Professional Documents
Culture Documents
SECOND EDITION
ENGINEERS AND SCIENTISTS
DIFFERENTIAL EQUATIONS FOR
LINEAR PARTIAL
HANDBOOK OF
HANDBOOK OF
LINEAR PARTIAL
DIFFERENTIAL EQUATIONS FOR
ENGINEERS AND SCIENTISTS
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
K18874
w w w. c rc p r e s s . c o m
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
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
v
vi C ONTENTS
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
3.2.2 Equation of the Form ∂w 1 ∂w
∂t = a ∂r 2 + r ∂r + Φ(r, t) . . . . . . . . . . . . . . . . . 294
∂2w
3.2.3 Equation of the Form ∂w 2 ∂w
∂t = a ∂r 2 + r ∂r . . . . . . . . . . . . . . . . . . . . . . . . . 298
∂2w
3.2.4 Equation of the Form ∂w 2 ∂w
∂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
3.3.4 Equations of the Form ∂w ∂ w ∂w
∂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
3.3.6 Equations of the Form ∂w ∂ w ∂w
∂t = f (x) ∂x2 + g(x, t) ∂x + h(x, t)w . . . . . . . . . 329
2
3.3.7 Equations of the Form ∂w ∂ w ∂w
∂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
3.4.2 Equations of the Form ∂w ∂w
∂t = a ∂x2 + f (x, t) ∂x . . . . . . . . . . . . . . . . . . . . . 341
∂2w
3.4.3 Equations of the Form ∂w ∂w
∂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
3.6.1 Equations of the Form ∂w ∂w
∂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
3.8.2 Equations of the Form ∂w ∂w
∂t = a ∂x2 + f (x, t) ∂x . . . . . . . . . . . . . . . . . . . . . 363
∂2w
3.8.3 Equations of the Form ∂w ∂w
∂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
4.2.2 Problems in Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442
4.2.3 Axisymmetric Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445
4.3 Other Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
4.3.1 Equations Containing Arbitrary Parameters . . . . . . . . . . . . . . . . . . . . . . . . . 455
4.3.2 Equations Containing Arbitrary Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 457
5 Second-Order Parabolic Equations with Three or More Space Variables 463
5.1 Heat Equation ∂w
∂t = a∆3 w . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463
5.1.1 Problems in Cartesian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 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
5.2.1 Problems in Cartesian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522
5.2.2 Problems in Cylindrical Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 528
5.2.3 Problems in Spherical Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534
5.3 Other Equations with Three Space Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 537
5.3.1 Equations Containing Arbitrary Parameters . . . . . . . . . . . . . . . . . . . . . . . . . 537
5.3.2 Equations Containing Arbitrary Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 539
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
5.4.3 Equations Containing Arbitrary Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 549
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
6.1.5 Equation of the Form ∂t2
= 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
7.1.1 Problems in Cartesian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633
7.1.2 Problems in Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 639
7.1.3 Axisymmetric Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
2
7.2 Nonhomogeneous Wave Equation ∂∂tw 2
2 = a ∆2 w + Φ(x, y, t) . . . . . . . . . . . . . . . 651
7.2.1 Problems in Cartesian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 651
7.2.2 Problems in Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 653
7.2.3 Axisymmetric Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 656
2
7.3 Equations of the Form ∂∂tw 2
2 = a ∆2 w − bw + Φ(x, y, t) . . . . . . . . . . . . . . . . . . . . . 659
7.3.1 Problems in Cartesian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 659
7.3.2 Problems in Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 664
7.3.3 Axisymmetric Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 670
2
7.4 Telegraph Equation ∂∂tw ∂w 2
2 + k ∂t = a ∆2 w − bw + Φ(x, y, t) . . . . . . . . . . . . . . . . . 676
7.4.1 Problems in Cartesian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 676
7.4.2 Problems in Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 681
7.4.3 Axisymmetric Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 688
7.5 Other Equations with Two Space Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693
x 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
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
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
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
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
Latin Characters
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
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
k=0 satisfy the equation y ′′ +(a−2q cos 2x)y =
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
k=0 satisfy the equation y ′′ +(a−2q cos 2x)y =
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
k=0 satisfy the equation y ′′ +(a−2q cos 2x)y =
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
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),
∂w ∂w
1. a +b = 0.
∂x ∂y
∂w ∂w
2. a + (bx + c) = 0.
∂x ∂y
∂w ∂w
3. + (ax + by + c) = 0.
∂x ∂y
3
4 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂w ∂w
4. ax + by = 0.
∂x ∂y
∂w ∂w
5. ay + bx = 0.
∂x ∂y
∂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
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
∂w ∂w
9. + (ax2 + bx + c) = 0.
∂x ∂y
∂w ∂w
10. + (ay 2 + by + c) = 0.
∂x ∂y
2 2ay + b
Ξ=x− √ arctan √ .
4ac − b2 4ac − b2
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.
6 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂w ∂w
15. + (y 2 + axy + a) = 0.
∂x ∂y
This is a special case of equation 1.1.1.60 with n = 1.
∂w ∂w
16. + (y 2 + axy − abx − b2 ) = 0.
∂x ∂y
This is a special case of equation 1.1.1.61 with n = 1.
∂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
This is a special case of equation 1.1.1.76 with n = 1.
∂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)
1.1. Equations of the Form f (x, y) ∂w
∂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
8 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
∂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
∂w ∂w ∂w
g1 + g2 + g3 = 0,
∂ξ1 ∂ξ2 ∂ξ3
∂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
∂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
∂w ∂w
48. (ax + by 3 ) − (cx3 + ay) = 0.
∂x ∂y
∂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
√ ∂w √ ∂w
54. (ax + b y ) − (c x + ay) = 0.
∂x ∂y
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
⊙ Literature: E. Kamke (1965).
∂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
1◦ . Principal integral for n 6= −1:
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
12 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂w ∂w
61. + (y 2 + axn y − abxn − b2 ) = 0.
∂x ∂y
∂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
x2ab a
Ξ= m
+ x2ab−m .
y − bx 2ab − m
xm
Ξ= + a ln x.
y − bxm
∂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
1.1. Equations of the Form f (x, y) ∂w
∂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
Ξ= exp x +a x exp x dx.
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
Ξ= exp x +a x exp x dx.
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
14 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
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
x2m
Ξ= + 2a ln |x|.
(y + bxm )2
1.1. Equations of the Form f (x, y) ∂w
∂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
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 .
16 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
⊙ Literature: A. D. Polyanin and V. F. Zaitsev (1996).
∂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
⊙ Literature: A. D. Polyanin and V. F. Zaitsev (1996).
∂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
⊙ Literature: A. D. Polyanin and V. F. Zaitsev (1996).
1.1. Equations of the Form f (x, y) ∂w
∂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
∂w ∂w
83. x + ax2n+1 y 3 + (bx − n)y + cx1−n = 0.
∂x ∂y
Principal integral: Z
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
Principal integral: Z
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
1◦ . Principal integral for n 6= 1:
Z
dv xn−1 y
Ξ= − , v= .
av m + bv k n−1 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
∂w ∂w
90. axm + by n = 0.
∂x ∂y
b
Ξ= x1−m + a ln |y|.
m−1
Ξ = b ln |x| − a ln |y|.
∂w ∂w
91. axn + (by + cxm ) = 0.
∂x ∂y
1◦ . Principal integral for n 6= 1:
Z
−F c b
Ξ=e y− e−F xm−n dx, F = x1−n .
a a(1 − n)
c am−b
Ξ = x−b/a y − x a .
am − b
(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
20 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
∂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
1.1. Equations of the Form f (x, y) ∂w
∂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.
22 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
∂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
∂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
24 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
26 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
Principal integral: Z
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
1.1. Equations of the Form f (x, y) ∂w
∂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
28 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
1.1. Equations of the Form f (x, y) ∂w
∂x
+ g(x, y) ∂w
∂y
=0 29
∂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).
∂w ∂w
9. + a cosh(λy) = 0.
∂x ∂y
∂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).
1.1. Equations of the Form f (x, y) ∂w
∂x
+ g(x, y) ∂w
∂y
=0 31
∂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.
32 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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)
∂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)
1.1. Equations of the Form f (x, y) ∂w
∂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)
∂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
Principal integral: Z
dz
Ξ=x− , z = x + λy.
1 + aλ lnk z
∂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
34 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
1.1. Equations of the Form f (x, y) ∂w
∂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
36 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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)
∂w ∂w
2. + a sink (λy) + b = 0.
∂x ∂y
Z
dy
Principal integral: Ξ = x − k
.
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
38 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
∂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)
1.1. Equations of the Form f (x, y) ∂w
∂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
40 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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) .
cosk (λy) cosk (λy)
∂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) .
cosk (λy) cosk (λy)
∂w ∂w
28. + a tank (λy) + b = 0.
∂x ∂y
Z
dy
Principal integral: Ξ = x − k
.
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)
1.1. Equations of the Form f (x, y) ∂w
∂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) .
tank (λy) tank (λy)
42 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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) .
tank (λy) tank (λy)
∂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
Principal integral: Ξ = x − k
.
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).
1.1. Equations of the Form f (x, y) ∂w
∂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)
∂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
∂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
44 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
1.1. Equations of the Form f (x, y) ∂w
∂x
+ g(x, y) ∂w
∂y
=0 45
∂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
Principal integral: Ξ = x − .
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)
46 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
∂w ∂w
14. + a arccosk (λy) + b = 0.
∂x ∂y
Z
dy
Principal integral: Ξ = x − .
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
1.1. Equations of the Form f (x, y) ∂w
∂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
48 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
∂w ∂w
26. + a arctank (λy) + b = 0.
∂x ∂y
Z
dy
Principal integral: Ξ = x − .
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)
1.1. Equations of the Form f (x, y) ∂w
∂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
∂w ∂w
38. + a arccotk (λy) + b = 0.
∂x ∂y
Z
dy
Principal integral: Ξ = x − .
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
50 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
1.1. Equations of the Form f (x, y) ∂w
∂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
∂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.
∂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
1.1. Equations of the Form f (x, y) ∂w
∂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
∂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
∂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
1◦ . Principal integral for a > 0:
Z
e−λx y √
Ξ = arctan √ − 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
1.1. Equations of the Form f (x, y) ∂w
∂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
1◦ . Principal integral for a > 0:
2 Z
e−λx y √ 2
Ξ = arctan √ − a eλx f dx.
a
2◦ . Principal integral for a < 0:
2 Z
e−λx y p 2
Ξ = arctanh p + |a| eλx f dx.
|a|
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂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 .
∂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
∂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)
⊙ Literature: V. F. Zaitsev and A. D. Polyanin (1996).
1.1. Equations of the Form f (x, y) ∂w
∂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)
∂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
58 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
1.1. Equations of the Form f (x, y) ∂w
∂x
+ g(x, y) ∂w
∂y
=0 59
∂w ∂w
47. + (f y 2 + gx′ y + af e2g ) = 0.
∂x ∂y
1◦ . Principal integral for a > 0:
Z
e−g y √
Ξ = arctan √ − 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
∂w ∂w
1. f (x) + g(y) = 0.
∂x ∂y
Z Z
dx dy
Principal integral: Ξ = − .
f (x) g(y)
⊙ Literature: E. Kamke (1965).
∂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
8. + f (y + axn + b) − anxn−1 = 0.
∂x ∂y
Principal integral: Z
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
⊙ Literature: A. D. Polyanin and V. F. Zaitsev (1996).
∂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
⊙ Literature: V. F. Zaitsev and A. D. Polyanin (1996).
1.1. Equations of the Form f (x, y) ∂w
∂x
+ g(x, y) ∂w
∂y
=0 61
∂w ∂w
11. + e−λx f (eλx y) = 0.
∂x ∂y
Principal integral: Z
dv
Ξ= − x, v = eλx y.
f (v) + λv
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂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
Principal integral: Z
dv
Ξ= − x, v = y + aeλx + b.
f (v)
62 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
∂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)
⊙ Literature: V. F. Zaitsev and A. D. Polyanin (1996).
1.1. Equations of the Form f (x, y) ∂w
∂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)
⊙ Literature: V. F. Zaitsev and A. D. Polyanin (1996).
∂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)
64 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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.
⊙ Literature: E. Kamke (1965).
∂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
∂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)
1.1. Equations of the Form f (x, y) ∂w
∂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)
66 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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)
∂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
⊙ Literature: E. Kamke (1965).
∂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
68 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
1.2. Equations of the Form f (x, y) ∂w
∂x
+ g(x, y) ∂w
∂y
= h(x, y) 69
∂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)
70 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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.
1.2. Equations of the Form f (x, y) ∂w
∂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 ,
and h(x, y) = sxp y q + d.
∂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
72 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
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)
∂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
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+δy + k.
1.2. Equations of the Form f (x, y) ∂w
∂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
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).
1.2. Equations of the Form f (x, y) ∂w
∂x
+ g(x, y) ∂w
∂y
= h(x, y) 75
∂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
1.2. Equations of the Form f (x, y) ∂w
∂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).
78 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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)
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
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
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
82 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
∂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
1.2. Equations of the Form f (x, y) ∂w
∂x
+ g(x, y) ∂w
∂y
= h(x, y) 83
∂w ∂w
7. a +b = c arccos(λx + βy).
∂x ∂y
1◦ . General solution for aλ + bβ 6= 0:
c h p i
w= (λx + βy) arccos(λx + βy) − 1 − (λx + βy)2 + Φ(bx − ay).
aλ + bβ
∂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
1◦ . General solution for aλ + bβ 6= 0:
c 1 2
w= (λx + βy) arctan(λx + βy) − ln 1 + (λx + βy) + Φ(bx − ay).
aλ + bβ 2
∂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
1. a +b= f (x).
∂x ∂y
Z
1
General solution: w = f (x) dx + Φ(bx − ay).
a
⊙ Literature: E. Kamke (1965).
∂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
∂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
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂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)
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂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)
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
1.2. Equations of the Form f (x, y) ∂w
∂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)
∂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.
∂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).
∂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
1.2. Equations of the Form f (x, y) ∂w
∂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
⊙ Literature: V. F. Zaitsev and A. D. Polyanin (1996).
∂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
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂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.
⊙ Literature: V. F. Zaitsev and A. D. Polyanin (1996).
∂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)
∂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
In the integration, u is considered a parameter.
∂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
In the integration, u is considered a parameter.
⊙ Literature: V. F. Zaitsev and A. D. Polyanin (1996).
∂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.
⊙ Literature: V. F. Zaitsev and A. D. Polyanin (1996).
∂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
⊙ Literature: V. F. Zaitsev and A. D. Polyanin (1996).
∂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)
The transformation ξ = dx, η = dy leads to an equation of the form
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)
1.3. Equations of the Form f (x, y) ∂w
∂x
+ g(x, y) ∂w
∂y
= h(x, y)w 91
∂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).
⊙ Literature: E. Kamke (1965).
∂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.
92 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
∂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)
94 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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.
1.3. Equations of the Form f (x, y) ∂w
∂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
This is a special case of equation 1.3.7.33 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 = 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
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
96 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
This is a special case of equation 1.3.7.35 with f (x) = aeβx , g1 (x) = beγx , g0 (x) = c, and
h(x, y) = seµx + keδy + p.
∂w ∂w
8. aeβx + beγx + ceλy = seµx+δy + k w.
∂x ∂y
This is a special case of equation 1.3.7.35 with f (x) = aeβx , g1 (x) = beγx , g0 (x) = c, and
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
1.3. Equations of the Form f (x, y) ∂w
∂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
98 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
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
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
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
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)
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.
102 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
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
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
General solution: w = exp
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
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
1.3. Equations of the Form f (x, y) ∂w
∂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) .
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
∂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).
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
1.3. Equations of the Form f (x, y) ∂w
∂x
+ g(x, y) ∂w
∂y
= h(x, y)w 107
∂w ∂w
12. a +b = c arctan(λx + βy)w.
∂x ∂y
1◦ . General solution for aλ + bβ 6= 0:
c(λx + βy) ln[1 + (λx + βy)2 ]
w = exp arctan(λx + βy) − Φ(bx − ay).
aλ + bβ 2(aλ + bβ)
∂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
17. a +b = c arccot(λx + βy)w.
∂x ∂y
1◦ . General solution for aλ + bβ 6= 0:
c(λx + βy) ln[1 + (λx + βy)2 ]
w = exp arccot(λx + βy) + Φ(bx − ay).
aλ + bβ 2(aλ + bβ)
∂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 .
2(λx + βy) (λx + βy) x
∂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).
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.
1.3. Equations of the Form f (x, y) ∂w
∂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
∂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
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
110 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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)
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂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)
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂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)
∂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)
The transformation ξ = dx, η = dy leads to an equation of the form
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).
∂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
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂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
where u = y a x−b . In the integration, u is considered a parameter.
∂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
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
In the integration, u is considered a parameter.
∂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
In the integration, u is considered a parameter.
∂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.
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂w ∂w
34. f (x) + g1 (x)y + g0 (x)y k = h(x, y)w.
∂x ∂y
For k = 1, see equation 1.3.7.32. For k 6= 1, the substitution ξ = y 1−k leads to an equation
of the form 1.3.7.33:
∂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
leads to an equation of the form 1.3.7.33:
∂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)
The transformation ξ = dx, η = dy leads to an equation of the form
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)
114 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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)
1.4. Equations of the Form f (x, y) ∂w
∂x
+ g(x, y) ∂w
∂y
= h1 (x, y)w + h0 (x, y) 115
γ(αx + b) β(αy + d)
w=− 2
− + eαx/b Φ (cy + d)e−cx/b .
α α(α − c)
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
∂w ∂w
8. ay + (b1 x + b0 ) = (c1 x + c0 )w + s1 x + s0 .
∂x ∂y
∂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
116 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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 .
1.4. Equations of the Form f (x, y) ∂w
∂x
+ g(x, y) ∂w
∂y
= h1 (x, y)w + h0 (x, y) 117
∂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
where u = y a x−b . In the integration, u is considered a parameter.
∂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
This is a special case of equation 1.4.7.22 with f (x) = axn , g1 (x) = bxm , g0 (x) = cxk ,
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 .
120 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
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
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
where u = y/x. In the integration, u is considered a parameter.
∂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
16. a tanhn (λx) + b tanhm (µx) = c tanhk (νy)w + p tanhs (βx).
∂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 (νy), and F (x, y) = p tanhs (βx).
∂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
This is a special case of equation 1.4.7.22 with f (x) = a sinhn (λx), g1 (x) ≡ 0, g0 (x) =
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
This is a special case of equation 1.4.7.22 with f (x) = a tanhn (λx), g1 (x) ≡ 0, g0 (x) =
b cothm (µx), h(x, y) = c tanhk (νy), and F (x, y) = p coths (βx).
∂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
1.4. Equations of the Form f (x, y) ∂w
∂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
where u = y/x. In the integration, u is considered a parameter.
∂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
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 (νy), and F (x, y) = p sins (βx).
∂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
c/a k −(a+c)/a b/a −b/a
−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
where u = y/x. In the integration, u is considered a parameter.
∂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
18. ax + by = cw + k tan(λx + µy).
∂x ∂y
Z x
c/a k −(a+c)/a b/a −b/a
−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
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 (νy), and F (x, y) = p tans (βx).
∂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).
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).
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).
132 F IRST-O RDER E QUATIONS WITH T WO I NDEPENDENT VARIABLES
∂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
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
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).
1.4. Equations of the Form f (x, y) ∂w
∂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
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
where u = bx − ay. In the integration, u is considered a parameter.
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂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
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
where u = y a x−b . In the integration, u is considered a parameter.
∂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)
In the integration, u is considered a parameter.
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂w ∂w
23. f (x) + g1 (x)y + g0 (x)y k = h(x, y)w + F (x, y).
∂x ∂y
For k = 1, see equation 1.4.7.21. For k 6= 1, the substitution ξ = y 1−k leads to an equation
of the form 1.4.7.22:
∂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
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.
∂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
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
∂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 .
∂w ∂w ∂w
10. b2 cy + a2 cx − ab(ax + by) = 0.
∂x ∂y ∂z
∂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)
∂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
∂w ∂w ∂w
16. bc(by − 2cz) + ac(3cz − ax) + ab(2ax − 3by) = 0.
∂x ∂y ∂z
∂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:
∂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.
⊙ Literature: E. Kamke (1965).
∂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:
∂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
∂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
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.
2.1. Equations of the Form f (x, y, z) ∂w
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 143
u1 = αx + βy + γz.
3. If equation (1) has three different roots, s1 , s2 , and s3 , other than zero, then the
integrals
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 ).
⊙ Literature: E. Kamke (1965).
∂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
∂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 .
144 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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.
⊙ Literature: E. Kamke (1965).
∂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.
⊙ Literature: E. Kamke (1965).
∂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
Integral basis: u1 = , u2 = .
z x
⊙ Literature: E. Kamke (1965).
2.1. Equations of the Form f (x, y, z) ∂w
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 145
∂w ∂w ∂w
35. by 2 − axy + cxz = 0.
∂x ∂y ∂z
∂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
∂w ∂w ∂w
39. cxz − cyz + (ax2 + by 2 ) = 0.
∂x ∂y ∂z
∂w ∂w ∂w
40. xz + yz + (ax2 + ay 2 + bz 2 ) = 0.
∂x ∂y ∂z
y a(x2 + y 2 ) + (b − 1)z 2
Integral basis: u1 = , u2 = .
x (x2 + y 2 )b
⊙ Literature: E. Kamke (1965).
∂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
Integral basis: u1 = , u2 = .
x y
∂w ∂w ∂w
42. bcyz + acxz + abxy = 0.
∂x ∂y ∂z
∂w ∂w ∂w
43. bc(x2 − a2 ) + c(bxy + acz) + b(cxz + aby) = 0.
∂x ∂y ∂z
by + cz by − cz
Integral basis: u1 = , u2 = .
x−a x+a
⊙ Literature: E. Kamke (1965).
∂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
146 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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 .
⊙ Literature: E. Kamke (1965).
∂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.
⊙ Literature: E. Kamke (1965).
∂w ∂w ∂w
53. bc2 y 2 z + ac2 xz 2 − abxy 2 = 0.
∂x ∂y ∂z
∂w ∂w ∂w
54. x(by 2 − cz 2 ) + y(cz 2 − ax2 ) + z(ax2 − by 2 ) = 0.
∂x ∂y ∂z
ax2 + by 2 + cz 2 2ax2 + by 2
Integral basis: u1 = , u2 = .
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
Integral basis: u1 = , u2 = .
x yz
∂w ∂w ∂w
59. x(cz 4 − by 4 ) + y(ax4 − 2cz 4 ) + z(2by 4 − ax4 ) = 0.
∂x ∂y ∂z
∂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.
⊙ Literature: E. Kamke (1965).
∂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 .
⊙ Literature: E. Kamke (1965).
148 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
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:
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
∂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
∂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
In the integration, u1 is considered a parameter.
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 .
⊙ Literature: E. Kamke (1965).
∂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 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
2.1. Equations of the Form f (x, y, z) ∂w
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 153
∂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
∂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
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
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
∂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).
2.1. Equations of the Form f (x, y, z) ∂w
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 157
∂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
∂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
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)
2.1. Equations of the Form f (x, y, z) ∂w
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 159
∂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
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
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
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).
2.1. Equations of the Form f (x, y, z) ∂w
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 161
∂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)
Integral basis: u1 = , u2 = .
sin(µy) sin(νz)
∂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)
Integral basis: u1 = , u2 = .
cos(µy) cos(νz)
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.
162 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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).
2.1. Equations of the Form f (x, y, z) ∂w
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 163
∂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.
164 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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
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
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)
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂w ∂w ∂w
5. + f1 (x)y + f2 (x) + g1 (x)z + 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
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
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.
166 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
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).
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂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) .
y − a cosh(λx) λ
∂w ∂w ∂w
8. + f1 (x)y + f2 (x)y k + g1 (x)z + g2 (x)z m = 0.
∂x ∂y ∂z
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
∂w ∂w ∂w
9. + f1 (x)y + f2 (x)y k + g1 (x) + g2 (x)eλz = 0.
∂x ∂y ∂z
∂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
∂w ∂w ∂w
− λ f1 (x)ξ + f2 (x) − β g1 (x)η + g2 (x) = 0.
∂x ∂ξ ∂η
2.1. Equations of the Form f (x, y, z) ∂w
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 167
∂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
In the integration, u1 is considered a parameter.
∂w ∂w ∂w
12. + f1 (x)y + f2 (x) + g1 (y)z + g2 (y) = 0.
∂x ∂y ∂z
Integral basis:
Z Z
u1 = yF (x) − f2 (x)F (x) dx, F (x) = exp − f1 (x) dx ,
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.
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂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
u2 = + E dx, E = exp sinh(λx) .
y − a cosh(λx) λ
∂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)
168 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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)
∂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
∂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
where x0 may be taken as arbitrary.
⊙ Literature: E. Kamke (1965).
∂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
2.1. Equations of the Form f (x, y, z) ∂w
∂x
+ g(x, y, z) ∂w
∂y
+ h(x, y, z) ∂w
∂z
=0 169
∂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
u1 = yF (x) − f2 (x)F (x) dx, F (x) = exp − f1 (x) dx ,
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.
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂w ∂w ∂w
24. + f1 (x)y + f2 (x)y k + g(x, y)z + h(x, y)z m = 0.
∂x ∂y ∂z
∂w ∂w ∂w
25. + f1 (x)y + f2 (x)y k + g(x, y) + h(x, y)eλz = 0.
∂x ∂y ∂z
∂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 .
170 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂w ∂w ∂w
26. + f1 (x) + f2 (x)eλy + g(x, y)z + h(x, y)z k = 0.
∂x ∂y ∂z
∂w ∂w ∂w
27. + f1 (x) + f2 (x)eλy + g(x, y) + h(x, y)eβz = 0.
∂x ∂y ∂z
∂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)
The transformation ξ = dx, η = dy leads to an equation of the form
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)
The transformation ξ = dx, η = dy leads to an equation of the form
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)
◆ 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
∂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
∂w ∂w ∂w
5. + (ay + k1 x + k0 ) + (bz + s1 x + s0 ) = c1 x + c0 .
∂x ∂y ∂z
∂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
∂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
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
∂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
∂w ∂w ∂w
12. + (ay + k1 x2 + k0 ) + (bz + s1 x2 + s0 ) = c 1 x2 + c 0 .
∂x ∂y ∂z
∂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
∂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
∂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
∂w ∂w ∂w
19. +a +b = xyz.
∂x ∂y ∂z
∂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
∂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
∂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
( √ √
by + ab z exp − ab x if ab > 0,
u1 = by 2 − az 2 , u2 = p p p
by cos |ab| x + |ab| z sin |ab| x if ab < 0.
∂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
√ √
|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
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
176 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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,
∂w ∂w ∂w
31. + a1 xn1 y + b1 xm1 + (a2 xn2 y + b2 xm2 = c2 xk2 y + c1 xk1 z.
∂x ∂y ∂z
This is a special case of equation 2.2.7.4 with f1 (x) = a1 xn1 , f2 (x) = b1 xm1 , g1 (x) =
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
This is a special case of equation 2.2.7.5 with f1 (x) = a1 xn1 , f2 (x) = b1 xm1 , g1 (x) =
a2 xn2 , g2 (x) = b2 xm2 , h2 (x) = c2 xk2 , h1 (x) = c1 xk1 , and h0 (x) = 0.
∂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.
2.2. Equations of the Form f1 ∂w
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 , fn = fn (x, y, z) 177
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
This is a special case of equation 2.2.7.9 with f1 (x) = A1 eα1 x , f2 (x) = B1 eν1 x , g1 (x) =
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
1 αx 1 βy βy − αx 1 −γz
where u1 = − e + e and u2 = αx βy
+ e . In the integration,
aα bβ bβe − aαe cγ
u1 is considered a parameter.
178 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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
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
∂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 ,
h1 (x) = keβx , and h0 (x) = seγx .
∂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
∂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 .
2.2. Equations of the Form f1 ∂w
∂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
This is a special case of equation 2.2.7.7 with f1 (x) = a1 eλ1 x , f2 (x) = b1 eβ1 x , g1 (x) =
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
This is a special case of equation 2.2.7.8 with f1 (x) = a1 eβ1 x , f2 (x) = b1 eγ1 x , g1 (x) =
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
This is a special case of equation 2.2.7.9 with f1 (x) = a1 xn , f2 (x) = b1 xm , g1 (x) = a2 xk ,
g2 (x) = b2 xl , and h(x) = cxs .
∂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
180 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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
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
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).
2.2. Equations of the Form f1 ∂w
∂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).
182 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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).
2.2. Equations of the Form f1 ∂w
∂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
1 λx 1 βy 1 λx 1 γz
u1 = lntanh − lntanh , u2 = lntanh − lntanh .
aλ 2 bβ 2 aλ 2 cγ 2
∂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).
184 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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).
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
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) = 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
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
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).
186 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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.
2.2. Equations of the Form f1 ∂w
∂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).
188 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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
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 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
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 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).
This is a special case of equation 2.2.7.10 in which f (x) = a1 tann1 (λ1 x), g(y) =
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.
2.2. Equations of the Form f1 ∂w
∂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
This is a special case of equation 2.2.7.12 with f1 (x) = a, f2 (x) = b, f3 (x) = c arcsinn (λx),
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
This is a special case of equation 2.2.7.12 with f1 (x) = a, f2 (x) = b arcsinn (λx), f3 (x) = 1,
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
8. a1 +a2 +a3 = b1 arccos(λ1 x)+b2 arccos(λ2 y)+b3 arccos(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 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
This is a special case of equation 2.2.7.12 with f1 (x) = a, f2 (x) = b, f3 (x) = c arccosn (λx),
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
This is a special case of equation 2.2.7.12 with f1 (x) = a, f2 (x) = b arccosn (λx), f3 (x) = 1,
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.
190 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂w ∂w ∂w
13. a1 +a2 +a3 = b1 arctan(λ1 x)+b2 arctan(λ2 y)+b3 arctan(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 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
This is a special case of equation 2.2.7.12 with f1 (x) = a, f2 (x) = b, f3 (x) = c arctann (λx),
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
This is a special case of equation 2.2.7.12 with f1 (x) = a, f2 (x) = b arctann (λx), f3 (x) = 1,
f4 (x) = s arctanm (γx), g(y) = 1, and h(z) = c arctank (βz).
∂w ∂w ∂w
18. a1 + a2 + a3 = b1 arccot(λ1 x) + b2 arccot(λ2 y) + b3 arccot(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 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
This is a special case of equation 2.2.7.12 with f1 (x) = a, f2 (x) = b, f3 (x) = c arccot n (λx),
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.
2.2. Equations of the Form f1 ∂w
∂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
This is a special case of equation 2.2.7.12 with f1 (x) = a, f2 (x) = b arccotn (λx), f3 (x) = 1,
f4 (x) = s arccotm (γx), g(y) = 1, and h(z) = c arccot k (βz).
∂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.
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
192 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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).
Particular solution:
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.
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂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
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).
194 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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
This is a special case of equation 2.2.7.24 with g1 (x, y) = g1 (x), g2 (x, y) = g2 (y), and
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
This is a special case of equation 2.2.7.25 with g1 (x, y) = g1 (y), g2 (x, y) = g2 (x), and
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
This is a special case of equation 2.2.7.26 with g1 (x, y) = g1 (x), g2 (x, y) = g2 (y), and
h(x, y, z) = h1 (x) + h2 (y).
∂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
2.2. Equations of the Form f1 ∂w
∂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
u1 = x y , u2 = b − t−1 f xy −a/b ta/b , t dt.
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.
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂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◦ . For k 6= 1 and m = 1, the substitution ξ = y 1−k leads to an equation of the form
2.2.7.22.
3◦ . For k = m = 1, see equation 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
leads to an equation of the form 2.2.7.22:
∂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
196 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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
The transformation ξ = e−λy , η= z 1−k leads to an equation of the form 2.2.7.22:
∂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
The transformation ξ = e−λy , η = e−βz leads to an equation of the form 2.2.7.22:
∂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)
The transformation ξ = dx, η = dy leads to an equation of the form
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)
The transformation ξ = dx, η = dy leads to an equation of the form
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)
where h̄1 (ξ, η) ≡ , h̄2 (ξ, η) ≡ , and h̄3 (ξ, η, z) ≡ .
f2 (x)g1 (y) f2 (x)g1 (y) f2 (x)g1 (y)
◆ 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.
2.3. Equations of the Form f1 ∂w
∂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
( √ √
by + ab z exp − ab x if ab > 0,
u1 = by 2 − az 2 , u2 = p p p
by cos |ab| x + |ab| z sin |ab| x if ab < 0.
∂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
∂w ∂w ∂w
5. + (ay + k1 x + k0 ) + (bz + s1 x + s0 ) = (c1 x + c0 )w.
∂x ∂y ∂z
∂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
∂w ∂w ∂w
8. abx + b(ay + bz) + a(ay − bz) = cw.
∂x ∂y ∂z
∂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
∂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
∂w ∂w ∂w
12. + (ay + k1 x2 + k0 ) + (bz + s1 x2 + s0 ) = (c1 x2 + c0 )w.
∂x ∂y ∂z
∂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
200 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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
( √ √
by + ab z exp − ab x if ab > 0,
u1 = by 2 − az 2 , u2 = p p p
by cos |ab| x + |ab| z sin |ab| x if ab < 0.
∂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)
2.3. Equations of the Form f1 ∂w
∂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
( √ √
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
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
General solution: w = exp x Φ(u1 , u2 ), where
n
√ √
|x| ab by − ab z if ab > 0,
2 2 √
u1 = by − az , u2 = √
−ab −ab z
|x| exp − arctan if ab < 0.
by
∂w ∂w ∂w
29. abx + b(ay + bz) + a(ay − bz) = cxn w.
∂x ∂y ∂z
c n
General solution: w = exp 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 = exp x Φ a2 y 2 − 2abyz − b2 z 2 .
abn
202 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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,
∂w ∂w ∂w
31. + a1 xn1 y+b1 xm1 +(a2 xn2 y+b2 xm2 = (c2 xk2 y+c1 xk1 z)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.4 with f1 (x) = a1 xn1 , f2 (x) = b1 xm1 , g1 (x) =
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)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.5 with f1 (x) = a1 xn1 , f2 (x) = b1 xm1 , g1 (x) =
a2 xn2 , g2 (x) = b2 xm2 , h2 (x) = c2 xk2 , h1 (x) = c1 xk1 , and h0 (x) = 0.
∂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.
2.3. Equations of the Form f1 ∂w
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w, fn = fn (x, y, z) 203
∂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
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
∂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 ,
h1 (x) = keβx , and h0 (x) = seγx .
∂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 ,
h1 (x) = keβx , and h0 (x) = seγx .
∂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
∂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 .
2.3. Equations of the Form f1 ∂w
∂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
This is a special case of equation 2.3.7.7 with f1 (x) = a1 eλ1 x , f2 (x) = b1 eβ1 x , g1 (x) =
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
This is a special case of equation 2.3.7.8 with f1 (x) = a1 eβ1 x , f2 (x) = b1 eγ1 x , g1 (x) =
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
∂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
∂w ∂w ∂w
3. + a sinhn (βx) + b sinhk (λx) = c sinhm (γx)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.1 with f (x) = a sinhn (βx), g(x) = b sinhk (λx),
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
206 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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
∂w ∂w ∂w
8. + a coshn (βx) + b coshk (λx) = c coshm (γx)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.1 with f (x) = a coshn (βx), g(x) = b coshk (λx),
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).
2.3. Equations of the Form f1 ∂w
∂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
∂w ∂w ∂w
13. + a tanhn (βx) + b tanhk (λx) = c tanhm (γx)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.1 with f (x) = a tanhn (βx), g(x) = b tanhk (λx),
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
∂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 .
208 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂w ∂w ∂w
18. a +b + c coth(βz) = [k coth(λx) + s coth(γy)]w.
∂x ∂y ∂z
∂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
1 λx 1 βy 1 λx 1 γz
u1 = lntanh − lntanh , u2 = lntanh − lntanh .
aλ 2 bβ 2 aλ 2 cγ 2
∂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).
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).
210 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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
This is a special case of equation 2.3.7.22 with f1 (x) = 0, f2 (x) = a lnn (λx), g1 (x, y) = 0,
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
(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
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),
h2 (x) = h1 (x) = 0, and h0 (x) = c sink (γx).
2.3. Equations of the Form f1 ∂w
∂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
This is a special case of equation 2.3.7.1 with f (x) = a cosn (βx), g(x) = b cosk (λx),
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).
212 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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
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
This is a special case of equation 2.3.7.1 with f (x) = a tann (βx), g(x) = b tank (λx),
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
This is a special case of equation 2.3.7.10 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).
2.3. Equations of the Form f1 ∂w
∂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
This is a special case of equation 2.3.7.1 with f (x) = a cotn (βx), g(x) = b cotk (λx),
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),
h2 (x) = h1 (x) = 0, and h0 (x) = c sink (γx).
∂w ∂w ∂w
24. + a cosn (λx) + b sinm (βy) = c cosk (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 cosn (λx), g1 (x, y) = 0,
g2 (x, y) = b sinm (βy), and h(x, y, z) = c cosk (γy) + s sinl (µz).
214 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂w ∂w ∂w
25. + a cosn (λx) + b tanm (βy) = c cosk (γy) + s tanl (µz) w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.22 with f1 (x) = 0, f2 (x) = a cosn (λx), g1 (x, y) = 0,
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.
This is a special case of equation 2.3.7.10 in which f (x) = a1 tann1 (λ1 x), g(y) =
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
2. a1 + a2 + a3
∂x ∂y
∂z
= b1 arcsin(λ1 x) + b2 arcsin(λ2 y) + b3 arcsin(λ3 z) w.
This is a special case of equation 2.3.7.10 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)w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.12 with f1 (x) = a, f2 (x) = b, f3 (x) = c arcsinn (λx),
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
This is a special case of equation 2.3.7.12 with f1 (x) = a, f2 (x) = b arcsinn (λx), f3 (x) = 1,
f4 (x) = s arcsinm (γx), g(y) = 1, and h(z) = c arcsink (βz).
2.3. Equations of the Form f1 ∂w
∂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.
This is a special case of equation 2.3.7.10 with f (x) = a1 , g(y) = a2 , h(z) = a3 , ϕ(x) =
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
This is a special case of equation 2.3.7.12 with f1 (x) = a, f2 (x) = b, f3 (x) = c arccosn (λx),
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
This is a special case of equation 2.3.7.12 with f1 (x) = a, f2 (x) = b arccosn (λx), f3 (x) = 1,
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 .
216 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂w ∂w ∂w
14. a1 + a2 + a3
∂x ∂y
∂z
= b1 arctan(λ1 x) + b2 arctan(λ2 y) + b3 arctan(λ3 z) w.
This is a special case of equation 2.3.7.10 with f (x) = a1 , g(y) = a2 , h(z) = a3 , ϕ(x) =
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
This is a special case of equation 2.3.7.12 with f1 (x) = a, f2 (x) = b, f3 (x) = c arctann (λx),
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
This is a special case of equation 2.3.7.12 with f1 (x) = a, f2 (x) = b arctann (λx), f3 (x) = 1,
f4 (x) = s arctanm (γx), g(y) = 1, and h(z) = c arctank (βz).
∂w ∂w ∂w
19. a1 + a2 + a3
∂x ∂y ∂z
= b1 arccot(λ1 x) + b2 arccot(λ2 y) + b3 arccot(λ3 z) w.
This is a special case of equation 2.3.7.10 with f (x) = a1 , g(y) = a2 , h(z) = a3 , ϕ(x) =
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
This is a special case of equation 2.3.7.12 with f1 (x) = a, f2 (x) = b, f3 (x) = c arccot n (λx),
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
This is a special case of equation 2.3.7.12 with f1 (x) = a, f2 (x) = b arccotn (λx), f3 (x) = 1,
f4 (x) = s arccotm (γx), g(y) = 1, and h(z) = c arccot k (βz).
2.3. Equations of the Form f1 ∂w
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w, fn = fn (x, y, z) 217
∂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.
Particular solution:
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.
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
218 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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.
Particular solution:
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.
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂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
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
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
This is a special case of equation 2.3.7.22 with g1 (x, y) = g1 (x), g2 (x, y) = g2 (y), and
h(x, y, z) = h1 (x) + h2 (y).
220 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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.
This is a special case of equation 2.3.7.23 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) w.
∂x ∂y ∂z
This is a special case of equation 2.3.7.24 with g1 (x, y) = g1 (x), g2 (x, y) = g2 (y), and
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
This is a special case of equation 2.3.7.25 with g1 (x, y) = g1 (y), g2 (x, y) = g2 (x), and
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
This is a special case of equation 2.3.7.26 with g1 (x, y) = g1 (x), g2 (x, y) = g2 (y), and
h(x, y, z) = h1 (x) + h2 (y).
∂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
u1 = x y , u2 = b − t−1 f xy −a/b ta/b , t dt.
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.
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂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◦ . For k 6= 1 and m = 1, the substitution ξ = y 1−k leads to an equation of the form
2.3.7.22.
3◦ . For k = m = 1, see equation 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
leads to an equation of the form 2.3.7.22:
∂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
222 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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
The transformation ξ = e−λy , η= z 1−k leads to an equation of the form 2.3.7.22:
∂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
The transformation ξ = e−λy , η = e−βz leads to an equation of the form 2.3.7.22:
∂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)
The transformation ξ = dx, η = dy leads to an equation of the form
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)
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 =s(x, y, z)w.
∂x ∂y ∂z
Z Z
f2 (x) g1 (y)
The transformation ξ = dx, η = dy leads to an equation of the form
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)
where h̄1 (ξ, η) ≡ , h̄2 (ξ, η) ≡ , and h̄3 (ξ, η, z) ≡ .
f2 (x)g1 (y) f2 (x)g1 (y) f2 (x)g1 (y)
◆ 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.
2.4. Equations of the Form f1 ∂w
∂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
∂w ∂w ∂w
4. + (b1 x + b0 ) + (c1 y + c0 ) = aw + s1 x + s0 .
∂x ∂y ∂z
∂w ∂w ∂w
5. + (ay + k1 x + k0 ) + (bz + n1 x + n0 ) = (c1 x + c0 )w + s1 x + s0 .
∂x ∂y ∂z
∂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
224 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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
11. + (b1 x2 + b0 ) + (c1 y 2 + c0 ) = aw + s1 x2 + s0 .
∂x ∂y ∂z
∂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
2.4. Equations of the Form f1 ∂w
∂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
23. a + by + cz = kxn w + sxm .
∂x ∂y ∂z
1◦ . General solution for n 6= −1:
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)
where
( √ √
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
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 .
2.4. Equations of the Form f1 ∂w
∂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
√ √
|x| ab by − ab z if ab > 0,
√
u1 = by 2 − az 2 , u2 = √
−ab −ab z
|x| exp − arctan if ab < 0.
by
where
√ √ √ √
u1 = by + ( 2 − 1)cz |x|− 2 , u2 = by − ( 2 + 1)cz |x| 2 .
∂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
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 .
2.4. Equations of the Form f1 ∂w
∂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,
∂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 .
230 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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 .
This is a special case of equation 2.4.7.21 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 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 .
This is a special case of equation 2.4.7.21 with f1 (x) = a1 eλ1 x , f2 (x) = a2 eλ2 x , g1 (x, y) =
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
3. + a sinhn (βx) + b sinhk (λx) = cw + s sinhm (µx).
∂x ∂y ∂z
This is a special case of equation 2.4.7.1 with f (x) = a sinhn (βx), g(x) = b sinhk (λx),
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
This is a special case of equation 2.4.7.1 with f (x) = a coshn (βx), g(x) = b coshk (λx),
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).
232 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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
This is a special case of equation 2.4.7.1 with f (x) = a tanhn (βx), g(x) = b tanhk (λx),
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).
2.4. Equations of the Form f1 ∂w
∂x
+ f2 ∂w
∂y
+ f3 ∂w
∂z
= f4 w + f5 , fn = fn (x, y, z) 233
∂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
This is a special case of equation 2.4.7.1 with f (x) = a cothn (βx), g(x) = b cothk (λx),
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
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
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).
2.4. Equations of the Form f1 ∂w
∂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).
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
(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)
236 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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
2.4. Equations of the Form f1 ∂w
∂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
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
This is a special case of equation 2.4.7.1 with f (x) = a tann (βx), g(x) = b tank (λx),
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)
2.4. Equations of the Form f1 ∂w
∂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
This is a special case of equation 2.4.7.1 with f (x) = a cotn (βx), g(x) = b cotk (λx),
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
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.
2.4. Equations of the Form f1 ∂w
∂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).
242 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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.
2.4. Equations of the Form f1 ∂w
∂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
General solution: w = exp h(x) dx Φ(u1 , u2 ) + p(x) exp − h(x) dx dx ,
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
General solution: w = exp h(x) dx Φ(u1 , u2 ) + p(x) exp − h(x) dx dx ,
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
General solution: w = exp h(x) dx Φ(u1 , u2 ) + p(x) exp − h(x) dx dx ,
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).
244 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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
u2 = + E dx, E = exp sinh(λx) .
y − a cosh(λx) λ
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
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
This is a special case of equation 2.4.7.22 in which g1 (x, y) = g1 (x), g2 (x, y) = g2 (y),
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
This is a special case of equation 2.4.7.24 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
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.
246 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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
where x0 may be taken as arbitrary.
∂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.
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
∂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◦ . For k 6= 1 and m = 1, the substitution ξ = y 1−k leads to an equation of the form
2.4.7.20.
3◦ . For k = m = 1, see equation 2.4.7.20.
2.4. Equations of the Form f1 ∂w
∂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)
The transformation ξ = dx, η = dy leads to an equation of the form
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)
248 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
∂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)
The transformation ξ = dx, η = dy leads to an equation of the form
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)
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
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).
2.4. Equations of the Form f1 ∂w
∂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
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
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
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
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 .
∆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.
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 ,
∂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
∂w ∂w ∂w ∂w
3. + ax2 + bx3 + cx4 = 0.
∂x1 ∂x2 ∂x3 ∂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:
∂w ∂w ∂w
5. αβγx1 + βγ(βx3 + γx4 ) + αγ(αx2 + γx4 )
∂x1 ∂x2 ∂x3
∂w
+ αβ(αx2 + βx3 ) = 0.
∂x4
Integral basis:
∂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
⊙ Literature: E. Kamke (1965).
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
2.4. Equations of the Form f1 ∂w
∂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
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
n
X
∂w ∂w
14. + a k xm
1
k
= bw + c.
∂x1 ∂xk
k=2
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 .
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
This is a special case of equation 2.4.9.37.
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
This is a special case of equation 2.4.9.37.
2.4. Equations of the Form f1 ∂w
∂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.
256 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
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.
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.
2.4. Equations of the Form f1 ∂w
∂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
whose coefficients are defined as fk (x1 , x2 , x3 , . . . , xk−1 ) ≡ f¯k (x1 , u2 , x3 , . . . , xk−1 ), etc.
258 F IRST-O RDER E QUATIONS WITH T HREE OR M ORE I NDEPENDENT VARIABLES
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 .
⊙ Literature: A. D. Polyanin, V. F. Zaitsev, and A. Moussiaux (2002).
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 ),
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).
⊙ Literature: E. Kamke (1965).
Chapter 3
Second-Order
Parabolic Equations
with One Space Variable
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
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).
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).
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).
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
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
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
◮ Domain: 0 ≤ x ≤ l. Third boundary value problem (k1 > 0 and k2 > 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 )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
⊙ Literature: A. G. Butkovskiy (1979), H. S. Carslaw and J. C. Jaeger (1984).
268 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
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
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 −∞
3.1. Constant Coefficient Equations 271
where
1 (x − ξ)2
G(x, ξ, t) = √ exp − .
2 πat 4at
⊙ Literature: A. G. Butkovskiy (1979), H. S. Carslaw and J. C. Jaeger (1984).
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
⊙ Literature: A. G. Butkovskiy (1979), H. S. Carslaw and J. C. Jaeger (1984).
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
⊙ Literature: A. G. Butkovskiy (1979), H. S. Carslaw and J. C. Jaeger (1984).
272 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
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
∞
1 2X nπx nπξ an2 π 2 t
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 − .
2 πat n=−∞ 4at 4at
The first series converges rapidly at large t and the second series at small t.
⊙ Literature: A. G. Butkovskiy (1979), B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
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
The first series converges rapidly at large t and the second series at small t.
2◦ . The following conditions are prescribed:
The first series converges rapidly at large t and the second series at small t.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980), A. V. Bitsadze and D. F. Kalini-
chenko (1985).
3.1. Constant Coefficient Equations 275
∂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.
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
2◦ . Fundamental solution:
1 x2
E (x, t) = √ exp − + bt .
2 πat 4at
where
1 (x − ξ)2
G(x, ξ, t) = √ exp − + bt .
2 πat 4at
⊙ Literature: A. G. Butkovskiy (1979).
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).
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)
− 2k exp − − kη dη .
0 4at
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
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
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
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
The solution w(x, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞
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 .
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
⊙ Literature: A. G. Butkovskiy (1979).
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
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
280 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂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.
∂u ∂2u
= a 2 + exp(βt + µx)Φ(x, t), (2)
∂t ∂x
which is considered in Section 3.1.2 in detail.
3.1. Constant Coefficient Equations 281
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 ∞
w(x, t) = f (ξ)G(x, ξ, t) dξ + Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ,
−∞ 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
282 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
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 < ∞.
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,
(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 +
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
3.1. Constant Coefficient Equations 283
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
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
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
The solution w(x, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞
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.
2◦ . Fundamental solution:
1 (x + bt)2
E (x, t) = √ exp − + ct .
2 πat 4at
∂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
∂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 ∞
w(x, t) = f (ξ)G(x, ξ, t) dξ + Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ,
−∞ 0 −∞
where
1 b b2 (x − ξ)2
G(x, ξ, t) = √ exp (ξ − x) + c − t− .
2 πat 2a 4a 4at
286 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
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
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
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
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
The solution w(x, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
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
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.
⊙ Literature: H. S. Carslaw and J. C. Jaeger (1984), A. D. Polyanin, A. V. Vyazmin, A. I. Zhurov, and
D. A. Kazenin (1998).
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.
290 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
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
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
3.2. Heat Equation with Axial or Central Symmetry and Related Equations 291
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
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
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
The numerical values of the first five roots µn = µn (s) can be found in Abramowitz and
Stegun (1964).
⊙ Literature: A. V. Lykov (1967), H. S. Carslaw and J. C. Jaeger (1984).
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
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,
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.
3.2. Heat Equation with Axial or Central Symmetry and Related Equations 295
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).
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).
⊙ Literature: A. V. Lykov (1967), H. S. Carslaw and J. C. Jaeger (1984).
296 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
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 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
The numerical values of the first six roots µn can be found in Carslaw and Jaeger (1984).
⊙ Literature: A. V. Lykov (1967), H. S. Carslaw and J. C. Jaeger (1984).
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) = ,
∂ξ ξ=R1 ∂ξ ξ=R2
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
3.2. Heat Equation with Axial or Central Symmetry and Related Equations 297
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).
w = f (r) at t = 0
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 ∞
Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ,
0 R
∂w
∂2w 2 ∂w
3.2.3 Equation of the Form =a +
∂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.
3.2. Heat Equation with Axial or Central Symmetry and Related Equations 299
∂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.
300 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
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),
n=1
an (2n + 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.
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
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
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
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
The numerical values of the first six roots µn can be found in Carslaw and Jaeger (1984).
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
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) .
∂ξ ξ=R1 ∂ξ ξ=R2
3.2. Heat Equation with Axial or Central Symmetry and Related Equations 303
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
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,
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 )],
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
304 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
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,
w = f (r) at t = 0
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
⊙ Literature: H. S. Carslaw and J. C. Jaeger (1984).
3.2. Heat Equation with Axial or Central Symmetry and Related Equations 305
∂w
∂2w 2 ∂w
3.2.4 Equation of the Form =a + + Φ(r, t)
∂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.
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
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
306 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
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
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,
∞
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
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
3.2. Heat Equation with Axial or Central Symmetry and Related Equations 307
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) .
∂ξ ξ=R1 ∂ξ ξ=R2
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 )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 ).
⊙ Literature: H. S. Carslaw and J. C. Jaeger (1984).
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
Φ(ξ, τ )G(r, ξ, t − τ ) dξ dτ,
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.
3.2. Heat Equation with Axial or Central Symmetry and Related Equations 309
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
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).
X∞
1 n n d2 1 − 2β d
w(x, t) = f (x) + t L [f (x)], L≡ + ,
n=1
n! dx2 x dx
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 that contains only even
powers.
A solution containing an arbitrary function of time:
∞
X 1 (n)
w(x, t) = g(t) + x2n gt (t),
n=1
4n n! (1 − β)(2 − β) . . . (n − β)
310 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
where g(t) is any infinitely differentiable function. This solution is bounded at x = 0 and
possesses the properties
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
Here, γ(β, z) is the incomplete gamma function and Γ(β) = γ(β, ∞) is the gamma function.
⊙ 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.
∂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
2◦ . The transformation
w(x, t) = u(z, t) exp btx + 13 ab2 t3 + ct , z = x + abt2
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.
1◦ . Particular solutions (A and µ are arbitrary constants):
√
√ 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
∂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
leads to the constant coefficient equation ∂τ u = ∂zz u, which is considered in Section 3.1.1.
∂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.
The following conditions are prescribed:
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
3◦ . The transformation
∂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
leads to an equation of the form 3.3.1.5:
∂u ∂2u k
= + z −2 u.
∂τ ∂z 2 a
See also W. Miller, Jr. (1977).
∂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
∂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.
316 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂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.
1◦ . Particular solutions (A and µ are arbitrary constants):
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.
3.3. Equations Containing Power Functions and Arbitrary Parameters 317
2◦ . The transformation
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
This is a special case of equation 3.8.1.3 with f (t) = bt + c.
1◦ . Particular solutions (A and µ are arbitrary constants):
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 .
2◦ . The transformation
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 .
2◦ . The transformation
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
318 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂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
This is a special case of equation 3.8.2.1 with f (t) = bt + c.
1◦ . Particular solutions (A, B, and µ are arbitrary constants):
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.
3.3. Equations Containing Power Functions and Arbitrary Parameters 319
∂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.
1◦ . Particular solutions (A, B, and µ are arbitrary constants):
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.
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.
3◦ . Domain: −∞ < x < ∞. Cauchy problem.
An initial condition is prescribed:
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
∂w ∂2w ∂w
4. =a + x(bt + c) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.2.3 with f (t) = bt + c.
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.
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 ∂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
1
τ= t2b+1 , z = xtb ,
2b + 1
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.
3.3. Equations Containing Power Functions and Arbitrary Parameters 321
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),
∂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
∂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.
322 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂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◦ . The substitution
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.
2◦ . The substitution
1 2
w(x, t) = u(x, t) exp 2 Ax + Bx + Ct
∂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
This is a special case of equation 3.8.7.5.
∂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
1◦ . The transformation
∂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.
3.3. Equations Containing Power Functions and Arbitrary Parameters 325
X∞
(at)n n d2
w(x, t) = f (x) + L [f (x)], L≡x ,
n=1
n! dx2
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.
where g(t) is any infinitely differentiable function and A is an arbitrary number. This
solution possesses the properties
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
∂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 .
326 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
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
∂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
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.
3.3. Equations Containing Power Functions and Arbitrary Parameters 327
TABLE 3.2
Particular solutions of equation 3.3.4.7 for different values
of the determining parameters (µ is an arbitrary number)
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
1. = ax2 + bx + cw.
∂t ∂x2 ∂x
1◦ . Particular solutions (A, B, and µ are arbitrary constants):
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.
2◦ . The transformation
a−b
w(x, t) = |x|n exp (c − an2 )t u(z, t), z = ln |x|, n= ,
2a
leads to a constant coefficient equation, ∂t u = a∂zz u, which is considered in Section 3.1.1.
∂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
328 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂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
3.3. Equations Containing Power Functions and Arbitrary Parameters 329
∂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
which is considered in Section 3.1.3.
∂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.
1◦ . Particular solutions (A, B, and µ are arbitrary constants):
∂w ∂2w
3. = (x2 + a2 )2 + Φ(x, t).
∂t ∂x2
Domain: 0 ≤ x ≤ l. First boundary value problem.
The following conditions are prescribed:
∂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
which is considered in Section 3.1.4.
∂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.
3.3. Equations Containing Power Functions and Arbitrary Parameters 331
∂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.
1◦ . Particular solutions (A, B, and µ are arbitrary constants):
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.
332 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
w = w0 at t = 0 (initial condition),
w = w1 at x = 0 (boundary condition),
w → w0 at x → ∞ (boundary condition),
∂w ∂2w 1 − 2k ∂w 1
= + , k= ,
∂τ ∂ξ 2 ξ ∂ξ n+1
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
∂v n−1 ∂ 2 v
= an2 ξ n . (2)
∂t ∂ξ 2
3.3. Equations Containing Power Functions and Arbitrary Parameters 333
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
2◦ . The transformation
∂w ∂2w ∂w
9. = a x2m 2 + mx2m−1 .
∂t ∂x ∂x
∂w ∂2w ∂w
10. = axn + x(btm + c) .
∂t ∂x2 ∂x
∂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 .
3.3. Equations Containing Power Functions and Arbitrary Parameters 335
∂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.
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.
336 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
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.
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 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
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.
λm = 4m + 1.68 (m = 0, 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).
338 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂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)
∂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
∂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
∂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 .
2◦ . The transformation
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
where Jν (ξ) and Yν (ξ) are the Bessel functions of the first and second kind, respectively.
340 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂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.
1◦ . Particular solutions (A and λ are arbitrary constants):
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
2◦ . The transformation
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
∂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.
3.4. Equations Containing Exponential Functions and Arbitrary Parameters 341
∂w ∂2w
7. =a + (beβt + ceλt )w.
∂t ∂x2
∂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
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
3.4.2 Equations of the Form =a + f (x, t)
∂t ∂x2 ∂x
∂w ∂2w ∂w
1. =a + (beβt + c) .
∂t ∂x2 ∂x
b βt
2◦ . On passing from t, x to the new variables t, z = x + e + ct, we obtain a constant
β
coefficient equation, ∂t w = a∂zz w, which is considered in Section 3.1.1.
∂w ∂2w ∂w
2. =a + (beβx + c) .
∂t ∂x2 ∂x
∂w ∂2w ∂w
= aβ 2 z 2 2 + βz(bz + c + aβ) .
∂t ∂z ∂z
3.4. Equations Containing Exponential Functions and Arbitrary Parameters 343
∂w ∂2w ∂w
3. =a + x(beβt + c) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.2.3 with f (t) = beβt + c.
1◦ . Particular solutions (A, B, and λ are arbitrary constants):
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.
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 ∂2w ∂w
4. =a + (beβt + ceλt ) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.2.1 with f (t) = beβt + ceλt .
1◦ . Particular solutions (A, B, and µ are arbitrary constants):
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
coefficient equation, ∂t w = a∂zz w, which is considered in Section 3.1.1.
∂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 .
344 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂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
This is a special case of equation 3.8.2.3 with f (t) = beβt + ceλt .
1◦ . Particular solutions (A, B, and µ are arbitrary constants):
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.
∂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
3.4. Equations Containing Exponential Functions and Arbitrary Parameters 345
∂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.
346 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂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
This is a special case of equation 3.8.4.5 with f (t) = beβt + c.
∂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
1◦ . Particular solutions (A, B, and µ are arbitrary constants):
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
3.4. Equations Containing Exponential Functions and Arbitrary Parameters 347
∂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,
348 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
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
where f (x) is any infinitely differentiable function. This solution satisfies the initial con-
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 .
∂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
∂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
This is a special case of equation 3.8.1.3 with f (t) = b coshk ωt + c.
∂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
This is a special case of equation 3.8.2.1 with f (t) = b coshk ωt + c.
350 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂w ∂2w ∂w
6. =a + x(b coshk ωt + c) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.2.3 with f (t) = b coshk ωt + c.
∂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
∂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
This is a special case of equation 3.8.4.2 with f (t) = b coshk ωt + c.
∂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
This is a special case of equation 3.8.4.5 with f (t) = b coshk ωt + c.
∂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
This is a special case of equation 3.8.1.3 with f (t) = b sinhk ωt + c.
∂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.
3.5. Equations Containing Hyperbolic Functions and Arbitrary Parameters 351
∂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
∂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
∂w ∂2w
1. =a + (b tanhk ωt + c)w.
∂t ∂x2
∂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.
352 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂w ∂2w
3. =a + x(b tanhk ωt + c)w.
∂t ∂x2
This is a special case of equation 3.8.1.3 with f (t) = b tanhk ωt + c.
∂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
This is a special case of equation 3.8.2.1 with f (t) = b tanhk ωt + c.
∂w ∂2w ∂w
6. =a + x(b tanhk ωt + c) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.2.3 with f (t) = b tanhk ωt + c.
∂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
∂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
This is a special case of equation 3.8.4.2 with f (t) = b tanhk ωt + c.
∂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
This is a special case of equation 3.8.4.5 with f (t) = b tanhk ωt + c.
∂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.
3.5. Equations Containing Hyperbolic Functions and Arbitrary Parameters 353
∂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
∂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
∂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
∂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
∂w ∂2w
4. =a + (−bx2 + c lnk t)w.
∂t ∂x2
∂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
This is a special case of equation 3.8.2.3 with f (t) = b ln t + c.
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
considered in Section 3.1.1.
∂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
This is a special case of equation 3.8.8.3 with n(t) = a, f (t) = g(t) = s(t) = p(t) = 0,
and h(t) = b. See also equation 3.8.6.5.
∂w ∂2w
8. = ax2 + btk ln2 x w.
∂t ∂x2
This is a special case of equation 3.8.8.3 with n(t) = a, f (t) = g(t) = s(t) = p(t) = 0,
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
This is a special case of equation 3.8.4.5 with f (t) = b ln t + c.
356 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂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
This is a special case of equation 3.8.1.3 with f (t) = b cosk ωt + c.
∂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
This is a special case of equation 3.8.2.1 with f (t) = b cosk ωt + c.
∂w ∂2w ∂w
6. =a + x(b cosk ωt + c) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.2.3 with f (t) = b cosk ωt + c.
∂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
This is a special case of equation 3.8.4.2 with f (t) = b cosk ωt + c.
∂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
This is a special case of equation 3.8.4.5 with f (t) = b cosk ωt + c.
∂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
This is a special case of equation 3.8.1.3 with f (t) = b sink ωt + c.
∂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
This is a special case of equation 3.8.2.1 with f (t) = b sink ωt + c.
∂w ∂2w ∂w
6. =a + x(b sink ωt + c) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.2.3 with f (t) = b sink ωt + c.
∂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
∂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
This is a special case of equation 3.8.4.2 with f (t) = b sink ωt + c.
358 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂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
This is a special case of equation 3.8.4.5 with f (t) = b sink ωt + c.
∂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
This is a special case of equation 3.8.1.3 with f (t) = b tank ωt + c.
∂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
This is a special case of equation 3.8.2.1 with f (t) = b tank ωt + c.
∂w ∂2w ∂w
6. =a + x(b tank ωt + c) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.2.3 with f (t) = b tank ωt + c.
∂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
∂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
This is a special case of equation 3.8.4.2 with f (t) = b tank ωt + c.
∂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
This is a special case of equation 3.8.4.5 with f (t) = b tank ωt + c.
∂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
This is a special case of equation 3.8.1.3 with f (t) = b cotk ωt + c.
∂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
This is a special case of equation 3.8.2.1 with f (t) = b cotk ωt + c.
∂w ∂2w ∂w
6. =a + x(b cotk ωt + c) .
∂t ∂x2 ∂x
This is a special case of equation 3.8.2.3 with f (t) = b cotk ωt + c.
∂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.
360 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂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
∂w ∂2w
1. =a + f (t)w.
∂t ∂x2
1◦ . Particular solutions (A, B, and λ are arbitrary constants):
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
1◦ . Particular solutions (A and λ are arbitrary constants):
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 .
2◦ . The transformation
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
considered in Section 3.1.1.
∂w ∂2w
4. =a + x2 f (t)w.
∂t ∂x2
This is a special case of equation 3.8.7.5.
∂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),
2◦ . The transformation
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
considered in Section 3.1.1.
∂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
∂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
2◦ . The transformation
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
leads to an equation of the form 3.8.1.6:
∂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
3.8. Equations Containing Arbitrary Functions 363
∂w ∂2w
9. =a + x2 f (t) + xg(t) + h(t) w.
∂t ∂x2
This is a special case of equation 3.8.7.5.
∂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
3.8.2 Equations of the Form =a + f (x, t)
∂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
2◦ . On passing from t, x to the new variables t, z = x + f (t) dt, we obtain a constant
coefficient equation, ∂t w = a∂zz w, which is considered in Section 3.1.1.
∂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.
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.
364 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
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.
N
X
w(x, t) = Cn wn (x, t) (Cn are arbitrary constants)
n=0
∂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).
1◦ . Particular solutions (A, B, and λ are arbitrary constants):
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.
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.
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 t = 0 (initial condition),
w = w1 at x = 0 (boundary condition),
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.
366 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂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 t = 0 (initial condition),
w = w1 at x = 0 (boundary condition),
w → w0 at x → ∞ (boundary condition)
∂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.
3.8. Equations Containing Arbitrary Functions 367
∂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 ,
∂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
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
∂w ∂2w ∂w
3.8.3 Equations of the Form =a + f (x, t) + g(x, t)w
∂t ∂x2 ∂x
∂w ∂2w ∂w
1. =a + f (t) + g(t)w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.7.3.
1◦ . Particular solutions (A, B, and λ are arbitrary constants):
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) .
2◦ . The transformation
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
1◦ . Particular solutions (A and B are arbitrary constants):
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)
∂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
where F (t) = B exp f (t) dt , leads to a constant coefficient equation, ∂τ u = a∂zz u,
which is considered in Section 3.1.1.
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
∂w ∂2w ∂w
6. =a + [xf (t) + g(t)] + h(t)w.
∂t ∂x2 ∂x
Z
where F (t) = B exp f (t) dt , leads to a constant coefficient equation, ∂τ u = a∂zz u,
which is considered in Section 3.1.1.
∂w ∂2w ∂w
7. =a + [xf (t) + g(t)] + [xh(t) + s(t)]w.
∂t ∂x2 ∂x
∂w ∂2w ∂w
8. =a + [xf (t) + g(t)] + [x2 h(t) + xs(t) + p(t)]w.
∂t ∂x2 ∂x
∂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
This is a special case of equation 3.8.8.1.
∂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|
∂u ∂2u ∂u
=a 2 −a .
∂τ ∂z ∂z
∂w ∂2w
3. = ax2 + ln x f (t)w.
∂t ∂x2
This is a special case of equation 3.8.8.3.
3.8. Equations Containing Arbitrary Functions 371
∂w ∂2w
4. = ax2 + ln2 x f (t)w.
∂t ∂x2
This is a special case of equation 3.8.8.3.
∂w ∂2w ∂w
5. = axn + xf (t) .
∂t ∂x2 ∂x
1◦ . Particular solutions (A and B are arbitrary constants):
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 .
where Z
F (t) = exp f (t) dt ,
∂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
372 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂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
This is a special case of equation 3.8.6.4.
1◦ . The equation has particular solutions of the form
where the function u(x) is determined by solving the following linear ordinary differential
equation with parameter λ:
∂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
3.8. Equations Containing Arbitrary Functions 373
∂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.
1◦ . The equation has particular solutions of the form
where the function u(x) is determined by solving the following linear ordinary differential
equation with parameter λ:
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)
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,
374 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
where the prime stands for the differentiation with respect to x. Integrating these equations
successively in order of decreasing number i, we obtain
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
3.8. Equations Containing Arbitrary Functions 375
with the aid of the integral operator Lf that is defined by relation (5) in equation 3.8.6.1.
376 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
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.
1◦ . Particular solutions (A and B are arbitrary constants):
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)
3.8. Equations Containing Arbitrary Functions 377
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.
N
X
w(x, t) = Cn wn (x, t) (Cn are arbitrary numbers),
n=0
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.
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)
where the function u(x) is determined by solving the following linear ordinary differential
equation with parameter λ:
f (x)u′′xx + g(x)u′x + h(x) + λ u = 0.
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
3.8. Equations Containing Arbitrary Functions 381
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
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
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
∂w ∂ 2w ∂w
6. = f (x) + g(x) + h(t)w.
∂t ∂x2 ∂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
1◦ . Particular solutions (A, B, and λ are arbitrary constants):
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 .
3.8. Equations Containing Arbitrary Functions 383
for the function w(τ, z) we obtain a constant coefficient equation, ∂τ w = ∂zz w, which is
considered in Section 3.1.1.
∂w ∂ 2w ∂w
2. = f (t) + xg(t) .
∂t ∂x2 ∂x
1◦ . Particular solutions (A, B, and λ are arbitrary constants):
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 .
for the function w(τ, z) we obtain a constant coefficient equation, ∂τ w = ∂zz w, which is
considered in Section 3.1.1.
∂w ∂ 2w ∂w
3. = f (t) + g(t) + h(t)w.
∂t ∂x2 ∂x
This is a special case of equation 3.8.7.4.
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.
384 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
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 ,
∂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
3.8. Equations Containing Arbitrary Functions 385
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
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
Let us perform the transformation
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 ,
∂w ∂ 2w ∂w
2. = x2 f (t) + xg(t) + h(t)w.
∂t ∂x2 ∂x
∂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
1◦ . Particular solutions (A, B, and λ are arbitrary constants):
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
2◦ . The transformation
Z Z
1
w(x, t) = x exp g(t) dt u(ξ, τ ), ξ= , τ= f (t) dt
x
∂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
∂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
which is considered in Section 3.1.3.
∂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):
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 .
3.8. Equations Containing Arbitrary Functions 389
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.
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
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 .
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.
3.8. Equations Containing Arbitrary Functions 391
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
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
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 .
2◦ . 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
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
∂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).
The transformation (A and B are any numbers)
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)
which is considered in Section 3.1.1.
⊙ 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).
The transformation (A and B are any numbers)
Z n+1 Z
h (x) g(x)
t= dx + A, z = yh(x), where h(x) = B exp − dx ,
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
x = 0, w = w0 ; y = 0, w = w1 ; y → ∞, w → w0
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.
⊙ Literature: A. D. Polyanin, A. V. Vyazmin, A. I. Zhurov, and D. A. Kazenin (1998).
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.
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)
∂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
⊙ Literature: A. D. Polyanin, A. V. Vyazmin, A. I. Zhurov, and D. A. Kazenin (1998).
396 S ECOND -O RDER PARABOLIC E QUATIONS WITH O NE S PACE VARIABLE
h i
∂w b ∂w ∂2w
7. f (x)y n−1 + g(x)y n − = .
∂x y2 ∂y ∂y
The transformation (A and B are any numbers)
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).
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.
3.9. Equations of Special Form 397
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).
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+ ).
⊙ Literature: W. Miller, Jr. (1977).
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= ,
~
Eigenvalues:
r !
~2 1 8mU0 a2
En = − (s − n)2 , s= −1 + 1+ , 0 ≤ n < s.
2ma2 2 ~2
3.9. Equations of Special Form 399
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).
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
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
◮ 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.
Fundamental solution:
1 x2 + y 2
E (x, y, t) = exp − .
4πat 4at
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
considered in Section 3.1.1.
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.
⊙ Literature: W. Miller, Jr. (1977).
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
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
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
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
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 = ξt + a/t, R = − 14 (ξ 2 + η 2 )t
15 Airy function Airy function
y = ηt + b/t + 12 (aξ + bη)/t
∂w
4.1. Heat Equation ∂t
= 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
406 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
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
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, 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
The solution w(x, y, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
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 ∞
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
⊙ Literature: A. G. Butkovskiy (1979), H. S. Carslaw and J. C. Jaeger (1984).
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, 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
408 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
The solution w(x, y, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
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
⊙ Literature: H. S. Carslaw and J. C. Jaeger (1984).
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, 0, t − τ ) dξ dτ,
0 0
∂w
4.1. Heat Equation ∂t
= 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
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, 0, η, t − τ ) dη dτ
0 0
Z tZ ∞
∂
+a g2 (ξ, τ ) G(x, y, ξ, η, t − τ ) dξ dτ,
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
× exp − + exp −
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
Solution:
Z ∞Z l
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 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=l
Z tZ l
∂
+a g3 (ξ, τ ) G(x, y, ξ, η, t − τ ) dξ dτ,
0 0 ∂η η=0
where
G(x, y, ξ, η, t) = G1 (x, ξ, t) G2 (y, η, t),
∞
2X nπx nπξ an2 π 2 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
Example 4.6. The initial temperature is uniform, f (x, y) = w0 . The boundary is maintained at
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
where
The solution w(x, y, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
Here
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
and the µn are positive roots of the transcendental equation = 2 .
µ µ − k1 k2
Solution:
Z ∞Z l
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 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=l
Z tZ l
−a g3 (ξ, τ ) G(x, y, ξ, 0, t − τ ) dξ dτ,
0 0
where
Solution:
Z ∞Z l
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, 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, η, t − τ ) dξ dτ,
0 0 ∂η η=0
where
Example 4.7. The initial temperature is uniform, f (x, y) = w0 . The boundary is maintained at
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
⊙ Literature: H. S. Carslaw and J. C. Jaeger (1984).
Solution:
Z l1Z l2
w(x, y, t) = f (ξ, η) G(x, y, ξ, η, t) dη dξ
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
Solution:
Z l1Z l2
w(x, y, t) = f (ξ, η) G(x, y, ξ, η, t) dη dξ
0 0
Z tZ
l2
∂
+a g1 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=0
Z t Z l2
∂
−a g2 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
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
+a g4 (ξ, τ ) G(x, y, ξ, l2 , t − τ ) dξ dτ,
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
w(x, y, t) = f (ξ, η) G(x, y, ξ, η, t) dη dξ
0 0
Z t Z l2
∂
+a g1 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
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
∂
+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, l2 , t − τ ) dξ dτ,
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
One-dimensional problems with axial symmetry that have solutions of the form w =
w(r, t) are considered in Section 3.2.1.
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
4.1. Heat Equation ∂t
= a∆2 w 417
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.
⊙ Literature: H. S. Carslaw and J. C. Jaeger (1984).
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, 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.
418 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
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
4.1. Heat Equation ∂t
= a∆2 w 419
Solution:
Z 2πZ R2
w(r, ϕ, t) = f (ξ, η) G(r, ϕ, ξ, η, t)ξ dξ dη
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
Solution:
Z 2πZ R2
w(r, ϕ, t) = f (ξ, η) G(r, ϕ, ξ, η, t)ξ dξ dη
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
420 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
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
µnm are positive roots of the transcendental equation
µJn′ (µR1 ) − k1 Jn (µR1 ) µYn′ (µR2 ) + k2 Yn (µR2 )
= µYn′ (µR1 ) − k1 Yn (µR1 ) µJn′ (µR2 ) + k2 Jn (µR2 ) .
Solution:
Z ϕ0Z ∞
w(r, ϕ, t) = f (ξ, η) G(r, ϕ, ξ, η, t)ξ dξ dη
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
Solution:
Z ϕ0Z ∞
w(r, ϕ, t) = f (ξ, η) G(r, ϕ, ξ, η, t)ξ dξ dη
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
Solution:
Z ϕ0Z R
w(r, ϕ, t) = f (ξ, η) G(r, ϕ, ξ, η, t)ξ dξ dη
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 ∂
+a g2 (ξ, τ ) G(r, ϕ, ξ, η, t − τ ) dξ dτ
0 0 ξ ∂η η=0
Z tZ R
1 ∂
−a g3 (ξ, τ ) G(r, ϕ, ξ, η, t − τ ) dξ dτ.
0 0 ξ ∂η η=ϕ0
422 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
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.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980), H. S. Carslaw and J. C. Jaeger
(1984).
Solution:
Z ϕ0Z R
w(r, ϕ, t) = f (ξ, η) G(r, ϕ, ξ, η, t)ξ dξ dη
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.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
Some problems for this domain were studied in Budak, Samarskii, and Tikhonov (1980).
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
4.1. Heat Equation ∂t
= 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.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980), H. S. Carslaw and J. C. Jaeger
(1984).
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).
426 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
Here,
where the 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).
2◦ . A semiinfinite circular cylinder is considered. The following conditions are prescribed:
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, R, η, t − τ ) dη dτ
0 0
Z tZ R
∂
+ 2πa ξg2 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ.
0 0 ∂η η=0
Here,
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).
3◦ . A semiinfinite circular cylinder is considered. The following conditions are prescribed:
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
1 µ2n µn r µn ξ aµ2n t
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
where the µn are positive roots of the transcendental equation
µJ1 (µ) − kRJ0 (µ) = 0.
Example 4.12. The initial temperature is the same at every point of the cylinder, f (r, z) = w0 .
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
⊙ Literature: H. S. Carslaw and J. C. Jaeger (1984).
Here,
where the 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).
Example 4.14. The initial temperature is the same at every point of the cylinder, f (r, z) = w0 .
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).
⊙ Literature: H. S. Carslaw and J. C. Jaeger (1984).
Solution:
Z lZ R
w(r, z, t) = 2π ξf (ξ, η) G(r, z, ξ, η, t) dξ dη
0 0
Z tZ l
+ 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
Z tZ R
+ 2πa ξg3 (ξ, τ ) G(r, z, ξ, l, t − τ ) dξ dτ.
0 0
430 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
Here,
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).
The solution w(r, z, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
Solution:
Z lZ R
w(r, z, t) = 2π ξf (ξ, η) G(r, z, ξ, η, t) dξ dη
0 0
Z tZ l
∂
− 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, 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,
where the 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).
2◦ . A circular cylinder of finite length is considered. The following conditions are pre-
scribed:
w = f (r, z) at t = 0 (initial condition),
∂r w = g1 (z, t) at r = R (boundary condition),
w = g2 (r, t) at z = 0 (boundary condition),
w = g3 (r, t) at z = l (boundary condition).
Solution:
Z lZ R
w(r, z, t) = 2π ξf (ξ, η) G(r, z, ξ, η, t) dξ dη
0 0
Z tZ l
+ 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, 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
432 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
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 J02 (µn ) R R R
∞
2X nπz nπη an2 π 2 t
G2 (z, η, t) = sin sin exp − ,
l l l l2
n=1
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).
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.
A hollow circular cylinder of finite length is considered. The following conditions are
prescribed:
w = f (r, z) at t = 0 (initial condition),
∂r w = g1 (z, t) at r = R1 (boundary condition),
∂r w = g2 (z, t) at r = R2 (boundary condition),
∂z w = g3 (r, t) at z = 0 (boundary condition),
∂z w = g4 (r, t) at z = l (boundary condition).
Solution:
Z lZ R2
w(r, z, t) = 2π ξf (ξ, η) G(r, z, ξ, η, t) dξ dη
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
− 2πa ξg3 (ξ, τ ) G(r, z, ξ, 0, t − τ ) dξ dτ
0 R1
Z tZ R2
+ 2πa ξg4 (ξ, τ ) G(r, z, ξ, l, t − τ ) dξ dτ.
0 R1
Here,
where Jk (µ) and Yk (µ) are Bessel functions of order k = 0, 1 and the µn are positive roots
of the transcendental equation
A hollow circular cylinder of finite length is considered. The following conditions are
prescribed:
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
w = f (x, y) at t = 0.
Solution:
Z ∞Z ∞
w(x, y, t) = f (ξ, η) G(x, y, ξ, η, t) dξ dη
−∞ −∞
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
Solution:
Z ∞Z ∞
w(x, y, t) = f (ξ, η) G(x, y, ξ, η, t) dη dξ
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
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, 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
436 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
The solution w(x, y, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
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 ∞
w(x, y, t) = f (ξ, η) G(x, y, ξ, η, t) dξ dη
0 0
Z tZ ∞ h ∂ i
+a g1 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=0
Z tZ ∞ h ∂ i
+a g2 (ξ, τ ) G(x, y, ξ, η, t − τ ) dξ dτ
0 0 ∂η η=0
Z t Z ∞Z ∞
+ Φ(ξ, η, τ ) G(x, y, ξ, η, t − τ ) dξ dη dτ,
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
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, 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 ∞
+ Φ(ξ, η, τ ) G(x, y, ξ, η, t − τ ) dξ dη dτ,
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
The solution w(x, y, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
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
438 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
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
The solution w(r, z, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
tan(λl) k2 + k3
µJ1 (µ) − k1 RJ0 (µ) = 0, = 2 .
λ λ − k2 k3
440 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
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 ∞
Φ(ξ, η, τ ) G(x, y, ξ, η, t − τ ) dη dξ dτ,
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
Φ(ξ, η, τ ) G(x, y, ξ, η, t − τ ) dη dξ dτ,
0 0 0
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
The solution w(x, y, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
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 .
λm 2λm λm +k42 2λ2m 2 λm
tan(µl1 ) k1 + k2 tan(λl2 ) k3 + k4
= 2 , = 2 .
µ µ − k1 k2 λ λ − k3 k4
442 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
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
Φ(ξ, η, τ ) G(x, y, ξ, η, t − τ ) dη dξ dτ,
0 0 0
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.
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
4.2. Heat Equation with a Source ∂t
= a∆2 w + Φ(x, y, t) 443
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.
⊙ Literature: H. S. Carslaw and J. C. Jaeger (1984).
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 )
with the additional term
Z tZ ϕ0Z R
Φ(ξ, η, τ ) G(r, ϕ, ξ, η, t − τ )ξ dξ dη dτ, (3)
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 )
with the additional term (3); the Green’s function is also taken from Section 4.1.2.
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
Z t Z ∞Z R
+ 2π ξΦ(ξ, η, τ ) G(r, z, ξ, η, t − τ ) dξ dη dτ.
0 0 0
446 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
Here,
where the 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).
⊙ Literature: H. S. Carslaw and J. C. Jaeger (1984).
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, 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
Z t Z ∞Z R
+ 2π ξΦ(ξ, η, τ ) G(r, z, ξ, η, t − τ ) dξ dη dτ.
0 0 0
Here,
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).
∂w
4.2. Heat Equation with a Source ∂t
= a∆2 w + Φ(x, y, t) 447
The solution w(r, z, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
Here, J0 (µ) is the zeroth Bessel function, and the µn are positive roots of the transcendental
equation
µJ1 (µ) − k1 RJ0 (µ) = 0.
⊙ Literature: H. S. Carslaw and J. C. Jaeger (1984).
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, 0, t − τ ) dξ dτ
0 0
Z tZ ∞Z R
+ 2π ξΦ(ξ, η, τ ) G(r, z, ξ, η, t − τ ) dξ dη dτ.
0 0 0
448 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
Here,
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, R, η, t − τ ) dη dτ
0 0
Z tZ R
∂
+ 2πa ξg2 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ
0 0 ∂η η=0
Z tZ ∞Z R
+ 2π ξΦ(ξ, η, τ ) G(r, z, ξ, η, t − τ ) dξ dη dτ.
0 0 0
Here,
The solution w(r, z, t) is determined by the formula in Item 2◦ (see above) where
Here,
where the 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).
450 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
A circular cylinder of finite length is considered. The following conditions are prescribed:
Solution:
Z lZ R
w(r, z, t) = 2π ξf (ξ, η) G(r, z, ξ, η, t) dξ dη
0 0
Z tZ l
+ 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
Z tZ R
+ 2πa ξg3 (ξ, τ ) G(r, z, ξ, l, t − τ ) dξ dτ
0 0
Z tZ l Z R
+ 2π ξΦ(ξ, η, τ ) G(r, z, ξ, η, t − τ ) dξ dη dτ.
0 0 0
Here,
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).
A circular cylinder of finite length is considered. The following conditions are prescribed:
The solution w(r, z, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
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) = 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 ,
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
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 1
J 2 (µn ) R R R
∞
1 2X nπz nπη an2 π 2 t
G2 (z, η, t) = + cos cos exp − ,
l l n=1 l l l2
452 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
where the 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).
2◦ . A circular cylinder of finite length is considered. The following conditions are pre-
scribed:
w = f (r, z) at t = 0 (initial condition),
∂r w = g1 (z, t) at r = R (boundary condition),
w = g2 (r, t) at z = 0 (boundary condition),
w = g3 (r, t) at z = l (boundary condition).
Solution:
Z lZ R
w(r, z, t) = 2π ξf (ξ, η) G(r, z, ξ, η, t) dξ dη
0 0
Z tZ l
+ 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, 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
Z tZ lZ 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),
∞
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
∞
2X nπz nπη an2 π 2 t
G2 (z, η, t) = sin sin exp − ,
l n=1 l l l2
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).
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
A hollow circular cylinder of finite length is considered. The following conditions are
prescribed:
w = f (r, z) at t = 0 (initial condition),
∂r w = g1 (z, t) at r = R1 (boundary condition),
∂r w = g2 (z, t) at r = R2 (boundary condition),
∂z w = g3 (r, t) at z = 0 (boundary condition),
∂z w = g4 (r, t) at z = l (boundary condition).
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.
A hollow circular cylinder of finite length is considered. The following conditions are
prescribed:
Solution:
Z lZ R2
w(r, z, t) = 2π ξf (ξ, η) G(r, z, ξ, η, t) dξ dη
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
454 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
Z tZ R2
− 2πa ξg3 (ξ, τ ) G(r, z, ξ, 0, t − τ ) dξ dτ
0 R1
Z t Z R2
+ 2πa ξg4 (ξ, τ ) G(r, z, ξ, l, t − τ ) dξ dτ
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
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
transcendental equation
k1 J0 (λR1 ) + λJ1 (λR1 ) k2 Y0 (λR2 ) − λY1 (λR2 )
− k2 J0 (λR2 ) − λJ1 (λR2 ) k1 Y0 (λR1 ) + λY1 (λR1 ) = 0,
Solution:
Z lZ R2
w(r, z, t) = 2π ξf (ξ, η) G(r, z, ξ, η, t) dξ dη
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 h ∂ i
+ 2πa ξg3 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ
0 R1 ∂η η=0
Z t Z R2 h ∂ i
− 2πa ξg4 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ
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.
leads to the two-dimensional heat equation ∂τ u = a ∂ξξ u + ∂ηη u .
See also Niederer (1973) and Boyer (1974).
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.
See also Niederer (1973) and Boyer (1974).
∂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.
4.3. Other Equations 457
∂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.
Fundamental solution:
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).
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
458 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
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)
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.
An initial condition is prescribed:
w = f (x, y) at t = 0.
4.3. Other Equations 459
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
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
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
∂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,
∂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
∂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:
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)
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.
462 S ECOND -O RDER PARABOLIC E QUATIONS WITH T WO S PACE VARIABLES
∂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)
∂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
5.1.1 Problems in Cartesian Coordinates
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.
◮ Particular solutions:
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
Fundamental solution:
1 x2 + y 2 + z 2
E (x, y, z, t) = exp − .
8(πat)3/2 4at
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
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
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, 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 − .
8(πat)3/2 4at 4at 4at
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 −
8(πat)3/2 4at 4at 4at
√ 2 x+ξ √
−2k πat exp k at+k(x+ξ) erfc √ +k at .
2 at
Solution:
Z lZ ∞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 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
Solution:
Z lZ ∞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 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
468 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
The solution w(x, y, z, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where
∞
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 .
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
Solution:
Z lZ ∞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 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, l, t − τ ) dξ dη dτ,
0 −∞ −∞
∂w
5.1. Heat Equation ∂t
= 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
Solution:
Z lZ ∞Z ∞
w(x, y, z, t) = f (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
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 ∞
∂
+a g2 (ξ, η, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dτ
0 0 −∞ ∂ζ ζ=0
Z t Z ∞Z ∞
∂
−a g3 (ξ, η, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dτ,
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
× sin sin exp − .
l l l2
n=1
470 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
tan(µl) k2 + k3
with the µn being positive roots of the transcendental equation = 2 .
µ µ − k2 k3
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
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
× sin sin exp − .
n=1
l l l2
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
⊙ Literature: H. S. Carslaw and J. C. Jaeger (1984).
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, 0, η, ζ, t − τ ) dη dζ dτ
0 0 0
Z tZ ∞Z ∞
−a g2 (ξ, ζ, τ ) G(x, y, z, ξ, 0, ζ, t − τ ) dξ dζ dτ
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
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 t Z ∞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 ∞
−a g2 (ξ, ζ, τ ) G(x, y, z, ξ, 0, ζ, t − τ ) dξ dζ dτ
0 0 0
Z t Z ∞Z ∞
−a g3 (ξ, η, τ ) G(x, y, z, ξ, η, 0, t − τ ) dξ dη dτ,
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
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 t Z ∞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 ∞
∂
+a g2 (ξ, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ∂η η=0
Z t Z ∞Z ∞
−a g3 (ξ, η, τ ) G(x, y, z, ξ, η, 0, t − τ ) dξ dη dτ,
0 0 0
∂w
5.1. Heat Equation ∂t
= 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
Solution:
Z ∞ Z l2Z l1
w(x, y, z, t) = f (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
−∞ 0 0
Z t Z ∞ Z l2
∂
+a g1 (η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dη dζ dτ
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
∂
+a g3 (ξ, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dζ dτ
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
476 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
Solution:
Z ∞ Z l2Z l1
w(x, y, z, t) = f (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
−∞ 0 0
Z t Z ∞ Z l2
−a g1 (η, ζ, τ ) G(x, y, z, 0, η, ζ, t − τ ) dη dζ dτ
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
−a g3 (ξ, ζ, τ ) G(x, y, z, ξ, 0, ζ, t − τ ) dξ dζ dτ
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
The solution w(x, y, z, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) 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∞ ∞
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 ,
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
tan(µl1 ) k1 + k2 tan(λl2 ) k3 + k4
= 2 , = 2 .
µ µ − k1 k2 λ λ − k3 k4
Solution:
Z ∞ Z l2Z l1
w(x, y, z, t) = f (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
−∞ 0 0
Z tZ ∞ Z l2
∂
+a g1 (η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dη dζ dτ
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
−a g3 (ξ, ζ, τ ) G(x, y, z, ξ, 0, ζ, t − τ ) dξ dζ dτ
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
478 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
Solution:
Z ∞Z l2Z l1
w(x, y, z, t) = f (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
0
Z t Z0 0
∞Z l2
∂
+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 l2
∂
−a g2 (η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=l1
Z t Z ∞Z l1
∂
+a g3 (ξ, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dζ dτ
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
∂
+a g5 (ξ, η, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dτ,
0 0 0 ∂ζ ζ=0
where
Solution:
Z ∞Z l2Z l1
w(x, y, z, t) = f (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
0 0 0
Z tZ ∞Z l2
−a g1 (η, ζ, τ ) G(x, y, z, 0, η, ζ, t − τ ) dη dζ dτ
0 0 0
Z tZ ∞Z l2
+a g2 (η, ζ, τ ) G(x, y, z, l1 , η, ζ, t − τ ) dη dζ dτ
0 0 0
Z tZ ∞Z l1
−a g3 (ξ, ζ, τ ) G(x, y, z, ξ, 0, ζ, t − τ ) dξ dζ dτ
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
−a g5 (ξ, η, τ ) G(x, y, z, ξ, η, 0, t − τ ) dξ dη dτ,
0 0 0
where
Solution:
Z ∞Z l2Z l1
w(x, y, z, t) = f (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
0 0 0
Z tZ ∞Z l2
∂
+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 l2
∂
−a g2 (η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=l1
Z t Z ∞Z l1
∂
+a g3 (ξ, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dζ dτ
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
−a g5 (ξ, η, τ ) G(x, y, z, ξ, η, 0, t − τ ) dξ dη dτ,
0 0 0
where
Solution:
Z ∞Z l2Z l1
w(x, y, z, t) = f (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
0 0 0
Z tZ ∞Z l2
−a g1 (η, ζ, τ ) G(x, y, z, 0, η, ζ, t − τ ) dη dζ dτ
0 0 0
Z tZ ∞Z l2
+a g2 (η, ζ, τ ) G(x, y, z, l1 , η, ζ, t − τ ) dη dζ dτ
0 0 0
Z tZ ∞Z l1
−a g3 (ξ, ζ, τ ) G(x, y, z, ξ, 0, ζ, t − τ ) dξ dζ dτ
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 tZ l2Z l1
∂
+a g5 (ξ, η, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dτ,
0 0 0 ∂ζ ζ=0
where
Solution:
Z l3Z l2Z l1
w(x, y, z, t) = f (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
0 0 0
Z t Z l3Z l2
∂
+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 l3Z l2
∂
−a g2 (η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=l1
Z t Z l3Z l1
∂
+a g3 (ξ, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ∂η η=0
Z t Z l3Z 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
∂
+a 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
∂
−a g6 (ξ, η, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dτ,
0 0 0 ∂ζ ζ=l3
where
Solution:
Z l3Z l2Z l1
w(x, y, z, t) = f (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
0 0 0
Z t Z l3Z l2
−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
+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
−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
+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
−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
+a g6 (ξ, η, τ ) G(x, y, z, ξ, η, l3 , t − τ ) dξ dη dτ,
0 0 0
where
G(x, y, z, ξ, η, ζ, t) = G1 (x, ξ, t) G2 (y, η, t) G3 (z, ζ, t),
X∞
1 πnx πnξ π 2 n2 at
G1 (x, ξ, t) = 1+2 cos cos exp − ,
l1
n=1
l1 l1 l12
X∞
1 πny πnη π 2 n2 at
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
The solution w(x, y, z, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where
Here,
k1 k2 µ2n +k12
2 k1 l1 k12
ϕn(x) = cos(µnx)+ sin(µnx), kϕnk = 2 2 + + 1+ 2 ,
µn 2µn µn +k22 2µ2n 2 µn
k3 2 k4 λ2m +k32 k3 l2 k32
ψm(y) = cos(λmy)+ sin(λmy), kψmk = 2 2 + + 1+ 2 ,
λm 2λm λm +k42 2λ2m 2 λm
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
Solution:
Z l3Z l2Z l1
w(x, y, z, t) = f (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
0 0 0
Z t Z l3Z l2
∂
+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 l3Z l2
∂
−a g2 (η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=l1
Z t Z l3Z l1
∂
+a g3 (ξ, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ∂η η=0
Z t Z l3Z 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
−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
+a g6 (ξ, η, τ ) G(x, y, z, ξ, η, l3 , t − τ ) dξ dη dτ,
0 0 0
where
Solution:
Z l3Z l2Z l1
w(x, y, z, t) = f (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
0 0 0
Z t Z l3Z l2
∂
+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 l3Z l2
∂
−a g2 (η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=l1
Z t Z l3Z l2
−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
+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
−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
+a g6 (ξ, η, τ ) G(x, y, z, ξ, η, l3 , t − τ ) dξ dη dτ,
0 0 0
where
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.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
∂w
5.1. Heat Equation ∂t
= a∆3 w 489
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.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
Here, the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation
µJn′ (µR) + kJn (µR) = 0.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
490 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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 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,
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.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
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 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
5.1. Heat Equation ∂t
= a∆3 w 491
Here,
where the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation Jn′ (µR) = 0.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
The solution w(r, ϕ, z, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where
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
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 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 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 (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, . . . ,
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.
2◦ . A semiinfinite circular cylinder is considered. The following conditions are prescribed:
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 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, ζ, t − τ ) dξ dη dτ.
0 0 0 ∂ζ ζ=0
Here,
where the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation Jn′ (µR) = 0.
∂w
5.1. Heat Equation ∂t
= a∆3 w 493
A circular cylinder of finite length is considered. The following conditions are prescribed:
Solution:
Z lZ 2πZ R
w(r, ϕ, z, t) = ξf (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t) dξ dη dζ
0 0 0
Z t Z l 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
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,
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.
A circular cylinder of finite length is considered. The following conditions are prescribed:
Solution:
Z lZ 2πZ R
w(r, ϕ, z, t) = ξf (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t) dξ dη dζ
0 0 0
Z t Z l Z 2π
+ aR g1 (η, ζ, τ ) G(r, ϕ, z, R, η, ζ, t − τ ) dη dζ dτ
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,
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 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, . . . ,
where the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation Jn′ (µR) = 0.
The solution w(r, ϕ, z, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where
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
5.1. Heat Equation ∂t
= a∆3 w 495
1◦ . A circular cylinder of finite length is considered. The following conditions are pre-
scribed:
Solution:
Z lZ 2πZ R
w(r, ϕ, z, t) = ξf (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t) dξ dη dζ
0 0 0
Z t Z l 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 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,
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.
2◦ . A circular cylinder of finite length is considered. The following conditions are pre-
scribed:
Solution:
Z lZ 2πZ R
w(r, ϕ, z, t) = ξf (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t) dξ dη dζ
0 0 0
Z t Z l 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, 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,
where the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation Jn′ (µR) = 0.
An infinite hollow circular cylinder is considered. The following conditions are prescribed:
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
5.1. Heat Equation ∂t
= 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 )
Zn (µnm r) = Jn (µnm R1 )Yn (µnm r) − Yn (µnm R1 )Jn (µnm r),
where the Jn (r) and Yn (r) are Bessel functions and the µnm are positive roots of the
transcendental equation
Jn (µR1 )Yn (µR2 ) − Yn (µR1 )Jn (µR2 ) = 0.
The solution w(r, ϕ, z, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where
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
µnm are positive roots of the transcendental equation
µJn′ (µR1 ) − k1 Jn (µR1 ) µYn′ (µR2 ) + k2 Yn (µR2 )
= µYn′ (µR1 ) − k1 Yn (µR1 ) µJn′ (µR2 ) + k2 Jn (µR2 ) .
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
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 )
Zn (µnm r) = Jn (µnm R1 )Yn (µnm r) − Yn (µnm R1 )Jn (µnm r),
where the Jn (r) and Yn (r) are Bessel functions and the µnm are positive roots of the
transcendental equation
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),
500 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
A semiinfinite hollow circular cylinder is considered. The following conditions are pre-
scribed:
The solution w(r, ϕ, z, t) is determined by the formula from the previous paragraph (for
the second boundary value problem) in which
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
µJn′ (µR1 ) − k1 Jn (µR1 ) µYn′ (µR2 ) + k2 Yn (µR2 )
= µYn′ (µR1 ) − k1 Yn (µR1 ) µJn′ (µR2 ) + k2 Jn (µR2 ) .
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 0 R1
Z t Z ∞Z 2π
∂
+ aR1 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 ∞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, 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
∞ ∞
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 )
Zn (µnm r) = Jn (µnm R1 )Yn (µnm r) − Yn (µnm R1 )Jn (µnm r),
where the Jn (r) and Yn (r) are Bessel functions and the µnm are positive roots of the
transcendental equation
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 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, ζ, t − τ ) ξ dξ dη dτ.
0 0 R1 ∂ζ ζ=0
502 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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 (the
prime denotes a derivative with respect to the argument); and the µnm are positive roots of
the transcendental equation
A circular cylinder of finite length is considered. The following conditions are prescribed:
Solution:
Z lZ 2πZ R2
w(r, ϕ, z, t) = f (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 R1
Z t Z l Z 2π
∂
+ aR1 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
∂
− 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
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
5.1. Heat Equation ∂t
= 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
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 )
Zn (µnm r) = Jn (µnm R1 )Yn (µnm r) − Yn (µnm R1 )Jn (µnm r),
where the Jn (r) and Yn (r) are Bessel functions and the µnm are positive roots of the
transcendental equation
A circular cylinder of finite length is considered. The following conditions are prescribed:
Solution:
Z lZ 2πZ R2
w(r, ϕ, z, t) = f (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 R1
Z tZ lZ 2π
− aR1 g1 (η, ζ, τ ) G(r, ϕ, z, R1 , η, ζ, t − τ ) dη dζ dτ
0 0 0
Z tZ lZ 2π
+ aR2 g2 (η, ζ, τ ) G(r, ϕ, z, R2 , η, ζ, t − τ ) dη dζ dτ
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
504 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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)
∞ ∞
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 (the
prime denotes a derivative with respect to the argument); and the µnm are positive roots of
the transcendental equation
Jn′ (µR1 )Yn′ (µR2 ) − Yn′ (µR1 )Jn′ (µR2 ) = 0.
tan(λl) k3 + k4
where the λs are positive roots of the transcendental equation = 2 .
λ λ − k3 k4
1◦ . A circular cylinder of finite length is considered. The following conditions are pre-
scribed:
w = f (r, ϕ, z) at t = 0 (initial condition),
w = g1 (ϕ, z, t) at r = R1 (boundary condition),
w = g2 (ϕ, z, t) at r = R2 (boundary condition),
∂z w = g3 (r, ϕ, t) at z = 0 (boundary condition),
∂z w = g4 (r, ϕ, t) at z = l (boundary condition).
Solution:
Z lZ 2πZ R2
w(r, ϕ, z, t) = f (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 R1
Z tZ l Z 2π
∂
+ aR1 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
∂
− 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, 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,
∞ 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),
where the Jn (r) and Yn (r) are Bessel functions and the µnm are positive roots of the
transcendental equation
2◦ . A circular cylinder of finite length is considered. The following conditions are pre-
scribed:
w = f (r, ϕ, z) at t = 0 (initial condition),
∂r w = g1 (ϕ, z, t) at r = R1 (boundary condition),
∂r w = g2 (ϕ, z, t) at r = R2 (boundary condition),
w = g3 (r, ϕ, t) at z = 0 (boundary condition),
w = g4 (r, ϕ, t) at z = l (boundary condition).
Solution:
Z lZ 2πZ R2
w(r, ϕ, z, t) = f (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 R1
Z t Z l Z 2π
− aR1 g1 (η, ζ, τ ) G(r, ϕ, z, R1 , η, ζ, t − τ ) dη dζ dτ
0 0 0
Z tZ l 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, 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
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)
∞ ∞
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 (the
prime denotes a derivative with respect to the argument); and the µnm are positive roots of
the transcendental equation
Solution:
Z ∞ Z ϕ0Z ∞
w(r, ϕ, z, t) = f (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
−∞ 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
G(r, ϕ, z, ξ, η, ζ, t) = √ exp − G1 (r, ϕ, ξ, η, t),
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
Solution:
Z ∞ Z ϕ0Z ∞
w(r, ϕ, z, t) = f (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
−∞ 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
G(r, ϕ, z, ξ, η, ζ, t) = √ exp − G1 (r, ϕ, ξ, η, t),
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
+ Inπ/ϕ0 cos cos ,
n=1
2at ϕ0 ϕ0
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 ∞
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
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 2
1 r +ξ 1 rξ
G1(r, ϕ, ξ, η, t) = exp − I0
aϕ0t 4at 2 2at
X∞
rξ nπϕ nπη
+ Inπ/ϕ0 cos cos ,
n=1
2at ϕ 0 ϕ 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 ∞
1 r + ξ2 X rξ nπϕ nπη
G1 (r, ϕ, ξ, η, t) = exp − Inπ/ϕ0 sin sin ,
aϕ0 t 4at 2at ϕ0 ϕ0
n=1
Solution:
Z ∞Z ϕ0Z ∞
w(r, ϕ, z, t) = f (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 0
Z tZ ∞Z ∞
−a g1 (ξ, ζ, τ ) G(r, ϕ, z, ξ, 0, ζ, t − τ ) dξ dζ dτ
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 t Z ϕ0Z ∞
∂
+a g3 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) ξ dξ dη dτ.
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 2
1 r +ξ 1 rξ
G1(r, ϕ, ξ, η, t) = exp − I0
aϕ0t 4at 2 2at
X∞
rξ nπϕ nπη
+ Inπ/ϕ0 cos cos ,
n=1
2at ϕ 0 ϕ 0
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 ∞
1 r + ξ2 X rξ nπϕ nπη
G1 (r, ϕ, ξ, η, t) = exp − Inπ/ϕ0 sin sin ,
aϕ0 t 4at 2at ϕ0 ϕ0
n=1
Solution:
Z lZ ϕ0Z ∞
w(r, ϕ, z, t) = f (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 0
Z tZ l Z ∞
−a g1 (ξ, ζ, τ ) G(r, ϕ, z, ξ, 0, ζ, t − τ ) dξ dζ dτ
0 0 0
Z tZ l Z ∞
+a g2 (ξ, ζ, τ ) G(r, ϕ, z, ξ, ϕ0 , ζ, t − τ ) dξ dζ dτ
0 0 0
Z tZ ϕ0Z ∞
−a g3 (ξ, η, τ ) G(r, ϕ, z, ξ, η, 0, t − τ )ξ dξ dη dτ
0 0 0
Z tZ ϕ0Z ∞
+a g4 (ξ, η, τ ) G(r, ϕ, z, ξ, η, l, t − τ )ξ dξ dη dτ.
0 0 0
Here,
1◦ . A wedge domain of finite thickness is considered. The following conditions are pre-
scribed:
w = f (r, ϕ, z) at t = 0 (initial condition),
w = g1 (r, z, t) at ϕ = 0 (boundary condition),
w = g2 (r, z, t) at ϕ = ϕ0 (boundary condition),
∂z w = g3 (r, ϕ, t) at z = 0 (boundary condition),
∂z w = g4 (r, ϕ, t) at z = l (boundary condition).
Solution:
Z lZ ϕ0Z ∞
w(r, ϕ, z, t) = f (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 0
Z tZ lZ ∞
1 ∂
+a g1 (ξ, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ξ ∂η η=0
Z tZ lZ ∞
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 ∞
−a g3 (ξ, η, τ ) G(r, ϕ, z, ξ, η, 0, t − τ )ξ dξ dη dτ
0 0 0
Z t Z ϕ0Z ∞
+a g4 (ξ, η, τ ) G(r, ϕ, z, ξ, η, l, t − τ )ξ dξ dη dτ.
0 0 0
Here,
2◦ . A wedge domain of finite thickness is considered. The following conditions are pre-
scribed:
w = f (r, ϕ, z) at t = 0 (initial condition),
r −1 ∂ϕ w = g1 (r, z, t) at ϕ = 0 (boundary condition),
−1
r ∂ϕ w = g2 (r, z, t) at ϕ = ϕ0 (boundary condition),
w = g3 (r, ϕ, t) at z = 0 (boundary condition),
w = g4 (r, ϕ, t) at z = l (boundary condition).
∂w
5.1. Heat Equation ∂t
= a∆3 w 513
Solution:
Z lZ ϕ0Z ∞
w(r, ϕ, z, t) = f (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 0
Z tZ l Z ∞
−a g1 (ξ, ζ, τ ) G(r, ϕ, z, ξ, 0, ζ, t − τ ) dξ dζ dτ
0 0 0
Z tZ l Z ∞
+a g2 (ξ, ζ, τ ) G(r, ϕ, z, ξ, ϕ0 , ζ, t − τ ) dξ dζ dτ
0 0 0
Z t Z ϕ0Z ∞
∂
+a g3 (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 ∞
∂
−a g4 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) ξ dξ dη dτ.
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
+ Inπ/ϕ0 cos cos ,
2at ϕ0 ϕ0
n=1
Solution:
Z ∞ Z ϕ0Z R
w(r, ϕ, z, t) = f (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
−∞ 0 0
Z t Z ∞ Z ϕ0
∂
− aR g1 (η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 −∞ 0 ∂ξ ξ=R
Z tZ ∞ Z R
1 ∂
+a g2 (ξ, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dζ dτ
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
514 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
Here,
1 (z − ζ)2
G(r, ϕ, z, ξ, η, ζ, t) = √ exp − G1 (r, ϕ, ξ, η, t),
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-
dental equation Jnπ/ϕ0 (µR) = 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
∞ ∞
4 X X Jnπ/ϕ0(µnmr)Jnπ/ϕ0(µnmξ)
G1(r, ϕ, ξ, η, t) = 2 ′
R ϕ0 n=1m=1 [Jnπ/ϕ 0
(µnmR)]2
nπϕ nπη
×sin sin exp(−µ2nmat),
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.
∂w
5.1. Heat Equation ∂t
= a∆3 w 515
Solution:
Z lZ ϕ0Z R
w(r, ϕ, z, t) = f (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 0
Z t Z l Z ϕ0
∂
− aR g1 (η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=R
Z tZ l Z R
1 ∂
+a g2 (ξ, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ξ ∂η η=0
Z tZ l Z R
1 ∂
−a g3 (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 R
∂
+a 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
∂
−a g5 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) ξ dξ dη dτ.
0 0 0 ∂ζ ζ=l
Here,
where the Jnπ/ϕ0 (r) are Bessel functions and the µnm are positive roots of the transcen-
dental equation Jnπ/ϕ0 (µR) = 0.
A cylindrical sector of finite thickness is considered. The following conditions are pre-
scribed:
w = f (r, ϕ, z) at t = 0 (initial condition),
w = g1 (ϕ, z, t) at r = R (boundary condition),
w = g2 (r, z, t) at ϕ = 0 (boundary condition),
w = g3 (r, z, t) at ϕ = ϕ0 (boundary condition),
∂z w = g4 (r, ϕ, t) at z = 0 (boundary condition),
∂z w = g5 (r, ϕ, t) at z = l (boundary condition).
∂w
5.1. Heat Equation ∂t
= a∆3 w 517
Solution:
Z lZ ϕ0Z R
w(r, ϕ, z, t) = f (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 0
Z t Z l Z ϕ0
∂
− aR g1 (η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=R
Z tZ l Z R
1 ∂
+a g2 (ξ, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ξ ∂η η=0
Z tZ l Z R
1 ∂
−a g3 (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 R
−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
+a g5 (ξ, η, τ ) G(r, ϕ, z, ξ, η, l, t − τ )ξ dξ dη dτ.
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 ′
R ϕ0 n=1m=1 [Jnπ/ϕ 0
(µnmR)]2
nπϕ nπη
×sin sin exp(−µ2nmat),
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.
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.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980), H. S. Carslaw and J. C. Jaeger
(1984).
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
5.1. Heat Equation ∂t
= 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.
The solution w(r, θ, ϕ, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where
∞ 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.
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
520 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
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
Zn+1/2 (λR2 ) = 0.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
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, 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
5.1. Heat Equation ∂t
= 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).
The solution w(r, θ, ϕ, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where
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
522 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
The integrals that determine the coefficients Bnms can be expressed in terms of the
Bessel functions and their derivatives.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
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
∂w
5.2 Heat Equation with Source = a∆3 w + Φ(x, y, z, t)
∂t
5.2.1 Problems in Cartesian Coordinates
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 ∞
w(x, y, z, t) = f (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
−∞ −∞ −∞
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
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
Z tZ ∞ Z ∞ Z ∞
+ Φ(ξ, η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dζ dτ,
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 − .
8(πat)3/2 4at 4at 4at
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
taken from Section 5.1.1.
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
taken from Section 5.1.1.
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
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
taken from Section 5.1.1.
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
taken from Section 5.1.1.
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.
⊙ Literature: A. G. Butkovskiy (1979), H. S. Carslaw and J. C. Jaeger (1984).
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
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.
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
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
is also taken from Section 5.1.1.
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
is also taken from Section 5.1.1.
∂w
5.2. Heat Equation with Source ∂t
= a∆3 w + Φ(x, y, z, t) 527
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
Φ(ξ, η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dζ dτ, (5)
0 0 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.
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
function is also taken from Section 5.1.1.
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
is also taken from Section 5.1.1.
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.
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
Φ(ξ, η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dζ dτ, (6)
0 0 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.
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
is also taken from Section 5.1.1.
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.
528 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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.
⊙ Literature: A. G. Butkovskiy (1979), H. S. Carslaw and J. C. Jaeger (1984).
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
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 (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.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
∂w
5.2. Heat Equation with Source ∂t
= a∆3 w + Φ(x, y, z, t) 529
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, 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,
where the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation Jn′ (µR) = 0.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
The solution w(r, ϕ, z, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where
(
1 (z −ζ)2 1 for n = 0,
G2(z, ζ, t) = √ exp − , An =
2 πat 4at 2 for n = 1, 2, . . .
530 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
Here, the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation
µJn′ (µR) + kJn (µR) = 0.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
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
taken from Section 5.1.2.
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
taken from Section 5.1.2.
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
function is also taken from Section 5.1.2.
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
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
function is also taken from Section 5.1.2.
∂w
5.2. Heat Equation with Source ∂t
= 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
function is also taken from Section 5.1.2.
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
Green’s function is also taken from Section 5.1.2.
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
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
function is also taken from Section 5.1.2.
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
function is also taken from Section 5.1.2.
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
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
function is also taken from Section 5.1.2.
532 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
function is also taken from Section 5.1.2.
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
Green’s function is also taken from Section 5.1.2.
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
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
function is also taken from Section 5.1.2.
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
function is also taken from Section 5.1.2.
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);
the Green’s function is also taken from Section 5.1.2.
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
∂w
5.2. Heat Equation with Source ∂t
= 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
function is also taken from Section 5.1.2.
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
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
function is also taken from Section 5.1.2.
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.
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
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
function is also taken from Section 5.1.2.
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.
534 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
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
Φ(ξ, η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ )ξ dξ dη dζ dτ, (9)
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
function is also taken from Section 5.1.2.
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
Φ(ξ, η, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ )ξ dξ dη dζ dτ, (10)
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
function is also taken from Section 5.1.2.
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.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980), H. S. Carslaw and J. C. Jaeger
(1984).
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 .
2 for k 6= 0, (n + k)! R2 λ2nm − n(n + 1) Jn+1/2 (λnm R)
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.
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
taken from Section 5.1.3.
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
taken from Section 5.1.3.
∂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.
Fundamental solution:
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
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 −∞ −∞ −∞
5.3. Other Equations with Three Space Variables 539
where
∂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
with in Section 5.1.1.
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
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).
5.3. Other Equations with Three Space Variables 541
∂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.
An initial condition is prescribed:
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
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 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
542 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
∂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
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).
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.
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
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
544 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
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).
The solution of the third boundary value problem is given by formulas (5) and (2),
where the λn and un (r) 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(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).
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.
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 )
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
where the following notation is used:
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
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
548 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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 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.
Fundamental solution:
i m n/2 m 2 πn
E (x, t) = − exp i |x| − i , |x|2 = x21 + · · · + x2n .
~ 2π~t 2~t 4
⊙ Literature: V. S. Vladimirov, V. P. Mikhailov, A. A. Vasharin et al. (1974).
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.
An initial condition is prescribed:
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 .
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)
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)
The Green’s function can be represented as the product (1) of the corresponding one-
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
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
where the functions fk (t), g(t), hk (t), and rk (t) are given by the following expressions
552 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
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:
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}.
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
Gk = δ(xk − yk ) at t = τ, (3)
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:
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)
The Green’s function can be represented as the product (1) of the corresponding one-
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 = f (x) at t=0 (initial condition),
w = g(x, t) for x ∈ S (boundary condition).
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
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
5.4. Equations with n Space Variables 555
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.
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 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 = f (x) at t = 0 (initial condition),
∂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
w(x, t) = Φ(y, τ ) G(x, y, t − τ ) dVy dτ + f (y) G(x, y, t) dVy
0 V V
Z tZ
+ g(y, τ ) G(x, y, t − τ ) dSy dτ. (6)
0 S
556 S ECOND -O RDER PARABOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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:
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
557
558 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
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).
w = f (x) at t = 0,
∂t w = g(x) at t = 0.
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
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).
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:
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).
w = A cos(ωt + γ) at x = 0, w → 0 as x → 0.
Solution:
w = Ae−λx cos(ωt − βx + γ),
where √ 1/2 √ 2
ω 2 + b2 − b ω + b2 + b 1/2
λ= , β= .
2a 2a
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
6.1. Constant Coefficient Equations 561
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
The initial velocities of the string points are zero, g(x) = 0.
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:
w = f (x) at t = 0 (initial condition),
∂t w = g(x) at t = 0 (initial condition),
∂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).
562 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
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:
x − at = 0, x + at = 0, x − at = 2c, x + at = 2b.
⊙ Literature: A. V. Bitsadze and D. F. Kalinichenko (1985).
∂2w ∂2w
6.1.2 Equations of the Form = a2 + Φ(x, t)
∂t2 ∂x2
◮ Domain: −∞ < x < ∞. Cauchy problem.
Initial conditions are prescribed:
w = f (x) at t = 0,
∂t w = g(x) at t = 0.
564 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
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 )
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 )
Solution:
1
w(x, t) = w1 (x, t) + w2 (x, t),
2a
6.1. Constant Coefficient Equations 565
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 )
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
566 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
where
t
∞
2 X1 nπx nπξ nπat
G(x, ξ, t) = + cos cos sin .
l 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 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.
w1 = Aw(x + C1 , ±t + C2 ),
w2 = Aw(−x + C1 , ±t + C2 ),
x − vt t − va−2 x
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)
√ √ √
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)
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).
6.1. Constant Coefficient Equations 569
w = f (x) at t = 0,
∂t w = g(x) at t = 0.
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.
⊙ Literature: B. M. Budak, A. N. Tikhonov, and A. A. Samarskii (1980).
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
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
570 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
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.
⊙ Literature: A. G. Butkovskiy (1979).
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
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
The solution w(x, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞ 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
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
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)
G(x, ξ, t) = sin(λn x) sin(λn ξ) p , λn = .
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
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
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
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
⊙ Literature: A. G. Butkovskiy (1979).
6.1. Constant Coefficient Equations 573
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
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
⊙ Literature: A. G. Butkovskiy (1979).
The solution w(x, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞
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
574 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂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
which is discussed in Section 6.1.3.
where J0 (z) and J1 (z) are Bessel functions of the first kind.
⊙ Literature: A. N. Tikhonov and A. A. Samarskii (1990).
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 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.
⊙ Literature: A. G. Butkovskiy (1979).
576 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
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.
The solution w(x, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞ √
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
where the µn are positive roots of the transcendental equation
tan(µl) 4a4 (k1 + k2 )
= 4 2 .
µ 4a µ − (2a2 k1 + b)(2a2 k2 − b)
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).
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
578 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
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
problem for 0 ≤ r ≤ R).
⊙ Literature: M. M. Smirnov (1975), B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
The solution w(r, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞
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
The numerical values of the first six roots λn can be found in Carslaw and Jaeger (1984);
see also Abramowitz and Stegun (1964).
6.2. Wave Equations with Axial or Central Symmetry 579
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
∂
w(r, t) = 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
580 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
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
transcendental equation
The numerical values of the first five roots λn = λn (s) can be found in Abramowitz and
Stegun (1964).
The solution w(r, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞
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
transcendental equation
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
6.2.2 Equation of the Form =a + + Φ(r, t)
∂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
6.2. Wave Equations with Axial or Central Symmetry 581
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)
with the additional term (1); the Green’s function is also taken from Section 6.2.1.
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.
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
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
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 )
with the additional term (2); the Green’s function is also taken from Section 6.2.1.
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:
∂2u 2
2∂ u
= a ,
∂t2 ∂r 2
which is discussed in Section 6.1.1.
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:
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
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
6.2. Wave Equations with Axial or Central Symmetry 583
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
problem for 0 ≤ r ≤ R).
µ cot µ + kR − 1 = 0.
The numerical values of the first six roots µn can be found in Carslaw and Jaeger (1984).
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
G(r, ξ, t) = sin sin sin .
πar n R2 − R1 R2 − R1 R2 − R1
n=1
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, 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 )].
The solution w(r, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
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
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 )
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)
with the additional term (1); the Green’s function is also taken from Section 6.2.3.
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)
and expression (1); the Green’s function is also taken from Section 6.2.3.
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
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 )
and expression (2); the Green’s function is also taken from Section 6.2.3.
∂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
∂
w(r, t) = f0 (ξ)G(r, ξ, t) dξ + f1 (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
6.2. Wave Equations with Axial or Central Symmetry 587
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
of the first ten µn are specified in Section 3.2.1 (see the first boundary value problem for
0 ≤ r ≤ R).
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
where the µn are positive zeros of the first-order Bessel function, J1 (µ) = 0. The numerical
values of the first ten µn are specified in Section 3.2.1 (see the second boundary value
problem for 0 ≤ r ≤ R).
The solution w(r, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞ √
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
588 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
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.
The numerical values of the first five roots µn = µn (s) 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, 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
transcendental equation
The numerical values of the first five roots µn = µn (s) can be found in Abramowitz and
Stegun (1964).
∂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.
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
The solution w(r, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞ √
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
592 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
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)],
where the λn are positive roots of the transcendental equation
(λ2 R1 R2 + 1) tan[λ(R2 − R1 )] − λ(R2 − R1 ) = 0.
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
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).
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
problem for 0 ≤ r ≤ R).
4◦ . Domain: 0 ≤ x ≤ l. Third boundary value problem.
The following conditions are prescribed:
The numerical values of the first six roots µn can be found in Carslaw and Jaeger (1984).
∂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).
6.3. Equations Containing Power Functions and Arbitrary Parameters 595
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, 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
where 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).
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
596 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
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
where 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 ≤ x ≤ l. Third boundary value problem.
The following conditions are prescribed:
The numerical values of the first six roots µn can be found in Abramowitz and Stegun
(1964) and Carslaw and Jaeger (1984).
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
∂2w ∂2w ∂w
5. = (ax + b) +a + cw + Φ(x, t).
∂t2 ∂x2 ∂x
∂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
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
′′ +(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,
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
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
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.
The following conditions are prescribed:
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
6.3. Equations Containing Power Functions and Arbitrary Parameters 599
Solution:
Z tZ a
w(x, t) = Φ(ξ, τ )G(x, ξ, t − τ ) dξ dτ
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
⊙ Literature: A. G. Butkovskiy (1979).
∂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.
The following conditions are prescribed:
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
⊙ Literature: M. M. Smirnov (1975).
∂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
leads to an equation of the form 6.5.3.1:
∂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.
6.3. Equations Containing Power Functions and Arbitrary Parameters 601
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.
⊙ Literature: M. M. Smirnov (1975).
∂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
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.
The following conditions are prescribed:
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
⊙ Literature: A. G. Butkovskiy (1979).
∂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.
6.3. Equations Containing Power Functions and Arbitrary Parameters 603
∂2w ∂ ∂w
9. = a2 xm + Φ(x, t).
∂t2 ∂x ∂x
1◦ .Domain: 0 ≤ x ≤ l. First boundary value problem.
The following conditions are prescribed:
1.1. Case 0 < m < 1:
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 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 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
∂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
leads to a similar equation
∂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.
6.3. Equations Containing Power Functions and Arbitrary Parameters 605
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
1◦ . Domain: −∞ < x < ∞. Cauchy problem.
Initial conditions are prescribed:
w = f (x) at t = 0,
∂t w = g(x) at t = 0.
606 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
m−2
∂2w ∂2w ∂w
12. = tm + bt 2 , m ≥ 2.
∂t2 ∂x2 ∂x
Domain: −∞ < x < ∞. Cauchy problem.
Initial conditions are prescribed:
w = f (x) at t = 0,
∂t w = g(x) at t = 0.
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.
∂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
which is considered in Section 6.1.3.
2◦ . Fundamental solution:
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.
⊙ Literature: V. S. Vladimirov, V. P. Mikhailov, A. A. Vasharin et al. (1974).
608 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
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 )
where I0 (z) and I1 (z) are modified Bessel functions of the first kind.
4◦ . Domain: 0 ≤ x ≤ l. First boundary value problem.
The following conditions are prescribed:
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
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
6.4. Equations Containing the First Time Derivative 609
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
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:
w = A cos(ωt + γ) at x = 0, w→0 as x → ∞.
Solution:
w = Ae−λx cos(ωt − βx + γ),
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
which is considered in Section 6.1.3.
2◦ . Fundamental solutions:
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.
610 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
w = f (x) at t = 0,
∂t w = g(x) at t = 0.
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
6.4. Equations Containing the First Time Derivative 611
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
⊙ Literature: A. G. Butkovskiy (1979).
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, 0, t − τ ) dτ + a g2 (τ )G(x, l, t − τ ) dτ.
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
6.4. Equations Containing the First Time Derivative 615
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
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
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
∂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
where 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 value problem for
0 ≤ r ≤ R).
3◦ . Domain: 0 ≤ r ≤ R. Second boundary value problem.
The following conditions are prescribed:
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
Here, λn = a2 µ2n R−2 +b− 14 k2 and the µn are positive roots of the transcendental equation
The numerical values of the first six roots µn can be found in Abramowitz and Stegun
(1964) and Carslaw and Jaeger (1984).
∂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
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, 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
2◦ . Domain: 0 ≤ x ≤ l. First boundary value problem.
The following conditions are prescribed:
w = f0 (x) at t = 0 (initial condition),
∂t w = f1 (x) at t = 0 (initial condition),
w = g(t) at x = l (boundary condition),
w 6= ∞ at x = 0 (boundedness condition).
620 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
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
value problem for 0 ≤ r ≤ R).
3◦ . Domain: 0 ≤ x ≤ l. Second boundary value problem.
The following conditions are prescribed:
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
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.
The numerical values of the first six roots µn can be found in Abramowitz and Stegun
(1964) and Carslaw and Jaeger (1984).
∂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.
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
∂2w 2a ∂w ∂2w
2. + = − b2 w.
∂t2 t ∂t ∂x2
Domain: −∞ < x < ∞. Cauchy problem.
Initial conditions are prescribed:
w = f (x) at t = 0,
2a
t ∂t w = g(x) at t = 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
Domain: −∞ < x < ∞. Cauchy problem.
Initial conditions are prescribed:
w = f (x) at t = 0,
t2a ∂t w = g(x) at t = 0.
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
∂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)
∂2w ∂w 2
2∂ w ∂w
2
+ (k − 1) = a 2
+ (b − a2 ) + cw.
∂τ ∂τ ∂ξ ∂ξ
6.5. Equations Containing Arbitrary Functions 623
∂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.
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
which is discussed in Section 6.1.5.
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 .
624 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
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
3◦ . If the conditions
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.
6.5. Equations Containing Arbitrary Functions 625
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.
w = g1 (t) at x = x1 ,
w = g2 (t) at x = x2
∂x w = g1 (t) at x = x1 ,
∂x w = g2 (t) at x = x2
w = g1 (t) at x = x1 ,
∂x w = g2 (t) at x = x2
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
is given by relations (3) and (4) with
∂
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).
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
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
6.5. Equations Containing Arbitrary Functions 627
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:
The functions Un = Un (t, τ ) are determined by solving the Cauchy problem for the linear
ordinary differential equation
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.
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
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
628 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
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).
ψn (x) = An x + Bn ,
Z x
ψk (ξ)
ψk−1 (x) = Ak x + Bk + 2k(2k + 1) (x − ξ) dξ,
a f (ξ)
where C1 , C2 , and λ are arbitrary constants, and the function H = H(x) is determined by
the ordinary differential equation [f (x)Hx′ ]′x − λ2 H = 0.
3◦ . Separable particular solution:
where C1 , C2 , and λ are arbitrary constants, and the function Z = Z(x) is determined by
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
630 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH O NE S PACE VARIABLE
∂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)
6.5. Equations Containing Arbitrary Functions 631
∂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
Sections 15.1.1 and 15.1.3.
∂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
with Two Space Variables
∂ 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
1◦ . Particular solutions:
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 ,
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
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.
⊙ Literature: A. V. Bitsadze and D. F. Kalinichenko (1985).
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
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).
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 ∂
+a g1 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=0
Z t Z l2
2 ∂
−a g2 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=l1
Z t Z l1
2 ∂
+a g3 (ξ, τ ) G(x, y, ξ, η, t − τ ) dξ dτ
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).
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
− 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.
The solution w(x, y, t) is determined by the formula in the previous paragraph (for the
638 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T WO S PACE VARIABLES
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
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 t Z l2
2 ∂
+a g1 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=0
Z t Z l2
2 ∂
−a g2 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
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
+ a2 g4 (ξ, τ )G(x, y, ξ, l2 , t − τ ) dξ dτ,
0 0
∂2w
7.1. Wave Equation ∂t2
= 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.
2◦ . A rectangle is considered. The following conditions are prescribed:
w = f0 (x, y) at t = 0 (initial condition),
∂t w = f1 (x, y) at t = 0 (initial condition),
w = g1 (y, t) at x = 0 (boundary condition),
∂x w = g2 (y, t) at x = l1 (boundary condition),
w = g3 (x, t) at y = 0 (boundary condition),
∂y w = g4 (x, t) at y = l2 (boundary 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 ∂
+a g1 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
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 ∂
+a g3 (ξ, τ ) G(x, y, ξ, η, t − τ ) dξ dτ
0 0 ∂η η=0
Z tZ l1
+ a2 g4 (ξ, τ )G(x, y, ξ, l2 , t − τ ) dξ dτ,
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
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, . . .),
where the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation Jn′ (µR) = 0.
⊙ Literature: A. G. Butkovskiy (1979), B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
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 ,
1 for 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 the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the
transcendental equation
Solution:
Z 2πZ R2
∂
w(r, ϕ, t) = f0 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
∂t 0 R1
Z 2πZ R2
+ f1 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
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
the µnm are positive roots of the transcendental equation
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
where the Jnπ/ϕ0 (r) are Bessel functions and the µnm are positive roots of the transcen-
dental equation Jnπ/ϕ0 (µR) = 0.
Solution:
Z ϕ0Z R
∂
w(r, ϕ, t) = f0 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
∂t 0 0
Z ϕ0Z R
+ f1 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
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
where the Jnπ/ϕ0 (r) are Bessel functions and the µnm are positive roots of the transcen-
′
dental equation Jnπ/ϕ (µR) = 0.
0
∂2w
7.1. Wave Equation ∂t2
= a 2 ∆2 w 645
Solution:
Z ϕ0Z R
∂
w(r, ϕ, t) = f0 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
∂t 0 0
Z ϕ0Z R
+ f1 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
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
µJs′ n (µR) + kJsn (µR) = 0.
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.
646 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T WO S PACE VARIABLES
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
√
mπz mπη sin at λnm
× 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
values of the first ten roots µn are specified in Section 3.2.1 (see the second boundary value
problem for 0 ≤ r ≤ R).
The solution w(r, z, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞ ∞ √
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
648 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T WO S PACE VARIABLES
tan(βl) k2 + k3
µJ1 (µ) − k1 RJ0 (µ) = 0, = 2 .
β β − k2 k3
1◦ . A circular cylinder of finite length is considered. The following conditions are pre-
scribed:
w = f0 (r, z) at t = 0 (initial condition),
∂t w = f1 (r, z) at t = 0 (initial condition),
w = g1 (z, t) at r = R (boundary condition),
∂z w = g2 (r, t) at z = 0 (boundary condition),
∂z w = g3 (r, t) at z = l (boundary condition).
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
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
√
mπz mπη sin at λnm
× cos cos √ ,
l l a λnm
(
µ2n π 2 m2 1 for m = 0,
λnm = 2 + 2
, Am =
R l 2 for m > 0,
where the µn are positive zeros of the Bessel function, J0 (µ) = 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).
2◦ . A circular cylinder of finite length is considered. The following conditions are pre-
∂2w
7.1. Wave Equation ∂t2
= a 2 ∆2 w 649
scribed:
w = f0 (r, z) at t = 0 (initial condition),
∂t w = f1 (r, z) at t = 0 (initial condition),
∂r w = g1 (z, t) at r = R (boundary condition),
w = g2 (r, t) at z = 0 (boundary condition),
w = g3 (r, t) at z = l (boundary condition).
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
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,
∞ ∞
4ξ X X 1 µn r µn ξ
G(r, z, ξ, η, t) = 2 2 (µ ) J0 J0
R l J0 n R R
n=0 m=1
√
mπz mπη sin at λnm
× sin sin √ ,
l l a λnm
µ2n π 2 m2
λnm = + ,
R2 l2
where the µn are zeros of the first-order Bessel function, J1 (µ) = 0 (µ0 = 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).
A hollow circular cylinder of finite length is considered. The following conditions are
prescribed:
w = f0 (r, z) at t = 0 (initial condition),
∂t w = f1 (r, z) at t = 0 (initial condition),
w = g1 (z, t) at r = R1 (boundary condition),
w = g2 (z, t) at r = R2 (boundary condition),
w = g3 (r, t) at z = 0 (boundary condition),
w = g4 (r, t) at z = l (boundary condition).
650 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T WO S PACE VARIABLES
Solution:
Z lZ R2 Z lZ R2
∂
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 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.
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
transcendental equation
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 )
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
is also taken from Section 7.1.1.
⊙ Literature: A. G. Butkovskiy (1979), 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 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
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
⊙ 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
is also taken from Section 7.1.2.
⊙ Literature: A. G. Butkovskiy (1979), 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 third boundary
value problem for 0 ≤ r ≤ R, 0 ≤ ϕ ≤ 2π) with the additional term (1); the Green’s function
is also taken from Section 7.1.2.
⊙ Literature: A. G. Butkovskiy (1979), 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 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, (2)
0 0 R1
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
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
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
2◦ . A circular cylinder of finite length is considered. The following conditions are pre-
scribed:
w = f0 (r, z) at t = 0 (initial condition),
∂t w = f1 (r, z) at t = 0 (initial condition),
∂r w = g1 (z, t) at r = R (boundary condition),
w = g2 (r, t) at z = 0 (boundary condition),
w = g3 (r, t) at z = l (boundary condition).
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);
the Green’s function is also taken from Section 7.1.3.
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
∂ 2w
7.3 Equations of the Form = a2∆2 w − bw + Φ(x, y, t)
∂t2
7.3.1 Problems in Cartesian Coordinates
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
w = f (x, y) at t = 0,
∂t w = g(x, y) at t = 0.
p
where ρ = (x − ξ)2 + (y − η)2 .
660 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T WO S PACE VARIABLES
p
where ρ = (x − ξ)2 + (y − η)2 .
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
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 ∂
+a g1 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=0
Z t Z l2
2 ∂
−a g2 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=l1
Z t Z l1
2 ∂
+a g3 (ξ, τ ) G(x, y, ξ, η, t − τ ) dξ dτ
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
7.3. Equations of the Form ∂t2
= 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.
662 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T WO S PACE VARIABLES
The solution w(x, y, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞ 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
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 t Z l2
2 ∂
+a g1 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=0
Z t Z l2
2 ∂
−a g2 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
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
+ Φ(ξ, η, τ )G(x, y, ξ, η, t − τ ) dη dξ dτ,
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
(
nπ mπ p 1 for m = 0,
pn = , qm = , λnm = a2 p2n + a2 qm2 + b, Am =
l1 l2 2 for m 6= 0.
Solution:
Z Z
∂ l1 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 t Z l2
2 ∂
+a g1 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=0
Z t Z l2
+ a2 g2 (η, τ )G(x, y, l1 , η, t − τ ) dη dτ
0 0
664 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T WO S PACE VARIABLES
Z tZ l1
2 ∂
+a g3 (ξ, τ ) G(x, y, ξ, η, t − τ ) dξ dτ
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
+ Φ(ξ, η, τ )G(x, y, ξ, η, t − τ ) dη dξ dτ,
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
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 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
7.3. Equations of the Form ∂t2
= 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,
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.
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.
The solution w(r, ϕ, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞ ∞ 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
µJn′ (µR) + kJn (µR) = 0.
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 ,
1 for n6= 0, Jn(µnmR1)−Jn(µnmR2)
Zn(µnmr)=Jn(µnmR1)Yn(µnmr)−Yn(µnmR1)Jn(µnmr),
where the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the
transcendental equation
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
Jn′ (µR1 )Yn′ (µR2 ) − Yn′ (µR1 )Jn′ (µR2 ) = 0.
The solution w(r, ϕ, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞ ∞
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
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
where the Jnπ/ϕ0 (r) are Bessel functions and the µnm are positive roots of the transcen-
p
dental equation Jnπ/ϕ0 (µR) = 0, and λnm = a2 µ2nm + b.
∂2w
7.3. Equations of the Form ∂t2
= a2 ∆2 w − bw + Φ(x, y, t) 669
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
+ Φ(ξ, η, τ )G(r, ϕ, ξ, η, t − τ )ξ dξ dη dτ.
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
where the Jnπ/ϕ0 (r) are Bessel functions and the µnm are positive roots of the transcen-
′
dental equation Jnπ/ϕ (µR) = 0.
0
Solution:
Z ϕ0Z R
∂
w(r, ϕ, t) = f0 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
∂t 0 0
Z ϕ0Z R
+ f1 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
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
+ Φ(ξ, η, τ )G(r, ϕ, ξ, η, t − τ )ξ dξ dη dτ.
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
where the Jsn (r) are Bessel functions and the µnm are positive roots of the transcendental
equation
µJs′ n (µR) + kJsn (µR) = 0.
In the solutions 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.
A circular cylinder of finite length is considered. The following conditions are prescribed:
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
∂
− 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
where 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 value problem for
0 ≤ r ≤ R).
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
+ a2 g1 (η, τ )G(r, z, R, η, t − τ ) dη dτ
0 0
672 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T WO S PACE VARIABLES
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
+ Φ(ξ, η, τ )G(r, z, ξ, η, t − τ ) dξ dη dτ.
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,
where the µn are zeros of the first-order Bessel function, J1 (µ) = 0 (µ0 = 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).
1◦ . A circular cylinder of finite length is considered. The following conditions are pre-
scribed:
w = f0 (r, z) at t = 0 (initial condition),
∂t w = f1 (r, z) at t = 0 (initial condition),
w = g1 (z, t) at r = R (boundary condition),
∂z w = g2 (r, t) at z = 0 (boundary condition),
∂z w = g3 (r, t) at z = l (boundary 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
− 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
+ Φ(ξ, η, τ )G(r, z, ξ, η, t − τ ) dξ dη dτ.
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,
where 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 value problem for
0 ≤ r ≤ R).
2◦ . A circular cylinder of finite length is considered. The following conditions are pre-
674 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T WO S PACE VARIABLES
scribed:
w = f0 (r, z) at t = 0 (initial condition),
∂t w = f1 (r, z) at t = 0 (initial condition),
∂r w = g1 (z, t) at r = R (boundary condition),
w = g2 (r, t) at z = 0 (boundary condition),
w = g3 (r, t) at z = l (boundary condition).
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 ∂
+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
Z tZ l Z R
+ Φ(ξ, η, τ )G(r, z, ξ, η, t − τ ) dξ dη dτ.
0 0 0
Here,
∞ ∞
4ξ X X 1 µn r µn ξ
G(r, z, ξ, η, t) = 2 J0 J0
R l n=0 m=1 J02 (µn ) R R
√
mπz mπη sin t λnm
× sin sin √ ,
l l λnm
a2 µ2n a2 π 2 m2
λnm = 2
+ + b,
R l2
where the µn are zeros of the first-order Bessel function, J1 (µ) = 0 (µ0 = 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).
Solution:
Z l Z R2
∂
w(r, z, t) = f0 (ξ, η)G(r, z, ξ, η, t) dξ dη
∂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 ∂
−a g2 (η, τ ) G(r, z, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R2
Z t Z R2
2 ∂
+a g3 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ
0 R1 ∂η η=0
Z t Z R2
2 ∂
−a g4 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ
0 R1 ∂η η=l
Z t Z l Z R2
+ Φ(ξ, η, τ )G(r, z, ξ, η, t − τ ) dξ dη dτ.
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
√
mπz mπη sin t λnm
× 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
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.
A hollow circular cylinder of finite length is considered. The following conditions are
prescribed:
w = f0 (r, z) at t = 0 (initial condition),
∂t w = f1 (r, z) at t = 0 (initial condition),
∂r w = g1 (z, t) at r = R1 (boundary condition),
∂r w = g2 (z, t) at r = R2 (boundary condition),
∂z w = g3 (r, t) at z = 0 (boundary condition),
∂z w = g4 (r, t) at z = l (boundary condition).
676 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T WO S PACE VARIABLES
Solution:
Z l Z R2
∂
w(r, z, t) = f0 (ξ, η)G(r, z, ξ, η, t) dξ dη
∂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
+a g4 (ξ, τ )G(r, z, ξ, l, t − τ ) dξ dτ
0 R1
Z tZ l Z R2
+ Φ(ξ, η, τ )G(r, z, ξ, η, t − τ ) dξ dη dτ.
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
transcendental equation
J1 (µ)Y1 (sµ) − J1 (sµ)Y1 (µ) = 0.
◮ Fundamental solutions.
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
⊙ Literature: V. S. Vladimirov, V. P. Mikhailov, A. A. Vasharin et al. (1974).
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 .
678 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T WO S PACE VARIABLES
Solution:
Z Z
∂ l1 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) + kf0 (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 t Z l2
+ a2 g2 (η, τ )G(x, y, l1 , η, t − τ ) dη dτ
0 0
Z tZ l1
− a2 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 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.
The solution w(x, y, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞ 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
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.
2◦ . A rectangle is considered. The following conditions are prescribed:
w = f0 (x, y) at t=0 (initial condition),
∂t w = f1 (x, y) at t=0 (initial condition),
w = g1 (y, t) at x=0 (boundary condition),
∂x w = g2 (y, t) at x = l1 (boundary condition),
w = g3 (x, t) at y=0 (boundary condition),
∂y w = g4 (x, t) at y = l2 (boundary 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) + kf0 (ξ, η) G(x, y, ξ, η, t) dη dξ
0
Z t0Z l2
2 ∂
+a g1 (η, τ ) G(x, y, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=0
Z t Z l2
+ a2 g2 (η, τ )G(x, y, l1 , η, t − τ ) dη dτ
0 0
Z t Z l1
2 ∂
+a g3 (ξ, τ ) G(x, y, ξ, η, t − τ ) dξ dτ
0 0 ∂η η=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
+ Φ(ξ, η, τ )G(x, y, ξ, η, t − τ ) dη dξ dτ,
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
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) + 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
+ Φ(ξ, η, τ )G(r, ϕ, ξ, η, t − τ )ξ dξ dη dτ.
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, . . .),
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.
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) + kf0 (ξ, η) G(r, ϕ, ξ, η, t)ξ dξ dη
0 0
Z t Z 2π
+ a2 R 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
∂2 w
7.4. Telegraph Equation ∂t2
+ 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
where the Jn (ξ) are Bessel functions and the µm are positive roots of the transcendental
equation Jn′ (µR) = 0.
The solution w(r, ϕ, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞ 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
∂
w(r, ϕ, t) = f0 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
∂t 0 R1
Z 2πZ R2
+ f1 (ξ, η) + kf0 (ξ, η) G(r, ϕ, ξ, η, t)ξ dξ dη
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
+ Φ(ξ, η, τ )G(r, ϕ, ξ, η, t − τ )ξ dξ dη dτ.
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 ,
1 for n6= 0, Jn(µnmR1)−Jn(µnmR2)
Zn(µnmr)=Jn(µnmR1)Yn(µnmr)−Yn(µnmR1)Jn(µnmr), λnm =a2µ2nm+b− 41 k 2,
where the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the
transcendental equation
Solution:
Z 2πZ R2
∂
w(r, ϕ, t) = f0 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
∂t 0 R1
Z 2πZ R2
+ f1 (ξ, η) + kf0 (ξ, η) G(r, ϕ, ξ, η, t)ξ dξ dη
0 R1
Z tZ 2π
− a2 R1 g1 (η, τ )G(r, ϕ, R1 , η, t − τ ) dη dτ
0 0
∂2 w
7.4. Telegraph Equation ∂t2
+ 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
+ Φ(ξ, η, τ )G(r, ϕ, ξ, η, t − τ )ξ dξ dη dτ.
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,
the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the transcen-
dental equation
Jn′ (µR1 )Yn′ (µR2 ) − Yn′ (µR1 )Jn′ (µR2 ) = 0.
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),
where the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the
transcendental equation
′
µJn (µR1 ) − s1 Jn (µR1 ) µYn′ (µR2 ) + s2 Yn (µR2 )
= µYn′ (µR1 ) − s1 Yn (µR1 ) µJn′ (µR2 ) + s2 Jn (µR2 ) .
686 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T WO S PACE VARIABLES
Solution:
Z ϕ0Z R
∂
w(r, ϕ, t) = f0 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
∂t 0 0
Z ϕ0Z R
+ f1 (ξ, η) + kf0 (ξ, η) G(r, ϕ, ξ, η, t)ξ dξ dη
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 ∂
+a 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
+ Φ(ξ, η, τ )G(r, ϕ, ξ, η, t − τ )ξ dξ dη dτ.
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
where the Jnπ/ϕ0 (r) are Bessel functions and the µnm are positive roots of the transcen-
dental equation Jnπ/ϕ0 (µR) = 0.
Solution:
Z ϕ0Z R
∂
w(r, ϕ, t) = f0 (ξ, η)G(r, ϕ, ξ, η, t)ξ dξ dη
∂t 0 0
Z ϕ0Z R
+ f1 (ξ, η) + kf0 (ξ, η) G(r, ϕ, ξ, η, t)ξ dξ dη
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
+ Φ(ξ, η, τ )G(r, ϕ, ξ, η, t − τ )ξ dξ dη dτ.
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
where the Jnπ/ϕ0 (r) are Bessel functions and the µnm are positive roots of the transcen-
′
dental equation Jnπ/ϕ (µR) = 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
where the Jsn (r) are Bessel functions and the µnm are positive roots of the transcendental
equation
µJs′ n (µR) + βJsn (µR) = 0.
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
where 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 value problem for
0 ≤ r ≤ R).
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) + kf0 (ξ, η) G(r, z, ξ, η, t) dξ dη
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
+a g3 (ξ, τ )G(r, z, ξ, l, t − τ ) dξ dτ
0 0
Z tZ lZ R
+ Φ(ξ, η, τ )G(r, z, ξ, η, t − τ ) dξ dη dτ.
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
√ #
mπz mπη sin t λnm
× cos cos √ ,
l l λnm
690 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T WO S PACE VARIABLES
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
values of the first ten roots µn are specified in Section 3.2.1 (see the second boundary value
problem for 0 ≤ r ≤ R).
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
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) + kf0 (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
− 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
+ Φ(ξ, η, τ )G(r, z, ξ, η, t − τ ) dξ dη dτ.
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
mπz mπη sin(λnm t)
× 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,
2◦ . A circular cylinder of finite length is considered. The following conditions are pre-
scribed:
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) + kf0 (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, 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
Z tZ lZ R
+ Φ(ξ, η, τ )G(r, z, ξ, η, t − τ ) dξ dη dτ.
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
mπz mπη sin(λnm t)
× sin sin ,
l l λnm
r
a2 µ2n a2 π 2 m 2 k2
λnm = + + b − ,
R2 l2 4
where the µn are zeros of the first-order Bessel function, J1 (µ) = 0 (µ0 = 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).
A hollow circular cylinder of finite length is considered. The following conditions are
prescribed:
w = f0 (r, z) at t = 0 (initial condition),
∂t w = f1 (r, z) at t = 0 (initial condition),
w = g1 (z, t) at r = R1 (boundary condition),
w = g2 (z, t) at r = R2 (boundary condition),
w = g3 (r, t) at z = 0 (boundary condition),
w = g4 (r, t) at z = l (boundary condition).
7.5. Other Equations with Two Space Variables 693
Solution:
Z l Z R2
∂
w(r, z, t) = f0 (ξ, η)G(r, z, ξ, η, t) dξ dη
∂t 0 R1
Z l Z R2
+ f1 (ξ, η) + kf0 (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 ∂
−a g2 (η, τ ) G(r, z, ξ, η, t − τ ) dη dτ
0 0 ∂ξ ξ=R2
Z t Z R2
2 ∂
+a g3 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ
0 R1 ∂η η=0
Z t Z R2
2 ∂
−a g4 (ξ, τ ) G(r, z, ξ, η, t − τ ) dξ dτ
0 R1 ∂η η=l
Z t Z l Z R2
+ Φ(ξ, η, τ )G(r, z, ξ, η, t − τ ) dξ dη dτ.
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 )
√
mπz mπη sin t λnm
× 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
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.
w = f (x, y) at t = 0,
tm/2 ∂t w = g(x, y) at t = 0.
where Ct = {ρ2 ≤ km
2 t2−m } is the circle with center at (x, y) and radius k t1/km .
m
⊙ Literature: M. M. Smirnov (1975).
Chapter 8
Second-Order
Hyperbolic Equations with
Three or More Space Variables
∂ 2w
8.1 Wave Equation = a2∆3 w
∂t2
8.1.1 Problems in Cartesian Coordinates
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.
695
696 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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.
⊙ Literature: A. V. Bitsadze and D. F. Kalinichenko (1985).
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
8.1. Wave Equation ∂t2
= a 2 ∆3 w 697
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
− 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
8.1. Wave Equation ∂t2
= a 2 ∆3 w 699
The solution w(x, y, z, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where
∞ ∞ ∞
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 )
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 ∂
+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 l3Z l2
2 ∂
−a g2 (η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=l1
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
+ 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.
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 ∂
+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 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, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ∂η η=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
∂
+ 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
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.
702 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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 ∂
+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
2 ∂
−a ξg3 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dη dτ.
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,
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.
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, 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,
where the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation Jn′ (µR) = 0.
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
1◦ . A circular cylinder of finite length is considered. The following conditions are pre-
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
∂
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
−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
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
√ Jn (µnm r)Jn (µnm ξ)
πaR l n=0 m=1
k=0
[Jn (µnm R)]2 λnmk
′
p
kπz kπζ
× cos[n(ϕ − η)] cos cos sin at λnmk ,
l l
(
k2 π2 1 for n = 0,
λnmk = µ2nm + 2 , An =
l 2 for n > 0,
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.
2◦ . A circular cylinder of finite length is considered. The following conditions are pre-
∂2w
8.1. Wave Equation ∂t2
= a 2 ∆3 w 705
scribed:
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 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 ∂
+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
2 ∂
−a ξg3 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dη dτ.
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,
λnmk = µ2nm + 2 , An =
l 2 for n > 0,
where the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation Jn′ (µR) = 0.
706 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
A hollow circular cylinder of finite length is considered. The following conditions are
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 ∂
+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
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
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),
where the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the
transcendental equation
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
√
kπz kπζ sin at λnmk
×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);
the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the transcen-
dental equation
Jn′ (µR1 )Yn′ (µR2 ) − Yn′ (µR1 )Jn′ (µR2 ) = 0.
708 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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 ,
λs 2λ2s λ2s + k42 2λs 2 λs
where the Jn (r) and Yn (r) are Bessel functions; the µnm are positive roots of the transcen-
dental equation
′
µJn (µR1 ) − k1 Jn (µR1 ) µYn′ (µR2 ) + k2 Yn (µR2 )
= 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
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 t Z 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
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
+a g4 (ξ, η, τ )G(r, ϕ, z, ξ, η, l, t − τ )ξ dξ dη dτ.
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
√
kπz kπζ sin at λnmk
×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),
where the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the
transcendental equation
2◦ . A hollow circular cylinder of finite length is considered. The following conditions are
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 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
+ a R2 g2 (η, ζ, τ )G(r, ϕ, z, R2 , η, ζ, t − τ ) dη dζ dτ
Z t Z0 2πZ0 R02
2 ∂
+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
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
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
√
kπz kπζ sin at λnmk
×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
Znm (r) = Jn′ (µnm R1 )Yn (µnm r) − Yn′ (µnm R1 )Jn (µnm r);
the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the transcen-
dental equation
Jn′ (µR1 )Yn′ (µR2 ) − Yn′ (µR1 )Jn′ (µR2 ) = 0.
Solution:
Z l Z ϕ0Z R
∂
w(r, ϕ, z, t) = f0 (ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
∂t 0 0 0
Z l Z ϕ0Z R
+ f1 (ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 0
Z t Z l Z ϕ0
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 tZ lZ R
2 1 ∂
+a g2 (ξ, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ξ ∂η η=0
Z tZ lZ R
2 1 ∂
−a g3 (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 R
2 ∂
+a 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 ∂
−a g5 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) ξ dξ dη dτ.
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
where the Jnπ/ϕ0 (r) are Bessel functions and the µnm are positive roots of the transcen-
dental equation Jnπ/ϕ0 (µR) = 0.
A cylindrical sector of finite thickness is considered. The following conditions are pre-
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),
w = g2 (r, z, t) at ϕ = 0 (boundary condition),
w = g3 (r, z, t) at ϕ = ϕ0 (boundary condition),
∂z w = g4 (r, ϕ, t) at z = 0 (boundary condition),
∂z w = g5 (r, ϕ, t) at z = l (boundary condition).
712 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
Solution:
Z l Z ϕ0Z R
∂
w(r, ϕ, z, t) = f0 (ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
∂t 0 0 0
Z l Z ϕ0Z R
+ f1 (ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 0
Z t Z l Z ϕ0
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 tZ lZ R
2 1 ∂
+a g2 (ξ, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ξ ∂η η=0
Z tZ lZ R
2 1 ∂
−a g3 (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 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
+a g5 (ξ, η, τ )G(r, ϕ, z, ξ, η, l, t − τ )ξ dξ dη dτ.
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.
One-dimensional problems with central symmetry that have solutions w = w(r, t) are
considered in Section 6.2.3.
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
G(r, θ, ϕ, ξ, η, ζ, t) = √ Ak Bnmk Jn+1/2 (λnm r)Jn+1/2 (λnm ξ)
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
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
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.
Here, 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.
⊙ Literature: M. M. Smirnov (1975).
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,
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),
( 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 )
where 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
and the λnm are positive roots of the transcendental equation Zn+1/2 (λR2 ) = 0.
Solution:
Z 2πZ π Z R2
∂
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 R1
Z 2πZ π Z R2
+ f1 (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
− 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
716 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
Ak = Bnmk = rZn+1/2 (λnm r) dr,
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
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
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
Green’s function is also taken from Section 8.1.2.
∂2w
8.2. Nonhomogeneous Wave Equation ∂t2
= 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
Green’s function is also taken from Section 8.1.2.
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.
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.
which allows for the equation’s nonhomogeneity; this term is the solution of the nonhomo-
geneous equation with homogeneous initial and boundary conditions.
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
is also taken from Section 8.1.3.
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
taken from Section 8.1.3.
◮ Fundamental solutions.
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
8.3. Equations of the Form ∂t2
= 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
w = f (x, y, z) at t = 0,
∂t w = g(x, y, z) at t = 0.
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
where
nπ mπ kπ
αn = , βm = , γk = , λnmk = a2 (α2n + βm
2
+ γk2 ) + b.
l1 l2 l3
∂2 w
8.3. Equations of the Form ∂t2
= a2 ∆3 w − bw + Φ(x, y, z, t) 723
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
−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
−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
+ 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
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
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.
724 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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.
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 ∂
+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 l3Z l2
2 ∂
−a g2 (η, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dη dζ dτ
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
+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
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
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
2◦ . A rectangular parallelepiped is considered. The following conditions are prescribed:
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),
∂x w = g2 (y, z, t) at x = l1 (boundary condition),
w = g3 (x, z, t) at y = 0 (boundary condition),
∂y w = g4 (x, z, t) at y = l2 (boundary condition),
w = g5 (x, y, t) at z = 0 (boundary condition),
∂z w = g6 (x, y, t) at z = l3 (boundary condition).
726 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
∂
+ 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 ∂
+a g3 (ξ, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ∂η η=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, ζ, t − τ ) dξ dη dτ
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
+ Φ(ξ, η, ζ, τ )G(x, y, z, ξ, η, ζ, t − τ ) dξ dη dζ dτ.
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.
One-dimensional problems with axial symmetry that have solutions w = w(r, t) are
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
8.3. Equations of the Form ∂t2
= a2 ∆3 w − bw + Φ(x, y, z, t) 727
A circular cylinder of finite length is considered. The following conditions are prescribed:
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 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 ∂
+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
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
√ Jn (µnm r)Jn (µnm ξ)
πR l
n=0 m=1 k=1
[Jn (µnm R)]2 λnmk
′
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.
728 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
where the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation Jn′ (µR) = 0.
The solution w(r, ϕ, z, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where
∞ ∞ ∞
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
1◦ . A circular cylinder of finite length is considered. The following conditions are pre-
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
∂
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
−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
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
730 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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,
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.
2◦ . A circular cylinder of finite length is considered. The following conditions are pre-
scribed:
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 tZ lZ 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 ∂
+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
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
∂2 w
8.3. Equations of the Form ∂t2
= 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,
where the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation Jn′ (µR) = 0.
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 t λnmk
× cos[n(ϕ − η)] sin sin √ ,
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);
the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the transcen-
dental equation
Jn (µR1 )Yn (µR2 ) − Yn (µR1 )Jn (µR2 ) = 0.
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
√
kπz kπζ sin t λnmk
× 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
Znm (r) = Jn′ (µnm R1 )Yn (µnm r) − Yn′ (µnm R1 )Jn (µnm r);
the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the transcen-
dental equation
Jn′ (µR1 )Yn′ (µR2 ) − Yn′ (µR1 )Jn′ (µR2 ) = 0.
The solution w(r, ϕ, z, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where
∞ ∞ ∞
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)
− µnmYn′ (µnmR1)−k1Yn(µnmR1) Jn(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
734 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
where the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the
transcendental equation
′
µJn (µR1 ) − k1 Jn (µR1 ) µYn′ (µR2 ) + k2 Yn (µR2 )
= 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
Here,
∞ ∞ ∞
π X X X An Ak µ2nm Jn2 (µnm R2 )
G(r, ϕ, z, ξ, η, ζ, t) = Znm (r)Znm (ξ)
4l J 2 (µ R ) − Jn2 (µnm R2 )
n=0 m=1 k=0 n nm 1
√
kπz kπζ sin t λnmk
× cos[n(ϕ − η)] cos cos √ ,
l l λnmk
∂2 w
8.3. Equations of the Form ∂t2
= 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);
the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the transcen-
dental equation
Jn (µR1 )Yn (µR2 ) − Yn (µR1 )Jn (µR2 ) = 0.
2◦ . A hollow circular cylinder of finite length is considered. The following conditions are
prescribed:
w = f0 (r, ϕ, z) at t = 0 (initial condition),
∂t w = f1 (r, ϕ, z) at t = 0 (initial condition),
∂r w = g1 (ϕ, z, t) at r = R1 (boundary condition),
∂r w = g2 (ϕ, z, t) at r = R2 (boundary 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
∂
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 lZ 2π
− a2 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
+ 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 ∂
+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
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,
∞ √
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
√
kπz kπζ sin t λnmk
×cos[n(ϕ−η)] sin sin √ ,
l l λnmk
736 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
Znm (r) = Jn′ (µnm R1 )Yn (µnm r) − Yn′ (µnm R1 )Jn (µnm r);
the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the transcen-
dental equation
Jn′ (µR1 )Yn′ (µR2 ) − Yn′ (µR1 )Jn′ (µR2 ) = 0.
A cylindrical sector of finite thickness is considered. The following conditions are pre-
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),
w = g2 (r, z, t) at ϕ = 0 (boundary condition),
w = g3 (r, z, t) at ϕ = ϕ0 (boundary condition),
w = g4 (r, ϕ, t) at z = 0 (boundary condition),
w = g5 (r, ϕ, t) at z = l (boundary condition).
Solution:
Z l Z ϕ0Z R
∂
w(r, ϕ, z, t) = f0 (ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
∂t 0 0 0
Z l Z ϕ0Z R
+ f1 (ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 0
Z t Z l Z ϕ0
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 tZ lZ R
2 1 ∂
+a g2 (ξ, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ξ ∂η η=0
Z tZ lZ R
2 1 ∂
−a g3 (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 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 ∂
−a g5 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) ξ dξ dη dτ
0 0 0 ∂ζ ζ=l
Z t Z l 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 0
∂2 w
8.3. Equations of the Form ∂t2
= 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
where the Jnπ/ϕ0 (r) are Bessel functions and the µnm are positive roots of the transcen-
dental equation Jnπ/ϕ0 (µR) = 0.
A cylindrical sector of finite thickness is considered. The following conditions are pre-
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),
w = g2 (r, z, t) at ϕ = 0 (boundary condition),
w = g3 (r, z, t) at ϕ = ϕ0 (boundary condition),
∂z w = g4 (r, ϕ, t) at z = 0 (boundary condition),
∂z w = g5 (r, ϕ, t) at z = l (boundary condition).
Solution:
Z l Z ϕ0Z R
∂
w(r, ϕ, z, t) = f0 (ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
∂t 0 0 0
Z l Z ϕ0Z R
+ f1 (ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
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 ∂
+a g2 (ξ, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ξ ∂η η=0
Z tZ lZ R
2 1 ∂
−a g3 (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 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
+ a2 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
+ Φ(ξ, η, ζ, τ )G(r, ϕ, z, ξ, η, ζ, t − τ )ξ dξ dη dζ dτ.
0 0 0 0
738 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
positive roots of the transcendental equation Jnπ/ϕ0 (µR) = 0.
One-dimensional problems with central symmetry that have solutions of the form w =
w(r, t) are treated in Section 6.2.6.
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
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
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
p
× Pnk (cos θ)Pnk (cos η) cos[k(ϕ − ζ)] sin t a2 λ2nm + b ,
∂2 w
8.3. Equations of the Form ∂t2
= 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
and the λnm are positive roots of the transcendental equation Jn+1/2 (λR) = 0.
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
+ 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 ;
2 for k 6= 0, (n + k)! R2 λ2nm − n(n + 1) Jn+1/2 (λnm R)
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.
740 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
The solution w(r, θ, ϕ, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where
∞ 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
above), and the λnm are positive roots of the transcendental equation
′
λRJn+1/2 (λR) + kR − 12 Jn+1/2 (λR) = 0.
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,
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),
( 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 )]
where 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
and the λnm are positive roots of the transcendental equation Zn+1/2 (λR2 ) = 0.
Solution:
Z 2πZ π Z R2
∂
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 R1
Z 2πZ π Z R2
+ f1 (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
− 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
742 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
Ak = Bnmk = rZn+1/2 (λnm r) dr,
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
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
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
∂
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) + kf0 (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
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, ξ, η, ζ, t − τ ) dη dζ dτ
0 0 0 ∂ξ ξ=l1
Z t Z l3Z l1
2 ∂
+a g3 (ξ, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ∂η η=0
Z t Z l3Z l1
2 ∂
−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
2 ∂
+a 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, 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,
∞ ∞ ∞
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
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
Z l3Z0 l2Z0 l1 0
+ f1 (ξ, η, ζ) + kf0 (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
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
+ 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
Z0 t Z0 l3Z0 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
Z t Z0 l3Z0 0
Z
l2 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 ∞
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.
The solution w(x, y, z, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where
∞ 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).
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
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
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
∞ ∞ ∞
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
∂
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) + kf0 (ξ, η, ζ) G(x, y, z, ξ, η, ζ, t) dξ dη dζ
0 0 0
Z t Z l3Z l2
2 ∂
+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 l3Z l2
+ a2 g2 (η, ζ, τ )G(x, y, z, l1 , η, ζ, t − τ ) dη dζ dτ
0 0 0
748 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
Z t Z l3Z l1
2 ∂
+a g3 (ξ, ζ, τ ) G(x, y, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ∂η η=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, 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
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 ∞
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
One-dimensional problems with axial symmetry that have solutions w = w(r, t) are
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.
A circular cylinder of finite length is considered. The following conditions are prescribed:
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
+ 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 ∂
−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 ∂
+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
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,
∞ ∞ ∞
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
× cos[n(ϕ − η)] sin sin √ ,
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.
A circular cylinder of finite length is considered. The following conditions are prescribed:
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
+ 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
+ a2 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
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
+ ξΦ(ξ, η, ζ, τ )G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dη dζ dτ.
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,
where the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation Jn′ (µR) = 0.
The solution w(r, ϕ, z, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where
∞ ∞ ∞
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
8.4. Telegraph Equation ∂t2
+ k ∂w
∂t
= a2 ∆3 w − bw + Φ(x, y, z, t) 751
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,
752 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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.
2◦ . A circular cylinder of finite length is considered. The following conditions are pre-
scribed:
w = f0 (r, ϕ, z) at t = 0 (initial condition),
∂t w = f1 (r, ϕ, z) at t = 0 (initial condition),
∂r w = g1 (ϕ, z, t) at r = R (boundary condition),
w = g2 (r, ϕ, t) at z = 0 (boundary condition),
w = g3 (r, ϕ, t) at z = l (boundary 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
+ 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
+ a2 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, ζ, t − τ ) dξ dη dτ
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 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
√
sπz sπζ sin t λnms
×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
where the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation Jn′ (µR) = 0.
∂2w
8.4. Telegraph Equation ∂t2
+ k ∂w
∂t
= a2 ∆3 w − bw + Φ(x, y, z, t) 753
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
√
sπz sπζ sin t λnms
× 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
Znm (r) = Jn (µnm R1 )Yn (µnm r) − Yn (µnm R1 )Jn (µnm r),
where the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the
transcendental equation
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
√
sπz sπζ sin t λnms
× cos[n(ϕ − η)] cos cos √ ,
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
Znm (r) = Jn′ (µnm R1 )Yn (µnm r) − Yn′ (µnm R1 )Jn (µnm r);
∂2w
8.4. Telegraph Equation ∂t2
+ k ∂w
∂t
= a2 ∆3 w − bw + Φ(x, y, z, t) 755
the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the transcen-
dental equation
Jn′ (µR1 )Yn′ (µR2 ) − Yn′ (µR1 )Jn′ (µR2 ) = 0.
A hollow circular cylinder of finite length is considered. The following conditions are
prescribed:
The solution w(r, ϕ, z, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where
∞ ∞ ∞
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
756 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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) + kf0 (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 R1
Z tZ lZ 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
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
2
−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
+ a2 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
+ Φ(ξ, η, ζ, τ )G(r, ϕ, z, ξ, η, ζ, t − τ )ξ dξ dη dζ dτ.
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
√
sπz sπζ sin t λnms
× 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
Znm (r) = Jn (µnm R1 )Yn (µnm r) − Yn (µnm R1 )Jn (µnm r),
where the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the
transcendental equation
2◦ . A hollow circular cylinder of finite length is considered. The following conditions are
prescribed:
w = f0 (r, ϕ, z) at t = 0 (initial condition),
∂t w = f1 (r, ϕ, z) at t = 0 (initial condition),
∂r w = g1 (ϕ, z, t) at r = R1 (boundary condition),
∂r w = g2 (ϕ, z, t) at r = R2 (boundary 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
∂
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) + kf0 (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 R1
Z tZ lZ 2π
− a2 R1 g1 (η, ζ, τ )G(r, ϕ, z, R1 , η, ζ, t − τ ) dη dζ dτ
0 0 0
Z tZ lZ 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, ζ, t − τ ) ξ dξ dη dτ
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,
∞ √
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
√
sπz sπζ sin t λnms
× cos[n(ϕ − η)] sin sin √ ,
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
Znm (r) = Jn′ (µnm R1 )Yn (µnm r) − Yn′ (µnm R1 )Jn (µnm r);
the Jn (r) and Yn (r) are Bessel functions, and the µnm are positive roots of the transcen-
dental equation
Jn′ (µR1 )Yn′ (µR2 ) − Yn′ (µR1 )Jn′ (µR2 ) = 0.
758 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
A cylindrical sector of finite thickness is considered. The following conditions are pre-
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),
w = g2 (r, z, t) at ϕ = 0 (boundary condition),
w = g3 (r, z, t) at ϕ = ϕ0 (boundary condition),
w = g4 (r, ϕ, t) at z = 0 (boundary condition),
w = g5 (r, ϕ, t) at z = l (boundary condition).
Solution:
Z l Z ϕ0Z R
∂
w(r, ϕ, z, t) = f0 (ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
∂t 0 0 0
Z l Z ϕ0Z R
+ f1 (ξ, η, ζ) + kf0 (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 0
Z t Z l Z ϕ0
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 tZ l Z R
2 1 ∂
+a g2 (ξ, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ξ ∂η η=0
Z tZ l Z R
2 1 ∂
−a g3 (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 R
2 ∂
+a 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 ∂
−a g5 (ξ, η, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) ξ dξ dη dτ
0 0 0 ∂ζ ζ=l
Z t Z l 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 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
where the Jnπ/ϕ0 (r) are Bessel functions and the µnm are positive roots of the transcen-
dental equation Jnπ/ϕ0 (µR) = 0.
∂2w
8.4. Telegraph Equation ∂t2
+ k ∂w
∂t
= a2 ∆3 w − bw + Φ(x, y, z, t) 759
A cylindrical sector of finite thickness is considered. The following conditions are pre-
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),
w = g2 (r, z, t) at ϕ = 0 (boundary condition),
w = g3 (r, z, t) at ϕ = ϕ0 (boundary condition),
∂z w = g4 (r, ϕ, t) at z = 0 (boundary condition),
∂z w = g5 (r, ϕ, t) at z = l (boundary condition).
Solution:
Z l Z ϕ0Z R
∂
w(r, ϕ, z, t) = f0 (ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
∂t 0 0 0
Z l Z ϕ0Z R
+ f1 (ξ, η, ζ) + kf0 (ξ, η, ζ) G(r, ϕ, z, ξ, η, ζ, t)ξ dξ dη dζ
0 0 0
Z t Z l Z ϕ0
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 tZ l Z R
2 1 ∂
+a g2 (ξ, ζ, τ ) G(r, ϕ, z, ξ, η, ζ, t − τ ) dξ dζ dτ
0 0 0 ξ ∂η η=0
Z tZ l Z R
2 1 ∂
−a g3 (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 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
+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
+ Φ(ξ, η, ζ, τ )G(r, ϕ, z, ξ, η, ζ, t − τ )ξ dξ dη dζ dτ.
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
positive roots of the transcendental equation Jnπ/ϕ0 (µR) = 0.
760 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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)+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
+ 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 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
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 Jn+1/2 (λR) = 0.
∂2w
8.4. Telegraph Equation ∂t2
+ k ∂w
∂t
= a2 ∆3 w − bw + Φ(x, y, z, t) 761
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)+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
+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
+ 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
√ ∞ ∞ 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)
Here, the Jn+1/2 (r) are Bessel functions, the Pns (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.
The solution w(r, θ, ϕ, t) is determined by the formula in the previous paragraph (for
762 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
Here, the Jn+1/2 (r) are Bessel functions, the Pns (µ) are associated Legendre functions
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 Zn+1/2 (λR2 ) = 0.
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
764 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
where the Jn+1/2 (r) and Yn+1/2 (r) are Bessel functions, the Pns (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
The solution w(r, θ, ϕ, t) is determined by the formula in the previous paragraph (for
the second boundary value problem) where
∞ ∞ n
e−kt/2 X X X Al
G(r, θ, ϕ, ξ, η, ζ, t) = √ Zn+1/2 (λnm r)Zn+1/2 (λnm ξ)
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-
dre functions (see above), and the λnm are positive roots of the transcendental equation
′ 1
λZn+1/2 (λR2 ) + s2 − Zn+1/2 (λR2 ) = 0.
2R2
8.5. Other Equations with Three Space Variables 765
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.
w = f0 (r) at t = 0,
(1)
∂t w = f1 (r) at t = 0
766 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
where the λn and un (r) are the eigenvalues and corresponding eigenfunctions of the Sturm–
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 .
General properties of the Sturm–Liouville problem (5)–(6):
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
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).
∂u
+ k(r)u = 0 for r ∈ S. (9)
∂N
768 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
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.
⊙ Literature: V. S. Vladimirov (1988).
◮ Properties of solutions.
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.
8.6. Equations with n Space Variables 769
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
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).
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.
770 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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).
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
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,
772 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
The summation is performed over the indices satisfying the condition s1 + · · · + sn > 0;
the term corresponding to s1 = · · · = sn = 0 is singled out.
where
X∞ X ∞ ∞
X
2n
G(x, y, t) = ... sin(λs1 x1 ) sin(λs2 x2 ) . . . sin(λsn xn )
l1 l2 . . . ln
s1 =1 s2 =1 sn =1
p
sin at λ2s1 + · · · + λ2sn
× sin(λs1 y1 ) sin(λs2 y2 ) . . . sin(λsn yn ) p ,
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
◮ Domain: −∞ < xk < ∞; k = 1, . . . , n. Cauchy problem.
Initial conditions are prescribed:
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 ).
⊙ Literature: 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 ∂
+a gk (y, τ ) G(x, y, t − τ ) dSy(k) dτ
0 S (k) ∂yk y =0
k=1 k
Xn Z tZ
∂
− a2 hk (y, τ ) G(x, y, t − τ ) dSy(k) dτ,
0 S (k) ∂yk yk =lk
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
s1 =1 s2 =1 sn =1
p
sin t a2 (λ2s1 + · · · + λ2sn ) + b
× sin(λs1 y1 ) sin(λs2 y2 ) . . . sin(λsn yn ) p ,
a2 (λ2s1 + · · · + λ2sn ) + b
where
s1 π s2 π sn π
λs1 = , λs2 = , ..., λsn = .
l1 l2 ln
Xn Z t Z
+ a2 hk (y, τ )G(x, y, t − τ ) y dSy(k) dτ.
k =lk
k=1 0 S (k)
8.6. Equations with n Space Variables 775
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,
The solution w(x, t) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞ 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,
λsm (bm cm + λ2sm )(bm + cm )
ϕsm = arctan , Esm = lm + , m = 1, 2, . . . , n;
lm (b2m + λ2sm )(c2m + λ2sm )
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 ∂
+a gk (y, τ ) G(x, y, t − τ ) dSy(k) dτ
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
G(x, y, t) = ... sin(λs1 x1 ) sin(λs2 x2 ) . . . sin(λsn xn )
l1 l2 . . . ln
s1 =1 s2 =1 sn =1
p
sin t a2 (λ2s1 + · · · + λ2sn ) + b
× sin(λs1 y1 ) sin(λs2 y2 ) . . . sin(λsn yn ) p ,
a2 (λ2s1 + · · · + λ2sn ) + b
where
∂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
Solution:
Z tZ
w(x, t) = Φ(y, τ )G(x, y, t − τ ) dVy dτ
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 ∂
+a gk (y, τ ) G(x, y, t − τ ) dSy(k) dτ
(k) ∂y k
k=1 0 S yk =0
Xn Z t Z
∂
− a2 hk (y, τ ) G(x, y, t − τ ) dSy(k) dτ,
0 S (k) ∂y k yk =lk
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:
∞ ∞ ∞
2n e−βt/2 X X X
G(x, y, t) = ... sin(λs1 x1 ) sin(λs2 x2 ) . . . sin(λsn xn )
l1 l2 . . . ln
s1 =1 s2 =1 sn =1
p
sin t a2 (λ2s1 + · · · + λ2sn ) + b − β 2/4
× sin(λs1 y1 ) sin(λs2 y2 ) . . . sin(λsn yn ) p ,
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.
The following conditions are prescribed:
w = f0 (x) at t = 0 (initial condition),
∂t w = f1 (x) at t = 0 (initial condition),
∂xk w = gk (x, t) at xk = 0 (boundary conditions),
∂xk w = hk (x, t) at xk = lk (boundary conditions).
Solution:
Z tZ
w(x, t) = Φ(y, τ )G(x, y, t − τ ) dVy dτ
Z0 V
Z
+ f0 (y)G(x, y, t) dVy + f1 (y) + βf0 (y) G(x, y, t) dVy
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)
Xn Z tZ
+ a2 hk (y, τ )G(x, y, t − τ ) y dSy(k) dτ.
k =lk
k=1 0 S (k)
778 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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,
∞ 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,
Solution:
Z tZ
w(x, t) = Φ(y, τ )G(x, y, t − τ ) dVy dτ
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 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
G(x, y, t) = ... sin(λs1 x1 ) sin(λs2 x2 ) . . . sin(λsn xn )
l1 l2 . . . ln s =1 s =1
1 2 sn =1 p
sin t a2 (λ2s1 + · · · + λ2sn ) + b − β 2/4
× sin(λs1 y1 ) sin(λs2 y2 ) . . . sin(λsn yn ) p ,
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
∂2w n − 1 ∂w
3. + = ∆n w.
∂t2 t ∂t
Darboux equation. Cauchy problem.
Initial conditions are prescribed:
w = f (x) at t = 0,
∂t w = 0 at t = 0.
780 S ECOND -O RDER H YPERBOLIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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).
⊙ Literature: R. Courant and D. Hilbert (1989).
Chapter 9
Second-Order
Elliptic Equations
with Two Space Variables
781
782 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
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.
∂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).
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
∂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 −∞
∗
More rigorously, Σ must satisfy the Lyapunov condition [see Babich, Kapilevich, Mikhlin et al. (1964)
and Tikhonov and Samarskii (1990)].
784 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
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)
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)
∂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 −∞
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)
⊙ Literature: H. S. Carslaw and J. C. Jaeger (1984).
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
∂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
∂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.
∂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.
9.1. Laplace Equation ∆2 w = 0 787
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
⊙ Literature: M. M. Smirnov (1975).
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
⊙ Literature: M. M. Smirnov (1975).
◮ Particular solutions:
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.
The condition
w = f (ϕ) at r=R
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
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 → ∞.
9.1. Laplace Equation ∆2 w = 0 789
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).
where the coefficients an and bn are defined by the same relations as in the inner problem,
and C is an arbitrary constant.
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964).
790 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
The condition
∂r w + kw = f (ϕ) at r = R
where the coefficients a0 , an , and bn are defined by the same relations as in the inner
problem.
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964).
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
(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
⊙ Literature: M. M. Smirnov (1975).
∂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
∂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
792 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
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:
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
9.1. Laplace Equation ∆2 w = 0 793
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).
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.
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964).
∆2 w = −Φ(r) (1)
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
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]
∂y 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, 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]
where C is an arbitrary constant.
∂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
second boundary value problem) where
∞
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
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 −∞
9.2. Poisson Equation ∆2 w = −Φ(x) 799
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
where
1 cosh[π(x − ξ)/a] − cos[π(y + η)/a]
G(x, y, ξ, η) = ln
4π cosh[π(x − ξ)/a] − cos[π(y − η)/a]
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).
(k2 + k3 )µ
Here, the µn are positive roots of the transcendental equation tan(µa) = .
µ2 − k2 k3
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
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
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
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
9.2. Poisson Equation ∆2 w = −Φ(x) 803
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.
⊙ Literature: A. G. Butkovskiy (1979).
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
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
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
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
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 ]
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. Poisson Equation ∆2 w = −Φ(x) 807
◮ 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
1 cosh[π(x − ξ)/a] − cos[π(y + η)/a]
G(x, y, ξ, η) = ln .
4π cosh[π(x − ξ)/a] − cos[π(y − η)/a]
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|
808 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
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
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
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
and A. N. Tikhonov (1974).
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
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)
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
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
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.
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.
⊙ Literature: A. N. Tikhonov and A. A. Samarskii (1990).
9.3. Helmholtz Equation ∆2 w + λw = −Φ(x) 813
∂2w ∂2w
+ + λw = −Φ(x, y).
∂x2 ∂y 2
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 −∞
∂y w = f (x) at y = 0.
9.3. Helmholtz Equation ∆2 w + λw = −Φ(x) 815
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)).
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
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 .
∂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 ∞
+ Φ(ξ, η)G(x, y, ξ, η) dξ dη.
0 0
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 ∞
+ Φ(ξ, η)G(x, y, ξ, η) dξ dη.
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
∂y 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, 0) dξ + f2 (ξ)G(x, y, ξ, a) dξ
Z a−∞
Z ∞ −∞
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.
∂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
second boundary value problem) where
∞
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
Here, the µn are positive roots of the transcendental equation tan(µa) = .
µ2 − k1 k2
818 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
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
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
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.
The solution w(x, y) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞
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
Here, the µn are positive roots of the transcendental equation tan(µa) = .
µ2 − k2 k3
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.
820 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
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 ∞
∂
− 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
∞
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.
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.
∂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.
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.
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
,
where the µn and νm are positive roots of the transcendental equations
(k1 + k2 )µ (k3 + k4 )ν
tan(µa) = , tan(νb) = .
µ2 − k1 k2 ν 2 − k3 k4
Eigenfunctions:
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
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
x = 0, y = 0, y = a − x.
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
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964).
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.
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
∂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.
(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.
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), B. M. Budak, A. A. Samarskii,
and A. N. Tikhonov (1980).
826 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
∂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.
w = 0 at r = R1 , w = 0 at r = R2 .
Eigenvalues:
λnm = µ2nm ; n = 0, 1, 2, . . . ; m = 1, 2, 3, . . .
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.
∂r w = 0 at r = R1 , ∂r w = 0 at r = R2 .
Eigenvalues:
λnm = µ2nm ; n = 0, 1, 2, . . . ; m = 0, 1, 2, . . .
∂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, . . . ;
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.
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
The square of the norm of an eigenfunction is given by
αR2 h ′ i2
kwnm k2 = J nπ (µnm ) .
4 α
∂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
9.3. Helmholtz Equation ∆2 w + λw = −Φ(x) 829
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
The square of the norm of an eigenfunction is given by
2 αR2 n2 h i2 αR2
kwnm k = (1 + δn0 ) 1 − 2 J nπ (µnm ) , kw00 k2 = .
4 µnm α 2
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
w=0 at r = R1 , w = 0 at r = R2 ,
w=0 at ϕ = 0, w = 0 at ϕ = α.
Eigenvalues:
λnm = µ2nm ,
830 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
∂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).
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
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).
TABLE 9.4
∂2w ∂2 w
Transformations reducing equation 9.4.1.3 to the Helmholtz equation ∂ξ2
+ ∂η 2
= bw
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
∂2w ∂2w
5. + = keax+by w.
∂x2 ∂y 2
The transformation
ξ = ax + by, η = bx − ay
leads to an equation of the form 9.4.1.4:
∂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),
834 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
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.
2◦ . The transformation
z = 12 (x2 − y 2 ), ζ = xy
leads to a similar equation
√
∂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,
where C is an arbitrary constant.
∂2w ∂2w
9. + = [f (ax + by) + g(bx − ay)]w.
∂x2 ∂y 2
The transformation
ξ = ax + by, η = bx − ay
leads to an equation of the form 9.4.1.8:
∂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
leads to an equation of the form 9.4.1.8:
∂2w ∂2w
+ = [f (2z) + g(ζ)]w.
∂z 2 ∂ζ 2
9.4. Other Equations 835
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)
836 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
where
(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:
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
∂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)
∂2U ∂2U
2
+ 2
= γ + 14 α2 + 14 β 2 U,
∂x ∂y
∂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.
9.4. Other Equations 837
1◦ . Particular solutions:
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 .
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
∂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.
1◦ . Particular solutions:
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◦ ).
9.4. Other Equations 839
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).
840 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
∂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.
1◦ . Particular solutions:
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◦ ).
9.4. Other Equations 841
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
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
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.
842 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
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
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
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. Other Equations 843
′′ A ′ 4 − nm
wξξ + w = 0, A= . (1)
ξ ξ (2 − n)(2 − m)
where ϕ(x) and ψ(y) are determined by the following second-order linear ordinary differ-
ential equations (A1 is an arbitrary constant):
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
where C1 and C2 are arbitrary constants.
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◦ . 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 . (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
The general solution of equation (2) is given by
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,
∂ ∂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◦ . 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 .
′′ A ′
wξξ + w = Bw, (1)
ξ ξ
where
4 − nm 4c
A= , B= .
(2 − n)(2 − m) ab(2 − n)2 (2 − m)2
The general solution of equation (1) is given by
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),
where ϕ(x) and ψ(y) are determined by the following second-order linear ordinary differ-
ential equations (A1 is an arbitrary constant):
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
where f (x) and g(y) are determined by the following second-order linear ordinary differ-
ential equations (A2 is an arbitrary constant):
The solutions of equations (3) are expressed in terms of the Bessel functions (or modified
Bessel functions).
846 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
where ϕ(x) and ψ(y) are determined by the following second-order linear ordinary differ-
ential equations (A1 is an arbitrary constant):
′′ 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 .
848 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
where C1 , C2 , and λ are arbitrary constants, and the function H = H(x) is determined by
the ordinary differential equation [f (x)Hx′ ]′x + λ2 H = 0.
3◦ . Separable particular solution:
where C1 , C2 , and λ are arbitrary constants, and the function Z = Z(x) is determined by
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
1◦ . Particular solutions:
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 ϕ(x) and ψ(y) are determined by the following second-order linear ordinary differ-
ential equations (A is an arbitrary constant):
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
This is a two-dimensional equation of the heat and mass transfer theory with a linear source
in an inhomogeneous anisotropic medium. The functions f = f (x) and g = g(y) are the
principal thermal diffusivities.
1◦ . There are multiplicative separable particular solutions of the form
where ϕ(x) and ψ(y) are determined by the following second-order linear ordinary differ-
ential equations (A is an arbitrary constant):
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
where Φ(x) and Ψ(y) are determined by the following second-order linear ordinary differ-
ential equations (C is an arbitrary constant):
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.
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964).
∂2w ∂2w
2. yn + = 0.
∂x2 ∂y 2
1◦ . Particular solutions:
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)
where A, B, C, and D are arbitrary constants.
2◦ . Particular solutions with even powers of x:
m
X
w= ϕk (y)x2k ,
k=0
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.
854 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
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
1◦ . Particular solutions:
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 , C2 , and λ are arbitrary constants, and the function H = H(x) is determined by
the ordinary differential equation Hxx′′ + λ2 f (x)H = 0.
where C1 , C2 , and λ are arbitrary constants, and the function Z = Z(x) is determined by
′′ − λ2 f (x)Z = 0.
the ordinary differential equation Zxx
9.4. Other Equations 855
∂2w ∂2w ∂w
a(x) 2
+ 2
+ b(x) + c(x)w = −Φ(x, y) (1)
∂x ∂y ∂x
s1 ∂x w − k1 w = f1 (y) at x = x1 ,
(2)
s2 ∂x w + k2 w = f2 (y) at x = x2 ,
856 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T WO S PACE VARIABLES
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
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
3◦ . If the conditions
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.
9.4. Other Equations 857
TABLE 9.5 √
The functions Ψn in (3) for various boundary conditions. Notation: βn = 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
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
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
Three or More Space Variables
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 = 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
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964).
∂z w = f (x, y) at z = 0.
10.1. Laplace Equation ∆3 w = 0 861
Solution:
Z ∞Z ∞
1 f (ξ, η) dξ dη
w(x, y, z) = − p + C,
2π −∞ −∞ (x − ξ)2 + (y − η)2 + z 2
where C is an arbitrary constant.
⊙ Literature: V. S. Vladimirov, V. P. Mikhailov, A. A. Vasharin et al. (1974).
◆ 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.
862 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
◮ Particular solutions:
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.
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
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,
◆ For the solution of other boundary value problems for the three-dimensional Laplace
equation in the cylindrical coordinate system, see Section 10.2.3 for Φ ≡ 0.
◮ Particular solutions:
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
w = f (θ, ϕ) at r = R.
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
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).
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.
866 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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).
◆ For the solution of other boundary value problems for the three-dimensional Laplace
equation in the spherical coordinate system, see Section 10.2.4 for Φ ≡ 0.
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
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
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
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}.
10.2. Poisson Equation ∆3 w + Φ(x) = 0 869
∆3 w + Φ(r) = 0 (1)
w = f (r) for r ∈ S
∂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}
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
by the following conditions:
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|
870 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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).
where
p p
R− = (x − ξ)2 + (y − η)2 + (z − ζ)2 , R+ = (x − ξ)2 + (y − η)2 + (z + ζ)2 .
⊙ Literature: A. G. Butkovskiy (1979), B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
∂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
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 ∞
+ Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dξ dη dζ,
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
⊙ Literature: V. S. Vladimirov, V. P. Mikhailov, A. A. Vasharin et al. (1974), A. G. Butkovskiy (1979).
10.2. Poisson Equation ∆3 w + Φ(x) = 0 873
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 ∞
+ Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dξ dη dζ,
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
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 −∞ −∞
874 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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 .
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, ζ) dξ dη
−∞ −∞ ∂ζ ζ=0
Z ∞Z ∞
+ f2 (ξ, η)G(x, y, z, ξ, η, a) dξ dη
Z−∞Z
a ∞
−∞
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 − (−1)n ζ − 2na]2 ,
p
rn2 = (x − ξ)2 + (y − η)2 + [z + (−1)n ζ − 2na]2 .
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 ∞
+ Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dξ dη dζ.
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 .
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
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.
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 −∞
876 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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 .
∂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 ∞
+ Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dζ dη dξ.
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
10.2. Poisson Equation ∆3 w + Φ(x) = 0 877
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
tan(µa) = , tan(νb) = .
µ2 − k1 k2 ν 2 − k3 k4
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 ∞
∂
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, 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 ∞
+ 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) 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
2◦ . An infinite cylindrical domain of a rectangular cross-section is considered. Boundary
conditions are prescribed:
w = f1 (y, z) at x = 0, w = f2 (y, z) at x = a,
∂y w = f3 (x, z) at y = 0, ∂y w = f4 (x, z) at y = b.
878 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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, 0, ζ) dζ dξ
0 −∞
Z aZ ∞
+ f4 (ξ, ζ)G(x, y, z, ξ, b, ζ) dζ dξ
0 −∞
Z aZ bZ ∞
+ Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dζ dη dξ.
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
(
nπ mπ p 1 for m = 0,
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.
Green’s function:
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
10.2. Poisson Equation ∆3 w + Φ(x) = 0 879
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 .
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.
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.
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 .
880 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
∂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,
∂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
tan(µa) = , tan(νb) = .
µ2 − k1 k2 ν 2 − k3 k4
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,
∂z 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) cosh(βnm ζ) for z > ζ ≥ 0,
Hnm (z, ζ) =
exp(−βnm ζ) cosh(βnm z) for ζ > z ≥ 0.
10.2. Poisson Equation ∆3 w + Φ(x) = 0 881
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 .
∂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,
∂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 ,
λm 2λm λm +k42 2λ2m 2 λm
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
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.
∂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 ∞
+ Φ(ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ)ξ dζ dη dξ.
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
where the Jnπ/ϕ0 (r) are Bessel functions and the µnm are positive roots of the transcen-
dental equation Jnπ/ϕ0 (µR) = 0.
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
positive roots of the transcendental equation Jnπ/ϕ0 (µR) = 0.
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).
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.
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), B. M. Budak, A. A. Samarskii,
and A. N. Tikhonov (1980).
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
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),
( 2
1 for k = 0, (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 )
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.
10.2. Poisson Equation ∆3 w + Φ(x) = 0 891
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.
⊙ Literature: V. S. Vladimirov, V. P. Mikhailov, A. A. Vasharin et al. (1974), B. M. Budak, A. A. Samarskii,
and A. N. Tikhonov (1980).
892 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
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
894 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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 .
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1984).
∂
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
boundary value problem.
2◦ . The solution of the second boundary value problem with the boundary condition
∂w
= f (r) for r∈S
∂N
can be represented in the form
Z Z
w(r) = Φ(ρ)G(r, ρ) dVρ + f (ρ)G(r, ρ) dSρ . (3)
V S
10.3. Helmholtz Equation ∆3 w + λw = −Φ(x) 895
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
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.
The principle of limit absorption and the principle of limit amplitude are also employed
to separate a single solution.
⊙ Literature: A. N. Tikhonov and A. A. Samarskii (1990).
∆3 w + λw = −Φ(r) (7)
∂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
∗
Here and later in this section, the cross-section D is assumed to have finite dimensions.
10.3. Helmholtz Equation ∆3 w + λw = −Φ(x) 897
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
898 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
Relations (11) and (15) involve the eigenfunctions un and eigenvalues µn of the two-dimen-
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.
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).
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
3◦ . The Green’s function of the three-dimensional third boundary value problem for equa-
tion (7) with the boundary conditions
∂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).
Relations (11) and (18) involve the eigenfunctions un and eigenvalues µn of the two-dimen-
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-
tion (7) with the boundary conditions
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)
Relations (11) and (19) involve the eigenfunctions un and eigenvalues µn of the two-dimen-
sional first boundary value problem (10) with s = 0 and k = 1.
The Green’s functions of other mixed boundary value problems can be constructed
likewise.
900 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
This solution was obtained taking into account the radiation condition at infinity (see Sec-
tion 10.3.2, condition (6)).
⊙ Literature: A. N. Tikhonov and A. A. Samarskii (1990).
10.3. Helmholtz Equation ∆3 w + λw = −Φ(x) 901
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 ∞
+ Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dξ dη dζ.
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 .
∂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 ∞
+ Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dξ dη dζ.
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 .
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 ∞
+ Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dξ dη dζ.
0 0 −∞
∂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
∞ ∞
+ Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dξ dη dζ.
0 0 −∞
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 −∞ −∞
∂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 ∞
+ Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dξ dη dζ.
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
⊙ Literature: A. N. Tikhonov and A. A. Samarskii (1990).
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
(
nπ mπ p 1 for n = 0,
pn = , qm = 2 2
, βnm = pn + qm − λ, An =
a b 2 for n 6= 0.
∂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.
The solution w(x, y, z) is determined by the formula in the previous paragraph (for the
second boundary value problem) where
∞ ∞
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
tan(µa) = , tan(νb) = .
µ2 − k1 k2 ν 2 − k3 k4
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.
906 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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, 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 ∞
+ 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) 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
2◦ . An infinite cylindrical domain of a rectangular cross-section is considered. Boundary
conditions are prescribed:
w = f1 (y, z) at x = 0, w = f2 (y, z) at x = a,
∂y w = f3 (x, z) at y = 0, ∂y 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, 0, ζ) dζ dξ
0 −∞
Z aZ ∞
+ f4 (ξ, ζ)G(x, y, z, ξ, b, ζ) dζ dξ
0 −∞
Z aZ bZ ∞
+ Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dζ dη dξ.
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
(
nπ mπ p 1 for m = 0,
pn = , qm = , βnm = p2n + qm 2 − λ, Am =
a b 2 for m 6= 0.
10.3. Helmholtz Equation ∆3 w + λw = −Φ(x) 907
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.
Solution:
Z bZ ∞
∂
w(x, y, z) = f1 (η, ζ) G(x, y, z, ξ, η, ζ) dζ dη
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 ∞
+ Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dζ dη dξ.
0 0 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 βnm
n=1 m=1
nπ mπ p
pn = , qm = , βnm = p2n + qm 2 − λ,
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.
∂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,
∂z w = f5 (x, y) at z = 0.
908 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
(
nπ mπ p 1 for n = 0,
pn = , qm = , βnm = p2n + qm 2 − λ, An =
a b 2 for n 6= 0,
(
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 − 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,
∂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
second boundary value problem) where
∞ 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,
10.3. Helmholtz Equation ∆3 w + λw = −Φ(x) 909
(k1 + k2 )µ (k3 + k4 )ν
tan(µa) = , tan(νb) = .
µ2 − k1 k2 ν 2 − k3 k4
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,
∂z w = f5 (x, y) at z = 0.
Solution:
Z bZ ∞
∂
w(x, y, z) = f1 (η, ζ) G(x, y, z, ξ, η, ζ) dζ dη
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 ∞
+ Φ(ξ, η, ζ)G(x, y, z, ξ, η, ζ) dζ dη dξ.
0 0 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 βnm
n=1 m=1
nπ mπ p
pn = , qm = , βnm = p2n + qm 2 − λ,
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.
910 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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,
(
nπ mπ p 1 for n = 0,
pn = , qm = , βnm = p2n + qm 2 − λ, An =
a b 2 for n =
6 0,
(
exp(−βnm z) sinh(βnm ζ) for z > ζ ≥ 0,
Hnm (z, ζ) =
exp(−βnm ζ) sinh(βnm z) for ζ > z ≥ 0.
◆ 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 10.3.2.
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.
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)
πn πm πk
pn = , qm = , sk = .
a b c
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), B. M. Budak, A. A. Samarskii,
and A. N. Tikhonov (1980).
∂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,
∂z w − k5 w = f5 (x, y) at z = 0, ∂z w + k6 w = f6 (x, y) at z = c.
λnml = µ2n + νm
2
+ σl2 ; n, m, l = 1, 2, 3, . . .
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 .
The square of the norm of an eigenfunction is defined as
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)
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
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
π2 2 2
π2 k2
λnmk = (n + m) + m + 2 ; n, m, k = 0, 1, 2, . . .
a2 c
Eigenfunctions:
π πmy n πmx π πkz
wnmk = cos (n+m)x cos −(−1) cos cos (n+m)y cos .
a a a a c
π2 2 2
π2 k2
λnmk = (n + m) + m + 2 ; n, m = 1, 2, 3, . . . ; k = 0, 1, 2, . . .
a2 c
10.3. Helmholtz Equation ∆3 w + λw = −Φ(x) 915
Eigenfunctions:
π πmy n πmx π πkz
wnmk = sin (n+m)x sin −(−1) sin sin (n+m)y cos .
a a a a c
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
√ √
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.
916 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
∂r w = f1 (ϕ, z) at r = R, ∂z 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 ∞
+ Φ(ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ)ξ dζ dη dξ.
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,
βnm = µ2nm − λ, An =
2 for n 6= 0,
where the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation Jn′ (µR) = 0.
∂r w + k1 w = f (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, ξ, η, 0)ξ dξ dη
0 0
Z R Z 2πZ ∞
+ Φ(ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ)ξ dζ dη dξ.
0 0 0
10.3. Helmholtz Equation ∆3 w + λw = −Φ(x) 919
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,
βnm = µ2nm − λ, An =
2 for n 6= 0,
where the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation
µJn′ (µR) + k1 Jn (µR) = 0.
w = f1 (ϕ, z) at r = R, ∂z 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, 0)ξ dξ dη
0 0
Z R Z 2πZ ∞
+ Φ(ξ, η, ζ)G(r, ϕ, z, ξ, η, ζ)ξ dζ dη dξ.
0 0 0
Here,
∞ ∞
1 X X An Jn (µnm r)Jn (µnm ξ)
G(r, ϕ, z, ξ, η, ζ) = 2 cos[n(ϕ − η)]Fnm (z, ζ),
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,
where the Jn (ξ) are Bessel functions and the µnm are positive roots of the transcendental
equation Jn (µR) = 0.
◆ 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 10.3.2.
920 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
The square of the norm of an eigenfunction is defined as
(
(1) (2) πR2 a 2 1 for n = m,
kwnmk k2 = kwnmk k2 = (1 + δn0 ) Jn′ (µnm ) , δnm =
4 0 for n 6= m.
∂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
10.3. Helmholtz Equation ∆3 w + λw = −Φ(x) 921
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
π2 k2
λnmk = + µ2nm ; n = 0, 1, 2, . . . ; m, k = 1, 2, 3, . . .
a2
922 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
The square of the norm of an eigenfunction is defined as
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.
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.
The square of the norm of an eigenfunction is defined as
2
(1) (2) a(1+δn0)(1+δk0) n2 Jn′ (µnmR1) n2
kwnmk k2 = kwnmkk2 = 1− − 1− ,
πµ2nm R22µ2nm Jn′ (µnmR2) R12µ2nm
where δn0 is the Kronecker delta.
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964), B. M. Budak, A. A. Samarskii,
and A. N. Tikhonov (1980).
10.3. Helmholtz Equation ∆3 w + λw = −Φ(x) 923
◮ Particular solutions:
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
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
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).
◆ 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 10.3.2.
∂r w = 0 at r = R.
Eigenvalues:
µ2nk
λ00 = 0, λnk = ; n = 0, 1, 2, . . . ; k = 1, 2, 3, . . .
R2
10.3. Helmholtz Equation ∆3 w + λw = −Φ(x) 927
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 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
930 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
Eigenvalues:
a2
λn = − ; n = 1, 2, 3, . . .
4n2
932 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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).
This equation is encountered in problems of convective heat and mass transfer in a simple
shear flow.
Fundamental solution:
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
This equation is encountered in problems of convective heat and mass transfer in a straining
flow.
Fundamental solution:
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)
10.4. Other Equations with Three Space Variables 933
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),
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.
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.
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 → ∞; 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.
10.4. Other Equations with Three Space Variables 935
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
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).
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).
◮ Particular solutions.
1◦ . Fundamental solution:
1 2π n/2
E (x) = − , σn = (n ≥ 3).
(n − 2)σn |x|n−2 Γ(n/2)
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).
10.5. Equations with n Space Variables 937
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
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
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.
A rectangular parallelepiped is considered. Boundary conditions are prescribed:
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
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.
10.5. Equations with n Space Variables 939
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.
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.
Fundamental solution:
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.
⊙ Literature: V. M. Babich, M. B. Kapilevich, S. G. Mikhlin et al. (1964).
940 S ECOND -O RDER E LLIPTIC E QUATIONS WITH T HREE OR M ORE S PACE VARIABLES
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
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).
⊙ Literature: M. M. Smirnov (1975).
Chapter 11
∂w ∂ 3w
1. + = Φ(x, t).
∂t ∂x3
∞
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
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.
4◦ . Fundamental solution:
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.
5◦ . Domain: −∞ < x < ∞. Cauchy problem.
An initial condition is prescribed:
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 −∞
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:
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.
1◦ . Fundamental solution:
Z ∞ Z
1 2 3 ixu 1 ∞
E = exp[(−bu + iau )t]e du = exp(−bu2 ) cos(au3 t + xu) du.
2π −∞ π 0
2◦ . Domain: −∞ < x < ∞. Cauchy problem.
An initial condition is prescribed:
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 −∞
∂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
944 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
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
11.1. Third-Order Partial Differential Equations 945
∂2u ∂3u ∂u
= + (4ac − b2 ) .
∂t2 ∂z 3 ∂z
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◦ .
3◦ . Fundamental solution:
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)
w = 0 at x = 0.
Solution:
Z ∞
w(x, t) = Ee (x − y, t) − Ee (x + y, t) [f (y) − bf ′′ (y)] dy,
0
∂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
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).
Boundary conditions are prescribed:
∂x w = 0 at x = 0, ∂x w = 0 at x = l.
11.1. Third-Order Partial Differential Equations 947
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).
Boundary conditions are prescribed:
w = 0 at x = 0, ∂x w = 0 at x = l.
Solution:
∞
X aβn2 t
w(x, t) = An sin(βn x) exp − ,
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).
Boundary conditions are prescribed:
∂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
948 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
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 < ∞
with the boundary conditions
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
where ϑ(t) is the Heaviside unit step function and −∞ < t < ∞.
11.1. Third-Order Partial Differential Equations 949
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.
∂w ∂ 2w ∂3w
2. −a −b = Φ(x, t).
∂t ∂x2 ∂t∂x2
1◦ . Domain: −∞ < x < ∞. Cauchy problem.
An initial condition is prescribed:
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.
950 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
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. Third-Order Partial Differential Equations 951
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
952 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
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.
4◦ . Domain: 0 ≤ x ≤ l. Second boundary value problem.
The following conditions are prescribed:
w = f (x) at t = 0 (initial condition),
∂t w = g(x) at t = 0 (initial condition),
∂x w = 0 at x = 0 (boundary condition),
∂x w = 0 at x = l (boundary condition).
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.
2◦ . Solutions periodic in t:
w(x, t) = e−λx A cos(ωt − µx) + B sin(ωt − µx) ,
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 )
Here one simultaneously takes only the upper or only the lower signs; A and B are arbitrary
constants.
11.1. Third-Order Partial Differential Equations 953
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 → ∞,
where ϑ(t) is the Heaviside unit step function and −∞ < t < ∞.
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
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
∂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
where A, B, and β are arbitrary constants.
954 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
2◦ . Solutions periodic in t:
w(x, t) = e−λx A cos(ωt − µx) + B sin(ωt − µx) ,
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 )
Here one simultaneously takes only the upper or only the lower signs; A and B are arbitrary
constants.
3◦ . The substitution w = e−λt u, where λ is a root of the quadratic equation ελ2 −σλ−k = 0,
leads to an equation of the form 11.1.4.2:
∂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.
An initial condition is prescribed:
w = f (x) at t = 0.
Solution:
Z
w= Ee (x − y, t) f (y) − b∆y f (y) dVy , dVy = dy1 . . . dyn ,
Rn
4◦ . Consider an open bounded domain V with boundary S. The following conditions are
prescribed:
w = f (x) at t = 0 (initial condition),
∂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 → ∞,
∂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.
An initial condition is prescribed:
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
∂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 = f (x) at t = 0 (initial condition),
∂t w = g(x) at t = 0 (initial condition),
∂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
w(x, t) = Ak ψk1 (t) + Bk ψk2 (t) uk (x),
k=1
where
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
where f (x) is an arbitrary infinitely differentiable function. This solution satisfies the initial
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.
1◦ . Fundamental solution:
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 ∞
w(x, t) = E (x − y, t)f (y) dy + E (x − y, t − τ )Φ(y, τ ) dy dτ.
−∞ 0 −∞
w = f (x) at t=0
958 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
and various homogeneous boundary conditions. The solution can be represented in terms
of the Green’s function as
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
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.
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
11.2. Fourth-Order One-Dimensional Nonstationary Equations 959
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:
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
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 tan(λl) − tanh(λl) = 0.
960 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
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
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);
∂2w ∂4w
11.2.2 Equation of the Form + a2 =0
∂t2 ∂x4
This equation is encountered when studying transverse vibrations of elastic bars.
11.2. Fourth-Order One-Dimensional Nonstationary Equations 961
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
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.
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 −∞
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.
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
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. 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
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
w = ∂xx w = 0 at x = 0, w = ∂xx w = 0 at x = l.
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
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)
964 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
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◦ .
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.
Boundary conditions are prescribed:
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
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 tan(λl) − tanh(λl) = 0.
5◦ . Domain: 0 ≤ x ≤ l. One end of the bar is clamped and the other is free.
Boundary conditions are prescribed:
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) = sin(λ2n at),
al λ2n 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.
11.2. Fourth-Order One-Dimensional Nonstationary Equations 965
6◦ . Domain: 0 ≤ x ≤ l. One end of the bar is hinged and the other is free.
Boundary conditions are prescribed:
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)
and the λn are positive roots of the transcendental equation tan(λl) − tanh(λl) = 0.
7◦ . Domain: 0 ≤ x ≤ l. The first and third derivatives are prescribed at the ends:
∂x w = ∂xxx w = 0 at x = 0, ∂x w = ∂xxx w = 0 at x = l.
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
w = ∂xx w = 0 at x = 0, ∂x w = ∂xxx w = 0 at x = l.
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.
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
and various homogeneous boundary conditions. The solution can be represented in terms
of the Green’s function as
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
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.
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 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
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.
11.2. Fourth-Order One-Dimensional Nonstationary Equations 967
2◦ . Domain: 0 ≤ x ≤ l. The function and its second derivative are prescribed at the ends:
w = ∂xx w = 0 at x = 0, w = ∂xx w = 0 at x = l.
Green’s function:
∞ p
2X sin t a2 λ4n + k πn
G(x, ξ, t) = sin(λn x) sin(λn ξ) p , λn = .
l a2 λ4n + k l
n=1
3◦ . Domain: 0 ≤ x ≤ l. The first and third derivatives are prescribed at the ends:
∂x w = ∂xxx w = 0 at x = 0, ∂x w = ∂xxx w = 0 at x = l.
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
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 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
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 tan(λl) − tanh(λl) = 0.
968 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
6◦ . Domain: 0 ≤ x ≤ l. Mixed boundary conditions are prescribed at the ends (case 2):
w = ∂x w = 0 at x = 0, ∂xx w = ∂xxx w = 0 at x = l.
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
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.
7◦ . Domain: 0 ≤ x ≤ l. Mixed boundary conditions are prescribed at the ends (case 3):
w = ∂xx w = 0 at x = 0, ∂x w = ∂xxx w = 0 at x = l.
Green’s function:
∞ p
2X sin t a2 λ4n + k π(2n + 1)
G(x, ξ, t) = sin(λn x) sin(λn ξ) p , λn = .
l a2 λ4n + k 2l
n=0
8◦ . Domain: 0 ≤ x ≤ l. Mixed boundary conditions are prescribed at the ends (case 4):
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.
∂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
which is discussed in Section 11.2.1.
11.2. Fourth-Order One-Dimensional Nonstationary Equations 969
∂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
which is discussed in Section 11.2.1.
∂w ∂4w
4. = (ax2 + bx + c)4 .
∂t ∂x4
This is a special case of equation 11.7.4.5 with k = 1 and n = 4.
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
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
which is discussed in Section 11.2.4.
970 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
∂2w ∂4w
7. = ax8 .
∂t2 ∂x4
∂2w ∂4w
8. = (ax2 + bx + c)4 .
∂t2 ∂x4
∂4w ∂4w
9. − = 0.
∂t4 ∂x4
1◦ . Particular solution:
w(x, t) = f (x − t) + g(x + t),
2◦ . General solution:
w = w1 + w2 ,
∂ 2 w1 ∂ 2 w1 ∂ 2 w2 ∂ 2 w2
− = 0, + = 0.
∂t2 ∂x2 ∂t2 ∂x2
3◦ . Fundamental solution:
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.
−∞
(v) For t → 0,
Ee (x, t) = 12 te−|x| 1 + O(t2 ) .
3◦ . Domain: −∞ < x < ∞. Cauchy problem.
Initial conditions are prescribed:
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.
The following conditions are prescribed:
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
an = f0 (ξ) sin(βn ξ) dξ, bn = f1 (ξ) sin(βn ξ) dξ.
l 0 l 0
8◦ . Domain: 0 ≤ x ≤ l. Mixed boundary value problem.
The following conditions are prescribed:
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
an = f0 (ξ) sin(βn ξ) dξ, bn = f1 (ξ) sin(βn ξ) dξ.
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).
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
w = f0 (x) at t = 0,
∂t w = f1 (x) at t = 0,
∂x w + ∂ttx w = g(t) at x = 0.
11.2. Fourth-Order One-Dimensional Nonstationary Equations 975
The solution of the boundary value problem is given by the formula in Item 5◦ of
equation 11.2.6.1 with the additional term
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
equation 11.2.6.1 with the additional term
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).
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 −∞ −∞
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.
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 ).
11.2. Fourth-Order One-Dimensional Nonstationary Equations 977
By substituting these relations into the equation and the dynamic boundary conditions (3),
we arrive at the nonclassical Sturm–Liouville boundary value problem
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
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.
For the geometric boundary conditions (2), one should set M1 = M2 = 0 in formulas
(8) and (9).
978 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
For the geometric boundary conditions (2), one should set K1 = K2 = 0 in formulas
(10) and (11).
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
For the geometric boundary conditions (2), one should set K1 = K2 = 0 in the above
formula for w(x, t).
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.
980 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
where A, B, C, D, and β are arbitrary constants. (It is assumed in all cases that the
radicands are positive.)
2◦ . Fundamental solution:
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.
3◦ . Domain: −∞ < x < ∞. Cauchy problem.
Initial conditions are prescribed:
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.
982 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
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
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.
The following conditions are prescribed:
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.
∂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.
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.
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).
∂2 ∂2 ∂2 ∂2
5. − a2 − b2 w = 0.
∂t2 ∂x2 ∂t2 ∂x2
1◦ . General solution with a 6= b:
w(x, 0) = ϕ0 (x), wt (x, 0) = ϕ1 (x), wtt (x, 0) = ϕ2 (x), wttt (x, 0) = ϕ3 (x). (3)
986 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
and various homogeneous boundary conditions. The solution can be represented in terms
of the Green’s function as
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
4◦ . The second and third derivatives are prescribed on the sides of the rectangle:
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
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.
11.3. Two-Dimensional Nonstationary Fourth-Order Equations 989
2◦ . Particular solutions:
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),
◮ Cauchy problem.
1◦ . Domain: −∞ < x < ∞, −∞ < y < ∞. Cauchy problem.
Initial conditions are prescribed:
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).
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).
990 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
and various homogeneous boundary conditions. The solution can be represented in terms
of the Green’s function as
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
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 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.
w = ∂xx w = 0 at x = 0, w = ∂xx w = 0 at x = l1 ,
∂y w = ∂yyy w = 0 at y = 0, ∂y w = ∂yyy w = 0 at y = l2 .
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.
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).
⊙ Literature: B. M. Budak, A. A. Samarskii, and A. N. Tikhonov (1980).
∂2w
11.3.3 Equation of the Form + a2 ∆∆w + kw = Φ(x, y, t)
∂t2
◮ Solutions of boundary value problems in terms of the Green’s function.
Consider boundary value problems in the rectangular domain 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 with
the general initial conditions
and various homogeneous boundary conditions. The solution can be represented in terms
of the Green’s function as
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
∂2w
∂4w ∂4w
2
11.3.4 Equation of the Form +a + + kw = Φ(x, y, t)
∂t2 ∂x4 ∂y 4
◮ Solutions of boundary value problems in terms of the Green’s function.
Consider boundary value problems in the rectangular domain 0 ≤ x ≤ l1 , 0 ≤ y ≤ l2 with
the general initial conditions
and various homogeneous boundary conditions. The solution can be represented in terms
of the Green’s function as
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
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
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 .
994 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
Green’s function:
∞ ∞
4 XX sin(λnm t)
G(x, y, ξ, η, t) = sin(pn x) sin(qm y) sin(pn ξ) sin(qm η) ,
l1 l2 λnm
n=1 m=1
πn πm p
pn = , qm = , λnm = a2 (p4n + qm 4 ) + k.
l1 l2
3◦ . The first and third derivatives are prescribed on the sides of the rectangle:
∂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 t)
G(x, y, ξ, η, t) = εn εm cos(pn x) cos(qm y) cos(pn ξ) cos(qm η) ,
l1 l2 λnm
n=0 m=0
(
πn πm p 1 for n = 0,
pn = , qm = 2 4 4
, λnm = a (pn + qm ) + k, εn =
l1 l2 2 for n 6= 0.
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 ′ .
⊙ Literature: S. A. Gabov and A. G. Sveshnikov (1990).
996 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
∂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 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 ′ .
⊙ Literature: S. A. Gabov and A. G. Sveshnikov (1990).
∂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.
1◦ . Particular solutions:
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
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
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
w = f (x, y, z) at t = 0, ∂t w = g(x, y, z) at t = 0.
Boundary conditions:
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 ,
w = ∂zz w = 0 at z = 0, w = ∂zz w = 0 at z = l3 .
998 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
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.
Initial conditions are prescribed:
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 .
⊙ Literature: V. S. Vladimirov, V. P. Mikhailov, A. A. Vasharin et al. (1974).
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
1000 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
∞ 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
∞ X
X ∞ X
∞
wp (x, t) = ϕnmk (t)unmk (x),
n=1 m=1 k=1
11.4. Three- and n-Dimensional Nonstationary Fourth-Order Equations 1001
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
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
∂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).
1◦ . The general solution of the homogeneous equation with Φ ≡ 0:
w = w1 + w2 ,
∂ 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
∞
wp (x, t) = ϕnmk (t)unmk (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
unmk (x) = sin sin sin ,
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
1002 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
∂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.
1◦ . Particular solutions:
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|
3◦ . Fundamental solution:
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
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
w = f (r) at t = 0,
∂t w = g(r) at t = 0,
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 ′ .
⊙ Literature: S. A. Gabov and A. G. Sveshnikov (1990).
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
2◦ . Particular solutions:
4◦ . Fundamental solution:
1 2 p
Ee (x, y) = r ln r, r= x2 + y 2 . (1)
8π
11.5. Fourth-Order Stationary Equations 1005
where u1 and u2 are arbitrary functions satisfying the Laplace equation ∆uk = 0 (k = 1, 2).
⊙ Literature: A. N. Tikhonov and A. A. Samarskii (1990).
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.
⊙ Literature: A. V. Bitsadze and D. F. Kalinichenko (1985).
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
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)
∂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
11.5. Fourth-Order Stationary Equations 1007
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 ,
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),
1 dn 2 dm
Pn (x) = (x − 1)n , Pnm (x) = (1 − x2 )m/2 Pn (x).
n! 2n dxn dxm
1008 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
1◦ . Fundamental solution:
1 p 2
Ee (x, y, z) = − x + y2 + z2 . (2)
8π
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.
⊙ Literature: A. V. Bitsadze and D. F. Kalinichenko (1985).
◮ n-dimensional equation.
1◦ . Particular solutions:
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 u1 and u2 are arbitrary functions satisfying the n-dimensional Laplace equation
∆n um = 0 (m = 1, 2).
⊙ Literature: A. V. Bitsadze and D. F. Kalinichenko (1985).
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−
2
R+ = (x − ξ)2 + (y + η)2 , 2
R− = (x − ξ)2 + (y − η)2 .
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) ,
1014 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
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
The above formulas were obtained with the aid of generalized (nonorthogonal) polar
coordinates ρ, ϕ defined by
∂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 ,
w = 0 at y = 0, ∂y w = f (x) at y = 0.
11.5. Fourth-Order Stationary Equations 1015
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
⊙ Literature: G. E. Shilov (1965).
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
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 pn and qm are positive roots of the transcendental equations
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 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
1016 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
∂4w ∂4w
11.5.5 Equation of the Form + + kw = Φ(x, y)
∂x4 ∂y 4
◮ Particular solutions of the homogeneous equation (Φ ≡ 0):
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
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,
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 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.)
11.5. Fourth-Order Stationary Equations 1019
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).
a sin ξ a sinh η
ρ= , z= .
cosh η − cos ξ cosh η − cos ξ
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).
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).
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).
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
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
are set, then the solution of Eq. (3) is sought in the form
Each term in (6) satisfies Eq. (3), and the functions ϕm−1 , ϕm−2 , . . . , ϕ0 are determined
successively from the linear system
∂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
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
where s1 , . . . , sn are nonnegative integers. From now on, we adopt the notation
∂ s1 ∂ sn ∂ s1 +···+sn
... ≡ .
∂x1 ∂xn ∂xs11 . . . ∂xsnn
11.6. Higher-Order Linear Equations with Constant Coefficients 1023
∂ ∂ k ∂ ∂
Pk b ,...,b = b Pk ,..., , b 6= 0 is an arbitrary constant.
∂x1 ∂xn ∂x1 ∂xn
X
Pk (ω1 , . . . , ωn ) ≡ as1 ,...,sn ω1s1 . . . ωnsn 6= 0 if ω ∈ Rn , |ω| =
6 0.
X n
X
∂ ∂ ∂ s1 +···+sn w
Pk ,..., w≡ as1 ,...,sn s1 = 0, si = k, (1)
∂x1 ∂xn ∂x1 . . . ∂xsnn
i=1
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
∂ ∂
Pk ,..., w=0 (2)
∂x1 ∂xn
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.
1024 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
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.
P (λ, ω1 , . . . , ωn ) = 0
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
∂mw X ∂ ∂k w
m−1
= pk i , (1)
∂tm ∂x ∂tk
k=0
1026 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
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:
∂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.
11.6. Higher-Order Linear Equations with Constant Coefficients 1027
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.
−∞
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
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.
⊙ Literature: I. M. Gelfand and G. E. Shilov (1959), S. G. Krein (1972).
Pk (ω1 , . . . , ωn ) = 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 f (x) is an arbitrary infinitely differentiable function. This solution satisfies the initial
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.
Fundamental solution:
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 ,
∂w ∂ 2n+1 w
2. +a = Φ(x, t), n = 1, 2, . . . .
∂t ∂x2n+1
1◦ . Particular solutions of the homogeneous equation with Φ ≡ 0:
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 f (x) is an arbitrary infinitely differentiable function. This solution satisfies the initial
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.
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
1032 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
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
4◦ . Domain: −∞ < x < ∞. Cauchy problem.
An initial condition is prescribed:
w = f (x) at t = 0.
n
X
∂w ∂ 2k w
3. + ak = Φ(x, t).
∂t ∂x2k
k=0
1◦ . Particular solutions of the homogeneous equation with Φ(x, t) = 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
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.
Fundamental solution:
Z ∞ Z
1 1 ∞
E (x, t) = exp −tP (ξ) + ixξ dξ = exp −tP (ξ) cos(xξ) dξ. (2)
2π −∞ π 0
11.6. Higher-Order Linear Equations with Constant Coefficients 1033
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 −∞
n
X
∂w ∂ 2k+1 w
4. + ak = Φ(x, t).
∂t ∂x2k+1
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.
Fundamental solution:
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 ∞
w(x, t) = E (x − y, t)f (y) dy + E (x − y, t − τ )Φ(y, τ ) dy dτ,
−∞ 0 −∞
n
X
∂w ∂kw
5. + ak = Φ(x, t).
∂t ∂xk
k=0
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 ,
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
2◦ . The condition
2k≤n
X
P (ξ) = (−1)k a2k ξ 2k > 0
k=0
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 −∞
n
X
∂w ∂3w ∂ 2k w
6. −b + ak = Φ(x, t).
∂t ∂t∂x2 ∂x2k
k=0
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 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).
1036 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
w = w1 + w2 ,
∂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
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 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.
Fundamental solution:
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 ,
w = f (x) at t = 0,
∂t w = g(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)f (y) dy + E (x − y, t)g(y) dy.
∂t −∞ −∞
11.6. Higher-Order Linear Equations with Constant Coefficients 1037
n
X
∂2w ∂ 2k w
2. + ak = Φ(x, t).
∂t2 ∂x2k
k=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.
Fundamental solution:
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 ∞
w(x, t) = E (x − y, t − τ )Φ(y, τ ) dy dτ
0
Z−∞
∞ Z ∞
∂
+ E (x − y, t)f (y) dy + E (x − y, t)g(y) dy.
∂t −∞ −∞
1038 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
n
X
∂2w ∂4w ∂ 2k w
3. −b + ak = Φ(x, t).
∂t2 ∂t2 ∂x2 ∂x2k
k=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
be satisfied for all real ξ. The condition is satisfied if, say, ak = (−1)k ck with ck ≥ 0 for
all k.
Fundamental solution:
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. Higher-Order Linear Equations with Constant Coefficients 1039
n−1
X
w(x, t) = tk fk (x + at),
k=0
n−1
X
w(x, t) = xk fk (x + at),
k=0
tn−1
E (x, t) = δ(x + at).
(n − 1)!
∂ ∂2 n
2. − w = 0, n = 1, 2, . . .
∂t ∂x2
1◦ . General solution (two representations):
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.
2◦ . Fundamental solution:
1 n−3/2 x2
E (x, t) = √ t exp − .
2 π (n − 1)! 4t
Solution: Z ∞
w(x, t) = f (ξ)E (x − ξ, t) dξ.
−∞
∂2 ∂2 n
3. − w = 0, n = 1, 2, . . .
∂t2 ∂x2
1◦ . General solution (two representations):
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.
2◦ . Fundamental solution:
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
⊙ Literature: G. E. Shilov (1965).
∂2 ∂2 n
4. + w = 0, n = 1, 2, . . .
∂x2 ∂y 2
This is the polyharmonic equation of order n with two independent variables.
1◦ . General solution (two representations):
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.
Boundary conditions are prescribed:
∂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
See also Example 11.7 in Section 11.6.4.
⊙ Literature: G. E. Shilov (1965), L. D. Faddeev (1998).
11.6. Higher-Order Linear Equations with Constant Coefficients 1041
∂2 ∂2 n
5. + w = Φ(x, y), n = 1, 2, . . .
∂x2 ∂y 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
Particular solutions:
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
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.
⊙ Literature: G. E. Shilov (1965).
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
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
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
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
∂ 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
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.)
∂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
(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.
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
11.7. Higher-Order Linear Equations with Variable Coefficients 1047
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).
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
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
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
1048 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
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
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
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
w(x, t) = f (y)G(x, y, t, 0) dy + Φ(y, τ )G(x, y, t, τ ) dy dτ
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).
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
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
w(x, t) = f (y)G(x, y, t, 0) dy + Φ(y, τ )G(x, y, t, τ ) dy dτ
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).
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.
11.7. Higher-Order Linear Equations with Variable Coefficients 1051
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.
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
1052 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
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
⊙ Literature for Section 11.7.2: A. G. Butkovskiy (1979, 1982), A. D. Polyanin (2000a, 2002).
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 ).
11.7. Higher-Order Linear Equations with Variable Coefficients 1053
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).
∂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
∂2E
− Lx,t [E ] = 0,
∂t2
∂E
E t=τ = 0, = δ(x − y),
∂t t=τ
∂2w ∂w
2
+ ψ(x, t) − Lx,t [w] = Φ(x, t)
∂t ∂t
1054 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
∂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
1◦ . Particular solutions of the homogeneous equation with Φ(x, t) = 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
11.7. Higher-Order Linear Equations with Variable Coefficients 1055
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 −∞
∂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.
1056 H IGHER -O RDER PARTIAL D IFFERENTIAL E QUATIONS
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.
1◦ . General solution:
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.
2◦ . Fundamental solution:
tn−1
En (x, t) = E1 (x, t),
(n − 1)!
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)!
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 .
11.7. Higher-Order Linear Equations with Variable Coefficients 1057
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
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.
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
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,
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 ,
1062 S YSTEMS OF L INEAR PARTIAL D IFFERENTIAL E QUATIONS
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
∂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).
1064 S YSTEMS OF L INEAR PARTIAL D IFFERENTIAL E QUATIONS
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
where λ1 and λ2 are roots of the quadratic equation
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).
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
∂ 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
L[θ1 ] = λ1 θ1 , L[θ2 ] = λ2 θ2 .
1066 S YSTEMS OF L INEAR PARTIAL D IFFERENTIAL E QUATIONS
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.
curl u = 0 (1)
wy − vz = 0, uz − wx = 0, vx − uy = 0 (2)
u = ∇ϕ or u = ϕx , v = ϕy , w = ϕz ,
◮ Nonhomogeneous equation.
Consider the nonhomogeneous vector equation
u = ϕx , v = ϕy + η, w = ϕz + ζ. (6)
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).
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)
∆ϕ = 0.
1068 S YSTEMS OF L INEAR PARTIAL D IFFERENTIAL E QUATIONS
◮ 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 ′ .
∆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′ |
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.
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,
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.
1070 S YSTEMS OF L INEAR PARTIAL D IFFERENTIAL E QUATIONS
2◦ . The scalar and vector functions γ and ω in (15), for example, can be sought in the form
By substituting (18) into (15) and by taking into account the identity
∆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′ |
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 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:
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
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
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
∂2 1 ∂ 1 ∂2 ∂2
∆≡ + + + .
∂r 2 r ∂r r 2 ∂ϕ2 ∂z 2
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
12.6. Elasticity Equations 1073
∂ 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.
u = ∇ϕ + curl Ψ, (8)
◮ Cauchy–Kovalevskaya solution.
Any solution of Eqs. (1) with f = 0 can be represented in the form
2 1 [w] = 0. (11)
Here and in the following, the d’Alembert operators 1 and 2 are given by
If ut 6≡ 0, then the general solution of Eq. (10) can be represented as the sum
w = w1 + w2 ,
1 [w1 ] = 0, 2 [w2 ] = 0.
1 ϕ = 0, 2 ξ = 0, 2 η = 0. (13)
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).
12.6. Elasticity Equations 1075
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.)
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.
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.
Any solution of Eqs. (1) can be represented in the form (10), where the vector function w
satisfies the equation
2 1 [w] = f.
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.
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
represented in the form
∆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)
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 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.
1078 S YSTEMS OF L INEAR PARTIAL D IFFERENTIAL E QUATIONS
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,
◮ Invariant transformations.
The Stokes equations (1)–(2) are invariant under the transformation
u + curl Ψ◦ ,
u=e p = pe,
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.
u|x∈S = 0. (4)
We point out that no conditions on the pressure p are needed in the above-stated prob-
lems for the Stokes equations.
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.
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:
Remark 12.8. For uz = 0, the cylindrical coordinates r and ϕ are also used as the polar coordi-
nates on the xy-plane.
1080 S YSTEMS OF L INEAR PARTIAL D IFFERENTIAL E QUATIONS
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
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 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. Stokes Equations for Viscous Incompressible Fluids 1081
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)
∆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.
u = ∇ϕ + curl Ψ, p = p0 − ϕt , (12)
◮ 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)
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).
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.
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)
1084 S YSTEMS OF L INEAR PARTIAL D IFFERENTIAL E QUATIONS
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.
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
1086 S YSTEMS OF L INEAR PARTIAL D IFFERENTIAL E QUATIONS
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
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).
|u − a| ∼ ε, |p − p0 | ∼ ε, |f| ∼ ε,
u + curl Ψ◦ ,
u=e p = pe,
∆p = div f.
12.8. Oseen Equations for Viscous Incompressible Fluids 1087
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)
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 some
solutions of the convective heat equation
Every solution of the Oseen equations (1) with f = 0 can be represented by the formulas
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
∆θe − k2 θe = 0, ∆ψe − k 2 ψe = 0.
1088 S YSTEMS OF L INEAR PARTIAL D IFFERENTIAL E QUATIONS
where p0 = p0 (t) is an arbitrary function, the vector function Ψ satisfies the equation
Ψt + (a · ∇)Ψ − ν∆Ψ = ω,
where p0 (t) is an arbitrary function and the vector function U satisfies the equation
where the vector function v = (v1 , v2 , v3 ) satisfies the independent linear nonhomogeneous
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).
|u − a| ∼ ε, |p − p0 + a′t · x| ∼ ε, |f| ∼ ε,
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).
u + curl Ψ◦ ,
u=e p = pe,
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.
u = ∇ϕ + v, p = p0 − τ ϕtt − ϕt ,
(2)
v = (v1 , v2 , v3 ), v1 = ψy(1) , v2 = −ψx(1) + ψz(2) , v3 = −ψy(2) ,
1090 S YSTEMS OF L INEAR PARTIAL D IFFERENTIAL E QUATIONS
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 telegraph equation
Any solution of the Maxwell equations (1) with f = 0 can be represented by the formulas
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)
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.
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)
where p0 = p0 (t) is an arbitrary function, the vector function Ψ satisfies the equation
τ Ψtt + Ψt − ν∆Ψ = ω,
where p0 (t) is an arbitrary function and the vector function U satisfies the equation
where the vector function v = (v1 , v2 , v3 ) satisfies the independent linear nonhomogeneous
equation
τ vtt + vt − ∆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 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).
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.
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).
u = ∇ϕ + v, p = p0 − L[ϕ],
(2)
v = (v1 , v2 , v3 ), v1 = ψy(1) , v2 = −ψx(1) + ψz(2) , v3 = −ψy(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
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 equation
L[ψ] = 0, (3)
and the function ϕ = ϕ(x, t) is a solution of the Laplace equation ∆ϕ = 0.
◮ 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],
where p0 (t) is an arbitrary function and the vector function U satisfies the equation
∆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].
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
ptt − c2 ∆p = 0
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)
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.
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
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
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).
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).
◮ 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).
1098 S YSTEMS OF L INEAR PARTIAL D IFFERENTIAL E QUATIONS
where ψ (1) = ψ (1) (x, t) and ψ (2) = ψ (2) (x, t) are some solutions of the heat-type equation
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)
1102 S YSTEMS OF L INEAR PARTIAL D IFFERENTIAL E QUATIONS
◮ Toroidal–poloidal decomposition.
Each solution of the thermoelasticity equations (2)–(3) with f = 0 can be represented in the
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
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).
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
where the vector function w and the scalar function ϕ satisfy the equations
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),
where ψ (1) = ψ (1) (x, t) and ψ (2) = ψ (2) (x, t) are some solutions of the wave equation
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
◮ Toroidal–poloidal decomposition.
Each solution of the nondissipative thermoelasticity equations (2)–(3) with f = 0 can be
represented in the 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
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).
Ψtt − c22 ∆Ψ = ω,
ϕtttt − [a + c21 + (αβ)/ρ]∆ϕtt + ac21 ∆∆ϕ = γtt − a∆γ.
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 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),
Any solution of the homogeneous viscoelasticity equations (2) with f = 0 can also be
represented as
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
u = [∂t2 − c21 ∆ − (2ν + κ)∂t ∆][w] + ∇[c21 − c22 + (ν + κ)∂t ][div w], (8)
w = w1 + w2 ,
u = ∇ϕ + curl(aξ + bη), a · b 6= 0,
u = ∇ϕ + curl(aξ + xη), |a| =
6 0,
⊙ Literature for Section 12.16: M. E. Gurtin and E. Sternberg (1962), A. D. Polyanin and S. A. Lychev
(2014a,b).
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,
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.
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
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.
1110 S YSTEMS OF L INEAR PARTIAL D IFFERENTIAL E QUATIONS
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
represented in the form
where the functions ϕ0 (x) and ϕ1 (x) satisfied the Laplace and Poisson equations
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).
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
◮ Invariant transformations.
1◦ . The vector equation (1) is invariant under the transformation
u + curl Ψ◦ ,
u=e
u=e
u + ∇Φ,
(L + K1 ∆)[Φ] = 0.
(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
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
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).
1112 S YSTEMS OF L INEAR PARTIAL D IFFERENTIAL E QUATIONS
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)
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
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)
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.
u = Λϕ + v, v = (v1 , v2 , v3 ),
(13)
v1 = Ω2 ψ (1) , v2 = −Ω1 ψ (1) + Ω3 ψ (2) , v3 = −Ω2 ψ (2) .
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.
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 ,
where w1 and w2 are arbitrary solutions of the two simpler equations
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
(2014a,b), S. A. Lychev and A. D. Polyanin (2015).
◮ Invariant transformations.
1◦ . The coupled equations (1)–(2) are invariant under the transformation
u + curl Ψ◦ ,
u=e p = pe,
where Ψ◦ is an arbitrary solution of the equation L[Ψ◦ ] = 0.
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
where the functions ψ (1) , ψ (2) , and ϕ satisfy the independent equations
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)
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
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)
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)
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
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.
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,
and the functions ϕ = ϕ(x, t) and p = p(x, t) are determined from the system of equations
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).
u = ∇ϕ + v, p = p, (8)
where the vector function v = (v1 , v2 , v3 ) satisfies the independent linear nonhomogeneous
equation
L[v] = f − ∇G (9)
and the functions ϕ and p are described by the system of equations
Analytical Methods
Chapter 13
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
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)
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.
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.
∂w ∂w
+a = 0. (13.1.1.5)
∂x ∂y
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
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
w = s(y) at x = x0 , (13.1.1.7)
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 .
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).
1126 M ETHODS FOR F IRST-O RDER L INEAR PDE S
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.
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 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 ∂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
ξ = C1 , kξ 2 = C2 . (13.1.2.12)
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.
w = Φ(u1 , u2 , . . . , un−1 ),
dx1 dxn dw
=···= = , (13.2.1.6)
f1 (x1 , . . . , xn ) fn (x1 , . . . , xn ) g(x1 , . . . , xn )w + h(x1 , . . . , xn )
Φ(u1 , u2 , . . . , un ) = 0,
∂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),
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
∂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).
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,
and p
a dy − b − b2 − ac dx = 0, (14.1.1.6)
ϕ(x, y) = C.
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.
ξ = t + z, η =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,
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)
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
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 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
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
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
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).
1140 S ECOND -O RDER L INEAR PDE S : C LASSIFICATION , P ROBLEMS , PARTICULAR S OLUTIONS
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
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
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)
1142 S ECOND -O RDER L INEAR PDE S : C LASSIFICATION , P ROBLEMS , PARTICULAR S OLUTIONS
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.
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
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:
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.
⊙ Literature for Section 14.2: S. G. Mikhlin (1970), V. S. Vladimirov (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. Constanda (2002), A. D. Polyanin (2002),
A. D. Polyanin and A. V. Manzhirov (2007).
1144 S ECOND -O RDER L INEAR PDE S : C LASSIFICATION , P ROBLEMS , PARTICULAR S OLUTIONS
TABLE 14.3
Boundary conditions for various boundary value problems specified by parabolic
and hyperbolic equations in two independent variables (x1 ≤ x ≤ x2 )
w = A1 w1 + A2 w2 + · · · + Ak wk (14.3.1.2)
14.3. Properties and Particular Solutions of Linear Equations 1145
∂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 ∂2w
e ∂k w
e
, , ..., , ...
∂x1 ∂x21 ∂xk1
∂w
e ∂we ∂2w
e ∂2w
e ∂k w
e
, , ..., , , ..., , ...
∂t ∂x1 ∂x22 ∂t∂x1 ∂xk3
∂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)
∂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
14.3. Properties and Particular Solutions of Linear Equations 1147
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).
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.
1148 S ECOND -O RDER L INEAR PDE S : C LASSIFICATION , P ROBLEMS , PARTICULAR S OLUTIONS
TABLE 14.4
Homogeneous linear partial differential equations that admit multiplicative separable solutions
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).
14.3. Properties and Particular Solutions of Linear Equations 1149
where f (x) is an arbitrary infinitely differentiable function. This solution satisfies the initial
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
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).
1150 S ECOND -O RDER L INEAR PDE S : C LASSIFICATION , P ROBLEMS , PARTICULAR S OLUTIONS
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
1152 S ECOND -O RDER L INEAR PDE S : C LASSIFICATION , P ROBLEMS , PARTICULAR S OLUTIONS
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
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
∂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
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:
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.
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.
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. Separation of Variables (Fourier Method) 1157
∂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
1158 S EPARATION OF VARIABLES AND I NTEGRAL T RANSFORM M ETHODS
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:
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.
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.
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:
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
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
1160 S EPARATION OF VARIABLES AND I NTEGRAL T RANSFORM M ETHODS
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)
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
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
with the boundary conditions
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
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”
∂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
1164 S EPARATION OF VARIABLES AND I NTEGRAL T RANSFORM M ETHODS
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
w(x, t) = An sin(βn x) exp − ,
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
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 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
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
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.,
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
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),
TABLE 15.2
Main properties of the Laplace transform
TABLE 15.3
Laplace transforms of some functions
~=w
Finding the transform w ~ (x, p)
Figure 15.2: Solution procedure for linear boundary value problems using the Laplace
transform.
◮ 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 .
15.2. Integral Transform Method 1169
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
with the boundary conditions
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 = w0 at x = 0 (boundary 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
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
1170 S EPARATION OF VARIABLES AND I NTEGRAL T RANSFORM M ETHODS
∂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).
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).
15.2. Integral Transform Method 1171
TABLE 15.4
Main properties of the Fourier transform
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 ).
1172 S EPARATION OF VARIABLES AND I NTEGRAL T RANSFORM M ETHODS
∂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
with the initial conditions
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
15.2. Integral Transform Method 1173
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
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 −∞
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
1174 S EPARATION OF VARIABLES AND I NTEGRAL T RANSFORM M ETHODS
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 → ∞.
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).
15.2. Integral Transform Method 1175
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
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
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)
15.2. Integral Transform Method 1177
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)
with the boundary conditions
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
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
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
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.
15.2. Integral Transform Method 1179
◮ Hankel transform.
1◦ . The Hankel transform is defined as follows:
Z ∞
e
f ν (u) = xJν (ux)f (x) dx, 0 < u < ∞, (15.2.4.4)
0
where ν > − 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→∞
TABLE 15.6
Summary table of integral transforms
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
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
where δ(xk ) is the one-dimensional Dirac delta function, Φ(x) is an arbitrary continuous
function, and dVy = dy1 . . . dyn .
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
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).
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.
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
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|
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
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-
lyanin (2002), A. D. Polyanin and A. V. Manzhirov (2007).
16.2. Representation of the Solution of the Cauchy Problem via the Fundamental Solution 1185
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
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)
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.
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)
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
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
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.)
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
16.2. Representation of the Solution of the Cauchy Problem via the Fundamental Solution 1187
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 ).
◮ 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)
can be represented as
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,
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:
∂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
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
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 .
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
can be represented as the sum
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).
◮ 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 ∂w
L[w] ≡ a 2
+b + Mx [w] = Φ(x, t) (16.2.4.1)
∂t ∂t
with the initial conditions
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).
◮ 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.
16.2. Representation of the Solution of the Cauchy Problem via the Fundamental Solution 1195
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).
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
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| → ∞.
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
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
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),
A. D. Polyanin and A. V. Manzhirov (2007).
Chapter 17
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)
∂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
1199
1200 B OUNDARY VALUE P ROBLEMS . G REEN ’ S F UNCTION
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(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)
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.
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
1202 B OUNDARY VALUE P ROBLEMS . G REEN ’ S F UNCTION
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)
∂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)
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
3◦ . If the conditions
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.
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
∂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
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)
1206 B OUNDARY VALUE P ROBLEMS . G REEN ’ S F UNCTION
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).
17.2. Boundary Value Problems for Hyperbolic Equations with One Space Variable 1207
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
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
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 .
1208 B OUNDARY VALUE P ROBLEMS . G REEN ’ S F UNCTION
The functions Λ1 (x, t) and Λ2 (x, t) that occur in the integrands of the last two terms in
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.
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
ordinary differential equation
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.
The functions Λ1 (x, t) and Λ2 (x, t) that occur in the integrands of the last two terms in
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.
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
∂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
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)
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
1212 B OUNDARY VALUE P ROBLEMS . G REEN ’ S F UNCTION
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).
1214 B OUNDARY VALUE P ROBLEMS . G REEN ’ S F UNCTION
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.
⊙ Literature for Section 17.2: 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), 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).
∂2w ∂2w ∂w
a(x) 2
+ 2
+ b(x) + c(x)w = −Φ(x, y) (17.3.1.1)
∂x ∂y ∂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.
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
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
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
3◦ . If the conditions
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
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
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
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.
⊙ Literature for Section 17.3: 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), 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).
1218 B OUNDARY VALUE P ROBLEMS . G REEN ’ S F UNCTION
w = f (x) at t = 0, (17.4.1.3)
In the general case, Γx is a first-order linear differential operator in the space coordinates
with coefficients depending on x and t.
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
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
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)
1220 B OUNDARY VALUE P ROBLEMS . G REEN ’ S F UNCTION
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
w = f0 (x) at t = 0, (17.4.2.2)
∂t w = f1 (x) at t=0 (17.4.2.3)
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,
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).
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
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).
1222 B OUNDARY VALUE P ROBLEMS . G REEN ’ S F UNCTION
∂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).
TABLE 17.8
Statements of boundary value problems for equations of various types
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.
⊙ Literature for Section 17.4: S. G. Mikhlin (1970), V. S. Vladimirov (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. Constanda (2002), A. D. Polyanin (2002),
A. D. Polyanin and A. V. Manzhirov (2007).
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).
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
w = f (x) at t = 0. (17.5.2.2)
∂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
∂Gm
− Lm [Gm ] = 0 (m = 1, . . . , n)
∂t
with the initial conditions
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.
1226 B OUNDARY VALUE P ROBLEMS . G REEN ’ S F UNCTION
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
∂2w
Lx [w] + = Φ(x, z), (17.5.3.1)
∂z 2
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.
1228 B OUNDARY VALUE P ROBLEMS . G REEN ’ S F UNCTION
w = f (x) at z = 0.
Green’s function:
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:
1
− p .
4π (x1 − y1 ) + (x2 − y2 )2 + (x3 + y3 )2
2
∂z w = f (x) at z = 0.
Green’s function:
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
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.
1230 B OUNDARY VALUE P ROBLEMS . G REEN ’ S F UNCTION
∂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.
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 )
17.5. Construction of the Green’s Functions. General Formulas and Relations 1231
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),
A. D. Polyanin and A. V. Manzhirov (2007).
Chapter 18
Duhamel’s Principles.
Some Transformations
∂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)
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)
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 = ∞.
∂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,
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.
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.
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)
1236 D UHAMEL’ S P RINCIPLES . S OME T RANSFORMATIONS
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
∂t ∂xi ∂xj ∂xi
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.
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-
lyanin (2002), A. D. Polyanin and A. V. Manzhirov (2007).
18.2. Transformations Simplifying Initial and Boundary Conditions 1237
can be reduced to a linear problem with homogeneous boundary conditions. To this end,
one should make the change of variable
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)
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.
1238 D UHAMEL’ S P RINCIPLES . S OME T RANSFORMATIONS
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).
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,
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 )
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
p is the temperature
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;
p is the temperature
Linear viscoelasticity, −(µ0 + λ0 ) div u
11 ρutt − µ0 ∆u Absent
see Eqs. (1) −µ1 ∆ut −(µ1 + λ1 ) div ut
in Section 12.16
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).
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)
div v = 0 (19.1.2.9)
◮ Invariant transformations.
1◦ . Equations (19.1.2.5)–(19.1.2.6) are invariant under the transformation
u + curl Ψ◦ ,
u=e p = pe,
1
u=e
u + ∇Φ, p = pe − (L + K1 ∆)[Φ],
σ
1244 S YSTEMS OF L INEAR C OUPLED PDE S . D ECOMPOSITION M ETHODS
u=e
u + ∇Φ, p = pe + Θ,
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
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).
1246 S YSTEMS OF L INEAR C OUPLED PDE S . D ECOMPOSITION M ETHODS
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
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
where the functions ϕ0 (x) and ϕ1 (x) satisfied the Laplace and Poisson equations
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.
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)
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).
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) ,
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
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),
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
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)
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
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
Here and in what follows, the d’Alembert operators 1 and 2 are defined by
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
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).
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)
1252 S YSTEMS OF L INEAR C OUPLED PDE S . D ECOMPOSITION M ETHODS
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
2 1 [w] = f. (19.3.1.8)
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.
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
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.
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)
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] .
⊙ 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
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
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
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
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
◮ 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.
1258 S OME A SYMPTOTIC R ESULTS AND F ORMULAS
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
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
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
and
e t) = Φ(x
S(p, e 0 (p, t), t). (20.2.1.17)
1260 S OME A SYMPTOTIC R ESULTS AND F ORMULAS
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.
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
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.
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.
1262 S OME A SYMPTOTIC R ESULTS AND F ORMULAS
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
◮ 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∗ ).
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
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
Elements of Theory
of Generalized Functions
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.
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.
2
exp − a for |x| < a,
ϕa (x) = a2 − x2
0 for |x| ≥ a.
1265
1266 E LEMENTS OF T HEORY OF G ENERALIZED F UNCTIONS
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
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
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:
(f ′ , ϕ) = −(f, ϕ′ ). (21.1.4.1)
1268 E LEMENTS OF T HEORY OF G ENERALIZED F UNCTIONS
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
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.
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
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
for some m ≥ 0. Then this function defines a regular functional f ∈ J ′ by formula (21.1.3.1),
where ϕ(x) ∈ J .
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.
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)
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.
1272 E LEMENTS OF T HEORY OF G ENERALIZED F UNCTIONS
∂ |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
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
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
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
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.
1277
1278 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M APLE
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
semicolon. In these chapters, we follow this tradition in every example and problem.
1280 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M APLE
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).
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.
1282 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M APLE
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).
Consider the most important functions for finding all possible analytical solutions of a
given problem for PDEs.
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.
1284 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M APLE
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
is an exact solution of the Poisson equation (with a special right-hand side) as follows:
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:
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:
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:
22.2. Analytical Solutions and Their Visualizations 1285
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);
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;
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
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;
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.
(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);
22.2. Analytical Solutions and Their Visualizations 1287
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);
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.
1288 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M APLE
Table 22.1.
Maple predefined functions and exact solutions of various types of linear PDEs
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-
22.2. Analytical Solutions and Their Visualizations 1289
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.).
where {x ∈ R,t ≥ 0}, we can construct the direction vector fields for the given first-order linear
PDE as follows:
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):
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,
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/xn ) = 0 or u(x, y) = xn g (y/x) as follows:
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).
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
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:
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
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.
The same result can be obtained with the principal part coefficient matrix as follows:
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:
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));
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:
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:
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 ,
u1 (x,t) = f (xet ) et
We show that the implicit form of the solution (or parametric representation of the solution) of this
Cauchy problem has the Maple form
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-
Figure 22.2 Characteristic curves for ut − xux = u and the solution u2 (x,t) (for f (x) = x).
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)
and verify that this solution is an exact solution of the Laplace equation as follows:
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:
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. Analytical Solutions and Their Visualizations 1297
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
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:
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.
22.2. Analytical Solutions and Their Visualizations 1299
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:
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)
1300 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M APLE
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:
⊙ 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).
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)
1302 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M APLE
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;
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
Finally, we verify (symbolically, numerically, and graphically) that this solution satisfies the linear
heat equation and the initial and boundary conditions as follows:
22.3. Analytical Solutions of Mathematical Problems 1303
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
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:
1304 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M APLE
Example 22.28. Linear heat equation. Separable solution. Fourier series. Initial-boundary
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
linear and homogeneous, by the superposition principle, the infinite series
∞ 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. Analytical Solutions of Mathematical Problems 1305
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.
U = C1 e−sx/c + C2 esx/c .
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:
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);
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},
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 ,
Green (1828)].
22.3. Analytical Solutions of Mathematical Problems 1307
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).
1308 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M APLE
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:
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
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
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).
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.
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,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
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);
1312 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M APLE
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)));
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.
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
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)
(to specify BC at the left/right boundary). Numerical BCs are not required. Morton and Mayers (1995)
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)
(to specify BC at the left/right boundary). Numerical BCs are not required. Morton and Mayers (1995)
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)
Numerical BCs are required depending upon the order of the PDE in space. Morton and Mayers (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)
Numerical BCs are required depending upon the order of the PDE in space. Morton and Mayers (1995)
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-
1314 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M APLE
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.
22.4. Numerical Solutions and Their Visualizations 1315
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]):
p4:=Sol1:-plot(u-rhs(SolF),t=0.3,numpoints=NP,color=L1[2]):
p5:=Sol1:-plot(u-rhs(SolF),t=0.5,numpoints=NP,color=L1[3]): display({p3,p4,p5});
22.4. Numerical Solutions and Their Visualizations 1317
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);
Sol1:-animate(u(x,t),x=xR,t=tR,frames=NF,numpoints=NP,thickness=3);
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);
p3:=Sol1:-plot(u-rhs(SolF),t=0.2,numpoints=NP,color=L1[1]):
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,
1318 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M APLE
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:
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
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:
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:
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
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:
1320 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M APLE
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
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)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:
Finally, we compare the approximate numerical solution with the exact solution of this problem,
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:
22.4. Numerical Solutions and Their Visualizations 1321
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)):
for i from 1 to NX1 do
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
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:
od; od;
printf(` i X(i) U(X(i),NT)\n`);
for i from 1 to NX1 do
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:
Such boundary value problems describe a steady-state process u(x, y) in a bounded rectangular
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:
22.4. Numerical Solutions and Their Visualizations 1323
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:
22.4. Numerical Solutions and Their Visualizations 1325
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);
standing wave profile at some time (e.g., t = 20), and visualize the solutions in 3D (GU3D, GV3D) as
follows:
⊙ 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
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
1327
1328 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
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).
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.
23.1. Introduction 1331
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.
1332 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
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]
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:
23.2. Analytical Solutions and Their Visualizations 1333
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).
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.
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.
1334 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
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
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]]]
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
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}
23.2. Analytical Solutions and Their Visualizations 1335
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 .
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)
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).
1336 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
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
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
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,
where {x ∈ R, y ∈ R}, we can prove that the general solution of the given first-order PDE takes the
form y u y
f , n = 0 or u(x, y) = xn g
x x x
as follows:
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).
1338 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
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
where α(ξ, η) = C̃/à and β(ξ, η) = G̃/à with à = 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
{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
the first canonical form and the second canonical form for hyperbolic PDEs, respectively.
Example 23.10. Linear second-order equation. Classification. Canonical forms. Considering
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
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}]
1340 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
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]
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).
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]]]
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
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
The characteristic curves for f (x) = cos x are presented in Fig. 23.1.
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:
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
and verify that this solution satisfies the given Poisson problem (testPDE, testBCs) as follows:
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
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]]}
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.
1346 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
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
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}
23.3. Analytical Solutions of Mathematical Problems 1347
Example 23.18. Linear first-order system. Exact solution. Considering the linear first-order
uncoupled system with two independent variables and two dependent variables,
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).
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
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)
1348 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
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]}
{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
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
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
23.3. Analytical Solutions of Mathematical Problems 1349
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:
Example 23.22. Linear heat equation. Separable solution. Fourier series. Initial-boundary
value problem.
Consider the second-order linear parabolic PDE with the initial and boundary conditions
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
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
1350 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
which satisfies the original heat equation and the boundary conditions. Since the heat equation is
linear and homogeneous, by the superposition principle, the infinite series
∞ nπx 2 kt
u(x,t) = ∑ An sin e−(nπ/L)
n=1 L
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]
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 .
U = C1 esx/c + C2 e−sx/c .
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)
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
l2=Limit[sol1[[1,2]],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}
1352 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
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}
l1=Limit[sol1[[1,1]],x->Infinity,Assumptions->{c>0,s>0}]//Simplify
l2=Limit[sol1[[1,2]],x->Infinity,Assumptions->{c>0,s>0}]//Simplify
{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]
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}:
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 ,
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)].
23.3. Analytical Solutions of Mathematical Problems 1353
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 ]
{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
1354 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
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
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
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}}
23.3. Analytical Solutions of Mathematical Problems 1355
1.0
0.8
0.6
0.5
- 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}]
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}]
1356 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
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).
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.
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.
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:
<<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}]]
1358 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
NDSolve[{PDE,IC,BC},u,{x,x1,x2},{t,t1,t2},Method->{m,subOps}]
Options[NDSolve‘MethodOfLines]
Method->{"MethodOfLines","SpatialDiscretization"->{"TensorProductGrid",
"MinPoints"->val,"MaxPoints"->val,"MaxStepSize"->val,"PrecisionGoal"->val,
"DifferenceOrder"->val}}
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:
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}]
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).
The final FD scheme for the linear heat equation is
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]
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
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}
23.4. Numerical Solutions and Their Visualizations 1361
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}],
PlotStyle->{Blue,Thickness[0.01]},AxesLabel->{"X","U"}];
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,
1362 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
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
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}],
PlotStyle->{Blue,Thickness[0.01]},AxesLabel->{"X","U"}];
grs=N[Table[g[j],{j,0,nNT}]]; ListAnimate[grs]
23.4. Numerical Solutions and Their Visualizations 1363
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
Such boundary value problems describe a steady-state process u(x, y) in a bounded rectangular
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]
1364 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
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:
{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,
23.4. Numerical Solutions and Their Visualizations 1365
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):
{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]
1366 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH M ATHEMATICA
⊙ 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
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.
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
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
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
For linear parabolic and hyperbolic PDEs, the coefficients c, a, f , and d can depend on
time t.
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-
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
24.1. Introduction 1371
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],
24.1. Introduction 1373
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;
1374 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH MATLAB
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).
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)].
sol=pdepe(m,@PDEfun,@ICfun,@BCfun,xMesh,tSpan,ops)
uk=sol(j,:,k); [uOut,DuOutDx]=pdeval(m,xMesh,uk,xOut)
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.
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');
24.2. Numerical Solutions of Linear PDEs 1377
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.
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:
clear all; close all; echo on; format long; m=0; a=0; b=10; Nx=101;
t0=0; tf=10; Nt=101; x=linspace(a,b,Nx); t=linspace(t0,tf,Nt);
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.
24.2. Numerical Solutions of Linear PDEs 1379
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.
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.
Solution regions for drawing solution domains, solving 2D PDE within the solution do-
main, visualizing the results (in 3D).
pdetool starts the PDE Toolbox that invokes the graphical user interface (GUI).
1380 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH MATLAB
R
Draw, Boundary, PDE, Mesh, Solve, Plot are various icons (or modes) in the GUI menu
corresponding to different stages of the solution process.
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 :
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.
24.2. Numerical Solutions of Linear PDEs 1381
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
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
24.2. Numerical Solutions of Linear PDEs 1383
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:
Remark. The M-file LinearPoisson1.m can be observed by type LinearPoisson1 and modified us-
ing MATLAB editor.
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 .
1384 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH MATLAB
R
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:
Example 24.5. Linear Euler equation. Method of characteristics. Cauchy problem. Consider
the initial value problem for the linear Euler equation
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|.
and plot the characteristic curves and the solution (whose graph is an integral surface). See Fig. 24.4.
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:
Remark. It should be noted that the Maple function PDEplot (see Chapter 15) plots solutions of
Cauchy problems for general scalar first-order PDEs.
24.2. Numerical Solutions of Linear PDEs 1385
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
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.
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.
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.
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}:
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.
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).
1388 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH MATLAB
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).
• 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.
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.
• 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).
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}:
where F(x, y) = 2π2 sin(πx) sin(πy) and u(x, y) = sin(πx) sin(πy) is the analytical solution of this
boundary value problem.
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.
24.2. Numerical Solutions of Linear PDEs 1391
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.
5◦ . We specify some parameters for solving a linear PDE and obtain the numerical solution.
• 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).
1392 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH MATLAB
R
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.
x j = xL + ( j − 1)h, tn = t0 + (n − 1)k,
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
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 ,
1394 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH MATLAB
R
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 ).
We construct an approximate numerical solution of the initial-boundary value problem by ap-
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);
hold on; plot(x,u(:,1),'k-',x,u(:,50),'k--',x,u(:,NT),'k:','LineWidth',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);
G=plot(x,u(:,1),'LineWidth',3,'erase','xor');
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]);
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)');
24.3. Constructing Finite Difference Approximations 1395
x j = xL + ( j − 1)h, tn = nk,
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
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(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, 0.049, 0.1');
legend('t=0','t=0.049','t=0.1');
figure(3);
G=plot(x,u(:,1),'LineWidth',3,'erase','xor');
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]);
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)');
⊙ 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).
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
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:
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.
1398 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH MATLAB
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)
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:
clear all; close all; echo on; format long; m=0; a=0; b=1; Nx=31;
t0=0; tf=2; Nt=101; x=linspace(a,b,Nx); t=linspace(t0,tf,Nt);
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,'FontSize',14); set(gca,'FontName','Arial');
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,'FontSize',14); set(gca,'FontName','Arial');
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.
24.4. Numerical Solutions of Systems of Linear PDEs 1399
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
−∇ · (c ⊗ ∇u) + au = f,
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 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.
1400 L INEAR PARTIAL D IFFERENTIAL E QUATIONS WITH MATLAB
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).
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:
• 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.
• 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 ).
• 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
α(α − 1) . . . (α − n + 1)
where Cαn = are binomial coefficients.
n!
1405
1406 E LEMENTARY F UNCTIONS AND T HEIR P ROPERTIES
ax = ex ln a .
The inequality
(
x1 > x2 if a > 1,
ax1 > ax2 ⇐⇒
x1 < x2 if 0 < a < 1.
y = loga x ⇐⇒ x = ay ,
where a > 0, a 6= 1.
Basic properties of the logarithmic function:
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
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
√
Here r = a2 + b2 , sin ϕ = a/r, cos ϕ = b/r, sin ψ = b/r, and cos ψ = a/r.
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)!
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
1410 E LEMENTARY F UNCTIONS AND T HEIR P ROPERTIES
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
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, . . .
π π
arctan(tan x) = x − nπ if nπ − 2 < x < nπ + 2,
arccot(cot x) = x − nπ if nπ < x < (n + 1)π.
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
1414 E LEMENTARY F UNCTIONS AND T HEIR P ROPERTIES
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.
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 = 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
[(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
⊙ 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
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
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
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
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
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
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
1428 F INITE S UMS AND I NFINITE S ERIES
∞
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
∞
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
26.4. Infinite Functional Series 1431
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
∞
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
26.4. Infinite Functional Series 1433
∞
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
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
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
1438 I NDEFINITE AND D EFINITE I NTEGRALS
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
27.1. Indefinite Integrals 1439
◮ 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
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
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 .
⋆ 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)
√ √
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
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)
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
Z
19. x2 ln(x2 + a2 ) dx = 1
3 x3 ln(x2 + a2 ) − 23 x3 + 2a2 x − 2a3 arctan(x/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
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
1450 I NDEFINITE AND D EFINITE I NTEGRALS
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
◮ 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
27.1. Indefinite Integrals 1451
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
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.
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
27.2. Definite Integrals 1455
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, . . .
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
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
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 ∞ 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
1460 I NDEFINITE AND D EFINITE I NTEGRALS
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
a, b > 0 and C(z) and S(z) are Fresnel integrals.
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
a, b > 0 and C(z) and S(z) are Fresnel integrals.
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
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
1463
1464 I NTEGRAL T RANSFORMS
Z ∞
No. Original function, f (x) Laplace transform, fe(p) = e−px f (x) dx
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)
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
1466 I NTEGRAL T RANSFORMS
Z ∞
No. Original function, f (x) Laplace transform, fe(p) = e−px f (x) dx
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
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. Tables of Laplace Transforms 1467
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
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
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
1468 I NTEGRAL T RANSFORMS
Z ∞
No. Original function, f (x) Laplace transform, fe(p) = e−px f (x) dx
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. Tables of Laplace Transforms 1469
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
1470 I NTEGRAL T RANSFORMS
Z ∞
No. Original function, f (x) Laplace transform, fe(p) = e−px f (x) dx
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).
−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
No. Laplace transform, fe(p) Inverse transform, f (x) = epx fe(p) dp
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
1472 I NTEGRAL T RANSFORMS
Z c+i∞
1
No. Laplace transform, fe(p) Inverse transform, f (x) = epx fe(p) dp
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
Z c+i∞
1
No. Laplace transform, fe(p) Inverse transform, f (x) = epx fe(p) dp
2πi c−i∞
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
1474 I NTEGRAL T RANSFORMS
Z c+i∞
1
No. Laplace transform, fe(p) Inverse transform, f (x) = epx fe(p) dp
2πi c−i∞
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
28.2. Tables of Inverse Laplace Transforms 1475
Z c+i∞
1
No. Laplace transform, fe(p) Inverse transform, f (x) = epx fe(p) dp
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)
ak = a cos ϕk , bk = a sin ϕk , ϕk =
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)
ak = a cos ϕk , bk = a sin ϕk , ϕk =
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
ak = a cos ϕk , bk = a sin ϕk , ϕk =
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
1476 I NTEGRAL T RANSFORMS
Z c+i∞
1
No. Laplace transform, fe(p) Inverse transform, f (x) = epx fe(p) dp
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
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
28.2. Tables of Inverse Laplace Transforms 1477
Z c+i∞
1
No. Laplace transform, fe(p) Inverse transform, f (x) = epx fe(p) dp
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)
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
1478 I NTEGRAL T RANSFORMS
Z c+i∞
1
No. Laplace transform, fe(p) Inverse transform, f (x) = epx fe(p) dp
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
No. Laplace transform, fe(p) Inverse transform, f (x) = epx fe(p) dp
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
No. Laplace transform, fe(p) Inverse transform, f (x) = epx fe(p) dp
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
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
28.2. Tables of Inverse Laplace Transforms 1481
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
No. Laplace transform, fe(p) Inverse transform, f (x) = epx fe(p) dp
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).
Z ∞
No. Original function, f (x) Cosine transform, fˇc (u) = f (x) cos(ux) dx
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
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
1484 I NTEGRAL T RANSFORMS
Z ∞
No. Original function, f (x) Cosine transform, fˇc (u) = f (x) cos(ux) dx
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
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| < b
cosh(bx) b cos πab−1 + cosh πb−1 u
1 π sinh a−1 bu
4
cosh(ax) + cos b a sin b sinh πa−1 u
r b2 − u2 abu
1 π
5 exp −ax2 cosh(bx), 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| < b
sinh(bx) 2b cos πab−1 + cosh πb−1 u
1
8 tanh(ax), a>0 ln coth 1
4
πa−1 u
x
28.3. Tables of Fourier Cosine Transforms 1485
Z ∞
No. Original function, f (x) Cosine transform, fˇc (u) = f (x) cos(ux) dx
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)
1 u
1 Ei(−ax) − arctan
u a
0 if 0 < u < a,
2 Ci(ax) π
− 2u if a < u
1 u+a
3 si(ax) − ln , u 6= a
2u u−a
28.3. Tables of Fourier Cosine Transforms 1487
Z ∞
No. Original function, f (x) Cosine transform, fˇc (u) = f (x) cos(ux) dx
0
√ 1 if 0 < u < a,
4 J0 (ax), a>0 a2 − u2
0 if a < u
cos[ν arcsin(u/a)]
√ if 0 < u < a,
a2 − u2
5 Jν (ax), 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 < u < a,
1
6 Jν (ax), a > 0, ν > 0 aν cos(πν/2)
x
√ ν if a < u
ν u + u2 − a2
√ ν−1/2
π a2 − u2
if 0 < u < a,
7 x−ν Jν (ax), a > 0, ν > − 12 (2a)ν Γ ν + 12
0 if a < u
0
√
if 0 < u < a,
xν+1 Jν (ax), 2ν+1 π aν u
8
a > 0, −1 < ν < − 12 Γ −ν − 1 (u2 − a2 ν+3/2 if a < u
2
2
√ 1 a
9 J0 a x , 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 < u < a,
12 J0 a x2 + b2 a2 − u2
0 if a < u
0 if 0 < u < a,
13 Y0 (ax), a>0 1
−√ 2 if a < u
u − a2
0
√
if 0 < u < a,
14 xν Yν (ax), a > 0, |ν| < 1 (2a)ν π
− Γ 1 − ν (u2 − a2 ν+1/2 if a < u
2
2
p π p
15 K0 a x 2 + b2 , a, b > 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).
1488 I NTEGRAL T RANSFORMS
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)
1490 I NTEGRAL T RANSFORMS
Z ∞
No. Original function, f (x) Sine transform, fˇs (u) = f (x) sin(ux) dx
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
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
sin(ax) 1 u + a
1 , a>0 ln
x 2 u−a
(
1
sin(ax) 2
πu if 0 < u < a,
2 , a>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
28.4. Tables of Fourier Sine Transforms 1491
Z ∞
No. Original function, f (x) Sine transform, fˇs (u) = f (x) sin(ux) dx
0
(
sin(ax)
1
2
πb−1 e−ab sinh(bu) if 0 < u < a,
4 , a, b > 0
x 2 + b2 1
2
πb−1 e−bu sinh(ab) if u > a
sin(πx) sin u if 0 < u < π,
5
1 − x2 0 if u > 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 < u < 2a,
1
8 sin2 (ax), a>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 < u < a − b,
1 π
11 sin(ax) sin(bx), a ≥ b > 0 if a − b < u < a + b,
x 4
0 if a + b < u
a √
π a √
12 sin , a>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 < u < a,
cos(ax) 1
π if u = a,
15 , a>0 4
x
1
2
π if a < u
where ξ = u2 − a2
1 arcsin(u/a) if 0 < u < a,
6 J0 (ax), a > 0, ν > 0
x π/2 if a < u
−1
ν sin ν arcsin(u/a) if 0 < u < a,
1
7 Jν (ax), a > 0, ν > −1 aν sin(πν/2)
x
√ ν if a < u
ν u + u2 − a2
0 √ if 0 < u < a,
8 xν Jν (ax), a > 0, −1 < ν < 1 π(2a)ν
2 ν+1/2 if a < u
Γ 12 − ν u2 − a2
!
2u
9 x−1 e−ax J0 (bx), a>0 arcsin p p
(u + b)2 + a2 + (u − b)2 + a2
10
J0 (ax) b−1 sinh(bu)K0 (ab) if 0 < u < a,
, a, b > 0
x 2 + b2 0 if a < u
xJ0 (ax) 0 if 0 < u < a,
11 , a, b > 0
x 2 + b2
1
2
πe−bu I0 (ab) if a < u
√
xJ2n+1/2 (ax)
, (−1)n sinh(bu)K2n+1/2 (ab) if 0 < u < a,
12 x 2 + b2
a, b > 0, n = 0, 1, 2, . . . 0 if a < u
xν Jν (ax)
13
, bν−1 sinh(bu)Kν (ab) if 0 < u < a,
x 2 + b2
a, b > 0, −1 < ν < 5 0 if a < u
2
x1−ν Jν (ax)
, 0 if 0 < u < a,
14 x 2 + b2 1
πb−ν e−bu Iν (ab) if a < u
a, b > 0, ν > − 32 2
2
√ 1 a
15 J0 a x , a>0 cos
u 4u
28.4. Tables of Fourier Sine Transforms 1493
Z ∞
No. Original function, f (x) Sine transform, fˇs (u) = f (x) sin(ux) dx
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 < u < a,
18 Y0 (ax), a>0 √
2 ln u − u2 − a2 − ln a
√ if a < u
π u2 − a2
(
0 if 0 < u < a,
19 Y1 (ax), a>0 u
− √ if a < u
a u2 − a2
√
ln u + u2 + a2 − ln a
20 K0 (ax), 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
Using these formulas, one can express x, y, z in terms of the curvilinear coordinates
x1 , x2 , x3 as follows:
∂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.
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
∗
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
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 .
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
∂f ∂f ∂f
(~v · ∇)f = vx + vy + vz .
∂x ∂y ∂z
Gradient of a vector w
~ along a vector ~v :
Curl of a vector ~v :
∂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
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).
1500 C URVILINEAR C OORDINATES , V ECTORS , O PERATORS , AND D IFFERENTIAL R ELATIONS
Direct and inverse transformations of the components of a vector ~v = vx~i + vy~j + vz~k =
vρ~iρ + vϕ~iϕ + vz~iz :
Curl of a vector ~v :
1 ∂vz ∂vϕ ~ ∂vρ ∂vz ~ 1 ∂(ρvϕ ) ∂vρ ~
curl ~v ≡ ∇ × ~v = − iρ + − iϕ + − iz .
ρ ∂ϕ ∂z ∂z ∂ρ ρ ∂ρ ∂ϕ
Laplacian of a scalar f :
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.
Coordinate surfaces:
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).
∂f ~ 1 ∂f ~ 1 ∂f ~
∇f = ir + iθ + iϕ .
∂r r ∂θ r sin θ ∂ϕ
Divergence of a vector ~v :
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
Gradient of a vector w
~ along a vector ~v :
Curl of a vector ~v :
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
Laplacian of a scalar f :
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
1◦ . Transformations of coordinates:
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
y = x tan ϕ (half-planes through the z-axis).
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
Divergence of a vector ~v :
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 )
Gradient of a vector w
~ along a vector ~v :
Laplacian of a scalar f :
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
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
y = x tan ϕ (half-planes through the z-axis).
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
1504 C URVILINEAR C OORDINATES , V ECTORS , O PERATORS , AND D IFFERENTIAL R ELATIONS
Divergence of a vector ~v :
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 )
Gradient of a scalar f along a vector ~v :
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
Gradient of a vector w
~ along a vector ~v :
~ = (~v · ∇)wσ~iσ + (~v · ∇)wτ~iτ + (~v · ∇)wϕ~iϕ .
(~v · ∇)w
Laplacian of a scalar f :
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
σ2 − τ 2 σ2 − τ 2
gσσ = a2 , gτ τ = a2 , gzz = 1.
σ2 − 1 1 − τ2
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 < ∞).
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.
Coordinate surfaces:
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 )
1506 C URVILINEAR C OORDINATES , V ECTORS , O PERATORS , AND D IFFERENTIAL R ELATIONS
gσσ = gτ τ = σ 2 + τ 2 , gzz = 1.
Laplacian of a scalar f :
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.
x2 + y 2
= 2z + σ 2 (paraboloids of revolution),
σ2
x2 + y 2
= −2z + τ 2 (paraboloids of revolution),
τ2
y = x tan ϕ (half-planes through the z-axis).
gσσ = gτ τ = σ 2 + τ 2 , gϕϕ = σ 2 τ 2 .
Laplacian of a scalar f :
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. Other Special Orthogonal Coordinates 1507
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
y = x tan ϕ (half-planes through the z-axis).
a2 a2 sinh2 τ
gσσ = gτ τ = , gϕϕ = .
(cosh τ − cos σ)2 (cosh τ − cos σ)2
Laplacian of a scalar f :
(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
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!
General case:
Γ(a + 1)
Cab = , where Γ(x) is the gamma function.
Γ(b + 1)Γ(a − b + 1)
Properties:
1509
1510 S PECIAL F UNCTIONS AND T HEIR P ROPERTIES
◮ 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
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
where n = 0, 1, . . .
Numerical values:
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, . . .
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:
∞ ∞
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
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
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
◮ 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
Asymptotic expansion as x → 1:
∞
X lnk x
li(x) = C + ln |ln x| + .
k! k
k=1
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:
◮ Expansions as x → 0 and x → ∞.
Asymptotic expansion as x → ∞:
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
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.
◮ Expansions as x → 0 and x → ∞.
Asymptotic expansion as x → ∞:
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, . . .
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
◮ Expansions as x → 0 and x → ∞.
Asymptotic expansion as x → ∞:
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
Simplest properties:
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
√
Γ(z) = 2π e−z z z−1/2 1 + 1 −1
12 z + 1 −2
288 z + O(z −3 ) (|arg z| < π).
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
n−1
X
ψ(1) = −C, ψ(n) = −C + k−1 (n = 2, 3, . . . ).
k=1
1518 S PECIAL F UNCTIONS AND T HEIR P ROPERTIES
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
∞
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
Definition:
Z 1
B(x, y) = tx−1 (1 − t)y−1 dt,
0
Γ(x)Γ(y)
B(x, y) = .
Γ(x + y)
◮ Some properties.
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:
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
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)
1520 S PECIAL F UNCTIONS AND T HEIR P ROPERTIES
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
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).
x2 yxx
′′
+ xyx′ + (x2 − ν 2 )y = 0
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.
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),
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.
2 2
W (Jν , J−ν ) = − sin(πν), , W (Jν , Yν ) =
πx πx