BEGINNING PARTIAL DIFFERENTIAL EQUATIONS

BEGINNING PARTIAL DIFFERENTIAL EQUATIONS

PETER V. O’NEIL
University of Alabama at Birmingham

A Wiley-Interscience Publication JOHN WILEY & SONS, INC. New York • Chichester • Weinheim • Brisbane • Singapore • Toronto

222 Rosewood Drive. electronic. E-Mail: PERMREQ@WILEY. All rights reserved. without either the prior written permission of the Publisher. Requests to the Publisher for permission should be addressed to the Permissions Department. p. Published simultaneously in Canada. ISBN 0-471-23887-2 (cloth : alk. (978) 750-8400. No part of this publication may be reproduced. John Wiley & Sons.. Inc. or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center. mechanical. QA377. Beginning partial differential equations / Peter V. Library of Congress Cataloging-in-Publication Data: O’Neil.353 — dc21 Printed in the United States of America 10 9 8 7 6 5 4 3 2 1 . New York. I. fax (978) 750-4744. Differential equations. photocopying. MA 01923.This book is printed on acid-free paper. scanning or otherwise. fax (212) 850-6008. (212) 850-6011. stored in a retrieval system or transmitted in any form or by any means. paper) 1. except as permitted under Section 107 or 108 of the 1976 United States Copyright Act.054 1999 98-44621 515 . Copyright 1999 by John Wiley & Sons. Title. Danvers. cm. Peter V. 605 Third Avenue. ‘‘A Wiley-Interscience publication.’’ Includes index. O’Neil. recording. NY 10158-0012. Partial.COM. Inc.

9 v .5. 35 Canonical Form of the Elliptic Equation. 53 The Second Order Cauchy Problem.9.1 2.7 2. 54 Characteristics and the Cauchy Problem. 72 2. 70 2.9.3 Laplace’s Equation. 58 Characteristics as Carriers of Discontinuities.5.1 1. 5 The Significance of Characteristics.1 The General Cauchy Problem. 14 The Quasi-Linear Equation and the Method of Characteristics.4 2.6 2.5 Classification. or the Potential Equation.2 The Cauchy-Kowalevski Theorem. 70 2. 73 29 2.3 An Example.1 The Wave Equation. 20 ix 1 Linear Second Order Partial Differential Equations 2.2 The Heat Equation.2 1. 31 Canonical Form of the Parabolic Equation. 68 The Cauchy-Kowalevski Theorem.4 2 Preliminary Notation and Concepts.8 2.3 1.3 2.9.5. 45 2. 29 Canonical Form of the Hyperbolic Equation. 49 2. 45 2. 1 The Linear Equation.CONTENTS Preface 1 First Order Partial Differential Equations 1.2 2. 39 Canonical Forms and Equations of Mathematical Physics.

6 A Nonhomogeneous Wave Equation.10 A Nonhomogeneous Problem on a Bounded Interval. 155 4 The Wave Equation 158 88 4. 256 4. 251 4.2 Dirichlet’s Formula. 96 3.7 The Complex Fourier Integral and the Fourier Transform. 183 4.3. 213 4. 86 ¨ 3 Elements of Fourier Analysis 3.1 Fourier Cosine and Sine Integrals.12 A Wave Equation in Two Space Dimensions.3.1 Solution by Fourier Transform. 134 3.9. 115 3. 263 .2 d’Alembert’s Solution as a Sum of Forward and Backward Waves. 232 4. 131 3. 158 4.13 Kirchhoff/Poisson Solution in 3-Space and Huygens’ Principle. 200 4.1 Fourier Sine Series. 188 4. 88 3. 148 3.1 The Cauchy Problem and d’Alembert’s Solution.9.6 The Fourier Integral.4 The Wave Equation on a Half-Line.9 Fourier Sine and Cosine Transforms.2 Fourier Cosine Series.3.14 Hadamard’s Method of Descent. 191 4. 248 4. 138 3.3 Convergence of Fourier Series.11 Fourier Integral Solution of the Cauchy Problem.9. 208 4. 102 3.2 The Fourier Series of a Function.3.1 Periodic Functions.9 Fourier Series Solution on a Bounded Interval.1 Why Fourier Series?. 119 3. 98 3. 116 3.vi CONTENTS 2.3.4.4 Convergence of the Fourier Series.5 Bessel’s Inequality and Uniform Convergence.8 Convolution.1. 127 3. 195 4.8 A General Problem on a Bounded Interval.5 A Nonhomogeneous Initial-Boundary Value Problem on a HalfLine. 76 2. 168 4. 178 4.3 Domain of Dependence and the Characteristic Triangle.4 Proof of the Cauchy-Kowalevski Theorem for Linear Systems.6.5 Change of Scale. 108 3.4 Sine and Cosine Expansions.11.1 Verification of the Solution.7 The Wave Equation on a Bounded Interval. 92 3.5 Hormander’s Theorem.3 The Riemann-Lebesgue Lemma.1 An Alternate Derivation of d’Alembert’s Formula. 100 3.4. 96 3. 245 4. 165 4.

3 5. 348 6.5. 351 6.1 Heat Conduction in an Infinite Bar. 312 5.CONTENTS vii 5 The Heat Equation 5. 276 5.1 Green’s Function for a Sphere.4. 359 6.2 A Representation Theorem in R2. 302 5.3 Ends of the Bar at Different Temperatures. 346 6.4 Representation Theorems.14. 287 5.8 Dirichlet Problem for a Rectangle. 403 6. 268 The Weak Maximum Principle.9 Dirichlet Problem for a Disk.2 The Cauchy Problem for Laplace’s Equation Is Ill Posed. 407 . 396 6.1 A Representation Theorem in R3.15.3 A Problem Involving a Nonhomogeneous Boundary Condition.11. 376 6.2 5.2 Bilinear Transformations.3 Some Harmonic Functions. 361 6. 271 Solutions on Bounded Intervals.6 The Nonhomogeneous Heat Equation. 347 6.1 Solution by Fourier Integral. 384 6.2 Harmonic Functions in R3. 384 6. 353 6. 387 6.4. 402 6.4.4. 393 6. 400 6.10 Poisson’s Integral Representation for a Disk.15 Conformal Mapping Techniques.4 Diffusion of Charges in a Transistor.7 The Heat Equation in Several Space Variables.2 Temperature in a Bar With Insulated Ends.5.3.11 Dirichlet Problem for the Upper Half Plane.4. 276 5.7 Uniqueness and Continuous Dependence of Solutions on Boundary Data.6 The Maximum Principle. 323 5.5 Solutions on Unbounded Domains. 370 6. 353 6.1 The Setting of the Problems.11.1 Ends of the Bar at Temperature Zero.2 Solution by Fourier Transform.5.1 Harmonic Functions in the Plane. 267 Initial and Boundary Conditions. 282 5. 302 5. 337 6. 368 6. 321 5. 363 6. 332 267 6 Dirichlet and Neumann Problems 337 6. 293 5.13 Dirichlet Problem for a Cube.12 Dirichlet Problem for the Right Quarter Plane.4 The Cauchy Problem for the Heat Equation Is Ill Posed. 391 6.15.4.14 Green’s Function for a Dirichlet Problem.1 Review of Conformal Mappings. 373 6. 386 6.1 5.14.3.2 The Method of Electrostatic Images.2 Heat Conduction in a Semi-Infinite Bar.5 The Mean Value Property.

444 Neumann Problem for the Upper Half Plane.18 6. 422 Existence of a Solution of the Dirichlet Problem. 451 References for Further Reading.15.16.3 The Idea Underlying an Existence Theorem for the Dirichlet Problem.19 6.3 7. 427 6.2 An Example in the Plane.16. 448 451 Conclusion 7.viii CONTENTS 6.15.15.5 Solution of Dirichlet Problems by Conformal Mapping. 455 Answers to Selected Exercises. 437 Neumann Problem for a Rectangle.4 Historical Perspective.17 6.16. 412 6. 430 6. 420 6.2 7. 432 6. 454 Glossary.1 An Example in 3-Space.1 7.20 7 6. 434 6.16.16 6. 427 6.4 An Integral Solution of the Dirichlet Problem for a Disk.4 Barrier Functions and an Existence Theorem.3 Construction of Conformal Mappings Between Domains. 441 Neumann Problem for a Disk. 435 The Neumann Problem.16.5 Dirichlet’s Principle. 456 Index 497 .

Many of these come with hints for the development of a proof or a derivation. we also point out physical interpretations. Although the emphasis is on the mathematics. canonical forms. Other exercises provide additional information about partial differential equations. for those who wish it. Answers for many of these are included at the end of the book. differential geometry. Some are computational. for those circumstances in which an initial-boundary value problem models a physical setting. more abstract elements of partial differential equations. well-posed problems. ranging from routine to challenging. Sometimes we can visualize a solution as a surface. Partial differential equations invite graphical representation and experimentation. The exercises have two purposes. or analytic number theory. to the modern. or extensions of the material of the text. ix . as well as an entry point. Topics include characteristics. Because this is an introductory treatment. such as the effect of the specific heat of a metal bar on the way it conducts heat energy. We can also experiment with the influence of different parameters on the physical phenomenon represented by the partial differential equation. and both are important for later work in related areas. Further. we attempt a balance between theory and technique. some partial differential equations model physical phenomena. Computational facility and theoretical understanding reinforce each other. as well as techniques for writing expressions for solutions. Our objective is to provide an introduction to this important field of mathematics.PREFACE This book is intended as a first course in partial differential equations. and it is interesting and instructive to couple mathematics with physical intuition by observing the evolution with time of a solution that represents heat conduction or wave motion. such as mathematical physics. and properties of solutions.

and the editorial staff of John Wiley & Sons for their professionalism in bringing the project to completion. Occasional mention is made of uniform convergence. contour integration to evaluate certain integrals. Preliminary versions of this book have been tested at the University of Alabama at Birmingham over the past five years. Readers who do not have access to suitable hardware and software can skip over these exercises. Access to a computational program. Parts of some exercises pursue such issues. Several discussions (the use of conformal mappings to solve the Dirichlet problem. Such access is not a prerequisite to reading this book. and Euler’s formula) use complex numbers and complex function theory. topics from the standard post-calculus course in elementary ordinary differential equations. . vector calculus (theorems of Green and Gauss). but some exercises inviting computer study and experimentation have been included for those who have it. Topics from Fourier analysis are not assumed as background and are included. convergence of Fourier series. ACKNOWLEDGMENTS I would like to thank my colleagues at UAB for helpful conversations and suggestions. several classes of students for tolerating the use of preliminary versions of this book in the form of course notes. such as MAPLE or Mathematica. and other issues of interest in partial differential equations. and convergence of series and improper integrals.x PREFACE or the way the tension and density of a guitar string influence the way it vibrates. such as MAPLE. PREREQUISITES The reader should be familiar with standard properties of real-valued functions of n real variables. is useful both in performing calculations and in studying properties of solutions. and require the use of a computational package.

uxy = 2u/ y x. Such equations are of interest in mathematics (for example.1 PRELIMINARY NOTATION AND CONCEPTS A partial differential equation is an equation that contains at least one partial derivative. In this notion. respectively. y. y w x y z 3 w x4 3 w z u 2 3 w u3 cos(xu) xyzu w =0 x and 2 h x2 2 h y2 2 h = f(x. For example. economics. and other areas. engineering. and so on. There is some economy in using subscripts to denote partial derivatives.1 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 1. uxx = 2u/ x 2. in the study of curvature and surfaces) and in modeling phenomena in the sciences. 1 . u x 4 x u = xuy2. For the moment we will be concerned with developing the vocabulary and notation with which to engage in a discussion of partial differential equations. ecology. The partial differential equations listed above can be written. ux = u/ x. z) z2 are partial differential equations.

As an example of a solution. y). consider the equation uxx 9uyy = 0.1 we obtain 4ux 3uy u= e x/4 f(3x 4y) 12e 12e x/4 f (3x 4y) 4y) e x/4 x/4 f (3x f(3x 4y) = 0 for all (x.2 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS ux wxxxx and hxx hyy 3wuz wuuu xuy = xuy2. and for the independent variables confined to a specified set of values. Often we seek a solution satisfying certain conditions. Upon substitution into equation 1. z). A solution of a partial differential equation is any function that satisfies the equation. y) = e x/4 3uy u=0 (1. the equation 4ux has solution u(x. We will verify this by substituting u(x. (1. As another example. y. cos(xu)wzyx xyzuwx = 0 hzz = f(x.2) . Chain rule differentiations yield ux = = and uy = 4e x/4 1 e 4 1 e 4 x/4 f(3x f(3x 4y) 4y) e x/4 d d(3x x/4 4y) [ f(3x 4y)] d(3x dx 4y) x/4 3e f (3x 4y) f (3x 4y). y) into the partial differential equation.1) f(3x 4y) in which f can be any differentiable function of a single variable.

y.2 is of order two. For example. Equation 1. y. EXERCISE 1 Show that u(x. y) = sin(3x y) e 3x y (3x y)2 cos(3x y). This equation is nonlinear because of the cos(u). we could choose f(t) = sin(t) and g(t) = e t t2 cos(t) to obtain the solution u(x. z) ≠ (0.PRELIMINARY NOTATION AND CONCEPTS 3 It is routine to check that u(x. In view of the latitude in choosing f and g. uxuy and (ux)3 terms. 0.1 is of order one and equation 1. being linear in its second derivative terms uxx and uyy. z) = x is a solution of uxx uyy 1 2 y2 z2 uzz = 0 for (x. x 2uxx yuxy = u is a linear partial differential equation.1). the second order equation uxx 4yuyy (ux)3 uxuy = cos(u) is quasi-linear. The order of a partial differential equation is the order of the highest partial derivative occurring in the equation. For example. . We are now ready to begin studying first order partial differential equations. y) = f(3x y) g(3x y) is a solution for any twice differentiable functions f and g of a single variable.2 has infinitely many solutions (as does equation 1. An equation that is not linear is nonlinear. A partial differential equation is quasi-linear if it is linear in its highest order derivative terms. For example. 0). equation 1. and (uxx)1/2 4uyy = xu is nonlinear because of the (uxx)1/2 term. Of course. A partial differential equation is linear if it is linear in the unknown function and its partial derivatives. any linear equation is also quasi-linear. while x 2uxx yuxy = u2 is nonlinear because of the u2 term.

c is a positive constant. 0) = (x). if p is a continuously differentiable function of one variable. ut(x. y) = ln((x satisfies Laplace’s equation uxx except (x0. EXERCISE 3 Show that 1 ( (x 2 ct) (x ct)) 1 2c x ct u(x. y) of real numbers . then the first order partial differential equation ut = p(u)ux has solution implicitly defined by u(x. y0). t) = f(x ct) g(x ct) is a solution of utt = c2uxx for any twice differentiable functions f and g of one variable. t) = (x p(u)t). EXERCISE 4 Show that. 3. x0)2 (y y0)2) EXERCISE 5 uyy = 0 for all pairs (x. ut ut ut ut ut = = = = = kux. 0) = (x) for all real x. Use this idea to determine (perhaps implicitly) a solution of each of the following equations: 1. in which can be any continuously differentiable function of one variable. with k a nonzero constant uux cos(u)ux euux u sin(u)ux Show that u(x. 2. t) = (s) ds x ct is a solution of utt = c2uxx for any that is twice differentiable and that is differentiable for all real x. 4. c is a positive constant.4 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS EXERCISE 2 Show that u(x. 5. Show that this solution satisfies the conditions u(x.

9. . classify the equation as linear. v is also a solution for any numbers EXERCISE 7 In each of the following. nonlinear but quasi-linear. c. 5. 3. and f are continuous in some region of the plane. 2. or not quasi-linear. 1. in which f is continuous on the real line. 10. y)uy c(x. y). y)uxy c(x. t) when f(x) 1. and that a(x. (1. Hint: Use a change of variables and the standard result that e w2 dw = . Show that ut = kuxx for <x< . 8. y). 6. y)ux e(x. y)u = f(x. t > 0. b. y) and b(x. y)ux b(x. 1. t) = 1 2 kt e (x )2/4kt EXERCISE 8 f( ) d .2 THE LINEAR EQUATION Consider the linear first order partial differential equation in two independent variables: a(x.3) We assume that a. y)uy and g(x. Determine u(x. y) are not both zero for the same (x. y)uyy d(x. y)uxx Show that w Let v and w be solutions of b(x. y)u = 0. u2uxx x 2ux (x (x x 2uyy ux ux uux uxy uy /ux uy = cos(u) y2uy uxy = 2xy 2 y)ux uxy = 1 y)u2 2uy = 4y x yuxx = tan(u) u2 uxx = 4 y uxuy uy = 0 uxy = u2 u2 u2 sin(ux) = 0 x y = x2 Let k be a positive constant. Let u(x.THE LINEAR EQUATION 5 EXERCISE 6 a(x. 7. 4. .

we can. y) in some set of points in the x. b y)w (a x b y)w cw = f. y). Begin the search for a suitable transformation by computing chain rule derivatives: ux = w and uy = w y x w x w y . On . a . then. y) which transforms equation 1.3 to obtain: a(w Write this equation as (a x x w x ) b(w y w y ) cw = f. We will define this transformation in such a way that it is one-to-one. y) so that This is nearly in the form of equation 1. ). Substitute these into equation 1. at least for all (x. The key is to determine a change of variables = (x. = (1. y) in . y-plane. = (x. )w = F( .3. To insure this we will require that the Jacobian of the transformation does not vanish in : J= x x y y = x y y x ≠0 for (x.4) where w( . If y x b y =0 ≠ 0 this requires that x y = b .4 if we choose a for (x.5) (x.6 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS We will show how to solve equation 1. at least in theory. )).3 to the simpler linear equation w h( . y( . solve for x and y as functions of and . ) (1. ) = u(x( . y) in .

One simple choice is = (x.6 is called the characteristic equation of the linear equation 1. y)w c(x. a = (x. or characteristics. Thus far we have found that we can make the coefficient of w in the transformed equation 1. With this choice. the change of variables = x. y).5. y) (1. with c an arbitrary constant.3. Putting (x. We can now choose to suit our convenience and the condition that J ≠ 0. y) = c.5 vanish if we choose = (x.3. y).5 comes very close to the transformed equation 1. then d = implies that dy = dx This means that x y x dx y dy = 0 = b . = (x.3 to a(x.THE LINEAR EQUATION 7 Suppose for the moment that there is such an . The equation (x. dx a(x. Now from equation 1. y) = x. We will say more about these in the next section.6) Equation 1. y) . With this step alone equation 1. y)w = f(x. transforms equation 1. y) is an integral of the ordinary differential equation dy b(x. y) = c an equation defining the general solution of the characteristic equation 1. y) = c defines a family of curves in the plane called characteristic curves. y) = . of equation 1.4 we want to achieve. 1 J= x y 0 = y and this is nonzero in by previous assumption.6. with (x.

) = Example 1 C( . x 1/x = c are the characteristics of this partial differential .8 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS To complete the transformation to the form of equation 1. The characteristic equation is dy b y = = . y). and f (x. ) A( . y) in terms of and to obtain A( . 1 . ) p( . first write a(x. ) Consider the linear equation x 2ux yuy xyu = 1. y). y) = 1.4. ) and F( . This is an integral of the characteristic equation and we choose = (x. ) ≠ 0. restricting the variables to a set in which A( . y) = xy and f (x. y) = ln(y) Graphs of ln(y) equation. We will transform this equation to the simpler equation 1.3 with a(x. y) = x 2. y) = y. ) = .4 with h( . This is equation 1. )w C( . we have w C p w= A A and this is in the form of equation 1.4. )w = p( . y x integrate and rearrange terms to obtain ln(y) 1 =c x for y > 0 and x ≠ 0. ). b(x. c(x. Finally. c(x. A( . dx a x 2 Write 1 1 dy = 2 dx.

Compute ux = w and uy = w w =w 1 1 =w . x x x w y y The partial differential equation transforms to 2 w 1 2 w e 1/ w 1 e 1/ e 1/ w = 1. with w( . = ln(y) 1 . ) = u(x.THE LINEAR EQUATION 9 Upon choosing = x we have the Jacobian J= = 1 ≠0 y y as required. y). = ln(y) so ln(y) = and y=e Now apply the transformation = x. . 1 . Since = x. y e 1/ w =w w 1 x2 =w 1 2 1/ 1 .

in which g is any differentiable function of one variable.4. )d w= F( . . )d . with carried along as a parameter. w) = F( . )d w = F( .7) This is the general solution of the transformed equation (by general solution. )d . )e h( . )e h( . in any region of the . The point to transforming equation 1. )e h( . (1. We obtain e h( . )d w h( .3 to the form of equation 1. )e h( . )d F( . )e h( . ) as a linear first order ordinary differential equation in . -plane with and this has the form of equation 1. Then w 1 e 1/ w= 1 2 . Think of w h( . Following the method for ordinary differential equations. )w = F( . )d d g( )e h( . Since is being carried through this process as a parameter. )d . )d d g( ). )d h( . the constant of integration may depend on .10 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS or 2 w e 1/ w = 1.4 is that we can solve this transformed equation. ) = e h( . Then w( . Integrate with respect to . )d h( . ≠ 0. multiply the differential equation by e to obtain e Recognize this as (e h( .

Example 2 Consider the constant coefficient equation aux buy cu = 0 in which a. y). y). we find by a routine calculation that the partial differential equation transforms to aw or w Multiply this equation by e (c/a)d cw = 0 c w = 0.3 by substituting = (x. Assume that a ≠ 0. b. = bx ay. Now obtain the general solution of the original equation 1. With this transformation. . = (x. y) g( (x. to get c we c /a = 0. a e c /aw hence (e c /aw) = 0. a .THE LINEAR EQUATION 11 we mean one that contains an arbitrary function). and c are numbers. The characteristic equation is dy b = dx a with general solution defined by the equation bx Put = x.8) in which g is any differentiable function of one variable. or e c /a. This general solution will have the form u(x. y) = e (x. ay = c. The characteristics of this differential equation are the straight line graphs of bx ay = c.y) [M(x. (1. y))].

sin(x) = c. cos(x)uy u = xy. transform this solution back in terms of x and y: u(x. Example 3 Consider ux The characteristic equation is dy = cos(x) dx with general solution defined by y Put = x. y) = e cx/a g(bx ay). Now we have y= sin(x) = sin( ) and the partial differential equation transforms to w w= [ sin( )]. Finally. =y sin(x). Observe that the solution in Example 2 has the form specified by equation 1. Graphs of y sin(x) = c are the characteristics of this partial differential equation.8. This solution is readily verified by substitution into the partial differential equation. ) = e c /a g( ). . Then w( .12 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS Integrate with respect to to get e c /aw = g( ) in which g can be any differentiable function of one variable.

then we pick out the unique solution corresponding to the curve passing through (x0. If we require that y(x0) = y0. to obtain we = e e sin( ). the general solution of y d(x)y = p(x) contains an arbitrary constant. each surface corresponding to a choice of the arbitrary function g in equation . which is e .THE LINEAR EQUATION d 13 Multiply this equation by e e w Write this equation as . w( . In the former case.3. By contrast. (we ) = Integrate to obtain e e sin( ). then z = u(x. Graphs of the solutions obtained by making choices of the constant are curves in the (x. y0). Contrast the idea of the general solution for the linear first order ordinary differential equation with that for the linear first order partial differential equation. ) = ( Finally u(x. y) = (y sin(x))(x 1) 1 (sin( ) 2 cos( )) 1 cos( ) 2 e g( ). we = = e( Then e d 1) e sin( ) d 1 e (sin( ) 2 cos( )) 1 e cos( ) 2 g( ). in which g is any differentiable function of a single variable. if u is the general solution of the linear first order partial differential equation 1. y) plane. y) defines a family of surfaces in 3-space. 1) 1 x (sin(x) 2 cos(x)) 1 cos(x) 2 e xg(y sin(x)).

4. 1. 2. 12. 1. . (c) find the general solution of the transformed equation.3 THE SIGNIFICANCE OF CHARACTERISTICS In the preceding section we mentioned characteristics but did not actually do anything with them. Use the general solution to find the solution satisfying u(0. in which n is a nonnegative integer. Now we will investigate their significance. (d) find the general solution of the given equation. and (e) verify the solution by substituting it into the partial differential equation. 9. (a) solve the characteristic equation and sketch graphs of some of the characteristics. 11. 3. EXERCISE 9 For each of the following partial differential equations. 8.14 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 1. 5. y) = yn. 10.4 and obtain the transformed equation.8. In the next section we will investigate the kind of information that should be given in order to pick out one of these surfaces and determine a unique solution. (b) define a transformation of the partial differential equation to the form of equation 1. 6. beginning with an example that will be instructive for the point we want to make. 3ux 5uy xyu = 0 ux uy yu = 0 ux 4uy xu = x 2ux uy yu = 0 xux yuy u=x x 2ux 2uy xu = x 2 ux xuy = 4 x 2ux xyuy xu = x ux uy u=y ux y2uy yu = 0 ux yuy xu = 0 xux yuy 2=0 y EXERCISE 10 Find the general solution of ux (y 1)uy = 1 f(x)(y 2 1)u in which and are real numbers and f is continuous on the real line. 7.

suppose we choose as the x-axis and give values of u(x. and hence does not determine a unique solution. We need u(x. Now suppose we specify values of u(x. y) along a curve in the plane. g(3x 2y) in which g can be any differentiable function defined over the real line.THE SIGNIFICANCE OF CHARACTERISTICS 15 Consider 2ux The characteristic equation is dy 3 = dx 2 and the characteristics are the straight line graphs of 3x 2y = c. we find that this partial differential equation has general solution u(x. Using the method of Section 1. g(t) = e4t/3 sin(t/3). as occurs with ordinary differential equations. 0) = sin(x). Notice that simply specifying that the solution is to have a given value at a particular point does not uniquely determine g. say u(x. 0) = sin(x) on is u(x.2. y) = e =e 4x 4x g(3x) = sin(x) g(3x sin 2y) = e x 4x 4(3x 2y)/3 e sin 1 (3x 3 2y) 8y/3 2 y . 0) = e so g(3x) = e4x sin(x). To be specific for this example. y) at points on . This determines g and the solution satisfying the condition u(x. y) = e 4x 3uy 8u = 0. Putting t = 3x. 3 .

x) = x 4. Despite these two successes.16 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS In this example. x) = e and we can choose g(x) = x 4e 4x to obtain the unique solution u(x. To understand the significance of characteristics in the context of existence 4x g 3x 2 1 (3x 2 1) = x2 . and another choice no solution? The difference was that the x-axis and the line y = x are not characteristics of the partial differential equation. Now we must choose g so that e and this requires that g(1) = e 4xx 2. say u(x. and prescribe values u(x. 1 (3x 2 1) = x 2. x) = x 4. y) is to have along . y) = e 4x 4x g(x) = x 4 g(3x 2y) = e 8(x y)(3x 2y)4 satisfying u(x. Why did some choices of give a solution. y) on this line. This is impossible. hence there is no solution taking the value x 2 at points (x. From the general solution. this requires that u(x. while the line 3x 2y = 1 is a characteristic. not every curve in the plane can be used to determine g. specifying values of u along the x-axis uniquely determined the arbitrary function in the general solution. Next seek a solution having given values along the line y = x. say u x. and hence determined the unique solution of the partial differential equation having these given values. Suppose we choose to be the line 3x 2y = 1.

because we can choose for g any differentiable function such that g(k) = C. y))].y) [M(x.8 of the linear first order partial differential equation 1. . y) g( (x. so equation 1. go back to the general solution 1. if q(x) is of this form for some C.y(x)) [M(x. (1.9) in which C is constant. Suppose we prescribe u(x. y) are determined by the partial differential equation. y(x)) g(k)] [M(x. Let = x. y) = e (x. then there is no solution taking on these prescribed values on . The characteristics are =y x2 e. Example 4 Consider xux 2x 2uy u = x 2e x.10) First we will find the general solution. On the other hand.9 places a constraint on the given data function q(x). Now a characteristic is specified by (x. y) = k. y) = q(x) along a characteristic. (1. The characteristic equation is dy = 2x dx and this has general solution defined by y parabolas. If y = y(x) along this characteristic.y(x)) (x. If q(x) is not of this form for any constant C. y(x)) C]. and are not under our control.THE SIGNIFICANCE OF CHARACTERISTICS 17 and uniqueness of solutions.3. The functions M(x. then q(x) = e or q(x) = e (x. w= 2 x 2 = k. then there are infinitely many such solutions. y) and (x. This general solution is u(x. to obtain w which we write as w 1 w= e.

x 2 4) = xe x xg(4) = xe x x. Integrate with respect to to get 1 w=e g( ) so w= e The general solution of equation 1. We will now attempt to find solutions satisfying given conditions along various curves. y) = xe x xg(y x 2). This equation requires that g(4) = 1. There is no solution satisfying the requested condition. y) = xe x x on the parabola y = x2 4. Now we need u(x. Next suppose we want a solution such that u(x. We will need u(x.10 is u(x. and this is impossible.18 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS ( 1/ )d Multiply this equation by e . which is 1/ . x 2 4) = xe x xg(4) = sin(x). g( ). or 1 w =e. Notice that information is being specified along a characteristic. y) = sin(x) on the curve y = x2 4. We must be able to find a constant C such that xe x Cx = sin(x) for all x. This problem has infinitely many solutions because we can choose g to be any differentiable function of one var- . Suppose first we seek a solution such that u(x. to obtain 1 2 1 w w=e.

and the given information on the curve is called Cauchy data. say. 4x) = xe x xg(4x) = cos(x). the form of the data allows infinitely many solutions. y) = 2x on the ellipse 3y2 2x 2 = 4.3 taking on prescribed values on a given curve is called a Cauchy problem (for the linear equation). . solve the characteristic equation and sketch graphs of some of the characteristics. and attempt to find solutions satisfying the Cauchy data on the given curves. 3yux 2xuy = 0 (a) Find a solution satisfying u(x. 1.THE SIGNIFICANCE OF CHARACTERISTICS 19 iable such that g(4) = 1. Our examples suggest that we can expect a unique solution of a Cauchy problem if the curve is not characteristic. y) = cos(x) along the (noncharacteristic) parabola y = x 2 4x. (c) Find a solution satisfying u(x. (b) Find a solution satisfying u(x. y) = xe x xg(y x 2) cos = xe x 4x y 4 y x2 1 (y 4 x2 x 2)e ( y x 2)/4 . y) = 1 x 2 on the line y = x. Even though data is specified on a characteristic. x 2 This requires that g(4x) = Choose cos(t/4) g(t) = 4 t t t/4 e 4 cos(x) x xe x . t > 0. suppose we want a solution such that u(x. and no solution or infinitely many solutions if the curve is characteristic. The solution of the problem (for x > 0) is u(x. Finally. EXERCISE 11 For each of the following partial differential equations. for. y) = x 2 on the line y = x. Now we need u(x. find the general solution of the partial differential equation. The problem of finding a solution of equation 1.

1). y) = cos(x) on the line y = 3x. y. 6. y) = 1 on the parabola y = x . we can define a curve on the surface u = u(x. u)uy = h(x. y) = sin(x) on the line y = 0. (c) Find a solution satisfying u(x.20 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS 2. Given a solution u(x. and its projection into the x. . y2ux x 2uy = y2 2y on y3 = x 3 2 (a) Find a solution satisfying u(x. y) = e x on the line y = 6x 2. y-plane is (Figure 1. y) = x 3 on the curve 3y2 = 2x 3. In this context the curve in x. (b) Find a solution satisfying u(x. y) whose projection in the x.11) in which we seek a solution u as a function of the independent variables x and y. u) (1. y) = 5. y-plane is called a characteristic trace. y. (c) Find a solution satisfying u(x. y on y = x. y) = x 3 on the line x (b) Find a solution satisfying u(x. (c) Find a solution satisfying u(x. (b) Find a solution satisfying u(x. 1. y-plane. y) = x on y = 4x. 2y2 = 1.4 THE QUASI-LINEAR EQUATION AND THE METHOD OF CHARACTERISTICS Consider the first order quasi-linear partial differential equation f(x. u-space is also called a characteristic. ux 6uy = y (a) Find a solution satisfying u(x. y. This is linear in the highest partial derivatives (which in the first order case are of order one). 4yux uy yu = 0 2y = 3. (a) Find a solution satisfying u(x. (a) Find a solution satisfying u(x. 2 x on the curve y = x 2. y) = y on y = x. u)ux g(x. y) = x on the line y = 2x. y) = 2. y. (b) Find a solution satisfying u(x. 4ux 8uy u=1 4x on the line y = 6x. (c) Find a solution satisfying u(x. but may be nonlinear in u. 2 (b) Find a solution satisfying u(x. yux x 2uy = xy (c) Find a solution satisfying u(x. y) = 3. In the linear case we defined characteristics to be certain curves in the x. y) = 4x on the curve y = (1/3)x 3/2. y) = 1 4. y) = 2 on x (a) Find a solution satisfying u(x. y) of a linear equation and a characteristic .

projecting onto a characteristic For the quasi-linear equation 1.THE QUASI-LINEAR EQUATION 21 FIGURE 1. characteristics are curves in x.1 trace. y0). = g(x. y) is a solution of equation 1. y. y. u). u-space defined by dx dy = f(x. y(t0) = y0. so u0 = (x0. suppose the characteristic has parametric equations x = x(t).12) We will show that a solution u(x. y. . To see why this is true. y = y(t). dt (1. Observation One Suppose u = (x. A characteristic C on a solution surface.11 defining a surface . x(t0) = x0. Then the characteristic passing through P0 lies entirely on . u). u0) is a point on . u = u(t). u(t0) = u0 . We need two observations. y) of a quasi-linear equation may be interpreted as a surface made up of characteristics. y0. This observation can be used to obtain solutions containing given (noncharacteristic) curves. Then for some t0. u). y.11. dt dt du = h(x. and that P0 :(x0. and hence provides a way of solving the Cauchy problem when the partial differential equation is quasi-linear.

y). dx ds dy =f ds g x y x y . u). k dx dt dy dt y x y x f(x.11. u) g(x. we can use equations 1.12 to compute dt (x(t). y. u). dt Observation Two If we begin with an arbitrary (but noncharacteristic) curve and construct the family of characteristics passing through points of . u) = h(x. u) on . y. Suppose is parameterized by x = x(s). y(t)) = dt = Therefore u(t) = (x(t). then the resulting surface is the graph of a solution of the partial differential equation. u) = du . y = y(s). To see why this is true. y. But P0 is on the surface. = h(x.22 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS Because this curve is characteristic. u) = ds so is a solution. = g(x. y(t)) and the characteristic lies on . y. as in Figure 1. Then du = h(x. y. y. u) ds ds ds because the surface is made up of characteristics. dx dy du = f(x. At any (x. We want to show that is a solution of equation 1. so (x(t0). y. y(t)) for some constant k. z = z(s). Therefore u(t) = (x(t). y.2. assume that is the graph of u = (x. y(t0)) = u0 = u(t0) implies that k = 0.

not a surface. then the characteristic through each point of C would be just C itself. If we did so. Here are two illustrations of the method of characteristics. and this surface is the graph of the solution of the Cauchy problem. A solution surface formed by constructing characteristics through points These observations suggest the method of characteristics for solving the Cauchy problem when the partial differential equation is quasi-linear. Construct the characteristic through each point of . This strategy also suggests why we do not want to specify data along a characteristic C. = dt dt x. dt . u = 0.THE QUASI-LINEAR EQUATION 23 FIGURE 1. sin(x)) = 0. The characteristics of this partial differential equation are specified by dx dy = y.11 assuming prescribed values on a given curve that is not characteristic. Suppose we want the solution of equation 1. This means that we u(x. Example 5 We want the solution of yux that passes through the curve require xuy = e u given by y = sin(x).2 of . du = e u. This defines a surface in three space.

y sin(t) . y = sin(s)cos(t) s sin(t).24 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS From the first two of these equations we can write dy = dx or ydy with general solution (in terms of t) x = a cos(t) b sin(t). b = sin(s). Let P :(s. We use s as parameter on to distinguish between points on and points on characteristics. u = 0. and e 0 = 1 = 0 c. Now eliminate t and s from these equations. xdx = 0. We want to construct a characteristic through each point of (as in Figure 1. From du/dt = e u we obtain e u =t c. y = sin(s). From the first two equations. giving us a = s. b sin(t).2). e u =1 t. x . the characteristic intersecting at P has parametric equations x = s cos(t) sin(s)sin(t). and c = 1 at this point of intersection of with P. y = b cos(t) a sin(t). Therefore. Then at t = 0. The characteristics therefore have parametric representation x = a cos(t) Parameterize as x = s. sin(s). 0) be a point of and suppose a characteristic passes through P when t = 0 (this is just a scaling of the parameter). y with a and b constant. y = b = sin(s). s = x cos(t) and sin(s) = y cos(t) x sin(t). x = a = s. y = b cos(t) a sin(t). e u =c t.

We would like the solution passing through :x = s 2.13) =1 t implies that t = 1 e u) e u. Solution of yux xuy = e u containing the curve y = sin(x). = y. u = 0 satisfies this equation.3(b) shows the same surface from a different perspective. = sec(u). u = 0 . As in Example 5. (1. Figure 1.3(a) (Example 5). It is easy to check that y = sin(x). Example 6 Consider the quasi-linear equation xux yuy = sec(u). sin(u) = t c. we want to construct the characteristic through each point FIGURE 1. u = 0. y = sin(s). The characteristics satisfy dx dy du = x. y = Be t. and Figure 1.13) e u) e u)). dt dt dt which have solutions defined by x = Ae t. sin(x cos(1 y sin(1 e u)) = y cos(1 x sin(1 This equation implicitly defines the solution of the Cauchy problem.3(a) shows a graph of part of the surface (solution).THE QUASI-LINEAR EQUATION 25 Therefore sin(x cos(t) But e to get u y sin(t)) = y cos(t) x sin(t). Substitute this into equation (1.

3(b) (Example 5). We want to eliminate s and t from these equations.3(a) of . sin(s). then y = sin(s)e sin(u) so sin(s) = ye and s = arcsin(ye sin(u) sin(u) ). sin(0) = 0 = 0 so A = s 2. 0) at t = 0. y = B = sin(s).26 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS FIGURE 1. Another perspective of the solution surface shown in Figure 1. Then x = s 2e t. y = sin(s)e t and sin(u) = t. B = sin(s) and c = 0 at this point. c. Since t = sin(u). x = A = s 2. . Suppose a characteristic passes through Then at P :(s 2.

u = 0. 2. 6. yuy = u.4(a) Graph of the curve x = t 2.4(a) shows a graph of . xux ux ux ux ux ux ux x 3ux 9. xux FIGURE 1. u = 2. is given by y = 1 x. u = 0. xuy = 4. 8. 7. 4. u = 0. is given by y = x 2 2. is the curve defined by y = x 3. Figure 1. use the method of characteristics to find a solution of the partial differential equation that passes through the given curve . 2 y uy = cos(u). Graph part of the solution surface. 1. 3 y uy = sec(u).4(b) part of the solution surface.THE QUASI-LINEAR EQUATION 27 Then x = e sin(u)[arcsin(ye sin(u) )]2. is given by y = 1 is given by y = x x 2. 5. is given by y = 2x 1. 10. u = 0. u = 0. is given by x = y2. 3. u = 0. yuy = u. u = 4. and Figure 1. EXERCISE 12 For each of the following. u = 1. This equation implicitly defines u(x. ux yuy = sec(u). 1. u = 0 (Example 6). . y 2uy = 1. is given by y = x 2. u = 0. is given by x = y2 1. the solution surface contains the data curve . y) such that u(s 2. y = sin(t). uy = e u. sin(s)) = 0—that is. is given by y = 4x. uy = u2. y2uy = u. The solution may be implicitly defined.

Next use the idea of Exercise 4 to solve this problem. Solution of xux yuy = sec(u) containing the curve shown in Figure EXERCISE 13 of the problem Use the method of characteristics to show that the solution uux uy = 0 u(x. Compare these solutions.28 FIRST ORDER PARTIAL DIFFERENTIAL EQUATIONS FIGURE 1. 0) = f(x) is defined implicitly by the equation u(x. y = s. Hint: Think of as the curve x = s. y) = f(x u(x. y)y). u = f (s).4(a). .4(b) 1.

Sign up to vote on this title
UsefulNot useful