MATHEMATICAL METHODS IN PHYSICSZO i Lage 1496 MATHEMATICAL METHODS IN PHYSICS Ye wera Scientific Singapore « New Jersey * LondoneHong KongPublished by World Scientific Publishing Co. Pte. Ltd. PO Box 128, Farrer Road, Singapore 912805 USA office: Suite 1B, 1060 Main Street, River Edge, NJ 07661 UK office: 57 Shelton Street, Covent Garden, London WC2H SHE 1Data . D. Lindenbaum, P. om. ISBN 9810227604 1, Mathematical physics. 1. Title. QC20.1498 1996 530.1'5--de20 96-31604 cr British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library. Copyright © 1996 by World Scientific Publishing Co. Pte. Ltd. All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means, electronic or mechanical, including photocopying, recording or any information storage and retrieval system now known or to be invented, without written permission from the Publisher. For photocopying of material in this volume, please pay a copying fee through the Copyright ‘Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the publisher. This book is printed on acid-free paper. Printed in Singapore by Uto-PrintPreface This book comprises a set of class notes on a Columbia University (New York) 2-semester graduate course in Mathematical Methods in Physics. Although the course is normally taken by first-year MA or PhD physics majors, it can be used by senior undergraduates having the requisite mathematical background and majoring in physics, engineer- ing or other technically related fields. Some studies, not usually found in standard mathematical physics texts, are presented to complement the basic theories. These include: the theory of curves in space in Chapter 1; analytic continuation in Chapter 6; and retarded and advanced D-functions in Chapter 11. In addition, applications of the theories are inserted throughout the text where deemed appropriate. These include: moment of inertia in “Tensor Analysis”; Maxwell’s equations, magnetostatics, stress tensor, continuity equation, electrostatics, and heat flow in “Tensor Fields”; special relativity in “Matrix and Vector Algebra in N-Dimensional Space”; Fourier series, Hermitian operators, Legendre polynomials and spherical harmonics in “Hilbert Space”; electrostatics, hydrodynamics and Gamma function in the study of analytic functions in “Theory of Functions of a Complex Variable;” vibrating string, vibrating membrane and harmonic oscillator in “Theory of Ordinary Differential Equations”; age of the earth and temperature variation of the earth’s surface in “Heat Conduction”; and field due to a moving point charge (Liénard-Wiechert potentials) in “Wave Equations”. Lastly, Appendix A (Part I) and Appendix B (Part II) comprise approximately 150 pages of problems, with their solutions, which represent applications as well as extensions to the theory.vil Contents Preface... ee ee ee ee ee v PARTI Chapter 1 Vector Analysis 1.1 Vector Algebra 1.1.1. Rotation of Coordinate Axes . 1.1.2. The Inverse Rotation . . . 1.1.3 Definition of a Vector . . . 114 EqualiyolNedos 1.1.5 Addition (Subtraction) of Vectors... . . 1.1.6 Geometrical Interpretation of Vectors. . . 17 ane 11.8 11.9 144 1A Scalar Product ...... Example of Scalar Product Definition of a Scalar. . . 0 Unit Vectors . 11 Vector Product 1.2 Examples and Applications . 1.21 Central Forces ...... 1.2.2. Infinitesimal Rotation . . . BaaA 1.3 Theory of Curves in Space... 6... 2... eee eee eee 12 ONOoOnMaOaANS Chapter 2 Tensor Analysis 2.1 nth Rank Tensor 2.2 2nd-Rank Isotropic (Invariant) Tensor 2.3 Contraction 2.4 Outer Product Theorem 2.5 Srd-Rank Isotropic (Invariant) Tensor... . . 2.6 Examples and Applications 2.7 Geometrical Representation of Tensors 2.8 Moment of Inertia Tensor Chapter 3 Fields 3.1 Tensor Field 3.1.1 Definition of a Tensor Field 3.1.2 Operators of a Tensor Fieldvill 27 3.2 Gauss’ Theorem . coer 3.2.1 Applications in Physics... .. 3.3 Stokes’ Theorem 39 3.4. Connectivity of Space 40 3.5 Helmholtz Theorem 44 3.6 Equivalent Forms of Gauss’ and Stokes’ Theorems . . . 49 3.6.1 Gauss’ Theorem . 49 3.6.2 Stokes’ Theorem . . - 50 3.7 Maxwell's Equations 50 3.7.1 Magnetostatics . . . 50 3.7.2 Faraday's Law . . . 52 3.7.3 Wave Equations . 52 3.7.4. Physical Significance of Solution to Wave Equation. » 54 3.8 Curvilinear Orthogonal Coordinate Systems . 55 3.8.1 Vector Components... 2... eee eens 6a Chapter 4 Matrix and Vector Algebra in N-Dimensional Space 4.1 Algebra of N-Dimensional Complex Space 63 4.1.1. Rotations in N-Dimensional Space . 64 4.2 MatrixAlgebra . 2... 2 ee . 66 4.2.1 Commutation of Matrices 4.2.2 Matrix Inverse. . . 4.2.3 Special Matrices 4.3 Examples of Matrices 4.3.1 Orthogonal Matrix . 43.2. Special Relativity . 4.4 Tensor Analysis in N-Dimensional Space 83 4.4.1 Tensors and Pseudo Tensors . . . ae 4.4.2 Similarity Transformations . 88 4.5 Matrices in N-Dimensional Space 90 4.5.1 Diagonalization of a Matrix 4.5.2 Applications of Matrix Diagonalization - 96 4.6 Linear Independence and Completeness . 102 4.6.1 Orthonormality and Linear Independence 102 462 Completeness ......... 000000 103 Chapter 5 Hilbert Space 5.1 Definitions 2... ee eee 105 5.2 Weierstrass's Theorem... 6.1... eee ee eee 1085.3 Examples of Complete Orthonormal Sets ........---+ 11 6.3.1 FourlerSerles.... 0-2 ee ee eee eee eee 411 5.3.2 Generalization of a Hermitian Operator in Hilbert Space... 114 5.3.3. Fourier Integral Theorem; Fourier Transforms 5.3.4 Piecewise Continuous Functions . . . « 5.3.5 Legendre Polynomials... . 5.3.6 Multipole Expansion 5.3.7 Recursion Formulae for Legendre Polynomials deen eee 138 5.3.8 Spherical Harmonics Appendix A. Problems and Solutions ProbemsssA J 152 PART II Chapter 6 Theory of Functions of a Complex Variable 6.1 Theory of Complex Variables . . 233 6.1.1 Complex Numbers 233 6.1.2 Complex Variables : 233 6.1.3 Addition of Complex Variables. 234 6.1.4 Functions of a Complex Variable : 234 6.1.5 Calculus of Functions of Complex Variables. . 234 6.2 Analytic Functions . 235 6.3 Applications of Analytic Functions . 237 6.3.1. Electrostatics . . . 237 63.2 Hydrodynamics . . 241 6.4 Integral Calculus of Complex Variables 245 6.4.1 Cauchy Integral Theorem 246 6.4.2 Residue Theorem . 247 Taylor’s Theorem Laurent Theorem Singularities . . Liouville Theorem . Multiple-Valued Functions 6. 10 Theory of Residues .......-.- 6.10.1 Complex Integration 6.11 Analytic Continuation 6.11.1 Gamma Function (T) . .Chapter 7 Theory of Ordinary Differential Equations 7.1 Ordinary Differential Equations in Physics 276 TAA Vibrating String... ........000008 7.1.2 General Properties of the Linear Operator L. . . 278 7.1.3 The 3-Dimensional Schrédinger Equation .. . . 279 7.1.4 Separation of Variables... ... ef - 280 7.2 Ordinary Points and = Points ce 7.2.1 Iteration Method . 287 7.2.2 Power Series Method 292 7.3 Hermite Polynomials . 293 7.3.1 Hermite Polynomials and the Harmonic Oscillator... .. . 297 7.3.2 Hermite Polynomial Theorems: 299 7.4 Behavior of Solutions Near Singular Points . 303 7.5 Bessel Functions 31 7.5.1 Bessel’s Equation — Singularities 316 7.5.2 Bessel Functions — Theorems . 320 7.5.3 Bessel Functions vis-a-vis Sine and Cosine Functions... . 327 75.4 Hankel Functions . Chapter 8 Theory of Partial Differential Equations 8.1 Examples of Field Equations in Physics ........... 334 8.2 Theory of Characteristics . . . . . 336 8.2.1 The Characteristic Curve .. . . Raieaeae 339 8.2.2 The One-Dimensional Wave Equation 340 Chapter 9 Heat Conduction 9.1 Fundamental Equations . . . . .. 344 9.2 InfiniteMedium ........ . . 345 9.3 Semi-Infinite Medium... . . . 351 9.3.1 Age of the Earth 353 9.3.2 Temperature Variation of the Earth's Surtace | 355, Chapter 10 The Eigenvalue Problem 10.1 Eigenvalues and Eigentunctions ......... . 358 10.2 Harmonic Oscillator/Free Particle in a Sphere 10.3 The Variational Principle 10.3.1 Distribution of the Eigenvalues . . 10.3.2 Upper Bound to the Eigenvalues .xi Chapter 11 Wave Equations 11) eee 11.2 Retarded and Advanced D-Functions . . 11.3 Field Due to a Moving Point Charge 11.4 Finite Boundary Medium ..... . 11.5 Green's Function Method Applied to ‘Schrédinger’ 's Equation and to Heat Conduction... ........2.-0 +0055 396 Appendix B. Problems and Solutions Problem SetsA—E 6... 06 ee ee ee 398PARTI Chapter 1 Vector Analysis 1.1 Vector Algebra 1.1.1 Rotation of Coordinate Axes Following a rotation of coordinate axes (Fig. 1-1), the new coordinates xi x4 xf can be expressed most generally as f= Udy + Wake + is%s, Xf = Uard1 + ak + lasts, XS = Usd + Usk + sax, or, in summation form, 3 xt = Yupy. Fi a * fa Fig. 1-1 Since the length of a vector remains = unchanged under a coordinate rotation, ie. |r| = |r’, as ge eae = alu + ud + 18,) + xia + ube + wba) + (uly + uts + ws) + x(t + Uitte + Usits2) + xpxs(Ueriltis + Uailtis + Uailtas + Usiltss) + Xa (Wialtrs + Unallas + Us2llss).2 Vector Analysis Hence, (uh + ud + dh) = 1. dit + taste + Wan) = 0. (ua + we + ube) = 1. (nits + datas + Uailtss) = 0. (uis + ubs + us) = 1. (arattia + Uaattas + Usaltss) = 0. or, in summation form, Myla = de, where summation over iis implied, and where 8=0 for j#k, and B= 1 for j=k. 1.1.2 The Inverse Rotation ‘We had for the coordinate system rotation £ — E’, = Unt + Uke + Misr. = Udy + Ur, + Masts. XS = UgiX1 + 3X2 + Us3%s. 2 @ of = Yuyy. Fl For the inverse coordinate system rotation E’ + E, X= wart + vax7 + visxs. M2 = Wart + Ux? + vars. Xa = Vat! + Vsxz + V33t3. 3 @ am Yow. 1 However, since x] +13 +33 = x1? +9" +29", 3 3 @) Yuya = de, and Yoyoa = Be. ft atMethods in Mathematical Physics | From (1) and (2), we have xi = un(vuxl + vaxz + vsx3) + Ua(var! + Vaaxd + V29%3) + uys(Vsiat + Vsexd + Vast) = xf (uur + W221 + 13031) + 22 (UV12 + W2V22 + Ha3032) + x3 Un013 + Mr2023 + Wg033). Employing a similar procedure for x and x3, we have the result: MyWn * WyWartW3v31 = 1. Un V12 + Ur2V2tuaaVa2 = 1. Us1V13 + Us2VostUs3033 = 1. MyVr2 + Wy2Watt302 = 0. UV + Urad21FWas031 Us1V12 + Us2V22tUs3032 = 0. a0 + Us2W2iFWs3031 = 0. Mv13 + Uv23+W13033 = O. Un 013 + U22V25tll23033 = 0. or, in summation form (summation over i implied): @ ‘Comparing (4) with (3), we see that uj = vy Hence, 3 x= Yay. one A4 Vector Analysis 1.1.3 Definition of a Vector A vector A is defined as the totality of 3 quantities (41, 42, 43) which, under a rotation of axes 3 x= Yaw, A transform as do the coordinates themselves, i.e. (Ai, Az, 43) = (A', Ar’, As’), where 3 At = YuyAy. A Examples of Vectors (1) Position vector r(x1, x2, x3): since the x; are the coordinates themselves. de (dey dea deo dt| dt’ dt’ dt a _ét(én én &xy a | (2) Velocity vector v = r= } since time is invariant toa rotation. (3) Acceleration vector a = ar a? ae” ae? a 1.1.4 Equality of Vectors A= Bif and only if A; = Bj (i = 1,2, 3). If this is true in one coordinate system, then it is true in any other coordinate system, since 3 3 Ai = Duy; = YuyB = Fl ri Hence, to prove the equality of two vectors, one need only show it to be true in any convenient coordinate system. It then follows that all vector equations are invariant to rotations.Methods in Mathematical Physics | 1.1.5 Addition (Subtraction) of Vectors IfA and B are separately vectors, then YutA;+ B) = Yuyd; + YuyB; = Al + Bi, i Fl Fl which defines the transformation of a new vector formed from (A + B): (A; + Bi, Az + Ba, A3 + Bs). It follows from the linearity of the transformation that (A + B) = (B+ A). 1.1.6 Geometrical Interpretation of Vectors (a) A vector A:(A1, 42, As) can be represented by a line segment and a direction, Define r = rp — rg, where rp and rg are coordinate vectors of points P and Q, respectively (Fig. 1-2). Define A, = Xp Xe, Az ™ Ye-Yos As = Zp-ZQ, Fig. 1-2 so that rpg = A:(A1, Az, A3), where the set (A1, A2, As) transforms as a vector. Hence, if a quantity is a vector (Ai, A2, A3), it can be represented by a directed line segment. (b) Converse: if a quantity can be represented by a directed line segment, then it is a vector. 1.1.7 Scalar Product 3 The scalar product is defined as: AB = ) ABi. A Theorem: The scalar product of 2 vectors is a scalar (i. the same in a rotated frame).6 Vector Analysis 3 3 Proof: Under a rotation Ai -+ Ai’, Bi + Bi, and AB -» YAi'Bi . ft a : : A A : YAiBi = YuyAwaBs = LAB: Yuya = YABoe - Y AB. i et Jel il det BL 1.1.8 Example of Scalar Product In Mechanics: F = ma and the rate of doing work equals the rate of change of kinetic energy. 1.1.9 Definition of a Scalar Ascalar is defined to be a quantity which does not change under a rotation of coordinates. : Let B = A, so that B-A = A:A = )/A? = A”, where A is the magnitude of A. fm Example: if A=, r=Vi2+y°+Z, the = magnitude of the position vector r. B Since A:B does not change under a rotation, choose axes as shown in Fig. 1-3. Then A:(A, 0,0), B:(B cos6, B sin®, 0), and /?) A A‘B = AB cos®, where @ = (A, B), the angle between A and B. Fig. 1-3 a Since A-B = AB cos@ in this coordinate system, and since A-B is invariant, ASB = AB cos0 in all coordinate systems. If AB = 0 and both A and B ¥ 0, then @ = 90° and A is perpendicular to B.Methods in Mathematical Physics | 7 (Scalar Product) Theorem Converse: If the quantities (A1, A2, 3) satisfy 3 EAiBi~ a scalar for any arbitrary vector B, then A:(A1, 42, As) is a vector. 3 Proof: Since Bis a vector, Bi = Y'uyBj. A Sas, - Yau ft ~ Yar Sun = Duara» Ea i fl el ae (by hypothesis) by interchanging j and i (i.e. employing the dummy index rule). Hence, pd 3 E [4 - Baile ~ 0. But since Bis an arbitrary vector, Ai= I, uyA/, which z i is the law for the inverse transformation of the components of a vector. 1.1.10 Unit Vectors In the E system, the unit vectors io, jo, ko have the following properties: ig:(1,0,0) —jo:(0,1,0) —_e:(0,0,1) = kek = 1. too = jorbo = Kerio = 0. If we now rotate coordinates to the Z’ system (Fig. 1-4), as a result of which vector components transform as Yun, Fa we can find the components of io',jo', ko’ in the E system. Similarly, from the inverse transformation x 3 a= Yuey, x A we can find the components of i, jo, ko in the L’ system. Fig. 1-48 Vector Analysis 3 In the E’ system, io':(1,0,0), jo':(0,1,0), ko'(0,0,1). Then from xi = ajax, it follows that ee foto = My + U0 + Ws 0 = Un. Go’ fo = Mrz l + U22'0 + Us7x0 = Ur. fo'Kky = Uy3'1 + 230 + U33°0 = U3. Jobo = U0 + Uy] + 30 = uri. Jo'jo = Ui20 + U1 + Uy = Una. Joho = Wyy0 + Uas'l + Us3°0 = Urs. Kolin = yy + ry + usy 1 = Us). Keo'yjo = uir'0 + Uay'0 + us = usr. Koh = M30 + Uay0 + Us3'l = uss. Hence, io! = nike + Wize + Uiako. Jo’ = Uribe + Urajo + Ursko- Keo! = Ustiy + Uszfo + Usako. 3 Using the transformation x{ = Eup, we get = fyig’ = ul + W120 + 130 = un. Goijo! = ay + 220 + 2x0 = tai. Goth! = Us] + Us2°0 + 330 = usr. yy + uy2] + Wy3°0 = Uy2- p10 + ay] + Ug30 = Une. Joke’ = usv0 + usz'l + 330 = Usa. Kori’ = uy 0 + ui + tis] = hs. Koyjo! = 210 + Ur'O + urs" = Uns. Rocko! = Us10 + Us2°0 + Us] = uss.Methods in Mathematical Physics | 9 Hence, fp = uinio’ + Uryjo’ + Usiko'. Jo = Urrio’ + Urafo! + Usako’. Ko = Ursio’ + Uryo' + Ussko'. 1.1.11 Vector Product If A and B are vectors, we define a quantity (AxB):(e1, €2, €3), where £1 = A2B—AsBr. €2 = AsB,-AiBs. 3 = AiBr~ ABs. Problem: (a) Show that AxB is a vector, and (b) Find the magnitude and direction of (AxB). Solution: (@) To solve this, we need merely show that the three new quantities £1", €2', €3' are related to the old quantities 1, €2, €3 by the rule for the transformation of vector components. Take any arbitrary vector C and construct aac - (AxByC = [Ay Az Aa] = Y.Ci(AxB);. By By Bs eu If we can show the determinant to be a scalar, then AxB is a vector. In other words, we have to show that the determinant does not change under a rotation. The determinant, however, is merely the volume of a parallelo- piped formed on the vectors A, B and C, and this volume does not change under the rotation. (b) Choosing the system of coordinates shown in Fig. 1-5: A:(A,0,0). B:(B cos®,B sin®,0). C:(0,0,4B sin®). Fig. 1-5 Hence |(AxB)| = |C| = ABsin@ and A, B and AxB form a right-handed system.10 Vector Analysis 1.2 Examples and Applications 1.2.1 Central Forces (Refer to Fig. 1-6) 7 By Newton's law, F = mr, where HE =r. When F is patallel to r, we have the case of central forces (e.g., electrons in an atom, planets about the sun, etc.). o ‘We now show that in the case of central forces, the momentum vector p = mit = v always lies in a plane. Define a new yector, L = rxp, the angular momentum vector. Then L = rxp + rxp; but rxp=0 and Exp = rx(mr) = rxF = 0 for F parallel to r; hence L=0 or Lisa constant of the motion. But L is perpendicular to p by definition, so that p must lie in a plane about L (Fig. 1-7). Note: Given a problem wherein a particle is under the influence of central forces, it can always be reduced from a three-dimensional one to a two-dimensional one, i.e. its motion lies in a plane. 1.2.2 Infinitesimal Rotation (Refer to Fig. 1-8) Consider a rotation about the z-axis: Attime t= 0, r= 4r(0); at time = 81, r= (61), where r(5t) - (0) is generated by a rotation of z amount Sp about the z-axis. But this is the same 39 as a rotation of coordinate system by an amount 5¢ in the opposite direction (Fig. 1-9). K 0 (81) = x(0) cosde ~ (0) sindp. 7 981) = (0) cosd— + x(0) sindp. 2(61) = 2(0). x Fig 8Methods in Mathematical Physics | n For infinitesimal rotations, sindp = 59; cosdp = 1. Hence, y (61) = x(0) - y(OM(Sp) + O(p)*. Gt) = y(0) + (0X5) + Oo. 261) = 2(0). O(6¢)" denotes terms “of the order of” (6—)*. Let 5f — O and expand about x(0) and (0). 4(6f) = x(0) +30) dr +--+ y(6t) = yO) + (0) Br +++ Comparing the above expressions with those derived from Fig. 1-9, we have: ke [2] ye {2} tro. We define a vector @ having direction parallel to the axis of rotation and having magnitude [| = 8q/6t. In this system, @ = (0, 0, dp/1). Note that r = Or yields: X= i = -@y = - yp). y= by = on = x59). z=0, Fig. 1-9 If now @ is constant in time, what is the acceleration of the particle? od oe ae att = 7 @e) = Oxi + Oxr = Dx(Oxr) - Gr)o- o'r. (The last step involves a vector identity which is proven later — see Problems.) If we now hypothesize that, in addition, r@ = 0, (rotation perpendicular to the plane of motion), then a = - @”r, and is opposite in direction to r.12 Vector Analysis 1.3 Theory of Curves in Space (Refer to Fig. 1-10) Define a unit tangent vector f= v/v. Then z t a= 1-2 (06) = ite vie. But foto = 1. Hence, fo'to = 0 and io is perpendicular to fo. y The acceleration, therefore, has 2 components, So one along the tangent vector fo and one 7 perpendicular to fo. Fig. 1-10 Define a unit normal vector no ae, ° Vel Let s be defined as the length of the curve from some fiducial point so, so that v= 5. Then, since fo=f(), ty lol = vMethods in Mathematical Physics | 13 Physical Meaning of Radius of Curvature R (Figs. 1-11, 1-12, 1-13) Since f and fotdto are unit vectors, [dfol = |old® = d®, If ds is approximated by the arc of a circle, what will be the radius of curvature of the arc, Rare? fotdty dO _ {ato} _ ds, Re ds |ds| R° % Hence the Radius of Curvature is the value of the equivalent radius of a circle having an arc length ds. Fig. I-11 Define the binomial unit vector bo = foXNo. If a particle moves along a path such that it always lies in a plane, Bo will never change direction, being always perpendicular to that plane. If the curve of motion should leave the plane, bo will change direction, and we have torsion. fotdty 7 Fig. 1-12 Define torsion T ee. As the particle moves a distance ds along the curve, by turns through an R i: angle d0', where d0’ = |db|. Hence torsion is the rate of change of the fMg plane with respect to. Za) distance along the curve. Fig. 1-13 Define Curvature (Ist Curvature) Cy Torsion (2nd Curvature) CQ Total Curvature CHatG From the definition of the binomial unit vector Bo, bo = foXto * foXito. But ig is parallel to no; hence bo = foXMto, and is perpendicular to f, and since boxbo = 0, bo is perpendicular to bo. Therefore, bo is perpendicular to fy and to bo, and lies in the mo direction.14 Vector Analysis Problem: Show that C = Solution: My = boXbo. tae. tg ysl, Since dbp is in the no direction, Bont is parallel to by with magnitude Hence, 2° is a vector lying in the Bo, f plane, and : : dbo ds C= C+. 2 or,Chapter 2 Tensor Analysis 2.1 nth Rank Tensor A tensor of rank n, Ti,iz....i, is defined fo be the totality of 3" quantities which, under a rotation of axes, xj uixj, transforms as iT 3 Tijpesi, = Yuu ij Fl (a) For n= 1, we have a Ist-rank tensor, with 3! = 3 components, 7; (i=1, 2, 3). 3 T; = YujT}, identical with a vector. A (b) For n= 0, S’=S, and Sis a scalar. (©) For n= 2, we have a 2nd-rank tensor, with 3” = 9 components, Ty (i,j =1, 2, 3). 3 Ty = Yuan. art Example: Consider x; x;. We know this to transform as: 3 3 (xix)! = xix! = Youare ur = Yuu (ed. cot im Hence, x; 3; is a 2nd-rank tensor.16 Tensor Analysis 2.2 2nd-Rank Isotropic (Invariant) Tensor 1 for i=j Define dy = {f foe t#} 3 3 Yuan Buz 7 Yuan = by, i a which follows from the properties of the rotation transformation. Hence, not only is 34; a good 2nd-rank tensor (by virtue of its transformation properties), but it is an invariant tensor, i.e. its components are the same in any coordinate system. 2.3 Contraction Theorem: If Ti, 3 Yous is an nth rank tensor, then O; yields a tensor of rank (n ~ 2). 3 Proof: Define Ohigenig = Bh iThin-in ini=1 = 3 = Loi Yaivtinbliyyen tiga vaioondn ael jal 3 3 Se ae iat jet 3 Yi tis Bi Thom jal 3 = LB iste ts Tiadooni iat 3 = Ys Jat Hence, Ojyi, i, transforms like a tensor of rank (n ~ 2). i.) MieMethods in Mathematical Physics | 17 2.4 Outer Product Theorem Theorem: If Sjy,...i, is a tensor of rank n and Thy,...j., is a tensor of rank m, then Sisig---igTjja---vme 18 a tensor of rank (m + 1) with 3"*" components. cei jain) = Shyigewin Darin 3 2 [mii lain [Sraeo-aTi a Hence, Siyg---ipTiyasie transforms like a tensor of rank (m + n). By virtue of the above proof: If Ai and B; are components of a vector, then AiB; is a tensor of the 2nd rank and can be written as A;B; = Ty. It is not necessarily true, however, that any Tj can be written as the outer product of two vectors. Consider, for example, the expression involving two good Ist-rank tensors Ai and Bi: 3 3 Louw - Law, A Ai a scalar (tensor of rank 2 - 2 = 0). If Ai, By and C; are good Ist-rank tensors, then Ai,Bi,Ci, = Thani, a 3rd-rank tensor, We now perform a contraction on iz and i3 and on i and ig and compare. 3 3 Dis - YABC, |, (BC). : ; Vitis = LAG, = f & (AB) and we note that the two resulting expressions are quite different.18 Tensor Analysis 2.5 3rd-Rank Isotropic (Invariant) Tensor Define +1 for ijk an even permutation of 1, 2, 3 eye = {-1 for ijk an odd permutation of 1, 2,3 0 forall other cases Theorem: €;x is an invariant tensor of the 3rd rank. Proof: Form 3 Mia Mia Wis Veiamindtineian = |ir Mia Mia = Mi Mia Mi +1 for even permutation of i; i2 is = i(k’) x }-1 for odd permutation of i; iz is 0 inall other cases, i.e. 2 identical indices = Fits thus showing both transformation property and invariance of eyx. 2.6 Examples and Applications Vector (Cross) Product: If A; and B; are 1st-rank tensors, then AiBj is a 2nd-rank tensor and €jA/Byy is a Sth-rank tensor = Tein. Consider the 1st-rank tensor: 3 Yew AB = O:. A From the properties of €jx, 1 = ABs - A3B2, which is identical to the expression for the components of a cross product. Hence, O; = (AxB). The vector (cross) product can therefore be written as 3 (AxB) = Yew ,Be- dehMethods in Mathematical Physics | 19 Scalar Product: 3 3 AB = AB = YdyABy. ff A Oth Rank Tensor: 3 YTnew, im Ss a Oth-rank tensor with 1 component, = Th +Tas1 +Tat2 ~ Tas ~ Tix ~Taar- 2.7 Geometrical Representation of Tensors Problem: Given a tensor of 2nd rank. Construct a vector from it, i.e. represent a tensor geometrically. Solution: (a) Consider first an anti-symmetric tensor of 2nd rank, Ty, having components (in 3 dimensions, only 3 independent components). First form a 4th-rank tensor and then contract: 3 3 YT due = LY Tyen = Si, i Fi thereby showing that there exists a direct correspondence between vectors and 2nd-rank anti-symmetric tensors. Problem: Given a vector Aj. Construct a 2nd-rank tensor. Vein: = Su (= Sy). Hence, from a vector one can construct an anti-symmetric tensor of the 2nd rank. Now we wish to show the symmetry (anti-symmetry) property to be invariant, ie. if Sy=—-()Si, Sy = -C)Sh-20 Tensor Analysis 3 Sy = : A 7 Za Sy = +S_ = Sh = +5}. 2 a Si = YuaSn Su = Se = Sy = ~Si- Eat Su = 0 Si = +A3 Sis = ~ Ar Su = —A3 Sn = 0 Say = + Ay Sn = +A. Sx = -Ai Sy = 0 0 As -A Sy=|-4 0 AL]. A, -Ai 0 3 Given Ty: $1 = YewiTy. et S1 = Tx - Ts = 2 Tes. 2 = Tu Tis = 2 Ts. S3 = Ta- Ta = 2Tr. 2Ts Si = |2 Ta}. 272 (b) Consider a symmetric tensor of 2nd-rank, Ti = Tj. Can we represent it geometrically? Ty will have n + (n® ~ ny2 independent components (6 in 3 dimensions). If we try to form 3 Yew = Si, at then S;=0 [e.g. Sz = (€312T31 + €132Ti3) = €312(Ta1 ~ Ti3) = 0], so that we cannot form a vector from a 2nd-rank symmetric tensor.Methods in Mathematical Physics | 2 Consider instead: Ax’ + By’ +C2 + Dxy + Eyz + Fox = 1, a general quadratic surface with 6 constants. Hence, we can represent a symmetric 2nd-rank tensor by a quadratic surface. Theorem: Any symmetric 2nd-rank tensor Ty can be represented uniquely by a surface : Yin = +1, A where the sign is determined by the sign of the determinant det [Tj . (The sign convention is needed; otherwise, if, for example, Ty = -5i, then LTipunyj = 7 - y - 2 = 1, which does not exist in real space.) Under a rotation, Ty + Ty and ETiay + LTixi xj. The sign, since it is determined by the determinant, itself a scalar, is invariant to a rotation, and the sign convention holds under the transformation. Proof: We must show that the quantity £7j;;x; has proper transformation properties. We note that it is a scalar, hence if = + 1 in one system, it remains = +1 in every system. (We have already shown the sign con- vention to be invariant.) Furthermore, the physical surface itself is invariant toa rotation. , For the case )' x? = 1, what is the symmetric tensor corresponding to this surface? I 3 Evidently, Ty = 8y, |Tilis +, and Ty +0y. For the case ) x? ITid is -, and a Ty= by. Hence: Given a 2nd-rank symmetric tensor, one can find a quadratic surface. Given a quadratic surface, one can find a symmetric tensor of 2nd rank.22 Tensor Analysis 3 Principal Axes: For every surface ), Tixixj = + 1, there exists a rotation such that Ty ~+ Ady, where oa 2 A Yass’ = Yan? = 21, et n where the 2, are the eigenvalues of Ti. The new axes are called the principal axes. Arbitrary 2nd-Rank Tensors: What is the significance of an arbitrary tensor of the 2nd rank, Rij? Define Sy = Y2(Ry- Ri) sothat Sy = ~S5. Sj = 1/2(Ry+ Rj) sothat Sy = +S). Then Ry = Sy + Sj, the sum of a symmetric and an anti-symmetric 2nd- rank tensor. Hence: Any arbitrary 2nd-rank tensor can be represented by a vector (anti- symmetric part) and a quadratic surface (symmetric part). 2.8 Moment of Inertia Tensor For the body ilustrated in Fig. 2-1, angular momentum Lis defined as: L = fex(pvar). Substituting v = @xr, L = focrxaxnyar = fp [rs - non fr. In tensor notation, Le Solr 1 Oi Fig. 2-1Methods in Mathematical Physics | Define the symmetric, 2nd rank tensor Jj: ty = flPs-reifc so that 3 Vie, = Jo[Por- nox fer = Ls. A Rewriting, L = Lhe; ia Note that, in general, L is not in the same direction as @. For principal axes, however, Li = Aico. If no external torques act, then L = 0 and L is fixed in space while © precesses about it. Physical Significance of ly: If we define the moment of inertia of a body about an axis (Fig. 2-2) as T= foa'ar, where d is the perpendicular distance to the axis, and if ¢; (i=1, 2, 3) be unit vectors defined by their direction cosines, then a T= Yee, 7 which is a scalar invariant. Hence we can rotate to any desired coordinate system. To verify the result, Yee = Lee fol 61 ref. Choosing the z-axis along e:(0, 0, 1), we have T= exeslys = Sear - 2] = fodr. Fig. 2-2Chapter 3 Fields 3.1. Tensor Field 3.1.1 Definition of Tensor Field A tensor field of the nth rank, Ti, is the totality of 3" functions of space (iiz..in= 1,2,3), which for any given point in space (x12x3) constitutes an nth-rank tensor. Examples: n=0 => scalar field, one component, e.g., 9(r). n=1 = vector field, 3' = 3 components, e.g., E(r). Theorem: (Fundamental Theorem in Tensor Field Analysis) If Tips is a tensor field of the nth rank, then Zz Tri field of rank (n + 1). : i, defines a tensor Proof: a y a (a Boni] "oy Under a rotation of coordinates, 3 3 ax, xf = Yup; ot x= Yuga! = ay 7 A A 7Methods in Mathematical Physics | 25 = Yona tsotty Tew ot Ot which is the law for the transformation of a tensor of rank (n + 1). 3.1.2 Operators of a Tensor Field (@) Vp= gradient» = Vo = a a Ist-rank tensor, hence a vector. () V-A= divergence of A =« VA = r&, aecale Lx; (© VxA=curlof A = (VxA) = Yep 2, a ist-rank tensor, hence a Ox; vector. ik 3 @ Ag= Vp = Laplacian of p ~ Vp = os, ascalar. ro (© SA-V?A=Laplacian of A = [VA], = ym, a Ist-rank tensor, hence a vector. Vector Identities (1) diveurl=0 = —V-(VxA)=0 forany A. a OAL) _ CAL Yan [Zen . Yew axaxy - Ye Si anax, ou.26 Fields (2) curl grad =0 = Vx(Vp)=0 for any 9. os Ivx(voN], = Leman 3g? ~~ LOH 5e By? (3) Vx(VxA) = V(V-A)- VA. . é OAm| _ [vx(vxA)], = x 05. [Bem $4 | Am : z [ibn - Bnd] 3 oe 4 . nay = ree var [vcv-a)- VA] . (See Problems for the last step.) Tensor Fields in Electrostatics Scalar Field (Electrostatic potential g): In 3-dimensional space, @ = const. defines lines of constant electrostatic potential, i.e. equipotentials. @) For a point charge, p = ¢/r, p = const. gives equi- potential spherical surfaces about the source e (Fig. 3-1). Vector Field (Electrostatic field E): A line of force is that line such that E is tangent to the line at that point (Fig. 3-2). This means that E = ads, or = adn. , = ody. = adry. Fig. 3-2Methods in Mathematical Physics | 27 Thus the analytical expression of a line of force is given by where ds is a line element:(dn1, dr2, dr3). Lines of force are arbitrarily drawn such that the magnitude of E equals the number of lines per unit area perpendicular to E (Fig. 3-3). If E = -V@, what is the relationship between the equipotential surfaces and the lines of force Fig. 3-3 at the surface @ = constant? ca If dr is on the equipotential eae borer surface, then ier : dp = Vode = Yee an = -Edr = so that E is perpendicular to dr. IH Hence, at the surface @ = const., the lines of force are perpendicular to the equipotential surface. 2 Fig. 3-4 maps the field lines and the equipotential surfaces for the case of two equal and oppositely charged point charges. Fig, 3-4 field lines 3.2 Gauss’ Theorem Let Vin Fig, 3-5 include only the shaded volume. Theorem: a XY. Se, Tae where dS:(d51, dS2, dS3) is parallel to the outward drawn normal and has magnitude equal to |S}. Fig. 3-5Proof: Consider the infinitesimal cubical volume element dxydradz3 shown in Fig. 3-6. Yen Toone = fo Tapoa de + f = STitenia deadss + f|Tas = Stiegl Tieng ats + f(t) Paral Jd +f-n. HPP AS, + [Pa Sa + [Trees ds = YS Tiens aS. 7 i + Ti] ) dere (EHD F ‘We have thus shown Gauss’ theorem G, c to be true for the particular volume element dr illustrated in Fig. 3-6. Fi 1 If we now condider volume Vto (~dradz3|0,0) 1 be divided up into small cubical , volume elements, the surface integrations of neighboring cubes will cancel each other out, thus leaving us with a surface integration over the outer surface. Fig. 3-6 (dxaflxs,0,0) eS | _Jp YA Special Cases: : : OT; e e (@ WTisa vector ela 7: DIZ ar = frre ~ fra = fres. Hence, Sra - Seas.Methods in Mathematical Physics | 29 (b) If Ty is a 2nd-rank tensor, i dr = frydss. 7 5 How do we use Gauss’ theorem for the case f If we define Ryjyigmig ® ByTit - Lean, ee Ria Then, applying Gauss’ theotem: a -fy2 : Sg avnate = [ES Raw PSP = LfoaTine nid: = PT nd}. Hence, for i # iji2....in, nos 7 Jan™ fr . 3.2.1 Applications in Physics ds;. (@) Equation of Continuity: Consider a fluid of density p and velocity v, and consider a volume V, fixed in space, not moving with the fluid. (p and v are evaluated at a particular point in space.) Ll) At any instant the flux of fluid passing outward through an element of surface dS (Fig. 3-7) is that contained in an oblique cylinder of height v and base dS. The mass then will be given by pdt = pv-dS. Hence, the mass of fluid flowing outward per unit time through the entire closed surface is given by fpves, which, by Gauss" theorem, must equal [V-pvdr. y a Fig. 3-730 Fields But by the law of conservation of mass, the time rate of change of mass within V, which is expressed by a = [2 Sfee Sf — must equal the negative of the outgoing flux, i.e. 82 a = - fy. J aa Sr pvr. [For a positive outward flux, there must be a decrease in mass.] Since the above holds for all volume elements dr, eS ot Vv = 0 at each point of the fluid. This is the equation of continuity, and is seen to be a consequence of the law of conservation of mass, and holds whether or not the fluid is compressible. Fluid incompressibility is expressed by ae 7 & +Vpv=0. But, since V-py = pV-v+Vpw, Ba vipv - 2 vows pv -0 = Vv-0, the condition of incompressibility. If the fluid density is uniform throughout, p is independent of position, ie. Vp ~0, while if it is incompressible, V-v = 0, so that Sovas Srevas = flovv+ vpvidr = 0, 5 y Thus, Jovas = 0, : which says that the net outward flux is zero, of the mass leaving S per unit time equals the mass entering per unit time.Methods in Mathematical Physics | 31 (b) Magnetostatics: Theorem: If V-H = 0 ina region R, then lines of force of H are continuous inR. Proof: Consider a tube formed of lines of force parallel to H with end surfaces 51 and S2 perpendicular to H (Fig 3-8). Was = fitas = firds + [itas = fv-Har - 0, rs - Jes - fess -f by hypothesis. H Hence, the number of lines of force through Ba S Si (negative) plus the number of lines of i force through S2 (positive) = -Ni + N2 = 0. Since the volume can be chosen as small as Fig. 3-8 desired, still remaining in region R, the lines of force must be continuous, i.e. lines of force must either form closed curves or extend from -°° to +o, Thus, if a line of force is created or annihilated at a point, a source exists at that point where lines of field fail to remain continuous, i.e. V-A = 0 there. As an example: V-E = 4p, anda charge exists at the discontinuity. (c) Stress Tensor: as Consider a deformable body immersed in a ye medium such as a gas, liquid ot solid. What are the forces exerted on the body? In general, there will be two kinds of force acting on the body in V Gig. 3-9). (a) Body force: frac — electrical, gravitational — Fig. 3-9 v some function of charge density, mass density, etc., expressed by action at a distance. (b) Surface force: that force exerted across the surface of the medium, i.e. some function of the particular medium expressed by an action through the medium, We seek the form that such a force takes.Let fi = the force exerted on dS by the exterior of V upon the interior of V. For small areas, most generally, fi- Yas, where we neglect terms of order dS}, and where f= 0 if dS = 0, ie. a linear combination. Since fi and dS; are vectors, Ty is a 2nd-rank tensor, called the stress tensor. The total force is given by Soaae ~ fra + {nas = frac afte where the last term follows from Gauss’s theorem. Since the above must be true for any volume element dr, a Ft+Y 7, = pa, 7 y a TY 7 Pm yielding the equation of motion of deformable bodies. Physical Interpretation of Ty: Si = TydS; + Ty2dSz + TrrdSs. Si = TadS, + TrzdSz + TrdS3. Ss = TsdS, + T2dSz + TadS3. If we choose dS parallel to the x-axis (dS = iedS), then f= TudS. fy = TudS. f = Tnd8. In this case, Ti1 represents tension or compression (force in the direction of dS) while Tz1 and Ts, represent shearing forces. From considerations of rotational stability of the medium, it can be shown that Ty = Tj if we wish to avoid angular accelerations. Hence, Tj is a 2nd- rank symmetric tensor, and one can therefore diagonalize it to Tj = Ady.Methods in Mathematical Physics | 33 For an element of area dS parallel to the x-axis, r= dS.) a> = tension a ‘<0 = compression Once the principal axes are found, the stress tensor represents a tension or compression along these principal axes. For any other axes, Ty represents shearing forces. Example: For an inviscid fluid (no viscosity), Ty = ~p8y, which represents a pressure regardless of choice of axes. Bo, - = -[vpl,, ox so that the on of motion becomes pa; = Fi- - > pa = F-Vp. Let v(r, 1) represent the velocity of a given particle. What is the relation between a and v? Is it a = v? The answer is no, since a = v applies toa fixed point in space. Here, we consider a point in the medium and seek the velocity at that point at a particular instant in time. We have r at time f= f and r + vor at time f= fo + dt. Then abt = v(r+vbt, f+5t) - v(r, fo) - we [ |e “alee +, a ov ov + Mone Mans ant....-vE) Ox, WW, P. wv WwW. SW Ov ar Ot Box aa a+ Poors = Fp Ot (WVVOr. Thus, Euler’s equation, to second order in 6¢ and 6x;, becomes ov = F- a +0 vy] = F-Vp.34 Fields (d Electrostatics: Let o = charge density. For a point charge at the origin, what is o over all space? Forr #0, o(r) = 0,where fo(r)dr = e. all, space We see that 0(4) is a singular function, becoming infinite at the origin, but such that fo(r)dr = e. We introduce the Dirac 5-function, such that all space B(x) = BO)E@) : 8(x) = 0 forx#0 o(r) = ed(r)4 5, 5(x)dx = 1 fora,b>O oe aa ala Consider a function like (I/a/m)e*/ , the Gaussian shown in Fig. 3-10. If we let the width approach zero, while the height approaches infinity in such a manner that the * Fig. 3-10 enclosed area remains equal to 1, we approach a delta-function, Such a delta-function is a continuous function whose width we let approach zero while the area is kept constant and equal to 1. Returning to the problem above: What is the potential and field due to this point charge at points away from the singularity? From Coulomb’s law, é oe e-f = EB=-Ve- Sn. VE=0 = Ap= Vp = VVp = -VE=0.Methods in Mathematical Physics | 35 To evaluate V-E at the origin, where E is not defined, we consider a volume of radius e enclosing the charge, and imagine that V-E is defined over the volume. From Gauss’ law, a ere edS ved = feas - fas = fas - fF fren fees - {ipa - $58 - §5 2 = $2 ~ are. te Remembering that o(r) = e5%r) and that focoae =e, it follows, by ‘comparison: Space VE = 4ned(r). = Ag = -4ned%(r). If we have some Gaussian distribution of charge, and then let this distri- bution shrink to zero in a manner described above, then, since the cl distribution becomes a delta-function, it will be found that V-E = 4ne5°(r). Generalizing: (a) For a discrete distribution (Fig. 3-11), the potential at point P is given by Vp = a R=|-rl. a : (b) Fora continuous distribution p(r’), the = potential at point P is given by r ve) = [JP ar, i O Fig. 3-1 where the limits are such as to include all of p(r’). If p(r’) should include point charges at point a, for example, we introduce the Dirac delta-function, namely, the limiting form of Fig, 3-12, given by 3e-a) = (ee ifx=a where. Joce- one - lifa ©, SS (1+ Ar)redS > Oas r+ 0 by virtue of ™ (Fig. 3-14). ce . funds’ = fa + aryreedQ = fo™(1 + reydQ 3 > = 4ne™(1+Ae) > 4m ase>0. Hence we see that [¥-Vdr = 4n over the region excluding the singularity, whereas had we applied Gauss’ theorem blindly, the result would be zero.Methods in Mathematical Physics | 39 3.3 Stokes’ Theorem Let 5 be a two-dimensional surface bounded by a closed curve s. Then dS is a surface element of S parallel to the normal of S, and ds is a line element of s parallel to the tangent of curve s. Choose directions such that the right- hand screw rule applies (Fig. 3-15). ds Theorem: If A is any vector point function, then amis fads = fivxayas. , 5 (See problems for the case of arbitrary nth rank tensor.) ae Proof: Consider an infinitesimal rectangle. Since the equations are scalar, we can choose a convenient 2 system of coordinates (Fig. 3-16). In this coordinate © ¢ |/<«—B system, dS: (0,0,dxidx2). Y lax aA; 24, i vxa)ds = [2 - A bnde, a Sc on J on ena Soeur a ths, ashe, Fig. 3-16 ~ fArlees ~ fail feet Ae ~ ards - fade fae + foes 4 oe . - Sate * fade + fender + fade - § ‘Acds over the complete rectangle. For any arbitrary surface (not necessarily 2-dimensional), the surface can always be divided into small rectangles (Fig. 3-17). We have s proven Stokes’ theorem for any arbitrary rectangle; hence it will be true for all the rectangles of S. For neighboring rectangles, the line integrals over the common paths cancel and we are left with the line integral of A-ds’ over the zig-zag path shown. Fig. 3-1740 Fields Consider any small section as shown in Fig. 3-18. a ‘We would like to show that as the section gets smaller, de}, + \ds, > Ads’ — JA-ds. J i) : f £ Consider Fidxy + Fodxj , where F:(F\, F2). Note that for small sections, ds:(dxy',dxz). Hence, Fig. 3-18 Fide{ + Fodeg + Feds, so that Sra - Sras. 3.4 Connectivity of Space Definition (1): If any point in a region R can be reached from any other point in R through a continuous motion in R, then R is defined to be a connected region. A connected region has a single closed outer boundary S. Definition (2): If any closed curve in a connected region Ri can be made, through continuous motion and deformation in R1 , to coincide with any other closed curve in Ri , then R, is defined to be a singly connected region of type 1. Any region of type 1 that is not single connected is defined to be multiply connected of type 1. Definition (3): If the outer boundary S of a connected region Rp can be made, through continuous deformation, to shrink to zero, then Ro is a singly connected region of type 2. Any region of type 2 that is not singly connected is defined to multiply connected of type 2. Examples: Sphere in a sphere (Fig. 3-19): We have a connected region between the two concentric spheres. (@) Singly connected of type 1, since curve 1 can be made to coincide with curve 2. (b) Multiply connected of type 2, since the outer boundary S cannot be made to shrink to zero.Methods in Mathematical Physics | a Donut (Fig. 3-20): (@) Singly connected of type 2, since curve 1 cannot, be made to coincide with curve 2. (b) Multiply connected of type 1, since the outer boundary S can be made to shrink to S’, and S’ can be made to shrink to zero. Fig. 3-20 Theorem: If XE = 0 ina singly connected region (type 1 and 2), as ina sphere, then there exists a scalar function @ in R such that E = Vp. . Proof: Define @(r) = fitsds, where o is arbitrarily r ) chosen (Fig. 3-21). Using Stokes" theorem, we show that ¢(r) does not depend on the path but only on r. fe. ds = - fovyas - O by hypothesis, hence fds ~ fe ds and fe) -fe. ds isa function oy Be or) = - fe. ds =» dor) = Eds and E(r) = Vo(r). We could also show this by considering the following: If {fede = g(x), then Fig. 3-21 982) ~ 4), [pers ~ ats) - (0). Fe lreas - $- te) e601 F00. Since ds:(des,dradsy) and E:(E%,£2,E3), fe) = fcryds = Joedn + Bad + Bd). Ban, Ban Bog — B- V9. Oxy42 Fields ‘Suppose now that the region is singly connected type 1, multiply connected type 2, as a sphere in a sphere. We can make this region all singly connected by a proper cut, as shown in Fig. 3-22. Then @ exists within the region R’, which excludes all shaded regions. The question now becomes: what is the behavior of @ across the shaded boundary? If pis continuous across the boundary, then it can be shrunk to zero. Theorem: If S‘is completely enclosed in S, then a function @ exists every- where in R= R’ + ||, where R isa singly connected type 1, multiply con- nected type 2 region. Proof: Consider feds - [vxends - feds + feds. ioe tal a sess, a Since E is finite and continuous, with continuous derivatives in R, as we shrink the shaded region to zero, fE‘ds + 0. Hence, BIAL E-ds = [VxE-ndS = 9.-9., FO where o- feeds and g. = fea. If S‘is completely enclosed in S, we set 5" to be a cap 53 (Fig. 3-23), on which VE = O everywhere. Hence, sera =0=9-@. “s, Therefore, ~. = p- and @ is continuous across the cut.Methods in Mathematical Physics | 43 It is of interest to consider what happens if we choose $3 to cut through S‘, within which VxE # 0 (Fig. 3-24). Consider firas = feas - fowe:. StS; From Gauss’ theorem, fees = fvxeas = +S, (VxE)dt = 0. Hence, {ress + [reas =0. But {reas =0 = fvxEds = 0, in which case @, = p- and again @ is continuous across the cut, and there- fore defined in the entire region surrounding S’. Physically, we can see the above result by considering a region 51, within which VxE 0 and outside of which VxE = 0. Defining j = VxE, 0 outside 5, j+0 within S, Within 51, V-j = 0, and we have already shown that if V-j = 0 ina region, then lines of j are continuous (Fig. 3-25), Then we are considering foxeas - fas. 3 : Although j # 0 within 51, we get negative and positive contributions to j-dS and the surface integral vanishes. We showed this result above using Gauss" theorem. Fig. 3-25 We now wish to restate the above important theorem, together with some of its uses.Theorem: cS (1) If VxE + 0 only within a finite region S’ Ze -— | - . which is singly connected, types 1 and2, | | i then outside the region, you can represent ~~ E-Vo. (2) If the region S’ extends to infinity, we can introduce a cut, extending to infinity, as shown in Figs. 3-26 and 3-27, so that @ is defined everywhere outside of S’, but is discontinuous across the cut. Magnetostatics: If His the magnetic field and j the current density, then VxH = (4njyc = V-(VxH) = (41/0)V'J = 0. V-j=0 implies continuous lines of j, or no monopoles, i.e. current loops. For a region 51, within which j # 0 (ie. VH # 0), we have current loops inside (Fig. 3-28). Outside $1 we can represent H = Vep. Electrostatics: Within S1, VxE # 0, so that outside of $1, E = -V¢. Hydrodynamics: cut Fig. 3-26 Fig. 3-28 If v is the velocity at a point, and Vxv # 0 only ina finite region, then outside this region, v = Ve (“potential flow”). 3.5 Helmholtz Theorem Any vector field F can be represented by a scalar function @ and a vector function A, such that F = -Vp+VxA, where VA - Proof: Let F approach zero faster than r°, i.e. F< oh usr, where n > 0. paMethods in Mathematical Physics | 45 Then G = [FE ae exists and ~ 0 asr +, Using Dirac delta-function properties, where V, denotes derivatives with respect to variable r, VG = St iea na) Fee = ~ fand°-r Fd’ = - anF(). But VxVxG = V(V'G) - VG, so that Fey) ~ - LIV(v.G) - vvxG]. Define SiG of: of gee ae ee An GG = Se Then F(r) = -Vp + VxA and V-A= 7 V-(VxG) = 0. Let us now evaluate V,-G(r). on = Vf Lae eee eG) = vif eae = fre a Rode 1 , Paine -fr Fr) de + [OR ae ei ea : [Beha as +f SRD oe VER) Sen rr irri since the first term on the right side of the equal sign vanishes, by virtue of the behavior of F as r +=,46 Fields Hence, VeF(r’ anp = VG) = [TR ae. Let us now evaluate VxG(r). VxG(n) = freq noe - fea a fro FRB XE (r')dr’ 0 irr But fovsuarar » [Yen ge mae 7 Y fae [Zeme}e = LJ Yeiuads, = fasxua = - fuaxas. 34 $ . Hence ; F ee ie and 4A. = VxG() = ee a. ‘Substituting the expressions for @ and A into F = -Vp + VxA, 2 If we let (V-F) = - Vp = 4np and (¥*E) = VeVxA = 4ni/c, then e- Seo dv’ represents the solution of the scalar form of Poisson's equation and A = pa ja” represents the solution ofthe vector form of a1 present Poisson’s equation.Methods in Mathematical Physics | 47 Corollary (i): If V3p=0 and @ + as r+ then p =0. r Proof: If F = - V@, then V-F = -V"@ = 0 by hypothesis, so that ert Supa 0. Note that V*F = Vo = 0, so that A = aL ee v orl The restriction dee on F, eed us to use Helmholtz’ theorem, is here satisfied, since if@ < ta, then F< Ge, as =, 7 Corollary (ii): If F does not approach 0 as r + ©, but vor - 48 >< then Helmholtz’ theorem still holds. - ? Proof: Define A(r) = afi =r dy. VeA(r) = ae wienae od fe dO) ys A PHD dr ren edie ae = Af HO gg 4 LEVIED gp fae ea But c Wei a = 0 and [+0 asr + $ Hence, V,A(r) = 0, so that VxVxA = V(V-A) - V2A = - V2A. From the definition of A above, V7A = (4n)/e= VXF = YxtF - VxA] = Hence, F- VxA = - Vo and F = -Vp+VxA.Corollary (iii): Consider a sphere of radius R. Define F(®) = FO) for |r| R. Ir-rl Hence, V,A(r) = 0. ‘We have the vector identity, VxVxA = V(V-A) — VA . But since V-A = 0, VxVxA = - V°A = 4nj/c from the definition of A and from Poisson’s equation. Hence, =O inside R an WelF-VxAL = “TU'-H) Lo outside R Then within a singly connected region R, such as a sphere, F - VxA = -V@, and since the region Ris arbitrary, we may make it as large as we please, thus proving the theorem for every point except at infinity, which can be included if |j| -> (1/7"*), which is usually the case.Mathematical Methods in Physics | 49 Corollary (iv): F = - Vp+ VxA’ where V-A’ equals any given function (not necessarily 0). Proof: We know that F can be written as F = -V@ + VxA, where V-A = 0. Let us define x such that Vx = f and construct the vector A’ from the vectors A and Vx. Define A’ = A+ Vx = V-A'=V-A+ Vy =f and VxA’ = VxA. If we choose a Yo such that V7yo = 0, and define a new vector B such that Be A+ Vyo,then VB = VA+Vy. = VA. VxB = VxA + Vx(Vy0) = VxA. Thus we see that the vector potentials A and B are not unique, i.e. F = -Vp+V*B = -Vp + VxA. 3.6 Equivalent Forms of Gauss’ and Stokes’ Theorems 3.6.1 Gauss’ Theorem (refer to problems) In its general form, a J Bx isd é x fea Aude = LJ Lewtuas, In its special form, 8 = iP Tydt = ¥ fru.as.The above reduce to Sree - fess. for T= tensor of Oth rank, i. a scalar. Jrrade - _faxas and frac - -facs, for T= tensor of Ist rank, i.e. a vector. 3.6.2 Stokes’ Theorem (refer to problems) In its general form, Lew a ThiindSi = Lr via ds 0 Lfewas TundS; = YfTis.ds. These reduce to fave 7 Seas, for T= tensor of Oth rank, ie. a scalar. foxaas = fads, $ . for T= tensor of Ist rank, i.e. a vector. 3.7 Maxwell’s Equations 3.7.1 Magnetostatics Define a vector potential 1 fiw a Aw) we aeMathematical Methods in Physics | 51 VxA = Since -1 (e-r) v -- Vir-r'l = - : ir-FI a ord ee yxa = b —- : Ir-rP cs Ampere showed experimentally that the magnetic field Hp (Fig. 3-29) due to current jdt’ is given by r he J ier -r' ig ir cr -r'P Fig. 3-29 Hence,H = VxA = V-H = 0. Note that VxH = Vx(VxA) = V(V-A) - VA, and if V-j=0, then V-A = 0, since 1 i. ha a. - ih ~iSe A a as rr] > Oasr+e, —l pitas es|r-r| Thus, VxH = - VA = 4nj/c, which defines magnetostatics. 0? In general, Vj #0. Consider the Op/at. Thus, the static case What is the significance of V: continuity equation V-pv = -dp/ét or V: implies 2p/at = 0, which in turn implies V-j Let us therefore write VxH = 4nj/c + x; where 7’= 0 when things do not change with time. Thus, V-(VxH) = 0 = (4/c)V-j + V-Z! in other words, va’ -S yy - 4/22 = Laanp) _ LaVE) _ VE ce c iH . Of) cor cor ¢ Hence, we choose %'= E/c, so that VxH = 4nj/c + E/c, where E/c is denoted the displacement current. For constant electric fields, i.e. no time variations, E = 0.52 Fields 3.7.2, Faraday’s Law as Faraday’s law states (Fig. 3-30) that - -12 fy, fea ~~ eG fe. Applying Stokes" theorem to the above, Fig.3-30 1a ic fred +4 Slee - Hence, = Lon VxE = ca’ ‘Summarizing Maxwell's equations for electrodynamics: (1) VE = 4np, the electric field arising from a charge distribution p. (2) V-H = 0, implying the non-existence of magnetic monopoles. @) VxE = -Hyc, arising from Faraday’s law of electromagnetic induction. (4) VxH = 4nj/e+ Fc, arising from Ampere’s law and the equation of continuity. 3.7.3. Wave Equations If V-H = 0 everywhere, then we know (see Problems) that (2a) ~H = VxA, where V-A can equal any function we please. Substituting (2a) into (3), wE--lyxd vefE+tal - ¢ ¢Mathematical Methods in Physics | 53 But we have shown that if the curl of a vector vanishes, then the vector can be expressed as the gradient of a scalar. Hence, Ga) E-=-Vo-Aé. Substituting (3a) into (1), we get (5) -Ap-(W-AYe = np. [If we choose the “Lorentz gauge,” such that V-A = ~@/c, then we have Vp Gc? = -4np] Substituting (2a) and (3a) into (4) we get Vx(VxA) - : -Vp - 7 A+vvasqg ~ 8, and, using the Lorentz gauge again: © aa-ba - -45. ¢ ¢ 2 above wave equations as o OA = 4. Op = -4np. Defining the D' Alembertian operator O? = A - + z, we can rewrite the Consider the homogeneous case (i.e. within a vacuum), for which p = j = 0. ‘The wave equations then become 7A = Q?@ = 0, and Maxwell’s equations become VH =0. VE =0. VxE = -Hfe =~ Vx(VxE) = -VxBc = -Ec? = -V°E. VxH = Efe = Vx(VxH) = VxE/c = -H? = -VH. Thus H and E satisfy the same equations as do A and @, namely: QH=0. GE-0.Thus, ina vacuum the function x = @, Ai, Hi, Ej all satisfy O°y = 0. Let x = fig), where z= 1 - (uo'rYc, and where fis an arbitrary function of z and uo is a unit vector in any direction. %. fe. dix) ax, dz dx dz)" a-ae 36 ay de ¢ K- Fx_ bf oy a a deat dee or de 2 Since Yu? = 1, we have Oy = a 3} = 0, thereby showing 7 ho that x = f(2) satisfies O?y = 0. If x1 and y2 are solutions, then by virtue of the linear operations V and /@r, the sum of 11 and 2 is also a solution. Hence, the most general solution of Oy = 0 is given by 2 Ealt-*} ofa, 3.7.4 Physical Significance of Solution to Wave Equation Consider a particular solution, namely y = const. at t= fo. (Fig. 3-31). Wo Wallet) —— wrvillotbt) ——~ pe ho vewo FieS3t yey(o)Mathematical Methods in Physics | 55 We are implying that 2 = fo ~ (urrc = constant, which are clearly seen to be plane surfaces. Note that along z = const,, dz = 0, and dt = 0, so that uedr = 0 and uo is perpendicular to dr. Let a later time t = f+ 8 be repre- sented by the dotted lines, as shown in Fig. 3-31. Then at this later time (fixed at t= ty + 51), dz=0, ie. the same surface is under consideration at a later time. deus ¢ dz = 0 = dt- = |dr| = cde, where c = phase velocity, i. that with which the planes move. For a sinusoidal function, sin Wao (k-r - of), where @ = clk] (Fig. 3-32). For r parallel to k, ay, Fig. 3-32 w= vo Sr - on). 2-dimensional analysis (Fig. 3-33): a Ata fixed point, things repeat after a time T= (2nYo, and A)\ at a given time, things repeat after a distance 4 = (2nYk, the wavelength. @ = angular velocity of phase. Fig. 3-33 k = wave number (the number of waves in a distance 2n). 3.8 Curvilinear Orthogonal Coordinate Systems A. Spherical Coordinates A point is determined by (r, ©, p) (Fig. 3-34). Ata point P we have the orthogonal system formed by the triad shown in Fig. 3-35. This system is called orthogonal because the triad formed at every point by unit vectors in the direction of increasing r, 8 and @ is orthogonal. This is reflected in the fact that ds? contains no mixed coordinate differentials.56 Fields ds = dP + 7d0? + Psin’Odg’. B. Cylindrical Coordinates A point is determined by (p, z, p) (Fig. 3-36). Ata point P we have the orthogonal system formed by the triad shown in Fig. 3-37. * Fig. 3-36 In this case, ds* = dp” + p*dg” + dz”. In general the transformation from cartesian rectangular to orthogonal curvilinear coordinates is given by x1 x2 x3 = 1 E2 3, where 41 = %i(Er Ea Es)5 2 = x0(Er 2 bs); xs = r0(Es Be ). on ay oe a ee Ox; o dn = Fa eat ce ae Ox, = ne diy = Say + 2 oP ay. In the x-coordinate system, de = Yad = Yanan.Mathematical Methods in Physics | 57 Since we can write dx; = LEE the transformed ds” becomes: a = 8 eae Pencicts where Bn = By In the expression ds” = Y’ gudtjd& one would naturally expect to get cross dk terms such as dE ds. However, if ds* = ' h?dt?, then the E-system defines an orthogonal system, and if dx;dxij # 0, then we do not have an orthogonal system. For spherical coordinates: fer &=-0 B= bmp =e Baz For any set of orthogonal curvilinear coordinates (Fig. 3-38), we have then ds = Ditae - Leuat?, he IndE where a - Ele): Ind) Fig. 3-3858 Fields 3.8.1 Vector Components To enable us to obtain vector components in orthogonal curvilinear coor- dinate systems, we define (Migp = Au ps where e; is a unit vector in the ith direction, i.e. tangent to the ith coor- dinate. Direction i is such that dE; = dE = 0. Gradient (V) voy. - (F]. where s is a distance. But along direction i, dé; = d& = 0 and ds = hiSE:. Hence, 1a ea ere CD) = 7 oe, Spherical Coordinates (VO), = 5, oe - , 138 1_ ob Cy 06 7 OV = oa et 39%" sind Op °" Lo (WD) = Find OpMathematical Methods in Physics | 59 Divergence (V-A) G One could plug in A and V- in the iojoko c B system and write out V-A involving transfor- mations from x to E systems. Instead, we derive V-A applying Gauss’ divergence theorem as follows: D E In general, [V-Adt = A-dS, and let us Indé& A Indby a s Fig. 3-39 consider a small volume 8r (Fig. 3-39). Then, Br = Hdl = blah dad. SAAS )arncy = ~AtlahsdE ad (at 8). JA-dS)u cron = +Arhrhsdéadés (at &1 + dy). The sum of these tems yields ay Ailalo ae. Doing the same for dS2 and dS3 merely involves cyclical permutation of indices. Hence, va = 1b [MAulls) , Osh) ea .déxds, Trnaet oe or — x(a AArhishs) , Ashley "Tahal 8: a ae, |Spherical Coordinates Ay hy hy 1 r = Iuka = Psind. rsind a — 1 [Av?sin®) | a(Aarsin®) aan} vA - sl oC _ 1a), 1 Aosind) , 1 A45) Par —rsin® 30 rand ap * Cylindrical Coordinates mal In =p = Iilahs = p. Ay =1 vA 5 [PG + Sele, EO pl op op oz = 1OApp) , 1 As) , O42p) ep op p op o Laplacian (V*®) Vb = VVO 1 fa a a ~ Fata m (VPdatn)* 3p (Pato) + (Veh. Using expressions previously developed for V&, we have: Spherical Coordinates ia 1 af, 9b), 1 ae s(r ar)" Fea al =o)" Poin’ Og?Mathematical Methods in Physics | 61 Cylindrical Coordinates 1a o® ao - 2 (08 ade. 28 P) pap ar” Curl (VxA) ‘As we did in the case of the divergence, we derive VA applying Stokes’ theorem, Srxaas 7 § ‘Ads, to an infinitesimal surface area S that is parallel to the ith axis. Setting |S|= . | ‘VxA); = lim A-ds. (om = inf Construct a surface so that dS is parallel to axis 1 (Fig. 3-40), and consider Acds. Fig. 3-40 Contribution from AB: AzhzdE, (at Es). Contribution from BC: AshsdE; (at E; + dE). Contribution from CD: ~Azhad®2 (at E3 + dE). Contribution from DA: —AshsdEs (at E2). Hence, faras - [ze an Oh) ols) a. Subutitting "= hgh, (UA). = [oe ata) ~ 3p (Aaa) Employing cyclic interchange, (PxAY: Ga [ae ahd) — 3 (4Spherical Coordinates 1 a... é (¥xA), = False (rsino,) 2a). a E (7xAYy = =p [ay 4) 5p sin) (VxA), = le (7A) 35 4} Cylindrical Coordinates (WA), = 5 [5p 4d ~ 5 (040). ole wa), = [Fap- 3 40). ojo 0 (Vea), = alae (PA,) 54)Chapter 4 Matrix and Vector Algebra in N-Dimensional Space 4.1 Algebra of N-Dimensional Complex Space The generalization of a radial vector, rpg , in Plere...%n) 3-dimensional space is given by Tpg :(1-Y1,42-Y2 - -.-En-Yn) Fig. 4-1), where xiand y; are complex. [Acomplex number z can be written as z = a + bi with a and b real and i? = ~1. ais called the real part and b the imaginary part of z. The complex conjugate of zis 2* = a ~ bi, and when multiplied by z yields z*z = (a - bia + bi)= a" + 20) We define [rpg the square of the distance between, PO) Pand Q, as y Ieeo? = Yor-wre-y 20. Consider the 1-dimensional complex space : (Fig. 4-2), where z = x + iy. Then rpg:(2- z) Qe) and Fig. 4-2 Ie? = @-Z)*@-z) = @-XP +O -yP- But, since this is ds” for a 2-dimensional real space, we see that there is an analogy between a 1-dimensional complex space and a 2-dimensional real space.64 N-Dimensional Space 4.1.1 Rotations in N-Dimensional Space Fora rotation of coordinate axes, wherein a point P remains fixed but its components are referred to a new set of coordinate axes, we have the case P(x1%2...tN) + P'(x1'x2'...N’). As discussed earlier, such a rotation is defined by : Theorem: If under a rotation x/ = Yow the quantity dq? = [rpg = Sor yore yi) is an invariant, then the transformation matrix uy fl satisfies Suytua = 3j, where 5,¢is the familiar Kronecker delta. Uis = said to be a unitary matrix and the transformation is said to be unitary. Proof: Applying the rotation definition to any points P and Q: N N @: = Yury = owt = Lusty. eae Fl N N P: = Yup = xt = Yujtse A mn By hypothesis, x y Yor - xa -y) - L@-y*es-y)- oe fl Dieta’ + y’*y! - Ota! +x/*y)) 7 = Dita + ty - Ova +3"). 7 But Yx'*x' = Yixitx; and Y'y'*y’ = Lyityi, since these represent the 7 7 7 7 squares of distances from the origin to points P and Q in the primed and unprimed systems and, by hypothesis, are equal.Methods in Mathematical Physics | 65 Our expression therefore reduces to: Ny Ny Loves + arty) - Lowa ary fF fa Evaluating the left side: Ny NEN Ny Ypiex’ +xi*y/] = ip Uy" Win Yi Xe + Yuytun xF"Ye a ote ie wy = Yiuytus byt + 3¢yd- oe Evaluating the right side: x x Ybvy + ty = Ldy bya + y*yd- fl Ted Combining: xm a P| [uy*ua] ~ By raat =0. ot |e Ikis from the above expression that we wish to derive our result, namely, Youu = Be. fi Choosing xx = 8: and yj = 8jy, so that x¢* = xx and yj* = yy, (1) becomes: + y @) [Bora ~ Bi N Yloa.*u] = | = 0. Choosing x= iB: and yj= Bj, 80 that xx" = -xx and yj* = yj, (1) becomes: : : ® [mrrOn)] | Yara a66 N-Dimensional Space We note that equations (2) and (3) are of the form A + B= O and A - B=0, so that A = B= 0. We have, therefore, Ny Yuutua,= du, ot where jo and ky are arbitrary. 4.2 Matrix Algebra Definition: M is defined to be a mxn (rowsxcolumns) matrix if M has mxn elements and if these elements are arranged in the form Mu M2 M3 - + + Min AG Mae a Mm Mn2 > °° * Min Definition: If Ais an mxn matrix and B is an nx/ matrix, then the product A’Bis defined to be an mx! matrix, and 2 j=l 2 | (A-B)y = YAnBy, i Since B-A has no meaning unless / = m, it is not always the case that B-A = AB. The usefulness of this matrix notation is easily seen if one considers any linear transformation given by x =). Mjxj. Denote the (m1) matrix ft x, the (nx1) matrix x, and the square matrix My, respectively, by: on! Mt Mu Mn Mis > > Min aa! Boy MuMn + °° + Min Qa) | Oa) we In! Xn Mm Mr * °° * MmMethods in Mathematical Physics | 67 ‘We can then write the transformation as x’ = Mx, a matrix equation which embodies all n equations of the given linear transformation. As in all of algebra, we will have need for certain convenient notations. Therefore, let us introduce some new definitions. Transpose: (A)y = Aj. IfA isa mxn matrix, A is a nxm matrix. Complex Conjugate: (A")y = (Ay)". If A is mxn, A* is mxn also. Adjoint: (A*)y = (Aj)" = (Ay = (@’)y. If A is mxn, A’ is nxm. Matrices that are well defined, though not necessarily commutable, obey the following rules: [1] Associative Law: (A-B)'C = A(B-C). Proof: [(4-B) Cy = Yancy = YAiBaCy = YAdBacy) = YAKBOy = ABO. 7 [2] Transpose of Product: (A°B) = BA. Proof: (AB); = (AB) = YAndu = LOuay = (BA)y. [3] Complex Conjugate of Product: (A-B) = A™-B’. Proof: (4:B)'y = (4B) = Vaiss - Laue = A'B). [4] Adjoint of Product: (A-B)" = B*-A*. Proof: [(4:B)'}y = (4B) Iy = (A'B')y = BAY); = BAY.68 N-Dimensional Space " We saw before that )) Muxk could be written as x’ = Mx. If we take the i transpose of both sides, we have ¥ = XM, where X= (xy x2 my). X= (tx... Mn). Mu: May My + * > Mn MnMn - ** * Mr MinMin * > * * Mm In the future we shall be dealing only with nxn, m1 and 1xn matrices. Henee, let us adopt the following notations: XYABCM will denote arbitrary nxn matrices. 6 will denote particular 2x2 matrices. xyZW@ will denote nx1 (column) matrices. An important additional matrix is the unit matrix I, defined as Iy = By. Then for any matrix A, FA = A',since (FA)a = Yuu = An. 7 (De = YAude = An 7 Furthermore, I= I= 1" = I’. 4.2.1 Commutation of Matrices Definition: If A-B = B‘A, then A commutes with B. Definition: If 0. is a pure number, (aA)y = aAy.Methods in Mathematical Physics | As an example of non-commutation, consider the matrices describing electron spin, the so-called Pauli spin matrices, defined as: oft oP a ae Employing matrix multiplication, a0, = (9!) = -ic,. oo. = 10 =f) 0) 0, =|; |= ir 00 = 00, = (; H| = ios. 0,0, = It now becomes convenient to define: (M+ Ny = My+Ny. (aM)y = aMy. (x+y) = mat Ya, where a. is a pure number. Then, from the 6 relations above, we have 0,0,+ 0,6, = 0 = 010; + 0/0; = 2y, where i,j > x,y, 2. Note:70 N-Dimensional Space 4.2.2. Matrix Inverse Does such a thing as a matrix inverse exist? Since this is very important, we shall need to discuss some major theorems in algebra before we can answer the above question. x a x Theorem (1): If Y Ayx;= 0, and if (x) =| - |¥ 0, det|A| = 0, ie. the only Fl : solution for 4; is det |A] = 0. - Theorem (2): If det|A| #0, and if Ax = 0, then x = 0. Theorem (3): If det|A| +0, and if AB = 0, then B= 0. Proof: Consider )/AyBjx = 0, and fix a particular (). Then Buy is identi- cal with a column Matrix x, i.e. Bu Bx Buy =| * | = a. Bnk But, from Theorem (2), x= 0, and since this holds for all k, it holds for all By. Hence, B = 0. Theorem (4): If det|A| #0, and ify # 0, then there exists an x # 0 such that Ax=y. Theorem (5): If det |A| # 0, there exists an A” such that A"'A = AA! = I. Proof: From Theorem (4), there exists an x/ (nx1 column matrix), such that 0 0 Ay = ; , where the jth element is 1. 0Methods in Mathematical Physics | 1 In other words, 1 0 0 0 1 0 0 0 0 Ax'=|0|, A? =]o|, Ax =/0}. : : 1 0 0 0 Define Moxtxts + at i100 o cA oe oo] 010---0 a So 1001] ole geek Oo00---1 Consider the expression AA”! = J. Multiply both sides of the equation on the right by A, obtaining AA'A = IA =A = A(AT-D = 0. But det|4] #0 = A'A=J. We have thereby shown that if det|4| 0, A'A=AA™'=1. If on the other hand, det|A| =0, we cannot, in general, find the inverse of A. 4.2.3 Special Matrices ‘When dealing with n-dimensional space, we considered the transformation (rotation) x/ = Yuipxj, where Yx/"x/ = Yixixi, ie. ds? is an invariant. mi 5 7 From this we proceeded to show that Yujue = de = UU=r. 7 We ask the following: Does it follow from the above that UU" = I, ie. is it true that U* = U'? We must then ask ourselves whether or not det|U] = 0.72 N-Dimensional Space Theorem: If U*U = I, then det{U| = e®, where @ is real, and UU" = U'U=1. Proof: We first employ some properties of determinants. Lemma (1) det{A| = det|A|. Lemma (2) det|A’| = det|A’|. = det|A| = [det/Af’. Lemma (3) det|AB| = det|A|-det|B}. We proceed to prove Lemma (3), (1) and (2) being fairly evident. det|A| = Levu, Aas Aa where iia, mand (-1)"is +1/-1 if i1i2,...-5in is an evenfodd permutation of 1,2, wn. Then det{A|-det[B| = Y(-1)""Ay, Aas... Ani, Bu, Bay. - Brie a Let us introduce the following symbolic permutation notation: 123 | (eee es 2 oe my it cal iy inis+ ++ in || jijris + jn iy in iss+ in} [Rika kgs ++ kn |” I 0 where the k; are defined from the ji, ie. there is a direct one-to-one correspondence between forms I and II. 1 2 3-+m)_ (123+ 10 101 76---- 7. det|Al-det|B] = Y'(-1)" Au, Bip, Ani, Bis, r where (-1)"* = (-1)"*, so that P(+)and P{-) + P-). PA+) and P{+) + P+). PA+)and P{-) + P(-). PA-) and P(-) > Pi(+). Thus, det|AB] = det/A|-det|B}.Methods in Mathematical Physics | 73 Hence, from det|U* U|= det = 1, it follows that det|U"U| = det|U|det|U] = (det|Up" det|U] = (det|U)* = Hence, the most general form of det|U] is det|U|= e® #0. Definition: If UU* = U"U = I, then Uis a unitary matrix. Theorem: If x’ = Ux and Uis unitary, then x’*x’ = x*x. Proof: We have already shown that if U is unitary (i.e. Yujuie = 8), then length? is invariant, which is to say that Xx" i = Yan. Employing mattix algebra, = Ux = xx = x'U'Ux = x(UUx = xx. When you insist that length? remain unchanged, ie. )yujuit = Be, you are i in essence specifying a unitary transformation, x/' =)” uyjxj. This is what is done in Quantum Mechanics, where length is defined by (x°x)””. In the study of Special Relativity, however, we define length by (x), and we demand that Zx remain unchanged, i.e. ). xx’ = Y xin. Theorem: If x’ = Vx and if xx’ = xx, then WW= W=I (ie. V=V"). Proof of Converse: x’ = Vxand W=W=1 = 2x =iVVx=ix. Proof of Theorem: It 2’ = Xx, that is the length of a vector, defined by pq? = (x= y)(x ~ y), remains unchanged, and x’ = Vx, then Xx ie = 0 = We-m%e=0 = XW-Dx=0. ‘We would like to show that VV - I= 0. We will show it to be the case for a specific x, although the relation actually holds for all x. (We are using a specific x to find the value of VV - I.)74 N-Dimensional Space Let x have as its only element the ith element, equal to unity, i.e. let 0 0 0 0 7 1 x=10) = xx = (0010--0)/0} = 1. 0 0 It follows then that (W- Du (W- Da: (W-Dsi x(W - Dx = (0010--0) . = (W-Di = 0. (W- Dwi Hence, x(VV - Dx = (VV - Di = 0 and all diagonal elements of (VV - 1) = 0. If we let x have as its only elements the ithand jth elements, both equal to unity, Le. let 0 (W- Dut W-Dy 0 (W- Dat (W- Dy 1 (W- D+ (W- Dy x=]0] = XW-Dx = (00101-0) : a 0 (W- Dui + (WD [iv - Dit V-D5+ W- Dx + V- Dy) Hence, WV - Dx = (W- Du + (W Dy + (W- Dy + (W- Dy = 0.Methods in Mathematical Physics | 75 But (WW — Du = (W - Dy = 0, so that (W-Dj+(W-Di= 0 = (W-D+(W-D - 0, and since (VW - 1) = (WV - D, it follows that (WV- 1) = 0 or W=I. Since WW = I, we wish to know whether it is true that VV = VV =I, i.e. is it true that V= "2 Multiply by V to the left of both sides of the expression VV = to get VV = Vior (I- V1)V = 0. We see that the only solution for = VV'is for det|| #0. Thus we investigate det|V. Since WV =I, det|7 =det|H| det! = = (det|V| )° = 1. Thus, det] = +1 and W- WW=1, where V= V1. Definition: A matrix Vhaving the property that VV = VV = Tis called an orthogonal matrix. In real space, a unitary transformation becomes identical to an orthogonal transformation. Why? A unitary transformation retains length”, where dyq= Lei ~ yd) @i-y) = @-y)'@- y). But if space is real, (x - y)* = (a= y)" = (= y) and dj = (x = y)(x - y), which is the length? definition for an orthogonal transformation. Furthermore, if Vis real, thenV=V = “= =1. Thus, W= V'V= VV =1, which defines a unitary matrix. Hence we have that a real, orthogonal matrix is a unitary matrix. Dual Role of a Matrix: Consider the transformation y = Ax (Fig. 4-3). (1) The transformation can be considered as a point | P(x) transformation within a given coordinate system. (2) The transformation can also be considered as a 20) coordinate transformation from the x coordinate system to the y coordinate system. Fig. 4-376 N-Dimensional Space In the physical applications we shall be dealing with, we usually require that the length of a vector be kept the same. Therefore, we have to know what we mean by length. We have fundamentally two definitions, one applying to Quantum Mechanics (QM) and the other to Special Relativity (SR). Definition 1(QM): dj = Y(xi-y)'i-y) = dy = @-y)@-y)- Definition IT (SR): djq = (x7 y(x-y). Summary [) Ifx’ = Ux, then U'U = UU" = Tif and only if x’*x’ = x°x. Necessity: xx’ = x°U'Ux = x'x. Sufficiency: See pg. 64, followed by showing that det|U] [] If’ = Vx, then WW = VV = Tif and only if %'x’ = Necessity: x'x’ = xVVx = xx. Sufficiency: See pg. 73, followed by showing that det|V| = 4.3 Examples of Matrices 4.3.1 Orthogonal Matrix Theorem: The most general form of an orthogonal matrix in 2-dimensional space is given by = { cosp sing) 9. ye _ (cosp sing -sing cos sing -cosp |’ where det|V| = +1 (or det|| = 1), and where p = 6 + ix, where ¢ is complex and 0 and y are real.Methods in Mathematical Physics | 7 E Vit Vi = 1. eee wean = (% Mallu Va) po vat +Va¥e = 0. Vay Via}| Vir Vin a A+ Vh = 1. If p = cos 'Vi1, so that Viz = cosp and Viz = sing, then Van, _Vu _ _cosp Va Vir sing” ‘The negative sign can be absorbed by having either V21 = sing, V22 = ~cosp ot V2 = cose, V21 = -sinep. Thus the reason for the two forms, Vand V’ Observe that y = { c08@ sing -sinp cos represents a 2-dimensional real coordinate rotation if @ is real (Fig. 4- [x = Xx i ing. = Ve fA 7 210089 + sing x! = -x sing + x, cosp. Since this represents a rotation, Vv de = Wel Now V’ can be obtained from V by a reflection, ie. V’ = RV = (0 iy where we denote R as the reflection matrix. Note then that x = Re yields 2y' = x1 and x2’ = ~22, ie. a mirror reflection about the x1-axis (Fig. 4-5). ] We speak of det{M = +1 for the “proper” orthogonal matrix, ae 1H and of det|V] = -1 for the “improper” orthogonal matrix. Theorem: tan! (1) If Vi and V3 ate orthogonal, then V;-V2 is orthogonal. Fig. 45 (2) If U; and Up are unitary, then U:-U2 is unitary.78 N-Dimensional Space Proof: (ViVaXVits) = Wis = I. a) (AVN) = VWaVaY, = 1. (U,U2)\(UiU2) = UU; Ur = 1. @ (UUsXUiU2) = UU2U2Ui = T- ‘The physical significance of the above is that if we perform a rotation from x tox followed by another rotation from x’ to x", we can accomplish the same resultant rotation by going directly from x to x” by a different rotation, ive. if x’ = Vix and x” = Vox’, then x” = V2Vix = x” = Vax, and V3 = Voi. 4.3.2. Special Relativity Special Relativity — a subject originally stimulated by a series of experi- ments carried out by Michelson and Morley, in which the velocity of light was found to be the same in all reference systems. Consider the case in Newtonian 2 X2" mechanics where system E’ moves . with velocity v with respect to > . system E (Fig. 4-6). : Since time is considered absolute, : a the Newtonian transformation is bee one 3 a defined by: : ‘ : Fig. 4-6 t= 03 Xl = X25 x3) = Hs. my! = x17 Ute According to Newtonian mechanics, the velocity of a body moving in the £’ system would appear to be increased by an amount v when observed in the E system. The velocity of light c, however, was observed to have remained the same in all coordinate systems; hence the above transformation is not the proper one. We must inquire into what is meant when we say that c, the velocity of light, remains constant.Methods in Mathematical Physics | 79 Consider the equation of a surface, x} + x3 + x3 - c’?’ = 0, which describes wave fronts, or spherical surfaces having radii which increase with time at the rate c. Now if x1? + x2” + x3 - 2°? = xf +33 +24 - 7, then we are implying that the velocity c is the same in the two systems. Hence, when we say that c is the same in both systems, E and ’, we are demanding that the length of a vector in 4-space remains unchanged, i.e. that space and time obey the following: ‘ 4 Ya? = De = Ry! = dan, eel where xq = ict, x4=-c7P, and x4? = -cr?. This implies that the transfor- mation given by x’ = Vx should be an orthogonal transformation, where x x! x af gall ov =|], xs xs ict ict’ Since we would like V to obey x2’ = x2 and x3' = x3, we proceed as follows: Ha! = Vaixs + Vaar2 + Vasts + Vaeta = Var = 1, Var = 0 = Vas = Vas. Xs! = Vay, + Va2x2 + Vaars + Vagty + Vaz = 1, Var = 0 = Vaz = Vag. a = Vit + Vat + Vises + Viets = Vin = Vis = 0, Vi & Via #0. Xa) = Vaixr + Varxn + Vaaxs + Vaarg Vaz = Via = 0, Var & Vig #0. Finally, then, we have: x = Vix + ictVig. xq = ict = Vary + ictVag. In the x1-1 example illustrated in Fig. 4-6, where we do not rotate axes x2 and x3, V constitutes a 2x2 matrix: y= {Yu Mis) _ [ cose sing Va Vas} [-sinp cos)? where det|| = 1 and where @ is complex; i.e. we have here a rotation in 2- dimensional complex space. To show this, note the following: 21’ is real while x4’ is pure imaginary, so that from x1’ = Vuxi + ictVi4 it80 N-Dimensional Space follows that Vii must be real and Via must be pure imaginary, while from x4’ = Vaixi + ictVaa, it follows that Va1 must be pure imaginary and Vag must be teal. Hence, cos@ is real and sing is pure imaginary. [In general, @ = a+ 10, but here, since @ is pure imaginary, @ = i8 (8 real).] Consider the following trigonometric identities: eo op ee ie oe = sinh?0 = cosh@- 1. Observe that for @ = i8, 9 oe ae isinhO, 2i so that cosep is real and sing is pure imaginary. Finally, then, cosh® 0 0 isinhO y-| 9 10 0 0 010 ~isinh© 0 0 cosh® Defining B = tanh®, cosh@ = 1 + sinh’@ = 1 + B’cosh’®, so that 1 cosh@ = sinh® = + and substituting into the expressions for x1‘ and 1’, aa’ = coshOx, + ict(isinh®) = ee = E Fisinhir, + ieifcosh®)] = Gee. The above two equations define the Lorentz Transformation.Methods in Mathematical Physics | 81 [The most general form of the 4-dimensional orthogonal transformation is obtained by a 3-dimensional spatial rotation performed upon V.] Ifx’ = Vx,then x= Va’, where cosh® 0 0 ~isinh@ _| 0 10 0 Fel oot isinhO 0 0 cosh® Hence, the inverse transformation is defined by aa | 27 oe vi-6? Consider a point P(x1'x2'xs'’) fixed in £'. What will be the velocity of point Pas observed in £? (2), - tbe a) 7 Hence, Thus we see that Bc = v is the relative velocity of E’ with respect to E. We rewrite the Lorentz Transformation and its inverse, substituting for B: 1. mcut oo t= vxyc? "Tae Vi-v¥e” xy’ + vr t+ oxic? 1 Ot tO YI-0y¥e* Y1-vye"82 N-Dimensional Space Consequences: (1) Consider two events occurring in E’ a time interval ty’ - 4’ apart, What will be the observed time interval in £? We must evaluate fg ~ t, where x1'(A) = x1(B); 32'(A) = 22'(B); x3'(A) = 23'(B). Then, fit fata Faget ue Example: The apparent increased lifetime of a meson in the lab system. (2) Consider a length L’ = x1'(A) ~ x1'(B) in ¥’ (Fig. 4-7), What will be the observed length L in the Z system? x,’ ~ xg'= L’ in’, and is independent of fy’ and t,’ since the ruler is fixed. When we ask, what is L in £, we must - oo |2 specify that 1, = tg, otherwise any length is possible when the measure- x ment is made, since the bar is moving x A with respect to E. 7 x Att, = tp: : Fig. 4-7 Oe te a Oe Waa Ga ulna! = lela). aay = GTA vy =) Cee a" v1-B ep = ¥1-B (x - 45')- Thus, x4 ~ xp = V1 — ? (x4' ~ xp') S (x4' ~ xp’) and therefore represents a contraction. Example: The electron’s electrostatic field for v = 0 _ yields equipotential surfaces, i.e. spherical surfaces - about the electron. When in motion at very high velocities, the equipotential surfaces appear to be Fig. 4-8 ellipsoidal (Fig. 4-8). Note that the expression x4 - xg = ¥ 1 — B? (x4' — xp’) $ (x,' - xp") could have been obtained from the inverse transformation, i.e. (x4' ~ xp’) = [G4 ~ x9) + v(t — 1) 1 - B?, where 1, = tp.Methods in Mathematical Physics | 4.4 Tensor Analysis in N-Dimensional Space 4.4.1 Tensors and Pseudo Tensors non: A {tensor res . Definition: Ds ‘do | of the nth rank, Ty... @ > 1,2,3,...5N) is one which, under an ie transformation x’ = Vx, transforms like ge sant “Vii Waitt Notes: (1) If det|H| = +1, there is no distinction between psudo tensor and tensor. (2) The set formed of coordinates xj, Xi, ... X}, is a tensor of the nth rank, since each coordinate transforms like a tensor. (3) dy is an invariant tensor of the 2nd rank, since ¥ Evavad - Lavin = in, = Bin: (4) ivi aig» Which = Tift. -inis an &¥@ permutation of 1,2,....2 and odd = 0 otherwise, is an invariant pseudo tensor of the nth rank. [Note that must = N for the case of e.] Proof: Consider Vin Via Vis + Vin ViaiVia + * Vin Vistiir.t, = det x Vip Vise Via Via84 N-Dimensional Space But Vi Vir Vas > Vin Vin Vin * + Vin | = Via Vin * + Vian Hence, multiplying through by det|V|, recalling that (det|V))* = +1, we have N det Y Vis Mise Mati diode = Pt Examples in 3-dimensional space (a) If A and B are vectors (tensors of the first rank), then AxB is a pseudo vector, or a pseudo tensor of the first rank. Proof: By definition, (AxBy! = Drew's Be 5 = Ydeti Y ViVinVientinn 2, Viodo, VipBy ik mn ° P = det Y, [> Vino Y, Vag Vea 4 r = det] YSncSwpVitinnoBy = det) Y VicinnAnBn ntl im 0 det(v| vi Yea = det] yy (ABy. Hence (AxB) is a pseudo vector by the law of transformation.Methods in Mathematical Physics | 85 (b) If A is a pseudo vector and B is a good vector, then (AxB) is a good vector. Proof: (AxB)’ = Yew’A/Be ik Laan Dv nVincn Lede), VinB, : —. (det? Y [vata Viti, r Imnop which, as above, reduces to +1) Vit [2 Cima) Thus, TL (AxB)’ = YViCAxBy, 7 satisfying the law of transformation of a good vector. Consider the inversion transformation in 3 dimensions expressed by 21! = -X1; X2' = —x2; x3' = -x3. The inversion transformation matrix Vis given by: -10 0 V=]0-1 0 |; detjy = -1. 00-1 [Note that for the case of 3 dimensions, an inversion is different from a rotation for which det|Vj.= +1, while for the case of 4 dimensions, an inversion is the same as a rotation, i.e. for odd dimensionality, an inversion is different from a rotation. A reflection, which changes the sign of one coordinate only, is always different from a rotation, regardless of the dimensionality.] Under the 3-dimensional inversion above, r + ~r and v —> ~v (derivatives of coordinates must also change sign), Hence r and v, which change sign, behave like good vectors. Therefore, (rxv); = Yeyirjoe > (rxvyi, E which doesn’t change sign because it is the product of two sign-changing components — hence, a pseudo vector.86 N-Dimensional Space Consider the graphical interpretation of an inversion (Fig. 4-9), where we have constructed the vector cross product (rxv) formed of two vectors rand v. We see that the angular momentum vector (rxv) is a pseudo vector, since (rxvy' = (rxv). eae Vectors and pseudo vectors in electromagnetic theory In electromagnetic theory (restricted to 3 dimensions), the following Maxwell equations apply: VE = 4np. V-H = 0. VxE = VxH = We perform an inversion and investigate the behavior of E and H, i.e. see if they are vectors or pseudo vectors. Before we can do this, we must know what happens to j and p under an inversion. We adopt the usual convention that charge remains unchanged under an inversion (non-conjugation of charge), i.e. if positive, it remains positive. ‘We also require that Maxwell’s equations be invariant to all transforma- tions, as is necessary by physical law. We have that j = pv, and since v changes sign, j must change sign; hence, j is a good vector. Furthermore, since @/ax’ —+ ~@/@rxr, the gradient operator, V = O/ax, changes sign and is also a good vector. Thus, under an inversion, Ve + -V- and Vx > -Vx, Since V-E = 4xp and p doesn’t change sign but V- does, E must change sign; hence E is a good vector. Since VxE remains unchanged, H and there- fore H remain unchanged; hence H is a pseudo vector. Thus, if system E undergoes an inversion to system E’, we have the following results: pe i-d {r = Meno vector} fe 4 | =Methods in Mathematical Physics | 87 According to Einstein, all physical laws must be invariant to 4-dimensional orthogonal transformations (i.e. Lorentz transformations). Hence, we must be able to express physical laws as good tensor relations. In 4 dimensions we have x1, x2, x3, x4 ict, where the x, are good vector components, as are the dy , commonly called the 4-vector. According to Newton, F = p, which is a good vector relationship in 3 dimensions. However, it doesn't hold for 4 dimensions, i.e. the velocity vector is no longer a good vector in 4 dimensions. Which vector should we replace it with? Define 4 ds =-Yade = -(dd + dd + dd) + de. mt Under a Lorentz orthogonal transformation V, ds is a scalar and remains invariant; hence oe : (@ & a #4), behaves as a good vector, and i dt dy transforms as( a i LY Note that ds/dt = V-(dxt + dx + dx) + Cdt/dt, so that defining the 3- dimensional velocity vas v1 = deyit; v2 = dufdt; vs = deat, we have v = Vot + 0} + of and ds/dt = Vc? v?. Then ee dat ee B= et aye iene ep aa i ds ds Ve=0 VI-p Henee, dx vr v2 U388 N-Dimensional Space ‘We now introduce the 4-momentum oe Pw lo ae which we see represents good vectors. We have Movi ime | iE moc? Pi = Vege ope ee where the p; are the generalized momenta of the particle and ps is the generalized energy of the particle. For small velocities (f << 1), the momenta behave like 3-dimensional momenta, since Pr = MoV} + ..5 Pz = MoV2 + ..., P3= MoV3, while pas i{mce + mv'/2c) = iE/e = E=mge?+ mv 2+... For a system at rest, v = 0, p1 = p2 = p3=0, Eo = moc” (rest energy), while for a system moving with velocity v > 0, p > mov, E> Ep. Let us define Mo, = where mt is the test mass and m the moving mass. Then, 5 2 Pi = mvz, P2 = Mv2, ps = mvs, ps = imc, E = mc’. 4.4.2. Similarity Transformations Consider a linear transformation in £ which transforms a point P(x) into another point Q(y) by the transformation A, ie. y= Ax. Now consider a transformation S which transforms into Z’, such that y’ = Sy and x’ = Sx. We ask! What transformation A’ in Z’ will accomplish the same transfor- mation in E’ as A did in £, i.e. we seek an A’ such that PG’) -> Q”), where yaa.Methods in Mathematical Physics | 89 Theorem: A’ = SAS" defines a similarity transformation. Proof: If y’ = A’x’, fy e 3 = Sy=A'Sx = S'Sy = SA'Sx. Hence, y = S1A’Sx. But y=Ax = Ax=S7A’Sx =» A=S7A'S. Therefore, multiplying on the left by S and on the right S!, we have: SAS" = SS'A’SS" = A’, If Sis a unitary transformation, it means that we have used a unitary matrix 5S (S* = S"'). If we use an orthogonal matrix S (5 = S~), then the similarity transformation is called an orthogonal transformation. Example in 2-dimensional real space (Fig. 4-10) Let A be a rotation in E of 90° and ler S be a \@ reflection from E to Z’. Find A’. a | Pe @) Under A: x» y. UnderS:x—x'andy>y. ,\A r Under A’: x’ + y’ Under A: a Fig. 4-10 yi = -rsina, = -x2. a eae: ye = reosa = x1. 1 oy Reflecting about the @ axis: fe) (0) a oe |) ee ee bie Vie ee i a iS ie = ye! i90 N-Dimensional Space Definition: If H = H’ (i.e. Hy = Hj), then H is a Hermitian matrix. [Note: Every real symmetric matrix is Hermitian]. Theorem: If Visa unitary matrix and H is Hermitian, then UHU™ is also Hermitian. Proof: (UHU"Y = (UHU')' = (U')'H'U' = UHU' = UHU". We have thereby shown that the Hermitian property is retained for unitary similarity transformations. [Note: since we have shown that the product of unitary matrices yields another unitary matrix, then the unitary property is retained under a unitary similarity transformation.] 4.5 Matrices in N-dimensional Space 4.5.1 Diagonalization of a Matrix Theorem: If A is either a Hermitian matrix or a unitary matrix, then there exists a unitary matrix U, such that m0 0 0 0m 0 . iy Se Ae een 900 [We have already dealt with diagonalizations for real symmetric tensors Ty in 3 dimensions, which are merely special cases of a Hermitian matrix.] Proof (i): For A Hermitian, i.e. A = A’. Consider the characteristic equation Ay = Ay, where is an nx1 matrix, an eigenvector of A, and where 2 is a number, the eigenvalue of A (at this stage not necessarily real).Methods in Mathematical Physics | 91 If the above is true, then we have that (A - ADy = 0, the only solution of which, for y #0, is given by Au-A An As + + * Aw An An-% As °° * Am Oe An = An . fe Pe, an nth order equation in A with solutions 41A2...An, the n eigenvalues of A. Corresponding to each eigenvalue A, there corresponds an eigenvector y'. Hence our characteristic equation can now be written as the set of n™ equations qa Ay For each A; we have a set of n equations. Choose, for example, 47 with its corresponding ’. We then have Anyi + Anyi... + Anya = Anyi. Anyi + Anyi +... + Aan = Anyi Ani + Any + ...+ Amn = Bayi From these equations we can solve for y’. Repeating the above procedure for each 2, we build up the nxn matrix wivi vi wvi wi a : Wn We now wish to normalize w’, i.e. choose w! such that (y')*y! = 1. To proceed, we will have need of the following Lemmas:92 N-Dimensional Space Lemma (1): All eigenvalues 4, are real. We multiply the characteristic equation (1) on the left by y** to get witay! = rw"! = A. Since A, is just a number, it follows that Ai =A;. Therefore, [v'av] = [vay] = wia'y! = yiayi = 4. Lemma (2): Two eigenvectors belonging to two different eigenvalues must be orthogonal. Consider: Q) Ay! = Ay, @) Ay = Ay. Taking the adjoint of (2), y'"A* = Ary’. But since A = A’, we have (4) "A = dy". Multiplying (3) by y"* on the left and (4) by w/ on the right: ) way = Ay". ©) y'Ay = Aw"y. Subtracting: MT — i- Aw") = 0. If A: # 2, then yw! = 0, which implies that the two particular eigenvectors are orthogonal. Note that every " = ey! obeys the choice w'*y! = 1. This means that the phase of the eigenvector is variable. [This comes under consideration when dealing with gauge transformations in Quantum Mechanics.] Lemma (3): If i= j, the Schmidt orthogonalization method can be used to find a set of orthogonal vectors from the set of unit vectors . Let A1 = Az = 2 and let the corresponding eigenvectors be g! and g”.Methods in Mathematical Physics | Then Ag! = Aig! = Ag! and Ag” = A2g? = Ag’, where @ + cp!, otherwise @’ and g! would satisfy the same equation and essentially be the same eigenvector. Define y' = g! and choose @!"g' = 1. 2 192 Define y? = [? -(v"'e' et ls , where N? is chosen to normalize y. Note then that Ito! wi'y! = php =i. a [e-(v'e ple -(v'e ee = [9 -e"e'v'][e-v'eo = [set oe" e'e" ee! oto 4 - [oo ole? Wi? = 1, since w* y’ # 0 and N? is chosen to normalize *. For any set of 2; = A, say Ay = Az = , we have a set of corresponding eigenvectors pi P2 93 - First define y' = @1 and choose 9191 = 1. Then define e- (v"'e)' py: 5 where N? is chosen to normalize y*. Then, as shown above, w""w” = 1, while yy" =| gige - (vie2}oio Js =0. Then define v=[2-(v'elv-(w"e)e'Ps, where N3 is chosen to normalize y*. Then, yy’ = 1, while y'*y? = [vies = (v"*es)v"'v? Eas (v'"ea}v"v" hs = 0, and yw ae [v*"es-(v"*es]v*w? - (wi"es)w*w! Ns = 0. we94 N-Dimensional Space To show how the vectors w? and y? were - 2 created geometrically: v ‘We first construct y” to be perpendicular to 1 (or y') (Fig. 4-11). Then since w= Oo A ery 2 - OA and since OA = projection of 2 on Fig. 4-11 91, OA = 92" pi = 1"@p2, we have wv? =[¢2-(o'e2}piv. Returning to our theorem, we have shown that, for any set of eigenvalues, the n eigenvectors can be so chosen that yy! = By. Let us choose U such that vi vi vi vi vi vi ele ‘ We Va Wh It then follows that wove we) [vive vs aw va vivi wa Uy -u-U = elo 6 wi wi Wn wi Wi Wn [Note then that tet ylty? : 10 ° cc Vy _loo 0 es el lee 5 illustrating that U is a unitary matrix.]Methods in Mathematical Physics | 95 To show that UAU" = Ady: divi Awi avi AU’ = ravi ravi wo BEY ana VS a duy"y' day"y? Any" y" My 0 0 0 uy 0 vel e 2 : dn We have already shown that if A is Hermitian and Uis unitary, then UAU* is still Hermitian, thus implying that the eigenvalues A, are real. For A real and Hermitian, since the A; are real, the y' can be anything. We will always choose the yy’ to be real. Proof (ii): For A unitary, i.e. AA* = 1, A” Consider the same characteristic equations as in Proof (i): (8) Ay! = Ay’. @) Ay’ = yy’. Although we can choose the y' such that '*y! = 1, the problem is to show that for i # j, wy! = 6y. This can be done as follows: Take the adjoint of (8) and apply it to (8) on the left: Atay! = wi wii. Thus, 74;=1 = [AP =1 = Ai=e, Oreal. Hence the eigenvalues of a unitary matrix are complex, with magnitude equal to unity.96 N-Dimensional Space (10) Ay! = e® yi. (i) ia’ =e" Applying (11) to the left side of (10): (12) hatay! = &%yty = yty!(e®- 1) = 0. But fori#j,0#@, 4#2j so that y*y'=0. The procedure follows now exactly as for Proof (i). For the case that A is unitary and real, since Ay! = Avy’: If © = 0 (A real), y! may be complex, but can also be real. If 8 #0 (A complex), y! must be complex. 4.5.2 Applications of Matrix Diagonalization We have shown that for every matrix A (whether unitary or Hermitian) we can find a unitary matrix U such that mo 0 Om 0 vaut=A=/9°....°], de where UU" = U'U= Example 1: Diagonalization of a real 2nd rank symmetric tensor Ty. [An example in real 3-dimensional space would be the moment of inertia tensor] Tu Ta +++ Tw Ta Tn + * > Tan Tia Tas * > Tn | Tin Tan © * TanMethods in Mathematical Physics | 97 Since Ty = Tj, T= T* and Tis Hermitian. For the characteristic equation, Ty! = Avy’, all the A are real; hence, we choose y to be real. We then have wivi vi viwi wi - U=|- +... + |isrel = U= (UY = (0) = U = 8, Wh Wa Wn so that wiwi owa|(Arwi wi awl) (4,0 0 vive ova |[awi - : Om 0 Ue Lele vi wi wh |[Aiwa Aewe — Anwn | (0 0 An where UU = UU = 1 and Vis teal. Hence we have shown that by performing a real rotation (U) in real space, any 2nd rank symmetric tensor can be diagonalized. Example 2: Stnall vibrations and normal modes in Mechanics Consider a system of n particles acting through each other by means of a potential in quadratic form. If q1 q2 ... qnare the generalized coordinates of a system of particles and V = Y-vjaig;is the potential, then the equation of motion is given by g mdi = ~Yvygy = 12y0)s i a system of n partial differential equations. m; is a scalar (for a complex system, m would correspond to some inertial property). In cartesian. coordinates, m = mass. The problem of solving these n differential equations is alleviated by noting that we can form a matrix equation from the above if we define x; = Vmiqi. Then VmiVmidgi = “Lo ain - He Lim98 N-Dimensional Space ‘The matrix equation becomes % = ~Tx, where Xi ole v; %=|- | and Ty - pe. 4 ¥mm;, Sn Note first that Tj is real. Since Vis real, let us choose vy to be symmetric. Actually, vy is not uniquely defined, since if one considers V = x1x2, we could have vj = 1 and v2i = 0, of vi2 = Y/2 and vai = 1/2. We choose this latter symmetric case. Thus, Tj is real and symmetric (jj = Tji = Ti), so that Tis Hermitian. Thus, there exists a unitary matrix U such that UTU* = A. Let y = Ux, where yu Yn Then, x= -Tx = Ux=-UT(U'U)x = ¥=-Ay = yi=~-)yi for every i. [Note that since vy doesn’t depend on time, and m; doesn’t depend on time, Ti is independent of time and U is independent of time, so that U = 0, which is what we have implied above.] The differential equations are now much simpler, and their solutions, repre- senting oscillating functions, are the normal modes of vibration y= Ae + Be, where the A; and B; depend on the initial conditions of position and velocity. Using the inverse transformation, x= Uy=U'y = xintermsoft, i : 9 Te 7 Gin terms of t.Methods in Mathematical Physics | 99 Example 3: Rotation in 2-dimensional real space. 7} Q+'y’) ‘The rotation matrix is given by . cos® =) Pon) sin® cos6 which can be seen by realizing that we can achieve x the above rotation by a rotation of axis through an Fig. 4-12 angle -@ (Fig. 4-12). Then since the 2-dimensional real space rotation matrix is given by ar = { 5059 sing -sinp cose |” we merely set (p = -@ to get A above. A is clearly an orthogonal matrix, thus a real unitary matrix (U* = = U"). and unprimed systems, respectively. We seek the eigenvalues and eigenvectors of A. We let z= x + iy and investigate the following (x’ = x cos® ~ y sin® and y’ = xsin® + y cos®): z= (et iy’) = ert iy) = ez. z= iy) = Mai) = CR. In complex notation 2’ = ¢2 represents a rotation of vector z, while z” = e °z' represents a rotation of the complex conjugate of this vector in the opposite direction (Fig. 4-13). Fig. 4-13 ‘We construct the following two eigenvectors from z and z’: aL ee |2e ae fae ae Vz Note that cos0-isin® e® a : 2 ¥2| _ -io| V2 -1 41 = | sind-+icos® ie®|"* | See v2 wT 2100 N-Dimensional Space = al roy om ‘ =| 2) = 2) dg, = ep. ~ie i v2 v2 ‘We see that 1 and (p2 do indeed constitute eigenvectors. [Note also that i = 2.] We now construct the matrix U* from the eigenvectors 1 and 2. fee fe yee oe 8 v2 12 V2 V2 Then, e® e _| v2 v2 aut =| vau' VE VE 18 and the eigenvalues are seen to be e“® and e®, Converse: any matrix B, with eigenvalues (diagonal elements) e~ and e” can be considered to be equivalent to a rotation in 2-dimensional real space. -0 Proof: If B = (“ | then we can find an A (and a U) such that é . ole. 0 _ (cos® -siné) B= vAU' -(8 Pll ae (as id Theorem: Any finite 3-dimensional rotation can be achieved by a single 2-dimensional rotation, if we can find the proper axis. Proof: Let the rotation in 3-dimensional space be specified by a 3x3 real, unitary matrix A. Then we can find a matrix U such that m0 0 UAU* = A=| 0 22 0 |, where det|UAU’| = det|A| = ArAzA3 = +1. 00%Methods in Mathematical Physics | 101 Since A is unitary, its eigenvalues are complex, such that [A?| = 1. A possible choice is 41 =e or e™*. If we let 1 = e® (04 0), then Ay! = ey’, where y' is an eigenvector corrsponding to the eigenvalue Ai = e®, It then follows that Ay” = ey", from which we see that e~” is also an eigenvalue of A. Hence, 42 = e"®, 43 = Land Ay’ = 1-y’, so that 0 00 e000 oe] ° vaU' =|0%20}-/06*0 oa |: 00%) [o o1 0 1 Since A is real, we choose w* to be real, and let it be the z-axis, i.e., But we have previously shown that any matrix with elements e” and e~® is equivalent to a 2-dimensional rotation in real space, so that cos -sin@ 0 UAU' = | sin® cos® 0}, OF OFT where 6 refers to the particular axis about which we are performing the 2-dimensional rotation. Any 4-dimensional rotation in real space can be written as 0 0 -i8 é Oe 0 °, 0 coo ooo e & which is equivalent to two 2-dimensional rotations. The eigenvalues operate in pairs, since if Ay' = ey’, where y’ is the eigen- vector for e®, then Ay” = ey" and e“® is the other eigenvalue.102 N-Dimensional Space For 5 dimensions, the rotation matrix becomes 10000 oe 000 00¢"0 0 0000 00 0 0e* Any rigid body can move from position A to position B by a translation plus a 3-dimensional rotation. We have just seen that we can accomplish this by moving point P to point P’ by a translation, then accomplish the 3-dimen- sional rotation by choosing a proper axis through point P’ and performing the equivalent 2-dimensional rotation. 4.6 Linear Independence and Completeness 4.6.1 Orthonormality and Linear Independence Definition: Let x'x°x° -- - x" represent m nx1 matrices, where m can be less than n, the dimensionality of space. $ j_ a only ifall a= 0 | 1 Yaw! i A x" are defined be linearly (eee. Theorem: If His Hermitian and nxn (H = H*), and its eigenvectors yly? ++ w" obey wi*y! = by, then )'aiy! = 0 will be true only when all a a; = 0, i. these eigenvectors are necessarily independent. 4 Proof: Multiplying )’ aiy! = 0 on the left by w”*, we have a wv Yaw! = Yea =a =0. Since the above holds for any j, we have proved the theorem, namely that orthonormality implies linear independence.Methods in Mathematical Physics | 103 Note that linear independence does not necessarily imply orthonormality. ‘We can, however, construct an orthonormal set from a set of m linearly independent vectors using the Schmidt orthogonalization method, as follows: Let (1 @2 «.. Pm be a linearly independent set. We then set p1°p1 = 1 (Le. we normalize 1), and let yi = pi. We now set wo = (‘p2- wi" p21}, where zis chosen to normalize wa, i.e. y2"w2 = 1. Observe that wr'W2 = (P1'@2 — Wi'gayi'Wi}N2 = 0. We now set ws = (‘ps — w2"@paw2 ~ wi" paws)N3, where N3 is chosen to normalize 3. Observe that Ways = (Wr'@s — Wo" awe ~ Wa'pavn Wi}Ns = 0. Wa" Ws = (2's — Wa" PaWa'Wa ~ Wi" psV2" Wi}Ns = 0. 4.6.2. Completeness Definition: The set of vectors y'y" : - - y" forms a complete set if every w can be expressed as we Yaw, where a; is a number. Theorem: Completeness implies linear independency. Proof: Define a matrix B and a vector a as follows: 1 a a104 N-Dimensional Space Wi = aii + aayi ++ +* + agyi. Wn = GiWat aan +++ * + ayn. If y is given, in order for the a; to exist, det|B] # 0. ‘Assume the set y' (i -+7) is a linear dependent set. Then, Soy! 0, where not all by = 0. fi If we let b then )' bay! = by! + bay? ++ + bay" = 0, or the set of equations, Fl byl + bayi +--+ + bai. = 0 bry} + bayd +--+ + Bayh. = 0 buyn + Bryn +++ * + Baym = 0 But if not all by = 0, then det | = 0, which is contrary to the completeness hypothesis. Hence, all b; = 0, and the set y'y" - - y" constitutes a linear independent set. Furthermore, the a; are uniquely determined by ai = y"*y. Converse: Any set of n linearly independent vectors (n = dimensionality of space) also form a complete set. Summary: Orthonormality (in n dimensions) > linear independence > completeness. ‘Completeness > linear independence (in n dimensions) > orthonormality (which can be constructed by the method of 4.6.1)Chapter 5 Hilbert Space 5.1 Definitions Hilbert space is the space of functions defined on the interval a < x < b. The algebra of Hilbert space is “almost” identical with that of N-dimensional space, but is concerned with functions. We will limit ourselves to functions which are continuous and single-valued. b Normalization of a function fi(x): In Hilbert space, if [fi'fide = 1, then ‘fi(2) is normalized, where the range of x is given by a 0, there exists an N(x) such that NG) @ be ~ Lass) If there exists a maximum finite N for all x in a Sx < b, the series is said to converge uniformly. In general, our approximating series will not satisfy this condition, but will satisfy the weaker condition, that of convergence in the mean: b-a 45 Fig. 5-3 ya and € > 0 (Fig. 5-3). Due to the linear relationship between x and y, if the set Ly y’y*. . . forms a complete set, then the set 1 x.2°x’, . . is also a complete set. Hence, without loss of generality, we will show the set of x’s to be complete within the intervalO O, there exists an no such that for all > no, If) ~ fx - n)| < ©, presupposes that f(x) be bounded, since the above continuity statement says that f(x) cannot be infinite and therefore must be bounded. 5.3.2 Generalization of a Hermitian Operator in Hilbert Space Consider the complete and orthonormal set of functions cos nx/L and sin nmx/L. Can we relate these to eigenfunctions of a “functional” Hermitian operator? We recall, by analogy, that the eigenvectors of a Hermitian operator in vector space were orthonormal. Definition: If », Ofsdx = lire ja] for any fi and fy in Hilbert space, then O is a Hermitian operator in Hilbert space. O may contain integrals, derivatives, functions, etc. In n-dimensional space, if H is Hermitian (i.e. H= H’ — Hy= Hj), then (viva) = Lovitidwad, where yi and y2 are Ge vectors in n-dimensional space. (Wily) = Yevoannwo, . Levineiovo, where we have interchanged dummy indices i and j. Thus, (wily) = Leventhivi 7 Leventtcvir, = (wily). Hence, wiHy2 = (y3H¥1)’ is the expression which becomes generalized into that of a Hermitian operator in Hilbert space. To show that if yiHw2 = (y3Hy1)' for all yi, then H is Hermitian, we choose the yj; such that 2 has only its 4th element (=1) and such that yi has only its 3rd element (=1).Hilbert Space 115, Thus, 0 0 0 0 0 1 we ={1] wi = {0}. Then we would have 0) Hi 0 Hy 0 Hy willy, = (0010..0)| Hy || 1] = (0010..0)] - | = Ha. Fn Similarly, : 0 His 0 Hs . x Hs) (yiHy,)" = (0001..0)] #, |]0] = (0001..0)| . | = Hes. Thus, by hypothesis, H34= Ha3 = H=H’. Let us now consider some familiar operators and check whether they are good Hermitian functional operators. age (aes bee ol fide feat, where fi and f; are chosen to vanish at the limits. Thus, due to the presence of the minus sign, O is not Hermitian.116 Methods in Mathematical Physics | ld @o-7a: 2 4, 3 fede 2 [Lf Lag = lL fee ‘edie - [toda -[hhedia, where again we have chosen fi and fz to vanish at the boundary. Thus, we see that due to the presence of the factor i, O is Hermitian. [In quantum mechanics ne is associated with the momentum of a particle.] (3) Any real function of x, ,(x), is a good Hermitian operator, since 4, b, : Jowyfiae - Licona ; We now proceed to show that the complete and orthonormal set comprising cos a and sin a is connected with a specific Hermitian operator. Let Hilbert space be formed only of continuous periodic functions of period 2L, ie. fle + 2L) = fl), and let us study the algebra of such a space. 2 . d’ Consider the operator 0 = -5: . ac i w at be Since the functions are periodic the first two expressions on the right vanish. Hence, en a : Kripa feszoa| 2 and O = a isa good functional Hermitian operator.Hilbert Space ate: a To evaluate the eigenfunctions of >, we set up the “eigenvector” equation, Off nn es -1, where 2. = -K2 Solutions are easily seen to bef } Since fx + 2L) = f(z), we must have sinks = sink(x + 2L), so that 2kL = 2nn,or k= mL, n= 0,1... Note that a ae are real and negative. We have thus shown that the eigenfunctions of the 2 4 form a complete and orthonormal set. kis real, and therefore k” and -k? are real. Hence, the eigenvalues of | Hermitian operator ae 5.3.3 Fourier Integral Theorem; Fourier Transforms Since my 4 ginny cr L 2 and sin L cos any function that is expressible in a series of sines and cosines can be expressed as a linear combination of e’""” and e""". Thus, yee le fe) us oe For n # no and for periodic functions of period 2L, wL aL flamed = 2 ean I 2s 2L Lila no)VL | , a [er ew] = 0 . Hence,118 Methods in Mathematical Physics | Let us introduce kn = nn/L, so that as An varies in steps of 1, Aky varies in steps of 1/L. Then, . ; jeer | Ak Fe) = Dative (3) where A(ky) = hueman As L + ©, Akn gets smaller and smaller, so that ky practically maps out a continuum. We would then have Ak) = reac and fi) = drone. J x If we define g(k) = FAW, we then have the Fourier Integral Theorem: a Sf) = Fe lewerax . s®) = 7a Sieve as. g(k) is called the Fourier transform of f(x), while f(c) is called the Fourier transform of g(k). In the above treatment, x and k were dimensionless variables running from -©° to +, ‘We want the expression for the Fourier Integral Theorem in 3-dimensional space. Consider a function of 2 variables, flxy), where y is regarded as a parameter. Applying the Fourier Integral Theorem to f(zy) as a function of x, we get: © for) = Fe Jeonre™at. @ — 20k) = rere ar.Hilbert Space 119 Regarding ky as a parameter in g(yk1): ©) a0) = afaaiteat. © Wiki) = efeonne ay. Substituting (3) into (1) and (2) into (4): ©) feo = A [fidence at. (4) hk) = aor linea, By a similar extension to 3 dimensions (3 variables), the Fourier Integral Theorem becomes: Soy) = a fe kikaks Jedd. a(iiks) = qa feore ded. Examples of use of Fourier transforms [1] Dirac 8-function In 1 dimension, B(x - 0) = Fe fewerat = @®= co B(x - xed Thus, 8(¢~ 30) = fe @el* ak. In3 dimensions, 8°(r - ro) = —s.120 Methods in Mathematical Physics | [2] Wave equation: Ap - 422 = 0, where @ = o(rt). ¢ Regarding fas a parameter, (1) = Toreador Peso that iz = Ain, Ag(rt) on™ ede ek, where we note that A operates on f(r), while g (kr) is independent of r. Thus, 8 ter _ ier, ee ike fer pike ae eke = ae kde", so that Ae' Ke**, where KP = |kf = -(2 +6 +). Hence, | 0) 2 Ag(rt) om Bek. fed a 1 fs Jitker), rt) amnlim Pk. a Substituting into the wave equation, Pei ola Setting dt = dxdydz, we perform the following operation: fi. a {fee (Kea + 00} =0, where ko is any fixed vector in k space. We previously had that B(k- ko) = alll (ks), so that our expression above reduces to Sffo'e-K[eb00 + F800} = Keto) +7 let) = 0. Butsince ko was an arbitrary vector, we have that (kt) = -k%"g(kt) = -w"g(ki), where @ = ke.Hilbert Space 121 Solutions of the above equation are easily seen to be g(kt) = Axe" + Bye™™, so that ort) = it Ave + Bye] Pk, Qr wll which we recognize as an expression for plane waves. It represents the most general solution of the wave equation, where Ay and By are any two functions of k. 5.3.4 Piecewise Continuous Functions Definition: A function f(x) is piecewise continuous if it is discontinuous only at a finite number of points, and if the amount of discontinuity is finite. Theorem: A piecewise continuous function can be represented by a Fourier series which converges to the function at all points where the function is continuous and which, at a point of discontinuity, xo, converges to the value ¥72[fes) +f). Proof: Let a function f(x) be piecewise continuous in a manner as shown in Fig. 5-6. At every point of discontinuity we can draw a curve connecting the points of discontinuity ~L i +L ig. 5-7). ae Since a Fourier series can be used for piecewise continuous functions, except at points of discontinuity, we let i | max 543) = $+ Yay cos ME + by sin BE : KO Fig. 5- continuous line represent a continuous function S,(x), which we make approach f(t) as close as we want over the continuous line above. Thus, Sil) > fee) except at the discontinuous point, while at the point of discontinuity, Sa) + 12[fe°) +f).122 Methods in Mathematical Physics | Observe that 1 fhe cos an iy ‘) cos L dx| Li in by = 7 Yeo sin ae are well defined since we can consider the integrals as divided into separate regions of integration. Define a function D(x) = D( + 2L) to be an odd function (Fig. 5-8) [D(-x) = -D()], where ¢1/2 D@) = 2 (L-x for 0 0. FG) =f) forx <0. Expanding F() inthe interval from-Lto+L, -L = x=0 4 Fig. 5-10 FO) = $+ Feu cos Fain : where a nx 1h nx Cn iJ Fe) cos 12+ 7 JC) cos ax and 1 1 Ac] d, = TPO sin Pde +7 4 fee sin a Replacing x by -x in the first integral of d, , we get Jre sin de Fo sin dx) Fes (- sin ae dt) = fre (sin which exactly cancels the second term of dy. Hence, when F(x) is an even function, there remains only a cosine series.126 Methods in Mathematical Physics | But between 0 and L, F(x) = f(), so that in the interval 0 0, ee. —sCse then F(x) is expressed only in terms of a sine series (similar argument to that above), and since F(x) = f(x) in the interval 0 < x < L, then a +L Se) = YBrsin™, m= where 2t nx By = = ffx) sin™ ax 1 L Fig. 5-11 5.3.5 Legendre Polynomials We have already seen (Section 5.2) that the set 1, x, 2”. . . forms a complete but not necessarily orthonormal set. Consider the interval -1 < x < +1 and let us use the Schmidt orthogonalization method to construct an ortho- normal set from the above complete one. To proceed, we define Pol. = fore - fac - EL -L Noting tn = [-] 2, we set vi = VIZ P, = V3Px ie. Py=x. = Severe -0, 1 and Po and P1 are orthogonal within the given range.Hilbert Space 127 Using the Schmidt method, =| {hvele- [broly Since w! = /@2)x, w° = V(I72) and |VG/2)xx7dx = 0, we take g = x’, so that 4 .) ro Nn? 2 ae : es cere ence ia y [? [frvz alma) [2 oh 3 @e-1). a 4 8 Nn WHE Sine fo? tae fox - 62 + nae 8 8? 5 9 2, 1 Hence, we take w= V572P2 = VS2(1/2)(3 ~ 1) = Po = (1/2)(3X - 1). Theorem: The general member of a series of Legendre polynomials is given by [(2n + 1/2]7P,(x) n= 0, 1, 2,...%, where “Is a — 7 nae a i. [Note that Po = 1, Pi =x, P2 = (3x" - 192, ete.] If the above is true, then [(2n + 1)2]P,(x) represents a complete (a linear combination of the x's) and orthonormal set. Proof: Consider the following integral: Hige-ofe- mod = a™ ay re e-] le Ge w& re rte,128 Methods in Mathematical Physics | Fors men m> 2m and gion ro — 1" =0. (b) Forn=m, a. 1)" = (2n)! eHilbert Space: 129 (2Yn) JPs@)Pa(x)de = b-1 mf - 2dr. “1 “1 To evaluate f( -2°)'"dr, first set x = cos®, obtaining: 1 fo ~¥Yde in’ 0(dcos0) = [eos0 sin™0] ~foose d(in"e) 2n cos’6 sin” "8 d0 = - fon sin” '9 do + fan sin”"'6 dO = an fin?” '0 a0- fon sin™ 9 a0, We also have that 4 Ja. -xy"ax = fsine(acos0) = ~ feine0 d= 4 : . : 4 Hence, sin*'9 d0 = 2n sin?” "0 do - fan sin” d0, or ° : 2 ap 19 - 28 Parlp yo - [20 _)(2n-2) 0. ams Odo = 37h fri "0.0 ( |e 0 do, 2n+1 ie, «ang qq - —20(2n- 2) 940 - Ga 1Nan- 1) : 22"n! (2n*D)Qn-1)...3° Hence, | = 5 (2nyi2™ 2-2"! Pracoratoxte bom (nt Gn* Ian- Da 2 * oe aT130 Methods in Mathematical Physics | Consider the following polynomial: ey = DE [op MOD) pn, HHIME-2M0-3) ps Pry tore 2er-1)* * 26@r-1er3)* -}. We would like to show that P,'(x) is identical to P,(x) = Evaluating: Ca Ce rere rfr-1yr-2e"S ee - gles a 2 i J 1 Lew. = (2rM2r-1(2r-2) ... (r# I)! - (2r-22r-3) ... (r=? 4g Mr=A2r~4)(2r-5) ... (r-3)e* 2 _ Mr-1(r-2)(2r-6)(2r-7) «.. (F-S* , 3! = Qn tx’ _ r2r-2) x"? | r{r-1)2r-4yix"* rt (r-2)! 21(r-4)! _ Hr-1fr-2)(2r-6)x"* 31(r-6)! ice (nt P-It? | Pr-1)(r-2(r-3)"* xr Tn BT arar-ty * 2@n(2r-1(2r-2)(2r-3) (2r)! { oe rr-te’? rr-1(r-2(r-3)x"* } 2(2r-1) 2-4(2r-1)(2r-3) ray, rer“ 1Xr-2)(r-3)2"* } 2(2r-1) 2-4(2r-1)(2r-3)Hilbert Space 131 Properties of Legendre polynomials 1] Since the set of Legendre polynomials was defined in the interval 1 $x$ +1, and since cos@ is also defined in this interval, we can express any function of cos® in a series of Legendre polynomials, i.e.: ficos®) = )f,P,(cos®), where Ghrcos0)P,(c0s0)d(cos0) = ¥ frrs(coseyPatcos0)(cos8) 1 wort 2] P,(1) = 1 for any n. Let x= 1+, where e > 0. Then x*- 1=(x+1)(x- 1)=e(2 +e). 1 a" »_ 1 4 aos Pa de ~ age Ot ' flee + oe P(x) = 1 PAL Taking derivatives of the bracketed expression on the right will always involve derivatives of e" multiplied by some factor and derivatives of (2 + ©)" multiplied by € raised to some power. This last set vanishes as e-0. oT ny g ow . Hence, since 2 [e (2 + e) 1. = n!2", we have 1 2"n! ‘We may have occasion to observe a complete and orthonormal set and may want to know whether this set is a set of Legendre polynomials (e.g. the two sets differ only in their normalization factors). We investigate Py(1)- foe the set is a set of Legendre polynomials. } P,(1) = =—n'2" = 1. If P,(1) # 1 the set is not a set of Legendre polynomials.132 Methods in Mathematical Physics | 5.3.6 Multipole Expansion — (example involving the use of Legendre polynomials in electrostatics) Consider a charge distribution p(1) in the region shown (Fig. 5-12). It is desired to find the potential at a point, P, far from and outside this charge distribution. The potential is given by = feo ve- Sqogat for all R > r in the distribution. It is desired to have an expression of the form Vr = ign pots Therefore, consider IR-r = (m+? 2rReoso)” = {1+ 5-2 oso] : ” If we set E= 1/R and x= cos0, then [R - r| = R(1+E?-2Ex) . Using Taylor's theorem, we expand in powers of E about point E = 0: (1+ 2-28)” - LE Zo+e-an)" |. foo a{Slre-2H)"| Eo 1 ~}1st-2t3) "or-2a| = 1+ P26) “ee -29-A(1+ 2-28) “2 eta a(t) |. et ce eractouncoHilbert Space 133 2 (148-28) = (Pac + Pay + P+...) Z : 1 ~ Pn(cos6)r" We strongly believe that =) a7 forall n, where th estongly belive hat et ~ F PACED foal, wher te P,(cos®) are the Legendre polynomials. [Thus far, we have only shown this to be true for n = 0, 1, 2.] If the above is true, then va = Sore), Paseo a . Sar +f ote) dr + f coro» pln) dt+... 0, 9 = 9, 1, a, On Rv Im where 00 = fo(e)a*r, the total charge, 0; = freos6p(r)d*, the dipole moment, and o = f £ (3cos"6 - 1)p(r)d’r, the quadrupole moment, etc. Theorem: 1 = F Peat Bron” Leer > Where the Py Ce) are identically the Legendre polynomials. — 1 RA” -wes-yree-a)” {J “32 ol] ole mex ae [R-r] ox ee | ar (" 2 =F ( "ements = 2 Hens, a) J134 Methods in Mathematical Physics | . -(5-5-4] IR-r la? ay a? J (R- td =0. {} we Pax)" Taking the Laplacian of = + = Yi QO and setting x = cos®, we have R-r sR r a ; A, [Ps(eos0)"] = true for any value of R. We now multiply through by R and then let R +22 (o that only the n ~ 0 term remains). We then get A, [Po(cos®)?°] = Multiplying through by R? and then letting R -+°° (so that only the n= 1 term remains) yields A, [Pi(cos®)r'] = 0. Repeating the process we ultimately obtain A,[P,(cos6)r"] = 0. Consider the Laplacian operator in spherical coordinates: ino 3 foe Psin®® ap?” io = 2 al an Fa Psin® al’ rT) of which only the first two terms operate on Py(cos0)r". im operator A, i.e. with respect to Cartesian coordinates of r. However, we are now going to refer A, [P,(cos6)r"] to spherical coordinates. We are allowed to do this since Ag is a scaler, herice the same in all coordinate systems. Operating on 7”: Observe that A, jg; only referred to Cartesian coordinate Laplacian or") 1 8 a} za <3 (n+ IP = n(n t Wr",Hilbert Space 135 Hence, operating on P,(cos0)r": A, [Pa(cos8)r"] = n(n + 1)r"*P,(cos®) te [sino ea 2} - 0. * sind 20 y iT alma + pscos0) + SI i [sn ing P| =0. If = e080, sin20 = 1 ~ c0s%0 = 1 - a2, dr = ~sindd0, and —2— = - 4. s : vand ned) dx Substituting into [1], we have: n(n + LPr(x) +50 - 2) =0. Rewriting, 2 glo- - 2) Fe) Pion) = n(n + 1)P,(x). Let us define a new operator H = #(0- ash so that equation [2] becomes B) HP,(x) = —n(n + I)P.(x), and the P,(x) are seen to be the eigensolutions of the operator H. We now show that H is intimately connected with the Laplacian operator. In fact, #lo-& }136 Methods in Mathematical Physics | Proof: fee & [0-9 ae - [va- ) ] -fu-9 2 ae =0- -fa- ete ay, [fy - -[fa- al -fa- ye ay, Thus, H is Hermitian. Theorem: If H is Hermitian for the interval a < x < b and if Hy = Ay and Hep = dg, then if A # 0’, Sv pdx = 0, ie. the eigensolutions are orthogonal. Proof: : ; @) — Je'nyar ~ afo'yar. () [fu : Iefollows from (a) and (that (2 — ‘fe 'ydr = 0. If His real, then it follows fom fy" Hydx = af 'ydx that since the left side is real and the integral on the right is real, 1 must be real. Thus, since 2’ is real, 2’ = 2" Hence, from the above, (A - 2 fo’ vax = 0. But by hypothesis, 4 # 2’, so ae = Oand the ee of a Hermitian operator for 1. 2 are orthogonal. [Note the similarity to the properties of a Hermitian matrix in n-dimensional space]Hilbert Space 197 As an example of another Hermitian operator, let us define the following: ie sine ap” toe al), CL aw ene] Theorem: fered - fer yesinsaedy = { fo'-1yWsinedsdg}". Proof: We have already shown (Sect. 5.3.2) that Je Sou «fo, Hence, (-?) is the sum of two Hermitian operators and is therefore itself a Hermitian operator. Summary: a ey > Lene, where the P,(x) are real polynomials in x. @) (b) The P,(x) are the eigensolutions of the Hermitian operator H, i.e. “4 HP.) = ~n(n+ 1)P,(2), such that [P,(2)P_(x)de = 0 for n¥ m. “1 (©) Forx=1, ata: Lea. But a5 57 Le = PI)= 1. (4) The P,(x) are polynomials in x of degree n. (©) The P,(2) are orthogonal sets of polynomials, i.e. J Py(x)Pa(x)dx = 0; their normalizing property is sustained by showing that P,(1) = 1.138 Methods in Mathematical Physics | 5.3.7 Recursion Formulae for Legendre Polynomials Consider a (1+ 8 ~ 26x)” where the P,(x) are the Legendre polynomials. Taking In of both sides, ~ Fin + -269 = 1[Zreve = YP, ‘We now perform two differentiations. Differentiating with respect to E: ; SY paone" z Zama +8 -269) « ES - = —. LPwe Equating coefficients of &", a ye (&Py ~ Pret) = Ye [G+ 1)Pyot + (~ 1)Py-t ~ 2xPy], and, since this must be true for any E ze 1 (m+ 1)Poet + (= 1)Py-t = 2xnPy +xPy ~ Poet u @ (m+ 1)Pai + nPy-1 = x(2n + 1)Py- Differentiating with respect to.x: a ¥ prone ge (PIN +B - 26x) = aE Tee Lrg? ‘We now equate coefficients of "1.Hilbert Space 139 Yer, = Yer! ee + Pi - 2xPy). Since this must be true for any == <1, (Pry + Ppl ~ 2xPy - Py = 0. All other recursion formulae found in the literature are obtained by linear combinations of appropriate derivatives of the above two formulae, I and II. 5.3.8 Spherical Harmonics (another example of a complete and orthonormal set) We have already shown in Section 5.3.3 that on the interval (-n, x) any function of @ can be expressed as fie) - YAne*™*, where the functions e* are related to the eigenfunctions of the Hermitian operator 37/9. We shall see that when we consider functions, F (6), defined on a spherical surface, we will get a series of spherical harmonics which are eigenfunc- tions of the Hermitian operator (-L”), where oe ee Sind 20 [sine x "sine Op?” Note the following: In 1 dim: 1, x, x’... forma complete set in the interval a © = (mc/e)r + E, where Eis a constant vector. Hence, (mc/e)r = rH + d - E, or (e/mc)[rxH + (d - E)}. r154 Problems and Solutions Setting io = (e/mc\d ~ E), we have f = (e/me)(rH) + io. If to is parallel to H and the z-axis is chosen along H, then fo = Vo = (00v.) and H = (00H). Thus, i j= ve, smci-utntent and r= iox + jay + kez, we have dk _eHy dy__eHs | _eHdx _ dy _ ie mee de me" me dt” de? or, é : dy (eH), _ a (Eely = 2 , 7 _ fA sin[(eH/me)t + po] a Se with solutions y eee - Also, 7 Vo Z= Vol + Zo = me dy __{mec) ,(_ eH) (elt We also have that x= - 79 =~ [oy oH isl or ao} resting in x= Asin( 22+ a) y= Aese( +} Z = Vol + Zo. r then is given by the expression = iAsin( j (eHt tS i sin 2 +) +i cos + 99]+ kilos ® Po), which we see is the sum of a constant rotation (first two terms on the right) and a uniform velocity (last term). 2 IV. The Radius of Curvature R is defined as R’ = ee : For the problem above, find expressions for R, Torsion C2 = Total Curvature C, where C= R? + C3. dbo andMathematical Methods in Physics | 155 From iy = v 2, we have that ds constant in time), so that [a _ _ [eHA (elt eHA .. (eHt Be ae cos +l] ile sinline +o] hw. telesales] bit = oF = (ft = jee 2. 42(eH) _ (eH); 2_ 2 - «(2 - (Zee ¥). Thus, e-( e a : abe dbo dbo - ti Torsion C2 = Oe la . We also have that bo fo = 6X no and bo = 1X Mtg + foX Ito. Since Mo ==, tte = =~, so that lol lol = [0X ito (fox tho = 0). a p= (eH/mc)t + @o, we have Stee leh: coc fn om i A cop tie me sing + ko ig = nelid = ~ + [ietyinc) A sing + jCeFY/mcy'A cose]. % = fadbol = 4L- iseH/me) A cose + jdeHH/me)'A sing]. = {tots ~ tina) + jets ~ tuts] + ke{ tiie ~ tah], where 713 = 0.- Problems and Solutions is =o (2) aie ol SE] acon] +1(22) (oe sing conpsin)| : 2 > bo = see, Asing + jo lie) toa ~ (C22: || - BA - (es dby|* _ (eHvo a fal ~(oeel coves eff ; a (0? ~ 08+ ob - (2 sales) ds Total CurvatureMathematical Methods in Physics | 157 Problem Set B I. If Ay Az Ag are the eigenvalues of a symmetrical tensor of 2nd rank, Ty, prove the following: @) Yn - Yn. () yn = YN. © Yate = rN. (©) det fj] = Aidads. (a) If As A2 As are the eigenvalues of a symmetric 2nd rank tensor, Ty, then in a particular reference frame, Ty ~ Tj =|0%0), 00 Ty = YunaayTi, a so that Yunets = Yous - Ye - Yu. at at T r Actually, Ti is a contraction of a 2nd rank tensor, hence a scalar, and 7 therefore a scalar in all coordinate systems and in particular in the system for which )Tix= YA. _ | 0) YT = LY uetertntng TT = YL BinBinTiiTn 7 iT bn tom = LdinBndeDandidn = Londidankn = Yoiaz = Yad. im Hence, ).TjTy = Yaz. 7 m158 Problems and Solutions We have here performed a double contraction on a 4th-rank tensor, yielding a scalar, hence a scalar in the primed system, in which yn Tj = Yay = Da. i (©) Following the procedure above, ), TijTjTu is essentially a triple contraction of a 6th-rank tensor, yielding a scalar. Hence, Lritets = LTTE TH = LAs Arpaddu os tk ik Panbirby YN. (@) Consider ee Yin Tin€inn = | Tr Tr Tpa| » im Ta Tia Tia and . E| Lettie = ¥ | th tp Tale = 6 deti7. _- | Tis Tee Tos Note that for any interchange of jk, the det changes sign, as does eye; hence the product | |-¢ retains its sign and value. Lae the sum E| |e yields 6 terms, each of which is det|Tj]. Thus, det\7;| = 6, Entn a Oth-rank tensor, i.e. a scalar, in every coordinate system. na by the invariant property, det\Tj| = det{Tj] = a1 2223. Il. Prove the following: 2) Sixt = BiBjnBin + BinBjBus + BinBjDim ~ Bind jin ~ B:BjBim ~ BindjnBer- (0) Yeti = Bidin- Binds. :Mathematical Methods in Physics | 159 © Yewem = 28u. vi @ Yenen = 6. m (@) Consider E_pinn = BiBjnBin + DinDnDe + BizDjDem ~ BinDjDen — BByuDie — Biju 3u Bin Bin = |1 8m S|. Bu Bin Sin . For /mn arranged in the order ijk, the product is + 1 (and both sides =+1. 2. For Imn arranged in such an order that an even permutation is required to bring it into the order ijk, the product is + 1 (and both sides = + 1). 3. For in arranged in such an order that an odd permutation is required to bring it into the order ijk, the product is - 1 (and both sides = - 1). The 6 possible combinations of ¢;x€inn cotresponding to the possible arrangements of Imn are: he) Lae ey ere ea hl J ken OidjnBin) (Bind jin) (Gindjxu) — (GiBjnBim) ~ (Gindjn Bus) — (Bind jDen) Noting the determinant above, we see that no two values /mn or ijk may be the same. Hence, only one of the 6 possibilities above can exist for given values of ijk Imn. Hence the sum of the 6 terms yields the desired expression.160 Problems and Solutions (&) Consider Yeincine = BiBjm — Bind. Setting n = k and summing over kin (a) above, Yeweins - Lodndu + Londsbu + Yos6 em = Youbibu - Loidpbm - Loubndu = 35Bin t Bindu + Bind — 38inds — B:B im ~ Bid in = 8dim ~ Sindjr- Eine eRe ' ' +1 ee Note that for any i and j we have 2 possible terms, J @itdjm) — (Gindz2) (c) Contracting expression (b) above with respect to j, i.e. setting m = j and summing over j, we have: Yenen = Yay - Yar = 36u- bu = 26y. i ao Note that for i # J, both sides vanish, while for i = I, we get 2 terms, (eqn)? + (um)? = 2. (4) Contracting expression (c) above with respect to i, ie setting /= i and summing over i, we have Yew = 2Yb, = 23 = 6. ‘e : Note that we have 6 terms to sum over: Yer2sy? + (e1a2)” + (€n13)? + (e231)? + (es12)? + (e320)? = 6. IIL. Prove the following: (a) [Ax(Bxc)] = EeseuBCn. (b) Ax(BxC) = B(A:C) - C(A'B).Mathematical Methods in Physics | 161 (@) [Ax(BxC)] = Yend(BxOy. = Lend PeundCn e = YeyseumsBiCm « kim (b) From (a) above, [Ax(BxC)], = tieetnd Bn = YeiemABiCn fim But since )eyeeink = 5u0jm ~ Sind z [Ax(Bxc)], = LOam = Bind DABICm j = YGAmCnBi- BinAsBiCn) = Y(AmnCmBi- AnBnCi) . a BAC) - C{A'B). Iv. (a) Find the moment of inertia tensor for a cube of side a and total mass m in any cartesian coordinate system with the center of the cube as its origin. The density of the cube is assumed to be uniform. [Hint: Find Iy for the simplest coordinate system] (b) Use the above tensor to find the moment of oe inertia about any axis which passes through the center of the cube, eo) (@) From Fig. A-2, ly = fodt(°by- nv, =. where p= m/a® and P= +y? +2 iy = 3 [Mfo? + Ddxayee : Fig. A-2 Ps SMI lo? + 2] xtyae = 3 So? + Pade 3}? " SJlo*+ eyez - a we162 Problems and Solutions b--5Moawe 50h Is = ~ B [Mfrcaeayae 0. 2 = In = ts = In = hs = In = 0. By symmetry, Ina = SMe D)dxdydz = fs be the system shows no preference for %,y orz axes. Hence, a -™ (b) To find the moment of inertia about an arbitrary axis through the origin, we have I = Lees where ee; are the direction cosines of a unit vector eg 2 2 Iying on the axis. Thus, J = re y= Ye = v. (a) Prove that if 4; is an invariant tensor of the Ist rank, then A; = (b) Prove that if Sy is an invariant anti-symmetric tensor of the 2nd rank, then Sj = 0. (©) Prove that any invariant symmetrical tensor of the 2nd rank must be of the form cby. (A) Prove that the tensor by is also the only form for any invariant tensor of the 2nd rank. (a) by geometrical argument If Aj is a 1st rank invariant tensor, it is a vector whose components do not change under a rotation. The only vector whose components do not change under a rotation is the null vector Aj = (a) by analytic argument ‘The invariant property is expressed as A; = A/, ot Af - Ai= But since A/ = )'uyAj and A; = ),3yAj, it follows that )- (ui - 6))4, = 0. 7 a iMathematical Methods in Physics | 163 Thus we have a series of 3 homogeneous equations in the 3 unknowns 4j: (un — DA + Ur2A2 + ta3d3 = 0. unAy + (u22 ~ 1)A2 + ur3A3 = 0. 131A) + Us2A2 + (Us3 ~ 1)A3 = 0. For a solution of these homogeneous equations such that A; # 0, the det}, = 0, where vy = (uy - 84). But we can always choose the rotation such that (uy 3) ¥ 0. Hence, the only solution of these equations is Aj = 0. (b) by geometrical argument If Sj is an anti-symmetrical invariant tensor of the 2nd rank, it can be represented by a vector, similarly invariant, where An = YewS; — Ax25z, 2513, 25,2). u The only invariant vector is the one whose components vanish, as in (a) above. Hence, Sy = 0 for i #j. However, by the anti-symmetry property, Sy=0. Thus, Sy = 0 for all i,j. (b) by analytic argument By the invariant property, Sy= Sif + Sif - Sy = 0. But we also have that Si = Yuna and Sy = YdudjSu. Thus, Y° (waujs ~ 845;)Su = 0, yielding ‘a ‘a ‘a 9 equations in the 9 unknowns Sj ; i and j are free to be separately 1, 2 and 3. For example, if i= 1 and j= 2, one of the 9 terms would look like the following: Mya Sir + Uy2teyS21 + UrslaSs1 + (Uritear ~ 1)Si2 + Ur2eS2 + UrsldrrSs2 + UritrsS13 + Uyali2sS23 + UisilasS33 = 05 and we get 8 more terms like this. The condition that any number 7 of homogeneous equations in n unknowns yield a non-zero solution for these equations is that the det |Cyx = 0, where Cyjte = inten ~ Bini.164 Problems and Solutions But, since we have performed arbitrary rotations, we can always find a rotation such that det|Cyel # 0. Therefore, the only solution of the homo- geneous equations is for Su = 0. (©\(4) by geometrical argument Any invariant symmetrical tensor of the 2nd rank can be represented by a quadratic surface )) Tipcixj = #1. The only invariant surface is a sphere, x2 +y? + 2? = const, which is represented in tensor notation by the surface Yedyry = i cby, where c is a constant. 1. Hence, the invariant symmetrical tensor must be of the form (c)(@) by analytic argument We already know that any 2nd rank symmetrical tensor can be diagonalized. Hence, Tj; = ‘by can be formed by a rotation from Tj’. The invaraent Property then states that Ty” ~ Tj ~ 0. We would like to show that aNecen, Ty = YuaujTel’, and therefore, since Ty’ = by or Tel’ = 0Bu, we have a 285 = Yunyr!du i V Yuasa! = Yair Ou = Young! — Yaaaa!- a) = 0. 7 ri 7 As before, since the transformation U is arbitrary, this implies that 2! = a/,Mathematical Methods in Physics | 165 Problem Set C L. If p and w are both scalar functions, show that (@) Veetw) = Ver Vy. (b) Vipy) = eVyt wo. @ (Mew = Zloty) = +k = veety) = Vervy. a © (Vow; = SO - yey — vey - oy +yV9. IL. If pis a scalar function and A and B are both vector functions, show that (a) V(A+B) = VA+VB. (vb) Vx(A +B) = VxA + VXB. (©) VPA) = A'Ve+@V-A. (d)_Vx(pA) = ~ AXV@ + pVxA. (©) V(AB) = (A-V)B + (B-V)A + Ax(V%B) + Bx(VxA). (D) V-(AxB) = B(VxA)- A(VxB). Vx(AxB) = A(V-B) - B(V-A) + (B-V)A - (A-V)B. e (a) V(A+B) = yea +B) = yz (Ai + Bi) Lax; max; OA; , OB) _ vy, ; 2 ara VA+VB. a oe () [V*(A +B) Yen 5 dt BD = den at = = [VxA], + [VxB].166 Problems and Solutions (© VPA = Le car = L [ado] = AVp+oV-A. a ade. op = Yew 5 (An = Ay @ Vrt@ay = Yeagy (om Yeale oA ae 32) = oA), +(VorA). © [WAB)]i = Zz LAB = rs Bt ee oe any MA, 4 OB: 2a) - Lae Ba, Fax, aga — (OA; _ OA: (OB; OB; OA; - Lp Pal (ge Saltese tase [onda 0,805) Si +4 Gee] * BNA. (AV -z & OA, OB, - Ee [Bret + a] + BVA + (AVB =F [ene See enn + om St a] + CB V+ (A i ary = Se Dene Me + Deh Fae y y + (B-V)A; + (A-V)B; = [Bx(VxA)]. + [Ax(V*B)]: + (B-V)As + (A‘V)B, © V(AxB) = Yigg OB - Le exam - Leas (AB) i OX aA, 2B, ~ Yee Be YenMathematical Methods in Physics | 167 Interchanging subscripts, OA; OB, , yo a ‘V-(AxB) Yew ox; Be Yee an Aj = Yevxayse - Y(V%B)A; = Be(VxA) - A-(V%B). 7 7 (@ [Vx(AxB)]: = a Les ; nA ee - Beas Budi) 5 Aba j . — : aa Om ex; O(ABi) a = (V-B)A; + (B'V)A; ~ (V-A)B; - (A'V)B;. IIL. If is the electrostatic potential and E the electric field (E = -V@), the electric force F acting on any volume V bounded by a closed surface S is F = J pEdr, where p is the electric charge density inside V(V7@ = -4np). Show that F can also be written as Fi = f Y, Tyd5), where Ty, the stress 3 tensor, is given by / ~ ps6 Ez a Solution: 2 mle ie, =-1fv. V'pEdt = ae VV) Edt 7 ye E)E;dr oo OE; al OEE) _ 5, OE: aed ES «al Hay168 Problems and Solutions Consider now YE; Et. (EW): i We have shown previously that V(A:B) = (A-V)B + (BV)A + Ax(VxB) + Bx(VxA). Setting B= A, VA? = 2(A:V)A + 2Ax(VxA), while for A = E, using VxE = -Vx(Vq) = 0, we have VE = AEV)E = (EVE OE; 1 tee, P52 ms EBay = & -1yy (Xe _1 aa | HAS Fl ax 2 Oy I -1/fy2leg-1 . anf bay BE 5 Euler. Finally, by Gauss’ theorem, -fy “a - wo fE Sa Efe where 4nTy = EiEj~ ge by. a ay Ey, and IV. Verify that at any point, the direction of E is along a principal axis of the above stress tensor Ty. Calculate the eigenvalues of this stress tensor and show that it represents a tension (+E”/8n) along the direction of E and a pressure (-E’/8n) normal to this direction. Solution: Consider 4n Ty = EiEj— £ dy from Problem III above. We can rotate to principal axes, since we know that any symmetrical tensor of the 2nd rank can be so diagonalized.Mathematical Methods in Physics | 169 E 4nTj = 4m YuauyTa = Yusuy (a8 - Fou) i 4 = Yuu Yost . 7 Duane, = 4nTj = E‘E/->3, 22 where E/'E; refer to the components of E in the new system of coordinates. Since for i¥ j, Tj = 0, it follows that E'E/ ~ 0. For E;Ej' = 0 the following are possible: E\E2 = 0; E1E3 = 0; ExE3 = 0. If E; # 0, then Ey = E3 = 0. If Ey + 0, then Ey = E3 = 0. If E3 + 0, then Ey = Ep = 0. Hence we see that along the principal axes, E has its component [E| along one of the axes, with its other two components equal to 0. We choose principal axes such that E:(E,0,0). Then, i=j=1 4nTy = 4m. = P-E2= Ep. i=j=2 4nTif = 4m. = 0-E2--Ep. i=j=3 4nTs = 4nd3 = 0-E2=-E2. Thus, Eo o m-|0 -£ 0 0 0 -E ‘We see that 41 = E’/8n represents a tension along E1, ot along E. To show that Tj represents a compression perpendicular to E we have to show that for any direction perpendicular to E, 42 and A3 do not change and are equal to-E°An.170 Problems and Solutions Let us, therefore, perform a rotation about the E axis, and go from the ijk system to the i system by a rotation through an angle o. Ty’ = YuitujmTun , which only has meaning in for 1= m. Fori#j, Uji T at + MaupTed + UsupTss Considering the rotation shown in Fig. A-3, Jojo! = cosa. = wz. he Joke! = c0s(90- a) = sina = us. Kiko! = cosa = uss. Jo'ky = c0s(90 +a) = - sina = urs. igi’ = Un = 1 lgfo = nfo! = totko = iotke’ = 0; un = us = 0. By virtue of these relations, for i + j, Ty! = wauaTo' + uisuTss'- Tiz' = Tr’ = 03 Ts" = Tai" = 0. Tes’ = Tad! = UraltyaTed + UztssTs3 = sina.cosa(-E°8n) ~ sinacosa(-E/8n) = 0. Fori=j, Ti" hTul = Mh. Tes! = uTrf + usTs§ = Da = Ancos’a + Aysin’a = A.(=As). Ty" = whyTai + uisTsi = Ba = Aasin’a. + Aycos*a, = As(=Aa). Hence under a rotation, 42 and A3 don’t change, and represent a compression (-E°Bn) perpendicular to E.Mathematical Methods in Physics | m V. Plot the equipotential surfaces and the lines of force for the electrostatic field generated by two equal point charges located at the points x = 1. Solution: A line of force is defined by xy = ~ &_D_& dE = PAE By symmetry, any section of the field by a plane including the x-axis will result in the -a,0 0 4,0 same appearance. Let us therefore take that Fig. A-4 section made by the xy-plane. We will plot the result for x > 0 only. The rest of the picture is then obviously apparent. From Fig. A-4, we have (x + a) (x + a) ee eeeae ceca i: Fee ara) [rear ay E, “Feel : [P+@-a Defining u = (x + aYy and v = (x- aYy, we have | ata). ge-a) [y+ @+ay [y+@-ay Hy * Fae? ya + E- + ba +@tay [> +(x-ay : + : Fa +e)? ya + v4? (+)? + (1+ v4)” BYU ++ v4? a& E view) +uievy® vite)? ++ FO + eM + vy?172 Problems and Solutions Fromx+a=uy and x- a= vy, wehave y= 77 and x 422), so that oo woe dv). dk = stato dv)+(u- v)(du + dv)] a 2a (ude ~ od. Hence, dy _ _do-du_ dx udv~vdu" Combining the two expressions for dy/dr: do=du_ | A-dydv _ (i+w ++ vy udv-vdu u-vdudo (1+)? +u(1 + 0°)?" du _ w+) - w+)” _ (ww +)? do ui +v)”-v(l+ v9) ~~ (u- v1 tv)? a+? Ce Separating variables and integrating: du dv : . an aoe In terms of x and y: (x +a) [tay +? + (ea) [ay ty}? = Setting a = 1, we have: sa teu a © Tay eaMathematical Methods in Physics | 173 ‘The equipotential surfaces are defined by = C’: a (eral ry? e-ar ey and, since a= 1, 2 o tla py ter)» When values of C and C’ are 0 os substituted into equations (a) and (&) above, we get the following plot of equipotential surfaces and lines of force. In Fig. A-5, solid lines represent lines of force while dotted lines represent equipotential surfaces. 2.0 Fig. A-5 VI. If @ is any scalar function and A(r) any vector function, prove that: @ Jvodr - foras. y 5 ® Sova = - faxas. $ (Vis any volume bounded by the closed surface S.)174 Problems and Solutions @ Srorar - Se. Gauss’ theorem states that rls Thai. jet = SXtun wd Si. TE we a si want the case that index i not appear in the tensot Thyj....je We have to find ts Thi. .ndt. We proceed by defining Ryii..ip = 5yTiyi..4- Then o y Zones S y oa and a é a fg tue Lf gy Pitinnude = DIZ Rasnadt, which, by Gauss’ theorem - Y fra. Yo fis = FP For the current problem, we set T'= @, yielding Seat - Jods, + fear ~ foas. ® forrarar 7 [Rene 7 D> jy tear, Defining Ry = )eypAk, we have - 2 = 4, Joona [ESYaa [Ens which, by Gauss" theorem = fRyds; = f Yewands; 3 37 . J Pewnuds, - -faxas. 4 feveaar = ~ faxas. 5Mathematical Methods in Physics | 175 Problem Set D I. Show that if a fluid is incompressible, then V-v = 0, where v is the velocity function of the fluid. Solution: By virtue of the conservation of mass (which says nothing about homogeneity ot incompressibility of the medium), we have 2. v.py = 24 py. ve at Vev = Fs pvv+ vpw = 0. If, in addition, the fluid is incompressible, then by definition, ae = 0, so that dp _ op, Op dk, epdy, Opa _9 _ Pia. Q a aad od xa? oe Substituting (2) into (1) we have V-v = 0. II. The stress tensor for a viscous fluid is known to be of the form = =p - 2 (w-vyby +u(See Sh where p is the pressure and yt the coefficient of viscosity. Show that the equation of motion can be expressed as, el where p is the fluid density and F the body force per unit volume. + wv] = F-Vp+ e V(V-v) + pV*v, It Solution: We saw in Chapter 3 that the equation of motion for deformable bodies is given by ye. Fh t where Tj is the stress tensor. ~-N_N, wae WH _ WY, an ath ar * Lag, 7 ae TO176 Problems and Solutions +o ve 2 s " a a ES Boromes(te dl _ 2 a Fv, o dy ax; 3 (Sa cu} Ee aa = Fi (Vp)i~ (2n/3)Vi(V-v) + pV70; + HVi(V-v) ~ (Vp) + (WA)VA(V-v)+ BV?0:. Collecting terms, 7 (5 + vy = F-Vp+ 7 V(V-v) + pVv. III. Prove that for any general tensor function of the nth rank: a Ss Thi... ,dSi = ii ig Si @ L fas, is ¥ ft it where S is a 2-dimensional surface bounded by the closed curve s. The indices it...in-1 may vary arbitrarily from 1 to 3; i.e. prove Stokes’ theorem for an arbitrary general tensor of the nth rank. ® D feng, Be sds, ~ § where kit...in can vary arbitrarily from 1 to 3. ist, © roxas ~ feads, where @ is a scalar function.Mathematical Methods in Physics | 7 (a) Consider the rectangle of sides 5x3 and Vx] + 24 , shown in Fig. A-6. x en 5 Tina: = f lax Tota Fey Tess} Sle Ts” Joe Pah ao}ese “Sle Pride te i Tit 1. }a5s. + Sas - ae] + Sn ae - 1 as] = S(t Aes ffl ns, “PL 4 SP tga Moe ff lag Pl J - finde fray +f: de, «fr de, 4 t . 2 . ‘ +f dey + fri.ds + fr, dx, +f dx, = $P.des + § Taide fT.dee = Ye § Ts.178 Problems and Solutions The arbitrary, unclosed surface, shown in Fig. A-7, can always be so sub- divided to get plane rectangles by passing planes parallel to the x-y plane and parallel to the x-z plane. Each rectangle, then, will be of the type discussed above, i.e. an infinitesimal Fig. A-7 rectangle of sides 8x3 and Vox + 6x3. For any two adjacent rectangles, Yf=0 the line integral over the common“ side vanishes (Fig. A-8). Hence, the sum of the line integrals over all rectangles comprising S reduces to the line integral over the small, remaining triangles of the type : shown in Fig. A-9. oo It only remains to show that for those triangles whose hypotenuses lie along curve s, ‘ Slides + [Pad = y Sti dss. But for infinitesimally small partitions of S, _ ds: (dxy, dx2) (Fig. A-10). Hence, a 4,2. DTA = Tint dts + Ta. : Fig. A-10 (b) Define a tensor Su, a a fe By Tas = J Leven ag Tad a - Ska, Swi, .idS) y § Sit - r Pout ids: = $T,.ids.- (The last two steps follow from part (a) above.) SueTi...ig-» SO thatMathematical Methods in Physics | 179 © S[vorss] : Y fon Seas = Theta - ~feds. (The last step follows from part (b) above.) IV. If His an arbitrary vector field and V-H = 0 everywhere, prove that H=VvxA, where . Aiceaae) = [Hews —fncaesonie. Ar(x1x2%3) ~ fiticuans As(xix2%3) Solution: Let fr(evias diy = GiGait'), 50 that SEF = Hitaean). Then : Jie ars = Gi(xix2%s) — Giaix20), and Zi finery! = Hy(xpn%s)- 0 = Hil). It follows then that oAn xs As Ox Ps a - 2 frttorsn yen = ~ Hiv). OAs _ 22 _ ax, oxy = Aims).180 Problems and Solutions Similarly: a - P fortensesrac’ - 2 fintessordes = Hix) - 0. i =0. i a = HiGvan). Proceeding similarly once again: i - ~ 2 fantasy - [2 HG )dey. = 2 fortes = E fsecasrde . Sz He xX2%3')drs' — Hy(x1220). on, Om it ox + ox, + Hs(x1x20). Oh , 22 __ OH te BuVH=-0 = a ae xy” GA, 0A OH; ’ Bf re any + Hy(r0) = Haren), and we have established that = 24s _dA2 Hy (x122%3) an” ax @A,_ aA HG) = 57 ag, | 7 ‘Hlsites) = Vea. 242 _ 0A, Ha(x1x2%3) = anonMathematical Methods in Physics | 181 Hence, given the fact that V-H = 0 everywhere, one can always find a vector A such that H = VxA. [This vector A is not unique, as will be discussed in Problem VI] V. If V-H = 0 everywhere except inside a finite region R and if in this region 8, [onde #0, then prove that it is impossible to have a vector function A such that H = VxA everywhere outside the region R. Solution: Let region R be bounded by S’, within which fv Har #0. Consider a surface S completely enclosing R, as shown in Fig. A-11. By Gauss’ theorem: frome - freas. Ro For any closed surface S, § VxAdS = JPeamdsi+ forands, s Fig. A-11 = fards+ fade = 0. e tae ~~ ee - te This last statement follows since we can always divide S into Sy and S2, each of which has the common curve s enclosing it (Fig. A-12). Hence, for any closed surface 5, $ VxA:dS = 0. s Fig. A-12 Program for proof: 1. Assume H can be given by VxA everywhere in R’. 2. Then show that this leads to a contradiction of the above, ie. fvxa-ds +0. 5 3. By this contradiction we know that H # VxA in R’. 4. We have demonstrated the proof.182 Problems and Solutions 1. H=VxAinR. 2, fvndr = fv-Har = finds = fvxacds + 0. | R — But this is a contradiction, since for any closed surface S, VxA-dS = 0. 4. Hence, H # VxA in R’. VL. Let H be a constant vector everywhere. Construct a scalar function @ and a vector function A such that H = V¢p and H = VxA with V-A = 0. ‘Are the functions and A unique? Solution: If H is to equal Vp and be a constant vector, then since the gradient involves the first derivative, p must be linear in its variables, i.e. representable by r. We can therefore construct from r and H the only possible scalar, namely, rH = @. (Note: Since V*H = 0 everywhere, H = Vop, where dp = Verds = H:ds; hence, p = Hr.) It is obvious that @ is unique only to within an additive constant, for if @2 be also a solution, H = V2 and V(p2 ~ $1) = 0, so that 2 = $1 + const. If His to equal VxA (which again involves first derivatives of A) and still be a constant, A must be linear in its variables, i.e. in r. We then can form from r and H the only possible vector, namely A = a(rxH), so that H = VxA = aVx(rxH) =a [nv - H(V'r) + (HV) — (r-V)H] : Since H is a constant vector, V-H = 0 and (r-V)H = 0, reducing the above to He a[-3H + H] =-2aH = a=-l2. Thus, A = - (1/2) rxH. Again, A is not unique, since we can find another vector B such that B = A + Vg’. Then, H = VB = VxA and V-B=V-A + V'9". Since V-A is to be zero, if we choose Vg! = 0, we have completely determined B. In fact, we can choose Ve = Ve (from the first part), ie. H= Vp = Vo". Then, B= A +H, VxB=VxA, and V-B=V-A.Mathematical Methods in Physics | 183 Problem Set E I. Show that if F1 = Vop and F2 = VA, then fFy-Frdr= 0 provided that as Ir] + ©, either |pF2] or |AxF| goes to zero faster than 1°. Solution: fevridr = frp vxade = SY oe Yen Mt = SE [Ztowand -0 5, ad}, But since y i (VxA);= V-(VxA) =0, Sima es Ie Leno ge ds, which, by virtue of Gauss’ theorem, = Df ocveayas, = fvxayds. Finally, then fremar = Fovxayds = f[vx(pA) + AxVe}as. os Jt a ‘We have already shown (Prob. V, Set D) that f 'VXxV-dS = 0, ie. that the surface integral of the curl of a vector over a closed surface vanishes. Hence, fFrhide = f(axvoyas = fo(0xayas. i 2 2 = As |r| > =, kp(VxA)|IaS| = |pF:||dS| > a 70. : = fener =0. ee I eo space184 Problems and Solutions IL. Prove that if V7 = 0 and if as |r| —> ~, @ goes to zero faster than r"”, then @ = 0 everywhere. Solution: Set F; = F2 = V@ in Prob. (1) above. Letting p —> const. r*, where s<-1/2, ie. letting s = -1/2 -— n, where n > 0, it a WF s| = Vel 2. Tava pred” awd? and since n > 0, {pF + 0 faster than r"?. Thus firma = Severe. The quantity (V@)" is positive definite, so that [(Veyar vanishes only for v Vep = 0,or@ = const. But we know that @ -+ 0 as |r| —* °°; hence, @ = 0 at infinity. But since @ = constant, p = 0 everywhere. IIL. The wave equation for a meson (satisfied by any particle with mass) is given by i Veo-ae-Ke =0. (a) Show that ¢ = poexp|+i(k-r - w)] is a solution in which go is an arbitrary constant and k is related to w by c*( + x’) - @” = 0. (®) What is the phase velocity of this wave? Solution: (a) From p= oexp[ti(ker - on], op ax ~ PeexPl I(tiks) ee _ nc ax? PeeRPL Ike = oexpl -k)} = Vp = -Rpexpl 1 = “Re. RPS Ie = poexpl ke)Mathematical Methods in Physics | 185 a ope sig); 28. 2 --#@ ar PeexPl Ii); ap 2 PPL 1 ae Hence, p- 4 2B - 2p 7 -e-$oe = 0. Thus, © = Goexp[+i(k-r ~ at)] will be a solution provided c7(? + x?) - a? (b) To find the phase velocity of the wave, let us construct wave surfaces at a particular instant of time fa (Fig. A-13). The wave surfaces are given by p= const., or what amounts to the same thing, z = k'r ~ @fy = const. c’. Fig. A-13 These are the equations of planes; the different planes within the family are determined by the parameter c’, the constant. ‘Assume for the moment that k is any vector in space, and let us move a distance dr along @ = 1, where z= 21 (constant). Then, along z= z1, dz=0, i.e. d(kr) = 0. We see from Fig. A-13 that the only way that d{k-r) can equal 0 is for k to be perpendicular to the wave surfaces. Hence, we correct our diagram to be as shown in Fig. A-14. | | exerary eo) Fig. A-14186 Problems and Solutions The phase velocity, or the velocity with which the wave surface moves, is given by b = |dr/dt. If now we allow the wave surface to move with time and observe the same wave front (say (1) after some time interval dt, then dz =Oagain. Thus, 0 = d(k-r) - wdt. But, since dk = 0, it follows that K-dr = dt. From Fig. A-14 we see that k-dr = kldr|. Thus, the phase velocity is given by [using the result from part (a) above] o- 8. ie. IV. Let Ax Ay Az be the components of vector A in cartesian coordinates, and Ap Ag Az the components of A in cylindrical coordinates. (@) Express Ay Ag A; in terms of A, Ay Az. (b) Using the above relations together with a direct change of variables in differentiation, show that AAs , Ay, Ws 18 (gy), 1 Be , OAs ex dy o& pop pop a& (a) From Fig. A-15, x = pcos. y = psing. fee From the above: Fig. A-ISMathematical Methods in Physics | 187 In curvilinear coordinates a vector is expressed by (A); = A-eoi, where 01 is a unit vector in the E; direction, i.e. Ap = A’eop, etc. But, since Cop = (cos, sinp, 0), og =(-sing, cosp, 0), €cz = (0, 0, 1), Qo a a i» = Axcosp + Aysing. Pvt ye vey = -A,sing + Aycosp. = 4, = a A Yess Vere Ar = Az. A, (b) We also have: 2A, _ 2A, ap , AAs Bp , As Be Op ax Op ax oz ox’ Ay Op , Ay Bp , Ay 2 ap ay" Op dy” oz dy" A, ap , OAs Op , AAs Bz _ AAs ez Op dz Op Oz a Gz az 2 off 2 From solution (a): fA,sinp = A,sinpcosp + Aysin*p. : = Ay = Aysing + Agcosp. = = -A,singcose + Aycos’9 | ef casa nd [A,cosp = A,cos"p + A,sinpcosp. . = A, = A,cos : i = -A,sin’g + A,singcose | eee As ( i al m cosp Se 7 sing “ cosp + ~Aysing + cosp “Ae a ~ Aycose ~ sing S}( - SP) 2 oe sete singcosp 24 — Ay, a rreos'g - $2 singeosy + > ee p p op188 Problems and Solutions V. Express Va and VxA in cylindrical (@) In cylindrical coordinates, yet n= p fy = 1 1 fAViPhehs) , AVehshs) , + eehito “Iihahs| 2 2 ots Ie Hence, in cylindrical coordinates, Mp = V4 = L[pVeh) , AVeb) , o(pV.b) Vb = V(Ve) ( ae "| ButMathematical Methods in Physics | 189 Hence, ro So} 20) 268) SoM} 5-35 &) Ingeneral, so that in cylindrical coordinates: VI Express V7A and (A-V)A in spherical coordinates. (a) VA = V(V-A) - Ux(VxA). For spherical coordinates, Bar b= B-0@ Ih-r B= @ hy = rsind We have shown (p. 60) that vA ’ 2 (4) +t EZ tsinoas) — a aea Problems and Solutions ‘We also know (p. 58) that vv-ay|, = 2A. i: VV-A)|. = 736 (V-A). VV-A)|, = ran oe ‘A). From page 62 we have Vx(VxA)|, = a {5 9. 1 sin(VA)y - 2 Srveay} lo a ae fe sin0(V%A)y 5 (weap. Vx(VxA)|, = rane (op (VxA),~ Z rinocvxay. yxcvxa)|, = 1G 1(VXA)o - wry. Substituting the appropriate expressions for (VxA),, (VXA)o and (VXA)p (page 62) yields mora ft “tet 2 Vx(WxA)|, = was Shas 1 (Aainta my “lBMathematical Methods in Physics | 191 Iron = [lanl oe “ar r ar|sind | op 1fA[_1 {Asin@Ay) aA + Gola ae Sl Combining: |val, = |VCv-A)|, - |Vx(VxA)], . |VA], = [V(V-A)], - |Vx(VxA)], - |val, = [¥cv-a)], - |VxcvxA)], - ‘The importance of the above result lies in the fact that we cannot compute V’A from A in the case of general orthogonal curvilinear coordinates as we could in cartesian coordinates. For instance, we ask: Is V7A given in spherical coordinates by the expressions: |VA|,=V74,2 |V°A], =V'40? |VAl, = V7Ay? The answer is NO! and we cite a simple example. Consider the vector A= igntjy the =r. Vr = iV'x + joV'y + koV'z = 0: if we use cartesian coordinates and apply |V°A|, = V74j. On the other hand, if we try the same in spherical coordinates: oo SEBS SE] 3 2 |v Thus, A = 22 40, (b) We employ the identity: V(AB) = Ax(VxB) + Bx(VxA) + (A:V)B + (B-V)A.192 Problems and Solutions Setting A = B, we get V(A”) = 2Ax(VXA) + 2(A-V)A, so that 2, a) (AYA = YOO — axcvxa), 2 A? = {magnitude of A)’, which is invariant to coordinate system transfor- mations. In cartesian coordinates, A” = A? + A3 + AZ. One can find A, Ae Ap in terms of Ar Ay Az by using the definition of a vector in the new r @ system, as follows: (A): = (A’éoi), where ois a unit vector in the ith direction, i.e. ear = (sinOcose, sin@sinp, cos®). e.0 = (cosOcose, cosOsing, ~sind). op = (~sing, cose, 0). and A, = A,sinOcos + A,sinOsing + A,cos8. ‘Ap = A,cosBcosp + AycosOsing ~ A,sind. Ag = ~A,sing + Aycosp. Solving for Ax Ay Az: Ax = A,sin®cosp + Agcos®sing - Aysing. ‘Ay = A,sinOsing + AqcosOsing + A,cosp. A; = A,cos® ~ Agsin®. Evaluating A? = A? + A} + A2, one finds it equal to A? + A3 + Ag. This result should not surprise us since, in spherical coordinates, A. = Alor + Aveoo + Aglop => AA = A? = A? + AR + AG. geo Therefore, [vay], = net + 43+ A].Mathematical Methods in Physics | 193 Regarding the second term in Eq, (1), we can solve for (AXB) by noting that in spherical coordinates, A = Arbor + Aveoo + Aglop- B = Breer + Botoo + Byeoy- qd AXB = (AoBy ~ AgBs)eur+ (AgB ~ ArBy)eso + (ArBo ~ AoB, Jem. Substituting the expressions for the gradient and the curl (pages 58 and 62, respectively) into Eq, (1) above, Jawal, - 2 [ato d+ 4] Ag (aA, _ a(rsin®Ay) Tr alae or }- JAA, - ¢S fat + ad +3] -{ Ay [2erinsas)- aca) 4 (eta 2A, Psind| 28 Op) or 00): |AWAL, ~ sang dol td +l] “bt : Ae [Aegpte - Acta) : Psind| 08 ap194 Problems and Solutions Mid-Term Exam Problem Set L. Vectors ABC are related to three other vectors abe through the following relations: A= Yq; B= Yb; G= LT, A A Fi where Ty is a symmetrical tensor of the 2nd rank with eigenvalues 41243. Prove that A(BxC) = Asdsdala(bxe)]. Solution: Note that A-(BxC) is a scalar and hence invariant to transfor- mations. Let us rotate to that particular system in which Ty = yA; , i.e. the principal axes. A(BXC) = AB XC!) = LA’) enB/Cr oe "YTin’cr’ - Lend rani Lrdabn' Ladner’ = Dewrarb/rcr = Drrarena’djcr’ im im = Yaa/Yephjbp'er = dudads a'(b’xe’) 7 ik = Arhads ar(bxe). The last statement follows since a-(bxc) is a scalar and thus invariant to the transformation. IL. If eo is a constant unit vector, prove that V = |}(eo'r)dS|, where Vis the volume of a region bounded by the closed surface S.Mathematical Methods in Physics | 195 Solution: Since eo is a constant unit vector, V= [fea But we note that v V(eor) = (€o°V)r + (r-V)eo + €%x(Vr) + rx(VXe,). Noting that 0 is a constant unit vector, (1-V)éo = (Vxr) = (Veo) = 0, so that V(eo'r) = (€0'V)r = €o. Hence, from Gauss” rio == [fe UL Let S be the area of a surface which lies completely on a plane and let fo be a constant unit vector lying on the same plane, Prove that = [feecrss where S is bounded by the closed curve s. Solution: Since toxdS = |flldS| sin(to,dS) = dS, and since V(to'r) = fo from Prob. I above, we have = |fexas| - [focoras)| - [Pen Z corns ~ [feonns, by Stokes” theorem.196 Problems and Solutions Problem Set G I. Pauli spin matrices are defined as: [(01) (02). (1 0 e110) |i 0) 10-1 (a) of = % (i=1,2,3) (0) 010; = dy +i) ee0% (ij varies from 1 to 3) t Prove: (6) a0; + oj0; = 2By Solutio @ (b) We first note that 010) = 8y + y 4x0 implies that 0102 = 103. 0201 = ~i03. 6203 = id. 0302 = ~i0. 0301 = i02. 0103 = ~i02. = = =. This, then, is what we have to show.Mathematical Methods in Physics | 197 ow = P30 3} -(6 5} =i 3] an (19 o301 = ( om OPS) (S]- a5 S]-m om S08) oof} Rt Thus, for i= j, 10) = dy, while fori #j, 010) = io {i for ijk an {esa permutation of 1, 2, 3.198 Problems and Solutions (©) Since for i+ j, 010)= iog and 010; = -iox, we have G19; + 0/0; = 0, while for i= j, 010) = I, so that 010) + 0/0; = 21. Therefore the expression 010; + 0j0; = 2by is satisfied. II. Hamilton's generalization of pure imaginaries ijk may be defined as i= -io. j= io. k = -i0s. Show that ijk obey the following rules of multiplication: (@) P=R=-1. Oy sk ky Solution: (-ioiX-ioy) = -of = -I. (-io.(-ioz) = -0F = =I. # = (-io3(-ios) = -03 = -I. ‘Thus we see that Hamilton's imaginaties ik obey the same rules as our “ordinary” imaginaries, i.e. i2= 1, where by -1 we mean (3 i} (b) ij = (-io1-io2) = -0102 = -io3 = k. Ji = (Cioaf-i0;) = -0201 = +i03 dk = (-iozX(-io3) = -0203 = -io; kj = (-ios-io2) = -0302 = tio, Ki = (~ios(-io1) = -030, = ~io2 ik = (-io;X-ios) = -103 = tion LIL A quaternion Q is defined to be Q = t+ ix + jy + kz, where txy zare real numbers. (Notice that all real numbers are quaternions.)Mathematical Methods in Physics | 199 Prove that: (@) If Qi and Q are quaternions, then Qi + Q2 and Q1-Q2 are quaternions. (®) QQis.a real number, where Q" is called the “complex conjugate” of Qand is given by Q* = 1 - ix - jy - kewhile Q= 1+ ix + jy + kz. (Note that Q” is also a quaternion. (©) For every quaternion Q there exists another quaternion Q” such that eg'=@'Q=1. Solution: (a) We have to show that Q: + Q and Q1-Q: can be expressed in the form ttixtyytie. TE Qi = t+ in + jy + ker and Q = 12+ x2 + jy + hep, then Qi + Qe = (tr +b) + (ha + ie) + (yr + fy) + (Kai + ke). But since (ix + ix2) is simply the addition of two matrices: 0 | 0 ol 0 -i@itx) 2) a 0 |*|-im o -i@itm) 0 = ~i0;(%1 + x2) = +i(%1 + 2). Hence, Qi + Q2 = (ti + 2) +i (x1 +22) + 01 + 2) +k (zi + 22), Which is clearly seen to be a quaternion. QrQe = (t+ iar + jy + kea(h + ie + yr + ha) = T+iX+j¥+kZ, where T = th-xm-yy2- tt = T isteal. X= xhtuhtyie-yu — X isreal. Y= yih+ yh +um- 2% = Y isteal. Z=ahtuhtxuy-xmy = Z isteal. Hence, Q1-Q: is a quaternion.200 Problems and Solutions (b) If Q* =2- ix jy - kz, then QO = - e- fy- ee itt jy +) 4 1?(real number) + (x*+ y** z?)(real number). = 17+ x y* 2? (real number). (c) We have seen that Q°-Q yields a real number. Let us so choose the quaternion Q" to resemble Q" yet yield the real number QQ = 1. In this way we will have formed Q”. Letting Q’= peek , where D = 17+ x?+ y?+ 2? we have go = okay eke) | Prateyted’ D D : =y-ke Be xt yt 7? Qe Ca Observe also that oo - owe ye kaye io kD | Thus, QQ = QQ = J, andQ'= ee IV. Let systems © and E’ both be inertial systems, and let system E move with a uniform velocity vo with respect to system E’ along the x1 direction (Fig. A-16). Ifa particle moves with a constant velocity v along the same direction, as observed in system Z, show that the same particle, if observed in the E system, would move with a velocity given by 1 DtV. v= Bve" a where c is the velocity of light. 1+Mathematical Methods in Physics | 201 Solution: The problem is to find v’, the velocity of point B, in system I’ (Fig. A-16). The Lorentz transformation between systems E and I’ is given by yw EER. EC) Fig, A-16 where B = ve/c. In system E, x = Xo + vt, and substituting: y= Rt@tvy , _ 1 vet oye? v1-B 1-F a) _ BD. =) _ Ltvvyc? ar), V1-6 v1-B Therefore, (a oe dr * 1+ vv? or Note that by combining 2 velocities, each of which is less than c, there results a velocity less than c, for if v = c~ k and vo = ¢ - @, where both k and @ are greater than 0 but less than c, gn © = kte- ole _e2e-k-w) a er SF aktah= 8) epg) + HE 1+———— (2c -k- 0) Note also that if we combine the velocity of light ¢ with any other velocity less than c, we still get a velocity v’ = c: - ote ete ltevy/c? (ct vye202 Problems and Solutions Following is an illustration in the . z e" use of the matrix method to solve v the problem of compounding two | x velocities, and which therefore x : represents another method of solving the given problem. ra Referring to Fig, A-I7, x= Vix’, while x” = Vax, so that x" = V2Vix', where V2 and V; are the following Lorentz transformations: —1_ 9) $b 00 tb v1-B v1-Bi 7 v1-6 0 10 0 0 10 0 n= V= 0 o1 0 0 01 0 a Bib, pol Y1- Br Y1- Bi VI- Ba v1- Be Bi = voc, B2= v/c, and the resulting transformation is given by W= V2V1, where Wa = —— + ie ¥(1~ B21 ~ Bi) (1 ~ B2)(1 ~ Bi) (a=sinak pi (1 + pipay fy = 2 - Bute "ce 1+ Bis" Thus, Vot vd 1+ vovfe? is the velocity of the moving system E” as observed in system E’. vs =Mathematical Methods in Physics | 203 V. If under a Lorentz transformation one regards the electromagnetic field E and Has forming an antisymmetric tensor of 2nd rank Fyy while the electric current J and charge density p form a 4-vector jy, then show that ‘Maxwell's equations can be written as ea 4nj, (a) Vee aa » Mah.) o (b) wep 3 Fup = 0, Lem 35, . where Fo = ~Fous Fotis, Fu=Hi, Funth ee ee ee ee and where jy = (5, ipc). Solution: 0 Hy -Hh ~iE du x | 0 Mh -ib , | a2 faba Fw! nn -m, 0-i&| # "|_| 7] 2 iE, iE, iE, 0 ipe ict a Any; (@) Consider Yan FH oe Forj=1: OF , Fis , Fu _ Anh Ox, — Ox3 OXs ca 196 at (gap, = 4, 1 Es c Ot c e oc a (2 ata) ar, Ox Ox, Ox; Ox, - Ox; Om) c Ot ee204 Problems and Solutions Forj=3: Ox, Ox, Oxy - (OH, Ah) 1 0B, _ 4h _ 4s , 1 OE } ec or = a. *e or” OF , Fea , Was _ Anja om Om Ox = ict. = (VED- desceres oS) ce (VE) = 4np. Thus, for j= 1, 2,3: (vary - 424 1 while for j= 4: (V-E) = 4np, and we have arrived at 2 of Maxwell's equations. (b) Consider em Ge =0. Fori=1: OFu_ Fn , Fa Fu, Fn OF a Ox. Ox, OX; Os Oxy Oy 1 (2-22), 29 8 _ 1am, al le a ole Getor OF 4 OF 3 , OF 13 OF ' OEs Ey) | 2i OH, _ - - 10m af S| Hh - 0, (VE), orMathematical Methods in Physics | 205 eee ae Fn _ OFn , Fu _OFu , OFu Fa _ 4 Ox Oxy Om Ok, x, OM : ' 21 OH ~1aih cor a: (HE), oe Fori=4: OFxs_ OF, Fa _ OF, OF _ OF _ Ox, Ox, Ox; 0x3) Ox. Om ' an” a2 * ax) Cee Thus, for i= 1,2,3: VXE = 1 Hf, while fori =4: VET = 0,and we have obtained the 2 remaining Maxwell's equations. Let us attempt to construct the anti-symmetric tensor Fyy and the 4-vector jy from Maxwell's equations. In free space, Maxwell's equations are given by: @® VE = 4np. di) VH-o0. ity) vx ~ -4 if. (vy wet - 21g, ae Since V-H = 0, we know that H can be expressed as the curl of a vector, i.e. H= VxA. Let us also specify that V-A = - 12, Substituting H = VxA into (ii): 41 a(VxA) c oO VxE = - =-lyxA = vx(E+ 3] -0. c c206 Problems and Solutions But if the cur! of a vector vanishes, we can write E+“ = -Ve, with va--122, Consider the 4-dimensional gradient operator (which we know to be a good 4-vector), having components: a-(2224)-2 ax’ dy’ Oz’ ic dt ax, To test for vector validity, we dot the expression into (J and observe whether we get a Lorentz invarient as a result. We proceed to do this for jy. j= 2, 82 , sy, he eae to orion an the last by virtue of the equation of continuity. Since the quantity 0 is Lorentz invarient, jy is a good 4-vector. Let us now perform the same test for the “4-vector” 7 =(A, ig). 2A, , 2A: , 1 dp 109 4 ahr 189 _ yy 41 de _ oe i ies Kserieialon Gar oclonme the last by virtue of our choice of V-A.. Once again we see that Vi is a = =-1 Ai _ 2, good 4-vector. From Hy=(WxA); and Ey=- 7S - 52: OA; _ OAy _ Vs _ Va Ge OA, OA, _ OV _ OVs A, = Hy "on ax ats ars" pw dhe he _ Vs _a% ae ay Om Ox” p= 28 10d _ fai aH farce on «ion «ou; _o% ay BN on, Ota” pp = oats Ox, Ox”Mathematical Methods in Physics | 207 Note that the terms on the right are components of an anti-symmetric tensor OVy OV, Fyy 2 2-2 Fy, where Fyy ar so that Fu = Fx = Fx = Fu = 0. Fr = He Fu =~ He. Fay = Hy. Fn = ~ He. 0 H, -H, -H, 0 Hy H, -H, 0 iE, iE, iE,208 Problems and Solutions Problem Set H I. Find the eigenvectors and eigenvalues of the Pauli spin matrices 01, oz and 03. Solution: (01 (On : 711 Ol: eae 0-1)" All. 1-k 0 oe | Afro - 0 ta ¥-1=0 ¥-1=0 1-4 =0 Mtl Me-l Matte nl Mata ol Hence, for all three spin matrices eigenvalues are +1 and -1. We then have: ofP! = AP, where i= 1, 2,3 and j= 1,2, and where we (3: of =a oo = aa oot! = a! (°o](ei)-»[)- (8) ae) (8) 6g)Mathematical Methods in Physics | 209 WG = 4) wis a eae vi) Hi, iVi= 4 [G Hence, als ele wie aI = a om vz ae ae y= at Y= a v2 cry wi = ntlan % 1 aS. II. The interaction of an electron with a magnetic field H is described by a hermitian matrix E, a 2me 2mo 240s where e, m, c, fare, respectively, the charge and mass of the electron, the velocity of light, and (Planck’s constant)/2n. (@) Show that o; transforms like a vector, ie. that EoiH depends only on 1a]. (b Find the eigenvalues of E for arbitrarily oriented magnetic field H.210 Problems and Solutions Solutions: (a) If we can show that © ojHj behaves like a scalar, then we will have proved that 0; behaves like a vector. pon aft aft a (ans a Since this is a hermitian matrix, there exists a unitary matrix, U, such that E oilti= A= Mby. Then, Hy + iH, -H3-% V = Hi+H}+Hi= Ho .=+[H], = -1Bl. aoe a4 Thus, under the unitary transformation U, E oid: > ; 2 1H], and E oil; depends only on, 2.0; behaves like a veetor upon transfor mation, so that G-H = £ oHj. (b) From (a) above: E+ sell (i o)*l0 h i fm Zain eh : me Hi * ita) a HsMathematical Methods in Physics | ait det |E - 48,4 = 0 yields: uote 2 4 - (tu (ee t+ Ha) = 0 a a 2me where H=VH}+ H+ HE. Ii. Let A and B be either unitary or hermitian matrices. Prove that the necessary and sufficient condition for AB = BA is that there exists a unitary matrix U such that UAU* and UBU’ are both diagonal matrices. Solution: Sufficiency Since A and B are both either unitary or hermitian, there exists a unitary matrix U which simultaneously reduces A and B to their respective diagonal forms A’ and B’, where A’ = UAU* and B’ = UBU". Since two diagonal matrices commute with each other, A’B’ = B’A’, or UAU'UBU' = UBU'UAU' = UABU' = UBAU’ = AB = BA. Necessity: Since AB = BA and both A and B are unitary or hermitian, the product AB remains unitary or hermitian; hence, it is possible to find a unitary matrix U such that qd) UABU’ = A, is a diagonal matrix. Multiplying both sides of (1) on the left by UA"U", we have UA*U*UABU* = UA*U*A, or (2) UBU' = UA'U'AL. ‘Multiplying equation (1) on the right by UA*U*, we have UABU* UA‘ U* = A,UA*U* = UBAU*UA*U* = A\UA*U* (since AB = BA), i.e. @) ~~ UBU" = A\UA'U’. Equations (2) and (3) show that UA*U* and Ay commute with each other. But, since A, is a diagonal matrix, UA*U* is necessarily diagonal (see next page). Thus, UAU* = (UA*U’)’ is also a diagonal matrix.212 Problems and Solutions If we repeat the same procedure, this time multiplying through by UB*U’, we get the result that UB* U" is necessarily diagonal, and that UBU* is a diagonal matrix as well. We need now only prove that if 2 matrices commute with each other and one is a diagonal matrix, then the other is necessarily diagonal. Set Dy = 4iBy (i.e. D a diagonal matrix), and let DA = AD. Then we have: (DAyy = (AD = YDipAy = YAipDos- ° : 4 Lrvidys = LAighpByy > edy = Avy = Aidyds- ° ' Ay=Ajby = Ais diagonal. IV. Prove that the result of any finite rotation in a real 4-dimensional space can be achieved by two 2-dimensional rotations. Furthermore, show that these two 2-dimensional rotations always commute with each other. Solution: Let A be the matrix in question. Since A is real and orthogonal, it is also a unitary matrix, and so AA‘ = 1, ie. A* = A”. One can therefore find a unitary transformation U such that UAU* = A (diagonal matrix). For proper rotations only, det [A] = +1, so that det |A] = det |UAU"| = det JA] = AidoAsay = +1. The eigenvalues of a unitary matrix are complex, and satisfy the expression IAP = Maz = 1. Setting a1 = e (6 real): Ay! = ey! me Ay = ey", Since A is a real, unitary matrix (A = A"), Ay = ey! = Ay" = &®y", and 42 = e® is also an eigenvalue.Mathematical Methods in Physics | 213 Setting A3 = e'? (g real): Ay = ety = Aty™ = ey”, Since A is a real, unitary matrix (A = A’), Ay = ey? = Ay" = ey", and 24 =e”? is also an eigenvalue. Therefore, 000 e® 0 ool(100 0 To) Ores 0.0 | 10 2700/1011 0, 0 0 oe oO 0 0 10//00e" o 00 0e* 0 0 01jsl000e* ‘We have already proven that any matrix with diagonal elements e”, e“® is equivalent to a 2-dimensional rotation. Hence we see that the 4-dimen- sional rotation in real space, A, is equivalent to two 2-dimensional rotations. To show that these rotations commute with each other, we need only note that they are diagonal matrices, and then show that all diagonal matrices commute with each other. Proof: If A and D ate diagonal matrices, Ay = Aiby and Dy = Dy. Then, (DA)y = YDaipAydyj = DAG; = DAB. ° (ADyy = YABipD,By = ADB; = DA’. 7 V. An infinitesimal rotation is one that is generated by a unitary (or orthogonal) matrix U, where U ~ I= € is an infinitesimal quantity. Show that if one neglects second or higher order quantities in €, then (@ &y = -& (Ey = -€i for U orthogonal) (6) _Allinfinitesimal rotations commute with each other. An infinitesimal rotation is a unitary (or orthogonal) transformation such that the components of a vector are “almost” the same in both sets of axes, i.e. the change is infinitesimal.214 Problems and Solutions Thus, for the 3-dimensional position vector r(x142%3): 3 xy = xt Our + Gun t Ets xf = at Ley mi 4 : 3 xf = L@y+ edy = Yn, Fi Fi where Uy = 5y + €&j —* U=1+ € defines the infinitesimal rotation. Solution: (@) For U unitary For U orthogonal u'=u ul-v [Note that UU! = (I+ €)(I- €) = 1+ €I- €I- © = 1, if we neglect terms of order €”. ] U=I+E U=ItE ut=1-e] | Ul=I-€ U=I+e e (b) If we neglect terms of order €?: U=I+E (1+ €\)(1 + €2)= 1+ €, + &2 + EE, ~ 1+, + 2. (1+ €,\1 + &)= 1+ 82 + + 2) ~ 1 +O + Eh. Since matrices commute with regard to addition, it follows that infinitesimal matrices commute, This implies that, contrary to finite rotations which do not necessarily commute, infinitesimal rotations do commute. In other words, since a rotation is specified by a direction (axis of rotation) and a magnitude (angle turned through) and since it obeys the commutative property in the addition of rotations (i.e. UiU2 = U2U1), then we can represent infinitesimal rotations by vectors.Mathematical Methods in Physics | 215 VI. Let H be a hermitian matrix and let e” be another matrix, defined as en = $a 5a Prove that e” is a unitary matrix. Solution: ead (ery - Setting H =H": any iH_ # in’ (ee) or ar at set so that (enter sagt pei}e aly Laid 8,6 vw a a ar )-#(%-aea) bli yi li 315! a. 2 ve 65! 2a! 3! x) o (rb! @-! For n odd, (~i)" = -(i") and the last term sum vanishes. For n even, we get a series of terms, the sum of which vanishes. The result, therefore, is [exp (i#)][exp (iH)]" = 1, which is the condition for a unitary matrix.216 Problems and Solutions Problem Set I I. Ann xn cyclic matrix C is defined as: Cie Cree Ce Cy Ce Ci + + + Cur Crt eee CG Cr Ge Prove that the eigenequation Cy = Asy* has 7 solutions, with 7 Yc. exp [i(m-1)2sn/n] . ot 1 exp (i2st/n)'1 exp (i2st/n)-2 exp (i2st/n) [n-1] Solution: Consider the first of the n equations formed from Cy* = As’, oral) a Q) Cr+ Caexp Best) , + Cyexp + Cyexp Giop a + Crexp ae = Ycn exp 2D, = avi, = ie. Cry? = Aswi true for the first equation.Mathematical Methods in Physics | 217 Multiply (1) on the right by exp (2ins/n)-1 to get Lewes pie LEST exp 2, a series in Cm multiplied by some exponential. The coefficients Cn-1 now have the same exponential powers as the previous set of Cm. What about terms C and C,? Foc; = C exp 2. ForG, = G exp 2st) = Cyexp(22ns) = Cy, yielding 2 Ca¥ Cr exp PE «6, exp PEO) +-C, exp PEO + + BaD) 6 gp BED) Cy2 exp -1eXp 5 ee - Y.cu exp PEED exp Bes = Ais = ie. Coa? = As = true for the second equation. In fact, every : i2ns(p) a : operation of exp"? upon Ae = Yu exp [i(m-1)2s1/n] serves to shift mel the coefficients C;, so that m — m ~ p. Setting p = n - 1 yields an equation of the form: Bet) as Crexp 22) 4 + n Crt Cyexp = + Cy exp Cuexp oust +Cexp Beste} = Sc. exp PEED on poe) = het which we recognize as the nth equation of Cy’ = Asy’, i.e. Cui! = Asi -218 Problems and Solutions We have therefore shown that 4s =); Cm exp [i(m-1)2sn/n] and = + exp (i2sm/n) 1 exp (i2sn/n)-2 satisfy the equation Cy’ = Agy*. exp (i2sr/n)[n-1] Il. Consider a one-dimensional, ring-shaped crystal having 7 lattice points. Let x; (i= 1...n) be the distance of the ith lattice point from its equilibrium position. The kinetic energy T and potential energy V of the entire crystal are given by on To Ft et ve £ Yaa - x) + £ (many. (@) Find the normal modes of this system. (b) Express each x; as an explicit function of time. Solution: KP k VF D fea ean tal] + [xh 2mm taf] = Lovey, where the matrix vy is given by: 1-420 0- -12 -Y2 1 -Y2 0 - 0 0-2 1 -Y- 0 pet -y20 0 -y2 1 Note that v is a cyclic matrix and, since it is also real and symmetrical, it is hermitian.Mathematical Methods in Physics | 219 The equations of motion of the system are mi, - -¥ on Sow, (=1,2...0). ft In matrix form, = Tx, where Ty= 2. Since Tis hermitian, one can perform a real rotation in real space to effect its diagonalization, i.e. one can find a U such that UTU = A. If we define q= Ux = q=U%(U independent of time), then UX = -UTx = -UNUU)x = -(UTU)q = -Aq = 4 = -Aq. Thus, for every 5, qs = —Asqs, where As are the eigenvalues of matrix T, ie. of matrix v/m. In problem I above, the eigenvalues of the cyclic matrix C, Asc, were found tobe dec = Sc. exp [i(m-1)2ns/n].. m Since the eigenvalues of the hermitian matrix T, Asr, are real, we take the real parts of Asc to get: 7 £¥\cucos 2nstm-1Yn, ™ where Ci = 1, C2=-1/2, Cn = -1/2, and all other C; = 0. ~ 008 TS 5 OD = EY cos 2 ccc: n m n Solutions of gs = —Asqs are therefore given by qs = Ase" + Bye '™, where «3 = im (1 - cos 2ns/n). The normal modes of vibration, then, are: Ge = Ag ATCT 4 gnc220 Problems and Solutions (b) The inverse transformation is obtained as follows: q- Ux = x= Uq = x= LU pa, A where cos 21/n cos 41/n > - - cos 41/n cos 81/n > * - : poles ++ + + cos 2ns(p-1Y/n cos 41/n cos 81/n + - - cos 21/n cos 41/n > * - and Xp = YUmae = Y.qscos 2n(p-1)s/n. a A Hence, 4p = Yioos 2ns(p-1y/n [Ay eM 2S +B, @ RT om TN), An TIT. Show that for 0 + Tan cosnx + ya sinnx, where a : A” Trout Ao = : frends. Be 2 frtowinnsds. We can write Ay as Vf t = [ronsen «resem. Replacing x by ~x in the first integral yields Q ° . JF(aeosn-2)a-2) = -[facosneds = Pjeveosnxar. Hence, 2 2 . An = 5 Percosneds. Ao = 2 Grae. B, = 0. (The last term follows since F(x) is even while sinnx is odd.) Evaluating: aa = fae -[- 287 2q- 2 = q(l-cost) = =. lo 2p [sintn+1yx - sin(n-1)e] An = 2 foinxcosnxde mm it se cestet 3 % ntl - if cos(n+ In _ soot oe) ml ont For n= 1, 3, 5,...(i.e. odd), An vanishes identically, since 1 - cos(n+1)n = 0. For n = 2, 4, 6,...(i.e. even), Zp -. 2a i cc w-1 (n+ 1Xn-1)"222 Problems and Solutions Hence, YiAn cos = Az cos2x + Ay costx +... I Finally then, = sine - 2 - 4/9082 , cosdx , cos6x a (ses (b) In the range -1 Bhs) = -CUE+ gtk). (The last statement follows since ko is arbitrary.) Solutions are therefore given by: g(kt) = Aye" + Bye, where w = V+ pb’ and Ay and By are functions of k and @. Substituting back into the original expression for @(r?): 1 rt) a ort) = amelie + Belk. Absorbing the x factor within Ay and By, we have: wet) = Spiel? + hela.Mathematical Methods in Physics | 225 Final Exam Problem Set L. The electric and magnetic fields are related through Maxwell’s equations: ven = 41,5. eae cc c In the above, j is the electric current density and c the velocity of light. Prove the following: 8 (E+) - feigre & sf a = sei de + few, where Vis the volume enclosed by the surface S. Solution: Note the following: ; : oo 5lEt#] > 5, E+ HH) - 2(EE+ HH). V-LExH] = rz (ExHli = Yan eaEjHe » Yew [ex ae no i] a oi - Dew ge Lar (VxE)H - (VX) E. Thus, @ -Sfale + le - -a Jee + Hild & fueanas = 5 feteandr = [(oxeH - var) ar 7 ~ jg fOEHY BELar- fear. Hence, Shale +B - finde +< foots.Problems and Solutions Il. Let H and H’ be any two hermitian matrices of equal rank. (@) Which of the following matrices are hermitian? 1. HH 2. HH + HH 3. (HH' - HH) 4. H" 5. 6. HHH he 8. ee” (0) What is the necessary and sufficient condition on H and H' for all the above matrices to be hermitian? Solution: (a) 1. (HH'Y = H"H’ = HH # HH, hence not hemitian. 2. (HH'+ HH) = H°H’ + H'H” = HH + HH’ hence hermitian. 3. HH’ + HH)’ = -i(H"H - HH") = -i(HH - HH’) = i(HH' + HH)’, hence hermitian. 4. HH'=1 = (HH') = HH = H"H =1. WH=1 = H'H=H'H = (HY-H)H=0. But if H” exists, det H #0 and H = H", hence hermitian. eee renee {e"] - ven EE EY . ene = e" hence hermitian.Mathematical Methods in Physics | 227 6. (WHY = HHH = W'HH, hence hemitian. 7. Define H" = H+ H’ (H” hermitian). Then[e”] = e” from 5; hence hermitian. 8. ee = AA’, where A and A’ are hermitian. But from 1, AA’ is not hermitian, so that ee is not hermitian. (b) The necessary and sufficient condition that HH’ and e”-e"” be hermitian is that H and H’ commute with each other. Sufficiency (commutation): HH' = HH. 1. (HH) = HH = HH = HH’. 2. [etet] =[e"] [et] = ete = eter. The last follows since each of the two terms is a hermitian matrix. Thus, they commute. Necessity (hermiticity): HH’ = (HH'Y 1. (HH! = HH! = H'H’ = HH. Thus, Hand H’ commute. 2. [ete] = [eT [et] = ete”. Thus, ee” = ele", and the two terms commute. TIL. Let H be a hermitian matrix and let U be defined by U = (- in) C+ it), where J is the unit matrix. (a) Prove that UH = HU. (b) Prove that Uis unitary. (0) Let Vbe a unitary matrix such that VHV*= A, where A is a diagonal matrix with Ay = 4y. Express explicitly the matrix elements of U, defined above, in terms of A; and Vj.228 Problems and Solutions Solution: (a) Noting that U~= (I~ ify '(I + iff) is a function of H only, ie. U= UE), it can be expanded in a power series in H. I+iHt ae 1+ 2iH - 2H? - 2iH? + 2H + 2iH?..., U and any function of H commutes with H. In this case, it is fairly obvious that UH = IH + 2iH? - 2H? - 2iH* + 2H°... = HU. Alternative (direct) method: UH = (I~ iH)" + iH = (1- iH) HUI + it). HU = H(l- iH) (I+ i). To show that H commutes with (I ~ iH)"', we assume it to be true and investigate the consequences. Let H(I - iH) ' = (1 - iH) 'H and multiply both sides on the right by (I- iH), yielding H = (1 - iH) 'HU - iH) = (1 - iH) (I - iE)H = H, an identity; hence our assumption holds, and UH = HU. (b) In power series notation, U! = I~ 2iH" - 2H"? + 21H”? + 2H" -2iH"...., = I~ 2iH - 2H? + 2iH? + 2H! - 2iH*... = i Hence, . _ [oiH I+in C0 eee emia Alternative (direct) method: U=(I- iH) (I+ iH) = (I~ iH)U = (I+ iH). Since from (a) above we know that U commutes with H, we have (I - iH)U = U(I- iH). Thus, U(I - iH) = 1+ iH. Taking the adjoint of both sides, (I - i)" U* = (I+ i)’, or (I+ iH)U* = I-iH.Mathematical Methods in Physics | 229 Multiplying the original expression for U on the left by the latter term: (1+ ENUU = (1 iH\I- iH) C+ i) = (C+ lU'U - 11 = 0. Since det [I + iH] #0, U'U =I. [If det |I + iH] = 0, (1 + iH) ' would not exist!] (©) VHV' = A, where Ay = Aiby. Applying the power series method again and using the fact that VV = I: VUV = I+ 2i(VHU') - 2(.VHV'(VHV’) ~2i(VHV \VHV \VHV’) +... = [+ 2iA ~ 2A? - 2id° + 204 + - pie “de = (-iA)UtiA). W=VUV = W= (I-iAyU+iA). [a- ww], = + ids = LG, mb ,)My = 85 + Biri ° ' Loin - iA)Wy = 94(1 + iA) > (1 iA)Wy = BL + 1A) ° W= VU = U=VWV = Uy = DV Woe. - Ls ba Bru = a Brn230 Problems and Solutions IV. The differential equation for the time variation of temperature T(rf) due to heat flow in an isotropic medium is given by oT op an 7 BT. where pis a positive constant. (@) Prove that the general solution of this equation is Tet) = ffi) exp (ier - pd k, where fle) is determined by the initial distribution of temperature at t= 0 through the relation 7 (ny Sk) = fi T(r, =0) e**dr. (b) If at r= 0, T(r, t=0) = 5°(r), show that at all later time r > 0: — Te) ~ Gun? exe aut} Solution: (a) Taking Fourier transforms: 1 " Tet) = apasem eek, where 1 ke atk) = aaestee ar, VIR) = ~ Gypelh ety ee. Substituting into the differential equation, we have: lesno Ae =0.Mathematical Methods in Physics | 231 Choose an arbitrary vector ko in k space and perform the following: freeffeaen + au), etek - Ue ee =0 4 f[eecn +802) ara. koe -0. = Retkan + HD -0. oat Since ko is arbitrary, we have (kt) = -pk’g(kt). Solutions are given by: atk) = foe, where fi(k) = g(k, t=0). Defining f(k) = fi(kY/(2n)”, we have: Tet) = ffi) exp (iker ~ pnd. (b) Atr=0, we have from (a) above, that Ter, 0) = fk) ea, and ur, er, Sik) = = = t=0)e "dr ce 5 (e™'dr = a upon setting T(r, t=0) = 8°(r). Hence, Tr) = aha peek (2x 1 sin Bdkdade - owl 2 aan e ie sinaaean232 Problems and Solutions Tort) 2 2_Tsinkr 7 'k (sinkr)dk = - at ark ano --5 2__ || sink ys ~festem ae ur(2n) t jo 7 1 fw “ak Je ™ coskrdk. From integral tables, fe"*™ cosbrdx = & tad =e os Setting a” = pt and b=r, fe™' coskrdk = —*— eA", Thus, — § un? Tet) = few ge 2 aL eran, 4un?t 2(u1)? (Any)?Part Il Chapter 6 Theory of Functions of a Complex Variable 6.1 Theory of Complex Variables 6.1.1 Complex Numbers Defined: w=a+ib; v=c+id. Equality: w = v if and only ifa=c and b=d. Addition: w+ v = (a+ ib)+(c+tid) = (atc)+i(bt+d) Division: = wor! = 2110 _ atibe~id _ actbd, , bead o ctid ctidc-id GaP ed 6.1.2. Complex Variables Acomplex variable Z = x + iy is defined in a z-plane formed by a real axis, and an imaginary axis Fig. 6-1). From the polar form of Z, i.e. from Z= fle” = pe” = pcos0 + ipsin®, and from Z = x + iy, we have that x= pcos0 and y = psin® satisfy the conditions for the equality of two complex variables. The inverse relationships are given by: p=x+y; O= tant, where p = magnitude or absolute value of Z , and 6 = modulus of Z.234 Theory of Functions of a Complex Variable 6.1.3. Addition of Complex Variables Referring to Fig. 6-2, Roa ie, where [R|* = + ff? - 2[Z]fjcos. and where 01 = 0 += 6 + (180 - p-a). RR = @+D@ +7) =-Ptpteee = a+ A (ere . corn) = et + fa + 2] cos(180 - a) = fA + fA - 2] cosa. Fig. 6-2 Hence, |Ri” 6.1.4 Functions of a Complex Variable f@, a function of the complex variable Z , can be considered a vector field in that one can associate a particular f@) with every Zin the Z plane. Since f@) is also a complex function, it can be expressed as: £@ = oxy) + Wey) = FO = ey) ivey). Similarly, in polar (or vector) form, f@) = f@le', where /@I= 19709) + WG) and O = tan”! OD Py)” 6.1.5 Calculus of Functions of Complex Variables We deal first with single-valued functions. 8 Consider f@) =Z= pe’®. From Fig. 6-3 we note that + 4n+8 f@ is unchanged by going around 2nn times. Hence, f@ == pe“®*?", where n is integer-valued, is single-valued. Fig. 63Methods in Mathematical Physics I! 235, Consider the function S@ = nz = in(pe?"”) = Inp + (0 + 2nz). Here we note that f(@) is multiple-valued, in fact. infinitely valued. In order to work with f@), we must make it single-valued (otherwise it would not be an analytic function — analyticity will be discussed later). We do this by introducing a cut, so that we limit values of ® to 0 < @ < 2n (Fig. 6-4). Since one is not permitted to cross the cut, n is restricted to be zero-valued and we can now speak of a single-valued function f@ in this cut plane. Derivatives: GO «jin fEt8D- $B) a zo az independent of how AZ + 0. The independence condition is necessary, since there are many possible directions for AZ —+ 0 (Fig. 6-5). Fig. 6-5 6.2 Analytic Functions A function f@ is analytic at a point Z if and only if the derivative exists at that point. A function f@ is analytic over a region R if f@) is analytic at every point Zin the region. [This tegion may turn out to be quite odd, con- sisting of loops, rings, etc.] Theorem: If f@ is analytic, and if it can be expressed in the form f@ = @(xy) + iy(xy): Se0y) _ Oyiy) — Spy) _ _ Sway) ax ayo ly ox These expressions are known as the Cauchy-Riemann conditions. Hence- forth we will drop the variables x and y in @ and y with the understanding that they are neverthess implied.236 Theory of Functions of a Complex Variable Proof: Since f@) is analytic, AZ + 0 in any direction. For the case where Az = Ax (Fig. 6-6), AD . 0. 20, 20 “oo ax” ax fax’ while for the case where AZ = i Ay (Fig. 6-7), 4). 190). (02,2) 2042 igs & i ow dy ey. oy ay Since df@/di exists and is independent of how AZ + 0, these two expressions should be equal; hence op , ov _ Ow _ oD ax" ae” ay ay’ “I which, according to our definition of equality of complex Fig, 6-7 numbers, implies that Sete) _ SyQy) Spay) _ _ Oyy) | ay ay ax Theorem: (Converse of above) If f@) can be expressed as + iy, and if @ and y obey the Cauchy-Riemann conditions, then: (2) f@ isa function of Z only (ie. not 2"). () f@ is analytic. Proof: Z=x+ iy. Z =x-iy. % 20,28 a oe ,d0 ax Yay” ay lay Forf =/@,2): a Lr ~ 28, ;2¥,Methods in Mathematical Physics It 237 @ F-L. wa Subtracting equation (2) from equation (1), we have: OF _ (82 _ Ov), ; (Ov, 28) - 2m Boi (EZ) - 0. if the Cauchy-Riemann conditions are satisfied. Hence, 2 = Oand f = f@ only. Adding equations (1) and (2), we have: Iz af _ (2, 2), ; (2¥_ 20) 23 (ES) (t-S): Applying the Cauchy-Riemann conditions: @ £94, ax Ox” of _ dy _ 0 O 2 xia Since all the derivatives on the right side exist, 0/62 exists and f@) is analytic. This completes the proof, since we have shown that //0Z exists as ‘we approach from the x-axis and from the y-axis, i.e. independent of how Az +0. 6.3 Applications of Analytic Functions 6.3.1 Electrostatics Consider a charge-free region (p = 0) in which the potentials on conducting surfaces are specified. (We are dealing with a static case for which VxE = 0, since H = 0.) We seek E and 9 in the neighborhood of the conductor.238 Theory of Functions of a Complex Variable Under the static conditions, Maxwell's equations for E are V-E = 0 and VxE = 0. We have shown (Methods in Mathematical Phycics I) that if the curl of a vector vanishes in a certain region, then this vector can be expressed as the gradient of a scalar. Hence, E = -V¢ (the negative sign is used to produce certain desired end results). Thus, V-Vp-=0 = V*p=0. Since V@-dr = dg, then along @ = const., do = O and V@ is perpendicular to the surface. Hence the gradient of a function forms a set of lines perpendicular to the ~ set of lines @ = const. (Fig. 6-8). = const. 2 — = const. 3 Consider now the analytic function = const. 1 {@ = 9 + iy, where @ is the electrostatic potential (E = -Vg). Then, ap a Fig. 6-8 ; (Op Oy ow EV = “lar ax * ay ay -— w = const. oy dy _ oy dy |g dy ax ox dy” since @ and y obey the Cauchy-Riemann conditions for an analytic function f@). Thus E-Vy = 0, which implies (see Fig. 6-9): Lines of E are perpendicular to lines of Vw. Lines of E are parallel to lines of y = const. Lines of E are perpendicular to lines of @ = const. Hence, lines of y = const. are perpendicular to lines of p = const. Fig. 6-9 In the Problems it is shown that if f@ is analytic and can be expressed as f@= 9+ iy, then Vp =0 and V’y = 0. Furthermore, since E is ‘everywhere on the surface perpendicular to the surface, and since @ = const. are lines perpendicular to E, then @p = const. on the surface. Hence if you specify an analytic function f@ = ¢ + iy, this is sufficient to determine @ (and from this -V@), since @ satisfies Vp = 0 and @ = const. on the surface of the conductor. [There is a method, known as the Schwartz-Christoffel method (not to be studied in this course) for generating any desired analytic function, given the conducting surface.]Methods in Mathematical Physics Il 239 Special cases: Case I: We state that f@) = 47" isa solution to the problem of Fig. 6-10. * Proof: Note first that f@) is analytic. Then ] express it in the form @ + iy. 7 7 Fig. 6-10 LO = 5 prem = 7 P (cosn6 + i sinné) = Ap" (sinn® ~ i cosn®), ao ‘w= const. so that p= Ap” sinn® and (parallel to E lines) y= —Ap"cosn®. Note that along 0=0,p=0 while along 0 = /n,p = 0. Hence, @ = 0 along the length of the conducting surface. Remembering that lines of @ ate perpendicular to lines of y, we can map the lines of @ and E as shown in Fig. 6-11. 9-0 Case II: As a special case of the above, set n = 1 (Fig. 6-12). Fig. 6-11 $ ~ 42 = 4 pe® = Apsin6 - iApcos® oa = Apsin® = Ay. Fig. 6-12 Hence, lines of p = const. are ; represented by y = const. fi — lines of while lines of y = const. a = const. are represented by x = const. (Fig. 6-13). \ ines of Furthermore, @ = const. ee oo. 0 Fig. 6-13 = ig. 6-1 =-vAy=-ay, ° *240 Theory of Functions of a Complex Variable Case III: Asa special ease, set n= 2. f@ = 4@ . 4 H(cos20 + isin20). p = Ap’sin26. y = ~Ap*cos28. Lines of ~ = const. are represented by Ap’sin20 = Ap”2sin@cos® = 2Axy = const. (hyperbolas). Lines of constant y are represented by ~Ap”cos20 = -Ap?(cos’@ - sin’@) = ~A(x” - y*) = const., or x* = y? + const. (equilateral hyperbolas) Fig. 6-14). e-0 Fig. 6-14 Case IV: For the case n = 1/2, 1/n = 2n and the problem reduces to finding the solution for a knife edge (Fig. 6-15). @ = ApMsin (92). Gporsss v = ~Ap”cos(0/2). Fig, 6-15 Fig. 6-16 Lines of p= constant are represented by psin” @/2 = constant, or p= constan(1 - cos®), while lines of constant y are represented by pcos”@/2 = constant, or p = constant((1 + cos®) (Fig. 6-16).Methods in Mathematical Physics II 241 6.3.2 Hydrodynamics This is the case in which there exists a body about which there flows an incompressible fluid. One further specifies that there be irrotational motion (no whirls). Incompressibility implies that dp/dt = 0, or op , 2p dr , ap dy, op de _ op, a Oxdt dydt ad or Since the equation of continuity demands that apfet+Vipv=0 = dp/et+Vpv+ pV = 0, it follows that the incompressibility condition says that V-v = 0. Vpv. For irrotational motion, Vxv=0 = v=-Ve, which from V-v = 0 implies V7 = 0. Note here the important difference from the electrostatic case. Whereas E was perpendicular to the conducting surface and ~ = constant on the surface, for the hydrodynamic case V is parallel to the boundary and |v|= constant on the boundary surface. Since v is parallel to lines of y = constant and perpendicular to lines of @ = constant, we again have lines of @ = constant perpendicular to lines of w = constant. V lines Mapping of lines of v and g: To solve the hydrodynamic problem we seek a @ satisfying V¢ = 0 with lines of y = constant parallel to the surface and lines of @ = constant per- pendicular to the surface, in addition to v = -V@, such that |v| = constant on the surface. (Fig. 6-17). Fig. 6-17242 Theory of Functions of a Complex Variable We claim the solution to the hydrodynamic problem illustrated in Fig. 6-18 to be S@ = AZ = Ap"cosnd + iAp’sinnO. @ = Ap"cosn®. w = Ap’sinnO. Fig. 6-18 Note that along the boundary surfaces y = 0, while for other regions of space, lines of constant y are as shown in Fig. 6-19. Lines of v are parallel to lines of constant y. p= const. Z| | y=const. (v lines) Lines are seen to be exactly reversed as compared to the electrostatic case. Fig. 6-19 Special Cases Case I: For case of n= 1, v lines f@ = Apcos® + iApsind. | @ = Apcos® = Ax. lines AN | EG EH ca Mapping of lines of @ and v is shown. Fig. 6-20 in Fig. 6-20. v = Apsind = Ay. v = -Vp = -Aio. Yea Case II: For the case n= 2, 1/n = 7/2, and f@ = Ap’cos26 + iAp’sin26. @ = Ap’cos20 = A(x” -y’). w = Ap’sin20 = 2Axy. Vv = -Vop = -A(2xio 2yjo)- Lines of ¢ = const. give equilateral hyperbolas while lines of y = const. give hyperbolas (Fig. 6-21).Methods in Mathematical Physics It 243 Case Ill: For n= 1/2, $@ = AZ” = Apcos0/2 + iAp“sin®/2. @ = Ap cos/2. y = Ap™sin0/2. Lines of ¢~ = constant = lines of p(1 + cos@) = constant, while lines of y = constant = lines of p(1 - cos®) = constant (Fig. 6-22). Lines of @ = const. eee 6a = 2p (Po cost 6, sin@/2). A =\e--45 > =, Mose At@=Oand r, vis in the ~po and +8, direction, respectively (i.e. the fluid goes whipping around the sharp comet). Consider a function f@ = InZ of the type previously discussed, which at first does not appear to be analytic, that is, its derivative does not exist everywhere, i.e. df@Ydz = 1/Z, which does not exist at the origin. We seek to evaluate HD) _ 5, {E+ 8) - FO & a aro244 Theory of Functions of a Complex Variable Considering Fig. 6-23: Z+ Az = pe®. fE+ AZ) = In(pe*) = Inp + id. S@ = In[p'e™] = np’ + i2n-9). P-8 @ — fE+AZ-f@ = In(o/p')+ 0+ p- 2m). Fle. 623 Introducing a cut at y = 0 so that AZ cannot ctoss the x-axis, we have from Fig. 6-24: S@+ AZ = In(p’e") = Inp" +iw. 95 S@ = In[p'e*] = np’ + i2n-g). © (&) {E+ AD-f@ = In(p7/')+ (w+ p-2n). In the limit as AZ —+ 0, (a) and (b) yield different values for df@/dz, so that the 2% function is not analytic. By introducing the cut at the x-axis (y= 0), we limit f@) to be single-valued and, in addition, we make f@) = In Z analytic inthe p’ = —j —— cut plane 0 < 6 < 2x. Some examples will now clarify this discussion. Fig. 6-24 Case IV (Electrostatics): Consider the case where two different conducting surfaces make angles with each other (Fig. 6-25). We claim that a solution to the problem is given by: §@ ~ 4inz = 41n (pe) = 4inp +40. as i insulation @ = Ad. w= -Alnp. Fig. 6-25 Lines of @ = const. are given by @ = const. while lines of y = const. are given by p = const.Methods in Mathematical Physics II 245 At =0,—p=0, and at 0 0,, = AOn- . L_# 9 lines 7 i Also, : ! 2 = E lines oe p 08 op Mapping of lines of E and @ J p= Aba is shown in Fig. 6-26. Case V (Hydrodynamics): oS The corresponding hydrodynamical case for which f@) = (A/)Inz represents a solution is given by observing that one wants y = constant to yield lines parallel to the surface boundary, while one wants lines of ¢ = constant to be perpendicular to the surface boundary. Observing these lines in the Electro- static case above, it is obvious that the corresponding physical problem is as shown in Fig. 6-27. At @=0, p= 0 while at 0 = 2n, p= 2nA. Note that although there is a discontinuity in @ at ® = 0, v=-Ve =— (4/p)Bo, and is continuous at all points. The cut at @ = 0 is necessary so as to be able to utilize £@ > ona. ee 6.4 Integral Calculus of Complex Variables ‘Gf@cz is defined as the line integral of a function of a complex variable T over a path I’ in the complex plane (Fig. 6-28). f@)dz is designated as the contour integral over the path Cin |” the direction shown (Fig. 6-29). ef Fig. 6-29 Fig. 6-28246 ‘Theory of Functions of a Complex Variable 6.4.1 Cauchy Integral Theorem Theorem: If f@) is analytic on ang within the contour C, then the line integral of f@) over C is zero, ie. f@dz = 0. Cc Proof: Since f@) is analytic, derivatives of @ and y exist. freode - fe + ivyar + id) - fede vay) + if cya + 90). The general expression for Stokes’ theorem (Math Physics 1) is given by YL feuZ ok S: - For Ta scalar, this reduces to Y fea Seas = feds fasxve = fous. YL Stisnds:. Note that the only component of dS that exists is the z component , i.e. dS = ddyko. It follows that = - fatxdy 22 . oy Seas = -fazay 2. Guy = fay SE Hence, foe - I-F- Shoo DF Se Blt = 0, since @ and w satisfy the C-R conditions. c Corollary 1: If f@ is analytic on and within a closed curve C, and if 21 and 2; are two points lying within the region R bounded by C, then J bende is independent of the path. Fig. 6-30 Proof: Referring to Fig. 6-30, by Cauchy’s theorem: fred - fred frome - 0 - fox - fox. i.e. independent of the path.Methods in Mathematical Physics II 247 Corollary 2: If a simple closed curve C? lies entirely in the region bounded by another closed curve C1, and if f@) is analytic on Ci and C and in the region between C; and C2, then pone Proof: Referring to Fig. 6-31, introduce a cut I such that the two (dotted) lines are infinitesimally close to one another. Then, by Cauchy’s theorem: fpeade + freee + free free =0. Hence, Fig. 631 fee - from. (Both C; and C2 are taken in the counter-clockwise direction.) 6.4.2 Residue Theorem Theorem: If f@) is analytic on and within a closed curve C, and if isa point lying entirely within the region bounded by C, then @-a)" Proof: From Corollary 2 above,f =f. Define e(Z) to be ta f@, e@ = @-a) a function of Z on the circle C; of radius 5 centered about point @ (Fig. 6-32). oo oo ligee® ~ fin P=y-1 « 0. Hence, e@)—0 as 80. Fig. 6-32248 Theory of Functions of a Complex Variable Substituting for f@): [os sme =a Max —2) wih “aa = x {I+a+m), where 1 0§ 5%. 0 = fecoee. m ~ seofez. ‘Term I: Since e(@) is analytic on C; and on C and in the region between C and C1, 1 =f edz =f e@dz. Since e()-+0 as 5-+0, e(@) is analytic over t the entire region within C. Hence, II = 0. ‘Term I: [can be evaluated using vector substitution (refer to Fig. 6-33): @-a) = be. dz = Bie*dd. a = @-a T= fog es = iffdo ~ 21/0. Fig. 6-33 Term II: Similarly, 1 m = seag dz = f@pife"do = flasfe"| = 0 Collecting terms, {28 - 10 - {28Methods in Mathematical Physics It 249 Theorem: If f@) is analytic on and within a closed curve C, and if Gis a point lying entirely within the region bounded by C, then f@ - ES G-a Proof: Define a function 8@ = p@E-a) = f@, @-a) defined on the circle C1. In addition, lim g(@) = 0. Substituting for f@: (@) + p@E- ay +f@E- 3} : G-ay = [I+m+m], where ad & fa-a Noha ey 1 = fea. n-f@ TermI: As 5-0, foc = {225 5» fae@iae = 0, since g@) + 0.as 8-+0. (Also, p@) is analytic within C1.) Term Il: (Z- “455 = Collecting terms, Af fot _ : foe, 2nid, @- ay ¢@-ay250 Theory of Functions of a Complex Variable Theorem: If f@) is analytic on and within a closed curve C, then f(Z) is analytic within C. [Note that one usually states conditions for the contour and the region enclosed, and ends up with a result that holds only within the region.] Proof: When one states that f'(@) is analytic one means that F@ ~ tinLED-LO AO h exists and is independent of how h -> 0 ig. 6-34). From the theorem proved above: fG+h) - nilge se . D> ab Oe f'@ = nh Ce Fig. 6-34 @-a-hy “cal im 1 f LQ2G-@) - A amfare ne-a@ ky anid icf S@dz flee |: i ee ioe Pasa be-w WG-2) Nea ew Since @ and (4 + h) are within the contour C, and since Z is evaluated on C, ~a- fi and @- 0? are finite, Hence, f ——/@2 is finite and G~- hy’ and & - @) are finite. ef Zaigoae has some maximum value. Therefore, in the limit as hi + 0, this term vanishes, and we are left with 1 f_[@a (ni) @- ay” f@ - independent of h.Methods in Mathematical Physics Il 251 Theorem: If f@) is analytic on and within a contour C surrounding a region R, then all of its derivatives exist in R and are analytic. £@-s5 mf a Proof: The proof follows from the previous theorem in like manner. 6.5 Taylor’s Theorem Theorem: If C isa circle of radius R with @ as center (Fig. 6-35), and if f@) is analytic on and within C, then {@ = 10+ G-ay@+@-77LOs.. - FL@g-ay, for all |(@-@| 0) about the origin (Fig. 6-36), f@) is analytic on and within C. Thus, f@ = (1 - 2) | will be expandable ina power series - L£Me- Yor. € c 5 i ao Fig. 6-36Methods in Mathematical Physics II 253 6.6 Laurent Theorem Theorem: If {@) is analytic on C and on C1 and in the region between them (Fig. 6-37), then for Z = (@ + h) @ lying in the region between C and C1): f@ = fa+h) Z — @) + =? “Ye . Le ay Fig. 6-37 where - + f_foa_ Oo Sf GO b= Gf YOR Note that if f@ is analytic within C, as well, then by Cauchy's theorem, bm = O since (F- a) is finite while f@) is analytic. Note also that in this case the cn reduce to f"(@Yn!, so that fark - ¥P@e-ay, which is identical to Taylor's theorem. Proof: Introduce a cut (Fig. 6-37), so that f@) is analytic over the entire curve (C+T-C,-T). Since f@) is now analytic over the contour and within the contour, we have a+) - wile a-h) “aha a-h)* ‘We now apply the expansion we used before in treating Taylor's theorem. For the first integral (F - @ > hi): N 1 _ Kr ii ean Lear ea ea254 Theory of Functions of a Complex Variable ee wl are ew wana a-hy ee fiat Tl aareaa|: y a+ hy = Y coh + Since the absolute value of a sum is always less than the sum of the absolute value, ae mil ga a-a Fre a-h) a a-A)| Since (f - @ - h) remains finite no matter how close to the boundary (a = h) gets, and since f( is analytic and bounded on C, we have that |—fOdt_| <4, Thus, since h< R, @-a-h) ze am ey sa fGa ea ~a"G-a Sond, fi 1 1 For the second integral ( > 7- i * Gaon hoGca) hoo gr Rh-@-al)" aaa ¢-a" fash afrmaly ” R'e-a- BD) wa fart aafray z ve syrt As above, on C1, (7 - @)**! = RE!" = RN"!, while {dil = Redd. fai een < andMethods in Mathematical Physics Il 255 Hence, since Ro N, then point Ziis a pole of order N. A “simple” pole is one of Fig. 6-38 order 1, in which case we have only b/(Z ~ @). (b) If /@ is single-valued and point @ is a singular point, and if the series bn/@ ~ @)" is infinite (i.e. one cannot find an N such that (a) above holds), then point @ is called an essential singularity of f@). (©) I£f@ can be expanded in a Laurent series about a point @, and if cn = 0 for all n > N and cw + O, then point a is defined as a zero of order N.256 Theory of Functions of a Complex Variable The behavior of f@) near an essential singularity can be quite different from the behavior near a pole of finite order, as Z —> 2. For a pole of finite order Nas Z > @: f@ — byz- a)". For an essential singularity, one might get the results 0, 1, , etc. Consider, for example, the function iy rama . ter o” which we note has an essential singularity at 2 the origin (none of the bn vanish). Approaching, the origin from 3 directions (Fig. 6-39): eg x © aty=0, Za: ee + om, Fig. 6-39 =o P Aty=0, Z-x: Me — 0. atx=0, Z=iy: |e] + [e| 3, 1 Consider the point infinity as a singular point. If one makes the transfor- mation Z—+ 1/2’, then the singularity at infinity is transformed to one of finite consideration. If we define f@) = f(/) = g(@), then the analytic behavior of g(@’) at Z’= 0 will give the analytic behavior of f@ at Z= =, thus yielding a direct correspondence between infinity and the origin. (a) IfZ’= 0 is a pole of order N for g(Z), ie. all bm vanish for m > N, then 7 = © is defined to be a pole of order N for f@, where JO - Ynz +S, and the principal part of f@ for a pole of order N is given by YS onz" (as 2 0,f@ bw2"). a (b) IfZ’= 0 is an essential singularity for g(Z), then Z = © is defined to be an essential singularity for f@). (©) IfZ’= Disa zero of order N for g(Z}, then Z = © is defined to be a zero of order N for f@).Methods in Mathematical Physics Il 257 6.8 Liouville Theorem Theorem: If f@) has no singularities anywhere, including © (i.e. f@)| < K everywhere), then f@) = constant. [This is equivalent to stating that the potential is finite and bounded in a charge-free region, in which case V is constant everywhere.] Referring to Fig. 6-40: c ee anid $0) = eh PS In the above, E is evaluated on the circle C. Fig. 6-40 We would like to show that f@ =f@). ©@-z Proof: Consider f@ - f@) = mite ~Z\E-D" Since by hypothesis f(@) is analytic everywhere, choose some convenient origin O and expand the oe such that, if R is the radius, then \€-2| and ((E-2Z)| > =. It follows that, as R + ©, 2nKRE-2Z] _ 4KE-21 _ 4 R R V@ ~ f) Ag oa €-D Corollary: If f@ has only a finite number of poles, including ©, then f@) is a rational function (ie. a finite fraction of finite polynomials). Proof: Let ai @ ... Gy and © be poles of order Ni No... Nv and Ne. Consider the principal part of f(@) (PP,) at one of the poles, say ds (s=1,2...n).258 Theory of Functions of a Complex Variable Similarly for the principal part of f@) at infinity: » = Ybuz. at Define the function > ee rim @-a) om Note that there are no poles of g(2), even at infinity, for at any Pole of f@, the principal part of f@) will be canceled by the corresponding 5 a> = @- ai and similarly for the pole at infinity. Furthermore, since the summations do not extend to infinity, we have no essential singularity either. In fact, the function g(2) has no singularity everywhere and thus by Liouville’s theorem, g(@) = constant. Hence, x. so ~ cr Fhe ~ You, st et @- and is therefore a rational function. Converse: Any rational function f@), expressible as N Yar f@-=—, Yer" has the property that for N # 0, the numerator has a pole of order N at infinity, while the zeros of the denominator yield poles of order N. Hence, any rational function has a finite number of poles (including infinity) or else is a constant. If f@) is not rational, then it must have an essential singularity, and vice versa.Methods in Mathematical Physics It 259 6.9 Multiple-Valued Functions Consider f@ = InZ = In(pe®) = Inp + i0, and refer to Fig. 6-41: At point P,8=0 — f@~=InR. P At point P, @=2n — f@=InR + 2ni. Fig. 6-41 Thus, point P is seen to be multiple-valued for f@) = InR + 2nik, where k=0,1,...% One can reduce the multiple-valued function to a single-valued one by the introduction of a cut extending from Z = 0 to infinity in such a manner that @ not be allowed to rotate more than 360°, Then, considering f@) at any point P (ig. 6-42), f@ = InR + 18; and is unique, since we cannot cross the cut in performing the rotation. The sacrifice R cut one makes in reducing a multiple-valued ae : ranch function to a single-valued one is that NG . oe point one loses continuity across the cut. ee Hence, for the function f@ = Inz, with a cut extending from Z = 0 to infinity, one now has a single-valued function, but one which is not analytic along the cut. In the remaining region, all properties of the single-valued function f@) are retained. If analytic in the remaining region, then one applies those previous theorems dealing with analytic functions. On the other hand, if certain poles exist, one can apply those theorems relating to poles, etc. The point Z= 0 is called a branch point. It represents a peculiar type of singularity in the domain of multiple- valued functions. Consider the function f@) = 2%, where «is not a positive integer (else f@) would be single-valued at @ = 0, 2n, 4n, etc.). ForZ= Re®, f@ = R°e™. Note that once again one has a multiple-valued function, since At point P,6=0 — f@ = R*. At point P, @=2n > f@ = R%e™™., Hence, a cut is again required, with the same considerations applying as above.260 Theory of Functions of a Complex Variable Riemann Surfaces Riemann _ Surface 1 Assume that continuity is demanded, i.e. one allows multiple values. For the case a= 1/2, f@ =2" = Re. Considering points (a) and (b) of Fig. 6-43: Fo = RY. oo In order for f@) to remain continuous, we must define so-called Riemann Surfaces, such that when @ reaches point (b), we move on to a new surface, Riemann Surface II, i.e. (by' is the continuation on Riemann Surface II of point (b) of Riemann Surface I (Fig. 6-44). P . Riemann S@oy = -R”. Surface IT fOr = RV = -RV°e? = -f@p. Oe = RVe™ = RY = fDey- After reaching point (b)’, one continues around to point (c), then returns to Riemann Surface I. In this way f@) remains continuous. The combination of the two Riemannian sheets yields a function f@ = 2"? which is single-valued on any one sheet, and is a continuously varying function. (by’ O=2n. ~ © Fig. 6-44 0-40 For a= Yn, f@ = 2 (n> 0), we will require n Riemannian sheets, since we will have to revolve n times until the same function returns, i.e. At(R,) f@ = Re. AtO=0, f@=R”. At = 2nn, f@ = RMeMC™ = RM If ais an irrational number, we would need an infinite number of Rieman- nian sheets. The In function similarly requires an infinite number of sheets, since f@ = InR + i, and no matter how many revolutions one makes, one can never revert back to f@) = InR, corresponding to @ = 0.Methods in Mathematical Physics I! 261 6.10 Theory of Residues Definition: If a function f@) has a pole Z of order n, and if where @(2) is analytic at Z = Z:, then by (the coefficient of AZ - Z1), is defined to be the residue of f@) at = 71. Theorem: If f@) has, as singularities, only poles 21, Z2,...2n inside a closed curve C, then Sf@dz = 21 YR, c oe where Rjis the residue of f@ at Z = 7) (j= 1,2,...n). Proof: Referting to Fig. 6-45: f-£-Ef-0 faye ON t A since f@ is analytic over the region 2) between Cand EG). Hence §- Lf. 2° ad Considering one of the Cj, say C1: Fig. 6-45 fro - forma} a G a Since ¢@) is analytic at Z ~ 2, the frst integral on the right vanishes, by Cauchy’s theorem. Setting Z - Z1 = ce®, dz ~ ice'd®, so that frese - fd - Par - = eee © )2niby = 2niRj for m = 1.262 Theory of Functions of a Complex Variable 6.10.1 Complex Integration 2 Type: fFicose, sin8)d0 Under what conditions can we use the theorem of residues? We want to convert this type of integral to a contour integration over the unit circle. Over the unit circle, [_] = 1,2 = e®, dz = ie'd8 = Zd®, We also have: cos = (e +e ®y2=(%+ 2/2; sind =(e* - ey2i=@- 1Zy2i. Po Thus, [F(cos®, sin6)d0 = if F [. ealé . d 1 1g Qe Kann 2 circle where f® - sate VA, sind = 3. @- va]. We then proceed to investigate 2, looking for the poles within the unit circle. Example (a) Jo*eoatra -sin®\d8 or (b) fem sin(n0 - sin0)d0 , where n is a positive integer. Observe that fe £98 -Kn8 - 509) has a real part identical to (a) and an imaginary patt identical to (b). Over the unit citele, 2" = €°™®, nz" 'az = ~ine™*a — eap = 1 e Thus, x cos - sind) ay _ 1 goostrisinn ZL petdz _ 1 i ea = | fen = Soe = Sse. cme ae Note that f@) has a pole at Z = 0. We seek the residue at this pole . The pro- cedure is to expand f@) and select the coefficient of the 1/Z term. 1F5Zz . 1 ri riba - Residue = 77 Thus ae je cos(n8 — sin®)d0 = = Jom*eincno ~ sin®)d0 = 0.Methods in Mathematical Physics II 263 Type lt: fe" Fede. This type of integration occurs quite often in the use of Fourier integrals: 1 i ees fe) aie lets dk; 8) = Ga ene ae. ‘We seek the requirements for transforming the given Type II integral into one involving contour integration. Jordan’s Lemma: Let C be a contour along which Z = Re®, where Ris a constant and where 8 runs from 0 to x. If |F@)| <|1/Z"| for all [2] > Ro (where m > O, even fractional), and k > 0, then Jmfe"r@e =0. Proof: Over path C, Z= Reos0+ iRsin®, so that [| = [ttn = ete = gi, For R> Ro, [ferrede| < fletreona < rere nan < af Ree Ne Since sin® > 0/2 for 0 < @ < 1/2 (Fig. 6-46), 6 \ z, "eH Rado Fig. 6-46 a ig. F@d| < 2 fe lf | <2) Case (a): For k > and finite, jim eR = 0, so that [[e#rexe| = OasR— ©, Thus, we choose the fc contour for which @ runs from 0 to x (Fig. 6-47). Fig. 6-47264 Theory of Functions of a Complex Variable Case (b): For k= 0, aod - [free |, ana it we require that ic ic F@I< || forall R> Ro, ” - Rdo [fetreaz <2f% 0 fore > 1. [A fractional value for e < 1 means that [frexe| = 222s R +0] Thus, for k= 0, we can use the same contour, ie, from 8 = 0 to m. Case (c): Fork <0 (say k= -a), ok i «ai, ee WE aRsind [ferred | < fern < 2f— ee e 3 ” < Bfemna «A. omy” Note that if we take the contour integral from @ = 0 to 7/2, we get spele*- 1] = © asR+—, However, if we take the contour integral from @ = 0 to ~1/2, we get feet] = OasR— oR" Hence for k < 0, choose the contour in the lower half-plane (Fig. 6-48). Fig. 6-48Methods in Mathematical Physics II 265 The last step follows since the sin function is even, i.e. replacing x by -x: “Pagan - ita ta Ca) Referring to Fig. 6-49: ~~... &@ __ 3” 1 as 1 abt J} Fig. 6-49 SiN ay = lim 2e > 0. way oo Hence, as € > 0, 1 "Fsinx sing sink 4. 2mm x ae a diss a. Note that the imaginary part of ieee is the integral we wish to evaluate. Therefore, let us consider fe & over the contour C (Fig. 6-50). e& CG a R ~~, and noting that = has no singulatities inthe region, ye £0 2 ze =0. -R € € +R < Fig. 6-50266 Theory of Functions of a Complex Variable Furthermore, since in this case k= 1 and F@) = 1/7, so that the Lemma is satisfied, we have that rr ife#-o - Tah e J] so that (in what follows Im means the imaginary part of) Fs as ~ 1 fsb gta a 7 te oni ‘Residue at = 0) om ae NIA Example (ii) Anatom starts to radiate at time 1 = 0 (its vector potential at that time is Ao). We seek the expression for the vector potential A (r?) at a later time ¢ at a distance r from the the atom. The solution is given by the expression: gece: 4m » Ale oM(@ — Do + mn” where @p = natural frequency of the atom; Y= lifetime of the atom (necessarily > 0); Ao = the vector potential at time t = 0 (constant vector). We consider the @ plane and note that 1/r(« ~ @o + iy) has a singularity at © =~ iy. (a) Fort<1/c (i.e. r > distance covered by the speed of light in a time t), eit) = with b> 0. Since F(w) = 1/o, the Lemma is satisfied, and if we choose the upper contour, then 1 fede iodo C(@- Oty)? - to. where C is as shown in Fig. 6-51. < > Fig. 6-51Methods in Mathematical Physics I 267 do = ani DR - 0, since there are no poles within the contour. This result implies that wave fronts cannot travel faster than the speed of light. (®) Fort > 1/c, we choose the lower half-plane, @ running from 0 to -1/2: again, from the Lemma, the integral over C vanishes, so that fem 2MiAe ,-iw4e-r6)- He) 8 Aef Te @. + Ayr? | which we recognize as the expression for a travelling plane wave with an amplitude 2niAq/r and with an exponential damping factor, i.e. for t > 1/c, and y > 0, we have a damping factor e %”), Type II: fereae, ° where a. is not an integer, F@ is analytic on the positive real axis, and where Z*F@) + OasZ + Oand f] + &. [Note that since 2 is multiple-valued (@* = R“e!®® and has different values at @ = O and @ = 2n provided ais not an integer), 2*' is multiple-valued as well] If F@ has a pole or an essential singularity on the real axis, then fas no meaning. It can still have meaning, though, if F@) is not analytic 0 on the real axis. Consider the Z plane. Note that 2 YF@ has a branch point at Z = 0 because Z* has a branch point at Z=0. Introducing a cut from Z= 0 B toZ = (Fig. 6-52): OL 2 fern B ana -{f y flere. Fig. 6-52268 Theory of Functions of a Complex Variable = fereje =0, PR x a as [| + &, since by hypothesis, [2"F@] = 0as[Z] +=. x - [lzrojae 0, as [| + ~, since by hypothesis, [2"F@] + O.as [Z| + ~. There remains f LR Od = fred fir reve. ; fa Along AB: 6 = 0, 2° PF@ = x*" F(x). Along A’B’: 6 = 2n, 2° YF@ = x% 1e*"F(w), Thus, fer @ da - fered fered, at , ferred - aftr. Example (i) jn 1 de 5 where 0< a < 1. Note that 1 Eid Z has pole at ete ee 0, and that -fy ros om , ie. we havea Type I integral. Thus, 1 F@d = f= Fa = = 2ni(-1)"! = 2nie™ = -2nie, and 1+x pe je dx _ ~2nie™ _ _2niMethods in Mathematical Physics Il 269 6.11 Analytic Continuation Suppose a function f@) is analytic at Z = Z (Fig. 6-53). Expanding f@ about point Z., 90 ~ Yaue-2y. If point Sis the nearest singularity, then the series has a radius of convergence equal to(-%). Fig. 6-53 The question to be considered is the following: Can one find a function which will be analytic over a region greater than the radius of convergence of f@? Consider a point Z; lying within region R (Fig. 6-53). According to Taylor's theorem, one can expand f@) in a Taylor series about point Z;. The radius of convergence will now be (5" - 21), where 5" is the nearest singularity to a. Thus, for a function having an isolated singularity, one can cover the entire domain, except for this isolated singularity, by the method of analytic continuation. For f@ = 11-2) and analytic at Z= 0, Le pezegers.. - Ee, valid only for Z| < 1. Let R (Fig. 6-54) be the region of convergence [Z| < 1. Fig. 6-54 Expanding f@) about the point x = -1/2 (p = 1,Z= e” on the circle): FO = Foy + fae + y+ ECE 12Y, - FLEE ay nt270 Theory of Functions of a Complex Variable Evaluating term a term: f@ = 2.4 J e+ +3 G+ pyr... [Note that the very same result could have been obtained from the following perfectly calculable quantities: fo-Yz. pew = Lewy. = Beer pea = Yacyay. = Ynene®. preva) = Yee). etc] Within the circle of convergence i] < 1, and rewriting: a 1 i 1 27,_ @+ipy PR 3 Ea le -3[.85, Gray, ery, xp Gy = Gy’ which is seen to be the same series expansion as the an above, namely, N 1 ®Y, \ Ee fo - YPC r : E+ 12)| \ a and which converges for | vA < lor for \ 7 B+ |< 32. 3 Fig. 6-55 Hence, the new radius of convergence = 3/2 (Fig. 6-55).Methods in Mathematical Physics It amt Theorem: If F; and F are analytic in a connected region R, and if Fi = Fz in region Ri , which is included in R, then F = Fz everywhere. Proof: Define F\ - Fz = f@) to be analytic in region R, with f@ = 0 in Ri (Fig. 6-56). Choose a point @ within Ri and draw a circle tangent to Spat any point, say B. Since f@) is analytic on and within curve C, we have that f@ = ).axZ- a)", = Fig. 6-56 Since point @ is in region Ri and f@ = 0 in Ri by hypothesis, all an = 0, : so that f@) = On C, or Fi = F2 within C. By selecting some new point bin region C, we extend by analytic continuation and find, in this manner, that f@ = OinR, of that F; = FyinR. What does this mean? Suppose that the analytic continuation of the function f@ = greets. = Te beyond | < 1 yielded a particular function G@). In a connected region that excludes Z= 1, G@ and (1 ~ 2) are analytic, yet are equal to each other within a restricted region [|< 1. The above theorem shows that this G@) must equal 1/(1 - 2) over the entire connected region. In fact, any method of analytic continuation will always yield [1 - 2). 6.11.1 Gamma Function (I) The Gamma function '@) is one that will be shown to behave like n! at certain isolated points. I'(Z) is defined as: T@ = Jomra. T@) converges for Re() > 0, and is of the Type III integral discussed above.272 Theory of Functions of a Complex Variable For Re) > €> 0, the integrand is uniformly convergent, and we can differentiate under the integral sign: . positive @ T@ = fe" dn ae, ae ES which exists for Re(Z) > € > 0. Hence, P@) is analytic over the positive (2) half-plane (Fig. 6-57). For Re(%) > € > 0, partial integration yields: rest ~ Jetta = |-e'# [ez feF a, ae But for Re(Z) > 0, |e“, -+ 0, so that T@+1) = zI@. Evaluating: TA) = fear = |e" T(2) = 1TQ) = T(3) = T(2+1) = 21) = 21 = 2! T4) = TG +1) = 31(2) = 32-1 = 3! Tr) fa ( Sage We now proceed to the problem of extending P(@ into the Re(Z) < 0 plane. Consider a point Z located in region I, the strip between Re(Z) = 0 and Re(z) = -1 (Fig. 6-58). + 1) then locates a point in the Re) > € region. But since '@ + 1) is defined and analytic in this region, '@) is analytic in shaded region I (excluding poinrZ = 0).Methods in Mathematical Physics Il 273 Repeating the process, locating a point in the second shaded region II, T@ = 217E+1) - =o re+2), (z+ 1) so that for Z # 0, Z # ~1 and for analytic, [(@) will be analytic, By this method we succeed in expanding the region in which I(@) is analytic to all but the points Z = 0, -1, -2, ... -m..., this analytic function ['(@) reducing, in the region Re(Z) > 0, to fe" Yat. > 3 To determine the behavior of [(@) at points -n, let Z=-n + where d is very small (Fig. 6-59). Note that the point (-n + 3) is related to a corresponding point in ~~” the region where I(@) is analytic. Fig. 6-59 T@ = 21@+1) = pled = = ——— ___reensty 2+ IZ + 2)...Z +n) 7 T@+1 fim Tn +8) = fim Cyn + 8+ 1). + OS Se ) ~ €y"Ta) _ 1" Cmycn+ )..CDb nib nid” Hence, at points -n, T(-n+8) = cy We seek an expression for I'@) at Re@) < 0, one that is not obtained by the use of the extended expression '@) = @+ 1)E. For Re(Z) > 0, we have P'(@) = fe Pdt. We consider the complex plane, noting the fact that e“ is not analytic in the plane.274 Theory of Functions of a Complex Variable We therefore introduce a cut extending from 0 to infinity (Fig. 6-60). @) fema = fe" a = T@, c : since along CD, 0 = 0 and7= file” = ff. P : fer ar = -Jererar - —fomnemerar - -2*r@, 4 : since along AB, 0 = 2n and 7) = iF P™E), ‘ _ % © fet Par = fe"e™ ptiaa = aft fexp(-Me")e""a9 = 0 = : 3 as [| + 0 for Re) > 0. Hence, Pat = (1-7 @), one “ined @ 1@~ a fea. e" £acp 1- We see then that I(@) is defined for all Re(Z) > 0 except for integral values of Z (i.e. when e” = 1). Since we want I(@) for Re) < 0 2S as well, let us consider the deformed path (DP) shown in Fig. 6-61. For this region, odee 7 LLY: Thus, equation (d) above holds for non-integral values of Z and for all Re(Z). Since the integrand is uniformly convergent, one can differentiate under the integral sign. CQ Fig. 6-61Methods in Mathematical Physics It 275 ; L_ paps 2H fren, TOQ-= Infdt+ ‘dt ON = ff a= em? £ = | fee ae 2H 2 aed all oO : and exists; hence '(@) is analytic over all Re(Z) except for integral Z values.Chapter 7 Theory of Ordinary Differential Equations 7.1 Ordinary Differential Equations in Physics 7.1.1 Vibrating String Consider a fragment of string of length ds = dx, where we assume the slope of the string to be fairly small (Fig. 7-1). If p(x) = linear density, T(x) = tension in the string, and F(xt)dx = external force in the y-direction acting on dr, then the total force acting on length dr is given by F(at)dx + [Tosae Sin” ~ T, sin6'] eae et - roe [8 rool ‘The equation of motion then becomes ocode ZY « [rea + 2 [72] ae ef a,a : foo oe 72h = F(t).Methods in Mathematical Physics II 277 If we define the operator L= [co £25 3 the equation of motion can be written as oe QQ) Lyat) = Ft). In order to reduce the above partial differential equation to one of “ordi- nary” form, we make use of the Fourier transform, as follows: FO) = 7 Joseae. yt) = Fe ficnerto. yt) _ Sol) io Pye) 1 Fe jot, os 7 jt do, ab oF (ne“do. Substituting into partial differential equation (1) above: Sf 0% p(s) - oy Sele) - Gaede =0. Multiplying by e“" and integrating over t from @'t = 0 to 2n: tat ff [wtecvees - $752) Gcnlee™dodt . J {epee . 2 rit - 6.0) bo do = 0. 4 @ [wee - 575. ]e = 6.00. Since detivatives with respect to w do not appear in equation (2), the partial derivatives become total derivatives, with « regarded as parameter, ie. iY) = Ge. (3) [-«" p(x) - £ =278 Theory of Ordinary Differential Equations Defining p@) = “A2m3) and 960) = na equation (3) becomes Bs pe) Bs ge - - $2, a known function of x. 7.1.2 General Properties of the Linear Operator £ (where Lmay contain partial derivatives of several variables) LY = F (known function). £(¥) +2) = Yi + LY. (a) If ¥; and 'P; both satisfy the inhomogeneous equation LY = F, then the difference V, — 'P; must satisfy the homogeneous equation LH, - 2) = 0. (b) If Y, satisfies the homogeneous equation LY, = 0 and ; satisfies the equation LY, = F, then the sum Y, + ¥; must satisfy L(Y, +4) = F. With the aid of these above seemingly trivial statements, we seek (i) the most general solution of L Yo = 0; (ii) a particular solution satisfying LY = F. The most general solution of the inhomogeneous equation LY = F will then be given by ¥ = Wo + V1. [Note that the above properties hold in general for any linear operator, not necessarily one only involving differential operations.] Define a Green’s function G(ex2...X9t1'x2'...%n!), Such that Lf G62 Hi ky!) = BC A11YB Ha ~ 2)--+ Gn = Hy’). This definition suggests a method for finding a general solution to L‘¥ = F, by first finding a particular solution Theorem: A particular solution for Las PRR. x) = Fen) is given by Wi(ete..Xn) = SoGvn...rentn FG. Xn! \eboy'dxy'.. din’.Methods in Mathematical Physics I! 279 Proof: Lisa linear operator, and thus can be taken inside the integral, so that Lagi EEX. Xn) = feces — 01)... Bay — 5.)F Gt diddy, = F(xi%...%n). Note the following: (1) Finding the general solution of LY, = 0 does not depend upon F. (2) Finding G such that LG = 8"(x; - x;') also does not depend upon F. Thus, the Green’s function method will always yield a particular function corresponding to a particular type of operator. This function can then be used with any F one pleases. For example, for the case of V"Y = F(xyz), G~ yr-r'. 7.1.3, The 3-Dimensional Schrédinger Equation For the case of central forces, V(r) is a function of r only. The total energy is expressed by H = z + V(P), where p = = The eauation of motion for particles is obtained by replacing p by # Vand Hby - 22, Thus, hoy _[ Wo | q@ a [ am Vt VO) Mr) {Compare equation (1) with the wave equation for electromagnetic waves, vp - 2 aa O and with the heat flow/diffusion equation, V?7— 2 Note the resemblance between the latter equation and the See equation when we set V(r) = 0 and convert the real t axis into an imaginary it axis. The characteristics of the solution of the heat flow equation will then apply directly to the Schrddinger equation.] The program now is to transform the Schrédinger equation, a partial differential equation, into an ordinary differential equation, and we do this by making use of the Fourier transform again.280 ‘Theory of Ordinary Differential Equations wor) = Fe Jeane de = Hr - a iaba(nedo. Substituting into equation (1): 2 7 Pe Foonerdo 7 zit * ve vo] enedo Multiplying by e““®", integrating from @’t = 0 to 27, and using the ortho- normality relations for e®, we have: 2 E®{r) = [= Ve vo) (rn), where we have set E = ~fiw. nw Defining the operator L = - >— V? + V(r) - E, we can write 2m L&dr) = 0, where L= Liaw. [Since there are no operations involving the parameter , we will hence- forth drop the subscript @.] ‘Thus we have transformed the Schrédinger partial differential equation (1) involving time to a three-dimensional, time-independent partial differential equation. 7.1.4 Separation of Variables In spherical coordinates, 1@[.4 1 i270. V=sL|(Ps S ral" i Fins 30°35 * Fats agi’ so that L&(r) = 0 becomes WL @ O) “eal” 2)s onal fd [le Os sind 00 s * Pein ZI 0Methods in Mathematical Physics Il 281 Multiplying through by 7”, we see that the first operator, | hol (3 0. a5 a5 a Vo) - E], is a function of r only, while the second operator, Ww ie ino + 1a ~ 2m [sind 20 © sin’® dg°” is a function of @ and @ only. Hence, we can write Lp = L,+ Lop. In general, if Layx:¥ = O and if Las = La + Ley, where Lr, is a function of x1 only and Ly, is a function of x2 only, then Y = F\(x1)-Fo(x2) will be a solution. [Later we will show it to be the most general solution.] Assuming such a solution Y = F(x1)-Fa(x2), Lol = Le + GM = FrlaFit FilaFa = 0. ing by Fy-Fa, Len? _ Lei, GPa heh eee which can be true only if Fi Shy; SB 6 Lr ear Gh ~ ary ; : Let us therefore attempt a solution (r) = R()¥@p). Substituting into (1): ehel lois 0: io & om ai xl J. Find 00 “I? 36 * Paine og? Zlew + VOM) - E G(r) = 0 We multiply by 7 and, following the prescription outlined above, divide by ROY@9).282 Theory of Ordinary Differential Equations Eee) vo - E|Re RO| 2m|? ar’ ai HWA fd 8 i 8 é "2m oleeoe ind 3 * 5 ae She - 0. ‘What we have here essentially is F(r) + Fg) = 0, which can be true only if FQ) = -0’ and F@p) =’. Thus, Ww oo de , Flees sin 39 * J Shoo x¥G9). We set A’ = - #°2/2m, so that 18 9 2 é - (sts 5 55 Sao a a9 Shoo A¥(09). A= -I([+ 1) are the eigenvalues, and _ (QL+ d= my” i. Yox(0n) = (SER) Pitoondhe™” are the eigenfunctions of the Laplacian operator in @ and @. Since F(r) = -A’ = Ah/2m = - (I+ 172m, we have L [P12 22), yey ela =- (d+ ay2m R@| 2m|? ar” ar i or, @ At2e2.™ SE [E- Vo]- OU + DIRE = 0. Por aw We now seek to show that ®(r) = R(r) (8g) represents the most general solution of [eve vo )- Foe) =0 2m : :Methods in Mathematical Physics II 283 Since the Yimn{®p) form a complete and orthonormal set, any function of © and ¢ can be expanded in a series of spherical harmonics, ie. (709) = LRn(Yin(00)- Substituting into the 3-dimensional, time-independent partial differential equation (1), and multiplying through by 7” Lrwemr[- Eb 2e S)+ 00 rato + J] mn -0 ; +E eae [ted wi 5+ ante Og? Since the Yim(@®) are eigensolutions of the second part of the partial differential equation above with eigenvalues -I(/ + 1), we have }' vo- Eat : + YF Ral A-DU+ DYn(O9) =0. dm, an Factoring, v a) [F [- om £832). ¥)- ]-MenMan =o Until now nothing has been said about V(r) in the proof of a general solution. If V7) is a function of r only, however, then the entire bracketed expression { 4 above is also a function of r only. Furthermore, since the ¥im(8) form an orthonormal set, ‘Jelo J¥in(Op)¥..(Op)sinOdd = Bi dnp,284 Theory of Ordinary Differential Equations Thus, Jarfsnodor.s(on 2 ¥6(00|? Pag 2r2)+v0- Eat : ; + faefsineaor (00 Ya(0) SP" Rate = a +v~y—E}+ - Loran, fr al 25 7] vo) | a) Jaan 0. V : 2 fr Fas 2 x +V)- =| fe |r -0. Since only derivatives of r are involved, we are left with ordinary and not partial derivatives, so that Ad (dR) 2ml yp gt Da _ 2 wh {v9 aaa “he 0. P dr Rewriting, Afp@R ,5,4R)_2m|yq)_ p+ pn Mle 5 [p28 oor dt | ve) — (nr) = 7 If we define ptr) =2 and a) = -2l v0 -£+ in Jv have (FBO s 9) BOs gone) = 0. In what follows we shall discuss more fully this second-order differential equation. But before doing so, consider the first-order differential equation: ao, u@)g(@) = v(x), where u(x) and v(2) are well-known functions.Methods in Mathematical Physics I! 285 ‘We make the following transformations: eG) = owen fur: v@) = Vedexp|-fur ae], where V(x) is known since u(x) and v(x) are known. We then have: BE. ACO 0 fucr’yde | - Gedutsenp|—futx’ydr |. B®. vene00 - O25 [Jasna = VGexp _fucoae : : Hence, se. V(x), a known function, and G(x) ~ fue ar + constant. 7.2 Ordinary Points and Singular Points Consider the differential equation 0 Bs pe Ms gee - 0. If we regard z as the complex variable Z and study f@ in the complex plane, we will know the behavior of f(@) along the real axis. Definition: A point in the Z plane which is a singularity of either p@) or g@) (or both) is called a singular point of the differential equation (1). Any point which is not a singular point is called an ordinary point. Program: We shall attempt to find solutions for ordinary points, and then study the analytic behavior of f@) near singular points. If Z is an ordinary point, one can always find a point Z near Zp such that p@) is analytic over the path from Z to Z (Fig. 7-2).286 Theory of Ordinary Differential Equations Define f@ = @@exp [apes] , where p@ is 2 analytic over the path. If f@) is analytic, then ®(2) is analytic. . Setting s@ = thee, we can write f@ = b@e. It then follows that Fig. 7-2 YF@ _ ®@ 20_ POgp 30 . 2Op0_ POLO a & 2 7 m2: #@ _ FEO yo an go P@YO_ [@dp@ a 2 2d 2 & - £2@ .0_ POO, _f@dp@, ge la a bt a F8@ .0_ PG, YO, _L04e@, . 0 EO, aE oS ara Collecting terms and substituting back into differential equation (1): 0 = £2. pg %s aeye eAiF%@ _ f@p'@_ f© dp@ , [ae ae 42° 2 az eg - gf F2@ _8@PO 9 4@ - oof fa e@) _ 2 HO. axa) or, @ o- #2O . ose, 2 where we define JQ) = ¢@ - 22-1400,Methods in Mathematical Physics Il 287 Note that all ordinary points of the differential equation (1) ate also ordinary points of equation (2), since J@) is analytic if p@) and q@) are analytic. We now proceed to investigate ® #20). oye) =o. 7.2.1 Iteration Method Consider the differential equation my « #22 weve =0, which differs from the equation under investigation in the appearance of A, whose meaning is such that |*@|,, = 0@). Assuming (later to be shown to be true) that &*(Z) can be expanded in a power series of 2: P@ = © +A.@ + VO. +... = Yr" S,@. Assuming also that one can differentiate within the summation: #90, 280 290... AD. +NI@ [BD + AO:@ + VOE® +...+ NOD +. j= Collecting terms in powers of A: P2@ [fen o +700.) + [a2 : s00.] a +x(/f@ ae at [2 + sea Hes + I@Pr- ie). which can be satisfied only if the coefficients of each A‘ =288 Theory of Ordinary Differential Equations PO. qn@Q-0. £2Os se p@-0. fe) + I@Pnri@ = 0 for n>1. a 2 fo = 0 for n=0. Note here the iteration method: . stat with P20, © Solve for B.(Z). . Inset 2,0) into £22. - - IJ@P@. © Solve for &:@), ®2@), etc., up to ®,(Z). © Knowing all ©,(2) (as n + ~), you now know ©*() from the power series expansion. Consider again equation (1): We must find *() which satisfies [rol.-4 [Se Fa where A and B are given boundary conditions, and also satisfies go, A QI® = 0. a Note that o(Z) = A + 3G 2) sashes 8D =O for n=0.Forn21 it follows from fom =~ J@Pn-1@ that a = - Preyer. %Methods in Mathematical Physics It 289 This latter expression is satisfied by 0,@ - -fe-z@e.@x. since . SO - (g-2ye yn], -fg E-ZVEV Vaz - 0- frewnene. ‘We have ©) = eZ) + Ab (Z) + A7h2() +... , from which |e'@| ||, only, and de"@)| d&.@) | a | | oun e since, for n> 1, @x(G.) = 0 and jee =o. % m Hence, boundary conditions |@*@|_ =A and pea satisfied by ®,@) = A + BE - 2). @) : b,@ = -fe-zyeVe.@ee. % % ‘We have thus found a ©*(Z) satisfying fem +2@*@J® = Oand also satisfying the proper boundary conditions. Let us now proceed to investigate the series (2) -Ev'e,@. Let the upper limit of J@) and (2) along the path Z > Z be U and V respectively, such that V@| < U and |Po@)I < V.290 Theory of Ordinary Differential Equations. Theorem: |®r@| < —— vee ze. Proof: The method of proof will be by induction, i.e. we assume it to be true for n ~ 1 and show it to be true for n. Then if true for n= 0, the proof is complete. From (3) above, I2.@I < -fie-2enie, «eel. % Since ,() is analytic in R, the path is immaterial. Hence we choose the straight path from ZtoZ (Fig. 7-3). Fig. 7-3 Then, since 7’ is a point on the path, [z~7'I < [z ~ Za. Assuming the theorem true for - 1: 1@,-1@)| < Vg - 2" 7 (n- 0! 1b.@) < 2-2) vento, «eee _ NE ZU" Fae a jie zeke Setting t = [z’ - Zol, so that dt = |dz'l, we have an UWG- 2) "Frag, . YUE-Z)" l@.@I s -1! ‘ ce (n- 1)IQn-1)° But for n 2 1, 2n - 12 n, so that - i.) < WEB and the theorem is true for n. For n= 0, o@)I < V. Hence the theorem is true for n = 0, thus completing the proof. To show that 2) can indeed be expanded in a power seties in A, consider = an the power series e** = yxE , which converges for all values of E, with a radius of convergence of A = =.Methods in Mathematical Physics II 291 Setting E = UC - 2)", we have from the theorem above, 1,@l < worn - woe A. Thus, )/2"@,() converges uniformly for any finite 2. Furthermore, m0, ®@ = ¥/2",@ defines an analytic function of Z with a radius of conver- gence equal to infinity. For every ordinary second-order differential equation, there is a solution @), an analytic function, at points near Z, the ordinary point, such that ©) has two arbitrary constants, namely the integration constants, and these constants may be fixed by fixing (@.) and’) = | In addition, this solution is unique, as we now show. & % Theorem: If the values of a) and g@) are fixed, then (@) is unique. Let ¢1(2) and g2(@) both be solutions, so that pi) = 2Zo) and $1'@o) = G20). We will prove that p1(Z) = p2(Z) not only at Z but at points near Zo. Proof: Define P(@) = p1@) - p2(Z), so that PZ) = piZ) ~ P2Zo) = O and ©) = pr) ~ $2’) = 0. Since pi) and p2(Z) are solutions, (2) is also a solution, i.e. 4 % @) . Gee] = 0. This follows since fo [pa WO ~ 4G = 0 at Z=%. Taking derivatives of (4) and evaluating at Zo, poP +70 222 + gene +20 “e'- 0, i since $20. #0) 9) 09t2=2.292 Theory of Ordinary Differential Equations =0. Since ® is analytic and can be expanded in a Taylor t series about Zo, P@ = HG) + E-Z)O'G) + o"G) +... - Y Pae-at yee By directly observing p@) and q@), one can determine at once the mini- mum value of the radius of convergence of the solution @). An ordinary point of the differential equation is defined as a regular point of the solution O@). 7.2.2. Power Series Method Concerning equation 0 Bx p@2+qave - 0, let & = Z- Zand perform the expansion f@) = E aye", where the radius of convergence is not greater than the distance of Z, to the nearest singularity. Since both p@) and q@) are analytic at Z and in the region near Zo, we can write P@ = Lont. q@ = Yant”. EF) PO. F wn 1)0,9"? ee Lan a,b". qI@f@ = Yont" Yow. ro ®- Soak" Yna,e", Substituting back into equation (1) and collecting terms involving &”, Ye (n+ 2X0 + Vana + Ya — m+ 1)PmQn-mes + | = 0.Methods in Mathematical Physics II 293 This can be true only if all the coefficients separately vanish. Thus, 2) (+20 Daya + Yn m+ DP nteemes + Yednddr-m = 0- Equations (2) are called difference equations, with parameter n running from 0 to infinity. Forn=0: 2az + apo + ago = 0. For n= 1: 643 + 2arpo + aspi + Aigo * 4091 ete. ‘We ask: What advantages do these difference equations offer over the differ- ential equations? By fixing ao and a; arbitrarily, case n = 0 determines a2; case n= 1 determines a3, etc. Thus, fl) = do,f’Go) = a1,f" Ge) = 2ar, etc, thus determining f@ = E ant". (a) Fot ay = 1 and a; = 0, solution fi(® satisfies fi(Z) = 1 and fi'(Z.) = 0. (b) For ay = 1 and ay = 0, solution fs(@) satisfies fs(Zs) = 0 and ff’ Gs) = 1. Thus, if boundary conditions require that fZ.) =A and f'@) = B, then we and can write f@) = Afi® + Bf@), since fG) = Afi) + Bf@o) = S' Go) = Afi) + Bfi%) = B. 7.3 Hermite Polynomials The Hermite equation, given by 41@_», F@, . Qe 2g *2M@ = 0, where p@ = -2Z and q@) = 2A are both analytic everywhere except at Z = for p@), will be studied by the power series method of the previous section. The solution f@) is regular at all points of space except at Z = ©, and is there- fore an entire function, i.e. analytic everywhere except at infinity, with an infinite radius of convergence. If f@) is a polynomial, it has a pole at infinity; if it is not a polynomial, it has an essential singularity at infinity.294 Theory of Ordinary Differential Equations Since f@ is an entire function, it can be expanded about the point Zo in a Taylor series. fo- Set. W- Fraz. 72 - Snag. £9. Yan Daz. = Substituting into (1) and collecting powers of Z": Ye (n+ 2)(n + Vaya - 2nay + Ady) = 0, which holds only if (n + 2)(n + 1)an+2 - 2ndn + 2Adn = O[n=0, 1... _ 2 Ag @) ae = Gane’ Forn=0: a; = —2@e, 2 - Adar ae Forn=1: a3 = 3 Forn=2: a, = 22-Mae ae : 22 Bie is a regular singular point, and the functions f(z) and fs(z) have at most a branch point, since s(s~ 1) + s(p-1) + q-2 = O with roots s1 and sz. (@) For si # 52 (positive integer), then fz) = (z ~ 20)* Yyax(z ~ 20)" yields two distinct solutions fi(z) and f(z). =o (©) For s1 ~ s2 = £ (positive integer), we have fi(z) = (2 ~ 20)" Yoan(z- 20)" =o mb for s1 the higher of the two roots, and fa(z) = (2 - 2)" )/ka(2 ~ 20)" + mo fy fi(2)n(z ~ 20), where if hy = 0 and s1 and s2 are positive integers, then fil) and fo(z) are both analytic. Behavior of solutions at infinity: Consider again « fo), pos gaye) - 0. Setting z= 1/2’ we study the behavior of f(z) at infinity by studying the behavior of f(1/2’) at 2’ = 0. Wyfg-a+8+$ ve fz’) = a+ be’ + cz” +..., and at z'=0, S01’) is regular; hence f(z) is regular at z = °. Ufe=azr bd = fuye)=F+ oe and f(1/2’) has a pole of order 2 at (z 2 ~0; hence fl) has a pole of onder Satz = =,310 Theory of Ordinary Differential Equations We write differential equation (1) as a function of 2’, setting z = 1/2’, noting de dz ga’ ED. AND MO - Al, y—9 ME a dt ed 2 de” po ~ - pee? LO. Collecting terms and substituting into (1): ad ad ad = eal 22 dio). pez odo, gee) = 0. 4 4f@), cae POU @), I@Fe). 9 dz? |2' a zt The terms that we now have to investigate are tory = 2. B@ PQ) = 7 EP 22- peor. 19) ~ giay, z gq) Case (a): If 2 - p@)z" and q(e)z" are analytic at z = ©, then f(2) is analytic at infinity, and infinity is an ordinary point. Case (b): u- PD tas a poe of order 1 at = 0 (ie. if -p@2' is analytic at z= 0) and aes" has a pole of order 2 at 2’ = 0 (ie. if g(yz” is analytic at z’ = 0), then f(z) is analytic at infinity (or has at most a branch point), and infinity is a regular singular point. If case (b) is not true, then infinity is an irregular singular point, and we have an essential singularity at infinity.Methods in Mathematical Physics II 3 7.5 Bessel Functions Bessel’s equation is given by 2 Note that p(z) = 1/z has a simple pole at z= 0, while q(2) = 1 - (2/2)" has a pole of second order at z= 0. Hence, z = 0 is a regular singular point for fiz). Note also that p(@)z is analytic at z = ©, but that q(2)z’ is not analytic at z=0, Hence, z = © is an irregular singular point, and both solutions have an essential singularity at z= ©, Behavior near z = 0: From p@) = + Po* piz +... and g(z) = 4-22) + a + qo te... We Zz see that p-1 = 1 and q-2 = -A?, The indicial equation becomes Ss-1)+s-W=0 = San mm = 4h Choosing 4 > 0, s1 = +A and so = -A, so that s1 ~ s2 = 2A. 1) For 2A + integer (i.e. 2. is not an integer), 2 Yaz" f@ =} "| = both solutions have branch points. 2? Yon" 2) For A= integer (i.e. 24 is an integer), z Yaz" (analytic) fo~}; "~ = branch point if h, # 0. z* Ybn2" + hz Yanz" nz312 Theory of Ordinary Differential Equations 3) For A= 1/2 integer (ie. 22 is an integer), Fa Yaz’ (branch point) f@= = branch point even if h, # 0. 7 Soe the Loe Inz The Bessel equation occurs quite often in physics due to its intimate relationship with the Laplacian operator. Consider the wave equation V°d(rt) - #21 9 Applying a Fourier transformation with time as parameter, (0) = feanedo > Yale) - Ze Jomne at. PH(r) va froraneao, Vib(r) = Seve ao, Substituting into the wave equation: 2, ow Je [Pec +o van do =0. Since the e'” represents an orthonormal set, Q V(r) + ke Palr) = 0, where 3 = 07/c?. For ko = 0, the wave equation reduces to Laplace’s equation, while for ko # 0, one has the time-independent Schrédinger equation. (A) Make the following cylindrical coordinate transformation: Y(pp2) = ¥ Laatoremeat.Methods in Mathematical Physics II V*V(pez) = ke¥(092) = Y, [oykte™e™ak. Combining and substituting into equation (2): ry; al on =, Roa = Employing the orthonormality properties of the e”””, we have 1d (doy mf), 2) = >a | a oO) =? where x? = 3 - °. Expanding this equation, feo), 1 flo), { 2m). _ dp? p dp ft mV 0. Setting xp =z, oz « Oana a 2 Ffo) so that Senet 4 313 which we immediately recognize as Bessel's equation with = m (integer).314 Theory of Ordinary Differential Equations Thus, a Yaz" (analytic) | s : 1z* Yby2" + hiz* Y'ay2" Inz (branch point) (B) Make the following spherical coordinate transformation: (reg) = LRa(r¥inf 0). VAP(rOp) = rae eo +L FH) Psin’'® ag” - Ess Juen *Elpikw alte ae) Fae ag P- M009) = DAR nl Yin(O9)- Combining and subsituting into equation (2): bea +f afr BO} 00 ye a9), _1_#Hee) Lltea™ ° es an’d ag? hn"? + Lee Rel MYinl 60) =0.Methods in Mathematical Physics I 315 We know from Part I, Section 5.3.8, that ete ¢ ao (sts (55) ae ae Fhaon K+ D¥in(69)- E[b ale] -ean- Parfum Employing the orthonormality relationships for the Yin(®@): EPSP) ean - PR - 0. z dr seting Ri) =, LQ, 2d. _ HHO, pn ao cs 2? * vr dr* dr 2 dr* aR) _ __ fr) _ 1? df), fo), 3r? df) Plas oa. eo a ae 2 dr - - £0, » £10, LO ea Collecting and substituting into equation (4): fo. #0, 1 gf), RFO_ me . y? 4 fdr? 4 dt, 1dO,[p GIP | © ALM elon Setting kor = z, HO). 4, LO oyq FPO 52 E10 ae ee ye316 Theory of Ordinary Differential Equations Substituting into equation (5): 4f@), Lge, {,_U+ yar he : OL |e On, which we immediately recognize as Bessel’s equation with 2 = 1+ 1/2, ie. half-integral 0. Thus, fiz) = 2 Y.ayz" (branch point) Ale) = 2 Yb2" +... (branch pointy Summarizing: ‘The Fourier transform was used to remove the time dependence from the wave equation. Then, by the technique of separation of variables, one arrives at Bessel’s equation. (a) In cylindrical coordinates, integer values of A arise from the m in e"®. (b) In spherical coordinates, half-integer values of A arise from the eigen- values (1+ 1/2). 7.5.1 Bessel’s Equation — Singularities Since p@) has a simple pole at z = 0 and q(z) has a pole of second order at z= 0,2 = Oisa regular singular point and fi(z) and /a(z) have at most branch points. In addition, since q(2)z” is not analytic at z = %, z = is an irregular singular point and both solutions have an essential singularity at Zee Expanding solutions of Bessel’s equation about the point z = 0: fe = 2 Yaz", where A= +/AJ. @ y pa i y . . Lar marn- vag aa 1H. Lov dae.Methods in Mathematical Physics II 317 ; eS Xf@) 2 @) Len Oezeee Substituting into Bessel’s equation and collecting in powers of (n + 4-2): Le"? (a, [a +nya+n-1)+(n+2)- 47] + 4-2) =0. Since this expression is analytic for all z # 0: an [+ M+ n-141)-27] + a2 = 4 Oyo (2h + ny" Qa = Setting a; = 0 and a, (arbitrary) = ay and substituting into (1): FQ) = 2 (do + ax2? + a2 A(2A+2)(20+4) Ae y -1y"e" Be 2:4--2m(2A42\(2A+4)—-(ZA42m) ? 4 with the added specification that f() = doz" for m= 0. Note that 2-4--2m = 2”m!, while (2A+2)(2A+4)~-(2A42m) = 2"(AIY(A+2)~-(Atm) = 2+ Mm)! _ ym TAtme1) a T+) * Thus, ] F (-D"G2y"TO+!) fo az’ YY miT(Atm+l)*318 ‘Theory of Ordinary Differential Equations 1 We ch =>—., so that le choose ao PTO+1) _ score fo = x miT(Atm*l) * This, then, is what we define to be the Bessel function Ji(2). Cle QB) S@ = Lar: If 1 is neither a half integer nor a full integer, so that 2A + integer, we have two independent solutions, i.e. Jsyy # J-iy), where both series start out with different powers of z. If, however, 24 is integet-valued, but the In term is absent (i.e. hy = 0): (a) For 4 = tn (na positive integer) @ Ine) = Yor m!(m+n)!” and the solution is analytic everywhere except at infinity. As for J-n(2), it exists only when I'(A+m+1) is at least equal to (1), i.e. Atm+1 > 0 because T(0) represents a pole. Hence, © 2 = Learn “ty miD(m-n+1) If we set m~ n= s,so that m=s +n, then Sula) = Loar oy =cy Lorn Oe (stn)\s! (stn)!s!” and except for the factor (~1)", J-n(2) is identical to Jq(z). Hence, for 2 integer-valued, we have only one solution. (b) For A non-integral valued, the general solution of Bessel’s equation is given by f@) = AJsny + BU-py.Methods in Mathematical Physics I 319 Question: How do we find the second solution f(z) for integral-valued 4? We could use the general method, namely, /s(2) = fi(z)0(2). Instead, we note that h, # 0 and thereby expect a In term, i.e. a branch point. Consider the expression J+j,(z) - (-1)"J-p(z), which — 0 as 4 — n (positive integer). For all A, both J+(z) and J-,(z) are solutions of Bessel’s equation. We therefore ask the following: Is « Sen((z) — =o 1Y'-p((2) © — ¥2) = jim 7 different from J,(z)? If it is, then we have found our second solution, since (© satisfies Bessel’s equation and is different from Jn(z). Evaluating for n = fi Le = Jo) _ , [2A ¥2) = jim x af oA |. From equation (3), 7 ante) (py""*E1)" Ing) - - i miT(Atmt+1) -pwre 1)" Or(a+m+1)y = om [Farm on | The bracketed expression can be written as 1 Qeme) TOtm o& «Tm saloon). Defining ‘Y(A+m+1) = a InT'(A+m+1): = oF 2D" Ing) , FH (2 cor Wontly 140 = 2 IED) TO Dl320 Theory of Ordinary Differential Equations Rewriting: ' Se (1) ¥la) = 21sankea) - 2 PND ‘We see that Yo(z) is certainly different from Jo(z); in fact it contains the expected In term plus an analytic term. 7.5.2 Bessel Function — Theorems The Bessel function = Sepp" 4@ = Lear rama’ is a solution of Bessel’s equation 2 @ #401 18@, [1-She - which we saw occurs quite often in physics because of its relationship to the Laplace operator. We seek to relate /,(z) to an integral in the complex domain, similar to what was done for Hermite polynomials; we do this because the integral representation furnishes us with much more information. Theorem: (for 4 = integer-valued n) ee [ . =| GB) Af) = fs, pien|t-4, [ar. The integral is taken about a circle surrounding f = 0 in the complex r plane. Proof: Method (A), Show that equation (3) satisfies equation (2). af) "ae )- Py gz 2 2ni J 20 pt sre t~ atMethods in Mathematical Physics It 321 FIG) nye n@argz ea a 2° 2 IZ) @ ni elt ag war ge nel] 2ni J |g? ot |e 1 dJ(z) _ ny 12) - LE Sen fren t ia 7 ani 9 20 pr *P Substituting into Bessel’s equation (2) for =n: a) wre. m2, 1 Froolt-E ja : ear ga o(t t- Zt) ae Hence, Ja(z) as given by the integral representation (3) satisfies Bessel’s equation (2). Method (B). Show that the expression for Ja(z) as given in equation (1) is represented by the contour integral expression (3). dt = — : Consider e iG For t # 0, this series converges uniformly a and absolutely, i.e. ‘has good properties. Substituting into equation (3): J = LONGO where the summation has been taken outside of the integral since the series is uniformly convergent. From the theorem of residues, : cs dé Nyy FOC 5 ae) 2nid pm! ~ (nem)! mi(ntm) ~ o-n322 Theory of Ordinary Differential Equations For A not integer-valued, we know that our solutions will contain branch points at t= 0, and branch cuts must be introduced. In such a case, the : a exp(t-z/A)) integral § 5 ra dt is a Cc not taken over a closed contour and hence does not vanish. Instead we choose a particular path of integration : C shown in Fig. 7-7. Fig. 7-7 For this path, gah pa exp(t- ZA0) | _ wy [2 t- 2/At) ay ve ° 2ni pt ni pr - Thus, we have here found one of many possible contours, such that even for 4 not an integer, the integral representation for Jn(z) still satisfies (2). Question: Is Fam ~ al ee nt [ If it is true, then from method (B) above, Hoe eat pele) , = Fae (car ego 5 Oni Pr & mpm . oy" 2 reenter’ and the integral representation would be identical with the Bessel function expansion. We seek then to prove that 1 edt TQAtm+l) al per where the path C is as shown in Fig. 7-7.Methods in Mathematical Physics II 323 Theorem: T(2)T( - 2) = Fang? Where Re(2) > Obut <1. Proof: Consider the shaded strip (Fig. 7-8) in the z-plane. Both I(z) and (1 - z) are analytic in the strip, as are wand sinnz. Then, by the theorem proved in Section 6.11.1, if T@)T(1 - z) = wsinnz in region I, it will be true for all regions of analytic continuation, ice. if F(2) = T(@)F(1 - 2), which can be analytically continued, and is analytic over all z except at cettain isolated points, and if F2(z) = 1sinnz, which is analytic at all values of z except for certain isolated points z = 1, 2,..., then by showing that Fi(z) = Fa(z) in the shaded region, F\(z) will equal F2(z) over the entire region. Let z' = z+1, so that T@M(1-z') = P+ )P(-2) = 2T(@) 7 T(1-z) = -F@Pd-2), thus relating points at 2’ = z+1 to points in the shaded region, where the expression is analytic. ‘The P'-function was defined (Section 6.11.1) as [(2) = feria for Re(z) > 0. Setting t= 27: Te ~ Je*ee onde = afore ae, T(1-2) = afe*s! ae = afer ye May, Te@ra-2 ~ Affe" apy Peay. Making the transformation x = rcos®, y = rsin®, we have x* + y=, (yt = cots? and [fdedy = [[rdrdd. Hence, T@L-z) = fe "reot8 Ydrd® = fi [= Jeo 1g9 = 2Jcotd* 'd0.324 Theory of Ordinary Differential Equations Letting E = cot8, dé = -csc*@d@ = dO =- cs Substituting once more: + rerd-2 = fee = fe oe Note that for Re(z) between 0 and 1, 2z is not an integer, so that [bs pe falls into the type III integral discussed in Sect. 6.10.1. To test this, we must see whether 241+?) + 0 as E + 0 and; and it does, by observation. Note also that pe M1+k) has a branch point at — = 0. Hence, let us consider the contour shown in Fig. 7-9. In Section 6.10.1 we showed that 1 [e-teftee To show that this holds for the contour of Fig. 7-9, note that over the large outer circle, E(1+E?) -- 0 as E + ©, while over the small circle, BA 1+E) + 0 as E — 0; hence, fee -fa- fone pe fine. This last step follows since the first integral on the right is evaluated at © = 0 while the second integral is evaluated at 6 = 7/2. Thus, rae - ol ee Evaluating residues: EP ert Resf be® -E ea | i le We must be careful here since i = e””, e°¥, e°¥, etc., while -i= &"”, 25%, 69972 otc, Within the limits of our integration, however, sine, and -i= "7,Methods in Mathematical Physics II 325 _ Gel etn gine hag linge Reine f e® i 2 2 2 ~2nie™(1 +e?) ~ + e™\1 — 2") 2ni x We have thus shown that T@U-2) = fee de = sinnz” ‘We return now to the problem of showing that 1 edt TQ+m+1) om per In Section 6.11.1 we showed that 1@) = tag fers"'as, t where the contour C is as shown in Fig. 7-10. From the expression T@)TU - 2) = ~2nie"/(1 - €?"), we have T@(1-e%) = a - fevas Setting = 1-z,326 Theory of Ordinary Differential Equations The contour C in the t-plane (Fig. 7-7) is to be compared with the contour C’inthes 2S plane (Fig. 7-10). We perform the inversion “yo 1 = -s so that contour C’ + contour ~C P ae Fig. 7-7 1 ef ae - pea, TE) 2nic(-1if nig #’ Cc co and by setting E = (A+m+1): ————— f oe s-plane TAtm+1) — 2nit perl? Fig. 7-10 thus completing the proof of the integral representation theorem. Recursion relations Theorem: Sale) * Saale) = Bane. Proof: Using the integral representation (3): pias SriZ) + Jnei(2) - on = (yt i J a a 1 +r La nea EI Oo tel ak OLE [ed ' 1 pale . ey : afe| iP |e pees] "oon [Pe |Methods in Mathematical Physics II 327 Theorem: 6) SQ. Bn = - -Sin@) = 2 Sho Sa@ ~ ios) Proof: on So - Bag - Wah pn rat = An@- Jani(2). From Theorem (4), 2 102) = rs) * Sra. qd 2) A n@- Jinl2) = 3 [i - Jou] 4 2G - Ji@) ~ Suna). [Note that Theorem (4) could have been proven using a power series method, but it is too clumsy. The integral representation demonstrates again its great usefulness.] 7.5.3 Bessel Functions vis-a-vis Sine and Cosine Functions Fo uee i cur o 4 miT(m+3/2)’ where I'(m+3/2) = (m+1/2)T (m+ 1/2) = (m+1/2(m-1/2) ~~ (1/2172). Setting y = 1"? in 1(1/2)= fer Yar: TU) = JerMar = aferaey = ay =vn.328 Theory of Ordinary Differential Equations Hence, _ m+ 1 2m-1)2m=3) 14 Zm(Bm=2)(2m~A) 2 ae ane 2"! _ Omir mn Substituting into (1): — aS Ente ayna . pei Jn@ = (a? Y er Te = (ana) z oe , = oa {e- Z ae wo Q Syal2) = (2/ny? Sz (fn) oe nm 2m 3m and is plotted in Fig. 7-11. . To evaluate J3p(z), we make use of recursion relation (5) for A = 1/2, i.e. Fig. 7-11 June) = 2y@y- FO yt = Esse - SH Thus, QB) Jyp(2) = V8nz [5 - cox. Next higher order half-integral Bessel functions are obtained by continuing to use recursion relation (5).Methods in Mathematical Physics I! 329 Jatt) » 35 fue) - PH ~ > [1apRe( - cosa] ea) al : (|? (cos [ese] = Te |e : Sinz - sn] Zz Thus, (4) Ieal2) = Pia sna| 5 - 1} 22) For A half-integer valued, we seek the asymptotic behavior of Jy(z) as z ©. The general behavior of J,(z) for A = 1/2 integer can be obtained using recursion relation (5), oD - Ane = - Jasi(2). If one neglects : Ja(2) asz—, Jasi(z) > — a) « Then, since Jya(z) = (2/n) sinzW/Z, Jae) + ~ (ny? SE. Jug) + ~ Of) 5 = apna ne = - na)" sina. Jnle) + - Ge Crema Pee) = Ona)" cose. Je) -+ - Fe ~ (2yne)? sina. From this point on, the cycle repeats, i.e. Joya(z) behaves similarly to Jya(2).330 Theory of Ordinary Differential Equations The behavior of Jq(z) as z -* © is seen to be similar to the behavior of sines and cosines in the 4 quadrants, i.e. Jy > + sinz; Syn + ~ 00825 Jy > ~ sinz 5 Jon -* + cose ; Jon > + sinz; ete. Thus, for 4 half-integer valued, as z+ ©, 5 (Ae) = na)" oos[2-7F-F]. ‘We will now show that equation (5) holds for any A. Consider Bessel’s equation £060 1810) |, 2g ‘To aid in the investigation of J4(z) as z + ©, one removes the I/z term by the following transformation. Setting J4(z) = g@/¥z , we have: dh) _ 1 dg@_ 2@ 1dh@ _ 1 dg@_ 2@ Qe ee 2 de HQ) _ 1 Fg@)__1_de@__1_ de®, 38@) ae za ee de ra a 4?" Substituting into Bessel’s equation: de 20, 20.20. 4 dk a cee 1 de) _1 dg@), 32@), 1 de Md 4? 4 a © £20()- CM - #20 94 - 0. ae The last step follows if one neglects the term ae for large z.Methods in Mathematical Physics It 331 Solutions of (6) are given by g(2) = axcos(z + bx) > Ja(z) = ‘We wish to show that a, = (2/n)"” (independent of 4) and b, =-=-=, a From recursion relation (5), neglecting for large z, we ie — BA) _ ayvre0s(z + bass) Sivi(2) = dz rz Thus, for large z, aynscos(z+ bn) am d aqsin(z + by) ae Hence, a) = ays) = Qyea «.. = 0 (independent of 2), and be -& + B (independent of 2). This last follows since exten) =m BF sn(s- +] =sin (¢ + by). We have shown, therefore, that for any A, Gy oy seoste— ina * Py : Compating (7) with (5), the behavior of Ji(2) for half-integer 2, we would like to say that a = (2/t)”” and B = -1/4. However, the result (5) has been shown to be true only for values of 4 succeeding by 1. In other words, for 21=%,%, ... 101%, ete., values of a and B do not change. We do not know whether these o: and B values are (2/n)”” and -7/4, respectively for A= 1/2. This result can be found in Modern Analysis (Whittaker and Watson, Cambridge University Press, 1952).332 Theory of Ordinary Differential Equations 7.5.4 Hankel Functions HAG) = ttle (2) - Jal2)}- QM Hig) = , [e"**A@ - Ja@)]- Hankel functions H}"(z) are obviously solutions of Bessel’s equation since Ja(z) and J-(2) are solutions. For integer valued, one understands Hy?@) tobe im Hy. For z real, H(z) is clearly the complex conjugate of H\(z). We saw in Sec- tion 7.5.1 that since g(z)z’ is not analytic at z = infinity, infinity is an irregular singular point, where both solutions (2) and J-x(2), and therefore Hy"(), have an essential singularity at z = infinity. Let 4 =n + €, where nis an integer. Then, as e+0 (or as An), sindx = sin(nnten) = ex cosnn = (~1)"er, and i 2 tO" = 1)" as e0. ti — him Gi We Fon), fin 0 - Cy im | | @ lim w?@ - Shim eo [zee 7 (UVa), We saw in equation (6), Section 7.5.1, that lim [HOCH ae exactly the second solution of Bessel’s equation for integral A. Hence, the Hankel functions are second solutions of Bessel’s equation for integral 2. Using the asymptotic behavior of Jn(z) from Section 7.5.3, equation (5): aie) = 2 Caney" [e*eoste - an/2~ H/)] as 8) - Bee rie one)" [e eos(z + Ax/2 - 1/4)].Methods in Mathematical Physics It Substituting 2cos(z — Ax/2 - 1/4) = el PV2IN4 4 griesih2siNd inte (3); 2H) = a v2 [ei (gina 4 evan) A) = Sage reed [eo [e! e ) i ie sina (givin 4 ican a (2/nz)” [e (a Dan 4 6 oe) = area [er (oP 09] + HIM] Ma sinAn (2/n2)*[e*™ sina} 2i = 26 (2/nz)e™ Thus, HAZ) = (fnayele MA. 4 ie) = Cea H3(2) is obtained by replacing i by ~i in Hi(z).Chapter 8 Theory of Partial Differential Equations 8.1 Examples of Field Equations in Physics Poisson’s equation (V(r) = -4np(r), where the charge distribution p(r) is given. For p(r) = 0, equation (1) reduces to Laplace's equation V(r) = 0. Boundary Conditions (Fig. 8-1) Either ®(r) or (rn on S is known. These boundary conditions are sufficient to determine (r) uniquely everywhere. Fig. 8-1 Wave Equati open boundary fave Equation . op ee lcs Q) Veer - L SPOD « 9, oe An example is the vibration of a string of constant density under constant tension. The equation for such a string in 1 dimension is Aix), (2) given at t= 0 ay 12 a tat — Fig. 8-2Methods in Mathematical Physics Il 335 Boundary Conditions (Fig. 8-2) Atx=Oandx=L, y=0 forall time. Attime t= 0, y=f(x) and y’= g(2) are given. These boundary conditions then determine a unique solution to the wave equation. Heat Conduction @) ¥en = 2.2889, where 4 > 0. Fora 1-dimensional rod, equation (3) - erp - oe, Boundary Conditions (Fig. 8-3) Atr=0, T= T[x,0)is given. open boundary ee TO) in time TL) Atx=0, T= (0,1) is given, and given given Atx=L, T= TL.) is given for all time. The solution of T for all times and positions along the 1-dimensional rod is then uniquely 0 determined. T(%,0) given In general, for second-order ordinary Fig. 83 differential equations, solutions are given by: SQ) = Afi(x) + BAG), where fi(x) and (x) are each solutions of the differential equation, and the two arbitrary constants A and B are determined by specifying: (a) fle=0) and f{x=0) as given. This corresponds to the boundary con- ditions of an open-boundary problem. () fl-0) and f(x=L) as given, ot flx=0) and f(x=L) as given. This cor- responds to the boundary conditions of a closed-boundary problem in one dimension, x = 0 tox = L.336 Theory of Partial Differential Equations For Poisson’s equation (1): Since time is not involved, one needs the value of f(r) at all points on the boundary. For the wave equation (2): Since the boundary is closed in 1 dimension, one needs f(x=0) and f(@v=L). On the other hand, since time is an open boundary (of second order), one needs y(x,t=0) and y'(x,t=0). For the heat conduction equation (3): Since the boundary is closed in 1 dimension, one needs T(x=0,t) and T(x=L,). Time, however, is open and of first order, so that the value of 7(x,=0) will suffice. The Schrodinger equation [Ew v}xteo = 42 Hen resembles the heat conduction equation in that the equation is of second order in space and first order in time. 8.2. Theory of Characteristics Consider any surface Son which © and[V¢], are known. Is the solution completely specified? Consider the following (restricting ourselves to 2 dimensions): ‘The most general second-order partial differential equation is given as 4 £20), 25 THO , EBD) , p 20) axdy ay? ay + EMD « Fey) = 0. ey Problem: Given ® and[V@], along PQ, the P segment of a curve (Fig. 8-4). Consider a point Oy) OGy) on PQ. Can one find ®(ay) in the neighborhood of point 0 In other words,can Lowy one find all the derivatives of (xy)? Fig. 8-4 QMethods in Mathematical Physics II 337 Solution: Given the initial conditions, ie. that © and [V], are known, [Vo lane is known, Performing aration one can then get [vo]> 2 ana [va]. Choose ds = (dx, dy) along PQ. Since 7 and & are known at Oy) and at ow'y,a[ 32), andal S| are known We then have Ox Oy Jpg ox Ox {oh (ol leh sun From equation (1): a(2@ -(2 -(2|- eo [ose eS Pe] 423.22. oft, CO Oy. The left-hand sides of the above 3 equations are known by the arguments above. Rewriting as a set of simultaneous equations: Fo 4, FH ob ae aay 9 *0 - [oe eb ob +2 4 Fe exdy ay? | 5 he" co, eb ae 4 andy 2B Fe. a ES Fe ax? oy? [ Je! ‘We get solutions when the determinant does not vanish, i.e. when dx ® 0 det ]0 dx dy| = Cdr’ +A dy -2Bdxdy + 0. A 2BC338 Theory of Partial Differential Equations ‘Thus, the requirement for knowing the values of fs Zz and ae is: (2) Cd? +A dy - 2Bdxdy + 0. Now that second-order derivatives are known, can one find third-order derivatives? Differentiating equation (1) (of which is a solution) with respect to x: Fo Pb (Fo | Fo ~Fb ob ASD BS * Sag PSE ESeay Fa As before: Fo) _ [a 2, ob a he le (3 |-2 a me” db) (86) (#0) | oo, 0 (22), . (= (S1- ao ao Setting up the simultaneous equations again: 428 2p 22 4c 28, = known. x ax’ay ~ dxdy” eb, Fb aS +d *O a2ay* ax oo, eb Oat ara 0 = known. +0 = known. ‘We get solutions when the determinant does not vanish, i.e. when A2BC dx dy 0 0 dx dy det = Ady -2Bdxdy+Cd¢ #0. By repeated processes, i.e. taking derivatives with respect tox and y, one can find 3”*"4/ax"dy" and, assuming the Taylor series to be convergent, one can find @ in the neighborhood of the point O(,y) on PQ.Methods in Mathematical Physics I 339 8.2.1 The Characteristic Curve We seek the curve expressed by the equation (1) Ady’ + Cade - 2B dxdy = 0. Rewriting: 2 4(2] +c-28(2) =0 = Ay?+C-2By' =0. Solving for y: yielding, in general, two families of curves, where the two integration constants are the parameters of the corresponding families. 2B + V4B? - 4AC Ty Special Cases (i) Elliptic type of partial differential equation, for which B’ < AC. In this case, dy/dx is complex, yielding two complex families, complex conjugates of each other. Example: Laplace’s equation 2& + Fe _ 0, where A = 1,C=1,B=0, and BP < AC. ee The boundary for the electrostatic problem will never be a characteristic curve, since for elliptic types the characteristic curves are complex. If ® is specified on the surface, it will be known everywhere inside. (ii) Hyperbolic type of partial differential equation, for which B? > AC. In this case we have two real families of curves. 2, 12 « o, where A=1,B=0,C=-1 ¢ ar Ean: Wve qution 22 (setting y = ct) and B’ > AC. Equation (1) becomes dy’ - de =0 — di/dy = #1 + dx/edt = #1 + x = ct + constant.340 Theory of Partial Differential Equations The two families of curves, given by x+ct=pandx-ct=A, are shown in Fig. 8-5. Note that the curves must inter- sect at right angles since the slopes are +1, respectively. (iii) Parabolic type of partial differential equation, for which B? = AC. In this case we have one family of curves dy/dx = B/A. Example: Heat conduction equation # =a & = 0, where A = 1,B=0, C=0. ‘The characteristic curves are determined = jt from Ady" = 0 — dt =0, or t= constant oo (Fig. 8-6). Once @ is assigned along the J i characteristic curve, ca is known, so that + ab 10 ox —z will be determined. Hence, or ax one can assign only & (not both ® and 86/an = a/en); otherwise one may over- determine the solution. Fig. 8-6 8.2.2 The One-Dimensional Wave Equation In the previous section we showed that for the hyperbolic type differential equation the characteristic curves are given by x + ct = y and x - ct =A, or x= (p+ AY2 and ct = (1 - AY2, so that a bod add =2(5+3] - ax Op OA. ac op aes “*la-a) - ae “Loe aaloe aMethods in Mathematical Physics II 341 Substituting into the wave equation: fo 1%@ _ | & oe ee | ain 4 2 ob | ob = = a 0 = SF = 0) only ® = fi) + Alu). o ab ob aa 0 = a: gtu)only => & = f(A) +f(u). Combining, ()— B(u,A) = FQ) + Gi). The D*Alembert solution of the 1-dimensional wave equation is given by 2) &(x+et, x-ct) = F(e-ct) + Geter), where x + ct represents a wave travelling in the -x direction and x ~ ct represents a wave travelling in the +x direction (c = wave velocity). For x from 0 to L, the characteristic curves are as shown in Fig. 8-7. Question: Under what conditions can F(x-ct) -— p=L and G(x-cr) be found, i.e. under what conditions can one find ®? ct (A) Suppose & is known along OA and OB. Can eae one arbitrarily assign 06/On and thereby find ® in the shaded domain? Consider the following: Along OA:®|,,- F(0) + G(u), where p runs from pt = 0 to p1 = L, is known. Along OB: ®|.,= F(A) + G(0), whete A runs from 4 = 0 to A = L, is known. Hence, in the shaded domain OACB, wherein 0 < A < L and 0< HSL, we now know F(0) + GQ) + FQ) + G(u). Therefore, & = FQ) + G(u) is known, where A and p1 take on their respective limits.342 Theory of Partial Differential Equations Since @ is now known in the shaded domain, one cannot arbitrarily assign normal detivatives at the characteristic curves (at least inside this domain), because the solution is already known, and one may, as mentioned before, overdetermine the solution. (B) Suppose (x,t=0) and <(z,t=0) are known for 0 2 u@)- wea Ge) = fuera Hence, in the interval0 $<’ L and O 0, then on physical grounds one would expect the solution & for all time from 1= 0 tor and for all x from x= 0 tox = L to be known. To show this mathematically, consider the following: In the desired domain, 4 = x - ct runs from A = L to A = -°, while p = x + ct runs from p = 0 to p = +e, From discussion (B), F(A) and G(u) are known. in the domain 0 < p/A < L. Hence, by hypothesis, (© F(-ct) + G(ct) is known for > 0. (1) F(L-ct) + G(L+e0) is known for t> 0. From (6), since G(j1) is known for argument p1 from 0 to L, F(-ct) is known from argument ct = 0 to ct = L, or F(A) from 4 = 0 to A= -L. From (7), since F(L-ct) is known from L-ct = 0 to L-ct = L (or from ct = L to ct = 0), then G(L+ct) = G(2L+ct-L) is known from argumental values of Lo 2L, i, G(u) from pt = L to p= 2L is known. Mapping procedure (Fig. 8-8) From (B), ® in region A is known. Knowing Gu) from p = 0 to L, FQ) from A = 0 to ~L is known. Knowing F() from 2 = 0 to L, G(u) from pi = L to p= 2L is known. Thus, & is known in region B. Knowing G(u) from p = L to 2L, F(A) from 2 = ~L to ~2L is known. Knowing F() from 4 = 0 to -L, G(u) from 1 = 2L to p = 3L is known. Thus, @ is known in region C. iH Continuing in the forward direction, regions shown in Fig. 8-8 will be 3L mapped out. Proceeding in the one backward direction will map out the aL remaining regions. forward | rf C|BIA aChapter 9 Heat Conduction 9.1 Fundamental Equations If x = thermal conductivity, E = energy content per unit volume, 5 = energy supplied per unit volume per unit time by some external agency, and c = specific heat per unit volume, then Q = -«VT, where Q is the heat flux, and E=cT. From the first law of thermodynamics, OE 7 TVQtS, where -V-Q represents the heat influx. Substituting for E and Q, we have: cl avres. ot Defining s = 7 and A=, we get or = V2 a AWT +s. a) Equation (1) is an inhomogeneous partial differential equation of the para- bolic type (see Section 8.2.1), and is of first-order in time and second-order in space. Theorem: If s is a given function of space and time, and T(r,t-0) is given atr=0, and if for > O either T(r,1) or (Vn at the boundary is given, then T(r,1) is uniquely determined for all > 0.Methods in Mathematical Physics It 345 Proof: Define A(r,t) = T(r,t) - To(r,t), where T;(r,1) and T2(r,1) both satisfy the same conditions above. From (1), oD AW?T(r,) = 5 and Shieh ~AV°72(r,2) = s. Subtracting, we tae AVF = 0, and after sauiiplying by F, we get 0! F -agvg = a(visva- 9). Integrating over all space, (2 dS - 2, ala at a [rv Fas af ‘dt. By hypothesis, either F = 0 or VFn = 0 on the boundary. Integrating over time: Linl - - a fafa i Jud pFa - fl2 Since F= Oat t=0, we ate left win |Z. ay negative quantity. Since - the integral on the left side is a positive quantity, this result is clearly impos- sible unless ¥ = 0. 9.2 Infinite Medium (A) HOMOGENEOUS SOLUTION (s = 0) Consider the homogeneous equation T 1m T _ yep-o, qd at AVT = 0. Performing a Fourier analysis, oie jer Q) Te.) ape lr kne"ak.346 Heat Conduction Substituting into (1): ani ars rer ecs}e a Multiplying by e“™’* for k’ an arbitrary vector in k space and integrating over all space: eae [ak (EMD, er ale = see ar 4 jew a MEF WA}B' -k)=0. qd uD, MRK) =0 = see = -2RFK)). The last step follows since k’ is arbitrary. Expressing F(kf) = fe *™, we have from (2): oe or Te) = Goa lWerek = Teo) ee a Thus, @ MW- = gardener, so that knowledge of 7(r,0) yields knowledge of f(k) and therefore T(r,!) for all time. Green's function (Go) If T(r,0) = 8%(r - ro) at = 0, then for > 0, T(r,f) = Gol - o,f). From (3) above, ek SL Jee- rete - mF I) - G oeMethods in Mathematical Physics II 347 Substituting into (2): Ten) = aor omen 1 je (xx) EM, iho ja-2)-BM, * Gay a3 i eo a Jeo Evaluating +2 2 i 7 fret ol ale) 5 = Gy)Mersan, and substituting, we have - 3, ad Te.) on (Wye ar. Thus, el Gale-robt) = mau? ‘ar in 3 dimensions. (4) : Go ae eee om (nan? . Properties of Go(|t-rol,t) for t>0 1, Ge = av?Ga , ic. G satisfies the homogeneous equation (1). Proof: For t > 0, T(r,t) = Go. Thus, since T(r) satisfies the homogeneous equation, so does Go. 2. Seate-rapnar =1. Proof: Joate-ranar - sll Poet ee er, sas dxdyde ag? [nay] (4x1 ma348 Heat Conduction 3. Ast + 0*, Go > 8°(r - ro). 4. If T(e,0) is known at = 0, then for #20, T(r) = [Go(r-r,)TW’ Oar’. Proof: @ 5 7 av} “Si 7 av? ete v.07 0dr =0 from property 1 above. Thus, T(t) satisfies the homogeneous equation. (b) Ast +0, T(r) + T(r,0) since Go(r-r’,t) + 8°(r-1’), so that the boundary conditions at t= 0 are satisfied. (c) Asr + ©, T(r,!) + O since Go(r-r’,t) — 0, so that the boundary conditions as r —+ © are satisfied. (B) INHOMOGENEOUS SOLUTION (s # 0) The procedure, in general, is to find a particular solution of the inhomo- geneous equation and the general solution of the homogeneous equation, and then combine to form the most general solution. Particular solution ‘Theorem: If one can find an inhomogeneous Green's function Go(t-To,t-to) such that (6) 5 7 av] Gaeta) = B(r-1.)8(-f), then a particular solution can always be written as: Ted ~ [Gar ese teed Proof: Define Go(r-ro,f-to) = Go(t-o,f-to) for t- fo > 0 and = 0 for t= t <0, where Go is the homogeneous Green's function. Making the transformation f' = t~ fo: Go = 0 for t’ < 0, so that the left- and right-hand sides of (5) are zero (the right-hand side by virtue of the 5-function property). Go = Go for f > 0, so that the left-hand side of (5) vanishes, while the right- hand side again vanishes by virtue of the 6-function property.Methods in Mathematical Physics Il 349 We now seek the behavior as’ 0*. Considering equation (5), from property 3 above, Go = Go > 8(r-0), while the right side ia another delta function that blows up as 5°(r-r0)8(7’ + 0+). Question: Do both sides have the same singularity? Consider, therefore, the following: far [2 -av" aera df B-rB) = Bene). From propetty 3 above, lim Go(-rof’) = 8°(r-r0), so that the left side of the equation becomes ne. q fe we GoAe—Togt’) - afer VWGAr-Tot’) = 1+ I. Term = |Gar-tof)| = Gule-roste) — B%e-r.)as e+ 0. Term II = -Ae(V"Go),-9. + 0 as + 0. Hence, both sides approach 5°(r—ro) as 1’ + 0+, and the proof is complete. ‘We have shown (5) to be satisfied for 1 < 0, for > 0 and for ’ + 0*. Note the dependence of the inhomogeneous Green’s function (Go) upon the homogeneous Green’s function (Go). This dependence hinges on the fact that there is a first-degree derivative in time. This will be seen again when the wave equation is analyzed. Given an initial temperature distribution T(r,0), one can consider this as being built up of many temperature pips, i.e. many 8°(r-r). The resultant T(r,1) for all > 0 can then be found by summing up the diffusion of these pips throughout the infinite medium, i.e. © — Tet) = JGor-r' NT’, O)ar’. If, instead, one is given an external supply of energy s(r,t) at > 0, then the diffusion of heat is expressed by QD Te) =f[G.0-r se Naerar, i.e. the diffusion takes place in a manner similar to that of a temperature distribution.350 Heat Conduction General solution Theorem: If s(r,t) for t > 0 and T(r,t=0) are given, then for > 0, ® Ten -feur-rare ody + Jarl garrs-roe ed where the first term on the right is the general solution of the homogeneous equation, and the second term is a particular solution of the inhomogeneous equation. Proof: (@® T(c,1) satisfies the inhomogeneous equation oo (3; - av }ren = ste. (ii) At r=, Go(r-r’,t) + 0, while Go(r-r’,1) = Go(r-r’,.) + 0 for t>Oand Go =O fort<0. Hence, T(r) > Oat r=%, (iii) If T(r,0) = 0 and s(r,t) = 0, then 7(r,1) = 0 everywhere and for all r. (iv) Ast + 0,708) + Tte.0)- Hence, T(¢,1) satisfies all the necessary boundary conditions. Consider the expression JafGeee enue nar. Since Go(r-r’-1’) = 0 for r > t, we can rewrite equation (8) as © Tep=JeKe-r ore ode + farfgarr.-es(e nde ‘We consider next the problem of the semi-infinite medium. Question: Why should one get, or even expect to get a new Green's function? ‘Answer: Reflections may occur at the boundaries, and the Green’s function should incorporate these boundary properties in the description.Methods in Mathematical Physics II 351 9.3 Semi-Infinite Medium Boundary conditions (Fig. 9-1): (i) Forall time #2 0, 7-0, = 0 (i.e. in contact with x20 aheat reservoir at temperature Ty=Oatx=0),and ? Tee==,1) = 0. (i) Forx 0, Ste) AVT (2) = sGct) given. (ii) 76,0) given. Fig. 9-1 Consider the following related diffusion problem for the case of an infinite medium (essentially an image problem). Assume that the above temperature distribution and source s(%,f) are reflected in the —x plane, with negative signs for s(x,t) and T(e,t). Then if s’(x,f) and T(t) refer to an infinite medium, T(x,t=0) = T(x,t=0) for x > O and = - T(x,t0) for x < 0. S'(6) = sG6,) for x > O and = - sx) for x <0. From equation (9), Section 9.2, for > 0: QTE) = fo-¥.NTe Od + farfgae-x ers sae. Since Go(x-x’,f’) and s'(2’,’) are even functions in (x - x’) while T(2’,0) is an odd function in x’: T(x) = -T(-x) = =T(0,) = 0. Thus, the solution of the semi-infinite medium problem is the solution of the infinite medium problem for the +x domain. In other words, for a semi- infinite medium, for x > 0: Q) Te) Flex ~ Golxtx’ D]T (t-O)dx + farf[ouc-xs-r) ~ Golaxta’ sr) fd.352 Heat Conduction Semi-infinite Green’s function Let | Tet] = 8(xo-x’) for x > 0 and x’ > 0. The Green’s function G(x %o,t) is defined from T(x) = Geto) = Golx-Xo) - Golx+%o,)- From equation (2): Ten) =[oo-x ore node + farlowxnrsw nue. [Note here that G(«,x’), a function in 3 variables, x, x’ and 1, goes over into a function of 2 variables, (x-2’) and £.] Note that equation (3) has the same form as equation (1) for the case of an infinite medium, except for the domain of integration and the form of the Green’s function. Define the semi-infinite Green’s function G(x’) = G(ex’,f) for 1> O and =O forr<0. Properties of G(x,x’,t) @ ( 7 avila. = 0. This follows since T(,t) satisfies the homo- geneous equation. Gp (5 - av geo. = 8(x-x’)8(1). For 1 < 0, both sides vanish, while for 1> 0, G~= Gand the left side vanishes from (i) above, while the right side vanishes since 8(t #0) = 0. ii) Ast > or, G&D) + B(x-x). (iv) TQ) = Jove NT t-0)dx! + facfosc-vs-ny (xt)de. = Osince Gl-x'1) - Gol’ Thus, the Green’s function G(x-x’,t) obeys the same boundary conditions as does the temperature. Importance of the Green's function: The Green's function is completely determined, in principle, by the particular type of boundary in the problem, independent of initial temperature distribution and independent of heat source. Once the form of the Green's function is known, one can plug any given s(x,t) and 7(r,0) into the problem in order to solve for T(x,f) at some later time t > 0.Methods in Mathematical Physics II 353 9.3.1 Age of the Earth (Treatment due to Lord Kelvin) Assumptions (i) Attime t= 0 the earth starts to condense. (ii) Non-curvature of the earth, i.e. a semi-infinite medium. (iii) T= 0 at the surface of the earth today. (A fair assumption, since it is small compared with T(¢ = 0) = 120°C.) (iv) Uniform temperature of 1200°C as the melting point of rock at f= 0. Given (i) T= 0) = 1200°C. (Actually somewhat greater due to pressure.) (ii) A= We = 0.005 cm’fec. 6 (iii) s= 5 = ae = 10" Keo, (iv) At present time, \F = 32% = 327/10%em. [S From equation (3), Section 9.3, Q) Tee) = Joc moyae + faloo-vpnwe nde. Assuming T(’,0) = To = 1200°C and constant, and s(’,’) = so = constant, TOs) = 1: fow-x.nae + sJarlee-r sna. Differentiating, and evaluating at x = 0: if |SGG-x'.1), fel Pa. @ i LC 7 to aeeres |354 Heat Conduction Using the properties of Ge’): |? |. BGoK-x's)| oo S ax e = | 2Gole-w') AGolety’, 9 Gale) ox ar’ ee Thus, OGD gy POLED gy 2, f Gee af Se ae = apa le he eee (many? Substituting into (2), Te eS x L (nar)? f or Ta_, 250ft | ae) i One now compares the value of |22 | as calenlated from (3) using known no constants with the known current value of |S 1 = 10° years = 3 x 10'Ssec: . Expecting a value of 0 = (w5-10°-3-10")” = (3750-10) = 3-7:10%, so that calculations from (3) yield er) 1200 2610n 10 ee | 21-10° 21-10" 3000’ while the current known value is ae se jo km 10%em 300° esMethods in Mathematical Physics II 355 When Lord Kelvin first computed t, he neglected so which we see contributes the greater portion to er L in the ratio SO - 5t0 1. 9.3.2 Temperature Variation of the Earth’s Surface In contrast with the previous problem for which the gradient O7/Ax at the surface of the earth extends through a distance of kilometers, we are here concerned with a gradient that extends through centimeters, or metets, i.e. we are seeking a daily or yearly variation in temperature, with boundary conditions that are functions of time. Boundary conditions @ T=A sinot at x = 0. (If not a sine function itself, then at least a Fourier series in sines.) (ii) T(&,1-0) given at r= 0. (ii) 5@,1) known for 12 0. Whenever the boundary temperature is a function of time, as in boundary condition (i), itis only necessary to find a particular solution Tp that satis- fies the homogeneous equation 0T,/6t = AV"Tp. Since Tp is given at x = 0, if we define T’= T'- Tp, then conditions for T° are: (iv) T’=Oatx=0. (v) T’ will satisfy OT/ot= AV°T" + S. (vi) T’ is known at t= 0. In this manner we have reduced an inhomogeneous boundary condition to its corresponding homogeneous boundary condition for a semi-infinite medium. We seek, then, a solution of the homogeneous equation: OT, or Let T(x=0,1) = Ae™, the im: our boundary conditions at x () = Aav'7,. ry part of which we seek in order to satisfy356 Heat Conduction Assuming a particular solution 7)(x,t) = F(x)e'®" and substituting into (1), we have iaF (x)e™ = 2V7F (xe, or oc @ SP. Bre = FOFw. Choosing @ > 0, solutions to (2) are given by @) FG) = Bexp[(in/4)(a/2)x] + C exp[-(iny4y(a/2)"x], where B and C have yet to be determined. Since e™ = (1+i//7, and since at x=, exp|(in/4)(o/)'7x] = 2, B must = 0, since F(@) = 0. Also, F(Q)=C=A, so that C= A. We finally have for Ty: Txt) = Aexp[-(iry4X(o/ 2)x] exp [iar] = Aexp[-(@/22)(1+i)x + ior] = Aexp[-(@/2A)7x] exp [i€eor-(a/22)7x] s Since we seek the imaginary part as the physical solution, 4) Txt) = Aexp[-(@/22)"7x] sin [or-(@/22)x]. [Note that 7}(x,t) has a part that decays in the +x direction, and includes phase changes that accompany varying changes in x. Note also that the equation 87)(z,t) /ot = AV77)(x,2) is a real equation when is real. If Tp is complex, one can always equate real and imaginary parts. The Schrédinger equation, however, is not real since 4 is complex.] Yearly variation For soil, 0.002cm? | _2n___2n Se Te 910 acc (ay 20)” = [273-107-410]? cm = (12-104) = 710% em,Methods in Mathematical Physics Il 357 Atx=0, Tp = Asinayt (Fig. 9-2a). T, Atx= 1 meter, (oy 2A)x = 0.7 = 1/4. ef si. Ty(2,) = (A/2)sin(wyt-1/4) (Fig. 9-2b). 0 Atx= 4 meter, (@20)?x=n. Fi exp [oy2a)*x] = 12". T,Ax,1) = (4/2)sin(wyt-n). Thus, at a depth of 4 meters, the earth is coldest in summer and hottest in winter (Fig. 9-2c). Daily variation Yearly and daily phase changes will be equal when Vaoaxa = Vay. or When Xa ve, 1 % (@a/o,)" = 59° Thus, x= 1 meter (yearly) —- x= Sem (daily) to get 1/4 phase change, and x=4 meters (yearly) = x= 20cm (daily) to get m phase change.Chapter 10 The Eigenvalue Problem 10.1 Eigenvalues and Eigenfunctions In Section 8.1 we considered the particular example of the wave equation type of differential equation, Qa We-=S =o. v For a one-dimensional vibrating string, © is the displacement and v” = T (tensionYp (density). For the case of 2 dimensions, we are dealing with a vibrating membrane. We seek the characteristic frequency of the string/ membrane. Assuming (r,t) = ‘P(r)e" and substituting into (1), we have mead Vrpyeitt 4 2 OF ior ae Setting A= w/v’, we get: 2) -VAM(r) = AM(r) Equation (2) is typical of an eigenvalue problem, where one regards -V? as the operator, ¥ as the eigenfunction, and 2 as the eigenvalue. (A) One-dimensional string For the one-dimensional vibrating string, equation (2) becomes FM) _ ae MO.Methods in Mathematical Physics II 359 Solutions are given by ‘P(x) = A sinkx + B coskx, where 4 =k? Since boundary conditions demand that = 0 at x = 0, we see that B= 0. In addition, since boundary conditions also demand that Y= 0 at x = L, it follows that A sinkL=0 =+ k= nn/L(n=0,1...%). Thus, due to the boundary conditions, 4 = k?= (n°x”y/L’ has certain discrete values, namely the eigenvalues. Substituting for k, we have M(x) = A sinnna/L. 52 ATX 1 2 Note that [¥2(a)de = fei Oe = “fa ~ cos2nmx/L)dx = ° 7 : Thus, since we want to normalize the Yn, we choose A = (2/L)”, yielding for Wx(2): QB) B(x) = (2/L)"sinnnx/L, where An = k7= (n° YL? (n= 1,2...2), satisfying —V¥,(x) = An'Pa(x). Characteristics of the eigenvalue problem 1. ~ d¥dx" is a good Hermitian operator for all functions with boundary conditions ¥,, = 0 at x = 0 and x = L, i.e. Ta fe 4 | fu [- Joa for (--& oa, where ® and are in the class of functions admitted by the boundary conditions of the problem. Proof: [ott fe oar - oof oun360 The Eigenvalue Problem 4 2, Pensa dE = Ban. 3. The ,(x) (n = 1,2...2°) form a complete and ortho- normal set for any function ‘P(x) satisfying W(x) = Oatx=Oandx=L. (B) Two-dimensional circular membrane Fig. 10-1 ‘The eigenvalue equation [from (2)] is given by -V"H(p,9) = A¥(0,9), where A = @7/v under the condition that '¥ = 0 at p = po (radius of the membrane) (Fig. 10-1). To determine the eigenvalues and the eigen- functions, we can expand 'Y(p,p) as ¥(p,9) = )fn(p)e'"®, since on a circle the e”®, or equivalently, the Fourier series, form a complete set, and any function g(p) can be expanded in a Fourier series. Substituting into the eigenvalue equation: ELS velo Multiplying by e“"® and integrating over the circle: probe (aPs bs) dp’ p dp Using the orthonormality properties of the e"” yields to), 1 do), am 4) -|fe) - 0. O wow fe) ae d d & Set Vip =2,so that 5 >= WRG = apt hag md @ becomesMethods in Mathematical Physics I! 361 Dividing through by 2 we artive at Bessel’s equation with integral m values: : Te, 1A (mga 0. The m are necessarily integers since they arose from expanding ¥(p,@) into a Fourier series. In Section 7.5.4 we discussed the case for m being integer- valued, for which the second solution of Bessel’s equation is given by the Hankel functions H}, or Hx . Thus, the most general solution of the above vibrating membrane differential equation (5) is Ju(2) = AJn(2) + BHn(2)- Since we want fin(z) to be finite at p = 0 (i.e. at z = 0), while on the other hand Hn(z) possesses a logarithmic singularity at z = 0, we must have B= 0. Thus, (© — fu(Z) = AJn(2). Since ¥(posp) = O at p = po for all g, it follows that Jn(/Xpo) = 0. Itis this boundary condition which gives 2 its discreteness property. We now seek the roots of Jn(z) = 0. For m = 0 (Fig. 10-2), roots are Zo1, Z02, 203, etc. For m # 0 (Fig. 10-3), roots are Zmn, i.e. Zm1, Zm2, Zm3, etc . The origin is not included as a root since at the origin, z = 0, or VAp = 0. If p # 0, then 2. must = 0, which, however, means no vibration. Hence, roots are Znn, where n = 1,2... Jof2) InZ) Fig. 10-2 Fig. 10-3362 The Eigenvalue Problem Corresponding to roots Zmn, there are the eigenvalues an = Can / 0)", and the corresponding eigenfunctions (8) Yna(059) = Aln( Tenn ple™*. Degeneracy For the case of +m and -m (but ¥ 0), if we recall that J+m(2) differs from J-m(z) only by the factor (~1)", then it follows that both J+m(z) and J-m(z) will have identical roots, giving rise to identical eigenvalues. The Yin(P,?)s however, do not depend only on Jn(z), so that we have different eigen- functions that give rise to the same eigenvalue, i.e. degeneracy. Tabulating the roots in increasing order, we have: Zo = 2.405; 2 = 3.832; z21 = 5.135; zon = 5.520; 231 = 6.379; ete. From A = @7/v”, we note that Amn has a frequency, ®mn = VRmn 0 = Zmn/Pos associated with it. Thus, the eigenvalues of the circular membrane are given by Zmn0 (ee We now seek the vibration patterns for the above characteristic frequencies. m=0, n=1 This is a non-degenerate case, with only one eigenfunction, Wo1 = AJo(201P/Po). Note that for p < po (i.e. within the boundary of the problem), zo1p/po < zo1, and Jo(z) has no roots for z < zo (Fig. 10.2). Hence there is anode only at the boundary, and the amplitude of the oscillation is the amplitude of Jo(z). (9) mn = m=1, n=1 Here we have degeneracy (m = #1). Weir = AUi(zup/po)e"®. C21) + 2-1) = Adi(ZP/po)cosp. C211) - G2DP-11) = Adi(Zup/po)sing. Since the left-hand sides of both equations are separately eigensolutions, Ii(211/po)cosep and J(z119/po)sing are also eigensolutions.Methods in Mathematical Physics I 363 In fact, any linear combination of these degenerate solutions will also be eigenfunctions. Hence, for m= #1, n= 1: Wxis(P.p) ~ Ji(2up/p.)sin(p-8), where 0 is an arbitrary constant. oe Question: What will the vibration patterns look like, and vy; where else but at the boundary will ¥ = 0? From the form of Y above, we see that Y = 0 at p = 0. ~ Since J(z) is zero at the origin (Fig. 10.3) and at the root zi (Le. when p = po), nodes occur at p = po for all @, and along the nodal line @ = @ (Fig. 10-4). Fig. 10-4 m= £2, n=1 In this case we have degeneracy again. t2ip ae, 2 nodal lines P22,1 = Al2(221p/po)e®. Writing as before: TY (2)M.21 + V-2,] = Ads(2219/p.)eos20. eo C2DP¥ a1 ~ V2] = Adse1P/po)sin29. xs Since J2(z21p/po)cos2p and J2(z21p/po)sin2 are eigensolutions, a linear combination of the two will Fig. 10-5 also be eigensolutions. Thus, 2.405p,/5520 W22,1(P,9) & Ja(Z21P/po)sin(2p-8), so that Ps2,1(p,p) = 0 at p= 0/2, or p= 0/2+ 7/2, which results in the two nodal lines shown in Fig. 10-5. m=0, n=2 As was the case for m = 0, there is Fig. 10-6 degeneracy. Wor = AJo(Z02P/Po) = Jo(Zo2P/Po) = 0 at z= Zo1, 02, etc. (Fig. 10-2), ie. Woz = 0 when 202P/Po = 202. = _P = Po, and ‘oz = 0 when 202P/po = 201, oF when p = zo1Po/z02 = 2.405po/5.520. The nodal lines for all ~ ate shown in Fig. 10-6. Nodal lines for m # 0 appear as a result of degeneracy. The location of the nodal line in these cases is extremely arbitrary, due to cylindrical symmetry.364 The Eigenvalue Problem 10.2 Harmonic Oscillator/Free Particle in a Sphere (A) HARMONIC OSCILLATOR In the — oscillator ae (Section 7.3.1), we saw that by setting p>-ih® andé -+ th Z in the Schrddinger equation, E= Ee Viwe ee (-£ Eveyly Substituting V = kx7/2, multiplying through by 2/fia = 2(m/k)""/M , and let- ting z = (Vmk/fn)” x, we arrive at ay 2EY ae aoe ye Setting fiz) = '¥(z)e"” results in the Hermite equation w ate. 2 19, maf) =0, where 26/ffia = (22+ 1) = E=(4+1/2)he. Solutions of the Hermite equation were given as f(z) = H(z), so that Y,(2) = e*H,(z), and unnor- alized solutions of the harmonic oscillator are given by Q — ¥(ax) = &**HA(0x). Since JHtsz\Hale)e* de = By", we have as normalized solutions: a @)— Wa(ox) = (al eH Ko2), 1 vihere 2 = ox = | :Methods in Mathematical Physics It 365 The eigenvalue equation is given by @ -£e- vly = Ebay 2m where () Ex = (n+ 12)(/m)”. * - Fig. 10-7 ©) Sala) Yom x)A(0X) = Ban. Oo 2 x Fig. 10-8 Fig. 10-9 Forn=0: Woolox) = Ae”, For n= 1: Yoi(ax) = Bre®*”, For n= 2: Wax) = Ce**(Dx - F). Note that the lowest eigenvalue, Eo, corresponds to the case of no node (Fig. 10-7), while the next highest eigenvalue, E1, corresponds to the case of 1 node (Fig. 10-8), and the next highest eigenvalue, Ez, corresponds to the case of 2 nodes (Fig. 10-9). These are to be compared with the case of the circular membrane. (B) FREE PARTICLE IN A SPHERE a V=0 fora free particle, Schrédinger’s equation becomes E vine) = EW(r), where Y(t) = Oat r= a (radius of sphere) and is ae, irweset e~ HF, we have for Schrédinger’s equation: -VY(r) = V(r) = AY(r).366 The Eigenvalue Problem Expanding ‘Y(r,0,g) in terms of the complete set of spherical harmonics: ae ¥r0.9) - YVR n(0.9)- be Be waroe) + Puce = FSF (Piao) : 2¥inf0.9) , 1 Halts] Eola in 20°" 08 Paint | + Len RO) =0. Substituting OF in( 8, FYin( 0, eS sing Mn) 2. ® 10+ 1¥in(0,0): Tg ee ee Ropaow = 0. Employing the orthonormality relations for ¥im(0,9), we have 1 dy dRir) Le : o Fare A fe )-0. Setting Rin) -A2, Substituting in (7): ap, 1 _ [0 oto - a 4 HD, 1HO, py) C2 BIO - 0.Methods in Mathematical Physics I! 367 If we now let z= kr, we have dQ), 1 fe) [ a : ® ae * za at z fe) 2 which we recognize as Bessel's equation with A= + 1/2. Solutions are given by Sk) = const. Jnyal2) + Rin) = const, ue) Roots of Ji+12(kr) = 0 are given by kin, where n= 1,2... = Ain=kin- Finally then, we have for the eigenfunctions: ©) Yaslr89) = const Ae” v.50). 10.3 The Variational Principle In Section 5.3.2 it was pointed out that a Hermitian operator H satisfies Su nea - [fe mea) where dr includes the range of the particular problem. igenvalue characteristics For HY = WP: (A) All the eigenvalues, , are real. Proof: Sv" H¥de = Af¥'¥dr, from which _ Serva ye - Pena Seva | fever cr (fi war) Seva (B) If (@): HY = AW and (b): HY’ = 2°V', then. fP""Wdr = 0, which says that eigenfunctions ‘¥’ and are orthogonal.368 The Eigenvalue Problem @ Fora #2’: Proof: Multiplying (a) by ‘Y”" and integrating: @ — fernvar = aferwar. Multiplying (b) by ‘Y” and integrating: ® frnva-rfewa - [few var) Substituting (fvnwar] = fre" var into (4): 6) Setmedr = xfiv'war. Subtracting (5) from (2), we have (A - ¥)f'#""Wadr = 0, so that f'¥"War = 0 ifhen, (ii) For = 2’: In this case we have degeneracy, since two eigenfunctions and" give rise to the same eigenvalues 4 = 2’, and Y and Y’ are not necessarily orthogonal. However, one can use the Schmidt orthogonalization method, discussed in Section 6.4.1, to make them orthogonal. Since the character- istics above hold for H real or complex, if H is a real Hermitian operator, and if HY = AXP, then HYP" = 2X8", Since this equation is satisfied for the real and imaginary parts, separately, of any complex eigenfunction, we will always choose ¥ to be real. Functionals F(Y) Definition: For every given function ¥, there is a unique value for the func- tional F(Y). fervar An example of a functional is the expression <‘P|H|¥> = f ont: edt Functional Variation 5F(¥) Definition: 5F(Y) = F(Y + VY) — F(#), where terms of order o~r) are neglected. Example (i): If FOP) = f¥ar, then BFW) = fovsevyar - far = [[2vow+owy er = 2fvordr.Methods in Mathematical Physics I 369 Example (ii): WFC) = - fev*war. Fe) = — Sverre nde + fovwyrac -- fervyynas + foveyar. ‘We saw in Section 10.1 that V” will be a good Hermitian operator if we set W =O at the surface boundary. Thus, FW) = ~ fevevar = Sivas where both expressions (a) and (b) represent the same functional. @ BFC) = - forvevar - fev'owd. Due to the Hermitian property of V”, fevover = favVWar, so that BFW) = -2 fowv'war. 2 feowvanar - 2 Jov-vvar 2 fowvey nds - 2 fow-vvar. (b) BFW) = 2fv¥-VEYdr As above, if 8 = 0 on the boundary, then 3F(#) = -2 fowv'War, in agreement with (a) above. 10.3.1 Distribution of the Eigenvalues Let HYY; = 4;, where H is the Hermitian operator for the particular Y; that satisfies the boundary conditions ; = 0 on the boundary. Let us arrange the A such that A1 < Az < A3 $24 S... ie. a spectrum of eigen- values. Define F(¥), the functional in question, as Yd fea where the domain of integration is the allowed domain of the problem (such as the length of a string, area of a membrane, etc.), and where ¥ satisfies the proper boundary conditions. (1) SPREE? =370 The Eigenvalue Problem Theorem (i): If HY; = AY; and if ¥ = V;, then = ri. Proof: yr cee ee elie "Sitar Svar Theorem (ii): If 8<‘Y|H\> = 0, then HY = Y, ie. ¥ is an eigen- function (since <|H}Y> is just a number), and is an eigenvalue. Proof: Spx wiomyar ofvrrvar foods fra fea fra B = 0. Since for Hermitian operators, fovHWar = fWHOdr, — afovrrvdr 2feivar fvovar Se ‘ Q) CHIH = Joa eer - ¥)dr] =0. Choosing a point rp within the domain, such that 6 = 8°(r - rp), we have HY = CYHNY>Y within the domain. Theorem (iii): If Y =; + 3, where 39 is an infinitesimal, it follows that HIND = As + Op), implying that <'PJHTP> is stationary at Ws. (For example, iff fla) and X= + dx, and if f= fl%o) + O(8x)*, then fis stationary at x = x9.) Theorem (iv): The minimum value of “P|H}P> is Au (the lowest eigen- value) for all ¥ satisfying the boundary conditions. Proof: If a minimum, then 3 = 0 and “P|H}Y> = i. Hence, choose the lowest Ay = 1.Methods in Mathematical Physics I 371 Consider the following cases: (1) Vibrating membrane with a fixed boundary at S. (2) Same vibrating membrane as in (1), except that constraints are added at certain areas 51, S2, 53, etc., so that ‘¥ = 0 there. For both (1) and (2) we have HY = AY, where A = «7/v? and H = -V?, Theorem (v): Regarding cases (1) and (2) above, (21) $ (At) 9)» where we are dealing with different wavefunctions, since boundary conditions for (1) and (2) are different... Proof: Since the boundary conditions of (2) include, and therefore satisfy the boundary conditions of (1): (Qu)gy is a minimum of <‘P{HIY> for all Y satisfying the boundary con- ditions of (2), thereby also satisfying the boundary conditions of (1). (Quy is a minimum of for all satisfying the boundary con- ditions of (1) only. [In Section 10.1 we discussed the vibrating circular membrane and showed that the roots Zmn for m # 0, i.e. equivalent to case (2) above, are greater than Zon for m = 0, ie. equivalent to case (1) above, and Amn & rn & Zan -] ‘Thus, the effect of constraints is to raise the minimum 41, and as a result, Ory S$ Ady Theorem (vi): Let HY; = 411, where A1 is the lowest eigenvalue. Then the lowest value of is 22 for all satisfying the boundary con- ditions and obeying the constraint condition [4dr = 0, i.e. orthogonal to w. Proof: We use the method of Lagrangian multipliers, ie. GB) BCHLHNE> + mBfwrHide = 0, where is regarded as arbitrary and 1 is fixed, i.e. d¥1 = 0. Substituting for 3<‘|H|P> from equation (2): 2 fea [Jowtery - cmunpow)ar] + mafovor, + vewar = 0.372 ‘The Eigenvalue Problem By hypothesis 8'¥; = 0, so that after setting m’ = fae, we get ie fea ‘Again choosing a point rp within the domain such that 8¥ = 8°(r ~ rp): [Jou(anw — WINDY + mY, jdt] 0. (4) HY = cannot be 21 since the corresponding W1 does not satisfy f¥2dr = 0. For all other 2,, however, can be found which satisfy ['¥Midr = 0. Thus, the next lowest is A. Theorem (vii): [Generalization of Theorem (vi)] ‘The lowest <‘P|HPY> is An for all ‘P satisfying the boundary conditions and (fuvidr = 0, where i= 1,2 ...n-1. For example, for ¥1,i= 1 and n= 2, implying 42; for 2,7 = 2 and n = 3, implying As, etc. Consider any set of functions fi :-« fa which may or may not satisfy the boundary conditions. Define m(f; - *- f,) = minimum<|H)P> for satisfy- ing the boundary conditions and f¥idt = 0, i= 1, 2...n. Theorem (viii): The maximum value of m(fi -- -fx) for all fi *-- fa is Anet- Proof: @ Wfi=V1, m(P1)= 42 > maximum[m(¥;)] = Aa. Theorem (vi)Methods in Mathematical Physics I 373 (ii) For any fi, choose = a1"; + a2'¥2 (where Y and 2 are eigen- functions of H with eigenvalues 41, A2) to be orthogonal to fi, ie. Sivas = Shartras+ fhastadr = 0 — & = fee We have Sats + anna + a:2)dt BIEN = =< Jae + a2e,yar so that, substituting PHU dr = ay PUA dr = Ady, oe aid + aida ata = 2. 2. and m(fi) = min PUAN! = min 24 * 292. < oy for arbitrary fi, and aj + az maximum m(fi) = 22. Theorem (ix): For the resonator problem discussed above, in which (2) refers to the addition of constraints, (22), (Aa)y » Le. all frequencies are raised due to the constraints. Proof: For any f, we seek min <“Y|H}¥> for satisfying the boundary con- ditions for (1) and for (2), and fi¥fdr = 0. Any satisfying the boundary conditions for (2) and f¥¥fér = 0 will also satisfy the boundary conditions for (1) and fitfdr = 0. Hence, [mcr = [mc], = max [mth | so that Y = O at x= Oand L, at y= Oand L, and atz=0O and L. Thus, VPaan(xyz) = (nh + r+ hyn? (sn myRx Many, Manz z sin“ sin~y lane =~ dan Pnn(292)- Hence x Anon = 73 be = lima=%. (ii) Consider a cube having an enclosed surface S. We require that '¥ = 0 on $ and in the region between S and the edges of the cube (Fig. 10-10). furvar femvar % ite SYED are = = war hdr ig. 10- £ i Fig. 10-10 Thus, CYIH}P> cave = “PHY v, integrating over a volume Vs of surface 5, for Y to satisfy (i). Since ¥ satisfying (i) constitutes a more restrictive boundary contin than salisying(), we have denosing yy 2 corresponding to min , ,and denoting 2, as heute corresponding to min &, since day 7 8,Methods in Mathematical Physics I! 375 Theorem: If -V¥, + V¥n = AaWn, where H=-V? + V, V isa function of space only, Yn = 0 on a closed surface S, and V 2 Vo (a finite constant) within S, then lim Ay = ©. Proof: (@ If V= Vo (const.), then —' ~V~, = (An - Vo)Pn. From our previous is however, | lim (An - Vo) = ©, and since Vo is a constant, we have Tim An = ne (ii) For V > Vo (inside S). Define H(E) = -V? + E(V- Vo) + Vo, so that dH/dé = V- Vo. We also have that H(+1) = -V? + V and H(0) = -V? + Vo. Consider H(E)¥n(E) = An(E)Wn(E), where & is regarded as a parameter. SO wo! 2 . a nee +e) He, a ce Multiplying by 'Y,(E) and integrating term (2) over the domain (taking into account that H is Hermitian), Siracancey ME? a ~ FURD ane) 146) de = dal) pee we a. Thus, since terms (2) and (4) are equal, term (1) must equal term (3), i.e. Setter vods = [0 wacoar - 282 Faire, and age) _ SMB Vode a& feo Since a 20, it follows that Ay(E=1) 2 An(E=0); but Ay(E=0) (Le. V= Vo) has no upper bound, so that A,(E=1) (i.e. V> Vo) has no upper bound.376 The Eigenvalue Problem Importance of there being no upper bound to the eigenvalues Theorem: If H¥q = hy'¥n and if lim Ay = ©, then the orthonormal set Wi WAY... forms a complete set for functions satisfying the boundary conditions. Proof: Define H = H - A1, so that HY, = (An - 01). Let be any function satisfying the boundary conditions, and define Fr = ¥- Yanhn, = where ay = f¥Yndr. It then follows that fFtwde = frPnde- Yan feed = frend - Lente = Sa iede Yew = fevede-fyvndr = 0. Hence, fF; ade =0, m= 1,2...1, ie. Fy is orthogonal tothe first n eigenfunctions. : Consider now [FnHFdt = [FnH¥dt ~ [Fn Yan%ndr. Eh Since F, is orthogonal to ‘W,, the second term on the right vanishes, so that Srittrcr = fr,tvdr = sy - Sow. fiva S = futvas - San firativar. Since H is Hermitian, = Sen fieati¥dr = Sn fietiade 10. 1) fener a = a Y cide ~ A). Thus Sratirade = fetvdr - Jaan - my). =Methods in Mathematical Physics I oT Recalling (Sect. 10.3.1) that 2 Anei ~ A, we have fettvar - Yan - 4) a _ reir: = tt [rea ria: 4 fervar fever Tria 2 Awd me fFidr < jee! ; Since f¥¥ar is independent ofn [given a , one caleulates f¥FAVAr], lim fF2dt = 0 = lim F, = 0. Hence, (1) Fre = ¥- Yann = 0, F and the Ym form a complete set. Short review plus additional remarks (If HWn = AnPn) (i) The ¥, constitute an orthonormal set. Gi) The , constitute a complete set if lim Aa =. 2 (iii) For H=- z 2 + # (a particular example of H = -V* + V), solutions were shown to be @ — W,(ox) = Hyoxe**?, where a = (Vmk/f), The eigenvalues are given by GB) Ex = An = (n+ Y2)hiw = (n+ Y2)hi(W/m)” + © as n> ©, Thus, for z real, ® Y= Lanaae®, for any function ¥ that vanishes at x = 4°, thereby making H Hermitian.378 The Eigenvalue Problem 2 (iv) For H=- = the V,,(@) are sinmp and cosmg, with eigenvalues Am = +m?, and as m — %, Aw + ©, Hence, sing and cosp form a complete set, so that for any function : 6) ¥(p) = Ln sinme + By cose). () IH = - Be Seine 2-5, te comesponding eigen functions ate Yin(0,@) with eigenvalues (I + 1) which + © as 1 + ©, Hence, the spherical harmonics form a complete set, and any function of and @ can be written: ©) — ¥(0,9) = YA, Yin(0,0)- dm, (vi) If H = -V? inside a sphere, then the eigenfunctions are given by 1) Vina 859) = const, 1A (0,9), and Hr 0.9) = Y Aine ED Yi). &Chapter 11 Wave Equations 11.1 Infinite Medium Consider the time-dependent wave equation a so-te- ¢ where & may be the component of a magnetic vector potential, the displace- ment of a vibrating solid, etc. Boundary conditions Att=0: ®(r,t=0) = u(r) and (r,t =0) = v(r) are known, Carrying out a Fourier analysis, @® 09 = yo free. @ art Substituting into (1): -=— fr, = or ste ude“"Pk = 0. Multiplying by e“°" and integrating over all space: wad fe ar more [*- 2 A : cfe-se JS hotesia-to = [2 +5-S)0 - 0.380 Wave Equations Since ko is arbitrary, the last equation becomes GF) _ de solutions of which are given by @ F@ = fe™+ ge", so that 4 &rp = ae Sli elke nek. q - KORO), Applying boundary conditions at ¢ = 1 jer P(r,0) = u(r) = are axle“ ‘dk. 6) : Hr) = v0) = Goa fite- a jerar ‘Taking inverse Fourier transforms of (5): 1 ae, Sat a= aor fe “er, © . 1 ike(fe- 2x) = amr Pp, Thus, given u(r) and v(r), fi andgy are known from (6), Fi() is known from (3), and ®(.,2) is known from (2) for all time. Green’s function (D-function) for the infinite medium @ fat = 0, 0(r,0) = u(r) = 0 and (6,0) = v(r) = 6°(r), then for alll time, (r,1) = D(r,) defines the D-function. Given the above boundary conditions, we have from (6): Sita 0. had - Foe Am 8 ~ SoeMethods in Mathematical Physics Il 381 Substituting for fi and gx into (3) and then for Fi() into (2): _1 ict _ nike) ie Hr) ards aawe eek, yielding the D-function o Di) = aes dk. (8n)'% Gi) Ifat £ = 0, 0(r,0) = u(r) = 5%(r) and (1,0) = v(r) = 0, then from (6): 1 1 fet Ge Ge - gx) = 0. 7 fe &~ Foe Then for all time, gilt g grit ter Org) = ae Tames’ tet. V jer eD(r,t) (8) oe re ‘Pk = oT Thus, for any arbitrary boundary conditions at = 0, i.e. (r) = u(r) and ®(r) = v(r), we have for any time t: O ry) = J[Per.nve + 2 ue) fee. Proof: We have already shown that D(r,t) and se) are particular solutions of the wave equation, under the boundary conditions specified above. Thus, @ Ab- de = 0. (ii) As 1 ~ 0, it follows from (9) and the definitions of D and aD/2r: (rt +0) = forr— rel = ue). &(e4 40) = v0).382 Wave Equations a aD{rs 0) = =) - 30), ae oe. ~ RE pein a + 0ast—+0, . Hence, ® satisfies the wave equation and the boundary conditions at t = 0. For the wave equation Ab - &/c = 0, if &(r,f0) and (r,f0) are known at t= Ig, then since the equation is of 2nd-order in time, @ should be known at all other times. In fact, for f > fo, (20) (ea) = fee vr rier) + PEEL) oeraa te, ve! where for the case of an infinite medium, 6 Since d°k = kUQxdk, where dQ is the solid angle surround- ing the k vector in k-space, if we choose the axes so that r lies along z (Fig. 11-1), then Fig. U-1 ferany = Jae fermsnaao 7 = Substituting into (7): _ angle etl [eet], Died = Sd Ge Dike ik 7 j gira) 4 gritrred __ gitt-et) reMethods in Mathematical Physics Il 383 This last equation can be expressed as IfD 1 an DED = Fe[Be-en-ar+ ed), which implies that the D-function vanishes everywhere Fig. 11-2 but at the points ¢ = r/c and t = -1/c, which in turn means Then that D/0t is zero everywhere but infinitesimally near aD/et t= 41/c. If Dis as shown in Fig. 11-2, then D/ét will be as shown in Fig. 11-3. The D-function can be pictured as a function that exists only on the surface of the cone r = +ct (Fig. 11-4). Summarizing D(r,t) characteristics Fig. U-3 - 1_fsinkct e** @ Den= 35. ee di) Dir) = ‘(r-ct) - S(r+ct)}. esl incr iii) D(r,0) = 0; 2s) -0. iy) PEM 00, iO) pies) 4,220) - 0, since D(r,) was constructed to be a particular solution of the wave equation. Fig. 1-4 (i) (ep) = Jee = = 1(e 4.) PEE) dead fee384 Wave Equations 11.2 Retarded and Advanced D-Functions Let p(t) be the source function, so that the inhomogeneous wave equation reads () A®-3 4 = ~4np(r,). e We define the retarded D-function, Dree(r,f) as: 4nc*D(r,t) for t> 0. Q) Dale) = 0 for t< 0. Theorem: 3) DDalr,t) = -4n5°(r) 5). Proof: (a) Fort<0: Drr=0, Dyer = 0 and 8(1)=0 = true fort<0. (b) Fort>0: QDrt= CD =0 and 8()=0 = true fort>0. (c) We now need to investigate the singularities of both sides of (3) at = 0. 130 Sadao) = Jeranete i) - J faz — = eADrsyine- 2S = [Dates| + -4nd%(r). Thus, as € + 0, the left-hand side of (3) approaches ~4n6°(r). Since from (3) the right-hand side satisfies §8'@ o4ndt = -473°(0), the theorem is proved.Methods in Mathematical Physics II 385 From equation (2) and equation (11) Section 11.1, for r > 0: : Dales) = ZE[o¢e - ct) (¢ + ed}, e ner Since for t > 0, 8(r + ct) = 0 always, we have @) Dar) = £8(r- ef). Theorem: If at t= to, ®(r,fo) and ¢(r,f0) ate known, and if p(1-) is given, then for t > fo: Mea) = f[ Dev 149) + PEPE wee she vw + fat [Dede s-t)pe fhe. * Proof: ( Since QD =0 and 2-2 cp - 010, Sect. 11.1 Qe - Jacferr [-4nd%(r - r') 3(¢- P)p(e')] = -4np(r.) ve for t > fo, where we have set 8°(r) = 8°(r - r’) and 8°(1) = 3(t- £). For t< fo: 8(¢- 1’) = 0, since 1’ runs from fo to ©. Thus, ®(r,1) satisfies the inhomogeneous wave equation for all t. Gi) Ast >t, (r,t) +? Sie DG br 1) 20 and Coe eee) or Jac foeter s-rypte' sae = Jae fDte-r’ s fo: since Déet = 0 for time < 0. Thus, as t + fo , B(c2) + Joe - P)OEs)dr’ - Hest). We386 Wave Equations Gil) Ast +> tm, O(n) +? ae OD(r ~ Ft +h) or 0, from (5): PD(r - r,t +t) =8(r-r’) and "> = (r-r’) lin Jaf Pate ee 0. © = Dasrt) = -4nc*D(r,1) for t< 0. Recalling from (ii) Sect. 11.1 that D(r,t) = Fe loer- ct) - (r+ ct)], and 0 ee Dues) = ehrte) fort<0. Theorem: (8) QDasr,t) = -408°(r) 3). Proof: (a) Fort <0: QDyav = -4nc*QD =0 and &(1)=0 — true fort <0. (b) Fort>0: QDsav=0 and 6()=0 => true fort >0. (©) We now need to investigate the singularities of both sides of (8) at t= 0. +20 Je 20.0609 ~Jearateo-4 Sf 9 Baked = ~eADumtt) ~ 2 pastes] + -4n8%(r). 25Methods in Mathematical Physics Il 387 Thus, as € -+ 0, the left-hand side of (8) approaches -4n5%(r). Since from (8) the right-hand side satisfies sep F-8'w BeAndt = -4n0°(H), the theorem is proved. Theorem: If at t= t, (fo) and (r,f0) are known, and if p(r,t) is also given, then for t < fo: B60) = f[D er soHbes9) PE axe'sa fe : a + fae [Dsle-v 1) peer. wy Proof: ae oD _@ @ Since QD=0 and QS >= 5 QD = 0 [(v), Sect. 11-1}: Qe = je [-4n8(¢- r)p(r')] = -4np(r.t) for f < fo, where again we have set 8°(r) = 6°(r- r’) and 51) = 8(- 7). For > fo: 8(t- 7) = 0, since f’ runs from -2° to to. Thus, (1) satisfies the inhomogeneous wave equation for all r. Gil) Ast >t, P(r.) +? Since D(r - ¥',t fo) = 0 and (9) ast + fy: OD(r = Ft ~t0) a 10) «33 — ¥'), we have from ' Jat [date s-t)pte tar = Jat [Dusle-e PO P(e P re’ = 0, since Daav = 0 for time > 0. Thus, as t > fo. , He) = [Oe - re )P'r’ = Br). t388 Wave Equations Gii) Ast +m, P(r) +7 SPle—v 5 td = 8%r-r) and forest) = 0. Thus, from (9) we have: Daler it) . sin Jac Se rr’ = 0. Hence, as t+ fo (6) > [Br - PIO so)dr’ = H(r,4). Ve 11.3 Field Due to a Moving Point Charge Let fo = - © and ®(r,f0) = ®(r,f0) = 0, i.e. let there be no initial disturbance. Then at any later time t, we have from equation (5), Sect. 11.2, %&) - Jo fonte-e-nowendr, which says that the disturbance at any later time ¢ is due only to the source function (1,2). We rewrite (1), using Dra = ¢ 8(r - ctYr. (rp) = fav far Laer -cbplr'y), oat where we have set &(r-ct) = 3{r-r'| - o(t-r)]. Define n = [r-r'|~ c(t-1), so that in the integration over time, dn = cdf’. Substituting and integrating: r Q ep) - fer fin 22 oem - fer Note thatthe form fe wow where f= 1- rT suggests the Poisson problem, ig e. although one investigates ® at a point r and at a time 1, the source is at point r’ atatimes = 1 E=F1 , which is the difference between time f and the time it takes the functional disturbance to travel a distance |r - r'| with velocity c (Fig. 11-5).Methods in Mathematical Physics It 389 Asc — ©, — t, resulting in Poisson's equation ‘A® =~ 4np, ic. instantaneous propagation. Returning to our problem, let p(r’,f) =e 8fe'-r(ry) refer to a moving source located at re(t) (specifying the trajectory of the source particle), such that for a fixed time f, focr'sydr’ Substituting into equation (2): ed[r'-rdr)] ror rer } SS ‘o Fig. 1-5 B) &(r) - fer baer If we define the vector 7 = r’ ~ ré(1’), equation (3) becomes ova F e400) = fen ET” [ier ge” i a0 where we have used the expression for the transformation of volume elements from r’-space to 7/-space, e.g. dv = (222 a dnodns, r "Uni ne ns ini 22 ans where Jj is the Jacobian of the transformation, ie. x oy’ 2! oq: On Oni Ox Oy’ Oz" On2 On2 On2|’ Oe By az ans Ons Ons By definition, =r’ - re(’) = dij = dr’ - ¢e(¢’)dt’. In addition, oe eeneen ¢ c390 Wave Equations ae = A Meee ede) dee - 2 c lr - ry” Thus, ye Bde = r'yde! Oe de eee Since we want J evaluated at a particular time t and location r, we choose the x-axis parallel to te = Ue. In component form, dqy = de ~ 2AE= XIE _ vy yet _ vale = zee " r-rel etre dno = dy’. dns = dz’. setting up as simultaneous equations: du = ae [1 Nay [- cl(r - r')] aren) Gee dn2= 0 + dy + 0. du= 0+ 0+ a. Hence the det of the transformation Jy (3) is given by +(3) op Pene) Bele) el(r- r')] el(r - r’)* “bee F) el(r - FD]Methods in Mathematical Physics II 391 Thus, from (4): r= Pen wr) r= a cr-r| rer] Since H = 0 implies r‘ = re(t’), we have e 6) Pr) = Ie-r0)) r-re(t’)” Physical interpretation of (5) A source at r¢(1’) moves along some path C, and at any instant has a velocity Ge. Ata point r and at time ¢ the disturbance is given by Ir- rel? where an additional term appears in the denominator. In order for the dis- turbance to be felt at time f, it must have been produced at re at some previous (retarded) time ¢ = 1- ¥ Fl where !—¥4l isthe time it takes to go from re to r. [ Note: If the wave equation is given by -14.-44 Ce and if Ser) = dO - rr)], then at point r and at time t, we have evr’) on™ ee (©) A(r,t) = at t= retarded time which are the so-called Liénard-Wiechart potentials.392 Wave Equations 11.4 Finite Boundary Medium Program: Try to find the Green’s function for the homogeneous case, V7 - &/c? = 0, and then deduce the Green's function for the inhomogeneous case. Then, knowing and & at time t= fo, & will be known at all other times. One does not perform a Fourier analysis, since the boundary in this case is not at infinity, and might be quite complicated. (A) Homogeneous Case Consider HY, = V3 Y= Ann, where Y, = 0 on surface S, and imagine that this eigenvalue problem has been solved, ie. that we know the entire set of eigensolutions and their respective corresponding eigenvalues. We will obtain the Green’s function from this eigenvalue problem. Saattede = Sroar. 4 fen-va adr _ ~fv-ce,V¥,)de + f(VY,)ae Sa Jura ~ Soe.ve,ynds + fv) Sear Siar” The last step follows since by hypothesis ¥, = 0 on S. Hence, Sevenrar = 20 Sea (The Ap correspond to wi , the frequencies of vibration, and must be > 0.) Define the Green's function G(r,r',): Gee = Leora) enactMethods in Mathematical Physics II 393 Note that if one sets Ws(r)¥a(0") =e"), sets Vn = k and lets the E -+ f, then G(r,r’,)) for finite media becomes similar to D(x,1’,) for infinite media [equation (7), Section 11.1]. Properties of G(r,r’t) @ Veer - 4 Fad « 0, Proof: 1) Sind Anct VG = LAr) EE sinVAnct | 16 2. Aas Pale) OE aon Gy) VOGEL & Gert) _ 9 or cat ao . Proof: SB ~ Swern ycos! Fact. VS ~ Y-nele Me Joos! Rnct 12 a rs - Loma ¥sCe eos! Tact. Gi) G(r’) = 0 and 2GEED - 9 for ron S ot for r on S, since by hypothesis, Y,(r) = 0 for ron S, and ¥,(r’) = 0 for r’ on S. (iv) G(ry’,0) = O and Patera = 0 by virtue of term sin/Iyct.394 Wave Equations (vy Soe.) nnd =8%(r-r’). Proof: sex) = DAO ¥Ae). Since the Y, form a complete (as well as orthonormal) set, any function that satisfies the boundary conditions can be expanded in a set of Yn. Thus, 8°(r- 4’) = Yan'¥n(r), from which Santee - ride = [Yane(e Pad = Yaiban = an 0G(r,r’,0) = me Beer’ Hence, am = Yn(r') er -r me (vi) If ®(r,to) and (1,40) are given, then at any later time f: @ Bes) = Stern bte2) + SEEM gee sy). Proof: : (a) Since V?G(r,"' t-te) - ae = 0 from (i) above, and since ¢ 2 OG(ryr'tmto) LF G(r’ tmto) _ 7 ee Sa ap 0 from (i) above, we have 2, 1 PHD _ VME!) GE O- ) Ast to, ®(r,) + P(ryfo) and &(r,2) > (r,fo) from (iv) and (v) above. (©) ForronS, (1,2) = 0 from (iii) above. Thus, (1,1) satisfies the wave equation and the boundary conditions. (B) Inhomogeneous Case For the inhomogeneous case, where a source p(r,f) is present, we define: (4nc*G(r,r’,t) for 1>0. Gules) = 0 for 1<0.Methods in Mathematical Physics It 395 (0 for t>0. Gade’ t) = -4ncG(rr'f) for 1< 0. Properties of Gaav(r,r’t) and _Grei(r,r’,t) @) [v5 Sore = ~dnd'(r- r'V60. ‘The method of proof follows what was done previously for Dreyadv (Section 11.2): we show it to be true for f < 0 and for t > 0, and then show that the singularities on both sides of (3) are the same as t —+ 0. It follows from (3) above, that when a source (r,t) is present, [v a a By’) = ~4npte), and for f > to and t < fo, O Oe) = (+ OE), where (5) &(r,t) = S[ow ioe + 20lsee) acrsa ft . Fort > fo, © een - far Jor Gatea’a-ro0ce.), ae MQ e&rp- far fee Gates-note. (These results are to be compared with equations (5) and (9), Section 11.2.)396 Wave Equations 11.5 Green’s Function Method Applied to Schrédinger’s Equation and to Heat Conduction Consider the eigenvalue equation HY=E a For the case of Schrodingr’s equation, iT =- a + V and g--4 (eigenvalue E is imaginary). For the case of heat conduction, H'= A and E =~ . (eigenvalue E is real). In both cases, operator H is real and Hermitian. Problem: Given ¥ = (r,0) at ¢= 0 in both cases. Find (r,t) at all other times. One supposes the particular eigenvalue problem HV, = AnPn to be solved, where the 'y satisfy certain boundary conditions. The Yr then form a com- plete set for the particular operator H for any ‘¥(r,) satisfying the boundary conditions, ic. W ¥ED = LAOKM, so that : HYD = YHOHV) = DOPAC). Ay 7 HY- BY = Lv [itor - %] = and since the Y, form an orthonormal set, oo HO 10) = Ho ~ accro("E, and @ Wes) = Lavero( 2) Ya. We seek the Green's function for the problem.Methods in Mathematical Physics Il 397 Boundary conditions require that G(r,r’,f) = 5-function at t= 0, and at a later time t, G(r,r'’,f) = P(r). Att=0,¥(r,0) = Yaar) = 8°(r - r'). Since ay = Pa(r’) [Sect. 11.4], it follows that = GB Ger’ = LY exp (| is the appropriate Green's function. Properties of G(r’) @) HG=kG. Proof: Hex A) = LAM ee exp (#: BGe's) = Dade Waceexp "E (i) Ast +0, Gar) > 8(r-r). (ii) For all time t: 4 Hen) = Sones HE (Note that this is of the required form for the homogeneous equation of heat conduction in the infinite medium, as discussed in Section 9.2, Green’s func- tion property 4.) Real versus imaginary eigenvalues For the Schrédinger equation, exp (Iu/é) —> exp (~iAnf/1) represents an oscillation in time, and since represents matter density which must be conserved, fdr = finite. For the case of heat conduction, exp Au/E) —» exp(—Adnt ). Since Man is positive, this represents an exponential decay in time for t > 0 and an exponential increase in time for t < 0, which agrees with physical inter- pretations.Appendix B Problems and Solutions Problem Set A I. IFf(2) is analytic and f(@) = @ + iy, show that both p and y obey Laplace's equation, i.e. show that Ap = 0 and Ay = 0. Solution: Since f(z) is analytic and is expressible as f(z) = @ + iy, then @ and must obey the Cauchy-Riemann equations ap cy op sey oy Oy or Evaluating Ag and Ay: Soe of ala) al a)” ae eee : - 2,2 2 a2),,.2 (22) - #2, fe. ax Oxdy dydx II. Given a conducting circular cylinder of radius ro, whose potential p on the surface has the value A. (@) Find f@), a solution to the problem (trial and error). () Show that f(2) satisfies the boundary conditions. (©) Sketch lines of E and ¢ = constant. Solution: Observe that we want @ to be a constant on the surface and lines of constant ¢p to be concentric with the cylinder. Hence @ = f(p) only. Also, since lines of constant y should be perpendicular to lines of constant p, we see that y = (8) only. This suggests the form f(z) = Blnz.Mathematical Methods in Physics I 399 (a) Let fie) = B nz = B Inp + iB. Then, p = B Inp and y = BO. Atp=Tro, 9=A,s0 that B= A/Inro. Thus, = A Inp. lise A A ~ fl) = a ine. (b) In cylindrical coordinates, a ral 2 oa Poe el oh =0 = Vp =0,@ =A onthe surface. Thus, boundary conditions ate satisfied. (©) Lines of E are parallel to lines of w=constant or @ = constant. E - - vp - -4-vinp Aa ins p and the mapping is as shown in Fig. B.1. IIL. Show that if f@) is analytic on and within C (Fig, B.2), and a is a point within Cy, then 1 play = Lf, a. 2niz (z- ay Fig. B.2400 Problems and Solutions Solution: Define a function g(z) = 0-5 pa), which is defined on the circle C1, of radius 5, and is analytic on and within C;. Then, as + 0, az) +0 sine im =H) fa} 0. ‘We have shown that if 2) is analytic on and within C, then 5 ay “fe where C1 is the circle of radius 6 surrounding point a. Thus, feo ay ae Ee ogee a fe fe or Since (z~ a) is finite over C1, while g(z) is analytic on and within C1, we have, by Cauchy's integral theorem, §25 Since (z - a) = de®, a len feng Loe” Hence, fede _ @- a. 1 (Z@-ay Thus, fay = Lf foe. a (-ay TV. Show that if fi) is analytic and if f(z) = ¢ + iy, then the Cauchy- Riemann equations in polar form are satisfied, i.e. op low 4g Lae __ oy ap p 00 poe apMathematical Methods in Physics I 401 Solution: Since f(z) is analytic and f(z) = @ + iy, @ and y obey the Cauchy- Riemann equations: oe OY ag ov ex ay ay ex" If we regard f(z) = f(,8), then p = p(p,6) and w = y(p,6). The Cauchy- Riemann equations then become: 2p ap , dp 30 _ dy dp , dy 28 @p ax" 00 dx dp dy OB dy and Fe eS See ap dy 80 dy dp ax 0 ax | But from p = (x? + y*) and @ = tan”!>, we have oo ae oye ap cos® and ay sind 2 __y _ sind 14 08 _ x _ cosd Ck oe ° ay Pp Pp Thus we have 2p 4 Op sind _ dy ay soso cos 1 cos sind + ‘ © x 20 pap P 2 cing + 22 2088 _ _ OY 6g, Sy sind @ 50+ BB 3p 9+ Sop” Multiply (1) by cos@ and (2) by sin® and add to get: (Op ~ Sy (costo int) Op ap 0s" + sin'®) = BES a oy. Pp 06" ae p 06402 Problems and Solutions Multiply (1) by ~ sin and (2) by cos@ and add to get: op (= + auto) 20 P oe lop oe ap (e05"8 + sin’8) aol op V. Iff() is analytic when [z - al < R, then for 0< r< R, show that . ifs Fla) ~ J veas, where f(z) = p + iy. Solution: Expanding fie) in a Taylor series about point a, LO nt S@) = fla)+ faye - a) +S @- ay + (-ayte Writing z - a = re, we have Froe'a - firme +f(ayre™ fo Pel? + -}o sof ree i, . : —] 7 fe feta + Play Jea0 a jean a +e LO pf awinggs = 0. nes 2 2% +m Hence J f(e)ed0 = 0. It follows that [froee = ff@e®ao = 0, 7 7 3 2 % 2 ot, fip-ivye®d0-0 = fge"®ao = i fye"®a0.Mathematical Methods in Physics I 403 We have shown in Problem 3 above that f(a) = 51 rhe 7 7 dE. Since (E- a) =e and dé = ire®d0, we have le + iy)ire - = p= fesigt «3 foetal Since, as we have just shown, these two integrals on the right side are equal, ro) = foe404 Problems and Solutions Problem Set B I. How many singularities do the following functions have? Discuss the nature of these singularities. (@) f@) =e (b) f@ = cosz (©) f@) = (@- a)(z- By @ fo-e* (&) f@ = Inz - a) Solution: (fe) = b= 1tree eben. -¥E. 213! An Setting w= 1/2, td yl aw) = f@) = ee Lae and we note that ¢(w) has an essential singularity at w = 0; hence, e” has an essential singularity at z = infinity. (b) f@ = cosz. Setting w = 1/z, FD w aw? atwt : which has an essential singularity at w= 0. Hence, cosz has an essential singularity at z= infinity. (© fi2)=(2- a) - Noting that there are branch points at a and b, we run a cut from a to 7° and from b to © (Fig. B.3).Mathematical Methods in Physics II 405 We see that we have a singularity at w = 0, hence f(z) has a singularity at z = infinity. Note also that we have a branch point at z = infinity. To determine the required number of Riemann sheets, note that (z - a) = pe, and not until @ = 6m is (z- a)” at (0 = 6n) = (z- a)” at=0. The best procedure is to choose a particular value: e.g. f(z) = (z- a)”. Setting (z - a) = -1, = e™ = & etc, CD = cy” = 2 SM... Note that the series starts to repeat at the 4th term — where e'” occurs again — i.e. the series is triple valued, and one needs 3 Riemann sheets. Setting (¢ - ) = -1, Cp ee ee (<1)? = PF = hd a glOivs — ltind _ pitins _ rains... Note that the series starts to repeat at the 6th term — where e”"¥* occurs again — i.e. the series is quintuple valued, and one needs 5 Riemann sheets. (@ fe)=e*. Setting z= Yw, . i gov) = eo = Loo which has an essential singularity at w = 0; hence e has an essential singularity at z = infinity.406 Problems and Solutions (©) fle) = In(z - a). Setting z= Yw, gw) = tnf55-4] - ni - In(1 - aw) - Inw. (w) has a branch point at w = 0; hence In(z - a) has a branch point at z= infinity. Since there is also a branch point at z = a, we introduce a cut extending from a to infinity. IL. A function which is analytic everywhere, except maybe at infinity, is defined to be an entire function (or an integral function). Prove that if an entire function is not a polynomial, then it must have an essential singularity at infinity. Solution: If a function F(2) is an integral function, then it can be expanded in a Taylor series about the point z= 0, ie. F(z) = Yianz". imo If F(2) is not a polynomial, then it must have a singularity at z= infinity, in fact an essential singularity, since the series goes to infinity with increasing powers of z. TIL. Find the poles and residues of the functions (a) (b) x= (c) zcotz. a sinhz* Solution: (a) The zeros of sinhz lie at values of z for which e* - e~ = erie = e7!*" ie. zeros at z= inn. Hence we are interested in points along the imaginary axis, namely 0, in, 2in, 3in, etc. 1 1 Setti = . For2=0:Seting 2=iw, > Sinh(iw) ~ Tsinw ‘Asw—0,sinw - w, so that lim aac Ald z=0is1. woo iw Zz Hence the residue atMathematical Methods in Physics II 407 For z= inn: Setting z= i(nn + w), sinhz = sinhi(an + w) = isin(nm + w) = i [sinnm cosw + cosnn sinw] = isinw(-1)" Itfollows that lim —— = —1 - ——1 prone the residue osinhz (-1)"iw — (-1)" [z- inn] pee atz= inn oy At Geo BP . +1 for even x. ~1 for odd x. = 3ix n=3 R=-1 mia : (b) Since cosz = <= £—, we see that zeros occur when e+ 1 = 0, of 2 when (e” - e+ i)=0 = e% = 4i, or c= el)*™), Hence we have zetos when z = (5/2) +m. Letting z= nn + (1/2) +, cosz = cos[nn + (1/2) + w] = cos[nn + (n/2)]eosw ~ sin[nn + (7/2)]sinw = (-1)""'sinw. ro wee Cost Claw. Claie ont wa Hence, the residue at = mn + 2 is. At z= n=0 R=-1 z=ntn2 n=1 R=+1 etc, z=2n+m2 n=2 R=-1 (©) zcotz = (2 coszsinz and zeros of sinz occur at z= nx (except n = 0). Setting z= nn + w (n #0), _ (n+ wycos(nn + w) _ _ (nn + w)cosw sin(nn + w) sinw Zcotz nk R= Residue at z= nmis nn. = nm) . nt lim zcotz = — woe w408 Problems and Solutions IV. Use contour integration to prove that dx _ x(2n)! i ees any co [EH sfiait where m and n are positive integers and n > m. Solution: de dx — mand both integers. Hence, +2 [oe jes $F ~ Pri Residues ®), where C is the complete upper half contour shown in Fig. B.S. AtZ" =-1, roots are 2" = e'™ = z= 6", where k= 1,3,5°++(2n- 1), yielding n roots in the upper half-plane. (For k= ~1, -3,-5 +++ -(2n- 1), we have n roots in the lower half-plane, but these are neglected as we are only concerned with the upper half-plane.)Mathematical Methods in Physics I ant DR = Rt Rat +R, eS ra (eZ, oe are, ree,” y= ern 2m = eX OD, oD R« erianas gta grains Genta * Ginn gin Dn 1yan =e Om-IO Yang gin ane Van og gf Qm-2nt1)M-1Vn 2D R = Yerorvee BAY gieCQn-be an Y 62nd Van, But since a . in-2n(2m-2ne1Y2n in2n-Qm-2n1) F (cmon f Loe ae Senet) 4 nm niyn ~ 4 gintinrivan we have in 2nm-2nt an, R = ~eenrnvr, Loe te n, e 1 - rem ivan 1 = gin @n-2ne1) ~ GeGm Van _ iene tyan * e412 Problems and Solutions Since 2n ~ 2m -1 is an odd number, RON IMD 2 gtiO-2MD 1 so that any R = Finally then, Ride ante st" ane V. Prove, wsing fo ie= = , that t Sta ve (a) Seose'de Srintax i Pg, Ed 0) feostane “ae = Solution: 2 A (@) Consider fe” dz over the closed path OBA shown in Fig. B.6. Since e is analytic within this closed Aan ion, f ede = 0 the fact that a ° Tegion, 9 é =0. [We use fact that along AO, Fig. B.6 z= pe” and dz = dpe™). ao a @ — Sefae = fear a) fre- jew" ePadp F io2. To? Je teMdy = el fer ‘dp. e=n4 4Mathematical Methods in Physics I 413 4 Gi) fra: z= pe®; dz = ipe"d; = p’e = p’cos20 + ip’sin28. Ia However, in the interval between 0 = 0 and 0 = 1/4, sin20 > 0, so that |e#ae| 0, then T@ = ¢ferea, where s is any complex number whose real part > 0. I(2) is defined as T@ = Jovan. Solution: Define a new variable v = st, necessarily complex by virtue of s being complex. Then dv = sdt and du/st = dy/t = dv/o.. It follows then that fo = feat . fe “ ae Jervray radius R Consider the integral f e"°v® 'dv , where the contour is taken in the complex v plane (Fig. B.8). Since the integrand is analytic within the region shown, fe"°v® "dv = 0, Hence, ELL =0. nD ® & & (a) Consider fore tee: (2) [fere"'ae ) For -1/2 < 05 < 7/2, cos0 > 0, andas R = ~, eRe RE , 0, radius r Fig. B.8 < [fessor ete Manet ) < [fer*rao| ; ) (b) Consider fer tae: (4) fers" ‘ao (4) . [fersowanesee mit 9 3 [ferro] : a)416 Problems and Solutions Asr +0, forRez>0, € 7? +0. (© Consider fev 'dv = fer at = TE). QM ‘oO ¢ : (@ Consider fev 'av = foster tur = - Jste*1 ar, since along @), - ‘0 0 = 0p, v = woe = tsce (i.e. real part of s > 0). Hence, foe + feroriao -0, » 3) T@ = “fee = fea. ) II. Consider the equation AY? + v(p)'¥ = 0, where v is a function of the cylindrical radial variable, p, only. Show that the general solution of this equation can be written as. V(082) = Y foarye*ak, where ®ni(p) satisfies the ordinary differential equation [ae tpdp 00k efh = Solution: In cylindrical coordinates, we have x = peos®, y= psin®, z= z. Ao > rary n hea [2] +(2) (2) = (cos’e + sine) = 1. of 8) RIT ones weve (2) Qf] =.Mathematical Methods in Physics I! 47 a= slesle Sp)" aol }- 198 ow op WW = 'P(pz) and can be written as ‘¥;(pz, 8), a function of 8, with p and z regarded as parameters. We know that any function of @ can be ex; in a Fourier series: (p62) = Yn (pz)e™. Employing the Fourier transform, regarding p as parameter: (02) = 73= Some (ode. Setting pnt (p) = ¥2m Gms (p), we have 1¥(p0z) = L Jom (p)e""™dk. A\¥(p62) - Seon iF 7 (e Pent . 3 ¥ Jrton (eye dk - ¥ ftom (pye""*dk, so that a+ pny = Yo for [2 2 (p 22a) pap’ ap awa 42 ae ~ 5D Jom reat - 2Y, fou oe ak +3 fow (p)ve"™"*dk = 0.418 Problems and Solutions (p82) will be a solution to AYY(p8z) + v(p)'P(p6z) = O when -R+ vt) fne( = 0, a Pp dp where the partial derivatives have been replaced by total derivatives, i.e. Pni(p) satisfies @iid mp, . [S Pt dp pF R vpn) 0. Il. The Schrédinger equation for an electron in a hydrogen-like atom and in an external electric field in the z-direction may be written as AYV(r) + w'P(r) = 0 (Stark effect), where u = a+ b/r + cz, with a, b and cas constants, and where ris the spherical radial variable. [Due to the term cz, wis not a function of r alone.] (a) Show that, by using parabolic coordinates, (E1 E2 @), defined as x= Exbrcosp, y = Ebasing, z = (Ei BY2, the differential equation may be written as 2 1 Of, 2 & = Ypalea) gare ee culge]eo (b) Expand ¥(r) into the Fourier series, Y(r) = J, bmn (EiE2)e"®. Give the explicit form of the differential equation satisfied by ®n. (©) Show that if the technique of separation of variables is used, i.e. On = fib) AE), then fi and f; satisfy the two differential equations: rales) sfoet +o -co( sahi-o, (= 1,2) where d is a new constant.Mathematical Methods in Physics I 419 Solution: (@) de = Bcospd + Ercospds ~Bitssingde. dy = Espa + Bing, + EEscosedp. de = Be ~ ba. ae = ab ds? + de = (coop + Esl ~ Esinpdp) + (sinpLadés + Eel + Eiicosydy) + (Eas ~ Eada) = (Babs + Eats) + Beg + (Bias — Eat) Rewriting, as = ail ei + eh] + ats [Eh +] + dg? [2t e]. ‘Thus, the new parabolic coordinates are seen to be an orthogonal set, for which hy VETEE ha VEE: in VETER = Bs. (One can now use the expression for AW in generalized coordinates, given by or rin) BE) The (ee lis & ih } 8 cecere) ee lee on 88] “meeralalaes er] 1) “para See Reel)420 Problems and Solutions Rearranging, erale ae.) egos] wale lar ra eaeee (ede Let -aeet ea)" - Le). Thus, u = a+26(% +8). Se - €8). Substituting into AY + uw = 0: aide) 2868 “ataylee gi 0-9] a ale Bega By = Revrriting, ‘uss Rl etait cyt ely - 0. (b) Expand in a Fourier series, Y= Leeattitoe®, and substitute into AY + WP = Oto get: Lele ae Se) Bae [ae] OLE IQ orien : Ler +5 (et 2b}oa] =0.Mathematical Methods in Physics I 421 Since this must hold for all ©, and since the e’"? form an orthogonal set, we have pal Se} ean oe lee “(Ge Bpo. m0, § (6a. =o ale ws a+ b- cyt $8] oct - 0. ft (© We try for a solution of the form ®n(E1E2) = fi(E1)-fo(E2) and substitute into the last equation of (b) above. After dividing by fi(Es):filEo): 1 Hh) t1 ao peat S) peel) e's) +a(Ei+e)+ 2045 (e-8) = 0. Rewriting, noting that the above partial derivatives are now total derivatives: pedled ptesd “pekeg) gee td-« This equation can be written as G(E1) + H(E2) = 0, which can hold only if G(E1) = -H(E2) = -d, a constant. Thus, beater g-entse de422 Problems and Solutions IV. Any inhomogeneous second-order linear differential equation can be written as Ly Ls aay = Fle), where p,q and F are known functions of z. (a) With the transformations fla) = aoen- area Fe) ~ Geex|- 4 show that the differential equation can be written as 48. Og - ae (2g = G2), and find the expression for J(z). (b) Define g(z) = Y.gn(z), where Boz) = Je - Z)G(z’)dz’ EA BA2) = ~ fle- 2 UU2gei(2e for n> 1. % Prove that g(z) defined above is a formal solution of d°g/dz” + (z)g = G(z)- (©) Prove that at a point z where J(z) and G(2) are analytic, by suitable definition of the line integrals appearing in go(z) and g,(z) expressions above, the sum. E gn(z) must exist. (In fact, it converges uniformly.) Solution: z (a) Setting s =~ 7 [plz’vdz’, we have fz) = g(z)e? and F(z) = Ge’. %Mathematical Methods in Physics II 423 After rearranging again and substituting into fing 7 + qf = FQ), we have ae 2, o {fe -28 £8 « ag] Geet = B+ sag - 6H, 2 where Je) = - 9-2 B+ (2). (&) Define g(2) = Yygnz), where = gol2) = S(z- 2)G(e'\ae! 2 ga(2) = ~ fle ze ees(e'Mde’ forn2 1. 7 “ome that one can differentiate inside the sum: =. - eS deo _ f & Scere ~ = = G®). : ~fRe-eerenserne = - faerenstevee’. WO % Thus, fin I Bn-(@) = y fe = -J(2) Ye. 1(Z) = ~J(z)g(2) ae a mi a? ee by hypothesis.424 Problems and Solutions Thus, ec we G2) - Ie, and g(z) = gn(z) is a formal solution to the equation 70 20). nage = 6@, provided, of course that \g,(z) converges uniformly. nso (©) Let z be the point for which J(z) and G(z) are analytic, i.e. let z be an ordinary point. Then there exists a region about zo wherein J(z) and G(2) are analytic. Over the path from z to Zo, as shown in Fig. B.9, J(z) and G(2) are analytic. It follows then that fie- 2G’ = gz) * and Bx(Z) = -Je ~ ZW )Bn-(2')dz" . Fig. B9 are also analytic, [Note that if go is analytic, then g1 is analytic, from which it follows that g> is analytic, etc.] We now prove the following theorem: Theorem: lent2$ we le~ zo. Proof: Let [(2)| < U over the line integral from z to ze and let |go(2)| < V. For n = 0, the theorem is true by hypothesis. We prove by the method of induction, i. assume true for n ~ 1, then show it to be true for n. yur! @-p! bona) < ki-2pr?. lena) < fiz Zhen eee. 7Mathematical Methods in Physics I 425 Since gp(z) is analytic in the region shown, the path from z to z is immaterial; hence take the straight path zo to z. For this path, [z - z'| < ~ Zo), 80 that ‘ len(2)l < eer pre za fie - 2°. ie Letting s = |z’ ~ zo], ds = |dz'|, so that po fr tag = VU =z aa (n= D1@n- 1)" oo Forn>1, 2n-12n, so that |gn(2)| < ete nl Setting Uz ~ zal” = E, g(z) = Yeu forms a series of terms, each one of which is less than the corresponding term of the power series for e°, i.e. ee D e which converges for all E. Hence, g(z) = Leuo converges and therefore exists.426 Problems and Solutions Problem Set D L. Let Hy(z) be the Hermite polynomials. Prove that (0) fPlatine™= ade = (0+ AY2NE Bayes + (+ DILVE Bayes. () fite*oae - AS Jane : o §fz (alfa » Powe. Solution: (a) The generating function for the Hermite polynomials is given by eft y fe ore yy Hales” Helos Sse) e ee ~ ke Patel p rch - os = arf — 5-1) + (s+ Dlde " (s+ den = rive y Ot it z mer, wey Pee ansne i my ree . wy eee n! Es myo ‘0Mathematical Methods in Physics II 427 Equating coefficients of fs": | itattue™ ue = RE Bagn VE bum, nim! =, so that SrtayHalce® ade = FE2"(N+ 1YBmgnn +1R2" MN + Bayer (b) As above, efite =>) Hee ents = 5 eo Hale LE Sincere’ Pee = [tne = fe gas = eafeerst [le-s-0? + acre ~ (oto fe. But fe oN 5 de = 4 | cern. -o| +t foro ay -= i aod . fee" astnde = (st) [-rern seme] = (stv. Summing, we have oF, Sincere Pee = [E+ (osniva] ES Looe 0 ve428 Problems and Solutions In the first term on the left, s and t occur with equal powers, namely for n= m. Comparing coefficients of (st)" on both sides of the equation: Ge ey one ar eae write Ze - ies Ty Thus, fitoe* 2a = CoD ove, eter Tthom or - fe ( - on) - fe [2k - 2H dz - ik [20H ~ Hyer ~ 2H] de = fe [Hn He fide = fate 2a afetattnne® d+ fibre ae . va Pata = 2(n+ 112" +2" (n+ vt} = Vm2"n! a 1 Il. The matrix elements of any operator O for a harmonic oscillator in quantum mechanics are defined to be = f¥,()OV-(x)de, where v2 2, W(x) = lara H(axe?, — where a and where k is the force constant of the harmonic oscillator.Mathematical Methods in Physics It 429 (a) Show that = Y, i.e. matrix behavior. () If(Ax)n and (Ap) are defined as v2 v2 (ax). = [np — ?] , prove that (Ax) (Ape = 2249 28. Solution: @) = Svsaecoredr =f a Hone dx - fae (ax), . J oan ae aot) 1 ‘2n+1 2n+1 = ant) a) ma ovn2'n! ( Pete er = P,P a(adede f fiosed (ox), Sara Te aing Monisbande” hor). 2 (0+ D)I2"B mea + (int + 1)!2" bye ] ol nv 22" nlm! J - yx = Six [ee we eal + [C01 + 1)!2"%w 2| = on2"2"n'm! Gaeta aloe @2"2""Inin + 1)! o2"2" "n(n - 1)! Hence, = Y. m0430 Problems and Solutions (b) From (a) above, = 5 + We also have: a . = i = Seiad: =J a HXox)e diax) = 0. Hence, ” (Ad) = eel 7 [ = la 2 hein 4 ingeane ie Multiplying numerator and denominator by a”: -jae [x Hox? ret (taoxe” ‘eai) Jee. Setting @ = ax, ade on] . 4 [-wa dy do? [aso] - do [e (-om +] - flee one-two] ~Ga|. diy ys 2 =¢ “Alo os ie 4] - 06° oy ~ Hon eon [Ha(t = 0°) + Flyer + @(20Hy ~ Hoss) ~ 200+ pi | = ee? [Halo? -2n-1) + Helo - «)] = &° PH -2n-1).Mathematical Methods in Physics I 431 Hence, = J er [Hscaxye gat? -2n -)}Hax - a oF rican (ax)hatax) a 2 + (nee J ricane aor) _ at? eps +1 2n+1 "nln — nl = Fel 2 2M n+ neta] 5 ee. = | —— nyane* tania d a Haorrye™ (ana om oh (ayn _d_ (ayn, diva H,(ax)e ax y Halaxye ‘d(ax) ; - ae Jit [centi(ca - Hala ox) - AS et _frttaane(axyatox) = a Gingoxst tere on). From the results of Problem I (a), the first term vanishes for m =, while the second term vanishes for n# n+ 1. Hence, = 0 and ee Pyon. Finally, then: 2 v2 _ (2m) 1 famed _2ntl, oh (aren = [ , } al - |= hy ok.432 Problems and Solutions IIL. The Laguerre’s equation is given by Gf l=zd ne. a2 df" % (a) What are the regular and irregular singular points? (b) Prove that one of the two independent solutions is an entire function and that the other one has a branch point at z = 0. (©) Show that if n is a positive integer, then one of the solutions can be written as N £9 = 40 = Sor 1 Se” (mt) (=m)! which is defined to be the Laguerte polynomial, with n = 0, 1, 2,..- Solution: (refer to Section 7.2) (a) Since pte) -1- Land q(2)= 7, wehave p-i=1, q-i=m, and q-2=0. Hence, pz = 1 - z and gz” = nz are analytic at z= 0, so that z=Oisa regular singular point of the Laguerre equation. To investigate z = ©, we consider 22 - pz’ = 22-2+2=2? +z and q(2)z* = nz’, both of which ate not analytic at z= °°. Hence, z= is a singular point, and therefore an irregular point, since both pz and qz” are not analytic at z = ©, (b) The indicial equation has the form Ss-l)+spitgg=0 = S-sts=0 = S= where for this problem q-2= 0 and p- solution, e.g. s= 0. In the vicinity of given by fiz) = Lane”. 1, There thus results only one , one independent solution is Since fi(2) is regular at all points except z = , fi(z) is an entire function with an infinite radius of convergence.Mathematical Methods in Physics I 433 Furthermore, for p having a pole of order 1 and qa pole of order 1, (m < 1, n $2), we have at worst a function with a branch point at z = 0; this function constitutes the second independent solution. © — fi@)= Lanz". Cie FiO . = Ent — Daae”?, 1 1290 . = Ema magi"). F fie) = Yondna™* Substituting into the Laguerre equation: Le" [(12)(m+ Dana + (m+2)aqne (M+ 1aqrs + Nayes] = 0. V (m*2)'aq2 = Gm(mt1-n) = anon = owe V mn Gast * Gan, Gn. If nis a positive integer the series will stop when m = n. SiR) = ot YZ + O22" +... + Og AZ + Oy" + AgerZ + ax[t—me +m ye _ n(n=1Xn-2)2 , | an : A a. : Cmte orrene) n434 Problems and Solutions For m = n, the last term will be that for which the power of z is m, namely oe au|t m+ MCD nn= Deena, oe 2| Choosing ao = (-1)"n!: c wrt C1 n(n- Dnt? ar A@ = Cin + SCD, Caran’ “Glen: © ON7n-0}! ie Dt Cicea array n!(n-n)! This can be expressed as file) = Ec w(2] = (am)! " If one now chooses a = n! rather than (-1)"n!, one obtains, after dividing through by (-1)": fe = Lene (“ Gm 7 A which is the Laguerre polynomial. IV. Show that the Laguetre polynomials satisfy the following equations: @) 4) > Zf[ea 0] ex(a je, where the closed contour is taken to include the origin in the t plane. ) a-9" ‘exp (7% z= con LO (6) (1 #20 2)Lan= Lot ~ Laer = 0Mathematical Methods in Physics I 435 Solution: (@) Define 42) = Sf Gay sexo( z= ja, so that Oye) = let Pear dz 2ni3 (AF exe (7 fa dy(2) _ 1 lad x de 2ni a-) ” p(s fe. yA2) sede i ae +(1-2) + nyn(Z) a aa: ~ ex (x a cat ~ad te lis te ~alacomelis | - (75) 2d, (0-9 22+ my¢a) =~ Since ae (ee |e dt [iS exo(7% ke 2 iLe., yn(2) satisfies the Laguerre equation. From residue theory, rth. a while £,(0) = n!, so that £,(0) = n!y,(0). Hence, £,(z) = n!y,(z),and emt ae LAO) * ap (-) aa lis Ji a= 1, (b) Since from residue theory, om 1 fer t Lover ‘dt = y,(z), it follows that436 Problems and Solutions per 4 (n-1)! * “t ler el : yea t-2) = ana SOS ce 1)! * Equating coefficients of r": (=2L@) _ Luiz) _ Laer) _ 2Luf2) , Loi) n! (I)! nt (n-1)! (n-2)!” and multiplying through by n!: (1 ~ 2+ 2n)L(Z) ~ 1 Ly-1(2) ~ Lai() = 0. V. The associated Laguerre polynomials are defined as Lxl2) = fo) Show that @) eg pred, "Hate =0. (&) Iff@) = 2" ™eL4(2), show that fe), 2 df), aa + aa[ Ett + (tn-2k+ 232-2] = 0.Mathematical Methods in Physics I 437 Solution: (a) The Laguerre equation reads z % + a-o% +ny = 0. Differentiating k times with respect to z: k=) 4.8, + BinB 0, & il 4, ¢y,,4y_ dy dy dy, dy (2) rie ea a aa ge 5 &=b Bogen hen n - 0 However, since n° $4 - £ deo, vehwe Sreri-oSa-p|de =0 (b) Let y* = Li(2), 50 that if fi) = 2 MeV Lig) = ey, then ¢ = lege 5 an regen _ } yer, 4 Meme , GED rien ] i yew -4y"e png _ (ee ED He 2, en yt 1 era2e + a yhleRythMn , on verre ED yore,438 Problems and Solutions £E. remin saeyensen peagen = oo sree 1 yee cet) ere 2 4 2 Dyer EEN? 4 Rp yhe Lh - ye | of 44 _ ou, mal 2 oe Uk ‘ tek ev ay? + (kt 1-ay" seb -rge-tyn |Z ey tee [Ex yl. Since from part (a) above, ay? + (k+ 1 - zy" + (n- Wyk =0, we have ee A fk-IVryt (1-2 - 22k + 22 + 4en) az : =-e7 df df 1 ohn Ee Ela W)+ tn 2k + 2 ZI = 0. VL. Prove that for any values of A and 1: (Refer to Sections 6.11.1 and 15.1) @ x TA+n+ 10 (pent) _ DQtpt2ntl) 2 ntn-m) TO DE (utm+) (nT tp tntl) ” (Hint: Compare coefficients of x" in (1+x)*"(1+x)""" with those in (aynMathematical Methods in Physics I 439 = F DT Otpt2nty@aynr ® see ~ ¥ nT (tpentE(A+n+]) imp" © Fa) Saal) + Jae) A@ = 2sindn ae Solution: . TA+n+1)T(p+ntl) @ » aig mre ee - 1 [Porn , rowae] , On! | TH+) Tat) m=0 T(ntat D0 (nt p+ 1(@-1)! T(u#2)F(n+A) met Se: DI Pqu+2)F(n+ay mond fs teneneeeeD ~ 2 fan wan (atA)\(ntp)! ut a f } @-D! [@A-DIGD! (tun)! ee 10. ha = Fy [emtwtne-1...(u41) + QoenyQen-1)...041)] + pep Lent rmn-n...0049))] se a —[o(arnaan-t.. 082))] +... The product (1+x)**"(1+x)"*" can be expressed as [A][B], where ll = [i + Grane + Geman De) + [aemaen pene A |: a440 Problems and Solutions [B]is identical to [A], except that A is replaced by 1. The coefficient of x" in this expansion product is seen to be: A [cmsmnn-1)..(041) + QoenyQeen-1)..0641)] 1 a an ie Ay [corp aematn-1)...0+2))] = ——_lPaenryruentt) A (m! (nm)! m+ LT (tm 1) * Consider the expression TAtpt+2nt1) _ At pt2nAtp+2n-1)...(Atp+ DT (A+p+1) a n! (ayTOtpent) (atpen)(Atptn- 1)... +n+ DE At +1) = CAtptany(atpt2n-1).. .Aatptnt] nl The coefficient of x" in the product expansion of (1+x)"***?" is given by: (Qnthtpy2nthtp-1)...2ntatp-[n-1)) _ _T(A+p+2n+1) nt (iT (At+ptntl) Hence, since the expansions on both sides of (a) are for the same quantity, namely, (1+x)*""(1+x)""" = (1+x)""""", we have that . TAtnt 0 (uentl) _ Ttpt2n-1) © (niXn-m) IP On-me DT (uemel) ~ (nT Otptne) ” (b) Since Oke, Pee A yitay adobe -S(z 1) ee LE} BTG@+k 1 24 H@ = L(5) D@tl’ ® iy (kee -1)t" AZ) Jul) = L x (3) LT Ot DT (atl 1) * reMathematical Methods in Physics I! 4at If n= k +1, then as n runs from 0 to infinity, J also runs from 0 to infinity, but at best equal to m and never greater. Thus, oe A Zz a 1 1010 ° YL) oY Ligagreanroreny However, from part (a) above, a y Bn-DITO+n- 11 (u+l1) = T(Atp+2n-1 (nT Atptnt DT (Atat 1E(utntl)” so that Aspen _ PF CDTOt Inge 1 AOA = Y nTQtptnt DT (tne) T(utntl)” (©) Setting p = A-1 and A = ~A and then p= -A+1 and = + in the results of part (b) above, we have 8 _cayremyar! Fatt Sands = y TP(n)T(n-2+ DOr) . y (-1"Tem+2,¢2"" © ny !T(m+2)0 (m+ A4 1) (m-242) Setting n =m +1: _ CD TOG)" Jates* Fon 4 = orn WTO) i y C1" Tam+2\e2r""* © (myn TP Gn-A42)0(4A4m) - y (1yTQm+2\¢2"""" © Gays) (m-2+2)T AM) *442 Problems and Solutions Combining, WO TOTU-ayra) +S ("TQam+2yy2yn"! [ 1 1 |: (m)1P(m+ DTP (m-2#2)(14A4m) | m+1 m+1, Sahat tanh = Hence, a zTU-Ar@) m But F(A)P(1-A) = Sinn (Section 7.5.2) = JeaJnitJ-audy = 2a VIL. Let Z,(z) = AJa(z) + BHX(2), where A and B are arbitrary constants. Prove that the general solutions of the following differential equations are: @ Hrd [Lally 0 = f= ZH) de za |z ND ime (b) Be wep~o - So Daa [EE P) Solution: (a) Let y = 2i/Z, so that A(2iVz) > Aly).Mathematical Methods in Physics I 443 a [L(y J, Aleh Lah [1-3 - z (22) I* ¥lae y dy since J,()) is a solution of Bessel’s equation. ‘Thus we have shown that J4(2iV2) is a solution of equation (a). Since the Hankel function Hi (Section 7.5.4) is a combination of Ja(2ivz) and J-(2i42), it too is a solution of @, Hence the most general solution of equation (a) is Za(z) = AJa(z) + BI(2). aie (by + bz"f = 0. aed WB im2yr wg, is yont2) — so that 7 Yb2". Letting w 2 we have = phon?) = wins) mrt 2m) gg dw mt z= whe), v2. yma), Af A and & Pay Consider f = Vz. Zaye [A eo) + VzA(y). TH, a Dh pe ynr) i. gee ae WOE warren 2 dy BF BIE, camyeneay) , La YBOn+2) cm-1ytmery ee a ae + aD wae) t, - thoy, com+tytns2)] , Ba [¥BCm+2) | cm-tyne2 bef ole + os me) | + [onary444 Problems and Solutions Rearranging: hwy - pwr) i (A ohn seutino yy = (Ut2dy Substituting w = “TE, 4f, my (Fa OAL, (pd ae et bt oe eh oy Al! Fomae Buitr=—5, 2 t + bef = bwin fw (Pde Ses dealt] = 0, since Jjm+2(y) is a solution of Bessel’s equation. We have thus shown that VE Lime ee vb | is a solution i + bz"f=0, and therefore that ae [e*™Ay) -Ja0)] " San 2 I) - GA VEIA0) is also a solution doe + b2"f=0. Hence, the most general solution of Ema ae es Ois ae 22 gor).Mathematical Methods in Physics I 445 (©) Show that f = 2°Z,(6z) is the most general solution of the differential equation af ,1-20 df (e =| : ee (Oe Solution: Let y = Bz, so that z = y/B and Ja(Bz) > Jay). dQ) _ ah) dy _ 5 Bo #HO) _ gp fh) a& dy d& i ae : Consider f = 2°Z,(6z). d on ag hh eo adh te oe ‘ht eB [y TA ty’ SI: po # = Fy [ata-y" +20)" 8+ y 2) 2 (4+ 20 dh, , a(a- Ds). “pela ya” ¥ flor ar | » _ 1}, 2 22). (p 28) Be EY = [pe eh = Bec since J(y) satisfies Bessel;’s equation. Hence, 2°J(B2) is a solution of equation [1], and therefore 2“H;(Bz) is a solution as well. Hence, the most general solution of equation [1] is 2*Z,(Bz).446 Problems and Solutions Problem Set E I. Consider the differential equation for heat conduction (Chapter 9), a = AAT‘ S, for a one-dimensional medium (x 2 0). Let the boundary condition at x = 0 be (@TY (6x) = 0 for all time (i.e. the medium is sur- rounded by an insulating wall at x~ 0. (a) Show that the Green’s function for the homogeneous equation (S = 0) is given by 1 | exe ea) Ger) = eagle a te ae |. (b) Find the Green’s function for the inhomogeneous equation. Solution: i : a [ wy ext] (a) G(r) = pele MB + eae | = Gola) + Golxrx', we would be able to say that for all time ¢ and for x 2 0, Tex) = [oexsnTw or. We must show that 7(x,1) satisfies proper boundary conditions. . am)| _ Flacaxn 0 =F ne But x's) Go(x-x')| , |2Galx+2's) ex lo ox |xo ax |xo = | 2Gberx.0| , |aGorrx' Clg | ere. : oe roy oy get [E-2]-0.Mathematical Methods in Physics I 447 Thus, [zz 0, and the boundary condition at x = 0 is satisfied. 0 Gi me) - AAT) = sé - ralocerionw onde. ext cont VaRIE AG(xx'31) = alt. Oa 2 2 | ex) Ce) ol 2 cee Mt on eel (yy ext ways OS eae 2 2 x egy eet -aye Gare oe 2 2 Ef aeeye 1), -P[oeye MWERMAGOx's1) = a ee ele laa or, exe _@uy AG's) _ ale ar +e a ot a Vanat 2 2 ae’) 7 foexy 1), foe ea Vane SS em oor xlte ™ |ae af = WARMAG(x,'31).448 Problems and Solutions Hence we see that sed = AAG(xx'j1) = 0, so that ED gam -0 = Ted - foewsyrwoav satisfies the heat conduction equation. (ill) Ast 0+, G(xx'st) = Gola-a') + Golxtx’) + B(x-x') for x’ > 0. Hence, Tut) = foa-x)TxO}d’ = THO), and a second boundary condition is satisfied. _@xy _ ax? (iv) Asx +=, GOxx'st) = au te 4 | + 0,sothat Tht) ~ 0. oe Thus, all boundary conditions are satisfied, (b) To find the Green’s function for the inhomogeneous equation (S # 0), we must find a Green’s function G(x-Xo,t~to) satisfying 1 (3. Aa}gta-sot-1) = B0-rb(t-), so that a particular solution may be written as Text) = far [Go-x -1)s' yd. Define G(x-Xof-to) = G(xx',1') for f > 0 (i.e. > 10). Gl&-Xof-fo) = 0 fort’ < 0 (ie. <1). ‘We must show that this G(-xo,t~te) satisfies equation [1].Mathematical Methods in Physics I 449 @ Fort’ >0(t>t0),G=G and (5 - rales = 0 from (ii) part (a), while 5(x-X0)8(t-fo) = 0 for f > fo. Hence the above condition is satisfied for > 0. i) For f <0(t< to), G= 0 while 8(x-x0)5(t-t0) = 0 for t< to. Hence the condition is satisfied for f < 0. (iii) Ast + fo, the right-hand side of equation [1] approaches 8(x-x0)6(0), while the left-hand side approaches 6(2-%o) + 8(x#%0) + B(x) for +9. ‘We must investigate whether both sides have the same singularity at t= fo. fu (§-2a]oe-x009 i furee-2040 = 8G-%). i far 292-00) ao!) ig afer geen) 7 lim AeAG—x-X,£) = lim Gare) - 0 = tim |Ga-n2) = B(x-%). Hence for t ~> fo, both sides approach the same singularity, namely (2-20). Summarizing, we have shown G(x-Xo,f-fo) to satisfy equation [1], so that a particular solution of the inhomogeneous equation is given by Tet) = fat [Gu-x -rysu'de. Since for ’ >t, 1-7 <0 and G=0, we can rewrite the particular solution as That) = Jer Foo-xs-nsenae, and the most general solution will then be given by Tet) = Jowe )T(x'0)dx’ + far Joorxs-r S(t) dx’.450 Problems and Solutions IL. Consider problem I above, where instead of the adiabatic boundary con- dition, one has at x = 0 and for all time, oT rm eetere ec aes where a. is a constant. (@) Find the corresponding Green’s function for the homogeneous and the inhomogeneous equation. (b) Express, explicitly, the temperature distribution for all > 0 in terms of its initial distribution at r = 0 and the source function S(t). Solution: (@) Suspecting that Go(x-x’,t) and Go(x+x',t) will not suffice to give (@TY(x), we try adding a continuous source function a(n), ie. we try G(x) = Glx-x') + CGLxtx,t) + formycae-nsdn. a 2 LR) GRIT Gl0r's) = (1+ Oe ar + forme a. aM +Ce a + Joiner 4m dn]. 1 (xy) _ axe ok (9) ° (1) Gx’) = Vaart Atx=0: From equation [1], i 2 Gx) Fl et | Vii | _ =(1- oa % Slacnye at dino. Note that hoMathematical Methods in Physics I 451 Thus, 2 flan a). - fed et a ax J,lacme nes J) 55 je Mo 2 |" -t| oat = -la(n)e ats em al aie = lace Bel + fe 2 ay ~*3 an 2 2 epee ne = -a(x'je 4 +fe ae oom an Thus, at x= 0, oo B) amr sein =e or {Sg ax) fe B20 ay, Since we want, as boundary condition at x = 0, that (@GY{(@x) + aG = 0, we have, combining equations [2] and [3]: fo eae fra-or — a(-x)]+ (1+ Chae” Sie + fer tx 22D) cy 2M 2 on ae + afa(n)e” Hi dq = 0. Since equation [4] must be true for all > 0, we collect terms having the same powers of t: @1-C=0 = C=1. ii) a(-x')- (1 +C)=0 = a(-x')=20. (ii) SSD + aa(n) =0 = a(n)= be. Atn =-x,a(-x)=20= be = b=2ae, so that a(n) = 20e%?, 2 2 2 [5] V4ntGx' =e 1 +e a +200 fe au edn.452 Problems and Solutions Properties of G(xx',t) (i) G(xx’,t) satisfies the boundary condition that at x = 0, a +aG=0. (It was deliberately constructed so.) ii) G(x’) satisfies the heat conduction equation, Se) = AAG’). Proof: By letting Gx’) = Ger’) + Glatt) and oe G'0x'f) = ae Je au edn, so that G(xx’,t) = G°(x-x',1) + G"(x+x’,0), it then follows that 2G _ 3 , aG" : " at ‘at a AG = 24G°+AG". Since we already know that e = AAG’, it remains for us to show that ag = AAG". 2 Fe ay 1_9G"_ al Qe & ” orl Vana 2 gees a dn —————_ - feo ae Vanat avanay” © a FE a xn - abalone tt 7 ae |Mathematical Methods in Physics Il 453 re _ oma je ae dy aac le 2ae ax? V4anat i 2 a f2|- ew on Sen ea edn : 2 ee ee, eal aaa]? EdD = Oe at” Thus, CoG. oe AAG’ ‘at AAG. (iii) Asx + =, G ~ 0, as required. (b) For S # 0, we must find a new Green’s function G(xx’,t) such that 18 (5, Aa]oeer.n ~ 8m. Define Gle-Xost-to) = Glxx’,) for > 0 (ie. 1> t). Gl-Xot-te) = 0 for’ <0 (ie. t< fe). To show that the new Glx-%,f-to) satisfies [6] above, we show it to be true for the following cases: @ Fort’ >0 (t> 1), G= G. But, from part (a) for S = 0, we showed that 2 - rae =0, while for 1’ #0, 8(x-x')0(¢) = 0, thus satisfying both sides of [6] for r' > 0.454 Problems and Solutions Gi) For’ <0 (t< 1), G=0, while for f # fo, 5(x-x')O(r’) = 0, thus satisfying both sides of [6] for r’ < 0. (iii) Ast > f& (f + 0), G20 Ga ‘To compare singularities on both sides of [6] as t -* fo, consider: "ad tim | Ga-xot-t0) tim [ Gox-zo,e) - AeA] = &x-%) lim, Joce-x)a¢0dr = B(x-%). Hence, both sides of [6] approach the same singularity, namely, B(x~%0). We have thus shown that G satisfies [6], and can therefore say that for all 1> 0, Text) = foo. )T(x,0)dx" + fe foox. ASC ds’. IIL. Consider a ring conductor of length a. The heat conduction equation is given by ar_ er, or ax where x is the length measured along the circumference. The boundary condition is given as T(x,1) = Txta,t), 5, (@) Prove that the Green’s function is anna Ge-x) = 4 1+ 29, cos HO = 2 ES Solution: We must show that G(x-2,f) has the following properties: (i) G(x-x satisfies the heat conduction equation for $= 0.Mathematical Methods in Physics I 455 Proof: @ ac | -1fpfce zante) on [2] a mt @ 0G 1 F228 sin A ans) an ala gtd yy _ feet Joy antag Zante) EM) 196 al a’ a x (i) Gx-x,0) at r= 0 is the delta function 8(x-x’). Proof: Ast 0 forx= 2’ (0,9 ~ tim 4] +2[c0s 2A 2) + 295 AOA? +. = dfisag siete.) = o, Ast — 0 forx# x’: Letting y = x-x', we have Foo.owy “n us, ‘Feo.o.y = 1 and G(x-x,0) behaves like the delta function (x-2’). “

You might also like

• The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  

  From Everand

  The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  Mark Manson

  Rating: 4 out of 5 stars

  4/5 (5783)
• The Yellow House: A Memoir (2019 National Book Award Winner)

  

  From Everand

  The Yellow House: A Memoir (2019 National Book Award Winner)

  Sarah M. Broom

  Rating: 4 out of 5 stars

  4/5 (98)
• Never Split the Difference: Negotiating As If Your Life Depended On It

  

  From Everand

  Never Split the Difference: Negotiating As If Your Life Depended On It

  Chris Voss

  Rating: 4.5 out of 5 stars

  4.5/5 (838)
• • Magazines
  • Podcasts
  • Sheet music
• Shoe Dog: A Memoir by the Creator of Nike

  

  From Everand

  Shoe Dog: A Memoir by the Creator of Nike

  Phil Knight

  Rating: 4.5 out of 5 stars

  4.5/5 (537)
• The Emperor of All Maladies: A Biography of Cancer

  

  From Everand

  The Emperor of All Maladies: A Biography of Cancer

  Siddhartha Mukherjee

  Rating: 4.5 out of 5 stars

  4.5/5 (271)
• Fear: Trump in the White House

  

  From Everand

  Fear: Trump in the White House

  Bob Woodward

  Rating: 3.5 out of 5 stars

  3.5/5 (738)
• Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  

  From Everand

  Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  Margot Lee Shetterly

  Rating: 4 out of 5 stars

  4/5 (890)
• The Little Book of Hygge: Danish Secrets to Happy Living

  

  From Everand

  The Little Book of Hygge: Danish Secrets to Happy Living

  Meik Wiking

  Rating: 3.5 out of 5 stars

  3.5/5 (399)
• Team of Rivals: The Political Genius of Abraham Lincoln

  

  From Everand

  Team of Rivals: The Political Genius of Abraham Lincoln

  Doris Kearns Goodwin

  Rating: 4.5 out of 5 stars

  4.5/5 (234)
• Yes Please

  

  From Everand

  Yes Please

  Amy Poehler

  Rating: 4 out of 5 stars

  4/5 (1888)
• Grit: The Power of Passion and Perseverance

  

  From Everand

  Grit: The Power of Passion and Perseverance

  Angela Duckworth

  Rating: 4 out of 5 stars

  4/5 (587)
• Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  

  From Everand

  Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  Gilbert King

  Rating: 4.5 out of 5 stars

  4.5/5 (265)
• A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  

  From Everand

  A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  Dave Eggers

  Rating: 3.5 out of 5 stars

  3.5/5 (231)
• On Fire: The (Burning) Case for a Green New Deal

  

  From Everand

  On Fire: The (Burning) Case for a Green New Deal

  Naomi Klein

  Rating: 4 out of 5 stars

  4/5 (72)
• Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  

  From Everand

  Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  Ashlee Vance

  Rating: 4.5 out of 5 stars

  4.5/5 (474)
• Principles: Life and Work

  

  From Everand

  Principles: Life and Work

  Ray Dalio

  Rating: 4 out of 5 stars

  4/5 (599)
• Rise of ISIS: A Threat We Can't Ignore

  

  From Everand

  Rise of ISIS: A Threat We Can't Ignore

  Jay Sekulow

  Rating: 3.5 out of 5 stars

  3.5/5 (137)
• The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  

  From Everand

  The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  Ben Horowitz

  Rating: 4.5 out of 5 stars

  4.5/5 (344)
• The Unwinding: An Inner History of the New America

  

  From Everand

  The Unwinding: An Inner History of the New America

  George Packer

  Rating: 4 out of 5 stars

  4/5 (45)
• Steve Jobs

  

  From Everand

  Steve Jobs

  Walter Isaacson

  Rating: 4.5 out of 5 stars

  4.5/5 (806)
• The World Is Flat 3.0: A Brief History of the Twenty-first Century

  

  From Everand

  The World Is Flat 3.0: A Brief History of the Twenty-first Century

  Thomas L. Friedman

  Rating: 3.5 out of 5 stars

  3.5/5 (2219)
• Angela's Ashes: A Memoir

  

  From Everand

  Angela's Ashes: A Memoir

  Frank McCourt

  Rating: 4.5 out of 5 stars

  4.5/5 (440)
• The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  

  From Everand

  The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  Brené Brown

  Rating: 4 out of 5 stars

  4/5 (1090)
• John Adams

  

  From Everand

  John Adams

  David McCullough

  Rating: 4.5 out of 5 stars

  4.5/5 (2409)
• Bad Feminist: Essays

  

  From Everand

  Bad Feminist: Essays

  Roxane Gay

  Rating: 4 out of 5 stars

  4/5 (1015)
• The Glass Castle: A Memoir

  

  From Everand

  The Glass Castle: A Memoir

  Jeannette Walls

  Rating: 4.5 out of 5 stars

  4.5/5 (1711)
• The Outsider: A Novel

  

  From Everand

  The Outsider: A Novel

  Stephen King

  Rating: 4 out of 5 stars

  4/5 (1800)
• The Woman in Cabin 10

  

  From Everand

  The Woman in Cabin 10

  Ruth Ware

  Rating: 3.5 out of 5 stars

  3.5/5 (2322)
• A Man Called Ove: A Novel

  

  From Everand

  A Man Called Ove: A Novel

  Fredrik Backman

  Rating: 4.5 out of 5 stars

  4.5/5 (4609)
• The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  

  From Everand

  The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  Viet Thanh Nguyen

  Rating: 4.5 out of 5 stars

  4.5/5 (119)
• The Light Between Oceans: A Novel

  

  From Everand

  The Light Between Oceans: A Novel

  M.L. Stedman

  Rating: 4.5 out of 5 stars

  4.5/5 (789)
• Brooklyn: A Novel

  

  From Everand

  Brooklyn: A Novel

  Colm Tóibín

  Rating: 3.5 out of 5 stars

  3.5/5 (1937)
• Wolf Hall: A Novel

  

  From Everand

  Wolf Hall: A Novel

  Hilary Mantel

  Rating: 4 out of 5 stars

  4/5 (3811)
• Manhattan Beach: A Novel

  

  From Everand

  Manhattan Beach: A Novel

  Jennifer Egan

  Rating: 3.5 out of 5 stars

  3.5/5 (791)
• Little Women

  

  From Everand

  Little Women

  Louisa May Alcott

  Rating: 4 out of 5 stars

  4/5 (104)
• The Perks of Being a Wallflower

  

  From Everand

  The Perks of Being a Wallflower

  Stephen Chbosky

  Rating: 4.5 out of 5 stars

  4.5/5 (2099)
• The Art of Racing in the Rain: A Novel

  

  From Everand

  The Art of Racing in the Rain: A Novel

  Garth Stein

  Rating: 4 out of 5 stars

  4/5 (4193)
• A Tree Grows in Brooklyn

  

  From Everand

  A Tree Grows in Brooklyn

  Betty Smith

  Rating: 4.5 out of 5 stars

  4.5/5 (1929)
• Her Body and Other Parties: Stories

  

  From Everand

  Her Body and Other Parties: Stories

  Carmen Maria Machado

  Rating: 4 out of 5 stars

  4/5 (821)
• Sing, Unburied, Sing: A Novel

  

  From Everand

  Sing, Unburied, Sing: A Novel

  Jesmyn Ward

  Rating: 4 out of 5 stars

  4/5 (1103)
• The Constant Gardener: A Novel

  

  From Everand

  The Constant Gardener: A Novel

  John le Carré

  Rating: 3.5 out of 5 stars

  3.5/5 (104)