SCHAUM’S OUTLINE SERIES THEORY AND PROBLEMS OF FOURIER ANALYSIS with applications to BOUNDARY VALUE PROBLEMS UT a el ea INCLUDING 205 SOLVED PROBLEMS SCHAUM’S OUTLINE SERIES McGRAW-HILL BOOK COMPANYSCHAUM’S OUTLINE OF THEORY AND PROBLEMS of FOURIER ANALYSIS with Applications to Boundary Value Problems by MURRAY R. SPIEGEL, Ph.D. Former Professor and Chairman of Mathematics Rensselaer Polytechnic Institute of Connecticut eit whe SCHAUM’S OUTLINE SERIES ‘McGRAW-HILL BOOK COMPANY New York, St, Louis, San Francisco, Diisseldorf, Johannesburg, Kuala Lumpur, London, Mexico, Montreal, New Delhi, Panama, Sao Paulo, Singapore, Sydney, and TorontoCopyright © 1974 by McGraw-Hill, Ine. All rights reserved. Printed in the United States of America. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of the publisher. 0-07-060219-0 3456789101112181415 SH SH 79876 Library of Congress Catalog Card Spiegel, Murray R, Schaum's outline of theory and problems of Fourier analysis. (Schaum's outline series) 1, Fourier analysis. 2. Boundary value problems. 1. Title, II. Title: Theory and problems of Fourier analysis. (QA403.5.866] s16/ 2433, 173-21520Preface In the early years of the 19th century the French mathematician J, B, J. Fourier in his researches on heat conduction was led to the remarkable discovery of certain trigonometric series which now bear his name. Since that time Fourier series, and generalizations to Fourier integrals and orthogonal series, have become an essential part of the background of scientists, engineers and mathematicians from both an ap- plied and theoretical point of view. ‘The purpose of this book is to present the fundamental concepts and applications of Fourier series, Fourier integrals and orthogonal functions (Bessel, Legendre, Hermite, and Laguerre functions, as well as others). ‘The book is designed to be used either as a textbook for a formal course in Fourier Analysis or as a comprehensive supplement to all current standard texts. It should be of considerable value to those taking courses in engineering, science or mathematics in which these important methods are frequently used. It should also prove useful as a book of reference to research workers employing Fourier methods or to those inter- ested in the field for self-study. Each chapter begins with a clear statement of pertinent definitions, principles and theorems, together with illustrative and other descriptive material. The solved prob- lems serve to illustrate and amplify the theory and to provide the repetition of basic principles so vital to effective learning. Numerous proofs of theorems and derivations of formulas are included among the solved problems. The large number of supple- mentary problems with answers serve as a complete review of the material of each chapter. Considerably more material has been included here than can be covered in most first courses. This has been done to make the book more flexible, to provide a more useful book of reference, and to stimulate further interest in the topics. I wish to take this opportunity to thank Henry Hayden and David Beckwith for their splendid cooperation. M. R. SPIEGEL January 1974Chapter CONTENTS BOUNDARY VALUE PROBLEMS ceeeceee Mathematical Formulation and Solution of Physical Problems. Definitions Per- taining to Partial Differential Equations. Linear Partial Differential Equa- tions. Some Important Partial Differential Equations. The Laplacian in Differ- ‘ent Coordinate Systems. Methods of Solving Boundary Value Problems. Page 1 Chapter FOURIER SERIES AND APPLICATIONS .... 5 ‘The Need for Fourier Series. Periodie Functions. Piecewise Continuous Func tions. Definition of Fourier Series, Dirichlet Conditions, Odd and Even Functions. Half-Range Fourier Sine or Cosine Series, Parseval’s Identity. Uniform Convergence. Integration and Differentiation of Fourier Series, Com= plex Notation for Fourier Series, Double Fourier Series. Applications of Fourier Series. Chapter ORTHOGONAL FUNCTIONS ..... cee pee Definitions Involving Orthogonal Functions. Orthonormal Sets. Orthogonality with Respect to a Weight Function, Expansion of Functions in Orthonormal Series. Approximations in the Least-Squares Sense. Parseval’s Identity for Orthonormal Series. Completeness, Sturm-Liouville Systems. Eigenvalues and Eigenfunctions. The Gram-Schmidt Orthonormalization Process. Applications to Boundary Value Prob Chapter GAMMA, BETA AND OTHER SPECIAL FUNCTIONS .. Special Functions. The Gamma Function. Table of Values and Graph of the Gamma Funetion. Asymptotic Formula for I(n). Miscellaneous Results Involv- ing the Gamma Function. The Beta Function. Other Special Functions Asymptotic Series or Expansions. Chapter FOURIER INTEGRALS AND APPLICATIONS . 300 ‘The Need for Fourier Integrals. The Fourier Integral. Equivalent Forms of Fourier's Integral Theorem. Fourier Transforms. Fourier Sine and Cosine Transforms. Parseval's Identities for Fourier Integrals, The Convolution ‘Theorem for Fourier Transforms. Applications of Fourier Integrals and Trans- forms. Chapter BESSEL FUNCTIONS AND APPLICATIONS gq00n Bessel’s Differential Equation. The Method of Frobenius. Bessel Functions of the First Kind. Bessel Functions of the Second Kind. Generating Function for Jal). Recurrence Formulas. Functions Related to Bessel Functions, “Equa tions Transformable into Bessel’s Equation. Asymptotic Formulas for Bessel Functions. Zeros of Bessel Functions. Orthogonality of Bessel Functions of the First Kind. Series of Bessel Functions of the First Kind. Orthogonality and Series of Bessel Functions of the Second Kind. Solutions to Boundary Value Problems Using Bessel Functions. 7CONTENTS Page Chapter 7 LEGENDRE FUNCTIONS AND APPLICATIONS ...... 130 Legendre’s Differential Equation. Legendre Polynomials, Generating Function for Legendre Polynomials. Recurrence Formulas. Legendre Fanctions of the Second Kind, Orthogonality of Legendre Polynomials. Series of Legendre Polynomials. Associated Legendre Functions. Orthogonality of Associated Legendre Functions. Solutions,to Boundary. Value Problems Using Legendre Funetio Chapter § HERMITE, LAGUERRE AND OTHER ORTHOGONAL POLYNOMIALS ................. 154 Hermite's Differential Equation. Hermite Polynomials, Generating Function for Hermite Polynomials. Recurrence Formblas for Hermite Polynomials. Orthogonality of Hermite Polynomials. Series of Hermite Polynomials. La- ‘guerre's Differential Equation, Laguerre Polynomials, Some Important Prop- erties of Laguerre Polynomials. Miscellaneous Orthogonal Polynomials and ‘Their Properties. Appendix A. Uniqueness of Solutions . 167 Appendix BB Special Fourier Series ............ Sees 169 Appendix C Special Fourier Transforms 173 Appendix D) Tables of Values for Jo() and J:(e) 16 Appendix EZ Zeros of Bessel Functions... ut ANSWERS TO SUPPLEMENTARY PROBLEMS . ve 1D INDEX - 187Chapter 1 Boundary Value Problems MATHEMATICAL FORMULATION AND SOLUTION OF PHYSICAL PROBLEMS In solving problems of science and engineering the following steps are generally taken. 1, Mathematical formulation. To achieve such formulation we usually adopt mathematical ‘models which serve to approximate the real objects under investigation. Example 1. ‘To investigate the motion of the earth or other planet about the sun we can choose points as mathe- matical models of the sun and earth. On the other hand, if we wish to investigate the motion of the earth about its axis, the mathematical model cannot be a point but might be a sphere or even more accu- rately an ellipsoid. In the mathematical formulation we use known physical laws to set up equations describing the problem, If the laws are unknown we may even be led to set up experi- ‘ments in order to discover them. Example 2. In describing the motion of a planet about the sun we use Newton's laws to arrive at a differential ‘equation involving the distance of the planet from the sun at any time. 2 Mathematical solution. Once a problem has been successfully formulated in terms of equations, we need to solve them for the unknowns involved, subject to the various conditions which are given or implied in the physical problem. One important con- sideration is whether such solutions actually exist and, if they do exist, whether they are unique. In the attempt to find solutions, the need for new kinds of mathematical analysis — leading to new mathematical problems — may arise. Example 3. J.B. Fourier, in attempting to solve a problem in heat flow which he had formulated in terms of partial differential equations, was Ied to the mathematical problem of expansion of functions into series involving sines and cosines. ‘Such series, now called Fourier series, are of interest from the point of view of mathematical theory and in physical applications, as we shall see in Chapter 2. 3. Physical interpretation. After a solution has been obtained, it is useful to interpret it physically. Such interpretations may be of value in suggesting other kinds of problems, which could lead to new knowledge of a mathematical or physical nature. In this book we shall be mainly concerned with the mathematical formulation of physi- cal problems in terms of partial differential equations and with the solution of such equations by methods commonly called Fourier methods.2 BOUNDARY VALUE PROBLEMS [CHAP.1 ‘ DEFINITIONS PERTAINING TO PARTIAL DIFFERENTIAL EQUATIONS A partial differential equation is an equation containing an unknown function of two or more variables and its partial derivatives with respect to these variables. The order of a partial differential equation is the order of the highest derivative present. Example 4. au de ay ‘equation. Here w is the dependent variable while » and y are independent variables. = 22—y is a partial differential equation of order two, or a second-order partial differential A solution of a partial differential equation is any function which satisfies the equation identically. The general solution is a solution which contains a number of arbitrary independent functions equal to the order of the equation. A particular solution is one which can be obtained from the general solution by particu- lar choice of the arbitrary functions. Bxample 5. As seen by substitution, u = zty~J2y?+F(e) + G(y) is a solution of the partial differential equation of Example 4, Because it contains two arbitrary independent functions F(z) and G(y), it is the general solution. If in particular Fiz) =2 sin, Gly) = 8yt—6, we obtain the particular solution w= sty ~ fay + 2sine + 3A — 5 A singular solution is one which cannot be obtained from the general solution by par- ticular choice of the arbitrary functions. au _ (ow (ie -F(y)—[Ft)}? and n= 2/4 are solutions, The first is the general solution invelving one arbitrary function. Au ae second, which cannot be obtained from the ‘general solution by any choice of F(y), i'n singular solution. A boundary value problem involving a partial differential equation seeks all solutions of the equation which satisfy conditions called boundary conditions, Theorems relating to the existence and uniqueness of such solutions are called existence and uniqueness theorems. LINEAR PARTIAL DIFFERENTIAL EQUATIONS ‘The general linear partial differential equation of order two in two independent vari- ables has the form eu au au pat, ya aA Aap + Baggy + Che + D5 + BG, + Pu = 6 () where A,B,...,@ may depend on x and y but not on u. A second-order equation with independent variables x and y which does not have the “orm (1) is called nonlinear. If G = 0 identically the equation is called homogeneous, while if G0 it is called non- homogeneous. Generalizations to higher-order equations are easily made. Because of the nature of the solutions of (1), the equation is often classified as elliptic, hyperbolic, or parabolic according as B*—4AC is less than, greater than, or equal to zero, respectively.CHAP. 1] BOUNDARY VALU! PROBLEMS 8 SOME IMPORTANT PARTIAL DIFFERENTIAL EQUATIONS 1. Vibrating string equation This equation is applicable to the small transverse vibrations of a taut, flexible string, such as a violin string, initially located on the ¥ xis and set into motion (see Fig. 1-1). The function y(z,t) is the displacement of any point « of the string at time f. The constant ve) z a? = Ju, where + is the (constant) tension in the string and y is the (constant) mass per unit length of the string. It is assumed that no external forces act on the string and that, it vibrates only due to its elasticity. ‘The eqitation can easily be generalized to higher dimensions, as for example the vibrations of a membrane or drumhead in two dimensions. ™n two dimensions, the equation is ae »(2 , Be - o(S5+) ow at Here w(x, y,z,t) is the temperature at position (:,y, 2) in a solid at time ¢. The con- stant x, called the diffusivity, is equal to K/on, where the thermal conductivity K, the specific heat « and the density (mass per unit volume) ». are assumed constant. We call V'u the Laplacian of u; it is given in three-dimensional rectangular coordinates (4,2) by 2. Heat conduction equation «Vu au eu ay oF 3. Laplace's equation This equation occurs in many fields. In the theory of heat conduetion, for example, v is the steady-state temperature, ie. the temperature after a long time has elapsed, whose equation is obtained by putting au/at=0 in the heat conduction equation above. In the theory of gravitation or electricity v represents the gravitational or electric potential respectively. For this reason the equation is often called the potential equation. ‘The problem of solving Vv =0 inside a region R when v is some given function on the boundary of & is often called a Dirichlet problem, an 4, Longitudinal vibrations of a beam eT ‘This equation describes the motion of a beam (Fig. 1-2, page 4) which can vibrate longitudinally (i.e. in the x-direction) the vibrations being assumed small. The variable ‘u(x, t) is the longitudinal displacement from the equilibrium position of the cross section at #. The constant c* = E/s, where E' is the modulus of elasticity (stress divided by strain) and depends on the properties of the beam, »: is the density (mass per unit volume). Note that this equation is the same as that for a vibrating string.4 BOUNDARY VALUE PROBLEMS (CHAP. 1 ey 5. Transverse vibrations of a beam 3 ay +uth = 0 ‘This equation describes the motion of a beam (initially located ‘on the «-axis, see Fig. 1-3) which is vibrating transversely (i.e. perpendicular to the x-direction) assuming small vibrations. In this case y(z, 1) is the transverse displacement or deflection at any time t of any point x. The constant b*=EI/Ay, where B is the modulus of elasticity, J is the moment of inertia of any cross section about the z-axis, A is the area of cross section and y is the mass per unit length. In case an external transverse force F(z, t) is applied, the right-hand side of the equation is replaced by b°F (x, t)/EI. ey « Fig. 1-2 Fig.1-3 THE LAPLACIAN IN DIFFERENT COORDINATE SYSTEMS ‘The Laplacian Vu often arises in partial differential equations of science and engi- neering. Depending on the type of problem involved, the choice of coordinate system may be important in obtaining solutions. For example, if the problem involves a cylinder, it~ will often be convenient to use cylindrical coordinates; while if it involves a sphere, it will: be convenient to use spherical coordinates. ‘The Laplacian in cylindrical coordinates (p, #,2) (see Fig. 1-4) is given by eu leu ty = Fu, Law @ s af * pap A ‘The transformation equations between rectangular and cylindrical coordinates are = peoss, y= psing, #=% 6 where p20, 0 $<21, —» 0 also}, we have rafay\ _ ey my _ ey o- z a(#) = $4 or SE = att, where at = ru Write the boundary conditions for a vibrating string of length L for which (a) the ends ¢=0 and z=L are fixed, (b) the initial shape is given by f(a), (c) the initial velocity distribution is given by g(x), (d) the displacement at any point x at time t is bounded. (a) If the string is fixed at z= 0 and must he zero for all times t > 0, ie. 0,1) = 0, WL.) ~ 0 to = L, then the displacement y(x,) at 2 =0 and 2=L () Since the string has an initial shape given by f(z), we must have wle,0) = fe) o0 Write boundary conditions for a vibrating string for which (a) the end 2=0 is moving so that its displacement is given in terms of time by G(t), (b) the end x =L is not fixed but is free to move. (a) The displacement at x= 0 is given hy y(0,4). ‘Thus we have 0,0) = Ge t>o0CHAP. 1) BOUNDARY VALUE PROBLEMS 1 (8) If + is the tension, the transverse force acting at any point x is ue SL = neo Since the end x= L is free to move so that there is no force acting on it, the boundary condition is given by rust) = 0 or uallst) = 0 to 14. Suppose that in Problem 1.1 the tension in the string is variable, i.e. depends on the particular point taken. Denoting this tension by z(z), show that the equation for the vibrating string is fant ay alee] = ae In this case we write (2) of Problem 1.1 as eu) ny ON ee levar eles ‘so that the corresponding equation (4) is ~ ate) so OL oy . wae oe . ‘Thus, taking the limit as 4x0 (in which ease ¢0), we obtain a ay] ey aon] = after multiplying by 1.5. Show that the heat flux across a plane in a conducting medium is given by -Ke, where w is the temperature, m is a normal in a direction perpendicular to the plane and K is the thermal conductivity of the medium. Suppose we have two parallel planes I and II a dis- 1 ou tance An apart (Fig. 1-7), having temperatures u and us au, respectively. Then the heat flows from the plane of higher temperature to the plane of lower temperature Also, the amount of heat per unit area per unit time, called the heat flux, is directly proportional to the difference in ‘temperature Au and inversely proportional to the distance 4m. Thus we have « utau Meat tax trom toi = -K wy where K js the constant of proportionality, ealled the ther- ‘mat conductivity. The minus sign occars in (1) since if Au > 0 the heat flow actually takes place from II to I. 11 By taking the limit of (1) as am and thus Au approaches zero, we have as requir Moat fax sero plane = KE We somatine call the pan of which in vstor form is so that 2 can be wet Heat flux across plane 1 = -K'Vu oy 16, If the temperature at any point (¢,y,2) of a solid at time ¢ is u(z,y,z,t) and if K,o and » are respectively the thermal conductivity, specific heat and density of the solid, all assumed constant, show thatBOUNDARY VALUE PROBLEMS (CHAP. 1 9 = WV! where « = Kop Consider smal volume ment of he ald Va indicated in Fig. 1-8 and greatly enlarged tn Pip Lo By Proom 1 the tmnt of het pr Unltaren per unit Une entering the element throaph face PORS ie — 24, sted atthe poston x sie the aren of face PQRS i ay, th lta amount of het entering the element through face PQRS in time At is aes | Reca cece aeracainer earns os « Similarly, the amount of heat leaving the element through face NW2T is ax yaeae | cay indicates the derivative of w with respect to evaluated at 2+ Az. ‘The amount of heat which remains in the element is given by the amount entering minus the amount leaving, which is, from (1) and (2), {sal Bileae ~ AB) fovea ® In @ similar way we can show that the amounts of heat remaining in the clement due to heat transfer taking place in the y- and z-directions are given by el it Larne eEplsan ~ Hf esese ® snd {elu — #3 bocavar © respectively. ‘The total amount of heat gained by the element is given by the sum of (8), (4) and (6). This amount of heat serves to raise its temperature by the amount Au. Now, we know that the heat needed to raise the temperature of a mass m by Au ia given by ma Ax, where o is the specific heat. If the density of the solid is «, the mass is m= Ae Ay Az. Thus the quantity of heat given by the sum of (8), (4) and (6) is equal to on ar ay Az au @ If we now equate the sum of (8), (b) and (6) to (8), and divide by ax Ay Az At, we find = 1 S m xi] ae). pete — xe x Beleeae 7 ESL 4 [#Eheay —KSL 1 = opi az ay ar In hn Ln aay, 48 and ¢ al approach aro the above equation becomes 2 (at) . 2 (ga). 2(gm) jo (et) + §(kH) + (ee) = ot ” a ott Thin can be rewsiton as a % 2 9m wine «= E incated te pettyonAP. 1] BOUNDARY VALUE PROBLEMS 9 a. 18. Fig. 1-8 Fig.1-9 Work Problem 1.6 by using vector methods. Let V be an arbitrary volume lying within the solid, and let $ denote ite eurface (see Fig. 1-8), ‘The total flax of heat across S, or the quantity of heat leaving S per unit time, is Sf cxcices , where m is an outward-drawn unit normal to S. Thus the quantity of heat entering S per unit time is (iene (Wieteeee, fs by the divergence theorem. The heat contained in a volume V is given by Sffomav ‘Then the time rate of increase of heat is M4 afflomar = fff otter 7 ‘Equating the right-hand sides of (1) and (2), SUS [og 9-0] = 0 and since V is arbitrary, the integrand, assumed continuous, inust be identically zero, so that ee gs, . nt = VKIW) or if Kyo, are constants, K oo Bveve = nv @) a Show that for steady-state heat flow the heat conduction equation of Problem 1.6 or 1.7 reduces to Laplace's equation, y’u=0. . In the case of steady-state heat flow the temperature u does not depend on time ¢, s0 that Tins the equation 22 = V4 tacomes Vou = 0 du A thin bar of diffusivity « has its ends at 2=0 and x=L on the x-axis (see Fig. 1-10). Its lateral surface is insulated so that heat cannot enter or escape.10 BOUNDARY VALUE PROBLEMS [cHAP.1 (a) If the initial temperature is f(x) and the ends are kept at temperature zero, set up the boundary value problem. (b) Work part (a) if the end z= is insulated. (c) Work part (a) if the end = radiates into the surrounding medium, which is assumed to be at temperature wo, ‘a problem in one-dimensional heat. con- e the temperature can only depend on the position 2 at any time t and can thus be de- noted by ulz, 4). ‘The heat conduction equation is thus given by ue mT ae o<20 i) Fig.1410 (a) Since the ends are kept at temperature zero, we have (0,8) = 0, w(b,t) = 0 t>0 ® Since the initial temperature is f(x), we have u(e,0) = fle) O0 ) ‘The problem of solving (1) subject to conditions (2), (8) and (4) is the required boundary value problem, A problem exactly equivalent to that considered above is that of an infinite slab of conducting material bounded by the planes z=0 and z=, where the planes are kept at temperature zero and where the temperature distribution initially is f(z). (b) Hf the end #=L is insulated instead of being at temperature zero, then we must find @ replacement for the condition w(L,t)=0 in (2). To do this we note that if the end x =L is insulated then the flux at ¢ = Lis zero. Thus we have = 0 — orequivalently u,(L,1) ® which is the required boundary condition, (©) It is known from physical laws of heat transfer that the heat flux of radiation from one object at temperature U; to another abject at temperature Uz is given by a(U}—U3, where « is ‘constant and the temperatures Uy and Uz are given in absolute or Kelvin temperature which is the number of Celsius (centigrade) degrees plus 278. This law is often called Stefan's rradsation law. From this we obtain the boundary condition Kub,t) = alu) where wy = allt) © If 4 and up do not differ too greatly from each other, we ean write adh uh = (ayy — (0d + weg + maa + 0h) mmvel (G4) = Aui(uy — mo) since (uy/ug), (uy/u, (uy/ug) are approximately equal to 1. Us often referred to as Newton's law of cooling, we can write (6) as Kull, 0) = Aluy— wy) o this approximation, which is where f is a constant, CLASSIFICATION OF PARTIAL DIFFERENTIAL EQUATIONS 1.10. Determine whether each of the following partial differential equations is linear or nonlinear, state the order of each equation, and name the dependent and independent variables.CHAP. 1) BOUNDARY VALUE PROBLEMS cry (a) linear, order 2, dep. var. u, ind. var. x,t (e) ret, nonlinear, order 2, dep. var. W, ind. var. r,8,¢ 1.11. Classify each of the following equations as elliptic, hyperbolic or parabolic. #o © we oF u=y AaB 0 B2-44C =~4<0 and the equation is elliptic. (b) B?—4AC = 0 and the equation is parabolic. © Vat w= 4, A=, BHO, C=—-1 BP—4AC= Sat >0 and tho cqutionishy- perbolie atu, yeu - () ae + Sazay + Soe = te By A=1 B= c elliptic. Pu wu du (ae tus t Buse = 0 A=2, B=0, C=y; RA-4AC=~tay. Hence, in the region y > 0. the equation js eliptic; in the region 2y <0 the equation is hyperbolic; ify =0, the equation is parabolic. SOLUTIONS OF PARTIAL DIFFERENTIAL EQUATIONS 1.12 Show that u(x,t) =e" sin22 is a solution to the boundary value problem au _ gu ; ae 25 M0,t) = ule, t) ‘u(x, 0) = sin 2x From u(x,t) = e-M sin 22 we have u(0,t) = e 0 = 0, wet) = eM sine u(z,0) = e-®sin2x = sin 2x and the boundary conditions are satisfied. au at = 6st cos ge, BH = = BeMeoste, FF ‘Then aubstituting into the differential equation, we have Also = Be sin 2, 46-8 sin 22 Be“ sin 2x = 2(—4e-#* sin 22) which is an identity.12 BOUNDARY VALUE PROBLEMS [CHAP. 1 1.13. (a) Show that v= F(y~8z), where F is an arbitrary differentiable function, is a general solution of the equation av av az + Bay = 9 (0) Find the particular solution which satisfies the condition 0(0,y) = 4 siny. (a) Let y~32=u. Then v= F(u) and we _ wae je = uae = P83) = 8) = ™ _ maya = PW wy = sway = Few = Pw my gge Thos Bist = 0 Since the equation is of order one, the solution » arbitrary function, is a general solution. #(0) = Ply te), which involves one ©) v(2,v) = Fy—82). Then ¥(0,y) = Fly) = 4siny. But if Fly) = Fly~ 82) = 4 sin (y~ 82) 1s the required solution. iny, then v(2,y) = 1.14. (a) Show that y(z,t) = FQ@x+5t) + G(2x—5t) is a general solution of fy _ ogy ae = ape () Find a particular solution satisfying the conditions Y(0,t) = y(n, t) (a) Let 2¢45t=u, 22—5t= 0. Then y = Fw) + 6. 4 0, 2,0) = sin2z, 4(z,0) = 0 fy _ OF U3 W® _ pang + Gray Fl . HT TRS = Pee + oo-H = srw — sew a) 8 EF) — 56 2F au _ 580’ 90 open 1 FSP) — S610) = 5H — 59GB osm) + 256") ® oF me F'(u)(2) + G(w)(2) = 2F(u) + 26%v) (s) 2 jeri) s2qtey = 22, 2 4 996" av q Zerwszay = 2 Ms 2% — gery + 10%) ® From (2) and (6), 424 = 252 and the equation is satisfied, Since the equation is of order 2 and the solution involves two arbitrary functions, it is a general solution. (@) We have from y(z,0) = F(2z+5t) + G(2e—50), We,0) = F(z) + G@s) = sinde © Also ule.) 5F'2e+5t) — 5a(2x~6t) 20 that, ule, 0) = 5F"22) ~ 5622) = 0 o Differentiating (6), 2F'2r) + 2622) = 2 cose ” From (6), Fi@e) = Gee) ® ‘Then from (7), and (8), F'@s) = G22) = fcos2eCHAP. 1] BOUNDARY VALUE PROBLEMS 138 from which F(@a) = gsin®e + oy — GI@e) = Jsin2e + op ie. we,t) = fein +5) + Ysin @e— 51) +o + oy Using v(0,t)=0 or y(e,t)=0, eybe,=0 so that fain (2x-+5) + J sin(2z—6t) = sin 2x cos5t ule.) which can be checked as the required solution. METHODS OF FINDING SOLUTIONS OF PARTIAL DIFFERENTIAL EQUATIONS ation 2. = 1.15. (a) Solve the equation aeay = ty. (0) Find the particular solution for which 2(z,0) = 2%, 2(1,y) = cosy. (©) Write the equation as 2(%) = 22y, ‘Then integrating with repect to 2, we find # = govt Fw wo where Fi) is arbiters Interatng (1) with resect y, z= et Sra + Ge) () ‘ere a) in aritary, ‘The result 2) canbe writen 2 = az,y) = sey? + Hy) + Ge) @ which has two arbitrary (independent) functions and is therefore a general solution. (t) Since z(z,0) = 22, we have from (3) = HO) + Ge) or Ge) = at — 10) “ ‘Thus 2 = bP + Hy) + # - HO) © Sine s(1,y) = cosy, we have from (6) cosy = gu? + HW) +1-H0) or Hy) = cosy — put -1+ HO) (6) Using (@) in (6), we find the required solution = deh + cosy — ft + at 1 au . 116. Solve, ty + 2S = wt, Write the equation as Z[e2¥ +20] = 2%. Integrating with respect to 2, ie Mt ou = at oH 42 = gy AO eo = ate + FO M2 = at + Fi This isa linear equation having integrating factor ef! — gtter = gia? = Then Sew = 20 + rw Inenrting, ta = far s fiRIpae + mle) = jee + 0 + HE) and this is the required general solution. ; ‘i Fu, gu, you 1. Fi a uy phe 7. Find solutions of 55 + Brg, + BTR4 1.18, 119, 1.20. BOUNDARY VALUE PROBLEMS [CHAP.1 Assume u=es+b», Substituting In the given equation, we find (at +8ab-+2breort0y = 0 or a2 + Sab + 25% = 0 Then (a+ da+26)=0 and a a= 2), If @=—b, o-bthy = e4-2 isa solution for any value of b. If @=—2b, e-®rtty = ebty-20 ig a solution for any value of b. Since the equation is linear and homogeneous, sums of these solutions are solutions (Theorem 1). For example, 320-2 — 268-2 + 5¢"-2 ig a golution (among many others), and one is thus led to F(y~2) where F is arbitrary, which can be verified as a solut imilarly, Gly — 22), where G is arbitrary, is a solution. The general solution found by addition is then given by uw = Fy=2) + Gy—22) aug eu, tu 6a ox ay © ay* ou Find a general solution of (a) atte oot 2u, (b) aoe 0. (a) Let w= cot, Then 20496 = 2 @ = 258% ang elia-aboraie toy = gxetoimveey—B0) ig.a solution, Thus w= eF(2y~ 82) is a general solution, (H) Let w= east, ‘Then da ~ dab +62 = 0 and b 80 Fix 2y) is a solution. 20,24. From this u = e742 and By analogy with repeated roots for ordinary differential equations we might be led to believe xG(x+2y) or yGlz-+2y) to be another solution, and that this is in fact true is easy to verify. ‘Thus a general solution is u = Fe +2y)+ 2G(e+2y) or e = Fle +2y) + ule +29) au I Solve Sat Ga = lett, The homoreneous equation 2H by Pree 15900 To find yariar solution of the given equation assume x= ge**# where aan unknown constant ths athe meta of undeorminedcneicnts as in erdmay deren ssoatons = 0 has general solution u = Fle+iy) + G(e— iy) ae Wo find «= 2, 90 that the required general solution is = Fletiy) + Gle— iy) + 2etete eu eu Solve fa > 45s ‘The homogeneous equation has general golution = FQzty) + G2e—y To find particular solution, we would normally assume u = ae2t¥ as in Problem 1.19 but this assumed solution is already included in F(2x+y). Hence we assume as in ordinary differential equations that «= arel'v (or w= aye%*¥). Substituting, we find «= 4. ‘Then a general solution is = FQxty) + CQe—y) + yxerty SEPARATION OF VARIABLES 121. Solve the boundary value problem ou _ 4a aot Se = AGye OD) = BeCHAP. 1 BOUNDARY VALUE PROBLEMS 16 1.22, 1.23. by the method of separation of variables. Let w= XY in the given equation, where X depends only on 2 and ¥ depends only on y. Then XY = 4X¥' or XX = YUY av iay. where X’ = dX/dr and Y ‘Since X depends only on = and ¥ depends only on y and since x and y are independent vari- ables, each side must be a constant, say c. Then X’—4eX = 0, ¥'—e¥ =0, whose solutions are X= Actes, ¥ = Be, ‘A solution is thus given by ule,y) = XY = ABectty = Kecctsrs From the boundary condition, W(0,y) = Kev = Bem which is possible if and only if K=8 and e=—3, Then u(z,y) = 8e-%62+0 = Bento ig the required solution. Solve Problem 1.21 if u(0,y) = 8e-™ + 4e-™, ‘As before a solution is Keete*, Then Kyet(rts) and Kyetste+») are solutions and by the principle of superposition so also is their sum; ie. a solution is ua,y) = Kyte + Kyeatiet From the boundary condition, u(0,y) = Kyou + Kyeres = Be 4 dent which is possible if and only if K,=8, Kz =4, ¢=—8, e =—5. ‘Then u(z,y) = Se-42+0 4 de-Schet 9) = Be-I2e—By + Gom20e~54 jg the required solution. au _ au 2a u(z,0) = Ssindsx — 3sin8ex + 2sin10re, — |u(z,t)) < M Solve O<2<3, t>0, giventhat w(0,t) = W(3,¢) 0, where the last condition states that w is bounded for 0< 2 <3,t>0. Let w=XT. Then XI" =X"T and X"/X=T'/27. Bach side must be a constant, which we call x2, (If we use +2 the resulting solution obtained does not satisfy the boundedness condi- tion for real values of .)' ‘Then XU MX = 0, T+ 2NT = 0 ‘with solutions X = Ajcosde + Bysinds, T= eye A solution of the partial differential equation is thus given by ulest) = XP = ce-8(A, cosrx + By sind) = e-B*(A cop dx + B sin dz) Since 10, ¢) = 0, (A) 0 or A=0. Then u(e,t) = Be-® sine Since w(3,t)=0, Be sind =0. If B sin =0 or Sh= mz, X= mo/3, where m 0, the solution is identically zero, so we must ha 21, +2, ule,t) = Beamer gin ME Me) = Be i Also, by the principle of superposition, wle,t) = Byer tmit9 gin 5 yee gj is a solution. By the last boundary condition,16 1.24, BOUNDARY VALUE PROBLEMS (CHAP. 1 vte.0) = Brain + Ryan EE 4 = Ssindes — Bsn fee + 2sinioez whic a poinble and ely if 2, = 5, m= 1 By=—8, m= 24, By=2, y= 00. Subatitating these in (0) the rue slain i ule, t) = B=" gin dex — 3e~1280"t gin Sex + 2e~200** sin 10x (2) ‘This boundary value problem has the following interpretation as a heat flow problem. A bar whose surface is insulated (Fig. 1-11) has a length of 3 units and a diffusivity of 2 unite. If its ends ‘are kopt at temperature zero units and its initial temperature, u(z,0) = 6 sin dex ~ 9 sinBer + 2 sin 10rz, find the temperature at position x at time:t, ic. find u(x, t). (0,2) Fig. 141 oy ey . f Solve Gr= 165.2, 0<#<2, t>0, subject to the conditions y(0,t) = 0, y(2,t)=0, (2,0) =6 sinze—Bsindev, yi(z,0)=0, |y(x,t)| 0, given that u(0,t)=u(8,t)=0, w(e,0)= fe), ju(e, 0] @ S- 226 + 2 oy wo oe 0, M>0 atu au, atte (og + Vs ray * Mage = © SOLUTIONS OF PARTIAL DIFFERENTIAL EQUATIONS 134. Show that 2(2,y) = 4e~ cos 3y is a solution to the boundary value problem ae, ae oat 7 ay? 2(2,/2) = 0, 22,0) = de> av _ 9,38 135. (a) Show ve, aF(Q2 is a general solution of =: Be @ that o(z,y) = 2F x+y) eneral solution of 292 — 202 (®) Find a particular solution satisfying w(1,y) =v 136, Find a partial differential equation having general solution u = F(e—3y) + G2 + y). 137, Find a partial different 1 equation having general solution (@) 2 = ef yz), b) z= fe tw) + oz—2y) GENERAL SOLUTIONS OF PARTIAL DIFFERENTIAL EQUATIONS a 138, (a) Solve Bao. (0) Find the particular solution for which (2,0) = a+ 2-8, 2(2,4) = ay', ‘ae by 1 2CHAP. 1] BOUNDARY VALUE PROBLEMS 19 139, 140, al. 1a, Find general solutions of each of the following. ) me . e & Heed = on =0 @ oa = oF O get Ray ) Beg te _ gi wz _ Bt me Lg © Sa Panay Sat = 9 et ay TF Find general solutions of each of the following. wo Bait oe a SE = She ae @ Solve sii + SE = 8. iF Show that a general solution of ae 3 SEPARATION OF VARIABLES A. 1s, as. an, Solve each of the following boundary value problems by the method of sepai 2% 42% = 0, wa,0) = de® ) B= osu, ulz,0) = Best ere az = Pay © 0, let) au _ ou @ e-% 2,8) © nee eneeae mM Moy ule, et = Bete in = Mau, wle,0) = 100-* ~ 6 au tu x © Fa Fs. wo,9 = 0, ue = 0, wle,0) = 6 sinZZ+ 8 sinee Solve and give a physical interpretation to the boundary value problem Fe = Ft, 0,9 = 6,9 if (a) fle) = Gsinex, (0) fla) = 8 sin dex —2 sin Bex, e,0) = 0, (2,0) = fiz) (<2 <5, t>0) Solve St = Su = Eh au it u0,2 (8, = 0, u(e,0) = 2 sin ex — sin dea, Suppose that in Problem 1.24 we have y(z,0) = fiz), where 0<#<2, Show how the problem ccan be solved if we know how to expand f(z) in a series of sines, ‘Suppose that in Problem 1.25 the boundary conditions are u,(0,¢) = Show how the problem can be solved if we know how to expand (2) ‘a physical interpretation of this problem. . (8, ¢) = 0, u(x, 0) = fla). a series of cosines. GiveChapter 2 Fourier Series and Applications THE NEED FOR FOURIER SERIES In Problem 1.25, page 17, we saw that to obtain a solution to a particular boundary value problem we should need to know how to expand a function into a trigonometric series. In this chapter we shall investigate the theory of such series and shall use the theory to solve many boundary value problems. Since each term of the trigonometric series considered in Problem 1.25 is periodic, it is clear that if we are to expand functions in such series, the functions should also be periodic. We therefore turn now to the consideration of periodic functions. PERIODIC FUNCTIONS ‘A function f(a) is said to have a period P or to be periodic with period P if for all z, f(e+P) = f(z), where P is a positive constant. The least value of P > 0 is called the least period or simply the period o f(x). Example 1. ‘The function sins has periods 2x, 4s,6r,..., since. sin (z+ 2s), sin (e+ 4x), sin (e-+60), ... all equal sin 2, However, 2r is the least period or the period of sin. Example 2. ‘The period of sin nz or cos mz, where m is a positive integer, is 2x/n. Example 3. ‘The period of tan is =. Example 4 ‘A constant has any positive number as a period. Other examples of periodic functions are shown in the graphs of Fig. 2-1 z fis ; wetin : -crPrcr. @ ® © Fig. 21 20CHAP. 2) FOURIER SERIES AND APPLICATIONS 21 PIECEWISE CONTINUOUS FUNCTIONS ‘A function f(z) is said to be piecewise con- tinuous in an interval if (i) the interval can be divided into a finite number of subintervals in each of which f(z) is continuous and (ii) the limits of f(z) as 2 approaches the endpoints of each subinterval are finite. Another way of stating this is to say that a piecewise continu- ous funetion is one that has at most a finite number of finite discontinuities. An example of a piecewise continuous function is shown in Fig. 2.2. The functions of Fig. 2-1(a) and (c) are piecewise continuous. The function of Fig. 2-1(b) is continuous. Fig.22 The limit of (2) from the right or the right-hand limit of f(z) is often denoted by lim f(e+0 = f(a+0), where ¢>0. Similarly, the limit of f(x) from the left or the left- hand limit of f(z) is denoted by lim f(t—o = f(z—0), where «>0. The values f(x+0) and f(z~0) at the point « in Fig. 22 are astindicated. The fact that «>0 and <>0 is sometimes indicated briefly by «> 0+. Thus, for example, lim f+ = fle+0), dim fle—0) = fe—0). ent DEFINITION OF FOURIER SERIES Let f(a) be defined in the interval (-L,L) and determined outside of this interval by f(v+2L) = f(@), ie, assume that f(x) has the period 2L. The Fourier series or Fourier ex- pansion corresponding to f(z) is defined to be w= ne att B+ 3 (008% + Basin #) w where the Fourier coefficients a, and by are fa, = tS. F(@) cos "7 dx a n=0,1,2,... @) 1 on OE LS, (sine ae ‘Motivation for this definition is supplied in Problem 2.4. If f(z) nas the period 2L, the coefficients a, and b, can be determined equivalently from be = [a = tS” He) cos 7 ax } 1 n= 01,2... @) | ett gin MRE [& = ts. F(a) sin" de where ¢ is any real number. In the special case ¢=—L, (3) becomes (2). Note that the constant term in (1) is equal to $ = 3 S/, Merde, which is the mean of f(z) over a period. ao If L==, the series (1) and the coefficients (2) or (3) are particularly simple. The funetion in this case has the period 2x. It should be emphasized that the series (1) is only the series which corresponds to f(2). We do not know whether this series converges or even, if it does converge, whether it con-22 FOURIER SERIES AND APPLICATIONS [cHAP. 2 verges to f(z). This problem of convergence was examined by Dirichlet, who developed conditions for convergence of Fourier series which we now consider. DIRICHLET CONDITIONS : Suppose that (i) f(e) is defined and single-valued except possibly at a finite number of points in (—L,L) (ii) /(2) is periodic with period 2L (ili) f(z) and f’(z) are piecewise continuous in (—L,L) ‘Then the series (1) with coefficients (2) or (3) converges to (a) f(z) if « is a point of continuity (6) LE+OFIE=9) 6 « is a point of discontinuity For a proof see Problems 2.18-2.28. According to this result we can write = Hy Met in 72 fey = $+ (ancos 7 + d,sin"F*) a at any point of continuity x, However, if 2 is a point of discontinuity, then thedeft side is, replaced by 4[/(x +0) + f(a —0)], so that the series converges to the mean value of f(x +0) and f(z—0). ‘The conditions (i), (ii) and ({ii) imposed on f(x) are sufficient but not necessary, i.e. if the conditions are satisfied the convergence is guaranteed. However, if they are not satis- fied the series may or may not converge. The conditions above are generally satisfied in cases which arise in science or engineering. There are at present no known necessary and sufficient conditions for convergence of Fourier series. It is of interest that continuity of f(c) does not alone insure convergence of a Fourier series, ODD AND EVEN FUNCTIONS A function f(2) is called odd if f(-x)=—f(x). Thus 2°, 2°~82°+2s, sinz, tan3e are odd functions. A function f(z) is called even if f(-z)= f(x). ‘Thus a4, 2x*~4e*+5, cos, e +e-* are even functions. ‘The functions portrayed graphically in Fig. 2-1(a) and 2-1(b) are odd and even respec- tively, but that of Fig. 2-1(c) is neither odd nor even. In the Fourier series corresponding to an odd function, only sine terms can be present. In the Fourier series corresponding to an even function, only cosine terms (and possibly a constant, which we shall consider to be a cosine term) can be present. HALF-RANGE FOURIER SINE OR COSINE SERIES A half-range Fourier sine or cosine series is a series in which only sine terms or only cosine terms are present, respectively, When a half-range series corresponding to a givenCHAP. 2] FOURIER SERIES AND APPLICATIONS 23 funetion is desired, the function is generally defined in the interval (0,L) [which is half of the interval (—L,L), thus accounting for the name half-range| and then the function is specified as odd or even, so that it is clearly defined in the other half of the interval, namely (-L,0). In such case, we have a = 0, be = FL Me sin ™ 2 defor half-range sine series tJ, L (5) de a =3 f Fle) cos "Ede for half-range cosine series PARSEVAL’S IDENTITY states that LS, Maya = B+ Deroy ) if a, and , are the Fourier coefficients corresponding to f(x) and if f(x) satisfies the Dirichlet conditions. UNIFORM CONVERGENCE Suppose that we have an infinite series } w.(z). We define the Rth partial sum of the series to be the sum of the first R terms of the series, i. S,(e) 2 u,(2) ” Now by definition the infinite series is said to converge to f(x) in some interval if given any positive number «, there exists for each = in the interval a positive number N such that |S,(t)-f(2)| N 8) The number N depends in general not only on « but also on x. We call f(x) the sum of the series. An important case occurs when N depends on « but not on the value of « in the interval. In such case we say that the series converges uniformly or is uniformly convergent to f(z). ‘Two very important properties of uniformly convergent series are summarized in the following two theorems. Theorem 2-2: If each term of an infinite series is continuous in an interval (a,b) and the series is uniformly convergent to the sum /(z) in this interval, then 1, f(2) is also continuous in the interval 2, the series can be integrated term by term, ie. LS meo} ae = ZS wterae Oy Theorem 2-3: If each term of an infinite series has a derivative and the series of deriva- tives is uniformly convergent, then the series can be differentiated term by term, i. 4 Saw = z fue (10) ‘There are various ways of proving the uniform convergence of a series. The most obvious way is to actually find the sum S,(z) in closed form and then apply the definition directly. A second and most powerful way is to use a theorem called the Weierstrass M test.24 FOURIER SERIES AND APPLICATIONS [cHAP. 2 Theorem 24 (Weierstrass M test): If there exists a set of constants M,, n= 1,2,. such that for all z in an interval |ta(2)| SM, and if furthermore = My converges, then = u(z) converges uniformly in the interval. Incidently, the series is also absolutely convergent, ie. = |us(2)| converges, under these conditions. Bowmen The sevies 3 “52 converges uniformly in the interval (~ryx) fr, im fact, im any interval, since a set of constants iy = Un? can be found such that, 1 =i and INTEGRATION AND DIFFERENTIATION OF FOURIER SERIES Integration and differentiation of Fourier series can be justified by using Theorems 2-2 and 2-8, which hold for series in general. It must be emphasized, however, that those theorems provide sufficient conditions and are not necessary. The following theorem for integration is especially useful. Theorem 2-5: The Fourier series corresponding to f(z) may be integrated term by term from a to 2, and the resulting series will converge uniformly.to "#(w) au, provided that f(z) is piecewise continuous in —L SSL and both a and are in this interval COMPLEX NOTATION FOR FOURIER SERIES ‘Using Euler's identities, e# = cosd + ising, e# = cos — ising (a) where # is the imaginary unit such that i? in complex form as 1, the Fourier series for f(x) can be written f(a) cael (12) where = as. Haye-r=" de (13) In writing the equality (12), we are supposing that the Dirichlet conditions are satisfied and further that f(z) is continuous at x. If f(2) is discontinuous at z, the left side of (12) should be replaced by 1@*9) + fle—0) DOUBLE FOURIER SERIES ‘The idea of a Fourier series expansion for a function of a single variable x can be ex- tended to the case of functions of two variables x and y, i.e. f(x,y). For example, we can expand f(x,y) into a double Fourier sine series ney fe = 3S Baw sin sine (1) where Brn ood tS Fe, y) sin" sin de dy (15)CHAP. 2} FOURIER SERIES AND APPLICATIONS 25 Similar results can be obtained for cosine series or for series having both sines and cosines. ‘These ideas can be generalized to triple Fourier series, ete. APPLICATIONS OF FOURIER SERIES There are numerous applications of Fourier series to solutions of boundary value prob- lems. For example: 1. Heat flow. See Problems 2.25-2.29. 2 Laplace's equation. See Problems 2.30, 2.81. 8 Vibrating systems. See Problems 2.32, 2.33. Solved Problems FOURIER SERIES 21. Graph each of the following functions. 3B OK and L= =. Choosing = 0, we have = TG Heyes ae = tf xt eosnz de = (saz) can (=s922) + o(=8g™ = 4, aro . o(} Then fo) = sinne) for 0-< 9 © 2. But by the Diriehlet conditions, the series converges at x= 0 to }(0+424)= 258, Hence the desired result. ODD AND EVEN FUNCTIONS. HALF-RANGE FOURIER SERIES 28, Classify each of the following functions according as they are even, odd, or neither even nor odd. @ fe) = { OSS PG 2 -8<2<0 From Fig. 2-8 below itis seen that f(z) = fe) Fig.2830 29, FOURIER SERIES AND APPLICATIONS [CHAP. 2 ©) f@) = oe OSES Period = 2 From Fig.2-9 below it is seen that the function is neither even nor odd. fe) Fig.2.9 (0) fe) =2(10-2), 0<2<10, Period =10. From Fig. 2-10 below the function is seen to be even. fe) Fig.2-10 Show that an even function can have no sine terms in its Fourier expansion. Method 1. No sine terms appear if by =0, ®=1,2,5,.... To show this, let ws write = 1 py ant wo 28 ae 4 ("pay be = HS torantae = Ef sepantt ae sf pean D 1 wo make the transformation =u in the frst integral onthe right of (1), we obtain meg, — 1" et) a = 2" ew Beate = EL sew sin(—M) an = Ef sew sinte aw = ES" roa sin ae = Ef" se) sin ME de ) where we have used the fact that for an even function /(—u) = f(u) and in the last step that the dummy variable of integration u can be replaced hy any other syribl, in particular 2- ‘Thus fom (Gy using (2), we have be = ES Mersin" ae + ff se) sin" ae = 0 Method 2. Assuming convergence He) = Fa 3 (oot 5, ant) Then fen = B43 (« con® — b sin If fe) is even, f(-2) = f(a). Hence oo 2 ee +S (os conte +b, ninte® = H+3 (cn coats ~ by nt)CHAP. 2) FOURIER SERIES AND APPLICATIONS 31 He) = $+ 3 ay conte and s0 ZpasinZ= = ‘and no sine terms appear. ‘This method is weaker than Method 1 since convergence is assumed. In a similar manner we ean show that an odd function has no cosine terms (or constant term) fm ite Fourier expansion. 210. If f(x) is even, show that (a) ax= 7 f f(x) cos “P a = EGY pep costZde = EG" Moyo m2 ap = 1" pny con (2) au = 16" pe) conan feeds = Tf Aw ( Ti) au = ff, fen costa since by definition of an even funetion f(x) = f(w). ‘Then on = Eh teeta hf" pe cE te = Bf" He) coeBEce (@) This follows by Method 1 of Problem 2.9. 211. Expand f(x) =sinx, 0 (a) From Problem 2.12(a), “ 4 iat — Line 4 Lyn et — ...) w= f(a gente + pn ~~~) “ Integrating both sides from 0 to « (applying Theorem 2-5, page 24) and multiplying by 2, wwe find 1 ogg Bee Bex pons ae 0 SSE ) ®CHAP. 2] FOURIER SERIES AND APPLICATIONS 35 (B) To determine C in another way, note that (2) represents the Fourier cosine series for 2* in 0<2<2, Then since L=2 in this case, c= Sa df sede = Ef sae 4 3 ‘Then from the value of C in (a), we have (cae 2.17. Show that term by term differentiation of the series in Problem 2.12(a) is not vali ‘Term by term differentiation yields 2( cos = ~ cos "54 + cos 852 — Since the nth term of this series does not approach 0, the series does not converge for any value of x, CONVERGENCE OF FOURIER SERIES 218. Prove that 1, . _ sin(M +4)¢ (a) Jt cost +eos2t +--+ bose = BULA DE Lcrsin(+H)t 1 1° sin(M+atg, 1 ® oS, gsndt “= 2 FJ asngt “= 2° (@) We have cosnt sin Jt = flsin(n-+ $)t~ sin(n~ 4)t}. ‘Then summing from mn =1 to Af, sin Jelcos team Zt +--+ eon) = (sine sin Jo) + (in ft sin go sin OF + Je sin OW Hye] = pein (+ pe — sing) On dividing by sin f¢ and adding 4, the required result follows (©) Integrate the result in (a) from 0 to x and —r to 0 respectively. results, since the integrals of all the cosine terme are zero, This gives the required 219. Prove that lim ("" f(z)sinnzdx = lim {"" f(x) cosnzdz = 0 if f(2) is piecewise continuous. "~~ oD : This follows at once from Problem 2.15, since if the series “2+ $ (a3-+ #2) is convergent, lim a, = tim by = 0, 7 ‘The result ia sometimes called Riemann’s theorem. 2.20, Prove that Jim ff * 4(a) sin(M+3)zdz = 0 if f(a) is piecewise continuous. We have * S_f sinat + de ee Se sng conte de + fe eno inte te it once by using the result of Problem 2.19, with f(z) replaced b fiz) sin de and f(z) cos 42, respectively, which are piecewise continuous if flz) is. “ee 221, Assuming that L 2L= ‘The result can also be proved when the integration limits are a and } instead of —- and x. x, ie. that the Fourier series corresponding to f(z) has period show that36 FOURIER SERIES AND APPLICATIONS [cHaP. 2 S. Hevea) 7) ants Mt ap Sy) = f+ 3 le conne + by sin nz) sami Using the formulas for the Fourier coefficients with L= =, we have enenens + bysinns = (Ef so) cosmedneosnz + (EG” 100 sinme de)sin ne 27 feuyfoos mu cos nz + sin nu sin n2) du = 26" sorcante—ai ee ES” Horde a z Ft 3 lencosne + by sina) Also, Then Sule) = Ef" sf + 3 cons a} a = ET po teeawece Fain guar using Problem 218. Letting =a salt pe Sue) = Sf kee ae Sinee the integrand has period 2x, we can replace the interval ~*—2,r—z by any other interval of length 2r, in particalar —r, x. ‘Thus we obtain the required result. t, wehave 2.22. Prove that sya) - FEEDS IE—O) 2 1S" Mt IE) sin cars syeat a if” THe anaes From Problem 2.21, a sina sue) = Af" nee) SEEM ae AE yey OEM Maltpiying the integrals of Problem 2180) by e~0) and fle +0) resestivly, lb) 4 fe egg) HULA inet ne Fe EE eat Ef pet ae wy Subtracting (2) from (1) yields the required result, 2.28. If f(x) and f’(z) are piecewise continuous in (~z, z), prove that lim Sy(2) He+0) : Se The function E42) LE +0) i, vigcewise continuous in 0 are solutions. Thus we obtain the general solution P= Agoh + Bao © Since 1p, 9) must have period 2r in 9, we must have \ =m = 0,1,2,3,. Also, since u must be bounded at = 0, we must have By=0. ‘Thus us Pe = Age™(A, cosme + By me) = pm(A cosme + Bsinmg) By superposition, a solution is Ao, § wos) = +S omAp conmg + By sin my) from which uth) Ao, § Ft 4m come + By sin me) Then from the theory of Fourier series, Ay = Ef eg coume 1 Sl wemmete # 20 a 0 i m>0 precede = Nay itm 1 ao same ds 1 ag sinmp ds + 2h uesinme de = "2 a —coamey Toe: ag) = EM § Mm UN Reem) (sing + 4p sine + Jo sinby + - san (222) ) + z on making use of Problem 2.64.cuaP. 2] FOURIER SERIES AND APPLICATIONS a 2.29. A square plate with sides of unit length has its faces insulated and its sides kept at 0°C. If the initial temperature is specified, determine the sub- sequent temperature at any point of the plate. Choose a coordinate system as shown in Fig. 2-16. ‘Then the equation for the temperature u(z,y,£) at any point (x,y) at time ¢ is au _ (0, a a (S5+%) “ ‘The boundary conditions are given by ues, 0) < at u0,¥,1) = wlla,0) = W(2,0,1) = ae,1,) = 0 ule, y,0) = flou) where 0<2<1, 00, Fig. 2415 To solve the boundary value problem let w= XYT, whore X, ¥, T ave functions of , y, ¢ re spectively. Then (1) becomes xyr XYT + XY") Dividing by «XYT yields Ye y Since the left aide is a function of ¢ alone, while the right side is » function of # and y, we see that each side’ must be a constant, say ~-r# (which is needed for boundedness), Thus x rar =o = ® ‘The second equation can be written as Re x= 7¥ a” nd since the left side depends only on © while the right side depends only on y each side must be a constant, say ~y2 ‘Thus x4 eX = 0 Y" +08 )¥ = 0 o Solutions to the two equations in (3) and the first equation in (2), are given by X = a cos pe + by sinns, Y = ay con Vi®— ey + by sin VIF Fu, T= ayo It follows that a solution to (1) is given by wy + by sin VAP 2 Hage") lesa 1) = (ay eos ix + by sin ua)(ag cos VIP~ From the boundary condition u(0,y,t)=0 we see that a =0. From a(z,0,f)=0 we see that a, =0. ‘Thus the solution satisfying’ these two conditions is uley, 8) = Ber sin ue sin VEE y where we have written B= 6:60 From the boundary eondition_u(l,y,®) = 0 we see that «= ms, m= 1,28)... From We,1,0=0 we see that ViF~ P= nr, = 1,2,8,..., or X= Vme tan, It follows that a solution satisfying all the conditions except w(x, y, 0) fe,y) is given by (x,y, 8) = Bens hoo sin mew sin ney42 FOURIER SERIES AND APPLICATIONS [oHaP. 2 Now, by the superposition theorem we can arrive at the possible solution wend = 3 3 Byes sin mei sinney ® Letting ¢= 0 and using the condition u(x, y,0)= fleyy), we arzive at He) = 3, 3 Pa in mes sin ney As in Problem 2.24 we then find that, Bon = AS fovea ‘Thus the formal solution to our problem is given by (4), where the By, are determined from (6). ve sinnzy de dy ~ LAPLACE’S EQUATION 2.30, Suppose that the square plate of Problem 2.29 has three sides kept at temperature zero, while the fourth side is kept at temperature uw. Determine the steady-state temperature everywhere in the plate. Choose the side having temperature a to be the one where y= 1, a8 shown in Fig. 216. Since we wish the steady-state temperature 1s which doco not depend on time é the equation is obtained from (7) of Problem 2.29 hy setting dufat'= 0; he. Laplace’s equation in two dimensions au, ou matt a =o a ‘The boundary conditions are u(0,y) = u(y) = ule, 0) = 0, n(a, i) Fig.2.16 and ju(z,y)| 0 where (x,t) = displacement from a-axis at time ¢ (Fig. 2-22), Fig. 2-2244 FOURIER SERIES AND APPLICATIONS (CHAP. 2 Since the ends of the string are fixed at «=0 and 2=L, 0,0) = ib.) = 0 10 Since the itial shape of the string is given by flz), uir,0) = fle) O, if the period is =; and (b) compare with the result of Problem 2.37, explaining the similarities and differences if any. o OR © Zien a 4 BeV2 242 Show that 4+ =e. INTEGRATION AND DIFFERENTIATION OF FOURIER SERIES 243. (a) Show that for -2 <2 O Zw > ie48 2a. 248, 249, FOURIER SERIES AND APPLICATIONS [cHAP. 2 1 1 Show tht pig + gia Sea Show that @) & gaye Show that SOLUTIONS USING FOURIER SERIES 250, 252, 253. 255. 2st. (a) Solve the boundary value problem me. ge at Gat W0,t) = ult) = 0 ulz,0) = 252 where 0<2<4, £50. (8) Interpret physically the boundary value problem in (a) (a) Show that the solution of the boundary value problem ae = = ue,0) = fa Sa SE (0,8) = ule) = 0 ale,0) = fla) where 00, is given by wan = Lf" pode +? (Sew come fe er made (8) Interpret physically the boundary value problem in (a). Find the steady-state temperature in a bar whose ends are located at z=0 and x= 10, if these tends are kept at 150°C and 100°C respectively. ‘A circalar plate of unit radius (see Fig. 2-14, kept at temperature 120+ 60 cos 25. Find the 89) whose faces are insulated has its boundary sady-state temperature of the plate. Show that pang + dp? windy + pptaindg +--+ = gtan-s(Besine) and thus complete Problem 2.28. A string 2 ft long is stretched between two fixed points x=0 and 2=2. If the displacement of the string from the z-axis at £=0 is given by f(z) =0.032(2—2) and if the initial velocity is zero, find the displacement at any later time. ‘A square plate of side a has one side maintained at temperature f(x) and the others at zero, as indicated in Fig. 2-25. Show that the steady-state temperature at any point of the plate is given by wed = 3 [sac Sf 10 an tt ae] sin At sion tt Work Problem 2.56 if the sides are maintained at temperatures f,(2),0s(v);fa(2),gzlu), respectively. {Hint. Use the prinefple of superposition and the result of Problem 2.56.)CHAP. 2] FOURIER SERIES AND APPLICATIONS 49 258, 258. 8 263. Fig. 2-25 + Fig. 2-25 An infinitely long plate of width a (indicated by the shaded region of Fig, 2-26) has its two parallel sides maintained at temperature 0 and its other side at constant temperature ts. (a) Show that the steady-state temperature is given by in 82 4 Agen gin B22 Set 4 Neo gin SE a 75 a” = Mf. gin 8 4 Myre wey) = Se sin 24 de (8) Use Problem 2.54 to show that ule) Solve Problem 1.28 if the string has its eids fixed at =O and 2 = ment and velocity are given by /(2) and g(z) respectively. ‘A square plate (Fig. 2-27) having sides of unit length has y its edges fixed in the 2y-plane and is set into transverse vibration, (a) Show that the transverse displacement <(z,y,¢) of any point (x,y) at time ¢ is given by ae a = (8) vin oe tn, (0) Show that if the plate is given an initial shape f(x,y) and released with velocity g(eyy), then the displacement is given by Fig.227 26,00 = 33 [Aan 08 %mqtt + Bay i a] sin mes sin ney where Ane = 4 SL" fest) sinmes sinncy de dy Fan = GELS fea nn ee inane a ty and gq = eV a Work Problem 2.60 for a rectangular plate of sides b and e. Prove that the result for u(x,t) obtained in Problem 2.25 actually satisfies the partial differential ‘equation and the boundary conditions. Solve the boundary value problem ju te = Her eu | o0, (b) b=, and give physical inter- pretations. (@) Solve the boundary value problem ou a where (0,1) =0, w(Z,t)=0, u(z,0) = fla), jul, | 0 as M->, 80 that (15) is valid. We refer to (15) as Parseval's identity for orthonormal series of func- tions. In (6) of Chapter 2, page 23, we have obtained Parseval's identity for the special case of Fourier series. In the case where Ems>0 as M> =, ie. tim S [fe) -Sy(otae = 0 (16) we sometimes write Lim. S,(z) = f(@) an ‘This is read the limit in mean of Sy(z) as M > » equals f(z) or Sy(2) converges in the mean to f(z) as M> © and is equivalent to (16). STURM-LIOUVILLE SYSTEMS. EIGENVALUES AND EIGENFUNCTIONS A boundary value problem having the form . ale GE] + (aa) tary = 0 asad qu(a) + aya) = 0, Byy(b) + By"(b) = 0 (18)cHar. 3] ORTHOGONAL FUNCTIONS 55 where a, 8,8, are given constants; p(2),q(z),7(2) are given functions which we shall assume to be differentiable and A is an unspecified parameter independent of «, is called a Sturm-Liouville boundary value problem or Sturm-Liouville system. Such systems arise in practice on using the separation of variables method in solution of partial differential equations, In such case A is the “separation constant.” See Problem 3.14. A nontrivial solution of this system, ie. one which is not identically zero, exists in general only for a particular set of values of the parameter A. These values are called the characteristic values, or more often eigenvalues, of the system. The corresponding solu- tions are called characteristic functions or eigenfunctions of the system. In general to each eigenvalue there is one eigenfunction, although exceptions can occur. If p(z) and q(x) are real, then the eigenvalues are real. Also, the eigenfunctions form an orthogonal set with respect to the weight function r(z), which is generally taken as non- negative, ie. r(x) =0. It follows that by suitable normalization the set of functions can be made an orthonormal set with respect to r(x) in a=a=b, See Problems 3.8-8.11. THE GRAM-SCHMIDT ORTHONORMALIZATION PROCESS Given a finite or infinite set of linearly independent functions ¥,(*t), $4(2t), Y(t), -.. de= fined in an interval (a, b) it is possible to generate from these functions a set of orthonormal functions in (a,b). To do this we first consider a new set of functions obtained from the q(t) and given by CFDs Car Fi@) + eae ValZ) Cp Vil) + Cap Val) + esp Halt), (19) where the o’s are constants to be determined. We shall designate the functions in (19) by y(t), bal), sl@)y oe 6 We now choose the constants ¢,,,¢,y C+. $0 that the functions 4,(x), $4(2), 64(t), «-- are mutually orthogonal and also normalized in (a,b). ‘The process, known as the Gram- Schmidt orthonormalization process, is illustrated in Problem 3.12. An extension to the case where orthonormalization is with respect to a given weight function is easily made. APPLICATIONS TO BOUNDARY VALUE PROBLEMS In the course of solving boundary value problems using separation of variables we often arrive at Sturm-Liouville differential equations (see Problem 3.15, for example). The parameter A in these equations is the separation constant, and the values of A which are obtained represent the real eigenvalues. ‘The solution of the boundary value problem is then obtained in terms of the corresponding mutually orthogonal eigenfunctions. For an illustration which does not involve Fourier series, see Problem 3.13. Other illus- trations involving this general procedure will be given in later chapters.56 ORTHOGONAL FUNCTIONS (war. s Solved Problems ORTHOGONAL FUNCTIONS AND ORTHONORMAL SERIES 8.1. (@) Show that the set of functions 1, sinS®, cost, sin2Z, cos@3#, sin St, cos®z*, , siny, coaz, sin“Z~, cos"7, sin“7~, cosy form an orthogonal set in the interval (~D, L). (b) Determine the corresponding normalizing constants for the set in (a) so that the set is orthonormal in (—D, L). (a) This follows at once from the results of Problems 2.2 and 2.3, page 26. (®) By Problem 2.3, (sme de =, nef (ffante)ae = a Also, (Wide = 2b or ‘Thus the required orthonormal set is given by 1 ole 1 ee! E 3.2, Let {¢,(x)) be a set of functions which are mutually orthonormal in (a,b). Prove that if 5c, 4,(x) converges uniformly to f(x) in (a,b), then = S fas wae Multiplying both sides of « Fe) = 3 omenled © by #n(2) and integrating from a to 8, we have Sf er sntedds = 3 ex Sonla) onle) de “ where the interchange of integration and summation is justified by the fact that the series converges uniformly to f(z). Now since the functions {¢,(c)} are mutually orthonormal in (a, b), we have 0 men 1 oman rn(2) pyle) ds = { 0 that (2) becomes as es Sorento de = em ) as required. We call the coefieients cq given by (8) the generalized Fourier cveffcients corresponding to f(a) even though nothing may be known about the convergence of the series in (1). As in the case of Fourier series, convergence of 3 cyda(2) is then investigated using the coefficients (3). ‘The conditions of convergence depend of course on the types of orthonormal functions used. In the remainder of this book we shall be concerned with many examples of orthonormal functions and series,CHAP. 3} ORTHOGONAL FUNCTIONS 87 LEAST-SQUARES APPROXIMATIONS. PARSEVAL’S IDENTITY AND COMPLETENESS: 33. If Sz) = ¥a,4,(e), where 4,(2), n=1,2,..., is orthonormal in (a,2), prove that = S- where ¢, = S f(z) 4,(a)de are the generalized Fourier coefficients corresponding to f(x). * We have dz =f" yertas - 23 oc, + at fle) ~ Sula) = fe) — Son ale) By squaring we obtain x x (We) Sula)? = [flay]? ~ 2 a fle) bole) + SB, om on tlt) onl) Integrating both sides from a to 6 using a = (Tesi, — (aneraterde = 3 a wwe obtain S'ver-swentae = f"yentde - 23 et Bet ‘We have assumed that f(z) is sueh that all the above integrals exist 34. Show that S [fa)-Sy(aede = Senras + Beer - Bet This follows from Problem 3.3 by noting that Sierras 23 nee t Bot = fenrae + 3 ot -20e0 = [rere + 3 tora = feta + 3 eet ~ 3 ef 8.5. (a) Prove Theorem 3.1, page 54: The root mean square error is a minimum when the coefficients «, equal the Fourier coefficients ¢,. (b) What is the value of the root mean square error in this case? (@) From Problem 3.4 we have S10) swords = f'yerpae + 3 ea eot- 3 Now the root mean square error will be a minimum when the above is a minimum. However, it fs clear that the right-hand side is a minimum when (age)? = 0, Le. when ay = ¢y for all n. a58 36. 3a. ORTHOGONAL FUNCTIONS (CHAP. 3 (8) rom part (a) we see that the minimum value of the root mean square error is given by ff = sseyeac]” [Sverre ~ 3a)” Prove that if c, m=1,2,8,..., denote the generalized Fourier coefficients corre- sponding to f(z), then E, Se = Sree rom Problem 6 we see that sinc the root mean equrs error must be nonnegative Se = [eie ry ‘Then, taking the limit as Af > and noting that the right side does not depend on M, it follows that St sf ierrae ® ‘This inequality is often called Beseel’e inequality. ‘As a consequence of (2) we see that if the right side of (2) exists, then the series on the left, must converge. In the special ease where the equality holds in (2) we obtain Parseval's identity. inate Show that lim S Ha) 6,(e) de ‘nth term ef, and with it ¢,, must approach zero as n> ®, which is the required result. Note that this result for the special case of Fourier series is Riemann’s theorem (see Problem 2.19, page 5). STURM-LIOUVILLE SYSTEMS. EIGENVALUES AND EIGENFUNCTIONS 38. (a) Verify that the system y”+Ay=0, (0) =0, y(1) =0 is a Sturm-Liouville sys- tem. (0) Find the eigenvalues and eigenfunctions of the system. (c) Prove that the eigenfunctions are orthogonal in (0,1). (d) Find the corresponding set of normalized eigenfunctions. (e) Expand f(x) =1 ina series of these orthonormal functions. (@) The system is a special case of (18), pase 64, with p(x) =1, g(2) = 0, rie) =1,a=0, b= 1, a= 1, a= 0, By =1, fp =0 and thus is a Sturm-Liouville system. (®) The general solution of y"+Ay=0 is y= AcosViw+Bsin Viz, From the boundary condition y(0)=0 we have A=0, ie. y=BsinViz. From the boundary condition w(l) =0 we have BsinVX=0; since B cannot be zero (otherwise the solution will be iden- tically zero, i.e. trivial), we must have sinyi=0, Then Vi=mn, X= més, where m= 1,2,3,... are the required eigenvalues, ‘The eigenfunctions belonging to the eigenvalues = mx? can be designated by By sin mx, m= 1,23,..... Note that we exclude the value m=O or 4=0 as an eigenvalue, since the corresponding eigenfunction is zero. BeBe (oatn vee ~ cnt ee de = 0, men 2 ( m+ 6 SoD. unin —aiee _ sin tse)CHAP. 3} ORTHOGONAL FUNCTIONS 59 (@) The eigenfunctions will be orthonormal if Si Wasinmentee = 1, or By= V2, taking the BR Ba ie. if BR fsintmes de = SP f(a —costmea) de = positive square root, ‘Thus the set VZ sin mez, m=1,2,..., is an orthonormal set, (6) We must find constants ¢,¢)-- such that He) = 3 en tnled where f(2)=1, éal2)= VEsinmez, By the methods of Fourier series, te = fMeente de = VE fsinmes de = Wid sone Then the required series [Fourier series] is, assuming 0 <2 <1, 1= 3% ‘cos ms) 39. Show that the eigenvalues of a Sturm-Liouville system are real. We have Elro] + tae +rveqy = 0 “ Ayulb) + Bay) = 0 « Then assuming p(2),q(2),r(2),a1 aa, Ps» A2 are real, while \ and y may be complex, we have on taking the complex conjugate (represented by using a bar, a ae ela) + aay'(a) [ma di]4+ teereanee = 0 ® ayila) + asia) Built) + Byi/(0) = 0 wo Multiplying equation (1) by 7, (8) by v and subtracting, we find atter simplifying, Ff Pevs! — gv) = O— Veena ‘Then integrating from @ to b, we have 0-0 f'rowtdr = rors ann] = 0 6 -¢ r(z) = 0 and Is not identically zero in (a, ), the integral fon the left of (5) is positive and so )—1=0 or \=T, so that 2 is real 3.10. Show that the eigenfunctions belonging to two different eigenvalues are orthogonal with respect to r(z) in (a, b). If y; and yz are eigenfunctions belonging to the eigenvalues d, and A, respectively, ay FLO B] + tee +rrton = 0 w auns(@) + axxile) = 0, Bini) + Baxi) = 0 ®) avy fant] ev ernern = 0 - aue(a) + agus(a) = 0, Byug(0) + Aaug(d) = 0 wBu. ORTHOGONAL FUNCTIONS (CHAP. 3 ‘Then multiplying (1) by ua, (8) by 14 and subtracting, we find as in Problem 3.9, A pevuws—vvd) = O.-rdrledve Integrating from a to b, we have on using (2) and (4), Od fl remade = plenvind ws and since 2 72g. we have the required result £ reonwids = 0 Given the Sturm-Liouville system y”+Ay=0, y(0)=0, y(L) + sy(L) = 0, where p and L are given constants. Find (a) the eigenvalues and (b) the normalized eigen- functions of the system. (¢) Expand f(z), 0<20 (0,t) = 0, m{L,t) = 0, u(x.) = Fx), |ula,t)| < M leads to a Sturm-Liouville system. () Give a physical interpretation of the equa tion in (a). (c) How would you proceed to solve the boundary value problem in (a)? (©) Letting w= XP in the given equation, we find = 8 Ley ®X] 4 gone = 22 [aot] « noxr Then dividing by of2)XT yields re ee Eo hed [mf] + ne Setting each side equal to ~d, we find rears o o £fmoZ]+ me rome = 0 5 Also, from the conditions (0,1) =0 and u(L,t) =0 we are led to the conditions XO) = 0 XL) = 0 @ ‘The required Sturm-Liouville system is given by (2) and (3). Note that the Sturm) differential equation (2) corresponds to that of (13), page 54, if we choose y= X, plz) = Kle, ale) = Ma), ria) = (0), (8) By comparison with the derivation of the heat conduction equation on page 9 we see that ‘u(e,t) can be interpreted as the temperature at any point « at time t. In such case K(x) the (nonconstant) thermal conductivity and g(z) is the specific heat multiplied by the density. ‘The term h(x)u can represent the fact that a Newton's law of cooling type radiation into ‘medium of temperature zero is taking place at the surface of the bar, with a proportionality factor that depends on position, (c) From equation (2) subject to boundary conditions (8) we can find eigenvalues \, and normalized eigenfunctions X,(2), where m=1,2,3,.... Equation (1) gives T= ce," Thus a sohution obtained by superposition is uz) = 3 ene M*X(a) From the boundary condition u(z,0) (2) we have fe) = 3 eXqte)64 ORTHOGONAL FUNCTIONS (CHAP. 3 which leads to a = fi to xierae ‘Thus we obtain the solution ue, = 3 {froxma} Supplementary Problems ORTHOGONAL FUNCTIONS AND ORTHONORMAL SERIES 315. Given the functions ap, a; + a,2, a3 +a, + a522 where dy, stants so that these functions are mutually orthonorm: 5 are constants. Determine the con- in the interval (0, 1). 316. Generalize Problem 8.15 to arbitrary finite intervals. 317. (a) Show that the functions 1,1—z,2—4z+2% are mutually orthogonal in (0,=) with respect to the density function e+, (8) Obtain a mutually orthonormal set. 348, Give a vector interpretation to functions which are orthonormal with respect to a density or weight function, 3.19. (a) Show that the functions cos (ncos~!), m= 0,1,2,3,..., are mutually orthogonal in (~1,1) with respect to the weight function (1—2)~1/2, (b) Obtain'a mutually orthonormal set of these functions. 320. Show how to expand /(2) into a series with respect to the weight function w(z). | en ule), where gy(2) are mutually orthonormal in (2) 321, (a) Expand f(a) into a series having the form Sc, (s), where gy(s) are the mutually ortho- normal functions of Problem 3.19. (b) Discuss the relationship of the series in (a) to Fourier series. APPROXIMATIONS IN THE LEAST-SQUARES SENSE, PARSEVAL'S IDENTITY ‘AND COMPLETENESS: 822. Lot r be any three-dimensional vector. Show that (@) eC Se) (HCE DE+ (EW! = where r= r+r and discuss these with reference to Bessel’s inequality and Parseval’s identity. Compare with Problem 3.6, 328. Suppose that one term in any orthonormal series (such as @ Fourier series) is omitted. (2) Can ‘we expand an arbitrary function f(z) in the series? (b) Can Parseval’s identity be satisfied? (c) Can Beasel’s inequality be satisfied? Justify your answers. 13k) Fd opens oth het {7 f= oisine tints tsa «minima (©) What is the mean aquare error and root mean square error in approximating 2 by ¢, sin + easin2z + ey sin Sz, where cy cy, ¢g are the values obtained in (a)?CHAP. 3} ORTHOGONAL FUNCTIONS 65 (©) Suppose that it is desired to approximate 2 by a; sin + ap sin 2x + ay sin Sx + a, sin dz in the least-aquares sense in the interval (—r,x)- Are the values ay, ap, 9 the same a8 cy, Cy, cy of part (a)? Explain and discuss the significance of this. 825. Verify that Bessel’s inequality holds in Problem 3.24 26, Discuss the relationship of Problem 8.24 with the expansion of /(2) interval (—r, x) in a Fourier series in the 321, Prove that the set of orthonormal functions a(x), "= 1,2,8,..+4 cannot be complete in (a,6) if there exists some function /(z) different from zero which is orthogonal to all members of the set, Le. if Sree = 0 nau, 328, Is the converse of Problem 3.27 true? Explain. GRAM-SCHMIDT ORTHONORMALIZATION PROCESS 828 Verify that a continuation of the process in Problem 8.12 produces the indicated results for (x) and ¢(2). 830, Given the set of functions 1,2,2%,29,..., obtain from these a set of functions which are mutually orthonormal in (1,1) with respect’ to the weight funetion 2. 331, Work Problem 3.50 if the interval is (0, ©) and the weight function is es, The polynomials thus obtained are Laguerre polynomials. 482, Is it possible to use the Gram-Schmidt process to obtain from #, 1 orthonormal in (0,1)? Explain. x, 8+22 a set of functions STURM-LIOUVILLE SYSTEMS. EIGENVALUES AND EIGENFUNCTIONS 833. (a) Verify that the system y” +n = 0, v'(0) =0, vit) 1b) Find the eigenvalues and eigenfunctions of the system. 0 is @ Sturm-Liouville system. (©) Prove that the eigenfunctions are orthogonal and determine the corresponding orthonormal fonctions. 834, Work Problem 3.83, if the boundary conditions are (a) y(0) = 0, y'(1) 0) u'0) = 0, y'Ct) 445, Show that in Problem 3.11 we have 2 204+ 8) Ps = Tt DPB 4136, Show that any equation having the form ay(zly”” + ax(z)y! + in Sturm-Liouvile form as (2) + rag(z)]y = 0 can be written fo] + wer + ron = 0 with (a) = SH, gs) = Bow, He) = SV . ple) a) = Bre, re) = Bote) Discuss Problem 318 if the boundary conditions are replaced by u,(0,t)=hyu(0,0), u,(L,t) Ayu, 0.66 ORTHOGONAL FUNCTIONS [CHAP. 8 338, (a) Show that separation of variables in the boundary value problem’ wet = 2[ ro] + mow , uiLt) = 0, vle,0) = He), (2,0) = 0, ule, 8] < OF ive a physical interpretation of the equations in (a). 0,2) leads to a Sturm-Lionville system. (8) (@) How would you solve the boundary value problem? 339. Discuss Problem 8.88 if the boundary conditions y(0,t) = 0, y(L,t)=0 are replaced by y,(0,) = hyW00,9, ull, t) = hgulL,t), respectively. APPLICATIONS TO BOUNDARY VALUE PROBLEMS. 840. (a) Solve the boundary value problem ae ate Ms Ee o0 (0,0) and (b) interpret physically. (2,0) = fle), |ule,t)| 0 WO) = 0, 4410.) = 0, wiL,t) = 0, uel) = 0, wlz,0) = fle), |e, 0) < ME and (6) interpret physically. 3.43. Show that the solution of the boundary value problem mu Le = BE ocechero tg(0,t) = hu, t,t) = hulle), wle,0) = fle) where x, h and J are constants, is & «ts cou ya + ima wen = Ze 1 TAT soli con yet hain dye) le 2h where 2, are solutions of the equation tan Nl Give a physical interpretation. poe“Chapter 4 Gamma, Beta and Other Special Functions SPECIAL FUNCTIONS 7 In the process of obtaining solutions to boundary value problems we often arrive at special functions. In this chapter we shall survey some special functions that will be em- ployed in later chapters. If desired, the student may skip this chapter, returning to it should the need arise, ‘THE GAMMA FUNCTION The gamma function, denoted by T(n), is defined by rin) = fete de w which is convergent for » > 0. A recurrence formula for the gamma funetion is T(n+1) = nr(n) (2) where T(1)=1 (see Problem 4.1). From (2), In) can be determined for all n>0 when the values for 1= <2 (or any other interval of unit length) are known (see table on page 68). In particular if n is a positive integer, then T(n+1) =n! n= 1,2,3,... (3) For this reason T(n) is sometimes called the factorial function. Examples. rQ) = 11 =1, 16) = re. It can be shown (Problem 4.4) that rd) = VF (4) The recurrence relation (2) is a difference equation which has (1) as a solution. By taking (1) as the definition of T(n) for n>0, we can generalize the gamma function to n <0 by use of (2) in the form ~ @) See Problem 4.7, for example. The process is called analytic continuation. 6768 GAMMA, BETA AND OTHER SPECIAL FUNCTIONS [CHAP. 4 TABLE OF VALUES AND GRAPH OF THE GAMMA FUNCTION Tm cs 1.00 11,0000 " 1.10 0.9514 # 1.20 0.9182, * 1.30 0.8975 4 tom mst ic . tse osm ae opeeee tin os : iio oane ia oasie * tao one Nc ban tome ASYMPTOTIC FORMULA FOR rn) If nis large, the computational difficulties inherent in a direct calculation of 1(n) are apparent, A useful result in such case is supplied by the relation T(n+1) = VBannrememnnt? — <0< 1 (6) For most practical purposes the last factor, which is very close to 1 for large , can be omitted. If n is an integer, we can write nl ~ yBrinve™ O) where ~ means “is approximately equal to for large n”. This is sometimes called Stirling's factorial approximation (or asymptotic formula) for n!. MISCELLANEOUS RESULTS INVOLVING THE GAMMA FUNCTION 1 ra@yrd-2) = gh 4, 1) = Ve as in (4). BT a) re +4) = Varn) This is called the duplication formula for the gamma function. In particular if 3. re)r(x +3) r (= + 2). . a(e + no) = msir~me (Baym Tma) ‘The duplication formula is a special case of this with m = 2. . 1 1 __189_ 4. C cae ae eh { * jae * 2a8e* ~ BI,B0e } This is called Stirling’s asymptotic series for the gamma function. The series in braces is an asymptotic series as defined on page 70. 5 rm) = S e*Ingdr = ~y where y is Euler's constant and is defined as lv 1 tim (a4+g+d+--- 43, -tnm) = 05772156...CHAP. 4 GAMMA, BETA AND OTHER SPECIAL FUNCTIONS 69 p+) ere eet) Tet ~ 1tet eto tp oy THE BETA FUNCTION The beta function, denoted by B(m,n), is defined by Bim,n) = S 2m-1(1— 2)" de (8) which is convergent for m>0, n>0. ‘The beta funetion is connected with the gamma function according to the relation 1(m) (0) Bom.n) = emdetad ® See Problem 4.12. Using (4) we ean define B(m,n) for m<0, n<0. Many integrals can be evaluated in terms of beta or gamma funetions. Two useful results are SP sint-19 cos19 a0 = yB(m,m = FEIN (10) valid for m>0 and n>0 (see Problems 4.11 and 4.14) and S Zhe = rra-» = O0; (6) Tat 1)=n!, n=1,2,8,.... (a) ree) - Scere = lim {eerie +) 1 feria = fm ae a 9 90) = [Ceres = } re) = 1rd) = In general, ru +1) = ese im EQUI) = aP(%). Then ra) = 2r@) aot ra) = 3r(3) = 3-21 = at 16) £8, ¢ ert) 42. Evaluate c) TES) ay nt (a) = St arg os o ry) — € cO) 16. B15CHAP. 4] GAMMA, BETA AND OTHER SPECIAL FUNCTIONS u 44. 4a. erg) _ s@arg _ + srg) srg 8 () [wer de (®) Let 22 =. ‘Then the integral becomes Soe -¢ Prove that T(J) = V. Woe rq) = fUeutesée = af rean antetine #48 1 follows one tray = {2 fe anh fe fe tanh = Af (eee aude Changing to polar coordinates (p,¢), where w= p cos ¢, v =p sing, the last integral becomes Evaluate (@) f° Vie" dy, () f sao f ome (@) Letting y® = 2, the integral becomes [Fern jeiite = [mere = inp 0) fra = fens ae = Fesemtaa, tet ingle = and the integral ‘becomes Sree (©) Let —Inz Then 2 =e" When becomes = 1 2Vaind 7h Lye aims avin The integral eiterds = when # Lu fe = [Teme = ap =F Evaluate xe" dz, where m,n, 4 are positive constants. LOT} = ae Evaluate (a) 1(-1/2), (b) 1(-5/2). ‘We use the generalization to negative values defined by r(x) = 2@+0)2 48. 49. 4.10. GAMMA, BETA AND OTHER SPECIAL FUNCTIONS (CHAP. 4 (a) Letting =~}, 1-12) = rae) __, 2) = ave. (8) Letting » = -9/2, r(-9/2) using (a). ve, r(=3/2) Then 1(-8/2) = "OH (cat Prove that ff" zn z)nde = where 1 is a positive integer and m>~1. Letting 2=e-¥%, the integral becomes (-1)* [yrenmtidy, If (m-+t)y =x, this last integral becomes Pg On gy = Up yy = Det CD areas = ene Soe = BaD = GehES Prove that Let 1 = -£ = a! ‘Thos £ co) Integration with respect to p yields or ry But C Thu He,0) required, FE) on letting 2 = 0x2, A particle is attracted toward a fixed point O with a force inversely proportional to its instantaneous distance from 0. If the particle is released from rest, find the time for it to reach 0. As time #0 lt the partie be at onthe sala at = @> 0 and Tet 0 be the origin. ‘hea by Newton te een ft were ie he mae of he pall and > 0a cotant of proportional Seo, the vecty of te pen 2 te _dede vet Gam te ty ot the patel, thy BE = a8 B= ofl na aa aaa molt = Eo Me tne te yon intestine, Since =O at ee, we fad’ ohne. ‘Tn Serene 2 = Bing ~ at (s)CHAP. 4) GAMMA, BETA AND OTHER SPECIAL FUNCTIONS B where the negative sign is chosen since 2 is decreasing as t increases, We thus find that the time T taken for t} Je to go from z=a to 2=0 is given by = fm (7 ae r= VES Se ® Letting nafs = or 200, this becomes 1s oY Rf itera = a fBnw = of THE BETA FUNCTION wn 4.1L. Prove that (a) B(m,n) =B(x,m), (b) Bim,n)=2 f sin™-*9 cos—19 de, (a) Using the transformation «=1—y, we have nina) = fletanaetae = faoweretay = flr aewetay = Baum (0) Uaing the tanaformatin = =n, we have Bina) = flemianartde = [%atom-voton-t 2am one de sin?™-1 9 cos®™—1 6 do ATE) pan > 0 412, Prove that — B(m,n) ea Similarly, rim) = 2 "y2-te-x* dy. Then rom) r(n) = “(fren ot ae) (fp te * av) =O feminine aeay ‘Transforming to polar coordinates, z = pcosg, v = psing, rine) = 7ST pase ne Paget a a ) fla semre) = arin {ene tte dy = Tmt) Bem = Pom+n) Bim,n) using the results of Problem 4.11. Hence the required result follows. The above argument can be made rigorous by using a limiting procedure,4 413. 414. 415. 4.16. GAMMA, BETA AND OTHER SPECIAL FUNCTIONS (CHAP. 4 Bvaluate (a) f2(1-2)'de, (6) S a © Suve=e ay. fo fieamande = m6 = MEME = ES (®) Letting 2 = 2v, the integral becomes 7 /2.1(8) r(1/2) bay? tp = V8 fe v)-vde = ay8Be,4) = ofr r0 2) Say (e) Letting y? = ate or y = aVz, the integral becomes a81(6/2) 18/2) _ ea zr) BB Gf, vea- areas = Fas, 1m) r(n) 2r(m+n) ‘This follows at once from Problems 4.11 and 4.12. Show that [" sin®™19 cost™-19 do m,n>0. Evaluate (a) S “into ds, (b) S "int 6 costo do, (c) S "cost 0 dé. (a) Let 2m—1=6, 2n-1=0, ie m= 7/2, 0 /2, in Problem 4.14 rA/2) _ Se end) a2" ‘Then the required integral has the value PO2)r) _ 8 See eee Brat) =4, 2n—1=5, the required integral has the value 0) Sete de = 2 [conte 2r(v/a) 18/2) _ 3 Thus, letting 2m— =4 in Problem 4.14, the value is 28) ee inva do =f cosra do = (a) $-3--P—=D 5 it » is am even posi- integer, (b)“4°8--"(P—) it » is an odd positive integer. From Problem 4.14 with 2m—1=, n—1=0, we have 72 eae - Pa@+nIrQ) GD sinvo ds = eae (@) If p=2r, the integral equals re+ pra) _ “ary = a) = Gr=ner Serer Bye 2 = (0) Tf p=2r+1, the integral equals res DP) eat BFP e+ PrP dvs Intorhcates favo ay = (cowed, ax sen by tineCHAP. 4] GAMMA, BETA AND OTHER SPECIAL FUNCTIONS ccd 417. Evaluate (a) S “cost odo, (b) S sin’ @ costa d8, (c) S ‘sin’ 6 do, (0) From Problem 416 the integral ale = 55 [compare Problem 46) (8) The integral equals So stot ~ snd ayde =f" sinto ae — sine ao ‘The method of Problem 4.15(6) can also be used. (©) The given integral equals 4 {sine de = ( ~ a at 418. Given S Drs" = saps anps Where 00, p>0, then hE = = Sx) + Raz) where Sie) = fk — or + Cyr etd ern) =n} Balt) = (I wD: ptm) fee du du — S,(x)| = lim aX{R,(z)| = 0. - (®) Prove that lim olf (©) Explain the significance of the results in (b). (e) Integrating by parts, we have = fita = Stoo fs —@+DIpya 20 that Similarly Jp41 BES + m+ Wyse By continuing in this manner the requtred result follows, @) 1R,(@)| = plp+1) ++ orm f~ prt 5 pet) w+m fh” pari = MBt):-@+n) ‘Thus im P@+M-(4n) _ y ey oat ‘Tim 2*|R,(0)] (©) Because of the results in (b), we can say that Slee = fb gman .. o gat tha i.e, the series on the right is the asymptotic expansion of the funetion on the left. ) 4.25. Show that erf (2) ~ We have ert (a) which gives the required result,78 GAMMA, BBTA AND OTHER SPECIAL FUNCTIONS (CHAP. 4 Supplementary Problems THE GAMMA 10N 118) 118/2) 425. Evaluate (@) give ©) “gees Pua) rare) 6/2). tan ind (eres, [VE a, [vem a = aff oo. 429. Show that f $20, Provethat (@) 1) = f'(nt)" "as, n> 0 7 1\" T(p+1) 0) fle (nd)de = PPR, ta, erate ©) fora, « [emare, @ {Vee 432 Evaluate (a) 1-7/2), (8) 1-1/8). 43%, Prove that lim {P(2)| = © where m= 0,1,2,3, 431. Prove that if m ie a positive integer, M(—in +) = gf PR VE Hint: Expand e¥*)— ~~ in a power series and replace the lower limit of the integral by — THE BETA FUNCTION 431, Evaluate (a) B(S,5), (0) BIB/2,2), (¢) BI1/3, 2/8). to. Pad Wy flea ane, wo [VTREd, «fu meas 43, Evataate (o) [nase au, 0) Js Fee Luo, Prove that f Lan, Brains ) [ante conteds, 6) feat ae Lee Bran 0) fsnteas, 0) fee satecHaP. 4] GAMMA, BETA AND OTHER SPECIAL FUNCTIONS 9 43, at 445. 446. aa. 448, 448, 450. 451. 432. 453. Prove that (a) [" Vianede=2/VB; 6) f” tanrede = FaeckE, O 0, J wenn Prove that 22. [Hint: Differentiate with respect to b in Problem 4.45;] - 2(2- Show that ert (2) 2( 21 _ at Obtain the asymptotic expansion iz) ~ Show that (0) Si(—z) Sila), (B) Sil Obtain the asymptotic expansions Sia) ~ sine gy Show that J,“ = Fipyain rl)" O a (b) Graph f(z) and its Fourier transform for a= 3. (a) The Fourier transform of f(z) is Fo) =f" sored84 FOURIER INTEGRALS AND APPLICATIONS (CHAP. 5 () ‘The graphs of f(z) and F(a) for a= 8 are shown in Figs, 5-1 and’5-2 respectively. fe) Fe) 54. (a) Use the result of Problem 5. to evaluate S b) De che val "sine ay, 0) educe the value of f° nM (c) From Fourier’ integral theorem, if Fe = So fwemau then fle) = BS” Fleer da ‘Then from Problem 6.3, . 1 8 ear dae v2 wo - 0 |l>@ The left side of (1) is equal to Af snesotmatg, 5 4" semaines 4g - ‘The integrand in the second integral of (2) is odd and so the integral is zero. Then from (2) and (2), we have . 7 kie () If 2=0 and @=1 in the result of (a), we have fltees since the integrand is even, 5.5. (a) Find the Fourier cosine transform of f(z)=e-™, m>0. (0) Use the result in (a) to show that s Se a ae (p>0, B>0)CHAP. 5) FOURIER INTEGRALS AND APPLICATIONS 85 (a) The Fourier cosine transform of f(z) = em is by definition Fou) = [Temes cu + @ sin au) (b) From (12), page 81, we have fe = 2 ie, S Se = gem Replacing «by 2, = by p, and m by p, we have = con po ze [Soke = Bee prapro 58. Solve the integral equation "aysinands = (i7# O8@S1 S10) sinew 2 = {9 at a no = (Croamen = [50 828 eae eee 1) = 297 raeyanae de = Ef‘ ao sowie 2x = sinz) THE CONVOLUTION THEOREM 5.7. Prove the convolution theorem on page 82. We have by defi nn of the Fourier transform no = [tare a0) = floors A Then Fae) = ff srsieween dee a Let u+w =< in the double integral (2) which we wish to transform from the variables (u,0) to the variables (12). From advanced caleulus we know that 2h tu de ® dude Sot5a 5a. FOURIER INTEGRALS AND APPLICATIONS, (CHAP. 5 where the Jacobian of the transformation is given by jae aul awn _ |") _ [ao 1 wuz |e ae 04 au az ‘Thus (2) becomes L SH (eae dnd Sof poe ae Jae = F ir flu) ole —w) au FU * 9} wine peg = ff) ete=0) due the conan off and Fla) Gle) From this we have equivalently fo FM (@) Ge} cesFla) Gla) da Show that ft =9*f. Let z-u=v. Then fro fla—») oto) do Sto e-waw = = flowsa-va = oF Solve the integral equation = ole) + S$” vewre—w du where g(z) and r(z) are given. Suppose that the Fourier transforms of y(z), g(x) and r(z) exist, and denote them by Y(a), G{e) and R(q) respectively. Then, taking the Fourier transform of both sides of the given integral equation, we have by the convolution theorem Yo = G0 +¥eRe) Ye) = Se gal Gay a (ee Gee eee mew = fel) = Ef {5 a assuming this integral exists. 5.10. Solve for y(z) the integral equation "_y(u)du _ 1 Suthte = wie exactCHAP. 5) FOURIER INTEGRALS AND APPLICATIONS 87 We have fa) - PimF x} t. fon making use of Problem 5.5(5). ‘Then, taking the Fourier transform of both sides of the integral ‘equation, we find } = +febe} ror (he le Yo)Eem = Zetec Yq) = Leow 3 cosar dy = Thus ye) = Ef” omviedde = Eee oF PROOF OF THE FOURIER INTEGRAL THEOREM . 5.11, Present a heuristic demonstration of Fourier’s integral theorem by use of a limiting form of Fourier series. Tet te = Be 3 (% cos + by sin 32) wee an = EF, soocon tau ant a = EY pu sin 2 a ‘Then by substitution of these coefficients into (1) we find : se = ES" naan + Ee S50 cos E(u 2) du ie we ssune tat ff \u)| du converges, the first term on the right of (2) approaches zero as 1S pm eoeB anaes a This last step is not rigorous and makes the demonstration heuristic. Calling Sa = =/L, (8) ean be written fe) = dim, sa F(n de) w where we have written Fo) ©) But the limit (4) is equal to . fey =f Fede = « which is Fourie integral formula, ‘This demonstration merely provides a possible result, To be rigorous, we start with the double {integral in (¢) and examine the convergence. This method is considered in Problems 5125.15, 512 Provethat: (a) tim (SM? gy = (ott oo=n then tm fa = mf ieee eae arte88 FOURIER INTEGRALS AND APPLICATIONS (Har. 5 @) tet v= —y. Then tim J" 88a = Jim f" BBMdy = 5. 543. Riemann’s theorem states that if F(x) is piecewise continuous in (a,b), then lim L Fl2)sinex dz = 0 with a similar result for the cosine (see Problem 5.41). Use this to prove that @ tim {fer sina 4 = Efe+o) 7 2 4, ®) im f", fe+) where f(z) and /"(x) are assumed piecewise continuous [see condition 1. on page 80). zile-0) (@) Using Problem 5.12(a), itis seen that a proof of the given result amounts to proving that tim {gern ~ pero sina ° ‘This follows at once from Riemann’s theorem, because F(v) = eee is piecewise continuous in (0,L) since lim F(v) exists and f() is piecewise continuous. (8) A proof of this is analogous to that in part (a) if we make use of Problem 5.12(b). 5.14, If /(z) satisfies the additional condition that f” |f(2)|de converges, prove that i 7 a a @ iim f He +o) M2 dy =F 4240), (0) tim [fern Re dy = SH(a 0), (@) Wehave Sreroitstin = [pecotiitees (pero — w Lt es = Liner snerar + frero Bed Subtracting, {e+ »)— fle+ 0} {le +v) — fle +0) es ree Denoting the integrals in (8) by I, ,,[z and 1, respectively, we have I Now wos Sf verso" eet] pt Igt ly so that Ml = inl + Uhl + inl ® sine e+e) LS, eer vie Also sl since fee and" 82%ay bath converge we can che Eo Ire that = HIs| S/8._ Also, we ean choose « so large that |fy| $ «/3. ‘Then from (4) we have [I| S« for a and L sufficiently large, so that the required result follows.CHAP. 5} FOURIER INTEGRALS AND APPLICATIONS 89 (b) This result follows by reasoning exactly analogous to that in part (a) 5.15. Prove Fourier’s integral formula if f(z) satisfies the conditions stated on page 80. fat) +fle-0) We ma rove tat mn ff settee ede sine | $f snate—sa] =f Med whic omen, ftw By the Wel ru tt tor erate) converses hla nity foal an ‘We can show from this that the order of integration can be reve 1 aa fT _sdeosate-witu = E67 sonde foaled where we have let = as required. SOLUTIONS USING FOURIER INTEGRALS 5.16. A semi-infinite thin bar (x = 0) whose surface is insulated has an initial temperature equal to f(«). A temperature of zero is suddenly applied to the end 2=0 and maintained. (a) Set up the boundary value problem for the temperature u(x, ¢) at any point at time t. (b) Show that ue,t) = 1” SL soe" sina sin dada do (@) The boundary value problem is - 2>0, t>0 a u(e,0) = fle), 00,8) where the last condition is used since the temperature must be bounded for physical reasons. inet] 0 (Fig. 5-8) if » takes on the value f(x) on the z-axis, ‘The boundary value problem for the determina- tion of v(z, y) is given by te ye at tye = ° ‘v(z,0) = He), |vle,y)| 0 the'term in elvis unbounded as y->~; so that to keep (x,y) bounded we must have 4, = 0. This leads to the solution v(zy) = em M(A cosaw + B sin nz] ‘Since there is no restrietion on \, we ean replace A by A(\), B by B(x) and integrate over A to obtain vie) = "eriagy eae + BOY nie) a The boundary condition v(e,0) = fle) yields S, (40 cose + BO) sinaelan = pay Thus, from Fourier's integral theorem we find 40) = EFC Hw conway, — Ba) = Putting these in (1) we have finally: sia) BTC tiene aay i 5.20. Show that the solution to Problem 5.19 can be written in the form ven) = ES” tas ae Write the result (2) of Problem 5.19 as 26 Mo [fe ewite-aa a fe ‘Then by elementary Integration we have 0) SF meow -9a = potog ® 0 that (1) becomes vem = 2S" aw cc)92 FOURIER INTEGRALS AND APPLICATIONS (CHAP. 5 SOLUTIONS BY USE OF FOURIER TRANSFORMS 5.21. By taking the Fourier transform with respect to the variable z, show that @ #2) = aro, & #(E w) = FF) (2) By definition we have on using integration by parts: r(#) = = iaFe) where we suppose that v-> 0 as > te, (8) Let w= du/ta in part (a) then 0) gag (Bt) = Gal Fe x(2a) = ae(3z) = Gere Then if we formally replace w by » we have +(23) = GaP Fe) =e FO) provided that v and 59 as 2+ +». ae In general we can show that (a0) ' ne) (ayn FO wt it oF So as cote (©) By definition 5.22. (a) Use Fourier transforms to solve the boundary value problem oe oe = Se = «SS, ule, 0) = fle), |u(est)| < where —©0. (b) Give a physical interpretation. (a) Taking the Fourier transform with respect to x of both sides of the given partial differential equation and using results (6) and (2) of Problem 6.21, we have a Frm = aver o ‘where we have written the total derivative since Flu) depends only on ¢ and not on 2, Solving the ordinary differential equation (1) for Flu), we obtain Flu) = Comvats @ or more explicitly Flulz,t)) = Cennate @CHAP. 5] FOURIER INTEGRALS AND APPLICATIONS 98 523. Putting ¢=0 in (8) we see that Flute, 0) = FU} = C « s0 that (2) becomes . Flu) = Fifjenea"e ) We can now apply the convolution theorem, By Problem 4.9, page 72, ‘a0 7 -eme =f pore es Set owt = Ff Hence uz.) = fla) tatu de o (eum nO) brome ee (ay = Lf” eene-20/m) B (8) The problem is that of determining the temperature in a thin infinite bar whose surface is ingulated and whose initial temperature is fl). An infinite string is given an initial displacement y(z,0) = f(z) and then released. Determine its displacement at any later time t. ‘The boundary value problem is ey pay ae oo wo we,0) = lz), y0) = 0, |y(e,t)] 0, Letting y = XT" in (1) we find in the usual manner that a solution satisfying the second bound- ary condition in (2) is given by we,t) = (A cosdx + B sin x) cos rat By assuming that A and B are functions of \ and integrating from A=0 to © we then arrive ‘at the possible solution vt) = ff becare + 6 sna cnraten ® fo) = ff wore + 9 snr hen few fom U) An pe hat 40) =2f"fereasede, Bo) = 167 Hep anreae wr where we ave changed te dummy vara rom 219 Substitution of (4) into () yields vee = 267 fT sore nen + cnn siny] eannt doch = ES J sor coeate=o crate = EGS sntennereen n+ ceneet— opera94 FOURIER INTEGRALS AND APPLICATIONS [oHaP. 5 where in the last step we have used the trigonometric identity cos A cos B }leos (A +8) + cos (A ~ BY] with A=ale~2) and B= rat By interchanging the order of integration, the result ean be written ve) = EL” fH comnts +at— waver ae + EGF easnee at 0 aver 6 But we know from Foutier's integral theorem (equation (3), page 80] that ne) = EL” Ss cosne- ava ‘Then, replacing = by #++at and zat respectively, we see that (5) can be written vit) = Biletay + feat) ® which isthe required solution, Supplementary Problems ‘THE FOURIER INTEGRAL AND FOURIER TRANSFORMS 524. 5.25. sat. 5.28. [tle [ah 0 by using the result in (a) (c) Explain from the viewpoint of Fourier’s integral theorem why the result in (b) does not hold for m= 0. Solve for y(z) the integral equation Cunsinwte = fa ister ° 0 tee and verify the solution by direct substitution.CHAP. 5] FOURIER INTEGRALS AND APPLICATIONS 95 5.28, If Fla) is the Fourier transform of f(z) show that it is possible to find a constant ¢ so that Fle) = f(a) = 0 PARSEVAL'S IDENTITY 530, Bratoate (@) fH, 6) [EES ty use of Parseval's entity. [Hint Use the Fourier sine and cosine transforms of ¢~*, «> 0.) S21, Use Problem 528 to show that (a) f° (= S82) "ae Sat. Showshat (“te ana 533, Prove the results given by (a) equation (13), page 82; (b) equation (14), page 82. 534, Establish the results of equations (15), (16), (17) and (18) on page 82. CONVOLUTION THEOREM fl ieica 538, Verity the convolution theorem for the funetions fle) = o(2) = {9 aio. 536." Verify the convolution theorem for the functions f(e) = gl) = e~*. S41. Solve the integtal equation f'" y(u)y(z—u) du = e-*, 538 Provethat f*(g th) = f*g + fth. 538, Prove that /*(g*h) = (/*9)*h. PROOF OF FOURIER INTEGRAL THEOREM SA By interchanging he order of integration inf, fe sny ded, rove tnt and thus complete the proof in Problem 5.12 SAL. Let n be any real number. Is Fourier's integral theorem valid for f(z) = e-#"? Explain, SOLUTIONS USING FOURIER INTEGRALS 542, An infinite thin bar (—= <2<) whose surface is insulated has an initial temperature given by uy [al 0) has an initial temperature given by f(z) = uge-*. If the plane face («=0) is insulated show that the temperature at any point x at any time t is Viti we.) = oye 4ee) Solve and physically interpret the following boundary value problem: 0 y>0 “1 <0 cu MeO = Ty goo Meal OO iy Problem 6.4, th Show that it w(e,0) = {0 25) in Potten sy then Mg gga wen = 4 Bente 0 act Work Problem 6.44 if ua,0) 1 ft 0, y>0 has one edge =~ 0 kept at potential zero and the other edge y=0 kept at potential /(2). (a) Show that the potential at any point (x,y) is given by 1” 1 1 7] von = Tf vo lsaare ~ wearee] ©) Ue fe) =, show tat o,9) = Ztas-rd Verify that the result obtained in Problem 6.18 is actually a solution of the corresponding boundary value problem. ‘The lines y=0 and y=a in the ay-plane (see Fig. ¥ 5.4)-are kept at potentials 0 and f(z) respectively. 0,0 ‘Show that the potential at points (x,y) between these x z Tien 1 sien by ee fe ee - wes) Ee, it * hay = FS, So 10 BEM con nea aur rea ‘An infinite string coineiang with the -axis is given an intial shape f(z) and an initial velocity (2). Assuming that gravity is neglected, show that the displacement of any point « of the string at time t is given by ves) = Eitan +sa—an) +E pada ‘Work Problem 5.50 if gravity is taken into account, A. semi-infinite cantilever beam (x >0) clamped at s = 0. is given an initial shape f(c) and released. Find the resulting displacement at any later time t,Chapter 6 Bessel Functions and Applications BESSEL’S DIFFERENTIAL EQUATION Bessel functions arise as solutions of the differential equation aty” + ay’ + (any = 0 n=0 @ which is called Bessel’s differential equation. The general solution of (1) is given by (at) + CY a (a) (2) y ‘The solution Jn(z), which has a finite limit as «x approaches zero, is called a Bessel function of the first kind of order n. The solution Yq(z), which has no finite limit (ie. is unbounded) as # approaches zero, is called a Bessel function of the second kind of order n or a Neumann function. If the independent variable « in (1) is changed to Ax, where A is a constant, the resulting equation is ay” + xy! + (Nx ny = 0 (3) with general solution Y = eida(A) + ¢2¥a(A2) D) ‘The differential equation (1) or (8) is obtained, for example, from Laplace's equation Viv =0 expressed in cylindrical coordinates (,$,2). See Problem 6.1, THE METHOD OF FROBENIUS An important method for obtaining solutions of differential equations such as Bessel's equation is known as the method of Frobenius. In this method we assume a solution of the form y= 3 eave (5) where c,=0 for k<0, so that (5) actually begins with the term involving co which is assumed different from zero. By substituting (5) into a given differential equation we can obtain an equation for the constant (called an indicial equation), as well as equations which can be used to determine the constants cx. The process is illustrated in Problem 6.3. BESSEL FUNCTIONS OF THE FIRST KIND We define the Bessel function of the first kind of order n as Jee) = ore aft Bante * \ 6) 9798 BESSEL FUNCTIONS AND APPLICATIONS (CHAP. 6 or ie) = & eRe ” where T( +1) is the gamma funetion (Chapter 4). If n is a positive integer, T(n +1) = T(1)=1. For »=0, (6) becomes a x I(x) = 1 +op-opet @) ‘The series (6) or (7) converges for all z. Graphs of Jo(a) and J,(z) are shown in Fig. 6-1. If » is half an odd integer, Ja(x) can be ex- pressed in terms of sines and cosines. See Prob- Jems 6.6 and 6.9. A funetion J-.(2), n>0, can be defined by re- placing n by —n in (6) or (7). If n is an integer, then we can show that (see Problem 6.5) Jona) = (-1)a(2) (9) Fig. 6-1 If n is not an integer, Jn(z) and J-»(2) are linearly independent, and for this case the general solution of (1) is y = Ady (2) + BU -9(2) n#0,1,2,8,... (0) BESSEL FUNCTIONS OF THE SECOND KIND We shall define the Bessel function of the second kind of order n as { Ja(x) cos m: = 04 0,1,2,8, 006 Yale) = (1) n=0,1,2,8,... For the case where n= 0,1,2,3,... we obtain the following series expansion for Ya(z): 2 1 R= HD) (aay Yala) = F{ln (2/2) + yale) — = yew Wen we (12) -1 5 -1 a yyy (rl2yetn 5S, ( DMO) + 800 by where y= 0.5772156... is Euler’s constant (page 68) and o) = 145+ 9(0) = 0 (13) Graphs of the functions Yo(x) and Yi (2) are shown in Fig. 6-2. Note that these functions, as well as all the functions Y,(z) where > 0, are un- bounded at « = 0. ? If n is half an odd integer Y,(z) can be ex- pressed in terms of trigonometric functions. Fig.6-2CHAP. 6] BESSEL FUNCTIONS AND APPLICATIONS 99 GENERATING FUNCTION FOR J,(z) The function el) = SY alee (4) is called the generating function for Bessel functions of the first kind of integral order. It is very useful in obtaining properties of these functions for integer values of n—proper- ties which ean then often be proved for ali values of n. RECURRENCE FORMULAS ‘The following results are valid for all values of n. 1 Savs(a) = 2BSu(@) = Junale) Ha) = Ride-r(2) — Jars @)] 3 sua) = WSa(at) ~ 2Tuss(t) ada) = tn nd n(zt) Hier) = 2-12) 6 A le-ea(a)] = —a-nes(2) If n is an integer these can be proved by using the generating function. Note that results 8. and 4. are respectively equivalent to 5. and 6. ‘The functions Ya(z) satisfy exactly the same formulas, where Y_(z) replaces J (2). FUNCTIONS RELATED TO BESSEL FUNCTIONS 1. Hankel functions of the first and second kinds are defined respectively by Hyu'(@) = Jala) +i¥alz), He (x) = In(x) —i¥a(x) (15) 2 Modified Bessel functions. The modified Bessel function of the first kind of order n is defined as n(x) = inf) = e°9T, (iar) (16) If n is an integer, Lala) = In(z) «a but if mis not an integer, In(x) and I-s(z) are linearly independent. ‘The modified Bessel function of the second kind of order n is defined as Ton(2) — Inf one ae] HA O12, ca lim 3 Pn hay) = 0,1,2,3, nel sinpe 12,8, These functions satisfy the differential equation ay" + ay — (ttn = 0 (9)100 BESSEL FUNCTIONS AND APPLICATIONS [CHAP. 6 and the general solution of this equation is ¥ = eidn(2) + eK a(e) (20) if n40,1,2,8,..., cea Y= Als(a) + Bl-n(2) (et) Graphs of the functions Io(z), (2), Ko(z), Ki(a) are shown in Figs. 6-3 and 6-4. y Kye) Ke) Fig. Fig.6-4 3% Ber, Bei, Ker, Kei functions. ‘The functions Ber, (x) and Bei, (z) are respectively the real and imaginary parts of Ja(i%*x), where i®? = 4 = (\/3/2)(-1+ i), ie, In(i2x) = Beta(t) + i Bein(z) (22) ‘The functions Ker,(z) and Kei, (x) are respectively the real and imaginary parts of eK, (i822), where ii? = evi = (V/B/2\(1 +4), ie. emieK (ix) = Kery (a) + iKeia (a) (23) The functions are useful in connection with the equation wy” + ay — (xt try = 0 (24) which arises in elec equation is il engineering and other fields. ‘The general solution of this Y= cada(i%x) + cok (ia) (25) If n= 0 we often denote Bers (z), Bei, (x), Kers (2), Kei, (t) by Ber (z),Bei (x), Ker (2), Kei (z), respectively. The graphs of these functions are shown in Figs. 6-5 and 6-6. y Fig.6-5CHAP. 6] BESSEL FUNCTIONS AND APPLICATIONS 101 EQUATIONS TRANSFORMABLE INTO BESSEL’'S EQUATION ‘The equation ay + (2k+ lay’ + (ax + By = 0 (26) where k, a, 7, # are constants, has the general solution y = a [eWer(ax’lr) + c¥er(ax"/r)] (27) where «=F —. If #=0 the equation is an Euler or Cauchy equation (see Problem 6.79) and has solution y= 2 Nest eye“) (8) ASYMPTOTIC FORMULAS FOR BESSEL FUNCTIONS For large values of we have the following asymptotic formulas: Tula) ~ [Ecos (= -3-%), Ya(x) ~ Zan(e-3-™) (29) ZEROS OF BESSEL FUNCTIONS We can show that if m is any real number, J,(x)=0 has an infinite number of roots which are all real. The difference between successive roots approaches + as the roots increase in value. This can be seen from (29). We can also show that the roots of Jn(z) = 0 [the zeros of J,(x)] lie between those of Jx-1(z)=0 and Je+i(z)=0. Similar remarks can be made for Y,(x). For a table giving zeros of Bessel functions see Appendix E, page 177. ORTHOGONALITY OF BESSEL FUNCTIONS OF THE FIRST KIND If A and » are two different constants, we can show (see Problem 6.28) that Stoo) dule2) dx = sols) Fata) e aplivitay (80) while (see Problem 6.24) Sona = Zito + (1-2)20] (on From (30) we can see that if A and » are any two different roots of the equation RJa(a) + Sadn(a) = 0 (82) where R and S are constants, then S Jn(2) In(ur) de = 0 (83) which states that the functions V#Jq(\2) and V2 Js(u2) are orthogonal in (0,1). Note that as special cases of ($2) A and can be any two different roots of Ju(z)=0 or of Ji(2) =0. We can also say that the functions Jn(Az), Jn(ut) are orthogonal with respect to the density or weight function «.102 BESSEL FUNCTIONS AND APPLICATIONS (CHAP. 6 SERIES OF BESSEL FUNCTIONS OF THE FIRST KIND As in the case of Fourier series, we can show that if f(x) and /’(z) are piecewise con- tinuous then at every point of continuity of f(x) in the interval of 0 at ‘Then equation (12) of Problem 6.1 which ean be written eP ap Ge + eG + Ot wP = 0 becomes () ty @pe + @-ay = 0 or ty” + xy + (Py 0 as required.104 63. BESSEL FUNCTIONS AND APPLICATIONS (CHAP. 6 Use the method of Frobenius to find series solutions of Bessel’s differential equation ay’ +2y' + (et njy = 0. Assuming a solution of the form y = 3 eqztt® where k goes from —= to © and c= 0 for <0, we have (tiny = Seek t0t2 — Satake = Sey gett® — Saree ay! = Det peer? sy = Zk aylk+ B= Deets ‘Then by addition, DUH MRA Aee + +A + ona — wget = 0 and since the coefficients of the 2t*# must be zero, we find (e+ AP mle + e-2 = 0 o Letting k= 0 in (1) we obtain, since ey =0, the indicial equation (#?—n%)¢y = 0; or assuming cy) +0, p2=nt, Then there are two cases, given by ‘and p=. "We shall consider first the case =n and obtain the second ease by replacing n by —*. Case: p=. In this case (1) becomes Kn Key + eee = 0 ® Putting &=1,2,8,4,... successively in (2), we have Bonar = Gata) ~ TA VW H’ =O ee Thus the required series is Y= cat + ogett? + cent + a = eles mena“) ® Case 2: p= — On replacing n by —n in Case 1, we find v eel gp a5 t ] ® Now if m=O, both of these series are identical. If n= 1,2, However, if 1 0,1,2,... the two series can be shown to be linearly independent, and so for this ‘ease the general solution is “ ae ay * aaa] Fant + Desf = ‘The cases where m= 0,1,2,3,... are treated later (see Problems 6.17 and 6.18). The first series in (5), with suitable choice of multiplicative constant, provides the definition of In(2) given by (6), page 87. BESSEL FUNCTIONS OF THE FIRST KIND 6A, Using the definition (6) of Jx(z) given on page 97, show that if n*0,1,2,..., then the general solution of Bessel’s equation is y = AJn(z) + BJ—a(2).CHAP. 6} BESSEL FUNCTIONS AND APPLICATIONS 105 66. 67. Note that the definition of Jn(2) on page 97 agrees with the series of Case 1 in Problem 6.3, apart from a constant factor depending only on x. It follows that the result (5) ean be written ¥ = AJq(2)+BJ_q(2) for the cases *0,1,2,. (a) Prove that J-a(z) =(-1)Va(t) for n=1,2,3, (b) Use (a) to explain why AJ, (x) +BJ-»(2) is not the general solution of Bessel’s equation for integer values of n. (@) Replacing n by —n in (6) or the equivalent (7) on page 98, we have (ayrer2y TeRtr Jon(2) y (Aria/ay-m 42" S oayrle/a)-n + or Peatrey + 3, erenty tay +r Now since I-w+r+1) is infinite for_r = 0,1, Letting r=n+k in the second sum, it becomes ym —1, the first sum on the right is zero, Se Se 3 apie @+OIEFD ~ Fy Fe tke De (6) From (a) it follows that for integer values of », J_,(z) and J,(z) are linearly dependent and 80 AJ,(z) + BJ_y(2) cannot be a general solution of Bessel's equation. If n is not an integer, then we can show that J_n(x) and J(z) are linearly independent, 20 that AJ,(z) + BJ_»(2) is fa general solution (see Problem Prove (a) Jia(x) (0) J-rn(@) = fafa (a2) (@ yale) = ie + B78) Vary ~ ier * = RN __ (alae (as ~ Trees . wrayt fy ~ od = Gl) sine (awe ttt Br “awe ® - S eprerey-vase _ (eyay-2 tanya - ® Joint) 2 ries = ray ~ Trem * Bre } = Prove that for all n: @ Lene = = & irate a © even a106 68. 69, 6.10. BESSEL FUNCTIONS AND APPLICATIONS [CHAP. 6 ag (ayer = Hy (oayren tem TAI w FFD ® feo = Prove that for all n: (©) Fila) = Fa-x(2)—Iuss(ay, (0) Tunsla) + Jass(a) = 2 Fae). From Problem 6a), 2"F{(z) + ne*-\Jq(z) = 2Jy_ (0) or sDg(2) + Syl) = dy sl) « Prom Problem 6.7(b), 2-"Sj(2) — m2" “"Iy(2) = 2-2) or ADL) — ndyle) > dy 2) @ (0) Adding (1) and (2) and dividing by 2 gives MY =F yer) ~ Juy i] (®) Subtracting (2) from (1) and dividing by « gives Intl) + Insite) = Be Show that (a) Jua(x) = - (2 zee) () J-salay = Vz (came sees) (@) From Problems 6,8(6) and 6.6 we have on letting » /2, Jonl) = Pale) ~ Finis) = af 2(#22—cons) = af E(sineqecoe (©) From Problems 68(8) and 6.6-we have on letting =~}, Jani) = ~9[2(esaet sas) Evaluate the integrals (a) 2'Jy-.(2) ) f BO) ae, From Problem 6.7, (0 Stenson = edyseh Then favgitelde = ay 6 © ftengtey = eval. then f Betas = -enage) +eCHAP. 6) BESSEL FUNCTIONS AND APPLICATIONS 107 GAL. Evaluate (a) f2(a)dz, (6) f 2a) de. (@) Method 1, Integration by parts gives Satine = fonienierey = Healey) — f ee colzeds) = eine) = 2 f ayer wy(a) — DNy(a) +6 Method 2. We have, using Jy(s) =—J(2) [Problem 6.718) Sanwa = -femae = ~{om-f te) ae} Srvserde = fetanerd) = atesioy — f wr terteede) Senwee = -faiid = fen - f sate) ae} = -24y(e) + 2ety(0) Then fi atiylayde = —2Adgle) + Ale (a) — 2{-z¥o() + 2e(2))] + = (8x2 —x8\Jo(x) + (4x — 16x )J, (x) % Sererae =f a[e-uyteae] = Herat) = f (ene Sina Note that we define the Bessel function of the second kind if n is not an ineger by In(e) cosne — Jan(2) ¥a(@) — (8) ‘The expression Info) cose = Jule) becomes an “indeterminate” of the form 0/0 for the cage when n is an integer. ‘This is because for an integer m we have cosms=(-1)" and J.(x) = (—1)"J,(2) [see Problem 6.5]. This “indeterminate form” can be evaluated by using, Utfnptats rae, Le. Fol cose — Jy] ‘Ip(z) cos pr — J-pl2¥ in ee] = Bn Pe] ‘This motivates the definition (12) on page 98, 6.18. Use Problem 6.17 to obtain the general solution of Bessel's equation for n = 0. In this case we must evaluate Jp) con pr — Jp(2V Pe] ® Using L'Hospital’s rule (differentiating the numerator and denominator with respect to p), we find for the limit in (1) (0J,/@p) cos pe py [ etree = ago) Pade ee) where the notation indicates that we are to take the partial derivatives of Jp(z) and J_p(2) with respect to p and then put p=0. Since aJ_y/2(—p) = ~aJ_y/@p, the required limit is also equal to au, ? we tr of tenapeer | Z © a{ePeea} ® lays Now it we tt GEIB — 6, then in = (p-+20) nC) Innip-er-+1) ao that der entiation with respect to p gives 1a _ _ Mptrey) Gop In@) Tip+r+i) ‘Then for p= RP term — MEER a ie+D) reFi)CHAP. 6] BESSEL FUNCTIONS AND APPLICATIONS ut Using (2) and (8), we have 2S (urea rr+a) ER Green 6” ~ Ren) " 2 one + roe + 2[S- FeO 4D + gl tet where the last series is obtained on using the result 6. on page 69. This last series is the series for Yo(z). We can in a similar manner obtain the series (12), page 98, for ¥q(2) where m is an integer. The general solution if » is an integer is then given by y= evJ,(2) + €2¥ n(2)- FUNCTIONS RELATED TO BESSEL FUNCTIONS 6.19, Prove that the recurrence formula for the modified Bessel function of the first kind 1n(2) is Tnvi(a) Tn-1(z) — 22 12) From Problem 6.8(6) we have Jner(e) = Bplay ~ ya) oy Replace z by ix to obtain 7 Inert) = Bate) — Jeaatisd ® Now by definition I,(2) Julia) oF Jn(iz) = i*Iq(2), 0 that (2) becomes 2in fg (2) inlga) — -1T4(2) Dividing by "+1 then gives the required result 6.20. If nis not an integer, show that @ He @ = en dn(2) 2 e7Jn(2) — J=n(2) fanne—*(() He (@) sine (@) By definition of H{?%(a) and ¥,(2) (see pages 99 and 98 respectively) we have HP @) = Jae) + Ya) = Jute) + [22% ane = i] Jnl) sin ne + Wyle) c08 ne (2) Jnl) (008 tr [eee ee] = Lalaeerste @) Since HiP(2) = J4(2)—4¥,(2), we Sind on replacing i by —i in the result of part (a), ry an(e) = eedn(2) _ end (2) ~ Jonla) Hq) = Tal) te) dnl) = Jnl) =1-# 4 58 -... 621. Showthat (a) Ber(2) = 1 ~ 35 + slags - () Bei) = S- 5% = 2” See + ppeRIT2 BESSEL FUNCTIONS AND APPLICATIONS (CHAP. 6 We have Pat a, Gay U8, GE sets) = z a ~ aiaiee + eee raft yet ite _ a gt ~ Ree? * Bees = 1¢@ + ot 2 me pee * Bee zt Fa 2 \ (: ~ page * ee ) a G weet") the required result follows on noting that Jo(i¥%s) = Ber(s) +i Bel(#) and equating real and imaginary parts. Note that the subscript zero has been omitted from Berg (x) and Beig (2). EQUATIONS TRANSFORMABLE INTO BESSEL'S EQUATION 6.22, Find the general solution of the equation xy"+y’+ay = 0. ‘The equation can be written as zty’'+ 2y'+azy =0 and is a special case of equation (26), page 101, where k=0, @=Va, r=1/2, #=0. Then the solution as given by (27), page 101, is v= ey(@Vae) + og¥o(2VGx) ORTHOGONALITY OF BESSEL FUNCTIONS 628, Provethat f" eJa(\) Ju(ya) de w20) Fle) = Mele) if avn From (8) and (4), page 97, we see that 14 =J,(\4) and ys=Jq(s2) are solutions of the equations Bay + avi + Om, = 0, atu + ah + (tate ‘Multiplying the first equation by yz, the second by y, and subtracting, we find ° Uva! + Averill = 9 Which on division by z can be written as = Zivevt— wil + wai vwil = Mw or Stevan) = enn Then by integrating and omitting the constait of integration, 239) f evade = zlvavi- mu) or, using yi =Jq02), ¥2=Jq(ea) and dividing by st—a* 0, S102 te = BDI Au) Tet) — wy Ot) Jeu) Mae 1 Jule) T0) — byl) Ton ‘Thos Sfst00 ann de = Wee) = ott which is equivalent to the required result 624. Prove that Sf 20a) ae af ya 2) 3 [vee + (1 =) sa] . Let 47-2 im the result of Problem 6.28. Then, using Hospital's rule, we find Si ettony de = yy Mil zOY— dno) — ood MIB) = InQY TA) = nd JNO)CHAP. 6 BESSEL FUNCTIONS AND APPLICATIONS 18 But since 2257/0) + 9/0) + 0Q2—n8)J,() = 0, we find on solving for Jj) and substituting, ahoade = $[ 720) + ( 2)t0] Prove that if A and . are any two different roots of the equation RJ, (2) + SxJa(z) = where R and S are constants, then Sf =Iot02) Fol) dx = 0 ise, VEJn(A2) and VFJu(ux) are orthogonal in (0,1). Since d and 4 are roots of RJq(2) + SxJ;(2) = 0, we have BIg0) + SAO) = 0, RI ql) + SeIh) = 0 Oy ‘Then since R and S are not both zero we find from (1), wdn(0) Ine) — Iu(ud Tua) = 0 ‘and so from Problem 6.23 we have the required result, S02 4ulo0) de = 0 SERIES OF BESSEL FUNCTIONS OF THE FIRST KIND 626. If f(a) = Y ApIa(Aox), O Rada Oad = 3 SOLUTIONS USING BESSEL FUNCTIONS OF THE FIRST KIND 628. A circular plate of unit radius (see Fig. 6-7) has its plane faces insulated. If the initial temperature is F(p) and if the rim is kept at temperature zero, find the temper- ature of the plate at any time. -e the temperature is independent of g, the boundary value problem for determining au aug 1 ae ae (es 2) % u(1,@ = 0, tle,0) = Fle), |ulp, | < Me Let w= P(p) T(t) = PT in equation (1). Then or = oon spr) or dividing by «PT, r T from which reat =o matmanp = ‘These have general solutions T= ee", P =" Aydg0p) + B,Yo0) Fig.6-7 Since u=PT is bounded at p= 0, B,=0. Then lp.) = Ae-™* Fy (ro) where A= Aye, From the first boundary condition, ul) = Ae-™5) = 0 from which Jo(3)=0 and 4=n,d3,-.. are the positive roots, ‘Thus a solution is tWlo,8) = Ae BT On) 1,2,3, By superposition, a solution is M,t) = 3 Ane-O'Jo (qo)CHAP. 6] BESSEL FUNCTIONS AND APPLICATIONS 15 From the second boundary condition, ule, 0) = Fie) = ‘Then from Problem 6.26 with »=0 we have 2 An = BEG S, (One de [EK te t000 06] eatncoo} ® which can be established as the required solution, Note that this solution also gives the temperature of an infinitely long solid eylinder whose convex surface is kept at temperature zero and whose initial temperature is Fp). and so 629. A solid conducting cylinder of unit height and radius and with diffusivity « is ini- tially at temperature f(p,) (see Fig. 6-8), The entire surface is suddenly lowered to temperature zero and kept at this tem- perature. Find the temperature at any point of the cylinder at any subsequent time, Since there is no g-dependence, as is evident from symmetry, the heat conduction equation Is ou ‘au, Law, ue # - (tess) o where w= up,2,). The boundary conditions are given by Fig. 6-8 We, 2,0) = flo.2) 40,0,8) = 0, lo, 1,8) = 0,” wllz 8 = 0 lulose] 0. ‘To solve this boundary value problem let U = PZT = P(s)Z(2) T(t) in (1) to obtain Por = (war + par + par) ‘Then dividing by »,(2)a, (4) = ~) rs oa Thus (6) becomes a = Gl (2) _ eft y(? ‘a ” n= Mite (I) _ wl mp ) Now from formula 8 page 99, with m=1 and the corresponding formula involving Y, for m= 1, wwe have from (7) @) ‘Using these in (8) we thus find Vioe 2223) »(Yevera) - v2) 1 (28 vera) | O inteCHAP. 6) BESSEL FUNCTIONS AND APPLICATIONS 128 Supplementary Problems BESSEL FUNCTIONS OF THE FIRST KIND ; ae Py se 635. (a) Show that Jy(e) = 3-35 + Sloe — sis 4 gence is —2 cece, (6) Show that Jj(2) and verify that the interval of conver- ) (0 Show tnat fate) = alte 637, Evaluate (a) Jg;(z) and (#) J_sa(e) in terms of sines and cosines, 638, ind Ja(2) in terms of Jo(x) and Jy(#). 639, Prove that (0) Ji(2) = }ldq-2(@) ~ 2Iy(2) + Jyy 22) ©) Ta) = Wy al@) — By (2) + na (2) ~ Fu al@)] ct ceeate tee ess cm Brame) fant, @ fend @ f annde Sal) Gy, Gat, Evaluate (@) fnYzrae, 0) f° 642, Evaluate f Jp) sinz dx. 2einne 643. Verity directly the result J(2) J—q(2) ~ J! (2) Ju(2) for @ nek ad GENERATING FUNCTION AND MISCELLANEOUS RESULTS 64, Use the generating function to prove that Ji) = 4(J,-x(2) + Jyrs(e)] for the case where x is an integer. 645, Use the generating function to work Problem 6.89 for the case where m is an integer. 645, Show that (0) 1 = Jo(2) + 25y(2) + 2dg(x) + © Ile) ~ Jala) + Isle) ~ Je) +o = Paine G47. Show that ZJ,(o) = Jae) ~ 244(0) + BJe(2) — 648 Show that Ja(e) : 40. storsiat @ ["eends w ‘Jo(e #in 6) cos sin 9 do wa Ae),124 650. 651. 652. 653, BESSEL FUNCTIONS AND APPLICATIONS [cHaP. 6 Show that f"Jo(e)ae = 2S Jaevster Soviet 1) [Trentbede = 0) [Cem aibarde = Show that f“Jg(e)dz = 1. Prove that |J,(2)| 1 for all integers n. Is the result true if m is not an integer? BESSEL FUNCTIONS OF THE SECOND KIND 654. 655. 656. 6st. 658, 659. 650. Show that (a) Yussl2) = 2EY,(@)— ysl, @) Yate) = $[%a1l2) — Yavill Explain why the recurrence formulas for J,(z) on page 99 hold if J,(x) is replaced by Y¥q(2) Prove that Yg(x) = -¥y(a). Evaluate (a) Yin(2), @) Yosa(2l, (© Yale), (@ Y¥-sale). Prove that J,(2) ¥gle) — Jue) ¥qie) = 2, ratme @ forsee, © free, 0 f W2ae Prove the result (11), page 98. FUNCTIONS RELATED TO BESSEL FUNCTIONS 6s. 662. 663. 661. 665. 6.66, 661, Bnet pee Show that [y(2) = 1+ Show that (a) Iy(z) = HTqas(2) + Tn vale)), (0) alge) = alyy(2) ~ nlx). Show that efl'*2) = $ ray in the generating function for Iy(2) ” ‘cosh (2 sin ¢) da, Show that Iy(2) Show that (a) sinh = 2[f\(2) +12) +++] (B) cosh = I(x) + late) + Show that (0) faxed = of 2(6 0) Teale) = af Eons), (a) Show that Kyyy(e) = Ky—s(2) +22K,(2). (0) Explain why the functions Ky(2) satisty the same recurrence formulas as Iy(z). *CHAP. 6] BESSEL FUNCTIONS AND APPLICATIONS 125 668. Give asymptotic formulas for (a) Hix), (b) Hye). 658. Show tat (@) Berge) = 3, AUB cos (MEH )-, : 5 _(2/2rn ‘an + 2k 6) Beige) = 3 pO BE sin (M2 He 6170. Show that Ker (2) = ~{In(o/2) + 7) Ber (2) + F Bel () + 1 — asytgepn— ce EQUATIONS TRANSFORMABLE INTO BESSEL'S EQUATION 671. Prove that (27), page 101, is a solution of (20). 612. Solve dzy" + 4y' ty = 0. 613. Solve (a) xy" + 2y' + oy @ x ety = 0 674, Solve" Hey [Hint, Let 0 = 675. (a) Show by direct substitution that y= Jo(2V@) is a solution of xy” +y'+y=0 and (b) write the general solution. 676, (a) Show by direct substitution that y= VJ,5(#9") is a solution of "+ 2y =0 and (0) wi the general solution. 6277. (a) Show that Bessel’s equation 2%" +2y' 4 (22—n!)y = 0 can be transformed into fey 4 (i where y= w/VZ. (b) Discuss the ease where n= =1/2. (#) Discuss the case where = is large and explain the connection with the asymptotic formulas fon page 101. 678. Solve ty!" —ay' +aty = 0, 679. Show that the equation (26) on page 101 has the solution (28) if «= 0. choose p appropriately, or make the transformation 2 = et] [Hint. Let y=2" and ORTHOGONAL SERIES OF BESSEL FUNCTIONS 680. Is the result of Problem 6.27, page 118, valid for —1=2=1% Justify your answer. est. show that f suRaedz = S200) + s,,00)) — Fal02) Ins rz) + € 682. Prove the results (34) and (85), page 102. 683. Show that tet 2 3 he " & AB Ses? where ry are the positive roots of J9(X) Jie) 684. Show a Boe ce that sar Way) ane where ry are the positive roots of Jy(x) = 0.126 BESSEL FUNCTIONS AND APPLICATIONS (ouaP. 6 S 2G— Aya) 685, Show that es 35 tes where hy are the postive roots of (Jy() = 0 S 25-4) JoQy2) ow tha ee § TO eed 6.86. Show that PF Bhs ty) ‘where hy are the postive roots of Jp(3) = 0 Jolo2) S _ relate) 687. Show that te 7 2eo ke where hy are the positive roots of Jg(i) = 0 688. If fle) = S Apolye) where Jy(,) =0, p=1,2,8,..., show that Sewerd = 3 Avion Compare with Parseval's identity for Fourier series. 689. Use Problems 6.84 and 6.88 to show that where d, are the positive roots of J9(X) 690, Derive the results (a) (85) on page 102, (¥) (36) on page 102, and (e) (97) on page 102. SOLUTIONS USING BESSEL FUNCTIONS 691. The temperature of a long solid circular cylinder of unit radius is initially zero. At t=0 the surface is given a constant temperature wy which is then maintained. Show that the temperature of the cylinder is given by } where dy, = 1,2,3,..., are the positive roots of Jo(A) = 0 and « is the diffusivity. 692. Show that if F(p) = ug(1—p%), then the temperature of the plate of Problem 6.28 is given by § J0(np) Jan) 2, a) = tong 3, gO ate 693. A cylinder O0, wis bounded, and 9,9) up, 9,0) = p 60889, 46,450) = 0 (@) Give a physical interpretation to the solution. Solve and interpret the boundary value problem a(,u) 2 # aelfae) = Ge Ha), ude 0) = 0, vit, t) 251 t>0. given that y(2, 0) © and y(2,t) is bounded for (@) Work Problem 6.93 if the end += 0 is kept at temperature f(p; 4). (B) Determine the temper- ature in the special case where f(p,#) = 0 cos 4. (@) Work Problem 6.98 if there is radiation obeying Newton’s law of cooling at the end 2 = 0. A chain of constant mass per unit length is suspended vertically o| from one end as indicated in Fig. 6-1. If the chain is displaced 7 slightly at time t= 0 so that its shape is given by flz), 0< 2 0 0,2) » wiL,t) = 0, vle,0) = fle), y(z.0) wie, | 0 is given by e = Mela, 8) e+ at) Show that ifm n Im (det) Tn Oz) a ” " Se te = SE Ue) Aloe) — Ino TE02)) + € Tix) Detuce te nterat f 722% ee ny wing tng prseare in thereof Problem 610% In (a) Show that, {Se = ete we Explain how the Sturm-Liouville theory of Chapter 9 can be used to arrive at various results involving Bessel functions obtained in this chapter. A cylinder of unit height and radius (see Fig. 6-8, page 115) has its top surface kept at tem- erature uy and the other surfaces at temperature zero, Show that the steady-state temperature at any point is given by & (sinh ds2)JoCo) uo, 2) Bo 2 Oy sink hd, 00) where dy are'the positive roots of Jo(A) = 0 ‘Work Problem 6.29 if the base 2 =1 is insulated.CHAP. 6] BESSEL FUNCTIONS. AND APPLICATIONS 129 6118. 61, 6.115, 6.116. eat. 6.18, 6.119, 6.120, Work Problem 6.29 if the convex surface is insulated. ‘Work Problem 6.29 if the bases :=0 and z=1 are kept at constant temperatures 1, and tt; respectively. [Hint Let u(p,2,t) = v(o,2,)+ w(o,2) and choose w{p,2) appropriately, noting: that physically it represents the steady-state solution.) Show how Problem 6.29 ean be solved if the radius of the eylinder is a while the height is h. Work Problem 6.29 if the initial temperature is /(p, $52). y A membrane has the form of the region bounded by two concentric circles of radii a and > as shown in Fig. 6-12, (@) Show that the frequencies of the various modes of vibration are given by Dan er A where ris the tension per unit length, u is the mast er unit area, and gq are rots of the equation Innit) Vy (08) ~ Jy QB)¥ yh) = 0 (®) Find the displacement at any time of any point of the membrane if the membrane is given an initial shape and then released. Fig. 612 A metal conducting pipe of diffusivity « has inner radius «@, outer radius b and height h. A eo- ordinate system is chosen so that one of the bases lies in the «y-plane and the axis of the pipe is chosen to he the z-axis. If the initial temperature of the pipe is fiz), a

1, the leading coefficients are taken so that the recurrence formulas for P,(z) above apply also Q,(z). - ORTHOGONALITY OF LEGENDRE POLYNOMIALS ‘The following results are fundamental: S, Pw (2) Pa() de 0 ifmen (8) " 2 Si Pear = ar © The first shows that any two different Legendre polynomials are orthogonal in the interval -ln, Px'(x)=0. The functions Q*(z) are unbounded for The differential equation (12) is obtained from Laplace’s equation Vu in spherical coordinates (r, 4,4). See Problem 7.21. expressed ORTHOGONALITY OF ASSOCIATED LEGENDRE FUNCTIONS As in the case of Legendre polynomials, the Legendre functions Pr(2) are orthogonal in -1<9<1, ie S : PM2)PM(e)de = 0 nek (16) We also have J, (PRR d = BET (n= myl a) Using these, we can expand a function f(z) in a series of the form fle) = ZAPre) (18) SOLUTIONS TO BOUNDARY VALUE PROBLEMS USING LEGENDRE FUNCTIONS Various boundary value problems can be solved by use of Legendre functions. See Problems 7.18-7.20 and 7.28-7.30.CHAP. 7} LEGENDRE FUNCTIONS AND APPLICATIONS 188 Solved Problems LEGENDRE’S DIFFERENTIAL EQUATION a. 72. Ro, where R depends only on r and @ depends only on ¢, in Laplace's expressed in spherical coordinates, show that R and © satisfy the By letting equation yu equations OR up = d/, #) 2 9 eG t+ org + eR = 0 aa(sine Gq) — M (singe = 0 Laplace's equation in spherical coordinates is given by 1 af ,au a af aw 1 ote 22(+%) + sin 3 (sneit) + Paints 992 ° a See (4), page 5. If w is independent of 4, then the equation can be written ° oy 7 (sme) Letting only on #, we have #8 8(all) Hang 3 £88) + By Aone Multiplying by r2, dividing by Ro and rearranging, we find La(,.a@) _ 4 (in 9 2 Ra” ar) = ~oaine ae"? Ge, Since one side depends only on r while the other depends only on ¢, it follows that each side must be a constant, say —\%. Then we have 1 4 = ® ana ; - aime a Dt ® which can he rewritten reapectively as eR , yah GBs Bs yee = 0 © aang! : and Hlomedt) — weinne = 0 @ as required. Show that the solution for the R-equation in Problem 7.1 can be written as B R= ant oa where a= —n(n +1). ‘The R-equation of Problem 7.1 i an ihe B sie = 0134 13. TA. LEGENDRE FUNCTIONS AND APPLICATIONS (CHAP. 7 This is an Euler or Cauchy equation and can be solved by letting R= and determining p. Alternatively, comparison with (28) and (28), page 101, for the ease where 2 =r, y=R, k=}, 2=0, B=) shows that the general solution is R= aarVnO 4 By-Ve or R= Arias VIR + gee Vn w This solution ean be simplified if we write 14 .ff- = -$+ bo» @ 0 that iy 5 -i-¥ = @ Tn such case (1) becomes R= Amt w Multiplying equations (2) and (#) together leads to x= —nln43) 6) Show that the o-equation (6) of Problem 7.1 becomes Legendre’s differential equation (1), Page 180, on making the transformation = cos 0. Using the value Xt = ~n(n +1) from (8) of Problem 7.2 in the e-equation (¢) of Problem 7.1, 48 becomes £(sne dt) + nov+tysinos = 0 co) we We now let £= cose in this equation, Then de _ dod do de dy de sneer ae 20 aint ‘Thus sine = —smtet = Ve since sin? = 1 ~coste 2. It follows that = 4fe- 2] = [evi = ffi] ~ 40 y_ yao! = afe-n@]€ = gfo-et!]une a Using this in (2) and canceling the factor sin, we obtain af ya _ Hlo-o8] + nmtne = 0 Cy Replacing @ by y and § by 2, and carrying out the indicated differentiation, yields the required Legendre equation (L-aty" — Bey’ + aintdy = 0 w Use the method of Frobenius to find series solutions of Legendre’s differential equa- tion (1—2%)y"" — 2ey’ + n(u+1)y = 0. Assuming a solution of the form 2 tow and c,=0 for k<0, we have Sect*# where the summation index k goes fromCHAP. 7] LEGENDRE FUNCTIONS AND APPLICATIONS 186 nin tty = Saint deer 2ny’ = B-Wk+ Alea? a-20y" Ze pet p—Neyerre-t — B (e+ pyle+ p= Dead? VU+ PME B+ Nog past — DY (e+ Alle -+ B—Aeyet*h ‘Then by addition, ZG+ P+ ME+ A+ Degen — e+ A+ 2 D)oy ~ 2+ ploy + mint Delet+® = 0 and since the coeficient of 2+? must be zero, we find (e+ B+ 2k + B+ Veysg + [n(n +1) —(k+ Akt B+Viqe = 0 a Letting k= -2 we obtain, since 0, the indical equation p(?—T)q)=0 of, assuming co, B=0 ort Case: In this case (1) becomes (DIF Devsa + fume) — MEDI = 0 ® Putting k= —1,0,1,2,5,... In succesion, we find that, is anbitrary while tg = MED, wen, = Betoaeeny eee Since we have a solution with two arbitrary constants, we need not consider Case 2: For an even integer m0, the first of the above series terminates and gives a polynomial solution. For an odd integer n> 0, the second series terminates and gives a polynomial solution. ‘Thus for any integer = 0 the equation has polynomial solutions. If = 0,1,2,3, for example, ‘we obtain from (3) the polynomials 5z?\\ as atic, a which are, apart from a multiplicative constant, the Legendre polynomials P,(2). ‘This multipli- ative constant is chosen s0 that Py (1) = 1. ‘The series solution in (3) which does not terminate can be shown to diverge for z= <1. This second solution, which is unbounded for = +1 or equivalently for ¢ = 0,z, is called a Legendre function of the second kind and is denoted by Q,(z). It follows that the general solution of Legendre's differential equation can be written as ¥ = e Pale) + Qq(2) In case m is not an integer both series solutions are unbounded for x = 7.5. Show that a solution of Laplace's equation V'u=0 which is independent of ¢ is given by us (4+ Ft )iAPae + B:0.(9)136 LEGENDRE FUNCTIONS AND APPLICATIONS (CHAP. 7 This result follows at once from Problems 7.1 through 7.4 since w= R@ where B, R= aes Se and the general solution of the @-equation (Legendre’s equation) is written in terms of two linearly Independent solutions P,(@) and Qq(@) as 8 = APA) + BiQa(0) ‘The functions P,(¢) and Q,(@) are the Legendre functions of the first and second kinds respectively. LEGENDRE POLYNOMIALS 76. Derive formula (3), page 130, for the Legendre polynomials, From (2) of Problem 7.4 we see that if k=m then e,;2=0 and thus ¢yy4=0, en+e= , Then letting k = n—2,n—4,... we find from (2) of Problem 7.4, “ m=, m= 2n=3), nln =1)(n=2(n—8) , ma = aan jem amd = 2-4n=Den=3) This leads to the polynomial solutions _ MOD gy = In=2)in—3) gs, vs fen agent + taggers] ‘The Legendre polynomials P, (z) are defined by choosing @n=WGn—3)-+-B41 ‘This choice is made in order that P, (1) 1a Beni da By Problem 7.6 the Legendre polynomials are given by 7.7. Derive Rodrigue’s formula P(x) (1), PA) ex — MOD gaan 4, MN Win = B8—8) 3a = 1) 2-aan— Tn 3) Now integrating this » times from 0 to 2, we obtain on 2)(2n—3)- ni which can be written (2n=1)@n~3)---3+ Bn Ge= NEe= Bei which proves that Pale) GENERATING FUNCTION 1 yi-2at FE Using the binomial theorem 78. Prove that to = 1+ pe + MODs 4 MPRDO=B yCHAP. 7] LEGENDRE FUNCTIONS AND APPLICATIONS 137 we have = (bee 9) Vi- Batt 18 eee + BEE ee gh Bd 1 = 1+ deen + and the coefficient of ¢* in this expansion is _ 18+5-+-@n—3) 1s3e5. B46 (1) payy—2 Ta? (n= 2\n—3) 2 ra which ean be written as ie. P(x). ‘The required result thus follows. RECURRENCE FORMULAS FOR LEGENDRE POLYNOMIALS, 79. Provethat Paei(2) = 282) apy(a) — By Pasa). From the generating funetion of Problem 7.8 we have 1 2 Vette = 3, Pam a Differentiating with respect to ¢, . aot 4 Tate 2 mln Multiplying by 1—2et-+ ¢, Fi Now the left side of (2) can be written in terms of (1) and we have t me 3a sertemrsene ® Bt orayn = 3 a-2et+ amps By2Palerr — 3 Pacers = SE apyiayent — & anaPytalen +S mPy(aderts Bquating the cooficients of ton each side, we find @P Q(x) — Pyy(z) = (m+ 1)Pas (2) — 2neP, (x) + (m—1)P,_4(@) which yields the required result. 7.10. Given that Po(z)=1, Pilz) = 2, find (a) P.(z) and (6) Ps(2). Using the recurrence formula of Problem 79, we have on letting m= 1, Pie) = GaP) — dre = $e - 5 = poet Similarly letting » = 2, Py) = 8138 LEGENDRE FUNCTIONS AND APPLICATIONS (CHAP. 7 LEGENDRE FUNCTIONS OF THE SECOND KIND 711. Obtain the results (6) and (7), page 181, for the Legendre functions of the second kind in the case where n is a non-negative integer. ‘The Legendre functions of the second kind are the ser do not terminate. From (3) of Problem 7.4 we see that if n nate is lutions of Legendre’s equation which ‘even the series which does not termi- pe — GEDA 5 (n= in = Bie OVE D og i BP while if m is odd the series which does not terminate is nin+1) 9 4 nln— 20 “a ut 1 Bee 7 ‘These series solutions, apart from multiplicative constants, provide definitions for Legendre func- tions of the second kind and are given by (@) and (7) on page 131. The multiplicative constants are chosen so that the Legendre functions of the second kind will satisfy the same recurrence for- malas (page 131) as the Legendre polynomials. 7.12. Obtain the Legendre functions of the second kind (a) Qu(z), (b) Qu(x), and (c) Q.(z). (@) From (6), page 131, we have if» Qt) = 2 wb + US5B46 Bort = in(Lte = gmli-e, where we have used the expansion In (I+) = u— ui/2-+u¥/3—wi/4+ () Prom (7), page 131, we have if n= 1, Qe) -f — BBs ANE og MEDEDAIG oo 4. } - 2 - 24 } -1 (©) The recurrence formulas for Qy(z) are identical with those of Pa(z). ‘Then from Problem 7.9, 2nd Quel) = WET Tl) ~ EG Anil) Putting n=1, we have on using parts (a) and (b), Qe) = G0.) ~ gay = (BP4)m(L2 2)-% i-z)~ 2 ORTHOGONALITY OF LEGENDRE POLYNOMIALS 7.13. Prove that S Pa(z) Pala) dz = 0 if men, Since Py (zx), Py (2) satisfy Legendre's equation, (1=29P% — 22P%, + mim + Py (A 29Py — 22PL + nn ayP, = ° ° ‘Then multiplying the frst equation by Py, the second equation hy Py and subtracting, we find (1 29[P, Pe — PPel] ~ 2e[PyPa—PyPZ] = [nln-+ 1) ~ m(m+1)]PgPyCHAP. 7] LEGENDRE FUNCTIONS AND APPLICATIONS 139 which can be written (1-298 (P,Pq— PaPal — 221PaPn— PnP n(n 1) — m+ ))PaPa or (1210, Py ~ PaPS)) = [nin $1) — MOF AYP RP ‘Thus by integrating we have " net) — min +1] f* PgledPg@)de = A= 29)P Pm PaP Si rawrreyte = a TA, Provethat f" {Paa)iPdz = sey From the generating fanetion Pye Vers = OB ‘we have on squaring both sides, Bd nonern ‘Then by integrating from —1 to 1 we have Sirmtics - 33 {fleaerrmae on Using the result of Problem 7.18 on the right side and performing the integration on the left side, ~fma-aerel = 3 { f' teaeyras} om “ (42) = 3, {f° wacorras} om e Set - 3 {Sf eawact a Equating coefficients of t2, it follows that * ye = 2 Siatorrae = ah SERIES OF LEGENDRE POLYNOMIALS TAS. If fle) = ¥ AsPi(a), ~1<2<1, show that A = PRED (Pala) in) de Multiplying the given series by P,(2) and integrating from —1 to 1, we have on using Prob- lems 7.18 and 7.14,140 LEGENDRE FUNCTIONS AND APPLICATIONS (OHAP. 7 Si Patertieyde = 3 Ay [Pal Pale) de + 2Am = An finite = itty ‘Then as required, 2m Si Pte fey ae Oa Casel: OS a, In this case we must choose A =0 in (2) since otherwise the solution becomes unbounded rove, Then v=BP,(@)/r*t and by superposition we are led to consider the solution By v= Sah ” As in Case 1, this must reduce to the potential on the z-axis for 2 : 2, Jap eet ‘Thus, to find B, we must expand the left side in inverse powers of =. Again we use the bino- ‘ial theorem to obtain (8) aee(eg)y = Bef. -3(2) + Comparison of (8) and (9) leads to ] ) B= tae, = 8, B= tne), By = 8, By = te Using these in (7) we then find 2rae see[ roto) — 2(8)"Palese) + £:2(2)'Pycose) — ] co where r> a. ASSOCIATED LEGENDRE FUNCTIONS 721. Show how Legendre’s associated differential equation (12), page 132, is obtained from Laplace's equation Y*u = 0 expressed in spherical coordinates (r, 0,4). In this case we must modify the results obtained in Problem 7.1 by including the ¢-dependence, ‘Then letting u= Rew in (1) of Problem 7.1 we obtain od od, a8 es SA #) + wy Hsin e) 4 aw = 9 i Mlttying by 7, dividing by Rob and rearranging, we find “1 a/ ae 1 4/,,,20)_ 1 ee Ra Y ar, Taine wer? we Since one side depends only on r, while the other depends only on # and 4, it follows that each side ‘must be a constant, say —\2, Then we have 14/28) _ yy pa(e@) = ® 2 (ang 1 &e . and a a ( $) Fanta age ™ a ‘The equation (2) is identical with (2) in Problem 7.1, so that we have as solution according to Problem 7.2 R= A, aw where we use Xt = —n(n-+1)CHAP. 7) LEGENDRE FUNCTIONS AND APPLICATIONS 145 If now we multiply equation (8) by sin*¢ and rearrange, it ean be written as a @ dol ‘Since one side depends only on # while the other side depends only on ¢ each side must be a con- stant, say —m?. ‘Then we have lee © de® (in 9 Pe mile = sine (nine) + ful) si ° o Ge + mw = 0 @ If we now make the transformation {= cos @ in equation (5) we find as in Problem 7.2 that it ean be written as 2 [(y — xyae [min - = a-eg[a-e $2] + ime na ° Dividing by 1—¢ the equation becomes ae Ge — 268 + [nin n-7@]o = 0 o Which is Legendre’ associated diferential equation (12) on page 192 if we replace © by y and Eby ‘The general solution of (7) is shown in Problem 7.22 to be © = Apr + Bre oy where ¢ = cos# and a am enn ° Pre = a-eme Ee) ® me = (Let ee = a-e Z aw i) We call PR) and Qe) associated Legendre functions of the first and second kinds respectively. ‘The general solution of (6) is = Agecosms + By sinme a If the function u(r, 6,9) is to be periodic of period 2r in g, we must have m equal to an integer, which we take as positive. For the case m=0 the solution u(r,¢,¢) Is independent of ¢ and reduces to that given in Problem 7.5. 7.22. (a) Show that if m is a positive integer and u, is any solution of Legendre’s differen- tial equation, then d”z,/dx™ is a solution of Legendre'’s associated differential equation, (b) Obtain the general solution of Legendre’s associated equation. (a) If Legendre's differential equation has the solution 1, then we must have (La a8ugy = Qew, + n(n Duy = 0 By differentiating this equation m times and letting off = d™a/dz" we obtain a on dog Tee Pmt ade a-ay Ge + Inn +2) ~ mem+ Doe o In this equation we now let vf! = (1—#!)7y. Then it becomes (= 28)" ~ [20m-+ ted — at) + dpe(h — aly! + Wen + Lp! + (Ap? — 2p)at — 2p + [nln +1) — mim + Naty = 0146 123. 124, 1.25. LEGENDRE FUNCTIONS AND APPLICATIONS [CHAP.7 If we now choose _p = —m/2, this equation becomes after dividing, by 1— x? a-aty" = ary + [nine maly = 0 w which is Logendre's associated differential equation. Since sf! = (1—24)~"/ty, it follows that y= (290%, or e aa ate Te @ ye ¢ a is a solution of (1). (2) Since the general solution of Legendre’s equation is ¢Py(z) + ¢:Q,(2), we can show that the general solution of Legendre's associated differential equation is y= Pita) + Qi) a) a) = yun Pn rm ymin tn f where Pre) = Gsm, me) = atm ” Obtain the associated Legendre functions (a) P;(2), (b) P3(x), (c) Ps(z), (d) Q3(a). @ Pie = a-syndege = a-enad = satay ® Phe = a-aerZ re = a-a 8 = 162 — 168 sta) = dane. pyle) = 0. Note that in general Pole) =0 if m>m (eo) Pax) a yas Pala) 0. Note that in general P,'(<) = 0 if (® Using Problem 7.12(¢) we find Qe) a-22 2 Q,(0 a-ang f aya[ Sy, (Lbs) 4 922 = (ay) (Ems) 5 ail Verify that Pi(z) is a solution of Legendre’s associated equation (12), page 182, for m=2, n=3. By Problem 7.23, Pi(x) = 15z—16z%. Substituting this in the equation we find after simplifying, =ay-ans = ae08—c009 + [125 and so P¥(2) is a solution. ]ase—a525) = 0 Verify the result (16), page 132, for the functions P(x) and Pi(z). We have from Problem 7.23(a), P3(x) = 3x(1 2%). Also, 528 1622 18 (1 ay ri = a-ettre = a-ayrd(' ‘Then Storie a = f'Fa-wu = 0CHAP. 7] LEGENDRE FUNCTIONS AND APPLICATIONS wt 7.26, Verify the result (17), page 182, for the function P2(z). Since Pla) = sz(t ~ 2912, rt * 2)/' 36 12 Si pherde = 9 fi ea-aar = z] a. 8 Now according to (17), page 182, the required result should be 2 e+! _ 2st _ Wie = ea = 5 so that the verification is achieved. 7.27. Expand vo(1—2") in a series of the form > AxPY(2) where vo is a constant and m= We must find Ay, k= 0,1,2,..., 50 that vol — 22) = AgP3(x) + AyPH2) + AgPHe) + ++ w Method 1, ‘ Pie) = aa) Pyle Since Prey = 0-9 Zt we have a . @ [oe 1 Pie) = 0, Pie) = 0, Phe = a-9 S(* sa—s), Pla = 0-9 8%) = saa, ‘Then (1) becomes voll 22) = BAL 22) + IGA ge a8) ve By comparing coefficients on each side we see that this can be satisfied if 34, =, 15A,=0 and Ay=0 for k>3. Thus we have voll 22) = PPHe) @ so that the required expansion consists of only one term. Method 2. If f@) = 3, A,PLG@), then on multiplying by PE(2) and integrating from —1 to 1 we obtain Si perrte ce = 3a Sf rrerere ae Using (£4) and (17), page 192, we see that the right side reduces to the single term 2 (ntm)t Bed (=m An eat DRM! 6 ye) PE de 0 that Boe mye If f(2) = vo(l—2%) and m= 2, then Gat aya—o! ft An Btn +2)! vg(l = 22) PR) de Using this we can show that Ay = w/3, A, Method 1. 0, As= and so we obtain the result (2) as in148 INDRE FUNCTIONS AND APPLICATIONS [cHap.7 7.28. Show that a solution to Laplace's equation Y#v = 0 in spherical coordinates is given by / Bu apm " 7 ® (Ao + jr) [4ePS (e088) + B2Q2(c089)][As cos mp + Bs sin mg ‘This follows at once from Problems 7.21 and 7.22 since w= Rew where By R= Are Sh @ = AgPMcos 6) + BQM(c0s 6) © = Aycosme + Bs sin me 729. Suppose that the surface of the sphere of Problem 7.18 is kept at potential vosin®@ cos2g. Determine the potential (a) inside and (b) outside the surface. <1. (a) Interior Potential, 0 Since v is bounded at r= 0 we must choose By =0 in the solution as 7.28. Also since v is bounded at #=0 and x, we must choose B,=0. Then a bounded solution is given by ‘o(r,0,¢4) = MPM (eos 0A cos mg + B sin ms) Since m and n can be any non-negative integers we can replace A by Ayyy B by Byy and then, using the superposition principle, sum over m and x to obtain the solution 6056.9) = BS rrPM con oi'Any con me + Bus sin me) ‘Now the boundary potential is given by v(1, 6,4) = vp sin?s cos 2g ® By comparison of (2) with i819) = SB, PMcos odin cos me + Byy sin ms) obtained from (1) with r=1. it is seen that we must have Byy=0 for alll m and Aj, =0 for m#2. Hence, (8) becomes 1,49) = 3 AaPi(e0s #) cose Comparison with (2) then shows that we must have vasint@ = % AsPZ (cos) or using cose = ¢ . wt) = ZB Aneto = AmP§) + AnPT) + APH + + w We have already obtained this expansion in Problem 7.27, from which we see that A, while all other coefficients are zero, It thus follows from’(1) that. v3, vir,69) = PrePH(coss) cos2s = vot? sin? # cos 2s 6 (®) Exterior Potential, r > 1. Since v must be bounded as r= in this case and is also bounded at choose A; = 0, By=0 in the solution of Problem 7.28. Thus a solution is and =, we Pai(cos 8) (rs 0,8) ST (A cos ms + B sin mp)CHAP. 7) LEGENDRE FUNCTIONS AND APPLICATIONS 149 or by superposition vr, 8,6) (Aggy 608 mg + By sin ms) CO) Using the fact that v(1,4,g) = vy sin?s cos2p we again find m =2, By, =O which leads to equation (4) of part (a). As before we then find Az, = vy/3, while all other coefficients are zero, leading to the required solution ur, eo = a heint 9 cos 26 ” It is easy to check that the above are the required solutions by direct substitution. 7.80. Solve Problem 7.18 if the surface potential is /(0, $). As in Problem 7.29 we are Jed to the following solutions inside and outside the sphere: Inside the sphere, 0S r <1 vir.6.0) = SS rPcos odin c08m9 + Bay sin mg) ® Outside the sphere, r > 1 SS Pr(cos 8,8) BS Pee agg cos md + By iT ng) @ For the case r=1 both of these lead to Ha) = 3, 3 PMcose)(Aga cosmg + By sin ms) ‘This is equivalent to the expansion Fo) = BE PLA gn cosme + Byy sins) co) where {= cos#, Let us write this as Foo = 3,c.Pr@ “ where (Anu cosms + Bry sin me) cy ‘As in Method 2 of Problem 7.27 we find from (1) Qn+1)in—m! ¢* you c= MENAM! Fre genoa We also soe from (5) that Am, and By, are simply the Fourier coefficients obtained by expansion of C, (which is a function of ¢) in a Fourier series. Using the methods of Fourier series it follows that! Aon 2 Cade dn = 3 r meds Pan = 2S casinmede ‘Combining these results we see that Ay, = 2EEDG— mt S. (a eee den = ma)!150 LEGENDRE FUNCTIONS AND APPLICATIONS (CHAP. 7 while for m= 1,2,8, 4, Gael at " f FG 0) PLU) cos me a de = Be(n + m)! = @n+1n—m! ct ¢™, Ba = oat SS, FG, 9) PR) sin mg dg dp Using these results in (1) and (2) we obtain the required solutions. Supplementary Problems LEGENDRE POLYNOMIALS 731 132. 133. 138, 135. 736. 131. 138. 729. 140, TAL. Use Rodrigue’s formula (4), page 180, to verify the formulas for P(x), P, (2), ..., Ps(2), on page 180. Obtain the formulas for Py(z) and Ps(a) using a recurrence formula, Bratuate (©) f'2Ps) de, ©) f° Prlaitas, Co) J Pale) Pyle) a. Show that (a) P,(1) = 1 © P10) = 0 ©) PA) = C4 Pagld) = Cap a6 Gn) for n= Use the generating function to prove that Pj. () — Pia(2) = (n+1) Py(a). Prove that (a) Py, (2) — 2Pj(2) (+L) PQ), (6) 2Ph(z)— PLa(e) = mPy(a), swowthat $ reosn = Lael. Stow hat (0) Pains) = 4480830, (@) Pylon) = (8 cos 0 +5 cos 34). Show that Pre) = 4s (tat eon +16 a6) Show from the generating function that (a) Py(1 1, @) Py(-1) = (ay, Show thar 3 SF LEGENDRE FUNCTIONS OF THE SECOND KIND 142. 13. Ta. Prove that the series (6) and (7) on page 181 which are nonterminating are convergent for —1b 785. A solid uniform circular dise of radius @ and mass M is located in the 2y-plane with center at the origin. Show that the gravitational potential at any point of the plane is given by. Py(eos 0) + 3(E) Paleo 0 ~ gha(2) Palcorn 163 (x Pas FL ~ 4(¢)'Pacesn + (2) ralcme) ~ EES(2) rycorn + -] + itr>a, ASSOCIATED LEGENDRE FUNCTIONS 756. Find (a) P3(2), (6) Pilz), (@) PR). 151. Find (@) Qi), (®) QHe).152 LEGENDRE FUNCTIONS AND APPLICATIONS [CHAP.7 that the expressions for Pf(2) and Qi(e) are solutions of the corresponding diferential equation and thus write the general solution. 158. (0) m 759. Verity formulas (16) and (17), page 182, for the case where (a) m=1, »=1, 1 m=, [= 1. 760. Obtain a generating function for P2(2). 761. Use the generating function to obtain results (16) and (17) on page 192. 782. Show how to expand f(z) in a series of the form A,PE\2) and illustrate by using the cases (@) f(@)=2, m=2 and (6) fle)=20-2), m tity in each case. ‘Verify the corresponding Parseval’s iden- 163. Work Problem 7.18 if the potential on the surface is vy sin’ # cos 6 cos 3¢. MISCELLANEOUS PROBLEMS 781. Show that Pale) YP camer — H@=BI~— 21 where [7/2] is the largest intoger = 165. Show that Prix) = (@ + Vz? =1 cos u)" du Ute nel od ate) Pte rege Boe Siamese = {men aye 781. Show that 1 Bina +1) neo Stems) ae = ee n=0 1 0) Show tint fl anPso)de = 0 if men ori m=n ts on ol pave nee (8) Show that . (+ 2p)!T+p) Siverioe = sete tetn D for any non-negative integers m and p. 7.69. Show that # solution of the wave equation ny = Lev acai a ot depending on 7,0, and, but not on gs given by Vo = [ArInssyalerle) + ByIn—1p2(ur/e))[AgPq(cos 6) + ByQq(cos 6)](As cos wt + By sin of]cuar. 7) \LEGENDRE FUNCTIONS AND APPLICATIONS 158 10. mm. 172. 13. ‘Work Problem 7.69 if there is also ¢-depondence, ‘A heat-conducting region is bounded by two concentric spheres of radii a and b (a <6) which hhave their surfaces maintained at constant temperatures u, and up respectively. Find the steady- state temperature at any point of the region. Interpret Problem 7.18 as a temperature problem. Obtain a solution similar to that given in Problem 7.69 for the heat conduction equation au Bs ave where u depends on 7,0, and t but not on ¢.Chapter 8 Hermite, Laguerre and Other Orthogonal Polynomials HERMITE’S DIFFERENTIAL EQUATION. HERMITE POLYNOMIALS An important equation which arises in problems of physics is called Hermite’s differen- tial equation; it is given by ¥’ —2xy’ + 2ny = 0 @ where m=0,1,2,3,.... ‘The equation (1) has polynomial solutions called Hermite polynomials given by Rodrigue’s formula Hilt) = @ The first few Hermite polynomials are Hal) Hix) = 20, Hox) = 4a*—2, Hel) = 8x — 120 (3) Note that Ha(z) is a polynomial of degree n. GENERATING FUNCTION FOR HERMITE POLYNOMIALS ‘The generating function for Hermite polynomials is given by ‘ = Hal) = Sart “ ‘This result is useful in obtaining many properties of (2) RECURRENCE FORMULAS FOR HERMITE POLYNOMIALS We can show (see Problems 8.2 and 8.20) that the Hermite polynomials satisfy the re- currence formulas Hyss(t) = 22Ha() — 2nFys(@) (6) Hilat) = 2nHy-(2) (6) Starting with Ho(z) = 1, Hi(e) = 2x, we ean use (5) to obtain higher-degree Hermite poly- nomials. ORTHOGONALITY OF HERMITE POLYNOMIALS We can show (see Problem 8.4) that S "Hy(ct) Haz) dx = 0 man 3) 164CHAP. 8] HERMITE, LAGUERRE AND OTHER ORTHOGONAL POLYNOMIALS 155 so that the Hermite polynomials are mutually orthogonal with respect to the weight or density function e~*. In the case where m =n we can show (see Problem 8.4) that the left side of (7) becomes Sietmiayae = eye (8) From this we can normalize the Hermite polynomials so as to obtain an orthonormal set. SERIES OF HERMITE POLYNOMIALS Using the orthogonality of the Hermite polynomials it is possible to expand a function in a series having the form flr) = Agel) + AiHi(2) + AaH(z) + +++ (9) where An a S. See Problem 8.6. *f(0) Hala) dx (10) Bnlva In general such series expansions are possible when f(z) and f(x) are piecewise con- tinuous. LAGUERRE’S DIFFERENTIAL EQUATION. LAGUERRE POLYNOMIALS Another differential equation of importance in physics is Laguerre's differential equation given by ay” +(1—a)y' + ny = 0 (11) where 1=0,1,2,3,.... This equation has polynomial solutions called Laguerre polynomials given by Le) = oft which is also referred to as Rodrigue’s formula for the Laguerre polynomials. (z"e-*) (12) The first few Laguerre polynomials are Lo) = 1, Laz) = 1-2, Lola) = at- de +2, Lal) 182 +92t— a? (13) Note that La(2) is a polynomial of degree n. SOME IMPORTANT PROPERTIES OF LAGUERRE POLYNOMIALS In the following we list some properties of the Laguerre polynomials. 1. Generating function. enn = 10), a 2 Recurrence formulas. s(2) = (2n+1—2)La(2) — n*Ln—(2) (15) In(@) ~ mLin-s(at) + nn-s(2) = 0 (16) wn a) = nal) — n*Ln— (2) (an156 HERMITE, LAGUERRE AND OTHER ORTHOGONAL POLYNOMIALS 3. Orthogonality. oo _ f 0 ifmen Sle tated nied = tee if m=n 4, Series expansions. it fle) = Aolale) + Asks(2) + Aala(e) + 1 then 4 = Gaye S e-#f (a) La(x) de (CHAP. 8 (as) (a9) (20) MISCELLANEOUS ORTHOGONAL POLYNOMIALS AND THEIR PROPERTIES ‘There are many other examples of orthogonal polynomials. Some of the more important ones, together with their properties, are given in the following list, 1, Associated Laguerre polynomials Ly (z). These are polynomials defined by a) = A, En(z) = Gym En(a) and satisfying the equation ay’ + (m+1—2)y + (a—my = 0 If m>n then Ly(z) =0. We have a Smect@ ede = 0 pan 1 ame"=(Ln (2)? dz @ x i 2. Chebyshev polynomials 74(z). ‘These are polynomials defined by 1 (2) = cos(neos-t2) = am — (eva-2) + ( and satisfying the differential equation (= tjy” ~ ay + n'y = 0 where n= 0,1,2,.... A recurrence formula for T,(2) is given by Tevi(t) = 20Pa(z) — Tr1(2) and the generating funetion is 1-te 5 Tamera = 2M We also have men S Ratt Tnlz)Talt) 5 _ og vi-a + (Ta(2)}* [= Se fs A (et) (e2) (23) (24) ent sao (25) (26) en (28) (29) (80)CHAP.8) HERMITE, LAGUERRE AND OTHER ORTHOGONAL POLYNOMIALS 157 Solved Problems HERMITE POLYNOMIALS 81. Use the generating function for the Hermite polynomials to find (a) Hol), (6) Ha(x), (¢) Ha(x), (d) Holz). Werhave how be wet BO ase mine + Be + Bln 4 Now ot Se enw) + RES BE, Bee = tenet annne + (Ee + Comparing the two series, we have Holz) = 1, Hye) = 22, Hylz) = det 2, Hye) = 8x9 — ae 82. Provethat H(t) = 2nHe-s(2). newt 2 3 HO, oe 5 Meer 2 3 Me, Equating coeficients of 6 on both sides, 2H, (2) Hye) 1 _ Ba) FI) ge aria) = onal a nw 83. Prove that Hy(z) (lye" ae (@ *) We have ett gt tent en aM gens The Ber) is But Be =o et oF 84. Provethat f e-"Hw(e) Hala) de = 5 Haleyt Wehave ett = a Multiplying these,158 85. HERMITE, LAGUERRE AND OTHER ORTHOGONAL POLYNOMIALS (CHAP. 8 Multiplying by e~** and integrating from —® to =, flew e = 3 3 8 oH (2) Hla) de Now the left side is equal to en [erode = ox feta = eye = ve 3 MEE By equating coefficients the required result follows. ‘The result i entigis) He) de = 0 men can also be proved by using a method similar to that of Problem 713, page 188 (see Problem 8.24). Show that the Hermite polynomials satisfy the differential equation 0 From (5) and (6), page 154, we have on eliminating H,_1(2) Hyysle) = 2aHty(e) — Hye) w wy" — 2xy’ + 2ny Differentiating both sides we have Hyss@) = BaHy(2) + 2H gle) — Hea) @ But from (0), page 154, we have on replacing n by +1 Hise) = 2+) Hye) ) ‘Using (8) in (2) we then find on simplifying: Hye) — BeHiy(2) + 2nHi,(2) = 0 Which is the required result ‘We can also proceed as in Problem 8.25 (@ If fle) = S ArHi(x) show that Ay DRIVE S61) Hole de. (b) Expand 2* in a series of Hermite polynomials, : (@) ff) = B Arky(e) then on multiplying both sides by ¢-#H,(c) and integrating term by term from —= to © (assuming this to be possible) we arrive at Siete tieyte = Saf comer te Slvonomes = {0 ee ‘Thus (1) becomes eNfle) Hla) dz = Ante mane Soe Hale) de ® which yields the required result on replacing m by i, or Ay =CHAP. 8] () 87. (a) (b) @ HERMITE, LAGUERRE AND OTHER ORTHOGONAL POLYNOMIALS: 159 ‘We must find coefficients Ay, k= 1,2,8,..., such that a= 3 Ame ® Method 1. ‘The expansion (3) ean be written 28 = Agta) + Agi (2) + Aghia) + Agha) + > ® or 28 = Agll) + Ay(Bx) + Aglda? 2) + AgB2?— 122) +--+ © Since Hy(x) is a polynomial of degree k we see that we must have Ay=0, As=0, Ag=0, «205 ‘otherwise the left side of (5) is a polynomial of degree 3 while the right side would be a poly- nomial of degree greater than 3, Thus we have from (6) g9 = (Ay 2Ay) + @Ay—12Ag)e + dAge? + BAe? ‘Then equating coefficients of like powers of « on both sides we find B4y = 1, 44, = 0, 24,— 124, = 0, Ay—24, = 0 ‘from which 3 Ay = 0, Ay > " é cos Thus (2) becomes yao 1 = Fae) + Fhe) which is the required expansion. chee. Sings + kaye) = Le + e100) = » Mate 2 he coin A fn) agen by - 1 * postesHi (2) de ae = age Stemi as obtained in part (a) with /(2) = 2. Putting k= 0,1,2,8,4,... and integrating we then find 1 a= 4-3, ee ek a ao, and we are led to the same result as in Method 1. In general, for expansion of polynomials the first of the above methods will be easier and faster. Write Parseval’s identity corresponding to the series expansion f(z) = > AxFia(z). Verify the result of part (a) for the case where f(z) = 2°, = We ean obtain Parseval’s identity formally by first squaring both sides of f(2) = 3 A,Fy(e) tool s (et = BS, Acdptile) H62) ‘Then multiplying by ¢~* and integrating from —= to « we find Sleeve = 3 3d, Seamer nye) de160 HERMITE, LAGUERRE AND OTHER ORTHOGONAL POLYNOMIALS [cHaP. 8 Making use of the results of Problem 8.4 this can be written as Senta: = ve 3 meat which is Parseval’s identity for the Hermite polynomials. (b) From Problem 8.6 it follows that if f(x) = 29 then Ay=0, Ay=4, 42= 0, Ay= a, 4g=0, As=0, .... Thus Parseval's identity becomes eosetiide = Vee Ng? +2600) Now the right side reduces to 15Vz/8. The let side is 2 wtewae = SP uterean r® = Dave = B = By ‘Thus Parsoval's identity is verified. where we have made the transformation 2 = LAGUERRE POLYNOMIALS 88. Determine the Laguerre polynoniials (a) Zo(z), (b) Li(2), (¢) La(x), (d) La(2). We have Ly(e) = eZ (ere-2). Then (@) Loz) = 1 ©) Le) = eben = 1-2 © te) = Been = 2- att @ bye) = 6 — 182 + 922 ~ 29 89. Prove that the Laguerre polynomials L,(2) are orthogonal in (0,) with respect to the weight function e~*. From Laguerre’s differential equation we have for any two Laguerre polynomials (2) and LA), aly + (L=2)L, + Ly = 0 xb, + Q—x)L, + wb, = 0 Multiplying these equations by iy and Ly respectively and subtracting, we find allyl, — Ligh) + (1 2)[Lkig — Legh] = (n= m\lly (= mln or Etat, - bali] + Lgbig — Lgl) = Multiplying by the integrating factor (Steniede this can be written as lee lglg — Eli (n= me~*DglingCHAP. 8} HERMITE, LAGUERRE AND OTHER ORTHOGONAL POLYNOMIALS. 8.10. Sl. so that by integrating from 0 to #, sett Im) Sos gled Ea) de = 20-*Ulabiy~ bll Thus if m¥m, 7 Sf ernie ine) de = 0 which proves the required result. Prove that Ln+i(t) = (2n+1—2) La(z) — n?Lns(@). ‘The generating function for the Laguerre polynomials is, yap & Ela) Tt > Be « Differentiating both sides with respect to ¢ yields @ Za-ob@e- 3 which ean be written 3 bee male If we now equate coefficients of ¢ on both sides of this equation we find . Hale) (0+ Digs s(@) Brie) mS @ ey at wh Multiplying by n! and simplifying we then obtain, as required, Lgesle) = n+ —2) Egle) ~ Ly y(2) Expand 2° + x*~ 82 +2 ina series of Laguerre polynomials, ie. ) Axla(z). ‘We shall use a method similar to Method 1 of Problem 8.6(6). Since we must expand a poly- nomial of degree 3 we need only take terms up to Lx(2). ‘Thos Aglig(2) + AqLy(z) + Aglg(a) + Agls(a) Using the results of Problem 88 this can be written wba B2 42 = (Ag tAy tPA, +6Ay) — (A, + 4Ay + IBAGle + (Ay + 9AQ)a? — Ags Bata 42 ‘Then, equating like powers of « on both sides we have A+ A, +242 + 6g Ay + 4A, + 18Ay Ast 9g Solving these we find Ag = 4, = -19, A, = 10, Ay = 1 Then the required expansion is BaP Be $2 = Tafa) — 18Ly(2) + 10L4(2) — L4(2) We can also work the problem by using (19) and (20), page 156.162 HERMITE, LAGUERRE AND OTHER ORTHOGONAL POLYNOMIALS [CHAP. 8 MISCELLANEOUS ORTHOGONAL POLYNOMIALS 8.12, 813. 814, 8.15. Obtain the associated Laguerre polynomials (a) Li(z), (b) Lie), (c) Lie), (d) Li(a). (@ te) = fie = Pe-derey = 22-4 © He = Zig = Be-s+e = 2 © Be = Sie = Se-w+0%-2 = 1 -o fm = tye = My n0 (the) = Bitge) = 0. tn general 15(@) = 0 if m>m Verity the result (24), page 156, for m=1, » ‘We must show that f * ge(LMee))ede = Now since L(x) 22~4 by Problem 8.12(a) we have 4rd) ~ 1608) + 16r@) 4) — 162) + 160) = 8 SS ze-*(2— 4) de ze-tde 0 that the result is verified Verify the result (2), page 156, with m=2, n=2, p=8. We must show that - S ste*Li{x) Li(e)dx = 0 Since Lite Liz) = 18-62 by Problem 8.12(a) and (b) respectively the integral is [tera — 0 f were aer(a) — 121(4) flees) eoae see!) — 1281) = 0 as required. Verify that Li(x) satisfies the differential equation (22), page 156, in the special case m=2, 1=8. From Problem 8.12(¢) we have Li(x)=18—6z. The differential equation (22), page 156, with ay"+(@—ay+y = 0 Substituting y =18~6z in this equation we have (0) + (8—a)(-6) + 18—6z = 0 which is an identity. Thus Li(st) satisfies the differential equation,CHAP. 8] HERMITE, LAGUERRE AND OTHER ORTHOGONAL POLYNOMIALS 168 816. Show that the Chebyshev polynomial T(z) is given by Tilt) = 2 - (jera-#) + (t)ea-a7 - (e)e-ta- We have by definition T,(2) = cos (n cost 2) Let w= cose so that 2 =cosu. Then T,(z) = cosmu. Now by De oivre’s theorem (cosu +isinuy" = cosmu + isinnu Thus cos nu is the real part of (cose +issinw)®. But this expansion is, by the binomial theorem, teas + ("Yeoswe-t@sinw) + (2) and the real part of this is given by (s)oner-ntdnat + cooru = (2)eoer-tusintn + (7)ocar-tu intr = Then sine cosu = x and sintu = 1—a?, this becomes a (= a2) + (jaa aoe 817, Find (a) To(z) and (b) Ts(z). Using Problem 8.16 we find for n=2 and n=$ respectively: 2 @ Mle) = #-(j)eu-» = at (lat) = Ber om = 2-()eaey = tanta = oo Another method. Since Ty(2) = co80=1, 7,(2) = cos (cos! 2) = x we have from the recurrence formula (27), age 156, on putting m=1 and n=2 respectively, Ty(e) = B2Ty(e) — Tole) = 2x4 Ty(e) = BeTy(2) — Ty(@) = 2e(2x?—1) — 2 = Ae — Be 818. Verify that T(x) = cos(n cos~!x) satisfies the differential equation (1-29y" — ay + n'y = 0 for the case = 3. From Problem 8.17(b), 5(2) = 424—S2 and the differential equation for» = (= 2y" ~ ay toy = 0 ‘Then if y= 42-82 the left side becomes (1 22)(242) — 2(122?— 8) + 9(4x8—32) = 0 so that the differential equation reduces to an identity.164 HERMITE, LAGUERRE AND OTHER ORTHOGONAL POLYNOMIALS (CHAP. 8 Supplementary Problems HERMITE POLYNOMIALS 819, Use Rodrigue’s formula (2), page 154, to obtain the Hermite polynomials Ho(2), Hy(2), Ha(z), Ha(2). 820, Use the generating function to obtain the recurrence formula (6) on page 154, and obtain Hy(e), Hylz) given that Ho(z) = 1, Hy(z) = 22, tat, show areay hat) fee oto te = 8. 822 Bratuate fate 2Hy() de. yn) 823, Show that H,,(0) = "6 824. Prove the result (7), page 154, by using a method similar to that in Problem 7.13, pages 188 and 139. 825. Work Problem 8.5, page 158, by using (a) Rodrigue’s formula, (b) the method of Frobenius. 826. (o) Expand (2) = #°—32t +22 in a series of the form 3, AuHiy(2). (6) Verity Parseval’s identity for the function in part (a). 827. Find the general solution of Hermite's differential equation for the cases (a) n=0 and (0) LAGUERRE POLYNOMIALS 828. Find L(x) and show that it satisfies Laguerre’s equation (11), page 185, for n= 4. 829, Use the generating function to obtain the recurrence formula (16) on page 155. 830, Use formula (15) to determine L,(z), Ly(2) and Ly(z) if we define Lq(z)=0 when » and L,(2) = 1 when n= 0. 831, Show that nL,_y(+) = mb,_(2) — Lala). sa Provethat f"eebsaytae = (th 83%. Prove the results (19) and (20), page 156. tae epen fo) = at-bat Be 8 an of he fom _S, Ast ass, musts Fasc etty fr Pron 83 tsi, mad te ener sin of Lauer erential qatin for = Ast. tain Lage diferent quton (1, age 185, om the ceeating fens 2), 8g IscELLANBOUS ORTHOGONAL POLYNOMIALS 838, Find (a) Liz), (0) Lie). 859, Verify the results (23) and (¢4), page 156, for m=2, n= 3. 840. Verify that Li(x) satisfies the differential equation (22), page 156, in the special case m = 2, 1CHAP. 8}, HERMITE, LAGUERRE AND OTHER ORTHOGONAL POLYNOMIALS 165 Bal. 842. 843. 8a. 8.46, sat. 848. Evaluate ff" ste-2t%(e) de Show that a generating function for the associated Laguerre polynomials is given by omeni-o |S Be aor eh ke Ye Solve Chebyshev's differential equation (26), page 156, for the ease where » Find (@) T(z) and) Ts(2). Expand f(a) = 9+ 22 4e-+2 in a serien of Chebyshev polynomials 3, AxTe(2) (a) Write Parseval's identity corresponding to the expansion of (2) in a series of Chebyshev poly- nomials and (b) verify the identity by using the function of Problem 8.45. Prove the recurrence formula (27), page 156. Prove the results (29) and ($0) on page 156. MISCELLANEOUS PROBLEMS 84s, 850, Bat, 832. 853, 854, 855. 856. 8st. (a) Find the general solution of Hermite’s differential equation. (6) Write the general solution for the cases where n=1 and n=2, (Hint: Let y=vH,(z) and determine v so that Hermite’ equation is satisfied.) In quantum mechanics the Schroedinger equation for a harmonic oscillator is given by Oy, rtm ae toe (B-fex2w = 0 where E, m, k,« are constants, Show that solutions of this equation are given by 0 = CyHalalayerstet Vie orpeyz ‘The differential equation is a Sturm-Liouville differential equation whose eigenvalues and eigenfunc- tions are given by H and y respectively. where =0,1,2,3,... and (a) Find the general solution of Laguerre’s differential equation. (b) Write the general solution for the cases m=1 and m=2, (Hint: Let y= vLq(z). See also Problem 849.] Prove the results (18) on page 158 by using the generating function. (a) Show that Laguerre’s associated differential equation (22), page 156, is obtained by differenti- sting Laguerre’s equation (11) m times with respect to x, and thus (8) show that a solution ig ae, /de™. Prove the results (28) and (24) on page 156. (a) Find the general solution of Chebyshev's differential equation. (b) Write the general solution for the cases n=1 and m=2 (Hint: Let y= oTy(z).) Discuss the theory of (a) Hermite polyno polynomi ials, (b) Laguerre polynomials, (c) associated Laguerre and (d) Chebyshev polynomials from the viewpoint of Sturm-Liouville theory. Discuss the relationship between the expansion of a function in Fourier series and in Chebyshev polynomials.Appendix A Uniqueness of Solutions A proof establishing the uniqueness of solutions to boundary value problems can often be accomplished by assuming the existence of two solutions and then arriving at a contra- diction. We illustrate the procedure by an example involving heat conduction. Consider a finite closed region R having surface S. Suppose that the initial tempera- ture inside ® and the surface temperature are specified. Then the boundary value problem for the temperature u(z,y,2,t) at any point (x,y,z) at time ¢ is given by au GH = xvi inside R w u(x, ¥,2,0) = f(z,y,2) at points (z,y,2) of R (2) ‘ult, ¥,2,t) = g(2,y,2,t) at points (x,y,z) of S @) ‘We shall assume that all functions are at least differentiable at points of ® and S. Assume the existence of two different solutions, say «1 and 1s, of the above boundary value problem. Then letting U=«a-ws we find that U satisfes the boundary value problem @ = ey inside R “) Ule,y,2,0) = 0 at points of R 6) U(2,y,zt) = 0 at points of $ O) Let us now consider Wo = FSS Ves ordeayas ” where the integration is performed over the region . Using (5) we see that wo) =0 (8) Also from (7) we have ® = SSS uvandya: = « S{S ova ae aves ) where we have used (4). ‘We now make use of Green's theorem to show that SS UyUdedyds = Sf vas - SIS) + (B+ eae (29) 167168 UNIQUENESS OF SOLUTIONS where 7 is a unit outward-drawn normal to S. Since U=0 on S the first integral on the right of (10) is zero and we have [froin = - ST) + B+ (B) ewe an ‘Thus we have from . ~- SSUY + BY +E It follows from this that dW/dt = 0, ie, W is a noningreasing function of ¢, and, in view of (8), that W(t) 0. But from (7) we see that W(t) =0. Thus it follows that W(t) =0 identically. Now if U(x, y,,t) is not zero at a point of & it follows by its continuity that there will be a neighborhood of the point in which it is not zero. ‘Then the integral in (7) would have to be greater than zero, ie. W(t)>0. ‘This contradiction with W(t)=0 shows that U(e, 4% t) must be identically zero, which shows that wu: = w: and the solution is unique. ) Jae dy de (a2)Appendix. B Special Fourier Series 1 0e fe 1 mente co Be cs gin — Hit go Co 12) ParF la) PE os or) wt os tp(a=t time (=) 178174 SPECIAL FOURIER TRANSFORMS SPECIAL FOURIER COSINE TRANSFORMS Aa) Fela) e IS cs ate o 10 wectente lo cmt tan of) 13 aa att see (nr/2) , O : C22 at 5 cae wae C24 ent =a note sin tan“) 2s artes ta 26 wot? VE gg etn a C27 oe z= —n se (nz/2) 5 C28 © BEG) Ob 32 tan-1 (2/b) me cas exes Fiend 6-34Appendix D Tables of Values for J,(x) and J,(x) e| 0 1 2 3 4 5 6 1 9 0. | 1.0000 9975 990097789604 93859120812 2075 1.| 7052 719667116201 500951184554 3080 2818 2,| 2239 10081104. ——.0555 2248 3. | ~2601 343 4018 4,| -s971 -s887 8768 —.2610 2097 5. 0758 ~.0088 1220 o.| 1506 17732017228 24ag 260127402851 2981 7.) 8001 «290129512882 ames 26682162846 1944 a] 717 4475 1222-0960 06920490146, 0858 a —1877 —768 2000-2218 2408 «| 0 1 2 8 4 6 6 7 9 0. | 0000 .0499 0995 1960 3290 4058 1] 4601 47094988 5419 878 B12 2] .sTe7 50835560 5202 4416 8154 a 2613 1792 0538 —0272 4“ Sut 5 8241 2951 6 0349 a| = 0543, 1096 1813 2192 8, | 246 2476 2680 2708 2697 2559 9. | 2053 25m 274 1816 1168 0884Appendix. E Zeros of Bessel Functions ‘The following table lists the frst few positive roots of Jq(x) =0 and Jie! Note that for all ceases listed successive large roots differ approximately by x — 3.14160. n=1 n=2 ” 4 » 6 240s8 = 8.8817 .1855 68802 7.5883 BTS (9.9961 5.5201 «7.0155 B47 ©7610» 11.0847 «12.8886 3.5803, 80587 10.1785 11.6198 «18.0152 14.8725 15.7002 17.0028 11.7015 19,8237 14.7960 «16.2285 17.6160 18,9801 20.8208, 14.9809 164706. «17.9598 19.4094 20.8269 R.2178 «23.5861 18.0711 19.6159 21.1170 22.8827 24.0190 25.4808 26,8202 0.0000 1.8412 «8.0542 42012, SITS wISG 7.801, 88517 5.3514 G.7061 O12 wz 10,5199 11.7349) T0155 8.6363 9.0605 1.3459 ©6819 «= s.0872 «15.2082 Sule) 101785 14,7060 «18.1704 14.5859 15.9641 «17.8128 «18.6874 18.8237 14.8636 «16.8475 17.7888 19.1960 20.5755 21.9917 164705 18.0155 19.6129 20.9725 aa.do10 29.8098 25.1899 qTAnswers to Supplementary Problems CHAPTER 1 127, u(0,t) = Ty, u(L,t) Ts w(x, 0) = f(x), |u{z,t)| 4, parabolic it 22+ ¥2 (A elliptic if M <1, hyperbolic if M> 1, parabolic if M 135. @) ety 2 au, Hu ue, Bt 5 Be ° ae az ate. te @e 13, (a) 2 + 8H = ee (b) oat ~ Sanay ae 0 188. (a) rz = F(z) + Gy) (6) wz = 26 + 2 + byt — 68 138 @) «= let) +G@—y (@) 2 = Fe+y) + Gy—2) 0) 4 = Fy ~20) (2 = Fety) +2Gie+y) () u = Fle + iy) + Gle— iy) 2 - a ‘140. (a) Fu-2) +o () uw = Fly) + 2G(y) + 2H(y) + Ky — 22) + = y= Fens cesy—6 We = Mery t citernZany+tomy ‘LAL. Fle + iy) + Glo — iy) + aH (wt iy) + 2J(x— iy) + (2 + N/A LAB. (a) w= dev-a02 (ew = Bete (0) w= B08 20a w= to0rs-m (0) u = 26-9 sin 3x — de-100t gin Be (a) « = 6e-**/* sin (wx/2) + 8e-** sin ox (0) w= 60m og BOE — gy tg hE @ vy = Zeinez sinzet ©) vy = Zsinter sin det ~ 2 sin dex sin rt 145, = eM e—t singe — @-1% sin dx) 179180 CHAPTER 234. 236. 2a. 238, 239, 22, 253, 255. 251, 250, S (cos mz) 5. ure n sin 2nz 1 Ow 3 [es to sine in Hatt ane + Bf Ha) sin cos | ANSWERS TO SUPPLEMENTARY PROBLEMS © 5 att sin (20 ; a) en D coats — Se sin tt} (© 2 = 0,=10,220,. (@) © = =3,20,215,..58 + cos 8% 4 } BOG (6) 0) = fe) Same answer as 2.37. 23 {JF 9 = 0 sin 22 ae eres gig BE att 120 ++ 60)? eos 2p 1 pq Qn—Dee .,(2n—tset 3 [eactas Ste sin ae] sin 2 sin 2H + 3 ew Senate 28 sinh (ay) £ -(y) sin = au sin, = sinh == iat Rie+en + Hea) +f ow du261 268. 264, 265. 2.66, 2.0, am 27, 276, ANSWERS TO SUPPLEMENTARY PROBLEMS 181 2eust) = 3S 3 [Ann co Amat + Bru it hget] sin MEE gin RH where Ban = iL Sern sin ME sin 2H doy, ars SS oem sin 2 sin dear, and 2 ues) = ve) —F where we) ga =a ules) = where west) = weet) = wtes = o6) + BS {fre oto) sin tgae} esnteat inte B where + Ba-em Bae Same as 2.69 but with ¥(2) ven = v0) + FS {feos ana sin Ean where ye) = ey uleyys2) = 3, 3, Byyy Sin mre sin nay sinh Ym? +n rz vce tan = ara S [te en mee a ey urs) a 3) fg He) sin ag} in inte182 ANSWERS TO SUPPLEMENTARY PROBLEMS CHAPTER 3 35. (a) —V, a2 = 2V3, a; = V5, a = (0) 1, VB@e~1, VB (6st 6241) —8V5, a5 =6V5 Bat, @) 4, tne, LB dat a, 0) afEesoenrts) = a3, fEonoomts) (ana...) 320, 42) = 3 cstala) where ey = f(a) 2) a6) de ga. fa) = B+ 3 eapate where «=a S912 400 cos corrtay dr = af2 Ff seo) come ate 3a, (@) 2-39) VORF WV aan, VB, 4[s, 4[E (252). te te normatzes Legend polynomials B81 1-2, HO— det at, 46-18 +922 28) 288. (6) eigenvalues me, eigenfunctions By sin mes, where m= 1,2, (@ VBsin mez, m=1,2, @n=1)e (m= ee @m—Vee a, (a) COMED |p, ggg 2M MEE, YE cog BME, m= 1,2... 0) me, Ba sinmes, VEsin mes, m= 1, 240, ae, taf Sf 1 on Bee conaatatt (®) Heat conduction in an infinite strip of width L, with one side at 0°, the other side insulated, and intial temperature distribution given by f(z). = 2 STC sey cin nD re 5, n= Heat aa) vest) = FS [fsa sin AED vu | coe Oa () Vibrating string with end 2 = 0 fxed, end # = L fee, initial shape /(e), initial speed zero CHAPTER 4 428, (a) 20, @) 10/005, (@) gor 4st, (@) 1105, () 4/15, (@) DIVE aa. 438, (@) 160, 0) 2, (0) Be 428 439. (a) Ite, © 41, (@) 24, () 87128, © br) 4, (@) 85/256, (0) Bes 432, (@) Gsv=)/108, @) —ari2/ 442, (@) 16/15, @) 8/105ANSWERS TO SUPPLEMENTARY PROBLEMS: 188 CHAPTER 5 sa. a) BE 50. @ Et 5.5, cora—sina) — 0) SE sa. Mw) = ® ow acoso sine wuny <2 tant 526, @ =782 gy Hm Bal. ulesn) = Beant sm.) ts . 55. ula) 528, yle) = (2+ Boose — 4 con2x)/ee 56, utes) CHAPTER 6 637. (a) 2[2 = sing ha C08 | oy fZ[eeet gases) oan (85%) ate ~ S40 640, (a) 2Us(e) +e (B) Bol) BI) (0) a2 (2) + date) — F Jol dr ty aay Se) e hte) GAL. (a) 6V2I,V2) ~ BVH IQR) + € @ 3 -Ay sf Iola) de 642, aJo(z) sine — 2J,(2) cosx + ¢ tat.) Eee afZsne © -fE (sine +282) fZ(s2 ~ coe) 659. (a) 28¥g(x) te (0) Yala) — 2Y, (x)/e + (0 He Yie) ~ Ge ¥e@) ~ gE Vae +f f Yotwrae ss, te-ra-mm —) fEeuranmm 622, y = Adg(Vz) + BYyWV2) a7, (a) y = Senet Boos Ov = VE AS a2 + BF 1a(h2)) 644. y = Adg(et) + BYg(er) 6.75. (b) y = AJo(2¥x) + BYo(2Vz) 6.76, (0) y = AVE Ia a(G2¥9) + BYE I-13) 698, y = Ady(e; + BeY\(2)184 6.96. 6a. 6113, 67. ANSWERS TO SUPPLEMENTARY PROBLEMS ule, 4s0) = 3, AyJa(rgo) c08 99 cos dyt where Ay are the positive roots of Jy(h) = 0 and BIAS ~ 8)Jo (Au) — Ges) + 8) A NFEOW) _ & JolwvF) wed = 3 Hoo cos (fat) S He) Jo(\yV%) dx where Jo(y) & & Jaap sinh lea) @ winged) = ie: = (Aa sin ng + Bax cos m) where Aan = SS oflore%n 0) sin ng dos Baa = fff ntottan cen dete and Jn(xa) = 0 2 S& Aix) sinh xo — 2) ©) woes) = 33 “Five sinha Beene ex where Be = Gu[ addons ~ 3 Joon de] and Js 0) Me) = B.S eege MO glean) sin kee where Cem a oe (ry (p, 2) sin kez Iz Biscy J, J, rome Po2) in hrs dd and Ja(rq) =0 et 2,850) = 3, pAmotllhmap) C08 Nyact + 3, 3 Ann cose + Brn sin ms)tauae) C05 att where usO\nop) = Inna) YuChmal) ~ Jn (rent) Yo nabs han = I” (otlovernihnns cone dodo, J" foe, 29m ms) sin ded, elaine]? dp and ¢= ValaANSWERS TO SUPPLEMENTARY PROBLEMS CHAPTER 7 732. Pye) = Hea—anm0 +3609 P(e) = Ee — 1008+ 6228 783, (a) 0 (6) 2/5 (e) 0 eal bat waa fe) = ety i(1E2) - tv = ast ofa dm(2=2)] OnE ane eel 145. —SPo(2) + Pyle) ~ Pryce) + Brae) le) +5 Pye) 2 1A6, Pola) + 7P ila) + Pole) — Fg Pala) +--+ 148 (@) 9 = my + Boy cos 0 v= tee 7.49. pet 1 — rp, 4.) v= Me cos0] Boar _ Palos wv = Sepa) CHAPTER 8 81. 4, B6, 422, 89018 $28 do!—2, 828 Ite 822. VF if n=0, 27 if n= O otherwise 826. (a) ~$Holx) + FHylz) — BHy(2) + pHy(e) tn Wyn atea forte y= eaten f See 828, Life) = 24~ Bb + Rat — 8 + at 750. 71. 156. 151. 1163. 12, ra, 830, 8a. 838, Bal. 83, 185 m(i-4) mir where ¥ >a is the distance from the center of the sphere (@) 30-29) 5 2— 2124) 0) ~ 3-242 2124) (©) —A052(1 — 2799 -v=a[tn(22) +725] T=, a ae @ Inside,» = vor sin®¢ cos # cos 36 Outside, v = sin? cos ¢ cos 39 au, ably 9) a +o 1 = [Ade ale) + Bln 2207] + [AaPa(c0s #) + BsQq (cos #)]e-™* Le) = 2 Ly@) = 6 ~ 182 + 928 — 28 de tat 2Lgle) — BLy( a) + 6Ls(2) — Lyle) v= atef Se Oy = ota teft—o f tate Ha) = 144 — 960 + 1208 Lie) = 1296 + 6002 — 6022 180 w= Asintet B186 ANSWERS TO SUPPLEMENTARY PROBLEMS 84 (a) Tye) = Bet Bt +1 B51 (0) y = Ala(a) + BLA) f eee (6) Tyla) = 1625 — 202 + Bx ° - de kas ner sar—ieryeysriney MEO Ato) BI) Fe eda. (Ha) Sas, (0) y = AN) + BH(2) f°INDEX Absolute convergence, 24, 89 ‘Absolute temperature, 10 Acceleration due to gravity, 4, 17 vibrating string under, 17, 98. Analytic continuation of gamma function, 67, 11, 72 Angular momentum, 121 Approximations to functions, 34 in the least-squares sense, 58, 54, 57, 58, 64, 65 Arbitrary functions, 2, 12 Associated Laguerre polynomials, 156, 162, 165 Associated Legendre funetions, 192, 144-150 differential equation for, 182, 144-146 ‘expansion of functions in s Legendre functions orthogonality of, 132 Parseval’s identity and, 152 potential problems involving, 148-150 series of, 182 solutions of Laplace’s equation using, 148 Associative law of algebra, 82, 86 Asymptotic formulas, for Bessel functions, 101, 125 for gamma function, 68, 76 Asymptotic series or expansions, 70, 16, 77, 79, Bar, heat flow in, 9, 10, 17 assuming Newton's law of cooling, 17 infinite, 95, 96 semi-infinite, 89, 90 Beam, vibrating, 3, 4 Ber and Bei functions, 100, 111, 112, 125 graphs of, 100 Bessel functions, 97-129 asymptotic formula for, 101, 125 differential equation for, 97 of the first kind, 97, 98, 104-109, 123 functions related to, 99, 100, 111, 112, 124, 195, gamma function and, 98 generating function for, 99, 108, 109, 128, 124 graphs of, 98, 100 Hankel functions in terms of, 99, 111, 125 integral representation for, 109 integrals involving, 106, 107 linear independence of, $8, 104, 105, 108 ied (see Modified Bessel functions] of order half an odd integer, 98, 105 orthogonality of, 101, 112, 113 recurrence formulas for, 99, 105, 106, 109, 111 of the second kind, 97, 109-111, 119-122 series of, 102, 118, 114 solutions using, 102, 114-122, 196-128, table of values for, 178 zeros of, 101, 177 Bessel’s differential equation, 97, 102-104, 108 equations transformable into, 101, 112, 125 187 Beasel's differential equation (cont,) general solution of, 97, 104, 105, 109, 110 's inequality, 34, 58, 65 for Fourier series, 24 for orthonormal series, 58 for vectors, 64 Beta function, 69, 78-76 connection of with gamma function, 69, 73-75 definition of, 69 evaluation of integrals in terms of, 69, 72-76 Binomial theorem, 136 Boundary conditions, 2, 1, 15 for heat conduction, 10 for radiation, 10 for vibrating string, 6 Boundary value problems, 1-19 Bessel functions and, 102, 114-119 [ece also Bessel functions} definition of, 2 Fourier integrals and, 82, 89-91 [see also Fourier integrals] Fourier series solutions for [eee Fourier series] involving heat conduction, 7-10 involving vibrating string, 6, 7 ‘mathematical formulation of, 1, 6-10 methods of solving, 5 physical interpretation of, 1, 16, 17, 19, 48, 49 solution of, 1, 11, 19, 38 solved by separation of variables [see Separation of variables) Boundedness condition, 6 Be Cantilever beam, 96 Cauchy or Euler differential equation, 40, 401, 134 Celsius degrees, 10 Centigrade degrees, 10 Chain, vibrations of, 127 Characteristic functions, 55 Characteristic values, 55 Chebyshev polynomials, 156, 163, 165, Gireular eylinder, heat flow in, 115-117, 126-128 Gireular disc, potential of, 151 Circular membrane, 8, 45, 46, 117-119, 126 Circular plate, temperature of, 39, 40, 48, 60, 114, 115 Classification of partial differential equations, 40, 11, 18 Commutative law of algebra, 82, 86 Complementary error function, 69 Completeness, 34, 54, 65 Complex notation for Fourier series, 24 Components of vectors, 52 Conducting medium, 7 Conduction of heat [see Heat conduetion equation] Conductivity, thermal, 3, 7, 62, 63188 Continuity, piecewise [ree Piecewise ‘continuous functions] “points of, 28, 80, 81 Continuous functions, 21, 25 piecewjse [ace Piecewise continuous function: Convergence, of Fourier series, 21-24, 27, 25-37 of orthonormal series, 53 of series of Bessel functions, 98 uniform [eee Uniform convergence] Convergence in mean, 54 Convolution theorem, 82, 95 proof of, 85, 86 Cosine and sine integrals, 69, 79 Fresnel [eee Fresnel sine and cosine integrals] Cosine and sine transforms, Fourier, 81 Cylinder, heat low in, 115-117, 126-128 Cylindrical coordinates, 4 Density, 3, 7, 8, 63 Density function {see Weight function} Dependent variables, 10, 11, 18 Derivative, partial, 2, 6 Differentiable function, 12 Differential equation, 1 Differentiation and integration, of Fourier series, 24, 47 of series, 23, 24 Diffusivity, 3, 8, 9 Dirichlet conditions, 22-24, 27-29, 102, 120 ‘and Bessel series, 102 and Fourier series, 22 and orthonormal series, 53 and Parsevals identity, 23 see also Parseval’s identity] ‘chlet problem, 5, 43 Discontinuities, 21, 25, 26, 28, 46, 80, 81, 102 Distributive law of algebra, 82 Double Fourier series, 24, 25, 97, 45, 46 Drumhead, vibrations of [ece Membrane] Duplication formula for gamma funetion, 68, 15,76 proof of, 75, 76 Rigenfunctions, 54, 55, 58.61, 65 orthogonality of for Sturm-Liouville systems, 55, 58-61, Eigenvalues, 54, 55, 68-61, 65, reality of for Sturm-Liouville systems, 59 Elasticity, modules of, 8 Electric potential, 8 [see aleo Potential] Elliptic partial differential equations, 2, 11, 18 Equilibrium position, 3, 45 Error function, 69, 77, 90, 95 Error, mean square, 24, 58, 57, 58, 64 Euler or Cauchy differential equation, 40, 101, 194 Euler's constant, 68, 78, 98 Even and odd extensions, 32 Even and odd functions, 22, 20-33 Existence of solutions, 1, 2, 167 theorems on, 2 Experiments, 1 Exponential integral, 69 Factorial function (sve Gamma function] Finality, principle of, 54 irst harmonie, 45, Fourier, 1 Fourier coefficients, 21, 27, 83, 24, 36, 117 ‘ealeulation of, 27-28, 31-33 for double Fourier series, 24, 25, 87, 45, 46 generalized, 53, 56, 58 motivation for definition of, 27 Fourier expansion [ace Fourier series} Fourier integrals, 80.96 [see also Fourier’s itegral theorem) applications of, 82 conditions for validity, 80 definition of, 80 need for, 80 for odd and even functions, 81 Parseval’s identity for, 82 solutions to problems using, 89-91, 95 Fourier series, 1, 20-51 ‘applications of, 20-51 Bessel’s inequality for, 2 ‘complex notation for, 24 ‘convergence of, 21-24, 27, 85-87 definition of, 21, 22, 27 half range, 22, 23, 29-88, 97 integration and differentiation of, 24, 27, 34, 35, need for, 17, 20 orthonormal series as generalizations of, 53 Parseval's identity for, 28, 93, 34 solutions using, 48 special, 169-172 used in summing series, Fourier sine and cosine transforms, 81 Fourier transforms, 81, 83-85 convolution theorem for, 82, 85, 86, 95 inverse, 81 solutions to problems by, 92-94 special, 173-175, Fourier’s integral theorem, $0 [see also Fourier integrals] condition for validity of, 80 ‘equivalent forms of, 80, 81, 83 proof of, 87-89 Frequency of normal modes, for drumhead, 119 for string, 44 Fresnel sine and cosine integrals, 69, 79 Frobenius, method of, 97, 104, 184, 135 Fundamental frequency or first harmonic, 45, 46 Fundamental mode, 45 Gamma function, 67-79 analytic continuation of, 67, 71, 72 asymptotic formula for, 68, 76 Bessel functions and, 98 connection of with beta function, 69, 73-75 definition of, 67 duplication formula for, 68, 75, 76 evaluation of integrals using, 69, 71-78 ‘graph of, 68 miscellaneous results involving, 68 recurrence formula for, 67, 70 Stirling’s asymptotic series for, 68, 76Gamma function (cont.) table of values of, 68 General solution of a partial differential equation, 2, 5, 12, 13, 14, 18, 19 Generalized Fourier coefficients, 58, 56, 58 Generating functions, for Bessel functions, 99, 108, 109, 123, 124 for Chebyshev polynomials, 156 for Hermite polynomials, 154, 157 for Laguerre polynomials, 155 for Legendre polynomials, 131, 126, 137, 129 Gradient, 7 Gram-Schmidt orthonormalization process, 55, 61, 62, 65 Gravitational potential, 8, 18, 143, 144 [sce also Potential] Laplace's equation and, 3 Gravity, finite string under, 17 infinite string under, 96 Green's theorem, 167 Half plane, solution of Laplace's equation in, 91 Half range Fourier sine or cosine series, 22, 28, 29-83, 37 Hankel functions, 99, 111, 125 Harmonic oscillator, 165 Harmonies, 45 Heat conduction equation, 3, 9, 10, 62 applications of, 1, 16, 25, 87-42, 89-94 Aerivation of, 7, 8 Laplace's equation as, 3 ‘one-dimensional, 10 for a thin bar, 9, 10, 62, 63, ‘uniqueness of solution to, 167, 168 vector method for deriving, 9 Heat flux,7 Hemisphere, heat flow in, 142 Hermite polynomials, 154, 155, 157-160, 164 differential equation for, 154, 158, 165 generating function for, 154, 187 orthogonality of, 154, 155 Parseval's identity for, 159, 160 recurrence formulas for, 154, 157 Rodrigue’s formula for, 154, 157 series of, 155, 158, 150 Hollow sphere, potential of, 141, 142, 151 Homogeneous partial differential equations, 8 10, 18 Hyperbolic partial differential equations, 2, 11,18 Independence [sce Linear independence] Independent variables, 10, 11, 18 Indicial equation, 97, 104, 195 Inertia, moment of, 4 Infinite-dimensional vector, 52 Infinite series, 23, 24 convergence of, 23 uniform convergence of, 28, 24 [see also Uniform convergence] Initial temperature, 10, 41 Insulated surface, 9, 16, 38, 39, 41, 62, 89 boundary condition for, 10 Integral equation, 85.87, 94, 95 189 Integration and differentiation, of Fourier series, 24, 27, 4, of series, 23, 24 Inverse Fourier transforms, 81 Jacobian, 86 Kelvin temperature, 10 Ker and Kei funetions, 100, 125 graphs of, 100 Kronecker's symbols, 52 Laguerre polynomials, 05, associated, 156, 162, 165 differential equation for, 155 ‘generating function for, 155 important properties of, 155, 156 orthogonality of, 156, 160, 161 recurrence formulas for, 185, 161 Rodrigue’s formula for, 155 series expansions involving, 156, 161 Laplace's equation, 3, 9, 18, 42, 43, 97 associated Legendre functions and, 182 Bessel functions and, 97, 102, 108 in eylindrieal coordinates, 4 Fourier integral solutions of, 91 Fourier series golutions of, 25, 42, 43, gravitational or electric potential and, 3, 18, 148, 144 [sce also Potential] Legendre functions and, 180, 198, 135, 196 in spherical coordinates, 4, 5 Laplacian, 3 in cylindrical coordinates, 4 in rectangular coordinates, in spherical coordinates, 4, 5 Least-squares sense, approximations in, 53, 54, 57, 58, 64, 65, Left and right hand limits, 21 Legendre functions, 130-183 [ace also Legendre polynomials] associated (see Associated Legendre functions] differential equation for, 190, 123-186 of first kind, 130 linear independence of, 136 ‘of second kind, 130, 131, 13 series of, 131, 192, 139-141, 15 Legendre polynomials, 62, 130-182, 135, 136-144, 150, 151 derivation of, 136 generating funetion for, 181, 126, 187, 189 orthogonality of, 131, 138, 139 recurrence formulas for, 181, 1 Rodrigue’s formula for, 130, 136 series of, 181, 132, ‘Legendre’s associated differential equation, 192, 144-140 general solution of, 192 Legendre’s differential equation, 130, 183-136 general solution of, 130 L'Hospital's rule, 110 Limit in mean, 54 Limits, right and left hand, 21 3 138190 Linear independence, of Bessel functions, 98, 104, 105, 108, of Legendre functions, 136 Linear partial differential equations, 2, 5, 10, 11, 13, 14, 18 Longitudinal vibrations of a beam, 2, 4 Mass, 3, 8, 72 ‘Mathematical formulation of physical problems, 1, 6-10 Mathematical models, 1 Mean of a funetion, 21 ‘Mean, limit in, 54 Mean’square error, 34, 58, 54, 57, 58, 64 Membrane, circular, 117-118, 126 equation for vibrations of, 3 square, 45, 46 ‘Models, mathematical, 1 Modes of vibration, 44-46, 119 Modified Bessel functions, 99, 100, 124 differential equation for, 99 graphs of, 100 ‘Modulus of elasticity, 3 Moment of inertia, 4 Momentum, angular, 121 Music, 44, 46 Natural or normal modes of vibration, 44-46, 119 Necessary conditions, 22 Neumann functions, 97 Newton's law, 1, 6,772 cof cooling, 10, 17, 62, 63, 127 Noise, 119) Nonhomogencous partial differential equations, 2, 10, 18, Nontrivial and trivial solutions, 55, 58 Normal, 7 outward-drawn, 9, 168, unit, 9, 168 Normal or natural modes of vibration, 44-46, 119 Normalization, 55 Normalized function, 52, 65 Normalized vector, 52 Odd and even extensions, 32 Odd and even funetions, 22, 29. Order, of Bessel functions, 97-9 of a partial differential equation, 2, 10, 11, 12,18 Ordinary differential equations, 5, 14 used to solve partial differential equations, 13, Orthogonal functions, 52, [ace aleo Orthogonality] definitions involving, 52, 53 eigenfunctions as, 55 [see also Orthogonal sets, 56 veetors, 52 Orthogonality, 52, 58 of Bessel functions, 101, 112, 113 of eigenfunctions, 58 Orthogonality (cont) of Hermite polynomials, 154, 155 of Laguerre polynomials, 156, 160, 161 ‘of Legendre polynomials, 131, 138, 139 with respect to weight function, 53 Orthonormal series, 83, 58, 56, 64 Bessel’s inequality for, 58 expansion of functions in, Pareoval’s identity for, 54 Orthonormal sets, 62, 53, 57 Overtones, 45 Parabolic partial differential equations, 2, 11, 18, Parseval’s identity, 28, 83, 84, 47, 48, 57, 58, 64, 65, 126, 152 for Fourier integrals, 82 for Fourier series, 25, 38, 34 for Hermite polynomials, 159, 160 for orthonormal series, 54 for vectors, 64 Partial derivatives, 2, 6 Partial differential equations, 1, 2 classification of, 10, 11, 18 definitions pertaining to, 2 elliptic, hyperbolic and parabolic, 2, 11, 18 existence and uniqueness theorems for, 1, 2, 167, 168 homogeneous and nonhomogeneous, 2, 10, 18 linear, 2, 5, 10, 11,18, 14, 18 order of, 2, 10, 1, 12, 15 solved as ordinary differential equations, 5, 13 some important, 3, 4 Partial sum, 23, 34 Particular solutions of partial differential equations, 2, 5, 12, 13, 18 obtained by separation of variables, as special cases of general solutions, 6, 12 Pendulum, vibrations of, 121, 122 Period, 20, 25, 28, 29-23, 35, 36 Teast, 20, of vibration, 45 Periodic functions, 20 Physical interpretation of boundary value problems, 1, 16,17, 19, 48, 49 Physical laws, 1 Piano string, 44 Piecewise continuous functions, 21, 34, 35, 96, 88 Polar coordinates, 39, 73 Potential, of circular disc, 151 of hollow sphere, 141, 142, 151 of ring, 143, 144 of sphere, 148-150 Potential equation, 3, 96 [see algo Laplace's equation} Quantum mechanies, 165 Radiation, heat conduction with, 62 heat flax of, 10 Stefan’s law of, 10 Rectangular coordinates, Laplacian in, 3 Rectangular plate, 49Recurrence formulas, for Bessel functions, 99, 108, 106, 109, 111, for Chebyshev polynomials, 156 for gamma fonction, 67, 70 for Hermite polynomials, 154, 157 for Laguerre polynomials, 155, 161 for Legendre functions, 131, 187 Riemann's theorem, 35, 88 Right and left hand limits, 21 Ring, potential of, 143, 144 Rodrigue's formula, for Hermite polynomials, 194, 157 for Laguerre polynomials, 165 for Legendre polynomials, 130, 136 ‘Root mean square error, 58, 54, 57, 58, 64 least or minimum, 54, 57, 58 Scalar product, 52 ‘Schroedinger equation, 165 Separation constant, 55 leading to Sturm-Liouville systems, 55, 62-64 Series, of Bessel functions, 119-122, 125 ‘of Hermite polynomials, 155, 158, 159 of Laguerre polynomials, 156, 161 of Legendre functions, 181, 182, 189-141, 151 Sine and cosine integrals, 69, 79 Fresnel, 69, 79 Sine and cosine transforms, Fourier, 81 Singular solution, 2 Slab, heat flow in, 10 Solutions of partial differential equations, 2, 5, 13, 14, 18 existence and uniqueness of, 1, 2, 167, 168 methods of finding, 13, 14, 18,'19 Spocial functions, 67-79 Specific heat, 8, 7, 8, Sphere, potential of, 48-150 Spherical coordinates, 4, 5 Square plate, temperature in, 41, 43, 48, 49 transverse vibrations of, 49 Steady-state heat flow, 9 Steady-state temperature, 3, 38, 42, 48, 50, 51 Stefan’s radiation law, 10 Stirling’s approximation for n!, 68 roof of, 76 Stirling’s asymptotic series for gamma function, 68,76 Strai String, vibrating, &, 6,17, 45, 48, 98, 94, 96 [see aiso Vibrating string equation] under gravity, 17, 96 Sturm-Liouville systems, 54, 5 Sufficient conditions, 22 ‘Sums of series, Fourier methods for finding, 38-35, Superposition principle, 5 61, 65, 66, 128 ‘Temperature, 3, 7, 9, 10, 16, 87-43 ‘Tension of a string, 3, 6, 45 191 ‘Thermal conductivity, 2, 7, 62, 68 Torque, 121 ‘Transformation equations, between rectangular and cylindrical coordinates, 4 between rectangular and spherical coordinates, 5 ‘Transverse vibrations, 3 of a beam, 4 of a membrane, 45 of a plate, 49 of a string, 3 ‘Trigonometric expansions, 17 Trivial and non-trivial solutions, 55, 58 Undetermined coefficients, method of, 14 Uniform convergence, 23, 24, 27, 38, 47, 89 of orthonormal series, 55 theorems on, 28, 24 Weierstrass M test for, 22, 24 Uniqueness of solutions, 1, 2 ‘proof of, 167, 108 theorems on, 2 Unit normal, 8, 168 Unit veetors, 5 Variables, 2, 10, 11, 18 Vector method for deriving heat conduetion equation, 9 Vectors, 52: Bessel’s inequality for, 64 ‘components of, 52 expansion of, 53 fanctions as, 52 infinitesdimensional, 52 normalized, 52 Parseval's identity for, 64 unit, 52 ‘Vibrating beam, 3, 4 Vibrating chain, 127 Vibrating membrane, 43, 46 Vibrating string equation, derivation of, 6 generalized to higher dimensions, with gravity term, 17 solution of by Fourier integrals, 93, 94 solution of by Fourier series, 43-45 ‘with variable density and tension, 7, 128 Vibrating systems, 25 [ace also Vibrating bea Vibrating string equation) Vibration, modes of, 44-46, 119 Violin string, 8, 44 6, 7, A246 ‘Wave equation, 152 Weierstrass M test, 23, 2¢, 89 Weight function, 53, 55, 64, 101 for Bessel functions, 101 for Hermite polynomials, 154, 155, 197, 158 for Laguerre polynomials, 156, 160, 161 Wronskian, 108 Zeros of Bessel functions, 101, 177