are arbitrary
solutions of the telegraph equation (3) and the function ϕ satisfies the Laplace equation
∆ϕ = 0.
12.9.3 Various Forms of Decompositions for the Maxwell Equations

with f 6= 0
Let the mass force in the Maxwell equations (1) be written in the form of the sum f =
∇γ + curl ω (see Section 12.5.3). Then the solution of system (1) can be represented by
the formulas
u = ∇ϕ + curl Ψ, p = p0 − τ ϕtt − ϕt + γ, (6)
τ Ψtt + Ψt − ν∆Ψ = ω,

12.10. Equations of Viscoelastic Incompressible Fluids (General Model) 1091

The solution of the Maxwell equations (1) can be represented in the form
u = − curl curl U ≡ ∆U − ∇ div U, p = p0 (t) + div(τ Utt + Ut − ν∆U),
∆(τ ∂tt + ∂t − ν∆)[U] = f.
◮ Incomplete symmetric decomposition of the Maxwell equations.

Each solution of the Maxwell equations (1) can also be represented by the formulas
u = ∇ϕ + v, p = −τ ϕtt − ϕt + ν∆ϕ + G, (7)
equation
τ vtt + vt − ∆v = f − ∇G (8)
By setting v1 = 0 and Gx = f1 in (7)–(8), we obtain an asymmetric representation of
solutions of the Maxwell equations (1).
⊙ Literature for Section 12.9: W. L. Wilkinson (1960), K.-C. Lee and D. A. Finlayson (1986), D. D. Joseph
(1990), K. R. Rajagopal (1993), R. Racke and J. Saal (2012), A. D. Polyanin and A. I. Zhurov (2013),
A. D. Polyanin and A. V. Vyazmin (2013b), A. D. Polyanin and S. A. Lychev (2014a).
12.10 Equations of Viscoelastic Incompressible Fluids

(General Model)
12.10.1 Linearized Equations of Viscoelastic Incompressible Fluids.
Some Models of Viscoelastic Fluids
◮ Vector form of linearized equations for the general model of viscoelastic incom-
pressible fluids.
In the general case, slow motions of incompressible viscoelastic fluids are described by the
coupled linear equations
L[u] = −∇p + f,
(1)
div u = 0,
where L is a linear operator whose properties are described in what follows.
Equations (1) of viscoelastic incompressible fluids are invariant under the transforma-
tion
u=e u + curl Ψ◦ , p = pe,
where Ψ◦ is an arbitrary solution of the equation L[Ψ◦ ] = 0.
Equations (1) of viscoelastic incompressible fluids are also invariant under the transfor-
mation
u=e u + ∇Φ, p = pe − L[Φ] + p0 (t),
where p0 (t) is an arbitrary function and Φ is an arbitrary solution of the Laplace equation
∆Φ = 0.
From (1), we obtain an equation for the pressure,
∆p = div f.
◮ Some models of viscoelastic incompressible fluids.

In general, the linearized rheological equation of state describing slow motions of any
isotropic viscoelastic incompressible fluid can be written as
M[σij ] = −pδij + K[eij ], (2)

∂ui ∂uj
where the σij are the stress tensor components, the eij = 12 ∂x j
+ ∂xi are the strain rate
tensor components (x1 = x, x2 = y, x3 = z), M and K are linear operators in t, and δij
is the Kronecker delta. The linear operator L in Eqs. (1) can be expressed in terms of M
and K as follows:
L[u] = M[ut ] − 12 K[∆u]. (3)
Usually, M and K are differential operators, but they can also be integral, integro-
differential, or differential-difference operators. For example, the linearized equations of
incompressible Oldroyd viscoelastic fluids are determined by the operators
M[σ] = σ + a1 σt , K[e] = 2νe + a2 et ,
where ν is the kinematic viscosity, and a1 and a2 are constants. (Maxwell fluids correspond
to the value a2 = 0 for a1 = τ .)
Table 12.2 lists examples of specific linear operators M and K that determine the equa-
tion of state (2) and are used in some models of incompressible viscoelastic fluids.
Remark 12.12. The fractional derivative of order q, 0 < q < 1 (see row 7 in Table 12.2), is
defined as Z t
[q] 1 d f (s) ds
ft (t) ≡ ,
Γ(1 − q) dt 0 (t − s)q
where Γ(z) is the gamma function. The function σ|t+τ in the last row of the table is evaluated at
t + τ (with a shift in time).
12.10.2 Various Forms of Decompositions for Equations

of Viscoelastic Incompressible Fluids with f = 0
Each solution of Eqs. (1) of viscoelastic fluids with f = 0 can be represented by the formulas
u = ∇ϕ + v, p = p0 − L[ϕ],
(2)
12.10. Equations of Viscoelastic Incompressible Fluids (General Model) 1093
Table 12.2: Linear operators used in various models of incompressible viscoelastic fluids
No. Operator M[σ] Operator K[e] Rheologic model
1 σ 2νe Newtonian
2 σ + τ σt 2νe Maxwell
3 σ + aσt 2νe + bet Oldroyd
4 σ + a1 σt + a2 σtt 2νe + bet Burgers
5 σ + a1 σt + a2 σtt 2νe + b1 et + b2 ett Burgers, generalized
Pm (n) Pk (n)
6 σ + n=1 an σt 2νe + n=1 bn et General differential
[q] [r] With fractional derivatives
7 σ + aσt 2νe + bet
of orders q and r
Rt Integro-differential
8 σ 2νe + a 0 e−λ(t−s) e|t=s ds
Oldroyd
Rt Integro-differential
9 σ 2νe + 0
F (t − s)e|t=s ds
with difference kernel
10 σ|t+τ 2νe Difference-differential
solutions of the equation
L[ψ] = 0, (3)

Each solution of Eqs. (1) of viscoelastic fluids for f = 0 can be represented by the formulas
u = ∇ϕ + curl Ψ, p = p0 − L[ϕ], (4)
where p0 = p0 (t) is an arbitrary function, Ψ = (Ψ1 , Ψ2 , Ψ3 ) is a solution of the vector
equation
L[Ψ] = 0, (5)
12.10.3 Various Forms of Decompositions for Equations

of Viscoelastic Incompressible Fluids with f 6= 0
Let the mass force in the equations (1) of viscoelastic fluids be represented as the sum
f = ∇γ + curl ω (see Section 12.5.3). Then the solution of system (1) can be wtitten as
u = ∇ϕ + curl Ψ, p = p0 − L[ϕ] + γ, (6)
L[Ψ] = ω,
◮ Decomposition not requiring the mass force to split into components (the first ap-
proach).
The solution of Eqs. (1) of viscoelastic fluids can be represented in the form
u = − curl curl U ≡ ∆U − ∇ div U, p = p0 (t) + L[div U],
∆L[U] = f.
◮ Decomposition not requiring the mass force to split into components (the second
approach).
The solution of Eqs. (1) of viscoelastic fluids can be represented in the form

u = ∆U + ∇ ϕ − (LH + 1)[div U] , p = p0 (t) − ρL ϕ − (LH + 1)[div U] ,
where p0 (t) is an arbitrary function, H is an arbitrary constant coefficient linear differential
operator, and the vector and scalar functions U and Φ satisfy the equations
∆L[U] = f, ∆Φ = H[div f].
◮ Incomplete symmetric decomposition.

Each solution of Eqs. (1) of viscoelastic fluids can also be represented by the formulas
u = ∇ϕ + v, p = −L[ϕ] + G, (7)
equation
L[v] = f − ∇G (8)
By setting v1 = 0 and Gx = f1 in (7)–(8), we obtain asymmetric representation of
solutions of Eqs. (1).
⊙ Literature for Section 12.10: J. G. Oldroyd (1956), W. L. Wilkinson (1960), K.-C. Lee and D. A. Finlayson
(1986), D. D. Joseph (1990), K. R. Rajagopal (1993), I. C. Christov (2010), D. Tong (2010), A. D. Polyanin
and A. I. Zhurov (2013), A. D. Polyanin and A. V. Vyazmin (2013b), A. D. Polyanin and S. A. Lychev (2014a).
12.11 Linearized Equations for Inviscid Compressible

Barotropic Fluids
12.11.1 Vector Form of Equations without Mass Forces. Some
Remarks
The equations describing slow motions of an inviscid compressible barotropic fluid without
mass forces have the form
ρ0 ut + ∇p = 0, (1)
2
pt + ρ0 c div u = 0, (2)
12.11. Linearized Equations for Inviscid Compressible Barotropic Fluids 1095
where u is the fluid velocity, ρ0 is the unperturbed density, p is the pressure, and c is the
speed of sound (which is a constant).
System (1)–(2) can be reduced to the vector equation
utt − c2 ∇ div u = 0
for the velocity and the scalar wave equation
ptt − c2 ∆p = 0
for the pressure.
12.11.2 Decompositions of Equations for Inviscid Compressible

Barotropic Fluid
Each solution of Eqs. (1)–(2) can be represented in the form
u = ∇ϕ + v, p = −ϕt ,
(3)
v = (v1 , v2 , v3 ), v1 = ψy(1) , v2 = −ψx(1) + ψz(2) , v3 = −ρ0 ψy(2) ,
where ψ (1) = ψ (1) (x) and ψ (2) = ψ (2) (x) are arbitrary functions depending only on the
spatial variables and the function ϕ = ϕ(x, t) is a solution of the wave equation
ϕtt − c2 ∆ϕ = 0. (4)

Each solution of Eqs. (1)–(2) can be represented by the formulas
u = ∇ϕ + curl Ψ, p = −ρ0 ϕt , (5)
where Ψ = Ψ(x) is an arbitrary vector function depending only on the spatial variables
and the function ϕ is a solution of the scalar equation (4).
Remark 12.13. See also Section 12.12 with ν = κ = 0.
12.11.3 Vector Form of Equations with Mass Forces

The density and pressure distribution in an immovable compressible barotropic fluid with
the equation of state p = p(ρ) in the presence of a stationary potential mass force f = f(x) =
(f1 , f2 , f3 ) (e.g., the force of gravity) is described by the equations
u = 0, ∇p0 = ρ0 f, p0 = p0 (ρ0 ). (6)
The linearization of the full equations of an inviscid compressible barotropic fluid in a

neighborhood of the steady-state solution (6) on the basis of the relations
u = εv, p = p0 + εp̄, ρ = ρ0 + ερ̄,

where ε is a small parameter, leads to the equations
vt + ∇q = 0, q = p̄/ρ0 , (7)
qt + c2 div v + f · v = 0. (8)
p
Here c = p′ρ |ρ=ρ0 is the local speed of sound, which depends on the spatial variables in
this case.
By eliminating the function q from (7)–(8), we arrive at the equation
vtt − ∇(c2 div v + f · v) = 0 (9)
for the fluid velocity.

The solution of system (7)–(8) has the form
v = ∇ϕ + w(x), q = q0 − ϕt , (10)
where q0 is an arbitrary constant, the function ϕ = ϕ(x, t) satisfies the wave type equation
ϕtt = c2 ∆ϕ + f · ∇ϕ, (11)
and the stationary vector function w = w(x) is a solution of the linear first-order PDE
c2 div w + f · w = 0. (12)
The underdetermined equation (12) can readily be integrated. (Two arbitrary compo-
nents of the vector v are specified arbitrarily, and the remaining component satisfies a linear
first-order ODE in which the first two spatial coordinates occur as parameters.) For exam-
ple, the general solution of Eq. (12) can be represented in the form
Z
w = (w1 , w2 , w3 ), w1 = E A − F E −1 dx , w2 = w2 (x), w3 = w3 (x),
Z
F = (w2 )y + (w3 )z + c (f2 w2 + f3 w3 ), E = exp − f1 c−2 dx ,
−2
where w1 (x), w2 (x), and A = A(y, z) are arbitrary functions and x = (x, y, z).
⊙ Literature for Section 12.11: H. Lamb (1945), L. V. Ovsyannikov (1981).
12.12 Stokes Equations for Viscous Compressible

Barotropic Fluids
12.12.1 Linearized Equations of Viscous Compressible Barotropic
Fluids
The equations describing slow motions of a viscous compressible barotropic fluid have the
form
ut = −∇p + ν∆u + (ν + κ)∇ div u + f, (1)

2
pt + c div u = 0, (2)
12.12. Stokes Equations for Viscous Compressible Barotropic Fluids 1097
where u = (u1 , u2 , u3 ) is the fluid velocity, t is time, ν = µ/ρ0 , κ = λ/ρ0 , µ and λ are
the dynamic viscosities, ρ0 is the unperturbed density, p is the ratio of the pressure to the
unperturbed density (i.e., it is introduced in a different way from that in Section 12.11), c
is the speed of sound, and f = (f1 , f2 , f3 ) is the mass force.
Equations (1)–(2) describe the evolution of small perturbations of velocity and pres-
sure near the stationary solution by linearizing the full nonlinear equations of motion of a
viscous compressible barotropic fluid on the basis of the order-of-magnitude relations
|u| ∼ ε, |p − p0 | ∼ ε, |ρ − ρ0 | ∼ ε, |f| ∼ ε,
where ε is a small parameter and p = p(ρ).
In the limit case of c2 → ∞, Eqs. (1)–(2) become the Stokes equations for the viscous
incompressible fluid (see Eqs. (1)–(2) in Section 12.7).
12.12.2 Decompositions of Equations of Viscous Compressible

Barotropic Fluid with f = 0
u = ∇ϕ + v, p = −ϕt + (2ν + κ)∆ϕ,
(3)
where ψ (1) = ψ (1) (x, t) and ψ (2) = ψ (2) (x, t) are some solutions of the heat equation
ψt − ν∆ψ = 0 (4)
and the function ϕ = ϕ(x, t) is a solution of the third-order equation
ϕtt − (2ν + κ)∆ϕt − c2 ∆ϕ = 0. (5)

Each solution of Eqs. (1)–(2) for f = 0 can be represented by the formulas
u = ∇ϕ + curl Ψ, p = −ϕt + (2ν + κ)∆ϕ, (6)
where Ψ = (Ψ1 , Ψ2 , Ψ3 ) is a solution of the vector heat equation
Ψt − ν∆Ψ = 0 (7)
◮ Special types of decompositions using representations of the vector u via two scalar
functions.
Other representations of solutions for equations (1)–(2) of a viscous compressible barotro-
pic fluid with f = 0 can be obtained with the use of formulas of the form (6)
u = ∇ϕ + curl Ψn , p = −ϕt + (2ν + κ)∆ϕ (8)
and Table 12.1, where the functions ψ (1) = ψ (1) (x, t) and ψ (2) = ψ (2) (x, t) are arbitrary
solutions of the heat equation (4) and the function ϕ satisfies Eq. (5).
12.12.3 Decompositions of Equations of a Viscous Compressible

Barotropic Fluid with f 6= 0
Let the mass force in Eqs. (1)–(2) be represented as the sum f = ∇γ + curl ω (see Sec-
tion 12.5.3). Then the solution of system (1)–(2) can be represented by the formulas
u = ∇ϕ + curl Ψ, p = −ϕt + (2ν + κ)∆ϕ + γ, (9)
where the vector function Ψ satisfies the equation
Ψt − ν∆Ψ = ω
and the function ϕ is a solution of the equation
ϕtt − (2ν + κ)∆ϕt − c2 ∆ϕ = γt .
◮ Decomposition not requiring the force to split into components.

The solution of system (1)–(2) can be represented in the form

u = [∂t2 − (2ν + κ)∂t ∆ − c2 ∆][w] + ∇ ϕ + [c2 + (ν + κ)∂t ][div w] ,
(10)
p = −[∂t − (2ν + κ)∆][ϕ] − c2 (∂t − ν∆)[div w],
where the vector and scalar functions w and ϕ satisfy the independent equations
(∂t − ν∆)[∂t2 − (2ν + κ)∂t ∆ − c2 ∆][w] = f, (11)
[∂t2 − (2ν + κ)∂t ∆ − c2 ∆][ϕ] = 0. (12)
The representation of the solution in the form (10)–(12) corresponds to a decomposition of
the third order.
Remark 12.14. The general solution of the homogeneous equation (11) with f = 0 can be
represented as the sum
w = w1 + w2 ,
(∂t − ν∆)[w1 ] = 0, [∂t2 − (2ν + κ)∂t ∆ − c2 ∆][w2 ] = 0.
12.12.4 Reduction to One Vector Equation and Its Decompositions

◮ Reduction to one vector equation.
Let us reduce system (1)–(2) of coupled equations to a single vector equation. To this end,
we set
u = wt , p = −c2 div w, (13)
where w is the new unknown vector function. By substituting (13) into (1), we obtain the
equation
wtt − ν∆wt − ∇[c2 div w + (ν + κ) div wt ] = f. (14)
Equation (2) is satisfied by the expressions (13) identically.
12.12. Stokes Equations for Viscous Compressible Barotropic Fluids 1099
◮ Decomposition based on the Stokes–Helmholtz representation of the mass force.

Let the mass force in Eqs. (14) be represented as the sum f = ∇γ + curl ω. Then the
solution of system (14) admits the representation
w = ∇ϕ e
e + curl Ψ,
e satisfies the equation
where the vector function Ψ
e t − ν∆Ψ
Ψ e =ω
and the function ϕ
e is a solution of the equation
ϕ et − c2 ∆ϕ
ett − (2ν + κ)∆ϕ e = γ.

The solution of system (14) can be represented in the form

w = ∂t2 − (2ν + κ)∂t ∆ − c2 ∆ [e v] + ∇ c2 + (ν + κ)∂t ][div e
v], (15)
where the vector function e
v satisfies the equation

∂t2 − ν∂t ∆ ∂t2 − (2ν + κ)∂t ∆ − c2 ∆ [e
v] = f. (16)
One can reduce the order of this equation by integrating with respect to t.
Let us differentiate (15) with respect to t and take into account the first relation in (13).
As a result, we obtain the following representation of the velocity:

u = ∂t2 − (2ν + κ)∂t ∆ − c2 ∆ [ζ] + ∇ c2 + (ν + κ)∂t ][div ζ], ζ = e vt ,
where the vector function ζ satisfies the equation

∂t − ν∂∆ ∂t2 − (2ν + κ)∂t ∆ − c2 ∆ [ζ] = f.
12.12.5 Independent Equations for u and p

Let us differentiate Eq. (1) with respect to t and then eliminate pt with the use of Eq. (2).
As a result, we obtain the independent equation

utt − ν∆ut = ∇ c2 div u + (ν + κ) div ut + ft (17)
for the fluid velocity.
Let us apply the operator div to Eq. (1) and then eliminate div u with the use of Eq. (2).
As a result, we obtain the independent equation
ptt − (2ν + κ)∆pt − c2 ∆p = −c2 div f (18)
for the pressure.
The solutions of Eqs. (17) and (18) do not give a solution of the original system (1)–(2)
in general, because (17) and (18) were obtained by an application of various differential
operators to the original model, which makes the class of solutions broader. One can dis-
pense with the spurious solutions by using the original equations (1)–(2), which can be
interpreted as differential constraints for Eqs. (17) and (18).
⊙ Literature for Section 12.12: P. B. Mucha and W. M. Zajaczkowski (2002), N. A. Gusev (2011), I. I. Lipa-
tov and A. D. Polyanin (2013), A. D. Polyanin and S. A. Lychev (2014a,b).
12.13 Oseen Equations for Viscous Compressible

Barotropic Fluids
12.13.1 Vector Form of Equations. Some Remarks
The Oseen equations for viscous compressible barotropic fluids have the form
ut + (a · ∇)u = −∇p + ν∆u + (ν + κ)∇ div u + f, (1)

pt + c2 div u = 0, (2)
where a = (a1 , a2 , a3 ) and the remaining notation is as in Section 12.12.

Equations (1)–(2) describe the evolution of small perturbations of velocity and pressure
near a stationary solution with the use of the linearization of the full nonlinear equations
of motion of viscous compressible barotropic fluids on the basis of the order-of-magnitude
relations
|u − a| ∼ ε, |p − p0 | ∼ ε, |ρ − ρ0 | ∼ ε, |f| ∼ ε,
where ε is a small parameter and p = p(ρ).
12.13.2 Decompositions of Equations with f = 0

u = ∇ϕ + v, p = −ϕt − (a · ∇)ϕ + (2ν + κ)∆ϕ,

(3)
where ψ (1) = ψ (1) (x, t) and ψ (2) = ψ (2) (x, t) are some solutions of the heat-type equation
ψt + (a · ∇)ψ − ν∆ψ = 0 (4)
and the function ϕ = ϕ(x, t) is a solution of the third-order equation
ϕtt + (a · ∇)ϕt − (2ν + κ)∆ϕt − c2 ∆ϕ = 0. (5)

Each solution of Eqs. (1)–(2) with f = 0 can be represented by the formulas
u = ∇ϕ + curl Ψ, p = −ϕt − (a · ∇)ϕ + (2ν + κ)∆ϕ, (6)
where Ψ = (Ψ1 , Ψ2 , Ψ3 ) is a solution of the vector heat-type equation
Ψt + (a · ∇)Ψ − ν∆Ψ = 0, (7)

12.14. Equations of Thermoelasticity 1101
12.13.3 Decomposition of Equations with f 6= 0

Let the mass force in Eqs. (1)–(2) be represented as the sum f = ∇γ + curl ω (see Sec-
tion 12.5.3). Then the solution of system (1)–(2) can be represented by the formulas
u = ∇ϕ + curl Ψ, p = −ϕt − (a · ∇)ϕ + (2ν + κ)∆ϕ + γ, (8)
Ψt + (a · ∇)Ψ − ν∆Ψ = ω,
ϕtt + (a · ∇)ϕt − (2ν + κ)∆ϕt − c2 ∆ϕ = γt .
⊙ Literature for Section 12.13: I. I. Lipatov and A. D. Polyanin (2013), A. D. Polyanin and S. A. Lychev
(2014a,b).
12.14 Equations of Thermoelasticity

12.14.1 Vector Form of Thermoelasticity Equations
Coupled thermoelasticity equations have the form
ρutt = µ∆u + (λ + µ)∇ div u − α∇T + ρf, (1)
Tt = a∆T − β(div u)t , (2)
where u is the displacement vector, t is time, ρ is the material density, µ is the shear
modulus (or the modulus of elasticity of the second kind), λ is the Lamé coefficient, T is
temperature, α and β are the thermomechanical moduli, a is the thermal diffusivity, and f
is the mass force.
We also use the following modified representation of the vector equation (1):
utt = c22 ∆u + (c21 − c22 )∇ div u − (α/ρ)∇T + f, (3)
p p
where c1 = (2µ + λ)/ρ and c2 = µ/ρ are the longitudinal and transverse wave ve-
locities.
12.14.2 Decompositions of Thermoelasticity Equations with f = 0

Each solution of the thermoelasticity equations (2)–(3) with f = 0 can be represented in the
form
ρ 2
u = ∇ϕ + v, T = c1 ∆ϕ − ϕtt ,
α (4)
v = (v1 , v2 , v3 ), v1 = ψy , v2 = −ψx(1) + ψz(2) , v3 = −ψy(2) ,
(1)
where ψ (1) = ψ (1) (x, t) and ψ (2) = ψ (2) (x, t) are some solutions of the wave equations
ψtt − c22 ∆ψ = 0 (5)
and the function ϕ = ϕ(x, t) is a solution of the fourth-order equation
ϕttt − [c21 + (αβ)/ρ]∆ϕt − a∆ϕtt + ac21 ∆∆ϕ = 0. (6)

form
ρ 2
u = ∇ϕ + curl Ψ, T = c1 ∆ϕ − ϕtt , (7)
α
where Ψ = (Ψ1 , Ψ2 , Ψ3 ) is a solution of the vector wave equation
Ψtt − c22 ∆Ψ = 0 (8)
and the function ϕ is a solution of the fourth-order equation (6).
◮ Toroidal–poloidal decomposition.
form
ρ 2
u = ∇ϕ + curl(xξ) + curl curl(xη), T = c1 ∆ϕ − ϕtt , x = (x, y, z),
α
where ξ and η are scalar functions satisfying the wave equations
ξtt − c22 ∆ξ = 0, ηtt − c22 ∆η = 0 (9)

The solutions of Eqs. (2)–(3) with f = 0 can also be represented in the following two forms:
u = ∇ϕ + curl(aξ + bη), a · b 6= 0,
u = ∇ϕ + curl(aξ + xη), |a| =
6 0,
where the functions ξ and η are solutions of the wave equations (9) and the function ϕ is a
solution of the fourth-order equation (6).
12.14.3 Decompositions of Thermoelasticity Equations with f 6= 0

Let the mass force be represented as the sum f = ∇γ + curl ω (see Section 12.5.3). Then
the solution of the thermoelasticity equations (2)–(3) admits the representation
ρ 2
u = ∇ϕ + curl Ψ, T = c1 ∆ϕ − ϕtt + γ ,
α
where the functions Ψ and ϕ satisfy the independent equations
Ψtt − c22 ∆Ψ = ω,
ϕttt − [c21 + (αβ)/ρ]∆ϕt − a∆ϕtt + ac21 ∆∆ϕ = γt − a∆γ.
12.15. Nondissipative Thermoelasticity Equations (the Green–Naghdi Model) 1103
The solution of system (2)–(3) can be represented by the formulas
u = [(∂t − a∆)(∂t2 − c21 ∆) − (αβ/ρ)∂t ∆][w]

+ ∇ ϕ + [(αβ/ρ)∂t + (c21 − c22 )(∂t − a∆)][div w] ,
ρ
T = − (∂t2 − c21 ∆)[ϕ] − β∂t (∂t2 − c22 ∆)[div w],
α
where the vector function w and the scalar function ϕ satisfy the equations
(∂t2 − c22 ∆)[(∂t − a∆)(∂t2 − c21 ∆) − (αβ/ρ)∂t ∆][w] = f, (10)

[(∂t − a∆)(∂t2 − c21 ∆) − (αβ/ρ)∂t ∆][ϕ] = 0. (11)
If ut 6≡ 0, then the general solution of equation (10) with f = 0 can be represented as

the sum
w = w1 + w2 ,
where w1 and w2 are arbitrary solutions of the simpler two equations
(∂t2 − c22 ∆)[w1 ] = 0,

[(∂t − a∆)(∂t2 − c21 ∆) − (αβ/ρ)∂t ∆][w2 ] = 0.
⊙ Literature for Section 12.14: A. D. Polyanin and S. A. Lychev (2014a,b).
12.15 Nondissipative Thermoelasticity Equations

(the Green–Naghdi Model)
12.15.1 Vector Form of the Nondissipative Thermoelasticity
Equations
The coupled nondissipative thermoelasticity equations of the Green–Naghdi model have
the form
ρutt = µ∆u + (λ + µ)∇ div u − α∇T + ρf, (1)

Ttt = a∆T − β(div u)tt . (2)
Equations (1)–(2) differ from the ordinary thermoelasticity equations (see Eqs. (1)–(2) in
Section 12.14) in the second equation, which for the Green–Naghdi model contains second
time derivatives.
Below we use a modified form of the vector equation (1),
utt = c22 ∆u + (c21 − c22 )∇ div u − (α/ρ)∇T + f. (3)

12.15.2 Decompositions of the Nondissipative Thermoelasticity

Equations with f = 0
Each solution of the nondissipative thermoelasticity equations (2)–(3) with f = 0 can be
ρ 2
u = ∇ϕ + v, T = c1 ∆ϕ − ϕtt ,
α (4)
where ψ (1) = ψ (1) (x, t) and ψ (2) = ψ (2) (x, t) are some solutions of the wave equation
ψtt − c22 ∆ψ = 0, (5)
and the function ϕ = ϕ(x, t) is a solution of the fourth-order equation
ϕtttt − [a + c21 + (αβ)/ρ]∆ϕtt + ac21 ∆∆ϕ = 0. (6)
Equation (6) can be represented in the form
(∂t2 − k1 ∆)(∂t2 − k2 ∆)[ϕ] = 0, (7)
where k1 and k2 are the roots of the quadratic equation
k2 − [a + c21 + (αβ)/ρ]k + ac21 = 0. (8)
For ϕt 6≡ 0, the general solution of the fourth-order equation (7) can be represented as the
sum
ϕ = ϕ1 + ϕ2 ,
where ϕ1 and ϕ2 are arbitrary solutions of the wave equations
(∂t2 − k1 ∆)ϕ1 = 0, (∂t2 − k2 ∆)ϕ2 = 0.

ρ 2
u = ∇ϕ + curl Ψ, T = c1 ∆ϕ − ϕtt , (9)
α
where Ψ = (Ψ1 , Ψ2 , Ψ3 ) is a solution of the wave vector equation
Ψtt − c22 ∆Ψ = 0, (10)

12.15. Nondissipative Thermoelasticity Equations (the Green–Naghdi Model) 1105
◮ Toroidal–poloidal decomposition.
ρ 2
u = ∇ϕ + curl(xξ) + curl curl(xη), T = c1 ∆ϕ − ϕtt , x = (x, y, z),
α
where ξ and η are scalar functions satisfying the wave equations
ξtt − c22 ∆ξ = 0, ηtt − c22 ∆η = 0 (11)

The solutions of Eqs. (2)–(3) with f = 0 can also be represented in the following two forms:
u = ∇ϕ + curl(aξ + bη), a · b 6= 0,
u = ∇ϕ + curl(aξ + xη), |a| =
6 0,
where the functions ξ and η are solutions of the wave equations (5) and the function ϕ is a
solution of the fourth-order equation (7).
12.15.3 Decompositions of Thermoelasticity Equations with f 6= 0

Let the mass force be represented as the sum f = ∇γ +curl ω (see Section 12.5.3). Then the
solution of the nondissipative thermoelasticity equations (2)–(3) admits the representation
ρ 2
u = ∇ϕ + curl Ψ, T = c1 ∆ϕ − ϕtt + γ ,
α
where the functions Ψ and ϕ satisfy the independent equations
Ψtt − c22 ∆Ψ = ω,
ϕtttt − [a + c21 + (αβ)/ρ]∆ϕtt + ac21 ∆∆ϕ = γtt − a∆γ.

The solution of system (2)–(3) can be represented by the formulas
u = [(∂t2 − a∆)(∂t2 − c21 ∆) − (αβ/ρ)∂t2 ∆][w]

+ ∇ ϕ + [(αβ/ρ)∂t2 + (c21 − c22 )(∂t2 − a∆)][div w] ,
ρ
T = − (∂t2 − c21 ∆)[ϕ] − β∂t2 (∂t2 − c22 ∆)[div w],
α
where the vector function w and the scalar function ϕ satisfy the equations
(∂t2 − c22 ∆)[(∂t2 − a∆)(∂t2 − c21 ∆) − (αβ/ρ)∂t2 ∆][w] = f,

(12)
[(∂t − a∆)(∂t2 − c21 ∆) − (αβ/ρ)∂t2 ∆][ϕ] = 0.
If ut 6≡ 0, then the general solution of the first equation (12) with f = 0 can be represented
as the sum
w = w1 + w2 + w3 ,
where w1 , w2 , and w3 are arbitrary solutions of the three wave equations
(∂t2 − k1 ∆)[w1 ] = 0, (∂t2 − k2 ∆)[w2 ] = 0, (∂t2 − c22 ∆)[w3 ] = 0.
Here k1 and k2 are the roots of the quadratic equation (8).

⊙ Literature for Section 12.15: A. E. Green and P. M. Naghdi (1993), A. D. Polyanin and S. A. Lychev
(2014a,b), S. A. Lychev and A. D. Polyanin (2015).
12.16 Viscoelasticity Equations

12.16.1 Vector Form of Viscoelasticity Equations
Coupled viscoelasticity equations have the form
ρutt − µ0 ∆u − µ1 ∆ut − ∇[(λ0 + µ0 ) div u + (λ1 + µ1 ) div ut ] = ρf, (1)
where λ0 and µ0 are the elastic Lamé moduli, λ1 and µ1 are the viscosity coefficients,
and the remaining notation is the same as in the elasticity equations (see Eqs. (1) in Sec-
tion 12.6).
In what follows, we use a modified form of the vector equation (1),
utt − c22 ∆u − ν∆ut − ∇[(c21 − c22 ) div u + (ν + κ) div ut ] = f, (2)

p p
where c1 = (2µ0 + λ0 )/ρ, c2 = µ0 /ρ, ν = µ1 /ρ, and κ = λ1 /ρ.
12.16.2 Decompositions of Viscoelasticity Equations with f = 0

Any solution of the homogeneous viscoelasticity equations (2) with f = 0 can also be
represented as
u1 = ϕx + ψy(1) , u2 = ϕy − ψx(1) + ψz(2) , u3 = ϕz − ψy(2) , (3)
where the functions ϕ = ϕ(x, y, z), ψ (1) = ψ (1) (x, y, z), and ψ (2) = ψ (2) (x, y, z) are
solutions of the third-order equations
ϕtt − c21 ∆ϕ − (2ν + κ)∆ϕt = 0, (4)

(1) (1) (2) (2)
ψtt − c22 ∆ψ (1) − ν∆ψt = 0, ψtt − c22 ∆ψ (2) − ν∆ψt = 0. (5)
12.16. Viscoelasticity Equations 1107

Each solution of the viscoelasticity equations (2) with f = 0 can be represented by the
formulas
u = ∇ϕ + curl Ψ, (6)
where Ψ = (Ψ1 , Ψ2 , Ψ3 ) is a solution of the third-order vector equation
Ψtt − c22 ∆Ψ − ν∆Ψt = 0 (7)
◮ Solution of the Cauchy–Kovalevskaya type.

Each solution of the viscoelasticity equations (2) with f = 0 can be represented in the form
u = [∂t2 − c21 ∆ − (2ν + κ)∂t ∆][w] + ∇[c21 − c22 + (ν + κ)∂t ][div w], (8)
(∂t2 − c22 ∆ − ν∂t ∆)[∂t2 − c21 ∆ − (2ν + κ)∂t ∆][w] = 0. (9)
If ut 6≡ 0, then the general solution of equation (9) can be represented as
w = w1 + w2 ,
(∂t2 − c22 ∆ − ν∂t ∆)[w1 ] = 0, [∂t2 − c21 ∆ − (2ν + κ)∂t ∆][w2 ] = 0.
◮ Chadwick–Trowbridge solution (toroidal–poloidal decomposition).

Each solution of the viscoelasticity equations (2) with f = 0 can be represented in the form
u = ∇ϕ + curl(xξ) + curl curl(xη), x = (x, y, z),
where ϕ, ξ, and η are scalar functions satisfying the equations
ϕtt − c21 ∆ϕ − (2ν + κ)∆ϕt = 0, (10)

ξtt − c22 ∆ξ − ν∆ξt = 0, ηtt − c22 ∆η − ν∆ηt = 0. (11)

The solutions of Eqs. (2) with f = 0 can also be represented in the following two forms:
u = ∇ϕ + curl(aξ + bη), a · b 6= 0,
u = ∇ϕ + curl(aξ + xη), |a| =
6 0,
where ϕ, ξ, and η are scalar functions satisfying Eqs. (10)–(11).

12.16.3 Various Forms of Decompositions for Viscoelasticity

Equations with f 6= 0
◮ Lamé type decomposition of the elasticity equations.
Assume that the mass force in the viscoelasticity equations (2) is represented as the sum of
potential and solenoidal components (it is the Stokes–Helmholtz representation of the mass
force; see Section 12.5.3),
f = ∇γ + curl ω.
In this case, the displacement can be represented by formula (6), where the scalar function
ϕ and the vector function Ψ satisfy the nonhomogeneous third-order equations
ϕtt − c21 ∆ϕ − (2ν + κ)∆ϕt = γ,

Ψtt − c22 ∆Ψ − ν∆Ψt = ω.
◮ Decomposition of Cauchy–Kovalevskaya type not requiring the force to split into

components.
Each solution of the viscoelasticity equations (2) can be represented in the form (8), where
the vector function w satisfies the equation
(∂t2 − c22 ∆ − ν∂t ∆)[∂t2 − c21 ∆ − (2ν + κ)∂t ∆][w] = f.
⊙ Literature for Section 12.16: M. E. Gurtin and E. Sternberg (1962), A. D. Polyanin and S. A. Lychev
(2014a,b).
12.17 Maxwell Equations (Electromagnetic Field

Equations)
12.17.1 Maxwell Equations in a Medium and Constitutive Relations
The Maxwell equations in a medium are written as
curl E = −Bt ,
curl H = j + Dt ,
(1)
div D = ρ,
div B = 0,
where E is the electric field intensity, H is the magnetic field intensity, D is the electric
induction, B is the magnetic induction, t is time, j is the current density, and ρ is the charge
density.
The first and fourth Maxwell equations are satisfied identically if one sets
B = curl Ψ, E = −∇Φ − Ψt ,
where the scalar potential Φ and vector potential Ψ are chosen arbitrarily.
12.17. Maxwell Equations (Electromagnetic Field Equations) 1109
The second and third equations in (1) imply the continuity equation
ρt + div j = 0,
which expresses the charge conservation law.

According to experimental data, the vector fields E, H, D, B, and j are not indepen-
dent and should be supplemented with the material equations (constitutive relations) of the
medium. For isotropic linear media, these equations are
D = εE, B = µH, j = λE, (2)
where ε and µ are the dielectric permittivity and the magnetic permeability of the medium
and λ is the conductance of the medium. For a perfect dielectric, i.e., for a medium that
does not conduct current at all, one should set λ = 0.
12.17.2 Some Transformations and Solutions of the Maxwell

Equations
1◦ . By substituting (2) into the Maxwell equations (1), we obtain
curl E = −µHt ,
curl H = λE + εEt ,
(3)
div E = ρ/ε,
div H = 0.
Here the vectors E and H are the unknowns, and the electric charge density is assumed to
be a given function, ρ = ρ(x, t). System (3) is overdetermined, because it consists of 8
scalar equations for the 6 unknown components of the vector functions E and H.
Remark 12.15. It follows from the second and third equations in (3) that the charge density
distribution cannot be arbitrary and must have the special form
ρ = f1 (x) exp(−λt/ε), (4)
where f1 (x) is an arbitrary function.

The general solution of the Maxwell equations (3) with ρ = 0 is
H = − curl curl Ψ, E = µ curl Ψt ,
µ(εΨtt + λΨt ) − ∆Ψ = 0.
2◦ . By successively eliminating the vectors H and E from system (3), we obtain two inde-
pendent overdetermined subsystems:
For the magnetic field intensity,
µ(εHtt + λHt ) − ∆H = 0,
(5)
div H = 0.
For the electric field intensity,

µ(εEtt + λEt ) − ∆E + ∇ div E = 0,
(6)
div E = ρ/ε.
The general solution of Eqs. (5) for the magnetic field intensity can be expressed via
two scalar functions,
H = (H1 , H2 , H3 ), H1 = ψy(1) , H2 = −ψx(1) + ψz(2) , H3 = −ψy(2) , (7)
where ψ (1) and ψ (2) are arbitrary solutions of the equation
L[ψ] = 0, L = µ(ε∂t2 + λ∂t ) − ∆. (8)
The general solution of Eqs. (5) can also be expressed via the vector ψ (i.e., via three
scalar functions)
H = curl ψ, L[ψ] = 0, (9)
where the operator L is defined in (8).
3◦ . For the overdetermined system (6) for the electric field intensity to be solvable, the
consistency conditions (4) should be satisfied. If this is the case, then the solution can be
E = ∇ϕ + curl ψ, ϕ = ϕ0 (x) + ϕ1 (x) exp(−λt/ε), (10)
where the functions ϕ0 (x) and ϕ1 (x) satisfied the Laplace and Poisson equations
∆ϕ0 = 0, ∆ϕ1 = f1 (x)/ε,
the vector ψ satisfies the equation L[ψ] = 0, and the operator L is defined in (8).
⊙ Literature for Section 12.17: V. I. Smirnov (1974, Vol. 2, pp. 375–377), L. D. Landau and E. M. Lifshitz
(1980), M. Schwartz (1987), D. J. Griffiths (1999), F. Melia (2001), W. Benenson, J. W. Harris, H. Stocker, and
H. Lutz (2002), R. Fitzpatrick (2008), A. D. Polyanin and A. I. Chernoutsan (2011).
12.18 Vector Equations of General Form

12.18.1 Vector Equations Containing Operators div and ∇
◮ Class of vector equations considered.
Consider the vector equations
L[u] + ∇K[div u] = f, (1)
which are systems of three coupled scalar equations for u = (u1 , u2 , u3 ). Here we assume
that L and K are arbitrary constant coefficient linear differential operators.
The vector equation (1) is a generalization of the elasticity and viscoelasticity equa-
tions considered in Sections 12.6 and 12.16 (see also equation (17) for the fluid velocity in
Section 12.12). The more complex system considered below in Section 12.19 also can be
reduced to Eq. (1).
12.18. Vector Equations of General Form 1111
1◦ . The vector equation (1) is invariant under the transformation
u + curl Ψ◦ ,
u=e
where the vector function Ψ◦ is an arbitrary solution of the equation L[Ψ◦ ] = 0.

2◦ . The vector equation (1) is invariant under the transformation
u=e
u + ∇Φ,
where the scalar function Φ is an arbitrary solution of the equation
(L + K1 ∆)[Φ] = 0.
12.18.2 Decompositions of the Homogeneous Vector Equation

1◦ . Each solution of the homogeneous equation (1) with f = 0 can be represented in the
form
u = ∇ϕ + v, v = (v1 , v2 , v3 ),
(2)
v1 = ψy(1) , v2 = −ψx(1) + ψz(2) , v3 = −ψy(2) .
Since div v = 0, we have three independent equations for the unknown functions ϕ, ψ (1) ,
and ψ (2) ,
(L + ∆K)[ϕ] = 0, (3)
L[ψ (1) ] = 0, L[ψ (2) ] = 0. (4)
2◦ . Each solution of the homogeneous equation (1) with f = 0 can be represented in the
form
u = ∇ϕ + curl Ψ, (5)
where the scalar function ϕ and the vector function Ψ satisfy the independent equations
(L + ∆K)[ϕ] = 0, L[Ψ] = 0. (6)
The representation (5) of the solution contains one scalar function more than the repre-
sentation (2).
3◦ . Other representations of solutions of the vector equation (1) with f = 0 have the form
u = ∇ϕ + curl Ψn , n = 1, . . . , 7, (7)
where the scalar function ϕ satisfies Eq. (2) and the vector function Ψn can be expressed
via the two scalar functions ψ (1) and ψ (2) by the formulas indicated in Table 12.1. It is
assumed in rows 4–7 of Table 12.1 that the operator L in system (1) has the special form
L[u] = L1 [u] + L2 [∆u],
where L1 and L2 are linear differential operators in time t. The functions ψ (1) and ψ (2) are
described by independent (and identical) equations (4).
4◦ . The solution of system (1) with f = 0 can also be represented in the following form
(see the proof in Section 19.3.1):

u = L + ∆K [w] − ∇K[div w], (8)

L L + ∆K [w] = 0. (9)
The vector equation (9) consists of three independent scalar equations.

The general solution of the homogeneous equation (9) for ∆K 6= const L can be repre-
sented as the sum
w = w1 + w2 ,

L[w1 ] = 0, (L + ∆K [w2 ] = 0.
12.18.3 Decompositions of the Nonhomogeneous Vector Equation

1◦ . For the vector function f in Eq. (1), we use the representation
f = ∇γ + h, h = (θy(1) , −θx(1) + θz(2) , −θy(2) ),
where the formulas for the functions γ, θ (1) , and θ (2) are given in Section 12.5.3. Then the
solution of Eq. (1) admits the representation (2), where the scalar function ϕ and the stream
functions ψ (1) and ψ (2) satisfy the independent equations
(L + ∆K)[ϕ] = γ, L[ψ (1) ] = θ (1) , L[ψ (21) ] = θ (2) .
2◦ . Assume that the mass force f in the vector equation (1) is represented as the sum
f = ∇γ + curl ω (see Section 12.5.3). Then the solution of vector equation (1) admits
the representation (5), where the scalar function ϕ and the vector function Ψ satisfy the
independent equations
(L + ∆K)[ϕ] = γ, L[Ψ] = ω. (10)
3◦ . The solution of the vector equation (1) can be represented without decomposing the
vector f into components in the form (8), where the vector function w satisfies the equation

L L + ∆K [w] = f. (11)
The vector equation (11) consists of three independent scalar equations.

4◦ . The solution of the vector equation (1) can also be represented in the form

u = L + ∆K [w] + ∇ ϕ − K[div w] − L(Q · w) , (11a)
where Q = (Q1 , Q2 , Q3 ), Qn are arbitrary constants or arbitrary constant coefficient linear

differential operators with respect to x, y, z, and t, the vector function w satisfies Eq. (11),

L + ∆K [ϕ] = Q · f. (11b)
12.18. Vector Equations of General Form 1113
Note two special cases of the operator Ω,
Q = a, Q · w = a · w,
Q = ∇, Q · w = div w,
where a is an arbitrary constant vector. (With a = e1 +e2 +e3 , we have Q·w = w1 +w2 +w3 .)
The representation of the solution of the vector equation (1) in the form (8), (11) is a
special case of the representation (11a)–(11b) with Q = 0 and ϕ = 0.
12.18.4 Vector Equations Containing More General Operators

◮ Class of vector equations considered.
Consider the vector equation
L[u] + ΛK[Ω · u] = f, (12)
where Λ = (Λ1 , Λ2 , Λ3 ); Ω = (Ω1 , Ω2 , Ω3 ); L, K, Λn , and Ωn are constant coefficient
linear differential operators with respect to t, x, y, and z. In the special case Λ = Ω = ∇,
system (12) becomes system (1).
◮ Decompositions of the homogeneous vector equation.

1◦ . We seek the solutions of the vector equation (12) with f = 0 in the form
u = Λϕ + v, v = (v1 , v2 , v3 ),
(13)
v1 = Ω2 ψ (1) , v2 = −Ω1 ψ (1) + Ω3 ψ (2) , v3 = −Ω2 ψ (2) .
Since Ω · v = 0, we obtain two independent identical homogeneous equations for the

functions ψ (1) and ψ (2) ,
L [ψ (1) ] = 0, L [ψ (2) ] = 0, (14)
and an independent equation for the function ϕ,
L[ϕ] + K[Ω · Λϕ] = 0. (15)
Remark 12.16. The functions ψ (1) and ψ (2) in (13) are generalized counterparts of the stream
functions in hydrodynamics; see Remark 12.9 in Section 12.7.
2◦ . See also the next paragraph with f = 0.
◮ Decompositions of the nonhomogeneous vector equation.

The solution of the vector equation (12) can be represented in the form

u = L + KΩ · Λ [w] − ΛK[Ω · w], (16)

L L + KΩ · Λ [w] = f. (17)
The vector equation (17) consists of three independent scalar equations; i.e., we have
obtained complete decomposition of the original system (12) in this case. In the special
case of Λ = Ω = ∇, formulas (16) and Eq. (17) become (8) and (11), respectively.
The general solution of the homogeneous equation (17) with f = 0 and KΩ·Λ 6= const L
can be represented in the form of the sum
w = w1 + w2 ,

L[w1 ] = 0, L + KΩ · Λ [w2 ] = 0.
⊙ Literature for Section 12.19: A. D. Polyanin and A. I. Zhurov (2013), A. D. Polyanin and S. A. Lychev
12.19 General Systems Involving Vector and Scalar

Equations: Part I
12.19.1 Systems Containing Operators div and ∇
◮ Class of systems considered.
Consider the systems consisting of one vector and one scalar equation of the form

L[u] + ∇ σp + K1 [div u] = f, (1)
M1 [p] + M2 [div u] = f4 , (2)
where u = (u1 , u2 , u3 ) and p are the unknown functions, L, K1 , M1 , and M2 are arbitrary
constant coefficient linear differential operators with respect to t, x, y, and z, σ is a constant,
and f = (f1 , f2 , f3 ) and f4 are given functions.
Systems of coupled equations of the form (1)–(2) often occur in continuum mechanics
and physics; see Sections 12.6–12.16 and 12.17.2. In the degenerate case of σ = 0, Eq. (1) is
independent of p and coincides with the vector equation for u considered in Section 12.18.
(In this case, the function p is found from Eq. (2) after u has been determined.) In what
follows, we consider the nondegenerate case of σ 6= 0.
1◦ . The coupled equations (1)–(2) are invariant under the transformation
u + curl Ψ◦ ,
u=e p = pe,
2◦ . The coupled equations (1)–(2) are invariant under the transformation
1
u=e
u + ∇Φ, p = pe −
(L + K1 ∆)[Φ],
σ
where the function Φ is an arbitrary solution of the equation
(M1 L + M1 K1 ∆ − σM2 ∆)[Φ] = 0.
12.19. General Systems Involving Vector and Scalar Equations: Part I 1115
12.19.2 Decompositions of Systems with Homogeneous Vector

Equation
1◦ . Each solution of system (1)–(2) with f = 0 can be represented in the form
1
u = ∇ϕ + v, p = − L [ϕ] + K1 [∆ϕ] ,
σ (3)
where the functions ψ (1) , ψ (2) , and ϕ satisfy the independent equations
L [ψ (1) ] = 0, L [ψ (2) ] = 0, (4)

M1 L [ϕ] + (M1 K1 − σM2 ) [∆ϕ] = −σf4 . (5)
The solution representation (3) contains one unknown function less than the original
system (1)–(2).
Remark 12.17. If M1 = 0, then an arbitrary function p0 (t) can be added on the right-hand side
of the second formula in (3) for p.
2◦ . Each solution of system (1)–(2) with f = 0 can be represented in the form
1
u = ∇ϕ + curl Ψ, p=− L [ϕ] + K1 [∆ϕ] , (6)
σ
where the vector function Ψ = (Ψ1 , Ψ2 , Ψ3 ) satisfies the equation
L[Ψ] = 0 (7)
and the scalar function ϕ is determined from Eq. (5).

3◦ . See also Item 3◦ in Section 12.19.3 with f = 0.
12.19.3 Decompositions of Systems with Nonhomogeneous Vector

Equation
1◦ . For the vector function f in Eq. (1) we use the representation
f = ∇γ + h, h = (θy(1) , −θx(1) + θz(2) , −θy(2) ),
where the formulas for finding the functions γ, θ (1) , and θ (2) are given in Section 12.5.3.
Then the solution of system (1)–(2) admits the representation
1
u = ∇ϕ + v, p=γ − L [ϕ] − K1 [∆ϕ] ,
σ
v = (v1 , v2 , v3 ), v1 = ψy , v2 = −ψx(1) + ψz(2) , v3 = −ψy(2) ,
(1)
where the stream functions ψ (1) and ψ (2) satisfy the independent equations
L[ψ (1) ] = θ (1) , L[ψ (21) ] = θ (2)

and the scalar function ϕ is described by the independent equation

−M1 L [ϕ] + (σM2 − M1 K1 ) [∆ϕ] = σf4 − M1 [γ]. (8)
2◦ . Assume that the mass force f in the vector equation (1) is represented as the sum
f = ∇γ + curl ω (see Section 12.5.3). Then the solution of system (1)–(2) admits the
representation (6), where the vector function Ψ satisfies the equation
L[Ψ] = ω (9)
and the scalar function ϕ is described by the independent equation (8).
3◦ . The solution of system (1)–(2) can also be represented without splitting the vector f
into components as follows (for the proof, see Section 19.3.2):

u = (M1 L + ∆M1 K1 − σ∆M2 )[w] + ∇ ϕ + (σM2 − M1 K1 )[div w] ,
1 (10)
p = − L + ∆K1 [ϕ] − M2 L[div w],
σ
where the vector and scalar functions w and ϕ satisfy the independent equations
L(M1 L + ∆M1 K1 − σ∆M2 )[w] = f, (11)
(M1 L + ∆M1 K1 − σ∆M2 )[ϕ] = −σf4 . (12)
Remark 12.18. The solution of the homogeneous equation (11) can be represented in the form
of the sum w = w1 + w2 , where w1 and w2 are arbitrary solutions of the two simpler equations
L[w1 ] = 0, (M1 L + ∆M1 K1 − σ∆M2 )[w2 ] = 0.
Remark 12.19. For f4 = 0, one can set ϕ = 0 in formulas (10) and Eq. (12) without loss of
generality.
12.19.4 Equations for u and p. Reduction to One Vector Equation

◮ Independent equations for u and p.
1◦ . Let M1 6= 0. We apply the operator M1 to and then eliminate M1 [p] from Eq. (1) with
the use of Eq. (2). As a result, we obtain an equation for u,

M1 L[u] + ∇ (K1 M1 − σM2 )[div u] = M1 [f] − σ∇f4 . (13)
2◦ . By applying the operator div to Eq. (1), after elementary transformations we obtain
(L + ∆K1 )[div u] + σ∆p = div f. (14)
Let us apply the operator (L + ∆K1 ) to Eq. (2) and the operator −M2 to Eq. (14) and then
add the resulting relations. Then we obtain an equation for p,

LM1 + ∆K1 M1 − σ∆M2 [p] = (L + ∆K1 )[f4 ] − M2 [div f]. (15)
The solutions of the independent equations (13) and (15) do not in general give a solu-
tion of the original system (1)–(2), because (13) and (15) were obtained by an application
of various differential operators to the original equations, which results in a broader set of
solutions. The spurious solutions can be discarded with the use of the original equations
(1)–(2), which can be interpreted as differential constraints for Eqs. (13) and (15).
12.19. General Systems Involving Vector and Scalar Equations: Part I 1117
◮ Reduction of the system with f4 = 0 to one vector equation.

Let us reduce system (1)–(2) to one vector equation. To this end, set
u = M1 [ξ], p = −M2 [div ξ], (16)
where ξ is the new unknown vector function. By substituting (16) into (1), we obtain the
equation
LM1 [ξ] + ∇ K1 M1 − σM2 [div ξ] = f. (17)
Equation (2) with f4 = 0 is identically satisfied after the substitution of the expressions (16).
Vector equations of the form (17) are considered in Section 12.18. By using the re-
sults of this section, we arrive at the following representation of the solution of the vector
equation (17):

ξ = LM1 + ∆K1 M1 − σ∆M2 [η] + ∇ (σM2 − K1 M1 [div η] , (18)
where the vector function η satisfies the equation

LM1 LM1 + ∆K1 M1 − σ∆M2 [η] = f, (19)
which consists of three independent scalar equations for the components of the vector η.
We point out that the above-described full decomposition of the original system con-
sisting of the four equations (1)–(2) for f4 = 0 provides a description of the solution via
the three components of the vector η. (The solution was expressed via four functions in
Section 12.19.3, Item 3◦ .)
Let us represent the solution (18) and Eq. (19) in terms of the original unknown u =
M1 [ξ] (see formula (16)). To this end, we apply the operator M1 to (18) and introduce the
new unknown η e = M1 [η]. As a result, we obtain

u = LM1 + ∆K1 M1 − σ∆M2 [e η ] + ∇ (σM2 − K1 M1 [div η e] . (20)
e = M1 [η] can be reduced to the form

Equation (19) in terms of η

L LM1 + ∆K1 M1 − σ∆M2 [e η ] = f. (21)
η should be replaced by w), coincides with the first for-

Formula (20), up to notation (e
mula in (10) with ϕ = 0; Eqs. (21) and (11) coincide in this case. We see that for f4 = 0
one can set ϕ = 0 in formulas (10) without loss of generality.
12.19.5 Systems Containing More General Operators

Consider the system

L[u] + Λ σp + K1 [Ω · u] = 0, (22)
M1 [p] + M2 [Ω · u] = f4 , (23)
where Λ = (Λ1 , Λ2 , Λ3 ); Ω = (Ω1 , Ω2 , Ω3 ); L, K1 , M1 , M2 , Λn , and Ωn are constant

coefficient linear differential operators with respect to t, x, y, and z; and σ is a constant. In
the special case of Λ = Ω = ∇, system (22)–(23) becomes system (1)–(2) with f = 0.
In the degenerate case of σ = 0, Eq. (22) was considered in Section 12.18.4. In what
follows, we assume that σ 6= 0.
The solution of system (22)–(23) can be represented in the form
1
u = Λϕ + v, v = (v1 , v2 , v3 ), p = − L [ϕ] + K1 [Ω · Λϕ] ,
σ (24)
(1) (1) (2) (2)
v1 = Ω2 ψ , v2 = −Ω1 ψ + Ω3 ψ , v3 = −Ω2 ψ ,
where the functions ψ (1) and ψ (2) are arbitrary solutions of the two identical independent
equations
L [ψ (1) ] = 0, L [ψ (2) ] = 0,
and the function ϕ is described by the equation
M1 L[ϕ] + (M1 K1 − σM2 )[Ω · Λϕ] = −σf4 .

⊙ Literature for Section 12.19: A. D. Polyanin and A. I. Zhurov (2013), A. D. Polyanin and S. A. Lychev
12.20 General Systems Involving Vector and Scalar

Equations: Part II
12.20.1 Class of Systems Considered
Consider a system of coupled equations of the form
L[u] + ∇K[u, p] = f, (1)

M[u, p] = f4 , (2)
where u = u(x, t) = (u1 , u2 , u3 ) and p = p(x, t) are the unknown functions, x = (x, y, z),
f = f(x, t) = (f1 , f2 , f3 ) and f4 = f4 (x, t) are given functions, and L, K, and M are constant
coefficient linear differential operators with respect to x, y, z, and t.
Remark 12.20. The coefficients of the operator L may depend on time t, and the coefficients of
the operators K and M may depend on all independent variables x, y, z, and t.
12.20.2 Asymmetric Decomposition

Each solution of system (1)–(2) can be represented in the form
u = ∇ϕ + e2 v2 + e3 v3 , p = p, (3)
where e2 and e3 are the unit vectors of the y- and z-axes of the Cartesian coordinate system,
the two scalar functions v2 = v2 (x, t) and v3 = v3 (x, t) satisfy two independent linear
equations of the same type,
L [v2 ] = f2 − ∂y F, L [v3 ] = f3 − ∂z F, (4)

Z x
F = F (x, t) = F (x, y, z, t) = f1 (x1 , y, z, t) dx1 , (5)
0
12.20. General Systems Involving Vector and Scalar Equations: Part II 1119
and the functions ϕ = ϕ(x, t) and p = p(x, t) are determined from the system of equations
L [ϕ] + K [∇ϕ + e2 v2 + e3 v3 , p] = F, (6)

M [∇ϕ + e2 v2 + e3 v3 , p] = f4 . (7)
To avoid introducing excessive new notation, in (3) and in the following we write p = p
instead of p = pe, where pe is the new unknown function.
System (4)–(7) consists of two independent linear equations (4) and a subsystem of
two coupled equations (6)–(7) and is substantially simpler than the original system of four
coupled linear equations (1)–(2).
12.20.3 Symmetric Decomposition

Each solution of system (1)–(2) can also be represented in the symmetric form
u = ∇ϕ + v, p = p, (8)
equation
L[v] = f − ∇G (9)
and the functions ϕ and p are described by the system of equations
L [ϕ] + K [∇ϕ + v, p] = G, (10)

M [∇ϕ + v, p] = f4 . (11)
Equations (9)–(11) contain the arbitrary scalar function G = G(x, t).

Remark 12.21. The representation of the components of the vector (8) contains one extra (addi-
tional) function as compared with the representation (3). This permits one to simplify Eqs. (9)–(11)
by imposing an additional condition on the components v1 , v2 , v3 of the vector function v and by
choosing an appropriate function G. In particular, without loss in generality, one can set f1 = Gx
and v1 = 0 in (8)–(11), which gives the representation (3)–(7) with F = G.
Remark 12.22. All coupled equations of continuum mechanics and physics, as well as the more
general equations considered in Sections 12.6–12.19, are special cases of system (1)–(2).
⊙ Literature for Section 12.20: A. D. Polyanin and A. I. Zhurov (2013), I. I. Lipatov and A. D. Polyanin
(2013), A. D. Polyanin and S. A. Lychev (2014a,b).
Part II
Analytical Methods
Chapter 13
Methods for First-Order

Linear PDEs
13.1 Linear PDEs with Two Independent Variables

13.1.1 Special First-Order Linear PDEs with Two Independent
Variables
◮ Physical interpretation and the characteristic equation.
Consider a first-order linear homogeneous partial differential equation with two indepen-
dent variables of the special form
∂w ∂w
f (x, y) + g(x, y) = 0. (13.1.1.1)
∂x ∂y
Equation (13.1.1.1) describes a steady-state distribution of the concentration of a sub-

stance in a plane flow (without regard to diffusion). Moreover, it is assumed that the fluid
velocity components along the x- and y-axes are specified by the functions f and g.
The transformation
w = Ψ(u),
where Ψ(u) is an arbitrary function (Ψ 6≡ const), preserves the form of Eq. (13.1.1.1).
The first-order ordinary differential equation
dx dy
= (13.1.1.2)
f (x, y) g(x, y)
is called the characteristic equation corresponding to the partial differential equation

(13.1.1.1). The integral curves of Eq. (13.1.1.2) are called characteristics.
Remark 13.1. Suppose that the variables x and y belong to a domain V . Let the functions
f (x, y) and g(x, y) be continuously differentiable with respect to both x and y in V , and let
f 2 (x, y) + g 2 (x, y) 6= 0 in V . Then there exists a unique characteristic through each point of V .
1123
1124 M ETHODS FOR F IRST-O RDER L INEAR PDE S
◮ Formula for the general solution. General solutions of some first-order PDEs.
Let the general solution of the characteristic equation (13.1.1.2) be given by
Ξ(x, y) = C, (13.1.1.3)
where C is an arbitrary constant. Then the general solution of Eq. (13.1.1.1) has the form
w = Φ(Ξ), (13.1.1.4)
where Φ = Φ(Ξ) is an arbitrary function. The left-hand side Ξ(x, y) of Eq. (13.1.1.3) is
called a first integral of the partial differential equation (13.1.1.1).
Remark 13.2. In formulas (13.1.1.3) and (13.1.1.4), one can take an arbitrary nonconstant par-
ticular solution of Eq. (13.1.1.1) for Ξ.
Example 13.1. Consider the linear constant coefficient equation
∂w ∂w
+a = 0. (13.1.1.5)
∂x ∂y
The characteristic equation for this equation is
dx dy
= .
1 a
The general solution of the characteristic equation is y − ax = C. Thus, the general solution of the
original first-order PDE (13.1.1.5) can be expressed via an arbitrary function Φ of one variable as
w = Φ(y − ax). (13.1.1.6)
It is a traveling wave solution.
Table 13.1 lists general solutions of some linear first-order partial differential equations
in two independent variables of the form (13.1.1.1).
TABLE 13.1
General solutions of some first-order partial differential equations of the form (13.1.1.1); Φ(u) is
an arbitrary function. The subscripts x and y indicate the corresponding partial derivatives
No. Equations General solutions Notation

R R
1 wx + [f (x)y + g(x)]wy = 0 w = Φ e−F y − e−F g(x) dx F = f (x) dx
R R
2 wx + [f (x)y + g(x)y k ]wy = 0 w = Φ e−F y 1−k − (1 − k) e−F g(x) dx F = (1 − k) f (x) dx
R R
3 wx + [f (x)eλy + g(x)]wy = 0 w = Φ e−λy E + λ f (x)E dx E = exp λ g(x) dx
R R dy
4 f (x)wx + g(y)wy = 0 w = Φ fdx (x)
− g(y)
13.1. Linear PDEs with Two Independent Variables 1125
◮ Classical Cauchy problem.

Find a solution w = w(x, y) of Eq. (13.1.1.1) satisfying the condition
w = s(y) at x = x0 , (13.1.1.7)
where s(y) is a given function.

The solution of the Cauchy problem (also called the initial value problem) can be ob-
tained from the general solution (13.1.1.4). Substituting the initial data (13.1.1.7) into
Eq. (13.1.1.4), we have
s(y) = Φ Ξ(x0 , y) .
This relation serves to determine the function Φ.
Example 13.2. Find a solution of the Cauchy problem for Eq. (13.1.1.5) with the initial condi-
tion
w = yk at x = 1. (13.1.1.8)
Substituting the initial data (13.1.1.8) into the general solution (13.1.1.6) yields
y k = Φ(y − a).
This allows us to conclude that Φ(u) = (u + a)k . Substituting this expression into Eq. (13.1.1.6),
we find the solution of the Cauchy problem in the form
w = (y − ax + a)k .
◮ Physical interpretation of the classical Cauchy problem.

Let x and y be spatial coordinates, and let w be the concentration of a substance. It is
assumed that the concentration distribution is described by the steady-state transport equa-
tion (13.1.1.1) and the concentration profile (13.1.1.7) is specified at the input cross-section
x = x0 . The concentration w = w(x, y) is to be determined in the flow after the input cross-
section (for x ≥ x0 ).
Another, nonsteady-state interpretation of the Cauchy problem is possible. Let x be
time, let y be the spatial coordinate, and let w be the concentration (f ≡ 1). It is as-
sumed that the concentration distribution is described by the nonsteady-state transport
equation (13.1.1.1) and that the concentration profile (13.1.1.7) is prescribed at the ini-
tial instant x = x0 . The concentration w = w(x, y) is to be determined for all subsequent
instants of time (x ≥ x0 ).
◮ Generalized Cauchy problem.

Find a solution w = w(x, y) of Eq. (13.1.1.1) with the initial conditions
x = s1 (ξ), y = s2 (ξ), w = s3 (ξ), (13.1.1.9)
where ξ is a parameter (α ≤ ξ ≤ β), s1 = s1 (ξ) and s2 = s2 (ξ) are given functions, and
|s′1 | + |s′2 | =
6 0.
A geometric interpretation of this problem is as follows: find an integral surface of
Eq. (13.1.1.1) passing through the parametrically defined line (13.1.1.9).
The solution of the generalized Cauchy problem can be obtained from the general so-
lution (13.1.1.4) by substituting the initial data (13.1.1.9) into it.
The classical Cauchy problem (13.1.1.1), (13.1.1.7) can be represented as the general-
ized Cauchy problem (13.1.1.1), (13.1.1.9) if one rewrites the initial condition (13.1.1.7) in
the parametric form
x = x0 , y = ξ, w = s(ξ). (13.1.1.10)
Remark 13.3. In the statement of the Cauchy problem (13.1.1.1), (13.1.1.9), the plane curve
x = s1 (ξ), y = s2 (ξ) is assumed not to be tangent to any characteristic at any point; that is,
f (s1 , s2 )s′2 − g(s1 , s2 )s′1 6= 0.
Remark 13.4. If the curve x = s1 (ξ), y = s2 (ξ) is a characteristic, then
(a) The Cauchy problem has no solution if s3 (ξ) 6≡ const.
(b) The Cauchy problem has infinitely many solutions if s3 (ξ) ≡ const.
13.1.2 General First-Order Linear PDE with Two Independent

Variables
◮ The representation of the general solution via particular solutions.
In the general case, a first-order linear nonhomogeneous equation with two independent
variables has the form
∂w ∂w
f (x, y) + g(x, y) = h1 (x, y)w + h0 (x, y). (13.1.2.1)
∂x ∂y
The special case h0 = h1 = 0 is discussed in Section 13.1.1.
The general solution of the linear nonhomogeneous equation (13.1.2.1) can be repre-
sented as the sum of any particular solution of this equation and the general solution of
the corresponding homogeneous equation (with h0 ≡ 0). In what follows, we give a more
detailed statement about the representation of the general solution via particular solutions.
The general solution of Eq. (13.1.2.1) can be represented in the form
w = w2 + w1 Φ(w0 ),
where w0 = w0 (x, y) is any nonconstant particular solution of the “truncated” homoge-
neous equation (13.1.2.1) with h0 = h1 = 0; w1 = w1 (x, y) is a nontrivial particular
solution of the truncated homogeneous equation (13.1.2.1) with h0 = 0; w2 = w2 (x, y) is
a particular solution of the nonhomogeneous equation (13.1.2.1); and Φ(w0 ) is an arbitrary
function.
◮ Solution by using the characteristic system.

Given two distinct (functionally independent) integrals,
u1 (x, y, w) = C1 , u2 (x, y, w) = C2 (13.1.2.2)
of the characteristic system
dx dy dw
= = , (13.1.2.3)
f (x, y) g(x, y) h1 (x, y)w + h0 (x, y)
13.1. Linear PDEs with Two Independent Variables 1127
the general solution of the nonhomogeneous equation (13.1.2.1) is defined by
Φ(u1 , u2 ) = 0, (13.1.2.4)
where Φ is an arbitrary function of two variables. With Eq. (13.1.2.4) solved for u1 or u2 ,
we often specify the general solution in the form
uk = Ψ(u3−k ),
where k = 1, 2 and Ψ(u) is an arbitrary function of one variable.

Remark 13.5. In the general case, the integrals (13.1.2.2) can be represented in the form
Ξ(x, y) = C1 , p1 (x, y)w + p0 (x, y) = C2 ,
where Ξ(x, y) is a first integral of the partial differential equation (13.1.1.1) (or Eq. (13.1.2.1) with
h0 = h1 = 0) and p0 (x, y) and p1 (x, y) are some functions.
Remark 13.6. The degenerate case h0 = h1 = 0 in (13.1.2.3) corresponds to the simplest integral
w = C2 in (13.1.2.2).
Example 13.3. Consider the equation
∂w ∂w
+ ax = bw.
∂x ∂y
The corresponding characteristic system
dx dy dw
= = ,
1 ax bw
has two independent integrals
y − 12 ax2 = C1 , we−bx = C2 .
Therefore, the general solution of the equation in question is expressed in terms of an arbitrary
function of two variables as Φ y − 12 ax2 , we−bx = 0. By solving this equation for the second
argument, we obtain the solution in explicit form,
w = ebx Ψ(y − 21 ax2 ), where Ψ(u) is an arbitrary function.
◮ Solution based on a change of variables.

Let a particular solution u = u(x, y) (first integral) of the corresponding truncated homo-
geneous equation
∂u ∂u
f (x, y) + g(x, y) =0 (u 6≡ const) (13.1.2.5)
∂x ∂y
be known. By switching from x and y to the new variables x and u = u(x, y) in equation
(13.1.2.1), we obtain
∂w
f¯(x, u) = h̄1 (x, u)w + h̄0 (x, u),
∂x
where f¯(x, u) = f (x, y), h̄1 (x, u) = h1 (x, y), and h̄0 (x, u) = h0 (x, y) are the coefficients
of the original equation (13.1.2.1) rewritten in terms of x and u.
Equation (13.1.2.5) can be treated as an ordinary differential equation for w = w(x)

with parameter u. The general solution of Eq. (13.1.2.5) has the form
Z Z
h̄0 (x, u) dx h̄1 (x, u)
w=E + Φ(u) , E = exp dx ,
f¯(x, u) E f¯(x, u)
where Φ is an arbitrary function; in the integration, u is considered a parameter. To find a
general integral of Eq. (13.1.2.1), one should compute the integrals in the last relation and
then return to the original variables x and y.
◮ Classical and generalized Cauchy problems. Solution methods.

1◦ . The classical and generalized Cauchy problems for Eq. (13.1.2.1) are stated in the same
manner as for the “truncated” equation (with h0 = h1 = 0); see Section 13.1.1.
2◦ . The solution of the Cauchy problem can be obtained from the expression for the general
solution, into which the initial data should be substituted.
3◦ . An alternative (and more convenient) approach to the solution of the Cauchy prob-
lem (13.1.1.1), (13.1.1.9) consists of several stages. First, two independent integrals in
(13.1.2.2) of the characteristic system (13.1.2.3) are determined. Then the initial data
(13.1.1.9) are substituted into the integrals (13.1.2.2),

u1 s1 (ξ), s2 (ξ), s3 (ξ) = C1 , u2 s1 (ξ), s2 (ξ), s3 (ξ) = C2 , (13.1.2.6)
to determine the constants of integration C1 and C2 . By eliminating C1 and C2 from

Eqs. (13.1.2.2) and (13.1.2.6), we have

u1 (x, y, w) = u1 s1 (ξ), s2 (ξ), s3 (ξ) ,
(13.1.2.7)
u2 (x, y, w) = u2 s1 (ξ), s2 (ξ), s3 (ξ) .
These relations are a parametric form of the solution of the Cauchy problem (13.1.1.1),
(13.1.1.9). By eliminating the parameter ξ from (13.1.2.7), one can obtain the solution in
explicit form.
Example 13.4. Consider the classical Cauchy problem for the equation
∂w ∂w
+a =b (13.1.2.8)
∂x ∂y
subject to the initial condition
w = ky 2 at x = 0. (13.1.2.9)
The characteristic system
dx dy dw
= = (13.1.2.10)
1 a b
corresponding to Eq. (13.1.2.8) has two independent integrals
y − ax = C1 , w − bx = C2 . (13.1.2.11)
Let us rewrite the initial condition (13.1.2.9) for the classical Cauchy problem in terms of the
generalized Cauchy problem:
x = 0, y = ξ, w = kξ 2 .
13.2. First-Order Linear PDEs with Three or More Independent Variables 1129
By substituting these initial conditions into the integrals (13.1.2.11), we obtain
ξ = C1 , kξ 2 = C2 . (13.1.2.12)
By eliminating C1 and C2 from Eqs. (13.1.2.11) and (13.1.2.12), we have
y − ax = ξ, w − bx = kξ 2 .
On eliminating the parameter ξ from these algebraic equations, we obtain the solution of the Cauchy
problem (13.1.2.8)–(13.1.2.9) in explicit form,
w = bx + k(y − ax)2 .
Example 13.5. Consider the generalized Cauchy problem for Eq. (13.1.2.8) with the initial
conditions
x = ξ, y = ξ, w = ξ2. (13.1.2.13)
The corresponding characteristic system (13.1.2.10) has two independent integrals (13.1.2.11).
Substituting the initial data (13.1.2.13) into the integrals (13.1.2.11) yields
(1 − a)ξ = C1 , ξ 2 − bξ = C2 . (13.1.2.14)
On eliminating the parameters C1 , C2 , and ξ from Eqs. (13.1.2.11) and (13.1.2.14), we obtain the
solution of the generalized Cauchy problem (13.1.2.8), (13.1.2.13),
2
y − ax y − ax
w = bx − b + .
1−a 1−a
This is the solution for a 6= 1. For a = 1, the straight line y = x = ξ (where the initial data are
specified) is a characteristic, and the generalized Cauchy problem in question has no solutions.
Remark 13.7. For the existence and uniqueness theorem for the solution of the Cauchy problem,
see the end of Section 13.2.
13.2 First-Order Linear PDEs with Three or More

Independent Variables
13.2.1 Characteristic System. General Solution
◮ Linear homogeneous equations. Characteristic system. General solution.
Consider a linear homogeneous equation with n independent variables of the form
n
X ∂w
fi (x1 , . . . , xn ) = 0. (13.2.1.1)
∂xi
i=1
If n − 1 functionally independent integrals
u1 (x1 , . . . , xn ) = C1 , u2 (x1 , . . . , xn ) = C2 , ..., un−1 (x1 , . . . , xn ) = Cn−1

(13.2.1.2)
(a basis of integrals) of the characteristic system
dx1 dx2 dxn
= = ··· = (13.2.1.3)
f1 (x1 , . . . , xn ) f2 (x1 , . . . , xn ) fn (x1 , . . . , xn )
are known, then the general solution of Eq. (13.2.1.1) is defined by
w = Φ(u1 , u2 , . . . , un−1 ),
where Φ is an arbitrary function of n − 1 variables.
◮ Reducing the number of independent variables.

Assume that one integral u(x1 , . . . , xn ) = C of system (13.2.1.3) is known. By passing
from x1 , . . . , xn−1 , xn to the new variables x1 , . . . , xn−1 , u, we arrive at a first-order
linear homogeneous partial differential equation with one independent variable fewer than
the original equation,
n−1
X ∂w
fei (x1 , . . . , xn−1 , u) = 0,
∂xi
i=1
where u plays the role of a parameter.
◮ Linear nonhomogeneous equations. Characteristic system. General solution.
Consider a linear nonhomogeneous equation with n independent variables of the general

form
X n
∂w
fi (x1 , . . . , xn ) = g(x1 , . . . , xn )w + h(x1 , . . . , xn ). (13.2.1.4)
∂xi
i=1
Given n functionally independent integrals
u1 (x1 , . . . , xn , w) = C1 , u2 (x1 , . . . , xn , w) = C2 , ..., un (x1 , . . . , xn , w) = Cn

(13.2.1.5)
(a basis of integrals) of the characteristic system
dx1 dxn dw
=···= = , (13.2.1.6)
f1 (x1 , . . . , xn ) fn (x1 , . . . , xn ) g(x1 , . . . , xn )w + h(x1 , . . . , xn )
the general solution of Eq. (13.2.1.4) is defined by
Φ(u1 , u2 , . . . , un ) = 0,
where Φ is an arbitrary function of n variables.
◮ Some formulas for the solution.

1◦ . Let an integral basis uk = uk (x1 , . . . , xn ) (k = 1, . . . , n − 1) of the corresponding
“truncated” homogeneous equation (13.2.1.1) be known. By passing from x1 , x2 , . . . , xn
to the new variables x1 , u1 , . . . , un−1 , we arrive at the linear equation
∂w
fe1 (x, u) = ge(x, u)w + e
h(x, u), x = x1 ,
∂x
13.2. First-Order Linear PDEs with Three or More Independent Variables 1131
which can be treated as a first-order linear ordinary differential equation for w = w(x) with
the parameter vector u = (u1 , u2 , . . . , un−1 ). By solving this equation, we obtain
Z e Z
h(x, u) dx ge(x, u)
w = E Φ(u) + , E = exp dx ,
fe1 (x, u) E fe1 (x, u)
where Φ is an arbitrary function. When computing the two integrals, the components of the
vector u are treated as parameters. To determine the general integral of Eq. (13.2.1.4), one
should return to the original variables x1 , . . . , xn after the integration has been performed.
2◦ . Let g ≡ 0. Given a basis u = (u1 , . . . , un−1 ) of integrals of the corresponding homoge-
neous equation (with h ≡ 0) and a particular solution w0 = w0 (x1 , . . . , xn ) of the original
nonhomogeneous equation, the general solution is defined by
w = w0 + Φ(u),
where Φ is an arbitrary function.
13.2.2 Cauchy Problems

◮ Classical and generalized Cauchy problems (initial value problems).
Consider two statements of the Cauchy problem.
1◦ . Generalized Cauchy problem. Find a solution w = w(x1 , . . . , xn ) of Eq. (13.2.1.4)
x1 = ϕ1 (ξ1 , . . . , ξn−1 ), ..., xn = ϕn (ξ1 , . . . , ξn−1 ),

w = ϕn+1 (ξ1 , . . . , ξn−1 ),
(13.2.2.1)
where the ξk are parameters (k = 1, . . . , n − 1) and the ϕm (ξ1 , . . . , ξn−1 ) are given func-
tions (m = 1, . . . , n + 1).
2◦ . Classical Cauchy problem. Find a solution w = w(x1 , . . . , xn ) of Eq. (13.2.1.4) with
the initial condition
w = ψ(x2 , . . . , xn ) at x1 = 0, (13.2.2.2)
where ψ(x2 , . . . , xn ) is a given function.
It is convenient to represent the classical Cauchy problem as a generalized Cauchy
problem by rewriting the initial condition (13.2.2.2) in the parametric form
x1 = 0, x2 = ξ 1 , ..., xn = ξn−1 , w = ψ(ξ1 , . . . , ξn−1 ).
◮ Solution procedure for the Cauchy problem.

The solution procedure for the Cauchy problem (13.2.1.4), (13.2.2.1) involves several steps.
First of all, one determines independent integrals (13.2.1.5) of the characteristic system
(13.2.1.6). After that, to find the constants of integration C1 , . . . , Cn , the initial data
(13.2.2.1) should be substituted into the integrals (13.2.1.5),
uk (ϕ1 , . . . , ϕn , ϕn+1 ) = Ck , where ϕm = ϕm (ξ1 , . . . , ξn−1 ), k = 1, . . . , n.

(13.2.2.3)
On eliminating C1 , . . . , Cn from Eqs. (13.2.1.5) and (13.2.2.3), one obtains
uk (x1 , . . . , xn , w) = uk (ϕ1 , . . . , ϕn , ϕn+1 ), k = 1, . . . , n, (13.2.2.4)
where ϕm = ϕm (ξ1 , . . . , ξn−1 ), m = 1, . . . , n + 1. Relations (13.2.2.4) are a parametric

form of the solution of the Cauchy problem (13.2.1.4), (13.2.2.1). In some cases, one may
succeed in eliminating the parameters ξ1 , . . . , ξn−1 to obtain the solution in explicit form.
◮ The existence and uniqueness theorem.

Let f1 = 1, and let the other coefficients f2 , . . . , fn , g, and h in Eq. (13.2.1.4), as well
as the function ψ in (13.2.2.2), be continuously differentiable functions of x1 , . . . , xn in a
domain V = {0 < x1 < a, −∞ < xi < ∞, i = 2, . . . , n}. Suppose that the following
inequality holds in V :
q q
f2 (x1 , . . . , xn ) + · · · + fn (x1 , . . . , xn ) ≤ k 1 + x22 + · · · + x2n ,
2 2 k = const.
Then there exists a unique continuously differentiable solution of the Cauchy problem
(13.2.1.4), (13.2.2.2); the existence of this solution is guaranteed in the entire domain V .
⊙ Literature for Chapter 13: E. Kamke (1965), I. G. Petrovsky (1991), H. Rhee, R. Aris, and N. R. Amund-
son (1986), R. Courant and D. Hilbert (1989), A. I. Subbotin (1991), D. Zwillinger (1998), A. D. Polyanin,
V. F. Zaitsev, and A. Moussiaux (2002), A. D. Polyanin and A. V. Manzhirov (2007).
Chapter 14
Second-Order Linear PDEs:

Classification, Problems,
and Particular Solutions
14.1 Classification of Second-Order Linear Partial

Differential Equations
14.1.1 Equations with Two Independent Variables
◮ Examples of equations encountered in applications.
Three basic types of partial differential equations are distinguished—parabolic, hyper-
bolic, and elliptic. The solutions of the equations pertaining to each of the types have
their own characteristic qualitative differences.
The simplest example of a parabolic equation is the heat equation
∂w ∂2w
− = 0, (14.1.1.1)
∂t ∂x2
where the variables t and x play the role of time and the spatial coordinate, respectively.
Note that Eq. (14.1.1.1) contains only one highest derivative term. Frequently encountered
particular solutions of Eq. (14.1.1.1) can be found in Section 3.1.1.
The simplest example of a hyperbolic equation is the wave equation
∂2w ∂2w
− = 0, (14.1.1.2)
∂t2 ∂x2
where the variables t and x play the role of time and the spatial coordinate, respectively.
Note that the highest derivative terms in Eq. (14.1.1.2) differ in sign. Frequently encoun-
tered particular solutions of Eq. (14.1.1.2) can be found in Section 6.1.1.
The simplest example of an elliptic equation is the Laplace equation
∂2w ∂2w
+ = 0, (14.1.1.3)
∂x2 ∂y 2
1133
1134 S ECOND -O RDER L INEAR PDE S : C LASSIFICATION , P ROBLEMS , PARTICULAR S OLUTIONS
where x and y play the role of the spatial coordinates. Note that the highest derivative
terms in Eq. (14.1.1.3) have like signs. Frequently encountered particular solutions of
Eq. (14.1.1.3) can be found in Section 9.1.1.
Any linear partial differential equation of the second-order with two independent vari-
ables can be reduced, by appropriate manipulations, to a simpler equation that has one of
the three highest derivative combinations specified above in examples (14.1.1.1), (14.1.1.2),
and (14.1.1.3).
◮ Types of equations. Characteristic equations.
Consider a second-order partial differential equation with two independent variables of the
general form

∂2w ∂2w ∂2w ∂w ∂w
a(x, y) + 2b(x, y) + c(x, y) = F x, y, w, , , (14.1.1.4)
∂x2 ∂x∂y ∂y 2 ∂x ∂y
where a, b, c are some functions of x and y that have continuous derivatives up to the
second order inclusive.∗
Given a point (x, y), Eq. (14.1.1.4) is said to be
parabolic if b2 − ac = 0,
hyperbolic if b2 − ac > 0,
elliptic if b2 − ac < 0
at this point.
To reduce Eq. (14.1.1.4) to canonical form, one should write out the characteristic equa-
tion
a (dy)2 − 2b dx dy + c (dx)2 = 0,
which splits into two equations

p
a dy − b + b2 − ac dx = 0 (14.1.1.5)
and p
a dy − b − b2 − ac dx = 0, (14.1.1.6)
and then find their general integrals.

Remark 14.1. The characteristic equations (14.1.1.5)–(14.1.1.6) can be used if a 6≡ 0. If a ≡ 0,
the simpler equations
dx = 0,
2b dy − c dx = 0
should be used; the first equation has the obvious general solution x = C.
∗
The right-hand side of Eq. (14.1.1.4) may be nonlinear. The classification and the procedure of reducing
such equations to a canonical form are only determined by the left-hand side of the equation.
14.1. Classification of Second-Order Linear Partial Differential Equations 1135
◮ Canonical form of parabolic equations (case b2 − ac = 0).

In this case, Eqs. (14.1.1.5) and (14.1.1.6) coincide and have a common general integral
ϕ(x, y) = C.
By passing from x, y to new independent variables ξ, η in accordance with the relations
ξ = ϕ(x, y), η = η(x, y),
where η = η(x, y) is any twice differentiable function that satisfies the condition of non-
D(ξ,η)
degeneracy of the Jacobian D(x,y) in the given domain, we reduce Eq. (14.1.1.4) to the
canonical form
∂2w ∂w ∂w
= F1 ξ, η, w, , . (14.1.1.7)
∂η 2 ∂ξ ∂η
For η one can take η = x or η = y.
It is obvious that the transformed equation (14.1.1.7) has only one highest-derivative
term, just as the heat equation (14.1.1.1).
Remark 14.2. In the degenerate case where the function F1 does not depend on the deriva-
tive ∂ξ w, Eq. (14.1.1.7) is an ordinary differential equation in the variable η, and ξ serves as a
parameter.
◮ Canonical forms of hyperbolic equations (case b2 − ac > 0).

The general integrals
ϕ(x, y) = C1 , ψ(x, y) = C2
of Eqs. (14.1.1.5) and (14.1.1.6) are real and distinct. These integrals determine two distinct
families of real characteristics.
1◦ . First canonical form. By passing from x, y to new independent variables ξ, η in accor-
dance with the relations
ξ = ϕ(x, y), η = ψ(x, y),
we reduce Eq. (14.1.1.4) to

∂2w ∂w ∂w
= F2 ξ, η, w, , . (14.1.1.8)
∂ξ∂η ∂ξ ∂η
This is the so-called first canonical form of a hyperbolic equation.

2◦ . Second canonical form. The transformation
ξ = t + z, η =t−z
brings Eq. (14.1.1.8) to another canonical form,

∂2w ∂2w ∂w ∂w
− = F3 t, z, w, , , (14.1.1.9)
∂t2 ∂z 2 ∂t ∂z
where F3 = 4F2 . This is the so-called second canonical form of a hyperbolic equation.
Apart from notation, the left-hand side of the last equation coincides with that of the wave
equation (14.1.1.2).
In some cases, reduction of an equation to canonical form permits one to find its general
solution.
Example 14.1. The equation
∂2w ∂ 2w
kx + =0
∂x2 ∂x∂y
is a special case of Eq. (14.1.1.4) with a = kx, b = 12 , c = 0, and F = 0. The characteristic equations
kx dy − dx = 0,
dy = 0
have the general integrals ky −ln |x| = C1 and y = C2 . Switching to the new independent variables
ξ = ky − ln |x|, η=y
reduces the original equation to the canonical form
∂2w ∂w
=k .
∂ξ∂η ∂ξ
Integrating with respect to ξ yields the linear first-order equation
∂w
= kw + f (η),
∂η
where f (η) is an arbitrary function. Its general solution is expressed as
Z
w = ekη g(ξ) + ekη e−kη f (η) dη,
where g(ξ) is an arbitrary function.

Remark 14.3. For higher-order hyperbolic equations, see Section 11.6.3.
◮ Canonical form of elliptic equations (case b2 − ac < 0).

In this case, the general integrals of Eqs. (14.1.1.5) and (14.1.1.6) are complex conjugate;
these determine two families of complex characteristics.
Let the general integral of Eq. (14.1.1.5) have the form
ϕ(x, y) + iψ(x, y) = C, i2 = −1,
where ϕ(x, y) and ψ(x, y) are real-valued functions.
By passing from x, y to new independent variables ξ, η in accordance with the relations
ξ = ϕ(x, y), η = ψ(x, y),
we reduce Eq. (14.1.1.4) to the canonical form

∂2w ∂2w ∂w ∂w
+ = F4 ξ, η, w, , .
∂ξ 2 ∂η 2 ∂ξ ∂η
Apart from notation, the left-hand side of the last equation coincides with that of the
Laplace equation (14.1.1.3).
Remark 14.4. For higher-order elliptic equations, see Section 11.6.2.
◮ Linear constant coefficient partial differential equations.

1◦ . When reduced to a canonical form, linear homogeneous constant coefficient partial
differential equations
∂2w ∂2w ∂2w ∂w ∂w

a 2
+ 2b + c 2
+p +q + sw = 0 (14.1.1.10)
∂x ∂x∂y ∂y ∂x ∂y
admit further simplifications. In general, the substitution
w(x, y) = exp(β1 ξ + β2 η)u(ξ, η) (14.1.1.11)
can be used. Here ξ and η are new variables used to reduce Eq. (14.1.1.10) to canonical
form (see above); the coefficients β1 and β2 in (14.1.1.11) are chosen so that there is only
one first derivative remaining in a parabolic equation or both first derivatives vanish in a
hyperbolic or an elliptic equation. For final results, see Table 14.1.
TABLE 14.1
Reduction of linear homogeneous constant coefficient partial differential equations (14.1.1.10)
using the transformation (14.1.1.11); the constants k and k1 are given by formulas (14.1.1.12)
Type of equation, Variables ξ and η in Coefficients β1 and β2 in Reduced

conditions on coefficients transformation (14.1.1.11) transformation (14.1.1.11) equation
Parabolic equation, c 4cs−q 2 q

ξ = − x, η = y β1 = , β2 = − uξ −uηη = 0
a = b = 0, c 6= 0, p 6= 0 p 4c2 2c
Parabolic equation,
b2 −ac = 0 ξ=
a(ay−bx)
, η=x 4as−p2 p uξ −uηη = 0
β1 = , β2 = −
(aq−bp 6= 0, |a|+|b| =
6 0) bp−aq 4a2 2a
√
Hyperbolic equation, ξ = ay− b+ D x, aq−bp p
√ β1,2 = ± √ uξη +ku = 0
a 6= 0, D = b2 −ac > 0 η = ay− b− D x 4aD 4a D
Hyperbolic equation, ξ = x, cp−2bq p

β1 = , β2 = − 2 uξη +k1u = 0
a = 0, b 6= 0 η = 2by−cx 4b2 4b
Elliptic equation, ξ = ay−bx, aq−bp p

p β1 = , β2 = − p uξξ +uηη +4ku = 0
D = b2 −ac < 0 η = |D| x 2aD 2a |D|
Ordinary
ξ = ay−bx,
differential equation, β1 = β2 = 0 awηη +pwη +sw = 0
η=x
b2 −ac = 0, aq−bp = 0
2◦ . The coefficients k and k1 in the reduced hyperbolic and elliptic equations (see the last
row in Table 14.1) are expressed as
2bpq − aq 2 − cp2 s s cp2 − 2bpq

k= − , k1 = + . (14.1.1.12)
16a(b2 − ac)2 4a(b2 − ac) 4b 2 16b4
If the coefficients in Eq. (14.1.1.10) satisfy the relation
2bpq − aq 2 − cp2 − 4s(b2 − ac) = 0,

then k = 0; in this case with a 6= 0, the general solution of the corresponding hyperbolic
equation has the form

w(x, y) = exp(β1 ξ + β2 η) f (ξ) + g(η) , D = b2 − ac > 0,
√ √
ξ = ay − b + D x, η = ay − b − D x,
aq − bp p aq − bp p
β1 = + √ , β2 = − √ ,
4aD 4a D 4aD 4a D
where f (ξ) and g(ξ) are arbitrary functions.
3◦ . In the degenerate case b2 − ac = 0, aq − bp = 0 (where the original equation is reduced

to an ordinary differential equation; see the last row in Table 14.1), the general solution of
Eq. (14.1.1.10) is expressed as
√ √
px x λ x λ
w = exp − f (ay − bx) exp + g(ay − bx) exp − if λ = p2− 4as > 0,
2a 2a 2a
p p
px x |λ| x |λ|
w = exp − f (ay − bx) sin + g(ay − bx) cos if λ = p2− 4as < 0,
2a 2a 2a

px
w = exp − f (ay − bx) + xg(ay − bx) if p2− 4as = 0,
2a
where f (z) and g(z) are arbitrary functions.
14.1.2 Equations with Many Independent Variables

Let us look at a second-order partial differential equation with n independent variables
x1 , . . . , xn of the form
n
X
∂2w ∂w ∂w
aij (x) = F x, w, ,..., , (14.1.2.1)
∂xi ∂xj ∂x1 ∂xn
i,j=1
where the aij are some functions that have continuous derivatives of order ≤ 2 with re-
spect to all variables and x = {x1 , . . . , xn }. (The right-hand side of Eq. (14.1.2.1) may be
nonlinear. The left-hand side only is required for the classification of this equation.)
At a point x = x0 , the following quadratic form is assigned to Eq. (14.1.2.1):
n
X
Q= aij (x0 )ξi ξj . (14.1.2.2)
i,j=1
An appropriate linear nonsingular transformation

n
X
ξi = βik ηk (i = 1, . . . , n) (14.1.2.3)
k=1
reduces the quadratic form (14.1.2.2) to the canonical form

n
X
Q= ci ηi2 , (14.1.2.4)
i=1
where the coefficients ci assume the values 1, −1, and 0. The number of negative and
zero coefficients in (14.1.2.4) does not depend on the way in which the quadratic form is
reduced to the canonical form.
Table 14.2 presents the basic criteria according to which equations with many indepen-
dent variables are classified.
TABLE 14.2
Classification of equations with many independent variables
Type of Eq. (14.1.2.1) at a point x = x0 Coefficients of the canonical form (14.1.2.4)
Parabolic (in the broad sense) At least one of the coefficients ci is zero
Hyperbolic (in the broad sense) All ci are nonzero and some ci differ in sign
Elliptic All ci are nonzero and have like signs
Suppose all coefficients of the highest derivatives in (14.1.2.1) are constant, aij = const.
By introducing the new independent variables y1 , . . . , yn in accordance with the formulas
Pn
yi = βik xk , where the βik are the coefficients of the linear transformation (14.1.2.3),
k=1
we reduce Eq. (14.1.2.1) to the canonical form
n
X
∂2w ∂w ∂w
ci 2 = F1 y, w, ,..., , (14.1.2.5)
∂yi ∂y1 ∂yn
i=1
where y = {y1 , . . . , yn }; the coefficients ci are the same as in the quadratic form (14.1.2.4).
Remark 14.5. Among the parabolic equations, it is conventional to distinguish the parabolic
equations in the narrow sense, i.e., the equations for which only one of the coefficients, ck , is zero,
while the other ci is the same, and in this case the right-hand side of Eq. (14.1.2.5) must contain the
first partial derivative with respect to yk .
Remark 14.6. In turn, the hyperbolic equations are divided into normal hyperbolic equations—
for which all ci but one have like signs—and ultrahyperbolic equations—for which there are two
or more positive ci and two or more negative ci .
Specific equations of parabolic, elliptic, and hyperbolic types will be discussed further
in Section 14.2.
⊙ Literature for Section 14.2: S. G. Mikhlin (1970), V. S. Vladimirov (1971, 1988), S. J. Farlow (1982),
R. Courant and D. Hilbert (1989), E. Zauderer (1989), A. N. Tikhonov and A. A. Samarskii (1990), I. G. Petrov-
sky (1991), W. A. Strauss (1992), C. Constanda (2002), A. D. Polyanin (2002), A. D. Polyanin and A. V. Man-
zhirov (2007).
14.2 Basic Problems of Mathematical Physics

14.2.1 Initial and Boundary Conditions. Cauchy Problem. Boundary
Value Problems
Every equation of mathematical physics describes infinitely many qualitatively similar phe-
nomena or processes. This follows from the fact that differential equations have infinitely
many particular solutions. The specific solution that describes the physical phenomenon
under study is separated from the set of particular solutions of the given differential equa-
tion by means of the initial and boundary conditions.
Throughout this section, we consider linear equations in the n-dimensional Euclidean
space Rn or in an open domain V ∈ Rn (exclusive of the boundary) with a sufficiently
smooth boundary S = ∂V .
◮ Second-order parabolic equations. Initial and boundary conditions.

In general, a linear second-order partial differential equation of the parabolic type with
n independent variables can be written as
∂w
− Lx [w] = Φ(x, t), (14.2.1.1)
∂t
where
n
X X n
∂2w ∂w
Lx [w] ≡ aij (x, t) + bi (x, t) + c(x, t)w, (14.2.1.2)
∂xi ∂xj ∂xi
i,j=1 i=1
n
X n
X
x = {x1 , . . . , xn }, aij (x, t)ξi ξj ≥ σ ξi2 , σ > 0.
i,j=1 i=1
Parabolic equations describe unsteady thermal, diffusion, and other phenomena depen-
dent on time t.
Equation (14.2.1.1) is said to be homogeneous if Φ(x, t) ≡ 0.
Cauchy problem (t ≥ 0, x ∈ Rn ). Find a function w that satisfies Eq. (14.2.1.1) for t > 0
and the initial condition
w = f (x) at t = 0. (14.2.1.3)
Boundary value problem∗ (t ≥ 0, x ∈ V ). Find a function w that satisfies Eq. (14.2.1.1)
for t > 0, the initial condition (14.2.1.3), and the boundary condition
Γx [w] = g(x, t) at x∈S (t > 0). (14.2.1.4)
In general, Γx is a first-order linear differential operator in the space variables x with co-
efficient depending on x and t. The basic types of boundary conditions are described in
Section 14.2.2.
The initial condition (14.2.1.3) is said to be homogeneous if f (x) ≡ 0. The boundary
condition (14.2.1.4) is said to be homogeneous if g(x, t) ≡ 0.
∗
Boundary value problems for parabolic and hyperbolic equations are sometimes called mixed or initial-
boundary value problems.
14.2. Basic Problems of Mathematical Physics 1141
◮ Second-order hyperbolic equations. Initial and boundary conditions.

Consider a second-order linear partial differential equation of the hyperbolic type with n
independent variables of the general form
∂2w ∂w
+ ϕ(x, t) − Lx [w] = Φ(x, t), (14.2.1.5)
∂t2 ∂t
where the linear differential operator Lx is defined in (14.2.1.2). Hyperbolic equations
describe unsteady wave processes, which depend on time t.
Equation (14.2.1.5) is said to be homogeneous if Φ(x, t) ≡ 0.
Cauchy problem (t ≥ 0, x ∈ Rn ). Find a function w that satisfies Eq. (14.2.1.5) for t > 0
and the initial conditions
w = f0 (x) at t = 0,
(14.2.1.6)
∂t w = f1 (x) at t = 0.
Boundary value problem (t ≥ 0, x ∈ V ). Find a function w that satisfies Eq. (14.2.1.5)
for t > 0, the initial conditions (14.2.1.6), and boundary condition (14.2.1.4).
The initial conditions (14.2.1.6) are said to be homogeneous if f0 (x) ≡ 0 and f1 (x) ≡ 0.
Generalized Cauchy problem. In the generalized Cauchy problem for a hyperbolic
equation with two independent variables, the values of the unknown function and its first
derivatives are prescribed on a curve in the (x, t) plane. Alternatively, the values of the
unknown function and its derivative along the normal to this curve may be prescribed. For
more details, see Section 17.2.4.
Goursat problem. On the characteristics of a hyperbolic equation with two indepen-
dent variables, the values of the unknown function w are prescribed; for details, see Sec-
tion 17.2.5.
◮ Second-order elliptic equations. Boundary conditions.

In general, a second-order linear partial differential equation of elliptic type with n inde-
pendent variables can be written as
Lx [w] = Φ(x), (14.2.1.7)
where
n
X X n
∂2w ∂w
Lx [w] ≡ aij (x) + bi (x) + c(x)w, (14.2.1.8)
∂xi ∂xj ∂xi
i,j=1 i=1
X n n
X
aij (x)ξi ξj ≥ σ ξi2 , σ > 0.
i,j=1 i=1
Elliptic equations describe steady-state thermal, diffusion, and other phenomena indepen-
dent of time t.
Equation (14.2.1.7) is said to be homogeneous if Φ(x) ≡ 0.
Boundary value problem. Find a function w that satisfies Eq. (14.2.1.7) and the bound-
ary condition
Γx [w] = g(x) at x ∈ S. (14.2.1.9)
In general, Γx is a first-order linear differential operator in the space variables x. The basic
types of boundary conditions are described below in Section 14.2.2.
The boundary condition (14.2.1.9) is said to be homogeneous if g(x) ≡ 0. The boundary
value problem (14.2.1.7)–(14.2.1.9) is said to be homogeneous if Φ ≡ 0 and g ≡ 0.
◮ Adjoint and self-adjoint differential operators. Green formula.

The linear differential operator
n
X n
∂2 X ∂
L∗x [w] ≡ aij (x)w − bi (x)w + c(x)w
∂xi ∂xj ∂xi
i,j=1 i=1
is called the adjoint of Lx ; see (14.2.1.8). One has (L∗x )∗ = Lx .

If the operator Lx coincides with L∗x , then it is said to be self-adjoint. A self-adjoint
operator can be reduced to the form
n
X
∂ ∂w
Lx [w] ≡ aij (x) + c(x)w. (14.2.1.10)
∂xi ∂xj
i,j=1
One has
n
X ∂Pi
uLx [w] − wL∗x [u] = ,
∂xi
i=1
where
n
X
∂w ∂
Pi = uaij −w (aij u) + bi uw.
∂xJ ∂xj
j=1
Let V be a bounded domain whose boundary is a piecewise smooth surface S. Then

the Green formula
Z Z X
n

uLx [w] − wL∗x [u] dV = Pi cos(ν, xi ) dS
V S i=1
holds, where dV = dx1 . . . dxn and ν is the outward normal direction on S.

Example 14.2. For the 3D Laplace operator, we have ∆ = ∆∗ , and the Green formula acquires
the form Z Z
∂w ∂u
u∆w − w∆u dV = u −w dS,
V S ∂ν ∂ν
where ∂/∂ν is the outward normal derivative. In applications, one often uses the special case of this
formula with u = 1.
14.2.2 First, Second, Third, and Mixed Boundary Value Problems

For any (parabolic, hyperbolic, and elliptic) second-order partial differential equations, it is
conventional to distinguish four basic types of boundary value problems depending on the
form of the boundary conditions (14.2.1.4) [see also the analogous condition (14.2.1.9)].
14.2. Basic Problems of Mathematical Physics 1143
For simplicity, here we confine ourselves to the case where the coefficients aij of equations
(14.2.1.1) and (14.2.1.5) have the special form
(
1 if i = j,
aij (x, t) = a(x, t)δij , δij =
0 if i 6= j.
This situation is rather frequent in applications; such coefficients are used to describe vari-
ous phenomena (processes) in isotropic media.
First boundary value problem. The function w(x, t) takes prescribed values at the bo-
undary S of the domain:
w(x, t) = g1 (x, t) for x ∈ S. (14.2.2.1)
Second boundary value problem. The derivative along the (outward) normal is pre-
scribed at the boundary S of the domain:
∂w
= g2 (x, t) for x ∈ S. (14.2.2.2)
∂N
In heat transfer problems, where w is temperature, the left-hand side of the boundary con-
dition (14.2.2.2) is proportional to the heat flux per unit area of the surface S.
Third boundary value problem. A linear relationship between the unknown function
and its normal derivative is prescribed at the boundary S of the domain:
∂w
+ k(x, t)w = g3 (x, t) for x ∈ S. (14.2.2.3)
∂N
Usually, it is assumed that k(x, t) = const. In mass transfer problems, where w is the
concentration, the boundary condition (14.2.2.3) with g3 ≡ 0 describes a surface chemical
reaction of the first order.
Mixed boundary value problems. Conditions of various types listed above are set at
various portions of the boundary S.
If g1 ≡ 0, g2 ≡ 0, or g3 ≡ 0, then the respective boundary conditions (14.2.2.1),
(14.2.2.2), (14.2.2.3) are said to be homogeneous.
Boundary conditions for various boundary value problems for parabolic and hyperbolic
equations in two independent variables x and t are displayed in Table 14.3. The equation
coefficients are assumed to be continuous, with the coefficients of the highest derivatives
being nonzero in the range x1 ≤ x ≤ x2 considered.
Remark 14.7. For elliptic equations, the first boundary value problem is often called the Dirich-
let problem, and the second boundary value problem is called the Neumann problem.
R. Leis (1986), A. A. Dezin (1987), R. Haberman (1987), R. Courant and D. Hilbert (1989), E. Zauderer (1989),
A. N. Tikhonov and A. A. Samarskii (1990), I. G. Petrovsky (1991), W. A. Strauss (1992), R. B. Guenther and
J. W. Lee (1996), D. Zwillinger (1998), I. Stakgold (2000), C. Constanda (2002), A. D. Polyanin (2002),
A. D. Polyanin and A. V. Manzhirov (2007).
TABLE 14.3
Boundary conditions for various boundary value problems specified by parabolic
and hyperbolic equations in two independent variables (x1 ≤ x ≤ x2 )
Type of problem Boundary condition at x = x1 Boundary condition at x = x2
First boundary value problem w = g1(t) w = g2(t)

Second boundary value problem ∂xw = g1(t) ∂xw = g2(t)
Third boundary value problem ∂xw + β1w = g1(t) (β1 < 0) ∂xw + β2w = g2(t) (β2 > 0)
Mixed boundary value problem w = g1(t) ∂xw = g2(t)
Mixed boundary value problem ∂xw = g1(t) w = g2(t)
14.3 Properties and Particular Solutions of Linear

Equations
14.3.1 Homogeneous Linear Equations. Basic Properties
of Particular Solutions
◮ Preliminary remarks.
For brevity, in this section a homogeneous linear partial differential equation will be written
as
L[w] = 0. (14.3.1.1)
For second-order linear parabolic and hyperbolic equations, the linear differential opera-
tor L[w] is defined by the left-hand side of Eqs. (14.2.1.1) and (14.2.1.5), respectively. It is
assumed that Eq. (14.3.1.1) is an arbitrary homogeneous linear partial differential equation
of any order in the variables t, x1 , . . . , xn with sufficiently smooth coefficients.
The linear operator L possesses the properties
L[w1 + w2 ] = L[w1 ] + L[w2 ],

L[Aw] = AL[w], A = const.
An arbitrary homogeneous linear equation (14.3.1.1) has the trivial solution w ≡ 0.
A function w is called a classical solution of Eq. (14.3.1.1) if w, when substituted
into (14.3.1.1), turns the equation into an identity and if all partial derivatives of w that
occur in (14.3.1.1) are continuous; the notion of a classical solution is directly linked to the
range of the independent variables. In what follows, we usually write “solution” instead of
“classical solution” for brevity.
◮ Use of particular solutions for the construction of other solutions.

Below are basic properties of particular solutions of homogeneous linear equations.
1◦ . Let w1 = w1 (x, t), w2 = w2 (x, t), . . . , wk = wk (x, t), where x = (x1 , . . . , xn ), be any
particular solutions of the homogeneous equation (14.3.1.1). Then the linear combination
w = A1 w1 + A2 w2 + · · · + Ak wk (14.3.1.2)
14.3. Properties and Particular Solutions of Linear Equations 1145
with arbitrary constants A1 , A2 , . . . , Ak is a solution of Eq. (14.3.1.1) as well; in physics,

this property is known as the linear superposition principle.
P∞Assume that {wk } is an infinite sequence of solutions of Eq. (14.3.1.1). Then the series
k=1 wk , irrespective of its convergence, is called a formal solution of (14.3.1.1). If the
solutions wk are classical, the series is uniformly convergent, and the sum of the series
has all the necessary partial derivatives, then the sum of the series is a classical solution of
Eq. (14.3.1.1).
2◦ . Let the coefficients of the linear differential operator L be independent of time t. If
Eq. (14.3.1.1) has a particular solution w e = w(x,
e t), then the function w(x,
e t + a), where a
is an arbitrary constant, is a solution of the equation as well.
If the coefficients of L are independent of only one space coordinate, say xk , and
e t), then the function w(x,
e t)|xk ⇒xk +b ,
where b is an arbitrary constant, is a solution of the equation as well.
3◦ . Let the coefficients of the linear differential operator L be independent of time t. If
e t), then the partial derivatives of w
e with
respect to time, ∗
∂we ∂2w e ∂k w
e
, 2
, . . . , k
, ...,
∂t ∂t ∂t
are solutions of Eq. (14.3.1.1) as well.
4◦ . Let the coefficients of the linear differential operator L be independent of the space
variables x1 , . . . , xn . If Eq. (14.3.1.1) has a particular solution w
e = w(x,
e t), then the partial
derivatives of w e with respect to the space coordinates
∂we ∂we ∂we ∂2w

e ∂2w e ∂ k+m w
e
, , , ..., , , ..., , ...
∂x1 ∂x2 ∂x3 ∂x21 ∂x1 ∂x2 ∂x2 ∂xm
k
3

If the coefficients of L are independent of only one space coordinate, say x1 , and
Eq. (14.3.1.1) has a particular solution w
e = w(x,
e t), then the partial derivatives
∂we ∂2w
e ∂k w
e
, , ..., , ...
∂x1 ∂x21 ∂xk1

5◦ . Let the coefficients of the linear differential operator L be constant and let Eq. (14.3.1.1)
have a particular solution w e = w(x,
e t). Then any particular derivatives of w e with respect to
time and the space coordinates (inclusive mixed derivatives)
∂w
e ∂we ∂2w
e ∂2w
e ∂k w
e
, , ..., , , ..., , ...
∂t ∂x1 ∂x22 ∂t∂x1 ∂xk3
are solutions of Eq. (14.3.1.1).

∗
Here and in what follows, it is assumed that the particular solution w
e is differentiable sufficiently many
times with respect to t and x1 , . . . , xn (or the parameters).
6◦ . Suppose that Eq. (14.3.1.1) has a particular solution dependent on a parameter µ, w e=

w(x,
e t; µ), and the coefficients of the linear differential operator L are independent of µ (but
can depend on time and the space coordinates). Then, by differentiating w e with respect to µ,
one obtains other solutions of Eq. (14.3.1.1),
∂w
e ∂2w
e ∂k w
e
, , ..., , ...
∂µ ∂µ2 ∂µk
Let some constants µ1 , . . . , µk belong to the range of the parameter µ. Then the sum
w = A1 w(x,
e t; µ1 ) + · · · + Ak w(x,
e t; µk ), (14.3.1.3)
where A1 , . . . , Ak are arbitrary constants, is a solution of the homogeneous linear equa-

tion (14.3.1.1) as well. The number of terms in sum (14.3.1.3) can be finite as well as
infinite.
7◦ . Another efficient way of constructing solutions involves the following. The particular
solution w(x,
e t; µ), which depends on the parameter µ (as before, it is assumed that the
coefficients of the linear differential operator L are independent of µ), is first multiplied by
an arbitrary function ϕ(µ). Then the resulting expression is integrated with respect to µ
over some interval [α, β]. Thus, one obtains a new function,
Z β
w(x,
e t; µ)ϕ(µ) dµ,
α
which is also a solution of the original homogeneous linear equation.

The properties listed in Items 1◦ – 7◦ enable one to use known particular solutions
to construct other particular solutions of homogeneous linear equations of mathematical
physics.
Example 14.3. The linear equation
∂w ∂2w
= a 2 + bw
∂t ∂x
has the particular solution
e t) = exp[µx + (aµ2 + b)t],
w(x,
where µ is an arbitrary constant. Differentiating this equation with respect to µ (see Item 6◦ ) yields
another solution
e t) = (x + 2aµt) exp[µx + (aµ2 + b)t].
w(x,
Example 14.4. The wave equation
∂2w ∂2w
2
= a2 2
∂t ∂x
has particular solutions
e1 (x, t) = eµ(x+at) ,
w e2 (x, t) = eµ(x−at) ,
w
where µ is an arbitrary constant. According to Item 7◦ , the equation also has more general solutions
of the form
Z α Z α
µ(x+at)
w
e1 (x, t) = ϕ1 (µ)e dµ = F1 (x+at), w
e2 (x, t) = ϕ2 (µ)eµ(x−at) dµ = F2 (x−at).
β β
Since the functions ϕ1 (µ) and ϕ2 (µ) can be chosen arbitrarily, we can assume the functions F1 (z1 )
and F2 (z2 ) to be arbitrary as well. (This can readily be verified by a straightforward substitution
into the original equation.)
By the linear superposition principle (see Item 1◦ ), one can add the resulting solutions. As a
result, we arrive at the d’Alembert formula
w = F1 (x + at) + F2 (x − at).
14.3.2 Separable Solutions. Solutions in the Form of Infinite Series

◮ Multiplicative and additive separable solutions.
1◦ . Many homogeneous linear partial differential equations have solutions that can be rep-
resented as products of functions depending on distinct arguments. Such solutions are re-
ferred to as multiplicative separable solutions; very commonly these solutions are briefly,
but less accurately, called just separable solutions.
Table 14.4 presents the most commonly encountered types of homogeneous linear dif-
ferential equations with many independent variables that admit exact separable solutions.
Linear combinations of particular solutions that correspond to various values of the sepa-
ration parameters, λ, β1 , . . . , βn , are solutions of the equations in question as well. For
brevity, the word “operator” is used to denote “linear differential operator.”
For a constant coefficient equation (see the first row in Table 14.4), the separation pa-
rameters must satisfy the algebraic equation
D(λ, β1 , . . . , βn ) = 0, (14.3.2.1)
which results from substituting the solution into Eq. (14.3.1.1). In physical applications,
Eq. (14.3.2.1) is usually referred to as the dispersion equation. Any n out of the n + 1
separation parameters in (14.3.2.1) can be treated as arbitrary.
Example 14.5. Consider the linear equation
∂2w ∂w ∂2w ∂w
2
+k = a2 2 + b + cw.
∂t ∂t ∂x ∂x
A particular solution is sought in the form
w = A exp(βx + λt).
This results in the dispersion equation λ2 + kλ = a2 β 2 + bβ + c, where one of the two parameters
β or λ can be treated as arbitrary.
For more complex multiplicative separable solutions of this equation and related equations, see
Section 15.1.1.
Note that constant coefficient equations also admit more sophisticated solutions; see the
second and third rows, the last column.
TABLE 14.4
Homogeneous linear partial differential equations that admit multiplicative separable solutions
No. Form of Eq. (14.3.1.1) Form of particular solutions
Equation coefficients are w(x, t) = A exp(λt + β1 x1 + · · · + βn xn ),

1
constant λ, β1 , . . . , βn are related by an algebraic equation
Equation coefficients are w(x, t) = eλt ψ(x),

2 independent of time t λ is an arbitrary constant, x = {x1 , . . . , xn }
Equation coefficients are independent w(x, t) = exp(β1 x1 + · · · + βn xn )ψ(t),

3
of the coordinates x1 , . . . , xn β1 , . . . , βn are arbitrary constants
Equation coefficients are independent w(x, t) = exp(β1x1 + · · · + βkxk)ψ(t, xk+1, . . . , xn),
4
of the coordinates x1 , . . . , xk β1 , . . . , βk are arbitrary constants
Lt [w] + Lx [w] = 0, w(x, t) = ϕ(t)ψ(x),

5 operator Lt depends only on t, ϕ(t) satisfies the equation Lt [ϕ] + λϕ = 0,
operator Lx depends only on x ψ(x) satisfies the equation Lx [ψ] − λψ = 0
w(x, t) = ϕ(t)ψ1 (x1 ) . . . ψn (xn ),

Lt [w] + L1 [w] + · · · + Ln [w] = 0,
ϕ(t) satisfies the equation Lt [ϕ] + λϕ = 0,
6 operator Lt depends only on t,
ψk (xk ) satisfies the equation Lk [ψk ] + βk ψk = 0,
operator Lk depends only on xk
λ + β1 + · · · + βn = 0
w(x, t) = ϕ(t)ψ1 (x1 ) . . . ψn (xn ),
P
n
f0 (x1 )Lt [w] + fk (x1 )Lk [w] = 0, Lt [ϕ] + λϕ = 0,
k=1 Lk [ψk ] + βk ψk = 0, k = 2, . . . , n,
7
operator Lt depends only on t, h P
n i
operator Lk depends only on xk f1 (x1 )L1 [ψ1 ] − λf0 (x1 ) + βk fk (x1 ) ψ1 = 0
k=2
∂w w(x, t) = ψ1 (x1 , t)ψ2 (x2 , t) . . . ψn (xn , t),

+ L1 [w] + · · · + Ln [w] = 0,
∂t ∂ψk
8 m
Pk ∂sw + Lk [ψk ] = λk (t)ψk , k = 1, . . . , n,
where Lk [w] = fks (xk , t) s ∂t
s=0 ∂xk λ1 (t) + λ2 (t) + · · · + λn (t) = 0
The eighth row of Table 14.4 presents the case of incomplete separation of variables
where the solution is separated with respect to the space variables x1 , . . . , xn , but is not
separated with respect to time t.
Remark 14.8. For stationary equations that do not depend on t, one should set λ = 0, Lt [w] ≡ 0,
and ϕ(t) ≡ 1 in rows 1, 6, and 7 of Table 14.4.
Remark 14.9. Multiplicative separable solutions play an important role in the theory of linear
partial differential equations; they are used for finding solutions of stationary and nonstationary
boundary value problems; see Chapter 15.
2◦ . Linear partial differential equations of the form
Lt [w] + Lx [w] + kw = f (x) + g(t),
where Lt is a linear differential operator that depends on only t and Lx is a linear differen-
tial operator that depends on only x, have solutions that can be represented as the sum of
functions depending on distinct arguments,
w = u(x) + v(t).
Such solutions are referred to as additive separable solutions.

Example 14.6. The equation in Example 14.5 admits an exact additive separable solution w =
u(x) + v(t) with u(x) and v(t) described by the linear constant coefficient ordinary differential
equations
a2 u′′xx + bu′x + cu = C,
′′
vtt + kvt′ − cv = C,
where C is an arbitrary constant, which are easy to integrate. A more general partial differential
equation with variable coefficients a = a(x), b = b(x), k = k(t), and c = const admits an additive
separable solution as well.
◮ Solutions in the form of infinite series in t.

1◦ . The equation
∂w
= M [w],
∂t
where M is an arbitrary linear differential operator of the second (or any) order that only
depends on the space variables, has the formal series solution
∞ k
X t
w(x, t) = f (x) + M k [f (x)], M k [f ] = M M k−1 [f ] ,
k!
k=1
condition w(x, 0) = f (x).
Example 14.7. Consider the heat equation
∂w ∂2w
=a 2.
∂t ∂x
2
∂
In this case we have M = a ∂x 2 . Therefore the formal series solution has the form
∞
X (at)k (2n) dm
w(x, t) = f (x) + f (x), fx(m) = f (x).
k! x dxm
k=1
If the function f (x) is a polynomial of degree n, then so is the solution. For example, by setting
f (x) = Ax2 + Bx + C, we obtain the particular solution
w(x, t) = A(x2 + 2at) + Bx + C.
2◦ . The equation
∂2w
= M [w],
∂t2
where M is a linear differential operator, just as in Item 1◦ , has a formal solution repre-
sented by the sum of two series as
∞
X ∞
X
t2k k t2k+1
w(x, t) = M [f (x)] + M k [g(x)],
(2k)! (2k + 1)!
k=0 k=0
where f (x) and g(x) are arbitrary infinitely differentiable functions. This solution satisfies
the initial conditions w(x, 0) = f (x) and ∂t w(x, 0) = g(x).
14.3.3 Nonhomogeneous Linear Equations and Their Properties

For brevity, we write a nonhomogeneous linear partial differential equation in the form
L[w] = Φ(x, t). (14.3.3.1)
The linear differential operator L is defined above (see the beginning of Section 14.3.1).
Below are the simplest properties of particular solutions of the nonhomogeneous equa-
tion (14.3.3.1).
1◦ . If weΦ (x, t) is a particular solution of the nonhomogeneous equation (14.3.3.1) and
we0 (x, t) is a particular solution of the corresponding homogeneous equation (14.3.1.1),
then the sum
Aw e0 (x, t) + w
eΦ (x, t),
where A is an arbitrary constant, is a solution of the nonhomogeneous equation (14.3.3.1)
as well. The following, more general statement holds: The general solution of the non-
homogeneous equation (14.3.3.1) is the sum of the general solution of the corresponding
homogeneous equation (14.3.1.1) and any particular solution of the nonhomogeneous equa-
tion (14.3.3.1).
2◦ . Suppose that w1 and w2 are solutions of nonhomogeneous linear equations with the
same left-hand side and distinct right-hand sides; i.e.,
L[w1 ] = Φ1 (x, t), L[w2 ] = Φ2 (x, t).
Then the function w = w1 + w2 is a solution of the equation
L[w] = Φ1 (x, t) + Φ2 (x, t).
14.3.4 General Solutions of Some Hyperbolic Equations

◮ D’Alembert’s solution of the wave equation.
The wave equation
∂2w 2
2∂ w
− a =0 (14.3.4.1)
∂t2 ∂x2
has the general solution
w = ϕ(x + at) + ψ(x − at), (14.3.4.2)
where ϕ(x) and ψ(x) are arbitrary twice continuously differentiable functions. The solu-
tion (14.3.4.2) has the physical interpretation of two traveling waves of arbitrary shape that
propagate to the left and to the right along the x-axis at a constant speed a (a > 0).
◮ Laplace cascade method for hyperbolic equation in two variables.

A general linear hyperbolic equation with two independent variables can be reduced to an
equation of the form (see Section 14.1.1):
∂2w ∂w ∂w
+ a(x, y) + b(x, y) + c(x, y)w = f (x, y). (14.3.4.3)
∂x∂y ∂x ∂y
Sometimes it is possible to obtain formulas determining all solutions of Eq. (14.3.4.3).
First consider two special cases.
1◦ . First special case. Suppose that the identity

∂a
g≡ + ab − c ≡ 0 (14.3.4.4)
∂x
is valid; for brevity, the arguments of the functions are omitted. Then Eq. (14.3.4.3) can be
rewritten in the form
∂u
+ bu = f, (14.3.4.5)
∂x
where
∂w
u= + aw. (14.3.4.6)
∂y
Equation (14.3.4.5) is a linear first-order ordinary differential equation in x for u (the vari-
able y appears in the equation as a parameter) and is easy to integrate. Further, substitut-
ing u into (14.3.4.6) yields a linear first-order ordinary differential equation in y for w (now
x appears in the equation as a parameter). On solving this equation, one obtains the general
solution of the original equation (14.3.4.3) subject to condition (14.3.4.4):
Z n Z h Z Z i Z Z o
w = exp − a dy ϕ(x)+ ψ(y)+ f exp b dx dx exp a dy− b dx dy ,
where ϕ(x) and ψ(y) are arbitrary functions.

2◦ . Second special case. Suppose the identity
∂b
h≡ + ab − c ≡ 0 (14.3.4.7)
∂y
holds true. Proceeding in the same ways as in the first special case, one obtains the general
solution of (14.3.4.3):
Z n Z h Z Z i Z Z o
w = exp − b dx ψ(y)+ ϕ(x)+ f exp a dy dy exp b dx− a dy dx .
3◦ . Laplace cascade method. If g 6= 0, consider the new equation of the form (14.3.4.3),
∂ 2 w1 ∂w1 ∂w1
L1 [w1 ] ≡ + a1 (x, y) + b1 (x, y) + c1 (x, y)w1 = f1 (x, y), (14.3.4.8)
∂x∂y ∂x ∂y
where

∂ ln g ∂a ∂b ∂ ln g ∂ ln g
a1 = a − , b1 = b, c1 = c − + −b , f1 = a− f.
∂y ∂x ∂y ∂y ∂y
Whenever a solution w1 is found, the corresponding solution of the original equation
(14.3.4.3) can be obtained by the formula

1 ∂w1
w= + bw1 − f .
g ∂x
For Eq. (14.3.4.8), the functions similar to (14.3.4.4) and (14.3.4.7) are expressed as
∂ 2 ln g
g1 = 2g − h − , h1 = h.
∂x∂y
If g1 ≡ 0, then the function w1 can be found using the technique described above. If g1 6≡ 0,
one proceeds to the construction, in the same way as above, of the equation L2 [w2 ] = f2 ,
and so on. If h 6≡ 0, then a similar chain of equations may be constructed: L∗1 [w1∗ ] = f1∗ ,
L∗2 [w2∗ ] = f2∗ , etc.
If at some step gk or hk vanishes, then it is possible to obtain the general solution of
Eq. (14.3.4.3).
Example 14.8. Consider the Euler–Darboux equation
∂ 2w α ∂w β ∂w
− + = 0.
∂x∂y x − y ∂x x − y ∂y
We will show that its general solution can be obtained if at least one of the numbers α or β is an
integer.
With the notation adopted for Eq. (14.3.4.3) and the function (14.3.4.4), we have
α β α(1 − β)
a(x, y) = − , b(x, y) = , c(x, y) = f (x, y) = 0, g= ,
x−y x−y (x − y)2
which means that g ≡ 0 if α = 0 or β = 1. If g 6≡ 0, then we construct Eq. (14.3.4.8), where
2+α β α+β
a1 = − , b1 = , c1 = − .
x−y x−y (x − y)2
If follows that
(1 + α)(2 − β)
g1 = ,
(x − y)2
and hence g1 ≡ 0 at α = −1 or β = 2. Similarly, it can be shown that gk ≡ 0 at α = −k or β = k + 1
(k = 0, 1, 2, . . . ). If we use the other sequence of auxiliary equations, L∗k [wk∗ ] = fk∗ , it can be
shown that the above holds for α = 1, 2, . . . and β = 0, −1, −2, . . .
⊙ Literature for Section 14.3: S. G. Mikhlin (1970), W. A. Strauss (1992), G. A. Korn and T. M. Korn (2000),
A. D. Polyanin (2002), A. D. Polyanin and A. V. Manzhirov (2007).
Chapter 15
Separation of Variables
and Integral Transform Methods
15.1 Separation of Variables (Fourier Method)

15.1.1 Description of Separation of Variables. General Stage
of Solution
◮ Scheme of solving boundary value problems by separation of variables.
Many linear problems of mathematical physics can be solved by separation of variables.

Figure 15.1 depicts the scheme of application of this method to solve boundary value prob-
lems for second-order homogeneous linear equations of the parabolic and hyperbolic type
with homogeneous boundary conditions and nonhomogeneous initial conditions. For sim-
plicity, problems with two independent variables x and t are considered with x1 ≤ x ≤ x2
and t ≥ 0.
The scheme presented in Fig. 15.1 applies to boundary value problems for second-order
linear homogeneous partial differential equations of the form
∂2w ∂w ∂2w ∂w
α(t) + β(t) = a(x) + b(x) + c(x) + γ(t) w (15.1.1.1)
∂t2 ∂t ∂x2 ∂x
with homogeneous linear boundary conditions
s1 ∂x w + k1 w = 0 at x = x1 ,
(15.1.1.2)
s2 ∂x w + k2 w = 0 at x = x2
and arbitrary initial conditions
w = f0 (x) at t = 0, (15.1.1.3)
∂t w = f1 (x) at t = 0. (15.1.1.4)
For parabolic equations, which correspond to α(t) ≡ 0 in (15.1.1.1), only the initial
condition (15.1.1.3) is set.
1153
1154 S EPARATION OF VARIABLES AND I NTEGRAL T RANSFORM M ETHODS
Figure 15.1: Scheme of solving linear boundary value problems by separation of variables
′′ , and all B = 0).
(for parabolic equations, the function F2 does not depend on ψtt n
15.1. Separation of Variables (Fourier Method) 1155
Below we consider the basic stages of separation of variables in more detail. We as-
sume that the coefficients of Eq. (15.1.1.1) and the boundary conditions (15.1.1.2) meet the
following requirements:
α(t), β(t), γ(t), a(x), b(x), c(x) are continuous functions,

α(t) ≥ 0, a(x) > 0, |s1 | + |k1 | > 0, |s2 | + |k2 | > 0.
Remark 15.1. Separation of variables is also used to solve linear boundary value problems
for elliptic equations of the form (15.1.1.1) with α(t) < 0 and a(x) > 0 and with the boundary
conditions (15.1.1.2) in x and similar boundary conditions in t. In this case, all results obtained for
the general stage of solution described below remain valid; for details, see Section 15.1.4.
Remark 15.2. In various applications, equations of the form (15.1.1.1) may arise with the co-
efficient b(x) going to infinity at the boundary, b(x) → ∞ as x → x1 , with the other coefficients
being continuous. In this case, the first boundary condition in (15.1.1.2) should be replaced with
a condition of boundedness of the solution as x → x1 . This may occur in spatial problems with
central or axial symmetry where the solution depends only on the radial coordinate.
◮ Derivation of equations and boundary conditions for particular solutions.

The approach is based on searching for particular solutions of Eq. (15.1.1.1) in the product
form
w(x, t) = ϕ(x) ψ(t). (15.1.1.5)
After separation of the variables and elementary manipulations, one arrives at the following
linear ordinary differential equations for the functions ϕ = ϕ(x) and ψ = ψ(t):
a(x)ϕ′′xx + b(x)ϕ′x + [λ + c(x)]ϕ = 0, (15.1.1.6)

′′
α(t)ψtt + β(t)ψt′ + [λ − γ(t)]ψ = 0. (15.1.1.7)
These equations contain a free parameter λ called the separation constant. With the no-
tation adopted in Fig. 15.1, Eqs. (15.1.1.6) and (15.1.1.7) can be rewritten as follows:
ϕF1 (x, ϕ, ϕ′x , ϕ′′xx ) + λϕ = 0 and ψF2 (t, ψ, ψt′ , ψtt
′′ ) + λψ = 0.
Substituting (15.1.1.5) into (15.1.1.2) yields the boundary conditions for ϕ = ϕ(x):
s1 ϕ′x + k1 ϕ = 0 at x = x1 ,
(15.1.1.8)
s2 ϕ′x + k2 ϕ = 0 at x = x2 .
The homogeneous linear ordinary differential equation (15.1.1.6) in conjunction with the
homogeneous linear boundary conditions (15.1.1.8) forms an eigenvalue problem.
◮ Solution of eigenvalue problems. Orthogonality of eigenfunctions.

Suppose that ϕe1 = ϕe1 (x, λ) and ϕ
e2 = ϕ
e2 (x, λ) are linearly independent particular solutions
of Eq. (15.1.1.6). Then the general solution of this equation can be represented as the linear
combination
ϕ = C1 ϕe1 (x, λ) + C2 ϕe2 (x, λ), (15.1.1.9)
Substituting the solution (15.1.1.9) into the boundary conditions (15.1.1.8) yields the
following homogeneous linear algebraic system of equations for C1 and C2 :
ε11 (λ)C1 + ε12 (λ)C2 = 0,
(15.1.1.10)
ε21 (λ)C1 + ε22 (λ)C2 = 0,

ej )′x + ki ϕ
where εij (λ) = si (ϕ ej x=x . For system (15.1.1.10) to have nontrivial solutions,
i
its determinant must be zero; we have
ε11 (λ)ε22 (λ) − ε12 (λ)ε21 (λ) = 0. (15.1.1.11)
By solving the transcendental equation (15.1.1.11) for λ, one obtains the eigenvalues λ =
λn , where n = 1, 2, . . . For these values of λ, Eq. (15.1.1.6) has nontrivial solutions,
ϕn (x) = ε12 (λn )ϕ
e1 (x, λn ) − ε11 (λn )ϕ
e2 (x, λn ), (15.1.1.12)
which are called eigenfunctions (these functions are defined up to a constant factor).
To facilitate the subsequent analysis, we represent Eq. (15.1.1.6) in the form
[p(x)ϕ′x ]′x + [λρ(x) − q(x)]ϕ = 0, (15.1.1.13)
where
Z
b(x) c(x) 1
p(x) = exp dx , q(x) = − p(x), ρ(x) = p(x). (15.1.1.14)
a(x) a(x) a(x)
It follows from the adopted assumptions that p(x), p′x (x), q(x), and ρ(x) are continuous
functions, with p(x) > 0 and ρ(x) > 0.
The eigenvalue problem (15.1.1.13), (15.1.1.8) is known to possess the following prop-
erties:
1. All eigenvalues λ1 , λ2 , . . . are real, and λn → ∞ as n → ∞; consequently, the
number of negative eigenvalues is finite.
2. The system of eigenfunctions ϕ1 (x), ϕ2 (x), . . . is orthogonal on the interval
x1 ≤ x ≤ x2 with weight ρ(x); i.e.,
Z x2
ρ(x)ϕn (x)ϕm (x) dx = 0 for n 6= m. (15.1.1.15)
x1
3. If
q(x) ≥ 0, s1 k1 ≤ 0, s2 k2 ≥ 0, (15.1.1.16)
there are no negative eigenvalues. If q ≡ 0 and k1 = k2 = 0, the least eigenvalue is
λ1 = 0 and the corresponding eigenfunction is ϕ1 = const. Otherwise, all eigenvalues are
positive, provided that conditions (15.1.1.16) are satisfied; the first inequality in (15.1.1.16)
is satisfied if c(x) ≤ 0.
Section 3.8.9 (see the text after the Sturm–Liouville problem (5)) presents some esti-
mates for the eigenvalues λn and eigenfunctions ϕn (x).
◆ The procedure for constructing solutions of nonstationary boundary value problems is
further different for parabolic and hyperbolic equations; see Sections 15.1.2 and 15.1.3
below for results (elliptic and higher-order equations are treated in Sections 15.1.4 and
15.1.5).
15.1.2 Problems for Parabolic Equations: Final Stage of Solution

◮ Series solutions of boundary value problems for parabolic equations.
Consider the problem for the parabolic equation
∂w ∂2w ∂w
= a(x) 2 + b(x) + c(x) + γ(t) w (15.1.2.1)
∂t ∂x ∂x
(this equation is obtained from (15.1.1.1) in the case α(t) ≡ 0 and β(t) = 1) with homoge-
neous linear boundary conditions (15.1.1.2) and the initial condition (15.1.1.3).
First, one seeks particular solutions of Eq. (15.1.2.1) in the product form (15.1.1.5),
where the function ϕ(x) is obtained by solving an eigenvalue problem for the ordinary
differential equation (15.1.1.6) with the boundary conditions (15.1.1.8). The solution of
Eq. (15.1.1.7) with α(t) ≡ 0 and β(t) = 1 corresponding to the eigenvalues λ = λn and
satisfying the normalizing conditions ψn (0) = 1 has the form
Z t
ψn (t) = exp −λn t + γ(ξ) dξ . (15.1.2.2)
0
Then the solution of the nonstationary boundary value problem (15.1.2.1), (15.1.1.2),
(15.1.1.3) is sought in the series form
∞
X
w(x, t) = An ϕn (x) ψn (t), (15.1.2.3)
n=1
where the An are arbitrary constants and the functions wn (x, t) = ϕn (x) ψn (t) are partic-
ular solutions of Eq. (15.1.2.1) satisfying the boundary conditions (15.1.1.2). By the linear
superposition principle, the series (15.1.2.3) is a solution of the original partial differential
equation as well and satisfies the boundary conditions.
To determine the coefficients An , we substitute the series (15.1.2.3) into the initial
condition (15.1.1.3), thus obtaining
∞
X
An ϕn (x) = f0 (x).
n=1
Multiplying this equation by ρ(x)ϕn (x), where the weight function ρ(x) is defined in
(15.1.1.14), integrating the resulting relation with respect to x over the interval x1 ≤ x ≤ x2 ,
and taking into account the properties (15.1.1.15), we find
Z x2 Z x2
1 2
An = ρ(x)ϕ n (x)f 0 (x) dx, kϕ n k = ρ(x)ϕ2n (x) dx. (15.1.2.4)
kϕn k2 x1 x1
Relations (15.1.2.3), (15.1.2.2), (15.1.2.4), and (15.1.1.12) give a formal solution of the
nonstationary boundary value problem (15.1.2.1), (15.1.1.2), (15.1.1.3).
Example 15.1. Consider the first (Dirichlet) boundary value problem on the interval 0 ≤ x ≤ l
for the heat equation
∂w ∂2w
= (15.1.2.5)
∂t ∂x2
with the general initial condition (15.1.1.3) and the homogeneous boundary conditions
w=0 at x = 0, w = 0 at x = l. (15.1.2.6)
The function ψ(t) in the particular solution (15.1.1.5) is found from (15.1.2.2), where γ(t) = 0:
ψn (t) = exp(−λn t). (15.1.2.7)
The functions ϕn (x) are determined by solving the eigenvalue problem (15.1.1.6), (15.1.1.8) with
a(x) = 1, b(x) = c(x) = 0, s1 = s2 = 0, k1 = k2 = 1, x1 = 0, and x2 = l:
ϕ′′xx + λϕ = 0; ϕ = 0 at x = 0, ϕ = 0 at x = l.
So we obtain the eigenfunctions and eigenvalues:

2
nπx nπ
ϕn (x) = sin , λn = , n = 1, 2, . . . (15.1.2.8)
l l
The solution of problem (15.1.2.5)–(15.1.2.6), (15.1.1.3) is given by formulas (15.1.2.3) and

(15.1.2.4). Taking into account the fact that kϕn k2 = l/2, we obtain
∞
X Z l
nπx n2 π 2 t 2 nπξ
w(x, t) = An sin exp − 2 , An = f0 (ξ) sin dξ. (15.1.2.9)
n=1
l l l 0 l
If the function f0 (x) is twice continuously differentiable and the compatibility conditions (see
below) are satisfied, then the series (15.1.2.9) is convergent and admits termwise differentiation,
once with respect to t and twice with respect to x. In this case, formula (15.1.2.9) gives the clas-
sical smooth solution of problem (15.1.2.5)–(15.1.2.6), (15.1.1.3). [If f0 (x) is not as smooth as
indicated or if the compatibility conditions are not met, then the series (15.1.2.9) may converge to a
discontinuous function, thus giving only a generalized solution.]
Remark 15.3. For the solution of linear nonhomogeneous parabolic equations with nonhomo-
geneous boundary conditions, see Section 17.1.
◮ Conditions of compatibility of initial and boundary conditions.
Suppose that the function w has a continuous derivative with respect to t and two con-
tinuous derivatives with respect to x and is a solution of problem (15.1.2.1), (15.1.1.2),
(15.1.1.3). Then the boundary conditions (15.1.1.2) and the initial condition (15.1.1.3)
must be consistent; namely, the following compatibility conditions must hold:
[s1 f0′ + k1 f0 ]x=x1 = 0, [s2 f0′ + k2 f0 ]x=x2 = 0. (15.1.2.10)
If s1 = 0 or s2 = 0, then the additional compatibility conditions
[a(x)f0′′ + b(x)f0′ ]x=x1 = 0 if s1 = 0,

(15.1.2.11)
[a(x)f0′′ + b(x)f0′ ]x=x2 = 0 if s2 = 0
should hold as well; the primes denote derivatives with respect to x.

15.1.3 Problems for Hyperbolic Equations: Final Stage of Solution

◮ Series solution of boundary value problems for hyperbolic equations.
For hyperbolic equations, the solution of the boundary value problem (15.1.1.1)–(15.1.1.4)
is sought in the series form
∞
X
w(x, t) = ϕn (x) An ψn1 (t) + Bn ψn2 (t) . (15.1.3.1)
n=1
Here An and Bn are arbitrary constants. The functions ψn1 (t) and ψn2 (t) are particular
solutions of the linear equation (15.1.1.7) for ψ (with λ = λn ) that satisfy the conditions
′ ′
ψn1 (0) = 1, ψn1 (0) = 0; ψn2 (0) = 0, ψn2 (0) = 1. (15.1.3.2)
The functions ϕn (x) and λn are determined by solving the eigenvalue problem (15.1.1.6),
(15.1.1.8).
Substituting solution (15.1.3.1) into the initial conditions (15.1.1.3)–(15.1.1.4) yields
∞
X ∞
X
An ϕn (x) = f0 (x), Bn ϕn (x) = f1 (x).
n=1 n=1
Multiplying these equations by ρ(x)ϕn (x), where the weight function ρ(x) is defined
in (15.1.1.14), then integrating the resulting relations with respect to x on the interval
x1 ≤ x ≤ x2 , and taking into account properties (15.1.1.15), we obtain the coefficients
of series (15.1.3.1) in the form
Z x2
1
An = ρ(x)ϕn (x)f0 (x) dx,
kϕn k2 x1
Z x2 (15.1.3.3)
1
Bn = ρ(x)ϕ n (x)f 1 (x) dx.
kϕn k2 x1
The quantity kϕn k is defined in (15.1.2.4).

Relations (15.1.3.1), (15.1.1.12), and (15.1.3.3) give a formal solution of the nonsta-
tionary boundary value problem (15.1.1.1)–(15.1.1.4) for α(t) > 0.
Example 15.2. Consider a mixed boundary value problem on the interval 0 ≤ x ≤ l for the wave
equation
∂2w ∂2w
= (15.1.3.4)
∂t2 ∂x2
with the general initial conditions (15.1.1.3)–(15.1.1.4) and the homogeneous boundary conditions
w=0 at x = 0, ∂x w = 0 at x = l. (15.1.3.5)
The functions ψn1 (t) and ψn2 (t) are determined by the linear equation [see (15.1.1.7) with
α(t) = 1, β(t) = γ(t) = 0, and λ = λn ]
′′
ψtt + λψ = 0
with the initial conditions (15.1.3.2). We find

p 1 p
ψn1 (t) = cos λn t , ψn2 (t) = √ sin λn t . (15.1.3.6)
λn
The functions ϕn (x) are determined by solving the eigenvalue problem (15.1.1.6), (15.1.1.8) with
a(x) = 1, b(x) = c(x) = 0, s1 = k2 = 0, s2 = k1 = 1, x1 = 0, and x2 = l:
ϕ′′xx + λϕ = 0; ϕ = 0 at x = 0, ϕ′x = 0 at x = l.
So we obtain the eigenfunctions and eigenvalues:
p π(2n − 1)
ϕn (x) = sin(µn x), µn = λn = , n = 1, 2, . . . (15.1.3.7)
2l
The solution of problem (15.1.3.4)–(15.1.3.5), (15.1.1.3)–(15.1.1.4) is given by formulas
(15.1.3.1) and (15.1.3.3). Taking into account the fact that kϕn k2 = l/2, we have
X∞ h i
1 π(2n − 1)
w(x, t) = An cos(µn t) + Bn sin(µn t) sin(µn x), µn = ,
n=1
µ n 2l
Z Z (15.1.3.8)
2 l 2 l
An = f0 (x) sin(µn x) dx, Bn = f1 (x) sin(µn x) dx.
l 0 l 0
If f0 (x) and f1 (x) have three and two continuous derivatives, respectively, and the compati-
bility conditions are met (see below), then the series (15.1.3.8) is convergent and admits double
termwise differentiation. In this case, formula (15.1.3.8) gives the classical smooth solution of
problem (15.1.3.4)–(15.1.3.5), (15.1.1.3)–(15.1.1.4).
Remark 15.4. For the solution of linear nonhomogeneous hyperbolic equations with nonhomo-
geneous boundary conditions, see Section 17.2.
◮ Conditions of compatibility of initial and boundary conditions.

Suppose w is a twice continuously differentiable solution of problem (15.1.1.1)–(15.1.1.4).
Then conditions (15.1.2.10) and (15.1.2.11) must hold. In addition, the following condi-
tions of compatibility of the boundary conditions (15.1.1.2) and initial condition (15.1.1.4)
must be satisfied:
[s1 f1′ + k1 f1 ]x=x1 = 0, [s2 f1′ + k2 f1 ]x=x2 = 0.
15.1.4 Solution of Boundary Value Problems for Elliptic Equations

◮ Solution of a special problem for elliptic equations.
Now consider a boundary value problem for the elliptic equation
∂2w ∂2w ∂w ∂w
a(x) 2
+ α(y) 2
+ b(x) + β(y) + c(x) + γ(y) w = 0 (15.1.4.1)
∂x ∂y ∂x ∂y
with homogeneous boundary conditions (15.1.1.2) in x and the following mixed (homoge-
neous and nonhomogeneous) boundary conditions in y:
σ1 ∂y w + ν1 w = 0 at y = y1 ,
(15.1.4.2)
σ2 ∂y w + ν2 w = f (x) at y = y2 .
We assume that the coefficients of Eq. (15.1.4.1) and of the boundary conditions (15.1.1.2)
and (15.1.4.2) meet the following requirements:
a(x), b(x), c(x), α(y), β(y), and γ(t) are continuous functions,
a(x) > 0, α(y) > 0, |s1 | + |k1 | > 0, |s2 | + |k2 | > 0, |σ1 | + |ν1 | > 0, |σ2 | + |ν2 | > 0.
The approach is based on searching for particular solutions of Eq. (15.1.4.1) in the product
form
w(x, y) = ϕ(x) ψ(y). (15.1.4.3)
As before, we first arrive at the eigenvalue problem (15.1.1.6), (15.1.1.8) for the function
ϕ = ϕ(x); the solution procedure is detailed in Section 15.1.1. Further on, we assume the
λn and ϕn (x) have been found. The functions ψn = ψn (y) are determined by solving the
linear ordinary differential equation
′′
α(y)ψyy + β(y)ψy′ + [γ(y) − λn ]ψ = 0 (15.1.4.4)
subject to the homogeneous boundary condition
σ1 ∂y ψ + ν1 ψ = 0 at y = y1 , (15.1.4.5)
which is a consequence of the first condition in (15.1.4.2). The functions ψn are determined
up to a constant factor.
Using the linear superposition principle, we seek the solution of the boundary value
problem (15.1.4.1), (15.1.4.2), (15.1.1.2) in the series form
∞
X
w(x, y) = An ϕn (x)ψn (y), (15.1.4.6)
n=1
where An are arbitrary constants. By construction, the series (15.1.4.6) will satisfy equa-
tion (15.1.4.1) with the boundary conditions (15.1.1.2) and the first boundary condition
in (15.1.4.2). To find the series coefficients An , we substitute (15.1.4.6) into the second
boundary condition (15.1.4.2) to obtain
∞
X
dψn
An Bn ϕn (x) = f (x), Bn = σ2 + ν2 ψn (y2 ). (15.1.4.7)
n=1
dy y=y2
Further, we follow the same procedure as in Section 15.1.2. Specifically, multiplying equa-
tion (15.1.4.7) by ρ(x)ϕn (x), then integrating the resulting relation with respect to x over
the interval x1 ≤ x ≤ x2 , and taking into account properties (15.1.1.15), we obtain
Z x2 Z x2
1 2
An = ρ(x)ϕn (x)f (x) dx, kϕn k = ρ(x)ϕ2n (x) dx, (15.1.4.8)
Bn kϕn k2 x1 x1
where the weight function ρ(x) is defined in (15.1.1.14).

Example 15.3. Consider the first (Dirichlet) boundary value problem for the Laplace equation
∂2w ∂2w
2
+ =0 (15.1.4.9)
∂x ∂y 2
w = 0 at x = 0, w=0 at x = l1 ;
(15.1.4.10)
w = 0 at y = 0, w = f (x) at y = l2
in a rectangular domain 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 .
Particular solutions of Eq. (15.1.4.9) are sought in the form (15.1.4.3). We have the following
eigenvalue problem for ϕ(x):
ϕ′′xx + λϕ = 0; ϕ = 0 at x = 0, ϕ = 0 at x = l1 .
On solving this problem, we find the eigenfunctions with respective eigenvalues,
p πn
ϕn (x) = sin(µn x), µn = λn = , n = 1, 2, . . . (15.1.4.11)
l1
The functions ψn = ψn (y) are determined by solving the following problem for a linear ordinary
differential equation with homogeneous boundary conditions:
′′
ψyy − λn ψ = 0; ψ = 0 at y = 0. (15.1.4.12)
It is a special case of problem (15.1.4.4)–(15.1.4.5) with α(y) = 1, β(y) = γ(y) = 0, σ1 = 0, and
ν1 = 1. The nontrivial solutions of problem (15.1.4.12) are expressed as
p πn
ψn (y) = sinh(µn y), µn = λn = , n = 1, 2, . . . (15.1.4.13)
l1
Using formulas (15.1.4.6), (15.1.4.8), (15.1.4.11), and (15.1.4.13) and taking into account the
relations Bn = ψn (l2 ) = sinh(µn l2 ), ρ(x) = 1, and kϕn k2 = l/2, we find the solution of the original
problem (15.1.4.9)–(15.1.4.10) in the form
X∞ Z l1
2 πn
w(x, y) = An sin(µn x) sinh(µn y), An = f (x) sin(µn x) dx, µn = .
n=1
l 1 sinh(µ n l 2 ) 0 l1
◮ Generalization to the case of nonhomogeneous boundary conditions.

Now consider the linear boundary value problem for the elliptic equation (15.1.4.1) with
general nonhomogeneous boundary conditions
s1 ∂x w + k1 w = f1 (y) at x = x1 , s2 ∂x w + k2 w = f2 (y) at x = x2 ,
σ1 ∂y w + ν1 w = f3 (x) at y = y1 , σ2 ∂y w + ν2 w = f4 (x) at y = y2 .
(15.1.4.14)
The solution of this problem is the sum of solutions of four simpler auxiliary problems
for Eq. (15.1.4.1), each corresponding to three homogeneous and one nonhomogeneous
boundary conditions in (15.1.4.14); see Table 15.1. Each auxiliary problem is solved by
the procedure given in Section 15.1.1, starting from the search for solutions in the form
of a product of functions with distinct arguments (15.1.4.3), determined by Eqs. (15.1.1.6)
and (15.1.4.4). The separation parameter λ is determined by the solution of an eigenvalue
problem with homogeneous boundary conditions; see Table 15.1. The solution of each of
the auxiliary problems is sought in the series form (15.1.4.6).
Remark 15.5. For the solution of linear nonhomogeneous elliptic equations with nonhomoge-
neous boundary conditions, see Section 17.3.
TABLE 15.1
Description of auxiliary problems for Eq. (15.1.4.1) and problems for associated
functions ϕ(x) and ψ(y) that determine particular solutions of the form (15.1.4.3).
The abbreviation HBC stands for “homogeneous boundary condition”
Functions vanishing Eigenvalue problem Another problem with one

Auxiliary
in the boundary with homogeneous homogeneous boundary
problem
conditions (15.1.4.14) boundary conditions condition (for λn found)
f2 (y) = f3 (x) = f4 (x) = 0, functions ψn (y) and values λn functions ϕn (x)
Problem 1
function f1 (y) prescribed to be determined satisfy an HBC at x = x2
f1 (y) = f3 (x) = f4 (x) = 0, functions ψn (y) and values λn functions ϕn (x)
Problem 2
function f2 (y) prescribed to be determined satisfy an HBC at x = x1
f1 (y) = f2 (y) = f4 (x) = 0, functions ϕn (x) and values λn functions ψn (y)
Problem 3
function f3 (x) prescribed to be determined satisfy an HBC at y = y2
f1 (y) = f2 (y) = f3 (x) = 0, functions ϕn (x) and values λn functions ψn (y)
Problem 4
function f4 (x) prescribed to be determined satisfy an HBC at y = y1
15.1.5 Solution of Boundary Value Problems for Higher-Order

Equations
Separation of variables can also be used to solve boundary value problems described by
linear higher-order PDEs. The main stages in the application of the method are the same as
for second-order PDEs. Let us illustrate this by specific examples.
Example 15.4. Consider the third-order equation
∂w ∂2w ∂2w
−a 2 −b =0 (0 < t < ∞, 0 < x < l) (15.1.5.1)
∂t ∂x ∂t∂x2
with the general initial condition (15.1.1.3) and the homogeneous boundary conditions (15.1.2.6).
Equation (15.1.5.1) occurs in filtration theory and hydrodynamics and becomes the heat equation
for b = 0.
The approach is based on searching particular solutions of Eq. (15.1.5.1) in the product form
w = ϕ(x)ψ(t), which yields the relation ϕψt′ − aϕ′′xx ψ − bϕ′′xx ψt′ = 0. We divide it by ϕ′′xx ψt′ and
after elementary manipulations obtain
ϕ ψ
= a ′ + b. (15.1.5.2)
ϕ′′xx ψt
The left-hand side of Eq. (15.1.5.2) depends only on x; the right-hand side, only on t. Hence,
for (15.1.5.2) to be true, its left- and right-hand sides must be equal to one and the same constant,
which we denote by −λ. As a result, we obtain the following linear ODEs for the functions ϕ and ψ:
λϕ′′xx + ϕ = 0, (15.1.5.3)
(b + λ)ψt′ + aϕ = 0. (15.1.5.4)
The function ϕ = ϕ(x) satisfying Eq. (15.1.5.3) should also satisfy the homogeneous boundary
conditions ϕ(0) = ϕ(l) = 0 (because we seek particular solutions w = ϕ(x)ψ(t) satisfying the
boundary conditions (15.1.2.6)). The solution of this eigenvalue problem has the form
1 πn
ϕn (x) = sin(βn x), λn = , βn = , n = 1, 2, . . . (15.1.5.5)
βn2 l
The solution of Eq. (15.1.5.4) corresponding to the eigenvalue λ = λn and satisfying the normalizing
condition ψn (0) = 1 has the form

aβn2 t
ψn (t) = exp − . (15.1.5.6)
1 + bβn2
The solution of the original boundary value problem is sought in the series form (15.1.2.3). To
determine the coefficients An , we substitute the series (15.1.2.3) into the initial condition (15.1.1.3).
An argument similar to that given in Example 15.1 permits eventually obtaining the desired solution
in the form
X∞
aβn2 t
n=1
1 + bβn2
Z
2 l πn
An = f0 (x) sin(βn x) dx, βn = .
l 0 l
Example 15.5. Consider the fourth-order equation
∂2w ∂4w
2
+ a2 4 = 0 (0 < t < ∞, 0 < x < l) (15.1.5.7)
∂t ∂x
with the general initial conditions (15.1.1.3)–(15.1.1.4) and the homogeneous boundary conditions
w = ∂xx w = 0 at x = 0, w = ∂xx w = 0 at x = l. (15.1.5.8)
Equation (15.1.5.7) is encountered when studying forced (transverse) vibrations of elastic rods; the
boundary conditions (15.1.5.8) correspond to the case in which both ends of the rod are hinged.
The search for particular solutions of Eq. (15.1.5.7) of the form w = ϕ(x)ψ(t) with subsequent
separation of variables leads to the ODEs
ϕ′′′′
xxxx − λϕ = 0, (15.1.5.9)
′′ 2
ψtt + a λψ = 0, (15.1.5.10)
where λ is the separation constant.

The general solution of Eq. (15.1.5.9) is
ϕ = C1 cos(βx) + C2 sin(βx) + C3 cosh(βx) + C4 sinh(βx), λ = β4, (15.1.5.11)
where C1 , C2 , C3 , and C4 are arbitrary constants. The functions ϕ = ϕ(x) should satisfy the
homogeneous boundary conditions ϕ(0) = ϕ′′ (0) = ϕ(l) = ϕ′′ (l) = 0, which follow from the form
of the particular solution w = ϕ(x)ψ(t) and the boundary conditions (15.1.5.8). The solution of
this eigenvalue problem is determined by formula (15.1.5.11) with C1 = C3 = C4 = 0 and has the
form
πn
ϕn (x) = sin(βn x), λn = βn4 , βn = , n = 1, 2, . . . (15.1.5.12)
l
The solutions of Eq. (15.1.5.10) corresponding to the eigenvalues λ = λn and satisfying the
normalizing conditions (15.1.3.2) have the form
1
ψn1 (t) = cos aβn2 t , ψn2 (t) = 2
sin aβn2 t . (15.1.5.13)
aβn
The solution of the original boundary value problem is sought in the series form (15.1.3.1).
To determine the coefficients An and Bn , we substitute the series (15.1.3.1) into the initial condi-
tions (15.1.1.3)–(15.1.1.4). An argument similar to that given in Section 15.1.3 permits eventually
15.2. Integral Transform Method 1165
obtaining the desired solution in the form

∞
X h 1 i
w(x, t) = sin(βn x) An cos(aβn2 t) + Bn 2 sin(aβn2 t) ,
n=1
aβn
Z l Z l
2 2 πn
An = f0 (x) sin(βn x) dx, Bn = f1 (x) sin(βn x) dx, βn = .
l 0 l 0 l
H. S. Carslaw and J. C. Jaeger (1984), R. Leis (1986), A. A. Dezin (1987), R. Haberman (1987), E. Zaud-
erer (1989), I. Sneddon (1972, 1995), A. N. Tikhonov and A. A. Samarskii (1990), W. A. Strauss (1992),
R. B. Guenther and J. W. Lee (1996), D. Zwillinger (1998), I. Stakgold (2000), C. Constanda (2002), A. D. Po-
lyanin (2002), A. D. Polyanin and A. V. Manzhirov (2007).
15.2 Integral Transform Method

Linear problems of mathematical physics are very often solved using various integral trans-
forms. The Laplace transform and the Fourier transform are in most common use. These
and some other integral transforms are discussed below (see also Chapter 28).
15.2.1 Laplace Transform and Its Application in Mathematical Physics

◮ Laplace and inverse Laplace transforms. Laplace transforms for derivatives.
1◦ . The Laplace transform of an arbitrary (complex-valued) function f (t) of a real vari-
able t (t ≥ 0) is defined by
Z ∞

fe(p) = L f (t) , where L f (t) ≡ e−pt f (t) dt, (15.2.1.1)
0
and p = s + iσ is a complex variable, i2 = −1.

The Laplace transform exists for any continuous or piecewise continuous function sat-
isfying the condition |f (t)| < M eσ0 t with some M > 0 and σ0 ≥ 0. In the following,
σ0 often means the greatest lower bound of the possible values of σ0 in this estimate; this
value is called the growth exponent of the function f (t).
2◦ . Given the transform fe(p), the function f (t) can be found by means of the inverse
Laplace transform
Z c+i∞
1
f (t) = L−1 fe(p) , where L−1 fe(p) ≡ fe(p)ept dp, (15.2.1.2)
2πi c−i∞
and the integration path is parallel to the imaginary axis and lies to the right of all singular-
ities of fe(p), which corresponds to c > σ0 .
The integral in the inversion formula (15.2.1.2) is understood in the sense of the Cauchy
principal value: Z Z
c+i∞ c+iω
fe(p)ept dp = lim fe(p)ept dp.
c−i∞ ω→∞ c−iω
In the domain t < 0, formula (15.2.1.2) gives f (t) ≡ 0.

Formula (15.2.1.2) holds for continuous functions. If f (t) has a (finite) jump disconti-
nuity at a point t = t0 > 0, then the left-hand side of (15.2.1.2) is equal to 12 [f (t0 − 0) +
f (t0 + 0)] at this point (for t0 = 0, the first term in the square brackets must be omitted).
3◦ . Consider the important case in which the transform is a rational function of the form
R(p)
fe(p) = ,
Q(p)
where Q(p) and R(p) are polynomials in the variable p and the degree of Q(p) exceeds
that of R(p).
Assume that the zeros of the denominator are simple; i.e.,
Q(p) ≡ const (p − λ1 )(p − λ2 ) . . . (p − λn ).
Then the inverse transform can be determined by the formula

n
X R(λk )
f (x) = exp(λk x),
Q′ (λk )
k=1
where the primes denote the derivatives.

If Q(p) has multiple zeros, i.e., if
Q(p) ≡ const (p − λ1 )s1 (p − λ2 )s2 . . . (p − λm )sm ,
then
m
X 1 dsk −1
f (x) = lim s −1
(p − λk )sk fe(p)epx .
(sk − 1)! p→s k dp k
k=1
◮ Main properties of the Laplace transform.
The main properties of the correspondence between functions and their Laplace transforms
are gathered in Table 15.2.
The Laplace transforms of some functions are listed in Table 15.3; for more detailed
tables, see Section 28.1 and the list of references at the end of this chapter.
Such tables are convenient to use when solving linear problems for partial differential
equations.
To solve nonstationary boundary value problems, the following Laplace transform for-
mulas for derivatives will be required:

L f ′ (t) = pfe(p) − f (0),
(15.2.1.3)
L f ′′ (t) = p2 fe(p) − pf (0) − f ′ (0),
where f (0) and f ′ (0) are the initial conditions.

TABLE 15.2
Main properties of the Laplace transform
No. Function Laplace transform Operation

1 af1 (t) + bf2 (t) afe1 (p) + bfe2 (p) Linearity
2 f (t/a), a > 0 afe(ap) Scaling
f (t − a), Shift of
3
f (ξ) ≡ 0 for ξ < 0 e−ap fe(p) the argument
Differentiation
4 tn f (t); n = 1, 2, . . . (−1)n fep(n) (p)
Z ∞ of the transform
1 Integration
5 f (t) fe(q) dq
t p of the transform
at Shift in
6 e f (t) fe(p − a) the complex plane
7 ft′ (t) pfe(p) − f (+0) Differentiation
Pn
(k−1)
8 ft
(n)
(t) pn fe(p) − pn−k ft (+0) Differentiation
k=1
d hm Pn
(k−1)
i
9 tm ft
(n)
(t), m = 1, 2, . . . (−1)m m pn fe(p) − pn−k ft (+0) Differentiation
dp k=1
m
dn m m n d
10 t f (t) , m ≥ n (−1) p fe(p) Differentiation
dt n dpm
Z t
fe(p)
11 f (z) dz Integration
0 p
Z t
12 f1 (z)f2 (t − z) dz fe1 (p)fe2 (p) Convolution
0
TABLE 15.3
Laplace transforms of some functions
No. Function, f (t) Laplace transform, fe(p) Remarks

1 1 1/p
n!
2 tn n = 1, 2, . . .
pn+1
3 ta Γ(a + 1)p−a−1 a > −1
−at −1
4 e (p + a)
5 a −bt
t e Γ(a + 1)(p + b)−a−1 a > −1
a
6 sinh(at)
p 2 − a2
p
7 cosh(at)
p 2 − a2
1 C = 0.5772 . . .
8 ln t − (ln p + C)
p is the Euler constant
a
9 sin(at)
p 2 + a2
p
10 cos(at)
p 2 + a2

a 1 √
11 erfc √ exp −a p a≥0
2 t p
1
12 J0 (at) p J0 (t) is the Bessel function
p 2 + a2
Original problem = {partial differential equation for w = w(x, t),

initial conditions, and boundary conditions
Application of the Laplace transform (13.6.1.1)

with respect to t
Problem for transform = {ordinary differential equation

~ ~
for w = w(x, p), boundary conditions
Solution of the ordinary differential equation

~=w
Finding the transform w ~ (x, p)
Application of the inverse

Laplace transform (13.6.1.2)
Obtaining solution to the original problem: w = w(x, t)
Figure 15.2: Solution procedure for linear boundary value problems using the Laplace
transform.
◮ Solution procedure for linear problems using the Laplace transform.

Figure 15.2 shows schematically how one can utilize the Laplace transforms to solve
boundary value problems for linear parabolic or hyperbolic equations with two indepen-
dent variables in the case where the equation coefficients are independent of t (the same
procedure can be applied for solving linear problems characterized by higher-order equa-
tions). Here and henceforth, the short notation w(x, e p) = L{w(x, t)} will be used; the
arguments of w e may be omitted.
It is significant that with the Laplace transform, the original problem for a partial differ-
ential equation is reduced to a simpler problem for an ordinary differential equation with
parameter p; the derivatives with respect to t are replaced by appropriate algebraic expres-
sions taking into account the initial conditions; see formulas (15.2.1.3).
◮ Solving linear problems for parabolic equations with the Laplace transform.
Consider a linear nonstationary boundary value problem for the parabolic equation
∂w ∂2w ∂w
= a(x) 2 + b(x) + c(x)w + Φ(x, t) (15.2.1.4)
∂t ∂x ∂x
with the initial condition (15.1.1.3) and the general nonhomogeneous boundary conditions
s1 ∂x w + k1 w = g1 (t) at x = x1 ,
(15.2.1.5)
s2 ∂x w + k2 w = g2 (t) at x = x2 .
The application of the Laplace transform results in the problem defined by the ordinary
differential equation in x (p is treated as a parameter)
∂2w
e ∂w
e e p) = 0
a(x) 2
+ b(x) + [c(x) − p]w
e + f0 (x) + Φ(x, (15.2.1.6)
∂x ∂x
s1 ∂x w
e + k1 w
e = ge1 (p) at x = x1 ,
(15.2.1.7)
s2 ∂x w
e + k2 w
e = ge2 (p) at x = x2 .

e p) = L Φ(x, t) and e
Notation employed: Φ(x, gn (p) = L gn (t) (n = 1, 2). On solving
problem (15.2.1.6)–(15.2.1.7), one should apply the inverse Laplace transform (15.2.1.2)
to the resulting solution w
e = w(x,
e p) to obtain the solution w = w(x, t) of the original
problem.
Example 15.6. Consider the first boundary value problem for the heat equation,
∂w ∂ 2w
= (x > 0, t > 0),
∂t ∂x2
w=0 at t = 0 (initial condition),
w→0 at x → ∞ (boundary condition).
We apply the Laplace transform with respect to t. Let us multiply the equation, the initial
condition, and the boundary conditions by e−pt and then integrate with respect to t from zero to
infinity. Taking into account the relations
L{∂t w} = pwe − w|t=0 = pw

e (the first property (15.2.1.3) and the initial condition are used),
L{w0 } = w0 L{1} = w0 /p (property 1 in Table 15.2 and the relation L{1} = 1/p
are used; see property 1 in Table 15.3),
we arrive at the following problem for a second-order linear ordinary differential equation with
parameter p:
′′
w
exx − pw
e = 0, (15.2.1.8)
w
e = w0 /p at x = 0 (boundary condition),
(15.2.1.9)
w
e→0 at x → ∞ (boundary condition).
√ √
e = A1 (p)e−x p + A2 (p)ex p . Using the
Integrating the equation yields the general solution w
boundary conditions, we determine the constants, A1 (p) = w0 /p and A2 (p) = 0. Thus, we have
w0 −x√p
w
e= e .
p
Let us apply the inverse Laplace transform to both sides of this relation. We refer to Table 15.3,
row 11 (where a must be replaced by x), to find the inverse transform of the right-hand side. Finally,
we obtain the solution of the original problem in the form

x
w = w0 erfc √ .
2 t
◮ Solving linear problems for hyperbolic equations by the Laplace transform.

Consider a linear nonstationary boundary value problem defined by the hyperbolic equation
∂2w ∂w ∂2w ∂w
2
+ ϕ(x) = a(x) 2
+ b(x) + c(x)w + Φ(x, t) (15.2.1.10)
∂t ∂t ∂x ∂x
with the initial conditions (15.1.1.3), (15.1.1.4) and general boundary conditions (15.2.1.5).
The application of the Laplace transform results in the problem defined by the ordinary
differential equation for x (p is treated as a parameter)
∂2we ∂we e p) = 0
a(x) 2
+ b(x) + [c(x) − p2 − pϕ(x)]w e + f0 (x)[p + ϕ(x)] + f1 (x) + Φ(x,
∂x ∂x
(15.2.1.11)
with the boundary conditions (15.2.1.7). On solving this problem, one should apply the
inverse Laplace transform to the resulting solution w
e = w(x,
e p).
15.2.2 Fourier Transform and Its Application in Mathematical Physics

◮ Standard and alternative forms of the Fourier transform.
1◦ . The Fourier transform is defined as follows:
Z ∞
1
fe(u) = F f (x) , where F f (x) ≡ √ f (x)e−iux dx, (15.2.2.1)
2π −∞
where i2 = −1. This relation is meaningful for any function f (x) absolutely integrable on
the interval (−∞, ∞).
Given fe(u), the function f (x) can be found by the inverse Fourier transform
Z ∞

−1 e

−1 e
1
f (x) = F f (u) , where F f (u) ≡ √ fe(u)eiux du, (15.2.2.2)
2π −∞
where the integral is understood in the sense of the Cauchy principal value.
Formula (15.2.2.2) holds for continuous functions. If f (x) has a (finite)jump discon-
tinuity at a point x = x0 , then the left-hand side of (15.2.2.2) is equal to 12 f (x0 − 0) +
f (x0 + 0) at this point.
Remark 15.6. It is natural to consider the Fourier transform in the class of complex-valued
functions of a real argument, because in the general case fe(u) takes complex values even for a real
function f (x).
2◦ . The main properties of the correspondence between functions and their Fourier trans-
forms are gathered in Table 15.4. The function f (x) and its derivatives are assumed to
vanish sufficiently rapidly as x → ∞. (For example, for properties 5 and 6 to hold one
needs the function f (x) and its derivatives to decay faster than |x|−1 ).
3◦ . Sometimes the alternative Fourier transform is used (and called merely the Fourier
transform), which corresponds to the renaming e−iux ⇋ eiux on the right-hand sides
of (15.2.2.1) and (15.2.2.2).
TABLE 15.4
Main properties of the Fourier transform
No. Function Fourier transform Operation
1 af1 (x) + bf2 (x) afe1 (u) + bfe2 (u) Linearity
2 f (x/a), a > 0 afe(au) Scaling

−iau e
3 f (x − a) e f (u) Shift
Differentiation
4 xn f (x); n = 1, 2, . . . in feu(n) (u)
of the transform
′′
5 fxx (x) −u2 fe(u) Differentiation
6 fx(n) (x) (iu)n fe(u) Differentiation
Z ∞
7 f1 (ξ)f2 (x − ξ) dξ fe1 (u)fe2 (u) Convolution
−∞
◮ Asymmetric form of the Fourier transform.

Often it is more convenient to define the Fourier transform by
Z ∞

ˇ
f (u) = F f (x) , where F f (x) ≡ f (x)e−iux dx. (15.2.2.3)
−∞
In this case, the Fourier inversion formula reads

Z ∞
1
f (x) = F −1
fˇ(u) , where F −1
fˇ(u) ≡ fˇ(u)eiux du. (15.2.2.4)
2π −∞
◮ n-dimensional Fourier transform.

1◦ . The Fourier transform admits n-dimensional generalization:
Z
1
fe(u) = f (x)e−i(u·x) dVx , (u · x) = u1 x1 + · · · + un xn , (15.2.2.5)
(2π)n/2 Rn
where f (x) = f (x1 , . . . , xn ), fe(u) = fe(u1 , . . . , un ), and dVx = dx1 . . . dxn .
The corresponding inversion formula is
Z
1
f (x) = n/2
fe(u)ei(u·x) dVu , dVu = du1 . . . dun . (15.2.2.6)
(2π) R n
2◦ . Often it is more convenient to define the Fourier transform by the asymmetric form of
the Fourier transform Z
ˇ
f (u) = f (x)e−i(u·x) dVx . (15.2.2.7)
Rn
In this case, the Fourier inversion formula has the form
Z
1
f (x) = fˇ(u)ei(u·x) dVu . (15.2.2.8)
(2π)n Rn
The Fourier transforms (15.2.2.5) and (15.2.2.7) are frequently used in the theory of
linear partial differential equations with constant coefficients (x ∈ Rn ).
◮ Solving linear problems of mathematical physics by the Fourier transform.

The Fourier transform is usually employed to solve boundary value problems for linear
partial differential equations whose coefficients are independent of the space variable x,
−∞ < x < ∞.
The scheme for solving linear boundary value problems with the help of the Fourier
transform is similar to that used in solving problems with the help of the Laplace transform.
With the Fourier transform, the derivatives with respect to x in the equation are replaced
by appropriate algebraic expressions; see Property 5 or 6 in Table 15.4. In the case of
two independent variables, the problem for a partial differential equation is reduced to a
simpler problem for an ordinary differential equation with parameter u. On solving the
latter problem, one determines the transform. After that, by applying the inverse Fourier
transform, one obtains the solution of the original boundary value problem.
Example 15.7. Consider the following Cauchy problem for the heat equation:
∂w ∂2w
= (−∞ < x < ∞),
∂t ∂x2
w = f (x) at t = 0 (initial condition).
We apply the Fourier transform with respect to the space variable x. Setting w e = F{w(x, t)} and
taking into account the relation F{∂xx w} = −u2 w e (see Property 5 in Table 15.4), we arrive at the
following problem for a linear first-order ordinary differential equation in t with parameter u:
et′ + u2 w
w e = 0,
e = fe(u) at t = 0,
w
where fe(u) is defined by (15.2.2.1). On solving this problem for the transform w,
e we find
2
e = fe(u)e−u t .
w
Let us apply the inversion formula to both sides of this equation. After some calculations, we obtain
the solution of the original problem in the form
Z ∞ Z ∞ Z ∞
1 e −u2 t iux 1 −iuξ 2
w= √ f (u)e e du = f (ξ)e dξ e−u t+iux du
2π −∞ 2π −∞ −∞
Z ∞ Z ∞ Z ∞
1 2 1 (x − ξ)2
= f (ξ) dξ e−u t+iu(x−ξ) du = √ f (ξ) exp − dξ.
2π −∞ −∞ 2πt −∞ 4t
Z ∞ √ 2
2 2
π b
At the last stage, we have used the relation exp −a u + bu du = exp .
−∞ |a| 4a2
Example 15.8. Consider the Cauchy problem for the equation of transverse vibrations of elastic
rods
∂2w 4
2∂ w
+ a =0 (15.2.2.9)
∂t2 ∂x4
w = f (x) at t = 0, ∂t w = 0 at t = 0. (15.2.2.10)
To solve problem (15.2.2.9)–(15.2.2.10), we use the asymmetric form of the Fourier trans-
form (15.2.2.3) with respect to the space variable x. By setting w̌ = F {w(x, t)} and by taking
into account the relation F {∂xxxxw} = u4 w̌ (see Property 6 with n = 4 in Table 15.4), we arrive at
the following problem for a linear second-order ordinary differential equation in t with parameter u:
w̌′′tt + a2 u4 w̌ = 0,
(15.2.2.11)
w̌ = F (u) at t = 0, w̌′t = 0 at t = 0,
where F (u) = F {f (x)}. The solution of problem (15.2.2.11) has the form
w̌ = F (u) cos(au2 t). (15.2.2.12)
We apply the inverse Fourier transform (15.2.2.4) to (15.2.2.12) and obtain, after easy transforma-
tions,
Z ∞
1
w= F (u) cos(au2 t)eiux du
2π −∞
Z ∞ Z ∞
1
= f (ξ)e−iuξ dξ cos(au2 t)eiux du
2π −∞ −∞
Z ∞ Z ∞
1
= f (ξ) cos(au2 t)eiu(x−ξ) du dξ
2π −∞ −∞
Z Z ∞
1 ∞
= f (ξ) cos(au2 t) cos[u(x − ξ)] du dξ.
π −∞ 0
The inner integral can be evaluated (see the tables of definite integrals in Section 27.2.5),
Z ∞ r
2 π (x − ξ)2 (x − ξ)2
cos(au t) cos[u(x − ξ)] du = cos + sin .
0 8at 4at 4at
As a result, we find the solution of the original problem (the Boussinesq solution) in the form
r Z ∞
1 (x − ξ)2 (x − ξ)2
w= f (ξ) cos + sin dξ
8πat −∞ 4at 4at
r Z ∞ 2 2
1 η η
= f (x − η) cos + sin dη
8πat −∞ 4at 4at
Z ∞ √
1
= √ f (x − 2 at ζ) cos(ζ 2 ) + sin(ζ 2 ) dζ.
2π −∞
15.2.3 Fourier Sine and Cosine Transforms

◮ Fourier sine transform.
1◦ . Let a function f (x) be integrable on the half-line 0 ≤ x < ∞. The Fourier sine
transform is defined by
r Z
2 ∞
fes (u) = f (x) sin(xu) dx, 0 < u < ∞. (15.2.3.1)
π 0
For given fes (u), the function f (x) can be found by means of the inverse Fourier sine
transform
r Z
2 ∞e
f (x) = f s (u) sin(xu) du, 0 < x < ∞. (15.2.3.2)
π 0

The Fourier sine transform (15.2.3.1) is briefly denoted by fes (u) = Fs f (x) .
It follows from formula (15.2.3.2) that the Fourier sine transform has the property
F2s = 1.
Parseval’s relation for the Fourier sine transform:
Z ∞ Z ∞

Fs f (x) Fs g(x) du = f (x)g(x) dx.
0 0
There are tables of Fourier sine transforms (see Section 28.4 and the references listed
at the end of the current chapter).
2◦ . Often it is more convenient to apply the asymmetric form of the Fourier sine transform
defined by the following two formulas:
Z ∞ Z
2 ∞ ˇ
fˇs (u) = f (x) sin(xu) dx, f (x) = f s (u) sin(xu) du. (15.2.3.3)
0 π 0
The asymmetric
form of Fourier sine transform (15.2.3.3) is briefly denoted by fˇs (u) =
F̌s f (x) .
Some properties of the Fourier sine transform:

F̌s f ′′ (x) = −u2 F̌s f (x) + uf (0),
d2n (15.2.3.4)
F̌s x2n f (x) = (−1)n 2n F̌s f (x) , n = 1, 2, . . .
du
Here f (x) and their derivatives are assumed to vanish sufficiently rapidly (for example,
exponentially) as x → ∞.
◮ Fourier cosine transform.

1◦ . Let a function f (x) be integrable on the half-line 0 ≤ x < ∞. The Fourier cosine
transform is defined by
r Z
2 ∞
fec (u) = f (x) cos(xu) dx, 0 < u < ∞. (15.2.3.5)
π 0
For given fec (u), the function can be found by means of the Fourier cosine inversion
formula r Z ∞
2
f (x) = fec (u) cos(xu) du, 0 < x < ∞. (15.2.3.6)
π 0

The Fourier cosine transform (15.2.3.5) is denoted for brevity by fec (u) = Fc f (x) .
It follows from formula (15.2.3.6) that the Fourier cosine transform has the property
F2c = 1.
Parseval’s relation for the Fourier cosine transform:
Z ∞ Z ∞

Fc f (x) Fc g(x) du = f (x)g(x) dx.
0 0
There are tables of the Fourier cosine transform (see Section 28.3 and the references
listed at the end of the current chapter).
2◦ . Often the asymmetric form of the Fourier cosine transform is applied, which is given
by the pair of formulas
Z ∞ Z
ˇ 2 ∞ ˇ
f c (u) = f (x) cos(xu) dx, f (x) = f c (u) cos(xu) du. (15.2.3.7)
0 π 0
The asymmetric
form of Fourier cosine transform (15.2.3.7) is briefly denoted by fˇc (u) =
F̌c f (x) .
Some properties of the Fourier cosine transform:

F̌c f ′′ (x) = −u2 F̌c f (x) − f ′ (0),
d2n (15.2.3.8)
F̌c x2n f (x) = (−1)n 2n F̌c f (x) , n = 1, 2, . . .
du
Here f (x) and their derivatives are assumed to vanish sufficiently rapidly (for example,
exponentially) as x → ∞.
◮ Solving mathematical physics problems by the Fourier sine and cosine transforms.
The Fourier sine and cosine transforms are usually employed to solve boundary value prob-
lems for linear partial differential equations whose coefficients are independent of the space
variable x, 0 ≤ x < ∞.
If a boundary condition of the first kind is given at x = 0, then one uses the Fourier sine
transform. If a boundary condition of the second kind is given at x = 0, then one uses the
Fourier cosine transform. Let us illustrate the solution construction procedure by specific
examples.
Example 15.9. Consider the heat equation
∂w ∂2w
−a 2 =0 (0 < t < ∞, 0 < x < ∞) (15.2.3.9)
∂t ∂x
with the initial and boundary conditions
w = 0 at t = 0, w = f (t) at x = 0, w→0 as x → ∞. (15.2.3.10)
It is assumed that a > 0.

To solve problem (15.2.3.9)–(15.2.3.10), we use the asymmetric form of the Fourier sine trans-
form (15.2.3.3). With regard to the first relation in (15.2.3.4) and the boundary conditions (15.2.3.10)
(we also assume that ∂x w → 0 as x → ∞), we find the Fourier sine transform of the second derivative
of the unknown function,
Z ∞
∂xx w sin(xu) dx = −u2 W (t, u) + uf (t), (15.2.3.11)
0

where W (t, u) = F̌s w . As a result, the second-order PDE (15.2.3.9) leads to the first-order
Wt′ + au2 W = auf (t) (15.2.3.12)
with the initial condition W = 0 at t = 0. The solution of Eq. (15.2.3.12) with this condition has
the form Z t
W (t, u) = a uf (τ ) exp[−au2 (t − τ )] dτ. (15.2.3.13)
0
By using the inverse transform (15.2.3.3), we obtain the solution of the original problem (15.2.3.9)–
(15.2.3.10),
Z Z t
2 ∞
w(x, t) = sin(xu)W (t, u) du = G(x, t − τ )f (τ ) dτ, (15.2.3.14)
π 0 0
where Z ∞
2a
G(x, t) = u sin(xu) exp(−au2 t) du.
π 0
The integral determining the function G can be computed (see the tables of definite integrals in
Section 27.2.5). As a result, we obtain

x x2
G(x, t) = √ exp − . (15.2.3.15)
2 πa t3/2 4at
Formulas (15.2.3.14)–(15.2.3.15) give the solution of problem (15.2.3.9)–(15.2.3.10).
Example 15.10. Consider the third-order equation
∂w ∂2w ∂2w
−a 2 −b =0 (0 < t < ∞, 0 < x < ∞) (15.2.3.16)
∂t ∂x ∂t∂x2
with the initial and boundary conditions (15.2.3.10). Equation (15.2.3.16) is a generalization of the
heat equation (15.2.3.9) (they coincide for b = 0) and arises in filtration theory and hydrodynam-
ics. In the following, we assume that a > 0, b ≥ 0, and the initial and boundary data satisfy the
consistency condition f (0) = 0.
To solve this problem, just as in the preceding example, we use the asymmetric form of the
Fourier sine transform (15.2.3.3). The transform of the second derivative is given by formula
(15.2.3.11); further, by differentiating this formula with respect to t, we find the Fourier sine trans-
form of the mixed derivative ∂xxt w. As a result, the third-order PDE (15.2.3.16) leads to the first-
order ordinary differential equation
(1 + bu2 )Wt′ + au2 W = u[af (t) + bft′ (t)] (15.2.3.17)
with the initial condition W = 0 at t = 0. The solution of Eq. (15.2.3.17) with this condition has
the form
Z t
u au2 (t − τ )
W (t, u) = 2
exp − [af (τ ) + bf ′ (τ )] dτ. (15.2.3.18)
0 1 + bu 1 + bu2
By using the inverse transform (15.2.3.3), we obtain the solution of the original problem (15.2.3.16),
(15.2.3.10),
Z Z t h i
2 ∞ b
w(x, t) = sin(xu)W (t, u) du = G(x, t − τ ) f (τ ) + f ′ (τ ) dτ,
π 0 0 a
Z (15.2.3.19)
2a ∞ u sin(xu) au2 t
G(x, t) = exp − du.
π 0 1 + bu2 1 + bu2
Example 15.11. Consider the following boundary value problem for the Helmholtz equation:
∂2w ∂2w
2
+ − k2 w = 0 (0 < x < ∞, 0 < y < a), (15.2.3.20)
∂x ∂y 2
∂x w = 0 at x = 0, w→0 as x → ∞, (15.2.3.21)
w = f (x) at y = 0, w=0 at y = a. (15.2.3.22)
To solve this problem, we use the asymmetric form of the Fourier cosine transform (15.2.3.7).
In view of the first relation in (15.2.3.8) and the boundary conditions
(15.2.3.21) (we also assume
that ∂x w → 0 as x → ∞), for the function W (y, u) = F̌c w we obtain the ordinary differential
equation
′′
Wyy − (u2 + k 2 )W = 0 (15.2.3.23)
W = fˇc (u) at y = 0, W =0 at y = a, (15.2.3.24)
where fˇc (u) is the asymmetric form of the Fourier cosine transform for the function f (x).
The general solution of Eq. (15.2.3.23) has the form
p p
W = C1 (u) exp −y u2 + k 2 + C2 (u) exp y u2 + k 2 .
By substituting this expression into the boundary conditions (15.2.3.23), we determine the coeffi-
cients C1 (u) and C2 (u). As a result, after simple transformations we find the solution of prob-
lem (15.2.3.23)–(15.2.3.24),
√
ˇ sinh (a − y) u2 + k 2
W = f c (u) √ . (15.2.3.25)
sinh a u2 + k 2
Now the inverse transform (15.2.3.7) provides the solution of the original problem (15.2.3.23)–
(15.2.3.24):
Z √
2 ∞ ˇ sinh (a − y) u2 + k 2
w(x, y) = f c (u) √ cos(xu) du
π 0 sinh a u2 + k 2
Z Z √
2 ∞ ∞ sinh (a − y) u2 + k 2
= f (ξ) √ cos(xu) cos(ξu) dξ du.
π 0 0 sinh a u2 + k 2
15.2.4 Mellin, Hankel, and Other Integral Transforms

◮ Mellin transform.
1◦ . Suppose that a function f (x) is defined for positive x and satisfies the conditions
Z 1 Z ∞
|f (x)| xσ1 −1 dx < ∞, |f (x)| xσ2 −1 dx < ∞
0 1
for some real numbers σ1 and σ2 , σ1 < σ2 .

The Mellin transform of f (x) is defined by
Z ∞
ˆ
f (s) = f (x)xs−1 dx, (15.2.4.1)
0
where s = σ + iτ is a complex variable (σ1 < σ < σ2 ).

Given fˆ(s), the function f (x) can be found by means of the inverse Mellin transform
Z σ+i∞
1
f (x) = fˆ(s)x−s ds (σ1 < σ < σ2 ), (15.2.4.2)
2πi σ−i∞
where the integration path is parallel to the imaginary axis of the complex plane s and the
integral is understood in the sense of the Cauchy principal value.
Formula (15.2.4.2) holds for continuous functions. If f (x) has a (finite) jump dis-
continuity
at a point x = x0 > 0, then the right-hand side of (15.2.4.2) evaluates to
1
2 f (x 0 − 0) + f (x0 + 0) at this point (for x0 = 0, the first term in the square brack-
ets must be omitted).
2◦ . The main properties of the correspondence between the functions and their Mellin
transforms are gathered in Table 15.5.
TABLE 15.5
Main properties of the Mellin transform
No. Function Mellin transform Operation
1 af1 (x) + bf2 (x) afˆ1 (s) + bfˆ2 (s) Linearity
2 f (ax), a > 0 a −s
fˆ(s) Scaling
Shift of the argument
3 xa f (x) fˆ(s + a)
of the transform
1 ˆ 1

4 f (x2 ) 2
f 2s Squared argument
Inversion of the argument

5 f (1/x) fˆ(−s)
of the transform
1 − s+λ s+λ
6 xλ f axβ , a > 0, β 6= 0 a β fˆ Power law transform
β β
7 fx′ (x) −(s − 1)fˆ(s − 1) Differentiation
8 xfx′ (x) −sfˆ(s) Differentiation

Γ(s) ˆ
9 fx(n) (x) (−1)n f (s − n) Multiple differentiation
Γ(s − n)
d n
10 x f (x) (−1)n sn fˆ(s) Multiple differentiation
dx
Z ∞
11 xα tβ f1 (xt)f2 (t) dt fˆ1 (s + α)fˆ2 (1 − s − α + β) Complicated integration
0
Z ∞ x
12 xα tβ f1 f2 (t) dt fˆ1 (s + α)fˆ2 (s + α + β + 1) Complicated integration
0 t
There are extensive tables of direct and inverse Mellin transforms (see the references
listed at the end of the current chapter), which are useful when solving specific differential
and integral equations.
3◦ . By M{f (x), s} we denote the Mellin transform (15.2.4.1), and by L{f (t), p} we de-
note the Laplace transform (15.2.1.1).
The Mellin transform is related to the Laplace transform by
M{f (x), s} = L{f (ex ), −s} + L{f (e−x ), s}. (15.2.4.3)
Formula (15.2.4.3) permits one to apply the much more common tables of direct and
inverse Laplace transforms.
◮ Hankel transform.
1◦ . The Hankel transform is defined as follows:
Z ∞
e
f ν (u) = xJν (ux)f (x) dx, 0 − 12 and Jν (x) is the Bessel function of the first kind of order ν (see Sec-
tion 30.6).
For given feν (u), the function f (x) can be found by means of the Hankel inversion
formula Z ∞
f (x) = uJν (ux)feν (u) du, 0 < x < ∞. (15.2.4.5)
0
Note that if f (x) = O(xα )as x → 0, where α + ν + 2 > 0, and f (x) = O(xβ ) as
3
x → ∞, where β + 2 < 0, then the integral (15.2.4.4) is convergent.
The inversion formula (15.2.4.5) holds for continuous functions. If f (x) has a (finite)
jump discontinuity at a point x = x0 , then the left-hand side of (15.2.4.5) is equal to
1
2 [f (x0 − 0) + f (x0 + 0)] at this point.
For brevity, we denote the Hankel transform (15.2.4.4) by feν (u) = Hν f (x) .
2◦ . It follows from formula (15.2.4.5) that the Hankel transform has the property Hν2 = 1.
Other properties of the Hankel transform:

1 u u
Hν f (x) = Hν−1 f (x) + Hν+1 f (x) ,
x 2ν 2ν
′ (ν − 1)u (ν + 1)u
Hν f (x) = Hν+1 f (x) − Hν−1 f (x) ,
2ν 2ν
1 ν 2
Hν f ′′ (x) + f ′ (x) − 2 f (x) = −u2 Hν f (x) .
x x
The conditions

lim xν f (x) = 0, lim xν+1 f ′ (x) = 0, lim x1/2 f (x) = 0, lim x1/2 f ′ (x) = 0
x→0 x→0 x→∞ x→∞
are assumed to hold for the last formula.

Parseval’s relation for the Hankel transform:
Z ∞ Z ∞
1
uHν f (x) Hν g(x) du = xf (x)g(x) dx, ν>− .
0 0 2
◮ Summary table of integral transforms.

Table 15.6 summarizes the integral transforms considered above and also lists some other
integral transforms; for the constraints imposed on the functions and parameters occurring
in the integrand, see the references given at the end of this section.
⊙ Literature for Section 15.2: H. Bateman and A. Erdélyi (1954), V. A. Ditkin and A. P. Prudnikov (1965),
S. G. Mikhlin (1970), J. W. Miles (1971), V. S. Vladimirov (1971, 1988), F. Oberhettinger (1972, 1974, 1980),
I. Sneddon (1972, 1995), F. Oberhettinger and L. Badii (1973), B. Davis (1978), R. Bellman and R. Roth
TABLE 15.6
Summary table of integral transforms
Integral transform Definition Inversion formula

Z ∞ Z c+i∞
Laplace 1
fe(p) = e−pxf (x) dx f (x) = epxfe(p) dp
transform 0 2πi c−i∞
Z ∞ Z c+i∞
Laplace–Carson 1 fe(p)
fe(p) = p e−pxf (x) dx f (x) = epx dp
transform 0 2πi c−i∞ p
Two-sided Z ∞ Z c+i∞
1
Laplace fe∗(p) = e−pxf (x) dx f (x) = epxfe∗(p) dp
−∞ 2πi c−i∞
transform
Z ∞ Z ∞
Fourier 1 1
fe(u) = √ e−iuxf (x) dx f (x) = √ eiuxfe(u) du
transform 2π −∞ 2π −∞
r Z ∞ r Z ∞
Fourier sine 2 2
transform fes(u) = sin(xu)f (x) dx f (x) = sin(xu)fes(u) du
π 0 π 0
r Z ∞ r Z ∞
Fourier cosine e 2 2
transform f c(u) = cos(xu)f (x) dx f (x) = cos(xu)fec(u) du
π 0 π 0
Z ∞ Z ∞
Hartley 1
feh(u) = (cos xu + sin xu)f (x) dx f (x) = (cos xu + sin xu)feh(u) du
transform −∞ 2π −∞
Z ∞ Z c+i∞
Mellin 1
fb(s) = xs−1f (x) dx f (x) = x−sfb(s) ds
transform 0 2πi c−i∞
Z ∞ Z ∞
Hankel
fbν (w) = xJν (xw)f (x) dx f (x) = wJν (xw)fbν (w) dw
transform 0 0
Z ∞√ Z ∞√
Y -transform Fν (u) = ux Yν (ux)f (x) dx f (x) = ux Hν (ux)Fν (u) du
0 0
Meijer r Z ∞ Z c+i∞√
2 √ 1 b ds
transform fb(s) = sx Kν (sx)f (x) dx f (x) = √ sx Iν (sx)f(s)
(K-transform) π 0 i 2π c−i∞
Kontorovich– Z ∞ Z ∞
2
Lebedev F (τ ) = Kiτ (x)f (x) dx f (x) = τ sinh(πτ )Kiτ (x)F (τ ) dτ
0 π 2x 0
transform
√
Notation : i = −1; Jµ(x) and Yµ(x) are the Bessel functions of the first and the second kind, respec-
tively; Iµ(x) and Kµ(x) are the modified Bessel functions of the first and the second kind, respectively; and
X∞
(−1)j (x/2)ν+2j+1
Hν (x) = is the Struve function.
j=0
Γ j + 32 Γ ν + j + 32
(1984), Yu. A. Brychkov and A. P. Prudnikov (1989), R. Courant and D. Hilbert (1989), J. R. Hanna and
J. H. Rowland (1990), A. P. Prudnikov, Yu. A. Brychkov, and O. I. Marichev (1992a, 1992b), A. Pinkus and
S. Zafrany (1997), D. Zwillinger (1998), R. Bracewell (1999), A. V. Manzhirov and A. D. Polyanin (1999),
A. D. Polyanin (2002), R. J. Beerends and H. G. ter Morschem, and J. C. van den Berg (2003), D. G. Duffy
(2004), L. Debnath and B. Bhatta (2007), A. D. Polyanin and A. V. Manzhirov (2007, 2008).
Chapter 16
Cauchy Problem.
Fundamental Solutions
16.1 Dirac Delta Function. Fundamental Solutions

16.1.1 Dirac Delta Function and Its Properties
◮ Properties of the one-dimensional Dirac delta function.
The Dirac delta function δ(x) is the singular generalized function (distribution) acting by
the rule Z ∞
ϕ(x)δ(x) dx = ϕ(0)
−∞
for an arbitrary function ϕ(x) continuous at the point x = 0.
The Dirac delta function plays an important role in the theory of linear PDEs. The
rigorous definition of this function as the limit of delta sequences of regular distributions,
as well as its physical interpretation, can be found in Chapter 21 (see also the references
therein).
Basic properties of the one-dimensional Dirac delta function:
1. f (x)δ(x − x0 ) = f (x0 )δ(x − x0 ),
2. δ(x) = δ(−x),
3. δ(cx) = |c|−1 δ(x), c = const (c 6= 0),
′
4. ϑ (x) = δ(x),
Z b (
f (x) if a < x < b,
5. f (y)δ(x − y) dy =
a 0 if x < a or x > b,
where f (x) is any continuous function and ϑ(x) is the Heaviside unit step function (ϑ = 0
for x ≤ 0 and ϑ = 1 for x > 0).
Let a continuous function f (x) have only simple zeros x1 , x2 , . . . , xm . Then
m
X δ(x − xk )
δ(f (x)) = .
|f ′ (xk )|
k=1
The number m of zeros can be infinite.
1181
1182 C AUCHY P ROBLEM . F UNDAMENTAL S OLUTIONS
Let the derivative f (n) (x) be continuous for a < x < b. Then
Z b
f (y)δ(n) (x − y) dy = f (n) (x), n = 1, 2, . . . ,
a
Z b
f (y)δ(n) (y − x) dy = (−1)n f (n) (x).
a
◮ Properties of the n-dimensional Dirac delta function.

The n-dimensional Dirac delta function possesses the following basic properties:
1. δ(x) = δ(x1 )δ(x2 ) . . . δ(xn ),
Z
2. Φ(y)δ(x − y) dVy = Φ(x),
Rn
where δ(xk ) is the one-dimensional Dirac delta function, Φ(x) is an arbitrary continuous
function, and dVy = dy1 . . . dyn .
16.1.2 Fundamental Solutions. Constructing Particular Solutions

◮ Fundamental solution of a differential operator.
Consider the linear nonhomogeneous equation
Lx [w] = Φ(x). (16.1.2.1)
Here x ∈ Rn and Lx is a linear differential operator of the second (or any) order of general
form whose coefficients may depend on x.
A generalized function (distribution) Ee = Ee (x, y) that satisfies the equation
Lx [Ee ] = δ(x − y) (16.1.2.2)
with a special right-hand side is called a fundamental solution corresponding to the op-
erator Lx . (Sometimes the function Ee will also be called the fundamental solution of
Eq. (16.1.2.1).) In (16.1.2.2), δ(x) is the n-dimensional Dirac delta function and the vector
quantity y = {y1 , . . . , yn } appears in Eq. (16.1.2.2) as an n-dimensional free parameter. It
is assumed that y ∈ Rn .
Remark 16.1. The fundamental solution Ee is not unique; it is defined up to an additive term
w0 = w0 (x) that is an arbitrary solution of the homogeneous equation Lx [w0 ] = 0.
For constant coefficient equations, a fundamental solution can be found by means of
the n-dimensional Fourier transform (see Section 15.2.2 and Example 16.1).
◮ Using fundamental solutions for constructing particular solutions.

The fundamental solution Ee = Ee (x, y) can be used to construct a particular solution of the
linear nonhomogeneous equation (16.1.2.1); this particular solution is expressed as follows:
Z
w(x) = Φ(y)Ee (x, y) dVy . (16.1.2.3)
Rn
16.1. Dirac Delta Function. Fundamental Solutions 1183
It is assumed that the function Φ(x) is continuous and sufficiently rapidly decays as
|x| → ∞, thus ensuring the convergence of the integral in (16.1.2.3) and the existence
of the derivatives determining the operator Lx .
For constant coefficient linear equations, it is customary to use the fundamental solution
Ee = Ee (x) satisfying Eq. (16.1.2.2) with y = 0. In this case, a particular solution of
Eq. (16.1.2.1) can be represented as
Z Z
w(x) = Φ(y)Ee (x − y) dVy = Φ(x − y)Ee (y) dVy . (16.1.2.4)
Rn Rn
Remark 16.2. The right-hand sides of Eqs. (16.1.2.1) and (16.1.2.2) are sometimes prefixed
with the minus sign. In this case, formula (16.1.2.3) remains valid. See also the footnote in Sec-
tion 17.4.3.
◮ Examples of fundamental solutions.

Example 16.1. Consider the 3D Poisson equation
∂2w ∂2w ∂2w

+ + = Φ(x1 , x2 , x3 ). (16.1.2.5)
∂x21 ∂x22 ∂x23
In terms of Eq. (16.1.2.1), we have Lx [w] = ∆w in this case, where ∆ is the 3D Laplace operator.
Let us find the fundamental solution of the Laplace operator. It satisfies the following Poisson
equation with a singular right-hand side:
∂ 2 Ee ∂ 2 Ee ∂ 2 Ee
2 + 2 + = δ(x1 ) δ(x2 ) δ(x3 ). (16.1.2.6)
∂x1 ∂x2 ∂x23
We use the asymmetric form of the 3D Fourier transform (15.2.2.7):

Z
Eˇe (u) = Ee (x)e−i(u·x) dVx , (u · x) = u1 x1 + u2 x2 + u3 x3 , dVx = dx1 dx2 dx3 .
R3
As a result, the PDE (16.1.2.6) is reduced to the simple algebraic equation −|u|2 Eˇe = 1, the solution
of which has the form
1
Eˇe = − 2 , |u|2 = u21 + u22 + u23 . (16.1.2.7)
|u|
Further, by applying the Fourier inversion formula (15.2.2.8), we obtain

Z
1 1 i(u·x)
Ee (x) = − e dVu , dVu = du1 du2 du3 . (16.1.2.8)
(2π)3 R3 |u|2
To compute the triple integral (16.1.2.8), we use spherical coordinates in the space of the variables
(u1 , u2 , u3 ), the azimuthal axis being the direction of the vector x. We have
u1 = ρ sin θ cos ϕ, u2 = ρ sin θ sin ϕ, u3 = ρ cos θ,

√
|u| = ρ, (u · x) = ρr cos θ, r = |x|, g = ρ2 sin θ.
The chain of computations given below, which starts from formula (16.1.2.8), permits one to find
the fundamental solution,
Z ∞ Z π Z 2π
1 1 irρ cos θ 2
Ee (x) = − e ρ sin θ dϕ dθ dρ
(2π)3 0 0 0 ρ2
Z ∞Z π Z ∞Z 1
1 irρ cos θ 1
=− e sin θ dθ dρ = − eirρµ dµ dρ
(2π)2 0 0 (2π)2 0 −1
Z ∞ Z ∞
1 sin(rρ) 1 sin λ 1
=− 2 dρ = − 2 dλ = − . (16.1.2.9)
2π r 0 ρ 2π r 0 λ 4πr
A particular solution of the Poisson equation (16.1.2.5) can be found with the use of formu-
las (16.1.2.4) and (16.1.2.9),
Z ∞Z ∞Z ∞
1 Φ(y1 , y2 , y3 ) dy1 dy2 dy3
w(x1 , x2 , x3 ) = − p .
4π −∞ −∞ −∞ (x1 − y1 )2 + (x2 − y2 )2 + (x3 − y3 )2
Remark 16.3. For the 2D Laplace operator
∂2w ∂2w
∆w = +
∂x21 ∂x22
the fundamental solution has the form
1 p
Ee (x1 , x2 ) = ln r, r = (x1 − y1 )2 + (x2 − y2 )2 .
2π
Example 16.2. The 2D Helmholtz operator
∂2w ∂2w
Lx [w] = 2 + + λw
∂x1 ∂x22
has the following fundamental solutions:
1
Ee (x1 , x2 ) = − K0 (kr) if λ = −k 2 < 0,
2π
i (2)
Ee (x1 , x2 ) = H (kr) if λ = k 2 > 0,
4 0
p
where r = (x1 − y1 )2 + (x2 − y2 )2 , K0 (z) is the modified Bessel function of the second kind,
(2)
H0 (z) is the Hankel function of the second kind of order 0, k > 0, and i2 = −1.
Example 16.3. The 3D Helmholtz operator
∂2w ∂2w ∂2w
Lx [w] = 2 + 2 + + λw
∂x1 ∂x2 ∂x23
has the following fundamental solutions:
1
Ee (x1 , x2 , x3 ) = − exp(−kr) if λ = −k 2 < 0,
4πr
1
Ee (x1 , x2 , x3 ) = − exp(−ikr) if λ = k 2 > 0,
4πr
p
where r = (x1 − y1 )2 + (x2 − y2 )2 + (x3 − y3 )2 , k > 0, and i2 = −1.
Remark 16.4. For fundamental solutions of higher-order elliptic and hyperbolic equations of
the general type, see Sections 11.6.2 and 11.6.3.
⊙ Literature for Section 16.1: G. E. Shilov (1965), I. M. Gelfand, G. E. Shilov (1959), V. S. Vladimirov
(1971, 1988), S. G. Krein (1972), V. S. Vladimirov, V. P. Mikhailov et al. (1974), A. G. Butkovskiy (1979,
1982), L. Hörmander (1983, 1990), R. P. Kanwal (1983), A. N. Tikhonov and A. A. Samarskii (1990), A. D. Po-
16.2. Representation of the Solution of the Cauchy Problem via the Fundamental Solution 1185
16.2 Representation of the Solution of the Cauchy

Problem via the Fundamental Solution
16.2.1 Cauchy Problem for Ordinary Differential Equations
◮ Representation of the fundamental solution of an ordinary differential operator.
Consider the linear ordinary differential equation
m
X (k) (k) dk w
L[w] ≡ ak (t)wt = f (t), wt ≡ . (16.2.1.1)
dtk
k=0
One can show that a fundamental solution of the operator L, i.e., a function satisfying
the equation L[Ee ] = δ(t), is given by the formula
Ee (t) = ϑ(t)Z(t), (16.2.1.2)
where ϑ(t) is the Heaviside unit step function and the function Z(t) satisfies the homoge-
neous equation L[Z] = 0 with the special initial conditions
(m−2) (m−1) 1
Z(0) = Zt′ (0) = · · · = Zt (0) = 0, Zt (0) = . (16.2.1.3)
am (0)
Example 16.4. Fundamental solutions of some operators:
Operator Fundamental solution

d
+a Ee (t) = ϑ(t)e−at ,
dt
d2 sin(at)
+ a2 Ee (t) = ϑ(t) ,
dt2 a
d2 1
− a2 Ee (t) = exp(−a|t|), a > 0.
dt2 2a
To obtain the last fundamental solution, one first uses formula (16.2.1.2), which gives Ee (t) =
ϑ(t) sinh(at)
a
1
. By subtracting the particular solution 2a exp(at) of the homogeneous equation from
this expression (recall that the fundamental solution is not unique in that one can add a solution of
the homogeneous equation to it; see Remark 16.1), we obtain the expression given above.
The fundamental solution Ee = Ee (t) can be used to construct a particular solution of

the linear nonhomogeneous equation (16.2.1.1) for arbitrary continuous f (t) (which should
vanish sufficiently rapidly as |t| → ∞); this particular solution is expressed as follows:
Z ∞
w(t) = f (τ )Ee (t − τ ) dτ. (16.2.1.4)
−∞
This formula, up to an obvious change in notation, is a special case of (16.1.2.4).

◮ Classical and generalized Cauchy problems for ordinary differential equations.
Consider the Cauchy problem for the constant coefficient linear equation (16.2.1.1) (with
ak = const) with the initial conditions
(k)
wt = bk at t = 0; k = 0, 1, . . . , m − 1, (16.2.1.5)
where the bk are some constants.

Let w(t) be the classical solution of the Cauchy problem for t > 0. We extend the
functions w(t) and f (t) by zero into the domain t < 0. Let us denote the extended functions
by w+ and f+ . We have
w+ = ϑ(t)w(t), f+ = ϑ(t)f (t). (16.2.1.6)
By successively differentiating the first relation in (16.2.1.6) and by taking into account the
formulas ϑ′t (t) = δ(t) and a(t)δ(t) = a(0)δ(t) and the initial conditions (16.2.1.5), we find
the derivatives
k−1
X
(k) (k) (k−j−1)
(w+ )′t = ϑ(t)wt′ (t) + b0 δ(t), ... , (w+ )t = ϑ(t)wt (t) + bj δt (t),
j=0
where k = 2, . . . , m. By using these formulas, we compute L[w+ ],
m−1
X m−1
X m−k−1
X
(k) (k)
L[w+ ] = ϑ(t)L[w] + ck δt (t) = f+ (t) + ck δt (t), ck = ak+j+1 bj .
k=0 k=0 j=0
We see that the function w+ satisfies the equation
m−1
X (k)
L[w+ ] = f+ (t) + ck δt (t) (16.2.1.7)
k=0
on R1 .
Thus, the classical Cauchy problem in the domain t ≥ 0 for Eq. (16.2.1.1) with the
initial conditions (16.2.1.5) can be reduced to the generalized Cauchy problem defined by
Eq. (16.2.1.7) in the domain −∞ < t < ∞. (The right-hand side of this equation contains
complete information on the initial conditions in the classical problem.)
◮ Representation of the solution of the Cauchy problem.
To construct the solution of Eq. (16.2.1.7), we use formula (16.2.1.4), in which f (t) should
P (k)
be replaced by f+ (t) + m−1 k=0 ck δt (t). We find the fundamental solution Ee of the op-
erator L by formula (16.2.1.2), where Z(t) is the solution of the homogeneous equation
L[Z] = 0 with the special initial conditions (16.2.1.3). We have

Z ∞ m−1
X
(k)
w+ = Ee (t − τ ) f+ (τ ) + ck δτ (τ ) dτ
−∞ k=0
Z ∞ m−1
X (k)
= Ee (t − τ )f+ (τ ) dτ + ck [Ee (t)]t
−∞ k=0
Z t m−1
X (k)
= ϑ(t) Z(t − τ )f (τ ) dτ + ϑ(t) ck Zt (t).
0 k=0
Here we have taken into account the relations

(k) (k)
[Ee (t)]t = ϑ(t)Zt (t) (k = 0, 1, . . . , n − 1),
which follow from (16.2.1.3). Thus, we have obtained a representation of the solution of the
Cauchy problem via the function Z(t) and the initial conditions. (Recall that the functions
w+ and w coincide in the domain t > 0.)
16.2.2 Cauchy Problem for Parabolic Equations

◮ Formula for the solution of the Cauchy problem. General case.
Let x = {x1 , . . . , xn } and y = {y1 , . . . , yn }, where x ∈ Rn and y ∈ Rn .
Consider a nonhomogeneous linear equation of the parabolic type with an arbitrary
right-hand side,
∂w
− Lx [w] = Φ(x, t), (16.2.2.1)
∂t
where t > 0; the second-order linear differential operator Lx does not contain derivatives
with respect to t and has the form (14.2.1.2).
The solution of the Cauchy problem for Eq. (16.2.2.1) with an arbitrary initial condition,
w = f (x) at t = 0, (16.2.2.2)
can be represented as the sum of two integrals,
Z tZ Z
w(x, t) = Φ(y, τ )E (x, y, t, τ ) dVy dτ + f (y)E (x, y, t, 0) dVy , (16.2.2.3)
0 Rn Rn
dVy = dy1 . . . dyn .
Here E = E (x, y, t, τ ) is the fundamental solution of the Cauchy problem. It satisfies the
homogeneous linear equation
∂E
− Lx [E ] = 0 (16.2.2.4)
∂t
for t > τ ≥ 0 and the nonhomogeneous initial condition of the special form
E = δ(x − y) at t = τ. (16.2.2.5)
The quantities τ and y appear in problem (16.2.2.4)–(16.2.2.5) as free parameters, and
δ(x) = δ(x1 ) . . . δ(xn ) is the n-dimensional Dirac delta function.
Remark 16.5. If the coefficients of the differential operator Lx in (16.2.2.4) are independent
of time t, then the fundamental solution of the Cauchy problem only depends on three arguments,
E (x, y, t, τ ) = E (x, y, t − τ ).
◮ Relation between the fundamental solutions Ee and E .

Let E = E (x, y, t, τ ) be the fundamental solution of the Cauchy problem satisfying the ho-
mogeneous linear equation (16.2.2.4) with the initial condition (16.2.2.5). Then the func-
tion
Ee (x, y, t, τ ) = ϑ(t − τ )E (x, y, t, τ ), (16.2.2.6)
where ϑ(t) is the Heaviside unit step function (ϑ = 0 for t ≤ 0 and ϑ = 1 for t > 0), is
a fundamental solution corresponding to the operator ∂t − Lx ; i.e., it satisfies the linear
equation
∂Ee
− Lx [Ee ] = δ(t − τ )δ(x − y)
∂t
with a singular right-hand side. Formula (16.2.2.6) can be proved by a straightforward
verification involving the computation of [∂t − Lx ][ϑ(t − τ )E ] with regard to the relations
ϑ′t (t) = δ(t) and E (x, t)δ(t) = E (x, 0)δ(t) and the initial condition (16.2.2.5).
In view of formula (16.2.2.6), we do not distinguish between the fundamental solutions
E and Ee for linear parabolic equations in what follows and omit the factor ϑ(t − τ ) in the
function Ee .
◮ Formula for the solution of the Cauchy problem. Constant coefficient PDEs.
For constant coefficient linear parabolic equations, one customarily uses the fundamental
solution of the Cauchy problem E = E (x, t) depending on only two arguments and satisfy-
ing Eq. (16.2.2.4) and the simpler initial condition (16.2.2.5) with y = 0 and τ = 0. The so-
lution of the Cauchy problem for Eq. (16.2.2.1) with an arbitrary initial condition (16.2.2.2)
Z tZ Z
w(x, t) = Φ(y, τ )E (x − y, t − τ ) dVy dτ + f (y)E (x − y, t) dVy . (16.2.2.7)
0 Rn Rn
Remark 16.6. For equations of the form (16.2.2.1), where Lx is a higher-order differential op-
erator in the space variables x1 , . . . , xn that does not contain t-derivatives, the fundamental so-
lution of the Cauchy problem E = E (x, y, t, τ ) also satisfies Eq. (16.2.2.4) with the initial condi-
tion (16.2.2.5). In this case, the solution of the Cauchy problem (16.2.2.1)–(16.2.2.2) can also be de-
termined by formula (16.2.2.3) (or, for constant coefficient linear equations, by formula (16.2.2.7)).
Example 16.5. Let us find the fundamental solution of the Cauchy problem for the linearized
Burgers–Korteweg–de Vries equation
∂E ∂ 3E ∂2E
+ a 3 − b 2 = 0,
∂t ∂x ∂x (16.2.2.8)
E = δ(x) at t = 0,
where a ≥ 0 and b ≥ 0 (|a| + |b| 6= 0).

We use the asymmetric form of the Fourier transform (15.2.2.3) with respect to the space vari-
able x. By setting E = F {E (x, t)} and by taking into account the relations F {∂xx E } = −u2 E
and F {∂xxx E } = −iu3E (see Property 5 and Property 6 with n = 3 in Table 15.4), we arrive at the
following problem for a linear first-order ordinary differential equation in t with parameter u:
Et′ + (bu2 − iau3 )E = 0,

(16.2.2.9)
E = 1 at t = 0.
The solution of this problem is

E = exp[(−bu2 + iau3 )t]. (16.2.2.10)
By applying the inverse Fourier transform (15.2.2.4) to (16.2.2.10), after simple transformations we
obtain Z ∞
1
E = exp[(−bu2 + iau3 )t]eixu du
2π −∞
Z (16.2.2.11)
1 ∞ 2 3
= exp(−bu ) cos(au t + xu) du.
π 0
For the special case of a = 0, we have the heat equation. The computation of the integral
in (16.2.2.11) gives
1 x2
E = √ exp − .
2 πbt 4bt
For the special case of b = 0, we have the linearized Korteweg–de Vries equation. The funda-
mental solution of the Cauchy problem acquires the form
Z
1 ∞ 1 x
E = cos(au3 t + xu) du = 1/3
Ai(z), z = ,
π 0 (3at) (3at)1/3
R∞
where Ai(z) = π1 0 cos 13 ξ 3 + zξ dξ is the Airy function.
◮ Fundamental solution allowing incomplete separation of variables.

Consider the special case where the differential operator Lx in Eq. (16.2.2.1) can be repre-
sented as the sum
Lx [w] = L1 [w] + · · · + Ln [w], (16.2.2.12)
where each term depends on a single space coordinate and time,
∂2w ∂w
Lk [w] ≡ ak (xk , t) 2 + bk (xk , t) + ck (xk , t)w, k = 1, . . . , n.
∂xk ∂x k
Equations of this form are often encountered in applications. The fundamental solution of
the Cauchy problem for the n-dimensional equation (16.2.2.1) with the operator (16.2.2.12)
can be represented in the product form
n
Y
E (x, y, t, τ ) = Ek (xk , yk , t, τ ), (16.2.2.13)
k=1
where Ek = Ek (xk , yk , t, τ ) are the fundamental solutions satisfying the one-dimensional

equations
∂Ek
− Lk [Ek ] = 0 (k = 1, . . . , n)
∂t
Ek = δ(xk − yk ) at t = τ.
In this case, the fundamental solution of the Cauchy problem (16.2.2.13) admits incom-
plete separation of variables; the fundamental solution is separated in the space variables
x1 , . . . , xn but not in time t.
Example 16.6. Consider the 2D heat equation
∂w ∂2w ∂2w
− 2 − = 0.
∂t ∂x1 ∂x22
The fundamental solutions of the Cauchy problems for the corresponding one-dimensional heat
equations are expressed as
Equations Fundamental solutions

2

∂w ∂ w 1 (x1 − y1 )2
− =0 =⇒ E1 (x1 , y1 , t, τ ) = p exp − ,
∂t ∂x21 2 π(t − τ ) 4(t − τ )

∂w ∂2w 1 (x2 − y2 )2
− =0 =⇒ E2 (x2 , y2 , t, τ ) = p exp − .
∂t ∂x22 2 π(t − τ ) 4(t − τ )
The product of E1 and E2 is the fundamental solution of the 2D heat equation,

1 (x1 − y1 )2 + (x2 − y2 )2
E (x1 , x2 , y1 , y2 , t, τ ) = exp − . (16.2.2.14)
4π(t − τ ) 4(t − τ )
In the above formulas for fundamental solutions, one can set y1 = y2 = τ = 0, denote the left-
hand side of (16.2.2.14) by E (x1 , x2 , t), and use formula (16.2.2.7) for constructing the solution of
the Cauchy problem.
Example 16.7. The fundamental solution of the Cauchy problem for the equation
n
∂w X ∂2w
− ak (t) 2 = 0, 0 < ak (t) < ∞,
∂t ∂xk
k=1
is given by formula (16.2.2.13) with

Z t
1 (xk − yk )2
Ek (xk , yk , t, τ ) = √ exp − , Tk = ak (η) dη.
2 πTk 4Tk τ
In the derivation of this formula, we have taken into account the fact that the corresponding one-
dimensional equations could be reduced to the ordinary constant coefficient heat equation by passing
from xk , t to the new variables xk , Tk .
16.2.3 Cauchy Problem for Hyperbolic Equations

◮ Formula for the solution of the Cauchy problem. General case.
Consider a nonhomogeneous linear equation of the hyperbolic type with an arbitrary right-
hand side,
∂2w ∂w
2
+ ϕ(x, t) − Lx [w] = Φ(x, t), (16.2.3.1)
∂t ∂t
where the second-order linear differential operator Lx is defined by relation (14.2.1.2) with
x ∈ Rn and t > 0.
The solution of the Cauchy problem for Eq. (16.2.3.1) with the general initial conditions
w = f0 (x) at t = 0,
(16.2.3.2)
∂t w = f1 (x) at t=0
Z tZ Z
∂
w(x, t) = Φ(y, τ )E (x, y, t, τ ) dVy dτ − f0 (y) E (x, y, t, τ ) dVy
0 Rn Rn ∂τ τ =0
Z

+ f1 (y) + f0 (y)ϕ(y, 0) E (x, y, t, 0) dVy , dVy = dy1 . . . dyn .
Rn
Here E = E (x, y, t, τ ) is the fundamental solution of the Cauchy problem, which satisfies,
for t > τ ≥ 0, the homogeneous linear equation
∂2E ∂E
+ ϕ(x, t) − Lx [E ] = 0 (16.2.3.3)
∂t2 ∂t
with the semihomogeneous initial conditions of special form
E =0 at t = τ,
(16.2.3.4)
∂t E = δ(x − y) at t = τ.
The quantities τ and y appear in problem (16.2.3.3)–(16.2.3.4) as free parameters (y ∈ Rn ).
Remark 16.7. If the coefficients of the differential operator Lx in (16.2.3.3) are independent
of time t, then the fundamental solution of the Cauchy problem depends on only three arguments,
∂
E (x, y, t, τ ) = E (x, y, t − τ ). Here ∂τ E (x, y, t, τ )τ =0 = − ∂t
∂
E (x, y, t).
◮ Relation between the fundamental solutions Ee and E .

Let E = E (x, y, t, τ ) be the fundamental solution of the Cauchy problem satisfying the
homogeneous linear equation (16.2.3.3) with the initial conditions (16.2.3.4). Then the
function
Ee (x, y, t, τ ) = ϑ(t − τ )E (x, y, t, τ ), (16.2.3.5)
where ϑ(t) is the Heaviside unit step function, is a fundamental solution corresponding to
the operator ∂tt + ϕ(x, t)∂t − Lx and satisfying the nonhomogeneous linear equation
∂ 2 Ee ∂Ee
2
+ ϕ(x, t) − Lx [Ee ] = δ(t − τ )δ(x − y)
∂t ∂t
with a singular right-hand side. Formula (16.2.3.5) can be proved by a straightforward
verification involving the computation of [∂tt + ϕ(x, t)∂t − Lx ][ϑ(t − τ )E ] with regard to
the relations ϑ′t (t − τ ) = δ(t − τ ) and a(x, t)δ(t − τ ) = a(x, τ )δ(t − τ ) and the initial
conditions (16.2.3.4).
In view of formula (16.2.3.5), we do not distinguish the fundamental solutions E and
Ee for linear hyperbolic equations in what follows and omit the factor ϑ(t − τ ) in the
function Ee .
◮ Formula for the solution of the Cauchy problem. Constant coefficient PDEs.
For constant coefficient linear hyperbolic equations, one customarily uses the fundamental
solution of the Cauchy problem E = E (x, t) depending on only two arguments and sat-
isfying Eq. (16.2.3.3) and the simpler initial conditions (16.2.3.4) with y = 0 and τ = 0.
In this case, the solution of the Cauchy problem for Eq. (16.2.3.1) with arbitrary initial
conditions (16.2.3.2) can be represented as
Z tZ Z
∂
w(x, t) = Φ(y, τ )E (x − y, t − τ ) dVy dτ + f0 (y)E (x − y, t) dVy
0 R n ∂t Rn
Z

+ f1 (y) + f0 (y)ϕ(y, 0) E (x − y, t) dVy . (16.2.3.6)
Rn
Example 16.8. For the one-, two-, and three-dimensional wave equations, the fundamental so-
lutions of the Cauchy problem have the forms
Equations Fundamental solutions
∂2w ∂2w 1
2
− =0 =⇒ E (x, t) = ϑ(t − |x|);
∂t ∂x2 2
∂2w ∂2w ∂2w ϑ(t − ρ)
− 2 − =0 =⇒ E (x1 , x2 , t) = p ;
∂t2 ∂x1 ∂x22 2π t2 − ρ2
∂2w ∂2w ∂2w ∂2w 1
2
− 2 − 2 − 2 = 0 =⇒ E (x1 , x2 , x3 , t) = δ t2 − r 2 ,
∂t ∂x1 ∂x2 ∂x3 2π
p
2 2
pϑ(z) is the Heaviside unit step function (ϑ = 0 for z ≤ 0 and ϑ = 1 for z > 0), ρ = x1 + x2 ,
where
2 2 2
r = x1 + x2 + x3 , and δ(z) is the Dirac delta function.
Example 16.9. The one-dimensional Klein–Gordon equation
∂2w ∂2w
= − bw
∂t2 ∂x2
has the following fundamental solutions of the Cauchy problem:
p
E (x, t) = 12 ϑ t − |x| J0 k t2 − x2 for b = k 2 > 0,
p
E (x, t) = 12 ϑ t − |x| I0 k t2 − x2 for b = −k 2 < 0,
where ϑ(z) is the Heaviside unit step function, J0 (z) is the Bessel function, I0 (z) is the modified
Bessel function, and k > 0.
Example 16.10. The 2D Klein–Gordon equation
∂2w ∂2w ∂2w

= 2 + − bw
∂t 2 ∂x1 ∂x22
has the following fundamental solutions of the Cauchy problem:
p
cos k t2 − ρ2
E (x1 , x2 , t) = ϑ(t − ρ) p for b = k 2 > 0,
2π t2 − ρ2
p
cosh k t2 − ρ2
E (x1 , x2 , t) = ϑ(t − ρ) p for b = −k 2 < 0,
2π t2 − ρ2
p
where ϑ(z) is the Heaviside unit step function and ρ = x21 + x22 .
16.2.4 Higher-Order Linear PDEs. Generalized Cauchy Problem

◮ Reduction of a classical Cauchy problem to a generalized Cauchy problem.
Let us describe a procedure for reducing a classical Cauchy problem to a generalized
Cauchy problem for linear partial differential equations of the form
∂2w ∂w
L[w] ≡ a 2
+b + Mx [w] = Φ(x, t) (16.2.4.1)
∂t ∂t
w = f0 (x) at t = 0, ∂t w = f1 (x) at t = 0. (16.2.4.2)
Here a and b are constants, x = (x1 , . . . , xn ) ∈ Rn , and Mx is a constant coefficient

linear differential operator of arbitrary order in the space variables x1 , . . . , xn and does not
contain t-derivatives.
Let w(x, t) be a classical solution of the Cauchy problem (16.2.4.1)–(16.2.4.2) for t > 0.
(We assume that this solution exists.) Let us extend the functions w(x, t) and Φ(x, t) by
zero into the domain t < 0. Denoting the extended functions by w+ and Φ+ , we have
w+ = ϑ(t)w(x, t), Φ+ = ϑ(t)Φ(x, t). (16.2.4.3)
By differentiating the first relation in (16.2.4.3) twice and by taking into account the for-
mulas ϑ′t (t) = δ(t) and ϕ(x, t)δ(t) = ϕ(x, 0)δ(t) and the initial conditions (16.2.4.2), we
find the t-derivatives
∂w+ ∂w ∂ 2 w+ ∂2w
= ϑ(t) + f0 (x)δ(t), = ϑ(t) + f1 (x)δ(t) + f0 (x)δt′ (t). (16.2.4.4)
∂t ∂t ∂t2 ∂t2
By using formulas (16.2.4.3) and (16.2.4.4) and the relation Mx [w+ ] = ϑ(t)Mx [w], we
compute L[w+ ],

L[w+ ] = ϑ(t)(a∂tt + b∂t + Mx )[w] + af1 (x) + bf0 (x) δ(t) + af0 (x)δt′ (t)

= ϑ(t)L[w] + af1 (x) + bf0 (x) δ(t) + af0 (x)δt′ (t)

= Φ+ (x, t) + af1 (x) + bf0 (x) δ(t) + af0 (x)δt′ (t).
It follows that the function w+ satisfies the equation

L[w+ ] = Φ+ (x, t) + af1 (x) + bf0 (x) δ(t) + af0 (x)δt′ (t). (16.2.4.5)
Thus, we have reduced the classical Cauchy problem in the domain {t ≥ 0, x ∈ Rn }

for Eq. (16.2.4.1) with the initial conditions (16.2.4.2) to the generalized Cauchy problem
determined by Eq. (16.2.4.5) in the domain Rn+1 = {−∞ < t < ∞, x ∈ Rn }. (The right-
hand side of this equation contains complete information about the initial conditions in the
classical problem.)
Remark 16.8. The solution of the generalized Cauchy problem for the n-dimensional evolution
equation into which (16.2.4.1) degenerates for a = 0 (in this case, only the first initial condition
in (16.2.4.2) should be retained) satisfies (16.2.4.5) with a = 0.
◮ Representation of the solution of the Cauchy problem via the fundamental solution.
Let a 6= 0, and let E = E (x, t) be the fundamental solution of the Cauchy problem satis-
fying the homogeneous linear equation (16.2.4.1) with Φ(x, t) = 0 and the initial condi-
tions (16.2.3.4) with y = 0 and τ = 0. Then the function
1
Ee (x, t) = ϑ(t)E (x, t), (16.2.4.6)
a
where ϑ(t) is the Heaviside unit step function, is a fundamental solution corresponding
to the operator L = a∂tt + b∂t + Mx and satisfying the nonhomogeneous linear equation
L[Ee ] = δ(t)δ(x).
The solution of the Cauchy problem can be obtained with the use of formula (16.1.2.4)
in which Rn should be replaced by Rn+1 with xn+1 = t and the function Φ(x) should be
replaced by the right-hand side of Eq. (16.2.4.5). In view of (16.2.4.6), we have the chain
of relations
Z Z ∞ n
w+ = Ee (x − y, t − τ ) Φ+ (y, τ ) + af1 (y) + bf0 (y) δ(τ )
Rn −∞
o Z Z ∞
+ af0 (y)δτ′ (τ ) dτ dVy = Ee (x − y, t − τ )Φ+ (y, τ ) dτ dVy
Rn −∞
Z Z
∂
+ Ee (x − y, t) af1 (y) + bf0 (y) dVy + Ee (x − y, t)af0 (y) dVy
Rn ∂t Rn
Z Z t
1
= ϑ(t) E (x − y, t − τ )Φ(y, τ ) dτ dVy
a Rn 0
Z Z
1 ∂
+ E (x − y, t) af1 (y) + bf0 (y) dVy + E (x − y, t)f0 (y) dVy .
a Rn ∂t Rn
Thus, we have obtained a representation of the solution of the Cauchy problem (16.2.4.1)–
(16.2.4.2) via the fundamental solution E (x, t) and the initial conditions. (Recall that the
functions w+ and w coincide in the domain t > 0.)
Example 16.11. The Cauchy problem for the fourth-order constant coefficient partial differen-
tial equation
∂2w
+ ∆∆w = Φ(x, t)
∂t2
with the initial conditions (16.2.4.2) is a special case of the Cauchy problem considered above for
Eq. (16.2.4.1) with a = 1, b = 0, and Mx [w] = ∆∆w.
◮ Cauchy problem for a more general class of linear PDEs with mixed derivatives.
Consider the Cauchy problem for a constant coefficient partial differential equation of the
form
∂2 ∂
L[w] ≡ 2
L2,x [w] + L1,x [w] + L0,x [w] = Φ(x, t) (t > 0) (16.2.4.7)
∂t ∂t
with the initial conditions (16.2.4.2). Here the Lm,x [w] (m = 0, 1, 2) are constant coef-
ficient linear differential operators of arbitrary order in the space variables x1 , . . . , xn and
do not contain t-derivatives.
Let us introduce the functions w+ and Φ+ extended by zero into the domain t < 0 by
formulas (16.2.4.3). By using the expressions (16.2.4.4) for the derivatives, we obtain
∂ ∂
L0,x [w+ ] = ϑ(t)L0,x [w], L1,x [w+ ] = ϑ(t) L1,x [w] + δ(t)L1,x [f0 (x)],
∂t ∂t
∂2 ∂2
L 2,x [w+ ] = ϑ(t) L2,x [w] + δ(t)L2,x [f1 (x)] + δt′ (t)L2,x [f0 (x)].
∂t2 ∂t2
We compute L[w+ ] by using these relations and formulas (16.2.4.3). As a result, we find
that the function w+ satisfies the equation

L[w+ ] = Φ+ (x, t) + δ(t) L2,x [f1 (x)] + L1,x [f0 (x)] + δt′ (t)L2,x [f0 (x)] (16.2.4.8)
in Rn+1 .
The solution of the Cauchy problem for equation (16.2.4.7) with the initial conditions
(16.2.4.2) can be obtained with the use of formula (16.1.2.4) with Rn replaced by Rn+1 ,
xn+1 = t, and the function Φ(x) replaced by the right-hand side of Eq. (16.2.4.8).
◮ Relation between the fundamental solutions Ee and E . General case.

Consider the Cauchy problem for the linear partial differential equation
m
X ∂k
L[w] ≡ Lk,x [w] = Φ(x, t), (16.2.4.9)
∂tk
k=0
more general than (16.2.4.7), with the initial conditions

wt=0 = f0 (x), ∂t wt=0 = f1 (x), . . . ,
(m−2) (m−1)
(16.2.4.10)
∂t w t=0 = fm−2 (x), ∂t w t=0 = fm−1 (x).
Here the Lk,x [w] are constant coefficient linear differential operators of arbitrary order in
the space variables x1 , . . . , xn and do not contain t-derivatives. We assume that the Cauchy
problem (16.2.4.9)–(16.2.4.10) is well posed.
Let E = E (x, t) be the fundamental solution of the Cauchy problem satisfying the
homogeneous linear equation (16.2.4.9) with Φ(x, t) = 0 and the initial conditions of the
special form
(m−2) (m−1)
E t=0 = 0, ∂t E t=0 = 0, . . . , ∂t E t=0 = 0, ∂t E t=0 = δ(x),
(16.2.4.11)
and let Ee = Ee (x, t) be a fundamental solution corresponding to L and satisfying the
nonhomogeneous linear equation L[Ee ] = δ(t)δ(x). Then
ϑ(t)E (x, t) = Lm,x [Ee (x, t)], (16.2.4.12)
where ϑ(t) is the Heaviside unit step function.

The solution of the Cauchy problem for the homogeneous equation (16.2.4.9) with ini-
tial conditions of the special form
(m−2) (m−1)
wt=0 = 0, ∂t wt=0 = 0, . . . , ∂t w t=0 = 0, ∂t w t=0 = fm−1 (x)
is given by the formula

Z tZ Z
w= Ee (x − y, t − τ )Φ(y, τ ) dVy dτ + E (x − y, t)fm−1 (y) dVy . (16.2.4.13)
0 Rn Rn
By taking into account relation (16.2.4.13) between the fundamental solutions, we can
successively reduce the second integral in (16.2.4.13) to the form
Z Z
E (x − y, t)fm−1 (y) dVy = E (y, t)fm−1 (x − y) dVy
Rn Z R n
Z
= Lm,y [Ee (y, t)]fm−1 (x − y) dVy = Ee (x − y, t)Lm,y [fm−1 (y)] dVy .
Rn Rn
Here we have assumed that the function fm−1 (x) sufficiently rapidly decays as |x| → ∞.
◮ On a third-order partial differential equation with mixed derivatives.

1◦ . The constant coefficient third-order partial differential equation
∂w ∂
L[w] ≡ − a∆w − b ∆w = Φ(x, t), (16.2.4.14)
∂t ∂t
which is a special case of Eq. (16.2.4.7) with
L2,x [w] = 0, L1,x [w] = (1 − b∆)[w], L0,x [w] = −a∆w, (16.2.4.15)
arises in filtration theory.

Consider the Cauchy problem for Eq. (16.2.4.14) with the initial condition
w = f0 (x) at t = 0. (16.2.4.16)
We use formulas (16.2.4.3) to introduce the functions w+ and Φ+ and then substitute the
expressions (16.2.4.15) into (16.2.4.8). As a result, we obtain the equation
L[w+ ] = Φ+ (x, t) + δ(t)(1 − b∆)[f0 (x)], (16.2.4.17)
which corresponds to the generalized Cauchy problem.

The fundamental solutions of the Cauchy problem and the operator L are related by the
formula
ϑ(t)E (x, t) = (1 − b∆)[Ee (x, t)], (16.2.4.18)
which follows from (16.2.4.12) (with m = 1) and (16.2.4.15).
The solution of the Cauchy problem for equation (16.2.4.14) with the initial conditions
(16.2.4.16) can be found with the use of formula (16.1.2.4) with Rn replaced by Rn+1 ,
xn+1 = t, and the function Φ(x) replaced by the right-hand side of Eq. (16.2.4.17). As a
result, we obtain
Z tZ Z

w= Ee (x − y, t − τ )Φ(y, τ ) dVy dτ + Ee (x − y, t) f0 (y) − b∆y f0 (y) dVy ,
0 Rn Rn
where ∆y is the Laplace operator in the integration variables y1 , . . . , yn .

2◦ . Let us find the fundamental solution Ee = Ee (x, t) of L satisfying Eq. (16.2.4.14)

with a singular right-hand side Φ(x, t) = δ(t)δ(x) for x ∈ R3 . We use the asymmetric
form (15.2.2.7) of the 3D Fourier transform in the space variables,
Z
ˇ
E e (u, t) = Ee (x, t)e−i(u·x) dVx , (u ·x) = u1 x1 + u2 x2 + u3 x3 , dVx = dx1 dx2 dx3 .
R3
As a result, we obtain the ordinary differential equation (Eˇe )′t +a|u|2 Eˇe +b|u|2 (Eˇe )′t = δ(t),
or
d ˇ a|u|2 ˇ δ(t)
Ee + 2
Ee = , |u|2 = u21 + u22 + u23 .
dt 1 + b|u| 1 + b|u|2
The solution of this equation has the form (see the fundamental solution of the first operator
in Example 16.4)

ˇ ϑ(t) a|u|2 t
Ee = exp − . (16.2.4.19)
1 + b|u|2 1 + b|u|2
Further, by applying the Fourier inversion formula (15.2.2.8), we obtain
Z
ϑ(t) ei(u·x) a|u|2 t
Ee (x, t) = exp − dVu , dVu = du1 du2 du3 .
(2π)3 R3 1 + b|u|2 1 + b|u|2
(16.2.4.20)
To compute the triple integral (16.2.4.20), just as in Example 16.1, we proceed to spher-
ical coordinates in the space of the variables (u1 , u2 , u3 ), the azimuthal axis being the
direction of the vector x. The chain of computations given below, which starts from for-
mula (16.2.4.20), permits one to find the fundamental solution
Z ∞ Z π Z 2π irρ cos θ
ϑ(t) e aρ2 t
Ee (x, t) = exp − ρ2 sin θ dϕ dθ dρ
(2π)3 0 0 0 1 + bρ2 1 + bρ2
Z ∞ Z π 2 irρ cos θ
ϑ(t) ρ e aρ2 t
= exp − sin θ dθ dρ
(2π)2 0 0 1 + bρ2 1 + bρ2
Z ∞ Z 1 2 irρµ
ϑ(t) ρ e aρ2 t
= exp − dµ dρ
(2π)2 0 −1 1 + bρ2 1 + bρ2
Z
ϑ(t) ∞ ρ sin(rρ) aρ2 t
= exp − dρ.
2π 2 r 0 1 + bρ2 1 + bρ2
3◦ . The fundamental solution Ee = Ee (x, t) of L satisfying Eq. (16.2.4.14) with a singular

right-hand side Φ(x, t) = δ(t)δ(x) for x ∈ R2 is determined in a similar way,
Z
ϑ(t) ei(u·x) a|u|2 t
Ee (x, t) = exp − du1 du2
(2π)2 R2 1 + b|u|2 1 + b|u|2
Z ∞ Z 2π irρ cos ϕ
ϑ(t) e aρ2 t
= exp − ρ dϕ dρ
(2π)2 0 0 1 + bρ2 1 + bρ2
Z
ϑ(t) ∞ ρJ0 (rρ) aρ2 t
= exp − dρ,
2π 0 1 + bρ2 1 + bρ2
where J0 (z) is the Bessel function. In the derivation of this formula, we have used the
relation Z 2π Z 2π
iz cos ϕ
e dϕ = cos(z cos ϕ) dϕ = 2πJ0 (z).
0 0
⊙ Literature for Section 16.2: G. E. Shilov (1965), V. S. Vladimirov (1971, 1988), S. G. Krein (1972),
V. S. Vladimirov, V. P. Mikhailov et al. (1974), A. G. Butkovskiy (1979, 1982), L. Hörmander (1983, 1990),
R. Courant and D. Hilbert (1989), A. N. Tikhonov and A. A. Samarskii (1990), A. D. Polyanin (2002),
Chapter 17
Boundary Value Problems.

Green’s Function
17.1 Boundary Value Problems for Parabolic Equations

with One Space Variable. Green’s Function
◮ Statement of the problem (t ≥ 0, x1 ≤ x ≤ x2 ).
In general, a nonhomogeneous linear differential equation of the parabolic type with vari-
able coefficients in one dimension can be written as
∂w
− Lx [w] = Φ(x, t), (17.1.1.1)
∂t
where
∂2w ∂w
Lx [w] ≡ a(x, t) 2
+ b(x, t) + c(x, t)w, a(x, t) > 0. (17.1.1.2)
∂x ∂x
Consider the nonstationary boundary value problem for Eq. (17.1.1.1) with an initial
condition of general form
w = f (x) at t = 0, (17.1.1.3)
and arbitrary nonhomogeneous linear boundary conditions
∂w
α1 + β1 w = g1 (t) at x = x1 , (17.1.1.4)
∂x
∂w
α2 + β2 w = g2 (t) at x = x2 . (17.1.1.5)
∂x
By appropriately choosing the coefficients α1 , α2 , β1 , and β2 in (17.1.1.4) and (17.1.1.5),

we obtain the first, second, third, and mixed boundary value problems for Eq. (17.1.1.1).
1199
1200 B OUNDARY VALUE P ROBLEMS . G REEN ’ S F UNCTION
◮ Representation of the problem solution in terms of the Green’s function.

The solution of the nonhomogeneous linear boundary value problem (17.1.1.1)–(17.1.1.5)
Z t Z x2 Z x2
w(x, t) = Φ(y, τ )G(x, y, t, τ ) dy dτ + f (y)G(x, y, t, 0) dy
0 x1 x1
Z t Z t
+ g1 (τ )a(x1 , τ )Λ1 (x, t, τ ) dτ + g2 (τ )a(x2 , τ )Λ2 (x, t, τ ) dτ. (17.1.1.6)
0 0
Here G(x, y, t, τ ) is the Green’s function that satisfies, for t > τ ≥ 0, the homogeneous
equation
∂G
− Lx [G] = 0 (17.1.1.7)
∂t
with the nonhomogeneous initial condition of special form
G = δ(x − y) at t = τ (17.1.1.8)

∂G
α1 + β1 G = 0 at x = x1 , (17.1.1.9)
∂x
∂G
α2 + β2 G = 0 at x = x2 . (17.1.1.10)
∂x
The quantities y and τ appear in problem (17.1.1.7)–(17.1.1.10) as free parameters with
x1 ≤ y ≤ x2 , and δ(x) is the Dirac delta function.
The initial condition (17.1.1.8) implies the limit relation
Z x2
f (x) = lim f (y)G(x, y, t, τ ) dy
t→τ x1
for any continuous function f = f (x).

The functions Λ1 (x, t, τ ) and Λ2 (x, t, τ ) involved in the integrands of the last two terms
in solution (17.1.1.6) can be expressed via the Green’s function G(x, y, t, τ ). The corre-
sponding formulas for Λm (x, t, τ ) are given in Table 17.1 for the basic types of boundary
value problems.
It is significant that the Green’s function G and the functions Λ1 , Λ2 are independent
of the functions Φ, f , g1 , and g2 that characterize various inhomogeneities of the boundary
value problem.
If the coefficients of Eq. (17.1.1.1)–(17.1.1.2) are independent of time t, i.e., if the
conditions
a = a(x), b = b(x), c = c(x) (17.1.1.11)
hold, then the Green’s function depends on only three arguments,
G(x, y, t, τ ) = G(x, y, t − τ ).
In this case, the functions Λm depend on only two arguments, Λm = Λm (x, t−τ ), m = 1, 2.
17.1. Boundary Value Problems for Parabolic Equations with One Space Variable 1201
TABLE 17.1
Expressions of the functions Λ1 (x, t, τ ) and Λ2 (x, t, τ ) involved
in the integrands of the last two terms in solution (17.1.1.6)
Type of problem Form of boundary conditions Functions Λm (x, t, τ )

First boundary value problem w = g1 (t) at x = x1 Λ1 (x, t, τ ) = ∂y G(x, y, t, τ )y=x
1
(α1 = α2 = 0, β1 = β2 = 1) w = g2 (t) at x = x2 Λ2 (x, t, τ ) = −∂y G(x, y, t, τ ) y=x2
Second boundary value problem ∂x w = g1 (t) at x = x1 Λ1 (x, t, τ ) = −G(x, x1 , t, τ )

(α1 = α2 = 1, β1 = β2 = 0) ∂x w = g2 (t) at x = x2 Λ2 (x, t, τ ) = G(x, x2 , t, τ )
Third boundary value problem ∂x w + β1 w = g1 (t) at x = x1 Λ1 (x, t, τ ) = −G(x, x1 , t, τ )

(α1 = α2 = 1, β1 < 0, β2 > 0) ∂x w + β2 w = g2 (t) at x = x2 Λ2 (x, t, τ ) = G(x, x2 , t, τ )

Mixed boundary value problem w = g1 (t) at x = x1 Λ1 (x, t, τ ) = ∂y G(x, y, t, τ )y=x
1
(α1 = β2 = 0, α2 = β1 = 1) ∂x w = g2 (t) at x = x2 Λ2 (x, t, τ ) = G(x, x2 , t, τ )
Mixed boundary value problem ∂x w = g1 (t) at x = x1 Λ1 (x, t, τ ) = −G(x, x1 , t, τ )

(α1 = β2 = 1, α2 = β1 = 0) w = g2 (t) at x = x2 Λ2 (x, t, τ ) = −∂y G(x, y, t, τ )y=x
2
Formula (17.1.1.6) remains valid for the problem with boundary conditions of the third
kind if β1 = β1 (t) and β2 = β2 (t). Here the relation between Λm (m = 1, 2) and the Green’s
function G is the same as that in the case of constants β1 and β2 ; the Green’s function itself
is now different.
Remark 17.1. In the first, second, and third boundary value problems that are considered on the
interval x1 ≤ x < ∞, a condition of boundedness of the solution as x → ∞ is set out. In this case,
the solution is calculated by formula (17.1.1.6) with Λ2 = 0 and Λ1 specified in Table 17.1.
◮ Formulas for calculating Green’s functions.

Consider the parabolic equation of a special form
∂w ∂2w ∂w
= a(x) 2 + b(x) + c(x) + γ(t) w + Φ(x, t), (17.1.1.12)
∂t ∂x ∂x
which is a special case of Eq. (17.1.1.1) with the operator (17.1.1.2), where a(x, t) = a(x),
b(x, t) = b(x), and c(x, t) = c(x) + γ(t). It is assumed that a(x) > 0.
The solution of the nonhomogeneous linear boundary value problem for Eq. (17.1.1.12)
subject to the initial and boundary conditions (17.1.1.3)–(17.1.1.5) is found by formula
(17.1.1.6), where the Green’s function is given by
∞
X Z t
ϕn (x)ϕn (y)
G(x, y, t, τ ) = ρ(y) exp −λn (t − τ ) + γ(ξ) dξ , (17.1.1.13)
kϕn k2 τ
n=1
where the λn and ϕn (x) are the eigenvalues and the corresponding eigenfunctions of the
Sturm–Liouville problem for the linear ordinary differential equation (15.1.1.6) with the
homogeneous linear boundary conditions (15.1.1.8) for sn = αn and kn = βn , and
Z Z x2
1 b(y) 2
ρ(y) = exp dy , kϕn k = ρ(x)ϕ2n (x) dx, (17.1.1.14)
a(y) a(y) x1
Table 17.2 lists Green’s functions for some problems for the nonhomogeneous heat
equation with a(x) = a = const and b(x) = c(x) = γ(t) ≡ 0 in (17.1.1.12). The function
G(x, y, t − τ ) in this table should be substituted for G(x, y, t, τ ) in formula (17.1.1.6).
TABLE 17.2
The Green’s functions for some boundary value problems for
2
the nonhomogeneous heat equation ∂w∂t
= a ∂∂xw2 + Φ(x, t)
Type of problem Green’s function, G(x, y, t)

First boundary
1 n h (x−y)2 i h (x+y)2 io
value problem G(x, y, t) = √ exp − −exp −
2 πat 4at 4at
(0 ≤ x < ∞)
Second boundary
1 n h (x−y)2 i h (x+y)2 io
value problem G(x, y, t) = √ exp − +exp −
2 πat 4at 4at
(0 ≤ x < ∞)
2 X nπx nπy an2π 2t
∞
G(x, y, t) = sin sin exp −
First boundary l n=1 l l l2
value problem X∞ n h (x−y+2nl)2 i h (x+y+2nl)2 io
1
(0 ≤ x ≤ l) = √ exp − −exp −
The first series converges rapidly at large t and the second series at small t
1 2X
∞ nπx nπy an2π 2t
G(x, y, t) = + cos cos exp −
Second boundary l l n=1 l l l2
value problem X∞ n h (x−y+2nl)2 i h (x+y+2nl)2 io
1
(0 ≤ x ≤ l) = √ exp − +exp −
The first series converges rapidly at large t and the second series at small t
2 X h π(2n + 1)x i h π(2n + 1)y i h aπ 2(2n + 1)2t i
∞
Mixed boundary
G(x, y, t) = sin sin exp −
value problem l n=0 2l 2l 4l2
(0 ≤ x ≤ l); ∞ n h (x − y + 2nl)2 i h (x + y + 2nl)2 io
1 X
w set at x = 0 and = √ (−1)n exp − − exp −
∂xw set at x = l 2 πat n=−∞ 4at 4at
17.1.2 Problems for Equation

h i
∂w ∂ ∂w
s(x) = p(x) − q(x)w + Φ(x, t)
∂t ∂x ∂x
◮ General formulas for solving nonhomogeneous boundary value problems.
Consider linear equations of the special form
∂w ∂ h ∂w i
s(x) = p(x) − q(x)w + Φ(x, t). (17.1.2.1)
∂t ∂x ∂x
They are often encountered in heat and mass transfer theory and chemical engineering
sciences. Throughout this subsection, we assume that the functions s, p, p′x , and q are
continuous and s > 0, p > 0, and x1 ≤ x ≤ x2 .
The solution of Eq. (17.1.2.1) under the initial condition (17.1.1.3) and the arbitrary
linear nonhomogeneous boundary conditions (17.1.1.4)–(17.1.1.5) can be represented as
17.1. Boundary Value Problems for Parabolic Equations with One Space Variable 1203
the sum
Z tZ x2 Z x2
w(x, t) = Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ + s(ξ)f (ξ)G(x, ξ, t) dξ
0 x1 x1
Z t Z t
(17.1.2.2)
+ p(x1 ) g1 (τ )Λ1 (x, t − τ ) dτ + p(x2 ) g2 (τ )Λ2 (x, t − τ ) dτ.
0 0
Here the modified Green’s function is given by
∞
X Z x2
yn (x)yn (ξ) 2
G(x, ξ, t) = exp(−λn t), kyn k = s(x)yn2 (x) dx, (17.1.2.3)
kyn k2 x1
n=1
where λn and yn (x) are the eigenvalues and the corresponding eigenfunctions of the fol-
lowing Sturm–Liouville problem for a second-order linear ordinary differential equation:
[p(x)yx′ ]′x + [λs(x) − q(x)]y = 0,
α1 yx′ + β1 y = 0 at x = x1 , (17.1.2.4)
α2 yx′ + β2 y = 0 at x = x2 .
The functions Λ1 (x, t) and Λ2 (x, t) appearing in the integrands of the last two terms in
solution (17.1.2.2) are expressed via the Green’s function (17.1.2.3). The corresponding
formulas are given in Table 17.3 for the basic types of boundary value problems.
TABLE 17.3
Expressions of the functions Λ1 (x, t) and Λ2 (x, t) involved in the integrands of the
last two terms in solutions (17.1.2.2) and (17.2.1.5); the modified Green’s function
G(x, ξ, t) for parabolic equations of the form (17.1.2.1) is found by formula (17.1.2.3),
and that for hyperbolic equations of the form (17.2.2.1), by formula (17.2.2.3)
Type of problem Form of boundary conditions Functions Λm (x, t)

First boundary value problem w = g1 (t) at x = x1 Λ1 (x, t) = ∂ξ G(x, ξ, t)ξ=x
1
(α1 = α2 = 0, β1 = β2 = 1) w = g2 (t) at x = x2 Λ2 (x, t) = −∂ξ G(x, ξ, t)ξ=x
2
Second boundary value problem ∂x w = g1 (t) at x = x1 Λ1 (x, t) = −G(x, x1 , t)

(α1 = α2 = 1, β1 = β2 = 0) ∂x w = g2 (t) at x = x2 Λ2 (x, t) = G(x, x2 , t)
Third boundary value problem ∂x w + β1 w = g1 (t) at x = x1 Λ1 (x, t) = −G(x, x1 , t)

(α1 = α2 = 1, β1 < 0, β2 > 0) ∂x w + β2 w = g2 (t) at x = x2 Λ2 (x, t) = G(x, x2 , t)

Mixed boundary value problem w = g1 (t) at x = x1 Λ1 (x, t) = ∂ξ G(x, ξ, t)ξ=x
1
(α1 = β2 = 0, α2 = β1 = 1) ∂x w = g2 (t) at x = x2 Λ2 (x, t) = G(x, x2 , t)
Mixed boundary value problem ∂x w = g1 (t) at x = x1 Λ1 (x, t) = −G(x, x1 , t)

(α1 = β2 = 1, α2 = β1 = 0) w = g2 (t) at x = x2 Λ2 (x, t) = −∂ξ G(x, ξ, t)ξ=x
2
◮ Properties of the Sturm–Liouville problem (17.1.2.4). Heat equation example.

1◦ . There are infinitely many eigenvalues. All eigenvalues are real and distinct and can be
ordered so that λ1 < λ2 < λ3 < · · · , with λn → ∞ as n → ∞ (therefore, there can exist
only finitely many negative eigenvalues). Each eigenvalue is of multiplicity 1.
2◦ . Distinct eigenfunctions yn (x) and ym (x) are orthogonal with weight s(x) on the inter-
val x1 ≤ x ≤ x2 : Z x2
s(x)yn (x)ym (x) dx = 0 for n 6= m.
x1
q(x) ≥ 0, α1 β1 ≤ 0, α2 β2 ≥ 0 (17.1.2.5)
are satisfied, there are no negative eigenvalues. If q ≡ 0 and β1 = β2 = 0, then λ1 = 0
is the least eigenvalue, with the corresponding eigenfunction ϕ1 = const. Otherwise, all
eigenvalues are positive, provided that conditions (17.1.2.5) are satisfied.
Other general and special properties of the Sturm–Liouville problem (17.1.2.4) are
given in Section 3.8.9; various asymptotic and approximate formulas for the eigenvalues
and eigenfunctions can be found there as well.
Example 17.1. Consider the first boundary value problem in the domain 0 ≤ x ≤ l for the heat
equation with a source
∂w ∂2w
= a 2 − bw
∂t ∂x
under the initial condition (17.1.1.3) and boundary conditions
w = g1 (t) at x = 0,
(17.1.2.6)
w = g2 (t) at x = l.
The above equation is a special case of Eq. (17.1.2.1) with s(x) = 1, p(x) = a, q(x) = b, and
Φ(x, t) = 0. The corresponding Sturm–Liouville problem (17.1.2.4) has the form
′′
ayxx + (λ − b)y = 0, y=0 at x = 0, y = 0 at x = l.
The eigenfunctions and eigenvalues are found to be

πnx aπ 2 n2
yn (x) = sin , λn = b + , n = 1, 2, . . .
l l2
Using formula (17.1.2.3) and taking into account the fact that kyn k2 = l/2, we obtain the Green’s
function
∞
2 −bt X πnx πnξ aπ 2 n2
G(x, ξ, t) = e sin sin exp − 2 t .
l n=1
l l l
By substituting this expression into (17.1.2.2) with p(x1 ) = p(x2 ) = s(ξ) = 1, x1 = 0, and x2 = l
and by taking into account the formulas

Λ1 (x, t) = ∂ξ G(x, ξ, t) , Λ2 (x, t) = −∂ξ G(x, ξ, t)
ξ=x1 ξ=x2
(see the first row in Table 17.3), one obtains the solution of the problem in question.
◆ The solutions of various boundary value problems for parabolic equations with one
space variable can be found in Chapter 3.
⊙ Literature for Section 17.1: P. M. Morse and H. Feshbach (1953, Vol. 2), S. G. Mikhlin (1970), V. S. Vladi-
mirov (1971, 1988), S. J. Farlow (1982), R. Leis (1986), A. A. Dezin (1987), R. Haberman (1987), H. S. Carslaw
and J. C. Jaeger (1984), E. Zauderer (1989), A. N. Tikhonov and A. A. Samarskii (1990), I. G. Petrovsky
(1991), W. A. Strauss (1992), R. B. Guenther and J. W. Lee (1996), D. Zwillinger (1998), I. Stakgold (2000),
C. Constanda (2002), A. D. Polyanin (2002), A. D. Polyanin and A. V. Manzhirov (2007).
17.2. Boundary Value Problems for Hyperbolic Equations with One Space Variable 1205
17.2 Boundary Value Problems for Hyperbolic Equations

with One Space Variable. Green’s Function. Goursat
Problem
◮ Statement of the problem (t ≥ 0, x1 ≤ x ≤ x2 ).
In general, a one-dimensional nonhomogeneous linear differential equation of hyperbolic
type with variable coefficients is written as
∂2w ∂w
2
+ σ(x, t) − Lx [w] = Φ(x, t), (17.2.1.1)
∂t ∂t
where the operator Lx [w] is defined by (17.1.1.2).
Consider the nonstationary boundary value problem for Eq. (17.2.1.1) with the initial
conditions
w = f0 (x) at t = 0,
(17.2.1.2)
∂t w = f1 (x) at t = 0
and arbitrary nonhomogeneous linear boundary conditions
∂w
α1 + β1 w = g1 (t) at x = x1 , (17.2.1.3)
∂x
∂w
α2 + β2 w = g2 (t) at x = x2 . (17.2.1.4)
∂x
◮ Representation of the problem solution in terms of the Green’s function.

The solution of problem (17.2.1.1)–(17.2.1.4) can be represented as the sum
Z tZ x2
w(x, t) = Φ(y, τ )G(x, y, t, τ ) dy dτ
0 x1
Z x2 Z x2
∂
− f0 (y) G(x, y, t, τ ) dy + f1 (y) + f0 (y)σ(y, 0) G(x, y, t, 0) dy
x1 ∂τ τ =0 x1
Z t Z t
+ g1 (τ )a(x1 , τ )Λ1 (x, t, τ ) dτ + g2 (τ )a(x2 , τ )Λ2 (x, t, τ ) dτ. (17.2.1.5)
0 0
Here the Green’s function G(x, y, t, τ ) is determined by solving the homogeneous equation
∂2G ∂G
2
+ σ(x, t) − Lx [G] = 0 (17.2.1.6)
∂t ∂t
with the semihomogeneous initial conditions
G=0 at t = τ, (17.2.1.7)
Gt = δ(x − y) at t = τ, (17.2.1.8)

∂G
α1 + β1 G = 0 at x = x1 , (17.2.1.9)
∂x
∂G
α2 + β2 G = 0 at x = x2 . (17.2.1.10)
∂x
The quantities y and τ appear in problem (17.2.1.6)–(17.2.1.10) as free parameters (x1 ≤
y ≤ x2 ), and δ(x) is the Dirac delta function.
The functions Λ1 (x, t, τ ) and Λ2 (x, t, τ ) involved in the integrands of the last two terms
in the solution (17.2.1.5) can be expressed via the Green’s function G(x, y, t, τ ). The cor-
responding formulas for Λm (x, t, τ ) are given in Table 17.1 for the main types of boundary
value problems.
It is significant that the Green’s function G and Λ1 , Λ2 are independent of the func-
tions Φ, f0 , f1 , g1 , and g2 that characterize various inhomogeneities of the boundary value
problem.
If the coefficients of Eq. (17.2.1.1) are independent of time t, then the Green’s function
depends on only three arguments, G(x, y, t, τ ) = G(x, y, t − τ ). In this case, one can set
∂
G(x, y, t, τ ) = − ∂
G(x, y, t) in the solution (17.2.1.5).
∂τ τ =0 ∂t
◮ Formulas for the calculation of Green’s functions.

Consider the hyperbolic equation of the special form
∂2w ∂w ∂2w ∂w
2
+ β(t) = a(x) 2
+ b(x) + c(x) + γ(t) w + Φ(x, t), (17.2.1.11)
∂t ∂t ∂x ∂x
which is a special case of Eq. (17.2.1.1) with the operator (17.1.1.2) where σ(x, t) = β(t),
a(x, t) = a(x), b(x, t) = b(x), and c(x, t) = c(x) + γ(t). We assume that a(x) > 0.
The solution of the nonhomogeneous linear boundary value problem for Eq. (17.2.1.11)
subject to the initial and boundary conditions (17.2.1.2)–(17.2.1.4) is found by formula
(17.2.1.5), where the Green’s function is given by
X∞
ϕn (x)ϕn (y)
G(x, y, t, τ ) = ρ(y) ψn (t, τ ). (17.2.1.12)
n=1
kϕn k2
Here the λn and ϕn (x) are the eigenvalues and the corresponding eigenfunctions of the
Sturm–Liouville problem for the linear ordinary differential equation (15.1.1.6) with the
homogeneous linear boundary conditions (15.1.1.8) for sn = αn and kn = βn ; ρ(y) and
kϕn k are determined by (17.1.1.14), and ψ = ψn (t, τ ) is the solution of Eq. (15.1.1.7) with
α(t) = 1 and λ = λn that satisfies the initial conditions
ψ = 0 at t = τ, ψt′ = 1 at
t = τ.
−1/2 1/2
In the special case β(t) = γ(t) = 0, we have ψn (t, τ ) = λn sin λn (t − τ ) .
Table 17.4 lists Green’s functions for some problems for the nonhomogeneous wave
equation, which corresponds to a(x) = a2 = const, b(x) = c(x) = β(t) = γ(t) ≡ 0
in (17.2.1.11). One should substitute G(x, y, t − τ ) from this table for G(x, y, t, τ ) into
formula (17.2.1.5).
TABLE 17.4
Green’s functions for some boundary value problems for the nonhomogeneous
2 2 ∂2 w
wave equation ∂∂tw2 = a ∂x2 + Φ(x, t) in the bounded domain 0 ≤ x ≤ l
Type of problem Green’s function

∞
2 X 1 nπx nπξ nπat
First boundary
G(x, ξ, t) = sin sin sin
value problem aπ n=1 n l l l
t
∞
2 X 1 nπx nπξ nπat
Second boundary
G(x, ξ, t) = + cos cos sin
value problem l aπ n=1 n l l l
Mixed boundary ∞
value problem; 2 X 1 π(2n + 1)
G(x, ξ, t) = sin(λn x) sin(λn ξ) sin(λn at), λn =
w is set at x = 0 and al n=1 λn 2l
∂x w is set at x = l
17.2.2 Problems for Equation

∂2w ∂ ∂w
s(x) 2 = p(x) − q(x)w + Φ(x, t)
∂t ∂x ∂x
Consider linear equations of the special form

∂2w ∂ ∂w
s(x) 2 = p(x) − q(x)w + Φ(x, t). (17.2.2.1)
∂t ∂x ∂x
It is assumed that the functions s, p, p′x , and q are continuous and the inequalities s > 0,
p > 0 hold for x1 ≤ x ≤ x2 .
The solution of Eq. (17.2.2.1) under the general initial conditions (17.2.1.2) and the
arbitrary linear nonhomogeneous boundary conditions (17.2.1.3)–(17.2.1.4) can be repre-
sented as the sum
Z t Z x2
0 x1
Z x2 Z x2
∂
+ s(ξ)f0 (ξ)G(x, ξ, t) dξ + s(ξ)f1 (ξ)G(x, ξ, t) dξ
∂t x1 x1
Z t Z t
+ p(x1 ) g1 (τ )Λ1 (x, t − τ ) dτ + p(x2 ) g2 (τ )Λ2 (x, t − τ ) dτ. (17.2.2.2)
0 0
Here the modified Green’s function is determined by
∞ √ Z x2
X yn (x)yn (ξ) sin t λn 2
G(x, ξ, t) = √
2 λ
, kyn k = s(x)yn2 (x) dx, (17.2.2.3)
n=1
ky n k n x 1
where the λn and yn (x) are the eigenvalues and the corresponding eigenfunctions of the
Sturm–Liouville problem for the second-order linear ordinary differential equation
[p(x)yx′ ]′x + [λs(x) − q(x)]y = 0,
α1 yx′ + β1 y = 0 at x = x1 , (17.2.2.4)
α2 yx′ + β2 y = 0 at x = x2 .
solution (17.2.2.2) can be expressed via the Green’s function of (17.2.2.3). The corre-
sponding formulas for Λm (x, t) are given in Table 17.3 for the basic types of boundary
value problems.
◮ Properties of the Sturm–Liouville problem. The Klein–Gordon equation.

The general and special properties of the Sturm–Liouville problem (17.2.2.4) are given
in Section 3.8.9; various asymptotic and approximate formulas for the eigenvalues and
eigenfunctions can be found there as well.
Example 17.2. Consider the second boundary value problem in the domain 0 ≤ x ≤ l for the
Klein–Gordon equation
∂2w 2
2∂ w
= a − bw
∂t2 ∂x2
under the initial conditions (17.2.1.2) and the boundary conditions
∂x w = g1 (t) at x = 0,
∂x w = g2 (t) at x = l.
The Klein–Gordon equation is a special case of Eq. (17.2.2.1) with s(x) = 1, p(x) = a2 , q(x) = b,
and Φ(x, t) = 0. The corresponding Sturm–Liouville problem (17.2.2.4) has the form
a2 yxx
′′
+ (λ − b)y = 0, yx′ = 0 at x = 0, yx′ = 0 at x = l.
The eigenfunctions and eigenvalues are found to be

πnx aπ 2 n2
yn+1 (x) = cos , λn+1 = b + , n = 0, 1, . . .
l l2
By using formula (17.2.2.4) and by taking into account the fact that ky1 k2 = l and kyn k2 = l/2
(n = 1, 2, . . .), we find the Green’s function
∞ p
1 √ 2X πnx πnξ sin t (aπn/l)2 + b
G(x, ξ, t) = √ sin t b + cos cos p .
l b l n=1 l l (aπn/l)2 + b
By substituting this expression into (17.2.2.2) with p(x1 ) = p(x2 ) = s(ξ) = 1, x1 = 0, and x2 = l
and by taking into account the formulas
Λ1 (x, t) = −G(x, x1 , t), Λ2 (x, t) = G(x, x2 , t)
(see the second row in Table 17.3), one obtains the solution of the problem in question.
◆ Solutions of various boundary value problems for hyperbolic equations with one space
variable can be found in Chapter 6.
17.2.3 Problems for Equationn h i o

∂2w ∂w ∂ ∂w
+ a(t) = b(t) p(x) − q(x)w + Φ(x, t)
∂t2 ∂t ∂x ∂x
Consider the generalized telegraph equation of the form
∂2w ∂w n ∂ h ∂w i o
+ a(t) = b(t) p(x) − q(x)w + Φ(x, t). (17.2.3.1)
∂t2 ∂t ∂x ∂x
It is assumed that the functions p, p′x , and q are continuous and p > 0 for x1 ≤ x ≤ x2 .
The solution of Eq. (17.2.3.1) under the general initial conditions (17.2.1.2) and the
arbitrary linear nonhomogeneous boundary conditions (17.2.1.3)–(17.2.1.4) can be repre-
sented as the sum
Z t Z x2
w(x, t) = Φ(ξ, τ )G(x, ξ, t, τ ) dξ dτ
0 x1
Z x2 Z x2
∂
− f0 (ξ) G(x, ξ, t, τ ) dξ + f1 (ξ) + a(0)f0 (ξ) G(x, ξ, t, 0) dξ
x1 ∂τ τ =0 x1
Z t Z t
+ p(x1 ) g1 (τ )b(τ )Λ1 (x, t, τ ) dτ + p(x2 ) g2 (τ )b(τ )Λ2 (x, t, τ ) dτ.
0 0
(17.2.3.2)
Here the Green’s function is determined by
X∞ Z x2
yn (x)yn (ξ) 2
G(x, ξ, t, τ ) = Un (t, τ ), kyn k = yn2 (x) dx, (17.2.3.3)
n=1
kyn k2 x1
where the λn and yn (x) are the eigenvalues and the corresponding eigenfunctions of the
Sturm–Liouville problem for the following second-order linear ordinary differential equa-
tion with homogeneous boundary conditions:
[p(x)yx′ ]′x + [λ − q(x)]y = 0,
α1 yx′ + β1 y = 0 at x = x1 , (17.2.3.4)
α2 yx′ + β2 y = 0 at x = x2 .
The functions Un = Un (t, τ ) are determined by solving the Cauchy problem for the linear
Un′′ + a(t)Un′ + λn b(t)Un = 0,
(17.2.3.5)
Un = 0, Un′
t=τ
= 1.
t=τ
The prime denotes the derivative with respect to t, and τ is the free parameter occurring in
the initial conditions.
solution (17.2.3.2) can be expressed via the Green’s function of (17.2.3.3). The correspond-
ing formulas will be specified below when studying specific boundary value problems.
The general and special properties of the Sturm–Liouville problem (17.2.3.4) are de-
tailed in Section 3.8.9. Asymptotic and approximate formulas for the eigenvalues and the
eigenfunctions can be found there as well.
◮ First, second, third, and mixed boundary value problems.

1◦ . First boundary value problem. The solution of Eq. (17.2.3.1) with the initial con-
ditions (17.2.1.2) and boundary conditions (17.2.1.3)–(17.2.1.4) for α1 = α2 = 0 and
β1 = β2 = 1 is given by relations (17.2.3.2) and (17.2.3.3), where
∂ ∂

Λ1 (x, t, τ ) = G(x, ξ, t, τ ) , Λ2 (x, t, τ ) = − G(x, ξ, t, τ ) .
2◦ . Second boundary value problem. The solution of Eq. (17.2.3.1) with the initial con-
ditions (17.2.1.2) and the boundary conditions (17.2.1.3)–(17.2.1.4) for α1 = α2 = 1 and
β1 = β2 = 0 is given by relations (17.2.3.2) and (17.2.3.3) with
3◦ . Third boundary value problem. The solution of Eq. (17.2.3.1) with the initial condi-
tions (17.2.1.2) and the boundary conditions (17.2.1.3)–(17.2.1.4) for α1 = α2 = 1 and
β1 β2 6= 0 is given by relations (17.2.3.2) and (17.2.3.3), in which
4◦ . Mixed boundary value problem. The solution of Eq. (17.2.3.1) with the initial condi-
tions (17.2.1.2) and the boundary conditions (17.2.1.3)–(17.2.1.4) for α1 = β2 = 0 and
α2 = β1 = 1 is given by relations (17.2.3.2) and (17.2.3.3) with
∂

Λ1 (x, t, τ ) = G(x, ξ, t, τ ) , Λ2 (x, t, τ ) = G(x, x2 , t, τ ).
∂ξ ξ=x1
5◦ . Mixed boundary value problem. The solution of Eq. (17.2.3.1) with the initial condi-
tions (17.2.1.2) and the boundary conditions (17.2.1.3)–(17.2.1.4) for α1 = β2 = 1 and
α2 = β1 = 0 is given by relations (17.2.3.2) and (17.2.3.3) with
∂

Λ1 (x, t, τ ) = −G(x, x1 , t, τ ), Λ2 (x, t, τ ) = − G(x, ξ, t, τ ) .
∂ξ ξ=x2
17.2.4 Generalized Cauchy Problem with Initial Conditions Set along

a Curve. Riemann Function
◮ Statement of the generalized Cauchy problem. Basic property of the solution.
Consider the general linear hyperbolic equation in two independent variables reduced to
the first canonical form (see Section 14.1.1)
∂2w ∂w ∂w
+ a(x, y) + b(x, y) + c(x, y)w = f (x, y), (17.2.4.1)
∂x∂y ∂x ∂y
where a(x, y), b(x, y), c(x, y), and f (x, y) are continuous functions.
Let a curve segment in the xy-plane be defined by
y = ϕ(x) (α ≤ x ≤ β), (17.2.4.2)
where ϕ(x) is continuously differentiable, with ϕ′ (x) 6= 0 and ϕ′ (x) 6= ∞.

The generalized Cauchy problem for Eq. (17.2.4.1) with initial conditions defined along
the curve (17.2.4.2) is stated as follows: find a solution of Eq. (17.2.4.1) that satisfies the
conditions
∂w ∂w
w(x, y)|y=ϕ(x) = g(x), = h1 (x), = h2 (x), (17.2.4.3)
∂x y=ϕ(x) ∂y y=ϕ(x)
where g(x), h1 (x), and h2 (x) are given continuous functions related by the compatibility
condition
gx′ (x) = h1 (x) + h2 (x)ϕ′x (x). (17.2.4.4)
Basic property of the generalized Cauchy problem: the value of the solution at any
point M (x0 , y0 ) depends only on the values of the functions g(x), h1 (x), and h2 (x) on the
arc AB cut off on the given curve (17.2.4.2) by the characteristics x = x0 and y = y0 , and
on the values of a(x, y), b(x, y), c(x, y), and f (x, y) in the curvilinear triangle AM B; see
Fig. 17.1. The influence domain of the solution at M (x0 , y0 ) is shaded for clarity.
y
y = φ(x)
M (x0, y0)
y0
A
characteristics
B x
O x0
Figure 17.1: Influence domain of the solution of the generalized Cauchy problem at a
point M .
Remark 17.2. Rather than setting two derivatives in the boundary conditions (17.2.4.3), it suf-
fices to set either of them, with the other being uniquely determined from the compatibility condi-
tion (17.2.4.4).
Remark 17.3. Instead of the last two boundary conditions in (17.2.4.3), the value of the deriva-
tive along the normal to the curve (17.2.4.2) can be used,

∂w 1 ∂w ∂w
≡ p − ϕ′x (x) = h3 (x). (17.2.4.5)
∂n y=ϕ(x) 1 + [ϕ′x (x)]2 ∂y ∂x y=ϕ(x)
Denoting wx |y=ϕ(x) = h1 (x) and wy |y=ϕ(x) = h2 (x), we have

p
h2 (x) − ϕ′x (x)h1 (x) = h3 (x) 1 + [ϕ′x (x)]2 . (17.2.4.6)
The functions h1 (x) and h2 (x) can be found from (17.2.4.4) and (17.2.4.6). Further, by substituting
their expressions into (17.2.4.3), one arrives at the standard statement of the generalized Cauchy
problem, where the compatibility condition for the initial data (17.2.4.4) will be satisfied automati-
cally.
◮ Riemann function.
The Riemann function R = R(x, y; x0 , y0 ) corresponding to Eq. (17.2.4.1) is defined as
the solution of the equation
∂2R ∂ h i ∂ h i
− a(x, y)R − b(x, y)R + c(x, y)R = 0 (17.2.4.7)
∂x∂y ∂x ∂y
with the conditions

Z y Z x
R = exp a(x0 , ξ) dξ at x = x0 , R = exp b(ξ, y0 ) dξ at y = y0
y0 x0
(17.2.4.8)
on the characteristics x = x0 and y = y0 . Here (x0 , y0 ) is an arbitrary point in the domain of
Eq. (17.2.4.1). The points x0 and y0 appear in problem (17.2.4.7)–(17.2.4.8) as parameters
in the boundary conditions only.
T HEOREM. If the functions a, b, c and the partial derivatives ax , by are continuous,
then the Riemann function R(x, y; x0 , y0 ) exists. Moreover, the function R(x0 , y0 , x, y)
obtained by swapping the parameters and the arguments is a solution of the homogeneous
equation (17.2.4.1) with f = 0.
Remark 17.4. It is significant that the Riemann function depends neither on the shape of the
curve (17.2.4.2) nor on the initial data set on it (17.2.4.3).
Example 17.3. The Riemann function for the equation wxy = 0 is just R ≡ 1.
Example 17.4. The Riemann function for the equation
wxy + cw = 0 (c = const) (17.2.4.9)
can be expressed via the Bessel function J0 (z) as
p
R = J0 4c(x0 − x)(y0 − y) .
Remark 17.5. Any linear constant coefficient partial differential equation of the hyperbolic type
in two independent variables can be reduced to an equation of the form (17.2.4.9); see the end of
Section 14.1.1.
◮ Solution of the generalized Cauchy problem via the Riemann function.

Given the Riemann function, the solution of the generalized Cauchy problem (17.2.4.1)–
(17.2.4.3) at any point (x0 , y0 ) can written as
Z
1 1 1 ∂w ∂R
w(x0 , y0 ) = (wR)A + (wR)B + R −w + 2bwR dx
2 2 2 AB ∂x ∂x
Z ZZ
1 ∂w ∂R
− R −w + 2awR dy + f R dx dy.
2 AB ∂y ∂y ∆AMB
The first two terms on the right-hand side are evaluated at the points A and B (see Fig. 17.1).
The third and fourth terms are curvilinear integrals over the arc AB; the arc is defined
by Eq. (17.2.4.2), and the integrands involve quantities defined by the initial conditions
(17.2.4.3). The last integral is taken over the curvilinear triangular domain AMB.
17.2.5 Goursat Problem (a Problem with Initial Data

on Characteristics)
◮ Statement of the Goursat problem. Basic property of the solution.
The Goursat problem for Eq. (17.2.4.1) is stated as follows: find a solution of Eq. (17.2.4.1)
with the conditions
w(x, y)|x=x1 = g(y), w(x, y)|y=y1 = h(x) (17.2.5.1)
on the characteristics, where g(y) and h(x) are given continuous functions that match each
other at the point of intersection of the characteristics, so that
g(y1 ) = h(x1 ).
Basic properties of the Goursat problem: the value of the solution at a point M (x0 , y0 )
depends only on the values of g(y) at the segment AN (which is part of the characteristic
x = x1 ), the values of h(x) at the segment BN (which is part of the characteristic y =
y1 ), and the values of the functions a(x, y), b(x, y), c(x, y), and f (x, y) in the rectangle
N AM B; see Fig. 17.2. The influence domain of the solution at the point M (x0 , y0 ) is
shaded for clarity.
y
M (x0, y0)
y0
A x = x1
characteristics
y1 y = y1
N (x1, y1) B
x
O x1 x0
Figure 17.2: Influence domain of the solution of the Goursat problem at a point M .
◮ Solution representation for the Goursat problem via the Riemann function.
Given the Riemann function (see Section 17.2.4), the solution of the Goursat problem
(17.2.4.1), (17.2.5.1) at any point (x0 , y0 ) can be written as
Z A Z B ZZ
′
′

w(x0 , y0 ) = (wR)N + R gy + bg dy + R hx + ah dx + f R dx dy.
N N NAMB
The first term on the right-hand side is evaluated at the point of intersection of the charac-
teristics (x1 , y1 ). The second and third terms are integrals along the characteristics y = y1
(x1 ≤ x ≤ x0 ) and x = x1 (y1 ≤ y ≤ y0 ); these involve the initial data of (17.2.5.1).
The last integral is taken over the rectangular domain N AM B defined by the inequalities
x1 ≤ x ≤ x0 , y 1 ≤ y ≤ y 0 .
The Goursat problem for hyperbolic equations reduced to the second canonical form
(see Section 14.1.1) can be treated in a similar way.
Example 17.5. Consider the Goursat problem for the wave equation
∂2w 2
2∂ w
− a =0
∂t2 ∂x2
with the boundary conditions prescribed on its characteristics
w = f (x) for x − at = 0 (0 ≤ x ≤ b),
(17.2.5.2)
w = g(x) for x + at = 0 (0 ≤ x ≤ c),
where f (0) = g(0).
By substituting the values set on the characteristics (17.2.5.2) into the general solution w =
ϕ(x − at) + ψ(x + at) of the wave equation, we arrive at a system of linear algebraic equations for
ϕ(x) and ψ(x). As a result, the solution of the Goursat problem is obtained in the form

x + at x − at
w(x, t) = f +g − f (0).
2 2
The solution propagation domain is the parallelogram bounded by the four lines
x − at = 0, x + at = 0, x − at = 2c, x + at = 2b.
mirov (1971, 1988), S. J. Farlow (1982), R. Leis (1986), A. A. Dezin (1987), R. Haberman (1987), R. Courant
and D. Hilbert (1989), E. Zauderer (1989), A. N. Tikhonov and A. A. Samarskii (1990), I. G. Petrovsky (1991),
W. A. Strauss (1992), R. B. Guenther and J. W. Lee (1996), D. Zwillinger (1998), I. Stakgold (2000), C. Con-
standa (2002), A. D. Polyanin (2002), A. D. Polyanin and A. V. Manzhirov (2007).
17.3 Boundary Value Problems for Elliptic Equations with

Two Space Variables
17.3.1 Problems and the Green’s Functions for Equation
∂2w ∂2w ∂w
a(x) + + b(x) + c(x)w = −Φ(x, y)
∂x2 ∂y 2 ∂x
◮ Statements of boundary value problems.
Consider 2D boundary value problems for the equation
∂2w ∂2w ∂w
a(x) 2
+ 2
+ b(x) + c(x)w = −Φ(x, y) (17.3.1.1)
∂x ∂y ∂x
with general boundary conditions in x,
∂w
α1 − β1 w = f1 (y) at x = x1 ,
∂x (17.3.1.2)
∂w
α2 + β2 w = f2 (y) at x = x2 ,
∂x
and various boundary conditions in y. It is assumed that the coefficients of Eq. (17.3.1.1)
and the boundary conditions (17.3.1.2) meet the requirements
a(x), b(x), c(x) are continuous (x1 ≤ x ≤ x2); a > 0, |α1 |+|β1 | > 0, |α2 |+|β2 | > 0.
◮ Relations for the Green’s function.

In the general case, the Green’s function can be represented as
X∞
un (x)un (ξ)
G(x, y, ξ, η) = ρ(ξ) Ψn (y, η; λn ). (17.3.1.3)
n=1
kun k2
17.3. Boundary Value Problems for Elliptic Equations with Two Space Variables 1215
Here
Z Z x2
1 b(x) 2
ρ(x) = exp dx , kun k = ρ(x)u2n (x) dx, (17.3.1.4)
a(x) a(x) x1
and the λn and un (x) are the eigenvalues and eigenfunctions of the homogeneous boundary
value problem for the ordinary differential equation
a(x)u′′xx + b(x)u′x + [λ + c(x)]u = 0, (17.3.1.5)

α1 u′x − β1 u = 0 at x = x1 , (17.3.1.6)
α2 u′x + β2 u = 0 at x = x2 . (17.3.1.7)
The functions Ψn = Ψn (y, η; λn ) for some boundary conditions in y are specified in Ta-
ble 17.5 (see also the extended Table 9.5). For unbounded domains, the condition of bound-
edness of the solution as y → ±∞ is set.
TABLE 17.5 √
The functions Ψn in (17.3.1.3) for various boundary conditions. Notation: σn = λn
Domain Boundary conditions Function Ψn (y, η; λn )
1 −σn |y−η|
−∞ < y < ∞ |w| < ∞ as y → ±∞ e
2σn
(
w=0 at y = 0, 1 e−σn y sinh(σn η) if y > η,
0≤y<∞
|w| < ∞ as y→∞ σn e−σn η sinh(σn y) if η > y
(
∂y w = 0 at y = 0, 1 e−σn y cosh(σn η) if y > η,
0≤y<∞
|w| < ∞ as y→∞ σn e−σn η cosh(σn y) if η > y
(
w=0 at y = 0, 1 sinh(σn η) sinh[σn (h − y)] if y > η,
0≤y≤h
w=0 at y=h σn sinh(σn h) sinh(σn y) sinh[σn (h − η)] if η > y
(
∂y w = 0 at y = 0, 1 cosh(σn η) cosh[σn (h − y)] if y > η,
0≤y≤h
∂y w = 0 at y=h σn sinh(σn h) cosh(σn y) cosh[σn (h − η)] if η > y
Equation (17.3.1.5) can be rewritten in self-adjoint form as
[p(x)u′x ]′x + [λρ(x) − q(x)]u = 0, (17.3.1.8)
where the functions p(x) and q(x) are given by

Z Z
b(x) c(x) b(x)
p(x) = exp dx , q(x) = − exp dx
a(x) a(x) a(x)
and ρ(x) is defined in (17.3.1.4).

The eigenvalue problem (17.3.1.8), (17.3.1.6), (17.3.1.7) possesses the following prop-
erties:
1◦ . All eigenvalues λ1 , λ2 , . . . are real and λn → ∞ as n → ∞.
2◦ . The system of eigenfunctions {u1 (x), u2 (x), . . . } is orthogonal on the interval x1 ≤

x ≤ x2 with weight ρ(x); that is,
Z x2
ρ(x)un (x)um (x) dx = 0 for n 6= m.
x1
q(x) ≥ 0, α1 β1 ≥ 0, α2 β2 ≥ 0 (17.3.1.9)
are satisfied, there are no negative eigenvalues. If q ≡ 0 and β1 = β2 = 0, then the least
eigenvalue is λ0 = 0 and the corresponding eigenfunction is u0 = const; in this case, the
summation in (17.3.1.3) should start from n = 0. In the other cases, if conditions (17.3.1.9)
are satisfied, all eigenvalues are positive; for example, the first inequality in (17.3.1.9) holds
if c(x) ≤ 0.
Section 3.8.9 presents some relations for estimating the eigenvalues λn and the eigen-
functions un (x).
Example 17.6. Consider the boundary value problem for the Laplace equation
∂2w ∂2w
2
+ =0
∂x ∂y 2
in the strip 0 ≤ x ≤ l, −∞ < y < ∞ with the mixed boundary conditions
∂w
w = f1 (y) at x = 0, = f2 (y) at x = l.
∂x
This equation is the special case of Eq. (17.3.1.1) with a(x) = 1 and b(x) = c(x) = Φ(x, t) = 0.
The corresponding Sturm–Liouville problem (17.3.1.5)–(17.3.1.7) is written as
u′′xx + λy = 0, u=0 at x = 0, u′x = 0 at x = l.
The eigenfunctions and eigenvalues are found as

π(2n − 1)x π 2 (2n − 1)2
un (x) = sin , λn = , n = 1, 2, . . .
l l2
By using formulas (17.3.1.3) and (17.3.1.4) and by taking into account the identities ρ(ξ) = 1 and
kyn k2 = l/2 (n = 1, 2, . . .) and the expression for Ψn in the first row in Table 17.5, we obtain the
Green’s function in the form
∞
1X 1 p π(2n − 1)
G(x, y, ξ, η) = sin(σn x) sin(σn ξ)e−σn |y−η| , σn = λn = .
l n=1 σn l
17.3.2 Representation of Solutions of Boundary Value Problems via

Green’s Functions
The solution of the first boundary value problem for Eq. (17.3.1.1) with the boundary con-
ditions
w = f1 (y) at x = x1 , w = f2 (y) at x = x2 ,
w = f3 (x) at y = 0, w = f4 (x) at y = h
17.3. Boundary Value Problems for Elliptic Equations with Two Space Variables 1217
can be expressed via the Green’s function (17.3.1.3) as

Z h
∂
w(x, y) = a(x1 ) f1 (η) G(x, y, ξ, η) dη
0 ∂ξ ξ=x1
Z h
∂
− a(x2 ) f2 (η) G(x, y, ξ, η) dη
0 ∂ξ ξ=x2
Z x2 Z x2
∂ ∂
+ f3 (ξ) G(x, y, ξ, η) dξ − f4 (ξ) G(x, y, ξ, η) dξ
x1 ∂η η=0 x1 ∂η η=h
Z x2Z h
x1 0
The solution of the second boundary value problem for Eq. (17.3.1.1) with the boundary
conditions
∂x w = f1 (y) at x = x1 , ∂x w = f2 (y) at x = x2 ,
∂y w = f3 (x) at y = 0, ∂y w = f4 (x) at y = h
can be expressed via the Green’s function (17.3.1.3) as

Z h Z h
w(x, y) = −a(x1 ) f1 (η)G(x, y, x1 , η) dη + a(x2 ) f2 (η)G(x, y, x2 , η) dη
0 0
Z x2 Z x2
− f3 (ξ)G(x, y, ξ, 0) dξ + f4 (ξ)G(x, y, ξ, h) dξ
x1 x1
Z x2Z h
x1 0
The solution of the third boundary value problem for Eq. (17.3.1.1) can be represented via
the Green’s function in the same way as the solution of the second boundary value problem.
(The Green’s function is now different.)
◆ The solutions of various boundary value problems for elliptic equations with two space
variables can be found in Chapter 9.
mirov (1971, 1988), S. J. Farlow (1982), R. Leis (1986), A. A. Dezin (1987), R. Haberman (1987), R. Courant
and D. Hilbert (1989), E. Zauderer (1989), A. N. Tikhonov and A. A. Samarskii (1990), I. G. Petrovsky (1991),
W. A. Strauss (1992), R. B. Guenther and J. W. Lee (1996), D. Zwillinger (1998), I. Stakgold (2000), C. Con-
standa (2002), A. D. Polyanin (2002), A. D. Polyanin and A. V. Manzhirov (2007).
17.4 Boundary Value Problems with Many Space

Variables. Representation of Solutions
via the Green’s Function
17.4.1 Problems for Parabolic Equations
◮ Statement of the problem.
In general, a nonhomogeneous linear differential equation of the parabolic type in n space
variables has the form
∂w
− Lx [w] = Φ(x, t), (17.4.1.1)
∂t
where
n
X X n
∂2w ∂w
Lx [w] ≡ aij (x, t) + bi (x, t) + c(x, t)w,
∂xi ∂xj ∂xi
i,j=1 i=1
n n (17.4.1.2)
X X
x = {x1 , . . . , xn }, aij (x, t)ξi ξj ≥ σ ξi2 , σ > 0.
i,j=1 i=1
Let V be some simply connected domain in Rn with sufficiently smooth boundary

S = ∂V . Consider the nonstationary boundary value problem for Eq. (17.4.1.1) in the
domain V with an arbitrary initial condition
w = f (x) at t = 0, (17.4.1.3)
and a nonhomogeneous linear boundary condition
Γx [w] = g(x, t) for x ∈ S. (17.4.1.4)
In the general case, Γx is a first-order linear differential operator in the space coordinates
with coefficients depending on x and t.
◮ Representation of the problem solution via the Green’s function.

The solution of the nonhomogeneous linear boundary value problem defined by (17.4.1.1)–
(17.4.1.4) can be represented as the sum
Z tZ Z
w(x, t) = Φ(y, τ )G(x, y, t, τ ) dVy dτ + f (y)G(x, y, t, 0) dVy
0 V V
Z tZ
+ g(y, τ )H(x, y, t, τ ) dSy dτ, (17.4.1.5)
0 S
where G(x, y, t, τ ) is the Green’s function; for t > τ ≥ 0, it satisfies the homogeneous
equation
∂G
− Lx [G] = 0 (17.4.1.6)
∂t
17.4. Boundary Value Problems with Many Space Variables. Green’s Function 1219
with the nonhomogeneous initial condition of the special form
G = δ(x − y) at t=τ (17.4.1.7)
and the homogeneous boundary condition
Γx [G] = 0 for x ∈ S. (17.4.1.8)
The vector y = {y1 , . . . , yn } appears in problem (17.4.1.6)–(17.4.1.8) as an n-dimensional

free parameter (y ∈ V ), and δ(x − y) = δ(x1 − y1 ) . . . δ(xn − yn ) is the n-dimensional
Dirac delta function. The Green’s function G is independent of the functions Φ, f , and g
that characterize various inhomogeneities of the boundary value problem. In (17.4.1.5), the
integration is performed everywhere with respect to y with dVy = dy1 . . . dyn .
The function H(x, y, t, τ ) involved in the integrand of the last term in solution (17.4.1.5)
can be expressed via the Green’s function G(x, y, t, τ ). The corresponding formulas for
H(x, y, t, τ ) are given in Table 17.6 for the three basic types of boundary value problems;
in the third boundary value problem, the coefficient k can depend on x and t. The boundary
conditions of the second and third kind, as well as the solution of the first boundary value
problem, involve operators of differentiation along the conormal of operator (17.4.1.2);
these operators act as follows:
n
X n
X
∂G ∂G ∂G ∂G
≡ aij (x, t)Nj , ≡ aij (y, τ )Nj , (17.4.1.9)
∂Mx ∂xi ∂My ∂yi
i,j=1 i,j=1
where N = {N1 , . . . , Nn } is the unit outward normal to the surface S. In the special case
where aii (x, t) = 1 and aij (x, t) = 0 for i 6= j, the operator (17.4.1.9) coincides with the
ordinary operator of differentiation along the outward normal to S.
TABLE 17.6
The form of the function H(x, y, t, τ ) for the basic types of nonstationary boundary value problems
Type of problem Form of boundary condition (17.4.1.4) Function H(x, y, t, τ )
First boundary ∂G
w = g(x, t) for x ∈ S H(x, y, t, τ ) = − (x, y, t, τ )
value problem ∂My
Second boundary ∂w
= g(x, t) for x ∈ S H(x, y, t, τ ) = G(x, y, t, τ )
value problem ∂Mx
Third boundary ∂w
+ kw = g(x, t) for x ∈ S H(x, y, t, τ ) = G(x, y, t, τ )
value problem ∂Mx
If the coefficient of Eq. (17.4.1.6) and the boundary condition (17.4.1.8) are indepen-
dent of t, then the Green’s function depends on only three arguments, G(x, y, t, τ ) =
G(x, y, t − τ ).
p
P
Remark 17.6. Let Si (i = 1, . . . , p) be distinct portions of the surface S such that S = Si ,
i=1
and let boundary conditions of various types be set on the Si ,
(i)
Γx [w] = gi (x, t) for x ∈ Si , i = 1, . . . , p. (17.4.1.10)
Then formula (17.4.1.5) remains valid but the last term in (17.4.1.5) should be replaced by the
sum
Xp Z tZ
gi (y, τ )Hi (x, y, t, τ ) dSy dτ. (17.4.1.11)
i=1 0 Si
17.4.2 Problems for Hyperbolic Equations

The general nonhomogeneous linear differential hyperbolic equation in n space variables
can be written as
∂2w ∂w
2
+ σ(x, t) − Lx [w] = Φ(x, t), (17.4.2.1)
∂t ∂t
where the operator Lx [w] is explicitly defined in (17.4.1.2).
Consider the nonstationary boundary value problem for Eq. (17.4.2.1) in the domain V
with arbitrary initial conditions
w = f0 (x) at t = 0, (17.4.2.2)
∂t w = f1 (x) at t=0 (17.4.2.3)
and the nonhomogeneous linear boundary condition (17.4.1.4).
◮ Representation of the solution via the Green’s function.

The solution of the nonhomogeneous linear boundary value problem defined by (17.4.2.1)–
(17.4.2.3), (17.4.1.4) can be represented as the sum
Z tZ Z
∂
w(x, t) = Φ(y, τ )G(x, y, t, τ ) dVy dτ − f0 (y) G(x, y, t, τ ) dVy
0 V V ∂τ τ =0
Z

+ f1 (y) + f0 (y)σ(y, 0) G(x, y, t, 0) dVy
V
Z tZ
+ g(y, τ )H(x, y, t, τ ) dSy dτ. (17.4.2.4)
0 S
Here G(x, y, t, τ ) is the Green’s function; for t > τ ≥ 0 it satisfies the homogeneous
equation
∂2G ∂G
+ σ(x, t) − Lx [G] = 0 (17.4.2.5)
∂t2 ∂t
with the semihomogeneous initial conditions
G=0 at t = τ,
∂t G = δ(x − y) at t = τ,
and the homogeneous boundary condition (17.4.1.8).

If the coefficients of Eq. (17.4.2.5) and the boundary condition (17.4.1.8) are indepen-
dent of time t, then the Green’s function depends on only three arguments, G(x, y, t, τ ) =
17.4. Boundary Value Problems with Many Space Variables. Green’s Function 1221

G(x, y, t − τ ). In this case, one can set ∂τ ∂
G(x, y, t, τ )τ =0 = − ∂t∂
G(x, y, t) in solu-
tion (17.4.2.4).
The function H(x, y, t, τ ) involved in the integrand of the last term in solution (17.4.2.4)
can be expressed via the Green’s function G(x, y, t, τ ). The corresponding formulas for H
are given in Table 17.6 for the three basic types of boundary value problems; in the third
boundary value problem, the coefficient k can depend on x and t.
p
P
Remark 17.7. Let Si (i = 1, . . . , p) be distinct portions of the surface S such that S = Si ,
i=1
and let boundary conditions of various types (17.4.1.10) be set on the Si . Then formula (17.4.2.4)
remains valid but the last term in (17.4.2.4) should be replaced by the sum (17.4.1.11).
17.4.3 Problems for Elliptic Equations

In general, a nonhomogeneous linear elliptic equation can be written as∗
−Lx [w] = Φ(x), (17.4.3.1)
where
n
X Xn
∂2w ∂w
Lx [w] ≡ aij (x) + bi (x) + c(x)w. (17.4.3.2)
∂xi ∂xj ∂xi
i,j=1 i=1
2D problems correspond to n = 2; 3D problems, to n = 3.

Consider Eq. (17.4.3.1)–(17.4.3.2) in a domain V and assume that the equation is sup-
plemented with the general linear boundary condition
Γx [w] = g(x) for x ∈ S. (17.4.3.3)
The solution of the stationary problem (17.4.3.1)–(17.4.3.3) can be obtained by passing

in (17.4.1.5) to the limit as t → ∞. To this end, one should start from Eq. (17.4.1.1), whose
coefficients are independent of t, and take the homogeneous initial condition (17.4.1.3)
with f (x) = 0 and the stationary boundary condition (17.4.1.4).
◮ Representation of the solution via the Green’s function.

The solution of the linear boundary value problem (17.4.3.1)–(17.4.3.3) can be represented
as the sum
Z Z
w(x) = Φ(y)G(x, y) dVy + g(y)H(x, y) dSy . (17.4.3.4)
V S
Here the Green’s function G(x, y) satisfies the nonhomogeneous equation of the special
form
−Lx [G] = δ(x − y) (17.4.3.5)
∗
In Sections 17.4.3, 17.4.4, and 17.5.1 for clarity in comparing the solutions of equations of various types,
the left-hand side of Eq. (17.4.3.1) is taken with the minus sign. For this choice of the sign, the parabolic and
hyperbolic equations (17.4.1.1) and (17.4.2.1) become the elliptic equation (17.4.3.1) if one sets ∂/∂t ≡ 0.
Note that Eqs. (14.2.1.7) and (16.1.2.1) differ in the sign of the left-hand side from Eq. (17.4.3.1).
with the homogeneous boundary condition
Γx [G] = 0 for x ∈ S. (17.4.3.6)
The vector y = {y1 , . . . , yn } appears in problem (17.4.3.5), (17.4.3.6) as an n-dimensional

free parameter (y ∈ V ). Note that G is independent of the functions Φ and g characterizing
various inhomogeneities of the original boundary value problem.
The function H(x, y) involved in the integrand of the second term in solution (17.4.3.4)
can be expressed via the Green’s function G(x, y). The corresponding formulas for H are
given in Table 17.7 for the three basic types of boundary value problems. The boundary
conditions of the second and third kind, as well as the solution of the first boundary value
problem, involve operators of differentiation along the conormal of the operator (17.4.3.2);
these operators are defined by (17.4.1.9); in this case, the coefficients aij depend on x alone.
TABLE 17.7
The form of the function H(x, y) involved in the integrand of the last term in
the solution (17.4.3.4) for the basic types of stationary boundary value problems
Type of problem Form of boundary condition (17.4.3.3) Function H(x, y)
∂G
First boundary value problem w = g(x) for x ∈ S H(x, y) = − (x, y)
∂My
∂w
Second boundary value problem = g(x) for x ∈ S H(x, y) = G(x, y)
∂Mx
∂w
Third boundary value problem + kw = g(x) for x ∈ S H(x, y) = G(x, y)
∂Mx
Remark 17.8. For the second boundary value problem with c(x) ≡ 0, the Green’s function thus
defined must not necessarily exist (e.g., see Section 9.1.1).
17.4.4 Comparison of the Solution Structures for Boundary Value

Problems for Equations of Various Types
Table 17.8 lists brief statements of boundary value problems for second-order equations of
elliptic, parabolic, and hyperbolic types. The coefficients of the differential operators Lx
and Γx in the space variables x1 , . . . , xn are assumed to be independent of time t; these
operators are the same for the problems under consideration.
Below are the respective general formulas defining the solutions of these problems with
zero initial conditions (f = f0 = f1 = 0):
Z Z

w0 (x) = Φ(y)G0 (x, y) dVy + g(y)H G0 (x, y) dSy ,
V S
Z tZ Z tZ

w1 (x, t) = Φ(y, τ )G1 (x, y, t − τ ) dVy dτ + g(y, τ )H G1 (x, y, t − τ ) dSy dτ,
0 V 0 S
Z tZ Z tZ

w2 (x, t) = Φ(y, τ )G2 (x, y, t − τ ) dVy dτ + g(y, τ )H G2 (x, y, t − τ ) dSy dτ,
0 V 0 S
17.5. Construction of the Green’s Functions. General Formulas and Relations 1223
TABLE 17.8
Statements of boundary value problems for equations of various types
Type of equation Form of equation Initial conditions Boundary conditions
Elliptic −Lx [w] = Φ(x) not set Γx [w] = g(x) for x ∈ S
Parabolic ∂t w − Lx [w] = Φ(x, t) w = f (x) at t = 0 Γx [w] = g(x, t) for x ∈ S
w = f0 (x) at t = 0,
Hyperbolic ∂tt w − Lx [w] = Φ(x, t) Γx [w] = g(x, t) for x ∈ S
∂t w = f1 (x) at t = 0
where the Gn are the Green’s functions, and the subscripts 0, 1, and 2 refer to the elliptic,
parabolic, and hyperbolic problem, respectively. All solutions involve the same opera-
tor H[G]; it is explicitly defined in Sections 17.4.1–17.4.3 (see also Sections 17.1.1 and
17.2.1) for various boundary conditions.
It is apparent that the solutions of the parabolic and hyperbolic problems with zero
initial conditions have the same structure. The structure of the solution of a problem for
a parabolic equation differs from that for an elliptic equation by the additional integration
with respect to t.
R. Leis (1986), A. A. Dezin (1987), R. Haberman (1987), R. Courant and D. Hilbert (1989), E. Zauderer (1989),
A. N. Tikhonov and A. A. Samarskii (1990), I. G. Petrovsky (1991), W. A. Strauss (1992), R. B. Guenther and
J. W. Lee (1996), D. Zwillinger (1998), I. Stakgold (2000), C. Constanda (2002), A. D. Polyanin (2002),
17.5 Construction of the Green’s Functions. General

Formulas and Relations
17.5.1 Green’s Functions of Boundary Value Problems for Equations
of Various Types in Bounded Domains
◮ Expressions of the Green’s function in terms of infinite series.
Table 17.9 lists the Green’s functions of boundary value problems for second-order equa-
tions of various types in a bounded domain V. It is assumed that Lx is a second-order linear
self-adjoint differential operator of the form (14.2.1.10) in the space variables x1 , . . . , xn ,
and Γx is a zero- or first-order linear boundary operator that can define a boundary con-
dition of the first, second, or third kind; the coefficients of the operators Lx and Γx can
depend on the space variables but are independent of time t. The coefficients λk and the
functions uk (x) are determined by solving the homogeneous eigenvalue problem
Lx [u] + λu = 0, (17.5.1.1)
Γx [u] = 0 for x ∈ S. (17.5.1.2)
It is apparent from Table 17.9 that, given the Green’s function in the problem for a
parabolic (or hyperbolic) equation, one can easily construct the Green’s functions of the
TABLE 17.9
Green’s functions of boundary value problems for equations of various types in bounded
domains. In all problems, the operators Lx and Γx are the same; x = {x1 , . . . , xn }
Initial and
Equation Green’s function
boundary conditions
∞
X
Elliptic equation Γx [w] = g(x) for x ∈ S uk (x)uk (y)
G(x, y) = , λk 6= 0
−Lx [w] = Φ(x) (no initial condition required) kuk k2 λk
k=1
∞
X
Parabolic equation w = f (x) at t=0 uk (x)uk (y)
G(x, y, t) = exp −λk t
∂t w − Lx [w] = Φ(x, t) Γx [w] = g(x, t) for x∈S kuk k2
k=1
w = f0 (x) at t=0 ∞
Hyperbolic equation X uk (x)uk (y) p
w = f1 (x) at t=0 G(x, y, t) = √ sin t λk
∂tt w − Lx [w] = Φ(x, t) 2
kuk k λk
k=1
Γx [w] = g(x, t) for x∈S
corresponding problems for elliptic and hyperbolic (or parabolic) equations. In particular,
the Green’s function of the problem for an elliptic equation can be expressed via the Green’s
function of the problem for a parabolic equation as follows:
Z ∞
G0 (x, y) = G1 (x, y, t) dt. (17.5.1.3)
0
Here the fact that all λk are positive is taken into account; for the second boundary value
problem, it is assumed that λ = 0 is not an eigenvalue of problem (17.5.1.1)–(17.5.1.2).
◮ Some remarks and generalizations.

Remark 17.9. Formula (17.5.1.3) can also be used if the domain V is infinite. In this case, one
should make sure that the integral on the right-hand side is convergent.
Remark 17.10. Suppose that the equations given in the first column of Table 17.9 contain
−Lx [w] − βw instead of −Lx [w] with β being a free parameter. Then the λk in the expressions
of the Green’s function in the third column of Table 17.9 should be replaced by λk − β; just as pre-
viously, the λk and uk (x) were determined by solving the eigenvalue problem (17.5.1.1)–(17.5.1.2).
Remark 17.11. The formulas for the Green’s functions presented in Table 17.9 will also hold
for boundary value problems described by equations of the fourth or higher orders in the space
variables, provided that the eigenvalue problem for Eq. (17.5.1.1) supplemented with appropriate
boundary conditions is self-adjoint.
17.5.2 Green’s Functions Admitting Incomplete Separation

of Variables
◮ Boundary value problems for rectangular domains.
1◦ . Consider the parabolic equation
∂w
= L1 [w] + · · · + Ln [w] + Φ(x, t), (17.5.2.1)
∂t
where each term Lm [w] is a second-order linear differential operator in only one space
variable xm with coefficients dependent on xm and t,
∂2w ∂w
Lm [w] ≡ am (xm , t) 2
+ bm (xm , t) + cm (xm , t)w, m = 1, . . . , n.
∂xm ∂xm
For Eq. (17.5.2.1), we set the initial condition of general form
w = f (x) at t = 0. (17.5.2.2)
Consider the domain V = {αm ≤ xm ≤ βm , m = 1, . . . , n}, which is an n-dimensional

parallelepiped. We set the following boundary conditions at the faces of the parallelepiped:
∂w
s(1)
m
(1)
+ km (1)
(t)w = gm (x, t) at xm = αm ,
∂xm
(17.5.2.3)
∂w
s(2)
m
(2)
+ km (2)
(t)w = gm (x, t) at xm = βm .
∂xm
(1) (2) (1) (1) (2)
By appropriately choosing the coefficients sm , sm and the functions km = km (t), km =
(2)
km (t), we can obtain boundary conditions of the first, second, or third kind. For infinite
domains, the boundary conditions corresponding to αm = −∞ or βm = ∞ are omitted.
2◦ . The Green’s function of the nonstationary n-dimensional boundary value problem
(17.5.2.1)–(17.5.2.3) can be represented in the product form
n
Y
G(x, y, t, τ ) = Gm (xm , ym , t, τ ), (17.5.2.4)
m=1
where the Green’s functions Gm = Gm (xm , ym , t, τ ) satisfy the one-dimensional equations
∂Gm
− Lm [Gm ] = 0 (m = 1, . . . , n)
∂t
Gm = δ(xm − ym ) at t=τ
∂Gm
s(1)
m
(1)
+ km (t)Gm = 0 at xm = αm ,
∂xm
∂Gm
s(2)
m
(2)
+ km (t)Gm = 0 at xm = βm .
∂xm
Here ym and τ are free parameters (αm ≤ ym ≤ βm and t ≥ τ ≥ 0), and δ(x) is the Dirac
delta function.
It can be seen that the Green’s function (17.5.2.4) admits incomplete separation of vari-
ables; it separates in the space variables x1 , . . . , xn but not in time t.
Example 17.7. Consider the boundary value problem for the nonhomogeneous 2D heat equation
∂w ∂2w ∂ 2w
= + + Φ(x1 , x2 , t)
∂t ∂x21 ∂x22
with initial condition (17.5.2.2) and the nonhomogeneous mixed boundary conditions
w = g1 (x2 , t) at x1 = 0, w = h1 (x2 , t) at x1 = l1 ;
∂w ∂w
= g2 (x1 , t) at x2 = 0, = h2 (x1 , t) at x2 = l2 .
∂x2 ∂x2
The Green’s functions of the corresponding homogeneous one-dimensional heat equations with
homogeneous boundary conditions are expressed as
Equations and boundary conditions Green’s functions
∞
X
∂w ∂2w 2 2
= , w = 0 at x1 = 0, l1 =⇒ G1 = sin(λm x1 ) sin(λm y1 )e−λm (t−τ ) ;
∂t ∂x21 l1 m=1
∞
∂w ∂2w ∂w 1 2X 2
= , = 0 at x2 = 0, l2 =⇒ G2 = + sin(σn x2 ) sin(σn y2 )e−σn (t−τ ) ,
∂t ∂x22 ∂x2 l2 l2 n=1
where λm = mπ/l1 and σn = nπ/l2 . Multiplying G1 and G2 together gives the Green’s function
for the original 2D problem:
∞
4 X 2
G(x1 , x2 , y1 , y2 , t, τ ) = sin(λm x1 ) sin(λm y1 )e−λm (t−τ )
l1 l2 m=1
∞
1 X 2
× + sin(σn x2 ) sin(σn y2 )e−σn (t−τ ) .
2 n=1
◮ Boundary value problems for an arbitrary cylindrical domain.

1◦ . Consider the parabolic equation
∂w
= Lx [w] + Mz [w] + Φ(x, z, t), (17.5.2.5)
∂t
where Lx is an arbitrary second-order linear differential operator in x1 , . . . , xn with coeffi-
cients depending on x and t and Mz is an arbitrary second-order linear differential operator
in z with coefficients depending on z and t.
For Eq. (17.5.2.5), we set the general initial condition (17.5.2.2), where f (x) should be
replaced by f (x, z).
We assume that the space variables belong to a cylindrical domain V = {x ∈ D,
z1 ≤ z ≤ z2 } with arbitrary cross-section D. We set the boundary conditions∗
Γ1 [w] = g1 (x, t) at z = z1 (x ∈ D),
Γ2 [w] = g2 (x, t) at z = z2 (x ∈ D), (17.5.2.6)
Γ3 [w] = g3 (x, z, t) for x ∈ ∂D (z1 ≤ z ≤ z2 ),
where the linear boundary operators Γk (k = 1, 2, 3) can define boundary conditions of the
first, second, or third kind; in the last case, the coefficients of the differential operators Γk
can depend on t.
∗
If z1 = −∞ or z2 = ∞, the corresponding boundary condition is to be omitted.
2◦ . The Green’s function of problem (17.5.2.5)–(17.5.2.6), (17.5.2.2) can be represented

in the product form
G(x, y, z, ζ, t, τ ) = GL (x, y, t, τ )GM (z, ζ, t, τ ), (17.5.2.7)
where GL = GL (x, y, t, τ ) and GM = GM (z, ζ, t, τ ) are auxiliary Green’s functions; these

can be determined from the following two simpler problems with fewer independent vari-
ables:
Problem on the cross-section D: Problem on the interval z1 ≤ z ≤ z2 :

 
 ∂GL  ∂GM

 ∂t = Lx [GL ] for x ∈ D, 
 ∂t = Mz [GM ] for z1 < z < z2 ,
 GL = δ(x − y) at t = τ,
  GM = δ(z − ζ)

at t = τ,
 
Γ3 [GL ] = 0 for x ∈ ∂D, Γk [GM ] = 0 at z = zk (k = 1, 2).
Here y, ζ, and τ are free parameters (y ∈ D, z1 ≤ ζ ≤ z2 , and t ≥ τ ≥ 0).

It can be seen that the Green’s function (17.5.2.7) admits incomplete separation of vari-
ables; it separates in the space variables x and z but not in time t.
17.5.3 Construction of Green’s Functions via Fundamental

Solutions
◮ Elliptic equations. Fundamental solution.
Consider the elliptic equation
∂2w
Lx [w] + = Φ(x, z), (17.5.3.1)
∂z 2
where x = {x1 , . . . , xn } ∈ Rn , z ∈ R1 , and Lx [w] is a linear differential operator that

depends on x1 , . . . , xn but is independent of z. For the subsequent analysis, it is significant
that the homogeneous equation (with Φ ≡ 0) does not change under the replacement of z
by −z and z by z + const.
Let Ee = Ee (x, y, z − ζ) be the fundamental solution of Eq. (17.5.3.1), which means
that
∂ 2 Ee
Lx [Ee ] + = δ(x − y)δ(z − ζ).
∂z 2
Here y = {y1 , . . . , yn } ∈ Rn and ζ ∈ R1 are free parameters.
The fundamental solution of Eq. (17.5.3.1) is an even function in the last argument; i.e.,
Ee (x, y, z) = Ee (x, y, −z).
Below we present relations that permit one to express the Green’s functions of some
boundary value problems for Eq. (17.5.3.1) via its fundamental solution.
◮ Domain: x ∈ Rn , 0 ≤ z < ∞. Problems for elliptic equations.

1◦ . First boundary value problem. The boundary condition:
w = f (x) at z = 0.
Green’s function:
G(x, y, z, ζ) = Ee (x, y, z − ζ) − Ee (x, y, z + ζ). (17.5.3.2)
Domain of the free parameters: y ∈ Rn and 0 ≤ ζ < ∞.

Example 17.8. Consider the first boundary value problem in the half-space −∞ < x1 , x2 < ∞,
0 ≤ x3 < ∞ for the 3D Laplace equation
∂2w ∂2w ∂2w

2 + 2 + =0
∂x1 ∂x2 ∂x23
under the boundary condition

w = f (x1 , x2 ) at x3 = 0.
The fundamental solution of the Laplace equation has the form
1
Ee = p .
4π (x1 − y1 ) + (x2 − y2 )2 + (x3 − y3 )2
2
In terms of the notation adopted for Eq. (17.5.3.1) and its fundamental solution, we have x3 = z,
y3 = ζ, and E = E (x1 , y1 , x2 , y2 , z − ζ). Using formula (17.5.3.2), we obtain the Green’s function
for the first boundary value problem in the half-space:
G(x1 , y1 , x2 , y2 , z, ζ) = Ee (x1 , y1 , x2 , y2 , z − ζ) − Ee (x1 , y1 , x2 , y2 , z + ζ)

1
= p
4π (x1 − y1 ) + (x2 − y2 )2 + (x3 − y3 )2
2
1
− p .
4π (x1 − y1 ) + (x2 − y2 )2 + (x3 + y3 )2
2
2◦ . Second boundary value problem. The boundary condition:
∂z w = f (x) at z = 0.
Green’s function:
G(x, y, z, ζ) = Ee (x, y, z − ζ) + Ee (x, y, z + ζ).
Example 17.9. The Green’s function of the second boundary value problem for the 3D Laplace
equation in the half-space −∞ < x1 , x2 < ∞, 0 ≤ x3 < ∞ is expressed as
1
G(x1 , y1 , x2 , y2 , z, ζ) = p
4π (x1 − y1 ) + (x2 − y2 )2 + (x3 − y3 )2
2
1
+ p .
4π (x1 − y1 )2 + (x2 − y2 )2 + (x3 + y3 )2
It is obtained using the same reasoning as in Example 17.8.

3◦ . Third boundary value problem. The boundary condition:

∂z w − kw = f (x) at z = 0.
Green’s function:
Z ∞
G(x, y, z, ζ) = Ee (x, y, z − ζ) + Ee (x, y, z + ζ) − 2k e−ks Ee (x, y, z + ζ + s) ds
Z0 ∞
= Ee (x, y, z − ζ) + Ee (x, y, z + ζ) − 2k e−k(σ−z−ζ) Ee (x, y, σ) dσ.
z+ζ
◮ Domain: x ∈ Rn , 0 ≤ z ≤ l. Problems for elliptic equations.

1◦ . First boundary value problem. Boundary conditions:
w = f1 (x) at z = 0, w = f2 (x) at z = l.
Green’s function:
∞
X
G(x, y, z, ζ) = Ee (x, y, z − ζ + 2nl) − Ee (x, y, z + ζ + 2nl) . (17.5.3.3)
n=−∞
Domain of the free parameters: y ∈ Rn and 0 ≤ ζ ≤ l.

2◦ . Second boundary value problem. Boundary conditions:
∂z w = f1 (x) at z = 0, ∂z w = f2 (x) at z = l.
Green’s function:
∞
X
G(x, y, z, ζ) = Ee (x, y, z − ζ + 2nl) + Ee (x, y, z + ζ + 2nl) . (17.5.3.4)
n=−∞
3◦ . Mixed boundary value problem. The unknown function and its derivative are pre-
scribed at the left and right end, respectively:
w = f1 (x) at z = 0, ∂z w = f2 (x) at z = l.
Green’s function:
X∞

G(x, y, z, ζ) = (−1)n Ee (x, y, z − ζ + 2nl) − Ee (x, y, z + ζ + 2nl) . (17.5.3.5)
n=−∞
4◦ .Mixed boundary value problem. The derivative and the unknown function itself are
prescribed at the left and right end, respectively:
∂z w = f1 (x) at z = 0, w = f2 (x) at z = l.
Green’s function:
X∞

G(x, y, z, ζ) = (−1)n Ee (x, y, z − ζ + 2nl) + Ee (x, y, z + ζ + 2nl) . (17.5.3.6)
n=−∞
Remark 17.12. One should make sure that the series (17.5.3.3)–(17.5.3.6) are convergent; in
particular, for the 3D Laplace equation, the series (17.5.3.3), (17.5.3.5), and (17.5.3.6) are conver-
gent, and the series (17.5.3.4) is divergent.
◮ Boundary value problems for parabolic equations.

Let x ∈ Rn , z ∈ R1 , and t ≥ 0. Consider the parabolic equation
∂w ∂2w
= Lx [w] + + Φ(x, z, t), (17.5.3.7)
∂t ∂z 2
where Lx is an arbitrary linear differential operator in x1 , . . . , xn with coefficients depend-
ing on x and t but independent of z.
Let E = E (x, y, z − ζ, t, τ ) be the fundamental solution of the Cauchy problem for
Eq. (17.5.3.7); i.e.,
∂E ∂2E
= Lx [E ] + for t > τ ≥ 0,
∂t ∂z 2
E = δ(x − y) δ(z − ζ) at t = τ.
Here y ∈ Rn , ζ ∈ R1 , and τ ≥ 0 are free parameters.

The fundamental solution of the Cauchy problem possesses the property
E (x, y, z, t, τ ) = E (x, y, −z, t, τ ).
TABLE 17.10
Representation of the Green’s functions of some nonstationary boundary
value problems via the fundamental solution of the Cauchy problem
Boundary value
Boundary conditions Green’s functions
problems
First problem
G = 0 at z = 0 G(x, y, z, ζ, t, τ ) = E (x, y, z − ζ, t, τ ) − E (x, y, z + ζ, t, τ )
x ∈ Rn , z ∈ R1
Second problem
∂z G = 0 at z = 0 G(x, y, z, ζ, t, τ ) = E (x, y, z − ζ, t, τ ) + E (x, y, z + ζ, t, τ )
x ∈ Rn , z ∈ R1
G(x, y, z, ζ, t, τ ) = E (x, y, z − ζ, t, τ ) + E (x, y, z + ζ, t, τ )
Third problem ∂z G − kG = 0 Z ∞
x ∈ Rn , z ∈ R1 at z = 0 − 2k e−ks E (x, y, z + ζ + s, t, τ ) ds
0
∞
P
First problem G = 0 at z = 0, G(x, y, z, ζ, t, τ ) = E (x, y, z − ζ + 2nl, t, τ )
n=−∞
x ∈ Rn , 0 ≤ z ≤ l G = 0 at z = l
− E (x, y, z + ζ + 2nl, t, τ )
∞
P
Second problem ∂z G = 0 at z = 0, G(x, y, z, ζ, t, τ ) = E (x, y, z − ζ + 2nl, t, τ )
n=−∞
x ∈ Rn , 0 ≤ z ≤ l ∂z G = 0 at z = l
+ E (x, y, z + ζ + 2nl, t, τ )
P
∞
Mixed problem G = 0 at z = 0, G(x, y, z, ζ, t, τ ) = (−1)n E (x, y, z − ζ + 2nl, t, τ )
n=−∞
x ∈ Rn , 0 ≤ z ≤ l ∂z G = 0 at z = l
− E (x, y, z + ζ + 2nl, t, τ )
P
∞
Mixed problem ∂z G = 0 at z = 0, G(x, y, z, ζ, t, τ ) = (−1)n E (x, y, z − ζ + 2nl, t, τ )
n=−∞
x ∈ Rn , 0 ≤ z ≤ l G = 0 at z = l
+ E (x, y, z + ζ + 2nl, t, τ )
Table 17.10 presents formulas that permit one to express the Green’s functions of some
nonstationary boundary value problems for Eq. (17.5.3.7) via the fundamental solution of
the Cauchy problem.
⊙ Literature for Section 17.5: S. G. Mikhlin (1970), V. S. Vladimirov (1971, 1988), A. G. Butkovskiy (1982),
S. J. Farlow (1982), R. Leis (1986), A. A. Dezin (1987), R. Haberman (1987), R. Courant and D. Hilbert
(1989), E. Zauderer (1989), A. N. Tikhonov and A. A. Samarskii (1990), W. A. Strauss (1992), R. B. Guenther
and J. W. Lee (1996), D. Zwillinger (1998), I. Stakgold (2000), C. Constanda (2002), A. D. Polyanin (2002),
Chapter 18
Duhamel’s Principles.
Some Transformations
18.1 Duhamel’s Principles in Nonstationary Problems

18.1.1 Problems for Homogeneous Linear Equations
◮ Parabolic equations with two independent variables.
Consider the problem for the homogeneous linear equation of parabolic type
∂w ∂2w ∂w
= a(x) 2 + b(x) + c(x)w (18.1.1.1)
∂t ∂x ∂x
with the homogeneous initial condition
w=0 at t = 0 (18.1.1.2)
and the boundary conditions
s1 ∂x w + k1 w = g(t) at x = x1 , (18.1.1.3)
s2 ∂x w + k2 w = 0 at x = x2 . (18.1.1.4)
By appropriately choosing the values of the coefficients s1 , s2 , k1 , and k2 in (18.1.1.3)

and (18.1.1.4), one can obtain the first, second, third, and mixed boundary value problems
for Eq. (18.1.1.1).
The solution of problem (18.1.1.1)–(18.1.1.4) with the nonstationary boundary condi-
tion (18.1.1.3) at x = x1 can be expressed by the formula (Duhamel’s first principle)
Z t Z t
∂ ∂u
w(x, t) = u(x, t − τ ) g(τ ) dτ = (x, t − τ ) g(τ ) dτ (18.1.1.5)
∂t 0 0 ∂t
via the solution u(x, t) of the auxiliary problem for Eq. (18.1.1.1) with the initial and
boundary conditions (18.1.1.2) and (18.1.1.4) for u instead of w and with the following
simpler stationary boundary condition at x = x1 :
s1 ∂x u + k1 u = 1 at x = x1 . (18.1.1.6)
1233
1234 D UHAMEL’ S P RINCIPLES . S OME T RANSFORMATIONS
Remark 18.1. A similar formula holds for the homogeneous boundary condition at x = x1 and
a nonhomogeneous nonstationary boundary condition at x = x2 .
Example 18.1. Consider the first boundary value problem for the heat equation
∂w ∂2w
= (18.1.1.7)
∂t ∂x2
with the homogeneous initial condition (18.1.1.2) and the boundary condition
w = g(t) at x = 0. (18.1.1.8)
(The second boundary condition is not required in this case; 0 ≤ x < ∞.)
First, consider the following auxiliary problem for the heat equation with the homogeneous
initial condition and a simpler boundary condition:
∂u ∂2u
= , u=0 at t = 0, u = 1 at x = 0.
∂t ∂x2
This problem has a self-similar solution of the form
w = w(z), z = xt−1/2 ,
where the function w(z) is determined by the following ordinary differential equation and boundary
conditions:
u′′zz + 12 zu′z = 0, u = 1 at z = 0, u = 0 at z = ∞.
Its solution is expressed as

z x
u(z) = erfc =⇒ u(x, t) = erfc √ ,
2 2 t
Z ∞
2
where erfc z = √ exp(−ξ 2 ) dξ is the complementary error function. By substituting the
π z
expression obtained for u(x, t) into (18.1.1.5), we obtain the solution of the first boundary value
problem for the heat equation (18.1.1.7) with the initial condition (18.1.1.2) and an arbitrary bound-
ary condition (18.1.1.8) in the form
Z t
x x2 g(τ ) dτ
w(x, t) = √ exp − .
2 π 0 4(t − τ ) (t − τ )3/2
◮ Hyperbolic equations with two independent variables.

Consider the problem for the homogeneous linear hyperbolic equation
∂2w ∂w ∂2w ∂w
2
+ ϕ(x) = a(x) 2
+ b(x) + c(x)w (18.1.1.9)
∂t ∂t ∂x ∂x
with the homogeneous initial conditions
w = 0 at t = 0,
(18.1.1.10)
∂t w = 0 at t = 0,
and the boundary conditions (18.1.1.3) and (18.1.1.4).

18.1. Duhamel’s Principles in Nonstationary Problems 1235
The solution of problem (18.1.1.9), (18.1.1.10), (18.1.1.3), (18.1.1.4) with the nonsta-
tionary boundary condition (18.1.1.3) at x = x1 can be expressed by formula (18.1.1.5)
in terms of the solution u(x, t) of the auxiliary problem for Eq. (18.1.1.9) with the initial
conditions (18.1.1.10) and the boundary condition (18.1.1.4) for u instead of w, and the
simpler stationary boundary condition (18.1.1.6) at x = x1 .
In this case, Remark 18.1 remains valid.
◮ Second-order equations with several independent variables.

Duhamel’s first principle can also be used to solve homogeneous linear equations of the
parabolic or hyperbolic type with many space variables,
n
X n
X
∂k w ∂2w ∂w
k
= aij (x) + bi (x) + c(x)w, (18.1.1.11)
∂t ∂xi ∂xj ∂xi
i,j=1 i=1
where k = 1, 2 and x = {x1 , . . . , xn }.

Let V be some bounded domain in Rn with sufficiently smooth surface S = ∂V . The
solution of the boundary value problem for Eq. (18.1.1.11) in V with the homogeneous
initial conditions (18.1.1.2) if k = 1 or (18.1.1.10) if k = 2 and with the nonhomogeneous
linear boundary condition
Γx [w] = g(t) for x ∈ S (18.1.1.12)
is given by
Z t Z t
∂ ∂u
w(x, t) = u(x, t − τ ) g(τ ) dτ = (x, t − τ ) g(τ ) dτ.
∂t 0 0 ∂t
Here u(x, t) is the solution of the auxiliary problem for Eq. (18.1.1.11) with the same initial
conditions, (18.1.1.2) or (18.1.1.10) for u instead of w and with the simpler stationary
boundary condition
Γx [u] = 1 for x ∈ S.
Note that Eq. (18.1.1.12) can represent a boundary condition of the first, second, or
third kind; the coefficients of the operator Γx are assumed to be independent of t.
18.1.2 Problems for Nonhomogeneous Linear Equations

◮ Parabolic equations.
The solution of the nonhomogeneous linear equation
n
X n
X
∂w ∂2w ∂w
= aij (x) + bi (x) + c(x)w + Φ(x, t)
i,j=1 i=1
with the homogeneous initial condition (18.1.1.2) and the homogeneous boundary condi-
tion
Γx [w] = 0 for x ∈ S (18.1.2.1)
can be represented in the form (Duhamel’s second principle)

Z t
w(x, t) = U (x, t − τ, τ ) dτ. (18.1.2.2)
0
Here U (x, t, τ ) is the solution of the auxiliary problem for the homogeneous equation
n
X n
X
∂U ∂2U ∂U
= aij (x) + bi (x) + c(x)U
i,j=1 i=1
with the boundary condition (18.1.2.1), in which w must be substituted by U , and the
nonhomogeneous initial condition
U = Φ(x, τ ) at t = 0,
where τ is a parameter.
Note that Eq. (18.1.2.1) can represent a boundary condition of the first, second, or third
kind; the coefficients of the operator Γx are assumed to be independent of t.
◮ Hyperbolic equations.
The solution of the nonhomogeneous linear equation

n
X n
X
∂2w ∂w ∂2w ∂w
2
+ ϕ(x) = aij (x) + bi (x) + c(x)w + Φ(x, t)
∂t ∂t ∂xi ∂xj ∂xi
i,j=1 i=1
with the homogeneous initial conditions (18.1.1.10) and homogeneous boundary condi-
tion (18.1.2.1) can be expressed by formula (18.1.2.2) in terms of the solution U = U (x, t, τ )
of the auxiliary problem for the homogeneous equation
n
X n
X
∂2U ∂U ∂2U ∂U
2
+ ϕ(x) = aij (x) + bi (x) + c(x)U
∂t ∂t ∂xi ∂xj ∂xi
i,j=1 i=1
with the homogeneous initial and boundary conditions (18.1.1.2) and (18.1.2.1), where w
should be replaced by U , and the nonhomogeneous initial condition
∂t U = Φ(x, τ ) at t = 0,
where τ is a parameter.
Note that (18.1.2.1) can represent a boundary condition of the first, second, or third
kind.
⊙ Literature for Section 18.1: R. Courant and D. Hilbert (1989), G. A. Korn and T. M. Korn (2000), A. D. Po-
18.2. Transformations Simplifying Initial and Boundary Conditions 1237
18.2 Transformations Simplifying Initial and Boundary

Conditions
18.2.1 Transformations That Lead to Homogeneous Boundary
Conditions
A linear problem with arbitrary nonhomogeneous boundary conditions,
(k)
Γx [w] = gk (x, t) for x ∈ Sk , (18.2.1.1)
can be reduced to a linear problem with homogeneous boundary conditions. To this end,
one should make the change of variable
w(x, t) = ψ(x, t) + u(x, t), (18.2.1.2)
where u is the new unknown function, ψ is any function that satisfies the nonhomogeneous
boundary conditions (18.2.1.1), and
(k)
Γx [ψ] = gk (x, t) for x ∈ Sk . (18.2.1.3)
Table 18.1 gives examples of such transformations for linear boundary value problems
with one space variable for parabolic and hyperbolic equations. In the third boundary value
problem, it is assumed that k1 < 0 and k2 > 0.
TABLE 18.1
Simple transformations of the form w(x, t) = ψ(x, t) + u(x, t) that lead to
homogeneous boundary conditions in problems with one space variable (0 ≤ x ≤ l)
No. Problems Boundary conditions Function ψ = ψ(x, t)
First boundary w = g1 (t) at x=0 x

1 ψ = g1 (t) + g2 (t) − g1 (t)
value problem w = g2 (t) at x=l l
Second boundary ∂x w = g1 (t) at x=0 x2

2 value problem ψ = xg1 (t) + g2 (t) − g1 (t)
∂x w = g2 (t) at x=l 2l
Third boundary ∂x w + k1 w = g1 (t) at x=0 (k2 x − 1 − k2 l)g1 (t) + (1 − k1 x)g2 (t)

3 ψ=
value problem ∂x w + k2 w = g2 (t) at x=l k2 − k1 − k1 k2 l
Mixed boundary w = g1 (t) at x=0

4 ψ = g1 (t) + xg2 (t)
value problem ∂x w = g2 (t) at x=l
Mixed boundary ∂x w = g1 (t) at x=0

5 ψ = (x − l)g1 (t) + g2 (t)
value problem w = g2 (t) at x=l
Note that the selection of the function ψ is of a purely algebraic nature and is not
related to the equation in question; there are infinitely many suitable functions ψ that satisfy
condition (18.2.1.3). Transformations of the form (18.2.1.2) can often be used at the first
stage of solving boundary value problems.
18.2.2 Transformations That Lead to Homogeneous Initial

and Boundary Conditions
A linear problem with nonhomogeneous initial and boundary conditions can be reduced
to a linear problem with homogeneous initial and boundary conditions. To this end, one
should introduce a new dependent variable u by formula (18.2.1.2), where the function ψ
should satisfy nonhomogeneous initial and boundary conditions.
Below we specify some simple functions ψ that can be used in the transformation
(18.2.1.2) to obtain boundary value problems with homogeneous initial and boundary con-
ditions. To be specific, we consider a parabolic equation with one space variable and the
general initial condition
w = f (x) at t = 0. (18.2.2.1)
1. First boundary value problem : the initial condition is (18.2.2.1), and the bound-
ary conditions are given in row 1 of Table 18.1. Suppose that the initial and boundary
conditions are compatible; i.e., f (0) = g1 (0) and f (l) = g2 (0). Then in the transforma-
tion (18.2.1.2) one can take
x
ψ(x, t) = f (x) + g1 (t) − g1 (0) + g2 (t) − g1 (t) + g1 (0) − g2 (0) .
l
2. Second boundary value problem : the initial condition is (18.2.2.1), and the bound-
conditions are compatible; i.e., f ′ (0) = g1 (0) and f ′ (l) = g2 (0). Then in the transforma-
tion (18.2.1.2) one can set
x2
ψ(x, t) = f (x) + x g1 (t) − g1 (0) + g2 (t) − g1 (t) + g1 (0) − g2 (0) .
2l
3. Third boundary value problem : the initial condition is (18.2.2.1), and the boundary
conditions are given in row 3 of Table 18.1. If the initial and boundary conditions are
compatible, then in the transformation (18.2.1.2) one can take
(k2 x − 1 − k2 l)[g1 (t) − g1 (0)] + (1 − k1 x)[g2 (t) − g2 (0)]
ψ(x, t) = f (x) + ,
k2 − k1 − k1 k2 l
where k1 < 0 and k2 > 0.
4. Mixed boundary value problem : the initial condition is (18.2.2.1), and the bound-
conditions are compatible; i.e., f (0) = g1 (0) and f ′ (l) = g2 (0). Then in the transforma-
tion (18.2.1.2) one can set

ψ(x, t) = f (x) + g1 (t) − g1 (0) + x g2 (t) − g2 (0) .
5. Mixed boundary value problem : the initial condition is (18.2.2.1), and the bound-
conditions are compatible; i.e., f ′ (0) = g1 (0) and f (l) = g2 (0). Then in the transforma-
tion (18.2.1.2) one can take

ψ(x, t) = f (x) + (x − l) g1 (t) − g1 (0) + g2 (t) − g2 (0).
⊙ Literature for Section 18.2: A. N. Tikhonov and A. A. Samarskii (1990), A. D. Polyanin (2002), A. D. Po-
lyanin and A. V. Manzhirov (2007).
Chapter 19
Systems of Linear Coupled PDEs.

Decomposition Methods
19.1 Asymmetric and Symmetric Decompositions

19.1.1 Asymmetric Decomposition. Order of Decomposition
◮ Preliminary remarks. Systems of coupled equations to be considered.
Linear systems of coupled equations arise in various fields of continuum mechanics and
physics (elasticity, thermoelasticity, porous elasticity, hydrodynamics of viscous and vis-
coelastic incompressible fluids as well as of viscous compressible barotropic fluids and
gases, electrodynamics, etc.).
For a linear system of coupled equations, splitting the system into a few simpler sub-
systems (or, best of all, reducing it to several independent equations) is considered a great
achievement. A representation of solutions of a system of coupled equations via solutions
of independent (uncoupled) equations will be referred to as a complete decomposition of
the original system; the representation of solutions of a system of coupled equations via
solutions of a few simpler equations (of which only part is independent) will be referred to
as an incomplete decomposition, or a partial decomposition.
Decomposition dramatically simplifies the qualitative study and interpretation of the
most important physical properties of coupled 3D equations, thus permitting an efficient
study of their wave and dissipative properties. Moreover, in a number of cases, decompo-
sition permits one to find exact analytical solutions of the corresponding boundary value
problems and initial-boundary value problems and substantially simplifies the application
of numerical methods, which permits one to use the corresponding standard routines for
simpler independent equations or subsystems.
In this section, we consider systems involving one vector equation and one scalar equa-
tion of the form
L[u] + ∇K[u, p] = f, (19.1.1.1)
M[u, p] = f4 , (19.1.1.2)
where u = u(x, t) = (u1 , u2 , u3 ) and p = p(x, t) are the unknown functions, x = (x, y, z),
f = f(x, t) = (f1 , f2 , f3 ) and f4 = f4 (x, t) are given functions, L, K and M are constant
1239
1240 S YSTEMS OF L INEAR C OUPLED PDE S . D ECOMPOSITION M ETHODS
coefficient linear differential operators in (x, y, z, t), (x, y, z) are Cartesian coordinates,
and ∇ = (∂x , ∂y , ∂z ) is the gradient operator.
In the special case of K[u, p] = K[u], system (19.1.1.1)–(19.1.1.2) may consist of a
single vector equation for u; the absence of the scalar equation (19.1.1.2) corresponds to
M[u, p] ≡ 0 and f4 ≡ 0.
Linear systems of coupled equations of the form (19.1.1.1)–(19.1.1.2) occur in several
other areas of continuum mechanics and physics (see the summary Table 19.1).
1◦ . Such systems describe the slow motion of viscous incompressible and viscous com-
pressible barotropic fluids (see rows 1–4 in Table 19.1), where u is the fluid velocity, p is
the pressure, ρ is the fluid density, ν is the kinematic viscosity, b is the unperturbed ve-
locity in the incoming flow, µ and λ are the dynamic coefficients of viscosity, and k is the
compressibility factor.
2◦ . The operators L and K corresponding to the Maxwell viscoelastic fluid equations (for
creeping flows) as well as to the linearized hyperbolic Navier–Stokes equations are given
in row 5 in Table 19.1. Here τ is the relaxation time, and the remaining notation is the same
as in row 1 (see above).
For row 6 in Table 19.1, the operator L corresponding to various models of incompress-
ible viscoelastic fluids is given in Table 12.2.
3◦ . Systems of the form (19.1.1.1)–(19.1.1.2) are widely used in the linear theory of elas-
ticity, thermoelasticity, and thermoviscoelasticity (see rows 7–13 in Table 19.1), where u is
the displacement field, ρ is the medium density, p = T is the temperature (in rows 8–10), β
is the thermomechanical modulus, λ and µ (λ0 and µ0 ) are the Lamé elastic moduli, λ1 and
µ1 are the viscosity coefficients, Iλ and Iµ are the relaxation integral operators, and a and
a are thermal diffusivities. The last column in row 9 is an equation based on the Cattaneo
e
hyperbolic heat conduction model, and the last column of row 10 corresponds to nondis-
sipative thermoelasticity of the Green–Naghdi type (see Section 12.15), which describes
the effect of the second sound. (This effect arises, for example, in liquid helium and some
solids at low temperatures.)
4◦ . Systems of the form (19.1.1.1)–(19.1.1.2) also occur in electrodynamics (see rows 14
and 15 in Table 19.1).
◮ Asymmetric decomposition. The order of decomposition

Every solution of system (19.1.1.1)–(19.1.1.2) can be represented in the form
u = ∇ϕ + e2 v2 + e3 v3 , p = p, (19.1.1.3)
where e2 and e3 are the unit vectors corresponding to the Cartesian coordinates y and z, two
scalar functions v2 = v2 (x, t) and v3 = v3 (x, t) satisfy two independent linear equations of
the same type,
L [v2 ] = f2 − ∂y F, L [v3 ] = f3 − ∂z F, (19.1.1.4)

Z x
F = F (x, t) = F (x, y, z, t) = f1 (x1 , y, z, t) dx1 , (19.1.1.5)
0
19.1. Asymmetric and Symmetric Decompositions 1241
TABLE 19.1
Various linear systems of the form (19.1.1.1)–(19.1.1.2), occurring
in continuum mechanics and physics. Notation: u = (u1 , u2 , u3 )
No. Equation names Operator L[u] Operator K[u, p] Equation (19.1.1.2)
Stokes (viscous
1 incompressible fluid), ut − ν∆u p/ρ div u = 0
see Eqs. (1)–(2)
in Section 12.7
Oseen (viscous
2 incompressible fluid), ut + bux − ν∆u p/ρ div u = 0
see Eqs. (1)
in Section 12.8
Stokes (viscous
3 compressible fluid), ρ0 ut − µ∆u p − (µ + λ) div u kpt + ρ0 div u = 0
see Eqs. (1)–(2)
in Section 12.12
Stokes (viscous
compressible fluid), −c2 div u
4 utt − ν∆ut Absent
see Eqs. (14) −(ν + κ) div ut
in Section 12.12
Maxwell (viscoelastic
5 incompressible fluid), τ utt + ut − ν∆u p/ρ div u = 0
see Eqs. (1)
in Section 12.9
Viscoelastic
6 L[u] p/ρ div u = 0
incompressible fluid (any operator)
(general model)
Navier
7 (linear elasticity), ρutt − µ∆u −(µ + λ) div u Absent
see Eqs. (1)
in Section 12.6
Thermoelasticity,
8 see Eqs. (1)–(2) ρutt − µ∆u αp − (µ + λ) div u pt = a∆p − β (div u)t
in Section 12.14;
p is the temperature
Thermoelasticity
(with hyperbolic τ ptt + pt
9 ρutt − µ∆u αp − (µ + λ) div u
heat conductivity); = a∆p − β (div u)t
Thermoelasticity
(Green–Naghdi model),
10 see Eqs. (1)–(2) ρutt − µ∆u αp − (µ + λ) div u ptt = e
a∆p − β (div u)tt
in Section 12.15;
Linear viscoelasticity, −(µ0 + λ0 ) div u
11 ρutt − µ0 ∆u Absent
see Eqs. (1) −µ1 ∆ut −(µ1 + λ1 ) div ut
in Section 12.16
12 Linear viscoelasticity ρutt − Iµ [∆u] −(Iµ + Iλ )[div u] Absent

(general model)
TABLE 19.1 (continued)

Various linear systems of the form (19.1.1.1)–(19.1.1.2), occurred
in continuum mechanics and physics. Notation: u = (u1 , u2 , u3 )
No. Equation names Operator L[u] Operator K[u, p] Equation (19.1.1.2)
Linear
13 thermoviscoelasticity ρutt − Iµ [∆u] αp − (Iµ + Iλ )[div u] pt = a∆p − β (div u)t
(general model)
Maxwell
14 (electrodynamics), µ(εutt + λut ) − ∆u 0 div u = 0
see Eqs. (5) with H = u
in Section 12.17
Maxwell
15 (electrodynamics), µ(εutt + λut ) − ∆u div u div u = ρ/ε
see Eqs. (6) with E = u
in Section 12.17
and the functions ϕ = ϕ(x, t) and p = p(x, t) are determined from the system of equations
L [ϕ] + K [∇ϕ + e2 v2 + e3 v3 , p] = F, (19.1.1.6)
M [∇ϕ + e2 v2 + e3 v3 , p] = f4 . (19.1.1.7)
To avoid excessive change of notation in (19.1.1.3) and in what follows, we write p = p
instead of p = pe, where pe is the new unknown function.
System (19.1.1.4)–(19.1.1.7) consists of two independent linear equations (19.1.1.4)
and the subsystem of two coupled equations (19.1.1.6)–(19.1.1.7) and is substantially sim-
pler than the original system of four coupled linear equations (19.1.1.1)–(19.1.1.2).
The representation (19.1.1.3) of components of the vector u includes first derivatives
of the new unknown function ϕ, which corresponds to a first-order decomposition. In the
general case, the decomposition order is defined as the maximum order of derivatives of the
new unknown functions occurring in the representation of the components of the vector u.
Formulas (19.1.1.3) and (19.1.1.5) and Eqs. (19.1.1.4), (19.1.1.6), and (19.1.1.7) deter-
mine a first-order partial decomposition of the linear system (19.1.1.1)–(19.1.1.2).
19.1.2 Symmetric Decomposition. Invariant Transformations

◮ Symmetric decomposition
Every solution of system (19.1.1.1)–(19.1.1.2) can also be represented in the symmetric
form
u = ∇ϕ + v, p = p, (19.1.2.1)
where the vector function v = (v1 , v2 , v3 ) satisfies the independent linear equation
L[v] = f − ∇G (19.1.2.2)
and the functions ϕ and p are described by the system of equations
L [ϕ] + K [∇ϕ + v, p] = G, (19.1.2.3)
M [∇ϕ + v, p] = f4 . (19.1.2.4)
19.1. Asymmetric and Symmetric Decompositions 1243
Equations (19.1.2.3)–(19.1.2.4) contain an arbitrary scalar function G = G(x, t).

Formulas (19.1.2.1) and Eqs. (19.1.2.2)–(19.1.2.4) determine a first-order partial de-
composition of the linear system (19.1.1.1)–(19.1.1.2).
Remark 19.1. The representation (19.1.2.1) of the components of the vector v contains one
extra (additional) unknown function compared with the representation (19.1.1.3). This permits
slightly simplifying Eqs. (19.1.2.2)–(19.1.2.4) by imposing an additional condition on the com-
ponents (v1 , v2 , v3 ) and by choosing an appropriate function G. In particular, without loss of gen-
erality one can set v1 = 0 and G = F (i.e., f1 = Gx ) in (19.1.2.2)–(19.1.2.4), which results in the
representation (19.1.1.3)–(19.1.1.7).
Linear equations of continuum mechanics and physics are special cases of system
(19.1.1.1)–(19.1.1.2) of the special form

L[u] + ∇ σp + K1 [div u] = f, (19.1.2.5)
M1 [p] + M2 [div u] = f4 , (19.1.2.6)
where σ is a constant and K1 , M1 , and M2 are linear differential, integral, or integro-

differential operators in t, x, y, and z.
Every solution of system (19.1.2.5)–(19.1.2.6) can be represented in the form (19.1.2.1)
with the vector v = (v1 , v2 , v3 ) satisfying the independent linear equation (19.1.2.2), the
functions ϕ and p described by the system of equations
L[ϕ] + σp + K1 [∆ϕ] + K1 [div v] = G, (19.1.2.7)

M1 [p] + M2 [∆ϕ] + M2 [div v] = f4 , (19.1.2.8)
and G = G(x, t) being an arbitrary function.

These equations contain the function div v. In what follows, using Remark 19.1, we
simplify (19.1.2.7)–(19.1.2.8) by imposing the additional condition
div v = 0 (19.1.2.9)
(a differential constraint) on the components of the vector v.

We point out that there are various ways to satisfy Eq. (19.1.2.9) (see Sections 19.2.1
and 19.2.2 below), which generate various types of decompositions of the system in ques-
tion and lead to various numbers of the new unknown functions.
1◦ . Equations (19.1.2.5)–(19.1.2.6) are invariant under the transformation
u + curl Ψ◦ ,
u=e p = pe,

2◦ . Equations (19.1.2.5)–(19.1.2.6) with σ 6= 0 are invariant under the transformation
1
u=e
u + ∇Φ, p = pe − (L + K1 ∆)[Φ],
σ
where the function Φ is an arbitrary solution of the equation
(M1 L + M1 K1 ∆ − σM2 ∆)[Φ] = 0.
3◦ . Equations (19.1.2.5)–(19.1.2.6) with σ = 0 are invariant under the transformation
u=e
u + ∇Φ, p = pe + Θ,
where the functions Φ and Θ are arbitrary solutions of the equations
(L + K1 ∆)[Φ] = 0, M1 [Θ] + M2 [Φ] = 0.
19.2 First-Order Decompositions. Examples

19.2.1 Systems of Linear PDEs without Mass Forces (f = 0)
◮ Decomposition on the basis of two stream functions.
In the absence of mass forces (f = 0), it is convenient to set G = 0 in (19.1.2.2) and

(19.1.2.7). In this case, the right-hand side of the vector equation (19.1.2.2) is zero, which
gives three identical homogeneous equations for the components v1 , v2 , and v3 .
Let ψ (1) and ψ (2) be arbitrary solutions of two identical linear homogeneous equations
L [ψ (1) ] = 0, L [ψ (2) ] = 0. (19.2.1.1)
Then the formulas
v1 = ψy(1) , v2 = −ψx(1) + ψz(2) , v3 = −ψy(2) , (19.2.1.2)
where ψ (1) and ψ (2) are arbitrary solutions of Eqs. (19.2.1.1), give a general representation
of the solution of the homogeneous vector equation (19.1.2.2) (with f = 0 and G = 0) and
the scalar equation (19.1.2.9).
Every solution of system (19.1.2.5)–(19.1.2.6) with f = 0 can be represented by formu-
las (19.1.2.1) and (19.2.1.2), where the functions ψ (1) and ψ (2) satisfy Eqs. (19.2.1.1) and
the functions ϕ and p are described by the equations
L [ϕ] + σp + K1 [∆ϕ] = 0, (19.2.1.3)

M1 [p] + M2 [∆ϕ] = f4 . (19.2.1.4)
These equations are a consequence of (19.1.2.7)–(19.1.2.9) with G = 0.

Remark 19.2. Formulas (19.2.1.2), where ψ (1) and ψ (2) are arbitrary functions, give the general
solution of Eq. (19.1.2.9). The functions ψ (1) and ψ (2) can be interpreted as two stream functions
that permit one to eliminate the continuity equation from the 3D incompressible fluid equations
(where the fluid velocity is denoted by v). In the special case of ψ (2) = 0, in (19.2.1.2) we obtain
the usual representation of the fluid velocity components for 2D planar flows (v3 = 0) with a single
stream function.
19.2. First-Order Decompositions. Examples 1245
For σ 6= 0, one can eliminate p from (19.2.1.4) with the use of (19.2.1.3) and obtain a
separate (independent) equation for ϕ,
−M1 L [ϕ] + (σM2 − M1 K1 ) [∆ϕ] = σf4 . (19.2.1.5)
Here p can be found without quadratures by the formula
1
p=− L [ϕ] + K1 [∆ϕ] . (19.2.1.6)
σ
Thus, the general solution of system (19.1.2.5)–(19.1.2.6) with f = 0 consisting of four
3D coupled equations can be expressed via the solution of three independent equations
(two identical equations of the form (19.2.1.1) for the stream functions ψ (1) and ψ (2) and
Eq. (19.2.1.5) for ϕ). The above-described representation of the solution corresponds to a
first-order complete decomposition of the original system.
Example 19.1. The magnetic field intensity H in a medium is described by the overdetermined
system of homogeneous equations (which are a corollary of the Maxwell equations)
µ(εHtt + λHt ) − ∆H = 0,
(19.2.1.7)
div H = 0,
where ε and µ are the dielectric constant and the magnetic permeability of the medium and λ is the
medium conductivity. (For a perfect dielectric, i.e., for a nonconducting medium, one should set
λ = 0.)
Equations (19.2.1.7) are a special case of the overdetermined system consisting of the homoge-
neous vector equation (19.1.2.2) and the scalar equation (19.1.2.9) with
v = H, L[v] = µ(εvtt + λvt ) − ∆v, f = 0, G = 0.
Hence the general solution of system (19.2.1.7) is determined by formulas (19.2.1.2), where ψ (1)
and ψ (2) are arbitrary solutions of Eqs. (19.2.1.1) with L[ψ] = µ(εψtt + λψt ) − ∆ψ.
Example 19.2. The Stokes equations describing slow motions of a viscous compressible baro-
tropic fluid in the absence of mass forces have the form
ρ0 ut = −∇p + µ∆u + (µ + λ)∇ div u,
(19.2.1.8)
pt + c2 ρ0 div u = 0,
where u = (u1 , u2 , u3 ) is the fluid velocity, ρ0 ispthe unperturbed density, p is the pressure, µ and λ
are the dynamic viscosity coefficients, and c = p′ (ρ) |ρ=ρ0 is the speed of sound.
Equations (19.2.1.8) are the special case of Eqs. (19.1.2.5)–(19.1.2.6) for
L[u] = ρ0 ut − µ∆u, K1 [q] = −(µ + λ)q, M1 [p] = pt , M2 [q] = ρ0 c2 q,
(19.2.1.9)
σ = 1, f = 0, f4 = 0.
Hence the solution of these equations can be represented by formulas (19.1.2.1) and (19.2.1.2),
where the unknown functions ψ (1) , ψ (2) , and ϕ satisfy the three independent equations
(1) (2)
ρ0 ψt − µ∆ψ (1) = 0, ρ0 ψt − µ∆ψ (2) = 0, (19.2.1.10)
2
ρ0 ϕtt − (2µ + λ)∆ϕt − ρ0 c ∆ϕ = 0 (19.2.1.11)
and the pressure p is expressed via ϕ by the formula
p = (2µ + λ)∆ϕ − ρ0 ϕt . (19.2.1.12)
To derive Eqs. (19.2.1.10)–(19.2.1.11) and formula (19.2.1.12), one substitutes relations (19.2.1.9)
into (19.2.1.1), (19.2.1.5), and (19.2.1.6).
◮ The summary table of equations for determining the functions p and ϕ.
Table 19.2 shows the final equations (19.1.2.3) and (19.1.2.4) with G = 0 which are stated
for the functions p and ϕ for systems (19.1.1.1)–(19.1.1.2) listed in Table 19.1 in the ab-
sence of mass forces (f = 0). The numbering of systems in Tables 19.1 and 19.2 is the
same. The components v1 , v2 , v3 of the vector function v are determined independently by
formulas (19.2.1.2), where ψ (1) and ψ (2) are arbitrary solutions of Eq. (19.2.1.1), and the
unknowns u1 , u2 , and u3 can be computed from Eq. (19.1.2.1). In the last column, we use
the notation ψ for both ψ (1) and ψ (2) .
In what follows, some comments and examples pertaining to Table 19.2 are considered.
1◦ . For viscous and viscoelastic incompressible fluids (see rows 1, 2 and 5, 6 in Table 19.2),
the pressure is determined without quadratures. In the right part of the corresponding for-
mulas for p, one can add an arbitrary function p0 (t) of time.
2◦ . For all systems listed in Table 19.2 and containing the pressure p, it is possible to obtain
a separate equation for ϕ and a formula for p.
3◦ . Let us simplify the equations for ψ given in line 4 of Table 19.2. To this end, we
integrate it with respect to time t. As a result, we obtain
ψt − ν∆ψ = q(t), (19.2.1.13)
where q(t) is an arbitrary function. The general solution of Eq. (19.2.1.13) can be repre-
sented as follows: Z
e
ψ = ψ + q(t) dt, (19.2.1.14)
where ψe is the general solution of the homogeneous equation (19.2.1.13) with q(t) ≡ 0. For-
mula (19.2.1.2) determining the components of the velocity (19.1.2.1) includes the partial
derivatives of ψ with respect to the spatial variables alone. The second term in (19.2.1.14),
which depends on time and does not depend on spatial coordinates, does not affect the final
result. Therefore, one can set q(t) ≡ 0 in Eq. (19.2.1.13) without loss of generality.
4◦ . For the solvability of the overdetermined system for the electric field (see Eqs. (6) in
Section 12.17) two equations for the function ϕ (see row 15 in Table 19.2) have to be
consistent,
εϕtt + λϕt = 0, ∆ϕ = ρ/ε. (19.2.1.15)
The consistency condition (19.2.1.15) specifies the permissible density distribution of the
electric charge, which is determined by formula (4) in Section 12.17. In this case, the
consistent solution of Eqs. (19.2.1.15) has the form
ϕ = ϕ0 (x) + ϕ1 (x) exp(−λt/ε),
where the functions ϕ0 (x) and ϕ1 (x) satisfied the Laplace and Poisson equations
∆ϕ0 = 0, ∆ϕ1 = f1 (x)/ε.

TABLE 19.2
Equations for functions p and ϕ for systems of the form (19.1.1.1)–(19.1.1.2), occurring in continuum
mechanics and physics. The operators L and K as well as equation (19.1.1.2) are described in Table 19.1.
No. Equation names Equation (19.1.2.3) Equation (19.1.2.4) Equations (19.2.1.1)
Stokes (viscous
1 incompressible p = −ρ(ϕt − ν∆ϕ) ∆ϕ = 0 ψt = ν∆ψ
fluid)
Oseen (viscous
2 p = −ρ(ϕt ∆ϕ = 0 ψt + bψx = ν∆ψ
incompressible + bϕx − ν∆ϕ)
fluid)
Stokes (viscous
p = −ρ0 ϕt
3 compressible kpt + ρ0 ∆ϕ = 0 ρ0 ψt = µ∆ψ
+ (2µ + λ)∆ϕ
fluid)
Stokes (viscous
ϕtt − c2 ∆ϕ
4 compressible Absent ψtt = ν∆ψt
− (2ν + κ)∆ϕt = 0
fluid)
Maxwell
(viscoelastic p = −ρ(τ ϕtt
5 ∆ϕ = 0 τ ψtt + ψt = ν∆ψ
incompressible + ϕt − ν∆ϕ)
fluid)
Viscoelastic
6 incompressible fluid p = −ρL[ϕ] ∆ϕ = 0 L[ψ] = 0
(general model)
7 Navier ρϕtt = (2µ + λ)∆ϕ Absent ρψtt = µ∆ψ

(linear elasticity)
Thermoelasticity; αp = −ρϕtt
8 pt = a∆p − β∆ϕt ρψtt = µ∆ψ
p is the temperature + (2µ + λ)∆ϕ
Thermoelasticity
(with hyperbolic αp = −ρϕtt τ ptt + pt
9 ρψtt = µ∆ψ
heat conductivity); + (2µ + λ)∆ϕ = a∆p − β∆ϕt
Thermoelasticity
(Green–Naghdi αp = −ρϕtt ptt − e
a∆p
10 ρψtt = µ∆ψ
model); + (2µ + λ)∆ϕ = −β∆ϕtt
Linear ρϕtt = (µ0 + 2λ0 )∆ϕ ρψtt = µ0 ∆ψ
11 Absent
viscoelasticity + (µ1 + 2λ1 )∆ϕt + µ1 ∆ψt
Linear
12 viscoelasticity ρϕtt = (Iµ + 2Iλ )[∆ϕ] Absent ρψtt = Iµ [∆ψ]
(general model)
Linear
13 αp = −ρϕtt pt = a∆p − β∆ϕt ρψtt = Iµ [∆ψ]
thermoviscoelasticity + (Iµ + 2Iλ )[∆ϕ]
(general model)
14 Maxwell Satisfied identically ϕ=0 µ(εψtt + λψt ) = ∆ψ

(electrodynamics)
15 Maxwell εϕtt + λϕt = 0 ∆ϕ = ρ/ε µ(εψtt + λψt ) = ∆ψ

(electrodynamics)
◮ Using the Stokes–Helmholtz representation of the vector u.

The solution is sought in the form
u = ∇ϕ + curl Ψ, p = p, (19.2.1.16)
where ϕ and Ψ are the new unknown scalar and vector function, respectively. In this case,
v = curl Ψ, and hence the additional condition (19.1.2.9) is satisfied.
By substituting (19.2.1.16) into system (19.1.2.5)–(19.1.2.6) with f = 0, we obtain the
vector equation
L[Ψ] = 0 (19.2.1.17)
and two equations (19.2.1.3)–(19.2.1.4).
The representation via the stream functions in the form (19.1.2.1), (19.2.1.2) is a special
case of the Stokes–Helmholtz decomposition (19.2.1.16) with one zero component of the
vector Ψ,
Ψ = (Ψ1 , 0, Ψ3 ), Ψ1 = ψ (2) , Ψ3 = ψ (1) . (19.2.1.18)
Therefore, the representation (19.1.2.1), (19.2.1.2) can be referred to as the truncated
Stokes–Helmholtz decomposition.
We point out that the Stokes–Helmholtz representation of the vector u leads to the vector
equation (19.2.1.17), which is equivalent to three independent scalar equations for its com-
ponents Ψ1 , Ψ2 , and Ψ3 , while the truncated Stokes–Helmholtz decomposition (19.1.2.1),
(19.2.1.2) gives two scalar equations for the stream functions ψ (1) and ψ (2) (i.e., fewer
equations).
◮ Special types of decompositions using representations of the vector u via three

scalar functions.
Let us describe four decompositions of system (19.1.2.5)–(19.1.2.6) with f = 0 based on
the representation of solutions with the use of the formulas
u = ∇ϕ + curl Ψn , p = p, (19.2.1.19)
where the vector functions Ψn (n = 1, 2, 3, 4) are expressed in a special way via two
scalar functions ψ (1) and ψ (2) ,
1. Ψ1 = aψ (1) + bψ (2) (a · b 6= 0),

2. Ψ2 = aψ (1) + curl(bψ (2) ),
3. Ψ3 = aψ (1) + xψ (2) ,
4. Ψ4 = xψ (1) + curl(xψ (2) ).
Here a and b are arbitrary nonzero constant vectors. The vector functions Ψ3 and Ψ4 can
be used for systems (19.1.2.5)–(19.1.2.6) with an operator L of the special form
L[u] = L1 [u] + L2 [∆u],

where L1 and L2 are linear differential operators in time t. For operators of this form,
the following two identities have been used in the derivation of the definitive results stated
below:
∆(xg) ≡ x∆g + 2∇g, curl[∆(xg)] ≡ curl(x∆g),
where g is an arbitrary scalar function.

For all solutions of system (19.1.2.5)–(19.1.2.6) based on formula (19.2.1.19) and for
four distinct representations of the vector function Ψn , the unknown functions ψ (1) and
ψ (2) satisfy two independent (identical) equations (19.2.1.1), and the functions ϕ and p are
described by Eqs. (19.2.1.3)–(19.2.1.4).
Note that the above-described four special representations of solutions determined by
formula (19.2.1.19) contain one unknown function less than the Stokes–Helmholtz repre-
sentation (19.2.1.16).
19.2.2 Systems of Linear PDEs with Mass Forces

◮ Simplest case of potential mass forces.
If the mass forces are potential, i.e., if
f = ∇F, (19.2.2.1)
then one should set G = F in (19.1.2.2)–(19.1.2.3). Then the right-hand sides of equations
(19.1.2.2) are zero, just as in the absence of mass forces.
The solution of system (19.1.2.5)–(19.1.2.6), (19.2.2.1) can be represented by the for-
mulas (19.1.2.1) and (19.2.1.2), where the functions ψ (1) and ψ (2) satisfy Eqs. (19.2.1.1)
and the functions ϕ and p are described by Eq. (19.2.1.3) whose right-hand side is supple-
mented with the function F and by Eq. (19.2.1.4).
Let us use the Stokes–Helmholtz representation of the mass force in the form of the sum of
a potential and a solenoidal vector,
f = ∇γ + curl ω. (19.2.2.2)
Remark 19.3. If the vector function f is given, then the scalar function γ and the vector function
ω can be sought, say, in the form
γ = div U, ω = − curl U. (19.2.2.3)
By substituting (19.2.2.3) into (19.2.2.2) and by taking into account the identity curl curl U =
∇ div U − ∆U, we obtain the Poisson vector equation
∆U = f (19.2.2.4)
for the Lamé vector potential U.

We seek a solution of system (19.1.2.5)–(19.1.2.6) in the form
u = ∇ϕ + curl Ψ, p = p. (19.2.2.5)
As a result, in view of (19.2.2.2), for the unknown variables ϕ, Ψ, and p we obtain the
equations
L[Ψ] = ω,
L[ϕ] + σp + K1 [∆ϕ] = γ, (19.2.2.6)
M1 [p] + M2 [∆ϕ] = f4 .
For σ 6= 0, one can eliminate p from the last two equations in (19.2.2.6) and obtain an
independent equation for ϕ. The function p can be expressed via ϕ without quadratures
with the help of the second equation in (19.2.2.6). Thus, in this case we have a complete
decomposition of system (19.1.2.5)–(19.1.2.6).
Example 19.3. The coupled equations of linear elastodynamics in the small-strain approxima-
tion for an isotropic medium read
ρutt = µ∆u + (λ + µ)∇ div u + ρf, (19.2.2.7)
where u = (u1 , u2 , u3 ) is the displacement vector determining the strain of the body, t is time,
ρ is the medium density, λ and µ are the Lamé elastic moduli (µ is the shear modulus), and f =
(f1 , f2 , f3 ) is the mass force.
The division by ρ reduces system (19.2.2.7) to the special case of system (19.1.2.5)–(19.1.2.6)
with
L[u] = utt − c22 ∆u, σ = 0, K1 [q] = (c22 − c21 )q, M1 = M2 = 0, f4 = 0, (19.2.2.8)

p
where c1 = (λ + 2µ)/ρ is the velocity of longitudinal waves p(or p-waves, or primary waves,
because these waves are registered first in geophysics) and c2 = µ/ρ is the velocity of transverse
waves (or s-waves, or secondary waves).
Let the mass force be written in the form of the Stokes–Helmholtz decomposition (19.2.2.2).
Then the solution of the coupled equations (19.2.2.7) can be represented by the first formula in
(19.2.2.5) for u. By substituting (19.2.2.8) into (19.2.2.6), we obtain inhomogeneous wave equa-
tions for the scalar and vector functions ϕ and Ψ,
1 [ϕ] = γ, 2 [Ψ] = ω. (19.2.2.9)
Here and in what follows, the d’Alembert operators 1 and 2 are defined by
1 ≡ ∂t2 − c21 ∆, 2 ≡ ∂t2 − c22 ∆. (19.2.2.10)
The first formula in (19.2.2.5) and Eqs. (19.2.2.9) are called the Green–Lamé representation and
provide a complete decomposition of system (19.2.2.7).
Example 19.4. The Stokes equations, which describe slow motions of a viscous incompressible
fluid, have the form
1
ut = − ∇p + ν∆u + f,
ρ (19.2.2.11)
div u = 0,
where u = (u1 , u2 , u3 ), u1 , u2 , and u3 are the fluid velocity components in the Cartesian co-
ordinates, t is time, ρ is the fluid density, p is the pressure, ν is the kinematic viscosity, and
f = (f1 , f2 , f3 ) is the mass force.
19.3. Higher-Order Decompositions 1251
System (19.2.2.11) can be rewritten in the form (19.1.2.5)–(19.1.2.6) with
L[u] = ut − ν∆u, σ = 1/ρ, K1 = M1 = 0, M2 [q] = q, f4 = 0. (19.2.2.12)
Assume that the mass force is given by the Stokes–Helmholtz representation (19.2.2.2). The
solution of the Stokes equations (19.2.2.11) can be represented in the form (19.2.2.5). By substitut-
ing (19.2.2.12) into the first and third equations in (19.2.2.6), we obtain independent equations for
Ψ and ϕ,
Ψt − ν∆Ψ = ω,
(19.2.2.13)
∆ϕ = 0.
We derive the formula
p = ρ(ν∆ϕ − ϕt + γ) = ρ(γ − ϕt ) (19.2.2.14)
expressing the pressure p via the function ϕ from the second equation in (19.2.2.6) with regard
to (19.2.2.12).
One can add an arbitrary function p0 (t) to the right-hand side of this formula.
Formulas (19.2.2.5) and (19.2.2.14) and Eqs. (19.2.2.13) determine a complete decomposition
of system (19.2.2.11).
19.3 Higher-Order Decompositions

19.3.1 Decomposition of Systems Consisting of One Vector
Equation
Consider the vector equation
L[u] + ∇K[div u] = f, (19.3.1.1)
which is a system of three coupled scalar equations. Here we assume that L and K are
linear differential operators with constant coefficients.
We seek a solution of the vector equation (19.3.1.1) in the form
u = Q1 [w] + ∇Q2 [div w], (19.3.1.2)
where w is the new unknown vector function and Q1 and Q2 are linear differential operators
with constant coefficients to be determined.
We substitute (19.3.1.2) into (19.3.1.1) with
regard to the relation div u = Q1 + ∆Q2 [div w] and obtain, after some transformations
and a rearrangement of terms to which the gradient operator is applied,

LQ1 [w] + ∇ KQ1 + (L + ∆K)Q2 [div w] = f. (19.3.1.3)
By equating the expression in brackets after the gradient symbol to zero, we obtain the
operator equation KQ1 + (L + ∆K)Q2 = 0, whose solution can be represented in the form
Q1 = L + ∆K, Q2 = −K.
By substituting these operators into (19.3.1.2) and (19.3.1.3), we obtain the following rep-
resentation of the solution of Eq. (19.3.1.1):

u = L + ∆K [w] − ∇K[div w], (19.3.1.4)

L L + ∆K [w] = f. (19.3.1.5)
The representation of the solution in the form (19.3.1.4) depends on three differential
operators L, K, and ∆. The order of these operators determines the order of the decom-
position. The representation (19.3.1.4) of the solution does not require the preliminary
decomposition (19.2.2.2) of the mass force f into potential and solenoidal parts. The vector
equation (19.3.1.5) consists of three independent scalar equations; i.e., we have obtained a
complete decomposition of the original system (19.3.1.1) in this case. Other equivalent rep-
resentations of solutions can be obtained from (19.3.1.4) and (19.3.1.5) by the substitution
w = R[e w], where R is a linear operator.
Example 19.5. The vector equation (19.2.2.7) of elasticity theory is the special case of equation
(19.3.1.1), in which the determining operators have the form
L[u] = utt − c22 ∆u, K[q] = (c22 − c21 )q. (19.3.1.6)
By substituting (19.3.1.6) into (19.3.1.4)–(19.3.1.5), we obtain the Cauchy–Kovalevskaya solution
u = 1 [w] + (c21 − c22 )∇ div w, (19.3.1.7)
2 1 [w] = f. (19.3.1.8)
The d’Alembert operators 1 and 2 are defined in (19.2.2.10).
Remark 19.4. The general solution of the homogeneous equation (19.3.1.5) (for f = 0) can be
represented as the sum w = w1 + w2 , where w1 and w2 are arbitrary solutions of the two simpler
equations

L[w1 ] = 0, (L + ∆K [w2 ] = 0.
19.3.2 Decomposition of Systems Consisting of a Vector Equation

and a Scalar Equation (the First Approach)
Consider the system consisting of the vector and scalar equations (19.1.2.5)–(19.1.2.6),
where σ 6= 0 and L, K1 , M1 , and M2 are linear differential operators with constant coeffi-
cients (M1 6≡ 0 and M2 6≡ 0).
We seek the solution in the form

u = Q1 [w] + ∇ ϕ + Q2 [div w] , p = Q3 [ϕ] + Q4 [div w], (19.3.2.1)
where w and ϕ are the new unknown vector and scalar functions and Q1 , Q2 , Q3 , and
Q4 are linear differential operators with constant coefficients to be determined. We substi-
tute (19.3.2.1) into (19.1.2.5)–(19.1.2.6), take into account the relation

div u = ∆ϕ + Q1 + ∆Q2 [div w],
19.3. Higher-Order Decompositions 1253
and obtain, after some transformations and a rearrangement of terms to which the gradient
operator is applied,
n
LQ1 [w] + ∇ σQ3 [ϕ] + L[ϕ] + ∆K1 [ϕ]
| {z }
o
+ σQ4 [div w] + LQ2 [div w] + K1 Q1 + ∆Q2 )[div w] = f,
| {z }

M1 Q3 [ϕ] + ∆M2 [ϕ] + M1 Q4 [div w] + M2 Q1 + ∆Q2 [div w] = g.
| {z }
By equating the sums of such terms (underbraced) containing ϕ and div w with zero, we
arrive at the equations
σQ3 [ϕ] + L[ϕ] + ∆K1 [ϕ] = 0,

M1 Q3 [ϕ] + ∆M2 [ϕ] = g,
σQ4 + LQ2 + K1 Q1 + ∆Q2 ) = 0,

M1 Q4 + M2 Q1 + ∆Q2 = 0.
Hence we find the operators
Q1 = M1 L + ∆M1 K1 − σ∆M2 , Q2 = σM2 − M1 K1 ,

1 (19.3.2.2)
Q3 = − L + ∆K1 , Q4 = −M2 L
σ
determining the form of the solution (19.3.2.1). As a result, we obtain independent equa-
tions for the unknown scalar and vector functions w and ϕ,
LQ1 [w] = f, (19.3.2.3)

Q1 [ϕ] = −σf4 . (19.3.2.4)
Remark 19.5. The general solution of the homogeneous equation (19.3.2.3) (for f = 0) can be
represented as the sum w = w1 + w2 , where w1 and w2 are arbitrary solutions of the two simpler
equations
L[w1 ] = 0, Q1 [w2 ] = 0.
Remark 19.6. Examples of decompositions based on formulas (19.3.2.1) can be found in Sec-
tions 12.12.3, 12.14.3, and 12.15.3, where specific systems of the form (19.1.2.5)–(19.1.2.6) are
considered.
19.3.3 Decomposition of Systems Consisting of a Vector Equation

and a Scalar Equation (the Second Approach)
We again consider a system of the form (19.1.2.5)–(19.1.2.6) with f4 = 0, where L, K1 ,
M1 , and M2 are constant coefficient linear differential operators (M1 6≡ 0 and M2 6≡ 0).
Let us reduce system (19.1.2.5)–(19.1.2.6) to a single equation of the form (19.3.1.1).
To this end, set
u = M1 [v], p = −M2 [div v], (19.3.3.1)
where v is the new unknown vector function. By substituting (19.3.3.1) into (19.1.2.5), we
obtain an equation of the form (19.3.1.1),

LM1 [v] + ∇ K1 M1 − σM2 [div v] = f. (19.3.3.2)
Equation (19.1.2.6) with f4 = 0 is satisfied identically if we substitute the expressions

(19.3.3.1) into it.
The comparison of (19.3.3.2) with (19.3.1.1) results in the following representation of
the solution of the vector equation (19.3.3.2):

v = LM1 + ∆K1 M1 − σ∆M2 [w] + ∇ (σM2 − K1 M1 [div w] , (19.3.3.3)

LM1 LM1 + ∆K1 M1 − σ∆M2 [w] = f, (19.3.3.4)
which consists of three independent scalar equations for the components of the vector w.
We point out that the above-described complete decomposition of the original system
consisting of four equations (19.1.2.5)–(19.1.2.6) for f4 = 0 gives a representation of the
solution via three components of the vector w.
Remark 19.7. The solution (19.3.3.3) and Eq. (19.3.3.4) can be represented in terms of the
original unknown u = M1 [v] (see (19.3.3.1)). To this end, let us apply the operator M1 to (19.3.3.3)
and introduce the new unknown variable w e = M1 [w]. As a result, we obtain

u = LM1 + ∆K1 M1 − σ∆M2 [e w] + ∇ (σM2 − K1 M1 [div w e] .
e = M1 [w] can be reduced to the simpler form

Equation (19.3.3.4) in terms of w

L LM1 + ∆K1 M1 − σ∆M2 [e w] = f.
◆ Examples of decompositions of various coupled linear PDEs can be found in Sec-

tions 12.6–12.17, where specific systems of continuum mechanics and physics are con-
sidered.
⊙ Literature for Chapter 19: S. Kowalevski (1885), H. Lamb (1945), P. F. Papkovich (1932), U. Neuber
(1934), M. G. Slobodianskii (1959), M. E. Gurtin and E. Sternberg (1962), P. Chadwick and E. A. Trowbridge
(1967), O. A. Ladyzhenskaya (1969), A. C. Eringen and E. S. Suhubi (1975), A. D. Polyanin and A. I. Zhurov
(2013), I. I. Lipatov and A. D. Polyanin (2013), A. D. Polyanin and S. A. Lychev (2014 a,b, 2015), S. A. Lychev
and A. D. Polyanin (2015).
Chapter 20
Some Asymptotic Results

and Formulas
This chapter serves as a very brief introduction to asymptotic methods, which can be used
if the differential equation, system of equations, or problem to be solved contains a small
or large parameter. Examples of such problems include:
1◦. The Schrödinger equation in quantum mechanics. The equation contains Planck’s
constant ~. If we are in a situation where Planck’s constant can be viewed as a
small parameter, then one can try to solve problems for the Schrödinger equation
asymptotically as ~ → 0. This is known as the semiclassical approximation.
2◦. The wave equation utt − c2 ∆u = 0 itself does not contain a small parameter. But one
often considers high-frequency solutions (for the case of constant coefficients, the
simplest solution of this sort is the plane wave u = A cos(ωt − kx + φ0 ), ω/k = c,
with large wave number k), and then the reciprocal wave number 1/k is a small
parameter.
The topic is very broad, and we only hint at two directions in asymptotic theory. One is
regular perturbation theory, where the terms containing the small parameter can be viewed
as lower-order terms. We consider the simplest problem of regular perturbation theory,
the problem on the eigenvalues of the perturbed operator (which often occurs in quantum
mechanics). In singular perturbation theory, the small parameter resides in the leading
part of the equations (e.g., it multiplies the derivatives). Of the variety of problems be-
longing there, we discuss semiclassical asymptotic solutions of the Schrödinger equation.
Finally, we present the stationary phase method, which is an important tool in obtaining
semiclassical asymptotic solutions.
For more detailed information and other types of problems, the reader is encouraged to
consult the literature cited at the end of the chapter.
1255
1256 S OME A SYMPTOTIC R ESULTS AND F ORMULAS
20.1 Regular Perturbation Theory Formulas

for the Eigenvalues
20.1.1 Statement of the Problem
Let A(ε) = A + εB, ε ∈ [0, 1], be a family of self-adjoint operators in a Hilbert space H
with inner product ( · , · ) and norm kuk = (u, u)1/2 . Assume that the unperturbed operator
A = A(0) has an simple isolated eigenvalue λ0 with the corresponding eigenvector ϕ0 ∈ H.
This means that
Aϕ0 = λ0 ϕ0 , kϕ0 k = 1, (20.1.1.1)
and any other vector u ∈ H satisfying Au = λ0 u is proportional to ϕ0 . Under certain

assumptions [e.g., see Kato (1995)] for sufficiently small ε, the perturbed operator A(ε)
has a simple isolated eigenvalue λ(ε) with the corresponding eigenvector ϕ(ε),
(A + εB)ϕ(ε) = λ(ε)ϕ(ε), kϕ(ε)k = 1, (20.1.1.2)
such that λ(ε) and ϕ(ε) smoothly depend on ε, λ(0) = λ0 , and ϕ(0) = ϕ0 . Given λ0 and
ϕ0 , the problem is to find the coefficients in the expansions
λ(ε) = λ0 + ελ1 + ε2 λ2 + · · · , ϕ(ε) = ϕ0 + εϕ1 + ε2 ϕ2 + · · · (20.1.1.3)
without actually solving the (presumably complicated) equation (20.1.1.2).
20.1.2 Formulas for the Coefficients of the Expansion

The first few coefficients in the expansions (20.1.1.3) can be successively step by step
computed by the following formulas.
Step 1. λ1 = (ϕ0 , Bϕ0 ), and ϕ1 is found from the equations
(A − λ0 )ϕ1 = (λ1 − B)ϕ0 , Re(ϕ1 , ϕ0 ) = 0. (20.1.2.1)
Step 2. λ2 = (ϕ1 , (B −λ1 )ϕ0 ), λ3 = (ϕ1 , (B −λ1 )ϕ1 ), and ϕ2 is found from the equations
1
(A − λ0 )ϕ2 = (λ1 − B)ϕ1 + λ2 ϕ0 , Re(ϕ2 , ϕ0 ) = − (ϕ1 , ϕ1 ). (20.1.2.2)
2
Step 3. λ4 = (ϕ2 , (B − λ1 )ϕ1 ) + λ2 (ϕ0 , ϕ2 ).

Remark 20.1. The solvability of the first equations in (20.1.2.1) and (20.1.2.2) follows from
the orthogonality of their right-hand sides to ϕ0 . If, for example, there exists an orthonormal basis
{e0 , e1 , e2 , . . . } of eigenvectors of A in H with the corresponding eigenvalues µj , µ0 = λ0 , e0 = ϕ0 ,
then the solutions of (20.1.2.1) and (20.1.2.2) can be written in the form
X∞ X∞
(ej , (λ1 − B)ϕ0 ) 1 (ej , (λ1 − B)ϕ1 )
ϕ1 = ej , ϕ2 = − (ϕ1 , ϕ1 )ϕ0 + ej .
j=1
µj − µ0 2 j=1
µj − µ0
20.2. Singular Perturbation Theory 1257
20.2 Singular Perturbation Theory

20.2.1 Cauchy Problem for the Schrödinger Equation
◮ Schrödinger equation.
For the nonstationary Schrödinger equation
∂ψ 2
−i~ b = 0,
+ Hψ b = − ~ ∆ + V (x),
H (20.2.1.1)
∂t 2m
describing the evolution of the ψ-function of a quantum particle of mass m in the potential
field V (x), consider the Cauchy problem
ψ|t=0 = ψ0 (x) (the initial state of the particle). (20.2.1.2)
◮ Exact solution for the quantum oscillator.

Let the quantum Hamiltonian H b have the special form

b = 1 2 2
H −~ ∆ + x , x2 = x21 + · · · + x2n . (20.2.1.3)
2
This Hamiltonian is known as the quantum oscillator; for simplicity, we assume that the
mass and the elasticity coefficient are equal to unity.
The Cauchy problem (20.2.1.1)–(20.2.1.2) for the Hamiltonian (20.2.1.3) has the fun-
damental solution
e−iπn/2 n i o
2 2
E (x, y, t, ~) = exp (x + y ) cos t − 2xy , t ∈ (0, π),
(2π~ sin t)n/2 2~ sin t
(20.2.1.4)
so that the exact solution of this Cauchy problem is given by
Z Z
ψ(x, t, ~) = · · · E (x, y, t, ~)ψ0 (y, ~) dy, (20.2.1.5)
where, as usual, dy = dy1 · · · dyn .
◮ Semiclassical approximation.
Now assume that we are interested in solutions for which the Planck constant ~ can be
viewed as a small parameter. This is the subject of the so-called semiclassical approxima-
tion. The simplest typical initial data for the semiclassical approximation has the form
i
ψ0 (x, ~) = e ~ S0 (x) ϕ0 (x) (20.2.1.6)
where S0 and ϕ0 are smooth functions, S0 is real-valued, and ϕ0 (x) is zero outside a
sufficiently large ball in Rn . Such a function is known as a “WKB element” (derived from
the names of Wentzel, Kramers, and Brillouin, who were apparently the first to consider
this approximation). It turns out that, at least for small t, the solution of the Schrödinger
equation (20.2.1.1) with this initial data can be sought asymptotically in the form
i
ψ(x, t, ~) = e ~ S(x,t) a(x, t). (20.2.1.7)
More precisely, the procedure is as follows.
1◦ . Consider the classical Hamiltonian
p2
H(x, p) = + V (x)
2m
b in (20.2.1.1) and solve the Cauchy problem
corresponding to the quantum Hamiltonian H
for the Hamiltonian system of ODEs
∂H p ∂H ∂V
x′t = (x, p) = , p′t = − (x, p) = − (x) (20.2.1.8)
∂p m ∂x ∂x
with the initial data

∂S0
x|t=0 = x0 , p|t=0 = (x0 ). (20.2.1.9)
∂x
Denote the solution by x = X(x0 , t), p = P (x0 , t).
2◦ . Construct the function
Z t X
n
∂H
Φ(x0 , t) = S0 (x0 ) + Pj (x0 , τ ) (X(x0 , τ ), P (x0 , τ ))
0 ∂pj
j=1

− H(X(x0 , τ ), P (x0 , τ )) dτ (20.2.1.10)
and the Jacobian

∂(X1 (x0 , t), . . . , Xn (x0 , t)) ∂Xi
J (x0 , t) =
≡ det . (20.2.1.11)
∂(x1 , . . . , xn ) ∂xj
Note that J (x0 , 0) = 1, and so the Jacobian is nonzero for sufficiently small t whenever
x0 ∈ supp(ϕ0 ), where supp(ϕ0 ) is the support of ϕ0 , that is, the closure of the set of points
x0 where ϕ( x0 ) is nonzero.
3◦ . Solve the equation
x = X(x0 , t) (20.2.1.12)
for x0 , thus obtaining the function x0 = x0 (x, t). (The equation is solvable by the implicit
function theorem provided that the Jacobian (20.2.1.11) is nonzero.) Set
S(x, t) = Φ(x0 (x, t), t). (20.2.1.13)
This function is called the action, or eikonal.

4◦ . Now set
i
e ~ S(x,t) ϕ0 (x0 (x, t))
ψ(x, t) = p . (20.2.1.14)
J (x0 (x, t), t)
This function, known as the WKB solution, gives the asymptotics of the solution of prob-
lem (20.2.1.1)–(20.2.1.2) with accuracy O(~).
Remark 20.2. This solution is not valid (or even defined) everywhere; on the contrary, it only
holds until the Jacobian is zero. The points (x0 , t) where the Jacobian is zero are known as focal
points. The corresponding points (x, t) = (X(x0 , t), t) are known as caustics (sometimes they are
also called focal points). See below about how to construct the asymptotics of the solution in a
neighborhood of caustics.
Example 20.1. Consider the Cauchy problem with the oscillator Hamiltonian (20.2.1.3) for
n = 1 and with the initial data (20.2.1.6) such that
1 2
S0 (x) = x .
2
The Cauchy problem for the Hamiltonian system has the form
x′t = p, p′t = −x, x|t=0 = p|t=0 = x0 ,
and its solution is given by

√ π √ π
X(x0 , t) = x0 (cos t+sin t) = 2x0 sin t+ , P (x0 , t) = x0 (cos t−sin t) = 2x0 sin −t .
4 4
Next,
√ π x2
J (x0 , t) =
2 sin t + , Φ(x0 , t) = 0 (1 − sin 2t).
4 2
Finally, we obtain the WKB asymptotic solution in the form
2
i x (1−sin 2t) x
e ~ sin t+cos t ϕ0
ψ(x, t) = √ sin t + cos t .
sin t + cos t
3π
This solution is valid for 0 ≤ t ≤ 4 − ε, where ε > 0 is arbitrary but fixed (independent of ~).
◮ Solution in a neighborhood of focal points.

As we have seen from the preceding example, the WKB solution formula fails to be true in a
neighborhood of the points where the Jacobian J(x0 , t) vanishes. Thus, we need a different
representation of the solution near focal points. We will describe this representation only
for n = 1 (and refer the reader to the cited literature for the general case). The procedure is
as follows.
1◦ . Let (x0∗ , t∗ ) be a focal point (that is, J (x0∗ , t∗ ) = 0). Then the Jacobian
∂P
J1 (x0 , t) = (x0 , t) (20.2.1.15)
∂x0
is necessarily nonzero at (x0∗ , t∗ ). Then, by the implicit function theorem, we can solve
the equation p = P (x0 , t) for x0 , thus obtaining a function x0 = x0 (p, t).
2◦ . Construct the functions
e 0 , t) = Φ(x0 , t) − P (x0 , t)X(x0 , t)
Φ(x (20.2.1.16)
and
e t) = Φ(x
S(p, e 0 (p, t), t). (20.2.1.17)
3◦ . For a function ϕ(x0 , t) supported in a neighborhood of the point (x0∗ , t∗ ), define the
function

i ϕ(x0 , t)
e 1/~
[Kφ](x, t, ~) = F̄p→x e ~ S(p,t)
p (x, t, ~), (20.2.1.18)
|J1 (x0 , t)| x0 =x0 (p,t)
1/~
where F̄p→x is the Fourier transform with parameter ~ defined below in Section 20.2.3.
That is what the solution should look like near focal points. So far, we have only one
problem: for this function to satisfy the equation, the function φ(x0 , t) should be indepen-
dent of t, but then the expression (20.2.1.18) is not defined, because the function x0 (p, t)
is not defined everywhere on the support of ϕ. This problem will be dealt with in the next
item.
◮ Pasting the local solutions together.

In fact, now we have two kinds of “local solutions,” one of the form (20.2.1.18) and the
other of the form (20.2.1.14). Neither of them, taken alone, is a solution (even asymptotic)
of our Cauchy problem. So we paste them together as follows.
1◦ . Cover the domain of the variables (x0 , t) by open disks Uj such that, for each j, either
J (x0 , t) 6= 0 in the entire Uj (these Uj will be called nonsingular charts) or J1 (x0 , t) 6= 0
in the entire Uj (these Uj will be called singular charts). Assume also that J (x0j , tj ) 6= 0,
where (x0j , tj ) is the center of the disk Uj .
2◦ . For each disk Uj , define an integer mj (the Maslov index of Uj ) as follows. Let
∂(X(x0 , t) + iεP (x0 , t))
Jε (x0 , t) = , ε > 0. (20.2.1.19)
∂x0
If Uj is nonsingular, set
Z tj
1 ∂Jε
mj = lim Jε−1 (x0j , τ ) (x0j , τ ) dτ. (20.2.1.20)
π ε→+0 0 ∂t
If Uj is singular, set
Z tj (
1 −1 ∂Jε 0 if J (x0j , tj )J (x0j , tj ) > 0,
mj = lim Jε (x0j , τ ) (x0j , τ ) dτ +
π ε→+0 0 ∂t 1 if J (x0j , tj )J (x0j , tj ) < 0.
(20.2.1.21)
3◦ . For any function ϕ(x0 , t) with support in a nonsingular disk Uj , set
i
π
−i 2 mj e ~ Φ(x0 (x,t),t) ϕ(x0 (x, t), t)
[Kj ϕ](x, t, ~) = e p . (20.2.1.22)
|J (x0 (x, t), t)|
For any function ϕ(x0 , t) with support in a singular disk Uj , set

π
−i 2 mj i e ϕ(x0 , t)
[Kj ϕ](x, t, ~) = e 1/~
F̄p→x e ~ Φ(x 0 ,t)
p (x, t, ~).
|J1 (x0 , t)| x0 =x0 (p,t)
(20.2.1.23)
Here the functions x0 (x, t) and x0 (p, t) give the solutions of the respective equations x =
X(x0 , t) and p = P (x0 , t) in the corresponding disks Uj .
4◦ . Now the asymptotics of the solution of the Cauchy problem (20.2.1.1), (20.2.1.6) is
given by the formula
X
ψ(x, t, ~) = Kj [ej (x0 , t)ϕ0 (x0 )](x, t, ~), (20.2.1.24)
j
where {ej } is a partition of unity subordinate to the cover {Uj }:

X
ej (x0 , t) ∈ C ∞ , supp ej ⊂ Uj , ej (x0 , t) = 1.
j
Here supp ej is the support of ej , i.e., the closure of the set of points where the function
ej (x0 , t) is nonzero.
The function thus constructed is independent of the partition of unity with accuracy
O(~). This follows from the fact that, for any function ϕ such that, simultaneously,
supp ϕ ⊂ Uj and supp ϕ ⊂ Uk , one has
Kj ϕ = Kk ϕ + O(~).
This can be proved by the stationary phase method presented in the next subsection.
20.2.2 Stationary Phase Method

◮ Simplest case.
Consider the integral Z
i
I(~) = e ~ Φ(θ) a(θ) dθ,
where Φ(θ) and a(θ) are smooth functions, Φ(θ) is real-valued, and a(θ) vanishes outside
some finite interval.∗ Let us study the behavior of this integral as ~ → 0.
1◦ . If Φ′ (θ) 6= 0 anywhere, then I(~) = O(~∞ ) (i.e., |I(~)| ≤ CN hN for any N > 0).
2◦ . Assume that Φ′ (θ∗ ) = 0 at some point θ∗ (such a point is called a stationary point),
Φ′′ (θ∗ ) 6= 0 (the stationary point is nondegenerate), and Φ′ (θ) 6= 0 for θ 6= θ∗ (there are no
other stationary points). Then I(~) has the asymptotics

√ iπ ′′ i a(θ∗ )
I(~) = 2π~e 4 sign Φ (θ∗ ) e ~ Φ(θ∗ ) p + O(~) , (20.2.2.1)
|Φ′′ (θ∗ )|
n
1, ξ>0,
where sign ξ = −1, ξ<0 is the sign function.
(1) (s)
3◦ . If Φ(θ) has several stationary points θ∗ , . . . , θ∗ , each of which is nondegenerate,
then the asymptotics of I(~) is given by the sum of contributions (20.2.2.1) of these sta-
tionary points:
 
√ X s (j)
iπ ′′ (j) i (j) a(θ∗ )
I(~) = 2π~ e 4 sign Φ (θ∗ ) e ~ Φ(θ∗ )  q + O(~) .
′′ (j)
j=1 |Φ (θ∗ )|
∗
The last requirement, which is only needed to guarantee the convergence of the integral, can be weakened.
R i 2
Example 20.2. Consider the integral e 2~ θ a(θ) dθ. The phase function Φ(θ) = θ2 /2 has the
′′
unique stationary point θ∗ = 0, and Φ (0) = 1. By (20.2.2.1),
Z √
i 2 iπ
e 2~ θ a(θ) dθ = 2π~e 4 a(0) + O(~3/2 ).
◮ Dependence on parameters.
The functions Φ and a often depend on parameters, Φ = Φ(x, θ) and a = a(x, θ), where
x is a one- or many-dimensional parameter varying in some domain in Rn . Consider the
parameter-dependent integral
Z
i
I(x, ~) = e ~ Φ(x,θ) a(x, θ) dθ,
Assume that, for some parameter value x = x∗ , the function Φ(x, θ) has a stationary point
θ∗ and this point is nondegenerate; that is,
∂Φ
Φ′θ (x, θ) ≡ (x, θ) = 0 (20.2.2.2)
∂θ
for x = x∗ and θ = θ∗ , and Φ′′θθ (x∗ , θ∗ ) 6= 0. Then, by the implicit function theorem,
equation (20.2.2.2) defines a unique smooth function θ = Θ(x) in a neighborhood of x∗
such that Θ(x∗ ) = θ∗ and there are no stationary points near (x∗ , θ∗ ) other than (x, Θ(x)).
Now if there are no other stationary points on the support of a(x, θ) (for example, if a is
supported in a small neighborhood of (x∗ , θ∗ )), then I(x, ~) has the asymptotics

√ iπ
sign Φ′′
i
(x∗ ,θ∗ ) ~ Φ(x,Θ(x)) a(x, Θ(x))
I(x, ~) = 2π~e 4 θθ e p + O(~) .
|Φ′′θθ (x, Θ(x))|
R i 2
Example 20.3. Consider the integral e ~ (θ /2−xθ)
a(θ) dθ. The stationary point equation reads
θ−x=0 =⇒ Θ(x) = x.
Moreover, Φ′′θθ (x, x) = 1, and we obtain
Z √
i θ2 iπ ix2
e ~ 2 −xθ a(θ) dθ = 2π~e 4 e− 2~ a(x) + O(~3/2 ).
◮ General case.
Now consider the general case of a multiple oscillatory integral.
Let Φ(x, θ) be a smooth real-valued function of the variables x = (x1 , . . . , xn ) ∈ Rn
m , let a(x, θ) be a smooth function of the same parameters such
and θ = (θ1 , . . . , θm ) ∈ Rp
that a(x, θ) = 0 for |θ| = θ12 + · · · + θm 2 > R, and let ~ > 0 be a positive parameter. The
function I(x, ~) defined as the parametric integral

Z Z
1 i
I(x, ~) = m/2
· · · e ~ Φ(x,θ) a(x, θ) dθ, (20.2.2.3)
(2π~)
where i is the imaginary unit and dθ = dθ1 · · · dθm , is called the oscillatory integral with
phase function Φ and amplitude a. We sometimes write I[Φ, a](x, ~) instead of I(x, ~) to
emphasize the dependence of the integral (20.2.2.3) on the functions Φ and a occurring in
its definition. The problem is to find the asymptotics of the integral (20.2.2.3) as ~ → 0.
◮ Stationary points.
A stationary point of the integral (20.2.2.3) (or of the phase function Φ(x, θ)) is a point
(x∗ , θ∗ ) such that
∂Φ
(x∗ , θ∗ ) = 0, j = 1, . . . , m. (20.2.2.4)
∂θj
A stationary point (x∗ , θ∗ ) is said to be nondegenerate if
2
∂ Φ
det Φ′′θθ (x∗ , θ∗ )
≡ det (x∗ , θ∗ )
∂θj ∂θk 6= 0. (20.2.2.5)
If (x∗ , θ∗ ) is a nondegenerate stationary point, then, by the implicit function theorem ap-
plied to system (20.2.2.4), there exists a smooth function θ = Θ(x) such that, in a small
neighborhood of (x∗ , θ∗ ), all stationary points of the phase function Φ(x, θ) have the form
(x, Θ(x)). In particular, θ∗ = Θ(x∗ ). We will denote the matrix on the left-hand side
in (20.2.2.5) by Φ′′θθ (x∗ , θ∗ ).
◮ Oscillatory integrals without stationary points.

Assume that there are no stationary points on the support supp a(x, θ) of the amplitude
a(x, θ). Then I(x, ~) = O(~∞ ). This means that I(x, ~) = O(~N ) for every N > 0,
however large.
◮ Oscillatory integrals with nondegenerate stationary points.

Assume that, for each x, the phase function Φ(x, θ) has a unique stationary point (x, Θ(x))
on the support supp a and this stationary point is nondegenerate, det Φ′′θθ (x, Θ(x)) 6= 0.
Then the integral (20.2.2.3) has the asymptotics
i
iπm
sign(Φ ′′ (x,θ)) e ~ Φ(x,θ) a(x, θ)
I(x, ~) = e 4 θθ p + O(~), (20.2.2.6)
| det(Φ′′θθ (x, θ))| θ=Θ(x)
where
sign(Φ′′θθ (x, θ)) = σ+ (Φ′′θθ (x, θ)) − σ− (Φ′′θθ (x, θ))
is the signature of the real symmetric matrix Φ′′θθ (x, θ), i.e., the difference between the
numbers of its positive and negative eigenvalues.
Example 20.4. Consider the integral
ZZ
1 i
I(x, ~) = e ~ p(x−y) χ(p)a(y) dy dp,
(2π~)n
where x = (x1 , . . . , xn ), p = (p1 , . . . , pn ), y = (y1 , . . . , yn ), and

n
X
p(x − y) = pj (xj − yj ).
j=1
Here m = 2n and θ = (y, p). The stationary point equations and the matrix Φ′′θθ have the form

′′ 0 −E
p = 0, x = y, Φθθ (x, y, p) = ,
−E 0
where E is the n×n identity matrix. The signature of this matrix is zero (it has exactly n positive and
n negative eigenvalues), and so the stationary phase formula (20.2.2.6) gives I(x, h) = χ(0)a(x) +
O(~). Now we take a function χ(p) such that χ(0) = 1, make the change of integration variable
p = ξ/~ and rewrite the result in the form
ZZ
1
eiξ(x−y) χ(~ξ)a(y) dy dξ = a(x) + O(~).
(2π)n
By letting ~ → 0, we obtain the Fourier transform inversion formula.
20.2.3 Fourier Transform with a Parameter

◮ Fourier transform.
When working with differential equations containing a small parameter ~, it is expedient
to use a version of the Fourier transform which contains that parameter as well. It is called
the 1/~-Fourier transform. This transform and its inverse in the case of one variable are
defined by
Z
1/~ e−iπ/4 i
[Fx→p u](p) = e− ~ px u(x) dx, (20.2.3.1)
(2π~)1/2
Z
1/~ eiπ/4 i
[F̄p→x v](x) = e ~ px v(p) dp. (20.2.3.2)
(2π~)1/2
The corresponding n-dimensional versions (x, p ∈ Rn ) are

Z Z
1/~ e−iπn/4 i eiπn/4 i
[Fx→p u](p) = e− ~ px u(x) dx, 1/~
[F̄p→x v](x) = e ~ px v(p) dp,
(2π~)n/2 (2π~)n/2
where now px = p1 x1 + · · · + pn xn , dx = dx1 · · · dxn , and dp = dp1 · · · dpn .
◮ Commutation with the operators x and −i~∂/∂x.

The 1/~-Fourier transform enjoys the usual commutation formulas
h ∂u i ∂
1/~ 1/~ 1/~
Fx→p − i~ (p) = p[Fx→p u](p), [Fx→p (xu)](p) = i~ [F 1/~ u](p).
∂x ∂p x→p
⊙ Literature for Chapter 20: J. Brüning, V. V. Grushin, and S. Yu. Dobrokhotov (2012), I. M. Gelfand (1989),
T. Kato (1995), V. P. Maslov (1972), V. P. Maslov (1976), V. P. Maslov (1994), A. S. Mishchenko, V. E. Shat-
alov, and B. Yu. Sternin (1990)
Chapter 21
Elements of Theory
of Generalized Functions
21.1 Generalized Functions of One Variable
21.1.1 Important Terminological Remark
The main subject of this chapter—generalized functions—has in fact two names in the
literature. One name, “generalized functions,” tends to be more popular among engineers
and physicists; it probably goes back to Dirac and was firmly introduced in practice by
Gelfand and Shilov. The other name, “distributions,” mostly preferred by mathematicians,
stems from the seminal work of Laurent Schwartz. We use both terms interchangeably but
prefer “distributions”—just because it is shorter, one word instead of two.
21.1.2 Test Function Space
Functions vanishing outside a finite interval are said to be compactly supported. The closure
of the set of points where a function ϕ(x) is nonzero is called the support of ϕ and is
denoted by supp ϕ. The set {ϕ(x)} of smooth compactly supported functions (sometimes
referred to as test functions) equipped with an appropriate convergence (see below) is called
the test function space and is denoted by K.
Example 21.1. The following function lies in K:
 2

exp − a for |x| < a,
ϕa (x) = a2 − x2

0 for |x| ≥ a.
It has the support supp ϕa (x) = {|x| ≤ a}.
A sequence of functions ϕn (x) ∈ K is said to converge to a function ϕ(x) ∈ K if (i) there

dm dm
exists a finite interval containing the supports of all ϕn (x); (ii) dx m ϕn (x) ⇒ dxm ϕ(x) as
K
n → ∞ for m = 0, 1, . . . . In this case, one writes ϕn (x) −→ ϕ(x) as n → ∞.
1265
1266 E LEMENTS OF T HEORY OF G ENERALIZED F UNCTIONS
21.1.3 Distribution Space. Dirac Delta Function

◮ Regular and singular distributions. Some theorems.
The distribution space, or space of generalized functions K′ , is introduced as follows. Ordi-
nary locally integrable functions f (x) are identified in K′ with continuous linear functionals
(referred to as regular distributions, or regular generalized functions) of the form
Z ∞
(f, ϕ) = f (x)ϕ(x) dx, (21.1.3.1)
−∞
where ϕ(x) ∈ K is arbitrary. Every regular distribution corresponds to a unique (up to

values of a set of measure zero) locally integrable function on R1 .
Continuous linear functionals on K that cannot be represented in the form (21.1.3.1)
with a locally integrable f are called singular generalized functions, or singular distribu-
tions.
The convergence in K′ is defined as the weak convergence of functionals: a sequence
of distributions fn (x) ∈ K′ converges to a distribution f (x) ∈ K′ if (fn , ϕ) → (f, ϕ) as
n → ∞ for each ϕ(x) ∈ K.
T HEOREM ON THE COMPLETENESS OF THE SPACE K′ . Let fn ∈ K′ be a sequence of
distributions such that the numerical sequence (fn , ϕ) converges as n → ∞ for each ϕ ∈ K.
Then the functional f defined by
(f, ϕ) = lim (fn , ϕ), ϕ ∈ K,

n→∞
is linear and continuous on K as well ; i.e., f ∈ K′ .

T HEOREM ON THE REPRESENTATION OF DISTRIBUTIONS. Each distribution f ∈ K′
is a weak limit of test functions fn ∈ K; i.e., the set K is dense in K′ .
◮ Dirac delta function and delta sequences.

The theorem on the representation of distributions permits one to construct singular distri-
butions with the use of appropriate sequences of test functions.
Example 21.2. Consider the sequence of integrable discontinuous functions
(
ε−1 for 0 ≤ x ≤ ε, 1
fn (x) = where ε = . (21.1.3.2)
0 for x < 0 or x > ε, n
Let us show that Z ∞

lim (fn , ϕ) = lim fn (x)ϕ(x) dx = ϕ(0). (21.1.3.3)
n→∞ n→∞ −∞
Indeed, it follows from the continuity of ϕ(x) that for an arbitrarily small ∆ > 0 there exists an
ε0 > 0 such that |ϕ(x) − ϕ(0)| < ∆ for |x| < ε0 . Then
Z ∞ Z Z Z
1 ε 1 ε ∆ ε
f (x)ϕ(x) dx − ϕ(0) = ϕ(x) − ϕ(0) dx ≤ |ϕ(x) − ϕ(0)| dx < dx = ∆
n ε ε ε
−∞ 0 0 0
for all ε = 1/n ≤ ε0 , as desired.

21.1. Generalized Functions of One Variable 1267
Thus, the weak limit of the sequence fn (x) as n → ∞ is the functional that takes each continuous
function ϕ(x) to the value ϕ(0) of that function at the point x = 0. This functional is denoted by
δ(x) and is called the Dirac delta function.
Symbolically, the action of the delta function is denoted by
(δ, ϕ) = ϕ(0). (21.1.3.4)
A sequence of functions fn (x) is called a delta sequence if it converges to the delta

function as n → ∞. The sequence (21.1.3.2) of discontinuous functions is a delta sequence.
The following sequences are examples of delta sequences of continuously differentiable
functions:

ε 1 x2 1 x 1
fn (x) = 2 2
, fn (x) = √ exp − , fn (x) = sin , ε = .
π(x + ε ) 2 πε 4ε πx ε n
Remark 21.1. One can use the following result to obtain various delta sequences. Let g(x) ≥ 0
be an arbitrary function satisfying the normalization condition
Z ∞
g(x) dx = 1.
−∞
Then the sequence of functions

gn (x) = ng(nx)
is a delta sequence as n → ∞.
Remark 21.2. Physical interpretation. In physics, the Dirac delta function δ(x) is often defined
as a function that vanishes for all real x 6= 0, is infinite at x = 0, and satisfies the normalization
condition Z ∞
δ(x) dx = 1.
−∞
This definition permits intuitively representing the spatial density of physical variables (mass,
charge, heat source intensity, force, etc.) lumped or applied at a single point in space. For example,
the delta function can be identified with the mass density distribution for which there is a unit lump
mass at the point x = 0 and the mass is zero at all other points. If there is a lump mass m at a point
x = x0 , then its density is mδ(x − x0 ).
The main properties of the one-dimensional Dirac delta function are as follows:
1. f (x)δ(x − x0 ) = f (x0 )δ(x − x0 ),

Z b (
f (x) if a < x < b,
2. f (y)δ(x − y) dy =
a 0 if x < a or x > b,
where f (x) is any continuous function.
21.1.4 Derivatives of Distributions. Some Formulas

◮ Definitions of the first and higher derivatives. Schwartz theorem.
The derivative of a distribution is by definition the functional f ′ given by the formula
(f ′ , ϕ) = −(f, ϕ′ ). (21.1.4.1)
One can readily verify that f ′ is a continuous linear functional. For functions f (x) dif-
ferentiable in the ordinary sense, relation (21.1.4.1) follows from the integration by parts
formula by virtue of (21.1.3.1), so that the two notions of derivative coincide in this case.
Distributions introduced in this manner are infinitely differentiable (in the sense of dis-
tributions), the kth derivative being given by the formula
(f (k) , ϕ) = (−1)k (f, ϕ(k) ). (21.1.4.2)

Example 21.3. Let us find the generalized derivative of the Heaviside unit step function
(
0 for x ≤ 0,
ϑ(x) = (21.1.4.3)
1 for x > 0.
Using formula (21.1.4.1) with regard to (21.1.3.1), we obtain

Z ∞ Z ∞
′ ′ ′
(ϑ , ϕ) = −(ϑ, ϕ ) = − ϑ(x)ϕ (x) dx = − ϕ′ (x) dx = ϕ(0). (21.1.4.4)
−∞ 0
The comparison of (21.1.4.4) with (21.1.3.4) gives
(ϑ′ , ϕ) = (δ, ϕ) =⇒ ϑ′ (x) = δ(x). (21.1.4.5)
Thus, the delta function can also be defined as the generalized derivative of the Heaviside unit step
function.
This example shows that the derivative of a regular distribution (= the Heaviside unit
step function) can be a singular distribution (= the Dirac delta function). Note that the
Heaviside unit step function itself can be defined as the derivative of the continuous function
that is zero for x ≤ 0 and is equal to x for x > 0. This fact admits the following important
generalization.
S CHWARTZ T HEOREM. Each function in K′ is the generalized derivative (of some or-
der) of some regular distribution specified by a continuous function.
◮ Generalized derivative of a function that has discontinuities of the first kind.

Assume that a function f (x) is continuous and differentiable everywhere except at a point
x0 , where it has a discontinuity of the first kind. The jump of f (x) at x0 is the number
[f ]x0 = f (x0 + 0) − f (x0 − 0),
where f (x0 + 0) = lim f (x) and f (x0 − 0) = lim f (x). The generalized
x→x0 , x>x0 x→x0 , x<x0
derivative of this function is computed by the formula
f ′ = {f ′ (x)} + [f ]x0 δ(x − x0 ), (21.1.4.6)
where {f ′ (x)} is the classical derivative (defined everywhere except for the point x0 ). The
generalized derivative of a function f (x) that has several isolated discontinuities of the first
kind at some points xk can be written as follows:
X
f ′ = {f ′ (x)} + [f ]xk δ(x − xk ).
k
21.1. Generalized Functions of One Variable 1269
◮ Integrals containing derivatives of the delta function.

Let the derivative f (n) (x) be continuous for a < x < b. Then
Z b
f (y)δ(n) (x − y) dy = f (n) (x), n = 1, 2, . . . ,
a
Z b
f (y)δ(n) (y − x) dy = (−1)n f (n) (x).
a
21.1.5 Operations on Distributions and Some Transformations

Operations on and transformations of distributions are first introduced for regular distri-
butions on the basis of formula (21.1.3.1) and are then extended by definition to singular
distributions (to construct which one can use sequences of regular distributions and the
completeness theorem).
◮ Addition and multiplication of distributions.

1◦ . Addition of distributions. One can add distributions:
(f1 + f2 , ϕ) = (f1 , ϕ) + (f2 , ϕ).
2◦ . Multiplication of distributions. Let h = h(x) be an infinitely differentiable function.
Then, by definition,
(hf, ϕ) = (f, hϕ).
Example 21.4. Let us prove that
h(x)δ(x) = h(0)δ(x). (21.1.5.1)

Indeed, (hδ, ϕ) = (δ, hϕ) = h(0)ϕ(0) = (h(0)δ, ϕ) for all ϕ ∈ K.
Remark 21.3. In the general case, one cannot introduce a well-defined product of a distribution
by a discontinuous function or, so much the more, by another distribution.
Note that the notion of direct product of distributions of distinct arguments is well defined in
the theory of distributions of several variables (see below).
◮ Linear and general changes of variables.

1◦ . Linear change of variables. Let a and b be constants (a 6= 0). By definition,
1 x − b
f (ax + b), ϕ(x) = f (x), ϕ .
|a| a
In particular, it follows that
δ(ax) = |a|−1 δ(x), a 6= 0.
2◦ . Change of variables of general form. Let t(x) be a strictly monotone infinitely differ-
entiable function; i.e., there exists an inverse function, y = t(x) =⇒ x = τ (y). Then
f (t(x)) is understood as the functional acting by the rule
ϕ(τ (x))
f (t(x)), ϕ(x) = f (x), ′ .
|t (τ (x))|
21.1.6 Tempered Distributions and Fourier Transform

◮ Tempered distributions.
The space J of rapidly decaying test functions contains all infinitely differentiable func-
tions ϕ(x) decaying, together with all of their derivatives, more rapidly than any power of
|x|−1 as |x| → ∞. The convergence in J is defined as follows: a sequence of functions
dm β dm
ϕn ∈ J converges to a function ϕ ∈ J if and only if xβ dx m ϕn (x) ⇒ x dxm ϕ(x) as
n → ∞ for any β and m = 0, 1, . . . .

We have the following consequences of these definitions: (i) J is a linear space; (ii)
K ⊂ J ; (iii) the convergence in K implies the convergence in J ; (iv) K is dense in J ,
which means that for each ϕ ∈ J there exists a sequence ϕn ∈ K such that ϕn → ϕ as
n → ∞.
The space J ′ of tempered distributions consists of continuous linear functionals on the
test function space J . The convergence in J ′ is defined as the weak convergence of a
sequence of functionals.
One can prove that (i) J ′ ⊂ K′ ; (ii) the convergence in J ′ implies the convergence
in K′ .
Example 21.5. Let us describe a broad class of tempered distributions. Let f (x) be a locally
integrable function of polynomial (tempered) growth at infinity, so that it satisfies the condition
Z ∞
|f (x)|(1 + |x|)−m dx < ∞
−∞
for some m ≥ 0. Then this function defines a regular functional f ∈ J ′ by formula (21.1.3.1),
where ϕ(x) ∈ J .
◮ Fourier transform of tempered distributions.

Since the test functions J are absolutely integrable on R1 , it follows that they have well-
defined Fourier transforms
Z ∞
1
ϕe (u) = F ϕ(x) , where F ϕ(x) ≡ √ ϕ(x)e−iux dx.
2π −∞
One can show that the Fourier transform F (as well as the inverse Fourier transform) is a
one-to-one transformation of J into J (in the class of complex-valued functions of real
argument of the form ϕ(x) = ϕ1 (x) + iϕ2 (x), where ϕ1,2 (x) belong to the space of rapidly
decaying test functions).
The Fourier transform
of an absolutely integrable function f (x) defines a generalized
function F {f }, ϕ ∈ J . By changing the order of integration, we obtain the chain of
equalities
Z ∞ Z ∞ Z ∞
1 −iux
F {f }, ϕ = F {f (x)}ϕ(u) du = √ f (x)e dx ϕ(u) du
−∞ −∞ 2π −∞
Z ∞ Z ∞ Z ∞
1 −iux
= √ ϕ(u)e dx f (x) dx = f (x)F {ϕ(u)} dx;
−∞ 2π −∞ −∞
21.2. Generalized Functions of Several Variables 1271
i.e.,

F {f }, ϕ = f, F {ϕ} .
The last relation is taken to be the definition of the Fourier transform of any tempered
distribution.
21.1.7 Generalized Solutions of Linear Ordinary Differential

Equations
Consider a linear ordinary differential equation
m
X dk w
L[w] ≡ ak (x)wx(k) = f (x), wx(k) ≡ . (21.1.7.1)
dxk
k=0
The adjoint of the operator L is the operator L∗ defined by the formula

m
X (k)
L∗ [w] ≡ (−1)k ak (x)w(x) x . (21.1.7.2)
k=0
One has the following identity (which can be obtained by integration by parts) for clas-
sical solutions having m continuous derivatives and satisfying Eq. (21.1.7.1):
Z ∞ Z ∞ Z ∞
∗ ∗
(w, L [ϕ]) = w(x)L [ϕ(x)] dx = L [w(x)]ϕ(x) dx = f (x)ϕ(x) dx = (f, ϕ);
−∞ −∞ −∞
i.e.,
(w, L∗ [ϕ]) = (f, ϕ). (21.1.7.3)
This identity is taken to be the definition of a generalized solution: a generalized solution

of the ordinary differential equation (21.1.7.1) is a functional w such that Eq. (21.1.7.3)
holds for each test function ϕ ∈ K.
21.2 Generalized Functions of Several Variables

This section gives selected information, most often used in the theory of linear PDEs, about
distributions of several variables.
21.2.1 Some Definitions. Partial Derivatives. Direct Product. Linear

Transformations
◮ Test function and distribution spaces of several variables.
The space K(Rn ) of test functions of several variables consists of compactly supported
(in every direction) functions ϕ = ϕ(x), where x = (x1 , . . . , xn ), that have all possible
continuous partial derivatives of any order.
The following is an example of a test function of several variables:

 2
exp − a for |x| < a,
ϕa (x) = a2 − x2

0 for |x| ≥ a,
p
where x2 = x21 + · · · + x2n and |x| = x21 + · · · + x2n .
The notion of distributions as continuous linear functionals over the test function space
can be defined without extra effort in the same way as this was done in the one-dimensional
case. In particular, a regular distribution f of several variables can be identified with a
continuous linear functional of the form
Z
(f, ϕ) = f (x)ϕ(x) dVx , dVx = dx1 . . . dxn .
Rn
The space of distributions of n variables is denoted by K′ (Rn ).
◮ Partial derivatives of distributions.

Let k = (k1 , . . . , kn ) be a vector with integer nonnegative components kj . By Dk f we
denote the partial derivative of a function f (x) = f (x1 , . . . , xn ) of order |k| = k1 +· · ·+kn ,
∂ |k| f (x1 , . . . , xn )
Dk f = , D0 f = f (x).
∂xk11 . . . ∂xknn
The partial derivatives of distributions are defined as follows:

Dk f, ϕ = (−1)|k| f, Dk ϕ
for each ϕ(x) ∈ K(Rn ).
◮ Linear transformations of distributions.

Let A be a nonsingular matrix (i.e., det A 6= 0), and let A−1 be the inverse matrix. Then the
distribution f (Ax + b), where b is an arbitrary constant vector, is defined as follows:
1
f (Ax + b), ϕ(x) = f (x), ϕ A−1 (x − b) .
|det A|
In particular, δ(λx) = λ−n δ(x), where λ 6= 0 is a constant.
◮ Direct product of distributions.

Let f (x) ∈ K′ (Rn ) and g(y) ∈ K′ (Rm ) be distributions of distinct variables. The product
of these distributions is the distribution f (x)g(y) ∈ K′ (Rn+m ) acting on test functions
ϕ(x, y) ∈ K(Rn+m ) by the rule

f (x)g(y), ϕ(x, y) = f (x), g(x), ϕ(x, y) ,
which is an analog of the Fubini theorem on the coincidence of repeated integrals with the
multiple integral.
21.2. Generalized Functions of Several Variables 1273
21.2.2 Dirac Delta Function. Generalized Solutions of Linear PDEs

◮ Properties of the n-dimensional Dirac delta function.
1◦ . The n-dimensional Dirac delta function possesses the following basic properties:
δ(x) = δ(x1 )δ(x2 ) . . . δ(xn ),

Z
(21.2.2.1)
Φ(y)δ(x − y) dVy = Φ(x),
Rn
where δ(xk ) is the one-dimensional Dirac delta function, Φ(x) is an arbitrary continuous
function, and dVy = dy1 . . . dyn .
2◦ . The two-dimensional Dirac delta function in polar coordinates ρ, ϕ:
1
δ(x − x∗ ) = δ(ρ − ρ∗ )δ(ϕ − ϕ∗ ).
ρ
3◦ . The three-dimensional Dirac delta function in cylindrical coordinates ρ, ϕ, z:
1
δ(x − x∗ ) = δ(ρ − ρ∗ )δ(ϕ − ϕ∗ )δ(z − z∗ ).
ρ
4◦ . The three-dimensional Dirac delta function in spherical coordinates r, θ, ϕ:
1
δ(x − x∗ ) = δ(r − r∗ )δ(θ − θ∗ )δ(ϕ − ϕ∗ ).
r 2 sin θ
◮ Generalized solutions of linear PDEs.

Consider the constant coefficient linear partial differential equation
X
L[w] ≡ ak Dk w = Φ(x). (21.2.2.2)
|k|≤m
The adjoint of L is the operator L∗ defined by the formula

X
L∗ [w] ≡ (−1)|k| ak Dk w.
|k|≤m
A general solution of the partial differential equation (21.2.2.2) is a distribution w that

satisfies the integral identity
(w, L∗ [ϕ]) = (Φ, ϕ)
for each test function ϕ ∈ K.
⊙ References for Chapter 21: L. Schwartz (1950, 1951), G. E. Shilov (1965), I. M. Gelfand and G. E. Shilov
(1959), V. S. Vladimirov (1971, 1988), S. G. Krein (1972), A. G. Butkovskiy (1979), L. Hörmander (1983,
1990), R. P. Kanwal (1983), Yu. A. Brychkov and A. P. Prudnikov (1989).
Part III
Symbolic and
Numerical Solutions
with Maple,
Mathematica,

R
and MATLAB
Inna Shingareva
Carlos Lizárraga-Celaya
Mathematics and Physics Departments
University of Sonora, Mexico
Chapter 22
Linear Partial Differential

Equations with Maple
22.1 Introduction
The theory of linear partial differential equations (PDEs) is one of the most important fields
of mathematics due to numerous applications in many branches of science and engineer-
ing. Linear PDEs have been a research subject for more than three centuries [see Debnath
(2007)], and nowadays they are an area of intensive mathematical and scientific research.
With the development of modern computers, supercomputers, computer algebra sys-
tems (such as Maple and Mathematica), and interactive software providing a programming
environment for scientific computations (such as MATLAB), using modern powerful com-
putational methods for analytical, symbolic, numerical, and graphical solutions of PDEs
has become commonplace in mathematical research [see Akritas (1989), Calmet and van
Hulzen (1983), Shingareva and Lizárraga-Celaya (2011), Davenport, Siret, and Tournier
(1993), and Wester (1999)].
In this chapter, following the most important ideas and methods, we propose and de-
velop new computer algebra ideas and methods to obtain analytical, symbolic, numeri-
cal, and graphical solutions for studying linear partial differential equations. We compute
analytical and numerical solutions in terms of predefined functions (which are an imple-
mentation of known methods for solving linear PDEs) and develop new procedures for
constructing new solutions using Maple. We show a very helpful role that computer alge-
bra systems play in the analytical derivation of numerical methods, computing numerical
solutions, and comparing numerical and analytical solutions.
22.1.1 Preliminary Remarks

In general, a partial differential equation can be written in the form
F(x, y, . . . , u, ux , uy , . . . , uxx , uxy , . . .) = 0, (22.1.1.1)
where x, y, . . . are independent variables, u is an unknown function (dependent variable) of
these independent variables, and ux , uy , . . . , uxx , uxy , . . . are partial derivatives of u. We use
subscripts on dependent variables to denote differentiation: ux = ∂u/∂x, uxy = ∂2 u/∂x∂y,
etc.
1277
1278 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M APLE
Equation (22.1.1.1) is defined in a domain D of the n-dimensional space Rn of the

independent variables x, y, . . .. We seek solutions of equation (22.1.1.1), i.e., functions
u = u(x, y, . . .) that satisfy equation (22.1.1.1) identically in D. By introducing additional
conditions, one can find various particular solutions.
For example, first- and second-order partial differential equations, say, in two indepen-
dent variables x = (x1 , x2 ) = (x, y), can be represented as
F(x, y, u, ux , uy ) = 0, F(x, y, u, ux , uy , uxx , uxy , uyy ) = 0, (22.1.1.2)
respectively. These equations are defined in a domain D, where (x, y) ∈ D ⊂ R2 , F is a

given function, and u = u(x, y) is the unknown function (dependent variable) of the inde-
pendent variables (x, y). These equations can be written in standard notation as
F(x, y, u, p, q) = 0, F(x, y, u, ux , uy , p, q, r) = 0, (22.1.1.3)
where p = ux and q = uy (for first-order PDEs) and p = uxx , q = uxy , and r = uyy (for
second-order PDEs).
A second-order linear partial differential equation in n independent variables can be
written in the general form
n n n
∑ ∑ Ai j ux x i j + ∑ Bi uxi + Fu = G, (22.1.1.4)
i=1 j=1 i=1
where Ai j (Ai j = A ji ), Bi , F, and G are functions of the n independent variables xi .

In this chapter, we use the following notation introduced by convention in Maple:
• Cn (n=1,2,...) for arbitrary constants

• Fn for arbitrary functions
• c[n] for arbitrary constants when separating the variables
• s for a parameter in the characteristic system
• &where for the solution structure
• ε for a Lie group parameter
We also use the following notation for Maple solutions:
• Eqn for equations (n=1,2,...)

• PDEn/ODEn for PDEs/ODEs
• Soln for solutions
• Trn for transformations
• Sysn for systems
• IC, BC, IBC for initial and/or boundary conditions
• Ln for lists of expressions
• Gn for graphs of solutions
22.1. Introduction 1279
22.1.2 Brief Introduction to Maple

Maple is a general-purpose computer algebra system (CAS) in which symbolic computa-
tion can readily be combined with exact and approximate (floating-point) numerical com-
putation as well as with arbitrary-precision numerical computation. Maple provides pow-
erful scientific graphics capabilities [for details, see Kreyszig (2000), Corless (1995), Heck
(2003), Richards (2002), Abel and Braselton (2005), Meade et al. (2009), Shingareva and
Lizárraga-Celaya (2009), etc.].
The first concept of Maple and its initial versions were developed by the Symbolic
Computation Group at the University of Waterloo in the early 1980s. The Maplesoft com-
pany was created in 1988. Maple was mainly developed in research labs at Waterloo Uni-
versity and at the University of Western Ontario [see Char et al. (1990), Geddes, Czapor,
and Labahn (1992)], with important contributions from worldwide research groups at other
universities.
The most important features of Maple are as follows: fast symbolic and numerical com-
putation and interactive visualization; easy usage; easy incorporation of new user-defined
capabilities; understandable open-source software development path; availability on al-
most all operating systems; powerful programming language, intuitive syntax, and easy
debugging; extensive library of mathematical functions and specialized packages; free re-
sources and the collaborative character of development (for example, see the Maple website
www.maplesoft.com (MWS), Maple Application Center (MWS/applications), Teacher Re-
source Center (MWS/TeacherResource), Student Help Center (MWS/studentcenter), and
Maple Community (MWS/community)).
Maple consists of three parts: the interface, the kernel (basic computational engine),
and the library. The interface and the kernel (written in C programming language) form a
smaller part of the system (they are loaded when a Maple session is started). The interface
handles the input of mathematical expressions, output display, plotting of functions, and
support of other user communication with the system. The user interface is the Maple
worksheet. The kernel interprets the user input, carries out the basic algebraic operations,
and deals with storage management. The library consists of two parts: the main library
and additional packages. The main library (written in the Maple programming language)
includes many functions in which most of the common mathematical knowledge of Maple
resides.
Basic concepts. The prompt symbol (>) serves for typing a Maple function; the semi-
colon (; ) or colon (:) symbol1 followed by pressing Enter at the end of a function serves
for evaluating the Maple function, displaying the result, and inserting a new prompt.
Maple contains a complete online help system and a command line help system, which
can be used, e.g., by typing ?NameOfFunction or help(NameOfFunction); or ?help; ref-
erence information can also be accessed by using the Help menu, by highlighting a function
and then pressing Ctrl-F1 or F1 or F2 (for Release ≥ 9), and by pressing Ctrl-F2.
Maple worksheets are files that keep track of the working process and organize it as a
collection of expandable groups (see ?worksheet and ?shortcut). It is recommended to
begin a new worksheet (or a new problem) with the restart statement for cleaning Maple
1 In earlier versions of Maple and in Classic Worksheet Maple, we have to end a function with a colon or
semicolon. In these chapters, we follow this tradition in every example and problem.
memory. All examples and problems presented in the book are assumed to begin with
restart.
Previous results (during a session) can be referred to with the symbols % (the last result),
%% (the next-to-last result), %%...% (k + 1 times) (the kth-to-last result).
Comments start with the sharp symbol # and include all subsequent characters until the
end of the line. Text can also be inserted with Insert → Text.
Incorrect response. If you get no response or an incorrect response, you may have
entered or executed a function incorrectly. Correct the function or interrupt the computation
(the Stop button in the Tool Bar menu).
Maple source code can be viewed for most of the functions, general and specialized
(package functions); e.g., interface(verboseproc=2); print(factor);
Palettes can be used for building or editing mathematical expressions without needing
to remember the Maple syntax.
The Maplet User Interface (for Release ≥ 8) consists of Maplet applications, which are
collections of windows, dialogs, and actions (see ?Maplets).
A number of specialized functions are available in various specialized packages (sub-
packages) (see ?index[package], with).
Numerical approximations: numerical approximation of expr to 10 significant digits,
evalf(expr); global change of precision, Digits:=n (see ?environment); local change
of the precision, evalf(expr,n); numerical approximation of expr using a binary hard-
ware floating-point system, evalhf(expr); performing numerical approximations using
the hardware or software floating-point systems, UseHardwareFloats:=value (for de-
tails, see ?UseHardwareFloats, ?environment).
22.1.3 Maple Language

Maple language is a high-level programming language, which is well structured and com-
prehensible. It supports a large collection of data structures, or Maple objects (functions,
sequences, sets, lists, arrays, tables, matrices, vectors, etc.), and operations on these objects
(type-testing, selection, composition, etc.). Maple procedures in the library are available in
readable form. The library can be supplemented with locally developed user programs and
packages.
Arithmetic operators: + - * / ˆ ; logic operators: and, or, xor, implies, not; rela-
tion operators: <, <=, >, >=, =, <>.
A variable name is a combination of letters, digits, and/or the underline symbol ( ),
starting from a letter; e.g., a12 new.
Abbreviations for the longer Maple functions or any expressions: alias, for example,
alias(H=Heaviside); diff(H(t),t); to remove this abbreviation, type alias(H=H);
Maple is case sensitive; i.e., it distinguishes between lowercase and uppercase letters;
e.g., evalf(Pi) and evalf(pi) are distinct commands.
Various reserved keywords, symbols, names, and functions; these words cannot be used
as variable names, e.g., operator keywords, additional language keywords, global names
that start with ( ) (see ?reserved, ?ininames, ?inifncs, ?names).
The assignment/unassignment operators: a variable can be “free” (with no assigned
value) or can be assigned any value (symbolic or numeric) by the assignment operators
a:=b or assign(a=b). To unassign (clear) an assigned variable (see ?:= and ?'), type,
e.g., x:='x', evaln(x), or unassign('x').
The difference between the (:=) and (=) operators is as follows: the operator A:=B is
used to assign B to the variable A, and the operator A=B is used to indicate equality (not as-
signment) between the left- and right-hand sides (see ?rhs), for example, Equation:=A=B;
Equation; rhs(Equation); lhs(Equation);
The range operator (..), an expression of type range expr1..expr2, for example,
a[i]$ i=1..9; plot(sin(x),x=-Pi..Pi);
Statements are keyboard input instructions that are executed by Maple (e.g., break, by,
do, end, for, function, if, proc, restart, return, save, while).
The statement separators semicolon (;) and colon (:). The result of a statement followed
with a semicolon (;) will be displayed, and it will not be displayed if it is followed by a
colon (:), e.g., plot(sin(x),x=0..Pi); plot(sin(x),x=0..Pi):
An expression is a valid statement and is formed as a combination of constants, vari-
ables, operators, and functions. Every expression is represented as a tree structure in which
each node (and leaf) has a particular data type. For the analysis of any node and branch,
the functions type, whattype, nops, and op can be used. A boolean expression is formed
with logical operators and relation operators.
An equation is represented using the binary operator (=) and has two operands, the
left-hand side, lhs, and the right-hand side, rhs.
Inequalities are represented using the relation operators and have two operands, the
left-hand side, lhs, and the right-hand side, rhs.
A string is a sequence of characters having no value other than itself; it cannot be as-
signed to, and it will always evaluate to itself. For instance, x:="string"; and sqrt(x);
is an invalid function. Names and strings can be used with the convert and printf func-
tions.
Maple is sensitive to types of brackets and quotes.
Types of brackets: parentheses for grouping expressions, (x+9)*3, for delimiting the
arguments of functions, sin(x); square brackets for constructing lists, [a,b,c], vectors,
matrices, arrays; curly brackets for constructing sets, {a,b,c}.
Types of quotes: forward-quotes to delay the evaluation of expression, 'x+9+1'; to
clear variables, x:='x'; back-quotes to form a symbol or a name, `the name:=7`; k:=5;
print(`the value of k is`,k); double quotes to create strings; and a single double
quote " to delimit strings.
Types of numbers: integer, rational, real, complex, and root, e.g., -55, 5/6, 3.4, -2.3e4,
Float(23,-45), 3-4*I, Complex(2/3,3); RootOf( Zˆ3-2,index=1);
Predefined constants: symbols for commonly used mathematical constants, gamma, Pi,
I, true, false, infinity, FAIL, exp(1) (for details, see ?ininames, ?constants).
Functions or function expressions have the form f(x) or expr(args) and represent
a function call, or an application of a function (or procedure) to arguments (args). Ac-
tive functions (beginning with a lowercase letter) are used for computing, e.g., diff, int,
limit. Inert functions (beginning with a capital letter) are used for showing steps in the
problem-solving process, e.g., Diff, Int, Limit.
Library functions (or predefined functions) and user-defined functions.
Predefined functions: most of the well-known functions are predefined by Maple, and
they are known to some Maple functions (e.g., diff, evalc, evalf, expand, series,
simplify). Numerous special functions are defined (see ?FunctionAdvisor).
User-defined functions: the functional operator (->) (see ?->); e.g., the function f (x) =
sin x is defined as f:=x->sin(x);
Alternative definitions of functions: unapply converts an expression to a function, and
a procedure is defined with proc.
Evaluation of function f (x) at x = a, {x = a, y = b}, e.g., f(a); subs(x=a, f(x));
eval(f(x),x=a);
In the Maple language, there are two forms of modularity: procedures and modules. A
procedure (see ?procedure) is a block of statements which one needs to use repeatedly. A
procedure can be used to define a function (if the function is too complicated to be written
with the use of the arrow operator) or to create a matrix, graph, or logical value. A module
(see ?module) is a generalization of the procedure concept. While a procedure groups
a sequence of statements into a single statement (block of statements), a module groups
related functions and data.
In the Maple language, there are essentially two control structures: the selection struc-
ture if and the repetition structure for.
Maple objects, sequences, lists, sets, tables, arrays, vectors, and matrices are used for
representing more complicated data. Sequences a1,a2,a3, lists [a1,a2,a3], and sets
{a1, a2, a3} are groups of expressions. Maple preserves the order and repetitions in se-
quences and lists and does not preserve them in sets. The order in sets can change during
a Maple session. A table is a group of expressions represented in tabular form. Each entry
has an index (an integer or an arbitrary expression) and a value (see ?table). An array is a
table with integer range of indices (see ?Array). In Maple, arrays can be of any dimension
(depending of computer memory). A vector is a one-dimensional array with a positive in-
teger range of indices (see ?vector, ?Vector). A matrix is a two-dimensional array with
positive integer ranges of indices (see ?matrix, ?Matrix).
⊙ References for Section 22.1: J. Calmet and J. A. van Hulzen (1983), A. G. Akritas (1989), B. W. Char,
K. O. Geddes, G. H. Gonnet, M. B. Monagan, and S. M. Watt (1990), K. O. Geddes, S. R. Czapor, and
G. Labahn (1992), J. H. Davenport, Y. Siret, and E. Tournier (1993), R. M. Corless (1995), M. J. Wester
(1999), E. Kreyszig (2000), D. Richards (2002), A. Heck (2003), M. L. Abel and J. P. Braselton (2005),
L. Debnath (2007), D. B. Meade, S. J. M. May, C-K. Cheung, and G. E. Keough (2009), I. K. Shingareva and
C. Lizárraga-Celaya (2009, 2011).
22.2 Analytical Solutions and Their Visualizations

22.2.1 Constructing Analytical Solutions in Terms of Predefined Functions
The computer algebra system Maple has various predefined functions based on symbolic
algorithms for constructing analytical solutions of linear PDEs [see a more detailed descrip-
tion in Cheb-Terrab and von Bulow (1995)]. The predefined functions implement known
methods for solving PDEs, and Maple allows solving linear equations and obtaining solu-
tions automatically (in terms of predefined functions) as well as developing new methods
and procedures for constructing new solutions.
22.2. Analytical Solutions and Their Visualizations 1283
Consider the most important functions for finding all possible analytical solutions of a
given problem for PDEs.
pdsolve(PDE); pdsolve(PDE,build); pdetest(Sol,PDE);

pdsolve(PDE,funcs,HINT=val,INTEGRATE,build,singsol=val);
infolevel[procname]:=val; pdsolve(PDEsys); with(PDEtools);
declare(funcs); dchange(rules,PDE); diff_table(funcs);
separability(PDE,func); casesplit(PDEs); charstrip(PDE, func);
Laplace(PDE,func,ops); Solve(sys, vars, ops); mapde(PDE, form, ops);
Infinitesimals(PDE); SymmetrySolutions(Sol, Infinitesimals, ops);
TWSolutions(PDE,ops);
Remark. HINT=val gives some hints; with build one can construct an explicit expression for
the indeterminate function func; with ‘+‘ or ‘*‘ one can construct a solution by separation
of variables (in the form of sum or product, respectively); with 'TWS' or 'TWS(MathFuncName)'
one can construct a traveling wave solution as a power series in tanh(ξ) or several mathe-
matical functions (including special functions), where ξ represents a linear combination of
the independent variables, etc.
pdsolve finding analytical solutions for a given partial differential equation PDE and for
systems of PDE
PDEtools a collection of functions for finding analytical solutions for PDE, e.g.,
declare declaring functions and derivatives on the screen for a simple, compact
display
casesplit splitting into cases and successively decoupling a system of differential
equations
separability determining under what conditions it is possible to obtain a complete
solution by separation of variables
charstrip finding the characteristic strip (for a first-order PDE)
Laplace solving a second-order linear PDE (in two independent variables) using
the Laplace method
Solve finding exact, series, or numerical solutions for systems of algebraic or dif-
ferential equations (including inequalities, initial and boundary conditions)
mapde converting a PDE into a PDE of different form
SymmetrySolutions finding point symmetry solutions
TWSolutions constructing traveling wave solutions (for autonomous PDEs and sys-
tems of them), etc.
Let us assume that we have obtained exact solutions and wish to verify whether these
solutions are exact solutions of given linear PDEs.
Example 22.1. Linear Poisson equation. Verification of solutions. For the two-dimensional
Poisson equation 7.2.2
n n
uxx + uyy + Φ(x, y) = 0, Φ(x, y) = ∑ ∑ ai j exp(bi x + c j y),
i=1 j=1
we verify that the solution,

n n
ai j
u(x,t)= − ∑ ∑ 2 2
exp(bi x + c j y),
i=1 j=1 bi + c j
is an exact solution of the Poisson equation (with a special right-hand side) as follows:
with(PDEtools); declare(u(x,y)); U:=diff_table(u(x,y));

PDE1:=U[x,x]+U[y,y]+sum(sum(a[i,j]*exp(b[i]*x+c[j]*y),j=1..n),i=1..n)=0;
Sol1:=u(x,y)=-sum(sum(a[i,j]/(b[i]ˆ2+c[j]ˆ2)*exp(b[i]*x+c[j]*y),j=1..n),i=1..n);
Test1:=simplify(combine(pdetest(Sol1,PDE1)));
where ai j , bi , c j (i = 1, . . . n, j = 1, . . . n) are arbitrary real constants.
Example 22.2. Linear heat equation. Verification of solutions. For the one-dimensional heat
equation 1.1.1
ut = kuxx ,
we verify that
1 x2
u(x,t) = √ exp − , x ∈ R, t > 0,
2 πkt 4kt
is the fundamental solution of this linear heat equation as follows:
with(PDEtools); declare(u(x,t)); U:=diff_table(u(x,t));

PDE1:=U[t]=k*U[x,x]; Sol1:=u(x,t)=1/sqrt(4*Pi*k*t)*exp(-xˆ2/(4*k*t));
Test1:=pdetest(Sol1,PDE1);
Here k is a constant.
Example 22.3. Linear wave equation. General solution. The general solution of the one-
dimensional wave equation 4.1.1
utt = c2 uxx
can be found and tested as follows:
with(PDEtools); declare(u(x,t)); PDE1:=diff(u(x,t),t$2)=cˆ2*diff(u(x,t),x$2);

Sol1:=pdsolve(PDE1); Test1:=pdetest(Sol1,PDE1);
where the Maple result reads:

Sol1 := u (x,t) = F1 (ct + x) + F2 (ct − x)
Example 22.4. Linear Euler equation. Method of characteristics. The first-order linear Euler
equation
xux + yuy = nu
can be solved by the method of characteristics as follows:
with(PDEtools): declare(u(x,y));
PDE1:=x*diff(u(x,y),x)+y*diff(u(x,y),y)=n*u(x,y);
sysCh:=charstrip(PDE1,u(x,y)); funcs:=indets(sysCh,Function);
Sol1:=dsolve(sysCh,funcs,explicit); GenSol1:=pdsolve(PDE1);
where the Maple result reads:

sysCh := {u s = nu ( s) , x s = x ( s) , y s = y ( s)} , funcs := {u ( s) , x ( s) , y ( s)}
y
Sol1 := {u ( s) = C3 en s , x ( s) = C2 e s , y ( s) = C1e s } , GenSol1 := u (x, y) = F1 xn
x
Remark. For the given equation (PDE1), we first obtain the characteristic system (depending on the
parameter s) via charstrip and then solve this system via dsolve. Additionally we find the general
solution of the linear Euler equation.
Example 22.5. Linear wave equation with two space variables. Analytical solutions. For the
wave equation with two space variables in the rectangular Cartesian system of coordinates,
utt = c2 (uxx + uyy ),
which arises in many physical problems and describes wave processes in various media (e.g., shal-
low water waves, linearized supersonic flow in a gas, sound waves in space, electrical oscillations
in a conductor, torsional vibration of a rod, longitudinal vibrations of a bar, waves in magnetohy-
drodynamics, etc.), we can construct various forms of analytical solutions of the wave equation,
e.g.,
with(PDEtools): declare(u(x,y,t)); infolevel[pdsolve]:=5;
PDE1:=diff(u(x,y,t),t$2)=cˆ2*(diff(u(x,y,t),x$2)+diff(u(x,y,t),y$2));
casesplit(PDE1); separability(PDE1,u(x,y,t)); separability(PDE1,u(x,y,t),`*`);
Sol1:=pdsolve(PDE1,HINT=`+`,INTEGRATE); Sol2:=pdsolve(PDE1,HINT=`*`,INTEGRATE);
Sol3:=pdsolve(PDE1,HINT='TWS'); Sol4:=pdsolve(PDE1,HINT='TWS(tan)');
for i from 1 to 4 do Test||i:=pdetest(Sol||i,PDE1) od;
with the Maple output

hnn c3 o
Sol1 := (u (x, y,t) = F1 (x) + F2 (y) + F3 (t)) &where F1 (x) = 1/2 − c2 + 2 x2 + C1 x + C2 ,
c
F2 (y) = 1/2 c2 y2 + C3 y + C4 , F3 (t) = 1/2 c3 t 2 + C5 t + C6
√ √
Sol2 := (u (x, y,t) = F1 (x) F2 (y) F3 (t)) &where F1 (x) = C1 e c1 x + C2 e− c1 x ,
√ √ √ √
F2 (y) = C3 e c2 y + C4 e− c2 y , F3 (t) = C5 sin c − c1 − c2 t + C6 cos c − c1 − c2 t
p 3 p
Sol3 := u (x, y,t) = tanh −t C22 + C32 c + x C2 + C3 y + C1 C8 + tanh −t C22 + C32 c + x C2 + C3 y + C1 C6 + C5
p 3 p
Sol4 := u (x, y,t) = tan −t C22 + C32 c + x C2 + C3 y + C1 C8 + tan −t C22 + C32 c + x C2 + C3 y + C1 C6 + C5
Remark. First, we can split the problem into all related regular cases (the general case and all pos-
sible singular cases). In our situation, we have the general case. Then we check the separability
conditions (additive, by default, and multiplicative): if there is a separable solution, the result is 0.
For our problem, we have 0 in both cases (additive and multiplicative). Finally, we construct ex-
act solutions: additive and multiplicative separable solutions with HINT=‘+‘ and HINT=‘*‘2 ; traveling
wave solutions with HINT='TWS' and HINT='TWS(tan)'.
2 Solving a PDE by separation of variables, we can indicate the option INTEGRATE for the integration of the
set of resulting ODEs.

If we try to construct separable solutions with the functional HINT, e.g., HINT=f(C1*x+C2*y+C3*t)+
g(C1*x+C2*y+C4*t), then Maple cannot find the solution of this form (the solution f (x sin a + y sin a + bt) +
g(x sin a+y sin a−bt)), and, according to the general strategy [see Cheb-Terrab and von Bulow (1995)], applies
another method and gives the result that coincides with one of the known solutions, i.e., the traveling wave
solution (Sol3). However, in a simpler case, e.g., with HINT=f(x,y), Maple finds the solution u(x, y,t) =
F1(y − Ix) + F2(y + Ix).
Example 22.6. Linear elastic beam equation. Separable and self-similar solutions By applying
predefined Maple functions to the linear elastic beam equation
utt + α2 uxxxx = 0,
we can find additive separable and self-similar solutions and verify that these solutions are exact
solutions:
with(PDEtools): declare(u(x,t));
PDE1:=diff(u(x,t),t$2)+alphaˆ2*diff(u(x,t),x$4);
Sol1:=pdsolve(PDE1); Sol2:=collect(pdsolve(PDE1,build),[cos,sin]);
Sol3:=pdsolve(PDE1,HINT=`+`,build); Sol4:=[SimilaritySolutions(PDE1)];
for i from 1 to 4 do Test||i:=pdetest(Sol||i,PDE1); od;
with the Maple output

hn c2 oi
Sol1 := (u (x,t) = F1 (x) + F2 (t)) &where F2t,t (t) = c2 , F1x,x,x,x (x) = −
α2

√
4 √
4 C2 C6 C1 C6 √
Sol2 := u (x,t) = C3 ei c1 x C6 + C4 e c1 x C6 + √ 4 c x + √4
i c1 x
cos (α c1 t)
e 1 e
√
4 √
4 C2 C5 C1 C5 √
+ C3 ei c1 x C5 + C4 e c1 x C5 + √ 4 c1 x
+ i√ 4 c x sin (α c1 t)
e e 1
1 c2 x4 1 1 1
Sol3 := u (x,t) = − + C1 x3 + C2 x2 + C3 x + c2 t 2 + C5t + C6 + C4
24 α2 6 2 2

1 1
Sol4 := u (x,t) = C1t + C2,u (x,t) = C1 x3 + C2 x2 + C3 x + C4,
 6 2  
√
 1 3 5 3 1 x4  1 2
u (x,t) = C3 hypergeom ,1 , , , ,− 2
x + C4 FresnelS q t
2 4 4 2 64 α2 t 2 2 √π√α t
x2
 √   
1 2
+ C2 FresnelC  √ √ q  t + C1t  t −1 
2 π α t
x2
Remark. We start from the function pdsolve(PDE1) (without options) and obtain the structure of the
exact solution (multiplicative separable solution) Sol1, for which we build an explicit expression
in Sol2. Then we find the additive separable solution Sol3 via the option HINT. And then we construct
similarity solutions in Sol4.
Example 22.7. Three-dimensional wave equation. Traveling wave solutions. By applying

Maple predefined functions to the three-dimensional wave equation
utt = c2 (uxx + uyy + uzz),
(where c is a parameter), we can find traveling wave solutions and verify that these solutions are
exact solutions of the given PDE as follows:
with(PDEtools): declare(u(x,y,z,t)); PDE1:=diff(u(x,y,z,t),t$2)=
cˆ2*(diff(u(x,y,z,t),x$2)+diff(u(x,y,z,t),y$2)+diff(u(x,y,z,t),z$2));
Sol1:=pdsolve(PDE1,HINT='TWS'); Test1:=pdetest(Sol1,PDE1);
with Maple results

p 3
Sol1 := u (x, y, z,t) = C9 tanh − C22 + C32 + C42 ct + C2 x + C3 y + C4 z + C1
p
+ C7 tanh − C22 + C32 + C42 ct + C2 x + C3 y + C4 z + C1 + C6
In general, hyperbolic linear PDEs can have interesting traveling wave solutions. However, for
nonlinear equations the effects of nonlinear and higher-order terms interact and can lead to interest-
ing solutions [see Shingareva and Lizárraga-Celaya (2011), Shingareva and Lizárraga-Celaya (pp.
1622–1765, 2012)]. By applying the function TWSolutions (with various options) of the package
PDEtools, one can construct various types of traveling wave solutions in Maple; for example, we
have
infolevel[TWSolutions]:=2; Fs:=TWSolutions(functions_allowed);
Sol21:=TWSolutions(PDE1,singsol=false,functions=Fs); N:=nops([Sol21]);
for i from 1 to N do op(i,[Sol21]) od;
Sol22:=TWSolutions(PDE1,singsol=false,parameters=[c],functions=[cot],
remove_redundant=true); Sol23:=TWSolutions(PDE1,function=identity,output=ODE);
with the Maple results

p 3
Sol22 := c = c,u (x,y,z,t) = cot − C22 + C32 + C42 ct + C2 x + C3 y + C4 z + C1 C9
p o
+ cot − C22 + C32 + C42 ct + C2 x + C3 y + C4 z + C1 C7 + C6 ...

Sol23 := C42 − c2 C12 + C22 + C32 uτ,τ ,{τ = C1 x + C2 y + C3 z + C4t + C5,u (τ) = u (x,y,z,t)}
Remark. By adding the function infolevel with values (0–5), we can obtain increasingly detailed in-
formation on each of the steps of the solution process. Then, with the aid of the predefined function
TWSolutions, we can find various types of traveling wave solutions (of PDEs and systems) according
to a list of functions Fs (instead of the tanh function, by default). The algorithm is based on the tanh-
function method and its various extensions [see Shingareva and Lizárraga-Celaya (2011)]. Includ-
ing the options parameters (splitting into cases with respect to the parameters), singsol (avoiding
singular solutions), functions (expanding in series of different functions), remove redundant=true
(removing all redundant solutions, i.e., all particular cases of more general solutions being con-
structed), we can obtain numerous traveling wave solutions (Sol21). For example, by specifying the
cot function (and other options), we can obtain various traveling wave solutions. (We present just
one solution in Sol22.)
Sometimes it is useful to obtain a reduction (without computing a solution) that converts a PDE
into an ODE form and the corresponding ODE.
By default, the function TWSolutions generates solutions of the form fi (τ)= ∑nk=0 i
Aik τk , where
j
the ni are finite, the Aik are constants with respect to x j , and τ = tanh(∑i=0 Ci xi ). In Sol23, we
obtain an ODE by using a simpler form, i.e., by choosing the identity as the function (the option
j
function=identity), where τ becomes τ = ∑i=0 Ci xi . In our case, we have u(x, y, z,t) = u(τ), where
τ = C1 x + C2y + C3 z + C4t + C5 , and the ODE acquires the form C42 − c2 C12 + C22 + C32 uτ,τ = 0
(Sol23).
Exact solutions of various types of linear PDEs. Finally, by applying Maple prede-
fined functions, we find exact solutions (general, traveling wave, separable, similarity, and
symmetry) and characteristic systems of various types of linear PDEs (where a, b, c are
arbitrary parameters). The results are presented in Table 22.1.
Table 22.1.
Maple predefined functions and exact solutions of various types of linear PDEs
No. Linear PDEs Exact solution Maple function
1 xux + yuy = u {u s = u ( s) , x s = x ( s) , y s = y ( s)} charstrip(Eq1,u(x,y))

ct + x ct − x
2 utt = c2 uxx u = F1 + F2 Laplace(Eq2,u(x,t))
c c
C1 sin(ct) + C2 cos(ct) − e−x

3 utt + c2 uxx = 0 u == SimilaritySolutions(Eq3,S[2])
e−x
4 ux − uy = u u = F1 (x + y) ex pdsolve(Eq4)
p
5 utt − c2 uxx + a2 u = 0 u = C4 sin C22 c2 + a2 t + C2 x + C1 pdsolve(Eq5,HINT=TWS(sin))
√ √ √
ax ax
sin √ C1− cos √ C2 b c x
6 ut + aux + buxxx = 0 u= c2 t+ C3+ C4+ b
√ b
− 2 pdsolve(Eq6,HINT=f(x)+g(t),ops)
a a
7 ux − uy = 1 u = x + F1 (x + y) pdsolve(Eq7,build)
n √ √ o
8 ut = kuxx u = C3 e2 c1 x C1 + C2 e c1 k(t− ε)− c1 x SymmetrySolutions(Sol80,S8[1])
hn oi
9 utt − c2 uxx + a2 u = 0 (u = f (x) g (t)) &where fxx = c1 f (x) , gtt (t) = c2 c1 g (t) − a2 g (t) pdsolve(Eq9,HINT=f(x)*g(t))
n o
10 utt = c2 uxx u= C7 (tanh ( C2 ct+ C2x+ C1))3 + C5 tanh ( C2ct+ C2x+ C1) + C4 TWSolutions(Eq10,singsol=false)

−y + x
11 x2 ux + y2 uy = (x + y)u u = F1 xy Solve(Eq11,u(x,y))
yx
where the linear partial differential equations and some additional solutions are defined in
Maple as follows:
with(PDEtools): declare(u(x,t));
Eq1:=x*diff(u(x,y),x)+y*diff(u(x,y),y)=u(x,y);
Sol1:=charstrip(Eq1,u(x,y));
Eq2:=diff(u(x,t),t$2)=cˆ2*diff(u(x,t),x$2); Sol2:=Laplace(Eq2,u(x,t));
Eq3:=diff(u(x,t),t$2)+cˆ2*diff(u(x,t),x$2)=0; Infinitesimals(Eq3);
S:=Infinitesimals(Eq3,specialize_Fn); Sol3:=SimilaritySolutions(Eq3,S[2]);
Eq4:=diff(u(x,y),x)-diff(u(x,y),y)=u(x,y); Sol4:=pdsolve(Eq4);
Eq5:=diff(u(x,t),t$2)-cˆ2*diff(u(x,t),x$2)+aˆ2*u(x,t)=0;
Sol5:=pdsolve(Eq5,HINT=TWS(sin));
Eq6:=diff(u(x,t),t)+a*diff(u(x,t),x)+b*diff(u(x,t),x$3)=0;
ops:=INTEGRATE,build; Sol6:=collect(pdsolve(Eq6,HINT=f(x)+g(t),ops),[a,b]);
Eq7:=diff(u(x,y),x)-diff(u(x,y),y)=1; Sol7:=pdsolve(Eq7,build);
Eq8:=diff(u(x,t),t)=k*diff(u(x,t),x$2); Sol80:=pdsolve(Eq8,build);
S8:=[Infinitesimals(Eq8)]; Sol8:=SymmetrySolutions(Sol80,S8[1]);
Eq9:=diff(u(x,t),t$2)-cˆ2*diff(u(x,t),x$2)+aˆ2*u(x,t)=0;
Sol9:=pdsolve(Eq9,HINT=f(x)*g(t));
Eq10:=diff(u(x,t),t$2)=cˆ2*diff(u(x,t),x$2);
Sol10:=[TWSolutions(Eq10,singsol=false)]; Sol10[3];
Eq11:=xˆ2*diff(u(x,y),x)+yˆ2*diff(u(x,y),y)=(x+y)*u(x,y);
Sol11:=Solve(Eq11,u(x,y));
Remark. In the above table, for the linearized KdV equation (Eq6), an additive separable so-
lution is presented. The automatic integration of the ODE found by separation of variables
is performed with the option INTEGRATE, and an explicit expression for the unknown func-
tion is constructed with the option build. The infinitesimals introduced for solving the third
and the eighth equations refer to the list containing the components of the infinitesimal
generator of a symmetry group. Note that the infinitesimals for equation 3 involve arbitrary
functions; therefore, to work with these infinitesimals, we add the option specialize Fn. The
solution of equation 8 contains an arbitrary constant ε (Lie group parameter). Equations 6
and 9 are solved by separation of variables; in this case, according to Maple notation, the
arbitrary constants are denoted by c1 and c2 (so as to distinguish them from the usual
arbitrary constants C1 , C2 , etc.).
22.2.2 Constructing General Solutions via the Method of Characteristics

Consider the method for finding general solutions of first-order linear equations, the La-
grange method of characteristics. This method allows reducing a PDE to a system of
ODEs along which the given PDE with some initial data (the Cauchy data) is integrable.
Once the system of ODEs is found, it can be solved along the characteristic curves and
transformed into a general solution of the original PDE. This method was first proposed by
Lagrange in 1772 and 1779 for solving first-order linear and nonlinear PDEs.
In general, a first-order PDE in two independent variables (x, y) can be written in the
form
F(x, y, u, ux , uy ) = 0 or F(x, y, u, p, q) = 0, (22.2.2.1)
where F is a given function, u = u(x, y) is the unknown function of the independent vari-
ables x and y ((x, y) ∈ D ⊂ R2 ), ux = p, and uy = q. Equation (22.2.2.1) is said to be linear
if the function F is linear in the variables u, ux , and uy and the coefficients of these variables
are functions of the independent variables x and y alone. The most general first-order linear
PDE has the form
A(x, y)ux + B(x, y)uy +C(x, y)u = G(x, y). (22.2.2.2)
The solution u(x, y) of Eq. (22.2.2.1) can be visualized geometrically as a surface, called
an integral surface in the (x, y, u)-space.
A direction vector field or the characteristic direction, (A, B,C), is a tangent vector to
the integral surface f (x, y, u)=0 at a point (x, y, u).
A characteristic curve is a curve in (x, y, u)-space such that the tangent at each point
coincides with the characteristic direction field (A, B,C).
The parametric equations of the characteristic curve can be written in the form x = x(t),
y = y(t), u = u(t), and the tangent vector to this curve (dx/dt, dy/dt, du/dt) is equal to
(A, B,C).
The characteristic equations of Eq. (22.2.2.2) in parametric form is the system of ODEs
dx/dt = A(x, y), dy/dt = B(x, y), du/dt = C(x, y).
The characteristic equations of Eq. (22.2.2.2) in nonparametric form are dx/A(x, y) =
dy/B(x, y) = du/C(x, y). The slopes of the characteristics are determined by the equation
dy/dx = B(x, y)/A(x, y).
Example 22.8. First-order linear equations. Direction vector fields. By applying predefined
Maple functions to a first-order linear PDE of the form
ux + ut + u = 0, xux + tut + (x2 + t 2 ) = 0,

where {x ∈ R,t ≥ 0}, we can construct the direction vector fields for the given first-order linear
PDE as follows:
with(plots): R1:=-Pi..Pi: N1:=10: V1:=[-24,72]; V2:=[-59,-29];

A1:=`THICK`;setoptions3d(fieldplot3d,grid=[N1,N1,N1],axes=boxed);
r1:=(x,t,u)->[1,1,-u]; r2:=(x,t,u)->[x,t,-(xˆ2+tˆ2)];
fieldplot3d(r1(x,t,u),x=R1,t=R1,u=R1,arrows=SLIM,orientation=V1);
fieldplot3d(r2(x,t,u),x=R1,t=R1,u=R1,arrows=A1,orientation=V2);
The general solution (or general integral) of the given first-order PDE is an equation of
the form
f (φ, ψ)=0, (22.2.2.3)
where f is an arbitrary function of the known functions, φ = φ(x, y, u) and ψ=ψ(x, y, u),
and provides a solution of this partial differential equation. The functions φ(x, y, u) = C1 ,
ψ(x, y, u) = C2 (where C1 and C2 are constants) are solution curves of the characteristic
equations or the families of characteristic curves of Eq. (22.2.2.2).
Example 22.9. First-order equation. General solution. By applying the method of characteris-
tics to the first-order linear PDE
xux + yuy = u,
where {x ∈ R, y ∈ R}, we can prove that the general solution of the given first-order PDE takes the
form f (y/x, u/x) = 0 or u(x, y) = xg (y/x):
with(PDEtools): declare(u(x,y)); U:=diff_table(u(x,y));

F:=x; G:=y; H:=U[]; PDE:=F*U[x]+G*U[y]=H; CharEqs:=[dx/F,dy/G,du/H];
Eq10:=map(exp,combine(int(1/G,y)-int(1/F,x)=C10) assuming x>0);
Eq20:=map(exp,combine(int(1/u,u)-int(1/F,x)=C20) assuming x>0);
Eq11:=subs(rhs(Eq10)=C1,Eq10); Eq21:=subs(rhs(Eq20)=C2,Eq20);
GenSol:=f(lhs(Eq11),lhs(Eq21))=0; Test1:=pdetest(subs(u=U[],GenSol),PDE);
pdsolve(PDE,u(x,y)); GenSol1:=op(2,lhs(GenSol))=g(op(1,lhs(GenSol)));
GenSol2:=isolate(GenSol1,u); Test2:=pdetest(subs(u=U[],GenSol2),PDE);
Remark. The system of characteristic equations dx/x = dy/y = du/u gives the integral surfaces
φ = y/x = C1 , ψ = u/x = C2 (Eq11, Eq21, and C1 , C2 are arbitrary constants); therefore, according
to Eq. (22.2.2.3), the general solution of the linear PDE is f (y/x, u/x) = 0 (GenSol), where f is
an arbitrary function. This solution can be verified (Test1). By yapplying
the predefined function
pdsolve, we can find the result in a different form: u (x, y) = xg . This form of general solution
x
can be obtained (GenSol2) and verified (Test2).
Example 22.10. Linear Euler equation. General solution. By applying the method of charac-
teristics to the linear Euler equation
xux + yuy = nu,
form f (y/x, u/xn ) = 0 or u(x, y) = xn g (y/x) as follows:
interface(showassumed=0): assume(x>0,y>0,u>0,n>0); with(PDEtools):

declare(u(x,y)); U:=diff_table(u(x,y)); F:=x; G:=y; H:=n*U[];
PDE:=F*U[x]+G*U[y]=H; CharEqs:=[dx/F,dy/G,du/H];
Eq10:=map(exp,combine(int(1/G,y)-int(1/F,x)=C10) assuming x>0);
Eq20:=map(exp,int(1/(n*u),u)-int(1/F,x)=C20);
Eq200:=simplify(map(combine,Eq20)); Eq11:=subs(rhs(Eq10)=C1,Eq10);
Eq21:=subs(rhs(Eq200)=C21,Eq200); Eq22:=simplify(lhs(Eq21)ˆn)=C2;
GenSol:=f(lhs(Eq11),lhs(Eq22))=0; Test1:=pdetest(subs(u=U[],GenSol),PDE);
pdsolve(PDE,u(x,y)); GenSol1:=op(2,lhs(GenSol))=g(op(1,lhs(GenSol)));
GenSol2:=simplify(isolate(GenSol1,u)); Test2:=pdetest(subs(u=U[],GenSol2),PDE);
Remark. The system of characteristic equations dx/x = dy/y = du/(nu) gives the integral surfaces:
φ = y/x = C1 , ψ = x−n u = C2 (Eq11, Eq22, and C1 , C2 are arbitrary constants); therefore, according
to Eq. (22.2.2.3), the general solution of the linear PDE is f (y/x, x−n u) = 0 (GenSol), where f is
an arbitrary function. This solution can be verified (Test1). By applying the predefined function
y
pdsolve, we can find the result in a different form: u (x, y) = xn g x . This form of general solution
can be obtained (GenSol2) and verified (Test2).
22.2.3 Constructing General Solutions via Transformations to Canonical

Forms
First, consider the general first-order linear partial differential equation
A(x, y)ux + B(x, y)uy +C(x, y)u = G(x, y).
By introducing a new transformation by the equations ξ = ξ(x, y), η = η(x, y),3 we can
transform the original equation into a canonical (or standard) form
uξ + α(ξ, η)u = β(ξ, η),
where α(ξ, η) = C̃/Ã and β(ξ, η) = G̃/Ã, Ã = uξx + Bξy, B̃ = Aηx + Bηy, and the equation
is Ãuξ + B̃uη + C̃u = G̃. This equation can be integrated, and the general solution of the
original equation can be found.
Example 22.11. Linear first-order equation. Canonical form. General solution. Considering
the linear first-order PDE
ux + uy = u,
where {x ∈ R, y ∈ R}, we can obtain the transformation ξ = −x + y, η = x (Eq xi, Eq eta) and
the canonical form uη = u (CanForm). By integrating this equation, we find the general solution
u(x, y) = ex F(−x + y) (where F(−x + y) is an arbitrary function). Finally, we can find and test the
general solution using predefined functions as follows:
with(VectorCalculus): with(PDEtools): declare(u(x,y));

U:=diff_table(u(x,y)); V:=diff_table(u(xi,eta)); A:=1; B:=1; C:=-1; G:=0;
PDE:=A*U[x]+B*U[y]+C*U[]=0; CharEqs:=[dx/A,dy/B,du/U[]];
Eq_xi:=xi=int(1/B,y)-int(1/A,x); Eq_eta:=eta=x;
TestJ:=diff(rhs(Eq_xi),x)*diff(rhs(Eq_eta),y)-diff(rhs(Eq_xi),y)*diff(rhs(Eq_eta),x);
Vx:= V[xi]*diff(rhs(Eq_xi),x)+V[eta]*diff(rhs(Eq_eta),x);
Vy:= V[xi]*diff(rhs(Eq_xi),y)+V[eta]*diff(rhs(Eq_eta),y);
CanForm:=A*Vx+B*Vy=-C*V[]; CanForm1:=CanForm/(rhs(CanForm));
Eq1:=int(lhs(CanForm1),eta)=int(rhs(CanForm1),eta)+ln(F(xi)); Eq2:=isolate(Eq1,V[]);
GenSol:=U[]=subs({Eq_xi,Eq_eta},rhs(Eq2)); GenSolPred:=pdsolve(PDE,U[]);
Test1:=pdetest(GenSolPred,PDE); Test2:=pdetest(GenSol,PDE);
3 Here the functions ξ and η are continuously differentiable and the Jacobian, J(x, y) = ξ η − ξ η , is
x y y x
nonzero in a domain D.
For linear second-order PDEs, we consider the classification of equations (that does not
depend on their solutions and is determined by the coefficients of the highest derivatives)
and the reduction of a given equation to appropriate canonical forms.
Let us introduce the new variables a=F p , b= 12 Fq , c=Fr , and calculate the discriminant
δ=b2 −ac at some point. Depending on the sign of the discriminant δ, the type of the
equation at a specific point can be parabolic (if δ=0), hyperbolic (if δ > 0), and elliptic (if
δ < 0). Let us call the equations
uy1 y2 = f1 (y1 , y2 , u, uy1 , uy2 ), uz1 z1 − uz2 z2 = f2 (z1 , z2 , u, uz1 , uz2 ),
respectively, the first canonical form and the second canonical form for hyperbolic PDEs.
Example 22.12. Linear second-order equation. Classification. Canonical forms. Considering
the linear second-order PDE
−2y2 uxx + 12 x2 uyy = 0,
where {x ∈ R, y ∈ R}, we can verify that this equation is hyperbolic everywhere (except at the point
x=0, y=0) and prove that a change of variables, ξ = − 12 x2 + y2 and η = 12 x2 + y2 , transforms the
PDE, respectively, to the first and second canonical forms:
vξ η − vη ξ 1 vλ vµ
vηξ + =0, vλλ − vµµ + − = 0.
2(η2 − ξ2 ) 2 λ µ
Classification. In the standard notation (22.1.1.3), this linear equation takes the form F1 =
−2y2 p + 12 x2 r = 0. The new variables are a= − 2y2 , b=0, c= 12 x2 (tr2(F1)) and the discriminant
δ=b2 −ac=x2 y2 (delta1) is positive except at the point x=0, y=0.
with(PDEtools): declare(u(x,y),F1(p,r,q)); U:=diff_table(u(x,y));

PDE1:=-2*yˆ2*U[x,x]+xˆ2*U[y,y]/2=0;
tr1:=(x,y,U)->{U[x,x]=p,U[y,y]=r,U[x,y]=q};
tr2:=F->{a=diff(lhs(F(p,q,r)),p),b=1/2*diff(lhs(F(p,q,r)),q),c=diff(lhs(F(p,q,r)),r)};
delta:=bˆ2-a*c; F1:=(p,r,q)->subs(tr1(x,y,U),PDE1); F1(p,r,q); tr2(F1);
delta1:=subs(tr2(F1),delta)-rhs(F1(p,r,q));
is(delta1,'positive'); coulditbe(delta1,'positive');
The same result can be obtained with the principal part coefficient matrix as follows:
interface(showassumed=0): assume(x<0 or x>0, y<0 or y>0);

with(LinearAlgebra): A1:=Matrix([[-2*yˆ2,0],[0,xˆ2/2]]);
D1:=Determinant(A1); is(D1,'negative'); coulditbe(D1,'negative');
Remark. Here we calculate the determinant D1 of the matrix A1. The PDEs can be classified accord-
ing to the eigenvalues of the matrix A1, i.e., depending on the sign of D1: if D1=0, parabolic, if D1<0,
hyperbolic, and D1>0, elliptic equations.
Canonical forms. We find a change of variables that transforms the PDE to the first and second
canonical forms as follows:
with(LinearAlgebra): with(VectorCalculus): with(PDEtools): declare(v(xi,eta));

interface(showassumed=0): vars:=x,y; varsN:=xi,eta;
Op1:=Expr->subs(y=y(x),Expr); Op2:=Expr->subs(y(x)=y,Expr);
m1:=simplify((-A1[1,2]+sqrt(-D1))/A1[1,1],radical,symbolic);
m2:=simplify((-A1[1,2]-sqrt(-D1))/A1[1,1],radical,symbolic);
Eq1:=dsolve(diff(y(x),x)=-Op1(m1),y(x));
Eq11:=lhs(Eq1[1])ˆ2=rhs(Eq1[1])ˆ2; Eq12:=solve(Eq11,_C1); g1:=Op2(Eq12);
Eq2:=dsolve(diff(y(x),x)=-Op1(m2),y(x)); Eq21:=lhs(Eq2[1])ˆ2=rhs(Eq2[1])ˆ2;
Eq22:=solve(Eq21,_C1); g2:=Op2(Eq22); Jg:=Jacobian(Vector(2,[g1,g2]),[vars]);
dv:=Gradient(v(varsN),[varsN]); ddv:=Hessian(v(varsN),[varsN]);
ddu:=Jgˆ%T.ddv.Jg+add(dv[i]*Hessian(g||i,[vars]),i=1..2);
Eq3:=simplify(Trace(A1.ddu))=0;
tr1:={isolate(subs(isolate(g1=xi,xˆ2),g2=eta),yˆ2),
isolate(subs(isolate(g1=xi,yˆ2),g2=eta),xˆ2)};
CanForm1:=collect(expand(subs(tr1,Eq3)),diff(v(varsN),varsN));
c1:=coeff(lhs(CanForm1),diff(v(varsN),varsN));
CanFormF:=collect(CanForm1/c1,diff(v(varsN),varsN));
CanForm2:=expand(expand(dchange({xi=lambda+mu,eta=mu-lambda},CanFormF))*(-4));

x2 uxx + 2xyuxy + y2uyy = 0,
where {x ∈ R, y ∈ R}; we can verify that this equation is parabolic everywhere and prove that the
canonical forms of the PDE are x2 vξξ = 0 and vξξ = 0, respectively.
with(LinearAlgebra): with(VectorCalculus): with(PDEtools): declare(v(xi,eta));

Op1:=Expr->subs(y=y(x),Expr); Op2:=Expr->subs(y(x)=y,Expr); vars:=x,y; varsN:=xi,eta;
A1:=Matrix([[xˆ2,x*y],[x*y,yˆ2]]); D1:=Determinant(A1);
m1:=simplify((-A1[1,2]+sqrt(-D1))/A1[1,1],radical,symbolic);
Eq1:=dsolve(diff(y(x),x)=-Op1(m1),y(x)); Eq11:=solve(Eq1,_C1);
g1:=Op2(Eq11); g2:=x; Jg:=Jacobian(Vector(2,[g1,g2]),[vars]);
dv:=Gradient(v(varsN),[varsN]); ddv:=Hessian(v(varsN),[varsN]);
ddu:=Jgˆ%T.ddv.Jg+add(dv[i]*Hessian(g||i,[vars]),i=1..2);
CanForm1:=simplify(Trace(A1.ddu))=0; CanForm2:=expand(CanForm1/xˆ2);
22.2.4 Constructing Analytical Solutions of Cauchy Problems
If we consider a mathematical problem in an unbounded domain, then the solution can be

determined uniquely by prescribing initial conditions. The corresponding problem is called
the initial value problem, or the Cauchy problem.
Mathematical problems can be well-posed or ill-posed. A mathematical problem is
well-posed if it possesses the following features: existence (there exists at least one solu-
tion), uniqueness (there exists at most one solution), and continuity (the solution depends
continuously on the data).
In Maple, it is possible to construct exact solutions for linear (and nonlinear) partial
differential equations subject to initial conditions with the aid of the predefined function
pdsolve. Let us solve some Cauchy problems.
Example 22.14. Linear first-order PDEs. Initial value problem (IVP). Exact solution. Consid-
ering the Cauchy problem for the linear first-order partial differential equation,
yux + xuy = u, u(x, 0) = F(x),

Figure 22.1 Characteristics for the linear equation ux − uy = 1.
and applying Maple predefined functions, we p can solve the Cauchy problem for this equation and
F( x2 − y2 )(x + y)
verify that the solution obtained, u(x, y) = p , is an exact solution of this initial
x2 − y2
value problem as follows:
with(PDEtools): declare(F(x)); PDE1:=y*diff(u(x,y),x)+x*diff(u(x,y),y)=u(x,y);

IC1:=u(x,0)=F(x); sys1:=[PDE1,IC1]; Sol:=pdsolve(sys1);
Test1:= simplify(pdetest(Sol,sys1),symbolic);
Using different Maple notation, we can obtain the same result:
with(PDEtools): declare(F(x)); U:=diff_table(u(x,y)): PDE1:=y*U[x]+x*U[y]=U[];

IC1:=eval(U[],t=0)=f(x); IC1:=eval(U[],y=0)=F(x); sys1:=[PDE1,IC1];
Sol:=pdsolve(sys1); Test1:= simplify(pdetest(Sol,sys1),symbolic);
Example 22.15. Linear first-order equation. Method of characteristics. Classical Cauchy

problem. Considering the first-order linear partial differential equation with the Cauchy data,
ux − uy = 1, u(x, 0) = xn (n ∈ N),
and applying the method of characteristics, we can obtain the general and particular solutions of
this equation,
u(x, y) = f (x + y) + x, u(x, y) = (y + x)n − y,
respectively, and plot the characteristic curves (see Fig. 22.1) as follows:
with(plots): xR:=0..9; yR:=0..9; ODE:=diff(U(x),x)=1; tr1:=U(x)=u; tr2:=y(x)=y;

Sol_Ch:=dsolve({ODE,U(0)=C[2]}); Eq_Ch:=diff(y(x),x)=-1;
Cur_Ch:=dsolve({Eq_Ch, y(0)=C[1]});
display([seq(plot([subs(C[1]=y,eval(y(x),Cur_Ch)),x,x=xR],color=blue,thickness=2),
y=yR)],view=[yR,xR],scaling=constrained);
Eq1:=subs(tr1,isolate(Sol_Ch,C[2])); Eq2:=subs(tr2,isolate(Cur_Ch,C[1]));
Eq3:=rhs(Eq1)=f(rhs(Eq2)); F:=unapply(lhs(subs({y=0,u=xˆn},Eq3)),x);
u:=solve(Eq3,u); Sol1:=simplify(subs( f(rhs(Eq2))=F(rhs(Eq2)),u));
The same result can be obtained applying predefined functions:
with(PDEtools): declare(v(x,y)); PDE:=diff(v(x,y),x)-diff(v(x,y),y)=1;

IC:=v(x,0)=xˆn; sys1:=[PDE,IC]; Sol2:=pdsolve(sys1);
Test:= simplify(pdetest(Sol2,sys1),symbolic);
Example 22.16. First-order linear equation. Method of characteristics. Classical Cauchy

problem. Let us solve the first-order linear equation with Cauchy data,
ut − xux = u, u(x, 0) = f (x),
by applying the method of characteristics. This equation can be obtained from the Fokker–Planck
equation [see Bluman et al. (2010)],4
ut = uxx + (xu)x ,
by neglecting the term uxx .

We can obtain the solution of this Cauchy problem
u1 (x,t) = f (xet ) et
and plot the characteristic curves as follows:
with(plots); interface(showassumed=0); assume(t>0); assume(X[0],constant);

tR:=0..6; xR:=-10..10; Ops1:=color=blue,thickness=2; Ops2:=view=[xR,tR];
tr1:=x=x(t); tr2:=x(t)=x; tr3:=f(X[0])=f1(X[0]); f1:=x->x;
ODE1:=diff(x(t),t)=-x(t); ODE2:=diff(U(t),t)=U(t);
Sol2:=dsolve(ODE2,U(t)); Sol21:=subs(_C1=C,rhs(Sol2));
IniCond:=u(X[0],0)=f(X[0]); Const2:=evala(subs(t=0,Sol21))=rhs(IniCond);
Sol22:=U(t)=subs(Const2,Sol21); ODE11:=combine(subs(Sol22,ODE1));
Sol1:=dsolve(ODE11,x(t)); X0:=expand(subs(t=0,Sol1));
Const1:=isolate(X0,_C1); Chars:=subs(Const1,x(0)=X[0],Sol1);
GenSol:=rhs(Sol22); Char1:=subs(tr2,tr3,Chars);
trX0:=simplify(isolate(Char1,X[0]));
u1:=unapply(subs(trX0,GenSol),x,t); u1(x,t);
u2:=unapply(subs(f(X[0])=f1(rhs(trX0)),GenSol),x,t); u2(x,t);
display([seq(plot([subs(X[0]=x,eval(x(t),Char1)),t,t=tR],Ops1),x=xR)],Ops2);
display([seq(plot([u2(x,t),t,t=tR],Ops1),x=xR)],Ops2);
We show that the implicit form of the solution (or parametric representation of the solution) of this
Cauchy problem has the Maple form
Sol22 := U (t) = f (X0 ) et Chars := x (t) = X0 e−t
The characteristic curves for f (x) = x are presented in Fig. 22.2.
Note that there are some difficulties in stating Cauchy problems for parabolic and ellip-
tic equations. By way of example, consider the Hadamard example [see Hadamard (1952)].
4 The Fokker–Planck equation arises in various applications of statistical mechanics; it describes the evolu-
tion of the probability distribution function.

Figure 22.2 Characteristic curves for ut − xux = u and the solution u2 (x,t) (for f (x) = x).
Example 22.17. Laplace equation. Hadamard example. Cauchy problem. Considering an

elliptic equation, e.g., the Laplace equation with the initial conditions proposed by Hadamard [see
Hadamard (1952)],
1
uxx + uyy = 0, u(x, 0) = 0, uy (x, 0) = An sin(nx), An = (p > 0),
np
we obtain the solution of this Cauchy problem (SolF)
u(x, y) = sinh(ny) sin(nx)n−1−p
and verify that this solution is an exact solution of the Laplace equation as follows:
PDE1:=diff(u(x,y),x$2)+diff(u(x,y),y$2)=0; A[n]:= 1/nˆp;

SGen:=pdsolve(PDE1,u(x,y),HINT=sin(n*x)*h(y),build); SGen1:=eval(subs(_c[1]=1,SGen));
Eq1:=eval(rhs(SGen1),y=0)=0; Eq2:=eval(diff(rhs(SGen1),y),y=0)=A[n]*sin(n*x);
trConst:=op(solve({Eq1,Eq2},[_C1,_C2]));
Sol1:=subs(trConst,SGen1); SolF:=simplify(convert(Sol1,trig));
Test1:=pdetest(SolF,PDE1);
However, the solution of this problem does not depend continuously on the initial data (i.e., a
small change in the initial data produces a small change in the solution). These difficulties arise in
connection with the Cauchy–Kowalevski theorem [for details, see Petrovsky (1991)]. This can be
shown as follows:
limit(A[n]*sin(n*x),n=infinity) assuming p>0;

for i from 1 to 30 by 10 do plot(subs({n=i,y=1,p=1},rhs(SolF)),x=0..10); od;
for i from 1 to 10 by 3 do plot(subs({n=i,x=1,p=1},rhs(SolF)),y=0..1); od;
plot(subs({n=9,x=1,p=1},rhs(SolF)),y=0..10);
The function (Eq2), which describes the initial data, tends to zero as n tends to infinity, i.e.,
lim An sin(nx) = 0. If n → ∞, the solution (SolF) represents oscillations for any y 6= 0 (the first
n→∞
sequence of plots). Also, if n → ∞, the amplitude of the solution tends to infinity as y tends to
infinity for any x 6= 0 (the second sequence of plots). If n is a fixed number, the solution tends to
infinity as y tends to infinity for any x 6= 0 (for which sin(nx) 6= 0). Therefore, this problem is an
example of an ill-posed problem.
22.2.5 Constructing Analytical Solutions of Boundary Value Problems

A boundary value problem is a problem of finding an unknown function that satisfies a
given partial differential equation and particular boundary conditions. Boundary value
problems are associated with partial differential equations of elliptic type and are consider-
ably more difficult to solve.
Boundary value problems can be divided into two classes: interior and exterior bound-
ary value problems. For exterior boundary value problems, part of the boundary is at infin-
ity and their solutions should satisfy an additional requirement (boundedness at infinity).
Interior boundary value problems can have various types of boundary conditions. The most
important types of boundary conditions are the Dirichlet condition (the solution u is given
on the boundary), the Neumann condition (∂u/∂n is given on the boundary), the mixed con-
dition or the Robin condition (∂u/∂n + hu is given on the boundary, or boundary conditions
of distinct types are given on distinct parts of the boundary),5 and the periodic boundary
conditions.
In Maple, exact solutions of linear partial differential equations with boundary condi-
tions can be constructed with the aid of the predefined function pdsolve (for details, see
?pdsolve[boundaryconditions]). Consider various types of boundary value problems.
Example 22.18. Laplace equation. Dirichlet boundary value problem (BVP). Considering the
Laplace equation and the Dirichlet boundary conditions on the rectangular region
uxx + uyy = 0, 0 < x < a, 0 < y < b;
2
u(x, 0) = x , u(x, b) = x − b2 ,
2
u(0, y) = −y2 , u(a, y) = a2 − y2 ,
we obtain the general solution (SGen) of this boundary value problem (by applying the option
HINT=g(x)+h(y) for additive separable form of solution)
1 1
u(x, y) = − c2 x2 + C1x + C2 + c2 y2 + C3y + C4.
2 2
Then, considering the given boundary conditions, we get the constants (trConst), obtain the final
form of the analytical solution
u(x,t) = x2 − y2 ,
and verify that this solution is an exact solution of the Laplace equation as follows:
PDE1:=diff(u(x,y),x$2)+diff(u(x,y),y$2)=0;
SGen:=pdsolve(PDE1,u(x,y),HINT=g(x)+h(y),build); SGen1:=eval(subs(_c[2]=-2,SGen));
Eq1:=eval(rhs(SGen1),y=0)=xˆ2; Eq2:=eval(rhs(SGen1),y=b)=xˆ2-bˆ2;
Eq3:=eval(rhs(SGen1),x=0)=-yˆ2; Eq4:=eval(rhs(SGen1),x=a)=aˆ2-yˆ2;
S1:=op(solve({Eq1,Eq2,Eq3,Eq4},[_C1,_C2,_C3,_C4]));
trConst:=[op(1..3,subs(_C4=0,S1)),_C4=0];
SolF:=subs(trConst,SGen1); Test1:=pdetest(SolF,PDE1);
Example 22.19. Linear Poisson equation. Dirichlet boundary value problem (BVP). Consider
the Poisson equation and the Dirichlet boundary conditions on the rectangular region
uxx + uyy = f (x, y), a < x < b, c < y < d;
u(x, c) = f1 (x), u(x, d) = f2 (x), u(a, y) = f3 (y), u(b, y) = f4 (y).
5 Here ∂u/∂n is the directional derivative of the solution u along the outward normal to the boundary, and
h is a given continuous function on the boundary.

Such boundary value problems describe a steady-state process u(x, y) in a bounded rectangular
object. Let us choose f (x, y) = cosx sin y, f1 (x) = f2 (x) = 0, f3 (y) = f4 (y) = − 12 sin y, a = 0,
b = π, c = 0, d = 2π. By applying Maple predefined functions, we obtain the analytical solution
u(x, y) = − 12 cos x sin y (Sol12) of the boundary value problem, visualize it in the given region, and
verify that this solution is an exact solution of the linear Poisson equation as follows:
with(linalg): with(PDEtools): f:=(x,y)->cos(x)*sin(y);

PDE1:=laplacian(u(x,y),[x,y])-f(x,y)=0;
Sol1:=pdsolve(PDE1,build); Test1:=pdetest(Sol1,PDE1);
Sol11:=unapply(subsop(1=0,2=0,rhs(Sol1)),x,y); Sol12:=expand(Sol11(x,y));
Sol11(x,0); Sol11(x,Pi); Sol11(0,y); Sol11(2*Pi,y);
plot3d(Sol12,x=0..Pi,y=0..2*Pi,shading=zhue);
22.2.6 Constructing Analytical Solutions of Initial-Boundary Value Problems

If a mathematical problem is to find an unknown function satisfying a PDE (defined at an
appropriate domain) and appropriate supplementary conditions (initial and boundary con-
ditions), this is known as an initial-boundary value problem. The boundary conditions de-
scribe the unknown function at prescribed boundary points. The initial condition prescribes
the unknown function at a certain initial time t (e.g., t = t0 , t = 0).
In Maple, it is possible to construct exact solutions for an increasing number of linear
(and nonlinear) partial differential equations subject to initial and boundary conditions with
the aid of the predefined function pdsolve (see ?pdsolve[boundaryconditions]). Let
us solve some initial-boundary value problems.
Example 22.20. Linear telegraph equation. Initial-boundary value problem (IBVP). Consider-
ing the initial-boundary value problem for the linear telegraph equation,
uxx = utt + ut − u, x > 0, t > 0;
x
u(x, 0) = e , ut (x, 0) = −2e , x
u(0,t) = e−2t , ux (0,t) = e−2t , x ≥ 0, t ≥ 0,
and applying Maple predefined functions, we obtain the solution of this initial-boundary value prob-
lem (Sol2)
u(x,t) = e−2t+x
and verify that this solution is an exact solution of the linear telegraph equation as follows:
with(PDEtools): declare(u(x,t)); U:=diff_table(u(x,t)):
PDE1:=U[x,x]=U[t,t]+U[t]-U[]; Sol1:=pdsolve({PDE1},{u(x,t)},build);
IC1:=u(x,0)=exp(x),D[2](u)(x,0)=-2*exp(x);
BC1:=u(0,t)=exp(-2*t),D[1](u)(0,t)=exp(-2*t); Sol2:=pdsolve([PDE1,IC1,BC1]);
T1:=pdetest(Sol2,PDE1); T2:=simplify(subs(t=0,Sol2));
T3:=D[2](u)(x,0)=simplify(subs(t=0,diff(rhs(Sol2),t)));
T4:=simplify(subs(x=0,Sol2)); T5:=D[1](u)(0,t)=simplify(subs(x=0,diff(rhs(Sol2),x)));
Example 22.21. Linear wave equation. Initial-boundary value problem (IBVP). Considering a
semiinfinite vibrating string with a fixed end, we solve the following initial-boundary value problem
for the linear wave equation:
utt = c2 uxx , x > 0, t > 0;
u(x, 0) = f (x), ut (x, 0) = g(x) u(0,t) = 0 x ≥ 0, t ≥ 0.
The boundary condition at x = 0 produces a wave moving to the right with velocity c. By apply-
ing Maple predefined functions with appropriate assumptions (c > 0, x > 0, t > 0), we obtain the
solution of this initial-boundary value problem (Sol2)
 Z

 1 1 ct+x
 [ f (ct + x) − f (ct − x)] + g(ζ), x < ct,
2 2c ct−x
u(x,t) = Z ct+x

 1 1
 [ f (−ct + x) + f (ct + x)] + g(ζ), x > ct,
2 2c −ct+x
and verify that this solution is an exact solution of the linear wave equation as follows:
with(PDEtools): declare(u(x,t)); U:=diff_table(u(x,t)): PDE1:=U[t,t]=cˆ2*U[x,x];

Sol1:=pdsolve({PDE1},{u(x,t)},build); IC1:=u(x,0)=f(x),D[2](u)(x,0)=g(x);
BC1:=u(0,t)=0; Sol2:=pdsolve([PDE1,IC1, BC1]) assuming c>0, x>0, t>0;
T1:=pdetest(u(x,t)=op(2,op(2,Sol2)),PDE1);
T2:=pdetest(u(x,t)=op(4,op(2,Sol2)),PDE1);
T3:=simplify(subs(t=0,Sol2)) assuming c>0, x>0;
T4:=D[2](u)(x,0)=simplify(subs(t=0,diff(rhs(Sol2),t))) assuming c>0, x>0;
T5:=simplify(subs(x=0,Sol2)) assuming c>0, t>0;
22.2.7 Constructing Analytical Solutions of Systems of Linear PDEs

Computer algebra system Maple has various predefined functions based on symbolic al-
gorithms for constructing analytical solutions of systems of linear PDEs. These functions
allow solving linear systems and obtaining solutions automatically as well as developing
new methods and procedures for constructing new solutions. Just as before, we consider
the most important functions for finding analytical solutions of a given nonlinear system.
pdsolve(PDESys,[DepVars],HINT=val,singsol=val,mindim=N,parameters=P);
pdetest(Sol,PDESys); pdsolve([PDESys,ICs],[DepVars],series,order=N);
with(PDEtools); separability(PDESys,DepVars); casesplit(PDESys);
splitsys(PDESys,DepVars); ConsistencyTest(PDESys,ops);
TWSolutions(PDESys,ops); SimilaritySolutions(PDESys,ops);
ReducedForm(PDESys1,PDESys2);
Remark. ICs are initial conditions; DepVars is a set or list of dependent variables, indicating
the solving ordering, where slight changes in the solving ordering can lead to different
solutions or make a system unsolvable (by pdsolve) a solvable system.
pdsolve finding analytical solutions for a given PDE system; relevant options are: HINT,
some hints; singsol, computing or not computing singular solutions; mindim, a
minimum dimension of the solution space; parameters, indicating the solving vari-
ables with less priority; series, computing formal power series solutions,
PDEtools a collection of functions for finding analytical solutions for PDESys, e.g.,
splitsys splitting a PDE system into subsets (each one with equations coupled
among themselves but not coupled to the equations of the other subsets)
ConsistencyTest testing whether a given system of equations is or is not consistent

ReducedForm reducing one given PDE system with respect to another given PDE
system, etc.
Example 22.22. Linear hyperbolic system. General solution. Consider the linear first-order
hyperbolic system
ut + k1 ux + k2 vx = 0, vt + k2 ux + k1 vx = 0,
where k1 and k2 are parameters (k1 , k2 ∈ R). By applying Maple predefined functions, we solve this
linear hyperbolic system as follows:
with(PDEtools): declare((u,v)(x,t));
U,V:=diff_table(u(x,t)),diff_table(v(x,t)):
Sys1:=[U[t]+k[1]*U[x]+k[2]*V[x]=0,V[t]+k[2]*U[x]+k[1]*V[x]=0];
Sol1:=collect(pdsolve(Sys1,[U[],V[]]),[t]); pdetest(Sol1,Sys1);
We obtain the general solution (Sol1) of this linear system, and in the Maple notation it reads:
Sol1:= {u(x,t)=− F2 ((k1−k2)t−x) + F1 ((k1+k2)t−x) + C1, v(x,t)= F1 ((k1+k2)t−x) + F2 ((k1−k2)t−x)}
It can be shown that the solution obtained consists of two independent parts, one propagating with
speed k1 + k2 and another with speed k1 − k2 .
By applying Maple predefined functions, we will show how to find analytical solutions of
the Cauchy problem and boundary value problem for linear second-order systems.
Example 22.23. Linear second-order system. Cauchy problem. Considering the linear second-
order system and the Cauchy data,
vt − uxx = 0, ut + vxx = 0, u(x, 0) = ex , v(x, 0) = e−x ,
and applying Maple predefined functions, we solve the Cauchy problem for this system and verify
that the solution obtained, u(x,t) = (ex − x − 1) cos(t) − e−x sin(t) + x + 1, v(x,t) = ex cos(t) + (ex −
x − 1) sin(t) (S2), is an exact solution of this initial value problem as follows:
with(PDEtools): declare((u,v)(x,t));
U,V:=diff_table(u(x,t)),diff_table(v(x,t)): Sys1:=V[t]-U[x,x]=0,
U[t]+V[x,x]=0; Sol1:=pdsolve([Sys1],[U[],V[]],build);
Sol2:=subs(_c[1]=1,Sol1); IC1:=eval(rhs(Sol2[1]),t=0)=exp(x);
IC2:=eval(rhs(Sol2[2]),t=0)=exp(-x);
trConst1:=_C6=1,_C4=1,_C3=0,_C1=0,_C7=1,_C8=1;
IC11:=subs(trConst1,IC1); IC12:=subs(trConst1,IC2);
trConst2:=op(solve({IC11,IC12}));
S20:=subs(trConst2,trConst1,Sol2);
S2:=map(collect,expand(S20),[cos,sin,exp]);
Test1:=pdetest(Sol2,[Sys1]);
Test2:=simplify(subs(trConst2,trConst1,{IC1,IC2}));
Example 22.24. Linear second-order system. Boundary value problem. Considering the linear
second-order system and the boundary conditions,
vt − uxx = 0, ut + vxx = 0,
u(0,t) = 2[cos(t) − sin(t)], v(0,t) = 2[cos(t) + sin(t)],
1 + e2 1 + e2
u(1,t) = [cos(t) − sin(t)], v(1,t) = [cos(t) + sin(t)],
e e
22.3. Analytical Solutions of Mathematical Problems 1301
and applying Maple predefined functions, we solve the boundary value problem for this system and
verify that the solution obtained (SolF),
1 + e2x 1 + e2x
u(x,t) = [cos(t) − sin(t)], v(x,t) = [cos(t) + sin(t)],
ex ex
is an exact solution of this initial value problem as follows:
with(PDEtools): declare((u,v)(x,t)); U,V:=diff_table(u(x,t)),diff_table(v(x,t)):

Sys1:=V[t]-U[x,x]=0, U[t]+V[x,x]=0; BC1:=u(0,t)=2*(cos(t)-sin(t));
BC2:=v(0,t)=2*(cos(t)+sin(t)); BC3:=u(1,t)=((1+exp(1)ˆ2)/exp(1))*(cos(t)-sin(t));
BC4:=v(1,t)=((1+exp(1)ˆ2)/exp(1))*(cos(t)+sin(t));
Sol1:=pdsolve([Sys1],[U[],V[]],build); Sol2:=subs(_c[1]=1,Sol1);
S1:=simplify(subs(BC1,BC2,subs(x=0,Sol2)));
S2:=simplify(subs(BC3,BC4,subs(x=1,Sol2))); Sys2:= S1 union S2;
trConst1:=_C3=0,_C1=0,_C4=1,_C6=1; Sys3:=subs(trConst1, Sys2);
vars3:=remove(has,indets(Sys3),t); trConst2:=solve(Sys3,vars3);
SolF0:=subs(trConst1,trConst2,Sol2);
SolF:=factor(map(collect,expand(SolF0),[cos,sin]));
Test1:=pdetest(SolF,[Sys1]); Test2:=seq(subs(x=0,rhs(SolF[i]))=rhs(BC||i),i=1..2);
Test3:=seq(subs(x=1,rhs(SolF[i]))=rhs(BC||(i+2)),i=1..2);
simplify([Test2]); factor(combine([Test3]));
⊙ References for Section 22.2: J. Hadamard (1952), I. Petrovsky (1991), E. S. Cheb-Terrab and K. von
Bulow (1995), G. W. Bluman, A. F. Cheviakov, and S. C. Anco (2010), I. K. Shingareva and C. Lizárraga-
Celaya (2011, 2012).
22.3 Analytical Solutions of Mathematical Problems

22.3.1 Constructing Separable Solutions
Separation of variables is one of the most important methods for solving linear PDEs,
in which the structure of a PDE allows us to seek multiplicative separable or additive
separable exact solutions, e.g., u(x,t) = φ(x) ◦ ψ(t) (where the multiplication or addition
is denoted by ◦). Numerous problems in linear partial differential equations can be solved
by separation of variables. This method has been generalized [see, e.g., recent papers by
Galaktionov (1990, 1995), Polyanin and Zhurov (1998), Polyanin and Manzhirov (2007)],
and nowadays it is one of the classical methods in mathematics and physics.
Let us start from first-order linear equations, which can be solved by separation of
variables without considering Fourier series.
Example 22.25. First-order linear equation. Separable solution. Cauchy problem. Consider
the first-order linear PDE and the Cauchy data
aux + buy = 0, u(0, y) = αe−βy ,
where a, b ∈ R are parameters. By applying separation of variables and by seeking exact solutions
in the form u(x,t) = φ(x)ψ(y), we arrive at the following equations (Eq61, Eq62):
aφ′x ψ′y
− = C1 , = C1 .
bφ(x) ψ(y)
with(PDEtools): declare(u(x,y),W(x,y),phi(x),psi(y));
interface(showassumed=0): assume(n,'integer',n>0): tr1:=phi(x)*psi(y);
PDE1:=u->a*diff(u(x,y),x) +b*diff(u(x,y),y); IC1:= u(0,y)=alpha*exp(-beta*y);
Eq2:=expand(PDE1(W)); Eq3:=expand(subs(W(x,y)=tr1,Eq2));
Eq4:=expand(Eq3/phi(x)/psi(y)); Eq5:=isolate(Eq4,psi(y));
Eq61:=rhs(Eq5)=_C1; Eq62:=lhs(Eq5)=_C1;
Then we seek exact solutions of these equations as follows:
Sol1:=dsolve(Eq61,phi(x)); Sol2:=dsolve(Eq62,psi(y));
GenSol:=u(x,y)=simplify(subs(Sol1,Sol2,tr1)); trC3:=_C2ˆ2=_C3;
GenSol1:=subs(trC3,GenSol); Eq8:=subs(x=0,rhs(GenSol1))=rhs(IC1);
trC13:=_C1=-beta,_C3=alpha; SolF:=subs(trC13,GenSol1);
Test1:=pdetest(SolF,PDE1(u)); Test2:=subs(x=0,rhs(SolF))=rhs(IC1);
We find that φ(x) = C2 e−C1 bx/a (Sol1), ψ(y) = C2 eC1 y (Sol2), and the exact solution acquires the
form (SolF)
u(x, y) = αe−β(ay−bx)/a .
Separation of variables combined with the linear superposition principle can be applied for
solving a large class of initial-boundary value problems for linear partial differential equa-
tions. According to the method, the partial differential equation is reduced to two ordinary
differential equations (for φ(x) and ψ(y)). A similar idea can be applied to equations in
several independent variables. This method is also known as the Fourier method or the
eigenfunction expansion method.
First, let us obtain a solution of the initial-boundary value problem with the aid of the
predefined Maple function pdsolve, which applies separation of variables and the Fourier
method.
Example 22.26. Linear heat equation. Separable solution. Fourier series. Initial-boundary
value problem.
Consider the linear heat equation with the initial and boundary conditions
ut = cuxx , 0 < x < L, t > 0,

u(x, 0) = f (x), u(0,t) = 0, u(L,t) = 0,
where c is a real parameter, L = 1, and f (x) = x3 .

If we apply the predefined Maple function pdsolve for the linear heat equation (which invokes
separation of variables in this case), we can obtain the multiplicative separable solution (Sol1)
√
c1 x c c1 t ec c1t
u(x,t) = C1 C3 e e + C2 C3 √
e c1 x
and solve the given initial-boundary value problem step by step.
If we apply the predefined Maple function pdsolve for the given initial-boundary value problem
(which invokes separation of variables and the Fourier method in this case), we can solve this
problem (automatically) and obtain the particular solution (SolF),
∞ 2(−1)1+k (πk)2 − 6

2
u(x,t) = ∑ 3
sin(πkx)e−c(πk) t .
k=1 (πk)
Finally, we verify (symbolically, numerically, and graphically) that this solution satisfies the linear
heat equation and the initial and boundary conditions as follows:
with(PDEtools): with(plots): declare(u(x,t)); U:=diff_table(u(x,t)):

PDE1:=U[t]=c*U[x,x]; Sol1:=pdsolve({PDE1},{u(x,t)},build); f:=x->xˆ3;
IC1:=u(x,0)=f(x); BC1:=u(0,t)=0, u(1,t)=0;
Sol2:=pdsolve([PDE1,IC1,BC1]); SolF:=simplify(subs(_Z1=k,Sol2));
T1:=pdetest(SolF,PDE1); T2:=simplify(subs(t=0,SolF));
T3:=simplify(subs(x=0,SolF)); T4:=simplify(subs(x=1,SolF));
G1:=plot(f(x),x=0..1,color=magenta):
G2:=plot(evalf(sum(op(1,rhs(T2)),k=1..100)),x=0..1,color=blue): display({G1,G2});
Example 22.27. Linear wave equation. Separable solution. Fourier series. Initial-boundary
value problem.
Consider the second-order linear hyperbolic PDE with the initial and boundary conditions
utt = c2 uxx , 0 < x < L, t > 0,

u(x, 0) = f (x), ut (x, 0) = g(x) (0 ≤ x ≤ L), u(0,t) = 0, u(L,t) = 0 (t ≥ 0),
where c2 = 14 , L = 1, f (x) = 0, g(x) = sin(x) − sin(3πx). This problem describes a vibrating string
(with constant tension T and density ρ, c2 = T /ρ) stretched along the x-axis from 0 to L and fixed
at the endpoints. The initial displacement f (x) is zero, and the initial velocity is g(x).
By separation of variables, i.e., by seeking the exact solution in the form u(x,t) = φ(x)ψ(t), we
obtain the two ODEs
φ′′xx − λφ = 0, ψtt′′ − λc2ψ = 0.
Then, separating the boundary conditions, we solve the eigenvalue problem and obtain the solution
nπct nπct nπx
u(x,t) = An cos + Bn sin sin ,
L L L
which satisfies the original wave equation and the boundary conditions. Since the wave equation is
linear and homogeneous, by the superposition principle, the infinite series
∞ nπct nπct nπx
u(x,t) = ∑ An cos
L
+ Bn sin
L
sin
L
n=1
is a solution as well. We assume the following properties of the solution: it converges and is
twice continuously differentiable with respect to x and t. Since each term of the series satisfies
the boundary conditions, it follows that the series satisfies these conditions. From the two initial
conditions, we can determine the constants An and Bn . By differentiating the solution with respect
to t, we obtain
∞ nπx ∞
nπc nπx
u(x, 0) = f (x) = ∑ An sin L
, ut (x, 0) = g(x) = ∑ Bn L
sin
L
.
n=1 n=1
These equations are satisfied if f (x) and g(x) can be represented by Fourier sine series. According
to the Fourier theory, the formulas for the coefficients An and Bn read
Z L nπx Z L nπx
2 2
An = f (x) sin dx, Bn = g(x) sin dx.
L 0 L nπc 0 L
Finally, we obtain the analytical and graphical solutions of the initial-boundary value problem
as follows:
with(plots): L:=1: c:=1/4: N:=10: IC1:=sin(x)-sin(3*Pi*x);

u:=(x,t)->Sum((An*cos(n*Pi*c*t/L)+Bn*sin(n*Pi*c*t/L))*sin(n*Pi*x/L),n=1..N);
u(x,t); An1:=solve(u(x,0)=0,An); EqIC:=expand(eval(diff(u(x,t),t),t=0));
Eq1:=expand(Bn*Pi*add(n*sin(n*Pi*x),n=1..N),sin);
for i from 1 to N do
C[i]:=select(has,Eq1,sin(i*Pi*x)); IC[i]:=select(has,IC1,sin(i*Pi*x));
B[i]:=solve(C[i]=IC[i],Bn);
end do:
U:=unapply(add(subs(Bn=B[n],An=An1,op(1,u(x,t))),n=1..N),x,t);
animate(U(x,t),x=0..L,t=0.1..5,frames=100,color=blue,thickness=3);
G:=animate(subs(t=j,U(x,t)),x=0..L,j=0.1..5,frames=12,tickmarks=[2,2],color=blue):
display(G);
value problem.
Consider the second-order linear parabolic PDE with the initial and boundary conditions
ut = kuxx , 0 < x < L, t > 0,
u(x, 0) = f (x), u(0,t) = 0, u(L,t) = 0,
4
where k = 1/30, L = 1, f (x) = sin (πx). This problem describes a homogeneous rod (of length L).
We assume that the rod is sufficiently thin (i.e., the heat is distributed equally over the cross section
at time t), the surface of the rod is insulated (i.e., there is no heat loss through the boundary), and the
temperature distribution of the rod is given by the solution of the initial-boundary value problem.
By separation of variables, i.e., by seeking the exact solution in the form u(x,t) = φ(x)ψ(t), we
obtain the two ODEs
φ′′xx + λ2φ = 0, ψt′ + λ2kψ = 0.
Then, by separating the boundary conditions, we solve the eigenvalue problem and obtain the solu-
tion nπx 2
u(x,t) = An sin e−(nπ/L) kt ,
L
which satisfies the original heat equation and the boundary conditions. Since the heat equation is
∞ nπx 2
u(x,t) = ∑ An sin e−(nπ/L) kt
n=1 L
is a solution as well. This solution satisfies the initial condition if
∞ nπx
u(x, 0) = f (x) = ∑ An sin ,
n=1 L
where the coefficients An are the Fourier coefficients:
Z nπx
2 L
An = f (x) sin dx.
L 0 L
Finally, by recursively computing the Fourier coefficients and the desired solution, we obtain a
formal series solution of the initial-boundary value problem and visualize it as follows:
with(plots): f:=x->sin(Pi*x)ˆ4; L:= 1; k:=1/30: N:=100: NF1:=30: NF2:=12: NT:=7:
A:=proc(n) option remember; evalf(2/L*Int(f(x)*sin(n*Pi*x/L),x=0..L)) end:
u:=proc(n) option remember; u(n-1)+A(n)*exp(-(n*Pi/L)ˆ2*k*t)*sin(n*Pi*x/L) end:
A(1); u(1):=A(1)*exp(-(Pi/L)ˆ2*k*t)*sin(Pi*x/L); u(N);
A:=animate(subs(t=j,u(N)),x=0..L,j=0..NT,frames=NF2,color=blue): display(A);
animate(subs(t=j,u(N)),x=0..L,j=0..NT,frames=NF1,color=blue);
22.3.2 Constructing Analytical Solutions via Integral Transform Methods

Integral transform methods are the most important methods for constructing analytical so-
lutions of mathematical problems (initial and/or boundary value problems) described by
linear partial differential equations.
The main idea of the methods is to transform the original mathematical problem into
a simpler form whose solutions can be obtained and then inverted (by applying the corre-
sponding inverse integral transform) for representing the solutions in terms of the original
variables. Formal definitions of the most important integral transforms, together with their
general forms and properties, can be found in Debnath (2007).
In Maple, integral transforms (e.g., Fourier, Hankel, Hilbert, Laplace, and Mellin inte-
gral transforms) can be studied with the aid of the inttrans package. Let us solve some
initial-boundary value problems.
Example 22.29. Linear wave equation. Laplace transform. Initial-boundary value problem.
utt = c2 uxx , 0 ≤ x < ∞, t > 0,

u(x, 0) = h(x), ut (x, 0) = g(x) (0 < x < ∞), u(0,t) = A f (t), u(x,t) → 0 as x → ∞,
where h(x) = 0, g(x) = 0, and A is a constant (A ∈ R). This problem describes transverse vibrations
of a semiinfinite string. The initial displacement h(x) and the initial velocity g(x) are zero; i.e., the
string is at rest in its equilibrium position.
Let U(x, s) be the Laplace transform of u(x,t). By transforming the equation of motion and by
substituting the initial conditions, we obtain (Eq4)
s2U − c2Uxx = 0.
The solution of this ordinary differential equation is (Sol)
U = C1 e−sx/c + C2 esx/c .
By transforming the boundary conditions, we obtain (BC3)
U(0, s) = AF(s), lim U(x, s) = 0,

x→0
According to the second condition, we have C2 = 0, and by applying the first condition, we obtain
(Sol3)
U(x, s) = AF(s)e−sx/c .
By applying the inverse Laplace transform, we obtain the solution (SolF)
(
ct − x ct − x 0, t < x/c,
u(x,t) = A H f or u(x,t) =
c c A f (ct − x)/c , t ≥ x/c,
where H is the Heaviside unit step function. This solution represents a wave propagating at a
velocity c with the characteristic x = ct.
Finally, we obtain the the displacement of a semiinfinite string as follows:
with(inttrans): with(PDEtools): interface(showassumed=0): assume(c>0,s>0);

declare(u(x,t),U(x));
Eq1:=diff(u(x,t),t$2)-cˆ2*diff(u(x,t),x$2); Eq2:=laplace(Eq1,t,s);
Eq3:=subs(laplace(u(x,t),t,s)=U(x),Eq2);
IC1:={u(x,0)=0,D[2](u)(x,0)=0}; IC2:=laplace(IC1,t,s);
BC1:=u(0,t)=A*f(t); BC2:=laplace(BC1,t,s);
BC3:=subs({laplace(u(0,t),t,s)=U(0),laplace(f(t),t,s)=F(s)},BC2);
Eq4:=subs(IC1,Eq3); Sol:=dsolve(Eq4,U(x)); Sol1:=rhs(Sol);
l1:=simplify(limit(op(1,Sol1),x=infinity));
l2:=simplify(limit(op(2,Sol1),x=infinity));
Sol2:=subs(_C1=0,Sol1); Sol3:=subs(_C2=op(2,BC3),Sol2);
U1:=subs(F(s)=laplace(f(t),t,s),Sol3);
SolF:=simplify(invlaplace(U1,s,t)) assuming t>0 and x>0;
NumericEventHandler(invalid_operation = `Heaviside/EventHandler`(value_at_zero = 1));
SolF1:=convert(SolF,piecewise,t);
22.3.3 Constructing Analytical Solutions in Terms of Green’s Functions

It is well known that the linear superposition principle is one of the most important methods
for representing solutions of linear PDEs with initial and/or boundary conditions in terms
of eigenfunctions or Green’s functions.6
As was seen earlier (Section 22.3.1), the infinite series representation of solutions of
mathematical problems (involving linear PDEs) can be obtained by applying the eigen-
function expansion method. The integral representation of solutions can be obtained by
applying the Green’s function method. Integral representations have some advantages over
infinite series representations (e.g., the description of the general analytical structure of a
solution and the evaluation of a solution).
Consider the linear Poisson equation in a volume V with surface S on which Dirichlet
boundary conditions are posed. The Green’s function G(X , X0 ) (X = (x, y, z) and X0 =
(x0 , y0 , z0 )) associated with the boundary value problem is a function of two variables X
(the position vector) and X0 (a fixed location) defined as the solution to
∇2 G(X , X0 ) = δ(X − X0 ) in V ; G(X , X0 ) = 0 on S.
If the volume V is the whole space, the Green’s function is called the fundamental solution.
Since G(X , X0 ) is symmetric, i.e., G(X , X0 ) = G(X0 , X ), this fact can serve to verify that
G(X , X0 ) is computed correctly.
Example 22.30. Linear Laplace equation. Green’s function. Boundary value problem.
Consider the Dirichlet boundary value problem for the Laplace equation on the semiinfinite
plane V = {y > 0},
∇2 u(x, y) = uxx + uyy = 0 in V, u(x, y) = f (x) on S,
where V = {y > 0} and S = {y = 0}. To construct the Green’s function and find the solution of the
boundary value problem, we apply the method of images; i.e., we seek a Green’s function G(X, X0 )
such that, for X, X0 ∈ V ,
G(X, X0 ) = v(X, X0 ) + w(X, X0 ), where ∇2 v(X, X0 ) = −δ(X − X0 ) and ∇2 w(X, X0 ) = 0.

6 The auxiliary function known today as Green’s function was first introduced by George Green in 1828 [see
Green (1828)].
Here X = (x, y), X0 = (x0 , y0 ), the function v(X, X0 ) is the free space Green’s function (does not
depend on the boundary conditions), and the function w(X, X0 ) satisfies the Laplace equation and the
boundary conditions (and is regular at X = X0 ); i.e., ∇2 w(X, X0 ) = 0 in V and w(X, X0 ) = −v(X, X0 )
(i.e., G(X, X0 ) = 0) on S for the Dirichlet boundary conditions.
It is well known that the 2D free space function v(X, X0 ) is
1
v(X, X0 ) = − ln (x − x0 )2 + (y − y0)2 .
4π
If to v(X, X0 ) we add the function
1
w(X, X0 ) = ln (x − x0 )2 + (y + y0)2 ,
4π
which satisfies the Laplace equation ∇2 w(X, X0 ) = 0 in V and is regular at x = x0 and y = y0 , then
we obtain the Green’s function (G11)

1 (x − x0 )2 + (y − y0)2
G(X, X0 ) = v(X, X0 ) + w(X, X0 ) = − ln .
4π (x − x0 )2 + (y + y0)2
Then, setting y = 0, we obtain G(X, X0 ) = 0 (Test1). The solution of the boundary value problem is
Z
∂G
u(x0 , y0 ) = − f (x) dS.
S ∂n
∂G
By computing for the boundary y = 0, we obtain the derivative (DGn)
∂n

∂G ∂G y0
=− =−
∂n S ∂y y=0 π[(x − x0)2 + y20 ]
and the solution (SolF) of the boundary value problem
Z ∞
y f (s)
u(x, y) = .
π −∞ (x − s)2 + y2
Finally, we verify that the Green’s function obtained is symmetric and visualize it (see Fig. 22.3) as
follows:
with(plots): with(Student[Precalculus]): interface(showassumed=0):
assume(x>0,y>0,x0>0,y0>0); Rx:=-10..10; Ry:=0..10; Ops1:=grid=[100,100];
v:=-1/(4*Pi)*ln((x-x0)ˆ2+(y-y0)ˆ2); w:=1/(4*Pi)*ln((x-x0)ˆ2+(y+y0)ˆ2);
G1:=CompleteSquare(combine(v+w));
G11:=op(1,G1)*op(3,G1)*(-1)*combine(op(2,G1)*(-1));
Test1:=eval(G11,y=0); DGn:=eval(-diff(G11,y),y=0);
Sol1:=u(x0,y0)=-int(DGn*f(x),x=-infinity..infinity);
SolF:=CompleteSquare(expand(subs(x=s,x0=x,y0=y,Sol1)));
G11; G2:=unapply(G11,x,y,x0,y0); G2(x,y,x0,y0); G2(x0,y0,x,y);
assume(x>x0,y>y0); is(G2(x,y,x0,y0)=G2(x0,y0,x,y));
assume(x0>x,y0>y); is(G2(x,y,x0,y0)=G2(x0,y0,x,y));
plot3d(G2(x,y,1,1),x=Rx,y=Ry,axes=boxed,orientation=[-40,55],color=grey,Ops1);
contourplot(G2(x,y,1,1),x=Rx,y=Ry,grid=[150,150],axes=boxed,contours=20);
Green’s functions can also be constructed by applying Laplace transforms [see Cole
et al. (2011)], the eigenfunction expansion method [see Debnath (2007), Polyanin and
Manzhirov (2007), Polyanin (2002)], and conformal mappings of the complex plane (for
solving 2D problems).
Figure 22.3 Green’s function G(X, X0 ) for the semiinfinite plane y > 0 and x0 = 1, y0 = 1.
Now consider an extension of the theory of Green’s functions, namely, the construction
of modified Green’s functions and solutions of initial-boundary value problems.
Example 22.31. Linear Klein–Gordon equation. Modified Green’s function. Initial-boundary
value problem.
Consider the initial-boundary value problem for the Klein–Gordon equation with the following
initial conditions and Neumann boundary conditions:
utt = a2 uxx − bu, x1 ≤ x ≤ x2 ,

u(x, 0) = f1 (x), ut (x, 0) = f2 (x), ux (0,t) = g1 (t), ux (L,t) = g2 (t),
where a and b are real parameters (a > 0, b > 0), x1 = 0, x2 = L, f1 (x)=1, f2 (x)=0, g1 (t)=1,
g2 (t)=0. The linear Klein–Gordon equation is a special case of the equation
s(x)utt = (p(x)ux )x − q(x)u + φ(x,t),
where s(x) = 1, p(x) = a2 , q(x) = b, and φ(x,t) = 0. By applying the eigenfunction expansion
method, we will construct the modified Green’s function. The corresponding Sturm–Liouville prob-
lem has the form
a2 φ′′xx + (λ − b)φ = 0, φ′x = 0 at x = 0, φ′x = 0 at x = L.
By solving this eigenvalue problem, we find the eigenvalues and the corresponding eigenfunctions
(EVal, EFun),
πnx πna 2
φn+1 (x) = cos , λn+1 = b + , n = 0, 1, . . . .
L L
According to Polyanin (2002),
∞
√
φn (x)φn (ξ) sin(t λn )
G(x, ξ,t) = ∑ √ ,
n=1 kφn k2 λn
Z x2
where kφn k2 = s(x)φ2n (x) dx. Then the modified Green’s function (G1) becomes
x1
√ ∞
p
sin(t b) 2 cos(πnx/L) cos(πnξ/L) sin(t b + (πna/L)2 )
G(x, ξ,t) = √ +∑ p ,
L b n=1 L b + (πna/L)2
22.4. Numerical Solutions and Their Visualizations 1309
which is a spectral representation of the Green’s function for this problem. To visualize the Green’s
function, we use finitely many terms of the series (GAppr).
Finally, by constructing the solution of the IBVP according to Polyanin (2002),
Z t Z x2 Z x2
∂
u(x,t) = Φ(ξ, τ)G(x, ξ,t −τ) dξ dτ+ s(ξ) f1 (ξ)G(x, ξ,t) dξ
0 x1 ∂t x1
Z x2 Z t Z t
+ s(ξ) f2 (ξ)G(x, ξ,t) dξ+ p(x1 ) g1 (τ)(−G(x, x1 ,t −τ)) dτ+ p(x2 ) g2 (τ)G(x, x2 ,t −τ) dτ,
x1 0 0
we obtain the solution to the given problem and visualize it using finitely many terms of the series
(SolF) for various values of t (see Fig. 22.4) as follows:
with(plots): with(LinearAlgebra): nInf:=n=1..infinity; s:=x->1; p:=x->aˆ2; q:=x->b;
Phi:=(x,t)->0; L1:=1; a1:=1; b1:=1; Rx:=0..L1; Rxi:=0..L1; t1:=10; x1:=0; x2:=L1;
TrInf:=infinity=29; f1:=x->1; f2:=x->0; g1:=t->1; g2:=t->0; Ops1:=grid=[150,150];
tauT:=tau=0..t; assume(L>0,a>0); assume(n::integer,n>0); interface(showassumed=0):
xiX:=xi=x1..x2; Sol1:=dsolve(aˆ2*diff(phi(x),x$2)+(lambda-b)*phi(x)=0, phi(x));
phi:=unapply(rhs(Sol1),x); phi(x);
BC1:=[eval(diff(phi(x),x),x=0)=0, eval(diff(phi(x),x),x=L)=0];
MCoef:=GenerateMatrix(BC1,[_C1,_C2])[1]; CharEq:=Determinant(MCoef)=0;
Sols:=[solve(CharEq,AllSolutions = true,lambda)];
Trlambda:=lambda=subs(_Z1=n,expand(Sols[2]));
MN:=map(xi->simplify(subs(Trlambda,xi)),MCoef); NS:=NullSpace(MN);
Funs:=simplify(subs(_C1=NS[1][1],_C2=NS[1][2],Trlambda,phi(x)));
EVal:=unapply(rhs(Trlambda),n); EFun:=unapply(Funs,n,x); EVal(n); EFun(n,x);
NEFun:=int(s(x)*EFun(n,x)ˆ2,x=0..L); NEFun0:=int(s(x)*EFun(0,x)ˆ2,x=0..L);
GTerm0:=EFun(0,x)*EFun(0,xi)*sin(t*sqrt(EVal(0)))/sqrt(EVal(0))/NEFun0;
G1:=GTerm0+sum(EFun(n,x)*EFun(n,xi)*sin(t*sqrt(EVal(n)))/sqrt(EVal(n))/NEFun,nInf);
GAppr:=unapply(subs(TrInf,G1),x,xi,t,L,a,b); GAppr(x,xi,t,L1,a1,b1):
plot3d(GAppr(x,xi,t1,L1,a1,b1),x=Rx,xi=Rxi,axes=frame, orientation=[20,45],
color=grey,style=patchcontour,grid=[100,100],resolution=600,numpoints=1000);
contourplot(GAppr(x,xi,t1,L1,a1,b1),x=Rx,xi=Rxi,axes=boxed,contours=20,Ops1);
G2:=unapply(G1,x,xi,t); G2(x,xi,t);
u:=Int(Int(Phi(xi,tau)*G2(x,xi,t-tau),xiX),tauT)
+diff(Int(s(xi)*f1(xi)*G2(x,xi,t),xiX),t)+Int(s(xi)*f2(xi)*G2(x,xi,t),xi=x1..x2)
+p(x1)*Int(g1(tau)*(-G2(x,x1,t-tau)),tauT)+p(x2)*Int(g2(tau)*G2(x,x2,t-tau),tauT);
u1:=subs(L=L1,a=a1,b=b1,infinity=3,u):
SolF:=unapply(collect(map(value,u1),[cos]),x,t);
plot3d(SolF(x,t),x=Rx,t=0..t1,axes=boxed, color=grey,style=patchcontour);
contourplot(SolF(x,t),x=Rx,t=0..t1,grid=[150,150],axes=boxed,contours=20);
for i from 0 to t1 do G||i:=plot(SolF(x,i),x=Rx,color=black); od:
display(seq(G||i,i=0..t1));
plot([SolF(x,0),SolF(x,1),SolF(x,5),SolF(x,t1)],x=Rx,color=black,
linestyle=[solid,dot,dash,dashdot],legend=[typeset("u(x,0)"),typeset("u(x,1)"),
typeset("u(x,5)"),typeset("u(x,10)")]);
⊙ References for Section 22.3: G. Green (1828), V. A. Galaktionov (1990, 1995), A. D. Polyanin and
A. I. Zhurov (1998), A. D. Polyanin (2002), A. D. Polyanin and A. V. Manzhirov (2007), L. Debnath (2007),
K. D. Cole, J. V. Beck, A. Haji-Sheikh, and B. Litkouhi (2011).
22.4 Numerical Solutions and Their Visualizations

It is well known that there are numerous linear partial differential equations (arising when
modeling real-world problems) and associated mathematical problems (even with simple
Figure 22.4 The solution u(x,t) of the IBVP for the linear Klein–Gordon equation for various values
of t, t ∈ [0, 10].
boundary and/or initial conditions) that cannot be solved analytically. Therefore, one has
to develop and apply approximation methods. Nowadays, approximation methods are be-
coming very important and useful in applications due to the increasing development of
modern computers, supercomputers, and computer algebra systems. In this section, some
of the most important approximation approaches to the solution of linear partial differential
equations and associated mathematical problems are discussed with the aid of the computer
algebra system Maple.
22.4.1 Constructing Numerical Solutions in Terms of Predefined Functions

First, consider the predefined Maple functions with the aid of which we can obtain approx-
imate numerical solutions when solving various linear time-based PDE problems. With the
aid of the predefined function pdsolve (the option numeric), we can numerically solve
initial-boundary value problems.
The predefined function pdsolve allows us to solve a single (higher-order) PDE and PDE
systems by the default methods or specify a particular method for solving a single PDE.
It is possible to pose Dirichlet, Neumann, Robin, or periodic boundary conditions.
infolevel[all]:=5; Sol:=pdsolve(PDEs,IBCs,numeric,funcs,ops);
Num_vals:=Sol:-value(); Sol:-plot3d(func,t=t0..t1,ops);
Num_vals(num1,num2); Sol:-animate(func,t=t0..t1,x=x0..x1,ops);
pdsolve(PDE,IBCs,numeric,numericalbcs=val,method=M1,startup=M2,ops);
pdsolve(PDE,ICs,series,order=num,ops);
with(PDEtools): PDEplot(PDE,ICs,ranges,ops);
pdsolve,numeric finding numerical solutions of a partial differential equation PDE or a

system of PDEs
plot3d,animate visualizing the numerical solution Sol obtained by pdsolve,numeric
value displaying numerical values of a numerical solution Sol
pdsolve,series finding formal power series solutions for systems of differential equa-
tions
PDEplot plotting the solution of a first-order linear (or nonlinear) partial differential equa-
tion
Remark. The solution obtained is represented as a module (similar to a procedure, with

the operator :-), which can be used for obtaining visualizations (plot, plot3d, animate,
animate3d) and numerical values (value); for more detail, see ?pdsolve[numeric].
Numerical and graphical solutions by default methods. Numerical solutions can be
obtained in Maple automatically (without specifying a numerical method) by the default
θ-methods [see Larsson and Thomée (2008), Morton and Mayers (1995)]. The θ-method
is a generalization of the known finite difference approximations (explicit and implicit)
by introducing a parameter θ (0 ≤ θ ≤ 1) and by taking a weighted average of the two
formulas, where the special case θ = 12 corresponds to the Crank–Nicolson method [see
Crank and Nicolson (1947)] and θ = 0 and θ = 1 are just the explicit and implicit methods,
respectively.
It should be noted that in Maple (Release 18 or earlier) one can numerically solve only
evolution equations via predefined functions.
Example 22.32. Linear wave equation. Initial-boundary value problem. Numerical, graphical,
and exact solutions. We find numerical, graphical, and exact solutions of the following initial-
boundary value problem for the linear wave equation:
utt − c2 uxx = 0, 0 < x < L, t > 0,

u(x, 0) = f1 (x), ut (x, 0) = f2 (x), u(0,t) = g1 (t), u(L,t) = g2 (t),
where c = 1/10, L = 1, f1 (x) = A sin(Bπx), f2 (x) = 0, g1 (t) = 0, g2 (t) = 0, A = 3, and B = 2. First,

we solve the this initial-boundary value problem numerically and visualize it as follows:
with(plots): with(PDEtools): declare(u(x,t)); A:=3: B:=2: c:=1/10: L:=1:

tR:=0..1; xR:=0..L; NF:=30; NP:=50; N:=3;
L1:=[red,blue,green]; L2:=[1/8,3/8,5/8]; Ops:=spacestep=1/256,timestep=1/60;
f1:=x->A*sin(B*Pi*x); f2:=x->0; g1:=t->0; g2:=t->0;
PDE1:=diff(u(x,t),t$2)-cˆ2*diff(u(x,t),x$2)=0;
IC:={u(x,0)=f1(x),D[2](u)(x,0)=f2(x)}; BC:={u(0,t)=g1(t),u(L,t)=g2(t)};
Sol1:=pdsolve(PDE1,IC union BC,numeric,u(x,t),time=t,range=0..L,Ops);
for i from 1 to N do G||i:=Sol1:-plot(t=L2[i],color=L1[i],numpoints=NP): od:
display({seq(G||i,i=1..N)}); Num_vals1:=Sol1:-value(); Num_vals1(1/2,1/8);
Sol1:-plot3d(u(x,t),t=tR,shading=zhue,axes=boxed);
Sol1:-animate(u(x,t),x=xR,t=tR,frames=NF,numpoints=NP,thickness=3);
Then we construct the exact solution u(x,t) = A sin(Bπ x) cos(cπt) (SolF) of this initial-boundary
value problem and verify that this solution is an exact solution of the problem (T1–T5) as follows:
SolEx:=pdsolve(PDE1,HINT=`*`,build);
Eq1:=expand(subs(t=0,rhs(SolEx)))=f1(x);
Eq2:=expand(subs(t=0,diff(rhs(SolEx),t)))=f2(x);
Eq3:=expand(subs(x=0,rhs(SolEx)))=g1(t);
Figure 22.5 The numerical solution u(x,t) of the IBVP for the linear wave equation and the corre-
sponding errors (between the exact and numerical solutions) at times t = 1/8, 3/8, 5/8.
Eq4:=expand(subs(x=L,rhs(SolEx)))=g2(t);
Sols:=[solve({Eq1,Eq2,Eq3,Eq4},{_C1,_C2,_C3,_C4,_c[1]})] assuming _c[1]<0;
Sol3:=subs(Sols[2],SolEx); SolF:=simplify(convert(Sol3,trig));
T1:= pdetest(u(x,t)=rhs(SolEx),PDE1); T2:=simplify(subs(t=0,Sols[2],SolEx));
T3:=D[2](u)(x,0)=simplify(subs(t=0,Sols[2],diff(rhs(SolEx),t)));
T4:=simplify(subs(x=0,Sols[2],rhs(SolEx)));
T5:=simplify(subs(x=L,Sols[2],rhs(SolEx)));
Finally, we compare the numerical and exact solutions obtained as follows:
Trt:=t=3/8; Trx:=x=1/2; p1:=Sol1:-plot(Trt,color=blue,thickness=2):

p2:=plot(subs(Trt,rhs(SolF)),x=xR,color=green,thickness=2):
display({p1,p2});
Num_vals1(1/2,3/8); evalf(subs(Trx,Trt,rhs(SolF)),16);
p3:=Sol1:-plot(u-rhs(SolF),t=1/8,numpoints=NP,color=L1[1]):
p4:=Sol1:-plot(u-rhs(SolF),Trt, numpoints=NP,color=L1[2]):
p5:=Sol1:-plot(u-rhs(SolF),t=5/8,numpoints=NP,color=L1[3]):
display({p3,p4,p5});
The numerical solution of this initial-boundary value problem and the corresponding errors
(between the exact and numerical solutions) at times t = 1/8, 3/8, 5/8 are shown in Fig. 22.5.
22.4.2 Numerical Methods Embedded in Maple

In Maple, one can obtain numerical solutions specifying one of the eleven classical meth-
ods, specifying numerical boundary conditions and finite difference schemes for two-stage
methods.
The classical numerical methods embedded in Maple are the forward-time forward-
space (backward-space) method, centered-time forward-space (backward-space) method,
backward-time forward-space (backward-space) method, forward-time centered-space (or
Euler) method, centered-time centered-space (or Crank–Nicolson) method, backward-time
centered-space (or backward Euler) method, box method, Lax–Friedrichs method, Lax–
Wendroff method, leapfrog method, and DuFort–Frankel method.
However, there is some restriction with respect to these classical methods: a single PDE
must be parabolic or hyperbolic and of first order in time. PDEs that are greater than first
order in time can be solved by converting to an equivalent first-order system.
More detailed information about numerical methods embedded in Maple is presented
in Table 22.2.
Table 22.2.
Numerical methods embedded in Maple with a brief description and some references
Numerical method Brief description References
FTime1Space[b] Explicit 1-stage method for first-order time/space PDEs. Accuracy: O(h, k). Lapidus and Pinder (1999)
FTime1Space[f] Stability: k<ah (a depends upon the problem). [b/f]: to describe right/left TW, LeVeque (2007)
(to specify BC at the left/right boundary). Numerical BCs are not required. Morton and Mayers (1995)
CTime1Space[b] Implicit 1-stage method for PDEs: F(u, ux , ut , uxt ) = 0. Accuracy: O(h, k2 ). Lapidus and Pinder (1999)
CTime1Space[f] Stability: unconditionally stable for many problems. [b/f]: to describe right/left TW, LeVeque (2007)
BTime1Space[b] Implicit 1-stage method for PDEs: F(u, ux , ut , uxt ) = 0. Accuracy: O(h, k). Lapidus and Pinder (1999)
BTime1Space[f] Stability: unconditionally stable for many problems. [b/f]: to describe right/left TW, LeVeque (2007)
Euler Explicit 1-stage method for PDEs: 1-order in t, n-order in space (no mixed derivative). Lapidus and Pinder (1999)
or FTimeCSpace Accuracy: O(h2 , k). Stability: some restriction on h and k. Strikwerda (2004)
Numerical BCs are required depending upon the order of the PDE in space. Morton and Mayers (1995)
CrankNicholson Implicit 1-stage method for PDEs: 1-order in t, n-order in space (no mixed derivative). Lapidus and Pinder (1999)
or CTimeCSpace Accuracy: O(h2 , k2 ). Stability: unconditionally stable for many problems. Thomas (1995)
BackwardEuler Implicit 1-stage method for PDEs: 1-order in t, n-order in space (no mixed derivative). Lapidus and Pinder (1999)
or BTimeCSpace Accuracy: O(h2 , k). Stability: unconditionally stable for many problems. Strikwerda (2004)
Box[b] Implicit 1-stage method for PDEs: 1-order in t, n-order in space (no mixed derivative). Strikwerda (2004)
Box[f] Accuracy: O(h2 , k2 ). Stability: unconditionally stable for many problems. LeVeque (2007)
Numerical BCs are required depending upon the order of the PDE in space. Larsson and Thomée (2008)
Explicit 1-stage method for PDEs: 1-order in time, odd-order in space Larsson and Thomée (2008)
LaxFriedrichs (no mixed derivative). Accuracy: O(h2 , k, M). Stability: restriction of the form LeVeque (2007)
k<ahP . Numerical BCs are required depending upon the order of the PDE in space. Strikwerda (2004)
Explicit 1-stage method for linear PDEs that are first-order in time and space. Larsson and Thomée (2008)
LaxWendroff Accuracy: O(h2 , k, h, k2 ). Stability: restriction of the form k<ah. LeVeque (2007)
Numerical BC is required so that one BC is specified for each boundary. Strikwerda (2004)
Explicit 2-stage method for PDEs: 1-order in time, n-order in space Morton and Mayers (1995)
Leapfrog (no mixed derivative). Accuracy: O(h2 , k2 ). Startup method. Stability: restriction. Strikwerda (2004)
Numerical BCs are required depending upon the order of the PDE in space. Thomas (1995)
Explicit 2-stage method for linear/nonlinear PDEs: 1-order in t, even-order in space Larsson and Thomée (2008)
DuFortFrankel (no mixed derivative). Accuracy: O(h2 , k2 , Q). Startup method is required. LeVeque (2007)
No numerical BCs are required. Stability: restriction of the form k < ahP . Strikwerda (2004)
The abbreviated Maple names of the embedded numerical methods are presented in the
first column of the table. The following abbreviations in the table are adopted: F, forward;
B, backward; C, centered; b, backward; f, forward; TW, traveling wave; BC(s), boundary
condition(s); h, space step; k, time step; P, the differential order of the PDE in the spatial
variable; ceil, smallest integer greater than or equal to a number; M = h2N /k, Q = k2 /h2N ,
and N = ceil( 12 P). For example, for the forward-time backward-space method, the abbre-
viated Maple name is FTime1Space[b] and the corresponding complete Maple name is
ForwardTime1Space[backward].7
The forward-time 1-space [forward/backward] method is an explicit 1-stage method
for finding solutions to first-order time/space PDEs.
The finite difference scheme corresponding to the forward-time 1-space [backward]
method and used to compute the value, e.g., at the mesh point (i, 1), can be obtained by
applying differencing to the PDE about the point (xi ,t j ) and by substituting the following
discretizations:
Ui,1 −Ui,0 Ui,0 −Ui−1,0
ut = , ux = , u = Ui,0 ,
k h
where we introduce the notation ui, j = u(xi ,t j ) for the exact solution and Ui, j for the discrete
approximation (with uniform spacing h and k), so that Ui, j ≈ u(xi ,t j ). In particular, Ui,0 =
u(xi ,t0 ) and Ui,1 = u(xi ,t1 ).
For the forward-time 1-space [forward] method the discretizations read
Ui,1 −Ui,0 Ui+1,0 −Ui,0
ut = , ux = , u = Ui,0 .
k h
The centered-time 1-space [forward/backward] method is an implicit 1-stage method
for finding solutions to PDEs containing the derivatives ux , ut , uxt .
The finite difference scheme corresponding to the centered-time 1-space [backward]
method and used to compute the value at the mesh point (i, 1) can be obtained by applying
differencing to the PDE about the point (xi ,t j + k/2) and by substituting the following
discretizations:
Ui,1 −Ui−1,1 −Ui,0 +Ui−1,0 Ui,1 −Ui,0 Ui,1 −Ui−1,1 +Ui,0 −Ui−1,0
uxt = , ut = , ux = ,
kh k 2h
Ui,1 +Ui,0
u= , x = xi , t = t j + k/2. (22.4.2.1)
2
For the centered-time 1-space [forward] method, the discretizations read
Ui+1,1 −Ui,1 −Ui+1,0 +Ui,0 Ui,1 −Ui,0 Ui+1,1 −Ui,1 +Ui+1,0 −Ui,0
uxt = , ut = , ux = ,
kh k 2h
Ui,1 +Ui,0
u= , x = xi , t = t j + k/2. (22.4.2.2)
2
The backward-time 1-space [forward/backward] method is an implicit 1-stage method
for finding solutions of PDEs containing the derivatives ux , ut , uxt .
The finite difference scheme corresponding to the backward-time 1-space [backward]
method and used to compute the value at the mesh point (i, 1) can be obtained by apply-
ing differencing to the PDE about the point (xi ,t j + k) and by substituting the following
discretizations:
Ui,1 −Ui−1,1 −Ui,0 +Ui−1,0 Ui,1 −Ui,0
uxt = , ut = ,
kh k
Ui,1 −Ui−1,1
ux = , u = Ui,1 , x = xi , t = t j + k. (22.4.2.3)
h
7 TheMaple names CrankNicholson, LaxFriedrichs, LaxWendroff, and DuFortFrankel correspond
to numerical methods known in scientific literature: Crank–Nicolson, Lax–Friedrichs, Lax–Wendroff, and
DuFort–Frankel.
For the backward-time 1-space [forward] method, the discretizations read

Ui+1,1 −Ui,1 −Ui+1,0 +Ui,0 Ui,1 −Ui,0
uxt = , ut = ,
kh k
Ui+1,1 −Ui,1 Ui,1 +Ui,0
ux = , u= , x = xi , t = t j + k. (22.4.2.4)
h 2
The forward-time centered-space (or Euler) method is an explicit 1-stage method for
finding solutions of PDEs that are first order in time and arbitrary order in space (with
no mixed partial derivatives). The finite difference scheme for computing the value at the
mesh point (i, 1) can be obtained by applying central differencing to the PDE about the
point (xi ,t j ).
The centered-time centered-space (or Crank–Nicolson) method is an implicit 1-stage
method for finding solutions of PDEs that are first order in time and arbitrary order in
space (with no mixed partial derivatives). The finite difference scheme for computing the
value at the mesh point (i, 1) can be obtained by applying central differencing to the PDE
about the point (xi ,t j + k/2).
The backward-time centered-space method is an implicit 1-stage method for finding
solutions of PDEs that are first order in time and arbitrary order in space (with mixed partial
derivatives). The finite difference scheme for computing the value at the mesh point (i, 1)
can be obtained by applying central differencing to the PDE about the point (xi ,t j + k).
The box method is an implicit 1-stage method for finding solutions of PDEs that are
first order in time and arbitrary order in space (with mixed partial derivatives). The finite
difference scheme for computing the value at the mesh point (i, 1) depends on the extra
boundary condition (or numerical boundary condition for even-order problems): if the extra
boundary condition is given on the left boundary, then the finite difference scheme can be
obtained by applying central differencing of the PDE about the point (xi − h/2,t j + k/2);
if the extra boundary condition is given on the right boundary, then the finite difference
scheme can be obtained by applying central differencing of the PDE about the point (xi +
h/2,t j + k/2).
The Lax–Friedrichs method is an explicit 1-stage method for finding solutions of PDEs
that are first order in time and odd order in space (with no mixed partial derivatives). The
finite difference scheme for computing the value at the mesh point (i, 1) is similar to that of
the Euler scheme, where the value Ui,0 is replaced by the interpolation of the value at Ui,0
using all other points U j,0 in the Euler stencil ( j 6= i).
The Lax–Wendroff method is an explicit 1-stage method for finding solutions of linear
PDEs that are first order in time and space. The finite difference scheme of the method can
be derived in a variety of ways, for example, by using the idea of multistep methods: at the
first step, by considering expressions for half the time step and by applying central differ-
encing to approximate the derivative ux , and at the second step, by using the Lax method
[for more details, e.g., see Richtmyer (1963), Strikwerda (2004), and Thomas (1995)].
The Leapfrog method is an explicit 2-stage method for finding solutions of PDEs that
are first order in time and arbitrary order in space (with no mixed partial derivatives). The
finite difference scheme for computing the value at the mesh point (i, 1) can be obtained by
applying central differencing to compute all spatial derivatives and the discretization
Ui,1 −Ui,−1
ut =
2k
for the time derivative.

The DuFort-Frankel method is an explicit 2-stage method for finding solutions of linear
PDEs that are first order in time and even order in space (with no mixed partial derivatives).
The finite difference scheme for computing the value at the mesh point (i, 1) is similar to
that of the Leapfrog scheme, where the value Ui,0 is replaced by the average of the values
Ui,1 and Ui,−1 for all occurrences in the scheme.
Example 22.33. Linear advection equation. Initial-boundary value problem. Numerical,
graphical, and exact solutions. Single-stage numerical method. Consider the following initial-
boundary value problem for the linear advection equation (or first-order wave equation, or first-order
hyperbolic equation):
ut + cux = 0; u(x, 0) = f (x), u(0,t) = g(t).
where f (x) = sin(x), g(t) = 0, and c is a real parameter (c > 0).
First, we obtain the numerical and graphical solutions of this initial-boundary value problem by
applying the single-stage explicit ForwardTime1SpaceBackward method as follows:
with(PDEtools): with(plots): declare(u(x,t)); NF:=30; NP:=100; L:=1; T:=1;
xR:=0..L; tR:=0..T; c:=0.7; St:=1/100; Sx:=1/100; N:=3;
Ops:=timestep=St,spacestep=Sx; L1:=[red,blue,magenta]; L2:=[0.1,0.15,0.2];
f:=x->sin(x); g:=t->0; M1:=ForwardTime1Space[backward];
PDE1:=diff(u(x,t),t)+c*diff(u(x,t),x)=0; IBC:={u(x,0)=f(x),u(0,t)=g(t)};
Sol1:=pdsolve(PDE1,IBC,numeric,time=t,range=xR,method=M1,Ops);
Num_vals1:=Sol1:-value(); Num_vals1(0.5,0.5);
for i from 1 to N do G||i:=Sol1:-plot(t=L2[i],color=L1[i],numpoints=NP*2): od:
display({seq(G||i,i=1..N)});
Sol1:-animate(u(x,t),x=xR,t=tR,view=[xR,tR],frames=NF,numpoints=NP,thickness=3);
Remark. We choose the backward method (ForwardTime1Space[backward]) for first-order linear

PDEs (where the boundary condition is given at the left boundary) that describe right-traveling
waves. We can find numerical and graphical solutions of this initial-boundary value problem, e.g.,
at times t = 0.1, 0.15, 0.2 (L2).
Then we construct the exact solution (SolF)
1428571429 x sinx
u(x,t) =
1000000000t + 1428571429 x
of this initial-boundary value problem and verify that this solution is an exact solution of the given
initial-boundary value problem (T1–T3) as follows:
SolEx:=pdsolve(PDE1,HINT=`+`,build);
Eq1:=expand(subs(t=0,rhs(SolEx)))=f(x); Eq2:=expand(subs(x=0,rhs(SolEx)))=g(t);
Sols:=solve({Eq1,Eq2},{_c[2],_C1,_C2}); SolF:=simplify(subs(Sols,SolEx));
T1:= pdetest(u(x,t)=rhs(SolEx),PDE1); T2:=simplify(subs(t=0,Sols,SolEx));
T3:=simplify(subs(x=0,Sols,SolEx));
Finally, by comparing the numerical and exact solutions obtained, we visualize the solutions
and the corresponding errors between them at various times (e.g., t = 0.1, 0.3, 0.5) as follows:
Trt:=t=0.1; Trx:=x=0.5; p1:=Sol1:-plot(Trt,color=blue,thickness=2):
p2:=plot(subs(Trt,rhs(SolF)),x=xR,color=green,thickness=2): display({p1,p2});
Num_vals1(0.5,0.1); evalf(subs(Trx,Trt,rhs(SolF)),16);
p3:=Sol1:-plot(u-rhs(SolF),t=0.1,numpoints=NP,color=L1[1]):
p5:=Sol1:-plot(u-rhs(SolF),t=0.5,numpoints=NP,color=L1[3]): display({p3,p4,p5});
Example 22.34. Linear heat equation. Initial-boundary value problem. Numerical, graphi-
cal, and exact solutions. Crank–Nicolson method. Consider the following initial-boundary value
problem for the linear one-dimensional heat equation:
ut = k uxx , 0 < x < L, t > 0, u(x, 0) = f (x), u(0,t) = g1 (t), u(L,t) = g2 (t),
where L = 1, k = 0.1, f (x) = x(1 − x), g1 (t) = 0, and g2 (t) = 0. By constructing the numerical and
graphical solutions of this problem, we apply one of the successful implicit finite difference schemes
based on six grid points, the CrankNicholson method [see Crank and Nicolson (1947)], as follows:
with(PDEtools): with(plots): declare(u(x,t)); NF:=30; NP:=100;
L:=1; k:=0.1; xR:=0..L; tR:=0..5; S:=1/300;
Ops:=timestep=S,spacestep=S; N:=2; L1:=[red,blue]; L2:=[0,0.5];
f:=x->x*(1-x); g1:=t->0; g2:=t->0;
PDE1:=diff(u(x,t),t)=k*diff(u(x,t),x$2); IBC:={u(x,0)=f(x),
u(0,t)=g1(t), u(L,t)=g2(t)}; M1:=CrankNicholson;
Sol1:=pdsolve(PDE1,IBC,numeric,method=M1,Ops);
Num_vals1:=Sol1:-value(); for i from 1 to N do
G||i:=Sol1:-plot(t=L2[i],color=L1[i],numpoints=NP*2): od:
display({seq(G||i,i=1..N)}); Num_vals1:=Sol1:-value();
Num_vals1(1/2,Pi);
Then we construct the exact solution (SolF)

e2x+0.1t − e0.1t
u(x,t) = x(1 − x)
e2x − 1
of this initial-boundary value problem and verify that this solution is an exact solution of the given
initial-boundary value problem (T1–T4) as follows:
SolEx:=pdsolve(PDE1,HINT=`*`,build); Eq1:=expand(subs(t=0,rhs(SolEx)))=f(x);
Eq2:=expand(subs(x=0,rhs(SolEx))=g1(t)); Eq3:=expand(subs(x=L,rhs(SolEx))=g2(t));
Sols:=solve({Eq1,Eq2},{_C1,_C2,_C3,_c[1]});
SolF:=simplify(subs(Sols,_c[1]=1,SolEx));
T1:= pdetest(u(x,t)=rhs(SolEx),PDE1);
T2:=expand(simplify(subs(Sols,t=0,_c[1]=1,SolEx)));
T3:=simplify(subs(Sols,x=0,_c[1]=1,rhs(SolEx)));
T4:=simplify(subs(Sols,x=L,_c[1]=1,rhs(SolEx)));
Finally, by comparing the numerical and exact solutions obtained, we visualize the solutions
and the corresponding errors between them at distinct times (e.g., t = 0.2, 0.4) and examine the
errors by constructing the error function as follows:
Trt:=t=0.1; Trx:=x=0.5; p1:=Sol1:-plot(Trt,color=blue,thickness=2):
p2:=plot(subs(Trt,rhs(SolF)),x=xR,color=green,thickness=2): display({p1,p2});
Num_vals1(0.5,0.1); evalf(subs(Trx,Trt,rhs(SolF)),16);
p4:=Sol1:-plot(u-rhs(SolF),t=0.4,numpoints=NP,color=L1[2]): display({p3,p4});
errfunc:=rhs(Sol1:-value(u(x, t)-rhs(SolF),t=0.1,output=listprocedure)[3]);
for i from 0.01 to 1 by 0.1 do abs(errfunc(i)); od;
Example 22.35. Linear advection equation. Initial-boundary value problem. Numerical and
graphical solutions. Du Fort–Frankel method. Startup method. Numerical boundary condition.
Consider the initial-boundary value problem for the linear advection equation
ut + c ux = 0, −L < x < L, t > 0, u(x, 0) = f (x), ux (−L,t) = 0,
where 

0, x < −1,
L = 10, c = 0.5, f (x) = x + 1, −1 ≤ x ≤ 0,


1, x > 0.
For constructing the numerical and graphical solutions of this problem, we apply the two-stage
explicit Du Fort–Frankel method. Therefore, we have to indicate how to compute the additional
stage required for two-stage methods, i.e., the startup option. Since the advection equation is
of odd order in space, we have to indicate a numerical boundary condition (NBC), which we can
choose to be an NBC that forces the value of the solution on the right boundary to be the same as
the value of the solution at the first interior point; i.e., u[1,n]-u[1,n-1]. Then we find the numerical
and graphical solutions of this initial-boundary value problem at times t = 0, 0.5, 0.9 as follows:
with(PDEtools): with(plots): declare(u(x,t)); NF:=30; NP:=100; c:=0.5;

xR:=-10..10; tR:=0..9; S:=1/300; Ops:=timestep=S,spacestep=S; N:=3;
L1:=[red,blue,magenta]; L2:=[0,0.5,0.9];
f:=x->piecewise(x<-1,0,x>=-1 and x<=0,x+1,1); M1:=DuFortFrankel;
PDE1:=diff(u(x,t),t)+c*diff(u(x,t),x)=0; IBC:={u(x,0)=f(x),(D[1](u))(-L,t)=0};
Sol1:=pdsolve(PDE1,IBC,type=numeric,time=t,range=xR,numericalbcs=u[1,n]-u[1,n-1],
method=M1,startup=Euler,Ops);
for i from 1 to N do
G||i:=Sol1:-plot(t=L2[i],color=L1[i],numpoints=NP,thickness=2):
od:
display({seq(G||i,i=1..N)});
22.4.3 Numerical Solutions of Initial-Boundary Value Problems

Now we show a helpful role of computer algebra systems for generating and applying var-
ious finite difference approximations for constructing numerical solutions of linear PDEs.
To approximate linear PDEs by finite differences, we have to generate a mesh (or grid)
in a domain D; e.g., D = {a < x < b, c < t < d}. The mesh can be of various types, e.g.,
rectangular, along the characteristics, polar, etc. We assume (for simplicity) that the sets of
lines of the mesh are equally spaced and the dependent variable in a given PDE is u(x,t).
We write h and k for the line spacings and define the mesh points as follows: Xi = a+ ih,
T j = c + jk (i = 0, . . . , NX , j = 0, . . . , NT ), and h = (b − a)/NX , k = (d − c)/NT . We
calculate approximations to the solution at these mesh points; these approximate values
will be denoted by Ui, j ≈ u(Xi , T j ). We approximate the derivatives in a given equation by
finite differences (of various types) and then solve the resulting difference equations.
Example 22.36. Linear heat equation. Initial-boundary value problem. Forward/backward
finite difference (FD) methods. Crank–Nicolson method. Consider the following initial-boundary
value problem for the linear heat equation:
ut = ν uxx , 0 < x < L, t > 0, u(x, 0) = f (x), u(0,t) = 0, u(L,t) = 0,
where f (x)= sin(πx/L), L = 1, and ν = 1. By applying the forward/backward difference methods
and the Crank–Nicolson method, we construct the approximate numerical solution of the given
initial-boundary value problem.
We generate the rectangular mesh X = ih, T = jk (i = 0, . . ., NX, j = 0, . . ., NT , h = L/NX,
k = T /NT ). We denote the approximate solution of u(x,t) at the mesh point (i, j) by Ui, j .
In the forward difference method, the second derivative uxx is replaced by a central difference
approximation (CDA) and the first derivative ut by a forward difference approximation (FWDA) as
follows:
Ui−1, j − 2Ui, j + Ui+1, j Ui, j+1 − Ui, j
uxx (xi ,t j ) ≈ , ut (xi ,t j ) ≈ .
h2 k
The final FD scheme for the linear heat equation is
Ui, j+1 = (1 − 2r)Ui, j + r(Ui+1, j + Ui−1, j ),
where r=νk/h2 . In this explicit FD scheme, the unknown value Ui, j+1 (at the ( j+1)st step) is
determined from the three known values Ui−1, j , Ui, j , and Ui+1, j (at the jth step). This FD scheme is
unstable for r>0.5. We find the numerical solution as follows:
with(plots): nu:=1: NX:=15: NT:=100: L:=1.: T:=0.2; h:=L/NX; k:=T/NT; r:=nu*k/hˆ2;

f:=x->evalf(sin(Pi*x)); for i from 0 to NX do X[i]:=i*h od:
IC:={seq(U(i,0)=f(X[i]),i=0..NX)};
BC:={seq(U(0,j)=0,j=0..NT), seq(U(NX,j)=0,j=0..NT)}: IBC:=IC union BC:
FD:=(i,j)->(1-2*r)*U(i,j)+r*(U(i+1,j)+U(i-1,j));
for j from 0 to NT do for i from 1 to NX-1 do U(i,j+1):=subs(IBC,FD(i,j)); od: od:
G:=j->plot([seq([X[i],subs(IBC,U(i,j))],i=0..NX)],color=blue):
Ops1:=thickness=3,labels=["X","U"];
display([seq(G(j),j=0..NT)],insequence=true,Ops1);
This FD scheme can be represented in the matrix form
Ui =MUi−1 ,
where U0 =( f (X1 ), . . . , f (XNX−1 )), and M is the NX×NX tridiagonal band matrix (with 1−2r along
the main diagonal, r along the first subdiagonals, and zeros elsewhere). We obtain the numerical
solution using this matrix representation of the FD scheme as follows:
with(plots): with(LinearAlgebra): nu:=1: NX:=40: NT:=800: L:=1.: T:=0.2;

h:=L/NX; k:=T/NT; r:=nu*k/hˆ2; NG:=90; interface(rtablesize=NX):
M:=BandMatrix([r,1-2*r,r],1,NX-1); f:=x->evalf(sin(Pi*x));
U0:=Vector([[seq(f(i*h),i=1..NX-1)]]);
for k from 1 to NG do
U||k:=M.U||(k-1); G||k:=plot([[0,0],seq([i/NX,U||k[i]],i=1..NX-1),[L,0]]);
od:
display([seq(G||i,i=1..NG)],insequence=true,thickness=3);
In the backward difference method, the second derivative uxx is replaced by a central difference
approximation (CDA) and the first derivative ut by a backward difference approximation (BWDA)
as follows:
Ui−1, j − 2Ui, j + Ui+1, j Ui, j − Ui, j−1
uxx (xi ,t j ) ≈ , ut (xi ,t j ) ≈ .
h2 k
The final FD scheme for the linear diffusion equation is
(1 + 2r)Ui, j − r(Ui+1, j + Ui−1, j ) − Ui, j−1 = 0,
where r=νk/h2 . In this implicit FD scheme, we have to solve these difference equations numeri-
cally at each of the interior mesh points at each jth step (where j=1, . . . , NT ) with the initial and
boundary conditions. This FD scheme is unconditionally stable. We calculate the approximate nu-
merical solution of the initial-boundary value problem by applying the backward difference method
as follows:
with(plots): nu:=1: NX:=50: NT:=50: L:=1.: T:=0.2; h:=L/NX; k:=T/NT; r:=nu*k/hˆ2;

for i from 0 to NX do X[i]:=i*h; od: f:=i->evalf(sin(Pi*X[i])):
IBC:={seq(U[i,0]=f(i),i=0..NX), seq(U[0,j]=0,j=0..NT),seq(U[NX,j]=0,j=0..NT)}:
Sol0:=IBC; FD:=(i,j)->(1+2*r)*U[i,j]-r*(U[i+1,j]+U[i-1,j])-U[i,j-1];
for j from 1 to NT do
Eqs||j:={seq(FD(i,j)=0,i=1..NX-1)}; Eqs1||j:=subs(Sol||(j-1),IBC,Eqs||j);
vars||j:={seq(U[i,j],i=1..NX-1)}; Sol||j:=fsolve(Eqs1||j,vars||j);
od:
G:=j->plot([seq([X[i],subs(Sol||j,IBC,U[i,j])],i=0..NX)],color=blue,thickness=3):
display([seq(G(j),j=0..NT)],insequence=true,labels=["X","U"]);
The Crank–Nicolson method is obtained by centered differencing in time about the point
(xi ,t j+1/2 ). To do so, we average the central difference approximations in space at time t j and
t j+1 as follows:

1 Ui−1, j+1 −2Ui, j+1 + Ui+1, j+1 Ui−1, j −2Ui, j +Ui+1, j
uxx (xi ,t j+1/2 ) ≈ + ,
2 h2 h2
Ui, j+1 − Ui, j
ut (xi ,t j+1/2 ) ≈ .
k
−rUi−1, j+1 + 2(1+r)Ui, j+1 − rUi+1, j+1 = rUi−1, j + 2(1 − r)Ui, j + rUi+1, j ,
where r=νk/h2 . In this FD scheme we have three unknown values of U at the ( j+1)st time step
and three known values at the jth time step. This FD scheme is unconditionally stable. We calculate
the approximate numerical solution of the initial-boundary value problem by applying the Crank–
Nicolson method [Crank and Nicolson (1947)] as follows:
with(PDEtools): with(plots): declare(v(x1,t1)); L:=1; T:=0.2: nu:=1:

NX:=20: NT:=20: NX1:=NX-1: NX2:=NX-2: h:=L/NX; k:=T/NT; r:=nu*k/(hˆ2);
SX:=h; ST:=k; tR:=0..T; xR:=0..L; NF:=30: NP:=100: tk:=0.2:
Ops1:=spacestep=SX,timestep=ST: F:=i->sin(Pi*i);
IC:={v(x1,0)=F(x1)}; BC:={v(0,t1)=0,v(L,t1)=0}; U[NX-1]:=0:
PDE1:=diff(v(x1,t1),t1)-nu*diff(v(x1,t1),x1$2)=0;
for i from 1 to NX1 do U[i-1]:=evalf(F(i*h)); od:
LM[0]:=1+r: UM[0]:=-r/(2*LM[0]):
for i from 2 to NX2 do LM[i-1]:=1+r+r*UM[i-2]/2; UM[i-1]:=-r/(2*LM[i-1]); od:
LM[NX1-1]:=1+r+0.5*r*UM[NX2-1]:
for j from 1 to NT do
t:=j*k; Z[0]:=((1-r)*U[0]+r*U[1]/2)/LM[0];
for i from 2 to NX1 do
Z[i-1]:=((1-r)*U[i-1]+0.5*r*(U[i]+U[i-2]+Z[i-2]))/LM[i-1];
od:
U[NX1-1]:=Z[NX1-1]:
for i1 to NX2 do i:=NX2-i1+1; U[i-1]:=Z[i-1]-UM[i-1]*U[i]; od:
od:
Finally, we compare the approximate numerical solution with the exact solution of this problem,
u(x,t) = exp(−π2t) sin(πx) at (xk ,tk ), 0 < x < L, t > 0.
Additionally, we obtain the numerical solution using the Maple predefined function pdsolve (with
the option method=CrankNicholson) and compare the resulting two numerical solutions with the
exact solution as follows:
M1:=CrankNicholson; ExSol:=(x,t)->exp(-Piˆ2*t)*sin(Pi*x):
printf(`Crank-Nicolson Method\n`); for i from 1 to NX1 do
X:=i*h;
printf(`%3d %11.8f %13.8f %13.8f,%13.8f\n`,i,X,U[i-1],evalf(ExSol(X,tk)),
U[i-1]-evalf(ExSol(X,tk)));
od:
NSol1:=pdsolve(PDE1,IC union BC,numeric,v(x1,t1),Ops1,time=t1,range=0..L,method=M1):
printf(`Crank-Nicolson Method\n`); vtk:=NSol1:-value(t1=tk,output=listprocedure):
vVal:=rhs(op(3,vtk)):
X1:=i*h:
printf(`%3d %11.8f %13.8f %13.8f,%13.8f\n`,i,evalf(X1),vVal(X1),
evalf(ExSol(X1,tk)),abs(vVal(X1)-evalf(ExSol(X1,tk))));
od:
Note that to ensure the coincidence of the numerical solution (using pdsolve, numeric, method)
with our solution (with FD scheme), one has to establish the coincidence between the parameters of
the two solutions: SX:=h, ST:=k, spacestep=SX, timestep=ST.
Example 22.37. Linear wave equation. Explicit difference methods. Consider the initial-
boundary value problem for the linear wave equation describing the motion of a fixed string:
utt = c2 uxx , 0 < x < L, t > 0, u(x, 0) = f1 (x), ut (x, 0) = f2 (x), u(0,t) = 0, u(L,t) = 0,
where f1 (x) = 0, f2 (x) = sin(4πx), L = 0.5, c = 1/(4π). By applying the explicit central finite
difference method, we construct the approximate numerical solution of the initial-boundary value
problem.
In the explicit central difference method, each second derivative is replaced by a central differ-
ence approximation as follows:
Ui−1, j − 2Ui, j + Ui+1, j Ui, j−1 − 2Ui, j + Ui, j+1
uxx (xi ,t j ) ≈ , utt (xi ,t j ) ≈ .
h2 k2
The FD scheme for the linear wave equation is
Ui, j+1 = 2(1 − r)Ui, j + r(Ui+1, j + Ui−1, j ) − Ui, j−1,
where r=(ck/h)2 . In this FD scheme, we have one unknown value Ui, j+1 that depends explicitly
on the four known values Ui, j , Ui+1, j , Ui−1, j , Ui, j−1 at the previous time steps ( j and j−1). To
start the process, we have to know the values of U at the time steps j=0 and j=1. Thus, we can
define the initial conditions at these time steps: Ui,0 = f 1(Xi ) and U(Xi , 0)t ≈(Ui,1 −Ui,0 )/k= f 2(Xi ),
Ui,1 = f 1(Xi )+k f 2(Xi ). This FD scheme is stable for r ≤ 1. We find the approximate numerical so-
lution of the initial-boundary value problem by applying the explicit central finite difference method
as follows:
c:=1/(4*Pi); L:=0.5; T:=1.5; NX:=40; NT:=40; NX1:=NX+1; NX2:=NX-1; NT1:=NT+1;

NT2:=NT-1; h:=L/NX; k:=T/NT; r:=evalf(c*k/h); F1:=i->0; F2:=i->sin(4*Pi*i);
for j from 2 to NT1 do U[0,j-1]:=0; U[NX1-1,j-1]:=0; od:
U[0,0]:=evalf(F1(0)); U[NX1-1,0]:=evalf(F1(L));
for i from 2 to NX do
U[i-1,0]:=F1(h*(i-1));
U[i-1,1]:=(1-rˆ2)*F1(h*(i-1))+rˆ2*(F1(i*h)+F1(h*(i-2)))/2+k*F2(h*(i-1));
od:
for j from 2 to NT do for i from 2 to NX do
U[i-1,j]:=evalf(2*(1-rˆ2)*U[i-1,j-1]+rˆ2*(U[i,j-1]+U[i-2,j-1])-U[i-1,j-2]);
od; od;
printf(` i X(i) U(X(i),NT)\n`);
X[i-1]:=(i-1)*h: printf(`%3d %11.8f %13.8f\n`,i,X[i-1],U[i-1,NT1-1]);
od:
Points:=[seq([X[i-1],U[i-1,NT1-1]],i=1..NX1)];
plot(Points,style=point,color=blue,symbol=circle);
We construct and visualize the same approximate numerical solution of the initial-boundary
value problem by applying the explicit central finite difference method and by following a different
style of programming as follows:
with(plots): c:=evalf(1/(4*Pi)); L:=0.5: T:=1.5: NX:=40: NT:=40:

h:=L/NX; k:=T/NT; r:=(c*k/h)ˆ2; f1:=x->0: f2:=x->evalf(sin(4*Pi*x)):
IC:={seq(U1(i,0)=f1(i*h),i=1..NX-1),
seq(U1(i,1)=f1(i*h)+k*f2(i*h),i=1..NX-1)}:
BC:={seq(U1(0,j)=0,j=0..NT),seq(U1(NX,j)=0,j=0..NT)}: IBC:=IC union BC:
FD:=(i,j)->2*(1-r)*U1(i,j)+r*(U1(i+1,j)+U1(i-1,j))-U1(i,j-1);
for j from 1 to NT-1 do
for i from 1 to NX-1 do
U1(i,j+1):=subs(IBC,FD(i,j));
od:
od:
G:=j->plot([seq([i*h,subs(IBC,U1(i,j))],i=0..NX)],color=blue):
display([seq(G(j),j=0..NT)],insequence=true,thickness=3,labels=["X","U"]);
22.4.4 Numerical Solutions of Boundary Value Problems

Since Maple (Release 18 or earlier) is only capable (via predefined functions) of numeri-
cally solving evolution equations, we apply finite difference methods for constructing nu-
merical and graphical solutions of boundary value problems for elliptic equations, e.g., the
linear Poisson equation. It should be noted that MATLAB can numerically solve various
boundary value problems for linear and nonlinear elliptic equations by applying predefined
functions and embedded methods, e.g., solve scalar linear and nonlinear elliptic PDEs and
systems of PDEs in two space dimensions as well as linear and nonlinear problems defined
on a more complicated geometry.
Example 22.38. Linear Poisson equation. Boundary value problem. Central difference scheme.
Consider the two-dimensional linear Poisson equation
uxx + uyy = f (x, y), D = {a ≤ x ≤ b, c ≤ y ≤ d},
object. Let us choose f (x, y) = sin x cos y, f1 (x) = f2 (x) = − 12 sin x, f3 (y) = f4 (y) = 0, a = 0, b = π,
c = 0, d = 2π. For the linear Poisson equation, by applying the explicit finite difference scheme,
we obtain the approximate numerical solution of the boundary value problem, visualize it in D, and
compare with the exact solution (which can be obtained via predefined functions) as follows:
By applying Maple’s predefined functions, we find the exact solution

u(x, y) = − 12 sin x cos y
of the boundary value problem for the linear Poisson equation (Sol32) and visualize it in D as
follows:
with(linalg): with(PDEtools): f:=(x,y)->sin(x)*cos(y);
PDE1:=laplacian(u(x,y),[x,y])-f(x,y)=0;
Sol1:=pdsolve(PDE1,build); Test1:=pdetest(Sol1,PDE1);
Sol11:=unapply(subsop(1=0,2=0,rhs(Sol1)),x,y);
Sol12:=expand(Sol11(x,y));
plot3d(Sol12,x=0..Pi,y=0..2*Pi,shading=zhue);
Sol11(x,0); Sol11(x,2*Pi); Sol11(0,y); Sol11(Pi,y);
Let us generate a rectangular mesh: x=a+ih, y=c+ jk (where i = 0, . . ., NX, j = 0, . . ., NY ,

h = (b − a)/NX, k = (d − c)/NY ). We denote the approximate solution of u(x, y) at the mesh
point (i, j) by Ui, j . The second derivatives in the Poisson equation are replaced by a central differ-
ence approximation (CDA) as follows:
Ui−1, j − 2Ui, j + Ui+1, j Ui, j−1 − 2Ui, j + Ui, j+1
uxx (xi , y j ) ≈ , uyy (xi , y j ) ≈ .
h2 k2
The FD scheme acquires the form
2(1 + r)Ui, j − Ui+1, j − Ui−1, j − rUi, j+1 − rUi, j−1 = sin(ih) cos( jk),
where r=(h/k)2 .
Finally, we construct the approximate numerical solution of the boundary value problem by ap-
plying the above explicit finite-difference scheme and plot the numerical solution inside the domain
as follows:
with(plots): a:=0; b:=Pi; c:=0; d:=2*Pi; NX:=20; NY:=20;
h:=(b-a)/NX; k:=(d-c)/NY; r:=(h/k)ˆ2; Ops1:=orientation=[50,70];
XY:=seq(x[i]=a+i*h,i=0..NX),seq(y[j]=c+j*k,j=0..NY);
FD:=(i,j)->2*(1+r)*U[i,j]-U[i+1,j]-U[i-1,j]-r*U[i,j+1]-r*U[i,j-1]
-cos(j*k)*sin(i*h)=0;
F1:=i->-1/2*sin(i*h); F2:=i->-1/2*sin(i*h); F3:=j->0; F4:=j->0;
BC:=seq(U[i,0]=F1(i),i=0..NX),seq(U[i,NY]=F2(i),i=0..NX),
seq(U[0,j]=F3(j),j=0..NY),seq(U[NX,j]=F4(j),j=0..NY);
Eqs:={seq(seq(FD(i,j),i=1..NX-1),j=1..NY-1)}: Eqs1:=subs(BC,Eqs):
vars:={seq(seq(U[i,j],i=1..NX-1),j=1..NY-1)}:
Sol:=evalf(fsolve(Eqs1,vars));
Points:=[seq(seq([x[i],y[j],U[i,j]],i=0..NX),j=0..NY)]:
Points1:=subs({XY,BC,op(Sol)},Points):
pointplot3d(Points1,symbol=solidsphere,shading=z,axes=frame,Ops1);
22.4.5 Numerical Solutions of Cauchy Problems

There are various methods to obtain approximate solutions of Cauchy problems involving
linear partial differential equations. The most convenient and accurate method for solv-
ing Cauchy problems for hyperbolic equations is the method of characteristics (see Sec-
tions 22.2.2, 22.2.4). Finite difference methods for hyperbolic equations are not very con-
venient (e.g., in the case of discontinuous initial data) [see Richtmyer and Morton (1994),
Collatz (1966)]. Let us solve some Cauchy problems by applying various methods.
Example 22.39. Linear inhomogeneous advection equation. Cauchy problem. Exact and ap-
proximate (power series) solutions. Consider the initial value problem for the linear inhomogeneous
advection equation
ut + c ux = F(x,t), −∞ < x < ∞, t > 0, u(x, 0) = f (x),
where f (x) and F(x,t) are the
functions
of generalform or the functions specified as f (x) = sin x,
(−x) 3 (−xt) 5
F(x,t) = xt and f (x) = cos e , F(x,t) = sin e .
We obtain the exact solution of the Cauchy problem; in the Maple notation it reads (Sol1)
Z x Z x−ct
1 1 1
u(x,t) = f (x − ct)c + G a, (ct + a − x) d a − G a, (ct + a − x) , d a ,
c c c
the exact solution (Sol2) of the Cauchy problem with the simple specified functions,
1 1
u(x,t) = − ct 3 + t 2 x − sin(ct − x),
6 2
and the approximate (power series) solution (e.g., up to the third order) (Sol1Ser) of the Cauchy
problem with the complicated specified functions,
1
u(x,t) = 3
−15c2 cos(1) x2t + 5c cos(1) x3 + 6c3 cos(1) + 6c2 sin(1) x .
6c
with(PDEtools); declare(u(x,t)); PDE1:=diff(u(x,t),t)+c*diff(u(x,t),x)=F(x,t);
IC1:=u(x,0)=f(x); Sol1:=pdsolve([PDE1,IC1],u(x,t)); Test1:=pdetest(Sol1,PDE1);
PDE2:=diff(u(x,t),t)+c*diff(u(x,t),x)=x*t; IC2:=u(x,0)=sin(x);
Sol2:=pdsolve([PDE2,IC2],u(x,t)); Test2:=pdetest(Sol2,PDE2);
Sol1Ser:=pdsolve([PDE1,IC1],series,u(x,t),order=3);
F:=(x,t)->sin(exp(-x*t)ˆ5); f:=x->cos(exp(-x)ˆ3); Sol1Ser; evalf(Sol1Ser);
Sol2Ser:=pdsolve([PDE2,IC2],series,u(x,t));
Example 22.40. Linear inhomogeneous advection equation. Cauchy problem. Method of char-
acteristics. Numerical and graphical solutions. Consider the initial value problem for the linear
inhomogeneous advection equation (as in the previous example)
ut + c ux = F(x,t), −∞ < x < ∞, t > 0, u(x, 0) = f (x),
where f (x) = sin x and F(x,t) = xt.
By applying the method of characteristics, we obtain the initial value ODE problem and its
solution (ODEs, Sol1),
dx(r) c dt(r) 1 dz(r) x(r)t(r)
= , = , = , x(1) = s, t(1) = 0, z(1) = h(s),
r r r r r r
1 1
x(r) = c ln r + s, t(r) = ln r, z(r) = c ln r3 + ln r2 s + h(s).
3 2
Finding z (in terms of x and t), we have (Sol2)
1 1
s = x − ct, r = et , z = − ct 3 + t 2 x + h(x − ct),
6 2
and the solution u(x,t) = − 16 ct 3 + 12 t 2 x + h(x − ct) (see the previous example), which satisfies the
original PDE and the initial data (Test1, Test2). Finally, we visualize the solution of the given
Cauchy problem (Fig. 22.6) by using numerical ODE integration methods (with the Maple prede-
fined function PDEplot). For example, we obtain the solution of the Cauchy problem for c = 2 and
h(s) = sin s as follows:
Figure 22.6 Numerical solution of the Cauchy problem ut + c ux = xt, u(x, 0) = sin x.
with(PDEtools): TrVars:=x(r)=x,t(r)=t,z(r)=z;
PDE1:=diff(u(x,t),t)+c*diff(u(x,t),x)=x*t;
ODEs:=diff(x(r),r)=c/r,diff(t(r),r)=1/r,diff(z(r),r)=x(r)*t(r)/r;
Sol1:=dsolve({ODEs,x(1)=s,t(1)=0,z(1)=h(s)},{x(r),t(r),z(r)});
Sol2:=op(solve(subs(TrVars, Sol1),[s,r,z])); Sol2[3];
W:=unapply(rhs(Sol2[3]),x,t);
Test1:=is(diff(W(x,t),t)+c*diff(W(x,t),x)=x*t); Test2:=is(W(s,0)=h(s));
c:=2; h:=s->sin(s);
PDEplot(PDE1,[s,0,h(s)],s=-3*Pi..3*Pi,x=-3*Pi..3*Pi,t=0..1,color=h(x),
orientation=[-65,58],numchar=15,axes=boxed,style=patchnogrid);
22.4.6 Numerical Solutions of Systems of Linear PDEs

In this section, we show how to obtain numerical and graphical solutions of systems of lin-
ear partial differential equations in Maple with the aid of the predefined function pdsolve.
Example 22.41. Linear first-order system (equivalent to the linear wave equations). Numerical
and graphical solutions. Introducing the dependent variables u(x,t) and v(x,t) according to the
formulas u = wt and v = c2 wx , we can transform the linear wave equation, wtt = c2 wxx , to the
equivalent system of linear first-order equations ut = vx , vt = c2 ux . Consider the initial-boundary
value problem for this linear first-order system (describing standing waves):
ut = vx , vt = c2 ux , 0 < x < L, t > 0,

u(x, 0) = f (x), v(x, 0) = 0, u(0,t) = 0, u(L,t) = 0,
where L = 2π, f (x) = sin x.

We obtain numerical solutions (Sol1, P1) of the given initial-boundary value problem, visualize
it at various times (G1–G3, GU, GV) and in parametric form (PU, PV), construct the animation of a
standing wave profile at some time (e.g., t = 20), and visualize the solutions in 3D (GU3D, GV3D) as
follows:
Digits:=30: with(PDEtools): with(plots):

Ops1:=numpoints=100: Ops2:=color=magenta: Ops3:=color=blue:
Ops4:=color="BlueViolet": Ops5:=axes=boxed,shading=zhue,orientation=[40,50];
c:=1/2; L:=evalf(2*Pi); a:=0: b:=L: Tf:=20;
U,V:=diff_table(u(x,t)),diff_table(v(x,t)): S:=1/30;
Ops:=spacestep=S,timestep=S; f:=x->sin(x); L1:=[0.1,0.2,0.5]; NL1:=nops(L1);
sys1:={U[t]=V[x],V[t]=cˆ2*U[x]}; IBC1:={u(x,0)=f(x),v(x,0)=0, u(0,t)=0,u(L,t)=0};
Sol1:=pdsolve(sys1,IBC1,[u,v],numeric,time=t,range=a..b,Ops);
P1:=Sol1:-value(u,t=0); P1(0.1,0);
for i from 1 to NL1 do G||i:=Sol1:-plot(t=L1[i],Ops1,Ops||(i+1)); od:
display({G1,G2,G3});
GU:=Sol1:-plot(u(x,t),t=Tf,Ops1,Ops2): GV:=Sol1:-plot(v(x,t),t=Tf,Ops1,Ops3):
display({GU,GV});
PU:=r->eval(u,(Sol1:-value(u,t=Tf))(r)): PV:=r->eval(v,(Sol1:-value(v,t=Tf))(r)):
plot([PU,PV,a..b],Ops2,labels=[u,v],Ops1);
Sol1:-animate(t=Tf,Ops1,color=blue,thickness=3);
GU3D:=Sol1:-plot3d(u(x,t),t=0..Tf): GV3D:=Sol1:-plot3d(v(x,t),t=0..Tf):
display(GU3D,Ops5); display(GV3D,Ops5);
⊙ References for Section 22.4: J. Crank and P. Nicolson (1947), L. O. Collatz (1966), R. D. Richtmyer
and K. W. Morton (1967), K. W. Morton and D. F. Mayers (1995), J. W. Thomas (1995), L. Lapidus and
G. F. Pinder (1999), L. Strikwerda (2004), R. J. LeVeque (2007), S. Larsson and V. Thomée (2008), I. K.
Shingareva and C. Lizárraga-Celaya (2011).
Chapter 23

Equations with Mathematica
23.1 Introduction
23.1.1 Some Notational Conventions
In this chapter, we use the following conventions introduced in Mathematica:
C[n] (n=1,2,...) stands for arbitrary constants or arbitrary functions.
In general, arbitrary parameters (e.g., F1 , F2 , . . . ) can be specified by applying the
option GeneratedParameters->(Subscript[F,#]&) of the predefined function DSolve.
We also introduce the following notation (where n=1,2,...) for Mathematica solu-
tions:
eqn for equations
pden/oden for PDEs/ODEs
trn for transformations
sysn for systems
ic, bc, ibc for initial and/or boundary conditions
listn for lists of expressions
gn for graphs of solutions
23.1.2 Brief Introduction to Mathematica

Mathematica is a general-purpose computer algebra system in which symbolic computation
can readily be combined with exact, approximate (floating-point), and arbitrary-precision
numerical computation. Mathematica provides powerful scientific graphics capabilities
[for details, see Bahder (1995), Getz and Helmstedt (2004), Gray (1994), Gray and Glynn
(1991), Green et al. (1994), Ross (1995), Shingareva and Lizárraga-Celaya (2009), Vve-
densky (1993), Zimmerman and Olness (1995), etc.].
The first concept of Mathematica and its first versions were developed by Stephen Wol-
fram in 1979–88. The Wolfram Research company, which continues to develop Mathema-
tica, was founded in 1987.
The most important features of Mathematica are fast symbolic, numerical, acoustic,
and parallel computation; static and dynamic computation, and interactive visualization;
1327
1328 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
incorporation of new user-defined capabilities; availability on almost all operating sys-

tems; powerful and logical programming language; extensive library of mathematical func-
tions and specialized packages; interactive mathematical typesetting system; free resources
(e.g., see the Mathematica Learning Center www.wolfram.com/support/learn, Wolfram
Demonstrations Project demonstrations.wolfram.com, and Wolfram Information Cen-
ter library.wolfram.com).
Mathematica consists of two basic parts: the kernel (computational engine) and the
interface (front end). These two parts are separate but communicate with each other via
the MathLink protocol. The kernel interprets user input and performs all computations.
The kernel assigns the labels In[number] to the input expression and Out[number] to the
output. These labels can be used for keeping the computation order. In this chapter, we do
not include these labels in the examples. The result of kernel operation can be viewed with
the function InputForm. The interface between the user and the kernel is called the front
end and is used to display the input and the output generated by the kernel. The medium of
the front end is the Mathematica notebook.
There are significant changes to numerous Mathematica functions incorporated in the
new versions. The most important differences between Release < 6 and Release ≥ 6 are
described in the literature [e.g., see Shingareva and Lizárraga-Celaya (2009)].1
Mathematica 10 (launched in 2014) is the first version based on the complete Wol-
fram Language and has over 700 new functions (e.g., finite element analysis, enhanced
PDEs, symbolic delay differential equations, hybrid differential equations, highly auto-
mated machine learning, integrated geometric computation, advanced geographic compu-
tation, expanded random process framework, integration with the Wolfram Cloud, introduc-
ing Mathematica Online version, and access to the expanded Wolfram Knowledgebase).
Basic concepts. If we type a Mathematica command and press the RightEnter key or
Shift+ Enter (or Enter to continue the command on the next line), Mathematica evalu-
ates the command, displays the result, and inserts a horizontal line (for the next input).
Mathematica contains many sources of online help, e.g., Wolfram Documentation Cen-
ter, Wolfram Demonstrations Project (for Release ≥ 6), Mathematica Virtual Book (for
Release ≥ 7), and the Help menu; one can mark a function and press F1; type ?func,
??func, Options[func]; use the symbols (?) and (*); e.g., ?Inv*, ?*Plot, or ?*our*.
Mathematica notebooks are electronic documents that may contain Mathematica out-
put, text, and graphics (see ?Notebook). It is possible to work with several notebooks
simultaneously. A Mathematica notebook consists of a list of cells. Cells are indicated by
brackets along the right edge of the notebook. Cells can contain subcells, and so on. The
kernel evaluates a notebook cell by cell. There are various types of cells: input cells (for
evaluation) and text cells (for comments); Title, Subtitle, Section, Subsection, etc., can be
found in the menu Format → Style.
Previous results (during a session) can be referred to with symbols % (the last result),
%% (the next-to-last result), and so on.
Comments can be included within the characters (*comments*).
Incorrect response: if some functions take an “infinite” computation time, you may
1A complete list of all changes can be found in the Documentation Center and on the Wolfram Web Site
www.wolfram.com.
have entered or executed the command incorrectly. To terminate a computation, you can
use Evaluation → Quit Kernel → Local.
Palettes can be used for building or editing mathematical expressions, texts, and graph-
ics and allow one to access the most common mathematical symbols by mouse clicks.
In Mathematica, there exist many specialized functions and modules that are not loaded
initially. They must be loaded separately from files in the Mathematica directory. These
files are of the form filename.m. The full name of a package consists of a context
and a short name and is written as context`short. To load a package correspond-
ing to a context, type <<context`. To get a list of all functions in a package, type
Names["context`*"].
Numerical approximations: N[expr], expr//N (numerical approximation of expr to
6 significant digits); N[expr,n], NumberForm[expr,n] (numerical approximation of the
expression to n significant digits); ScientificForm[expr,n] (scientific notation of nu-
merical approximation of expr to n significant digits).
23.1.3 Mathematica Language
Mathematica language is a very powerful programming language based on systems of

transformation rules and on functional, procedural, and object-oriented programming tech-
niques [see Maeder (1996)]. This distinguishes it from traditional programming languages.
It supports a large collection of data structures, or Mathematica objects (functions, se-
quences, sets, lists, arrays, tables, matrices, vectors, etc.), and operations on these objects
(type-testing, selection, composition, etc.). The library can be extended with custom pro-
grams and packages.
In Mathematica 10, a new concept has been introduced, namely, the complete Wolfram
Language. It has a vast depth of built-in algorithms and knowledge, all accessible auto-
matically through a unified symbolic language. The main idea of the Wolfram Language
is to build as much knowledge (about algorithms and the real world) as possible into the
language.
Symbols refers to tokens with specified names, e.g., expressions, functions, objects,
optional values, results, and argument names. The name of a symbol is a combination of
letters, digits, or certain special characters, not starting from a digit, e.g., a12new. Once
defined, a symbol retains its value until it is changed or removed.
Expression is a symbol that represents an ordinary Mathematica expression expr in
readable form. The head of expr can be obtained with Head[expr]. The structure and var-
ious forms of expr can be analyzed with TreeForm, FullForm[expr], InputForm[expr].
A boolean expression is formed with logical operators and relation operators.
Basic arithmetic operators and the corresponding functions:
+ - * / ˆ , Plus, Subtract, Minus, Times, Divide, and Power.
Logic and relation operators and their equivalent functions: &&, ||, !, =>, ==, != <,
>, <= >=, And, Or, Xor, Not, Implies, Equal, Unequal, Less, Greater, LessEqual, and
GreaterEqual.
Mathematica is case sensitive in that it distinguishes lowercase and uppercase letters;
e.g., Sin[Pi] and sin[Pi] are not the same. All predefined Mathematica functions begin
with a capital letter. Some functions (e.g., PlotPoints) use more than one capital. To
avoid conflicts, it is best to begin with a lowercase letter for all user-defined symbols.
The result of each calculation is displayed unless the output is suppressed by using a
semicolon (;), e.g., Plot[Sin[x],x,0,2*Pi]; a=9; b=3; c=a*b.
Patterns: Mathematica language is based on pattern matching. A pattern is an ex-
pression that contains an underscore character ( ). The pattern can stand for any ex-
pression. Patterns can be constructed from templates; e.g., x , x /;cond, pattern?test,
x :IniValue, xˆn , x ˆn , f[x ], f [x ].
Basic transformation rules: ->, :>, =, :=, ˆ:=, ˆ=.
The rule lhs->rhs transforms lhs to rhs. Mathematica regards the left-hand side
as a pattern. The rule lhs:>rhs transforms lhs to rhs, evaluating rhs only after the
rule is actually used. The assignment lhs=rhs (or Set) specifies that the rule lhs->rhs
should be used whenever it applies. The assignment lhs:=rhs (or SetDelayed) specifies
that lhs:>rhs should be used whenever it applies, i.e., lhs:=rhs does not evaluate rhs
immediately but leaves it unevaluated until the rule is actually called. The rule lhsˆ:=rhs
assigns rhs to be the delayed value of lhs, and associates the assignment with symbols
that occur at level one in lhs. The rule lhsˆ=rhs assigns rhs to be the value of lhs,
and associates the assignment with symbols that occur at level one in lhs. Transformation
rules are useful for making substitutions without making the definitions permanent and are
applied to an expression using the operator /. (ReplaceAll) or //. (ReplaceRepeated).
The difference between the operators (=) and (==) is as follows: the operator lhs=rhs
is used to assign rhs to lhs, and the equality operator lhs==rhs indicates equality (not
assignment) between lhs and rhs.
Unassignment of definitions:
Clear[symb], ClearAll[symb], Remove[symb], symb=.;
Clear["Global`*"]; ClearAll["Global`*"]; Remove["`*"];
(to clear all global symbols defined in a Mathematica session),
?symb, ?`* (to recall a symbol’s definition)
ClearAll["Global‘*"]; Remove["Global‘*"]; is a useful initialization to start
working on a problem.
An equation is represented using the binary operator == and has two operands, the
left-hand side lhs and the right-hand side rhs.
Inequalities are represented using relational operators and have two operands, the left-
hand side lhs and the right-hand side rhs.
A string is a sequence of characters having no value other than itself and can be used as
a label for graphs, tables, and other displays. The strings are enclosed within double-quotes,
e.g., "abc".
Data types: every expression is represented as a tree structure in which each node (and
leaf) has a particular data type. A variety of functions can be used for the analysis of
any node and branch, e.g., Length, Part, and a group of functions ending in the letter Q
(DigitQ, IntegerQ, etc.).
Types of brackets: parentheses for grouping, (x+9)*3; square brackets for function
arguments, Sin[x]; curly brackets for lists, {a,b,c}.
Types of quotes: back-quotes for context mark, format string character, number mark,
precision mark, and accuracy mark; double-quotes for strings.
Types of numbers: integer, rational, real, complex, and root; e.g., -5, 5/6, -2.3ˆ-4,
ScientificForm[-2.3ˆ-4], 3-4*I, Root[#ˆ2+#+1&,2].
Mathematical constants: symbols for definitions of selected mathematical constants,
e.g., Catalan, Degree, E, EulerGamma, I, Pi, Infinity, and GoldenRatio; for example,
{60Degree//N, N[E,30]}.
Two classes of functions: pure functions and functions defined in terms of a variable
(predefined and user-defined functions).
Pure functions are defined without a reference to any specific variable. The arguments
are labeled #1,#2,..., and an ampersand & is used at the end of definition. Most of the
mathematical functions are predefined. Mathematica includes all common special func-
tions of mathematical physics.
The names of mathematical functions are complete English words or the traditional
abbreviations (for a few very common functions), e.g., Conjugate and Mod. Mathematical
functions named after persons have names of the form PersonSymbol, for example, the
Legendre polynomials Pn (x), LegendreP[n,x].
User-defined functions are defined using the pattern x ; e.g., the function f (x) =expr
of one variable is defined as f[x ]:=expr;
Evaluation of a function or an expression without assigning a value can be performed
using the replacement operator /., e.g., f[a], expr/.x->a.
Function application: expr//func is equivalent to fun[expr].
A module is a local object that consists of several functions which one needs to use
repeatedly (see ?Module). A module can be used to define a function (if the function is
too complicated to write by using the notation f[x ]:=expr), to create a matrix, a graph,
a logical value, etc. Block is similar to Module; the main difference between them is that
Block treats the values assigned to symbols as local but the names as global, whereas
Module treats the names of local variables as local. With is similar to Module, the important
difference between them is that With uses local constants that are evaluated only once,
while Module uses local variables whose values may change many times.
In Mathematica language, there are two types of control structures: selection structures
If, Which, Switch and repetition structures Do, While, For.
Mathematica objects: lists are fundamental objects in Mathematica. All other objects
(e.g., sets, matrices, tables, vectors, arrays, tensors, and objects containing data of mixed
type) are represented as lists. A list is an ordered set of objects separated by commas
and enclosed in curly braces, {elements}, or defined with the function List[elements].
Nested lists are lists that contain other lists. There are many functions that manipulate lists,
and here we review some of the most basic ones. Sets are represented as lists. Vectors are
represented as lists; vectors are simple lists. Vectors can be expressed as single columns
with ColumnForm[list,horiz,vert]. Tables, matrices, and tensors are represented as
nested lists. There is no difference between the way they are stored: they can be generated
by using the functions MatrixForm[list] or TableForm[list] or by using the nested
list functions. Matrices and tables can also be conveniently generated by using the Palettes
or Insert menu. A matrix is a list of vectors. A tensor is a list of matrices of the same
dimension.
23.1.4 Dynamic Computation and Visualization in Mathematica Notebook

In Mathematica (for Release ≥ 6), a new kind of manipulation of Mathematica expres-
sions (e.g., computation and visualization), dynamic computation and visualization, has
been introduced, allowing the creation of dynamic and control interfaces of various types.
Numerous new functions for producing interactive elements (or various dynamic and con-
trol interfaces) have been developed within a Mathematica notebook (for more detail, see
the Documentation Center, “Introduction to Manipulate,” “Introduction to Dynamic,” “Dy-
namic and Control,” “Interactive Manipulation,” “How to: Build an Interactive Applica-
tion,” etc.).
Let us mention the most important of them:
Dynamic[expr] Slider[Dynamic[x]]
Slider[x,{x1,x2,xStep}] Manipulate[expr,{x,x1,x2,xStep}]
TabView[{expr1,expr2,...}] SlideView[{expr1,expr2,...}]
DynamicModule[{x=x0,...},expr] Manipulator[expr,{x,x1,x2}]
Animator[x,{x1,x2,dx}] Pane[expr]
Dynamic, DynamicModule representing an object that displays as a dynamically updated current

value of expr; the object can be interactively changed or edited.
Slider, Slider,Dynamic representing sliders of various configurations.
Manipulate, Manipulator generating a version of expr with controls added to allow interac-
tive manipulations of the value of x, etc.
Example 23.1. Linear first-order equation. Cauchy problem. Dynamic and control objects.
Let us create various dynamic and control objects, for example, for the exact solution u(x, y) =
(y + x)n − y of the Cauchy problem
ux − uy = 1, u(x, 0) = xn , n ∈ N,
as follows:
F[x_,y_,n_]:=(x+y)ˆn-y; {r=2, k=9} {Slider[Dynamic[y]],
Dynamic[Plot[F[x,y,k],{x,-r,r},PlotRange->{{-r,r},{-r,r}},
ImageSize->500]]}
DynamicModule[{y=0.5},{Slider[Dynamic[y]],Dynamic[Plot[F[x,y,k],{x,-r,r},
PlotRange->{{-r,r},{-r,r}},ImageSize->500]]}]
TabView[Table[Plot[F[x,y,k],{x,-r,r},PlotRange->{{-r,r},{-r,r}}],{y,-r,r,0.1}]]
SlideView[Table[Plot[F[x,y,k],{x,-r,r},PlotRange->{{-r,r},{-r,r}}],{y,-r,r,0.1}]]
Manipulate[Expand[F[x,y,k]],{k,3,10,1}]
Example 23.2. Linear first-order equation. Cauchy problem. Dynamic object without controls
and animation frame. Let us create a dynamic object without controls and an animation frame, for
example, for the exact solution u(x, y) = − 12 cos x sin y of the boundary value problem
uxx + uyy = cosx sin y (0 < x < π, 0 < y < 2π), (23.1.4.1)
1 1
u(x, 0) = 0, u(x, 2π) = 0, u(0, y) = − sin y, u(π, y) = − sin y, (23.1.4.2)
2 2
as follows:
F[x_,y_]:=-1/2*Cos[x]*Sin[y];
t=0; r=2*Pi;
Row[{Pane[Animator[Dynamic[y],{-r,r}],{0,0}],
Dynamic[Plot[F[x,y],{x,-r,r},
ImageSize->300,PlotRange->{{-r,r},{-r,r}}]]}]
⊙ References for Section 23.1: J. Calmet and J. A. van Hulzen (1983), A. G. Akritas (1989), T. Gray and
J. Glynn (1991), J. H. Davenport, Y. Siret, and E. Tournier (1993), D. D. Vvedensky (1993), J. W. Gray (1994),
E. Green, B. Evans, and J. Johnson (1994), T. B. Bahder (1995), C. C. Ross (1995), R. L. Zimmerman and
F. Olness (1995), M. J. Wester (1999), S. Wolfram (2002, 2003), C. Getz and J. Helmstedt (2004), L. Debnath
(2007), I. K. Shingareva and C. Lizárraga-Celaya (2009, 2011).
23.2 Analytical Solutions and Their Visualizations

It is well known that analytical solutions are often impossible in practice or investigations,
since numerical solutions only can be obtained. However, it is important to understand the
general theory in order to perform a sophisticated research (e.g., certain types of equations
need appropriate boundary conditions and solution methods, problems may be ill posed,
etc.). In this section, we consider various methods for constructing analytical solutions of
linear partial differential equations and their systems.
23.2.1 Constructing Analytical Solutions in Terms of Predefined Functions

In the computer algebra system Mathematica, analytical and symbolic solutions of a given
linear partial differential equation can be found with the aid of the predefined function
DSolve:
DSolve[PDE,u,{x1,..,xn}] DSolve[PDE,u[x1,...,xn],{x1,...,xn}]
DSolve[PDE, u[x1,...,xn], {x1,...,xn}, GeneratedParameters->C]
DSolve finding analytical solutions of a PDE for the function u with independent vari-
ables x1,...xn (“pure function” solution).
DSolve finding analytical solutions of a PDE for the function u[x1,...xn] with indepen-
dent variables x1,...xn.
DSolve, GeneratedParameters finding analytical solutions of a PDE for the function

u[x1,...xn] with independent variables x1,...xn and specifying the arbitrary
constants.
The function DSolve can solve the following classes of linear partial differential equa-
tions (with two or more independent variables and one dependent variable): most first-order
PDEs and a limited number of second-order PDEs.
First, let us assume that we have obtained exact solutions and we wish to verify whether
these solutions are exact solutions of given linear PDEs.
Example 23.3. Linear Poisson equation. Verification of solutions. For the two-dimensional
Poisson equation with a special right-hand side
n n
uxx + uyy = Φ(x, y), Φ(x, y) = − ∑ ∑ ai j exp(bi x + c j y),
i=1 j=1
we verify that
n n
ai j
u(x,t)= − ∑ ∑ 2 2
exp(bi x + c j y)
i=1 j=1 bi + c j
is its exact solution as follows:
pde1=D[u[x,y],{x,2}]+D[u[x,y],{y,2}]+Sum[a[i,j]*Exp[b[i]*x+c[j]*y],{j,1,n},{i,1,n}]==0
sol1=u->Function[{x,y},-Sum[a[i,j]/(b[i]ˆ2+c[j]ˆ2)*Exp[b[i]*x+c[j]*y],{j,n},{i,n}]]
test1=FullSimplify[pde1/.sol1/.n->9]
FullSimplify[sol2[[1,1]]-sol2[[2,2,1]]-sol2[[3,2,1]]]
where ai j , bi , and c j (i = 1, . . . n, j = 1, . . . n) are arbitrary real constants.
Example 23.4. Linear heat equation. Verification of solutions. For the one-dimensional heat
equation (1.1.1)
ut = kuxx ,
we verify that

1 x2
u(x,t) = √ exp − , x ∈ R, t > 0,
2 πkt 4kt
is the fundamental solution of this linear heat equation as follows:
pde1:=D[u[x,t],t]==k*D[u[x,t],{x,2}]
sol1=u->Function[{x,t},1/Sqrt[4*Pi*k*t]*Exp[-xˆ2/(4*k*t)]]
test1=FullSimplify[pde1/.sol1]
where k is a constant.
Example 23.5. Linear wave equation. General solution. The general solution of the one-
dimensional wave equation (4.1.1)
utt = c2 uxx
can be found and tested as follows:
pde1=D[u[x,t],{t,2}]-cˆ2*D[u[x,t],{x,2}]==0
sol1=Assuming[{c\[Element] Reals && c>0},
DSolve[pde1,u,{x,t}]//Simplify]//First
sol2=u->Rationalize[N[PowerExpand[sol1[[1,2]]]]]
test1=Cancel[PowerExpand[pde1/.u->sol1[[1,2]]]] Print["sol2=",sol2]
HoldForm[sol2]==sol2 sol3=u[x,t]/.sol2
sol4=(u[x,t]/.sol2)/.{Table[C[i][var_]->Subscript[F,i][var],{i,1,2}]}//FullSimplify
{sol1,sol2,sol3,sol4}
where the Mathematica result reads:

x x
sol1 = u → Function {x,t},C[1] t − √ +C[2] t + √ ,
c2 c2
h h x i h x ii
sol2 = u → Function {x,t},C[1] t − +C[2] t + ,
h c c
xi h xi
sol3 = C[1] t − +C[2] t + ,
n h c c
xi h x io
sol4 = F1 t − + F2 t + .
c c
According to the Mathematica notation, sol1 is a “pure function” solution for u(x,t) (where C[1]
and C[2] are arbitrary functions), sol2 is a “pure function” simplified solution, sol3 represents the
solution u(x,t), and sol4 represents the solution u(x,t) in a more convenient form, with arbitrary
functions F1 and F2 .
23.2.2 Constructing General Solutions via the Method of Characteristics
Consider a method for finding general solutions of first-order linear equations, the La-
grange method of characteristics. This method allows reducing a PDE to a system of
ODEs along which the given PDE with some initial data (Cauchy data) is integrable. Once
the system of ODEs is found, it can be solved along the characteristic curves and trans-
formed into a general solution of the original PDE. This method was originally proposed
by Lagrange in 1772 and 1779 for solving first-order linear and nonlinear PDEs.
In general, a first-order PDE in two independent variables (x, y) can be written in the
form
F(x, y, u, ux , uy ) = 0 or F(x, y, u, p, q) = 0, (23.2.2.1)
where F is a given function, u = u(x, y) is an unknown function of the independent variables

x and y ((x, y) ∈ D ⊂ R2 ), and ux = p, uy = q. Equation (23.2.2.1) is said to be linear if the
function F is linear in the variables u, ux , and uy and the coefficients of these variables are
functions of the independent variables x and y alone. The most general first-order linear
PDE has the form
A(x, y)ux + B(x, y)uy +C(x, y)u = G(x, y). (23.2.2.2)
The solution u(x, y) of Eq. (23.2.2.1) can be visualized geometrically as a surface, called
an integral surface in the (x, y, u)-space.
A direction vector field or the characteristic direction, (A, B,C), is a tangent vector to
the integral surface f (x, y, u)=0 at the point (x, y, u).
A characteristic curve is a curve in (x, y, u)-space such that the tangent at each point
coincides with the characteristic direction field (A, B,C).
The parametric equations of the characteristic curve can be written in the form x=x(t),
y=y(t), u=u(t), and the tangent vector to this curve (dx/dt, dy/dt, du/dt) is equal to
(A, B,C).
The characteristic equations of Eq. (23.2.2.2) in parametric form are the system of
ODEs dx/dt = A(x, y), dy/dt = B(x, y), du/dt = C(x, y).
The characteristic equations of Eq. (23.2.2.2) in nonparametric form are dx/A(x, y) =

dy/B(x, y) = du/C(x, y). The slope of the characteristics is determined by the equation
dy/dx = B(x, y)/A(x, y).
Example 23.6. First-order linear equations. Direction vector fields. By applying Mathematica
predefined functions to a first-order linear PDE of the form
ux + ut + u = 0, xux + tut + (x2 + t 2 ) = 0,
where {x ∈ R,t ≥ 0}, we can construct the direction vector fields for the given first-order linear
PDE as follows:
r1[x_,t_,u_]:={1,1,-u}; r2[x_,t_,u_]:={x,t,-(xˆ2+tˆ2)};
{n1=10, p=Pi, v1={1,-3,1}, v2={1,2,3}}
SetOptions[VectorPlot3D,VectorColorFunction->Hue,VectorPoints->
{n1,n1,n1},VectorStyle->Arrowheads[0.02],PlotRange->All];
VectorPlot3D[r1[x,t,u],{x,-p,p},{t,-p,p},{u,-p,p},ViewPoint->v1]
VectorPlot3D[r2[x,t,u],{x,-p,p},{t,-p,p},{u,-p,p},ViewPoint->v2]
The general solution (or general integral) of a given first-order PDE is an equation of
the form
f (φ, ψ)=0, (23.2.2.3)
where f is an arbitrary function of the known functions φ = φ(x, y, u) and ψ=ψ(x, y, u),
and provides a solution of this partial differential equation. The functions φ(x, y, u) = C1 ,
ψ(x, y, u) = C2 (where C1 and C2 are constants) are the solution curves of the characteristic
equations, or the families of characteristic curves of Eq. (23.2.2.2).
Example 23.7. First-order equation. General solution. By applying the method of characteris-
tics to the first-order linear PDE
xux + yuy = u,
where {x ∈ R, y ∈ R}, we can prove that the general solution of the given first-order PDE has the
form y u y
f , = 0 or u(x, y) = xg :
x x x
{fU=u[x,y], fF=x, fG=y, fH=u[x,y]} {pde=fF*D[fU,x]+fG*D[fU,y]==fH,

charEqs={dx/fF,dy/fG,du/fH}}
eq10=Map[Exp,Integrate[1/fG,y]-Integrate[1/fF,x]==C10]
eq20=Map[Exp,Integrate[1/u,u]-Integrate[1/fF,x]==C20]
{eq11=eq10/.eq10[[2]]->C1, eq21=eq20/.eq20[[2]]->C2}
genSol=f[eq11[[1]],eq21[[1]]]==0 {genSol1=(genSol/.u->u[x,y]),
gs1=(genSol/.u->u[x,y])[[1]]}
test1=Simplify[pde/.{u[x,y]->gs1}/.{D[u[x,y],x]->D[gs1,x]}/.
{D[u[x,y],y]->D[gs1,y]}/.-u[x,y]->-pde[[1]]]
Equal[genSol1, test1] {DSolve[pde,u,{x,y}],
genSol1=genSol[[1,2]]==g[genSol[[1,1]]]} {genSol2=Solve[genSol1,u],
gs2=genSol2[[1,1,2]]}
test2=Simplify[pde/.{u[x,y]->gs2}/.{D[u[x,y],x]->D[gs2,x]}/.{D[u[x,y],y]->D[gs2,y]}]
Remark. The system of characteristic equations dx/x = dy/y = du/u gives the integral surfaces
(eq11, eq21):
y u
φ = = C1 , ψ = = C2 ,
x x
where C1 and C2 are arbitrary constants. Hence, according to Eq. (23.2.2.3), the general solution of
the linear PDE is f (y/x, u/x) = 0 (genSol), where f is an arbitrary function. This solution can be
verified (test1). By applying the predefined function DSolve, we find the result in a different form:
y
u (x, y) = xg .
x
This form of the general solution can be obtained (genSol2) and verified (test2).
Example 23.8. Linear Euler equation. General solution. By applying the method of character-
istics to the linear Euler equation
xux + yuy = nu,
form y u y
f , n = 0 or u(x, y) = xn g
x x x
as follows:
{fU=u[x,y], fF=x, fG=y, fH=n*u[x,y]}

{pde=fF*D[fU,x]+fG*D[fU,y]==fH, charEqs={dx/fF,dy/fG,du/fH}} eq10=
Map[Exp,Integrate[1/fG,y]-Integrate[1/fF,x]==C10] eq20=
Map[Exp,Integrate[1/(n*u),u]-Integrate[1/fF,x]==C20]
{eq11=eq10/.eq10[[2]]->C1, eq21=eq20/.eq20[[2]]->C2,
eq22=PowerExpand[(eq21[[1]])ˆn==C2]}
genSol=f[eq11[[1]],eq22[[1]]]==0 {genSol1=(genSol/.u->u[x,y]),
gs1=(genSol/.u->u[x,y])[[1]]} {tr1=u[x,y]->gs1,
test11=FullSimplify[pde/.{tr1}/.{D[tr1,x]}/.{D[tr1,y]}],
test12=test11[[1]]->0, test13=Thread[(test11/.test12)/n,Equal], genSol1==test13}
{DSolve[pde,u,{x,y}], genSol1=genSol[[1,2]]==g[genSol[[1,1]]]}
{genSol2 = Solve[genSol1, u], gs2=genSol2[[1,1,2]],
tr2=u[x,y]->gs2} test2=Simplify[pde/.{tr2}/.{D[tr2,x]}/.{D[tr2,y]}]
The system of characteristic equations dx/x = dy/y = du/(nu) gives the integral surfaces (eq11,
eq22):
y u
φ = = C1 , ψ = n = C2 ,
x x
where C1 and C2 are arbitrary constants. Hence, according to Eq. (23.2.2.3), the general solution of
the linear PDE is f (y/x, x−n u) = 0 (genSol), where f is an arbitrary function. This solution can be
verified (test11, test12, test13). By applying the predefined function DSolve, we find the result in
a different form: y
u (x, y) = xn g .
x
This form of the general solution can be obtained (genSol2) and verified (test2).
23.2.3 Constructing General Solutions via Conversion to Canonical Forms

Consider the general first-order linear partial differential equation
A(x, y)ux + B(x, y)uy +C(x, y)u = G(x, y).
By introducing a new transformation by the equations ξ = ξ(x, y), η = η(x, y)2 we can
reduce the original equation to the canonical (or standard) form
uξ + α(ξ, η)u = β(ξ, η),
where α(ξ, η) = C̃/Ã and β(ξ, η) = G̃/Ã with Ã = uξx + Bξy and B̃ = Aηx + Bηy , so that the
equation is Ãuξ + B̃uη + C̃u = G̃. This equation can be integrated, and the general solution
of the original equation can be found.
Example 23.9. Linear first-order equation. Canonical form. General solution. Considering the
linear first-order PDE
ux + uy = u,
where {x ∈ R, y ∈ R}, we can obtain the transformation ξ = −x + y, η = x (eqxi, eqeta) and the
canonical form uη = u (canForm). By integrating this equation, we find the general solution
u(x, y) = ex F(−x + y),

where F(−x + y) is an arbitrary function. Finally, we can find and test the general solution using
predefined functions as follows:
{aA=1,bB=1,cC=-1,gG=0}
pde=aA*D[u[x,y],x]+bB*D[u[x,y],y]+cC*u[x,y]==0
charEqs={dx/aA,dy/bB,du/u[x,y]}
{eqxi=xi==Integrate[1/bB,y]-Integrate[1/aA,x], eqeta=eta==x}
testJ=D[eqxi[[2]],x]*D[eqeta[[2]],y]-D[eqxi[[2]],y]*D[eqeta[[2]],x]
ux=D[u[xi,eta],xi]*D[eqxi[[2]],x]+D[u[xi,eta],eta]*D[eqeta[[2]],x]
uy=D[u[xi,eta],xi]*D[eqxi[[2]],y]+D[u[xi,eta],eta]*D[eqeta[[2]],y]
{canForm=aA*ux+bB*uy==-cC*u[xi,eta],
canForm1=Thread[canForm/canForm[[2]],Equal]}
eq1=Integrate[canForm1[[1]],eta]==Integrate[canForm1[[2]],eta]+Log[F[xi]]
eq2=u[xi,eta]->(Solve[eq1,u[xi,eta]])[[1,1,2,1]]
genSol=u[x,y]==eq2[[2]]/.{eqxi//ToRules}/.{eqeta//ToRules}
{genSolPred=DSolve[pde,u,{x,y}], gSol=ToRules[genSol[[1,1]]]}
{test1=pde/.genSolPred, test2=pde/.gSol/.D[gSol,x]/.D[gSol,y]}
For linear second-order PDEs, we consider the classification of equations (which does
not depend on their solutions and is determined by the coefficients of the highest deriva-
tives) and the reduction of a given equation to appropriate canonical forms.
Let us introduce the new variables a = F p , b = 12 Fq , c = Fr , and calculate the discrim-
inant δ = b2 − ac at some point. Depending on the sign of the discriminant δ, the type of
equation at a specific point can be parabolic (if δ = 0), hyperbolic (if δ > 0), or elliptic (if
δ < 0). Let us call the following equations
uy1 y2 = f1 (y1 , y2 , u, uy1 , uy2 ), uz1 z1 − uz2 z2 = f2 (z1 , z2 , u, uz1 , uz2 )

2 Here the functions ξ and η are continuously differentiable and the Jacobian J(x, y) = ξ η − ξ η x is
x y y ,
nonzero in a domain D.
the first canonical form and the second canonical form for hyperbolic PDEs, respectively.
the linear second-order PDEs
−2y2 uxx + 12 x2 uyy = 0,
where {x ∈ R, y ∈ R}, we can verify that this equation is hyperbolic everywhere (except at the point
x=0, y=0) and prove that the change of variables ξ = − 12 x2 + y2 , η = 12 x2 + y2 transforms the PDE
to the first and second canonical forms, respectively,
vξ η − vη ξ 1 vλ vµ
vηξ + = 0, vλλ − vµµ + − = 0.
2(η2 − ξ2 ) 2 λ µ
1. Classification. In standard notation, this linear equation takes the form

1
F1 = −2y2 p + x2 r = 0.
2
The new variables are a = −2y2 , b = 0, c = 12 x2 (tr2(f1)) and the discriminant δ = b2 − ac = x2 y2
(delta1) is positive except at the point x = 0, y = 0.
pde1=-2*yˆ2*D[u[x,y],{x,2}]+xˆ2*D[u[x,y],{y,2}]/2==0
tr1[x_,y_,u_]:={D[u[x,y],{x,2}]->p,D[u[x,y],{y,2}]->r,D[u[x,y],{x,y}]->q};
tr2[f_]:={a->D[f[p,q,r][[1]],p],b->1/2*D[f[p,q,r][[1]],q],c->D[f[p,q,r][[1]],r]};
f1[p_,r_,q_]:=pde1/.tr1[x,y,u]; delta=bˆ2-a*c {f1[p,r,q], tr2[f1],
delta1=delta/.tr2[f1]-f1[p,r,q][[2]]} {Reduce[delta1>0],
FindInstance[delta1>0,{x,y}]}
The same result can be obtained with the principal part coefficient matrix as follows:
{a1={{-2*yˆ2,0},{0,xˆ2/2}},d1=Det[a1],Reduce[d1<0],
FindInstance[d1<0,{x,y}]}
Here we calculate the determinant d1 of the matrix a1. PDEs can be classified according to the
eigenvalues of the matrix a1, i.e., depending on the sign of d1: if d1=0, then it is parabolic; if d1<0,
then it is hyperbolic; and if d1>0, then it is elliptic.
2. Canonical forms. We find a change of variables that transforms the PDE to the first and
second canonical forms as follows:
jacobianM[f_List?VectorQ,x_List]:=Outer[D,f,x]/;Equal@@(Dimensions/@{f,x});
hessianH[f_,x_List?VectorQ]:=D[f,{x,2}];
gradF[f_,x_List?VectorQ]:=D[f,{x}]; op1[expr_]:=expr/.y->y[x];
op2[expr_]:=expr/.y[x]->y; {vars=Sequence[x,y],
varsN=Sequence[xi,eta]}
m1=Assuming[{x>0,y>0},Simplify[(-a1[[1,2]]+Sqrt[-d1])/a1[[1,1]]]]
m2=Assuming[{x>0,y>0},Simplify[(-a1[[1,2]]-Sqrt[-d1])/a1[[1,1]]]]
{eq1=DSolve[D[y[x],x]==-op1[m1],y[x],x],
eq11=eq1[[1,1,1]]ˆ2==eq1[[1,1,2]]ˆ2}
{eq12=Solve[eq11,C[1]][[1,1,2]], g[1]=Expand[op2[eq12]*2]}
{eq2=DSolve[D[y[x],x]==-op1[m2],y[x],x],
eq21=eq2[[1,1,1]]ˆ2==eq2[[1,1,2]]ˆ2}
{eq22=Solve[eq21,C[1]][[1,1,2]], g[2]=Expand[op2[eq22]*2]}
{jg=jacobianM[{g[1],g[2]},{vars}], dv=gradF[v[varsN],{varsN}]}
ddv=hessianH[v[varsN],{varsN}]
ddu=Transpose[jg].ddv.jg+Sum[dv[[i]]*hessianH[g[i],{vars}],{i,1,2}]
{eq3=Simplify[Tr[a1.ddu]]==0, tr0={yˆ2->Y,xˆ2->X},
tr01={Y->yˆ2,X->xˆ2}}
tr1=Flatten[{Expand[Solve[First[g[2]==eta/.{Solve[g[1]==xi/.tr0,X]/.tr01}/.
tr0],Y]/.tr01],Expand[Solve[First[g[1]==xi/.
{Solve[g[2]==eta/.tr0,Y]/.tr01}/.tr0],X]/.tr01]}]
nForm=Collect[Expand[eq3/.tr1],D[v[varsN],varsN]]
c1=Coefficient[nForm[[1]],D[v[varsN],varsN]]
normalFormF=Collect[Thread[nForm/c1,Equal],D[v[varsN],varsN]]
nF[x_,t_]:=D[D[u[x,t],x],t]+(2*t*D[u[x,t],x]-2*x*D[u[x,t],t])/(4*tˆ2-4*xˆ2)==0;
tr2={xi->lambda+mu,eta->mu-lambda}; nF[xi,eta]
nFT[v_]:=((Simplify[nF[xi,eta]/.u->Function[{xi,eta},
u[(xi-eta)/2,(xi+eta)/2]]])/.tr2//ExpandAll)/.{u->v}; canonicalForm=nFT[v]

x2 uxx + 2xyuxy + y2uyy = 0,
where {x ∈ R, y ∈ R}, we can verify that this equation is parabolic everywhere and prove that the
canonical forms of the PDE are x2 vξξ = 0 and vξξ = 0, respectively.
jacobianM[f_List?VectorQ,
x_List]:=Outer[D,f,x]/;Equal@@(Dimensions/@{f,x});
hessianH[f_,x_List?VectorQ]:=D[f,{x,2}];
gradF[f_,x_List?VectorQ]:=D[f,{x}]; op1[expr_]:=expr/.y->y[x];
op2[expr_]:=expr/.y[x]->y; {vars=Sequence[x,y],
varsN=Sequence[xi,eta]} {a1={{xˆ2,x*y},{x*y,yˆ2}}, d1=Det[a1]}
m1=Assuming[{x>0,y>0},Simplify[(-a1[[1,2]]+Sqrt[-d1])/a1[[1,1]]]]
eq1=DSolve[D[y[x],x]==-op1[m1],y[x],x]/.Rule->Equal//First
{eq11=Solve[eq1,C[1]][[1,1,2]], g[1]=op2[Eq11], g[2]=x}
{jg=jacobianM[{g[1],g[2]},{vars}], dv=gradF[v[varsN],{varsN}]}
ddv=hessianH[v[varsN],{varsN}]
ddu=Transpose[jg].ddv.jg+Sum[dv[[i]]*hessianH[g[i],{vars}],{i,1,2}]
{norF=Simplify[Tr[a1.ddu]]==0, canF=Thread[norF/xˆ2,Equal]//Expand}
23.2.4 Constructing Analytical Solutions of Cauchy Problems

If we consider a mathematical problem in an unbounded domain, the solution can be deter-
mined uniquely by prescribing initial conditions. The corresponding problem is called the
initial value problem or the Cauchy problem.
Mathematical problems can be well-posed or ill-posed problems. A mathematical prob-
lem is well posed if it possesses the following features: existence (there exists at least one
solution), uniqueness (there exists at most one solution), and continuity (the solution de-
pends continuously on the data).
In Mathematica, it is possible to construct exact solutions for some linear partial differ-
ential equations subject to initial conditions with the aid of the predefined function DSolve.
Let us solve some initial value problems.
Example 23.12. Linear first-order PDEs. Initial value problem (IVP). Exact solution. Consid-
ering the Cauchy problem for the linear first-order partial differential equation,
yux + xuy = u(x, y), u(x, 0) = F(x),
and applying Mathematica predefined functions, we can solve the Cauchy problem for this equation
and verify that of the two solutions obtained,
p p p
x2 − y2 F x2 − y2 (x + y)F x2 − y2
u(x, y) = , u(x, y) = p ,
x+y x2 − y2
the second one is an exact solution of this initial value problem as follows:
{pde1=y*D[u[x,y],x]+x*D[u[x,y],y]==u[x,y], ic1=u[x,0]==F[x],
sys1={pde1,ic1}} {sol1=DSolve[sys1,u[x,y],{x,y}],
sol2=DSolve[sys1,u[x,y],{x,y}]//Simplify,
sol3=ComplexExpand[DSolve[sys1,u[x,y],{x,y}]]//Simplify}
{s1=sol3[[1,1,2]], s2=sol3[[2,1,2]]}
t1PDEl=pde1[[1]]/.u[x,y]->s1/.D[u[x,y],x]->D[s1,x]/.D[u[x,y],y]->D[s1,y]//Simplify
t1PDEr=pde1[[2]]/.u[x,y]->s1//Simplify {t1PDEl===t1PDEr,
testIC=ic1/.u[x,0]->s1/.y->0//Simplify}
t2PDEl=pde1[[1]]/.u[x,y]->s2/.D[u[x,y],x]->D[s2,x]/.D[u[x,y],y]->D[s2,y]//Simplify
t2PDEr=pde1[[2]]/.u[x,y]->s2//Simplify {t2PDEl===t2PDEr,
testIC=ic1/.u[x,0]->s2/.y->0//Simplify}
Note that the predefined function DSolve always treats all variables as complex, so the general
form of the solution is complex (sol1). However, we have to obtain real solutions, so we can
add real initial conditions, some assumptions, and simplify the resulting solution (e.g., Simplify,
ComplexExpand, sol2, sol3).
Example 23.13. Linear first-order equation. Method of characteristics. Classical Cauchy

problem. Considering the first-order linear partial differential equation with the Cauchy data,
ux − uy = 1, u(x, 0) = xn (n ∈ N),
and applying the method of characteristics, we can obtain the general and particular solutions of
this equation,
u(x, y) = f (y + x) + x, u(x, y) = (y + x)n − y,
respectively, and plot the characteristic curves as follows:
SetOptions[ParametricPlot,ImageSize->200,PlotStyle->{Hue[0.7],Thickness[0.01]}];
{ode=D[uN[x],x]==1, solCh=DSolve[ode,uN[x],x], eqCh=D[y[x],x]==-1}
{curCh=DSolve[eqCh,y[x],x]//First, tr1=uN[x]->u, tr2=y[x]->y}
g=Table[ParametricPlot[{(curCh[[1,2]]/.C[1]->y),x},{x,0,9}],{y,0,9}];
Show[g,PlotRange->{{0,9},{0,9}},AspectRatio->1]
eq1=Solve[(solCh/.Rule->Equal)[[1]]/.tr1,C[1]]
eq2=Solve[(curCh/.Rule->Equal)[[1]]/.tr2,C[1]]
eq3=eq1[[1,1,2]]==f[eq2[[1,1,2]]]
fN[xN_]:=eq3[[1]]/.{y->0,u->xˆn}/.x->xN; fN[x]
sol1=Solve[eq3,u]/.f[eq2[[1,1,2]]]->fN[eq2[[1,1,2]]]
The same result can be obtained by applying the predefined functions:
{pde=D[v[x,y],x]-D[v[x,y],y]==1, ic=v[x,0]==xˆn, sys1={pde,ic}}

{sol2=DSolve[sys1,v,{x,y}], test=sys1/.sol2}
Example 23.14. First-order linear equation. Method of characteristics. Classical Cauchy

problem. Let us solve the first-order linear equation with the Cauchy data
ut − xux = u, u(x, 0) = f (x)
by the method of characteristics. This equation can be obtained from the Fokker–Planck equation
[see Bluman et al. (2010)]3
ut = uxx + (xu)x ,
by neglecting the term uxx .
We can obtain the solution of this Cauchy problem
u1 (x,t) = f (xet ) et
and plot the characteristic curves as follows:
SetOptions[ParametricPlot,ImageSize->300,PlotStyle->{Hue[0.7],Thickness[0.001]}];
{tF=Pi/2, xF=2*Pi, tr1=x->x[t], tr2=x[t]->x,
tr3=f[xN[0]]->f1[xN[0]]} f1[x_]:=Cos[x]; f1[x]
{ode1=D[x[t],t]==-x[t], ode2=D[uN[t],t]==uN[t],
sol2=DSolve[ode2,uN[t],t]} {sol21=sol2[[1,1,2]],
iniCond=u[xN[0],0]==f[xN[0]]} {const2=(sol21/.t->0)->iniCond[[2]],
sol22=uN[t]->(sol21/.const2)} {ode11=ode1/.sol22,
sol1=DSolve[ode11,x[t],t], xN0=sol1/.t->0//First}
{const1=Solve[xN0/.Rule->Equal,C[1]],
chars=sol1/.const1/.x[0]->xN[0]} {genSol=sol22[[2]],
char1=chars/.tr2/.tr3//First//First}
trxN0=Solve[char1/.Rule->Equal,xN[0]]
u1[xN_,tN_]:=genSol/.trxN0/.{x->xN,t->tN}; u1[x,t]
u2[x1_,t1_]:=genSol/.f[xN[0]]->f1[trxN0[[1,1,2]]]/.{x->x1,t->t1};
u2[x,t]
g1=Table[ParametricPlot[{(char1[[1,2]]/.xN[0]->x),t},{t,0,tF}],{x,-xF,xF}];
Show[g1,PlotRange->{{-xF,xF},{0,tF}},AspectRatio->1]
g2=Table[ParametricPlot[{u2[x,t],t},{t,0,tF}],{x,-xF,xF}];
Show[g2,PlotRange->{{-xF,xF},{0,tF}},AspectRatio->1]
We show that the implicit form of the solution (or parametric representation of the solution) of this
Cauchy problem in the Mathematica notation reads
sol22 := uN [t] = f [xN[0]] et chars := x [t] = xN[0] e−t
The characteristic curves for f (x) = cos x are presented in Fig. 23.1.
23.2.5 Constructing Analytical Solutions of Boundary Value Problems

In Mathematica, by applying the predefined function DSolve, it is possible to find analyt-
ical solutions for restricted classes of linear PDEs (e.g., linear first-order PDEs and linear
homogeneous second-order PDEs of the form auxx + buxy + cuyy = 0). However, in this
section, we solve a specific boundary value problem associated with linear PDEs that do
not belong to such restricted classes.
3 The Fokker–Planck equation arises in various applications of statistical mechanics; it describes the evolu-
tion of the probability distribution function.

t
t
1.4
1.4
1.2
1.2
1.0 1.0
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
x 0.0
-6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6
Figure 23.1 Characteristic curves for ut − xux = u and the solution u2 (x,t) (for f (x) = cos x).
Example 23.15. Linear Poisson equation with polynomial right-hand side. Polynomial bound-
ary conditions. Exact solution. Consider the Dirichlet boundary value problem of a special form,
i.e., the linear Poisson equation with polynomial right-hand side and polynomial boundary condi-
tions:
uxx + uyy = q(x, y), 0 < x < m, 0 < y < m,

u(0, y) = 0, u(m, y) = f (y), u(x, 0) = 0, u(x, m) = 0,
where m is a parameter (m ∈ Q, m > 0); e.g., m = 5/2. The right-hand side q(x, y) of the Poisson
equation is a polynomial of degree n of the form
n
q(x, y) = ∑ ak xn−k yk ,
k=0
where the coefficients ak are determined in the solution process. Polynomial boundary conditions
are given, e.g., f (y) = y3 − y.
Since it is not possible to solve such problems with the aid of the predefined function DSolve,
we propose a new method for solving Dirichlet boundary value problems of special form on the
basis of some ideas [see Muleshkov et al. (2002)] for constructing exact solutions of the Poisson
equation. This method can also be applied for another configurations of the problem, e.g., when
the coefficients ak are given and the boundary conditions are determined in the solution process.
Looking for a solution of the given Poisson problem in the form
n
u(x, y) = ∑ pk xn−k+2 yk ,
k=0
we find the coefficients ak and pk (trAP) that satisfy the boundary conditions and the Poisson
equation
2(50(48 + 125y − 48y2) − 125x(−25 + 8y2) + 4x2(−48 − 125y + 48y2))

a0 = 0, a1 = ,
25(−625 + 16x2y2 )
x(−15625x2 + 400x4y2 + 6250y3 + 8xy3(−48 − 125y + 48y2))
a2 = 1, a3 = ,
75y3 (−625 + 16x2y2 )
48 + 125y − 4(12 + 5x)y2
p0 = 0, p1 =
−1875 + 48x2y2
and the exact solution
x3 y(−5 + 2y)(3125y + 2x(−240 + y(−721 + 2y(−5 + 50x + 48y))))

,
150(−625 + 16x2y2 )
and verify that this solution satisfies the given Poisson problem (testPDE, testBCs) as follows:
{n=3, m=5/2, pde=D[u[x,y],{x,2}]+D[u[x,y],{y,2}]==q[x,y]}

q[x_,y_]:=Sum[a[k]*xˆ(n-k)*yˆk,{k,0,n}]; q[x,y]
u[x_,y_]:=Sum[p[k]*xˆ(n-k+2)*yˆk,{k,0,n}]; f[y_]:=yˆ3-y; {u[x,y],
f[y]} lRange={Range[0,n],Range[0,n]}
eq1=Tuples[lRange]/.Append[CoefficientRules[q[x,y],{x,y}],{_,_}->0]
eq2=Tuples[lRange]/.Append[CoefficientRules[pde[[1]],{x,y}],{_,_}->0]
{eq3=Thread[eq2==eq1], trPk=Solve[eq3]//First, u[x_,y_]=u[x,
y]/.trPk} Map[Simplify, {u[0,y],u[m,y],u[x,0],u[x,m]}]
{bc1=u[0,y]==0, bc2=u[m,y]==f[y], bc3=u[x,0]==0, bc4=u[x,m]==0}
{p0=Solve[bc3,p[0]]//First, a2=a[2]->1,
p10=Solve[bc2/.p0/.a2,p[1]]//First,
a1=Solve[bc4/.p0/.p10/.a2,a[1]]//Simplify//First, p1=p10/.a1//Simplify}
{bcs={u[0,y],u[m,y],u[x,0],u[x,m]},
trCoeff=Flatten[{a1,p0,p1,a2}]//Simplify} testBCs=Map[#1/.trCoeff&,
bcs]//Simplify {pde1=pde/.trCoeff//Simplify, a0=a[0]->0,
pde2=pde1/.a0//FullSimplify,
a3=Solve[pde2,a[3]]//First, pde/.a3/.a0/.trCoeff//FullSimplify}
trAP=Flatten[{a0,a1,a2,a3,p0,p1}]//Simplify
{testPDE=pde/.trAP//Simplify, solF=u[x, y]/.trAP//FullSimplify}
23.2.6 Constructing Analytical Solutions of Initial-Boundary Value Problems
In Mathematica, by applying the predefined function DSolve, it is possible to find analyt-

ical solutions for restricted classes of linear PDEs (e.g., linear first-order PDEs and linear
homogeneous second-order PDEs of the form auxx + buxy + cuyy = 0). However, in this
section we solve some specific boundary value problems associated with linear PDEs that
do not belong to such restricted classes.
If a mathematical problem consists in finding an unknown solution of a PDE (defined at
an appropriate domain) satisfying appropriate supplementary conditions (initial and bound-
ary conditions), this is known as an initial-boundary value problem. The boundary condi-
tions describe the unknown function at prescribed boundary points. The initial condition
prescribes the unknown function at a certain initial time t (e.g., t = t0 , t = 0).
In Mathematica, it is possible to construct exact solutions for a restricted number of lin-
ear partial differential equations with the aid of the predefined function DSolve (e.g., single
first-order equations and some linear homogeneous second-order equations). However, we
cannot solve initial-boundary value problems with the aid of the function DSolve even for
simple linear first-order or second-order equations, e.g.,
pde1={D[u[x,t],t]==D[u[x,t],x], u[x,0]==f[x], u[0,t]==g[t]}

pde2={D[u[x,t],t]==D[u[x,t],{x,2}],
u[x,0]==f[x],(D[u[x,t],t]/.t->0)==g[x],u[0,t]==0}
{DSolve[pde1,u,{x,t}], DSolve[pde2,u,{x,t}]}
In this section, we solve a specific initial-boundary value problem associated with linear
PDEs that do not belong to such restricted classes.
Example 23.16. Linear telegraph equation. Initial-boundary value problem (IBVP). Consider
the initial-boundary value problem for the linear telegraph equation
uxx = utt + ut − u, x > 0, t > 0,

u(x, 0) = e , ut (x, 0) = −2ex , u(0,t) = e−2t ,
x
ux (0,t) = e−2t , x ≥ 0, t ≥ 0.
It is not possible to solve such problems automatically via the predefined function DSolve, but by
using separation of variables [for more details, see Shingareva and Lizárraga-Celaya (2011)], we
can obtain the solution of this initial-boundary value problem (SolF)
u(x,t) = e−2t+x
and verify that this solution is an exact solution of the linear telegraph equation (testPDE, testIC1,
testIC2, testBC1, testBC1) as follows:
tr1=w[x,t]->phi[x]*psi[t]; trC=c->1;
tr1D[v_]:=Table[D[tr1,{v,i}],{i,1,2}];
pde1[u_]:=D[u,{x,2}]==D[u,{t,2}]+D[u,t]-u;
{eq2=Expand[pde1[w[x,t]]],
eq3=PowerExpand[eq2/.tr1/.tr1D[t]/.tr1D[x]],
eq4=Factor[Thread[eq3/(phi[x]*psi[t]),Equal]], eq5=Map[Simplify,eq4]}
{solPhi=DSolve[eq5[[1]]==cˆ2,phi,x],
solPsi=DSolve[eq5[[2]]==cˆ2,psi,t]}
solU=(phi[x]/.solPhi)*(psi[t]/.solPsi)//Simplify//First
Map[Simplify, {ic1=(solU/.t->0)==Exp[x],
ic2=(D[solU,t]/.t->0)==-2*Exp[x],
bc1=(solU/.x->0)==Exp[-2*t], bc2=(D[solU,x]/.x->0)==Exp[-2*t]}]
{sys1=Map[#/.trC&,{ic1,ic2,bc1,bc2}]//Simplify,
trCoeff=Solve[sys1,{C[1],C[2]}]} solF=solU/.trCoeff/.trC
{testPDE=pde1[solF]//FullSimplify, testIC1=(solF/.t->0)==ic1[[2]],
testIC2=(D[solF,t]/.t->0)==ic2[[2]], testBC1=(solF/.x->0)==bc1[[2]],
testBC2=(D[solF,x]/.x->0)==bc2[[2]]}
23.2.7 Constructing Analytical Solutions of Systems of Linear PDEs

In the computer algebra system Mathematica, analytical (symbolic) solutions of restricted
classes of systems of linear partial differential equations can be found with the aid of the
predefined function DSolve:
DSolve[PDESys,u, {x1,x2}] DSolve[PDEsys,u[x1,x2],{x1,x2}]

DSolve[PDEsys,{u1[x1,...,xn],...,un[x1,...,xn]},{x1,...,xn}]
DSolve, finding analytical solutions of a system of PDEs for the function u with two
independent variables x1,x2 (“pure function” solution).
DSolve, finding analytical solutions of a system of PDEs for the function u[x1,x2] with
independent variables x1,x2.
0.5
0.0 x
- 0.5
- 15 - 10 -5 0 5 10 15
Figure 23.2 Exact solution of the linear system ux = sin(x2 ) + y, uy = cos(y2 ) + x (y = 0, C1 = 0).
DSolve, finding analytical solutions of an uncoupled system of PDEs for the functions
u1[x1,...xn], . . . , un[x1,...xn] with independent variables x1,...xn.
The function DSolve can solve some simple classes of systems of linear partial differ-
ential equations, e.g., first-order systems with two independent and one dependent variable
and first-order uncoupled systems with two or more independent variables and two or more
dependent variables. By applying Mathematica predefined functions, we will show how to
find analytical solutions of such linear first-order systems.
Example 23.17. Linear first-order system. Exact solution. Considering the linear first-order
system with two independent variables and one dependent variable,
ux = sin(x2 ) + y, uy = cos(y2 ) + x,
and applying Mathematica predefined functions, we solve this system and verify that the solution
obtained (sol1),
p hp i p hp i
u(x, y) = xy + C1 + π/2 FresnelC 2/πy + π/2 FresnelS 2/πx
is an exact solution of this system as follows:
sys1={D[u[x,y],x]==Sin[xˆ2]+y, D[u[x,y],y]==Cos[yˆ2]+x}
sol1=DSolve[sys1,u[x,y],{x,y}]//First
Plot[sol1[[1,2]]/.y->0/.C[1]->0,{x,-15,15}]
The exact solution of this system for y = 0 and C1 = 0 is presented in Fig. 23.2.
By applying the elimination procedure, we present an alternative (equivalent) approach to solv-
ing this linear system as follows:
Off[Eliminate::ifun]; sys={D[u[x,y],x]==Sin[xˆ2]+y,
D[u[x,y],y]==Cos[yˆ2]+x}; {v=Eliminate[sys,y],
sol1=DSolve[v[[2]],u[x,y],{x,y}]//First} {u[x_,y_]=sol1[[1, 2]],
sys1=Simplify[sys]} {sol2=DSolve[sys1[[2]],C[1][y],y]//First,
u[x,y]/.sol2}
Example 23.18. Linear first-order system. Exact solution. Considering the linear first-order
uncoupled system with two independent variables and two dependent variables,
ut + ux = sin t cos x, vt + 2vx = sin x sin t,
and applying Mathematica predefined functions, we solve this system of PDEs and verify that the
solution obtained (sol3),
1 1 1 1
u(x,t) = − cos(x+t)+F1(t −x)+ x sin(t −x), v(x,t) = F2 (t −x/2)− sin(t −x)− sin(t +x),
4 2 2 6
is an exact solution of this linear system as follows:
sys1={D[u[x,t],t]+D[u[x,t],x]==Sin[t]*Cos[x],D[v[x,t],t]+2*D[v[x,t],x]==Sin[x]*Sin[t]}
sol1=DSolve[sys1,{u,v},{x,t}]//First test1=sys1/.sol1//FullSimplify
sol2={sol1[[1]]/.C[1]->F1,sol1[[2]]/.C[1]->F2}
Map[ExpandAll,{sol2[[1,2,2]],sol2[[2,2,2]]}]
Since the predefined function DSolve represents the two arbitrary functions as C[1], it is
necessary to specify new names of the arbitrary functions for each solution function (in our case,
F1 and F2).
⊙ References for Section 23.2: A. S. Muleshkov, M. A. Golberg, A. H.-D. Cheng, and C. S. Chen (2002),
G. W. Bluman, A. F. Cheviakov, and S. C. Anco (2010), I. K. Shingareva and C. Lizárraga-Celaya (2011).
23.3 Analytical Solutions of Mathematical Problems
23.3.1 Constructing Separable Solutions
Separation of variables is one of the most important methods for solving linear PDEs,
in which the structure of a PDE allows us to seek multiplicative separable or additive
separable exact solutions; e.g., u(x,t) = φ(x) ◦ ψ(t) (where the multiplication or addition is
denoted by ◦). Numerous problems in linear partial differential equations can be solved by
separation of variables. This method was recently generalized [e.g., see Galaktionov (1990,
1995), Polyanin and Zhurov (1998), Polyanin and Manzhirov (2007)] and nowadays it is a
classical method in mathematics and physics.
Let us start from first-order linear equations that can be solved by separation of variables
without considering Fourier series.
Example 23.20. First-order linear equation. Separable solution. Cauchy problem. Consider
the first-order linear PDE and the Cauchy data
aux + buy = 0, u(0, y) = αe−βy ,
where a, b ∈ R are parameters. By applying separation of variables and by seeking exact solutions
in the form u(x,t) = φ(x)ψ(y), we arrive at the following equations (Eq51, Eq52):
aφ′x ψ′y
= −C1 , = C1 .
bφ(x) ψ(y)
tr1=w[x,y]->phi[x]*psi[y]; pde1[u_]=a*D[u[x,y],x]+b*D[u[x,y],y]==0;
pde1[u] {ic1=u[0,y]==alpha*Exp[-beta*y], eq2=pde1[w]//Expand}
eq3=eq2/.w[x,y]->tr1/.D[tr1,x]/.D[tr1,y]//Expand
eq4=Thread[eq3/tr1[[2]]/b,Equal]//Expand {eq51=eq4[[1,1]]==-C[1],
eq52=eq4[[1,2]]==C[1]}
Then we seek exact solutions of these equations as follows:
{sol1=DSolve[eq51,phi[x],x]//First,
sol2=DSolve[eq52,psi[y],y]//First} genSol=Simplify[tr1/.sol1/.sol2]
{trC3=C[2]ˆ2->C[3], genSol1=genSol/.trC3}
{eq8=(genSol1[[2]]/.x->0)==ic1[[2]],
trC13={C[1]->-beta,C[3]->alpha}} solF=genSol1/.trC13//Factor
test1=pde1[w]/.solF/.D[solF,x]/.D[solF,y]//Expand
test2=(solF[[2]]/.x->0)==ic1[[2]]
We find that φ(x) = C2 e−C1 bx/a (sol1), ψ(y) = C2 eC1 y (sol2), and the exact solution acquires the
form (solF)
u(x, y) = αe−β(ay−bx)/a .
Separation of variables combined with the linear superposition principle can be applied
for solving a large class of initial-boundary value problems for linear partial differential
equations. According to the method, the partial differential equation is reduced to two or-
dinary differential equations (for φ(x) and ψ(y)). Similar ideas can be applied to equations
in several independent variables. This method is also known as the Fourier method or the
eigenfunction expansion method.
Example 23.21. Linear wave equation. Separable solution. Fourier series. Initial-boundary
value problem.
Consider a second-order linear hyperbolic PDE with the initial and boundary conditions
utt = c2 uxx , 0 < x < L, t > 0,

u(x, 0) = f (x), ut (x, 0) = g(x) (0 ≤ x ≤ L), u(0,t) = 0, u(L,t) = 0(t ≥ 0),
where c2 = 14 , L = 1, f (x) = 0, g(x) = sin(x) − sin(3πx). This problem describes a vibrating string
(with constant tension T and density ρ, c2 = T /ρ) stretched along the x-axis from 0 to L and fixed
at the endpoints. The initial displacement f (x) is zero, and the initial velocity is g(x).
By applying separation of variables, i.e., by seeking the exact solution in the form u(x,t) =
φ(x)ψ(t), we obtain the two ODEs
φ′′xx − λφ = 0, ψtt′′ − λc2ψ = 0.
Then, separating the boundary conditions, we solve the eigenvalue problem and obtain the solution
nπct nπct nπx
u(x,t) = An cos + Bn sin sin ,
L L L
which satisfies the original wave equation and the boundary conditions. Since the wave equation is
∞ nπct nπct nπx
u(x,t) = ∑ An cos
L
+ Bn sin
L
sin
L
n=1
is a solution as well. We assume the following properties of the solution: it converges and is
twice continuously differentiable with respect to x and t. Since each term of the series satisfies
the boundary conditions, it follows that the series satisfies these conditions. From the two initial
conditions, we can determine the constants An and Bn . By differentiating the solution with respect
to t, we obtain
∞ nπx ∞
nπc nπx
u(x, 0) = f (x) = ∑ An sin , ut (x, 0) = g(x) = ∑ Bn sin .
n=1 L n=1 L L
These equations are satisfied if f (x) and g(x) can be represented by Fourier sine series. According
to Fourier theory, the formulas for the coefficients An and Bn read
Z nπx Z L nπx
2 L 2
An = f (x) sin dx, Bn = g(x) sin dx.
L 0 L nπc 0 L
Finally, we obtain the analytical and graphical solutions of the initial-boundary value problem
as follows:
{lL=1, c=1/4, nN=10, ic1=Sin[x]-Sin[3*Pi*x]}

u[x_,t_]=Sum[(an*Cos[n*Pi*c*t/lL]+bn*Sin[n*Pi*c*t/lL])*Sin[n*Pi*x/lL],{n,1,nN}];
{u[x,t], an1=Solve[u[x,0]==0,an]//First,
eqIC=D[u[x,t],t]/.t->0//Expand}
eq1=bn*Pi*Sum[n*Sin[n*Pi*x],{n,1,nN}]//Expand
cC=Table[Select[eq1,MemberQ[#1,Sin[n*Pi*x]]&],{n,1,nN}]
iIC=Table[Select[ic1,MemberQ[#1,Sin[n*Pi*x]]&],{n,1,nN}]
bB=Table[Solve[cC[[n]]==iIC[[n]],bn]//First,{n,1,nN}]
{sol1=u[x,t]/.an1,
sol2=DeleteCases[Table[sol1[[i]]/.bB[[i]],{i,1,nN}],0]}
U[x_,t_]=sol2//First; U[x,t]
Animate[Plot[U[x,t],{x,0,lL},PlotRange->{-0.2,0.2},PlotStyle->Hue[0.7]],{t,0.1,5}]
gG=Partition[Table[Plot[Evaluate[U[x,t]],{x,0,lL},Frame->True,ImageSize->300,
PlotRange->{-0.2,0.2},DisplayFunction->Identity],{t,0.1,5,0.4}],4];
GraphicsGrid[gG]
value problem.
Consider the second-order linear parabolic PDE with the initial and boundary conditions
ut = kuxx , 0 < x < L, t > 0,

u(x, 0) = f (x), u(0,t) = 0, u(L,t) = 0,
where k = 1/30, L = 1, f (x) = sin4 (πx). This problem describes a homogeneous rod (of length L).
We assume that the rod is sufficiently thin (i.e., the heat is distributed equally over the cross-section
at time t), the surface of the rod is insulated (i.e., there is no heat loss through the boundary), and
the temperature distribution of the rod is given by the solution of the given initial-boundary value
problem.
By applying separation of variables, i.e., by seeking the exact solution in the form u(x,t) =
φ(x)ψ(t), we obtain the two ODEs
φ′′xx + λ2φ = 0, ψt′ + λ2kψ = 0.
Then, by separating the boundary conditions, we solve the eigenvalue problem and obtain the solu-
tion nπx 2
u(x,t) = An sin e−(nπ/L) kt ,
L
which satisfies the original heat equation and the boundary conditions. Since the heat equation is
∞ nπx 2 kt
u(x,t) = ∑ An sin e−(nπ/L)
n=1 L
is also a solution. This solution satisfies the initial condition if

∞ nπx
u(x, 0) = f (x) = ∑ An sin ,
n=1 L
where the coefficients An are the Fourier coefficients,

Z L nπx
2
An = f (x) sin dx.
L 0 L
Finally, by recursively computing the Fourier coefficients and the desired solution, we obtain a
formal series solution of the initial-boundary value problem and visualize it as follows:
SetOptions[Plot,ImageSize->300,PlotStyle->{Hue[0.7],Thickness[0.01]},
PlotRange->{0,1}]; f[x_]:=Sin[Pi*x]ˆ4; {lL=1,k=1/30,nN=20,nNT=5}
{aA=Table[0,{i,1,nN}], u=Table[0,{i,1,nN}]}
Do[aA[[i]]=(2./lL)*Integrate[f[x]*Sin[i*Pi*x/lL],{x,0,lL}]//N,{i,1,nN}];
aA[[1]] u[[1]]=aA[[1]]*Exp[-(Pi/lL)ˆ2*k*t]*Sin[Pi*x/lL]
Do[u[[i]]=u[[i-1]]+aA[[i]]*Exp[-(i*Pi/lL)ˆ2*k*t]*Sin[i*Pi*x/lL]//N,{i,2,nN}]
U[x_,t_]=u[[nN]]; U[x,t]
Animate[Plot[Evaluate[U[x,t]],{x,0,lL}],{t,0,nNT}]
gG=Partition[Table[Plot[Evaluate[U[x,t]],{x,0,lL},Frame->True,
DisplayFunction->Identity],{t,0,nNT,0.4}],4]; GraphicsGrid[gG]
23.3.2 Constructing Analytical Solutions via Integral Transform Methods

Integral transform methods are the most important methods for constructing analytical so-
lutions of mathematical problems (initial and/or boundary value problems) described by
linear partial differential equations.
The main idea of the methods consists in transforming the original mathematical prob-
lem to a simpler form whose solutions can be obtained and inverted (by applying inverse
integral transforms) for representing the solutions in terms of the original variables. The
formal definitions of the most important integral transforms and their general forms and
properties can be found in Debnath (2007).
In Mathematica, integral transforms (Fourier, Laplace, etc.) can be studied with the aid
of the predefined functions FourierTransform, InverseFourierTransform, LaplaceTransform,
InverseLaplaceTransform, etc. Let us solve an initial-boundary value problem for the linear
wave equation.
Example 23.23. Linear wave equation. Laplace transform. Initial-boundary value problem.
utt = c2 uxx , 0 ≤ x < ∞, t > 0,

u(x, 0) = h(x), ut (x, 0) = g(x) (0 < x < ∞), u(0,t) = a sint, u(x,t) → 0 as x → ∞,
where h(x) = 0, g(x) = 0, and a is constant (a ∈ R). This problem describes transverse vibrations
of a semiinfinite string. The initial displacement h(x) and the initial velocity g(x) are zero (i.e., the
string is at rest in its equilibrium position).
Let U(x, s) be the Laplace transform of u(x,t). By transforming the equation of motion and by
substituting the initial conditions, we obtain (eq3)
s2U = c2Uxx .
The solution of this ordinary differential equation is (sol)
U = C1 esx/c + C2 e−sx/c .
By transforming the boundary conditions, we obtain (bC2)

a
U(0, s) = , lim U(x, s) = 0,
1 + s2 x→0
According to the second condition, we have C2 = 0, and by applying the first condition, we obtain
(sol3)
a
U(x, s) = e−sx/c .
1 + s2
By applying the inverse Laplace transform, we obtain the solution (solF)
u(x,t) = a H (t − x/c)sin (t − x/c),
where H is the Heaviside unit step function. This solution represents a wave propagating at a
velocity c with the characteristic x = ct.
Finally, we obtain the displacement of the semiinfinite string as follows:
eq1=D[u[x,t],{t,2}]==cˆ2*D[u[x,t],{x,2}]
eq2=LaplaceTransform[eq1,t,s]/.{LaplaceTransform[u[x,t],t,s]->uU[x],
LaplaceTransform[D[u[x,t],{x,2}],t,s]->uU''[x]}
ic1={u[x,0]->0,(D[u[x,t],t]/.{t->0})->0}
ic2=Map[LaplaceTransform[#,t,s]&,ic1,{2}] {eq3=(eq2/.ic1),
bc1=u[0,t]==a*Sin[t]}
bc2=LaplaceTransform[bc1,t,s]/.{LaplaceTransform[u[0,t],t,s]->uU[0]}
{sol=DSolve[eq3,uU[x],x], sol1=uU[x]/.sol}
l1=Limit[sol1[[1,1]],x->Infinity,Assumptions->{c>0,s>0}]//Simplify
{sol2=sol1/.{C[1]->0}, sol3=sol2/.{C[2]->bc2[[2]]},
uU1=sol3//ExpandAll}
solF=Assuming[{s>0,c>0,x>0,t>0},InverseLaplaceTransform[uU1,s,t]]
Remark. It should be noted that if we consider the more general boundary condition u(0,t) = a f (t)
(instead of u(0,t) = a sint), then we can find the exact solution (SolF)
x x
u(x,t) = aH t − f t−
c c
in Mathematica 6, 7, 8 and we cannot find this form of solution in Mathematica 9, 10. The corre-
sponding Mathematica script is
eq1=D[u[x,t],{t,2}]==cˆ2*D[u[x,t],{x,2}]
eq2=LaplaceTransform[eq1,t,s]/.{LaplaceTransform[u[x,t],t,s]->u1[x],
LaplaceTransform[D[u[x,t],{x,2}],t,s]->u1''[x]}
ic1={u[x,0]->0,(D[u[x,t],t]/.{t->0})->0}
ic2=Map[LaplaceTransform[#,t,s]&,ic1,{2}] {eq3=(eq2/.ic1),
bc1=u[0,t]==a*f[t]}
bc2=LaplaceTransform[bc1,t,s]/.{LaplaceTransform[u[0,t],t,s]->u1[0],
LaplaceTransform[f[t],t,s]->f1[s]}
{sol=DSolve[eq3,u1[x],x], sol1=u1[x]/.sol}
{sol2=sol1/.{C[1]->0}, sol3=sol2/.{C[2]->bc2[[2]]}}
u2=sol3/.f1[s]->LaplaceTransform[f[t],t,s]
solF=InverseLaplaceTransform[u2,s,t]
23.3.3 Constructing Analytical Solutions in Terms of Green’s Functions

It is well known that the linear superposition principle is one of the most important methods
for representing solutions of linear PDEs with initial and/or boundary conditions in terms
of eigenfunctions or Green’s functions.4
As was seen earlier (Section 23.3.1), the infinite series representation of solutions of
mathematical problems (involving linear PDEs) can be obtained by applying the eigen-
function expansion method. The integral representation of solutions can be obtained by
applying the Green’s function method. Integral representations have some advantages over
infinite series representations (e.g., the description of the general analytical structure of the
solution and the evaluation of a solution).
Consider the linear Poisson equation in a volume V with surface S on which the Dirich-
let boundary conditions are imposed. The Green’s function G(X , X0 ) (X = (x, y, z) and
X0 = (x0 , y0 , z0 )) associated with the boundary value problem is a function of two variables
X (the position vector) and X0 (a fixed location) defined as the solution to
∇2 G(X , X0 ) = δ(X − X0 ) in V ; G(X , X0 ) = 0 on S.
If the volume V is the entire space, then the Green’s function is called the fundamental
solution. Since G(X , X0 ) is a symmetric function, G(X , X0 ) = G(X0 , X ) (under interchange
of the arguments), this fact can serve to verify that G(X , X0 ) is correctly calculated.
Example 23.24. Linear Laplace equation. Green’s function. Boundary value problem.
Consider the Dirichlet boundary value problem for the Laplace equation on the semiinfinite
plane V = {y > 0}:
∇2 u(x, y) = uxx + uyy = 0 in V, u(x, y) = f (x) on S,
where V = {y > 0} and S = {y = 0}. By applying the method of images, i.e., by seeking a Green’s
function G(X, X0 ) such that, in V ,
G(X, X0 ) = v(X, X0 ) + w(X, X0 ), where ∇2 v(X, X0 ) = −δ(X − X0 ) and ∇2 w(X, X0 ) = 0,
we will construct the Green’s function and find the solution of the boundary value problem. Here
X = (x, y), X0 = (x0 , y0 ), the function v(X, X0 ) is the free space Green’s function (does not depend
on the boundary conditions), the function w(X, X0 ) satisfies the Laplace equation and the boundary
4 Auxiliary functions now known as Green’s functions were first introduced by George Green in 1828 [see
Green (1828)].
conditions (and is regular at X = X0 ), i.e., ∇2 w(X, X0 ) = 0 in V and w(X, X0 ) = −v(X, X0 ) (i.e.,

G(X, X0 ) = 0) on S for the Dirichlet boundary conditions.
It is well known that the 2D free space function v(X, X0 ) is
1
v(X, X0 ) = − ln (x − x0 )2 + (y − y0)2 .
4π
If to v(X, X0 ) we add the function
1
w(X, X0 ) = ln (x − x0 )2 + (y + y0)2 ,
4π
which satisfies the Laplace equation ∇2 w(X, X0 ) = 0 in V and is regular at x = x0 and y = y0 , then
we obtain the Green’s function (g11)

1 (x − x0 )2 + (y − y0)2
G(X, X0 ) = v(X, X0 ) + w(X, X0 ) = − ln .
4π (x − x0 )2 + (y + y0)2
Then, by setting y = 0, we have G(X, X0 ) = 0 (test1). The solution of the boundary value problem
is Z
∂G
u(x0 , y0 ) = − f (x) dS.
S ∂n
∂G
By computing for the boundary y = 0, we obtain the derivative (dGn)
∂n

∂G ∂G y0
=− =−
∂n S ∂y y=0 π[(x − x0)2 + y20 ]
and the solution (solF) of the boundary value problem

Z ∞
y f (s)
u(x, y) = .
π −∞ (x − s)2 + y2
Finally, we verify that the Green’s function obtained is symmetric and visualize it as follows:
{v=-1/(4*Pi)*Log[(x-x0)ˆ2+(y-y0)ˆ2],
w=1/(4*Pi)*Log[(x-x0)ˆ2+(y+y0)ˆ2]} {z1=ExpandAll[v+w],
gG0=Simplify[z1,{z1[[1,3,1]]>0}]}
z2=Simplify[ExpandAll[gG0[[3]]*(-1)//PowerExpand],{z1[[1,3,1]]>0}]
z3=gG0[[1]]*gG0[[2]]*(-1)*z2
z33=(z3[[3,1,1]]//FullSimplify)*(z3[[3,1,2]]//FullSimplify)
{g11=z3[[1]]*z3[[2]]*Log[z33], test1=g11/.y->0,
dGn=-D[g11,y]/.y->0}
sol1=u[x0,y0]=-Integrate[dGn*f[x],{x,-Infinity,Infinity}]
solF=sol1/.x->s/.x0->x/.y0->y g2[x_,y_,x0_,y0_]=g11;
{g2[x,y,x0,y0], g2[x0,y0,x,y]}
Simplify[g2[x,y,x0,y0]-g2[x0,y0,x,y]//FullSimplify,{x>x0,y>y0}]
Simplify[g2[x,y,x0,y0]-g2[x0,y0,x,y]//FullSimplify,{x0>x,y0>y}]
Plot3D[g2[x,y,1,1],{x,-10,10},{y,0,10},PlotPoints->{100,100},
PlotRange->All,BoxRatios->{3,3,2},ViewPoint->{2,1,2}]
ContourPlot[g2[x,y,1,1],{x,-5,5},{y,0,8},PlotPoints->150,Contours->20]
Green’s functions can also be constructed by applying Laplace transforms [see Cole
et al. (2011)], the eigenfunction expansion method [see Debnath (2007), Polyanin and
Manzhirov (2007), Polyanin (2002)], and conformal mappings of the complex plane (for
solving 2D problems).
Now let us consider an extension of the theory of Green’s functions, namely, the con-
struction of modified Green’s functions and solutions of initial-boundary value problems.
Example 23.25. Linear Klein–Gordon equation. Modified Green’s function. Initial-boundary
value problem.
Consider the initial-boundary value problem for the Klein–Gordon equation with the following
initial conditions and Neumann boundary conditions:
utt = a2 uxx − bu x1 ≤ x ≤ x2 ,
u(x, 0) = f1 (x), ut (x, 0) = f2 (x), ux (0,t) = g1 (t), ux (L,t) = g2 (t),
where a and b are real parameters (a > 0, b > 0), x1 = 0, x2 = L, f1 (x)=1, f2 (x)=0, g1 (t)=1, and
g2 (t)=0. The linear Klein–Gordon equation is a special case of the equation
s(x)utt = (p(x)ux )x − q(x)u + φ(x,t),
where s(x) = 1, p(x) = a2 , q(x) = b, and Φ(x,t) = 0. By applying the eigenfunction expansion
method, we will construct the modified Green’s function. The corresponding Sturm–Liouville prob-
lem has the form
a2 φ′′xx + (λ − b)φ = 0, φ′x = 0 at x = 0, φ′x = 0 at x = L.
By solving this eigenvalue problem, we find the eigenvalues and the corresponding eigenfunctions
(eVal, eFun)
πnx πna 2
φn+1 (x) = cos , λn+1 = b + , n = 0, 1, . . . .
L L
According to Polyanin (2002),
∞
√
φn (x)φn (ξ) sin(t λn )
G(x, ξ,t) = ∑ √ ,
n=1 kφn k2 λn
Z x2
where kφn k2 = s(x)φ2n (x) dx, we construct the modified Green’s function (g1)
x1
√ ∞
p
sin(t b) 2 cos(πnx/L) cos(πnξ/L) sin(t b + (πna/L)2 )
G(x, ξ,t) = √ +∑ p ,
L b n=1 L b + (πna/L)2
i.e., a spectral representation of the Green’s function for this problem. To visualize the Green’s
function (see Fig. 23.3), we use finitely many terms of the series (gAppr).
SetOptions[Plot3D,Boxed->True,PlotPoints->{100,100},PlotRange->All,BoxRatios->{3,3,2},
ViewPoint->{2,1,2}]; SetOptions[ContourPlot,PlotPoints->100,Contours->20];
s[x_]= 1; p[x_]=aˆ2; q[x_]=b; pPhi[x_,t_]=0; f1[x_]=1; f2[x_]=0;
g1[t_]=1; g2[t_]=0;
{lL1=1,a1=1,b1=1,t1=10,x1=0,x2=lL1,Inf=29,Inf1=3}
assumps={a>0,lL>0,n\[Element] Integers,n>0}
sol1=DSolve[aˆ2*D[phi[x],{x,2}]+(lambda-b)*phi[x]==0,phi[x],x]//ExpToTrig//First
sol11=FullSimplify[sol1[[1,2,1]]+sol1[[1,2,3]],{lambda>b}]//ExpToTrig
sol12=FullSimplify[sol1[[1,2,2]]+sol1[[1,2,4]],{lambda>b}]//ExpToTrig
sol2=sol11[[1]]+sol12[[2]]/(-I) phi[x_]=sol2; {phi[x],
bc1={(D[phi[x],x]/.x->0)==0,(D[phi[x],x]/.x->lL)==0}}
1.0
0.8
0.6
0.5
0.0 0.0 0.4
- 0.5
0.5 0.2
- 1.0
0.0
0.5
1.0
0.0
1.0 0.0 0.2 0.4 0.6 0.8 1.0
Figure 23.3 The modified Green’s function G(x, ξ,t) for the initial-boundary value problem (n = 29).
{mCoef=Normal[CoefficientArrays[bc1,{C[2],C[1]}]],
mCoef[[2]]//MatrixForm} charEq=Det[mCoef[[2]]]==0//Factor
sols=Reduce[charEq && lL>0 && a>0 && b>0 && lambda>0 &&
lambda>b,lambda]
sols1=(sols[[2,2,5]]/.C[1]->0/.Pi->Pi*n)//FullSimplify
{trlambda=sols1/.Equal->Rule, mn=mCoef/.trlambda}
{mn1=Simplify[mn[[2]],assumps], ns=NullSpace[mn1]}
funs=Simplify[phi[x]/.C[1]->ns[[1,2]]/.C[2]->ns[[1,1]]/.trlambda,assumps]
{eVal[n_]=trlambda[[2]], eFun[n_,x_]=funs, eVal[n], eFun[n,x]}
nEFun=Simplify[Integrate[s[x]*eFun[n,x]ˆ2,{x,0,lL}],n\[Element]
Integers] nEFun0=Integrate[s[x]*eFun[0,x]ˆ2,{x,0,lL}]
gTerm0=eFun[0,x]*eFun[0,xi]*Sin[t*Sqrt[eVal[0]]]/Sqrt[eVal[0]]/nEFun0
gIntnd=eFun[n,x]*eFun[n,xi]*Sin[t*Sqrt[eVal[n]]]/Sqrt[eVal[n]]/nEFun
gAppr[x_,xi_,t_,lL_,a_,b_]=gTerm0+Sum[gIntnd,{n,1,Inf}];
gAppr[x,xi,t,lL1,a1,b1]
Plot3D[Evaluate[gAppr[x,xi,t1,lL1,a1,b1]],{x,0,lL1},{xi,0,lL1}]
ContourPlot[Evaluate[gAppr[x,xi,t1,lL1,a1,b1]],{x,0,lL1},{xi,0,lL1}]
Finally, by constructing the solution of the IBVP according to Polyanin (2002),

Z t Z x2 Z x2
∂
u(x,t) = Φ(ξ, τ)G(x, ξ,t − τ) dξ dτ + s(ξ) f1 (ξ)G(x, ξ,t) dξ
0 x1 ∂t x1
Z x2 Z t
+ s(ξ) f2 (ξ)G(x, ξ,t) dξ + p(x1 ) g1 (τ)(−G(x, x1 ,t − τ)) dτ
x1 0
Z t
+ p(x2 ) g2 (τ)G(x, x2 ,t − τ) dτ,
0
we obtain the solution of the given problem and visualize it by using finitely many terms of the
series (solF) for various values of t as follows:
gG1=(gTerm0+Sum[eFun[n,x]*eFun[n,xi]*Sin[t*Sqrt[eVal[n]]]/Sqrt[eVal[n]]/nEFun,
{n,1,Inf1}])/.lL->lL1/.a->a1/.b->b1
gG2[x_,xi_,t_]=gG1;
u=Integrate[Integrate[pPhi[xi,tau]*gG2[x,xi,t-tau],{xi,x1,x2}],{tau,0,t}]+
D[Integrate[s[xi]*f1[xi]*gG2[x,xi,t],{xi,x1,x2}],t]+Integrate[
s[xi]*f2[xi]*gG2[x,xi,t],{xi,x1,x2}]+p[x1]*Integrate[g1[tau]*(-gG2[x,x1,t-tau]),
{tau,0,t}]+p[x2]*Integrate[g2[tau]*gG2[x,x2,t-tau],{tau,0,t}]
solF[x_,t_]=Collect[u/.a->a1,{Cos[_]}]
Plot3D[Evaluate[solF[x,t]],{x,0,lL1},{t,0,t1}]
ContourPlot[Evaluate[solF[x,t]],{x,0,lL1},{t,0,t1}]
gs=Table[g[i]=Plot[Evaluate[solF[x,i]],{x,0,lL1},Frame->False,AspectRatio->1,
PlotStyle->{Blue,Thickness[0.005]},PlotPoints->100],{i,0,t1}];
Show[gs,PlotRange->All]
Plot[{solF[x,1],solF[x,3],solF[x,5],solF[x,t1]},{x,0,lL1},
PlotStyle->{{Blue,Thickness[0.008]},{Blue,Dotted,Thickness[0.005]},
{Blue,Dashed,Thickness[0.009]},{Blue,DotDashed,Thickness[0.008]}},
PlotPoints->100,PlotLegends->Placed[{"u[x,1]","u[x,3]","u[x,5]","u[x,10]"},Below]]
⊙ References for Section 23.3: G. Green (1828), V. A. Galaktionov (1990, 1995), A. D. Polyanin and
A. I. Zhurov (1998), A. D. Polyanin (2002), A. D. Polyanin and A. V. Manzhirov (2007), L. Debnath (2007),
K. D. Cole, J. V. Beck, A. Haji-Sheikh, and B. Litkouhi (2011).
23.4 Numerical Solutions and Their Visualizations
It is well known that there are numerous linear partial differential equations (arising from
modeling of real-world problems) and associated mathematical problems (even with simple
boundary and/or initial conditions) that cannot be solved analytically. Presently, it is neces-
sary to develop and apply approximation methods. Nowadays, approximation methods are
becoming very important and useful in applications due to the increasing development of
modern computers, supercomputers, and computer algebra systems. In this section, some
of the most important approximation approaches to the solution of linear partial differential
equations and associated mathematical problems are discussed with the aid of the computer
algebra system Mathematica.
23.4.1 Constructing Numerical Solutions in Terms of Predefined Functions
First, consider the predefined Mathematica functions with the aid of which we can obtain
approximate numerical solutions when solving various linear time-based PDE problems.
With the aid of the predefined function NDSolve, it is possible to obtain approximate nu-
merical solutions of various linear PDE problems (initial-boundary value problems). Note
that in Mathematica (Release ≤ 10) it is only possible to solve evolution equations numer-
ically by using NDSolve. It is possible to impose Dirichlet, Neumann, Robin, or periodic
boundary conditions.
NDSolve[{PDEs,IC,BC}, DepVars, IndVars, Ops] NDSolve[{PDE,IC,BC},

u, {x,x1,x2},{t,t1,t2},...] NDSolve[{PDEs,IC,BC}, {u1,...,un},
{x,x1,x2},{t,t1,t2},...]
NDSolve, finding numerical solutions to PDE problems (initial-boundary value problems),

where DepVars and IndVars are the dependent and independent variables, respec-
tively.
Now we schematically write out the numerical solution sol with the aid of NDSolve and
then use it to obtain various visualizations (e.g., Plot, Plot3D, Animate) and numerical
values (numVals).
sol=NDSolve[{PDE,IC,BC},u,{x,x1,x2},{t,t1,t2},ops]
Plot[Evaluate[u[x,tk]/.sol],{x,x1,x2},ops]
numVals=Evaluate[u[xk,tk]/.sol]
Plot3D[Evaluate[u[x,t]/.sol],{x,x1,x2},{t,t1,t2},ops]
Animate[Plot[Evaluate[u[x,t]/.sol],{x,x1,x2},ops],{t,t1,t2},ops]
Example 23.26. Linear wave equation. Initial-boundary value problem. Numerical, graphical,
and exact solutions. We find numerical, graphical, and exact solutions of the following initial-
boundary value problem for the linear wave equation:
utt − c2 uxx = 0, 0 < x < L, t > 0,

u(x, 0) = f1 (x), ut (x, 0) = f2 (x), u(0,t) = g1 (t), u(L,t) = g2 (t),
where c = 1/10, L = 1, f1 (x) = A sin(Bπx), f2 (x) = 0, g1 (t) = 0, g2 (t) = 0, A = 3, B = 2. First, we

construct the numerical solution of this initial-boundary value problem and visualize it as follows:
<<DifferentialEquations`InterpolatingFunctionAnatomy`
{aA=3,bB=2,lL=1,tF=1,s=1/100,nP=100,nN=3,lL1={Red,Blue,Green},lL2={1/8,3/8,5/8}}
SetOptions[Plot,ImageSize->500,PlotRange->{All,{-5,5}},PlotPoints->nP*2,
PlotStyle->{Blue,Thickness[0.01]}]; SetOptions[Plot3D,ImageSize->500,
PlotRange->All]; c=N[1/10]; f1[x_]=aA*Sin[bB*Pi*x]; f2[x_]=0; g1[t_]=0; g2[t]=0;
pde1=D[u[x,t],{t,2}]-cˆ2*D[u[x,t],{x,2}]==0
ic1={u[x,0]==f1[x],(D[u[x,t],t]/.t->0)==f2[x]}
bc1={u[0,t]==g1[t],u[lL,t]==g2[t]}
sol1=NDSolve[Flatten[{pde1,ic1,bc1}],u,{x,0,lL},{t,0,tF},MaxStepSize->s,
PrecisionGoal->2]; f1=u/.First[sol1];
Map[Length,InterpolatingFunctionCoordinates[f1]]
Do[g[i]=Plot[Evaluate[u[x,lL2[[i]]]/.sol1],{x,0,lL},PlotStyle->{lL1[[i]],
Thickness[0.01]}],{i,1,nN}]; Show[Table[g[i],{i,1,nN}]]
numVals=Evaluate[u[0.1,1/2]/.sol1]; Print[numVals];
Plot3D[Evaluate[u[x,t]/.sol1],{x,0,lL},{t,0,tF},ColorFunction->Function[
{x,y},Hue[x]],ViewPoint->{-1,2,1},ImageSize->500]
ContourPlot[Evaluate[u[x,t]/.sol1],{x,0,lL},{t,0,tF},ColorFunction->Hue,
ImageSize->300]
Animate[Plot[Evaluate[u[x,t]/.sol1,{x,0,lL}],PlotRange->{-3,3}],{t,0,tF,0.001},
AnimationRate->0.5]
Knowing the exact solution, u(x,t) = A sin(Bπ x) cos(cπt) (for details, see the previous chapter)
of this initial-boundary value problem, we can compare the numerical and exact solutions as follows:
uEx[x_,t_]=aA*Sin[bB*Pi*x]*Cos[c*Pi*t]; {trt=t->3/8,trx=x->1/2}
p1=Plot[Evaluate[u[x,t]/.sol1]/.trt,{x,0,lL},PlotStyle->{lL1[[1]],Thickness[0.001]},
ImageSize->500]; p2=Plot[Evaluate[uEx[x,lL2[[2]]]],{x,0,lL},PlotStyle->{lL1[[2]],
Thickness[0.001]},ImageSize->500]; Show[{p1,p2}]
numVals=Evaluate[u[1/2,3/8]/.sol1]; Print[numVals];
N[uEx[1/2.,3/8.],16]
Do[p[i+2]=Plot[Evaluate[(u[x,t]/.sol1/.t->lL2[[i]])-uEx[x,lL2[[i]]]],{x,0,lL},
PlotStyle->{lL1[[i]],Thickness[0.001]},PlotRange->All,ImageSize->500],{i,1,nN}];
Show[Table[p[i+2],{i,1,nN}]]
23.4.2 Numerical Methods Embedded in Mathematica

In Mathematica, approximate numerical solutions of various linear PDE problems (initial-
boundary value problems) can be obtained with the aid of the predefined function NDSolve,
which implements the method of lines. In addition, it is possible to explicitly specify
the method of lines and appropriate suboptions for this method for solving mathemati-
cal problems, i.e., the option Method, which can be written in general form as follows:
Method->{MethodName, MethodOptions}. For instance, we can specify MethodOfLines
and other suboptions (e.g., MinPoints for a smaller grid spacing, etc.).
NDSolve[{PDE,IC,BC},u,{x,x1,x2},{t,t1,t2},Method->{m,subOps}]
Options[NDSolve‘MethodOfLines]
NDSolve,Method finding numerical solutions to PDE problems by the method of lines

with some specific suboptions subOps.
Method The option Method and the most important suboptions:
Method->{"MethodOfLines","SpatialDiscretization"->{"TensorProductGrid",
"MinPoints"->val,"MaxPoints"->val,"MaxStepSize"->val,"PrecisionGoal"->val,
"DifferenceOrder"->val}}
By applying the method of lines, it is possible to solve restricted classes of mathemat-

ical problems: problems which require a temporal variable with an initial condition (in
at least one independent variable). For example, the method of lines cannot solve ellip-
tic equations. For numerous initial-boundary value problems, the method of lines is quite
effective, fast, and general (for more details, see the Documentation Center, “Numerical
Solution of Partial Differential Equations”).
Example 23.27. Linear heat equation. Initial-boundary value problem. Numerical, graphical,
and exact solutions. Method of Lines. Consider the following initial-boundary value problem for
the linear one-dimensional heat equation:
ut = k uxx , 0 < x < L, t > 0, u(x, 0) = f (x), u(0,t) = g1 (t), u(L,t) = g2 (t),
where L = 1, k = 0.1, f (x) = x(1 − x), g1 (t) = 0, and g2 (t) = 0. Constructing the numerical and
graphical solutions of this problem, we explicitly specify the method of lines with several suboptions
as follows:
f[x_]=x*(1-x); g1[t_]=0; g2[t_]=0;

{k=0.1,sS=1/300.,nP=100,nNF=4,lL=1,lL1={Red,Blue},lL2={0,0.5},nN=2}
SetOptions[Plot,ImageSize->500,PlotRange->All,PlotPoints->nP*2,
PlotStyle->{Blue,Thickness[0.01]}];
{pde1=D[u[x,t],t]-k*D[u[x,t],{x,2}]==0,
ic1={u[x,0]==f[x]}, bc1={u[0,t]==g1[t],u[lL,t]==g2[t]}}
sol1=NDSolve[{pde1,ic1,bc1},u,{x,0,lL},{t,0,nNF},AccuracyGoal->1,
PrecisionGoal->1,MaxStepSize->sS,MaxSteps->{300,Infinity},
Method->{"MethodOfLines","SpatialDiscretization"->{"TensorProductGrid"}}];
Do[g[i]=Plot[Evaluate[u[x,lL2[[i]]]/.sol1],{x,0,lL},PlotStyle->{lL1[[i]],
Thickness[0.01]}],{i,1,nN}]; Show[Table[g[i],{i,1,nN}]]
numVals1=Evaluate[u[1/2,Pi]/.sol1]; Print[numVals1];
Plot3D[Evaluate[u[x,t]/.sol1],{x,0,lL},{t,0,nNF},ColorFunction->Function[
{x,t},Hue[xˆ2+tˆ2]],BoxRatios->1,ViewPoint->{-1,2,1},ImageSize->300]
ContourPlot[Evaluate[u[x,t]/.sol1],{x,0,lL},{t,0,nNF},
ColorFunction->Hue,ImageSize->300]
Animate[Plot[Evaluate[u[x,t]/.sol1,{x,0,lL}],PlotRange->{0,0.3}],
{t,0,nNF},AnimationRate->0.5]
e2x+0.1t − e0.1t
Finally, knowing the exact solution, u(x,t) = x(1 − x) (for details, see the pre-
e2x − 1
vious chapter) of the initial-boundary value problem, we can compare the numerical and exact
solutions and analyze the corresponding errors between them as follows:
uEx[x_,t_]=x*(1-x)*(Exp[2*x+0.1*t]-Exp[0.1*t])/(Exp[2*x]-1);
{trt=t->0.1, trx=x->0.5, lL3={0.2,0.4}}
p1=Plot[Evaluate[u[x,t]/.sol1]/.trt,{x,0,lL},PlotStyle->{lL1[[1]],
Thickness[0.001]},ImageSize->500]; p2=Plot[Evaluate[uEx[x,lL2[[2]]]],{x,0,lL},
PlotStyle->{lL1[[2]],Thickness[0.001]},ImageSize->500]; Show[{p1,p2}]
numVals=Evaluate[u[0.5,0.1]/.sol1]; Print[numVals];
N[uEx[0.5,0.1],16]
Do[p[i+2]=Plot[Evaluate[(u[x,t]/.sol1/.t->lL3[[i]])-uEx[x,lL3[[i]]]],{x,0,lL},
PlotStyle->{lL1[[i]],Thickness[0.001]},PlotRange->All,ImageSize->500],{i,1,nN}];
Show[Table[p[i+2],{i,1,nN}]]
PaddedForm[Table[Abs[N[(u[x,t]/.sol1/.trt/.x->i)-uEx[i,0.1],16]],
{i,0.01,lL,0.1}]//TableForm,{16,10}]

Now let us show the helpful role of computer algebra systems for generating and apply-
ing various finite difference approximations for constructing numerical solutions of linear
PDEs.
To approximate linear PDEs by finite differences, we have to generate a mesh (or grid)
in a domain D, e.g., D = {a < x < b, c < t < d}. The mesh can be of various types, e.g.,
rectangular, along the characteristics, polar, etc. We assume (for simplicity) that the sets of
lines of the mesh are equally spaced and the dependent variable in a given PDE is u(x,t).
We write h and k for the line spacings and define the mesh points as follows: Xi = a+ ih,
T j = c + jk (i = 0, . . . , NX , j = 0, . . . , NT ) and h = (b − a)/NX , k = (d − c)/NT . We
calculate approximations to the solution at these mesh points; these approximate points
will be denoted by Ui, j ≈ u(Xi , T j ). We approximate the derivatives in a given equation by
Example 23.28. Linear heat equation. Initial-boundary value problem. Forward/backward FD
methods. Crank–Nicolson method. Consider the following initial-boundary value problem for the
linear heat equation:
ut = ν uxx , 0 < x < L, t > 0, u(x, 0) = f (x), u(0,t) = 0, u(L,t) = 0,
where f (x)= sin(πx/L), L = 1, ν = 1. By applying the forward/backward difference methods and
the Crank–Nicolson method, we construct an approximate numerical solution of the given initial-
We generate the rectangular mesh X = ih, T = jk (i = 0, . . ., NX, j = 0, . . ., NT , h = L/NX,
k = T /NT ). We denote the approximate solution of u(x,t) at the mesh point (i, j) by Ui, j .
In the forward difference method, the second derivative uxx is replaced by a central difference
approximation (CDA) and the first derivative ut by a forward difference approximation (FWDA).
Ui, j+1 = (1 − 2r)Ui, j + r(Ui+1, j + Ui−1, j ),
where r=νk/h2 . In this explicit FD scheme, the unknown value Ui, j+1 (at the ( j+1)th step) is
determined from the three known values Ui−1, j , Ui, j , and Ui+1, j (at the jth step). This FD scheme is
unstable for r>0.5. We find the numerical solution as follows:
SetOptions[ListPlot,ImageSize->300,PlotRange->{{0,1},{0,1}},Joined->True];
f[x_]:=N[Sin[Pi*x]];
{nu=1,nNX=15,nNT=100,lL=1,tT=0.2,h=lL/nNX,k=tT/nNT,r=nu*k/hˆ2}
Table[xX[i]=i*h,{i,0,nNX}]; ic=Table[uU[i,0]->f[xX[i]],{i,0,nNX}];
bc={Table[uU[0,j]->0,{j,0,nNT}],Table[uU[nNX,j]->0,{j,0,nNT}]};
ibc=Flatten[{ic,bc}]
fd[i_,j_]:=(1-2*r)*uU[i,j]+r*(uU[i+1,j]+uU[i-1,j]);
Do[uU[i,j+1]=fd[i,j]/.ibc,{j,0,nNT},{i,1,nNX-1}];
g[j_]:=ListPlot[Table[{xX[i],uU[i,j]/.ibc},{i,0,nNX}],
PlotStyle->{Blue,Thickness[0.01]},AxesLabel->{"X","U"}];
grs=Evaluate[Table[g[j],{j,0,nNT}]]; ListAnimate[grs]
This FD scheme can be represented in the matrix form
Ui =MUi−1 ,
where U0 =( f (X1 ), . . . , f (XNX−1 )), and M is an NX×NX tridiagonal band matrix (with 1−2r along
the main diagonal, r along the subdiagonal and superdiagonal, and zero everywhere else). We obtain
the numerical solution by using this matrix representation of the FD scheme as follows:
SetOptions[ListPlot,ImageSize->500,PlotStyle->{Blue,Thickness[0.01]},
PlotRange->{{0,1},{0,1}},Joined->True]; f[x_]:=N[Sin[Pi*x]];
{nu=1,nNX=40,nNT=800,lL=1,tT=0.2,h=lL/nNX,k=tT/nNT,r=nu*k/hˆ2,nNG=90}
mat=SparseArray[{Band[{2,1}]->r,Band[{1,1}]->1-2*r,Band[{1,2}]->r},{nNX-1,nNX-1}];
Print[MatrixForm[mat]]; uU[0]=Table[f[i*h],{i,1,nNX-1}]
Do[uU[k]=mat.uU[k-1]; gr[k]={{0,0}};
Do[gr[k]=Append[gr[k],{i/nNX,uU[k][[i]]}],{i,1,nNX-1}];
gr[k]=Append[gr[k],{lL,0}]; g[k]=ListPlot[gr[k]],{k,1,nNG}];
grs=Evaluate[Table[g[j],{j,1,nNG}]]; ListAnimate[grs]
In the backward difference method, the second derivative uxx is replaced by a central difference
approximation (CDA) and the first derivative ut by a backward difference approximation (BWDA).
The final FD scheme for the linear diffusion equation is
(1 + 2r)Ui, j − r(Ui+1, j + Ui−1, j ) − Ui, j−1 = 0,
where r = νk/h2 . In this implicit FD scheme, we have to solve these difference equations numeri-
cally at each of the internal mesh points at each jth step (where j = 1, . . . , NT ) with the initial and
boundary conditions. This FD scheme is unconditionally stable. We calculate the approximate nu-
merical solution of the initial-boundary value problem by applying the backward difference method
as follows:
SetOptions[ListPlot,PlotRange->{{0,lL},{0,lL}},Joined->True];
{nu=1,nNX=50,nNT=50,lL=1,tT=0.2,h=lL/nNX,k=tT/nNT,r=nu*k/hˆ2}
Table[xX[i]=i*h,{i,0,nNX}]; f[i_]:=N[Sin[Pi*xX[i]]];
ibc={Table[uU1[i,0]->f[i],{i,0,nNX}],Table[uU1[0,j]->0,{j,0,nNT}],
Table[uU1[nNX,j]->0,{j,0,nNT}]}//Flatten
sol[0]=ibc;
fd[i_,j_]:=(1+2*r)*uU1[i,j]-r*(uU1[i+1,j]+uU1[i-1,j])-uU1[i,j-1];
Do[eqs[j]=Table[Expand[fd[i,j]]==0,{i,1,nNX-1}];
eqs1[j]=eqs[j]/.sol[j-1]/.ibc; vars[j]=Table[uU1[i,j],{i,1,nNX-1}];
sol[j]=NSolve[eqs1[j],vars[j]],{j,1,nNT}];
g[j_]:=ListPlot[Table[{xX[i],uU1[i,j]/.Flatten[sol[j]]/.ibc},{i,0,nNX}],
grs=Evaluate[Table[g[j],{j,1,nNT}]]; ListAnimate[grs]
The Crank–Nicolson method is obtained by centered differencing in time about the point
(xi ,t j+1/2 ). To do so, we average the central difference approximations in space at time t j and
t j+1 . The final FD scheme for the linear heat equation is
−rUi−1, j+1 + 2(1+r)Ui, j+1 − rUi+1, j+1 = rUi−1, j + 2(1 − r)Ui, j + rUi+1, j ,
where r=νk/h2 . In this FD scheme, we have three unknown values of U at the ( j+1)th time step
and three known values at the jth time step. This FD scheme is unconditionally stable. We calculate
the approximate numerical solution of the initial-boundary value problem by applying the Crank–
Nicolson method [see Crank and Nicolson (1947)] as follows:
fF[i_]:=Sin[Pi*i]; {lL=1,tT=0.2,nu=1,nNX=20,nNT=20,nNX1=nNX-1,
nNX2=nNX-2,h=lL/nNX,k=tT/nNT,r=nu*k/(hˆ2),tk=0.2}
{ic={v[x1,0]==fF[x1]}, bc={v[0,t1]==0,v[lL,t1]==0}}
{lLM=Table[0,{i,0,nNX}],uU=Table[0,{i,0,nNX}],uUM=Table[0,{i,0,nNX}],
zZ=Table[0,{i,0,nNX}], uU[[nNX-1]]=0}
pde1=D[v[x1,t1],t1]-nu*D[v[x1,t1],{x1,2}]==0
Do[uU[[i-1]]=N[fF[i*h]],{i,1,nNX1}];
{lLM[[0]]=1+r,uUM[[0]]=-r/(2*lLM[[0]])}
Do[lLM[[i-1]]=1+r+r*uUM[[i-2]]/2;
uUM[[i-1]]=-r/(2*lLM[[i-1]]),{i,2,nNX2}];
lLM[[nNX1-1]]=1+r+0.5*r*uUM[[nNX2-1]] Do[t=j*k;
zZ[[0]]=((1-r)*uU[[0]]+r*uU[[1]]/2)/lLM[[0]];
Do[zZ[[i-1]]=((1-r)*uU[[i-1]]+0.5*r*(uU[[i]]+uU[[i-2]]+zZ[[i-2]]))/lLM[[i-1]],
{i,2,nNX1}]; uU[[nNX1-1]]=zZ[[nNX1-1]];
Do[i=nNX2-i1+1;
uU[[i-1]]=zZ[[i-1]]-uUM[[i-1]]*uU[[i]],{i1,1,nNX2}],{j,1,nNT}];
Finally, we compare the approximate numerical solution with the exact solution of this problem
u(x,t) = exp(−π2t) sin(πx) at (xk ,tk ), 0 < x < L, t > 0,
as follows:
nD=10; extSol[x1_,t1_]:=Exp[-Piˆ2*t]*Sin[Pi*x]/.{x->x1,t->t1};
Print["Crank-Nicolson Method"]; Do[xX=i*h; Print[i,"
",PaddedForm[N[xX,nD],{12,10}]," ",
PaddedForm[uU[[i-1]],{12,10}]," ", PaddedForm[N[extSol[xX,tk],nD],{12,10}]," ",
PaddedForm[uU[[i-1]]-N[extSol[xX,tk],nD],{12,10}]],{i,1,nNX1}];
Example 23.29. Linear wave equation. Explicit difference methods. Consider the initial-
boundary value problem for the linear wave equation describing the motion of a fixed string:
utt = c2 uxx , 0 < x < L, t > 0, u(x, 0) = f1 (x), ut (x, 0) = f2 (x), u(0,t) = 0, u(L,t) = 0,
where f1 (x) = 0, f2 (x) = sin(4πx), L = 0.5, and c = 1/(4π). By applying the explicit central finite
difference method, we construct the approximate numerical solution of the initial-boundary value
problem.
In the explicit central difference method, each second derivative is replaced by a central differ-
ence approximation. The FD scheme for the linear wave equation is
Ui, j+1 = 2(1 − r)Ui, j + r(Ui+1, j + Ui−1, j ) − Ui, j−1,
where r = (ck/h)2 . In this FD scheme, we have one unknown value Ui, j+1 that depends explicitly
on the four known values Ui, j , Ui+1, j , Ui−1, j , Ui, j−1 at the previous time steps ( j and j − 1). To start
the process, we have to know the values of U at the time steps j = 0 and j = 1. So we can define the
initial conditions at these time steps: Ui,0 = f 1(Xi ) and U(Xi , 0)t ≈ (Ui,1 −Ui,0 )/k = f 2(Xi ), Ui,1 =
f 1(Xi ) + k f 2(Xi ). This FD scheme is stable for r ≤ 1. We find the approximate numerical solution
of the initial-boundary value problem by applying the explicit central finite difference method as
follows:
SetOptions[ListPlot,PlotRange->All,Joined->False]; fF1[x_]:=0;
fF2[x_]:=Sin[4*Pi*x]; {c=N[1/(4*Pi)],lL=0.5,tT=1.5,nNX=40,
nNX1=nNX+1,nNT=40,nNT1=nNT+1,h=lL/nNX,k=tT/nNT,r=N[c*k/h]}
fF1i[i_]:=fF1[h*(i-1)]; fF2i[i_]:=fF2[h*(i-1)];
uU=Table[0,{nNT1},{nNX1}]; For[i=1,i<=nNT1,i++,uU[[i,1]]=fF1i[i]];
For[i=2,i<=nNT,i++,uU[[i,2]]=(1-rˆ2)*fF1i[i]+rˆ2*(fF1i[i+1]
+fF1i[i-1])/2+k*fF2i[i]];
For[j=3,j<=nNX1,j++, For[i=2,i<=nNT,i++,
uU[[i,j]]=2*(1-rˆ2)*uU[[i,j-1]]+rˆ2*(uU[[i+1,j-1]]+uU[[i-1,j-1]])
-uU[[i,j-2]]//N];];
Print[" i xX[i] uU[xX[i],nNT1]", "\n"];
For[i=1,i<=nNT1,i++,Print[PaddedForm[i,2],PaddedForm[h*(i-1),7]," ",
PaddedForm[uU[[i,nNT1]],10]]];
points=Table[{h*(i-1),uU[[i,nNT1]]},{i,1,nNT1}]
ListPlot[points,PlotStyle->{Blue,PointSize[0.02]}]
Print[NumberForm[TableForm[Transpose[Chop[uU]]],3]];
ListPlot3D[uU,ViewPoint->{3,1,3},ColorFunction->Hue]
We construct and visualize the same approximate numerical solution of the initial-boundary
value problem by applying the explicit central finite difference method and by following a different
style of programming as follows:
SetOptions[ListPlot,ImageSize->500,PlotRange->{{0,0.5},{-1,1}},Joined->True];
fF1[x_]:=0; fF2[x_]:=N[Sin[4*Pi*x]];
{c=N[1/(4*Pi)],lL=0.5,tT=1.5,nNX=40,nNT=40,h=lL/nNX,k=tT/nNT,r=(c*k/h)ˆ2}
ic={Table[uU1[i,0]->fF1[i*h],{i,1,nNX-1}],
Table[uU1[i,1]->fF1[i*h]+k*fF2[i*h],{i,1,nNX-1}]};
bc={Table[uU1[0,j]->0,{j,0,nNT}],Table[uU1[nNX,j]->0,{j,0,nNT}]};
ibc=Flatten[{ic,bc}]
fd[i_,j_]:=2*(1-r)*uU1[i,j]+r*(uU1[i+1,j]+uU1[i-1,j])-uU1[i,j-1];
Do[uU1[i,j+1]=fd[i,j]/.ibc,{j,1,nNT-1},{i,1,nNX-1}];
g[j_]:=ListPlot[Table[{i*h,uU1[i,j]/.ibc},{i,0,nNX}],
grs=N[Table[g[j],{j,0,nNT}]]; ListAnimate[grs]
Since in Mathematica (Release ≤ 10) it is only possible to solve evolution equations nu-
merically (via predefined functions), we apply finite difference methods and the method of
lines for constructing numerical and graphical solutions of boundary value problems for
elliptic equations, e.g., the linear Poisson equation.
Example 23.30. Linear Poisson equation. Boundary value problem. Central difference scheme.
Consider the two-dimensional linear Poisson equation
uxx + uyy = f (x, y), D = {a ≤ x ≤ b, c ≤ y ≤ d}
object. Let us choose f (x, y) = cos x sin y, f1 (x) = f2 (x) = 0, f3 (y) = f4 (y) = − 12 sin y, a = 0, b = π,
c = 0, and d = 2π. For the linear Poisson equation, by applying the explicit finite difference scheme,
we obtain the approximate numerical solution of the boundary value problem, visualize it in D, and
compare with the exact solution u(x, y) = − 12 cos x sin y (see the previous chapter).
Let us generate the rectangular mesh x = a + ih, y = c + jk (where i = 0, . . ., NX, j = 0, . . ., NY ,
h = (b − a)/NX, and k = (d − c)/NY ). We denote the approximate solution of u(x, y) at the mesh
point (i, j) by Ui, j . The second derivatives in the Poisson equation are replaced by the central
difference approximation (CDA). The FD scheme takes the form
2(1 + r)Ui, j − Ui+1, j − Ui−1, j − rUi, j+1 − rUi, j−1 = cos(ih) sin( jk),
where r = (h/k)2 .
We construct an approximate numerical solution of the boundary value problem by applying the
above explicit finite-difference scheme and plot the numerical solution inside the domain as follows:
SetAttributes[{x,y},NHoldAll]; {nD=10,a=0.,b=N[Pi],c=0.,d=N[2.*Pi],
nNX=20,nNY=260,h=N[(b-a)/nNX],k=N[(d-c)/nNY],r=N[(h/k)ˆ2]}
xX=Table[x[i]->a+i*h,{i,0,nNX}]//N;
yY=Table[y[j]->c+j*k,{j,0,nNY}]//N;
fd[i_,j_]=2.*(1+r)*uU[i,j]-uU[i+1,j]-uU[i-1,j]-r*uU[i,j+1]-r*uU[i,j-1]
-Cos[i*h]*Sin[j*k]; f1[i_]=0.; f2[i_]=0.; f3[j_]=N[-0.5*Sin[j*k]];
f4[j_] = N[-0.5*Sin[j*k]];
bc=Flatten[{Table[uU[i,0]->f1[i],{i,0,nNX}],Table[uU[i,nNY]->f2[i],
{i,0,nNX}],Table[uU[0,j]->f3[j],{j,0,nNY}],Table[uU[nNX,j]->f4[j],
{j,0,nNY}]}]; eqs=Flatten[Table[fd[i,j]==0.,{i,1,nNX-1},{j,1,nNY-1}]];
eqs1=Flatten[eqs/.bc]; vars=Flatten[Table[uU[i,j],{i,1,nNX-1},{j,1,nNY-1}]];
sol=NSolve[eqs1,vars,nD];
points=Table[{x[i],y[j],uU[i,j]},{i,0,nNX},{j,0,nNY}];
points1=Flatten[N[points/.xX/.yY/.bc/.sol[[1]],nD],1];
g3D=ListPointPlot3D[points1,PlotStyle->PointSize[0.01],BoxRatios->{1,1,1},
PlotStyle->{{PointSize[0.05]},{PointSize[0.05]}},PlotRange->All];
gCP=ListContourPlot[points1]; GraphicsRow[{g3D,gCP},ImageSize->500]
uEx[x_,y_]=-0.5*Cos[x]*Sin[y]
g1=Plot3D[uEx[x,y],{x,0.,Pi},{y,0.,2.*Pi},PlotRange->All,BoxRatios->{3,3,2}];
g2=ContourPlot[uEx[x,y],{x,0.,Pi},{y,0.,2.*Pi},PlotPoints->100,Contours->20];
GraphicsRow[{g1,g2},ImageSize->500]
We restate our problem by following the approach proposed by Silebi and Schiesser (1992), i.e.,
by adding the time derivative to the steady-state elliptic partial differential equation and by applying
the method of lines and the predefined function NDSolve. Finally, we obtain the steady-state solution
by increasing the time interval (e.g., tT = 50) as follows:
f1[x_]:=0.; f2[x_]:=0.; f3[y_]:=-N[0.5*Sin[y]];

f4[y_]:=-N[0.5*Sin[y]]; {a=0.,b=N[Pi],c=0.,d=N[2.*Pi],tT=50}
{pde1=D[u[x,y,t],t]==D[u[x,y,t],{x,2}]+D[u[x,y,t],{y,2}]-Cos[x]*Sin[y],
ic={u[x,y,0]==-N[0.5*Sin[y]]}, bc={u[x,c,t]==f1[x],u[x,d,t]==f2[x],
u[a,y,t]==f3[y],u[b,y,t]==f4[y]}}
sol1=NDSolve[{pde1,ic,bc},u,{x,a,b},{y,c,d},{t,0,tT},Method->
{"MethodOfLines","SpatialDiscretization"->{"TensorProductGrid",
"DifferenceOrder"->"Pseudospectral"}}]
g1=Plot3D[Evaluate[u[x,y,tT]/.sol1],{x,a,b},{y,c,d},PlotRange->All];
g2=ContourPlot[Evaluate[u[x,y,tT]/.sol1],{x,a,b},{y,c,d}];
GraphicsRow[{g1,g2},ImageSize->500]

There are various methods to obtain approximate solutions of Cauchy problems involv-
ing linear partial differential equations. The most convenient and accurate method for
solving Cauchy problems for hyperbolic equations is the method of characteristics (see
Sections 23.2.2, 23.2.4). The finite difference methods for hyperbolic equations are not
very convenient (e.g., in the case of discontinuous initial data) [see Richtmyer and Morton
(1994), Collatz (1966)].
Example 23.31. Linear inhomogeneous advection equation. Cauchy problem. Method of char-
acteristics. Graphical solutions. Consider the initial value problem for the linear inhomogeneous
advection equation
ut + c ux = F(x,t), −∞ < x < ∞, t > 0, u(x, 0) = f (x),
where f (x) = sin x and F(x,t) = xt.

By applying the method of characteristics, we obtain the initial value ODE problem and its
solution (odes, sol1):
dx(r) c dt(r) 1 dz(r) x(r)t(r)

= , = , = , x(1) = s, t(1) = 0, z(1) = h(s),
r r r r r r
1 1
x(r) = c ln r + s, t(r) = ln r, z(r) = c ln r3 + ln r2 s + h(s).
3 2
Finding z (in terms of x and t), we have (sol2)
1 1
s = x − ct, r = et , z = − ct 3 + t 2 x + h(x − ct),
6 2
and the solution u(x,t) = − 16 ct 3 + 12 t 2 x + h(x − ct) satisfies the original PDE and the initial data
(test1, test2). Finally, we visualize the solution of the given Cauchy problem (e.g., for c = 2 and
h(s) = sin s) as follows:
{trVars={x[r]->x,t[r]->t,z[r]->z},
pde1=D[u[x,t],t]+c*D[u[x,t],x]==x*t} odes={D[x[r],r]==c/r,
D[t[r],r]==1/r, D[z[r],r]==x[r]*t[r]/r} sol1=DSolve[{odes,x[1]==s,

t[1]==0, z[1]==h[s]},{x[r],t[r],z[r]},r]//First
sol2=Reduce[(sol1/.trVars)/.Rule->Equal,{s,r}]
{sol2[[2]],sol2[[3]],sol2[[4]],trPar=c->2} w[x_,t_]=sol2[[2,2]];
w[x, t] {test1=Equal[D[w[x,t],t]+c*D[w[x,t],x]==x*t],
test2=Equal[w[s,0]==h[s]]} h[s_]=Sin[s]; h[s]
Plot3D[Evaluate[w[x,t]/.trPar],{x,-3*Pi,3*Pi},{t,0,1},Mesh->False,ColorFunction->h[x],
PlotRange->All,BoxRatios->{1,1,1},PlotStyle->{{PointSize[0.05]},{PointSize[0.05]}}]
ContourPlot[Evaluate[w[x,t]/.trPar],{x,-3*Pi,3*Pi},{t,0,1},PlotPoints->100,
Contours->20,PlotRange->All]
23.4.6 Numerical Solutions of Systems of Linear PDEs
In this section, we show how to obtain numerical and graphical solutions of systems of
linear partial differential equations in Mathematica with the aid of the predefined function
NDSolve.
Example 23.32. Linear first-order system (equivalent to the linear wave equations). Numerical
and graphical solutions. Introducing the dependent variables u(x,t) and v(x,t) according to the
formulas u = wt and v = c2 wx , we can transform the linear wave equation, wtt = c2 wxx , to the
equivalent system of linear first-order equations ut = vx , vt = c2 ux . Consider the initial-boundary
value problem for this linear first-order system (describing standing waves):
ut = vx , vt = c2 ux , 0 < x < L, t > 0,

u(x, 0) = f (x), v(x, 0) = 0, u(0,t) = 0, u(L,t) = 0,
where L = 2π, f (x) = sin x.

We obtain a numerical solution (sol1, w, q) of the given initial-boundary value problem, visual-
ize it at various times (g1–g3, gu, gv) in parametric form, construct the animation of a standing wave
profile, and visualize the solutions in 3D as follows:
{c=0.5,lL=N[2*Pi],a=0,b=lL,tT=20,lL1={0.1,0.2,0.5},lL2={Magenta,Blue,Red},
nNL1=Length[lL1]}
f[x_]=Sin[x];
sys1={D[u[x,t],t]==D[v[x,t],x],D[v[x,t],t]==cˆ2*D[u[x,t],x]}
ibc1={u[x,0]==f[x],v[x,0]==0,u[0,t]==0,u[lL,t]==0}
{sol1=NDSolve[{sys1,ibc1},{u,v},{x,a,b},{t,0,tT}], w=u/.sol1[[1]],
q=v/.sol1[[1]]}
Plot3D[Evaluate[w[x,t]],{x,a,b},{t,0,tT},BoxRatios->1,PlotRange->All]
Plot3D[Evaluate[q[x,t]],{x,a,b},{t,0,tT},BoxRatios->1,PlotRange->All]
{w[0.1,0], q[0.1,0]}
Do[g[i]=Plot[Evaluate[w[x,lL1[[i]]]],{x,0,lL},PlotStyle->{lL2[[i]],
Thickness[0.001]}],{i,1,nNL1}]; Show[Table[g[i],{i,1,nNL1}],ImageSize->500]
gu=Plot[Evaluate[w[x,tT]],{x,0,lL},PlotStyle->{lL2[[1]],Thickness[0.007]}];
gv=Plot[Evaluate[q[x,tT]],{x,0,lL},PlotStyle->{lL2[[2]],Thickness[0.007]}];
Show[{gu,gv}]
ParametricPlot[{Evaluate[w[r,tT],q[r,tT]]},{r,a,b},AspectRatio->1]
Animate[Plot[w[x,t],{x,a,b},PlotRange->{-1,1},PlotStyle->Thickness[0.001],
PlotRange->All],{t,0,tT},AnimationRate->0.5]
⊙ References for Section 23.4: J. Crank and P. Nicolson (1947), L. O. Collatz (1966), R. D. Richtmyer
and K. W. Morton (1967), K. W. Morton and D. F. Mayers (1995), J. W. Thomas (1995), L. Lapidus and
G. F. Pinder (1999), L. Strikwerda (2004), R. J. LeVeque (2007), S. Larsson and V. Thomée (2008), I. K.
Shingareva and C. Lizárraga-Celaya (2011).
Chapter 24

Equations with MATLAB R
24.1 Introduction
In the previous two chapters, we paid special attention to analytical solutions of linear par-
tial differential equations and systems in the context of computer algebra systems Maple
and Mathematica. However, in real-world problems, the functions and data in PDE prob-
lems are often defined at discrete points and the equations are too complicated, so that it
is not possible to construct analytical solutions. Therefore, one has to study and develop
numerical approximation methods for linear PDEs [e.g., see Crank and Nicolson (1947),
Larsson and Thomée (2008), Lax (1968), LeVeque (2007), Li and Chen (2009)].
Nowadays, for this purpose one can use computers and supercomputers equipped with
convenient and powerful computational software such as MATLAB, an interactive pro-
gramming environment for scientific computing, which provides integrated numeric com-
putation and graphical visualization in a high-level programming language. MATLAB’s
excellent graphics capabilities can help one better understand the results and the solution
properties.
In this chapter, we turn our attention to numerical methods for solving linear partial
differential equations. Following the most important ideas and methods, we apply and
develop numerical methods to obtain numerical and graphical solutions of linear partial
differential equations. We compute numerical solutions via MATLAB predefined func-
tions (which implement well-known methods for solving partial differential equations) and
develop new procedures for constructing numerical solutions (e.g., by applying the method
of characteristics and finite difference approximations) with the aid of MATLAB.
24.1.1 Preliminary Remarks
In this chapter, we consider second-order linear partial differential equations in one and
two space dimensions. These PDEs are classified into three groups, elliptic, parabolic, and
hyperbolic equations. The most significant difference is the order of the derivative of the
unknown function u with respect to time t. If the derivative is absent, the PDE is elliptic.
The first and second derivatives correspond to parabolic and hyperbolic equations, respec-
1367
1368 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH MATLAB
R
tively. These three classes of equations are associated with equilibrium states, diffusion
states, and oscillating systems, respectively. Since analytical solutions of such PDEs with
initial and boundary conditions are usually difficult to find, we consider numerical methods
for solving these equations.
In general, an elliptic PDE can be written in the form
−div(c∇u) + au = f (x,t), (24.1.1.1)
where c, a, and f are scalar (complex-valued) functions defined on a bounded domain D,

x1 , x2 , . . . , xn ,t are independent variables, u is the unknown scalar (complex-valued) func-
tion (dependent variable) of these independent variables, u = u(x1 , x2 , . . . , xn ,t) = u(x,t),
∇u is the gradient of the unknown function u, and div(c∇u) is the divergence of c∇u. If c
is a constant, we can simplify this to div(c∇u) = c∆u, where ∆ is the Laplacian operator.
In this case, the elliptic PDE has the form
−c (ux1 x1 + ux2 x2 + · · · uxn xn ) u + au = f (x,t).
We will use the subscripts on dependent variables to denote differentiations; e.g., ux1 =
∂u/∂x1 , ux1 x1 = ∂2 u/∂x1 ∂x1 .
Equation (24.1.1.1) is defined in a bounded domain D of the n-dimensional space Rn in
the independent variables x1 , x2 , . . . , xn . We will find numerical solutions of Eq. (24.1.1.1),
i.e., functions u = u(x1 , x2 , . . . , xn ,t) that satisfy Eq. (24.1.1.1) in D and additional initial
and boundary conditions.
The general form of a parabolic PDE can be written as
dut − div(c∇u) + au = f (x,t), (24.1.1.2)
where d, c, and a are scalar functions. If c is a constant, then the equation can be simplified
as
dut − c (ux1 x1 + ux2 x2 + · · · uxn xn ) u + au = f (x,t).
The general form of a hyperbolic PDE can be written as
dutt − div(c∇u) + au = f (x,t). (24.1.1.3)
If c is a constant, then the equation can be simplified as
dutt − c (ux1 x1 + ux2 x2 + · · · uxn xn ) u + au = f (x,t).
For linear parabolic and hyperbolic PDEs, the coefficients c, a, f , and d can depend on
time t.
24.1.2 Brief Introduction to MATLAB

MATLAB (short for “matrix laboratory”) is not a general purpose programming language
like Maple and Mathematica.
MATLAB is an interactive programming environment that provides powerful high-
performance numerical computing, excellent graphics visualization, symbolic computing
capabilities, and capabilities for writing new software programs using a high-level pro-
gramming language.
Symbolic Math Toolbox (Release ≥ 4.9) is based on the muPAD symbolic kernel and
provides symbolic computations and variable-precision arithmetic. Earlier versions of Sym-
bolic Math Toolbox are based on the Maple symbolic kernel.
Simulink (short for “simulation and link”), also included in MATLAB, offers modeling,
simulation, and analysis of dynamical systems (e.g., signal processing, control, communi-
cations, etc.) under a graphical user interface (GUI) environment.
The first concept of MATLAB and the original version (written in Fortran) were devel-
oped by Prof. Cleve Moler at the University of New Mexico in the late 1970s to provide
his students with simple interactive access to LINPACK and EISPACK software (without
having to learn Fortran).1 Over the next several years, this original version of MATLAB
spread within the applied mathematics community. In early 1983, Jack Little, together
with Cleve Moler and Steve Bangert, developed a professional version of MATLAB (writ-
ten in C and integrated with graphics). The company MathWorks was created in 1984 and
headquartered in Natick, Massachusetts, to continue its development.
The most important features of MATLAB are as follows: interactive user interface;
a combination of comprehensive mathematical and graphics functions with a powerful
high-level language in an easy-to-use environment; fast numerical computation and vis-
ualization, especially for performing matrix operations [e.g., see Higham (2008)]; great
flexibility in data manipulation; symbolic computing capabilities via the Symbolic Math
Toolbox (Release < 4.9 or Release ≥ 4.9), based on the Maple or muPAD symbolic ker-
nel, respectively; easy usability; the basic data element is an array that does not require
dimensioning; a large library of functions for a broad range of applications; easy incor-
poration of new user-defined capabilities (toolboxes consisting of M-files and written for
specific applications); understandability and availability on almost all operating systems;
a powerful programming language with intuitive and concise syntax and easy debugging;
Simulink, as an integral part of MATLAB, provides modeling, simulation, and analysis
of dynamical systems; free resources (e.g., MathWorks Web Site www.mathworks.com,
MathWorks Education Web Site www.mathworks.com/education, and MATLAB group
comp.soft-sys.matlab, etc.).
MATLAB consists of five parts: Development Environment, a set of tools that facili-
tate using MATLAB functions and files (e.g., graphical user interfaces and the workspace);
Mathematical Function Library, a vast collection of computational algorithms; the MAT-
LAB language, a high-level matrix/array language (with flow control statements, functions,
data structures, input/output, and object-oriented programming features); the MATLAB
graphics system, which includes high-level functions (for 2D/3D data visualization, image
processing, animation, etc.) and low-level functions (for fully customizing the appearance
of graphics and constructing complete graphical user interfaces); Application Program In-
terface (API), a library for writing C and Fortran programs that interact with MATLAB.
Basic concepts. The prompt symbol >> serves to type a MATLAB function; typing a
statement and pressing Return or Enter at the end of the function serves to evaluate the
1 LINPACK and EISPACK are collections of Fortran subroutines developed by Cleve Moler and his col-
leagues for solving linear equations and eigenvalue problems, respectively.

R
MATLAB function, display the result, and insert a new prompt; the semicolon (; ) symbol
at the end of a MATLAB function serves only to evaluate the function without displaying
any result.
In contrast to Maple and Mathematica, it is not possible in MATLAB to move the cursor
to the desired line, but corrections for sufficiently simple problems can be made by pressing
the up- and down-arrow keys for scrolling through the list of functions (recently used) and
then left- and right-arrow keys for changing the text. Also, corrections can be made
by using copy/paste of the previous lines located in the Command Window or Command
History.
The previous result (during a session) can be referred to with the variable ans (the last
result). MATLAB prints the answer and assigns the value to ans, which can be used for
further calculations.
MATLAB has many forms of help: a complete online help system with tutorials and
reference information for all the functions; the command line help system, which can be
accessed by using the Help menu, by pressing F1, by selecting Help->Demos, and by enter-
ing Help and selecting Functions->Alphabetical List or Index, Search, MATLAB->
Mathematics; or by typing helpbrowser, lookfor (e.g., lookfor plot); by typing help
FunctionName, doc FunctionName, etc.
In MATLAB 7, a new feature for typing function names correctly has been added. It is
possible to type only the first few letters of the function and then to press the TAB key (to
observe all available functions and complete typing the function).
MATLAB desktop contains tools (graphical user interfaces) for managing files, vari-
ables, and applications. The default configuration of desktop includes various tools, e.g.,
Command Window, Command History, Current Directory, Workspace, Find Files (for
more details, see demo MATLAB desktop), etc. It is possible to modify the arrangement of
tools and documents.
For a new problem, it is best to begin with the statement clear all for cleaning all
variables from MATLAB’s memory. All examples and problems in the book are assumed
to begin with clear all.
A MATLAB program can be typed at the prompt >> or, alternatively (e.g., for more
complicated problems), by creating an M-file (with .m extension) using MATLAB editor
(or using another text editor). MATLAB editor is invoked by typing edit at the prompt.
M-files are files that contain code in the MATLAB language. There are two kinds of
M-files: script M-files (which do not accept input arguments or return output data) and
function M-files (which can accept input arguments and return output arguments).
In the process of working with various M-files, it is necessary to define the path, which
can be done by selecting File->Set Path->Add Folder or via the cd function.
The structure of a MATLAB program or source code is as follows: the main program
or script and the necessary user-defined functions. The execution starts by typing the file
name of the main program.
Incorrect response. If you get no response or an incorrect response, you may have
entered or executed a function incorrectly. To correct the function or interrupt the com-
putation by entering debug mode and setting breakpoints, select Debug->Open M-files
when Debugging and Debug->Stop if Errors/Warnings on the Desktop menu.
It is also possible to detect erroneous or unexpected behavior in a program with the aid
of MATLAB functions, e.g., break, warning, and error.

Palettes can be used, e.g., for building or editing graphs (Figure Palette), displaying
the names of GUI components (Component Palette), etc.
MATLAB graphical user interface development environment (GUIDE) provides a set
of tools for creating graphical user interfaces (GUI). These tools greatly simplify the pro-
cess of constructing GUI, e.g., customizing the layout of the GUI components (panels,
buttons, menus, etc.), and programming the GUI.
MATLAB consists of a family of add-on toolboxes, which are collections of MATLAB
functions (M-files) that extend the MATLAB environment to solve particular classes of
problems.
Toolboxes can be standard and specialized (see Contents in Help). Nowadays, a vast
number of specialized toolboxes are available. MATLAB can be augmented by a number
of toolboxes consisting of M-files and written for specific applications.
24.1.3 MATLAB Language
MATLAB language is a high-level procedural dynamic and imperative programming lan-

guage (similar to Fortran 77, C, and C++), with powerful matrix/array operations, control
statements, functions, data structures, input/output, and object-oriented programming fea-
tures. On the other hand, MATLAB language is an interpreted language, similar to Maple
and Mathematica [e.g., see Shingareva and Lizárraga-Celaya (2009)]; i.e., the instructions
are translated into machine language and executed in real time (one at a time). MATLAB
language allows programming-in-the-small (coding or creating programs for performing
small-scale tasks) and programming-in-the-large (creating complete large and complex ap-
plication programs). It supports a large collection of data structures or MATLAB classes
and operations among these classes.
In linear algebra, there exist two distinct types of operations with vectors/matrices: op-
erations based on the mathematical structure of vector spaces and element-by-element op-
erations on vectors/matrices as data arrays. This difference can be made in the name of the
operation or the name of the data structure. In MATLAB, separate operations are defined
(for matrix and array manipulation), but the data structures array and vector/matrix are
the same. Note that, for example, in Maple the situation is opposite: the operations are the
same, but the data structures are different.
Arithmetic operators: scalar operators (+ - * / ˆ ), matrix multiplication/power (*
ˆ ), array multiplication/power (.* .ˆ ), left/right matrix division (\ /), and array division
(./).
Logical operators: and (&), or (|), exclusive or (xor), not (˜).
Relational operators: less/greater than (< >), less/greater than or equal to (<= >=),
equal/not equal (== ˜=).
A variable name is a character string of letters, digits, and underscores that begins with
a letter and whose length is bounded by N=namelengthmax (e.g., N = 63). Punctuation
marks are not allowed (see genvarname function). Variable declaration is not necessary in
MATLAB, but all variables must be given initial values, e.g., a12 new=9. A variable can
change in the calculation process, e.g., from integer to real (and vice versa).
R
MATLAB is case sensitive; i.e., it distinguishes between lowercase and uppercase let-
ters, as in pi and Pi.
Various reserved keywords, symbols, names, and functions, for example, reserved key-
words and function names, cannot be used as variable names (see isvarname, which -all,
isreserved, iskeyword).
A string variable is enclosed in single quotes and belongs to the char class (e.g.,
x='string'), so that the function sin(x) is invalid. Strings can be used with converting,
formatting, and parsing functions (e.g., see cellstr, char, sprintf, fprintf, strfind,
findstr).
MATLAB provides three basic types of variables: local variables, global variables,
and persistent variables.
The operator “set equal to” (=). A variable in MATLAB (in contrast to Maple and
Mathematica) cannot be “free” (with no assigned value) and must be assigned any initial
value by the operator “set equal to” (=).
The difference between the operators “set equal to” (=) and “equal” (==). The operator
var=val is used to assign val to the variable var, and the operator val1==val2 compares
two values; e.g., A=3; B=3; A==B.
Statements are keyboard input instructions that are executed by MATLAB (e.g., for
i=1:N s=s+i*2; end). A MATLAB statement may begin at any position in a line and
may continue indefinitely in the same line, or may continue in the next line, by typing three
dots (...) at the end of the current line. White spaces between words in a statement are
ignored; a number cannot be split into two pieces separated by a space.
The statement separator semicolon (;) The result of a statement followed with a semi-
colon (;) will not be displayed. If the semicolon is omitted, the results will be printed on
the screen, e.g., x=-pi:pi/3:pi; and x=-pi:pi/3:pi.
Multiple statements in a line: two or more statements may be written in the same line
if they are separated with semicolons.
Comments can be included with the percent sign % and all characters following it up to
the end of a line. Comments at the start of a code have a special significance: they are used
by MATLAB to provide the entry for the help manual for a particular script. The block
comment operators %{ %} can be used for writing comments that require more than one
line.
An expression is a valid statement and is formed as a combination of numbers, variables,
operators and functions. The arithmetic operators have different precedences (increasing
precedence + - * / ˆ ). Precedence is altered by parentheses (the expressions within
parentheses are evaluated before the expressions outside the parentheses).
A boolean or logical expression is formed with logical and relational operators, e.g.,
x>0. Logical expressions are used in if, switch, and while statements. The logical values
true and false are represented by the numerical values 1 and 0, respectively.
A regular expression is a string of characters that defines a pattern; e.g., 'Math?e\w*'.
Regular and dynamic expressions can be used to search text for a group of words that
matches the pattern (e.g., for parsing or replacing a subset of characters within a text).
MATLAB is sensitive to types of brackets and quotes (see help paren, help punct).
Types of brackets:
Square brackets, [ ], for constructing vectors and matrices, such as A1=[1 2 3],
A2=[1, 2, 3], A3=[1, 2; 4, 5], and for multiple assignment statements, for example,
A4=[1,5;2,6], [L,U]=lu(A4).
Parentheses, ( ), for grouping expressions, (5+9)*3, for delimiting the arguments of
functions, sin(5), for vector and matrix elements, A1(2), A3(1,1), A2([1 2]), and in
logical expressions, A1(A1>2).
Braces, { }, for working with cell arrays, C1={int8(3) 2.59 'A'}, C1{1},
X(2,1)={[1 3; 4 6]}.
Dot-parentheses, .( ), for working with a structure via a dynamic field name, S.F1=1;
S.F2=2; F='F1'; val1=S.(F).
Quotes:
Forward-quotes, (' '), for creating strings, e.g., T='the name=7;' k=5; disp('the
value of k is'); disp(k).
Single forward-quote and dot single forward-quote, (' .'), for matrix transposition (the
complex conjugate/nonconjugate transpose of a matrix), A1=[1+i,i;-i,1-i]; A1'; A.'.
Types of numbers. Numbers are stored (by default) as double-precision floating point
(class double). To operate with integers, it is necessary to convert from double to the inte-
ger type (e.g., classes int8, int16, int32), x=int16(12.3), str='MATLAB', int8(str).
Mathematical operations that involve integers and floating-point numbers result in an in-
teger data type. Real numbers can be stored as double-precision floating point (by de-
fault) or single-precision floating point, e.g., x1=3.25, x2=single(x1), x3=double(x2)
(for details, see whos, isfloat, class). Complex numbers can be created as z1=1+2*i,
z2=complex(1,2). Rational numbers can be formed by setting the format to rational,
e.g., x=3.25; format rational x format. To check the current format setting, type
get(0,'format').
Predefined constants: symbols for commonly used mathematical constants, e.g., true,
false, pi, i, j, Inf, inf, NaN (not a number), exp(1), the Euler constant γ, -psi(1), eps.
In MATLAB, there are predefined functions and user-defined functions.
Predefined functions are divided into built-in functions and library functions.
Built-in functions are precompiled executable programs and run much more efficiently (see
help elfun, help elmat).
Library functions are stored as M-files (in libraries or toolboxes), which are available in
readable form (see which, type, exist). MATLAB can be supplemented with locally
user-developed M-files and toolboxes.
Many functions are overloaded (i.e., have an additional implementation of an existing
function) so that they handle distinct classes (e.g., which -all plot).
Numerous special functions are defined, e.g., help bessel, help specfun.
User-defined functions can be created as M-files (see help ’function’) or as anony-
mous functions.
User-defined function written in an M-file (with the extension .m) must contain only
one function. It is best to have the same name for the function name and the file name. The
process of creating functions is as follows: create and save an M-file using a text editor,
then call the function in the main program (or in Command Window).
Functions written in M-files have the form
function OArg=FunName(IArg); FunBody; ,or
function [OArg1,OArg2,...]=FunName(IArg1, IArg2,...); FunBody;
R
where OArg and IArg are output arguments and input arguments, respectively.
For example, the function y = sin x is defined as function f=SinFun(x); f=sin(x);
Evaluation of functions: FunName(Args).
For example, for the sine function we have cd('c:/mypath'); SinFun(pi/2); type
SinFun.
Anonymous functions serve to create simple functions without storing functions to files.
Anonymous functions can be constructed either in the Command Window or in any function
or script; e.g., the function f (x) = sin x is defined as f=@(x) sin(x); f(pi/2).
A function handle, @, is one of the standard MATLAB data types that provides call-
ing functions indirectly, e.g., calling a subfunction when outside the file that defines that
subfunction (see class function handle) .
Nested functions are allowed in MATLAB; i.e., one or more functions or subfunctions
within another function can be defined in MATLAB. In this case, the end statements are
necessary.
The MATLAB language has the following control structures: the selection structures
if, switch, try and the repetition structures for, while.
MATLAB does not have a module system in the traditional form: it has a system based
on storing scripts and functions in M-files and placing them into directories (see cd function
for changing the current directory, help ..).
MATLAB data structures or classes, vectors, matrices, and arrays, are used for repre-
senting more complicated data. There are 15 fundamental classes, which are the form of
a matrix or array: double, single, int8, uint8, int16, uint16, int32, uint32, int64,
uint64, char, logical, function handle, struct, and cell. The numerical values are
represented (by default) as floating-point double precision (float double). It is possible
to construct various composite data types (e.g., sequences, lists, sets, tables, etc.) using the
classes struct and cell.
Vectors are ordered lists of numbers separated by commas or spaces inside [ ]; no
dimensioning is required. But vector and array indices can only be positive and nonzero.
The notation X=[1:0.1:9] stands for a vector of numbers from 1 to 9 in increments of 0.1
(see help colon).
Matrices are rectangular arrays of numbers (row/column vectors are special cases of
matrices).
⊙ References for Section 24.1: J. Crank and P. Nicolson (1947), P. D. Lax (1968), R. J. LeVeque (2007),
S. Larsson and V. Thomée (2008), N. J. Higham (2008), J. Li and Y. T. Chen (2009), I. K. Shingareva and
C. Lizárraga-Celaya (2009).
24.2 Numerical Solutions of Linear PDEs
In this section, we consider the construction of numerical and graphical solutions of var-
ious mathematical problems (initial-boundary value problems, boundary value problems,
and initial value problems) using predefined MATLAB functions and default methods. In
particular, we construct numerical and graphical solutions of scalar linear PDEs in one
space dimension (e.g., the linear heat equation, the convection–diffusion equations, and the
24.2. Numerical Solutions of Linear PDEs 1375
linear Euler equation) and scalar linear PDEs in two space dimensions (e.g., the linear Pois-
son equation, the linear heat equation with radiation, and the linear nonhomogeneous wave
equation). Additionally, we construct numerical and graphical solutions of linear problems
defined on a more complicated geometry, which can consist of various solid objects, e.g.,
for the linearized Poisson–Boltzmann equation defined on an irregular domain [see Harries
(1998)].
24.2.1 Constructing Numerical Solutions via Predefined Functions

1D Partial Differential Equations
First, consider the predefined MATLAB functions with the aid of which we can obtain
approximate numerical solutions solving various linear PDE problems. Applying the pre-
defined function pdepe (provided in MATLAB PDEToolbox),2 we can numerically solve
initial-boundary value problems for a class of linear parabolic PDEs in one space variable x
and time t.3
Consider a single linear PDE in one space dimension. In this case, the predefined function
pdepe allows us to solve initial-boundary value problems for a single parabolic PDE in 1D
(see help pdepe). These PDEs involve an unknown function u that depends on a scalar
space variable x and a scalar time variable t.
In MATLAB notation, a general class of linear parabolic PDEs defined in the domain
D = {xL ≤ x ≤ xR , t0 ≤ t ≤ t f } is represented in the form

c(x,t)ut = x−m ∂x xm f (x,t, u, ux ) + s(x,t, u, ux ), (24.2.1.1)
where xL , xR , t0 , t f are given constants, f (x,t, u, ux ) is the flux, s(x,t, u, ux ) is the source
term, m ∈ {0, 1, 2} (which corresponds to slab, cylindrical, or spherical symmetry, respec-
tively), and c(x,t) > 0 for parabolic equations.
The initial condition at the initial time t = t0 has the form
u(x,t0 ) = u0 (x) (24.2.1.2)
and the boundary conditions at x = xL and x = xR (for t0 ≤ t ≤ t f ) have the form
p(xL ,t, u) + q(xL ,t) f (xL ,t, u, ux ) = 0, p(xR ,t, u) + q(xR ,t) f (xR ,t, u, ux ) = 0. (24.2.1.3)
One can impose the Dirichlet, Neumann, Robin, or periodic boundary conditions.
sol=pdepe(m,@PDEfun,@ICfun,@BCfun,xMesh,tSpan,ops)
uk=sol(j,:,k); [uOut,DuOutDx]=pdeval(m,xMesh,uk,xOut)
PDEfun, ICfun, BCfun are function handles.

[c,f,s]=PDEfun(x,t,u,DuDx) evaluates the quantities defining the partial differential
equation, where PDEfun is the function name and the functions c, f, s can be
calculated from the given arguments.
2 The name pdepe stands for parabolic-elliptic partial differential equations.
3A single elliptic PDE in one space dimension is treated as an ODE.
R
u0=ICfun(x) evaluates the initial conditions.
[pL,qL,pR,qR]=BCfun(xL,uL,xR,uR,t) evaluates the boundary conditions at time t.
uk approximates component k of the solution at time tSpan(j) and mesh points xMesh.
pdeval evaluates the function u and the derivative ux DuDx at the array of points xOut at
points that are not in xMesh; it stores them in uOut and DuOutDx, respectively.
The numerical solution obtained is represented as a three-dimensional array sol, where

sol(:,:,k) approximates component k of the solution, sol(i,:,k) approximates com-
ponent k of the solution at time tSpan(i) and mesh points xMesh(:), and sol(i,j,k)
approximates component k of the solution at time tSpan(i) and the mesh point xMesh(j).
Numerical solutions can be used for obtaining visualizations (solution surface, solution
profile, or animation), e.g., by using plot and surf (for more detail, see help graphics).
For more advanced applications, it is best to specify options that change the default in-
tegration parameters: error control and step size (for details, see odeset, RelTol, AbsTol,
NormControl, InitialStep, MaxStep).
Example 24.1. Linear heat equation. Let us find numerical and graphical solutions of the
following initial-boundary value problem for the linear heat equation (as in Problems 28 and 22 of
the Maple and Mathematica chapters, respectively):
ut = νuxx , u(x, 0) = A sin4 (πx), u(0,t) = 0, u(L,t) = 0,
in the domain D = {0 ≤ x ≤ L, 0 ≤ t ≤ T } (L = 1) with initial amplitude A = 1 and kinematic

viscosity ν = 1/30.
First, let us rewrite the heat equation with initial and boundary conditions in the form (24.2.1.1)–
(24.2.1.3). We create three separate function M-files for each set of functions, i.e., PDE1.m for the
functions c, f , and s, IC1.m for the function u(x, 0) = u0 (x), and BC1.m for the functions p and q.
The pdepe function will combine these M-files and construct the solution of the problem.
For the heat equation, we have m = 0, c(x,t) = 1, f (x,t, u, ux ) = νux , and s(x,t) = 0; for the
initial condition, we have u0 (x) = A sin4 (πx); and for the boundary conditions, we have p(0,t, u) =
u, q(0,t) = 0, p(L,t, u) = u, and q(L,t) = 0. These functions are specified in the function M-files as
follows:
function [c,f,s]=PDE1(x,t,u,DuDx)
nu=1/30; c=1; f=nu*DuDx; s=0;
function u0=IC1(x)
A=1; u0=A*sin(pi*x)ˆ4;
function [pL,qL,pR,qR]=BC1(xL,uL,xR,uR,t)
pL=uL; qL=0; pR=uR; qR=0;
Then, for constructing numerical and graphical solutions of this initial-boundary value problem
for the heat equation, we compose the script M-file SolPDE1 as follows:
clear all; close all; echo on; format long; m=0; a=0; b=1; Nx=101;
t0=0; tf=40; Nt=161; x=linspace(a,b,Nx); t=linspace(t0,tf,Nt);
sol=pdepe(m,@PDE1,@IC1,@BC1,x,t); u=sol(:,:,1); n=2:10; m=1:10;
u(m,n)' figure(1);
surf(x,t,u); rotate3d on; title('Heat equation. Surface plot of solution');
xlabel('Distance x'); ylabel('Time t');
Solution profile at t =0, 1/2, 1

k
1
t=0
0.9 t=1/2
t=1
0.8
0.7
0.6
u(x,tk) 0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
x
Figure 24.1 Surface profiles of the solution of the linear heat equation at tk = 0, 12 , 1.
figure(2);
plot(x,u(1,:),'k-',x,u(3,:),'k--',x,u(5,:),'k:','LineWidth',2);
set(gca,'FontSize',14); set(gca,'FontName','Arial'); set(gca,'LineWidth',1);
title('Solution profile at t_k=0, 1/2, 1');
xlabel('x'); ylabel('u(x,t_k)'); legend('t=0','t=1/2','t=1');
figure(3);
G=plot(x,u(1,:),'erase','xor');
for k=2:length(t) set(G,'xdata',x,'ydata',u(k,:)); pause(0.7); end
title('Animation of the solution'); xlabel('x'); ylabel('u(x,t_k)');
echo off
In this script M-file, we choose a grid of x and t values (linsolve functions), solve the linear
equation (pdepe function), and extract the solution u from sol as u=sol(:,:,1). Then we print some
numerical values of the solution, plot it as a surface with interactive rotation (the rotate3d function),
plot the surface profile at tk = 0, 1/2, 1, and produce animations of the surface profile for t ∈ [0, 40].
Before working with these new files that are located in a new working directory in the Command
Window, we can make this directory the current working directory in MATLAB by using cd function,
for example cd('c:/lpde'), or by selecting this directory in the Desktop->Current Directory, or by
adding it to the path of directories that MATLAB searches to find files File->SetPath->AddFolder, or by
using addpath('c:/lpde'). Finally, we run the file SolPDE1 in the Command Window.
The text and numerical values display in the same Command Window, but graphics appear
in separate graphics windows (Figure1, Figure2, and Figure3). Three surface profiles of the
solution obtained at times tk = 0, 12 , 1 are shown in Fig. 24.1.
Example 24.2. Linear convection–diffusion equation. Consider the linear convection–diffusion

equation of the form
ut = uxx − kux ,
where k ≥ 0 is a constant. We find numerical and graphical solutions of the equation under the
R
following boundary and initial conditions:



1, x < a,
ut = uxx − kux in D, u(0,t) = 0, u(L,t) = 0, u(x, 0) = w(x), a < x < b,


0, x > b,
where k = 1, L = 10, a and b are real parameters, and w(x) is a real-valued function; e.g., we take
a = −1, b = 1, w(x) = cos(x), and D = {0 ≤ x ≤ 10, 0 ≤ t ≤ 10}.
As before, we rewrite the linear convection–diffusion equation with initial and boundary con-
ditions in the form (24.2.1.1)–(24.2.1.3) and create three separate function M-files, PDE2.m for the
functions c, f , and s, IC2.m for the function u(x, 0) = u0 (x), and BC2.m for the functions p and q.
For the convection–diffusion equation, we have m = 0, c(x,t) = 1, f (x,t, u, ux ) = ux , and
s(x,t, u, ux ) = −kux .
For the boundary conditions, we have p(0,t, u) = u, q(0,t) = 0, p(L,t, u) = u, q(L,t) = 0, and
the discontinuous initial data u0 (x) defined above.
These functions are specified in the function M-files as follows:
function [c,f,s]=PDE2(x,t,u,DuDx)
c=1; f=DuDx; k=1; s=-k*DuDx;
function [pL,qL,pR,qR]=BC2(xL,uL,xR,uR,t)
pL=uL; qL=0; pR=uR; qR=0;
function u0=IC2(x)
a=-1; b=1;
if x<a u0=1; elseif x>a && x<b u0=cos(x); else u0=0; end
Then, for constructing numerical and graphical solutions of this initial-boundary value problem
for the linear convection–diffusion equation, we compose the script M-file SolPDE2 as follows:
sol=pdepe(m,@PDE2,@IC2,@BC2,x,t); u=sol(:,:,1); figure(1);
surf(x,t,u); rotate3d on;
title('Convection--diffusion equation. Surface plot of solution');
xlabel('Distance x'); ylabel('Time t');
figure(2);
plot(x,u(10,:),'k-',x,u(30,:),'k--',x,u(101,:),'k:','LineWidth',2);
set(gca,'FontSize',14); set(gca,'FontName','Arial');
set(gca,'LineWidth',1);
title('Solution profile at t_k=0.9, 2.9, 10')
xlabel('x'); ylabel('u(x,t_k)'); legend('t=0.9','t=2.9','t=10');
figure(3);
G=plot(x,u(10,:),'erase','xor');
for k=2:length(t) set(G,'xdata',x,'ydata',u(k,:)); pause(0.7); end
title('Animation of the solution'); xlabel('x'); ylabel('u(x,t_k)');
echo off
We choose a grid of x and t values (linsolve functions), solve the linear convection–diffusion
equation (pdepe function), and extract the solution u from sol as u=sol(:,:,1). Then we plot
the solution as a surface (with interactive rotation, rotate3d function), plot the surface profile at
tk = 0.9, 2.9, 10, and produce animations of the surface profile for t ∈ [0, 10]. Three surface profiles
of the solution obtained at times tk = 0.9, 2.9, 10 are shown in Fig. 24.2.
Solution profile at t =0.9, 2.9, 10

k
0.14
t=0.9
t=2.9
0.12 t=10
0.1
0.08
u(x,t )
k
0.06
0.04
0.02
0
0 2 4 6 8 10
x
Figure 24.2 Surface profile evolution of the convection–diffusion equation for tk = 0.9, 2.9, 10.
2D Partial Differential Equations

The PDE Toolbox allows one to solve 2D linear partial differential equations via a
graphical user interface (GUI). The solution domains can be complicated and drawn by GUI
tools using basic elements (circles, ellipses, rectangles, and polygons), their combinations,
and basic set operations (union, difference, and intersection).
A solution procedure contains four stages: PDE for describing a partial differential
equation, ICs for setting the initial conditions (ICs), BCs for setting the boundary conditions
(BCs), Domain for drawing the solution domain.
The PDE GUI has the following functions:
Menu system for calling functions directly from the menu items and toolbar buttons.
Toolbar for defining the solution domains, setting parameters in the PDE, solving the
PDE, visualizing the results.
Set formula for defining set operations (union, intersect, difference).
Solution regions for drawing solution domains, solving 2D PDE within the solution do-
main, visualizing the results (in 3D).
pdetool Draw Boundary PDE Mesh Solve Plot

pdeinit pdecirc pdeellip pdepoly pderect pdebound
assemb initmesh refinemesh assempde pdenonlin parabolic
hyperbolic pdeeig pdesurf pdeplot pdemesh pdecont
pdetool starts the PDE Toolbox that invokes the graphical user interface (GUI).
R
Draw, Boundary, PDE, Mesh, Solve, Plot are various icons (or modes) in the GUI menu
corresponding to different stages of the solution process.
pdeinit, pderect, pdebound, initmesh, assempde, pdenonlin, pdeplot are examples

of functions (which correspond to different modes of the GUI interface) for writing
programs.
For starting the PDE Toolbox, one should type pdetool in the Command Window (in
the MATLAB prompt). This invokes the graphical user interface (in a separate window),
which is appropriate for simple or standard PDE problems. For more complicated problems
and full control over the numerical process, it is best to write programs and call them from
the prompt or M-files.
The PDE Toolbox allows solving three classes of PDE (see Section 24.1.1) and the
eigenvalue problems defined on a bounded domain D ∈ R2 :
−∇ · (c∇u) + au = f , elliptic equations.

dut − ∇ · (c∇u) + au = f , parabolic equations.
dutt − ∇ · (c∇u) + au = f , hyperbolic equations.
−∇ · (c∇u) + au = λdu, eigenvalue problems.
Here c is the scalar coefficient, λ is an unknown eigenvalue, and a, d, f , and u are complex-
valued functions defined on D. For linear PDEs, the coefficients a, d, and c do not depend
on the unknown solution u. For hyperbolic and parabolic PDEs, the coefficients can depend
on time.
The PDE Toolbox allows one to define the following boundary conditions:

∂
h(x,t)u = r(x,t), (c∇u) + qu = g,
∂D ∂n ∂D
which, respectively, are referred to as the Dirichlet (or essential) boundary conditions, gen-
eralized Neumann (or natural) boundary conditions,4 and the mixed boundary conditions
(a combination of Dirichlet and generalized Neumann conditions). Here ∂D denotes the
∂
boundary of the solution domain, ∂n is the partial derivative of vector x in the normal di-
rection, and g, q, h, and r are complex-valued functions defined on the boundary ∂D. For
linear PDEs, the coefficients g, q, h, and r do not depend on the unknown solution u. For
eigenvalue problems, g = 0, r = 0, and for hyperbolic and parabolic PDEs, the coefficients
can depend on time.
Let us briefly describe the solution process for simple problems via the graphical user
interface (GUI). There are various icons (or modes) in the GUI menu that correspond to
various stages of the solution process:
Draw, for defining the domain D and the geometry (using the constructive solid geometry
(CSG) model paradigm), e.g., by combining various solid objects (rectangle, circle,
etc.) using set formulas.
4 In some contexts, the generalized Neumann boundary conditions are also referred to as the Robin boundary
conditions.
Boundary, for specifying the boundary conditions (e.g., distinct types of boundary condi-
tions on distinct boundary segments).
PDE, for interactively specifying the type of the PDE and the coefficients d, c, a, f (e.g.,
specifying the coefficients for each subdomain independently).
Mesh, for generating and plotting meshes and controlling the parameters of the automated
mesh generator.
Solve, for invoking and controlling the solvers (e.g., adaptive mode); for specifying the
initial conditions and the times for which the solution should be constructed (for
parabolic or hyperbolic PDEs); for specifying the interval in which to search for
eigenvalues (for eigenvalue problems).
Plot, for visualizing numerical solutions (inside the GUI and in separate figures); for si-
multaneously plotting three distinct solution properties (using color, height, and vec-
tor field plots); for plotting surfaces, meshes, contours, and arrows (quivers); for
producing animations of solutions (for parabolic or hyperbolic PDEs).
Remark. After solving a problem, it is possible to refine the mesh (by returning to Mesh) and
then solve the problem again and again. This can also be done by using an adaptive mesh
refiner and solver.
Example 24.3. Linear Poisson equation. Consider the boundary value problem for the two-
dimensional linear Poisson equation
uxx + uyy = x + y, u(x, 0) = 0, u(x, L2 ) = 0, u(0, y) = 0, u(L1 , y) = w1 (y),
describing a potential field u(x, y) in a bounded domain D = {0 ≤ x ≤ L1 , 0 ≤ y ≤ L2 }. Here L1 = 3,

L2 = 6, and w1 (y) = cos(y).
First, we solve this problem with the aid of the graphical user interface (GUI).
1◦ . We draw the domain D = {0 ≤ x ≤ 3, 0 ≤ y ≤ 6}. By selecting Options->Grid, we generate
the grid. Then, by selecting Options->Axes Limits, we change the size of the default domain;
i.e., we put the new size [−1, 4] for the x-axis and [−1, 8] for the y-axis, Apply, and Close. By
selecting rectangle-icon, we draw a rectangle with the aid of the mouse (click-drag); its name R1 is
introduced by MATLAB. The exact coordinates of the cursor appear in the top right of the window.
To check and correct our coordinates, we double-click on R1 and enter the exact coordinates.
2◦ . We specify the boundary conditions. If we select Boundary->Boundary Mode, the boundary
will appear as a red arrow, indicating the type of condition (red, blue, and green for Dirichlet, Neu-
mann, and mixed conditions) and the direction. Then, by double-clicking on the bottom horizontal
line, we set the first boundary condition u(x, 0) = 0 by introducing h = 1 and r = 0 in the corre-
sponding menu (since the Dirichlet boundary conditions have the form hu = r). Next, we specify
the remaining three boundary conditions. Then, by selecting File->Save As, we save this domain
as M-file LinearPoisson1.m.
3◦ . By selecting PDE->PDE Specification, we define the type of the PDE: Elliptic (the type
of the linear Poisson equation), c=-1, a=0, and f=x+y.
4◦ . By selecting Mesh->Initialize Mesh, we generate a triangular finite element mesh. Then,
by selecting a triangle-inside-a-triangle icon, we can refine the initial mesh.
5◦ . By selecting Solve->Parameters, we specify some parameters for solving the linear partial
differential equation: adaptive mode, maximum number of triangles, maximum number of refinements,
R
Color: u Height: u
0
2
0 −1
−2 −2
−4
−3
−6
6 −4
3
4
2
2 −5
1
0 0
Figure 24.3 Surface plot of the solution of the linear Poisson equation.
etc. Then, selecting Solve->Solve PDE, we obtain the numerical solution. By default, the solution
appears as a color-scale contour plot.
6◦ . By selecting Plot->Parameters, we can set various options for generating a figure, e.g.,
Height (3-D plot), Plot in x-y grid, Color, colormap (gray), and Show Mesh. Then, by selecting
the mesh and Edit->Current Object Properties, we can edit the plot obtained (in the MATLAB
graphics editor) and create the final plot shown in Fig. 24.3).
Remark. The MATLAB graphics editor has various options for creating an appropriate figure (e.g., interactive
rotation, Tools->Rotate 3D).
In addition, we show how to solve this problem in another way, i.e., by writing an M-file and
by applying the export facilities of the GUI. This can be very useful when solving complicated
problems.
1◦ . We type pdetool and open the M-file (created recently) LinearPoisson1.m. We have created
the domain D and defined boundary conditions using graphical user interface (GUI). Then we ex-
port the decomposed geometry (i.e., the domain D) and the boundary conditions to the workspace
by selecting Boundary->Export Decomposed Geometry, Boundary Conditions. By default, the de-
composed geometry and boundary conditions are described by g and d, respectively. In our case, g
is Decomposed Geometry Matrix and d is Boundary Condition Matrix.
2◦ . We create the triangular mesh using the graphical user interface (GUI). Then, selecting
Mesh -> Export Mesh, we export the mesh to the workspace. By default, the mesh is described by p,
e, and t. In our case, p is Point Matrix, e is Edge Matrix, and t is Triangle Matrix.
3◦ . Now we can solve the linear Poisson equation with the following M-file LinearPoisson2.m:
echo on; format long; g, b, p, e, t c=-1; a=0; f='x+y';
[p, e, t]=refinemesh(g, p, e, t); u=assempde(b, p, e, t, c, a, f);
pdeplot(p,e,t,'zdata',u,'mesh','on') echo off
In this script M-file, we use the export variables, g, b, p, e,t, the PDE coefficients, c=-1, a=0,
f=x+y, for solving PDEs with the aid of the assempde function. This is the basic Partial Differential
Equation Toolbox function, which assembles a PDE problem by using the finite element formu-
lation. Also, we refine the mesh, and the new mesh is returned using the same matrix variables.
Finally, we draw the solution and the mesh.
This problem can also be solved with the pdenonlin function and an additional parameter, the
tolerance, tol=1e−2, as follows:
echo on; format long; g, b, p, e, t c=-1; a=0; f='x+y'; tol=1e-2;

[p, e, t]=refinemesh(g, p, e, t); u=pdenonlin(b, p, e, t, c, a, f,
'Tol',tol) pdeplot(p, e, t, 'zdata', u, 'mesh', 'on') echo off
Remark. The M-file LinearPoisson1.m can be observed by type LinearPoisson1 and modified us-
ing MATLAB editor.
24.2.2 Numerical Methods Embedded in MATLAB
In MATLAB, partial differential equations or systems in one space dimension (1D) and
time, which arise in many applications, can be solved with the aid of the predefined function
pdepe (provided in MATLAB PDEToolbox).
Implementing the method of lines (in a general setting), the function pdepe converts
the PDE into an ODE using a second-order accurate spatial discretization [for details, see
Skeel and Berzins (1990), Schiesser and Griffiths (2009), Lee and Schiesser (2004)] (i.e.,
by replacing only the spatial derivatives with finite differences) based on a fixed set of
nodes or a mesh xMesh (represented as an array), where xMesh(1)=a, xMesh(end)=b,
and xMesh(i)<xMesh(i+1) (for i=2, . . . , end-1). Then the resulting ODEs are integrated
(with a stiff ODE solver ode15s) to obtain approximate solutions at times tSpan, where
tSpan(1)=t 0, tSpan(end)=t f, and tSpan(i)<tSpan(i+1) (for i=2, . . . , end-1).
Partial differential equations or systems in two space dimensions (2D) and time, which
arise in a wide variety of phenomena in all branches of science and engineering [e.g., see
Schiesser (1994), Cook, Malkus, and Plesha (1989)], can be solved with the aid of the
PDEToolbox, which allows one to define a PDE problem (geometry, boundary conditions,
and a PDE or a system of PDEs), numerically solve the problem (generate meshes, dis-
cretize the equations, and find approximation to the solution), and visualize the results.
PDEs are solved using the Finite Element Method (FEM).
In this section, we consider the method of characteristics, which allows us to reduce a PDE
to a system of ODEs along which the given PDE with some initial data (the Cauchy data)
is integrable. Once the system of ODEs is found, it can be solved along the characteristic
curves and transformed into a general solution for the original PDE [e.g., see Schiesser and
Griffiths (2009)].
Example 24.4. Linear first-order equation. Method of characteristics. Classical Cauchy prob-
lem. Consider the initial value problem
ux − uy = 1, u(x, 0) = xn .
R
By applying the method of characteristics, we can obtain
dU(x) dY (x)
= 1, U(x) = x + X0, = −1,
dx dx
and the characteristic curves are defined as Y (x) = −x + X0 , where x ∈ [x0 , x f ]. Also we can find the
solution of this Cauchy problem, u(x, y) = (x + y)n − y, and plot the characteristic curves as follows:
clear all; close all;

x=0:0.1:9; hold on; figure(1) for x0=0:1:9
ChCur=-x+x0; plot(ChCur,x,'k-','LineWidth',2); end hold off;
set(gca,'XLim',[0 9],'YLim',[0 9]);
Example 24.5. Linear Euler equation. Method of characteristics. Cauchy problem. Consider
the initial value problem for the linear Euler equation
xux + yuy = nu, u(x, 1) = 1 + e−|x| ,
where n = 1. By applying the method of characteristics, we obtain the characteristic ODEs with the
initial conditions
dX(t) dY (t) dU(t)
= X(t), = Y (t), = U(t),
dt dt dt
X(0) = s, Y (0) = 1, U(0) = 1 + e−|s|.
We can also find the solution of this Cauchy problem,

set , et , (1 + e−|s|)et ,
and plot the characteristic curves and the solution (whose graph is an integral surface). See Fig. 24.4.
s=linspace(-5,5,21); t=linspace(0,1,20); axesM=[-5 5 0 5 0 5];

f=[s; ones(size(s)); 1+exp(-abs(s))];
for i=1:length(s)
[t,x]=ode45(@CharODEs,t,f(:,i)');
X(i,:)=x(:,1); Y(i,:)=x(:,2); U(i,:)=x(:,3);
end
plot3(f(1,:),f(2,:),f(3,:),'k','LineWidth',2);
axis(axesM) hold on
X([find(X<axesM(1)|X>axesM(2)|Y<axesM(3)|Y>axesM(4))])=NaN;
surf(X,Y,U), shading flat plot3(X',Y',zeros(size(U')),'k')
xlabel x, ylabel y
We solve this Cauchy problem numerically with the aid of the MATLAB function ode45. This
function is the most widely used initial value problem solver for ODEs, where the variable-step
4/5 Runge–Kutta–Felhberg algorithm is implemented. We describe the ODEs using the following
function M-file:
function dxdt=CharODEs(t,x) dxdt=x;
Remark. It should be noted that the Maple function PDEplot (see Chapter 15) plots solutions of
Cauchy problems for general scalar first-order PDEs.
0
5
4 5
3
2 0
1
y 0 −5
x
Figure 24.4 Solution surface and characteristic curves for the linear Euler equation.
In this section, we apply the MATLAB PDEToolbox to solve some initial-boundary value
problems for parabolic and hyperbolic equations with the aid of the graphical user interface
(GUI). The procedure for solving linear parabolic and hyperbolic PDEs is similar to the
case of elliptic PDEs. Some modifications should be introduced with respect to the PDE
description (its coefficients, solve parameters, and visualizations).
For example, for parabolic PDEs, the solve parameters are time (a MATLAB vector of
times at which a solution of the parabolic PDE should be generated), the initial value u(t0 )
(the initial value can be a constant or a column vector of values on the nodes of the current
mesh), relative and absolute tolerance (relative and absolute tolerance parameters for the
ODE solver applied to solve the time-dependent part of the parabolic PDE problem). For
hyperbolic PDEs, the solve parameters are the same except for the initial values: these are
u(t0 ) and u′ (t0 ).
Example 24.6. Linear heat equation with radiation. Consider the initial-boundary value prob-
lem for the two-dimensional linear heat equation
ut = ν(uxx + uyy ) − µ(u − u0), 0 < x < L1 , 0 < y < L2 , t > 0,

u(x, y, 0) = w(x, y), u(0, y,t) = 0, u(L1 , y,t) = 0, u(x, 0,t) = 0, u(x, L2 ,t) = 0,
describing the temperature distribution u(x, y,t) in a rectangular plate D = {0 ≤ x ≤ L1 , 0 ≤ y ≤ L2 }

with radiation (from the surface). Here ν, µ, u0 are constants. Let L1 = L2 = 4, ν = 0.01, µ = 100,
u0 = 0.2, and w(x, y) = e−x cos y.
1◦ . We draw the domain D = {0 ≤ x ≤ 4, 0 ≤ y ≤ 4}:
Options->Grid: we generate the grid.

Options->Axes Limits: we change the size of the default domain, i.e., we put the new size
[0, 5] for the x-axis and [0, 5] for the y-axis, Apply, and Close.
R
Time=5000 Height: u
0.25
0.2
0.15
0.1
0.05
4
3 4
2 3
2
1 1
0 0
Figure 24.5 The mesh plot of the solution of the linear heat equation (t = 5000).
rectangle-icon: we draw a rectangle with the aid of the mouse (click-drag); its name R1 is
introduced by MATLAB. The exact coordinates of the cursor appear in the top right of
the window. To check and correct our coordinates, we double-click on R1 and enter the
exact coordinates.
2◦ . We specify the boundary conditions. If we select Boundary->Boundary Mode, the boundary will
appear as a red arrow, indicating the type of condition (Dirichlet conditions) and the direc-
tion. By double-clicking on the bottom horizontal line, we set the first boundary condition
u(x, 0,t) = 0 by introducing h = 1 and r = 0 in the corresponding menu.5 Next, we specify
the remaining three boundary conditions.
File->Save As: we save this domain as M-file LinearHeatRad2D.m.
3◦ . We specify the PDE: selecting PDE->PDE Specification, we define the type of the PDE:
Parabolic (the type of the linear heat equation) and the coefficients: c=0.01, a=100, f=20
( f = µ u0 = 100 × 0.2 = 20), and d=1.
4◦ . We generate a triangular finite element mesh: selecting Mesh->Initialize Mesh.
triangle-inside-a-triangle icon: we can refine the initial mesh.
5◦ . We specify parameters for solving the linear parabolic equation: Solve->Parameters. Time
(the time range): 0:100:5000; u(t0) (the initial condition): exp(-x).*cos(y); Relative
tolerance: 0.01, Absolute tolerance: 0.001.
Solve->Solve PDE: we obtain the numerical solution. By default, the solution appears as a
color-scale contour plot.
6◦ . We can set various options for generating a figure: Plot->Parameters, e.g., Height (3-D
plot), Plot in x-y grid, Color, colormap (gray), and Show Mesh.
5 The Dirichlet boundary conditions have the form hu = r.

By selecting the mesh and Edit->Current Object Properties, we can edit the plot ob-
tained (in the MATLAB graphics editor) and create the final plot shown in Fig. 24.5).
Example 24.7. Linear nonhomogeneous wave equation. Consider the initial-boundary value
problem for the two-dimensional linear wave equation
utt = ν2 (uxx + uyy ) + F(x, y,t), 0 < x < L1 , 0 < y < L2 , t > 0,
u(x, y, 0) = w1 (x, y), ut (x, y, 0) = w2 (x, y),
u(0, y,t) = 0, u(L1 , y,t) = 0, u(x, 0,t) = 0, u(x, L2 ,t) = 0,
describing the vibrations of a membrane of length L1 and width L2 . Let us find the displacement
function u(x, y,t). Let L1 = L2 = π, ν = 1, F(x, y,t) = xy sint, w1 (x, y) = cos x sin y, and w2 (x, y) =
cos y sin x.
1◦ . We draw the domain D = {0 ≤ x ≤ π, 0 ≤ y ≤ π}:
Options->Grid: we generate the grid.

Options->Axes Limits: we change the size of the default domain, i.e., we put the new size
[0, 4] for the x-axis and [0, 4] for the y-axis, Apply, and Close.
rectangle-icon: we draw a rectangle with the aid of the mouse (click-drag); its name R1 is
introduced by MATLAB. The exact coordinates of the cursor appear in the top right of
the window. To check and correct our coordinates, we double-click on R1 and enter the
exact coordinates (0, 0, pi, pi).
2◦ . We specify the boundary conditions: Boundary->Boundary Mode, the Dirichlet conditions. By

double-clicking on the bottom horizontal line, we set the first boundary condition u(x, 0,t) =
0 by introducing h = 1 and r = 0 in the corresponding menu. Next, we specify the remaining
three boundary conditions.
File->Save As: we save this domain as M-file LinearWave2D.m.
3◦ . We specify the PDE: by selecting PDE->PDE Specification, we define the type of the PDE:
Hyperbolic (since deal with a linear wave equation) and set the coefficients: c=1.0, a=0.0,
f=x.*y.*sin(t), and d=1.
4◦ . We generate a triangular finite element mesh by selecting Mesh->Initialize Mesh.
triangle-inside-a-triangle icon: we can refine the initial mesh.
5◦ . We specify parameters for solving the linear hyperbolic equation: Solve->Parameters. Time
(the time range): 0:7; u(t0) (the initial condition): cos(x).*sin(y); u’(t0) (the second
initial condition): cos(y).*sin(x); Relative tolerance: 0.01, Absolute tolerance: 0.001.
Solve->Solve PDE: we obtain the numerical solution. By default, the solution appears as a
color-scale contour plot.
6◦ . We can set various options for generating the figure: Plot->Parameters, e.g., Height (3-D
plot), Plot in x-y grid, Color, colormap (gray), and Show Mesh.
By selecting the mesh and Edit->Current Object Properties, we can edit the plot ob-
tained (in the MATLAB graphics editor) and create the final plot shown in Fig. 24.6).
R
Time=7 Height: u
−1
4
3 4
2 3
2
1 1
0 0
Figure 24.6 The mesh plot of the solution of the linear nonhomogeneous wave equation (t = 7).
The PDE Toolbox provided by MATLAB allows one to describe and work with more com-
plicated geometries, which can consist of various solid objects. Various geometric models
can be constructed interactively (using pdetool GUI) or via M-files (using pdegeom and
pdebound).
We can devise a Constructive Solid Geometry (CSG) model of the geometry by drawing
solid objects that can overlap. By default, each object is assigned a unique name (e.g., C1,
R1), which can be changed. There are four types of solid objects: circle, polygon, rectangle,
and ellipse, which represent the set of points inside an object. It is possible to move, rotate,
cut, and paste selected objects.
The CSG models are described by the Geometry Description Matrix gd. The current
gd matrix can be viewed in the main workspace by selecting the Draw->Export Geometry
Description, Set Formula, Labels. Each column in the gd matrix corresponds to an
object in the CSG model.
By default, the resulting geometric model is the union of all objects. However, the solid
objects can be combined by typing a set formula (which is displayed in the GUI and can be
changed, e.g., R1-C1-C2). The resulting geometrical model is the set of points for which
the set formula evaluates to true.
A set formula sf is expressed with the set of variables (that correspond to the name of
each object) and the operators + * - (that correspond to the set operations: union, inter-
section, and set difference, respectively).
Example 24.8. Linearized Poisson–Boltzmann equation. Constructive solid geometry. Con-

sider the following boundary value problem for the two-dimensional linearized Poisson–Boltzmann
equation defined on an irregular domain D = R1 − Ω:
uxx + uyy = νu − F(x, y), (24.2.5.1)

2 −x 2 −x
u(x, −1) = x + 1 + e sin(−1), u(x, 1) = x + 1 + e sin(1), (24.2.5.2)
ux (−1, y) = 0, uy (1, y) = 0, (24.2.5.3)

u = 0.1, u = 2.0,
∂E1 ∂E2
(24.2.5.4)
where ν is a positive constant (e.g., ν = 1) and F(x, y) = x2 + y2 + e−x sin(y).

R1 is a rectangle, R1 = [−1, 1]× [−1, 1] with boundary ∂R1 . Ω is an arbitrary multiply connected
domain within R1 with boundary ∂Ω. In our case, it is the interior region of the two ellipses E1
and E2 defined by the equations (x + 0.35)2/0.252 + (y + 0.35)2/0.32 − 1 = 0 and (x − 0.35)2/0.22 +
(y − 0.35)2/0.22 − 1 = 0.
The linearized Poisson–Boltzmann equation is defined outside Ω but within the rectangle R1 .
Along the boundary ∂Ω, the Dirichlet boundary condition (24.2.5.4) is defined. According
to different applications, different boundary conditions can be imposed along the boundary of the
rectangle R1 ; e.g., we set the Dirichlet boundary conditions (24.2.5.2) and the Neumann boundary
conditions (24.2.5.3).
The linearized Poisson–Boltzmann equation (24.2.5.1) has been used in many applications, e.g.,
models describing bio-molecular processes and electrostatic interactions between colloidal particles
[e.g., see Fogolari, Zuccato, Esposito, and Viglino (1999)].
We solve this problem with the aid of the graphical user interface (GUI) in the Generic Scalar
mode (by selecting Options->Application->Generic Scalar).
1◦ . We draw the domain D = R1 − Ω = R1 − E1 − E2 :
• Options->Grid and Options->Axes Limits: we generate the grid.

• We draw the rectangle R1, [−1, −1] × [1, 1].
• We draw the the ellipse E1 with the parameters X-center, −0.35, Y-center, −0.35,
A-semiaxes, 0.25, B-semiaxes, 0.3, Rotation, 0.
• We draw the the ellipse E2 with the parameters X-center, 0.35, Y-center, 0.35, A-semi
axes, 0.2, B-semiaxes, 0.2, and Rotation.
• We draw the domain D = R1 − Ω = R1 − E1 − E2 by typing the set formula R1-E1-E2.
2◦ . We specify the boundary conditions:
• If we select Boundary->Boundary Mode, the boundaries will appear as red arrows, indicat-
ing the type of conditions (i.e., Dirichlet conditions) and the direction.
• By double-clicking on the bottom horizontal line, we set the first boundary condition
in (24.2.5.2), u(x, −1) = x2 + 1 + e−x sin(−1), by introducing h=1 as well as
r=x.ˆ2+1+exp(-x).*sin(-1) in the corresponding menu.
• We specify the remaining Dirichlet boundary conditions on the upper horizontal line by
typing h=1 and r=x.ˆ2+1+exp(-x).*sin(1), and on the two ellipses E1 and E2 by typing,
respectively, h=1, r=0.1 and h=1, r=2.0.
• We specify the two remaining Neumann boundary conditions by typing q=0 and g=0.
• By selecting File->Save As, we save this domain as M-file LineqrizedPBEqGeom2D.m.
3◦ . By selecting PDE->PDE Specification, we define the type of the PDE:
• Elliptic, c=-1.0, a=-1.0, f=-(x.ˆ2+y.ˆ2+exp(-x).*sin(y)).

R
Height: u
−1
1
0.5 1
0 0.5
0
−0.5 −0.5
−1 −1
Figure 24.7 The mesh plot of the solution of the linearized Poisson–Boltzmann equation.
4◦ . By selecting Mesh->Initialize Mesh, we generate a triangular finite element mesh.
• By selecting Mesh->Refine Mesh, we can refine the initial mesh (e.g., two times).
• By selecting Mesh->Jiggle Mesh, we can jiggle the mesh (for improving the triangle qual-
ity).
5◦ . By selecting Solve->Parameters, we specify some parameters for solving a linear PDE:
• Adaptive mode, Maximum number of triangles, Maximum number of refinements.

• Triangle selection method, Relative tolerance, Refinement method.
• By selecting Solve->Solve PDE, we obtain the numerical solution.
6◦ . By selecting Plot->Parameters, we can set various options for generating a figure, e.g., Height
(3-D plot), Color, colormap (gray), Show Mesh.
• By selecting the mesh and Edit->Current Object Properties, we can edit the plot ob-
tained and create the final plot shown in Fig. 24.7.
Example 24.9. Linear Poisson equation. Comparison of analytical and numerical solutions.
Consider the following boundary value problem for the two-dimensional linear Poisson equation
defined on the domain D = {0 ≤ x ≤ 1, 0 ≤ y ≤ 1}:
− (uxx + uyy ) = F(x, y), 0 < x < 1, 0 < y < 1, (24.2.5.5)

u(x, 0) = 0, u(x, 1) = 0, u(0, y) = 0, u(1, y) = 0, (24.2.5.6)
where F(x, y) = 2π2 sin(πx) sin(πy) and u(x, y) = sin(πx) sin(πy) is the analytical solution of this
As before (see the previous examples), we solve this problem with the aid of the graphical user
interface (GUI) and then compare the numerical and analytical solutions.
Height: u−sin(pi*x).*sin(pi*y)
−3
x 10
0.5
−0.5
−1
−1.5
−2
1
1
0.8
0.5 0.6
0.4
0.2
0 0
Figure 24.8 The difference between the numerical and analytical solutions of the linear Poisson
equation.
1◦ . We draw the domain D = {0 ≤ x ≤ 1, 0 ≤ y ≤ 1}.
2◦ . We specify the homogeneous Dirichlet boundary conditions.
3◦ . We specify the PDE: PDE->PDE Specification:
• Elliptic, c=1.0, a=0.0, f=2*piˆ2.*sin(pi*x).*sin(pi*y).
4◦ . We generate a triangular finite element mesh.
5◦ . We specify some parameters for solving a linear PDE and obtain the numerical solution.
6◦ . We plot the difference between the numerical and analytical solutions:
• Plot->Parameters, change the entry u into user entry in the Height (3-D plot) row and
write u-sin(pi*x).*sin(pi*y) into the corresponding field in the User entry column.
• We set various options for generating a figure (Height (3-D plot), Color, colormap
(gray), Show Mesh).
• By selecting the mesh and Edit->Current Object Properties, we can edit the plot ob-
tained and create the final plot shown in Fig. 24.8.
⊙ References for Section 24.2: R. D. Cook, D. S. Malkus, and M. E. Plesha (1989), R. D. Skeel and M.
Berzins (1990), W. E. Schiesser (1994), D. Harries (1998), F. Fogolari, P. Zuccato, G. Esposito, and P. Viglino
(1999), H. J. Lee and W. E. Schiesser (2004), W. E. Schiesser and G. W. Griffiths (2009).
R
Solution profile at tk= 0.004, 0.2, 0.4

1
t=0.004
0.9 t=0.2
t=0.4
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
Figure 24.9 Linear diffusion–convection equation: solution profiles at tk = 0.004, 0.2, 0.4.
24.3 Constructing Finite Difference Approximations

Now we show the helpful role of MATLAB when applying and developing various finite
difference approximations to construct numerical solutions of initial value problems and
initial-boundary value problems for linear PDEs (e.g., parabolic and hyperbolic equations)
[for details, see Li and Chen (2009), Mathews and Fink (1999), Yang, Cao, Chung, and
Morris (2005)].
To approximate a linear PDE by finite differences, we have to generate a mesh in a
domain D; e.g., D = {xL ≤ x ≤ xR , t0 ≤ t ≤ t f }.
We assume (for simplicity) that the sets of lines of the mesh are equally spaced and the
dependent variable in the given PDE is u(x,t). We write h and k for the line spacings and
define the mesh points as follows:
x j = xL + ( j − 1)h, tn = t0 + (n − 1)k,
where j = 1, . . . , NX + 1, n = 1, . . . , NT + 1, h = (xR − xL )/NX , and k = (t f − t0 )/NT .

We calculate approximations to the solution at these mesh points; these approximations
will be denoted by u j,n ≈ u(x j ,tn ). We approximate the derivatives in the given equation by
24.3.1 Explicit Finite Difference Solutions
Let us study some numerical methods, namely, explicit and implicit finite difference meth-
ods, for solving linear partial differential equations.
Example 24.10. Linear diffusion–convection equation. Forward difference method. Consider
the initial-boundary value problem for the linear diffusion–convection equation
ut = ν uxx − µux , u(x, 0) = sin(πx), u(xL ,t) = 0, u(xR ,t) = 0
in the domain D = {xL ≤ x ≤ xR , t0 ≤ t ≤ t f }, where xL = 0, xR = 1, t0 = 0, t f = 0.4, ν = 0.009,

µ = 0.4.
24.3. Constructing Finite Difference Approximations 1393
Let us generate the rectangular mesh

x j = xL + ( j − 1)h, tn = nk,
where j = 1, . . . , NX + 1, n = 1, . . . , NT , h = (xR − xL )/NX, and k = t f /NT . We denote the approxi-
mate solution of u(x,t) at the mesh point ( j, n) by u j,n .
In the forward difference method, the second derivative uxx is replaced by the central difference
approximation (CDA), and the first derivatives ut and ux are replaced by the forward difference
approximation (FWDA). The finite difference scheme for the linear heat equation has the form
u j,n+1 = u j,n + r(u j+1,n − 2u j,n + u j−1,n) − (k/h)µ(u j+1,n + u j,n ),
where r = νk/h2 .
In this explicit finite difference scheme, the unknown value u j,n+1 (at the (n + 1)st step) is deter-
mined from three known values u j−1,n, u j,n , and u j+1,n (at the nth step).
We construct an approximate numerical solution of the initial-boundary value problem by ap-
plying the forward difference method, and plot (see Fig. 24.9) the solution profiles at various times
(for example, tk = 0.004, 0.2, 0.4) as follows:
clear all; close all; nu=0.009; L=1; xL=0; xR=1; NX=100; t0=0;
tf=0.4; NT=100; h=(xR-xL)/NX; k=(tf-t0)/NT; r=nu*k/hˆ2; mu=0.4;
x=xL:h:xR; u=zeros(NX+1,NT); f=sin(pi*x/L); for n=1:NT t=n*k;
if n==1
for j=2:NX u(j,n)=f(j)+r*(f(j+1)-2*f(j)+f(j-1))-k/h*mu*(f(j+1)-f(j)); end;
u(1,n)=0; u(NX+1,n)=0;
else
for j=2:NX
u(j,n)=u(j,n-1)+r*(u(j+1,n-1)-2*u(j,n-1)+u(j-1,n-1))-...
k/h*mu*(u(j+1,n-1)-u(j,n-1));
end;
u(1,n)=0; u(NX+1,n)=0;
end
end; T=1*k:k:NT*k; figure(1);
colormap(gray); surf(x,T,u'); xlabel('x'); ylabel('y'); zlabel('u');
title('Numerical solution of the linear diffusion--convection equation');
figure(2);
hold on; plot(x,u(:,1),'k-',x,u(:,50),'k--',x,u(:,NT),'k:','LineWidth',2);
title('Solution profile at t_k= 0.004, 0.2, 0.4');
legend('t=0.004','t=0.2','t=0.4');
figure(3);
G=plot(x,u(:,1),'LineWidth',3,'erase','xor');
for j=2:NT set(G,'xdata',x,'ydata',u(:,j)); pause(0.1); end
title('Animation of solution'); xlabel('x'); ylabel('u(x,t_k)');
Example 24.11. Linear nonhomogeneous wave equation. Central difference method. Consider
the initial-boundary value problem for the linear nonhomogeneous wave equation
utt = c2 uxx + F(x,t), u(x, 0) = f (x), ut (x, 0) = g(x), u(xL ,t) = 0, u(xR ,t) = 0
describing the motion of a fixed string in the domain D = {xL ≤ x ≤ xR , t0 ≤ t ≤ t f }. We take
xL = 0, xR = 0.5, t0 = 0, t f = 1.7, c = 1/4, F(x,t) = x sint, f (x) = 0, and g(x) = sin(4πx).
In the explicit central difference method, each second derivative is replaced by the central dif-
ference approximation (CDA). The finite difference scheme for the linear nonhomogeneous wave
equation has the form
u j,n+1 = 2(1 − r)u j,n + r(u j+1,n + u j−1,n) − u j,n−1 + j sin(n)(ck)2 ,
R
Solution profile at t = 0.0212, 1.0625, 1.7

k
0.3
t=0.0212
t=0.0625
0.2 t=1.7
0.1
−0.1
−0.2
−0.3
−0.4
0 0.1 0.2 0.3 0.4 0.5
Figure 24.10 Linear nonhomogeneous wave equation: solution profiles at tk = 0.0212, 1.0625, 1.7.
where r = (ck/h)2 . In this finite difference scheme, we have one unknown value u j,n+1 that depends
explicitly on the four known values u j,n , u j+1,n, u j−1,n, u j,n−1 at the previous time steps n and n − 1.
To start the process, we have to know the values of u at the time steps n = 0 and n = 1. Thus, we can
define the initial conditions at these time steps: u j,0 = f (x j ) and u(x j , 0)t ≈ (u j,1 − u j,0 )/k = g(x j ),
u j,1 = f (x j ) + kg(x j ).
plying the explicit finite difference method and plot the numerical solution in D as follows:
clear all; close all; c=0.25; xL=0; xR=0.5; NX=60; t0=0; tf=1.7;
NT=80; h=(xR-xL)/NX; k=(tf-t0)/NT; r=(c*k/h)ˆ2; x=xL:h:xR;
u=zeros(NX+1,NT); f=0; g=sin(4*pi*x); for n=1:NT
t=n*k;
if n==1
for j=2:NX u(j,n)=f; end; u(1,n)=0; u(NX+1,n)=0;
elseif n==2
for j=2:NX u(j,n)=f+k*g(j); end; u(1,n)=0; u(NX+1,n)=0;
else
for j=2:NX
u(j,n)=2*(1-r)*u(j,n-1)+r*(u(j+1,n-1)+u(j-1,n-1))-u(j,n-2)+j*sin(n)*(c*k)ˆ2;
end; u(1,n)=0; u(NX+1,n)=0;
end
end; T=1*k:k:NT*k; figure(1);
colormap(gray); surf(x,T,u'); xlabel('x'); ylabel('y'); zlabel('u');
title('Numerical solution of linear nonhomogeneous wave equation');
figure(2);
title('Solution profile at t_k= 0.0212, 1.0625, 1.7');
legend('t=0.0212','t=0.0625','t=1.7');
figure(3);
for n=1:NT uMax=max(u(:,j)); uMin=min(u(:,j)); end
Nu=2; delta=(uMax-uMin)/Nu; axis([xL,xR,uMin-delta,uMax+delta]);
title('Animation of solution'); xlabel('x'); ylabel('u(x,t_k)');
24.3. Constructing Finite Difference Approximations 1395
24.3.2 Implicit Finite Difference Solutions

In this section, we consider another class of finite difference methods for solving linear
equations, implicit methods. In particular, we will construct a finite difference scheme
of the backward Euler method (or the backward time centered space method) for solving
linear heat equations.
The backward Euler method is an implicit one-stage method for finding solutions of
PDEs that are first order in time and arbitrary order in space (with mixed partial derivatives).
Example 24.12. Linear heat equation. Backward Euler method. Consider the initial-boundary
value problem for the linear heat equation
ut = ν uxx , u(x, 0) = sin(πx), u(xL ,t) = 0, u(xR ,t) = 0,
in the domain D = {xL ≤ x ≤ xR , t0 ≤ t ≤ t f }, where xL = 0, xR = 1, t0 = 0, t f = 0.1, and ν = 1.

Let us generate the rectangular mesh
x j = xL + ( j − 1)h, tn = nk,
where j = 1, . . . , NX + 1, n = 1, . . . , NT , h = (xR − xL )/NX, and k = t f /NT . We denote the approxi-

mate solution of u(x,t) at the mesh point ( j, n) by u j,n .
In the backward Euler method, the second derivative uxx is replaced by the central difference ap-
proximation (CDA), and the first derivative ut is replaced by the backward difference approximation
(BWDA). The finite difference scheme for the linear heat equation has the form
u j,n−1 = −ru j−1,n + (1 + 2r)u j,n − ru j+1,n,
where r = νk/h2 .
If the values of u0,n and uNX,n at both endpoints are given from the Dirichlet boundary con-
ditions, then the above finite difference equation can be transformed into the following system of
equations:
    
u1,n−1 + ru0,n 1 + 2r −r 0 · 0 0 u1,n
 u2,n−1   −r 1 + 2r −r · 0 0   
     u2,n 
 u   0 −r 1 + 2r · 0 0   u 
 3,n−1 =   3,n  .
 ·   · · · · ·  · 
    
 uNX−2,n−1   0 0 0 · 1 + 2r −r  uNX−2,n 
uNX−1,n−1 + ruNX,n 0 0 0 · −r 1 + 2r uNX−1,n

plying the forward difference method and plot (see Fig. 24.11) the solution profiles at various times
(for example, tk = 0, 0.049, 0.1) as follows:
clear all; close all; nu=1; f=inline('sin(pi*x)','x');

g1=inline('0'); g2=inline('0'); xL=0; xR=1; t0=0; tf=0.1; NX=25;
NT=100; h=(xR-xL)/NX; k=(tf-t0)/NT; x=[0:NX]'*h; t=[0:NT]*k; for
j=1:NX+1, u(j,1)=f(x(j)); end for n=1:NT+1, u([1
NX+1],n)=[g1(t(n)); g2(t(n))]; end; r=nu*k/h/h; for j=1:NX-1
A(j,j)=1+2*r; if j>1, A(j-1,j)=-r; A(j,j-1)=-r; end; end for
i=2:NT+1
b=[r*u(1,i); zeros(NX-3,1); r*u(NX+1,i)]+u(2:NX,i-1); u(2:NX,i)=A\b;
end figure(1);
colormap(gray); surf(x,t,u'); xlabel('x'); ylabel('y'); zlabel('u');
title('Numerical solution of linear heat equation');
R
Solution profile at t = 0, 0.049, 0.1

k
1
t=0
0.9 t=0.049
t=0.1
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
Figure 24.11 Linear heat equation: solution profiles at tk = 0, 0.049, 0.1.
figure(2);
title('Solution profile at t_k= 0, 0.049, 0.1');
legend('t=0','t=0.049','t=0.1');
figure(3);
for n=1:NT uMax=max(u(:,j)); uMin=min(u(:,j)); end
Nu=2; delta=(uMax-uMin)/Nu; axis([xL,xR,uMin-delta,uMax+delta]);
title('Animation of solution');
xlabel('x'); ylabel('u(x,t_k)');
⊙ References for Section 24.3: J. H. Mathews and K. D. Fink (1999), Yang, Cao, Chung, and Morris
(2005), J. Li and Y. T. Chen (2009).
24.4 Numerical Solutions of Systems of Linear PDEs

The MATLAB PDE Toolbox can also deal with systems of partial differential equations.
Let us consider systems of linear partial differential equations in one space dimension and
two space dimensions [for details, see Young (1971), Lapidus and Pinder (1999), Knabner
and Angerman (2003), Larsson and Thomée (2008), Li and Chen (2009)].
24.4.1 Linear Systems of 1D PDEs

The PDE solver pdepe (provided in MATLAB PDE Toolbox) can be applied for numer-
ically solving general one-dimensional partial differential equations. Now we consider
systems of linear partial differential equations in one space dimension. In this case, the
predefined function pdepe allows us to solve initial-boundary value problems for systems
of 1D parabolic-elliptic PDEs. There must be at least one parabolic equation in the system
24.4. Numerical Solutions of Systems of Linear PDEs 1397
(see help pdepe). The class of 1D parabolic-elliptic PDEs defined in D = {a ≤ x ≤ b,

t0 ≤ t ≤ t f } to which the function pdepe can be applied has the form

C(x,t)ut = x−m ∂x xm f(x,t, u, ux ) + s(x,t, u, ux ), (24.4.1.1)
where u is the unknown vector function that depends on the scalar space variable x and the
scalar time variable t; the flux function f and the source function s are vector functions; the
integer m ∈ {0, 1, 2} corresponds to slab, cylindrical, and spherical symmetry, respectively;
the function C is a diagonal matrix whose diagonal entries are zero or positive (which
corresponds to elliptic or parabolic equations, respectively).6
The initial condition at t = t0 and for a ≤ x ≤ b and a given function u0 is defined as
follows:
u(x,t0 ) = u0 (x). (24.4.1.2)
The boundary conditions at x = a and x = b and for t0 ≤ t ≤ t f have the form
p(a,t, u) + q(a,t)f(a,t, u, ux ) = 0, p(b,t, u) + q(b,t)f(b,t, u, ux ) = 0, (24.4.1.3)
where p and q are given vector functions.

Since the predefined function pdepe implements a second-order spatial discretization
method based on the xMesh values, it follows that the choice of xMesh is important and
can affect the accuracy and cost of the numerical solution (e.g., it is best to define closely
spaced xMesh points for domains where the solution can vary rapidly with respect to x).
The time points in [t0 ,t f ] at which the solution is obtained are given in the vector tSpan,
where tSpan(1)=t 0, tSpan(end)=t f, and tSpan(i)<tSpan(i+1) (for i=2,. . . ,end-1).
Since the time integration in pdepe is performed by the stiff ODE solver ode15s, the actual
time step values are chosen dynamically and do not affect the accuracy and cost.
Example 24.13. Linear system of 1D parabolic equations. Consider the following system of
linear one-dimensional partial differential equations:
ut = α1 uxx − g(u − v),

vt = α2 vxx + g(u − v),
where u(x,t) and v(x,t) are the unknown functions and α1 and α2 are real constants. The function
g(u − v) is defined as g(u − v) = (u − v)(δ1 − δ2 ), where δ1 and δ2 are real constants.
We find numerical and graphical solutions of this linear system under the following boundary
and initial conditions:
ux (0,t) = 0, v(0,t) = 0, u(L,t) = 1, vx (L,t) = 0,

u(x, 0) = 1, v(x, 0) = 0.
To apply the pdepe function, we denote (u, v) by (u1 , u2 ) and rewrite the linear system with
initial and boundary conditions in the form (24.4.1.1)–(24.4.1.3) as follows:

1 0 ∂ u1 ∂ α1 u1x −g(u1 − u2 )
= +
0 1 ∂t u2 ∂x α2 u2x g(u1 − u2)
6 According to this notation, there must be at least one parabolic equation.
R
and

1 0 α1 u1x −g(u1 − u2 )
m = 0, C(x,t) = , f(x,t, u, ux ) = , s(x,t, u, ux ) = .
0 1 α2 u2x g(u1 − u2)
Then we choose D = {0 ≤ x ≤ L, 0 ≤ t ≤ T } (L = 1, T = 2), α1 = 0.01, α2 = 0.5, δ1 = 5,

α2 = −5.
Creating three separate function M-files, SYS1.m, ICSYS1.m, and BCSYS1.m, we specify all param-
eters and functions as follows:
function [c,f,s]=SYS1(x,t,u,DuDx)
alpha1=0.01; alpha2=0.5; h=u(1)-u(2); delta1=5.; delta2=-5.;
g=delta1*h-delta2*h;
c=[1;1]; f=[alpha1*DuDx(1); alpha2*DuDx(2)]; s=g*[-1;1];
function u0=ICSYS1(x)
u0(1)=1; u0(2)=0; u0=[u0(1),u0(2)];
function [pL,qL,pR,qR]=BCSYS1(xL,uL,xR,uR,t)
pL=[0;uL(2)]; qL=[1;0]; pR=[uR(1)-1;0]; qR=[0;1];
Then, for constructing numerical and graphical solutions of this initial-boundary value problem,
we compose a script M-file SolSYS1 as follows:
sol=pdepe(m,@SYS1,@ICSYS1,@BCSYS1,x,t); u1=sol(:,:,1);
u2=sol(:,:,2); figure; subplot(2,1,1); colormap(gray);
surf(x,t,u1); rotate3d on; view(-39,27);
set(gca,'LineWidth',1); title('u_1(x,t)'); xlabel('x');ylabel('t');
subplot(2,1,2); surf(x,t,u2); rotate3d on; view(-45,25);
set(gca,'LineWidth',1); title('u_2(x,t)'); xlabel('x');ylabel('t');
figure; subplot(2,1,1);
plot(x,u1(2,:),'k-',x,u1(21,:),'k--',x,u1(Nt,:),'k:','LineWidth',2);
title('Solution profile at t_k=0.02,0.4,2');
xlabel('x'); ylabel('u1(x,t_k)'); legend('t=0.02','t=0.4','t=2');
subplot(2,1,2);
plot(x,u2(2,:),'k-',x,u2(21,:),'k--',x,u2(Nt,:),'k:','LineWidth',2);
title('Solution profile at t_k=0.02,0.4,2');
xlabel('x'); ylabel('u2(x,t_k)'); legend('t=0.02','t=0.4','t=2');
figure; subplot(2,1,1); G=plot(x,u1(2,:),'erase','xor');
for k=2:length(t) set(G,'xdata',x,'ydata',u1(k,:)); pause(0.2); end
title('Animation of u_1(x,t_k)'); xlabel('x'); ylabel('u1(x,t_k)');
subplot(2,1,2); G=plot(x,u2(2,:),'erase','xor');
for k=2:length(t) set(G,'xdata',x,'ydata',u2(k,:)); pause(0.2); end
title('Animation of u_2(x,t_k)'); xlabel('x'); ylabel('u2(x,t_k)');
echo off
In this case, we extract the solution u from sol as u1=sol(:,:,1), u2=sol(:,:,2). Then we
plot the solution as a surface (with interactive rotation), plot the surface profile at tk = 0.02, 0.4, 2,
and produce animations of the surface profile for t ∈ [0, 2]. Three surface profiles of the solution
obtained at times tk = 0.02, 0.4, 2 are shown in Fig. 24.12.
Solution profile at tk=0.02,0.4,2

1
t=0.02
t=0.4
t=2
u1(x,tk)
0.5
0
0 0.2 0.4 0.6 0.8 1
x
Solution profile at tk=0.02,0.4,2
0.8
t=0.02
0.6 t=0.4
t=2
u2(x,tk)
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
x
Figure 24.12 Surface profiles of the solution at tk = 0.02, 0.4, 2.
24.4.2 Linear Systems of 2D PDEs

Now consider systems of N linear elliptic PDEs in two space dimensions (2D) and time
over the domain D, i.e., systems of the form
−∇ · (c ⊗ ∇u) + au = f,
where c is an N × N × 2 × 2 tensor, a is an N × N matrix, and f and u are column vectors

of length N, respectively, describing the flux and unknown functions. In the MATLAB
notation, ∇ · (c ⊗ ∇u) means the N × 1 matrix with (i, 1)st entry
N
∂ci j11 ∂u j ∂ci j12 ∂u j ∂ci j21 ∂u j ∂ci j22 ∂u j
∑ ∂x ∂x + ∂x ∂y + ∂y ∂x + ∂y ∂y .
j=1
In general, the boundary conditions are mixed (i.e., for each point on the boundary there
is a combination of Dirichlet and generalized Neumann conditions) and have the form
hu = r, n · (c ⊗ ∇u) + qu = g + h′ µ,
where h is an M × N matrix (M ≥ 0) for M Dirichlet conditions, h′ µ is a source in the

generalized Neumann condition, and µ are the Lagrange multipliers. The data h, r, q, and g
are to be specified. In the MATLAB notation, n · (c ⊗ ∇u) means the N × 1 matrix with
(i, 1)st entry
N
∂u j ∂u j ∂u j ∂u j
∑ cos α ci j11 ∂x + cos α ci j12 ∂y + sin α ci j21 ∂x + sin α ci j22 ∂y ,
j=1
where the outward normal vector on the boundary is n = (cos α, sin α).
The generalized Neumann boundary condition is defined for M = 0, the Dirichlet bound-
ary condition is defined for M = N, and the mixed boundary condition is defined for
0 < M < N. These systems can be solved with the aid of PDE Toolbox.
R
Height: u Height: v
5 6
0 4
−5 2
−10 0
−15 −2
−20 −4
6 6
3 3
4 4
2 2
2 2
1 1
0 0 0 0
Figure 24.13 Surface plot of the solutions u1 (x, y) and u2 (x, y).
Consider the special case of the two-dimensional system of the form
− ∇ · (c11 ∇u1 ) − ∇ · (c12 ∇u2 ) + a11 u1 + a12 u2 = f1 ,

− ∇ · (c21 ∇u1 ) − ∇ · (c22 ∇u2 ) + a21 u1 + a22 u2 = f2 . (24.4.2.1)
The Dirichlet boundary conditions are
h11 u1 + h12 u2 = r1 , h21 u1 + h22 u2 = r2 . (24.4.2.2)
The generalized Neumann boundary conditions are
n · (c11 ∇u1 ) + n · (c12 ∇u2 ) + q11 u1 + q12 u2 = g1 ,

n · (c21 ∇u1 ) + n · (c22 ∇u2 ) + q21 u1 + q22 u2 = g2 .
The mixed boundary conditions are
h11 u1 + h12 u2 = r1 ,
n · (c11 ∇u1 ) + n · (c12 ∇u2 ) + q11 u1 + q12 u2 = g1 + h11 µ,
n · (c21 ∇u1 ) + n · (c22 ∇u2 ) + q21 u1 + q22 u2 = g2 + h12 µ.
Example 24.14. System of linear elliptic equations. Consider the boundary value problem for
two linear Poisson equations in two space variables:
(u1 )xx + (u1)yy = u2 + g1(x, y),

(u2 )xx + (u2)yy = −u1 + g2(x, y),
u1 (x, 0) = 0, u1 (x, 6) = 0, u1 (0, y) = 0, u1 (3, y) = cos(y),
u2 (x, 0) = 0, u2 (x, 6) = 0, u2 (0, y) = 0, u2 (3, y) = sin(y),
describing a potential field u(x, y) in a bounded domain D = {0 ≤ x ≤ 3, 0 ≤ y ≤ 6}. Let g1 (x, y) =

x2 + y2 and g2 (x, y) = x2 − y2 .
We solve this problem with the aid of the graphical user interface (GUI).
1◦ . We draw the domain D = {0 ≤ x ≤ 3, 0 ≤ y ≤ 6}.

• By selecting Options->Application->Generic System, we indicate that we work with a

generic system, i.e., a system of two equations (it is possible to work with systems of
arbitrary dimension by writing programs).
2◦ . We specify the boundary conditions for a system of two equations.
• When selecting Boundary->Boundary Mode, the boundaries and their directions will ap-
pear as red arrows. According to equations (24.4.2.2), we set the Dirichlet boundary
conditions (e.g., u1 (x, 0) = 0 by introducing h11 = 1, h12 = 0, h21 = 0, h22 = 1 and
r1 = r2 = 0).
• By selecting File->Save As, we save this domain as M-file LinSys2D.m.
3◦ . By selecting PDE->PDE Specification, we define the type of the PDE system:
• Elliptic, the coefficients of the PDE system: c11 = −1, c12 = c21 = 0, c22 = −1, a11 =
0, a12 = −1, a21 = 1, a22 = 0, f1 = x2 + y2 , and f2 = x2 − y2 , or, according to the
MATLAB syntax, x.ˆ 2+y.ˆ 2 (with the array operation .ˆ ).
4◦ . By selecting Mesh->Initialize Mesh, we generate a triangular finite element mesh.

5◦ . By selecting Solve->Parameters, we specify some parameters for solving a linear system.
• By selecting Solve->Solve PDE, we obtain the numerical solution. By default, the first
solution u1 (x, y) appears (or u, in MATLAB notation). By selecting Plot -> Parameters
-> Property (v), we plot the second solution u2 (x, y).
6◦ . By selecting Plot->Parameters and by setting various options (Height (3-D plot), Plot in
x-y grid, Color, colormap (gray), Show Mesh), we create the final figures shown in Fig. 24.13.
⊙ References for Section 24.4: D. M. Young (1971), L. Lapidus and G. F. Pinder (1999), P. Knabner and
L. Angerman (2003), S. Larsson and V. Thomée (2008), J. Li and Y. T. Chen (2009).
Part IV
Tables and
Supplements
Chapter 25
Elementary Functions
and Their Properties
⋆ Throughout Chapter 25 it is assumed that n is a positive integer unless otherwise spec-

ified.
25.1 Power, Exponential, and Logarithmic Functions

25.1.1 Properties of the Power Function
Basic properties of the power function:
xα xβ = xα+β , (x1 x2 )α = xα1 xα2 , (xα )β = xαβ ,
for any α and β, where x > 0, x1 > 0, x2 > 0.

Differentiation and integration formulas:
 α+1
Z x
α ′ α−1 α +C if α 6= −1,
(x ) = αx , x dx = α + 1

ln |x| + C if α = −1.
The Taylor series expansion in a neighborhood of an arbitrary point:

∞
X
xα = Cαn x0α−n (x − x0 )n for |x − x0 | < |x0 |,
n=0
α(α − 1) . . . (α − n + 1)
where Cαn = are binomial coefficients.
n!
25.1.2 Properties of the Exponential Function

Basic properties of the exponential function:
ax1 ax2 = ax1 +x2 , ax bx = (ab)x , (ax1 )x2 = ax1 x2 ,
where a > 0 and b > 0.
1405
1406 E LEMENTARY F UNCTIONS AND T HEIR P ROPERTIES
Number e, base of natural (Napierian) logarithms, and the function ex :

1 n x n
e = lim 1 + = 2.718281 . . . , ex = lim 1 + .
n→∞ n n→∞ n
The formula for passing from an arbitrary base a to the base e of natural logarithms:
ax = ex ln a .
The inequality
(
x1 > x2 if a > 1,
ax1 > ax2 ⇐⇒
x1 < x2 if 0 < a < 1.
The limit relations for any a > 1 and b > 0:

ax
lim = ∞, lim ax |x|b = 0.
x→+∞ |x|b x→−∞

Z
(ex )′ = ex , ex dx = ex + C;
Z
ax
(ax )′ = ax ln a, ax dx = + C.
ln a
The expansion in power series:
X xk ∞
x x2 x3 xn
ex = 1 + + + + ··· + + ··· = .
1! 2! 3! n! k!
k=0
25.1.3 Properties of the Logarithmic Function

By definition, the logarithmic function is the inverse of the exponential function. The
following equivalence relation holds:
y = loga x ⇐⇒ x = ay ,
where a > 0, a 6= 1.
Basic properties of the logarithmic function:
aloga x = x, loga (x1 x2 ) = loga x1 + loga x2 ,

logb x
loga (xk ) = k loga x, loga x = ,
logb a
where x > 0, x1 > 0, x2 > 0, a > 0, a 6= 1, b > 0, b 6= 1.
The simplest inequality:
(
x1 > x2 if a > 1,
loga x1 > loga x2 ⇐⇒
x1 < x2 if 0 < a < 1.
25.2. Trigonometric Functions 1407
For any b > 0, the following limit relations hold:
loga x
lim = 0, lim xb loga x = 0.
x→+∞ xb x→+0
The logarithmic function with the base e (base of natural logarithms or Napierian base)
is denoted by
loge x = ln x,
1 n
where e = lim 1 + = 2.718281 . . .
n→∞ n
Formula for passing from an arbitrary base a to the Napierian base e:
ln x
loga x = .
ln a

Z
′ 1
(ln x) = , ln x dx = x ln x − x + C.
x
Expansion in power series:
X∞
x2 x3 x4 x5 xk
ln(1 + x) = x − + − + − ··· = (−1)k−1 , |x| < 1;
2 3 4 5 k
k=1
X∞
x+1 2 2 2 2 1
ln = + 3 + 5 + 7 + ··· = 2 , |x| > 1;
x−1 x 3x 5x 7x (2k − 1)x2k−1
k=1

x−1 2 x−1 3 2 x−1 5 2 x−1 7
ln x = 2 + + + + ···
x+1 3 x+1 5 x+1 7 x+1
∞
X 1 x − 1 2k−1
=2 , x > 0.
2k − 1 x + 1
k=1
25.2 Trigonometric Functions

25.2.1 Simplest Relations
sin2 x + cos2 x = 1, tan x cot x = 1,

sin(−x) = − sin x, cos(−x) = cos x,
sin x cos x
tan x = , cot x = ,
cos x sin x
tan(−x) = − tan x, cot(−x) = − cot x,
1 1
1 + tan2 x = , 1 + cot2 x = .
cos2 x sin2 x
25.2.2 Reduction Formulas

sin(x ± 2nπ) = sin x, cos(x ± 2nπ) = cos x,
n
sin(x ± nπ) = (−1) sin x, cos(x ± nπ) = (−1)n cos x,
2n + 1 2n + 1
sin x ± π = ±(−1)n cos x, cos x ± π = ∓(−1)n sin x,
2 √ 2 √
π 2 π 2
sin x ± = (sin x ± cos x), cos x ± = (cos x ∓ sin x),
4 2 4 2
tan(x ± nπ) = tan x, cot(x ± nπ) = cot x,
2n + 1 2n + 1
tan x ± π = − cot x, cot x ± π = − tan x,
2 2
π tan x ± 1 π cot x ∓ 1
tan x ± = , cot x ± = ,
4 1 ∓ tan x 4 1 ± cot x
where n = 1, 2, . . .
25.2.3 Relations between Trigonometric Functions of Single

Argument
p tan x 1
sin x = ± 1 − cos2 x = ± √ = ±√ ,
1 + tan2 x 1 + cot2 x
p 1 cot x
cos x = ± 1 − sin2 x = ± √ = ±√ ,
1 + tan2 x 1 + cot2 x
√
sin x 1 − cos2 x 1
tan x = ± p =± = ,
1 − sin2 x cos x cot x
p
1 − sin2 x cos x 1
cot x = ± = ±√ = .
sin x 1 − cos2 x tan x
The sign before the radical is determined by the quarter in which the argument takes its
values.
25.2.4 Addition and Subtraction of Trigonometric Functions

x + y x − y
sin x + sin y = 2 sin cos ,
2 2
x − y x + y
sin x − sin y = 2 sin cos ,
2 2
x + y x−y
cos x + cos y = 2 cos cos ,
2 2
x + y x − y
cos x − cos y = −2 sin sin ,
2 2
sin2 x − sin2 y = cos2 y − cos2 x = sin(x + y) sin(x − y),
sin2 x − cos2 y = − cos(x + y) cos(x − y),
sin(x ± y) sin(y ± x)
tan x ± tan y = , cot x ± cot y = ,
cos x cos y sin x sin y
a cos x + b sin x = r sin(x + ϕ) = r cos(x − ψ).
25.2. Trigonometric Functions 1409
√
Here r = a2 + b2 , sin ϕ = a/r, cos ϕ = b/r, sin ψ = b/r, and cos ψ = a/r.
25.2.5 Products of Trigonometric Functions
sin x sin y = 12 [cos(x − y) − cos(x + y)],

cos x cos y = 12 [cos(x − y) + cos(x + y)],
sin x cos y = 12 [sin(x − y) + sin(x + y)].
25.2.6 Powers of Trigonometric Functions
cos2 x = 1
2 cos 2x + 12 , sin2 x = − 12 cos 2x + 12 ,
cos3 x = 1
4 cos 3x + 3
4 cos x, sin3 x = − 14 sin 3x + 3
4 sin x,
cos4 x = 1
8 cos 4x + 1
2 cos 2x + 38 , sin4 x = 1
8 cos 4x − 1
2 cos 2x + 38 ,
cos5 x = 1
16 cos 5x + 5
16 cos 3x + 5
8 cos x, sin5 x = 1
16 sin 5x − 5
16 sin 3x + 5
8 sin x,
n−1
X
2n 1 k 1 n
cos x= C2n cos[2(n − k)x] + C ,
22n−1 22n 2n
k=0
n
2n+1 1 X k
cos x = 2n C2n+1 cos[(2n − 2k + 1)x],
2
k=0
n−1
X
1 1 n
sin2n x = (−1)n−k C2n
k
cos[2(n − k)x] + C ,
22n−1 22n 2n
k=0
n
2n+1 1 X
sin x = 2n (−1)n−k C2n+1
k
sin[(2n − 2k + 1)x].
2
k=0
k m!
Here n = 1, 2, . . . and Cm = are binomial coefficients (0! = 1).
k! (m − k)!
25.2.7 Addition Formulas
sin(x ± y) = sin x cos y ± cos x sin y, cos(x ± y) = cos x cos y ∓ sin x sin y,
tan x ± tan y 1 ∓ tan x tan y
tan(x ± y) = , cot(x ± y) = .
1 ∓ tan x tan y tan x ± tan y
25.2.8 Trigonometric Functions of Multiple Arguments

cos 2x = 2 cos2 x − 1 = 1 − 2 sin2 x, sin 2x = 2 sin x cos x,
3
cos 3x = −3 cos x + 4 cos x, sin 3x = 3 sin x − 4 sin3 x,
cos 4x = 1 − 8 cos2 x + 8 cos4 x, sin 4x = 4 cos x (sin x − 2 sin3 x),
cos 5x = 5 cos x − 20 cos3 x + 16 cos5 x, sin 5x = 5 sin x − 20 sin3 x + 16 sin5 x,
Xn
n2 (n2 − 1) . . . [n2 − (k − 1)2 ]
cos(2nx) = 1 + (−1)k 4k sin2k x,
(2k)!
k=1
X n
cos[(2n + 1)x] = cos x 1 + (−1)k
k=1

[(2n + 1)2 − 1][(2n + 1)2 − 32 ] . . . [(2n + 1)2 − (2k − 1)2 ] 2k
× sin x ,
(2k)!
Xn 2 2 2 2 2
k k (n − 1)(n − 2 ) . . . (n − k ) 2k−1
sin(2nx) = 2n cos x sin x + (−1) 4 sin x ,
(2k − 1)!
k=1
Xn
sin[(2n + 1)x] = (2n + 1) sin x + (−1)k
k=1

[(2n + 1)2 − 1][(2n + 1)2 − 32 ] . . . [(2n + 1)2 − (2k − 1)2 ] 2k+1
× sin x ,
(2k + 1)!
2 tan x 3 tan x − tan3 x 4 tan x − 4 tan3 x
tan 2x = 2 , tan 3x = 2 , tan 4x = ,
1 − tan x 1 − 3 tan x 1 − 6 tan2 x + tan4 x
where n = 1, 2, . . .
25.2.9 Trigonometric Functions of Half Argument
x 1 − cos x x 1 + cos x
sin2 = , cos2 = ,
2 2 2 2
x sin x 1 − cos x x sin x 1 + cos x
tan = = , cot = = ,
2 1 + cos x sin x 2 1 − cos x sin x
2 tan x2 1 − tan2 x2 2 tan x2
sin x = , cos x = , tan x = .
1 + tan2 x2 1 + tan2 x2 1 − tan2 x2
25.2.10 Differentiation Formulas

d sin x d cos x d tan x 1 d cot x 1
= cos x, = − sin x, = , =− 2 .
dx dx dx cos2 x dx sin x
25.2.11 Integration Formulas

Z Z
sin x dx = − cos x + C, cos x dx = sin x + C,
Z Z
tan x dx = − ln | cos x| + C, cot x dx = ln | sin x| + C,

25.3. Inverse Trigonometric Functions 1411
25.2.12 Expansion in Power Series

x2 x4 x6 x2n
cos x = 1 − + − + · · · + (−1)n +· · · (|x| < ∞),
2! 4! 6! (2n)!
x3 x5 x7 x2n+1
sin x = x − + − + · · · + (−1)n +· · · (|x| < ∞),
3! 5! 7! (2n + 1)!
x3 2x5 17x7 22n (22n − 1)|B2n | 2n−1
tan x = x + + + +· · ·+ x +· · · (|x| < π/2),
3 15 315 (2n)!

1 x x3 2x5 22n |B2n | 2n−1
cot x = − + + +· · ·+ x +· · · (0 < |x| < π),
x 3 45 945 (2n)!
1 x2 5x4 61x6 (−1)n E2n 2n
= 1+ + + +· · ·+ x +· · · (|x| < π/2),
cos x 2 24 720 (2n)!
1 1 x 7x3 (−1)n−1 2(22n−1 − 1)B2n 2n−1
= + + +· · ·+ x +· · · (0 < |x| < π),
sin x x 6 360 (2n)!
where Bn and En are Bernoulli and Euler numbers (see Sections 30.1.3 and 30.1.4).
25.2.13 Representation in the Form of Infinite Products

x2 x2 x2 x2
sin x = x 1 − 2 1− 2 1 − 2 ... 1 − 2 2 ...
π 4π 9π n π
2 2 2
4x 4x 4x 4x2
cos x = 1 − 2 1− 2 1− ... 1 − ...
π 9π 25π 2 (2n + 1)2 π 2
25.2.14 Euler and de Moivre Formulas. Relationship with Hyperbolic

Functions
ey+ix = ey (cos x + i sin x), (cos x + i sin x)n = cos(nx) + i sin(nx), i2 = −1,
sin(ix) = i sinh x, cos(ix) = cosh x, tan(ix) = i tanh x, cot(ix) = −i coth x.
25.3 Inverse Trigonometric Functions

25.3.1 Definitions of Inverse Trigonometric Functions
Inverse trigonometric functions (arc functions) are the functions that are inverse to the
trigonometric functions. Since the trigonometric functions sin x, cos x, tan x, cot x are
periodic, the corresponding inverse functions, denoted by Arcsin x, Arccos x, Arctan x,
Arccot x, are multivalued. The following relations define the multivalued inverse trigono-
metric functions:
sin Arcsin x = x, cos Arccos x = x,

tan Arctan x = x, cot Arccot x = x.
These functions admit the following verbal definitions: Arcsin x is the angle whose sine is
equal to x; Arccos x is the angle whose cosine is equal to x; Arctan x is the angle whose
tangent is equal to x; Arccot x is the angle whose cotangent is equal to x.
The principal (single-valued) branches of the inverse trigonometric functions are de-
noted by
arcsin x ≡ sin−1 x (arcsine is the inverse of sine),
arccos x ≡ cos−1 x (arccosine is the inverse of cosine),
arctan x ≡ tan−1 x (arctangent is the inverse of tangent),
arccot x ≡ cot−1 x (arccotangent is the inverse of cotangent)
and are determined by the inequalities
− π2 ≤ arcsin x ≤ π
2, 0 ≤ arccos x ≤ π (−1 ≤ x ≤ 1);
− π2 < arctan x < π2 , 0 < arccot x < π (−∞ < x < ∞).
The following equivalent relations can be taken as definitions of single-valued inverse
trigonometric functions:
π π
y = arcsin x, −1≤x≤1 ⇐⇒ x = sin y, −≤y≤ ;
2 2
y = arccos x, −1≤x≤1 ⇐⇒ x = cos y, 0 ≤ y ≤ π;
π π
y = arctan x, − ∞ < x < +∞ ⇐⇒ x = tan y, − <y< ;
2 2
y = arccot x, − ∞ < x < +∞ ⇐⇒ x = cot y, 0 < y < π.
The multivalued and the single-valued inverse trigonometric functions are related by
the formulas
Arcsin x = (−1)n arcsin x + πn,
Arccos x = ± arccos x + 2πn,
Arctan x = arctan x + πn,
Arccot x = arccot x + πn,
where n = 0, ±1, ±2, . . .
25.3.2 Simplest Formulas

sin(arcsin x) = x, cos(arccos x) = x,
tan(arctan x) = x, cot(arccot x) = x.
25.3.3 Some Properties

arcsin(−x) = − arcsin x, arccos(−x) = π − arccos x,
arctan(−x) = − arctan x, arccot(−x) = π − arccot x,
(
x − 2nπ if 2nπ − π2 ≤ x ≤ 2nπ + π2 ,
arcsin(sin x) =
−x + 2(n + 1)π if (2n + 1)π − π2 ≤ x ≤ 2(n + 1)π + π2 ,
(
x − 2nπ if 2nπ ≤ x ≤ (2n + 1)π,
arccos(cos x) =
−x + 2(n + 1)π if (2n + 1)π ≤ x ≤ 2(n + 1)π,
25.3. Inverse Trigonometric Functions 1413
π π
arctan(tan x) = x − nπ if nπ − 2 < x < nπ + 2,
arccot(cot x) = x − nπ if nπ < x < (n + 1)π.
25.3.4 Relations between Inverse Trigonometric Functions
arcsin x + arccos x = π2 , arctan x + arccot x = π2 ;

 √

 arccos 1 − x2 if 0 ≤ x ≤ 1,

 √

 2
− arccos 1 − x if −1 ≤ x ≤ 0,
arcsin x = arctan √ x
 if −1 < x < 1,

 √ 1 − x2

 2
arccot 1 − x − π if −1 ≤ x < 0;

 x
√

 arcsin 1 − x2 if 0 ≤ x ≤ 1,

 √

 2
π − arcsin
√
1 − x if −1 ≤ x ≤ 0,
arccos x = 1 − x2

 arctan if 0 < x ≤ 1,

 x

 x
arccot √ if −1 < x < 1;
1 − x2
 x

 arcsin √ for any x,

 1 + x2



 1
arccos √ if x ≥ 0,
arctan x = 1 + x2
 1

 − arccos √ if x ≤ 0,

 1 + x2


arccot 1

if x > 0;
 x
 1

 arcsin √ if x > 0,

 1 + x2


π − arcsin √ 1

if x < 0,
arccot x = 1 + x2

 1

 arctan if x > 0,

 x

π + arctan 1

if x < 0.
x
25.3.5 Addition and Subtraction of Inverse Trigonometric Functions
p p
arcsin x + arcsin y = arcsin x 1 − y 2 + y 1 − x2 for x2 + y 2 ≤ 1,
p
arccos x ± arccos y = ± arccos xy ∓ (1 − x2 )(1 − y 2 ) for x ± y ≥ 0,
x+y
arctan x + arctan y = arctan for xy < 1,
1 − xy
x−y
arctan x − arctan y = arctan for xy > −1.
1 + xy

d 1 d 1
arcsin x = √ , arccos x = − √ ,
dx 1 − x2 dx 1 − x2
d 1 d 1
arctan x = , arccot x = − .
dx 1 + x2 dx 1 + x2

Z p
arcsin x dx = x arcsin x + 1 − x2 + C,
Z p
arccos x dx = x arccos x − 1 − x2 + C,
Z
1
arctan x dx = x arctan x −ln(1 + x2 ) + C,
2
Z
1
arccot x dx = x arccot x + ln(1 + x2 ) + C,
2
1 x3 1 × 3 x5 1 × 3 × 5 x7
arcsin x = x + + + + ···
2 3 2×4 5 2×4×6 7
1 × 3 × · · · × (2n − 1) x2n+1
+ + ··· (|x| < 1),
2 × 4 × · · · × (2n) 2n + 1
x3 x5 x7 x2n−1
arctan x = x − + − + · · · + (−1)n−1 + ··· (|x| ≤ 1),
3 5 7 2n − 1
π 1 1 1 1
arctan x = − + 3 − 5 + · · · + (−1)n + · · · (|x| > 1).
2 x 3x 5x (2n − 1)x2n−1
The expansions for arccos x and arccot x can be obtained from the relations arccos x =
π π
2 − arcsin x and arccot x = 2 − arctan x.
25.4 Hyperbolic Functions

25.4.1 Definitions of Hyperbolic Functions
Hyperbolic functions are defined in terms of the exponential functions as follows:
ex − e−x ex + e−x
sinh x = , cosh x = ,
2 2
ex − e−x ex + e−x
tanh x = x , coth x = x .
e + e−x e − e−x
25.4. Hyperbolic Functions 1415

cosh2 x − sinh2 x = 1, tanh x coth x = 1,
sinh(−x) = − sinh x, cosh(−x) = cosh x,
sinh x cosh x
tanh x = , coth x = ,
cosh x sinh x
tanh(−x) = − tanh x, coth(−x) = − coth x,
1 1
1 − tanh2 x = , coth2 x − 1 = .
cosh2 x sinh2 x
25.4.3 Relations between Hyperbolic Functions of Single Argument

(x ≥ 0)
p tanh x 1
sinh x = cosh2 x − 1 = p = p ,
2
1 − tanh x coth2 x − 1
p 1 coth x
cosh x = sinh2 x + 1 = p = p ,
2
1 − tanh x coth2 x − 1
p
sinh x cosh2 x − 1 1
tanh x = p = = ,
2
sinh x + 1 cosh x coth x
p
sinh2 x + 1 cosh x 1
coth x = = p = .
sinh x cosh2 x − 1 tanh x
25.4.4 Addition and Subtraction of Hyperbolic Functions

x+y x − y
sinh x + sinh y = 2 sinh cosh ,
2 2
x−y x + y
sinh x − sinh y = 2 sinh cosh ,
2 2
x + y x − y
cosh x + cosh y = 2 cosh cosh ,
2 2
x + y x − y
cosh x − cosh y = 2 sinh sinh ,
2 2
sinh2 x − sinh2 y = cosh2 x − cosh2 y = sinh(x + y) sinh(x − y),
sinh2 x + cosh2 y = cosh(x + y) cosh(x − y),
(cosh x ± sinh x)n = cosh(nx) ± sinh(nx),
sinh(x ± y) sinh(x ± y)
tanh x ± tanh y = , coth x ± coth y = ± ,
cosh x cosh y sinh x sinh y
where n = 0, ±1, ±2, . . .
25.4.5 Products of Hyperbolic Functions

sinh x sinh y = 12 [cosh(x + y) − cosh(x − y)],
cosh x cosh y = 12 [cosh(x + y) + cosh(x − y)],
sinh x cosh y = 12 [sinh(x + y) + sinh(x − y)].
25.4.6 Powers of Hyperbolic Functions
cosh2 x = 1 1
2 cosh 2x + 2 , sinh2 x = 12 cosh 2x − 12 ,
cosh3 x = 1 3
4 cosh 3x + 4 cosh x, sinh3 x = 14 sinh 3x − 34 sinh x,
cosh4 x = 1 1 3
8 cosh 4x + 2 cosh 2x + 8 , sinh4 x = 18 cosh 4x − 12 cosh 2x + 38 ,
cosh5 x = 1 5 5
16 cosh 5x + 16 cosh 3x + 8 cosh x, sinh5 x = 16
1
sinh 5x − 165
sinh 3x + 58 sinh x,
n−1
1 X k 1 n
cosh2n x = 2n−1 C2n cosh[2(n − k)x] + 2n C2n ,
2 2
k=0
n
1 X k
cosh2n+1 x = 2n C2n+1 cosh[(2n − 2k + 1)x],
2
k=0
n−1
1 X (−1)n n
sinh2n x = 2n−1 (−1)k C2n
k
cosh[2(n − k)x] + 2n C2n ,
2 2
k=0
n
1 X
sinh2n+1 x = 2n (−1)k C2n+1
k
sinh[(2n − 2k + 1)x].
2
k=0
k are binomial coefficients.

Here n = 1, 2, . . . and Cm
25.4.7 Addition Formulas
sinh(x ± y) = sinh x cosh y ± sinh y cosh x,

cosh(x ± y) = cosh x cosh y ± sinh x sinh y,
tanh x ± tanh y coth x coth y ± 1
tanh(x ± y) = , coth(x ± y) = .
1 ± tanh x tanh y coth y ± coth x
25.4.8 Hyperbolic Functions of Multiple Argument
cosh 2x = 2 cosh2 x − 1, sinh 2x = 2 sinh x cosh x,

cosh 3x = −3 cosh x + 4 cosh3 x, sinh 3x = 3 sinh x + 4 sinh3 x,
cosh 4x = 1 − 8 cosh2 x + 8 cosh4 x, sinh 4x = 4 cosh x(sinh x + 2 sinh3 x),
cosh 5x = 5 cosh x − 20 cosh3 x + 16 cosh5 x, sinh 5x = 5 sinh x + 20 sinh3 x + 16 sinh5 x.
[n/2]
n−1 n n X (−1)k+1 k−2
cosh(nx) = 2 cosh x + Cn−k−2 2n−2k−2 (cosh x)n−2k−2 ,
2 k+1
k=0
[(n−1)/2]
X
sinh(nx) = sinh x 2n−k−1 Cn−k−1
k
(cosh x)n−2k−1 .
k=0
k are binomial coefficients and [A] stands for the integer part of the number A.
Here Cm
25.4. Hyperbolic Functions 1417
25.4.9 Hyperbolic Functions of Half Argument

r r
x cosh x − 1 x cosh x + 1
sinh = sign x , cosh = ,
2 2 2 2
x sinh x cosh x − 1 x sinh x cosh x + 1
tanh = = , coth = = .
2 cosh x + 1 sinh x 2 cosh x − 1 sinh x

d sinh x d cosh x
= cosh x, = sinh x,
dx dx
d tanh x 1 d coth x 1
= , =− .
dx cosh2 x dx sinh2 x

Z Z
sinh x dx = cosh x + C, cosh x dx = sinh x + C,
Z Z
tanh x dx = ln cosh x + C, coth x dx = ln | sinh x| + C,

x2 x4 x6 x2n
cosh x = 1+ + + +· · ·+ +· · · (|x| < ∞),
2! 4! 6! (2n)!
x3 x5 x7 x2n+1
sinh x = x+ + + +· · ·+ +· · · (|x| < ∞),
3! 5! 7! (2n+1)!
x3 2x5 17x7 22n (22n −1)|B2n |x2n−1
tanh x = x− + − +· · ·+(−1)n−1 +· · · (|x| < π/2),
3 15 315 (2n)!
1 x x3 2x5 22n |B2n |x2n−1
coth x = + − + −· · ·+(−1)n−1 +· · · (|x| < π),
x 3 45 945 (2n)!
1 x2 5x4 61x6 E2n 2n
= 1− + − +· · ·+ x +· · · (|x| < π/2),
cosh x 2 24 720 (2n)!
1 1 x 7x3 31x5 2(22n−1 −1)B2n 2n−1
= − + − +· · ·+ x +· · · (0 < |x| < π),
sinh x x 6 360 15120 (2n)!
where Bn and En are Bernoulli and Euler numbers (see Sections 30.1.3 and 30.1.4).
25.4.13 Representation in the Form of Infinite Products

x2 x2 x2 x2
sinh x = x 1 + 2 1+ 2 1 + 2 ... 1 + 2 2 ...
π 4π 9π n π
2
2
2

4x 4x 4x 4x2
cosh x = 1 + 2 1+ 2 1+ ... 1 + ...
π 9π 25π 2 (2n + 1)2 π 2
25.4.14 Relationship with Trigonometric Functions

sinh(ix) = i sin x, cosh(ix) = cos x, i2 = −1,
tanh(ix) = i tan x, coth(ix) = −i cot x.
25.5 Inverse Hyperbolic Functions

25.5.1 Definitions of Inverse Hyperbolic Functions
Inverse hyperbolic functions are the functions that are inverse to hyperbolic functions. The
following notation is used for inverse hyperbolic functions:
arcsinh x ≡ sinh−1 x (inverse of hyperbolic sine),
−1
arccosh x ≡ cosh x (inverse of hyperbolic cosine),
−1
arctanh x ≡ tanh x (inverse of hyperbolic tangent),
arccoth x ≡ coth−1 x (inverse of hyperbolic cotangent).
Inverse hyperbolic functions can be expressed in terms of logarithmic functions:
p p
arcsinh x = ln x + x2 + 1 (x is any); arccosh x = ln x + x2 − 1 (x ≥ 1);
1 1+x 1 x+1
arctanh x = ln (|x| < 1); arccoth x = ln (|x| > 1).
2 1−x 2 x−1
Here only one (principal) branch of the function arccosh x is listed, the function itself being
double-valued. In order to write out both branches of arccosh x, the symbol ± should be
placed before the logarithm on the right-hand side of the formula.

arcsinh(−x) = − arcsinh x, arctanh(−x) = − arctanh x, arccoth(−x) = − arccoth x.
25.5.3 Relations between Inverse Hyperbolic Functions

√ x
arcsinh x = arccosh x2 + 1 = arctanh √ ,
2
x +1
√
√ x2 − 1
2
arccosh x = arcsinh x − 1 = arctanh ,
x
x 1 1
arctanh x = arcsinh √ = arccosh √ = arccoth .
1−x 2 1−x 2 x
25.5.4 Addition and Subtraction of Inverse Hyperbolic Functions

p p
arcsinh x ± arcsinh y = arcsinh x 1 + y 2 ± y 1 + x2 ,
p
arccosh x ± arccosh y = arccosh xy ± (x2 − 1)(y 2 − 1) ,
p
arcsinh x ± arccosh y = arcsinh xy ± (x2 + 1)(y 2 − 1) ,
x±y xy ± 1
arctanh x ± arctanh y = arctanh , arctanh x ± arccoth y = arctanh .
1 ± xy y±x
25.5. Inverse Hyperbolic Functions 1419

d 1 d 1
arcsinh x = √ , arccosh x = √ ,
dx 2
x +1 dx 2
x −1
d 1 d 1
arctanh x = (x2 < 1), arccoth x = (x2 > 1).
dx 1 − x2 dx 1 − x2

Z p
arcsinh x dx = x arcsinh x − 1 + x2 + C,
Z p
arccosh x dx = x arccosh x − x2 − 1 + C,
Z
1
arctanh x dx = x arctanh x +ln(1 − x2 ) + C,
2
Z
1
arccoth x dx = x arccoth x + ln(x2 − 1) + C,
2

1 x3 1 × 3 x5
arcsinh x = x − + −···
2 3 2×4 5
1 × 3 × · · · × (2n − 1) x2n+1
+ (−1)n +··· (|x| < 1),
2 × 4 × · · · × (2n) 2n + 1
1 1 1×3 1
arcsinh x = ln(2x) + 2
+ +···
2 2x 2 × 4 4x4
1 × 3 × · · · × (2n − 1) 1
+ +··· (|x| > 1),
2 × 4 × · · · × (2n) 2nx2n
1 1 1×3 1
arccosh x = ln(2x) − 2
− −···
2 2x 2 × 4 4x4
1 × 3 × · · · × (2n − 1) 1
− −··· (|x| > 1),
2 × 4 × · · · × (2n) 2nx2n
x3 x5 x7 x2n+1
arctanh x = x + + + +···+ +··· (|x| < 1),
3 5 7 2n + 1
1 1 1 1 1
arccoth x = + 3 + 5 + 7 + · · · + +··· (|x| > 1).
x 3x 5x 7x (2n + 1)x2n+1
⊙ References for Chapter 25: M. Abramowitz and I. A. Stegun (1964), A. P. Prudnikov, Yu. A. Brychkov,
and O. I. Marichev (1986), D. G. Zill and J. M. Dewar (1990), M. Kline (1998), R. Courant and F. John
(1999), I. S. Gradshteyn and I. M. Ryzhik (2000), G. A. Korn and T. M. Korn (2000), C. H. Edwards and
D. Penney (2002), D. Zwillinger (2002), E. W. Weisstein (2003), I. N. Bronshtein and K. A. Semendyayev
(2004), M. Sullivan (2004), H. Anton, I. Bivens, and S. Davis (2005), R. Adams (2006).
Chapter 26
Finite Sums and Infinite Series
26.1 Finite Numerical Sums

26.1.1 Progressions
Arithmetic progression:
n−1
X bn(n − 1)
1. (a + bk) = an + .
2
k=0
Geometric progression:
Xn
qn − 1
2. aq k−1 = a .
q−1
k=1
Arithmetic-geometric progression:
n−1
X a(1 − q n ) − b(n − 1)q n bq(1 − q n−1 )
3. (a + bk)q k = + .
1−q (1 − q)2
k=0
P
26.1.2 Sums of Powers of Natural Numbers Having the Form km
n
X n(n + 1)
1. k= .
2
k=1
n
X 1
2. k2 = n(n + 1)(2n + 1).
6
k=1
n
X 1 2
3. k3 = n (n + 1)2 .
4
k=1
n
X 1
4. k4 = n(n + 1)(2n + 1)(3n2 + 3n − 1).
30
k=1
n
X 1 2
5. k5 = n (n + 1)2 (2n2 + 2n − 1).
12
k=1
Xn
nm+1 nm 1 1 1 3 1 5
6. km = + + Cm B2 nm−1 + Cm B4 nm−3 + Cm B6 nm−5 + · · · .
m+1 2 2 4 6
k=1
1421
1422 F INITE S UMS AND I NFINITE S ERIES
Here Cm k are binomial coefficients and B are Bernoulli numbers (see Section 30.1.3); the
2k
last term in the sum contains n or n2 .
P
26.1.3 Alternating Sums of Powers of Natural Numbers, (−1)k km
n
X hn − 1i
1. (−1)k k = (−1)n ; [m] stands for the integer part of m.
2
k=1
n
X n(n + 1)
2. (−1)k k2 = (−1)n .
2
k=1
Xn
1
3. (−1)k k3 = 1 + (−1)n (4n3 + 6n2 − 1) .
8
k=1
n
X 1
4. (−1)k k4 = (−1)n (n4 + 2n3 − n).
2
k=1
Xn
1
5. (−1)k k5 = −1 + (−1)n (2n5 + 5n4 − 5n2 + 1) .
4
k=1
26.1.4 Other Sums Containing Integers

n
X
1. (2k + 1) = (n + 1)2 .
k=0
Xn
1
2. (2k + 1)2 = (n + 1)(2n + 1)(2n + 3).
3
k=0
n
X 1
3. k(k + 1) = n(n + 1)(n + 2).
3
k=1
Xn
1
4. (k + a)(k + b) = n(n + 1)(2n + 1 + 3a + 3b) + nab.
6
k=1
n
X
5. k k! = (n + 1)! − 1.
k=1
Xn
6. (−1)k (2k + 1) = (−1)n (n + 1).
k=0
n
X 1
7. (−1)k (2k + 1)2 = 2(−1)n (n + 1)2 − 1 + (−1)n .
2
k=0
26.1.5 Sums Containing Binomial Coefficients

◆ Throughout Section 26.1.5, it is assumed that m = 1, 2, 3, . . .
26.1. Finite Numerical Sums 1423
n
X
1. Cnk = 2n .
k=0
n
X
m m+1
2. Cm+k = Cn+m+1 .
k=0
n
X
3. (−1)k Cm
k
= (−1)n Cm−1
n
.
k=0
n
X
4. (k + 1)Cnk = 2n−1 (n + 2).
k=0
n
X
5. (−1)k+1 kCnk = 0.
k=1
n
X Xn
(−1)k+1 k 1
6. Cn = .
k m=1
m
k=1
n
X (−1)k+1 k n
7. Cn = .
k+1 n+1
k=1
n
X 1 2n+1 − 1
8. Cnk = .
k+1 n+1
k=0
n
X ak+1 k (a + 1)n+1 − 1
9. Cn = .
k+1 n+1
k=0
Xp
p
10. Cnk Cm
p−k
= Cn+m ; m and p are natural numbers.
k=0
n−p
X (2n)!
11. Cnk Cnp+k = .
(n − p)! (n + p)!
k=0
n
X
12. (Cnk )2 = C2n
n
.
k=0
X2n
13. (−1)k (C2n
k 2
) = (−1)n C2n
n
.
k=0
2n+1
X
14. (−1)k (C2n+1
k
)2 = 0.
k=0
n
X (2n − 1)!
15. k(Cnk )2 = .
[(n − 1)!]2
k=1
26.1.6 Other Numerical Sums

n−1
X πk π
1. sin = cot .
n 2n
k=1
n
X πk n m 1
2. sin2m = 2m C2m + , m < 2n.
2n 2 2
k=1
n−1
X πk 1
3. (−1)k cosm = 1 − (−1)m+n , m = 0, 1, . . . , n − 1.
n 2
k=0
n−1
X πk n
4. (−1)k cosn = n−1 .
n 2
k=0
26.2 Finite Functional Sums

26.2.1 Sums Involving Hyperbolic Functions
n−1
X
n−1 sinh(nx/2)
1. sinh(kx + a) = sinh x+a .
2 sinh(x/2)
k=0
n−1
X
n−1 sinh(nx/2)
2. cosh(kx + a) = cosh x+a .
2 sinh(x/2)
k=0
n−1
X
k 1 x n 2n − 1
3. (−1) sinh(kx+a) = sinh a− +(−1) sinh x+a .
2 cosh(x/2) 2 2
k=0
n−1
X
k 1 x n 2n − 1
4. (−1) cosh(kx+a) = cosh a− +(−1) cosh x+a .
2 cosh(x/2) 2 2
k=0
n−1
X 1 n
5. k sinh(kx + a) = − n sinh[(n − 1)x + a]
k=1
sinh2 (x/2) o
− (n − 1) sinh(nx + a) − sinh a .
n−1
X 1 n
6. k cosh(kx + a) = − n cosh[(n − 1)x + a]
k=1
sinh2 (x/2) o
− (n − 1) cosh(nx + a) − cosh a .
n−1
X 1 n
7. (−1)k k sinh(kx + a) = 2 (−1)n−1 n sinh[(n − 1)x + a]
k=1
cosh (x/2) o
+ (−1)n−1 (n − 1) sinh(nx + a) − sinh a .
n−1
X 1 n
8. (−1)k k cosh(kx + a) = 2 (−1)n−1 n cosh[(n − 1)x + a]
k=1
cosh (x/2) o
+ (−1)n−1 (n − 1) cosh(nx + a) − cosh a .
Xn
k n n x nx
9. Cn sinh(kx + a) = 2 cosh sinh +a .
2 2
k=0
X n
k n n x nx
10. Cn cosh(kx + a) = 2 cosh cosh +a .
2 2
k=0
26.2. Finite Functional Sums 1425
n−1
X a sinh x − an sinh(nx) + an+1 sinh[(n − 1)x]
11. ak sinh(kx) = .
1 − 2a cosh x + a2
k=1
n−1
X 1 − a cosh x − an cosh(nx) + an+1 cosh[(n − 1)x]
12. ak cosh(kx) = .
1 − 2a cosh x + a2
k=0
n
X 1 x 1 x
13. k
tanh k = coth x − n coth n .
2 2 2 2
k=1
n−1
X
14. 2k tanh(2k x) = 2n coth(2n x) − coth x.
k=0
26.2.2 Sums Involving Trigonometric Functions

n
X
1. sin(2kx) = sin[(n + 1)x] sin(nx) cosec x.
k=1
Xn
2. cos(2kx) = sin[(n + 1)x] cos(nx) cosec x.
k=0
Xn
3. sin[(2k − 1)x] = sin2 (nx) cosec x.
k=1
n
X
4. cos[(2k − 1)x] = sin(nx) cos(nx) cosec x.
k=1
n−1
X
n−1 nx x
5. sin(kx + a) = sin x + a sin cosec .
2 2 2
k=0
n−1
X
n−1 nx x
6. cos(kx + a) = cos x + a sin cosec .
2 2 2
k=0
2n−1
X
k 2n − 1 x
7. (−1) cos(kx + a) = sin x + a sin(nx) sec .
2 2
k=0
n
X sin(2nx)
8. (−1)k+1 sin[(2k − 1)x] = (−1)n+1 .
2 cos x
k=1
n
X 1 cos[(2n + 1)x]
9. (−1)k cos(2kx) = − + (−1)n .
2 2 cos x
k=1
Xn
n cos[(n + 1)x] sin(nx)
10. sin2 (kx) = − .
2 2 sin x
k=1
n
X n cos[(n + 1)x] sin(nx)
11. cos2 (kx) = + .
2 2 sin x
k=1
n−1
X sin(2nx) n cos[(2n − 1)x]
12. k sin(2kx) = 2 − .
4 sin x 2 sin x
k=1
n−1
X n sin[(2n − 1)x] 1 − cos(2nx)
13. k cos(2kx) = − .
k=1
2 sin x 4 sin2 x
n−1
X a sin x − an sin(nx) + an+1 sin[(n − 1)x]
14. ak sin(kx) = .
1 − 2a cos x + a2
k=1
n−1
X 1 − a cos x − an cos(nx) + an+1 cos[(n − 1)x]
15. ak cos(kx) = .
1 − 2a cos x + a2
k=0
Xn
k n n x nx
16. Cn sin(kx + a) = 2 cos sin +a .
2 2
k=0
Xn
k n n x nx
17. Cn cos(kx + a) = 2 cos cos +a .
2 2
k=0
Xn
k k n n x nx πn
18. (−1) Cn sin(kx + a) = (−2) sin sin + +a .
2 2 2
k=0
Xn
x nx πn
19. (−1)k Cnk cos(kx + a) = (−2)n sinn cos + +a .
2 2 2
k=0
Xn 2 2
k 2 x 2 x
20. 2 sin k = 2 sin n − sin2 x.
n
2 2
k=1
n
X 1 x 1 x
21. k
tan k = n cot n − 2 cot(2x).
2 2 2 2
k=0
26.3 Infinite Numerical Series

26.3.1 Progressions
∞
X a
1. aq k = , |q| < 1.
1−q
k=0
∞
X a bq
2. (a + bk)q k = + , |q| < 1.
1−q (1 − q)2
k=0
26.3.2 Other Numerical Series

∞
X (−1)n
1. = ln 2.
n+1
n=0
X∞
(−1)n π
2. = .
2n + 1 4
n=0
X∞
1
3. = 1.
n=1
n(n + 1)
26.3. Infinite Numerical Series 1427
X∞
(−1)n
4. = 1 − 2 ln 2.
n=1
n(n + 1)
∞
X 1 3
5. = .
n=1
n(n + 2) 4
X∞
(−1)n 1
6. =− .
n=1
n(n + 2) 4
X∞
1 1
7. = .
(2n − 1)(2n + 1) 2
n=1
X∞
1 π2
8. 2
= .
n 6
n=1
X∞
(−1)n+1 π2
9. = .
n2 12
n=1
X∞
1 π2
10. = .
(2n − 1)2 8
n=1
X∞
1 π 1
11. = coth(πa) − 2 .
n2 +a2 2a 2a
n=1
X∞
1 π 1
12. =− cot(πa) + 2 .
n=1
n2 −a2 2a 2a
X∞
1 22n−1 π 2n
13. = |B2n |; the B2n are Bernoulli numbers (see Section 30.1.3).
k2n (2n)!
k=1
∞
X (−1)k+1 (22n−1 − 1)π 2n
14. = |B2n |; the B2n are Bernoulli numbers.
k2n (2n)!
k=1
∞
X 1 (22n−1 − 1)π 2n
15. 2n
= |B2n |; the B2n are Bernoulli numbers.
(2k − 1) 2(2n)!
k=1
∞
X 1
16. = ln 2.
k2k
k=1
∞
X (−1)k n2
17. = .
n2k n2 + 1
k=0
∞
X 1
18. = e = 2.71828 . . .
k!
k=0
∞
X (−1)k 1
19. = = 0.36787 . . .
k! e
k=0
∞
X k
20. = 1.
(k + 1)!
k=1
26.4 Infinite Functional Series

26.4.1 Power Series
∞
X 1
1. xk = , |x| < 1.
1−x
k=0
X∞
x
2. kxk = , |x| < 1.
(1 − x)2
k=1
X∞
x(x + 1)
3. k 2 xk = , |x| < 1.
(1 − x)3
k=1
X∞
x(1 + 4x + x2 )
4. k 3 xk = , |x| < 1.
(1 − x)4
k=1
∞
X d n 1
5. (±1)k kn xk = x , |x| < 1.
dx 1 ∓ x
k=0
∞
X xk
6. = − ln(1 − x), −1 ≤ x < 1.
k
k=1
∞
X xk
7. (−1)k−1 = ln(1 + x), |x| < 1.
k
k=1
∞
X x2k−1 1 1+x
8. = ln , |x| < 1.
2k − 1 2 1−x
k=1
X∞
x2k−1
9. (−1)k−1 = arctan x, |x| ≤ 1.
2k − 1
k=1
X∞ Z x
xk ln(1 − t)
10. 2
= − dt, |x| ≤ 1.
k 0 t
k=1
∞
X xk+1
11. = x + (1 − x) ln(1 − x), |x| ≤ 1.
k(k + 1)
k=1
∞
X xk+2 x x2 1
12. = + + (1 − x2 ) ln(1 − x), |x| ≤ 1.
k(k + 2) 2 4 2
k=1
∞
X xk
13. = ex , x is any number.
k!
k=0
∞
X x2k
14. = cosh x, x is any number.
(2k)!
k=0
∞
X x2k
15. (−1)k = cos x, x is any number.
(2k)!
k=0
X∞
x2k+1
16. = sinh x, x is any number.
(2k + 1)!
k=0
26.4. Infinite Functional Series 1429
∞
X x2k+1
17. (−1)k = sin x, x is any number.
(2k + 1)!
k=0
∞
X xk+1
18. = ex − 1, x is any number.
k! (k + 1)
k=0
∞
X xk+2
19. = (x − 1)ex + 1, x is any number.
k! (k + 2)
k=0
X∞ √
x2k+1 π
20. (−1)k = erf x, x is any number.
k! (2k + 1) 2
k=0
X∞ n
(k + a)n k d t
21. x = exp(at + xe ) , x is any number.
k! dtn t=0
k=0
∞
X 22k (22k − 1)|B2k | 2k−1
22. x = tan x; the B2k are Bernoulli numbers, |x| < π/2.
(2k)!
k=1
∞
X 22k (22k − 1)|B2k | 2k−1
23. (−1)k−1 x = tanh x; the B2k are Bernoulli numbers,
(2k)!
k=1
|x| < π/2.
X∞
22k |B2k | 1
24. x2k−1 = − cot x; the B2k are Bernoulli numbers, 0 < |x| < π.
(2k)! x
k=1
X∞
22k |B2k | 2k−1 1
25. (−1)k−1 x = coth x− ; the B2k are Bernoulli numbers, |x| < π.
(2k)! x
k=1
26.4.2 Trigonometric Series in One Variable Involving Sine

∞
X 1 1
1. sin(kx) = (π − x), 0 < x < 2π.
k 2
k=1
∞
X (−1)k−1 1
2. sin(kx) = x, −π < x < π.
k 2
k=1
∞
X ak a sin x
3. sin(kx) = arctan , 0 < x < 2π, |a| ≤ 1.
k 1 − a cos x
k=1
∞
X
1 π x x 2 x
4. sin(kx) = cos − sin ln cot , 0 < x < 2π.
2k + 1 4 2 2 4
k=0
X∞
(−1)k 1 x 2 x+π π x
5. sin(kx) = − cos ln cot − sin , −π < x < π.
2k + 1 4 2 4 4 2
k=0
X∞ Z x
1 t
6. 2
sin(kx) = − ln 2 sin dt, 0 ≤ x < π.
k 0 2
k=1
X∞ Z x
(−1)k t
7. sin(kx) = − ln 2 cos dt, −π < x < π.
k2 0 2
k=1
∞
X
1 x x
8. sin(kx) = (π − x) sin2 + sin x ln 2 sin , 0 ≤ x ≤ 2π.
k(k + 1) 2 2
k=1
X∞
(−1)k 2 x x
9. sin(kx) = −x cos + sin x ln 2 cos , −π ≤ x ≤ π.
k(k + 1) 2 2
k=1
X∞
k π
10. sin(kx) = sinh[a(π − x)], 0 < x < 2π.
k2 +a2 2 sinh(πa)
k=1
X∞
k π
11. (−1)k+1 sin(kx) = sinh(ax), −π < x < π.
k2 +a2 2 sinh(πa)
k=1
X∞
k π
12. sin(kx) = sin[a(π − x)], 0 < x < 2π.
k2 −a2 2 sin(πa)
k=1
X∞
k π
13. (−1)k+1 sin(kx) = sin(ax), −π < x < π.
k2 −a2 2 sin(πa)
k=1
X∞
k 1 1
14. (−1)k sin(kx) = sin x + x cos x, −π < x < π.
−1 k2 4 2
k=2
X∞
1 (−1)n−1 (2π)2n+1 x
15. 2n+1
sin(kx) = B2n+1 ,
k 2(2n + 1)! 2π
k=1
where 0 ≤ x ≤ 2π for n = 1, 2, . . . ; 0 < x < 2π for n = 0; and the Bn (x) are
Bernoulli polynomials (see Section 30.18.1).
X∞
(−1)k (−1)n−1 (2π)2n+1 x+π
16. sin(kx) = B 2n+1 ,
k2n+1 2(2n + 1)! 2π
k=1
where −π < x ≤ π for n = 0, 1, . . . ; the Bn (x) are Bernoulli polynomials.
X∞
1
17. sin(kx) = exp(cos x) sin(sin x), x is any number.
k!
k=1
X∞
(−1)k
18. sin(kx) = − exp(− cos x) sin(sin x), x is any number.
k!
k=1
X∞
1 x x
19. sin(kx) = sin sin sinh cos , x is any number.
(2k)! 2 2
k=0
X∞
(−1)k x x
20. sin(kx) = − sin cos sinh sin , x is any number.
(2k)! 2 2
k=0
∞
X ak
21. sin(kx) = exp(k cos x) sin(k sin x), |a| ≤ 1, x is any number.
k!
k=0
X∞
a sin x
22. ak sin(kx) = , |a| < 1, x is any number.
1 − 2a cos x + a2
k=0
X∞
a(1 − a2 ) sin x
23. kak sin(kx) = , |a| < 1, x is any number.
(1 − 2a cos x + a2 )2
k=1
X∞
1 1 x
24. sin(kx + a) = (π − x) cos a − ln 2 sin sin a, 0 < x < 2π.
k 2 2
k=1
X∞
(−1)k−1 1 x
25. sin(kx + a) = x cos a + ln 2 cos sin a, −π < x < π.
k 2 2
k=1
∞
X sin[(2k − 1)x] π
26. = , 0 < x < π.
2k − 1 4
k=1
X∞
k−1 sin[(2k − 1)x] 1 x π π π
27. (−1) = ln tan + , − <x< .
2k − 1 2 2 4 2 2
k=1
∞
X sin[(2k − 1)x] 1 2a sin x
28. a2k−1 = arctan , 0 < x < 2π, |a| ≤ 1.
2k − 1 2 1 − a2
k=1
X∞
sin[(2k − 1)x] 1 1 + 2a sin x + a2
29. (−1)k−1 a2k−1 = ln , 0 < x < π, |a| ≤ 1.
2k − 1 4 1 − 2a sin x + a2
k=1
X∞
k sin[(k + 1)x] 1 x
30. (−1) = sin x − x(1 + cos x) − sin x ln2 cos .
k(k + 1) 2 2
k=1
∞
X a(1 + a2 ) sin x
31. a2k+1 sin[(2k + 1)x] = , |a| < 1, x is any number.
(1 + a2 )2 − 4a2 cos2 x
k=0
∞
X a(1 − a2 ) sin x
32. (−1)k a2k+1 sin[(2k+1)x] = , |a| < 1, x is any number.
k=0
(1 + a2 )2 − 4a2 sin2 x
X∞
sin[2(k + 1)x]
33. = sin(2x)−(π−2x) sin2 x−sin x cos x ln(4 sin2 x), 0 ≤ x ≤ π.
k(k + 1)
k=1
∞
(
X 1
if − 12 π ≤ x ≤ 12 π,
34. (−1) k sin[(2k + 1)x]
= 4 πx
(2k + 1)2 1 1 3
4 π(π − x) if 2 π ≤ x ≤ 2 π.
k=1
26.4.3 Trigonometric Series in One Variable Involving Cosine

X∞
1 x
1. cos(kx) = − ln 2 sin , 0 < x < 2π.
k 2
k=1
X∞
(−1)k−1 x
2. cos(kx) = ln 2 cos , −π < x < π.
k 2
k=1
X∞
ak 1
3. cos(kx) = ln √ , 0 < x < 2π, |a| ≤ 1.
k 1 − 2a cos x + a2
k=1
∞
X
1 π x x 2 x
4. cos(kx) = sin + cos ln cot , 0 < x < 2π.
2k + 1 4 2 2 4
k=0
X∞
(−1)k 1 x 2 x+π π x
5. cos(kx) = − sin ln cot + cos , −π < x < π.
2k + 1 4 2 4 4 2
k=0
∞
X 1 1
6. cos(kx) = (3x2 − 6πx + 2π 2 ), 0 ≤ x ≤ 2π.
k2 12
k=1
∞
X (−1)k 1
7. cos(kx) = (3x2 − π 2 ), −π ≤ x ≤ π.
k2 12
k=1
X∞
1 1 2 x x
8. cos(kx) = (x − π) sin x − 2 sin ln 2 sin + 1, 0 ≤ x ≤ 2π.
k(k + 1) 2 2 2
k=1
X∞
(−1)k 1 2 x x
9. cos(kx) = − x sin x − 2 cos ln 2 cos + 1, −π ≤ x ≤ π.
k(k + 1) 2 2 2
k=1
X∞
1 π 1
10. cos(kx) = cosh[a(π − x)] − 2 , 0 ≤ x ≤ 2π.
k2 +a2 2a sinh(πa) 2a
k=1
X∞
1 π 1
11. cos(kx) = − cos[a(π − x)] + 2 , 0 ≤ x ≤ 2π.
k2 −a2 2a sin(πa) 2a
k=1
X∞
(−1)k 1 1 1
12. 2
cos(kx) = − cos x − x sin x, −π ≤ x ≤ π.
k −1 2 4 2
k=2
X∞
k 1 1 x
13. cos(kx) = − − cos x − cos x ln 2 sin , 0 < x < 2π.
k2 − 1 2 4 2
k=2
X∞
1 (−1)n−1 (2π)2n x
14. 2n
cos(kx) = B2n ,
k 2(2n)! 2π
k=1
where 0 ≤ x ≤ 2π for n = 1, 2, . . . ; the Bn (x) are Bernoulli polynomials (see
Section 30.18.1).
X∞
(−1)k (−1)n−1 (2π)2n x+π
15. cos(kx) = B2n ,
k2n 2(2n)! 2π
k=1
where −π ≤ x ≤ π for n = 1, 2, . . . ; the Bn (x) are Bernoulli polynomials.
X∞
1
16. cos(kx) = exp(cos x) cos(sin x), x is any number.
k!
k=0
X∞
(−1)k
17. cos(kx) = exp(− cos x) cos(sin x), x is any number.
k!
k=0
X∞
1 x x
18. cos(kx) = cos sin cosh cos , x is any number.
(2k)! 2 2
k=0
X∞
(−1)k x x
19. cos(kx) = cos cos cosh sin , x is any number.
(2k)! 2 2
k=0
∞
X ak
20. cos(kx) = exp(a cos x) cos(a sin x), |a| ≤ 1, x is any number.
k!
k=0
X∞
1 − a cos x
21. ak cos(kx) = , |a| < 1, x is any number.
1 − 2a cos x + a2
k=0
∞
X a(1 + a2 ) cos x − 2a2
22. kak cos(kx) = , |a| < 1, x is any number.
(1 − 2a cos x + a2 )2
k=1
X∞
1 1 x
23. cos(kx + a) = (x − π) sin a − ln 2 sin cos a, 0 < x < 2π.
k 2 2
k=1
X∞
(−1)k−1 1 x
24. cos(kx + a) = − x sin a + ln 2 cos cos a, −π < x < π.
k 2 2
k=1
∞
X cos[(2k − 1)x] 1 x
25. = ln cot , 0 < x < π.
2k − 1 2 2
k=1
X∞
cos[(2k − 1)x] π
26. (−1)k−1 = , 0 < x < π.
2k − 1 4
k=1
X∞
cos[(2k − 1)x] 1 1 + 2a cos x + a2
27. a2k−1 = ln , 0 < x < 2π, |a| ≤ 1.
2k − 1 4 1 − 2a cos x + a2
k=1
∞
X cos[(2k − 1)x] 1 2a cos x
28. (−1)k−1 a2k−1 = arctan , 0 < x < π, |a| ≤ 1.
2k − 1 2 1 − a2
k=1
X∞
cos[(2k − 1)x] π π
29. = − |x| , −π ≤ x ≤ π.
(2k − 1)2 4 2
k=1
∞
X
cos[(k + 1)x] 1 x
30. (−1)k
= cos x − x sin x − (1 + cos x) ln2 cos .
k(k + 1) 2 2
k=1
∞
X a(1 − a2 ) cos x
31. a2k+1 cos[(2k + 1)x] = , |a| < 1, x is any number.
(1 + a2 )2 − 4a2 cos2 x
k=0
X∞
a(1 + a2 ) cos x
32. (−1)k a2k+1 cos[(2k+1)x] = , |a| < 1, x is any number.
k=0
(1 + a2 )2 − 4a2 sin2 x
X∞
cos[2(k + 1)x] π
33. = cos(2x)− −x sin(2x)+sin2 x ln(4 sin2 x), 0 ≤ x ≤ π.
k(k + 1) 2
k=1
26.4.4 Trigonometric Series in Two Variables

X∞
1 1 x+y x − y
1. sin(kx) sin(ky) = lnsin cosec , x ± y 6= 0, 2π, 4π, . . .
k 2 2 2
k=1
X∞
(−1)k 1 x+y x − y
2. sin(kx) sin(ky) = lncos sec , x ± y 6= π, 3π, 5π, . . .
k 2 2 2
k=1
∞
(
X 1
3.
1
sin(kx) sin(ky) = 2 x(π − y) if −y ≤ x ≤ y,
Here 0 < y < π.
k 2 1
k=1 2 y(π − x) if y ≤ x ≤ 2π − y.
∞
X (−1)k+1 1
4. 2
sin(kx) sin(ky) = xy, |x ± y| ≤ π.
k 2
k=1
X∞
ak 1 4a sin2 [(x + y)/2] + (a − 1)2
5. sin(kx) sin(ky) = ln 2 2
, 0 < a < 1.
k 4 4a sin [(x − y)/2] + (a − 1)
k=1
∞
X 1 1 π
6. sin2 (kx) sin2 (ky) = πx, 0≤x≤y≤ .
k2 2 2
k=1
∞
X 1 1
7. cos(kx) cos(ky) = − ln2(cos x − cos y), x ± y 6= 0, 2π, 4π, . . .
k 2
k=1
X∞
(−1)k 1
8. cos(kx) cos(ky) = − ln2(cos x + cos y), x ± y 6= π, 3π, 5π, . . .
k 2
k=1

1
X∞
1 − 2
 if 0 < x < y,
9. 1
sin(kx) cos(ky) = 4 (π − 2y) if x = y, Here 0 < y < π.
k 
 1
k=1
2 (π − x) if y < x < π.
∞
(
X 1 1
3x2 + 3(y − π)2 − π 2 if 0 ≤ x ≤ y,
10. 12
cos(kx) cos(ky) = 1 2
k2 12 3y + 3(x − π) − π
2 2 if y ≤ x ≤ π.
k=1
Here 0 < y < π.
X∞
(−1)k
11. cos(kx) cos(ky)
k2 (
k=1 1
12 3(x2 + y 2 ) − π 2 if −(π − y) ≤ x ≤ π − y,
= 1 2 2 2

12 3(x − π) + 3(y − π) − π if π − y ≤ x ≤ π + y.
Here 0 < y < π.
⊙ References for Chapter 26: H. B. Dwight (1961), V. Mangulis (1965), E. R. Hansen (1975), I. S. Gradshteyn
and I. M. Ryzhik (2000), A. P. Prudnikov, Yu. A. Brychkov, and O. I. Marichev (1986), D. Zwillinger (2002).
Chapter 27
Indefinite and Definite Integrals
27.1 Indefinite Integrals

⋆ Throughout Section 27.1, the integration constant C is omitted for brevity.
27.1.1 Integrals Involving Rational Functions

◮ Integrals involving a + bx.
Z
dx 1
1. = ln |a + bx|.
a + bx b
Z
(a + bx)n+1
2. (a + bx)n dx = , n 6= −1.
b(n + 1)
Z
x dx 1
3. = 2 a + bx − a ln |a + bx| .
a + bx b
Z 2
x dx 1 h1 i
4. = 3 (a + bx)2 − 2a(a + bx) + a2 ln |a + bx| .
Z a + bx b 2
dx 1 a + bx
5. = − ln .
x(a + bx) a x
Z
dx 1 b a + bx
6. =− + 2 ln .
x2 (a + bx) ax a x
Z
x dx 1 a
7. = ln |a + bx| + .
(a + bx)2 b2 a + bx
Z
x2 dx 1 a2
8. = a + bx − 2a ln |a + bx| − .
(a + bx)2 b3 a + bx
Z
dx 1 1 a + bx
9. = − ln .
x(a + bx)2 a(a + bx) a2 x
Z
x dx 1h 1 a i
10. = − + .
(a + bx)3 b2 a + bx 2(a + bx)2
◮ Integrals involving a + x and b + x.

Z
a+x
1. dx = x + (a − b) ln |b + x|.
b+x
1435
1436 I NDEFINITE AND D EFINITE I NTEGRALS
Z b+x
dx 1
2. = ln , a 6= b. For a = b, see Integral 2 with n = −2
(a + x)(b + x) a−b a+x
in the paragraph ‘Integrals involving a + bx’ above.
Z
x dx 1
3. = a ln |a + x| − b ln |b + x| .
(a + x)(b + x) a−b
Z a + x
dx 1 1
4. 2
= + 2
ln .
(a + x)(b + x) (b − a)(b + x) (a − b) b+x
Z a + x
x dx b a
5. = − ln .
(a + x)(b + x)2 (a − b)(b + x) (a − b)2 b+x
Z
x2 dx b2 a2 b2 − 2ab
6. = + ln |a + x| + ln |b + x|.
(a + x)(b + x)2 (b − a)(b + x) (a − b)2 (b − a)2
Z 1 a + x
dx 1 1 2
7. 2 2
=− 2
+ + 3
ln .
(a + x) (b + x) (a − b) a + x b+x (a − b) b+x
Z a a + x
x dx 1 b a+b
8. = + + ln .
(a + x)2 (b + x)2 (a − b)2 a + x b+x (a − b)3 b+x
Z a2 a + x
x2 dx 1 b2 2ab
9. 2 2
= − 2
+ + 3
ln .
(a + x) (b + x) (a − b) a + x b+x (a − b) b+x
◮ Integrals involving a2 + x2 .
Z
dx 1 x
1. 2 2
= arctan .
a +x a a
Z
dx x 1 x
2. 2 2 2
= 2 2 2
+ 3 arctan .
(a + x ) 2a (a + x ) 2a a
Z
dx x 3x 3 x
3. 2 2 3
= 2 2 2 2
+ 4 2 2
+ 5 arctan .
(a + x ) 4a (a + x ) 8a (a + x ) 8a a
Z Z
dx x 2n − 1 dx
4. = + ; n = 1, 2, . . .
(a2 + x2 )n+1 2na2 (a2 + x2 )n 2na2 (a2 + x2 )n
Z
x dx 1
5. 2 2
= ln(a2 + x2 ).
a +x 2
Z
x dx 1
6. 2 2 2
=− .
(a + x ) 2(a + x2 )
2
Z
x dx 1
7. =− .
(a2 + x2 )3 4(a2 + x2 )2
Z
x dx 1
8. 2 2 n+1
=− ; n = 1, 2, . . .
(a + x ) 2n(a + x2 )n
2
Z
x2 dx x
9. 2 2
= x − a arctan .
a +x a
Z
x2 dx x 1 x
10. 2 2 2
=− 2 2
+ arctan .
(a + x ) 2(a + x ) 2a a
Z 2
x dx x x 1 x
11. 2 2 3
=− 2 2 2
+ 2 2 2
+ 3 arctan .
(a + x ) 4(a + x ) 8a (a + x ) 8a a
27.1. Indefinite Integrals 1437
Z Z
x2 dx x 1 dx
12. 2 2 n+1
=− 2 2 n
+ ; n = 1, 2, . . .
(a + x ) 2n(a + x ) 2n (a + x2 )n
2
Z
x3 dx x2 a2
13. = − ln(a2 + x2 ).
a2 + x2 2 2
Z
x3 dx a2 1
14. 2 2 2
= 2 2
+ ln(a2 + x2 ).
(a + x ) 2(a + x ) 2
Z 3
x dx 1 a2
15. = − + ; n = 2, 3, . . .
(a2 + x2 )n+1 2(n − 1)(a2 + x2 )n−1 2n(a2 + x2 )n
Z
dx 1 x2
16. = ln .
x(a2 + x2 ) 2a2 a2 + x2
Z
dx 1 1 x2
17. = + ln .
x(a2 + x2 )2 2a2 (a2 + x2 ) 2a4 a2 + x2
Z
dx 1 1 1 x2
18. = + + ln .
x(a2 + x2 )3 4a2 (a2 + x2 )2 2a4 (a2 + x2 ) 2a6 a2 + x2
Z
dx 1 1 x
19. = − 2 − 3 arctan .
x2 (a2 + x2 ) a x a a
Z
dx 1 x 3 x
20. =− 4 − 4 2 − 5 arctan .
x2 (a2 + x2 )2 a x 2a (a + x2 ) 2a a
Z
dx 1 1 1 x2
21. = − − − ln .
x3 (a2 + x2 )2 2a4 x2 2a4 (a2 + x2 ) a6 a2 + x2
Z
dx 1 x 7x 15 x
22. 2 2 2 3
=− 6 − 4 2 2 2
− 6 2 2
− 7 arctan .
x (a + x ) a x 4a (a + x ) 8a (a + x ) 8a a
Z
dx 1 1 1 3 x2
23. = − − − − ln .
x3 (a2 + x2 )3 2a6 x2 a6 (a2 + x2 ) 4a4 (a2 + x2 )2 2a8 a2 + x2
◮ Integrals involving a2 − x2 .
Z
dx 1 a + x
1. = ln .
a2 − x2 2a a−x
Z a + x
dx x 1
2. = + ln .
(a2 − x2 )2 2a2 (a2 − x2 ) 4a3 a−x
Z a + x
dx x 3x 3
3. = + + ln .
(a2 − x2 )3 4a2 (a2 − x2 )2 8a4 (a2 − x2 ) 16a5 a−x
Z Z
dx x 2n − 1 dx
4. 2 2 n+1
= 2 2 2 n
+ 2
; n = 1, 2, . . .
(a − x ) 2na (a − x ) 2na (a − x2 )n
2
Z
x dx 1
5. 2 2
= − ln |a2 − x2 |.
a −x 2
Z
x dx 1
6. 2 2 2
= .
(a − x ) 2(a − x2 )
2
Z
x dx 1
7. 2 2 3
= .
(a − x ) 4(a − x2 )2
2
Z
x dx 1
8. 2 2 n+1
= ; n = 1, 2, . . .
(a − x ) 2n(a − x2 )n
2
Z
x2 dx a a + x
9. = −x + ln .
a2 − x2 2 a−x
Z
x2 dx x 1 a + x
10. = − ln .
(a2 − x2 )2 2(a2 − x2 ) 4a a−x
Z a + x
x2 dx x x 1
11. = − − ln .
(a2 − x2 )3 4(a2 − x2 )2 8a2 (a2 − x2 ) 16a3 a−x
Z Z
x2 dx x 1 dx
12. 2 2 n+1
= 2 2 n
− ; n = 1, 2, . . .
(a − x ) 2n(a − x ) 2n (a − x2 )n
2
Z
x3 dx x2 a2
13. = − − ln |a2 − x2 |.
a2 − x2 2 2
Z
x3 dx a2 1
14. 2 2 2
= 2 2
+ ln |a2 − x2 |.
(a − x ) 2(a − x ) 2
Z 3
x dx 1 a2
15. = − + ; n = 2, 3, . . .
(a2 − x2 )n+1 2(n − 1)(a2 − x2 )n−1 2n(a2 − x2 )n
Z x2
dx 1
16. = 2 ln 2 .
x(a2 − x2 ) 2a a − x2
Z x2
dx 1 1
17. = + ln 2 .
x(a2 − x2 )2 2a2 (a2 − x2 ) 2a4 a − x2
Z x2
dx 1 1 1
18. = + + ln 2 .
x(a2 − x2 )3 4a2 (a2 − x2 )2 2a4 (a2 − x2 ) 2a6 a − x2
◮ Integrals involving a3 + x3 .
Z
dx 1 (a + x)2 1 2x − a
1. 3 3
= 2
ln 2 2
+ √ arctan √ .
a +x 6a a − ax + x 2
a 3 a 3
Z Z
dx x 2 dx
2. = 3 3 + 3 .
(a3 + x3 )2 3a (a + x3 ) 3a a3 + x3
Z
x dx 1 a2 − ax + x2 1 2x − a
3. 3 3
= ln 2
+ √ arctan √ .
a +x 6a (a + x) a 3 a 3
Z 2 Z
x dx x 1 x dx
4. 3 3 2
= 3 3 3
+ 3 .
(a + x ) 3a (a + x ) 3a a + x3
3
Z
x2 dx 1
5. = ln |a3 + x3 |.
a3 + x3 3
Z x3
dx 1
6. = ln 3 .
x(a3 + x3 ) 3a3 a + x3
Z x3
dx 1 1
7. 3 3 2
= 3 3 3
+ 6
ln 3 3 .
x(a + x ) 3a (a + x ) 3a a +x
Z Z
dx 1 1 x dx
8. 2 3 3
=− 3 − 3 .
x (a + x ) a x a a + x3
3
Z Z
dx 1 x2 4 x dx
9. 2 3 3 2
=− 6 − 6 3 3
− 6 .
x (a + x ) a x 3a (a + x ) 3a a + x3
3
◮ Integrals involving a3 − x3 .
Z
dx 1 a2 + ax + x2 1 2x + a
1. = ln + √ arctan √ .
a3 − x3 6a2 (a − x)2 2
a 3 a 3
Z Z
dx x 2 dx
2. = 3 3 + 3 .
(a3 − x3 )2 3a (a − x3 ) 3a a3 − x3
Z
x dx 1 a2 + ax + x2 1 2x + a
3. 3 3
= ln 2
− √ arctan √ .
a −x 6a (a − x) a 3 a 3
Z 2 Z
x dx x 1 x dx
4. 3 3 2
= 3 3 3
+ 3 .
(a − x ) 3a (a − x ) 3a a − x3
3
Z
x2 dx 1
5. = − ln |a3 − x3 |.
a3 − x3 3
Z x3
dx 1
6. 3 3
= 3
ln 3 3 .
x(a − x ) 3a a −x
Z x3
dx 1 1
7. = 3 3 + 6 ln 3 .
x(a3 − x3 )2 3a (a − x3 ) 3a a − x3
Z Z
dx 1 1 x dx
8. 2 3 3
=− 3 + 3 .
x (a − x ) a x a a − x3
3
Z Z
dx 1 x2 4 x dx
9. = − − + .
x2 (a3 − x3 )2 a6 x 3a6 (a3 − x3 ) 3a6 a3 − x3
◮ Integrals involving a4 ± x4 .
Z √ √
dx 1 a2 + ax 2 + x2 1 ax 2
1. = √ ln √ + √ arctan 2 .
a4 + x4 4a3 2 a2 − ax 2 + x2 2a3 2 a − x2
Z
x dx 1 x2
2. = arctan .
a4 + x4 2a2 a2
Z √ √
x2 dx 1 a2 + ax 2 + x2 1 ax 2
3. = − √ ln √ + √ arctan 2 .
a4 + x4 4a 2 a2 − ax 2 + x2 2a 2 a − x2
Z a + x
dx 1 1 x
4. 4 4
= 3
ln + 3 arctan .
a −x 4a a−x 2a a
Z 2 2
x dx 1 a + x
5. = 2 ln 2 .
a4 − x4 4a a − x2
Z
x2 dx 1 a + x 1 x
6. 4 4
= ln − arctan .
a −x 4a a−x 2a a
27.1.2 Integrals Involving Irrational Functions

◮ Integrals involving x1/2 .
Z
x1/2 dx 2 1/2 2a bx1/2
1. = x − arctan .
a2 + b2 x b2 b3 a
Z
x3/2 dx 2x3/2 2a2 x1/2 2a3 bx1/2
2. 2 2
= 2
− 4
+ 5 arctan .
a +b x 3b b b a
Z
x1/2 dx x1/2 1 bx1/2
3. = − + arctan .
(a2 + b2 x)2 b2 (a2 + b2 x) ab3 a
Z
x3/2 dx 2x3/2 3a2 x1/2 3a bx1/2
4. = + − arctan .
(a2 + b2 x)2 b2 (a2 + b2 x) b4 (a2 + b2 x) b5 a
Z
dx 2 bx1/2
5. = arctan .
(a2 + b2 x)x1/2 ab a
Z
dx 2 2b bx1/2
6. = − − arctan .
(a2 + b2 x)x3/2 a2 x1/2 a3 a
Z
dx x1/2 1 bx1/2
7. = + arctan .
(a2 + b2 x)2 x1/2 a2 (a2 + b2 x) a3 b a
Z
x1/2 dx 2 1/2 2a a + bx1/2
8. = − x + 3 ln .
a2 − b2 x b2 b a − bx1/2
Z
x3/2 dx 2x3/2 2a2 x1/2 a3 a + bx1/2
9. = − − + ln .
a2 − b2 x 3b2 b4 b5 a − bx1/2
Z a + bx1/2
x1/2 dx x1/2 1
10. = 2 2 − ln .
(a2 − b2 x)2 b (a − b2 x) 2ab3 a − bx 1/2
Z
x3/2 dx 3a2 x1/2 − 2b2 x3/2 3a a + bx1/2
11. = − ln .
(a2 − b2 x)2 b4 (a2 − b2 x) 2b5 a − bx1/2
Z
dx 1 a + bx1/2
12. = ln .
(a2 − b2 x)x1/2 ab a − bx1/2
Z
dx 2 b a + bx1/2
13. = − + ln .
(a2 − b2 x)x3/2 a2 x1/2 a3 a − bx1/2
Z a + bx1/2
dx x1/2 1
14. 2 2 2 1/2
= 2 2 2
+ 3
ln 1/2
.
(a − b x) x a (a − b x) 2a b a − bx
◮ Integrals involving (a + bx)p/2 .

Z
2
1. (a + bx)p/2 dx = (a + bx)(p+2)/2 .
b(p + 2)
Z
2 (a + bx)(p+4)/2 a(a + bx)(p+2)/2
2. x(a + bx)p/2 dx = 2 − .
b p+4 p+2
Z
2 p/2 2 (a+bx)(p+6)/2 2a(a+bx)(p+4)/2 a2 (a+bx)(p+2)/2
3. x (a+bx) dx = 3 − + .
b p+6 p+4 p+2
◮ Integrals involving (x2 + a2 )1/2.

Z
1 a2
1. (x2 + a2 )1/2 dx = x(a2 + x2 )1/2 + ln x + (x2 + a2 )1/2 .
2 2
Z
1
2. x(x2 + a2 )1/2 dx = (a2 + x2 )3/2 .
Z 3
1 3 3
3. (x2 + a2 )3/2 dx = x(a2 + x2 )3/2 + a2 x(a2 + x2 )1/2 + a4 lnx + (x2 + a2 )1/2 .
4 8 8
Z a + (x2 + a2 )1/2
1 2
4. (x + a2 )1/2 dx = (a2 + x2 )1/2 − a ln .
x x
Z
dx
5. √ = ln x + (x2 + a2 )1/2 .
x2 + a2
Z
x dx
6. √ = (x2 + a2 )1/2 .
x2 + a2
Z
7. (x2 + a2 )−3/2 dx = a−2 x(x2 + a2 )−1/2 .
◮ Integrals involving (x2 − a2 )1/2.

Z
1 a2
1. (x2 − a2 )1/2 dx = x(x2 − a2 )1/2 − ln x + (x2 − a2 )1/2 .
2 2
Z
1
2. x(x2 − a2 )1/2 dx = (x2 − a2 )3/2 .
3
Z
1 3 3
3. (x2 − a2 )3/2 dx = x(x2 − a2 )3/2 − a2 x(x2 − a2 )1/2 + a4 lnx + (x2 − a2 )1/2 .
4 8 8
Z a
1 2
4. (x − a2 )1/2 dx = (x2 − a2 )1/2 − a arccos .
x x
Z
dx
5. √ = lnx + (x2 − a2 )1/2 .
2
x −a 2
Z
x dx
6. √ = (x2 − a2 )1/2 .
x2 − a2
Z
7. (x2 − a2 )−3/2 dx = −a−2 x(x2 − a2 )−1/2 .
◮ Integrals involving (a2 − x2 )1/2.

Z
1 a2 x
1. (a2 − x2 )1/2 dx = x(a2 − x2 )1/2 + arcsin .
2 2 a
Z
1
2. x(a2 − x2 )1/2 dx = − (a2 − x2 )3/2 .
3
Z
1 3 3 x
3. (a2 − x2 )3/2 dx = x(a2 − x2 )3/2 + a2 x(a2 − x2 )1/2 + a4 arcsin .
4 8 8 a
Z 2 2 1/2
1 2 a + (a − x )
4. (a − x2 )1/2 dx = (a2 − x2 )1/2 − a ln .
x x
Z
dx x
5. √ = arcsin .
a2 − x2 a
Z
x dx
6. √ = −(a2 − x2 )1/2 .
a2 − x2
Z
7. (a2 − x2 )−3/2 dx = a−2 x(a2 − x2 )−1/2 .
◮ Integrals involving arbitrary powers. Reduction formulas.

Z
dx 1 xn
1. = ln n .
x(axn + b) bn ax + b
Z
dx 2 xn/2

2. √ = ln √ .
Z x x +a
n 2 an n
x +a +a2
dx 2
a
3. √ = arccos n/2 .
n
x x −a 2 an x
Z √
dx 2 ax2n + bxn
4. √ =− .
x ax2n + bxn bnxn
⋆ The parameters a, b, p, m, and n below in Integrals 5–8 can assume arbitrary val-
ues, except those for which denominators vanish in successive applications of a formula.
Notation: w = axn + b.
Z Z
m n p 1 m+1 p
5. x (ax + b) dx = x w + npb xm wp−1 dx .
Z m + np + 1 Z
1 h i
m n p
6. x (ax + b) dx = −xm+1 wp+1 + (m + n + np + 1) xm wp+1 dx .
Z bn(p + 1) Z
1 h i
7. xm (axn + b)p dx = xm+1 wp+1 − a(m + n + np + 1) xm+n wp dx .
Z b(m + 1) Z
1 h i
8. xm (axn + b)p dx = xm−n+1 wp+1 −b(m − n + 1) xm−n wp dx .
a(m + np + 1)
27.1.3 Integrals Involving Exponential Functions

Z
1 ax
1. eax dx = e .
Z a
a x
2. ax dx = .
ln a
Z
ax ax x 1
3. xe dx = e − 2 .
a a
Z 2
2 ax ax x 2x 2
4. x e dx = e − 2 + 3 .
a a a
Z
n ax ax 1 n n n−1 n(n−1) n−2 n−1 n! n n!
5. x e dx = e x − 2x + x −· · ·+(−1) x+(−1) n+1 ,
a a a3 an a
n = 1, 2, . . .
Z X n
(−1)k dk
6. Pn (x)eax dx = eax Pn (x), where Pn (x) is an arbitrary polynomial
ak+1 dxk
k=0
Zof degree n.
dx x 1
7. px
= − ln |a + bepx |.
a + be a  ap
r

 1 a
Z 
 √ arctan epx if ab > 0,
dx p ab b
8. = √
aepx + be−px 
 1 b + epx −ab

 √ ln √ if ab < 0.
2p −ab b − epx −ab
 √ √

 1 a + bepx − a
Z 
 p√a ln √ √ if a > 0,
dx a + bepx + a
9. √ = √
a + bepx 
 2 a + bepx

 √ arctan √ if a < 0.
p −a −a
27.1.4 Integrals Involving Hyperbolic Functions

◮ Integrals involving cosh x.
Z
1
1. cosh(a + bx) dx = sinh(a + bx).
b
Z
2. x cosh x dx = x sinh x − cosh x.
Z
3. x2 cosh x dx = (x2 + 2) sinh x − 2x cosh x.
Z X n 2k
2n x x2k−1
4. x cosh x dx = (2n)! sinh x − cosh x .
(2k)! (2k − 1)!
k=1
Z X n
2n+1 x2k+1 x2k
5. x cosh x dx = (2n + 1)! sinh x − cosh x .
(2k + 1)! (2k)!
k=0
Z Z
6. xp cosh x dx = xp sinh x − pxp−1 cosh x + p(p − 1) xp−2 cosh x dx.
Z
7. cosh2 x dx = 12 x + 14 sinh 2x.
Z
8. cosh3 x dx = sinh x + 13 sinh3 x.
Z n−1
x 1 X k sinh[2(n − k)x]
9. cosh2n x dx = C2n
n
+ C2n , n = 1, 2, . . .
22n 22n−1 2(n − k)
k=0
Z n n
1 X k sinh[(2n − 2k + 1)x] X
k sinh
2k+1
x
10. cosh2n+1 x dx = C 2n+1 = C n ,
22n 2n − 2k + 1 2k + 1
k=0 k=0
n = 1, 2, . . .
Z Z
p 1 p−1 p−1
11. cosh x dx = sinh x cosh x+ coshp−2 x dx.
p p
Z
1
12. cosh ax cosh bx dx = 2 (a cosh bx sinh ax − b cosh ax sinh bx).
a − b2
Z
dx 2
13. = arctan eax .
cosh ax a
Z
dx sinh x 1
14. 2n = 2n−1
cosh x 2n − 1 cosh x
n−1
X k

2 (n − 1)(n − 2) . . . (n − k) 1
+ , n = 1, 2, . . .
(2n − 3)(2n − 5) . . . (2n − 2k − 1) cosh2n−2k−1 x
k=1
Z
dx sinh x 1
15. 2n+1 =
cosh x 2n cosh2n x
n−1
X (2n − 1)(2n − 3) . . . (2n − 2k + 1)
1 (2n − 1)!!
+ k 2n−2k
+ arctan sinh x,
2 (n − 1)(n − 2) . . . (n − k) cosh x (2n)!!
k=1
n = 1, 2, . . .

 sign x b + a cosh x
Z 
 −√ arcsin if a2 < b2 ,
dx b2 − a2 a+√ b cosh x
16. = a + b + a2 − b2 tanh(x/2)
a + b cosh x  √ 1
 ln √ if a2 > b2 .
a2 − b2 a + b − a2 − b2 tanh(x/2)
◮ Integrals involving sinh x.

Z
1
1. sinh(a + bx) dx = cosh(a + bx).
b
Z
2. x sinh x dx = x cosh x − sinh x.
Z
3. x2 sinh x dx = (x2 + 2) cosh x − 2x sinh x.
Z X n Xn
2n x2k x2k−1
4. x sinh x dx = (2n)! cosh x − sinh x .
(2k)! (2k − 1)!
k=0 k=1
Z Xn 2k+1

2n+1 x x2k
5. x sinh x dx = (2n + 1)! cosh x − sinh x .
(2k + 1)! (2k)!
k=0
Z Z
p p p−1
6. x sinh x dx = x cosh x − px sinh x + p(p − 1) xp−2 sinh x dx.
Z
7. sinh2 x dx = − 12 x + 14 sinh 2x.
Z
8. sinh3 x dx = − cosh x + 13 cosh3 x.
Z n−1
x 1 X k sinh[2(n−k)x]
9. sinh2n x dx = (−1)n C2n
n
2n
+ 2n−1 (−1)k C2n , n = 1, 2, . . .
2 2 2(n−k)
k=0
Z n
X
1 cosh[(2n − 2k + 1)x]
10. sinh2n+1 x dx = (−1)k C2n+1
k
22n 2n − 2k + 1
k=0 Xn
cosh2k+1 x
= (−1)n+k Cnk , n = 1, 2, . . .
2k + 1
Z k=0 Z
1 p−1
11. sinhp x dx = sinhp−1 x cosh x − sinhp−2 x dx.
p p
Z
1
12. sinh ax sinh bx dx = 2 2
a cosh ax sinh bx − b cosh bx sinh ax .
a −b
Z
dx 1 ax
13. = lntanh .
sinh ax a 2
Z n−1
X
dx cosh x 1 k−1 Ank
14. = − + (−1) ,
sinh2n x 2n − 1 sinh2n−1 x k=1 sinh2n−2k−1 x
2k (n − 1)(n − 2) . . . (n − k)
Ank = n = 1, 2, . . .
(2n − 3)(2n − 5) . . . (2n − 2k − 1)
Z n−1
X
dx cosh x 1 k−1 Ank
15. = − + (−1)
sinh2n+1 x 2n sinh2n x k=1 sinh2n−2k x
(2n − 1)!! x (2n − 1)(2n − 3) . . . (2n − 2k + 1)
+ (−1)n ln tanh , Ank = ,
(2n)!! 2 2k (n − 1)(n − 2) . . . (n − k)
n = 1, 2, . . .
Z √
dx 1 a tanh(x/2) − b + a2 + b2
16. = √ ln √ .
a + b sinh x a2 + b2 a tanh(x/2) − b − a2 + b2
Z √
Ax + B sinh x B Ab − Ba a tanh(x/2) − b + a2 + b2
17. dx = x + √ ln √ .
a + b sinh x b b a2 + b2 a tanh(x/2) − b − a2 + b2
◮ Integrals involving tanh x or coth x.

Z
1. tanh x dx = ln cosh x.
Z
2. tanh2 x dx = x − tanh x.
Z
3. tanh3 x dx = − 12 tanh2 x + ln cosh x.
Z Xn
tanh2n−2k+1 x
4. tanh2n x dx = x − , n = 1, 2, . . .
2n − 2k + 1
k=1
Z X n Xn
(−1)k Cnk tanh2n−2k+2 x
5. tanh2n+1 x dx = ln cosh x − = ln cosh x − ,
k=1
2k cosh2k x k=1
2n − 2k + 2
Zn = 1, 2, . . . Z
p 1 p−1
6. tanh x dx = − tanh x + tanhp−2 x dx.
Z p − 1
7. coth x dx = ln |sinh x|.
Z
8. coth2 x dx = x − coth x.
Z
9. coth3 x dx = − 12 coth2 x + ln |sinh x|.
Z n
X
2n coth2n−2k+1 x
10. coth x dx = x − , n = 1, 2, . . .
2n − 2k + 1
k=1
Z n
X Xn
Cnk coth2n−2k+2 x
11. coth2n+1 x dx = ln |sinh x| − = ln |sinh x| − ,
k=1
2k sinh2k x k=1
2n − 2k + 2
nZ = 1, 2, . . . Z
1
12. cothp x dx = − cothp−1 x + cothp−2 x dx.
p−1
27.1.5 Integrals Involving Logarithmic Functions

Z
1. ln ax dx = x ln ax − x.
Z
2. x ln x dx = 12 x2 ln x − 14 x2 .

Z  1 xp+1 ln ax − 1
xp+1 if p 6= −1,
p (p + 1)2
3. x ln ax dx = p + 1
1 2
2 ln ax if p = −1.
Z
4. (ln x)2 dx = x(ln x)2 − 2x ln x + 2x.
Z
5. x(ln x)2 dx = 12 x2 (ln x)2 − 12 x2 ln x + 14 x2 .
 p+1
Z x 2 2xp+1 2xp+1
(ln x) − ln x + if p 6= −1,
6. xp (ln x)2 dx = p + 1 (p + 1)2 (p + 1)3
1 3
3 ln x if p = −1.
Z n
x X
7. (ln x)n dx = (−1)k (n + 1)n . . . (n − k + 1)(ln x)n−k , n = 1, 2, . . .
n+1
k=0
Z Z
8. (ln x)q dx = x(ln x)q − q (ln x)q−1 dx, q 6= −1.
Z m
xn+1 X (−1)k
9. xn (ln x)m dx = (m + 1)m . . . (m − k + 1)(ln x)m−k ,
m+1 (n + 1)k+1
k=0
n,
Z m = 1, 2, . . . Z
p q 1 p+1 q q
10. x (ln x) dx = x (ln x) − xp (ln x)q−1 dx, p, q 6= −1.
Z p + 1 p + 1
1
11. ln(a + bx) dx = (ax + b) ln(ax + b) − x.
b
Z
1 2 a2 1 x2 a
12. x ln(a + bx) dx = x − 2 ln(a + bx) − − x .
2 b 2 2 b
Z 3
3
1 3 a 1 x ax2 a2 x
13. x2 ln(a + bx) dx = x − 3 ln(a + bx) − − + 2 .
3 b 3 3 2b b
Z
ln x dx ln x 1 x
14. =− + ln .
(a + bx)2 b(a + bx) ab a + bx
Z
ln x dx ln x 1 1 x
15. 3
=− 2
+ + 2 ln .
(a + bx) 2b(a + bx) 2ab(a + bx) 2a b a + bx
 √ √
 2 √ √ a + bx + a
Z 
 (ln x − 2) a + bx + a ln √ √ if a > 0,
ln x dx b a + bx − a
16. √ = √
a + bx 
 2 √ √ a + bx
 (ln x − 2) a + bx + 2 −a arctan √ if a < 0.
b −a
Z
17. ln(x2 + a2 ) dx = x ln(x2 + a2 ) − 2x + 2a arctan(x/a).
Z

18. x ln(x2 + a2 ) dx = 12 (x2 + a2 ) ln(x2 + a2 ) − x2 .
Z

19. x2 ln(x2 + a2 ) dx = 1
3 x3 ln(x2 + a2 ) − 23 x3 + 2a2 x − 2a3 arctan(x/a) .
27.1.6 Integrals Involving Trigonometric Functions

◮ Integrals involving cos x (n = 1, 2, . . . ).
Z
1
1. cos(a + bx) dx = sin(a + bx).
b
Z
2. x cos x dx = cos x + x sin x.
Z
3. x2 cos x dx = 2x cos x + (x2 − 2) sin x.
Z X n 2n−2k n−1
X 2n−2k−1

2n k x k x
4. x cos x dx = (2n)! (−1) sin x+ (−1) cos x .
(2n−2k)! (2n−2k−1)!
k=0 k=0
Z Xn 2n−2k+1
x x2n−2k
5. x2n+1 cos x dx = (2n + 1)! (−1)k sin x + cos x .
(2n − 2k + 1)! (2n − 2k)!
k=0
Z Z
6. xp cos x dx = xp sin x + pxp−1 cos x − p(p − 1) xp−2 cos x dx.
Z
7. cos2 x dx = 12 x + 14 sin 2x.
Z
8. cos3 x dx = sin x − 13 sin3 x.
Z n−1
2n 1 n 1 X k sin[(2n − 2k)x]
9. cos x dx = 2n C2n x + 2n−1 C2n .
2 2 2n − 2k
k=0
Z n
2n+1 1 X k sin[(2n − 2k + 1)x]
10. cos x dx = 2n C2n+1 .
2 2n − 2k + 1
k=0
Z x
dx π
11. = lntan + .
cos x 2 4
Z
dx
12. = tan x.
cos2 x
Z
dx sin x 1 x π
13. = + ln tan + .
cos3 x 2 cos2 x 2 2 4
Z Z
dx sin x n−2 dx
14. n
= n−1
+ , n > 1.
cos x (n − 1) cos x n−1 cosn−2 x
Z X n−1
x dx (2n−2k)x sin x−cos x 2n−1 (n−1)!
15. = Ank + x tan x+ln |cos x| ,
cos2n x cos2n−2k+1 x (2n−1)!!
k=0
(2n−2)(2n−4) . . . (2n−2k+2) 1
Ank = .
(2n−1)(2n−3) . . . (2n−2k+3) (2n−2k+1)(2n−2k)
Z
sin (b − a)x sin (b + a)x
16. cos ax cos bx dx = + , a 6= ±b.
2(b − a) 2(b + a)

 2 (a − b) tan(x/2)
Z 
 √ arctan √ if a2 > b2 ,
dx 2
a −b 2
√ 2 a2 − b2

17. = b − a2 + (b − a) tan(x/2)
a + b cos x   1 if b2 > a2 .
√ 2 ln √
b − a2 b2 − a2 − (b − a) tan(x/2)
Z Z
dx b sin x a dx
18. 2
= 2 2
− 2 2
.
(a + b cos x) (b − a )(a + b cos x) b −a a + b cos x
Z
dx 1 a tan x
19. = √ arctan √ .
a2 + b2 cos2 x a a2 + b 2 a2 + b2

 1 a tan x
Z 
 √ 2 arctan √ if a2 > b2 ,
dx a a −b 2
√ a2 − b2
20. = b2 − a2 − a tan x
a2 − b2 cos2 x   1
 √ 2 ln √ if b2 > a2 .
2a b − a 2 b − a + a tan x
2 2
Z
ax ax b a
21. e cos bx dx = e sin bx + 2 cos bx .
a2 + b2 a + b2
Z
eax 2
22. eax cos2 x dx = 2 a cos2 x + 2 sin x cos x + .
a +4 a
Z Z
ax n eax cosn−1 x n(n − 1)
23. e cos x dx = (a cos x + n sin x) + 2 eax cosn−2 x dx.
a2 + n2 a + n2
◮ Integrals involving sin x (n = 1, 2, . . . ).

Z
1
1. sin(a + bx) dx = − cos(a + bx).
b
Z
2. x sin x dx = sin x − x cos x.
Z
3. x2 sin x dx = 2x sin x − (x2 − 2) cos x.
Z
4. x3 sin x dx = (3x2 − 6) sin x − (x3 − 6x) cos x.
Z X n 2n−2k n−1
X 2n−2k−1
2n k+1 x k x
5. x sin x dx = (2n)! (−1) cos x+ (−1) sin x .
(2n−2k)! (2n−2k−1)!
k=0 k=0
Z X n
2n+1 x2n−2k+1
6. x sin x dx = (2n + 1)! (−1)k+1 cos x
(2n − 2k + 1)!
k=0
2n−2k
k x
+ (−1) sin x .
(2n − 2k)!
Z Z
7. xp sin x dx = −xp cos x + pxp−1 sin x − p(p − 1) xp−2 sin x dx.
Z
8. sin2 x dx = 12 x − 14 sin 2x.
Z
9. x sin2 x dx = 14 x2 − 14 x sin 2x − 18 cos 2x.
Z
10. sin3 x dx = − cos x + 13 cos3 x.
Z n−1
2n 1 n (−1)n X k sin[(2n − 2k)x]
11. sin x dx = 2n C2n x + 2n−1 (−1)k C2n ,
2 2 2n − 2k
k=0
k m!
where Cm = are binomial coefficients (0! = 1).
k! (m − k)!
Z n
1 X cos[(2n − 2k + 1)x]
12. sin2n+1 x dx = 2n (−1)n+k+1 C2n+1 k
.
2 2n − 2k + 1
k=0
Z
dx x
13. = lntan .
Z sin x 2
dx
14. 2 = − cot x.
Z sin x
dx cos x 1 x
15. 3 = − + ln tan .
Z sin x 2 sin2 x 2 2 Z
dx cos x n−2 dx
16. n =− n−1 + , n > 1.
sin x (n − 1) sin x n−1 sinn−2 x
Z n−1
X
x dx sin x+(2n−2k)x cos x 2n−1 (n−1)!
17. 2n = − A nk 2n−2k+1
+ ln |sin x|−x cot x ,
sin x sin x (2n−1)!!
k=0
(2n−2)(2n−4) . . . (2n−2k+2) 1
Ank = .
(2n−1)(2n−3) . . . (2n−2k+3) (2n−2k+1)(2n−2k)
Z
sin[(b − a)x] sin[(b + a)x]
18. sin ax sin bx dx = − , a 6= ±b.
2(b − a) 2(b + a)

 2 b + a tan x/2
Z 
 √ arctan √ if a2 > b2 ,
dx a2 − b2 √ a 2 − b2

19. = b − b2 − a2 + a tan x/2
a + b sin x   1 if b2 > a2 .
√ 2 ln √
b − a2 b + b2 − a2 + a tan x/2
Z Z
dx b cos x a dx
20. = + .
(a + b sin x)2 (a2 − b2 )(a + b sin x) a2 − b2 a + b sin x
Z √
dx 1 a2 + b2 tan x
21. = √ arctan .
a2 + b2 sin2 x a a2 + b2 a
 √

 1 a2 − b2 tan x
Z  √ arctan if a2 > b2 ,
dx a a2 − b2 a
22. = √ 2
a2 − b2 sin2 x   √1 b − a2 tan x + a

ln √ if b2 > a2 .
2a b 2 − a2 b2 − a2 tan x − a
Z
sin x dx 1 k cos x
23. p = − arcsin √ .
1 + k 2
2 sin x k 1 + k2
Z p
sin x dx 1
24. p = − lnk cos x + 1 − k2 sin2 x .
1 − k2 sin2 x k
Z p cos x p 1 + k2 k cos x
25. sin x 1 + k2 sin2 x dx = − 1 + k2 sin2 x − arcsin √ .
2 2k 1 + k2
Z p cos x p
26. sin x 1 − k2 sin2 x dx = − 1 − k2 sin2 x
2
1 − k2 p
− ln k cos x + 1 − k2 sin2 x .
2k
Z
a b
27. eax sin bx dx = eax sin bx − cos bx .
a2 + b2 a2 + b2
Z
eax 2
28. eax sin2 x dx = 2 a sin2 x − 2 sin x cos x + .
a +4 a
Z Z
ax n eax sinn−1 x n(n − 1)
29. e sin x dx = (a sin x − n cos x) + 2 eax sinn−2 x dx.
a2 + n2 a + n2
◮ Integrals involving sin x and cos x.

Z
cos[(a + b)x] cos (a − b)x
1. sin ax cos bx dx = − − , a 6= ±b.
2(a + b) 2(a − b)
Z c
dx 1
2. = arctan tan ax .
b2 cos2 ax + c2 sin2 ax abc b
Z c tan ax + b
dx 1
3. = ln .
b2 cos2 ax − c2 sin2 ax 2abc c tan ax − b
Z n+m−1
X
dx k tan2k−2m+1 x
4. = C n+m−1 , n, m = 1, 2, . . .
cos2n x sin2m x k=0
2k − 2m + 1
Z n+m
X
dx m k tan2k−2m x
5. = C n+m ln |tan x|+ C n+m , n, m = 1, 2, . . .
cos2n+1 x sin2m+1 x k=0
2k−2m
◮ Reduction formulas.
⋆ The parameters p and q below can assume any values, except those for which the
denominators on the right-hand side vanish.
Z Z
sinp−1 x cosq+1 x p−1
1. sinp x cosq x dx = − + sinp−2 x cosq x dx.
p+q p+q
Z Z
sinp+1 x cosq−1 x q−1
2. sinp x cosq x dx = + sinp x cosq−2 x dx.
p+q p+q
Z
sinp−1 x cosq−1 x 2 q−1
3. sinp x cosq x dx = sin x −
p+q p+q −2Z
(p − 1)(q − 1)
+ sinp−2 x cosq−2 x dx.
(p + q)(p + q − 2)
Z Z
p q sinp+1 x cosq+1 x p+q+2
4. sin x cos x dx = + sinp+2 x cosq x dx.
p+1 p+1
Z Z
p q sinp+1 x cosq+1 x p+q+2
5. sin x cos x dx = − + sinp x cosq+2 x dx.
q+1 q+1
Z Z
p q sinp−1 x cosq+1 x p−1
6. sin x cos x dx = − + sinp−2 x cosq+2 x dx.
q+1 q+1
Z Z
p q sinp+1 x cosq−1 x q−1
7. sin x cos x dx = + sinp+2 x cosq−2 x dx.
p+1 p+1
◮ Integrals involving tan x and cot x.

Z
1. tan x dx = − ln |cos x|.
Z
2. tan2 x dx = tan x − x.
Z
3. tan3 x dx = 12 tan2 x + ln |cos x|.
Z Xn
2n n (−1)k (tan x)2n−2k+1
4. tan x dx = (−1) x − , n = 1, 2, . . .
2n − 2k + 1
k=1
Z Xn
2n+1 n+1 (−1)k (tan x)2n−2k+2
5. tan x dx = (−1) ln |cos x| − , n = 1, 2, . . .
2n − 2k + 2
k=1
Z
dx 1
6. = 2 2
ax + b ln |a cos x + b sin x| .
a + b tan x a +b
Z r
tan x dx 1 a
7. √ = √ arccos 1 − cos x , b > a, b > 0.
a + b tan 2x b − a b
Z
8. cot x dx = ln |sin x|.
Z
9. cot2 x dx = − cot x − x.
Z
10. cot3 x dx = − 12 cot2 x − ln |sin x|.
Z Xn
(−1)k (cot x)2n−2k+1
11. cot2n x dx = (−1)n x + , n = 1, 2, . . .
2n − 2k + 1
k=1
Z Xn
(−1)k (cot x)2n−2k+2
12. cot2n+1 x dx = (−1)n ln |sin x| + , n = 1, 2, . . .
2n − 2k + 2
k=1
Z
dx 1
13. = 2 2
ax − b ln |a sin x + b cos x| .
a + b cot x a +b
27.1.7 Integrals Involving Inverse Trigonometric Functions

Z
x x p
1. arcsin dx = x arcsin + a2 − x2 .
Z a a
x 2 x 2 p x
2. arcsin dx = x arcsin − 2x + 2 a2 − x2 arcsin .
Z a a a
x 1 x x p
3. x arcsin dx = (2x2 − a2 ) arcsin + a2 − x2 .
a 4 a 4
Z p
x x3 x 1
4. x2 arcsin dx = arcsin + (x2 + 2a2 ) a2 − x2 .
Z a 3 a 9
x x p
5. arccos dx = x arccos − a2 − x2 .
Z a a
x 2 x 2 p x
6. arccos dx = x arccos − 2x − 2 a2 − x2 arccos .
a a a
Z
x 1 x xp 2
7. x arccos dx = (2x2 − a2 ) arccos − a − x2 .
a 4 a 4
Z p
x x3 x 1
8. x2 arccos dx = arccos − (x2 + 2a2 ) a2 − x2 .
a 3 a 9
Z
x x a
9. arctan dx = x arctan − ln(a2 + x2 ).
a a 2
Z
x 1 x ax
10. x arctan dx = (x2 + a2 ) arctan − .
a 2 a 2
Z
x x3 x ax2 a3
11. x2 arctan dx = arctan − + ln(a2 + x2 ).
a 3 a 6 6
Z
x x a
12. arccot dx = x arccot + ln(a2 + x2 ).
a a 2
Z
x 1 x ax
13. x arccot dx = (x2 + a2 ) arccot + .
a 2 a 2
Z
x x3 x ax2 a3
14. x2 arccot dx = arccot + − ln(a2 + x2 ).
a 3 a 6 6
27.2 Definite Integrals

⋆ Throughout Section 27.2 it is assumed that n is a positive integer, unless otherwise
specified.
27.2.1 Integrals Involving Power-Law Functions

◮ Integrals over a finite interval.
Z 1 n X n
x dx n (−1)k
1. = (−1) ln 2 + .
0 x+1 k=1
k
Z 1
dx β
2. 2
= .
0 x + 2x cos β + 1 2 sin β
Z 1
xa + x−a dx π sin(aβ)
3. 2
= , |a| < 1, β 6= (2n + 1)π.
0 x + 2x cos β + 1 sin(πa) sin β
Z 1
πa(1 − a)
4. xa (1 − x)1−a dx = , −1 < a < 1.
0 2 sin(πa)
Z 1
dx π
5. a (1 − x)1−a
= , 0 < a < 1.
0 x sin(πa)
Z 1
xa dx πa
6. a
= , −1 < a < 1.
0 (1 − x) sin(πa)
Z 1
Γ(p)Γ(q)
7. xp−1 (1 − x)q−1 dx ≡ B(p, q) = , p, q > 0.
0 Γ(p + q)
Z 1
π
8. xp−1 (1 − xq )−p/q dx = , q > p > 0.
0 q sin(πp/q)
27.2. Definite Integrals 1453
Z 1
πp
9. xp+q−1 (1 − xq )−p/q dx = , q > p.
0 q 2 sin(πp/q)
Z 1
π
10. xq/p−1 (1 − xq )−1/p dx = , p > 1, q > 0.
0 q sin(π/p)
Z 1
xp−1 − x−p
11. dx = π cot(πp), |p| < 1.
0 1−x
Z 1
xp−1 − x−p π
12. dx = , |p| < 1.
0 1+x sin(πp)
Z 1
xp − x−p 1
13. dx = − π cot(πp), |p| < 1.
0 x−1 p
Z 1
xp − x−p 1 π
14. dx = − , |p| < 1.
0 1+x p sin(πp)
Z 1 πp 1
x1+p − x1−p π
15. dx = cot − , |p| < 1.
0 1 − x2 2 2 p
Z 1 1+p 1−p
x −x 1 π
16. 2
dx = − , |p| < 1.
0 1+x p 2 sin(πp/2)
Z 1
dx 2
17. p = arctan a.
0 2
(1 + a x)(1 − x) a
Z 1
dx 1 1+a
18. p = ln .
0 (1 − a2 x)(1 − x) a 1−a
Z 1
dx π
19. √ = √ , 1 < a.
(a − x) 1 − x 2 2
a −1
−1
Z 1
xn dx 2 (2n)!!
20. √ = , n = 1, 2, . . .
0 1−x (2n + 1)!!
Z 1
xn−1/2 dx π (2n − 1)!!
21. √ = , n = 1, 2, . . .
0 1−x (2n)!!
Z 1
x2n dx π 1 × 3 × · · · × (2n − 1)
22. √ = , n = 1, 2, . . .
0 1−x 2 2 2 × 4 × · · · × (2n)
Z 1
x2n+1 dx 2 × 4 × · · · × (2n)
23. √ = , n = 1, 2, . . .
0 1−x 2 1 × 3 × · · · × (2n + 1)
Z 1
xλ−1 dx π
24. λ
= λ
, 0 < λ < 1, a > −1.
0 (1 + ax)(1 − x) (1 + a) sin(πλ)
Z 1
xλ−1/2 dx −1/2
sin[(2λ − 1)k]
25. λ λ
= 2π Γ λ + 1
2 Γ 1 − λ cos2λ k ,
0 (1 + ax)
√
(1 − x) (2λ − 1) sin k
k = arctan a, − 12 < λ < 1, a > 0.
◮ Integrals over an infinite interval.

Z ∞
dx π
1. 2
= √ .
0 ax + b 2 ab
Z ∞
√
dx π 2
2. = .
0 x4 + 1 4
Z ∞
xa−1 dx π
3. = , 0 < a < 1.
0 x+1 sin(πa)
Z ∞
xλ−1 dx π(1 − λ)
4. = λ , 0 < λ < 2.
0 (1 + ax)2 a sin(πλ)
Z ∞
xλ−1 dx π(aλ−1 − bλ−1 )
5. = , 0 < λ < 2.
0 (x + a)(x + b) (b − a) sin(πλ)
Z ∞
xλ−1 (x + c) dx π a − c λ−1 b − c λ−1
6. = a + b , 0 < λ < 1.
0 (x + a)(x + b) sin(πλ) a − b b−a
Z ∞
xλ dx πλ(1 − λ)
7. 3
= , −1 < λ < 2.
0 (x + 1) 2 sin(πλ)
Z
∞
xλ−1 dx π bλ−2 − aλ−2
8. = , 0 < λ < 4.
0 (x2 + a2 )(x2 + b2 ) 2 a2 − b2 sin(πλ/2)
Z ∞
xp−1 − xq−1
9. dx = π[cot(πp) − cot(πq)], p, q > 0.
0 1−x
Z ∞ n
xλ−1 dx n πCλ−1 n (λ − 1)(λ − 2) . . . (λ − n)
10. n+1
= (−1) λ
, Cλ−1 = ,
0 (1 + ax) a sin(πλ) n!
0 < λ < n + 1.
Z ∞
xm dx m+1 (2n − 2m − 3)!! am−n+1/2
11. n+1/2
= 2 m! m+1
, a, b > 0, m < b− 12 ,
0 (a + bx) (2n − 1)!! b
n, m = 1, 2, . . .
Z ∞
dx π (2n − 3)!! 1
12. 2 2 n
= , n = 1, 2, . . .
0 (x + a ) 2 (2n − 2)!! a2n−1
Z ∞
(x + 1)λ−1 1 − a−λ
13. dx = , a > 0.
0 (x + a)λ+1 λ(a − 1)
Z ∞ a−1
x dx π
14. b
= , 0 < a ≤ b.
0 x +1 b sin(πa/b)
Z ∞ a−1
x dx π(a − b)
15. b + 1)2
= 2 , a < 2b.
0 (x b sin[π(a − b)/b]
Z ∞ √ 1−2λ Γ(λ − 1/2)
xλ−1/2 dx √ √
16. λ λ
= π a+ b , λ > 0.
0 (x + a) (x + b) Γ(λ)
Z ∞
1 − xa c−1 π sin A πa πc
17. b
x dx = , A= , C= ; a + c < b,
0 1 − x b sin C sin(A + C) b b
c > 0.
Z ∞
xa−1 dx 1 1 1

18. 2 )1−b
= 2 B 2 a, 1 − b − 2 a , 12 a + b < 1, a > 0.
0 (1 + x
Z ∞
x2m dx π(2m − 1)!! (2n − 2m − 3)!!
19. 2 n
= √ , a, b > 0, n > m + 1.
0 (ax + b) 2 (2n − 2)!! am bn−m−1 ab
Z ∞ 2m+1
x dx m! (n − m − 2)!
20. 2 n
= , ab > 0, n > m + 1 ≥ 1.
0 (ax + b) 2(n − 1)!am+1 bn−m−1
Z µ
∞
xµ−1 dx 1 µ
21. = B , ν − , p > 0, 0 < µ < pν.
0 (1 + axp )ν paµ/p p p
Z ∞ p
n nan+1
22. x2 + a2 − x dx = 2 , n = 2, 3, . . .
0 n −1
Z ∞
dx n
23. √ n = n−1 2 , n = 2, 3, . . .
0 x + x2 + a2 a (n − 1)
Z ∞ p n m! nan+m+1
24. xm x2 + a2 − x dx = ,
0 (n − m − 1)(n − m + 1) . . . (n + m + 1)
n, m = 1, 2, . . . , 0 ≤ m ≤ n − 2.
Z ∞
xm dx m! n
25. √ n = ,
0 x+ x +a 2 2 (n − m − 1)(n − m + 1) . . . (n + m + 1)an−m−1
n = 2, 3, . . .
27.2.2 Integrals Involving Exponential Functions

Z ∞
1
1. e−ax dx = , a > 0.
0 a
Z 1 n
X
n −ax n! −a n! 1
2. x e dx = −e , a > 0, n = 1, 2, . . .
0 an+1 k! an−k+1
k=0
Z ∞
n!
3. xn e−ax dx = n+1 , a > 0, n = 1, 2, . . .
0 a
Z ∞ −ax r
e π
4. √ dx = , a > 0.
0 x a
Z ∞
Γ(ν)
5. xν−1 e−µx dx = ν , µ, ν > 0.
µ
Z0 ∞
dx ln 2
6. = .
0 1 + eax a
Z ∞ 2n−1 2n B
x dx n−1 2π 2n
7. px
= (−1) , n = 1, 2, . . . ; the Bm are Bernoulli
0 e −1 p 4n
numbers (see Section 30.1.3).
Z ∞ 2n−1 2π 2n |B |
x dx 1−2n 2n
8. px
= (1 − 2 ) , n = 1, 2, . . . ; the Bm are Bernoulli
0 e +1 p 4n
numbers.
Z ∞ −px
e dx π
9. −qx
= , q > p > 0 or 0 > p > q.
−∞ 1 + e q sin(πp/q)
Z ∞ ax
e + e−ax π
10. bx −bx
dx = πa , b > a.
0 e +e 2b cos
2b
Z ∞ −px
e − e−qx π πp
11. −(p+q)x
dx = cot , p, q > 0.
0 1−e p+q p+q
Z ∞
ν 1 µ
12. 1 − e−βx e−µx dx = B ,ν + 1 .
0 β β
Z ∞
r
2
1 π
13. exp −ax dx = , a > 0.
0 2 a
Z ∞
n!
14. x2n+1 exp −ax2 dx = n+1 , a > 0, n = 1, 2, . . .
0 2a
Z ∞ √
2n 2
1 × 3 × · · · × (2n − 1) π
15. x exp −ax dx = , a > 0, n = 1, 2, . . .
0 2n+1 an+1/2
Z ∞ √ b2
2 2
π
16. exp −a x ± bx dx = exp .
−∞ |a| 4a2
Z ∞ r
2 b 1 π √
17. exp −ax − 2 dx = exp −2 ab , a, b > 0.
x 2 a
Z0 ∞
1 1
18. exp −xa dx = Γ , a > 0.
0 a a
27.2.3 Integrals Involving Hyperbolic Functions

Z ∞
dx π
1. = .
0 cosh ax 2|a|
 √
Z ∞ 
 2 b2 − a2
 √ arctan if |b| > |a|,
dx a+
2. = b2 − a2 √b
0 a + b cosh x   1 a + b + a2 − b2
√ ln √ if |b| < |a|.
a2 − b2 a + b − a2 + b2
Z ∞ 2n π 2n+1
x dx
3. = |E2n |, a > 0, the Em are Euler numbers
0 cosh ax 2a
(see Section 30.1.4).
Z ∞ 2n
x dx π 2n (22n −2)
4. = |B2n |, the Bm are Bernoulli numbers
0 cosh2 ax |a|(2a)2n
(see Section 30.1.3).
Z ∞
cosh ax π
5. dx = πa , b > |a|.
0 cosh bx 2b cos
2b
Z ∞ 2n
cosh ax π d 1
6. x2n dx = 2n 1
, b > |a|, n = 1, 2, . . .
0 cosh bx 2b da cos 2 πa/b
πa πb
Z ∞ cos cos
cosh ax cosh bx π 2c 2c
7. dx = πa πb , c > |a| + |b|.
0 cosh(cx) c
cos + cos
c c
Z ∞ 2
x dx π
8. = 2 , a > 0.
0 sinh ax 2a
Z ∞ √
dx 1 a + b + a2 + b2
9. = √ ln √ , ab 6= 0.
0 a + b sinh x a2 + b2 a + b − a2 + b2
Z ∞ πa
sinh ax π
10. dx = tan , b > |a|.
0 sinh bx 2b 2b
Z ∞ πa
sinh ax π d2n
11. x2n dx = tan , b > |a|, n = 1, 2, . . .
0 sinh bx 2b dx2n 2b
Z ∞
x2n π 2n
12. dx = |B2n |, a > 0; the Bm are Bernoulli numbers.
0 sinh2 ax a2n+1
27.2.4 Integrals Involving Logarithmic Functions

Z 1
1. xa−1 lnn x dx = (−1)n n! a−n−1 , a > 0, n = 1, 2, . . .
0
Z 1
ln x π2
2. dx = − .
0 x+1 12
Z 1 2 X n
xn ln x π (−1)k
3. dx = (−1)n+1 + , n = 1, 2, . . .
0 x+1 12 k2
k=1
Z 1
xµ−1 ln x πaµ−1
4. dx = ln a − π cot(πµ) , 0 < µ < 1.
0 x+a sin(πµ)
Z 1
5. |ln x|µ dx = Γ(µ + 1), µ > −1.
0
Z ∞
π
6. xµ−1 ln(1 + ax) dx = , −1 < µ < 0.
0 µaµ sin(πµ)
Z 2n
1
1 X (−1)k−1
7. x2n−1 ln(1 + x) dx = , n = 1, 2, . . .
0 2n k
k=1
Z 1 X (−1)k
2n+1
2n 1
8. x ln(1 + x) dx = ln 4 + , n = 0, 1, . . .
0 2n + 1 k
k=1
Z 1 X n
n−1/2 2 ln 2 4(−1)n (−1)k
9. x ln(1 + x) dx = + π− , n = 1, 2, . . .
0 2n + 1 2n + 1 2k + 1
k=0
Z ∞
a2 + x2
10. ln 2 dx = π(a − b), a, b > 0.
0 b + x2
Z ∞ p−1
x ln x π 2 cos(πp/q)
11. dx = − , 0 0, C = 0.5772 . . .
0 µ
27.2.5 Integrals Involving Trigonometric Functions

◮ Integrals over a finite interval.
Z π/2
π 1 × 3 × · · · × (2n − 1)
1. cos2n x dx = , n = 1, 2, . . .
0 2 2 × 4 × · · · × (2n)
Z π/2
2 × 4 × · · · × (2n)
2. cos2n+1 x dx = , n = 1, 2, . . .
0 1 × 3 × · · · × (2n + 1)
Z π/2 m−1
X
n (n − 2k + 1)(n − 2k + 3) . . . (n − 1) 1
3. x cos x dx = −
0 (n − 2k)(n − 2k + 2) . . . n n − 2k
k=0 

 π (2m − 2)!!
 if n = 2m − 1,
2 (2m − 1)!!
+ 2 m = 1, 2, . . .
 π (2m − 1)!! if n = 2m,


8 (2m)!!
Z
(2n−2k −1)!! (2k −1)!! a+b k
π X n
dx π
4. = √ ,
0 (a+b cos x)n+1 2n (a+b)n a2 −b2 k=0 (n−k)! k! a−b
a > |b|.
Z π/2
π 1 × 3 × · · · × (2n − 1)
5. sin2n x dx = , n = 1, 2, . . .
0 2 2 × 4 × · · · × (2n)
Z π/2
2 × 4 × · · · × (2n)
6. sin2n+1 x dx = , n = 1, 2, . . .
0 1 × 3 × · · · × (2n + 1)
Z π
π2 Γ(µ + 1)
7. x sinµ x dx = µ+1 2 , µ > −1.
0 2 Γ µ + 12
Z π/2
sin x dx 1 1+k
8. p = ln .
0 1 − k2 sin x2 2k 1−k
Z π/2
n! m!
9. sin2n+1 x cos2m+1 x dx = , n, m = 1, 2, . . .
0 2(n + m + 1)!
Z π/2

10. sinp−1 x cosq−1 x dx = 12 B 12 p, 12 q .
0
Z 2π
(2n − 1)!! 2 n
11. (a sin x + b cos x)2n dx = 2π a + b2 , n = 1, 2, . . .
0 (2n)!!
Z π (
sin x dx 2 if 0 ≤ a ≤ 1,
12. √ =
0 a2 + 1 − 2a cos x 2/a if 1 < a.
Z π/2
π
13. (tan x)±λ dx = , |λ| < 1.
0 2 cos 12 πλ
Z a
cos(xt) dt π
14. √ = J0 (ax), J0 (z) is the Bessel function (see Section 30.6).
2
a −t 2 2
Z0 a
t sin(xt) dt π
15. √ = aJ1 (ax), J1 (z) is the Bessel function.
0
2
a −t 2 2
Z 2π
16. cos(a cos x) dx = 2πJ0 (a), J0 (z) is the Bessel function.
0
Z 2π
17. sin(a cos x) dx = 0.
0

Z ∞ r
cos ax π
1. √ dx = , a > 0.
0 x 2a
Z ∞ b
cos ax − cos bx
2. dx = ln , ab 6= 0.
x a
Z0 ∞
cos ax − cos bx
3. dx = 12 π(b − a), a, b ≥ 0.
0 x2
Z ∞

4. xµ−1 cos ax dx = a−µ Γ(µ) cos 12 πµ , a > 0, 0 < µ < 1.
Z0 ∞
cos ax π −ab
5. 2 + x2
dx = e , a, b > 0.
0 b 2b
Z ∞ √ ab
cos ax π 2 ab ab
6. dx = exp − √ cos √ + sin √ , a, b > 0.
0 b4 + x4 4b3 2 2 2
Z ∞
cos ax π
7. 2 + x2 )2
dx = 3 (1 + ab)e−ab , a, b > 0.
0 (b 4b
Z ∞
cos ax dx π be−ac − ce−ab
8. = , a, b, c > 0.
0 (b2 + x2 )(c2 + x2 ) 2bc b2 − c2
Z ∞ r
2
1 π
9. cos ax dx = , a > 0.
2 2a
Z0 ∞
Γ(1/p) π
10. cos axp dx = 1/p
cos , a > 0, p > 1.
0 pa 2p
Z ∞
sin ax π
11. dx = sign a.
0 x 2
Z ∞ 2
sin ax π
12. 2
dx = |a|.
0 x 2
Z ∞ r
sin ax π
13. √ dx = , a > 0.
x 2a
Z0 ∞

14. xµ−1 sin ax dx = a−µ Γ(µ) sin 12 πµ , a > 0, 0 < µ < 1.
0
Z ∞ r
2
1 π
15. sin ax dx = , a > 0.
2 2a
Z0 ∞
Γ(1/p) π
16. sin axp dx = 1/p
sin , a > 0, p > 1.
0 pa 2p

π
Z ∞
sin x cos ax  2 if |a| < 1,

17. dx = π4 if |a| = 1,
0 x 

0 if 1 < |a|.
Z ∞
tan ax π
18. dx = sign a.
x 2
Z0 ∞
b
19. e−ax sin bx dx = 2 , a > 0.
a + b2
Z0 ∞
a
20. e−ax cos bx dx = 2 , a > 0.
a + b2
Z0 ∞
2ab
21. xe−ax sin bx dx = 2 , a > 0.
0 (a + b2 )2
Z ∞
a2 − b2
22. xe−ax cos bx dx = , a > 0.
0 (a2 + b2 )2
Z ∞
∂n b
23. xn e−ax sin bx dx = (−1)n n , a > 0, n = 1, 2, . . . .
0 ∂a a2 + b2
Z ∞ n
n −ax n ∂ a
24. x e cos bx dx = (−1) , a > 0, n = 1, 2, . . . .
0 ∂an a2 + b2
Z ∞ r b2
1 π
25. exp −ax2 cos bx dx = exp − , a > 0.
0 2 a 4a
Z ∞ √ b2
2
πb
26. x exp −ax sin bx dx = 3/2 exp − , a > 0.
0 4a 4a
Z ∞ r 2
2 π b2 b
27. cos(ax ) cos bx dx = cos + sin , a, b > 0.
0 8a 4a 4a
Z ∞ r 2 2 2 2
2 π b b b b
28. cos(ax ) sin bx dx = cos C − sin S ,
0 2a 4a 4a 4a 4a
a, b > 0 and C(z) and S(z) are Fresnel integrals.
Z ∞ r 2 2
2 π b b
29. sin(ax ) cos bx dx = cos − sin , a, b > 0.
0 8a 4a 4a
Z ∞ r 2 2 2 2
π b b b b
30. sin(ax2 ) sin bx dx = cos C + sin S ,
0 2a 4a 4a 4a 4a
Z ∞ 2 2 2
1 2 bπ b b √ b π
31. 2
sin(ax ) cos bx dx = S −C + πa sin + ,
0 x 2 4a 4a 4a 4
Z ∞ r 2
2 2 1 π a
32. (cos ax + sin ax) cos(b x ) dx = exp − , a, b > 0.
0 b 8 2b
Z ∞ r a2
2 2 1 π
33. cos ax + sin ax sin(b x ) dx = exp − , a, b > 0.
0 b 8 2b
27.2.6 Integrals Involving Bessel Functions

Z ∞
1
1. Jν (ax) dx = , a > 0, Re ν > −1.
0 a

Z ∞ √ 1 if x < t,
2. cos(xu)J0 (tu) du = t 2 − x2
0 
0 if x > t.

Z ∞ 0 if x < t,
3. sin(xu)J0 (tu) du = 1
0 √ if x > t.
x 2 − t2

Z ∞  1
 if x < t,
4. cos(xu)J1 (tu) du = t t
0 
− √ 2 √ if x > t.
x − t (x + x2 − t2 )
2
Z ∞
sin(tu)J0 (au) sinh(bt)
5. 2 2
du = K0 (ab), b > 0, 0 < t < a, K0 (z) is the modified
0 u +b b
Bessel function (see Section 30.7).
Z ∞
u sin(tu)J0 (au) π
6. 2 2
du = e−bt I0 (ab), b > 0, a < t < ∞, I0 (z) is the modified
0 u +b 2
Bessel
Z ∞ function.
sin(tu)J1 (au) π −bt
7. 2 2
du = e I1 (ab), b > 0, a < t < ∞, I1 (z) is the modified
0 u +b 2b
Bessel
Z ∞ function.
u sin(tu)J1 (au)
8. du = sinh(bt)K1 (ab), b > 0, 0 < t < a, K1 (z) is the modified
0 u2 + b2
Bessel function.
Z ∞
J (au) 1 − e−ab
9. √1 du = , a > 0, Re b > 0.
0 u2 + b2 ab
◮ Other integrals.
Z 1
J1 (x)
1. uJ0 (xu) du = .
0 x
Z a
J1 (bx) dx 1 − cos(ab)
2. √ = , a > 0.
0
2
a −x 2 ab
Z t
uJ0 (xu) du sin(xt)
3. √ = .
0 t2 − u2 x
Z ∞
J1 (xu) du sin(xt)
4. √ = , x > 0, t > 0.
t
2
u −t 2 x
⊙ References for Chapter 27: H. B. Dwight (1961), I. S. Gradshteyn and I. M. Ryzhik (2000), A. P. Prud-
nikov, Yu. A. Brychkov, and O. I. Marichev (1986, 1988), D. Zwillinger (2002), I. N. Bronshtein and K. A. Se-
mendyayev (2004).
Chapter 28
Integral Transforms
28.1 Tables of Laplace Transforms

28.1.1 General Formulas
Z ∞
No. Original function, f (x) Laplace transform, fe(p) = e−px f (x) dx
0
1 af1 (x) + bf2 (x) afe1 (p) + bfe2 (p)
2 f (x/a), a > 0 afe(ap)

0 if 0 < x < a,
3 e−ap fe(p)
f (x − a) if x > a
dn e
4 xn f (x); n = 1, 2, . . . (−1)n f (p)
dpn
Z ∞
1
5 f (x) fe(q) dq
x p
6 eax f (x) fe(p − a)

1 e

7 sinh(ax)f (x) 2
f (p − a) − fe(p + a)

1 e

8 cosh(ax)f (x) 2
f (p − a) + fe(p + a)

9 sin(ωx)f (x) − 2i fe(p − iω) − fe(p + iω) , i2 = −1

1 e

10 cos(ωx)f (x) 2
f (p − iω) + fe(p + iω) , i2 = −1
Z ∞
1 p2
11 f (x2 ) √ exp − 2 fe(t2 ) dt
π 0 4t
1 Z ∞ p
12 xa−1 f , a > −1 (t/p)a/2 Ja 2 pt fe(t) dt
x 0
Z ∞
13 f (a sinh x), a > 0 Jp (at)fe(t) dt
0
Z a
f (x + a) = f (x) 1
14 f (x)e−px dx
(periodic function) 1 − eap 0
Z a
f (x + a) = −f (x) 1
15 f (x)e−px dx
(antiperiodic function) 1 + e−ap 0
1463
1464 I NTEGRAL T RANSFORMS
Z ∞
0
16 fx′ (x) pfe(p) − f (+0)

n
X
17 fx(n) (x) pn fe(p) − pn−k fx(k−1) (+0)
k=1
d m n e
18 xm fx(n) (x), m ≥ n − p f (p)
dp
dn m dm e
19 x f (x) , m ≥ n (−1)m pn f (p)
dx n dpm
Z x
f (t) dt fe(p)
20
0 p
Z x
1 e
21 (x − t)f (t) dt f (p)
0 p2
Z x
22 (x − t)ν f (t) dt, ν > −1 Γ(ν + 1)p−ν−1 fe(p)
0
Z x
1 e
23 e−a(x−t) f (t) dt f (p)
0 p+a
Z x afe(p)
24 sinh a(x − t) f (t) dt
0 p 2 − a2
Z x afe(p)
25 sin a(x − t) f (t) dt
0 p 2 + a2
Z x
26 f1 (t)f2 (x − t) dt fe1 (p)fe2 (p)
0
Z x
Z ∞
1 1
27 f (t) dt fe(q) dq
0 t p p
Z ∞ Z
1 1 p e
28 f (t) dt f (q) dq
x t p 0
Z √
1∞ √ π 1
29 √ sin 2 xt f (t) dt √ fe
0 t p p p
Z ∞ √ √
1 π 1
30 √ cos 2 xt f (t) dt √ fe
x 0 p p
Z ∞ t2
1 1 √
31 √ exp − f (t) dt √ fe p
0 πx 4x p
Z ∞ t2
t √
32 √ exp − f (t) dt fe p
0 2 πx3 4x
Z x p
p
33 f (x) − a f x2 − t2 J1 (at) dt fe p 2 + a2
0
Z x p p
34 f (x) + a f x2 − t2 I1 (at) dt fe p 2 − a2
0
28.1. Tables of Laplace Transforms 1465
28.1.2 Expressions with Power-Law Functions

Z ∞
0
1
1 1
p
(
0 if 0 < x < a, 1 −ap
2 1 if a < x < b, e − e−bp
p
0 if b < x
1
3 x
p2
1
4 −eap Ei(−ap)
x+a
n!
5 xn , n = 1, 2, . . .
pn+1
√
1 · 3 . . . (2n − 1) π
6 xn−1/2 , n = 1, 2, . . .
2n pn+1/2
r
1 π ap √
7 √ e erfc ap
x+a p
√ r
x π √ √
8 − π aeap erfc ap
x+a p
√
9 (x + a)−3/2 2a−1/2 − 2(πp)1/2 eap erfc ap
√
10 x1/2 (x + a)−1 (π/p)1/2 − πa1/2 eap erfc ap
√
11 x−1/2 (x + a)−1 πa−1/2 eap erfc ap
12 xν , ν > −1 Γ(ν + 1)p−ν−1
ν
13 (x + a) , ν > −1 p−ν−1 e−ap Γ(ν + 1, ap)
14 xν (x + a)−1 , ν > −1 keap Γ(−ν, ap), k = aν Γ(ν + 1)
15 (x2 + 2ax)−1/2 (x + a) aeap K1 (ap)
28.1.3 Expressions with Exponential Functions

Z ∞
0
1 e−ax (p + a)−1
−ax
2 xe (p + a)−2
3 xν−1 e−ax , ν > 0 Γ(ν)(p + a)−ν
1 −ax
4 e − e−bx ln(p + b) − ln(p + a)
x
1 2
5 1 − e−ax (p + 2a) ln(p + 2a) + p ln p − 2(p + a) ln(p + a)
x2
√ 1
6 exp −ax2 , a > 0 (πb)1/2 exp bp2 erfc(p b ), a =
4b
√ 1
7 x exp −ax2 2b − 2π 1/2 b3/2 p erfc(p b ), a =
4b
p √
8 exp(−a/x), a ≥ 0 2 a/p K1 2 ap
√ 1
p √ √
9 x exp(−a/x), a ≥ 0 2
π/p3 1 + 2 ap exp −2 ap
1 p √
10 √ exp(−a/x), a ≥ 0 π/p exp −2 ap
x
Z ∞
0
1 p √
11 √ exp(−a/x), a>0 π/a exp −2 ap
x x
√
12 xν−1 exp(−a/x), a>0 2(a/p)ν/2 Kν 2 ap
√ p
13 exp −2 ax p−1 − (πa)1/2 p−3/2 ea/p erfc a/p
1 √ p
14 √ exp −2 ax (π/p)1/2 ea/p erfc a/p
x
28.1.4 Expressions with Hyperbolic Functions

Z ∞
0
a
1 sinh(ax)
p 2 − a2
2a2
2 sinh2 (ax)
p3 − 4a2 p
1 1 p+a
3 sinh(ax) ln
x 2 p−a

4 xν−1 sinh(ax), ν > −1 1
2
Γ(ν) (p − a)−ν − (p + a)−ν
√
√ πa a/p
5 sinh 2 ax √ e
p p
√ √ p
6 x sinh 2 ax π 1/2 p−5/2 12 p + a ea/p erf a/p − a1/2 p−2
1 √ p
7 √ sinh 2 ax π 1/2 p−1/2 ea/p erf a/p
x
1 √
8 √ sinh2 ax 1 1/2 −1/2
π p ea/p − 1
x 2
p
9 cosh(ax)
p 2 − a2
p2 − 2a2
10 cosh2 (ax)
p3 − 4a2 p

11 xν−1 cosh(ax), ν>0 1
2
Γ(ν) (p − a)−ν + (p + a)−ν
√ p
√ 1 πa
12 cosh 2 ax + √ ea/p erf a/p
p p p
√ √
13 x cosh 2 ax π 1/2 p−5/2 12 p + a ea/p
1 √
14 √ cosh 2 ax π 1/2 p−1/2 ea/p
x
1 √
15 √ cosh2 ax 1 1/2 −1/2
π p ea/p + 1
x 2
28.1.5 Expressions with Logarithmic Functions

Z ∞
0
1
− (ln p + C),
1 ln x p
C = 0.5772 . . . is the Euler constant
1
2 ln(1 + ax) − ep/a Ei(−p/a)
p
1
3 ln(x + a) ln a − eap Ei(−ap)
p
n! 1 1 1

1+ + + ··· + − ln p − C ,
4 xn ln x, n = 1, 2, . . . pn+1 2 3 n

1 p
5 √ ln x − π/p ln(4p) + C
x
kn 2 22

2+ 3
+
+ · · · + 2n−1
5
− ln(4p) − C ,
6 x n−1/2
ln x, n = 1, 2, . . . pn+1/2 √
π
kn = 1 · 3 · 5 . . . (2n − 1) n , C = 0.5772 . . .
2

Γ(ν)p−ν ψ(ν) − ln p ,
7 xν−1 ln x, ν>0
ψ(ν) is the logarithmic derivative of the gamma function
1
8 (ln x)2 (ln x + C)2 + 16 π 2 , C = 0.5772 . . .
p
ln(p + a) + C
9 e−ax ln x − , C = 0.5772 . . .
p+a
28.1.6 Expressions with Trigonometric Functions

Z ∞
0
a
1 sin(ax)
p 2 + a2
a πp
2 |sin(ax)|, a>0 coth
p 2 + a2 2a
a2n (2n)!
3 sin2n (ax), n = 1, 2, . . .
p p2 + (2a)2 p2 + (4a)2 . . . p2 + (2na)2
a2n+1 (2n + 1)!

4 sin2n+1 (ax), n = 1, 2, . . .
p 2 + a2 p2 + 32 a2 . . . p2 + (2n + 1)2 a2
n! pn+1 X a 2k+1
5 xn sin(ax), n = 1, 2, . . . n+1 (−1)k Cn+1
2k+1
p 2 + a2 p
0≤2k≤n
a
1 arctan
6 sin(ax)
x p
1
7 sin2 (ax) 1
4
ln 1 + 4a2 p−2
x
Z ∞
0
1
8 sin2 (ax) a arctan(2a/p) − 14 p ln 1 + 4a2 p−2
x2
√
√ πa −a/p
9 sin 2 ax √ e
p p
1 √ p
10 sin 2 ax π erf a/p
x
p
11 cos(ax)
p 2 + a2
p2 + 2a2
12 cos2 (ax)
p p2 + 4a2
n! pn+1 X a 2k
13 xn cos(ax), n = 1, 2, . . . n+1 (−1)k Cn+1
2k
p2+ a2 0≤2k≤n+1
p
1
14 1 − cos(ax) 1
2
ln 1 + a2 p−2
x
1 1 p 2 + b2
15 cos(ax) − cos(bx) ln 2
x 2 p + a2
√ √ 1 1/2 −5/2
16 x cos 2 ax 2
π p (p − 2a)e−a/p
1 √ p
17 √ cos 2 ax π/p e−a/p
x
2abp
18 sin(ax) sin(bx)
p2
+ (a + b)2 p2 + (a − b)2

b p 2 − a 2 + b2
19 cos(ax) sin(bx)
p2 + (a + b)2 p2 + (a − b)2

p p 2 + a 2 + b2
20 cos(ax) cos(bx)
p2 + (a + b)2 p2 + (a − b)2
ax cos(ax) − sin(ax) a
21 p arctan −a
x2 x
a
22 ebx sin(ax) (p − b)2 + a2
p−b
23 ebx cos(ax)
(p − b)2 + a2
2a2 p
24 sin(ax) sinh(ax)
p4 + 4a4

a p2 + 2a2
25 sin(ax) cosh(ax)
p4 + 4a4

a p2 − 2a2
26 cos(ax) sinh(ax)
p4 + 4a4
p3
27 cos(ax) cosh(ax)
p4 + 4a4
28.1.7 Expressions with Special Functions

Z ∞
0
1 1
1 erf(ax) exp b2 p2 erfc(bp), b=
p 2a
√
√ a
2 erf ax √
p p+a
√
√ a
3 eax erf ax √
p (p − a)
1
p 1 √
4 erf a/x 1 − exp − ap
2 p
√ √
√ p+a− a
5 erfc ax √
p p+a
√ 1
6 eax erfc ax √
p+ ap
1
p 1 √
7 erfc a/x exp − ap
2 p
1
8 Ci(x) ln(p2 + 1)
2p
1
9 Si(x) arccot p
p
1
10 Ei(−x) − ln(p + 1)
p
1
11 J0 (ax) p
p 2 + a2
aν
12 Jν (ax), ν > −1 p p ν
p 2 + a2 p + p 2 + a2
−n−1/2
13 xn Jn (ax), n = 1, 2, . . . 1 · 3 · 5 . . . (2n − 1)an p2 + a2
−ν−1/2
14 xν Jν (ax), ν > − 12 2ν π −1/2 Γ ν + 12 aν p2 + a2
−ν−3/2
15 xν+1 Jν (ax), ν > −1 2ν+1 π −1/2 Γ ν + 32 aν p p2 + a2
√ 1 −a/p
16 J0 2 ax e
p
√
√ √ a −a/p
17 xJ1 2 ax e
p2
√
18 xν/2 Jν 2 ax , ν > −1 aν/2 p−ν−1 e−a/p
1
19 I0 (ax) p
p 2 − a2
aν
20 Iν (ax), ν > −1 p p ν
p 2 − a2 p + p 2 − a2
−ν−1/2
21 xν Iν (ax), ν > − 12 2ν π −1/2 Γ ν + 12 aν p2 − a2
−ν−3/2
22 xν+1 Iν (ax), ν > −1 2ν+1 π −1/2 Γ ν + 32 aν p p2 − a2
Z ∞
0
√ 1 a/p
23 I0 2 ax e
p
1 √ 1
24 √ I1 2 ax √ ea/p − 1
x a
√
25 xν/2 Iν 2 ax , ν > −1 aν/2 p−ν−1 ea/p
2 arcsinh(p/a)
26 Y0 (ax) − p
π p 2 + a2
p
ln p + p2 − a2 − ln a
27 K0 (ax) p
p 2 − a2
⊙ Literature for Section 28.1: G. Doetsch (1950, 1956, 1958), H. Bateman and A. Erdélyi (1954), V. A.
Ditkin and A. P. Prudnikov (1965), F. Oberhettinger and L. Badii (1973), A. P. Prudnikov, Yu. A. Brychkov,
and O. I. Marichev (1992a, Vol. 4).
28.2 Tables of Inverse Laplace Transforms

Z c+i∞
1
No. Laplace transform, fe(p) Inverse transform, f (x) = epx fe(p) dp
2πi c−i∞
−ax
1 fe(p + a) e f (x)
1 x
2 fe(ap), a>0 f
a a
1 b x
3 fe(ap + b), a>0 exp − x f
a a a
4 fe(p − a) + fe(p + a) 2f (x) cosh(ax)
5 fe(p − a) − fe(p + a) 2f (x) sinh(ax)

0 if 0 ≤ x < a,
6 e−ap fe(p), a≥0
f (x − a) if a < x.
df (x)
7 pfe(p) if f (+0) = 0
dx
Z x
1e
8 f (p) f (t) dt
p 0
Z x
1 e
9 f (p) e−ax eat f (t) dt
p+a 0
Z x
1 e
10 f (p) (x − t)f (t) dt
p2 0
Z
fe(p) 1 x
11 1 − ea(x−t) f (t) dt
p(p + a) a 0
Z x
fe(p)
12 (x − t)e−a(x−t)f (t) dt
(p + a)2 0
Z x
fe(p) 1
13 e−a(x−t) − e−b(x−t) f (t) dt
(p + a)(p + b) b−a 0
28.2. Tables of Inverse Laplace Transforms 1471
Z c+i∞
1
2πi c−i∞
Z x
fe(p) 1
14 e−a(x−t) sin b(x − t) f (t) dt
(p + a)2 + b2 b 0
Z x
1 e 1
15 f (p), n = 1, 2, . . . (x − t)n−1 f (t) dt
pn (n − 1)! 0
Z x
16 fe1 (p)fe2 (p) f1 (t)f2 (x − t) dt
0
Z √
1 1 ∞
cos 2 xt
17 √ fe √ f (t) dt
p p 0 πx
Z ∞ √
1 1 sin 2 xt
18 √ fe √ f (t) dt
p p p 0 πt
Z ∞
1 e 1 √
19 2ν+1
f (x/t)ν J2ν 2 xt f (t) dt
p p 0
Z ∞
1 e 1 √
20 f J0 2 xt f (t) dt
p p 0
Z x
1e 1 p
21 f p+ J0 2 xt − t2 f (t) dt
p p 0
Z x
1 e a 1 x − t ν p
22 2ν+1
f p+ , − <ν≤0 J2ν 2 axt − at2 f (t) dt
p p 2 0 at
Z ∞ t2
√ t
23 fe p √ exp − f (t) dt
0 2 πx 3 4x
Z ∞ t2
1 √ 1
24 √ fe p √ exp − f (t) dt
p πx 0 4x
Z x h i
√ 1 t t2
25 fe p + p √ 3/2
exp − f (t) dt
2 π 0 (x − t) 4(x − t)
Z x p
p
26 fe p 2 + a2 f (x) − a f x2 − t2 J1 (at) dt
0
Z x p
p
27 fe p 2 − a2 f (x) + a f x2 − t2 I1 (at) dt
0
p Z
fe p2 + a2 x p
28 p J0 a x2 − t2 f (t) dt
p 2 + a2 0
p Z
fe p2 − a2 x p
29 p I0 a x2 − t2 f (t) dt
p 2 − a2 0
Z x p
p
30 fe (p + a)2 − b2 e−ax f (x) + be−ax f x2 − t2 I1 (bt) dt
0
Z ∞
xt−1
31 fe(ln p) f (t) dt
0 Γ(t)
Z ∞
1e xt
32 f (ln p) f (t) dt
p 0 Γ(t + 1)
33 fe(p − ia) + fe(p + ia), i2 = −1 2f (x) cos(ax)

34 i fe(p − ia) − fe(p + ia) , i2 = −1 2f (x) sin(ax)
dfe(p)
35 −xf (x)
dp
Z c+i∞
1
2πi c−i∞
dn fe(p)
36 (−x)n f (x)
dpn
dm fe(p) dn m
37 pn , m≥n (−1)m n
x f (x)
Z ∞ dp
m dx
fe(q) dq 1
38 f (x)
pZ x
Z ∞
1 p e f (t)
39 f (q) dq dt
p Z0
∞ Z x t
x
1 f (t)
40 fe(q) dq dt
p p 0 t
28.2.2 Expressions with Rational Functions

Z c+i∞
1
2πi c−i∞
1
1 1
p
1
2 e−ax
p+a
1
3 x
p2
1 1
4 1 − e−ax
p(p + a) a
1
5 xe−ax
(p + a)2
p
6
(p + a)2 (1 − ax)e−ax
1 1
7 sinh(ax)
p 2 − a2 a
p
8 cosh(ax)
p 2 − a2
1 1
9 e−bx − e−ax
(p + a)(p + b) a−b
p 1
10
(p + a)(p + b) ae−ax − be−bx
a−b
1 1
11 sin(ax)
p 2 + a2 a
p
12 cos(ax)
p 2 + a2
1 1 −bx
13 e sin(ax)
(p + b)2 + a2 a
p h b i
14
(p + b)2 + a2 e−bx cos(ax) − sin(ax)
a
1 1 2
15 x
p3 2
1 1 −ax
16 e + ax − 1
p2 (p + a) a2
1 1
17 a − b + be−ax − ae−bx
p(p + a)(p + b) ab(a − b)
1 1
18 1 − e−ax − axe−ax
p(p + a)2 a2
1 (c − b)e−ax + (a − c)e−bx + (b − a)e−cx
19
(p + a)(p + b)(p + c) (a − b)(b − c)(c − a)
Z c+i∞
1
2πi c−i∞
p a(b − c)e−ax + b(c − a)e−bx + c(a − b)e−cx

20
(p + a)(p + b)(p + c) (a − b)(b − c)(c − a)
p2 a2 (c − b)e−ax + b2 (a − c)e−bx + c2 (b − a)e−cx
21
(p + a)(p + b)(p + c) (a − b)(b − c)(c − a)
1 1 −ax
22 e − e−bx + (a − b)xe−bx
(p + a)(p + b)2 (a − b) 2
p 1
23 −ae−ax + [a + b(b − a)x e−bx
(p + a)(p + b)2 (a − b)2
p2 1 2 −ax
24 a e + b(b − 2a − b2 x + abx)e−bx
(p + a)(p + b)2 (a − b)2
1 1 2 −ax
25 x e
(p + a)3 2
p
26
(p + a)3 x 1 − 12 ax e−ax
p2
27 1 − 2ax + 12 a2 x2 e−ax
(p + a)3
1 1
28 1 − cos(ax)
p(p2 + a2 ) a 2
h i
1 1 −bx b
29 1 − e cos(ax) + sin(ax)
p (p + b)2 + a2 a 2 + b2 a
1 1 h a i
30 e−ax + sin(bx) − cos(bx)
(p + a)(p2 + b2 ) 2
a +b 2 b
p 1
31
(p + a)(p2 + b2 ) −ae−ax + a cos(bx) + b sin(bx)
a 2 + b2
p2 1 2 −ax
32 a e − ab sin(bx) + b2 cos(bx)
(p + a)(p2 + b2 ) a 2 + b2
e−ax − eax/2 √
1 cos(kx) − 3 sin(kx) ,
33 3a2√
p 3 + a3
k = 12 a 3
e−ax − eax/2 √
p − cos(kx) + 3 sin(kx) ,
34 3a
√
p 3 + a3
k = 12 a 3
p2 1 −ax
√
35 3
e + 23 eax/2 cos(kx), k= 1
2
a 3
p3 + a3
e−ax − e−bx cos(cx) + ke−bx sin(cx)
1 ,
36 (a − b)2 + c2
p + a) (p + b)2 + c2 ] a−b
k=
c
−ae−ax + ae−bx cos(cx) + ke−bx sin(cx)
p ,
37 (a − b)2 + c2
p + a) (p + b)2 + c2 ] 2 2
b + c − ab
k=
c
Z c+i∞
1
2πi c−i∞
a2e−ax + (b2 + c2 − 2ab)e−bx cos(cx) + ke−bx sin(cx)

,
p2 (a − b)2 + c2
38
p + a) (p + b)2 + c2 ] ab2 − b3
k = −ac − bc +
c
1 1 3
39 x
p4 6
1 1 1 1 2 1
40 − 2x+ x − 3 e−ax
p3 (p + a) a3 a 2a a
1 1 2
41 x 1 + e−ax + 3 e−ax − 1
p2 (p + a)2 a2 a
1 a+b 1 1 1
42 − 2 2 + x+ 2 e−ax + 2 e−bx
p2 (p + a)(p + b) a b ab a (b − a) b (a − b)
1 1 h 2 2 i
43 e−ax x + + e−bx x −
(p + a)2 (p + b)2 (a − b) 2 a−b a−b
1 1 3 −ax
44 x e
(p + a)4 6
p 1 2 −ax
45 (p + a)4 2x e − 16 ax3 e−ax
1 1
46 ax − sin(ax)
p2 (p2 + a2 ) a3
1 1
47 sinh(ax) − sin(ax)
p 4 − a4 2a3
p 1
48
p 4 − a4 cosh(ax) − cos(ax)
2a2
p2 1
49 sinh(ax) + sin(ax)
p 4 − a4 2a
p3 1
50 cosh(ax) + cos(ax)
p4 − a4 2
1 1 ax
51 √ cosh ξ sin ξ − sinh ξ cos ξ , ξ = √
p 4 + a4 a3 2 2
p 1 ax ax
52 sin √ sinh √
p 4 + a4 a2 2 2
p2 1 ax
53 √ cos ξ sinh ξ + sin ξ cosh ξ , ξ = √
p4 + a4 a 2 2
1 1
54 sin(ax) − ax cos(ax)
(p2 + a2 )2 2a3
p 1
55
(p2 + a2 )2 x sin(ax)
2a
p2 1
56 sin(ax) + ax cos(ax)
(p2 + a 2 )2 2a
p3
57 cos(ax) − 12 ax sin(ax)
(p2 + a 2 )2
1 1 −bx
58 e sin(ax) − ax cos(ax)
[(p + b)2 + a2 ]2 2a3
1 1 h1 1 i
59 sinh(ax) − sinh(bx)
(p2 − a2 )(p2 − b2 ) a 2 − b2 a b
Z c+i∞
1
2πi c−i∞
p cosh(ax) − cosh(bx)
60
(p2 − a2 )(p2 − b2 ) a 2 − b2
p2 a sinh(ax) − b sinh(bx)
61
(p2 − a2 )(p2 − b2 ) a 2 − b2
p3 a2 cosh(ax) − b2 cosh(bx)
62
(p2 − a2 )(p2
− b2 ) a 2 − b2
1 1 h 1 1 i
63 sin(ax) − sin(bx)
(p2 + a2 )(p2 + b2 ) 2
b −a a2 b
p cos(ax) − cos(bx)
64
(p2 + a2 )(p2 + b2 ) b2 − a 2
p2 −a sin(ax) + b sin(bx)
65
(p2 + a2 )(p2 + b2 ) b2 − a 2
p3 −a2 cos(ax) + b2 cos(bx)
66
(p2 + a2 )(p2 + b2 ) b2 − a 2
1 1
67 , n = 1, 2, . . . xn−1
pn (n − 1)!
1 1
68 , n = 1, 2, . . . xn−1 e−ax
(p + a)n (n − 1)!

1 a−n 1 − e−ax en (ax) ,
69 , n = 1, 2, . . . z zn
p(p + a)n en (z) = 1 + +··· +
1! n!
Xn
1
1 − 2n exp(ak x) ak cos(bk x) − bk sin(bk x) ,
70 , n = 1, 2, . . . na k=1
p2n + a2n π(2k − 1)
ak = a cos ϕk , bk = a sin ϕk , ϕk =
2n
n
1 1 X
sinh(ax) + exp(ak x)
1 na2n−1 na2n k=2
71 , n = 1, 2, . . .
p2n − a2n × ak cos(bk x) − bk sin(bk x) ,
π(k − 1)
n
X n
e−ax 2
− exp(ak x)
1 (2n + 1)a2n (2n + 1)a2n+1 k=1
72 , n = 0, 1, . . .
p2n+1 + a2n+1 × ak cos(bk x) − bk sin(bk x) ,
π(2k − 1)
2n + 1
ax X n
e 2
+ exp(ak x)
1 (2n + 1)a2n (2n + 1)a2n+1 k=1
73 , n = 0, 1, . . .
p2n+1 − a2n+1 × ak cos(bk x) − bk sin(bk x) ,
2πk
2n + 1
Q(p)
, Xn

P (p) Q(ak )
′
exp ak x
74 P (p) = (p − a1 ) . . . (p − an ); k=1
P (ak )
Q(p) is a polynomial of degree (prime stands for differentiation)
≤ n − 1; ai 6= aj if i 6= j
Z c+i∞
1
2πi c−i∞
Q(p)
, mk
n X
X
P (p) Φkl (ak )
xmk −l exp ak x ,
75 P (p) = (p − a1 )m1 . . . (p − an )mn ; k=1 l=1
(mk − l)! (l − 1)!

Q(p) is a polynomial of degree dl−1 Q(p) P (p)
< m1 + m2 + · · · + mn − 1; Φkl (p) = , Pk (p) =
dpl−1 Pk (p) (p − ak )mk
ai 6= aj if i 6= j
Q(p) + pR(p) n
, X Q(iak ) sin(ak x) + ak R(iak ) cos(ak x)
P (p) ,
ak Pk (iak )
76 P (p) = (p2 + a21 ) . . . (p2 + a2n ); k=1
P (p)
Q(p) and R(p) are polynomials Pm (p) = 2 , i2 = −1
of degree ≤ 2n − 2; al 6= aj , l 6= j p + a2m
28.2.3 Expressions with Square Roots

Z c+i∞
1
2πi c−i∞
1 1
1 √ √
p πx
√ p ebx − eax
2 p−a− p−b √
2 πx3
1 1
3 √ √ e−ax
p+a πx
r
p+a
4 −1 1
2
ae−ax/2 I1 1
2
ax + I0 1
2
ax
p
√
p+a e−ax
5 √ + (a − b)1/2 e−bx erf (a − b)1/2 x1/2
p+b πx
r
1 x
6 √ 2
p p π
1
7 √ (b − a)−1/2 e−ax erf (b − a)1/2 x1/2
(p + a) p + b
1 1 √
8 √ √ eax erf ax
p (p − a) a
1 √
9 a−3/2 eax erf ax − 2a−1 π −1/2 x1/2
p3/2 (p − a)
1 2 √
10 √
p+a π −1/2 x−1/2 − aea x
erfc a x
a √
11 √ 1 − ea
2
x
erfc a x
p p+a
1 2 √
12 √
p+a p ea x
erfc a x
1 2 √
13 √ √ 2 1 − √ (ax)1/2 + (1 − 2ax)eax erf ax − 1
p+ a π
1 1 1 ax √ 2 √
14 √ √ 2 + 2x − e erfc ax − √ x
p p+ a a a πa
Z c+i∞
1
2πi c−i∞
1 2 √
15 √ √ 2 2π −1/2 x1/2 − 2axea x
erfc a x
p p+a
1 2 √ 2 √
16 √ 3 √ (a2 x + 1) x − ax(2a2 x + 3)ea x erfc a x
p+a π
2n
17 p−n−1/2 , n = 1, 2, . . . √ xn−1/2
1 · 3 . . . (2n − 1) π
2n
18 (p + a)−n−1/2 √ xn−1/2 e−ax
1 · 3 . . . (2n − 1) π
1
19 p J0 (ax)
p2
+ a2
1
20 p I0 (ax)
p 2 − a2
1 1/2
21 p exp − 12 ax J0 (b − 14 a2 x
p2 + ap + b
p 1/2 1
22 p 2 + a2 − p √ sin(ax)
2πx3
p 1/2 √
1 2
23 p p 2 + a2 + p √ cos(ax)
p2
+a 2
πx
p 1/2 √
1 2
24 p p 2 − a2 + p √ cosh(ax)
2
p −a 2
πx
p −n
25 p 2 + a2 + p na−n x−1 Jn (ax)
p −n
26 p 2 − a2 + p na−n x−1 In (ax)
−n−1/2 (x/a)n Jn (ax)
27 p 2 + a2
1 · 3 · 5 . . . (2n − 1)
−n−1/2 (x/a)n In (ax)
28 p 2 − a2
1 · 3 · 5 . . . (2n − 1)
28.2.4 Expressions with Arbitrary Powers

Z c+i∞
1
2πi c−i∞
1
1 (p + a)−ν , ν > 0 xν−1 e−ax
Γ(ν)
−2ν ν
2 (p + a)1/2 + (p + b)1/2 , ν>0 x−1 exp − 12 (a + b)x Iν 12 (a − b)x
(a − b)ν
√
−ν π x ν−1/2 a+b a−b
3 (p + a)(p + b) , ν>0 exp − x Iν−1/2 x
Γ(ν) a − b 2 2
√
−ν−1/2 π
4 p 2 + a2 , ν > − 12 xν Jν (ax)
(2a)ν Γ(ν + 12 )
√
−ν−1/2 π
5 p 2 − a2 , ν > − 12 xν Iν (ax)
(2a)ν Γ(ν + 12 )
√
−ν−1/2 a π
6 p p 2 + a2 , ν>0 xν Jν−1 (ax)
(2a)ν Γ ν + 12
Z c+i∞
1
2πi c−i∞
√
a π
7 p p2 − a 2 −ν−1/2
, ν>0 xν Iν−1 (ax)
(2a)ν Γ ν + 12
−ν
(p2 + a2 )1/2 + p =
8 −2ν 2 2 1/2 ν νa−ν x−1 Jν (ax)
a (p + a ) −p , ν >0
2 −ν
(p − a2 )1/2 + p =
9 ν νa−ν x−1 Iν (ax)
a−2ν p − (p2 − a2 )1/2 , ν > 0
−ν
10 p (p2 + a2 )1/2 + p , ν>1 νa1−ν x−1 Jν−1 (ax) − ν(ν + 1)a−ν x−2 Jν (ax)
2 −ν
11 p (p − a ) 2 1/2
+p , ν>1 νa1−ν x−1 Iν−1 (ax) − ν(ν + 1)a−ν x−2 Iν (ax)
p −ν
2
p +a +p 2
12 p , ν > −1 a−ν Jν (ax)
p 2 + a2
p −ν
p 2 − a2 + p
13 p , ν > −1 a−ν Iν (ax)
p 2 − a2

Z c+i∞
1
2πi c−i∞
n
0 if 0 < x < a,
1 p−1 e−ap , a>0
1 if a < x
n
1 if 0 < x < a,
2 p−1 1 − e−ap , a>0
0 if a < x
(
0 if 0 < x < a,
3 p−1 e−ap − e−bp , 0≤a0
e−b(x−a) if a < x

0 if 0 < x < a,
6 p−ν e−ap , ν>0 (x − a)ν−1
 if a < x
Γ(ν)
−1
7 p−1 eap − 1 , a>0 f (x) = n if na < x < (n + 1)a; n = 0, 1, 2, . . .
r
a √
8 ea/p − 1 I1 2 ax
x
1 √
9 p−1/2 ea/p √ cosh 2 ax
πx
1 √
10 p−3/2 ea/p √ sinh 2 ax
πa
r
x √ 1 √
11 p−5/2 ea/p cosh 2 ax − √ sinh 2 ax
πa 2 πa 3
√
12 p−ν−1 ea/p , ν > −1 (x/a)ν/2 Iν (2 ax
r
a √
13 1 − e−a/p J1 2 ax
x
Z c+i∞
1
2πi c−i∞
1 √
14 p−1/2 e−a/p √ cos 2 ax
πx
1 √
15 p−3/2 e−a/p √ sin 2 ax
πa
r
1 √ x √
16 p−5/2 e−a/p √ sin 2 ax − cos 2 ax
2 πa 3 πa
√
17 p−ν−1 e−a/p , ν > −1 (x/a)ν/2 Jν (2 ax
√ a
√ a
18 exp − ap , a>0 √ x−3/2 exp −
2 π 4x
√ a
√ a
19 p exp − ap , a>0 √ (a − 6x)x−7/2 exp −
8 π 4x
1 √ √a
20 exp − ap , a ≥ 0 erfc √
p 2 x
√ √ 1 a
21 p exp − ap , a > 0 √ (a − 2x)x−5/2 exp −
4 π 4x
1 √ 1 a
22 √ exp − ap , a ≥ 0 √ exp −
p πx 4x
√ √
1 √ 2 x a √ a
23 √ exp − ap , a ≥ 0 √ exp − − a erfc √
p p π 4x 2 x
p
exp −k p2 + a2 0 √ if 0 < x < k,
24 p , k>0
p 2 + a2 J0 a x2 − k2 if k < x
p
exp −k p2 − a2 0 √ if 0 < x < k,
25 p , k>0
p 2 − a2 I0 a x2 − k 2 if k < x

Z c+i∞
1
2πi c−i∞
1 f (x) = 2n if a(2n − 1) < x < a(2n + 1);

1 , a>0
p sinh(ap) n = 0, 1, 2, . . . (x > 0)
1 f (x) = 2n(x − an) if a(2n − 1) < x < a(2n + 1);
2 , a>0
p2 sinh(ap) n = 0, 1, 2, . . . (x > 0)
sinh(a/p) 1 √ √
3 √ √ cosh 2 ax − cos 2 ax
p 2 πx
sinh(a/p) 1 √ √
4 √ √ sinh 2 ax − sin 2 ax
p p 2 πa
√ √
5 p−ν−1 sinh(a/p), ν > −2 1
2
(x/a)ν/2 Iν 2 ax − Jν 2 ax

0 if a(4n − 1) < x < a(4n + 1),
1 f (x) =
6 , a>0 2 if a(4n + 1) < x < a(4n + 3),
p cosh(ap)
n = 0, 1, 2, . . . (x > 0)
1 x − (−1)n (x − 2an) if 2n − 1 < x/a < 2n + 1;
7 , a>0
p2 cosh(ap) n = 0, 1, 2, . . . (x > 0)
Z c+i∞
1
2πi c−i∞
cosh(a/p) 1 √ √
8 √ √ cosh 2 ax + cos 2 ax
p 2 πx
cosh(a/p) 1 √ √
9 √ √ sinh 2 ax + sin 2 ax
p p 2 πa
√ √
10 p−ν−1 cosh(a/p), ν > −1 1
2
(x/a)ν/2 Iν 2 ax + Jν 2 ax
1 f (x) = (−1)n−1 if 2a(n − 1) < x < 2an,
11 tanh(ap), a > 0
p n = 1, 2, . . .
1 f (x) = (2n − 1) if 2a(n − 1) < x < 2an,
12 coth(ap), a > 0
p n = 1, 2, . . .
1
13 arccoth(p/a) sinh(ax)
x

Z c+i∞
1
2πi c−i∞
1
1 ln p − ln x − C, C = 0.5772 . . . is the Euler constant
p
1 1 1
xn
−n−1 1+ + + ··· + − ln x − C,
2 p ln p 2 3 n
n!

kn 2 + 23 + 25 + · · · + 2n−1 2
− ln(4x) − C xn−1/2 ,
3 p−n−1/2 ln p 2 n
kn = √ , C = 0.5772 . . .
1 · 3 · 5 . . . (2n − 1) π
1
xν−1 ψ(ν) − ln x ,
4 p−ν ln p, ν>0 Γ(ν)
ψ(ν) is the logarithmic derivative of the gamma function
1
5 (ln p)2 (ln x + C)2 − 16 π 2 , C = 0.5772 . . .
p
1
6 (ln p)2 x (ln x + C − 1)2 + 1 − 1 2
π
p2 6
ln(p + b)
7 e−ax ln(b − a) − Ei (a − b)x }
p+a
ln p 1 1
8 cos(ax) Si(ax) + sin(ax) ln a − Ci(ax)
p 2 + a2 a a
p ln p
9 cos(ax) ln a − Ci(ax) − sin(ax) Si(ax)
p 2 + a2
p+b 1 −ax
10 ln e − e−bx
p+a x
p 2 + b2 2
11 ln 2 cos(ax) − cos(bx)
p + a2 x
p 2 + b2 2
12 p ln 2 cos(bx) + bx sin(bx) − cos(ax) − ax sin(ax)
p + a2 x
(p + a)2 + k2 2
13 ln cos(kx)(e−bx − e−ax
(p + b)2 + k2 x
1p
1 a
14 p ln p 2 + a2 cos(ax) − 1 + sin(ax)
p x2 x
1p
1 a
15 p ln p 2 − a2 cosh(ax) − 1 − sinh(ax)
p x2 x

Z c+i∞
1
2πi c−i∞
sin(a/p) 1 √ √
1 √ √ sinh 2ax sin 2ax
p πx
sin(a/p) 1 √ √
2 √ √ cosh 2ax sin 2ax
p p πa
cos(a/p) 1 √ √
3 √ √ cosh 2ax cos 2ax
p πx
cos(a/p) 1 √ √
4 √ √ sinh 2ax cos 2ax
p p πa
1 √ √ 1 a
5 √ exp − ap sin ap √ sin
p πx 2x
1 √ √ 1 a
6 √ exp − ap cos ap √ cos
p πx 2x
a 1
7 arctan sin(ax)
p x
1 a
8 arctan Si(ax)
p p
a 1
9 p arctan − a ax cos(ax) − sin(ax)
p x 2
2ap 2 p
10 arctan 2 sin(ax) cos x a2 + b2
p + b2 x

Z c+i∞
1
2πi c−i∞

√ 1 x2
1 exp ap2 erfc p a √ exp −
πa 4a

1 √ x
2 exp ap2 erfc p a erf √
p 2 a
(0 if 0 < x < a,
√ √
3 erfc ap , a>0 a
√ if a < x
πx x − a
√
√ a
4 eap erfc ap √
π x (x + a)
1 ap √ 1
5 √ e erfc ap p
p π(x + a)
p 1 √
6 erf a/p sin 2 ax
πx
1 p 1 √
7 √ exp(a/p) erf a/p √ sinh 2 ax
p πx
1 p 1 √
8 √ exp(a/p) erfc a/p √ exp −2 ax
p πx

9 p−a γ(a, bp), a, b > 0 xa−1 if 0 < x < b,
0 if b < x
Z c+i∞
1
2πi c−i∞
√
10 γ(a, b/p), a>0 ba/2 xa/2−1 Ja 2 bx
−p
11 a γ(p, a) exp −ae−x

0 if 0 < x < a,
12 K0 (ap), a>0
(x2 − a2 )−1/2 if a < x

0
if 0 < x < a,
13 Kν (ap), a>0 cosh ν arccosh(x/a)
 √ if a < x
x 2 − a2

√ 1 a2
14 K0 a p exp −
2x 4x

1 √ 1 a2
15 √ K1 a p exp −
p a 4x
⊙ Literature for Section 28.2: G. Doetsch (1950, 1956, 1958), H. Bateman and A. Erdélyi (1954), I. I.
Hirschman and D. V. Widder (1955), V. A. Ditkin and A. P. Prudnikov (1965), A. P. Prudnikov, Yu. A. Brychkov,
and O. I. Marichev (1992b, Vol. 5).
28.3 Tables of Fourier Cosine Transforms

Z ∞
No. Original function, f (x) Cosine transform, fˇc (u) = f (x) cos(ux) dx
0
1 af1 (x) + bf2 (x) afˇ1c (u) + bfˇ2c (u)

1 ˇu
2 f (ax), a>0 fc
a a
2n
n d
3 x2n f (x), n = 1, 2, . . . (−1) fˇc (u)
du2n
Z ∞
d2n+1
4 x2n+1 f (ax), n = 0, 1, . . . (−1)n 2n+1 fˇs (u), fˇs (u) = f (x) sin(xu) dx
du 0
h
1 ˇ u+b u−b i
5 f (ax) cos(bx), a, b > 0 fc + fˇc
2a a a

Z ∞
0
n
1 if 0 < x < a, 1
1 sin(au)
0 if a < x u
(
x if 0 < x < 1,
4 u
2 2 − x if 1 < x < 2, cos u sin2
u2 2
0 if 2 < x
1
3 , a>0 − sin(au) si(au) − cos(au) Ci(au)
a+x
π −au
1 e (the integral is understood
4 , a>0 2a
a2 + x 2 in the sense of Cauchy principal value)
28.3. Tables of Fourier Cosine Transforms 1483
Z ∞
0
1 π sin(au)
5 , a>0
a2 − x 2 2u
a a
6
a2 + (b + x)2
+ 2
a + (b − x)2 πe−au cos(bu)
b+x b−x
7 + 2 πe−au sin(bu)
a2 + (b + x)2 a + (b − x)2

1 au π au
8 , a>0 1
2
πa−3 exp − √ sin + √
a4 + x 4 2 4 2
1 π ae−bu − be−au
9 , a, b > 0
(a2 + x2 )(b2 + x2 ) 2 ab(a2 − b2 )
2m
x
, π ∂n √ √
10 (x2 + a)n+1 (−1)n+m n
a1/ m e−u a
n, m = 1, 2, . . . ; n + 1 > m ≥ 0 2n! ∂a
r
1 π
11 √
x 2u

 √1 r
if 0 < x < a, π
12 x 2 C(au), C(u) is the Fresnel integral
 2u
0 if a<x

0 if 0 < x < a, r
1 π
13 1 − 2C(au) , C(u) is the Fresnel integral
√ if a<x 2u
x

0 if 0 < x < a, r
1 π
14 cos(au) − sin(au)
√ if a < x 2u
x−a
1
15 √ K0 (au)
a2 + x 2

√ 1 if 0 < x < a, π
16 a2 − x 2 J0 (au)
 2
0 if a < x
−ν

17 x , 0<ν<1 sin 1
2
πν Γ(1 − ν)uν−1

Z ∞
0
a
1 e−ax
a2 + u2
1 −ax 1 b2 + u2
2 e − e−bx ln 2
x 2 a + u2
√ −ax √ u
3 xe 1
2
π (a2 + u2 )−3/4 cos 32 arctan
a
Z ∞
0
r 1/2
1 π a + (a2 + u2 )1/2
4 √ e−ax
x 2 a2 + u2
an+1 n! X u 2k
5 xn e−ax , n = 1, 2, . . . (−1)k Cn+1
2k
(a2 + u2 )n+1 a
0≤2k≤n+1
∂n 1
n−1/2 −ax kn u √ ,
6 x e , n = 1, 2, . . . ∂an r pr−a p
where r = a2 + u2 , kn = (−1)n π/2
u
7 xν−1 e−ax Γ(ν)(a2 + u2 )−ν/2 cos ν arctan
a
x 1 π2
8 − 2
eax − 1 2u 2
2a sinh2 πa−1 u
11 1 1 1
9 − + x − ln 1 − e−2πu
x 2 x e −1 2
r
1 π u2
10 exp −ax2 exp −
2 a 4a
a r
1 √
π − 2au √ √
11 √ exp − e cos 2au − sin 2au
x x 2u
r
1 a π −√2au √
12 √ exp − e cos 2au
x x x a

Z ∞
0
1 π
1 , a>0 1

cosh(ax) 2a cosh 2
πa−1 u
1 πu
2 , a>0
2
cosh (ax) 2a2 sinh 12 πa−1 u
−1

cosh(ax) 1
π cos 2 πab cosh 12 πb−1 u
3 , |a| 0 exp cos
2 a 4a 2
x π2
6 2 1

sinh(ax) 4a2 cosh πa−1 u
2

sinh(ax) π sin πab−1
7 , |a| 0 ln coth 1
4
πa−1 u
x

Z ∞
0
n
ln x if 0 < x < 1, 1
1 − Si(u)
0 if 1 < x u
r
ln x π h πi
√ − ln(4u) + C + ,
2 2u 2
x
πν h π πν i
3 xν−1 ln x, 0 < ν < 1 Γ(ν) cos u−ν ψ(ν) − tan − ln u
2 2 2
a+x
2
4 ln , a > 0 cos(au) Si(au) − sin(au) Ci(au)
a−x u
π
5 ln 1 + a2 /x2 , a > 0 1 − e−au
u
a2 + x 2 π −bu
6 ln , a, b > 0 e − e−au
b2 + x 2 u
aC + 12 a ln(u2 + a2 ) + u arctan(u/a)
7 e−ax ln x, a>0 −
u2 + a2
a π
8 ln 1 + e−ax , a>0 −
2u2 2u sinh πa−1 u
a π
9 ln 1 − e−ax , a>0 − coth πa−1 u
2u2 2u

Z ∞
0

1

 2π if u < a,
sin(ax) 1
1 , a>0 π if u = a,
x 
4
0 if u > a
(u + a)−ν − |u + a|−ν sign(u − a)
2 xν−1 sin(ax), a > 0, |ν| < 1 π
4Γ(1 − ν) cos 12 πν
(
x sin(ax)
1
2
πe−ab cosh(bu) if u < a,
3 , a, b > 0
x 2 + b2 − 12 πe−bu sinh(ab) if u > a
(
sin(ax)
1
2
πb−2 1 − e −ab
cosh(bu) if u < a,
4 , a, b > 0
x(x2 + b2 ) 1
πb−2 e−bu sinh(ab) if u > a
2

1 a+u a−u
5 e−bx sin(ax), a, b > 0 +
2 (a + u)2 + b2 (a − u)2 + b2

1 1 a2
6 sin2 (ax), a>0 ln1 − 4 2
x 4 u
1
1 π(2a − u) if u < 2a,
7 sin2 (ax), a > 0 4
x2 0 if u > 2a
1 a π √
8 sin , a>0 J0 2 au
x x 2
Z ∞
0
r ab a2 + b2
1 √ √ π π
9 √ sin a x sin b x , a, b > 0 sin sin −
x u 2u 4u 4
r 2 2
π u u
10 sin ax2 , a>0 cos − sin
8a 4a 4a
√
π Au2 Bu2
exp − sin ϕ − ,
11 exp −ax2 sin bx2 , a>0 (A2 + B 2)1/4 A2 + B 2 A2 + B 2
A = 4a, B = 4b, ϕ = 12 arctan(b/a)
12
1 − cos(ax) 1 a2
, a>0 ln1 − 2
x 2 u
1
1 − cos(ax) 2
π(a − u) if u < a,
13 , a>0
x2 0 if u > a
ν−1

14 x cos(ax), a > 0, 0 < ν < 1 1
2
Γ(ν) cos 12 πν |u − a|−ν + (u + a)−ν
(
cos(ax)
1
2
πb−1 e−ab cosh(bu) if u < a,
15 , a, b > 0
x 2 + b2 1
2
πb−1 e−bu cosh(ab) if u > a

b 1 1
16 e−bx cos(ax), a, b > 0 +
2 (a + u)2 + b2 (a − u)2 + b2
r 2
1 √ π a π
17 √ cos a x sin +
x u 4u 4
r ab a2 + b2
1 √ √ π π
18 √ cos a x cos b x cos sin +
x u 2u 4u 4
r 2
1 π 2
a +u au
19 exp −bx2 cos(ax), b>0 exp − cosh
2 b 4b 2b
r
π
20 cos ax2 , a>0 cos 14 a−1 u2 + sin 14 a−1 u2
8a
√
π Au2 Bu2
exp − cos ϕ − ,
21 exp −ax2 cos bx2 , a>0 (A2 + B 2)1/4 A2 + B 2 A2 + B 2
A = 4a, B = 4b, ϕ = 12 arctan(b/a)

Z ∞
0
1 u
1 Ei(−ax) − arctan
u a

0 if 0 0 a2 − u2

0 if a 0, ν > −1  aν sin(πν/2)

− if a < u,
ξ(u + ξ) ν
p
where ξ = u2 − a2
 −1

 ν cos[ν arcsin(u/a)] if 0 0, ν > 0 aν cos(πν/2)
x 
 √ ν if a 0, ν > − 12 (2a)ν Γ ν + 12


0 if a 0, −1 < ν < − 12  Γ −ν − 1 (u2 − a2 ν+3/2 if a 0 sin
u 4u
2
1 √ 4 a
10 √ J1 a x , a>0 sin2
x a 8u
√ a ν
xν/2 Jν a x , a2 πν
11 1 u−ν−1 sin −
a > 0, −1 < ν < 2 2 4u 2
 √
p  cos b a2 − u2
√ if 0 0 1
−√ 2 if a 0, |ν| < 1 (2a)ν π
 − Γ 1 − ν (u2 − a2 ν+1/2 if a 0 √ exp −b u2 + a2
2
2 u +a 2
⊙ Literature for Section 28.3: G. Doetsch (1950, 1956, 1958), H. Bateman and A. Erdélyi (1954), V. A. Ditkin
and A. P. Prudnikov (1965), F. Oberhettinger (1980).
28.4 Tables of Fourier Sine Transforms

Z ∞
No. Original function, f (x) Sine transform, fˇs (u) = f (x) sin(ux) dx
0
1 af1 (x) + bf2 (x) afˇ1s (u) + bfˇ2s (u)

1 ˇu
2 f (ax), a>0 fs
a a
d2n
3 x2n f (x), n = 1, 2, . . . (−1)n 2n fˇs (u)
du
Z ∞
d2n+1
4 x2n+1 f (ax), n = 0, 1, . . . (−1)n+1 2n+1 fˇc (u), fˇc (u) = f (x) cos(xu) dx
du 0
h
1 ˇ u+b u−b i
5 f (ax) cos(bx), a, b > 0 fs + Fs
2a a a

Z ∞
0
n
1 if 0 < x < a, 1
1 1 − cos(au)
0 if a < x u
(
x if 0 < x < 1,
4 u
2 2 − x if 1 < x < 2, sin u sin2
u2 2
0 if 2 < x
1 π
3
x 2
1
4 , a>0 sin(au) Ci(au) − cos(au) si(au)
a+x
x π −au
5 , a>0 e
a2 + x 2 2
1 π
6 , a>0 1 − e−au
x(a2 + x2 ) 2a2
a a
7 a2 + (x − b)2
− 2
a + (x + b)2 πe−au sin(bu)
x+b x−b
8 − 2 πe−au cos(bu)
a2 + (x + b)2 a + (x − b)2
n−2
X (2n − k − 4)!
x πue−au
9 (x2 + a2 )n
, a > 0, n = 1, 2, . . .
2n−2 2n−3
(2au)k
2 (n − 1)! a k=0
k! (n − k − 2)!
x2m+1
, π ∂n √
10 (x2 + a)n+1 (−1)n+m am e−u a
n, m = 0, 1, . . . ; 0 ≤ m ≤ n 2n! ∂an
r
1 π
11 √
x 2u
1 √
12 √ 2πu
x x
13 x(a2 + x2 )−3/2 uK0 (au)
√ 1/2 r
a2 + x 2 − a π −au
14 √ e
a2 + x 2 2u
−ν

15 x , 0<ν<2 cos 12 πν Γ(1 − ν)uν−1
28.4. Tables of Fourier Sine Transforms 1489

Z ∞
0
u
1 e−ax , a>0
a2 + u2
a n+1 [n/2]
X 2k+1
2k+1 u
2 xn e−ax , a > 0, n = 1, 2, . . . n! 2 2
(−1)k Cn+1
a +u a
k=0
1 −ax u
3 e , a>0 arctan
x a
√ 3
√ −ax π 2 u
4 xe , a>0 (a + u2 )−3/4 sin arctan
2 2 a
r √
1 a 2 + u2 − a)1/2
√ e−ax , a>0 π
5 √
x 2 a2 + u2
1 √ p
6 √ e−ax , a>0 2π a2 + u2 − a)1/2
x x
r " √ 1/2 #
π ∂n a2 + u2 − a
7 xn−1/2 e−ax , a > 0, n = 1, 2, . . . (−1)n √
2 ∂an a2 + u2
u
8 xν−1 e−ax , a > 0, ν > −1 Γ(ν)(a2 + u2 )−ν/2 sin ν arctan
a
u u2 + b2 u u
9 x−2 e−ax − e−bx , a, b > 0 ln 2 + b arctan − a arctan
2 u + a2 b a
1 1 π
10 , a>0 −
eax + 1 2u 2a sinh(πu/a)
1 π πu 1
11 , a>0 coth −
eax − 1 2a a 2u
12 ex/2 − 12 tanh(πu)
ex − 1
√ u2
π
13 x exp −ax2 u exp −
4a3/2 4a
π u
1
14 exp −ax2 2
erf √
x 2 a
a r
1 π −√2au √ √
15 √ exp − e cos 2au + sin 2au
x x 2u
a r
1 π −√2au √
16 √ exp − e sin 2au
x x x a

Z ∞
0
1 π
1 , a>0 tanh 1
2
πa−1 u
sinh(ax) 2a

x π 2 sinh 12 πa−1 u
2 , a>0
sinh(ax) 4a2 cosh2 12 πa−1 u
2au
1 −bx 1
3 e sinh(ax), b > |a| 2
arctan
x u2 + b2 − a2
1
4 , a>0 arctan sinh 12 πa−1 u
x cosh(ax)
Z ∞
0
1 π
5 1 − tanh 1
ax , a>0 −
2 u a sinh πa−1 u
π 1
6 coth 1
2
ax − 1, a>0 coth πa−1 u −
a u

cosh(ax) π sinh πb−1 u
7 , |a| < b
sinh(bx) 2b cos πab−1 + cosh πb−1 u
−1

sinh(ax) 1
π sin 2 πab sinh 12 πb−1 u
8 , |a| < b
cosh(bx) b cos πab−1 + cosh πb−1 u

Z ∞
0
1
ln x if 0 < x < 1, Ci(u) − ln u − C ,
1 u
0 if 1 < x C = 0.5772 . . . is the Euler constant
ln x
2 − 12 π(ln u + C)
x
r
ln x π π
3 √ − ln(4u) + C −
x 2u 2
−ν

πu ψ(ν) + π2 cot πν − ln u
4 xν−1 ln x, |ν| < 1 2

2Γ(1 − ν) cos πν2
π
a+x
5 ln , a>0 sin(au)
a−x u
(x + b)2 + a2 2π −au
6 ln , a, b > 0 e sin(bu)
(x − b)2 + a2 u
a arctan(u/a) − 12 u ln(u2 + a2 ) − eC u
7 e−ax ln x, a>0
u2 + a2
u
1
8 ln 1 + a2 x2 , a>0 −π Ei −
x a

Z ∞
0
sin(ax) 1 u + a
1 , a>0 ln
x 2 u−a
(
1
sin(ax) 2
πu if 0 0 1
x2 2
πa if u > a
|u − a| − |u + a|−ν
−ν
3 xν−1 sin(ax), a > 0, −2 < ν < 1 π , ν 6= 0
4Γ(1 − ν) sin 12 πν
Z ∞
0
(
sin(ax)
1
2
πb−1 e−ab sinh(bu) if 0 0
x 2 + b2 1
2
πb−1 e−bu sinh(ab) if u > a

sin(πx) sin u if 0 π

a 1 1
6 e−ax sin(bx), a>0 − 2
2 a2 + (b − u)2 a + (b + u)2
1 (u + b)2 + a2
7 x−1 e−ax sin(bx), a>0 ln
4 (u − b)2 + a2

1

 4 π if 0 0 1
π if u = 2a,
x 
8
0 if u > 2a
1
1 (u + 2a) ln |u + 2a| + 14 (u − 2a) ln |u − 2a|
9 sin2 (ax), a>0 4
x2 − 12 u ln u
r bu
1 π u2 + b2
10 exp −ax2 sin(bx), a>0 exp − sinh
2 a 4a 2a

 0 if 0 0 if a − b 0 √ J1 2 au
x 2 u
a r
1 π √ √ √
13 √ sin , a>0 sin 2 au − cos 2 au + exp −2 au
x x 8u
r
√ √ π −3/2 a2
14 exp −a x sin a x , a > 0 a u exp −
8 2u

0
 if 0 0 4
x 
1
2
π if a 0, |ν| < 1
4Γ(1 − ν) cos 12 πν
( 1 −ab
x cos(ax) − 2 πe sinh(bu) if u < a,
17 , a, b > 0
x 2 + b2 1
2
πe−bu cosh(ab) if u > a

1 − cos(ax) u u2 − a2 a u + a
18 , a>0 ln + ln
x2 2 u 2 2 u−a
r 2
1 √ π a π
19 √ cos a x cos +
x u 4u 4
r ab 2
1 √ √ π a + b2 π
20 √ cos a x cos b x , a, b > 0 cos cos +
x u 2u 4u 4

Z ∞
0

1 u2
1 erfc(ax), a>0 1 − exp − 2
u 4a

1 u2
2 ci(ax), a>0 − ln1 − 2
2u a

0 if 0 0
− 12 πu−1 if a 0 1
√ 2 if a 0, ν > −2  ν
 a cos(πν/2)

if a < u,
ξ(u + ξ)p ν
where ξ = u2 − a2

1 arcsin(u/a) if 0 0, ν > 0
x π/2 if a 0, ν > −1 aν sin(πν/2)
x 
 √ ν if a 0, −1 < ν < 1 π(2a)ν
2  ν+1/2 if a 0 arcsin p p
(u + b)2 + a2 + (u − b)2 + a2

10
J0 (ax) b−1 sinh(bu)K0 (ab) if 0 0
x 2 + b2 0 if a 0
x 2 + b2
1
2
πe−bu I0 (ab) if a 0, n = 0, 1, 2, . . . 0 if a 0, −1 < ν < 5 0 if a 0, ν > − 32 2
2
√ 1 a
15 J0 a x , a>0 cos
u 4u
Z ∞
0
2
1 √ 2 a
16 √ J1 a x , a>0 sin
x a 4u
2
√ aν a πν
17 xν/2 Jν a x , a > 0, −2 < ν < 1
2
cos −
2ν uν+1 4u 2


 2 arcsin(u/a)
 π √a2 − u2
 if 0 0 √

 2 ln u − u2 − a2 − ln a

 √ if a 0 u
− √ if a 0 √
u2 + a2
πu
21 xK0 (ax), a>0
2(u2 + a2 )3/2
√
22 xν+1 Kν (ax), a > 0, ν > − 32 π (2a)ν Γ ν + 32 u(u2 + a2 )−ν−3/2
⊙ Literature for Section 28.4: G. Doetsch (1950, 1956, 1958), H. Bateman and A. Erdélyi (1954), I. I.
Hirschman and D. V. Widder (1955), V. A. Ditkin and A. P. Prudnikov (1965), F. Oberhettinger (1980).
Chapter 29
Curvilinear Coordinates, Vectors,

Operators, and Differential
Relations
29.1 Arbitrary Curvilinear Coordinate Systems

29.1.1 General Nonorthogonal Curvilinear Coordinates
◮ Metric tensor. Arc length and volume elements in curvilinear coordinates.
The curvilinear coordinates x1 , x2 , x3 are defined as functions of the rectangular Cartesian
coordinates x, y, z:
x1 = x1 (x, y, z), x2 = x2 (x, y, z), x3 = x3 (x, y, z).
Using these formulas, one can express x, y, z in terms of the curvilinear coordinates
x1 , x2 , x3 as follows:
x = x(x1 , x2 , x3 ), y = y(x1 , x2 , x3 ), z = z(x1 , x2 , x3 ).
The metric tensor components gij are determined by the formulas
∂x ∂x ∂y ∂y ∂z ∂z
gij (x1 , x2 , x3 ) = i j
+ i j
+ ;
∂x ∂x ∂x ∂x ∂xi ∂xj
gij (x1 , x2 , x3 ) = gji (x1 , x2 , x3 ); i, j = 1, 2, 3.
The arc length dl between close points (x, y, z) ≡ (x1 , x2 , x3 ) and (x+dx, y+dy, z+dz) ≡
(x1 + dx1 , x2 + dx2 , x3 + dx3 ) is expressed as
3 X
X 3
2 2 2 2
(dl) = (dx) + (dy) + (dz) = gij (x1 , x2 , x3 ) dxi dxj .
i=1 j=1
The volume of the elementary parallelepiped with vertices at the eight points (x1 , x2 , x3 ),
(x1 + dx1 , x2 , x3 ), (x1 , x2 + dx2 , x3 ), (x1 , x2 , x3 + dx3 ), (x1 + dx1 , x2 + dx2 , x3 ),
1495
1496 C URVILINEAR C OORDINATES , V ECTORS , O PERATORS , AND D IFFERENTIAL R ELATIONS
(x1 + dx1 , x2 , x3 + dx3 ), (x1 , x2 + dx2 , x3 + dx3 ), (x1 + dx1 , x2 + dx2 , x3 + dx3 ) is
given by
q
∂(x, y, z) 1 2 3
dV = dx dx dx = ± det |gij | dx1 dx2 dx3 .
∂(x1 , x2 , x3 )
Here the plus sign corresponds to the standard situation where the tangent vectors to the
coordinate lines x1 , x2 , x3 , pointing in the direction of growth of the respective coordinate,
form a right-handed triple, just as unit vectors ~i, ~j, ~k of a right-handed rectangular Cartesian
coordinate system.
◮ Vector components in Cartesian and curvilinear coordinate systems.
The unit vectors ~i, ~j, ~k of a rectangular Cartesian coordinate system∗ x, y, z and the unit
vectors ~i1 , ~i2 , ~i3 of a curvilinear coordinate system x1 , x2 , x3 are connected by the linear
relations

~in = √ 1 ∂x ~
i+
∂y ~
j+
∂z ~
k , n = 1, 2, 3;
gnn ∂xn ∂xn ∂xn
1 ∂x2 ~ ∂x3 ~
~i = √g11 ∂x ~i1 + √g22 √
i2 + g33 i3 ;
∂x ∂x ∂x
1 ∂x2 ~ ∂x3 ~
~j = √g11 ∂x ~i1 + √g22 √
i2 + g33 i3 ;
∂y ∂y ∂y
1 ∂x2 ~ ∂x3 ~
~k = √g11 ∂x ~i1 + √g22 √
i2 + g33 i3 .
∂z ∂z ∂z
In the general case, the vectors ~i1 , ~i2 , ~i3 are not orthogonal and change their direction from
point to point.
The components vx , vy , vz of a vector ~v in a rectangular Cartesian coordinate system
x, y, z and the components v1 , v2 , v3 of the same vector in a curvilinear coordinate system
x1 , x2 , x3 are related by
~v = vx~i + vy~j + vz ~k = v1~i1 + v2~i2 + v3~i3 ,

n
√ ∂x ∂xn ∂xn
vn = gnn vx + vy + vz , n = 1, 2, 3;
∂x ∂y ∂z
∂x v1 ∂x v2 ∂x v3
vx = 1 √ + 2 √ + √ ;
∂x g11 ∂x g22 ∂x3 g33
∂y v1 ∂y v2 ∂y v3
vy = 1 √ + 2 √ + √ ;
∂x g11 ∂x g22 ∂x3 g33
∂z v1 ∂z v2 ∂z v3
vz = √ + √ + √ .
∂x1 g11 ∂x2 g22 ∂x3 g33
∗
Here and henceforth the coordinate axes and the respective coordinates of points in space are denoted by
the same letters.
29.1. Arbitrary Curvilinear Coordinate Systems 1497
29.1.2 General Orthogonal Curvilinear Coordinates

◮ Orthogonal coordinates. Length, area, and volume elements.
A system of coordinates is orthogonal if
gij (x1 , x2 , x3 ) = 0 for i 6= j.
In this case the third invariant of the metric tensor is given by
g = det |gij | = g11 g22 g33 .
The Lamé coefficients Lk of orthogonal curvilinear coordinates are expressed in terms

of the components of the metric tensor as
r
√ ∂x 2 ∂y 2 ∂z 2
Li = gii = + + , i = 1, 2, 3.
∂xi ∂xi ∂xi
Arc length element:
p p
dl = (L1 dx1 )2 + (L2 dx2 )2 + (L3 dx3 )2 = g11 (dx1 )2 + g22 (dx2 )2 + g33 (dx3 )2 .
The area elements dsi of the respective coordinate surfaces xi = const are given by
√
ds1 = dl2 dl3 = L2 L3 dx2 dx3 = g22 g33 dx2 dx3 ,
√
ds2 = dl1 dl3 = L1 L3 dx1 dx3 = g11 g33 dx1 dx3 ,
√
ds3 = dl1 dl2 = L1 L2 dx1 dx2 = g11 g22 dx1 dx2 .
Volume element:
√
dV = L1 L2 L3 dx1 dx2 dx3 = g11 g22 g33 dx1 dx2 dx3 .
◮ Basic differential relations in orthogonal curvilinear coordinates.

In what follows, we present the basic differential operators in the orthogonal curvilinear
coordinates x1 , x2 , x3 . The corresponding unit vectors are denoted by ~i1 , ~i2 , ~i3 .
The gradient of a scalar f is expressed as
1 ∂f ~ 1 ∂f ~ 1 ∂f ~
grad f ≡ ∇f = √ i1 + √ i2 + √ i3 .
g11 ∂x1 g22 ∂x2 g33 ∂x3
Divergence of a vector ~v = v1~i1 + v2~i2 + v3~i3 :

r r r
1 ∂ g ∂ g ∂ g
div ~v ≡ ∇ · ~v = √ v1 + v2 + v3 .
g ∂x1 g11 ∂x2 g22 ∂x3 g33
Gradient of a scalar f along a vector ~v :

v1 ∂f v2 ∂f v3 ∂f
(~v · ∇)f = √ 1
+√ 2
+√ .
g11 ∂x g22 ∂x g33 ∂x3
Gradient of a vector w
~ along a vector ~v :
~ = ~i1 (~v · ∇)w1 + ~i2 (~v · ∇)w2 + ~i3 (~v · ∇)w3 .
(~v · ∇)w
Curl of a vector ~v :
√
g11 ∂ √ ∂ √
curl ~v ≡ ∇ × ~v = ~i1 √ v3 g33 − v2 g22
g ∂x2 ∂x3
√
g22 ∂ √ ∂ √
+ ~i2 √ v1 g11 − v3 g33
g ∂x3 ∂x1
√
~ g33 ∂ √ ∂ √
+ i3 √ v2 g22 − v1 g11 .
g ∂x1 ∂x2
Remark 29.1. Sometimes curl ~
v is denoted by rot ~v .
Laplace operator of a scalar f :
√ √ √
2 1 ∂ g ∂f ∂ g ∂f ∂ g ∂f
∆f ≡ ∇ f = √ + + .
g ∂x1 g11 ∂x1 ∂x2 g22 ∂x2 ∂x3 g33 ∂x3
◮ Some other differential relations.

Let f , g and ~v , ~u be arbitrary sufficiently smooth functions and vector functions. Then, the
following differential relations hold:
curl ∇f = 0,
div curl ~v = 0,
div ∇f = ∆f,
curl curl ~v = ∇ div ~v − ∆~v ,
div(f~v ) = f div ~v + ∇f · ~v ,
curl(~v × ~u) = ~u · curl ~v − ~v · curl ~u,
curl(f~v ) = ∇f × ~v + f curl ~v ,
∆(f g) = f ∆g + g∆f + 2∇f · ∇g,
∆(f~v) = f ∆~v + ~v ∆f + 2(∇f · ∇)~v .
29.2 Cartesian, Cylindrical, and Spherical Coordinate

Systems
29.2.1 Cartesian Coordinates
◮ A vector in Cartesian coordinates and the metric tensor components.
A vector ~v in rectangular Cartesian coordinates x, y, z:
~v = vx~i + vy~j + vz~k.
Metric tensor components:
√
gxx = gyy = gzz = 1, g = 1.
29.2. Cartesian, Cylindrical, and Spherical Coordinate Systems 1499
◮ Basic differential relations.

Gradient of a scalar f :
∂f ~ ∂f ~ ∂f ~
∇f = i+ j+ k.
∂x ∂y ∂z
Divergence of a vector ~v :
∂vx ∂vy ∂vz

div ~v ≡ ∇ · ~v = + + .
∂x ∂y ∂z
∂f ∂f ∂f
(~v · ∇)f = vx + vy + vz .
∂x ∂y ∂z
~ = ~i(~v · ∇)wx + ~j(~v · ∇)wy + ~k(~v · ∇)wz .

(~v · ∇)w

∂vz ∂vy ~ ∂vx ∂vz ~ ∂vy ∂vx ~
curl ~v ≡ ∇ × ~v = − i+ − j+ − k.
∂y ∂z ∂z ∂x ∂x ∂y
Laplacian of a scalar f :
∂2f ∂2f ∂2f
∆f = + + .
∂x2 ∂y 2 ∂z 2
29.2.2 Cylindrical Coordinates

◮ Transformations of coordinates and vectors. The metric tensor components.
The Cartesian coordinates are expressed in terms of the cylindrical ones as
x = ρ cos ϕ, y = ρ sin ϕ, z=z

(0 ≤ ρ < ∞, 0 ≤ ϕ < 2π, −∞ < z < ∞).
The cylindrical coordinates are expressed in terms of the Cartesian ones as

p
ρ = x2 + y 2 , tan ϕ = y/x, z = z (sin ϕ = y/ρ).
Remark 29.2. Sometimes (including this handbook) the cylindrical coordinate ρ is denoted
by r.
Coordinate surfaces:
x2 + y 2 = ρ2 (right circular cylinders with their axis coincident with the z-axis),
y = x tan ϕ (half-planes through the z-axis),
z=z (planes perpendicular to the z-axis).
Direct and inverse transformations of the components of a vector ~v = vx~i + vy~j + vz~k =
vρ~iρ + vϕ~iϕ + vz~iz :
vρ = vx cos ϕ + vy sin ϕ, vx = vρ cos ϕ − vϕ sin ϕ,

vϕ = −vx sin ϕ + vy cos ϕ, vy = vρ sin ϕ + vϕ cos ϕ,
vz = vz ; vz = vz .
√
gρρ = 1, gϕϕ = ρ2 , gzz = 1, g = ρ.

∂f ~ 1 ∂f ~ ∂f ~
∇f = iρ + iϕ + iz .
∂ρ ρ ∂ϕ ∂z
1 ∂(ρvρ ) 1 ∂vϕ ∂vz
div ~v ≡ ∇ · ~v = + + .
ρ ∂ρ ρ ∂ϕ ∂z
∂f vϕ ∂f ∂f
(~v · ∇)f = vρ + + vz .
∂ρ ρ ∂ϕ ∂z
~ = (~v · ∇)wρ~iρ + (~v · ∇)wϕ~iϕ + (~v · ∇)wz~iz .

(~v · ∇)w

1 ∂vz ∂vϕ ~ ∂vρ ∂vz ~ 1 ∂(ρvϕ ) ∂vρ ~
curl ~v ≡ ∇ × ~v = − iρ + − iϕ + − iz .
ρ ∂ϕ ∂z ∂z ∂ρ ρ ∂ρ ∂ϕ

1 ∂ ∂f 1 ∂2f ∂2f
∆f = ρ + + .
ρ ∂ρ ∂ρ ρ2 ∂ϕ2 ∂z 2
Remark 29.3. The cylindrical coordinates ρ, ϕ are also used as polar coordinates on the xy
plane.
29.2.3 Spherical Coordinates

◮ Transformations of coordinates and vectors. The metric tensor components.
The Cartesian coordinates are expressed in terms of the spherical ones as
x = r sin θ cos ϕ, y = r sin θ sin ϕ, z = r cos θ

(0 ≤ r < ∞, 0 ≤ θ ≤ π, 0 ≤ ϕ < 2π).
29.2. Cartesian, Cylindrical, and Spherical Coordinate Systems 1501
The spherical coordinates are expressed in terms of the Cartesian ones as

p z y y
r = x2 + y 2 + z 2 , θ = arccos , tan ϕ = sin ϕ = p .
r x x2 + y 2
x2 + y 2 + z 2 = r 2 (spheres),
x2 + y 2 − z 2 tan2 θ = 0 (circular cones),
y = x tan ϕ (half-planes through the z-axis).
Transformation of the components of a vector ~v = vx~i + vy~j + vz ~k = vr~ir + vθ~iθ + vϕ~iϕ :
vr = vx sin θ cos ϕ + vy sin θ sin ϕ + vz cos θ,

vθ = vx cos θ cos ϕ + vy cos θ sin ϕ − vz sin θ,
vϕ = −vx sin ϕ + vy cos ϕ.
The inverse transformation is
vx = vr sin θ cos ϕ + vθ cos θ cos ϕ − vϕ sin ϕ,

vy = vr sin θ sin ϕ + vθ cos θ sin ϕ + vϕ cos ϕ,
vz = vr cos θ − vθ sin θ.
The metric tensor components are

√
grr = 1, gθθ = r 2 , gϕϕ = r 2 sin2 θ, g = r 2 sin θ.

∂f ~ 1 ∂f ~ 1 ∂f ~
∇f = ir + iθ + iϕ .
∂r r ∂θ r sin θ ∂ϕ
1 ∂ 2 1 ∂ 1 ∂vϕ
div ~v ≡ ∇ · ~v = 2
r vr + sin θ vθ + .
r ∂r r sin θ ∂θ r sin ϕ ∂ϕ
∂f vθ ∂f vϕ ∂f
(~v · ∇)f = vr + + .
∂r r ∂θ r sin θ ∂ϕ
~ = (~v · ∇)wr~ir + (~v · ∇)wθ~iθ + (~v · ∇)wϕ~iϕ .

(~v · ∇)w

1 ∂(sin θ vϕ ) ∂vθ ~
curl ~v ≡ ∇ × ~v = − ir
r sin θ ∂θ ∂ϕ

1 1 ∂vr ∂(rvϕ ) ~ 1 ∂(rvθ ) ∂vr ~
+ − iθ + − iϕ .
r sin θ ∂ϕ ∂r r ∂r ∂θ

1 ∂ 2 ∂f 1 ∂ ∂f 1 ∂2f
∆f = 2 r + 2 sin θ + .
r ∂r ∂r r sin θ ∂θ ∂θ r 2 sin2 θ ∂ϕ2
29.3 Other Special Orthogonal Coordinates

29.3.1 Coordinates of a Prolate Ellipsoid of Revolution
◮ Transformations of coordinates. The metric tensor components.
1◦ . Transformations of coordinates:
x2 = a2 (σ 2 − 1)(1 − τ 2 ) cos2 ϕ, y 2 = a2 (σ 2 − 1)(1 − τ 2 ) sin2 ϕ, z = aστ

(σ ≥ 1 ≥ τ ≥ −1, 0 ≤ ϕ < 2π).
Coordinate surfaces (the z-axis is the axis of revolution):
x2 + y 2 z2
+ =1 (prolate ellipsoids of revolution),
a2 (σ 2 − 1) a2 σ 2
x2 + y 2 z2
+ =1 (hyperboloid of revolution of two sheets),
a2 (τ 2 − 1) a2 τ 2
σ2 − τ 2 σ2 − τ 2 √
gσσ = a2 , gτ τ = a2 , gϕϕ = a2 (σ 2 −1)(1−τ 2 ), g = a3 (σ 2 −τ 2 ).
σ2 − 1 1 − τ2
2◦ . Special coordinate system u, v, ϕ:
σ = cosh u, τ = cos v, ϕ=ϕ (0 ≤ u < ∞, 0 ≤ v ≤ π, 0 ≤ ϕ < 2π);

x = a sinh u sin v cos ϕ, y = a sinh u sin v sin ϕ, z = a cosh u cos v.
guu = gvv = a2 (sinh2 u + sin2 v), gϕϕ = a2 sinh2 u sin2 v.

29.3. Other Special Orthogonal Coordinates 1503

r r
1 σ 2 − 1 ∂f ~ 1 1 − τ 2 ∂f ~ 1 ∂f ~
∇f = 2 2
iσ + 2 2
iτ + p iϕ .
a σ − τ ∂σ a σ − τ ∂τ a (1 − τ )(σ − 1) ∂ϕ
2 2

1 ∂ h p 2 2 )(σ 2 − 1)
i
∇ · ~v = vσ (σ − τ
a(σ 2 − τ 2 ) ∂σ

∂ h p 2 2 2
i ∂ σ2 − τ 2
+ vτ (σ − τ )(1 − τ ) + vϕ p .
∂τ ∂ϕ (σ 2 − 1)(1 − τ 2 )

r r
vσ σ 2 − 1 ∂f vτ 1 − τ 2 ∂f vϕ ∂f
(~v · ∇)f = + + p .
a σ 2 − τ 2 ∂σ a σ 2 − τ 2 ∂τ 2 2
a (σ − 1)(1 − τ ) ∂ϕ
~ = (~v · ∇)wσ~iσ + (~v · ∇)wτ~iτ + (~v · ∇)wϕ~iϕ .

(~v · ∇)w

1 ∂ 2 ∂f ∂ 2 ∂f σ2 − τ 2 ∂2f
∆f = 2 2 (σ − 1) + (1 − τ ) + 2 .
a (σ − τ 2 ) ∂σ ∂σ ∂τ ∂τ (σ − 1)(1 − τ 2 ) ∂ϕ2
29.3.2 Coordinates of an Oblate Ellipsoid of Revolution

◮ Transformations of coordinates. The metric tensor components.
x2 = a2 (1 + σ 2 )(1 − τ 2 ) cos2 ϕ, y 2 = a2 (1 + σ 2 )(1 − τ 2 ) sin2 ϕ, z = aστ

(σ ≥ 0, −1 ≤ τ ≤ 1, 0 ≤ ϕ < 2π).
x2 + y 2 z2
+ =1 (oblate ellipsoids of revolution),
a2 (1 − σ 2 ) a2 σ 2
x2 + y 2 z2
− =1 (hyperboloid of revolution of one sheet),
a2 (1 − τ 2 ) a2 τ 2
Components of the metric tensor:
σ2 + τ 2 σ2 + τ 2 √
gσσ = a2 , gτ τ = a2 , gϕϕ = a2 (1+σ 2 )(1−τ 2 ), g = a3 (σ 2 +τ 2 ).
1 + σ2 1 − τ2
2◦ . Special coordinate system u, v, ϕ:

σ = sinh u, τ = cos v, ϕ=ϕ (0 ≤ u < ∞, 0 ≤ v ≤ π, 0 ≤ ϕ < 2π),
x = a cosh u sin v cos ϕ, y = a cosh u sin v sin ϕ, z = a sinh u cos v.
guu = gvv = a2 (sinh2 u + cos2 v), gϕϕ = a2 cosh2 u sin2 v.

r r
1 σ 2 + 1 ∂f ~ 1 1 − τ 2 ∂f ~ 1 ∂f ~
∇f = 2 2
iσ + 2 2
iτ + p iϕ .
a σ + τ ∂σ a σ + τ ∂τ a (1 − τ )(σ + 1) ∂ϕ
2 2

1 ∂ p
∇ · ~v = 2 2 2
vσ (σ + τ )(σ + 1)
a(σ 2 + τ 2 ) ∂σ

∂ p ∂ σ2 + τ 2
+ vτ (σ 2 + τ 2 )(1 − τ 2 ) + vϕ p .
∂τ ∂ϕ (σ 2 + 1)(1 − τ 2 )
r r
vσ σ 2 + 1 ∂f vτ 1 − τ 2 ∂f vϕ ∂f
(~v · ∇)f = 2 2
+ 2 2
+ p .
a σ + τ ∂σ a σ + τ ∂τ a (σ + 1)(1 − τ ) ∂ϕ
2 2
~ = (~v · ∇)wσ~iσ + (~v · ∇)wτ~iτ + (~v · ∇)wϕ~iϕ .
(~v · ∇)w

1 ∂ 2 ∂f ∂ 2 ∂f σ2 + τ 2 ∂2f
∆f = 2 2 (1 + σ ) + (1 − τ ) + .
a (σ + τ 2 ) ∂σ ∂σ ∂τ ∂τ (1 + σ 2 )(1 − τ 2 ) ∂ϕ2
29.3.3 Coordinates of an Elliptic Cylinder

x = aστ, y 2 = a2 (σ 2 − 1)(1 − τ 2 ), z=z
(σ ≥ 1, −1 ≤ τ ≤ 1, −∞ < z < ∞).
x2 y2
+ =1 (elliptic cylinders),
a2 σ 2 a2 (σ 2 − 1)
x2 y2
2 2
+ 2 2 =1 (hyperbolic cylinders),
a τ a (τ − 1)
z=z (planes parallel to the xy-plane).
σ2 − τ 2 σ2 − τ 2
gσσ = a2 , gτ τ = a2 , gzz = 1.
σ2 − 1 1 − τ2
2◦ . Special coordinate system u, v, z:
σ = cosh u, τ = cos v, z = z;
x = a cosh u cos v, y = a sinh u sin v, z=z
(0 ≤ u < ∞, 0 ≤ v ≤ π, −∞ < z < ∞).
guu = gvv = a2 (sinh2 u + sin2 v), gzz = 1.
3◦ . Laplacian:
2
1 ∂ f ∂2f ∂2f
∆f = 2 + +
a (sinh2 u + sin2 v) ∂u2 ∂v 2 ∂z 2
√ √
σ2 − 1 ∂ p 2 ∂f 1−τ2 ∂ p 2
∂f ∂2f
= 2 2 σ − 1 + 1 − τ + .
a (σ − τ 2 ) ∂σ ∂σ a2 (σ 2 − τ 2 ) ∂τ ∂τ ∂z 2
Remark 29.4. The elliptic cylinder coordinates σ, τ are also used as elliptic coordinates on the
plane xy.
29.3.4 Conical Coordinates

Transformations of coordinates:
uvw u2 (v 2 − a2 )(w2 − a2 ) u2 (v 2 − b2 )(w2 − b2 )

x=± , y2 = , z 2
=
ab a2 (a2 − b2 ) b2 (b2 − a2 )
(b2 > v 2 > a2 > w2 ).
x2 + y 2 + z 2 = u2 (spheres),
x2 y2 z2
+ 2 − 2 =0 (cones with their axes coincident with the z-axis),
v2 v − a2 b − v2
x2 y2 z2
− − =0 (cones with their axes coincident with the x-axis).
w2 a2 − w2 b2 − w2
u2 (v 2 − w2 ) u2 (v 2 − w2 )
guu = 1, gvv = , gww = .
(v 2 − a2 )(b2 − v 2 ) (a2 − w2 )(b2 − w2 )
29.3.5 Parabolic Cylinder Coordinates

1 2
x = στ, y= (τ − σ 2 ), z = z.
2
x2
= 2y + σ 2 (right parabolic cylinders with element parallel to the z-axis),
σ2
x2
= −2y + τ 2 (right parabolic cylinders with element parallel to the z-axis),
τ2
gσσ = gτ τ = σ 2 + τ 2 , gzz = 1.

1 ∂2f ∂2f ∂2f
∆f = 2 + + .
(σ + τ 2 ) ∂σ 2 ∂τ 2 ∂z 2
Remark 29.5. The parabolic cylinder coordinates σ, τ are also used as parabolic coordinates
on the plane xy.
29.3.6 Parabolic Coordinates

1 2
x = στ cos ϕ, y = στ sin ϕ, z= (τ − σ 2 ).
2
x2 + y 2
= 2z + σ 2 (paraboloids of revolution),
σ2
x2 + y 2
= −2z + τ 2 (paraboloids of revolution),
τ2
gσσ = gτ τ = σ 2 + τ 2 , gϕϕ = σ 2 τ 2 .

1 1 ∂ ∂f 1 ∂ ∂f 1 1 ∂2f
∆f = 2 σ + τ + + 2 .
(σ + τ 2 ) σ ∂σ ∂σ τ ∂τ ∂τ σ2 τ ∂ϕ2
29.3.7 Bicylindrical Coordinates

a sinh τ a sin σ
x= , y= , z = z.
cosh τ − cos σ cosh τ − cos σ
Coordinate surfaces (abbreviation RCC is “right circular cylinders”):
x2 + (y − a cot σ)2 = a2 (cot2 σ + 1) (RCC with element parallel to the z-axis),
2 2 2 2
(x − a coth τ ) + y = a (coth τ − 1) (RCC with element parallel to the z-axis),
a2
gσσ = gτ τ = , gzz = 1.
(cosh τ − cos σ)2
2
1 2 ∂ f ∂2f ∂2f
∆f = 2 (cosh τ − cos σ) + + .
a ∂σ 2 ∂τ 2 ∂z 2
Remark 29.6. The bicylindrical coordinates σ, τ are also used as bipolar coordinates on the
plane xy.
29.3.8 Bipolar Coordinates (in Space)

a sin σ cos ϕ a sin σ sin ϕ a sinh τ
x= , y= , z=
cosh τ − cos σ cosh τ − cos σ cosh τ − cos σ
(−∞ < τ < ∞, 0 ≤ σ < π, 0 ≤ ϕ < 2π).
a2
x2 + y 2 + (z − a coth τ )2 = (spheres with centers on the z-axis),
sinh2 τ
p a2
( x2 + y 2 − a cot σ)2 + z 2 = (surfaces obtained by revolution of circular arches
sin2 σ
(y − a cot σ)2 + z 2 = a2/ sin2 σ about the z-axis),
a2 a2 sin2 σ
gσσ = gτ τ = , gϕϕ = .
(cosh τ − cos σ)2 (cosh τ − cos σ)2

(cosh τ − cos σ)2 ∂ sin σ ∂f
∆f =
a2 sin σ ∂τ cosh τ − cos σ ∂τ

∂ sin σ ∂f 1 ∂2f
+ + .
∂σ cosh τ − cos σ ∂σ sin σ(cosh τ − cos σ) ∂ϕ2
29.3.9 Toroidal Coordinates

a sinh τ cos ϕ a sinh τ sin ϕ a sin σ
x= , y= , z=
cosh τ − cos σ cosh τ − cos σ cosh τ − cos σ
(−π ≤ σ ≤ π, 0 ≤ τ < ∞, 0 ≤ ϕ < 2π).
a2
x2 + y 2 + (z − a cot σ)2 = (spheres with centers on the z-axis),
sin2 σ
p a2
( x2 + y 2 − a coth τ )2 + z 2 = (tori with centers on the z-axis),
sinh2 τ
a2 a2 sinh2 τ
gσσ = gτ τ = , gϕϕ = .
(cosh τ − cos σ)2 (cosh τ − cos σ)2

(cosh τ − cos σ)2 ∂ sinh τ ∂f
∆f =
a2 sinh τ ∂σ cosh τ − cos σ ∂σ

∂ sinh τ ∂f 1 ∂2f
+ + .
∂τ cosh τ − cos σ ∂τ sinh τ (cosh τ − cos σ) ∂ϕ2
⊙ References for Chapter 29: P. M. Morse and H. Feshbach (1953), D. H. Menzel (1961), P. Moon and
D. E. Spencer (1988), G. A. Korn and T. M. Korn (2000), D. Zwillinger (2002), E. W. Weisstein (2003),
G. B. Arfken and H. J. Weber (2005), A. D. Polyanin and A. V. Manzhirov (2007).
Chapter 30
Special Functions
and Their Properties
⋆ Throughout Chapter 30 it is assumed that n is a positive integer unless otherwise spec-

ified.
30.1 Some Coefficients, Symbols, and Numbers

30.1.1 Binomial Coefficients
◮ Definitions.
n n!
Cnk = = , where k = 1, . . . , n;
k k! (n − k)!
a (−a)k a(a − 1) . . . (a − k + 1)
Ca0 = 1, Cak = = (−1)k = , where k = 1, 2, . . .
k k! k!
Here a is an arbitrary real number.
◮ Generalization. Some properties.
General case:
Γ(a + 1)
Cab = , where Γ(x) is the gamma function.
Γ(b + 1)Γ(a − b + 1)
Properties:
Ca0 = 1, Cnk = 0 for k = −1, −2, . . . or k > n,

a a−b b
Cab+1 = b
Ca−1 = C , Cab + Cab+1 = Ca+1 b+1
,
b+1 b+1 a
1509
1510 S PECIAL F UNCTIONS AND T HEIR P ROPERTIES
n (−1)n n (2n − 1)!!

C−1/2 = 2n
C2n = (−1)n ,
2 (2n)!!
n (−1)n−1 n−1 (−1)n−1 (2n − 3)!!
C1/2 = C 2n−2 = ,
n22n−1 n (2n − 2)!!
2n+1
Cn+1/2 = (−1)n 2−4n−1 C2n
n n
, C2n+1/2 = 2−2n C4n+1
2n
,
22n+1 22n (n−1)/2
Cn1/2 = n , Cnn/2 = C .
πC2n π n
30.1.2 Pochhammer Symbol

◮ Definition.
Γ(a + n) Γ(1 − a)
(a)n = a(a + 1) . . . (a + n − 1) = = (−1)n .
Γ(a) Γ(1 − a − n)
◮ Some properties (k = 1, 2, . . . ).
(n + k − 1)!
(a)0 = 1, (a)n+k = (a)n (a + n)k , (n)k = ,
(n − 1)!
Γ(a − n) (−1)n
(a)−n = = , where a 6= 1, . . . , n;
Γ(a) (1 − a)n
(2n)! (2n + 1)!
(1)n = n!, (1/2)n = 2−2n , (3/2)n = 2−2n ,
n! n!
(a)mk+nk (a)2n (a)k (a + k)n
(a + mk)nk = , (a + n)n = , (a + n)k = .
(a)mk (a)n (a)n
30.1.3 Bernoulli Numbers

◮ Definition.
The Bernoulli numbers are defined by the recurrence relation
n−1
X
B0 = 1, Cnk Bk = 0, n = 2, 3, . . .
k=0
Numerical values:
B0 = 1, B1 = − 12 , B2 = 16 , 1
B4 = − 30 , B6 = 1
42 ,
1
B8 = − 30 ,
5
B10 = 66 , ...; B2m+1 = 0 for m = 1, 2, . . .
All odd-numbered Bernoulli numbers but B1 are zero; all even-numbered Bernoulli num-
bers have alternating signs.
The Bernoulli numbers are the values of Bernoulli polynomials at x = 0: Bn = Bn (0).
30.1. Some Coefficients, Symbols, and Numbers 1511
◮ Generating function.
Generating function:
X ∞
x xn
= B n , |x| < 2π.
ex − 1 n!
n=0
This relation may be regarded as a definition of the Bernoulli numbers.

The following expansions may be used to calculate the Bernoulli numbers:
∞
X 22n (22n − 1) 2n π
tan x = |B2n | x , |x| < ;
n=1
(2n)! 2
∞
X 22n 2n−1
cot x = (−1)n B2n x , |x| < π.
(2n)!
n=0
30.1.4 Euler Numbers

◮ Definition.
The Euler numbers En are defined by the recurrence relation
n
X
2k
C2n E2k = 0 (even numbered),
k=0
E2n+1 = 0 (odd numbered),
where n = 0, 1, . . .
Numerical values:
E0 = 1, E2 = −1, E4 = 5, E6 = −61, E8 = 1385, E10 = −50251, ...,

E2n+1 = 0 for n = 0, 1, . . .
All Euler numbers are integers, the odd-numbered Euler numbers are zero, and the even-
numbered Euler numbers have alternating signs.
The Euler numbers are expressed via the values of Euler polynomials at x = 1/2:
En = 2n En (1/2), where n = 0, 1, . . .
◮ Generating function. Integral representation.

Generating function:
X ∞
ex xn
= En , |x| < 2π.
e2x + 1 n=0
n!
This relation may be regarded as a definition of the Euler numbers.

Representation via a definite integral:
Z ∞ 2n
n 2n+1 t dt
E2n = (−1) 2 .
0 cosh(πt)
30.2 Error Functions. Exponential and Logarithmic

Integrals
30.2.1 Error Function and Complementary Error Function
◮ Integral representations.
Definitions:
Z x
2
erf x = √ exp(−t2 ) dt (error function, also called probability integral),
π 0
Z ∞
2
erfc x = 1 − erf x = √ exp(−t2 ) dt (complementary error function).
π x
Properties:
erf(−x) = − erf x; erf(0) = 0, erf(∞) = 1; erfc(0) = 1, erfc(∞) = 0.
◮ Expansions as x → 0 and x → ∞. Definite integral.
Expansion of erf x into series in powers of x as x → 0:
∞ ∞
2 X x2k+1 2 X 2k x2k+1
erf x = √ (−1)k = √ exp −x2 .
π k! (2k + 1) π (2k + 1)!!
k=0 k=0
Asymptotic expansion of erfc x as x → ∞:
X
1 2
M −1 1
m 2 m −2M −1

erfc x = √ exp −x (−1) 2m+1 + O |x| , M = 1, 2, . . .
π x
m=0
Integral:
Z x
1 1
erf t dt = x erf x − + exp(−x2 ).
0 2 2
30.2.2 Exponential Integral

Definition:
Z x Z ∞ −t
et e
Ei(x) = dt = − dt for x < 0,
−∞ t −x t
Z −ε t Z x t
e e
Ei(x) = lim dt + dt for x > 0.
ε→+0 −∞ t ε t
30.2. Error Functions. Exponential and Logarithmic Integrals 1513
Other integral representations:

Z ∞
x sin t + t cos t
Ei(−x) = −e−x dt for x > 0,
0 x2 + t 2
Z ∞
−x x sin t − t cos t
Ei(−x) = e dt for x < 0,
x2 + t 2
Z 0∞
Ei(−x) = −x e−xt ln t dt for x > 0,
1
Z x t
e −1
Ei(x) = C + ln x + dt for x > 0,
0 t
where C = 0.5772 . . . is the Euler constant.
◮ Expansions as x → 0 and x → ∞.
Expansion into series in powers of x as x → 0:
 ∞

 X xk

 C + ln(−x) + if x < 0,
 k! k
Ei(x) = k=1
X∞ k

 x

C + ln x +

k! k
if x > 0.
k=1
Asymptotic expansion as x → ∞:
n
X (k − 1)! n!
Ei(−x) = e−x (−1)k + Rn , Rn < .
xk xn
k=1
30.2.3 Logarithmic Integral

Definition:
Z x
 dt

 if 0 < x < 1,
0 ln
tZ Z
li(x) == 1−ε
dt x
dt


 lim + if x > 1.
ε→+0 0 ln t 1+ε ln t
◮ Limiting properties. Relation to the exponential integral.

For small x,
x
li(x) ≈ .
ln(1/x)
For large x,
x
li(x) ≈ .
ln x
Asymptotic expansion as x → 1:
∞
X lnk x
li(x) = C + ln |ln x| + .
k! k
k=1
Relation to the exponential integral:
li x = Ei(ln x), x < 1;

x
li(e ) = Ei(x), x < 0.
30.3 Sine Integral and Cosine Integral. Fresnel Integrals

30.3.1 Sine Integral
◮ Integral representations. Properties.
Definition:
Z x Z ∞
sin t sin t π
Si(x) = dt, si(x) = − dt = Si(x) − .
0 t x t 2
Specific values:
π
Si(0) = 0, Si(∞) = , si(∞) = 0.
2
Properties:
Si(−x) = − Si(x), si(x) + si(−x) = −π, lim si(x) = −π.

x→−∞

∞
X (−1)k+1 x2k−1
Si(x) = .
(2k − 1) (2k − 1)!
k=1
M
X −1
(−1)m (2m)!
si(x) = − cos x 2m+1
+ O |x|−2M −1
x
m=0
NX
−1
(−1)m (2m − 1)! −2N

+ sin x + O |x| ,
m=1
x2m
where M, N = 1, 2, . . .
30.3. Sine Integral and Cosine Integral. Fresnel Integrals 1515
30.3.2 Cosine Integral

◮ Integral representation.
Definition: Z Z
∞ x
cos t cos t − 1
Ci(x) = − dt = C + ln x + dt,
x t 0 t
where C = 0.5772 . . . is the Euler constant.

∞
X (−1)k x2k
Ci(x) = C + ln x + .
2k (2k)!
k=1
M
X −1
(−1)m (2m − 1)! −2M

Ci(x) = cos x + O |x|
m=1
x2m
NX
−1
(−1)m (2m)! −2N −1

+ sin x + O |x| ,
x2m+1
m=0
where M, N = 1, 2, . . .
30.3.3 Fresnel Integrals

Definitions: r √
Z x Z x
1 sin t 2
S(x) = √ √ dt = sin t2 dt,
2π 0 t π 0
Z x r Z √
x
1 cos t 2
C(x) = √ √ dt = cos t2 dt.
2π 0 t π 0

r ∞
2 X (−1)k x2k+1
S(x) = x ,
π (4k + 3) (2k + 1)!
k=0
r ∞
2 X (−1)k x2k
C(x) = x .
π (4k + 1) (2k)!
k=0
1 cos x sin x
S(x) = −√ P (x) − √ Q(x),
2 2πx 2πx
1 sin x cos x
C(x) = + √ P (x) − √ Q(x),
2 2πx 2πx
1×3 1×3×5×7 1 1×3×5
P (x) = 1 − 2
+ 4
− · · · , Q(x) = − + ··· .
(2x) (2x) 2x (2x)3
30.4 Gamma Function, Psi Function, and Beta Function

30.4.1 Gamma Function
◮ Integral representations. Simplest properties.
The gamma function, Γ(z), is an analytic function of the complex argument z everywhere
except for the points z = 0, −1, −2, . . .
For Re z > 0, Z ∞
Γ(z) = tz−1 e−t dt.
0
For −(n + 1) < Re z < −n, where n = 0, 1, 2, . . . ,
Z ∞ Xn
−t (−1)m z−1
Γ(z) = e − t dt.
0 m!
m=0
Simplest properties:
Γ(z + 1) = zΓ(z), Γ(n + 1) = n!, Γ(1) = Γ(2) = 1.
Fractional values of the argument:

√
1 √ 1 π
Γ = π, Γ n+ = n (2n − 1)!!,
2 2 2
1 √ 1 n √π
n 2
Γ − = −2 π, Γ − n = (−1) .
2 2 (2n − 1)!!
◮ Euler, Stirling, and other formulas.

Euler formula:
n! nz
Γ(z) = lim (z 6= 0, −1, −2, . . . ).
n→∞ z(z + 1) . . . (z + n)
Symmetry formulas:
π π
Γ(z)Γ(−z) = − , Γ(z)Γ(1 − z) = ,
z sin(πz) sin(πz)
1 1 π
Γ +z Γ −z = .
2 2 cos(πz)
30.4. Gamma Function, Psi Function, and Beta Function 1517
Multiple argument formulas:
22z−1 1
Γ(2z) = √ Γ(z)Γ z + ,
π 2
33z−1/2 1 2
Γ(3z) = Γ(z)Γ z + Γ z+ ,
2π 3 3
Y
n−1
k
(1−n)/2 nz−1/2
Γ(nz) = (2π) n Γ z+ .
n
k=0
Asymptotic expansion (Stirling formula):
√
Γ(z) = 2π e−z z z−1/2 1 + 1 −1
12 z + 1 −2
288 z + O(z −3 ) (|arg z| < π).
30.4.2 Psi Function (Digamma Function)
◮ Definition. Integral representations.
Definition:
d ln Γ(z) Γ′ (z)
ψ(z) = = z .
dz Γ(z)
The psi function is the logarithmic derivative of the gamma function and is also called the
digamma function.
Integral representations (Re z > 0):
Z ∞
ψ(z) = e−t − (1 + t)−z t−1 dt,
0
Z ∞
−1
ψ(z) = ln z + t − (1 − e−t )−1 e−tz dt,
0
Z 1
1 − tz−1
ψ(z) = −C + dt,
0 1−t
where C = −ψ(1) = 0.5772 . . . is the Euler constant.

Values for integer argument:
n−1
X
ψ(1) = −C, ψ(n) = −C + k−1 (n = 2, 3, . . . ).
k=1
◮ Properties. Asymptotic expansion as z → ∞.
Functional relations:
1
ψ(z) − ψ(1 + z) = − ,
z
ψ(z) − ψ(1 − z) = −π cot(πz),
1
ψ(z) − ψ(−z) = −π cot(πz) − ,
z
1 1
ψ 2 + z − ψ 2 − z = π tan(πz),
1 X k
m−1
ψ(mz) = ln m + ψ z+ .
m m
k=0
Asymptotic expansion as z → ∞ (|arg z| < π):
∞
X B2n
1 1 1 1 1
ψ(z) = ln z − − 2
+ 4
− 6
+ · · · = ln z − − ,
2z 12z 120z 252z 2z 2nz 2n
n=1
where the B2n are Bernoulli numbers.
30.4.3 Beta Function

◮ Integral representation. Relationship with the gamma function.
Definition:
Z 1
B(x, y) = tx−1 (1 − t)y−1 dt,
0
where Re x > 0 and Re y > 0.

Relationship with the gamma function:
Γ(x)Γ(y)
B(x, y) = .
Γ(x + y)
◮ Some properties.
B(x, y) = B(y, x);

y y
B(x, y + 1) = B(x + 1, y) = B(x, y);
x x+y
π
B(x, 1 − x) = , 0 < x < 1;
sin(πx)
1 n−1 m−1
= mCn+m−1 = nCn+m−1 ,
B(n, m)
where n and m are positive integers.
30.5. Incomplete Gamma and Beta Functions 1519
30.5 Incomplete Gamma and Beta Functions

30.5.1 Incomplete Gamma Function
◮ Integral representations. Recurrence formulas.
Definitions: Z x
γ(α, x) = e−t tα−1 dt, Re α > 0,
0
Z ∞
Γ(α, x) = e−t tα−1 dt = Γ(α) − γ(α, x).
x
Recurrence formulas:
γ(α + 1, x) = αγ(α, x) − xα e−x ,

γ(α + 1, x) = (x + α)γ(α, x) + (1 − α)xγ(α − 1, x),
Γ(α + 1, x) = αΓ(α, x) + xα e−x .
Special cases:
X
n
−x xk
γ(n + 1, x) = n! 1 − e , n = 0, 1, . . . ;
k!
k=0
Xn
xk
Γ(n + 1, x) = n! e−x , n = 0, 1, . . . ;
k!
k=0
n
n−1
X
(−1) −x k k!
Γ(−n, x) = Γ(0, x) − e (−1) k+1 , n = 1, 2, . . .
n! x
k=0
◮ Expansions as x → 0 and x → ∞. Relation to other functions.
Asymptotic expansions as x → 0:
X∞
(−1)n xα+n
γ(α, x) = ,
n=0
n! (α + n)
∞
X (−1)n xα+n
Γ(α, x) = Γ(α) − .
n! (α + n)
n=0
Asymptotic expansions as x → ∞:
M
X −1
α−1 −x (1 − α)m −M

γ(α, x) = Γ(α) − x e + O |x| ,
(−x)m
m=0
M
X −1
α−1 −x (1 − α)m −M

Γ(α, x) = x e m
+ O |x| − 32 π < arg x < 32 π .
m=0
(−x)
Asymptotic formulas as α → ∞:
h 1 i Z x 1
√ √ 1
γ(x, α) = Γ(α) Φ 2 x − α − 1 + O √ , Φ(x) = √ exp − t2 dt;
α 2π −∞ 2
h √ 1 i x 1/3 1
γ(x, α) = Γ(α) Φ 3 α z + O , z= −1+ .
α α 9α
Representation of the error function, complementary error function, and exponential
integral in terms of the gamma functions:
1 1 1 1
erf x = √ γ , x2 , erfc x = √ Γ , x2 , Ei(−x) = −Γ(0, x).
π 2 π 2
30.5.2 Incomplete Beta Function

Definitions:
Z x
Bx (a, b)
Bx (a, b) = ta−1 (1 − t)b−1 dt, Ix (a, b) = ,
0 B(a, b)
where Re a > 0 and Re b > 0, and B(a, b) = B1 (a, b) is the beta function.
◮ Some properties.
Symmetry:
Ix (a, b) + I1−x (b, a) = 1.
Recurrence formulas:
Ix (a, b) = xIx (a − 1, b) + (1 − x)Ix (a, b − 1),
(a + b)Ix (a, b) = aIx (a + 1, b) + bIx (a, b + 1),
(a + b − ax)Ix (a, b) = a(1 − x)Ix (a + 1, b − 1) + bIx (a, b + 1).
30.6 Bessel Functions (Cylindrical Functions)

30.6.1 Definitions and Basic Formulas
◮ Bessel functions of the first and the second kind.
The Bessel function of the first kind, Jν (x), and the Bessel function of the second kind,
Yν (x) (also called the Neumann function), are solutions of the Bessel equation
x2 yxx
′′
+ xyx′ + (x2 − ν 2 )y = 0
and are defined by the formulas

∞
X (−1)k (x/2)ν+2k Jν (x) cos πν − J−ν (x)
Jν (x) = , Yν (x) = . (1)
k! Γ(ν + k + 1) sin πν
k=0
30.6. Bessel Functions (Cylindrical Functions) 1521
The formula for Yν (x) is valid for ν 6= 0, ±1, ±2, . . . (the cases ν 6= 0, ±1, ±2, . . . are
discussed in what follows).
The general solution of the Bessel equation has the form Zν (x) = C1 Jν (x) + C2 Yν (x)
and is called the cylinder function.
◮ Some formulas.
2νZν (x) = x[Zν−1 (x) + Zν+1 (x)],

d 1 hν i
Zν (x) = [Zν−1 (x) − Zν+1 (x)] = ± Zν (x) − Zν±1 (x) ,
dx 2 x
d ν ν d −ν
[x Zν (x)] = x Zν−1 (x), [x Zν (x)] = −x−ν Zν+1 (x),
dx dx
1 d n ν 1 d n −ν
[x Jν (x)] = xν−n Jν−n (x), [x Jν (x)] = (−1)n x−ν−n Jν+n (x),
x dx x dx
J−n (x) = (−1)n Jn (x), Y−n (x) = (−1)n Yn (x), n = 0, 1, 2, . . .
1
◮ Bessel functions for ν = ±n ± 2
, where n = 0, 1, 2, . . .
r r
2 2
J1/2 (x) = sin x, J−1/2 (x) = cos x,
πx πx
r r
2 1 2 1
J3/2 (x) = sin x − cos x , J−3/2 (x) = − cos x − sin x ,
πx x πx x
r [n/2]
2 nπ X (−1)k (n + 2k)!
Jn+1/2 (x) = sin x −
πx 2 (2k)! (n − 2k)! (2x)2k
k=0
[(n−1)/2]
nπ X (−1)k (n + 2k + 1)!
+ cos x − ,
2 (2k + 1)! (n − 2k − 1)! (2x)2k+1
k=0
r [n/2]
2 nπ X (−1)k (n + 2k)!
J−n−1/2 (x) = cos x +
πx 2 (2k)! (n − 2k)! (2x)2k
k=0
[(n−1)/2]
nπ X (−1)k (n + 2k + 1)!
− sin x + ,
2 (2k + 1)! (n − 2k − 1)! (2x)2k+1
k=0
r r
2 2
Y1/2 (x) = − cos x, Y−1/2 (x) = sin x,
πx πx
Yn+1/2 (x) = (−1)n+1 J−n−1/2 (x), Y−n−1/2 (x) = (−1)n Jn+1/2 (x),
where [A] is the integer part of the number A.

◮ Bessel functions for ν = ±n, where n = 0, 1, 2, . . .

Let ν = n be an arbitrary integer. The relations
J−n (x) = (−1)n Jn (x), Y−n (x) = (−1)n Yn (x)
are valid. The function Jn (x) is given by the first formula in (1) with ν = n, and Yn (x)
can be obtained from the second formula in (1) by proceeding to the limit ν → n. For
nonnegative n, Yn (x) can be represented in the form
1 X (n − k − 1)! 2 n−2k
n−1
2 x
Yn (x) = Jn (x) ln −
π 2 π k! x
k=0
1X
∞ x n+2k ψ(k + 1) + ψ(n + k + 1)
− (−1)k ,
π 2 k! (n + k)!
k=0
n−1
P
where ψ(1) = −C, ψ(n) = −C + k−1 , C = 0.5772 . . . is the Euler constant, and
k=1
ψ(x) = [ln Γ(x)]′x is the logarithmic derivative of the gamma function, also known as the
digamma function.
◮ Wronskians and similar formulas.
2 2
W (Jν , J−ν ) = − sin(πν), , W (Jν , Yν ) =
πx πx

Handbook of Linear Partial Differential Equations For Engineers and Scientists-Andrei D. Polyanin, Vladimir E. Nazaikinskii PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Handbook of Linear Partial Differential Equations For Engineers and Scientists-Andrei D. Polyanin, Vladimir E. Nazaikinskii PDF

Uploaded by

Copyright:

Available Formats

Mathematics, Engineering, Physics

New to the Second Edition

To accommodate different mathematical backgrounds, the authors avoid wherever Polyanin

K18874_cover.indd 1 12/8/15 9:10 AM

No claim to original U.S. Government works

International Standard Book Number-13: 978-1-4665-8149-4 (eBook - PDF)

Preface to the Second Edition xxv

Part I Exact Solutions 1

2 First-Order Equations with Three or More Independent Variables 139

8 Second-Order Hyperbolic Equations with Three or More Space Variables 695

10 Second-Order Elliptic Equations with Three or More Space Variables 859

12.7 Stokes Equations for Viscous Incompressible Fluids . . . . . . . . . . . . . . . . . . . . . . 1077

12.15 Nondissipative Thermoelasticity Equations (the Green–Naghdi Model) . . . . . 1103

Part II Analytical Methods 1121

14.3 Properties and Particular Solutions of Linear Equations . . . . . . . . . . . . . . . . . . . . 1144

21.1.6 Tempered Distributions and Fourier Transform . . . . . . . . . . . . . . . . . . . . 1270

Part III Symbolic and Numerical Solutions with Maple, Mathematica,

23.3 Analytical Solutions of Mathematical Problems . . . . . . . . . . . . . . . . . . . . . . . . . . 1347

Part IV Tables and Supplements 1403

25.2.13 Representation in the Form of Infinite Products . . . . . . . . . . . . . . . . . . . 1411

26.4 Infinite Functional Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1428

28.4 Tables of Fourier Sine Transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1488

30.6 Bessel Functions (Cylindrical Functions) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1520

30.18 Nonorthogonal Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1564

PREFACE TO THE FIRST EDITION

Andrei D. Polyanin, D.Sc., is an internationally renowned

Vladimir E. Nazaikinskii, D.Sc., is an actively working

curl u curl of a vector u, sometimes also denoted by rot u

Brief Notation for Derivatives

◆ For brevity, often only a principal integral

where Φ = Φ(Ξ) is an arbitrary function.

1.1.1 Equations Containing Power-Law Functions

General solution: w = Φ(bx − ay), where Φ is an arbitrary function.

Principal integral: Ξ = 12 bx2 + cx − ay.

Principal integral: Ξ = (abx + b2 y + a + bc)e−bx .

For a = b, this is a conoid equation. Principal integral: Ξ = |x|b |y|−a .

Principal integral: Ξ = bx2 − ay 2 .

Principal integral: Ξ = ay 2 + kx2 + 2(bxy + cy + sx).

(a1 − s)(b2 − s) = a2 b1 , (1)

2.2. If s = 0, then b2 = −a1 . We have

Ξ = a2 x2 − 2a1 xy − b1 y 2 + 2c2 x − 2c1 y.

⊙ Literature: E. Kamke (1965).

◮ Coefficients of equations are quadratic in x and y.

Principal integral: Ξ = 13 ax3 + 12 bx2 + cx − y.

1◦ . Principal integral for 4ac − b2 > 0:

2◦ . Principal integral for 4ac − b2 < 0:

Principal integral: Ξ = Ay 3 − Ex3 + 3(Bxy 2 + Dx2 y + ay − bx).

Hesse’s equation. The introduction of the homogeneous coordinates x = ξ2 /ξ1 , y = ξ3 /ξ1

where gn = an ξ1 + bn ξ2 + cn ξ3 (n = 1, 2, 3). See 2.1.1.21 for the solution of this equation.

◮ Coefficients of equations contain integer powers of x and y.

This is a special case of equation 1.1.7.3 with f (x) = bx2 .

This is a special case of equation 1.1.8.6 with f (z) = Az 3 .

Principal integral: Ξ = 12 ax2 y 2 + by + cx.

Principal integral: Ξ = axy + 14 by 4 + 14 cx4 .

◮ Coefficients of equations contain fractional powers.

1◦ . Principal integral for a 6= 1:

2◦ . Principal integral for a = 1:

Principal integral: Ξ = axy + 23 by 3/2 + 23 cx3/2 .

◆ See also equations in 1.1.1.56–1.1.1.111 for fractional values of exponents.

◮ Coefficients of equations contain arbitrary powers of x and y.

1◦ . Principal integral for n 6= −1:

2◦ . Principal integral for n = −1:

1◦ . Principal integral for n + m 6= −1: