A FRIENDLY APPROACH TO COMPLEX ANALYSIS Sara Maad Sasane • Amal Sasane , ►world Scientific A FRIENDLY APPROACH TO COMPLEX ANALYSIS A FRIENDLY APPROACH TO COMPLEX ANALYSIS Sara Maad Sasane Lund University, Sweden Amol Sasane London School of Economics, UK 1!) World Scientific NEW JERSEY • LONDON • SINGAPORE • BEIJING • SHANGHAI • HONG KONG • TAIPEI • CHENNAI Published by World Scientific Publishing Co. Pte. Ltd. 5 Toh Tuck Link, Singapore 596224 USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601 UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library. Credit for Image from Wikimedia Commons: Color plot of complex function (x2-1) * (x-2-I)2 / (x2+2+2I), hue represents the argument, sat and value represents the modulus (Pennission=CC-BY 2.5). Image creator: Claudio Rocchini Source: http:!/en.wikipedia.org/wiki/File:Color_complex__plot.jpg A FRIENDLY APPROACH TO COMPLEX ANALYSIS Copyright© 2014 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 ofmaterial 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. ISBN 978-981-4578-98-l ISBN 978-981-4578-99-8 (pbk) Printed in Singapore by World Scientific Printers. Preface We give an overview of what complex analysis is about and why it is im­ portant. As the student must have learnt the notion of a complex number at some point, we will use that familiarity in our discussion here. Later on, starting from Chapter 1 onwards, we will start things from scratch again. So the reader should not worry about being lost in this preface! What is Complex Analysis? In real analysis, one studies (rigorously) calculus in the setting of real num­ bers. Thus one studies concepts such as the convergence of real sequences, continuity of real-valued functions, differentiation and integration. Based on this, one might guess that in complex analysis, one studies similar con­ cepts in the setting of complex numbers. This is partly true, but it turns out that up to the point of studying differentiation, there are no new features in complex analysis as compared to the real analysis counterparts. But the subject of complex analysis departs radically from real analysis when one studies differentiation. Thus, complex analysis is not merely about doing analysis in the setting of complex numbers, but rather, much more specialized: Complex analysis is the study of "complex differentiable" functions. Recall that in real analysis, we say that a function f : JR: -+ JR: is differentiable at x0 E JR: if there exists a real number L such that lim f(x) - f(xo) = L, x➔xo X - Xo V A Friendly Approach to Complex Analysis vi that is, for every E > 0, there is a 8 > 0 such that whenever 0 < lx-xol < 8, there holds that x - x o) _ I f( ) f( x-xo LI < E. In other words, given any distance E, we can make the difference quotient f(x) - f(xo) x-xo lie within a distance of E from the real number L for all x sufficiently close to, but distinct from, xo. In the same way, we say that a function f : (C --+ (C is complex differen­ tiable at z0 E (C if there exists a complex number L such that z - zo lim f ( ) f( ) z - Zo z-+zo = L, that is, for every E > 0, there is a 8 > 0 such that whenever O < lz -zol < 8, there holds that I f(z) - J(zo) _ z-zo LI < E. The only change from the previous definition is that now the distances are measured with the complex absolute value, and so this is a straightforward looking generalization. But we will see that this innocent looking generalization is actually . quite deep, and the class of complex differentiable functions looks radically different from real differentiable functions. Here is an instance of this. x2 if X > 0, . Example 0.1. Let f: Ill--+ Ill be given by J(x) = { 2 -X I"f X < 0. f 0 Fig. 0.1 Graphs of the functions f and its derivative f'. vii Preface Then f is differentiable everywhere, and f'(x) = { 2x �fx:::::o, -2X IfX < 0. (0.1) Indeed, the above expressions for f'(x) are immediate when x =/- 0, and J'(0) = 0 can be seen as follows. For x =/- 0, I f(x) -f(O) - o = f(x) = El:_= = lxl Ix - 0I, I J I x -0 x lxl and so given E > 0, we can take J = o < Ix -01 < o, I E (> 0) and then we have that whenever f(x) - f(O) oJ I - I - = x 0 < J = E. x-0 However, it can be shown that f' is not differentiable at 0; see Exercise 0.1. This is visually obvious since f' has a corner at x = 0. Summarizing, we gave an example of an f : IR?. -+ IR?., which is differ­ entiable everywhere in IR?., but whose derivative f' is not differentiable on R In contrast, we will later learn that ifF : (C -+ (C is complex differen­ tiable function in (C, then it is infinitely many times complex differentiable! In particular, its complex derivative F' is also complex differentiable in C. Clearly this is an unexpected result ifall we are used to is real analysis. We will later learn that the reason this miracle takes place in complex analysis is that complex differentiability imposes some "rigidity" on the function which enables this phenomenon to occur. We will also see that this rigid­ ity is a consequence of the special geometric meaning of multiplication of complex numbers. ◊ a Exercise 0.1. Prove that J' : R ➔ R given by (0.1) is not differentiable at 0. Why study complex analysis? Although it might seem that complex analysis is just an exotic generaliza­ tion ofreal analysis, this is not so. Complex analysis is fundamental in all ofmathematics. In fact real analysis is actually inseparable with complex analysis, as we shall see, and complex analysis plays an important role in the applied sciences as well. Here is a list ofa few reasons to study complex analysis: viii A Friendly Approach to Complex Analysis (1) PDEs. Iff : (C-+ (C is a complex differentiable function in C, then we have two associated real-valued functions u, v : IR2 -+ IR, namely the real and imaginary parts off: for (x,y) E IR2, u(x,y) := Re(f(x,y)) and v(x,y) := Im(f(x,y)). (x,y) � v(x,y) - -y f(x,y) u(x,y) The real and imaginary parts u, v off. Fig. 0.2 It turns out that real and imaginary parts u, v satisfy an important basic PDE, called the Laplace equation: {P u Dou := fJ 2 x P + {fJ u2 = 0. y Similarly Dov = 0 in IR as well. The Laplace equation itselfis important because many problems in applications, for example, in physics, give rise to this equation. It occurs for instance in electrostatics, steady-. state heat conduction, incompressible fluid flow, Brownian motion, etc. (2) Real analysis. Using complex analysis, we can calculate some integrals in real analysis, for example 2 1 00 cosx ---2 dx or -oo 1 + x Note that the problem is set in the reals, but one can solve it using complex analysis. Moreover, sometimes complex analysis helps to clarify some matters in real analysis. Here is an example ofthis. Consider 1 f(x) := - -2 , x E IR\ {-1,1}. 1 -x Then f has a "singularity" at x = ±1, by which we mean that it is not defined there. It is, however defined in particular in the interval (-1,1). The geometric series 1 + x2 + x + x6 + ... 4 ix Preface convergesfor lx 2 1 < 1, or equivalentlyfor lxl < 1, and we have 1 =f(x)forxE(-1,1). l+x 2+x 4+x 6+-··=1 -x 2 From theformulafor f, it is not a surprise that the power series rep­ resentation of the function f is valid onlyfor xE(-1,1), since f itself has singularities atx = 1 and atx = -1. But now let us consider the new function g given by g(x) := 1 , l+ x2 xER The geometric series 1- x2+ x4 -x 6+- . . . convergesfor I- x2 I < 1, or equivalentlyfor lxl < 1, and we have 1 = g(x)forxE(-1,l) . 2 +x l So the power series representation of the function g is again valid only for xE(-1,1), despite there being no obvious reason from theformula for gfor the series to break down at the points x = -1 and x = +1. The mystery will be resolved later on in this book, and we need to look at the complex functions 1-x 2+x 4-x 6 +-···= 1 1 and G(z) = F(z) = 2 1_z 1 + z2 ( whose restriction to � are the functions f and g, respectively). In particular, G now has singularities at z = ±i, and we will see that what mattersfor the power series expansion to be valid is the biggest size of the disk we can consider with center at z = 0 which does not contain any singularity of G. G F Fig. 0.3 Singularities of F and G. x A Friendly Approach to Complex Analysis (3) Applications. Many tools used for solving problems in applications, such as the Fourier/Laplace/ z-transform, rely on complex function the­ ory. These tools in turn are useful for example to solve differential equations which arise from applications. Complex analysis plays an important in applied subjects such as mathematical physics and engi­ neering, for example in control theory, signal processing and so on. (4) Analytic number theory. Perhaps surprisingly, many questions about the natural numbers can be answered using complex analytic tools. For example, consider the Prime Number Theorem, which gives an asymptotic estimate on the number 1r(n) of primes less than n for large n: 71 (n) Theorem 0.1. (Prime Number Theorem) lim n ➔ oo n 1 og n) = 1. It turns out that one can give a proof of the Prime Number Theorem using complex analytic computations with a certain complex differ­ entiable function called the Riemann zeta function. Associated with the Riemann zeta function is also a famous unsolved problem in an­ alystic number theory, namely the Riemann Hypothesis, saying that all the "nontrivial" zeros of the Riemann zeta function lie on the line Re(s) = ½ in the complex plane. We will meet the Riemann zeta func­ tion in Exercise 4.5 later on. What will we learn in Complex Analysis The central object of study in this course will be I holomorphic functions in a domain I that is, complex differentiable functions f : D ➔ <C, where D is a "do­ main" (the precise meaning of what we mean by a domain will be given in Subsection 1.3.4). The bulk of the book is then in Chapters 2, 3 and 4, where we construct the following three lanterns to shed light on our central object of study, namely holomorphic functions in a domain: (1) The Cauchy-Riemann equations, (2) The Cauchy Integral Theorem, (3) Taylor series. xi Preface The core content of the book can be summarized in the following Main Theorem 1: Theorem 0.2. Let D be an op_en path connected set and let f : D --+ C. Then the following are equivalent: (1) For all z ED, f'(z) exists. (2) For all z ED and all n 2: 0, f( n )(z) exists. fJu fJv · ble an d · (3) u := Re(f), v := Im(f) are continuous ly d:Fr = , iJJerentia fJx {)y 8u 8v . inD. 8x {)y (4) For each simply connected subdomain S ofD, there exists a holomorphic F: S--+ <C such that F'(z) = f(z) for all z ES. (5) f is continuous on D and for all piecewise smooth closed paths , in each simply connected subdomain ofD, we have -=-- 1 f(z)dz = 0. (6) If {z E <C: lz - zol ::S; r} CD, then there is a unique sequence (cn )n ::::o in <C such that for all z with lz - zol CX) < r, f(z) = :�::::Cn(z - zot- 1 n=O ( ) f( = -. d( 1 i 1(-zol=r (1"'> - zo )n+l 21r ------------Furthermore, Cn and Cn = f( n )(zo) . n.1 1 Don't worry about the unfamiliar terms/notation here: that is what we will learn, besides the proof! xii A Friendly Approach to Complex Analysis Complex Analysis is not complex analysis! Indeed, it is not very complicated, and there isn't much analysis. The analysis is "softer" than real analysis: there are fewer deltas and epsilons and difficult estimates, once a few key properties of complex differentiable functions are established. The Main Theorem above tells us that the subject is radically different from Real Analysis. Indeed, we have seen that a real­ valued differentiable function on an open interval ( a, b) need not have a continuous derivative. In contrast, a complex differentiable function on an open subset of (C is infinitely many times differentiable! This happens because the special geometric meaning of complex multiplication implies that complex differentiable functions behave in a rather controlled manner locally infinitesimally, and aren't allowed to map points willy nilly. This controlled behaviour makes these functions rigid and we will see this in Section 2.3. Nevertheless there are enough of them to make the subject nontrivial and interesting! The intended audience These notes constitute a basic course in Complex Analysis, for students who have studied calculus in one and in several variables. The title of the book is meant to indicate that we aim to cover the bare bones of the subject with minimal prerequisites. The notes originated as lecture notes when the second author gave this course for third year students of the BSc programme in Mathematics and/with Economics. Acknowledgements Thanks are due to Raymond Mortini, Adam Ostaszewski and Rudolf Rupp for many useful comments. This book relies heavily on some of the sources mentioned in the bibliography. This applies also to the exercises. At some instances we have given detailed references in the section on notes at the end of each chapter, but no claim to originality is made in case there is a missing reference. Sara Maad Sasane and Amol Sasane, London and Lund, 2013 Contents Preface V . 1 Complex numbers and their geometry 1.1 1.2 . 1 3 1.4 1.5 The field of complex numbers . Geometric representation of complex numbers . Topology of (C ••.•.•••.•......•.. 1.3.1 Metric on (C .••....•....•.. Open discs, open sets, closed sets and compact 1.3.2 sets .............. 1.3.3 Convergence and continuity Domains .......... 1.3.4 The exponential function and kith The exponential exp z .. 1.4.1 1.4.2 Trigonometric functions 1.4.3 Logarithm function Notes 2. Complex differentiability 2.1 2.2 2.3 2.4 2.5 1 5 1 1 1 2 1 3 1 4 1 5 17 18 2 2 2 3 27 9 2 Complex differentiability . Cauchy-Riemann equations .......... Geometric meaning of the complex derivative The d-bar operator . Notes .................. . 3 Cauchy Integral Theorem and consequences . 3 1 1 Definition of the contour integral xiii 0 3 36 49 56 58 59 59 xiv A Friendly Approach to Complex Analysis 3.. 1 1 An important integral .......... Properties of contour integration ........ Fundamental Theorem of Contour Integration . The Cauchy Integral Theorem........ . . 3.4.1 Special case: simply connected domains . 3.4.2 What happens with nonholomorphic functions? 3.5 Existence of a primitive ............... 3.6 The Cauchy Integral Formula ............... 3.7 Holomorphic functions are infinitely differentiable.... 3.8 Liouville's Theorem; Fundamental Theorem of Algebra. 3.9 Morera's Theorem: converse to Cauchy's Integral Theorem. 3.10 Notes ....... 3.2 3.3 3.4 4. Taylor and Laurent series · Series ....... Power series ....................... 4.2.1 Power series and their region of convergence 4.2.2 Power series are holomorphic . 4.3 Taylor series...... Classification of zeros ....... 4.4 4.5 The Identity Theorem ....... 4.6 The Maximum Modulus Theorem. 4.7 Laurent series ............ 4.8 Classification of singularities.... . 4 8.1 Wild behaviour near essential singularities 4.9 Residue Theorem . 4.10 Notes .... 4.1 4.2 5. Harmonic functions 5.1 5.2 5.3 What is a harmonic function? What is the link between harmonic functions and holomorphic functions? ..................... Consequences of the two way traffic: holomorphic ++ harmonic ................. 5.3.1 Harmonic functions are smooth 5.3.2 Mean value property 5.3.3 Maximum Principle ....... 67 69 73 77 82 85 88 91 96 99 101 103 105 106 107 107 113 117 12 2 12 6 12 9 13 1 14 1 15 1 15 3 162 163 163 165 170 171 172 172 Contents 5.4 5.5 Uniqueness of solution for the Dirichlet problem , Notes . . . . . . . . . . . . . . . . . . . . . . . . xv 173 176 Solutions 177 Bibliography 269 Index 271 Chapter 1 Complex numbers and their geometry In this chapter, we set the stage for doing complex analysis. We study three main topics: (1) We will introduce the set of complex numbers, and their arithmetic, making <C into a field, "extending" the usual field of real numbers. (2) Points in <C can be depicted in the plane JR2, and we will see that the arithmetic in <C has geometric meaning in the plane. This correspon­ dence between <C and points in the plane also allows one to endow <C with the usual Euclidean topology of the plane. (3) Finally we will study a fundamental function in complex analysis, namely the exponential function. We also look at some elementary functions related to the exponential function, namely trigonometric functions and the logarithm. 1.1 The field of complex numbers By definition, a complex number is an ordered pair of real numbers. For example, (1, 0), (0, 1), (0, 0), (-¾, v2) are all complex numbers. The set JR x JR of all complex numbers is denoted by <C. Thus <C = { z = ( x, y) : x E JR and y E JR}. For a complex number z = (x, y) E <C, where x, y E JR, the real number xis called the real part of z, and y is called the imaginary part of z. 1 2 A Friendly Approach to Complex Analysis We define the operations of addition "+" and multiplication "•" on C by: (x1,Y1) + (x2,Y2) = (x1 +x2,Y1 + Y2), (x1,Y1) · (x2,Y2) = (x1x2- Y1Y2,X1Y2 + X2Y1), for complex numbers (x1,YI), (x2,Y2). With these operations,C is a field, that is, (Fl) (C,+) is an "Abelian group", (F2) (C \ {0}, •) is an Abelian group, and (F3) the distributive law holds: for a, b, c EC, (a+ b) · c =a· c + b · c. In (Fl), Abelian group just means that the operation + on C is associative, commutative, there exists an "identity element" (0,0), such that (x,y) + (0,0) = (x,y) = (0,0) + (x,y) for all (x,y), and every element (x,y) has an "additive inverse" (-x,-y): (x,y) + (-x,-y) = (0,0) = (-x, -y) + (x,y). Similarly, in (F2), the multiplicative identity is (1,0), and the multiplicative inverse of a complex number (x, y) EC \ { (0,0)} is given by ( X , -y ) x2 + y2 x2 + y2 · Exercise 1.1. Check that (1.1) is indeed the inverse of (1.1) (x, y) EC\ {(O, 0)}. We have: Proposition 1.1. (C, +, ·) is a field. � is "contained" in C. In fact, we can embed� inside C, and view� as a "subfield" of C, that is, one can show that the map 1-7 (x,0) X sending the real number x to the complex number (x,0) is an injective field homomorphism. This just means that the operations of addition and multiplication are preserved by this map, and distinct real numbers are sent to distinct complex numbers. � X X1 +x2 X1 · X2 1 0 I-+ I-+ I-+ I-+ I-+ C (x,0) (x1 + x2,0) = (x1,0) + (x2,0) (x1 · x2,0) = (x1,0) · (x2,0) (1,0) (0,0) 3 Complex numbers Thus we can view real numbers as if they are complex numbers via this identification. For example, the real number v'2 can be viewed as the complex number ( v'2, 0). If this makes one uneasy, one should note that we have been doing such identifications right from elementary school, where for instance, we identified integers with rational numbers, for example, Z 3 3 = l E Q, and we didn't lose sleep over it! But the advantage of working with (C is that while in IR there was no solution x E IR to the equation x 2 + 1 = 0, now with complex numbers we have (0,1) · (0,1) + (1,0) = (-1,0) + (1,0) = (0,0). If we give a special symbol, say i, to the number (0,1), then the above says that i2 + 1 = 0, where we have made the usual identification of the real numbers 1 and 0 with their corresponding complex numbers (1,0) and (0,0). Henceforth, for the complex number (x, y), where x, y are real, we write x + yi, since (x, y) = (x, 0) + (y, 0) • (0, 1) = x + yi . ._.,_, .._,_, .._,_, =Y =i =X As complex multiplication is commutative, in particular, yi = iy, and so we have x + yi = x + iy. Exercise 1.2. Let 0 E x, y are real. ( - 27r, 27r) . Express 1 + i tan 0 1 _ i tan 0 . m the form x + yi, where Historical development of complex numbers. Contrary to popular belief, historically, it wasn't the need for solving quadratic equations, but rather cubic equations, that led mathematicians to take complex numbers seriously. The gist of this is the following. Around the sixteenth century, one viewed solving equations like ax 2 +bx+c = 0 as the geometric problem of finding the intersection point of the parabola y = x2 with the line y = -bx - c. Based on this geometric interpretation, it was easy to dismiss the lack of solvability in reals of a quadratic such as x 2 + 1 = 0, since that just reflected the geometric fact that parabola y = x 2 did not meet the line y = -1. See the picture on the left in Figure 1. 1. 4 A Friendly Approach to Complex Analysis 0 -1 Fig. 1.1 Lack of solvability in reals of x2 = -1 versus the fact that x 3 = always has a real solution x. 3px + 2q Meanwhile, Cardano (1501-1576) gave a formula for solving the cubic x3 3px + 2q, namely, X = {/q + ✓q2 _ p3 + {/q _ = ✓q2 _ p3. For example, one can check that for the equation x3 = 6x + 6, with p = 2 and q = 3, this yields one solution to be x = ,W + �- However, note that by the Intermediate Value Theorem, the cubic y = x 3 always intersects the line y = 3px + 2q. See the picture on the right in Figure 1.1. But for an equation like x 3 = 15x + 4, that is when p = 5 and q = 2, we have q2 - p3 = -121 < 0, and so Cardano's formula fails with real arithmetic, but we do have a real root, namely x = 4: 43 = 64 = 60 + 4 = 15 · 4 + 4. Three decades after the appearance of Cardano's work, Bombelli suggested that maybe with the use of complex arithmetic, Cardano's formula would give the desired real root. So we may ask if x = {12 + lli + {/2 - lli � 4. One can check that (2+i) 3 = 2+11i and (2-i) 3 = 2-lli, so that the above does work with these values of the cube root. Thus Bombelli's work estab­ lished that even for real problems, complex arithmetic might be relevant. From then on, complex numbers entered mainstream mathematics. Exercise 1.3. A field IF is called ordered if there is a subset P c IF, called the set of positive elements of IF, satisfying the following: (Pl) For all x,y E P, x+y E P. (P2) For all x, y E P, x · y E P. (P3) For each x E P, one and only one of the following three cases is true: .3,0 -XE P. 1° X = 0. £0 XE P. 5 Complex numbers For example, the field of real numbers IR! is ordered, since P := (0, oo) is a set of positive elements of R (Once one has an ordered set of elements in a field, one can compare the elements of F by defining a relation > p in F by setting y > p x for x, y E F if y - x E P.) Show that C is not an ordered field. Hint: Consider x := i, and first look at x · x. Geometric representation of complex numbers 1.2 Since (C = JR2, we can identify complex numbers with points in the plane. See Figure l. 2. Y ,____.(x,y) (0, 0) Fig. 1.2 X The complex number x + iy in the complex plane. The complex plane is sometimes called the Argand1 plane. Exercise 1.4. Locate the following points in the complex plane: 0, cos 1, -;, i, -v'2i, 3 +ism 3. 7r . • 7r So we can identify (Cas a set with the points in the plane JR2. Do the field operations in (Chave some geometric meaning in the plane? We see below that this is indeed the case: addition in (C is vector addition in the plane, and multiplication in (Chas a special geometric meaning in the plane, which is explained below as well. Geometric meaning of complex addition. With this identification of complex numbers with points in the plane, it is clear that complex addi­ tion is just addition of vectors in JR2. By addition of vectors, we mean the 1 It is named after Jean-Robert Argand (1768-1822), although it was used earlier by Caspar Wessel (1745-1818). A Friendly Approach to Complex Analysis 6 usual way of combining two vectors, that is, by completing the parallelo­ gram formed by the line segments joining (0,0) to each of the two complex numbers as sides, and then taking the endpoint of the diagonal from (0,0) as the sum of the two given complex numbers. See Figure 1.3. 0 Fig. 1.3 Addition of complex numbers is vector addition in JR. 2 • Indeed, the middle picture shows that addition of z1 and z2 as vectors in the plane yields the correct x-coordinate of their sum as complex numbers by looking at the two congruent blue triangles. Similarly, by the rightmost picture shows that the y-coordinate is also correct. Geometric meaning of complex multiplication. We will now see the special geometric meaning of complex multiplication. In order to do this, it will be convenient to use polar coordinates. Thus, let the point (x, y) E JR2 have polar coordinates r:::: 0 and 0 E (-1r, 1r]. This means that the distance of the point (x,y) to (0,0) is r (:::: 0), and the angle made by the ray from (0,0) to (x, y) makes an angle of 0 with the positive real axis (the x-axis). (If (x,y) is itself (0,0), we set 0 = 0.) Y i------�(x, y) r 0 Fig. 1.4 X Polar coordinates (r, 0) of (x, y) E JR. 2 • 7 Complex numbers Then from the right-angled triangle shown in Figure 1.4, we have x = rcos0, y = rsin0. Thus we can express the complex number in terms of the polar coordinates (r,0): x + yi = rcos0 + ( rsin0)i = r(cos0 + isin0). Now we give the geometric interpretation of complex multiplication. For two complex numbers expressed in polar coordinates as z1 = r1(cos01 + isin01), z2 = r2(cos02 + isin02), we have that z1 · z2 = r1(cos01 + isin01) · r2(cos02 + isin02) = r1r2(cos01cos02 - sin01sin02 + i(cos01sin02 + cos02sin01) ) = r1r2(cos (01 + 02) + isin (01 + 02) ), using the trigonometric identities for angle addition. Thus z1 · z2 has the polar coordinates ( r1r2, 01+02). In other words, the angles z1 and z2 make with the positive real axis are added in order to get the angle z1 • z2 makes with the positive real axis, and the distances to the origin are multiplied to get the distance z1 · z2 has to the origin. See Figure 1.5. Fig. 1.5 Geometric meaning of complex multiplication: angles get added, distances to the origin get multiplied. As a special case, consider multiplication by cos a:: + isin a::, which is at a distance of 1 from the origin. Then from the above, we see that if z E <C, then z • (cos a:: + isin a::) is obtained by rotating the line joining O to z anticlockwise through an angle of a::. In particular, multiplying z by i. = 0 + i. • 1 = cos 2 °+ i sm 2 1r produces a counterclockwise rotation of 90 . • • 1r 8 A Friendly Approach to Complex Analysis ______ -------: · (coso:+isino:) 0 Fig. 1.6 iz 0 Multiplication by cos a+ i sin a produces an anticlockwise rotation through a. De Moivre's formula and nth roots. We have for all n E N (cos 0+isin0)n = cos(n0) +isin(n0). This is called de Moivre's formula. Exercise 1.5. Recover the trigonometric equality cos(30) = 4(cos 0)3 -3 cos 0 using de Moivre's formula. Exercise 1.6. Express (1+i) 10 in the form x+iy with real x, y without expanding! Exercise 1.7. By considering (2+i)(3+i), show that�= tan- 1 � +tan- 1 i. Exercise 1.8. Gaussian integers are complex numbers of the form m+in, where m, n are integers. Thus they are integral lattice points in the complex plane. Show that it is impossible to draw an equilateral triangle such that all vertices are Gaussian integers. Hint: Rotation of one of the sides should give the other. Recall that v'3 r/. Q. De Moivre's formula gives an easy way of finding the nth roots of a complex number z, that is,complex numbers w that satisfy w n = z. Indeed,we first write z = r(cos0+isin0) for some r:?: 0 and 0 E [0,21r). Now ifwn = z, where w = p(coso:+isino:),then w n = pn (cos(na) +isin(no:)) = r(cos0+isin0) = z, and so by equating the distance to the origin on both sides, we obtain pn = r. Hence p = y'r, as both p and r are nonnegative. On the other hand,the angle that wn makes with the positive real axis is no:, which is in the set{·•• , 0 - 41r, 0 - 21r, 0, 0+ 21r, 0+ 41r, 0+ 61r, •··},because the angle made by a nonzero z with the positive real axis is unique only up to integral multiples of 21r, that is, instead of 0, we could just as well have used 0+ 21rk for any integer k. See Figure 1.7. 9 Complex numbers z 0 + 21r etc. Fig. 1. 7 The angle made by z with the positive real axis. Thus we get that a E { � + 2: k : k E Z} , and this gives distinct w for 0 0 27r 0 21r 0 21r aE {-, -+-, -+2·-, ... , -+(n-1)·-} . nn n n n n n In particular, if z = 1, we get the nth roots of unity, which are located at the vertices of an n-sided regular polygon inscribed in a circle. See Figure 1.8. Fig. 1.8 The six 6th roots of unity. Exercise 1.9. Find all complex numbers w such that w4 the complex plane. = -1. Depict these in Exercise 1.10. Find all complex numbers z that satisfy z 6 - z 3 - 2 = 0. Exercise 1.11. If a, b, c are real numbers such that a + b + c = ab+ be+ ca, then they must be equal. Indeed, doubling both sides and rearranging gives (a - b) 2 + (b - c)2 + (c - a) 2 = 0, and since each summand is nonnegative, it must be the case that each is 0. On the other hand, now show that if a, b, c are complex numbers such that a2 + b2 + c2 = ab + be + ca, then they must lie on the vertices of an equilateral triangle in the complex plane. Explain the real case result in light of this fact. Hint: Calculate ((b - a)w + (b - c)) • ((b - a)w 2 + (b - c)), where w is a nonreal cube root of unity. 2 2 2 10 A Friendly Approach to Complex Analysis Exercise 1.12. The Binomial Theorem says that if a, bare real numbers and n EN, then are the binomial coefficients. The algebraic reasoning leading to this is equally valid if a, b are complex numbers. Show that ( 3n 0 ) + ( Hint: Find (1 + 1) 3n root of unity. 3n 3 ) + 3n ( ) 6 + ... + (3n 3n) + (1 + w) + (1 + w2 ) 3n 3n , = 23n + 2 . (-l) n 3 where w denotes a nonreal cube Exercise 1.13. Show, using the geometry of complex numbers, that the line seg­ ments joining the centers of opposite external squares described on sides of an arbitrary convex quadrilateral are perpendicular and have equal lengths. Absolute value and complex conjugate. The absolute value lzl of the complex number z = x + iy, where x, y E JR:, is defined by lzl = Jx2 + y2 . Note that by Pythagoras' Theorem, this is the distance of the complex number z to O in the complex plane. See the picture on the left of Figure 1.9. By expressing z1, z2 E C in terms of polar coordinates, or by a direct calculation, it is clear that lz1z2 I = lz1I · lz2 IExercise 1.14. Verify the property lz1z2 I Cartesian coordinates. iy Z lzl lxl Z = X +iy = X +iy X I YI -iy Z= X - iy Fig. 1.9 The absolute value of z is the distance of z to the origin, and the complex conjugate is obtained by reflecting z in the real axis. 11 Complex numbers The complex conjugate z of z = x + iy where x, y E IR?., is defined by z= x-iy. In the complex plane, z is obtained by reflecting the point corresponding to z in the real axis. See the picture on the right of Figure 1.9. From this geometric interpretation, convince yourself that for all z1, z2 E z1 + z2 = z1 + z2 and z1 · z2 The following properties are easy to check: Re(z) z=z, = --, z+z 2 = z1 · zz. Im(z) Exercise 1.15. Verify that the four equalities above hold. Exercise 1.16. Prove that for all z EC, lzl Give geometric interpretations of each. = lzl, re, z-z 2i = -.-. IRe(z)I :S: lzl and IIm(z)I :S: lzl. Exercise 1.17. If a, z E (C satisfy lal < 1 and lzl :S: 1, then prove that It� :S: 1. ;z I Exercise 1.18. Consider the polynomial p given by p(z) = eo + c1z + • • • + Cd Zd , where eo, c1, ..., Cd E � and Cd -I 0. Show that if w EC is such that p(w) = 0, then also p(w) = 0. l Exercise 1.19. Show that the area of the triangle formed by 0, a, b EC is 1 z1 z1 Exercise 1.20. Prove for any complex z1, z2 , zs that i <let [ 1 z2 z2 1 Z3 Z3 I I· Im ab) ; is real. Exercise 1.2 1. Show that for any two complex numbers z1, z2, there holds that lz1 + z2 1 2 + lz1 - z2 1 2 = 2(lz11 2 + lz2 / 2 ). What is the geometric interpretation of this equality? 1.3 Topology of re The concepts in ordinary calculus in the setting of IR?., like convergence of sequences, or continuity and differentiability of functions, all rely on the notion of closeness of points in IR?.. For example, when we talk about the convergence of a real sequence (an ) n EN to its limit L E IR?., we mean that given any positive E, there is a large enough index N such that beyond that index, the corresponding terms an all have a distance to L which is at most 12 A Friendly Approach to Complex Analysis E. This "distance of an to L" is taken as Ian - LI, and this is the length of the line segment joining the numbers an and L on the real number line. Now in order to do calculus with complex numbers, we need a notion of distance d(z1,z2 ) between for pairs of complex numbers (z1,z2 ), and the first order of business is to explain what this notion is. 1.3.1 Metric on C Since C is just JR.2, we use the usual Euclidean distance in JR.2 as the metric in C. Thus, for complex numbers z1 = x1 +iy1 and z2 = x2 +iy 2 , we have ✓ d(z1,z2 ) = (x1 - x 2 )2 +(Y1 - Y2 )2 = lz1 - z2 IBy Pythagoras's Theorem, this is the length of the line segment joining the points (x1,Y1),(x 2 ,Y 2 ) in lR.2 ; see Figure 1.10. I� Z2 = (x 2 ,Y2 ) lx1-x2 I d(z1,z2 ) :=lz1-z2 I ....._______.....,,.z1 = (x1,Y1) IY1 - Y2 I Fig. 1.10 The distance between z1 and z2 is the length of the segment joining z1 to z2. Using the geometric meaning of addition of complex numbers, and the well known result from Euclidean geometry that the sum of the lengths of any two sides of a triangle is at least as big as the length of the third side, we obtain the following triangle inequality for the absolute value: lz1 +z2 I::; lz1I +lz 2 I for z1,z2 EC. See Figure 1.11. The triangle inequality above can also be verified analyti­ cally by using the Cauchy-Schwarz inequality for real numbers x1,x2 ,y1,Y 2 : (xr +YI)(x� +y�) :2'.: (x1x2 +Y1Y2 ) 2 . 0 Fig. 1.11 Triangle inequality. 13 Complex numbers Exercise 1.22. Show that for all z1,z2EC, lz1-z2l 2: llz1I -lz2II­ Exercise 1.23. Sketch the following sets'in the complex plane: (1) {zEC:lz-(1-i)l=2}. (2) {zEC:lz-(1-i)I < 2}. (3) {zEC: 1 < lz-(1-i)I < 2}. (4) {zEC: Re(z-(1-i)) = 3}. (5) {zEC: IIm(z-(1-i))I (6) {zEC: lz-(1-i)I =lz-(1 + i)l}- (7) {zEC: lz-(1-i)I + (8) {zEC: lz-(1-i)I 1.3.2 < 3}. lz-(1 + i)I = 2}. + lz-(1 + i)I < 3}. Open discs, open sets, closed sets and compact sets In order to talk about sets of points near a given point, it will be convenient to introduce the following definitions. An open ball/disc D(zo,r) with center zo and radius r D(zo,r) := {z Ere: lz - zol < r}. > 0 is defined by U ofreis called open if for every z E U, there exists an Tz > 0 such D(z,rz ) C U. In other words, no matter what point we choose in U, A subset that there is always some "room" around that point comprising points only from U. D(z0, r) is an open set. D(z0, r) makes sense. Here are annulus Ar := { z Ere: r < lzl < 1} For example, it can be checked that an open disc So using the adjective open in the name for some more examples of open sets. The and the right half-plane IHI:= {z Ere: Re(z) > O} are open sets. It is also convenient to give a special name to sets whose complement is open, and these are called closed sets. One can also give a characterization of closed sets in terms of sequential convergence: A set and only if for every sequence (zn ) nE N in there holds that the limit LE F. F F c re is closed if which is convergent in re to L, 14 A subset for all z S of A Friendly Approach to Complex Analysis re is called bounded if there exists a M > 0 such that E S, JzJ ::; M. Thus S is contained in a big enough disc in the complex plane. A subset K c re is called compact if it is both closed and bounded. We will often use the known fact from real analysis that a real valued continuous function on a compact set possesses a maximizer and a minimizer. 1.3.3 Convergence and continuity We can also talk about convergent sequences in E re. A sequence (zn )n EN is said to be convergent with limit L if for every > 0, there exists an index holds that Jzn - LJ < N E N such that for every n > N, there E. It follows from the triangle inequality that for a convergent sequence the limit is unique, and we write lim Zn n➔oo Example 1.1. Let = L. z be a complex number with JzJ < 1. Then the sequence (zn )n EN converges to 0. Indeed, Jzn - OJ= Jzn J= JzJn = JJzJn - OJ, but as JzJ < ◊ 1, we know that Jzl n -+ 0 as n-+ oo. Exercise 1.24. Consider the polynomial p given by p(z) =co+ c1z+ • • • + cd zd , where eo, c1, ..., Cd E C and Cd =/- 0. Show that there exist M, R > 0 such that jp(z)I � Mlzl d for all z EC such that lzl > R. Exercise 1.25. Show that a sequence (zn ) nE N of complex numbers is convergent to L if and only if the two real sequences (Re(zn ))nE N and (Im(zn ))nE N are con­ vergent respectively to Re(L) and Im(L). Exercise 1.26. Show that a sequence (zn ) nE N of complex numbers is convergent to L if and only if (zn ) nE N converges to L. Exercise 1.27. Prove that C is complete, that is, every Cauchy sequence in C converges in C. (A sequence (zn )nE N is called a Cauchy sequence if for every E > 0, there is an index N E N such that for all indices m, n > N, there holds that Jzn - Zml < E.) re, re. Let S be a subset of z0 E S and f : S -+ continuous at z0 if for every E > 0, there exists a 8 Then f is said to be > 0 such that whenever < E. f is said to z ES satisfies Jz - zol < 8, there holds that lf(z) - f(zo)I be continuous if for every z E S, f is continuous at z. One can also give a characterization of continuity at a point in terms of convergent sequences: f : S -+ re is continuous at z0 E S if and only Complex numbers 15 if for every sequence (zn )nE N in S convergent to z0, there holds that the sequence (f(zn))nE N is convergent to f(zo). Example 1.2. Complex conjugation is continuous, that is, z 1-t z : (C --+ (C is continuous. Indeed, we have lz-zol = lz - zol = lz-zol for all z, zo E <C. This shows that complex conjugation is continuous at each z0 E <C, and so it is a continuous mapping. This is geometrically obvious, since complex conjugation is just reflection in the real axis, and so the image stays close to the reflected point if we are close to the point! Since (z) = z for all z E <C, complex conjugation is its own inverse. So complex conjugation is invertible with a continuous inverse. Thus com­ plex conjugation gives a homeomorphism (that is, a continuous bijective mapping with a continuous inverse) from (C to <C. ◊ Exercise 1.28. Prove that the map z 1.3.4 i-+ Re(z) : (('.-+ JR is continuous. Domains In the sequel, the notion of a path-connected open set will play an important role. By this we mean that we will prove results about our central object of study, namely functions f : D --+ (C that are complex differentiable at every point of the set D ( C <C), and it will turn out that for the validity of many of these theorems, we will need D to be a "nice" subset of <C, and not just any old subset of <C. Sets which satisfy this "niceness" assumption, stipulated precisely below, will be what we call a domain. We will call an open path-connected subset of, (C a domain. We already know what "open" means. Now let us explain what we mean by "path­ connectedness" . Definition 1. 1. (1) A path (or curve) in (C is a continuous function 1: [a, b] --+ <C. a b 16 A Friendly Approach to Complex Analysis (2) A stepwise path is a path 'Y : [a, b] -+ C such that there are points to = a < ti < · · · < tn < tn+l = b such that for each k = 0, l, ..., n, the restriction 'Y: [tk, tk+d -+ C is a path with either a constant real part or constant imaginary part. (3) An open set U C C is called path-connected if for every z1, z2 E U, there is a stepwise path 'Y: [a, b] -+ C such that 'Y(a) = z1, "f(b) = z2, and for all t E [a, bl, 'Y(t) E U. Actually in the above definition of path-connected open sets, the restriction that the paths being stepwise paths can be relaxed, that is, if we look at those open sets in which any two points can be joined merely by a path, then this class of open sets coincides with our path-connected sets. But this is an unnecessary diversion for us. So we just live with the definition we have given instead. Example 1.3. (1) The open unit disc lI)) := {z EC: lzl < 1} is a domain. (2) For r E (0, 1), the annulus Ar := {z EC: r < lzl < 1} is a domain. (3) The right half-plane IHI:= {z EC: Re(z) > 0} is a domain. IHI Fig. 1.12 The domains lill, Ar and IHI. On the other hand, the set S := {z E C : lzl -=/=- 1} =: C \ '][' is not a domain, since although it is open, it is not path-connected. Indeed, there 17 Complex numbers is no path joining, say O and 2: for if there were one such path 1, then by the Intermediate Value Theorem applied to the map t H l,(t)I : [ab]--+ , R, we see that since l,(a)I = 0 < 1 < 2 = l,(b)I, there must be at* E [ab] , such that l,(t*)I = 1, but then ,(t*) fj. S. ◊ Exercise 1.29. Show that the set domain. Exercise 1.30. Let also a domain. 1.4 D {z E (('.: Re(z) • Im(z) > 1} is open, but not a be a domain. Set D* := {z E (('.: z ED}. Show that D* is The exponential function and kith In this last section, we discuss some basic complex functions: the exponential function z H exp z, the trigonometric functions z H sin z, cos z, and the logarithm z H Logz. They will serve as counterparts to the familiar functions from calculus, to which they reduce when restricted to the real axis. In other words, when we restrict our functions to the argument z = x E R, then we get the usual real-valued functions xHe x , x H sinx, cosx, x H logx. So our definitions provide extensions of the usual real-valued counterparts; see Figure 1.13. Fig. 1.13 The real valued functions map points on the real line to the real line, but our definitions will give extensions of these to the shaded region, the complex plane. 18 A Friendly Approach to Complex Analysis We will see that these extensions have new and interesting properties in the complex domain that are not possessed by them when the argument is only allowed to be real. Also they will serve as important examples of complex differentiable functions: we will see later on that the exponential and trigonometric functions are complex differentiable everywhere in the complex plane, and the logarithm function is complex differentiable where it happens to be continuous. Let's begin with the exponential function. 1.4.1 The exponential exp z Definition 1.2 (The complex exponential). For z = x+iy E <C, where x, y are real, we define the complex exponential, denoted by exp z, as fol­ lows: expz = e x (cos y + i sin y). First we note that when y = 0 , the right hand side is simply e x . So our definition extends the usual exponential function (IR 3) x r-+ e x ( E IR). But this definition does appears to be mysterious. After all, z r-+ e Re(z) also gives an extension of the usual real exponential function. So why not use this simple definition instead? We define our exp function in the way we do, because we will see later on that this is the unique extension of the real exponential to the whole complex plane having the property that the extension is complex differentiable everywhere; see Example 4 .8 on page 1 28. In fact, just like the real counterpart, where we have that d x e = e x for all x E IR , dx we will later see that d expz = expz for all z E <C. dz So eventually we will learn that our mysterious looking definition is actually quite natural! Right now, let us check the following elementary properties: Proposition 1.2. (1) (2) (3) (4) (5) exp0 = e 0 (cos0 + i sin0) = 1 · (1 + i0) = 1 . For z1 , z2 E <C, exp(z1 + z2) = (expz1)(expz2). 1 For z E <C, expz-=/=- 0 , and (expz)- = exp(-z). For z E <C, exp(z + 21ri) = expz. For z E <C, I expzl = e Re(z) . 19 Complex numbers Proof. (2) If z1=x1 + iy1 and z2=x2 + iy2, then exp(z1 + z2) i = e(x1+x2 )+ (y1+v2 ) =ex1 +x2 (cos(y1 + Y2) + i sin(y1 + Y2)) = ex1 ex 2 (cos Y1 cos Y2 - sin Y1 sin Y2 + i (sin Y1 cos Y2 + cos Y1 sin Y2)) = ex1(cosy1 + i siny1)ex2 (cosy2 + i siny2)=(expz1)(expz2). (3) From the previous part, we see that 1=exp O=exp(z - z) =(expz)(exp(-z)), showing that expz =I- 0 and (expz)-1=exp(-z). Thus exp maps (C to the "punctured" plane (C \ { 0}. (4) We have exp(z + 21ri )=(expz)(exp(21ri )) =(expz) • e0 (cos(21r) + i sin(21r)) =(expz)·l·(l + i ·O)=expz. This shows that exp is "periodic in the y-direction" in the complex plane, with a period of 21r; see Figure 1.14. ' ' I I Z3 W2 -� W1 expzi exp 21r -� Wo z2 Z1 0 zo w_1 expwi '' ' '' ' Fig. 1. 14 21r-periodicity of exp in the y-direction. This phenomenon is not present in the x-direction, where just like in the real setting, the function x r--+ exp(x+iyo) (with Yo E IR fixed) is one-to-one. See Figure 1.15. 20 A Friendly Approach to Complex Analysis Fig. 1.15 x >-+ ex is one-to-one. J (5) For x, y E �, le x cosy+ ie x sin YI = e 2x ((cosy)2 + (siny)2 ) = e x . So I exp(x+iy)I = e x . This implies that exp maps vertical lines in the complex plane ( that is all points having a common real part) into circles ( that is all points having the same absolute value, in other words same distance to the origin). □ Proposition 1.2.(3) above shows that the map exp is not one-to-one, but rather, it is periodic with period 21ri. Figure 1.16 shows the effect of the mapping z f--t exp z on horizontal (fixed imaginary part y) and vertical lines (fixed real part x). This picture is arrived at by putting together the observations displayed in Figure 1.17. , 21r 37f I , I� I , -1 Fig. 1.16 0 1 � z ,-+ expz 2 7f 7f 2 0 2 The image of horizontal and vertical lines under the exponential map. 21 Complex numbers -oo-----+---+-__,,__+_oo exp � "e-(X)" "e+= " ---------6----'T--- e"(cos 0o+i sin0o) · +i0o exp � Xo+i· Xo Fig. 1.17 Bo exo (cos • +i sin·) exp � The image of horizontal and vertical lines under the exponential map. In Figure 1.16, we note that exp preserves the angle between the curves we have considered in its domain. Namely, the horizontal and vertical lines, which are mutually perpendicular, are mapped to circles and radial rays, which are also mutually perpendicular. We will see later on that this is no coincidence, and this property of "conformality", that is, of the preservation of angles between curves in the domain together with their "orientation" is something which is possessed by all complex differentiable functions in domains. Euler's formula. Note that for z =iy, where y is real, we have exp(iy) =cosy+i siny. This is the so-called Euler's formula. Hence the polar form of a complex number can now be rewritten as z =r(cos0+i sin 0) =r exp(i0). 22 A Friendly Approach to Complex Analysis 9 Exercise 1.31. Compute exp z for the following values of z: i ;, 3 + 1ri. Exercise 1.32. Find all z E (('. that satisfy exp z = 1ri. Exercise 1.33. Plot the curve ti-+ exp(it) : [O, 21r] -+ <C. Exercise 1.34. Describe the image of the line y = x under the exponential map z = x + iy i-+ exp z. Proceed as follows: Start with the parametric form x = t, y = t, and get an expression for the image curve in parametric form. Plot this curve, explaining what happens when t increases, and when t-+ ±oo. Exercise 1.35. Find the modulus and the real and imaginary parts of exp(z 2) and of exp (1/ z) in terms of the real and imaginary parts x, y of z = x + iy. 1.4.2 Trigonometric functions Just as we extended the real exponential function, we now extend the famil­ iar real trigonometric functions to complex trigonometric functions. From the Euler formula we established earlier, we have for real x that exp(ix) = cosx + i sinx and which gives exp(ix) exp(-ix) = cosx - i sinx, + exp( -ix) exp(ix) - exp(-ix) . and sm x = . 2 2i This prompts the following definitions: for z E C, we define exp(iz) + exp( -iz) exp(iz) - exp( -iz) . cosz = and smz = . 2 2i Clearly these definitions give extensions of the usual real trigonometric functions because when we put z = x, we get cos z = cos x and sin z = sin x, as we had seen above from Euler's formula. Several trigonometric identities continue to hold in the complex setting. For instance, cos(z1 +z2) = (cosz1)(cosz2) - (sinz1)(sinz2). Indeed, cos x = (cosz1)(cosz2) - (sinz1)(sinz2) exp(iz1) + exp(-iz1) exp(iz2) + exp(-iz2) = ( ) ( ) _ ( exp(iz1) 2 �t xp(-iz2) ) ( exp(iz2) 2 �t xp(-iz2) ) 2 exp(i (z1 +z2)) + 2 exp(-i (z1 +z2)) = COS(Z1 + Z2 ) • 4 Exercise 1.36. Show that sin(z1 + z2) = (sinz1)(cosz2) + (cosz1)(sinz2) for all z1, z2 E <C. = 23 Complex numbers Also, (sinz)2 + (cosz)2 = 1, since 2 2 exp(iz) - exp(- iz) exp(iz)+exp(-iz) . (sm z )2 + (cos z )2 = ( ) )+ ( 2i 2 exp(2iz) - 2+ exp(- 2iz) exp(2iz)+ 2+ exp(- 2iz) + -4 4 = 1. However, as opposed to the real trigonom etric functions which satisfy I sin xi ::; 1 and I cos xi ::; 1 for real x, z r-+ sinz and z r-+ cosz are not bounded. Indeed, for z = iy, where y is real, we have exp( ( y) )+ exp(- ( y)) exp(- y)+ exp(y) cos(iy) = _ _ _ z_· z_· _ _ _ _ _ _ _z·_ z·_ _ = _ _ _ _ _ _ _ _ _ _ 2 2 and so cos(iy) ➔+oo as y ➔ ±oo. Sim ilarly, since sin(iy) = 2 e-Y - eY 2i we also have that I sin(iy)I ➔ oo as y ➔ ±oo. We will see later that z r-+ cos z and z r-+ sin z are complex differentiable everywhere in the complex plane. Exercise 1.37. Show that cos z = ( cosx)(cosh y) - i(sinx)(sinh y) and for z = x+iy, where x, y are real, and cosh y I cos zl 2 = ( cosh y) 2 - (sinx)2 , eY-e-Y . := eY+e-Y and smh y := 2 2 Exercise 1.38. We know that the equation cosx = 3 has no real solution x. How­ ever, show that there are complex z that satisfy cosz = 3, and find them all. 1.4.3 Logarithm function In the real setting, given a positive y, logy E IR.is the unique real num ber such that e 10gy = y. Thus log : (0, oo) ➔IR.serves as the inverse of the function x r-+ e x : IR.➔ (0, oo ). See Figure 1.18. � e ---------------0---�-- logy � Fig. 1.18 0 The maps x ,-, e x : JR➔ (0, oo) and y ,-, logy: (0, oo) ➔ R A Friendly Approach to Complex Analysis 24 In the complex case, we know that exp : C--+ C\ {0}, and we now wonder if there is a "complex logarithm function" mapping C\ {0} to Cthat serves as an inverse to the complex exponential function. Given a z -/=- 0, we seek a complex number w such that exp w = z, and we would like to call this w the "complex logarithm of z". However, we have seen that the exponential function exp is 21r-periodic in the y-direction, and so the moment we find one w such that e w = z, we know that there are infinitely many others, since exp(w + 21rin) = exp w = z for all n E Z. Given this infinite choice, which w must we call the complex logarithm of z? We remedy this problem of nonuniqueness by just choosing a w that lies in a fixed particular horizontal strip of width 21r. Indeed, all possible nonzero complex numbers can be obtained as the exp of something lying in any such strip, and now for the purpose of defining the complex logarithm, we choose (somewhat arbitrarily), the strip§:= {z EC: -1r < Im(z) :S 1r}. See Figure 1.19. 7r -Jr Fig. 1.19 The strip § := JR x (-1r, 1r] is mapped by exp onto C \ {0}. This will give, as we shall show below, a unique w in the strip such that expw =:= z, and we will call this unique w the "principal logarithm of z", denoted by Log z. In order to do this, we will first introduce the notion of the principal argument of a nonzero complex number. Principal argument of a nonzero complex number. For z-/=- 0, let 0 be the unique real number in the interval (-1r, 1r] such that z = lzl(cos0+i sin0). This value of 0 is called the principal argument of z, and is denoted by Arg z. Here are a few examples: Arg(-1) = 1r, Arg(i) = i, . 7r Arg(-i) = --. 2 We note that if we start at a point on the positive real axis in the complex plane, and go around anticlockwise in a circle, then there is a sudden jump Arg(3) = 0, 25 Complex numbers in the value of the principal argument as we cross the negative real axis: on the negative real axis, the value of the principal argument is 1r, while just below the negative real axis, the principal argument is close to -1r. Exercise 1.39. Depict { z E (C: z =/ 0, � < IArg(z)I < i} in the complex plane. Now we are ready to define -the principal logarithm of nonzero complex numbers. Definition 1.3. The principal logarithm Log z of z -/- 0 is defined by Log z = log lzl +iArg z. First of all, let us observe that exp(Log z) =e 10glzl (cos(Arg z) +i sin(Arg z)) = lzl (cos(Arg z) +i sin(Arg z)) = z. This shows that exp :§ H C \ {0} is onto. Also, it is one-to-one, because if z1,z2 E § are such that expz1 = expz2, then exp(z1 -z2) = 1, and so z1 -z2 = 21rni, for some n E Z. But as z1,z2 E §, their imaginary parts must differ by a number< 21r. This implies that n = 0, and so z1 =z2. So the two maps exp :§➔ C \ {0} and Log : C \ {0}➔§ are inverses of each other. Of course, had we chosen to define the principal argument 0 of a nonzero z = lzle i0 to lie in a different interval (a, 21r+a] or [a, 21r+a) for some other a, we would have obtained a different well-defined notion of the logarithm (which would also be equally legitimate). But when we talk about the principal logarithm of z, in this book we will always mean log lzl +iArg z, with the principal argument Arg z E (-1r,1r]. Here is an example: Log(-i) = log I - ii+iArg(-i) = log1 - �2 i = 0 - �2 i = -�2 i. Continuity of Log in C \ (-oo, 0]. First of all, we remark that owing to the lack of continuity of Arg : C \ {0} ➔ (-1r,1r] across the negative real axis, also the function Log is not continuous on C \ {0}. It fails to be continuous at each point in (-oo, 0). Let us show the lack of continuity at -1. To this end, consider the sequence which converges to -1: = \ = lim exp (i(- 1r+�))= lim (-1) (cos�+i sin�)= -1(1 +i0) =-1. n➔ n n➔ n n 26 A Friendly Approach to Complex Analysis Also, Thus we have }11! Log (exp (i( - 1r + ¾))) = i(O- 1r) = -i1r Log (}11! exp (i( - 1r + ;) ) ) = Log(-1) = il-- i1r showing that Log is not continuous at -1 EC\ {O}. On the other hand, Log is continuous on the smaller set C\ (-oo, OJ. This just follows by observing that the principal argument Arg(z) is continuous in C\ (-oo, OJ. Indeed, the key thing is that if we take any complex number z0 not lying on (-oo, O], then there is some room around z0 not touching the negative real axis: we can always find a small enough r such that the disc D(z0 ,r) does not touch the line (-oo,OJ. Thus, given an E > 0, by shrinking r further if necessary, we can ensure that the points z in D ( zo,r) satisfy IArg(z) -Arg(zo)I < E. See Figure 1.20. . Arg � 7r Fig. 1.20 Arg(zo) -1r Continuity of the principal argument Arg in C \ ( -oo, O]. As both z f-t log lzl and Arg are continuous in C \ (-oo, OJ, it follows that Log is continuous there too. Using the continuity of Log on C\ (-oo, O], we will see later on that Log is complex differentiable in C\ ( -oo,O]. ab for a EC\ {O} and b EC. One can now also talk about a b , where a, b are complex numbers and a f=. 0, and we define the principal value of a b as a b := e bLog(a). Complex numbers 27 For example, the principal value of ii is exp(i • Log(i)) = exp(i(log Iii+iArg i)) = exp,(i (0+ii)) = e-i. Exercise 1.40. Find Log(l + i). Exercise 1.41. Find Log(-1) and Log(l). Show that Log(z2 ) isn't always equal to 2 • Log(z). Exercise 1.42. Find the image of the annulus { z E (C : 1 < izl < e} under the principal logarithm. Exercise 1.43. Find the principal value of (1 + i) 1 -i. 1.5 Notes The remark on the historical development of complex numbers is taken from [Needham (1997)]. Exercise 1.2 is taken from [Shastri (2000)]. Exercises 1.7, 1.7, 1.13, 1.19 are taken from [Needham (1997)]. Exercises 1.23 and 1.34 are taken from [Beck, Marchesi, Pixton, Sabalka (2008)]. Chapter 2 Complex differentiability In this chapter we will learn three main things: (1) The definition of complex differentiability, that is, given f : U -+ <C, where U is an open subset <C, and zo EU, we will learn the meaning of the statement "f is complex differentiable at z0 with complex derivative f'(zo)". . . 8u 8v 8u 8v (2) The Cauchy-Riemann equations: and By = - ox• ox = By These are PDEs that are satisfied by the real and imaginary parts u, v of a complex differentiable function f : U -+ <C wherever it is complex differentiable. (x,y) u � v(x,y) - - - f(x,y) ; 0 u(x,y) Fig. 2.1 The real and imaginary parts u, v off. Vice versa, we will learn that if these Cauchy-Riemann equations are satisfied everywhere in an open set and u, v are 0 1 , then f = u + iv is complex differentiable in U. (3) The geometric meaning of the complex derivative f'(z0): infinitesi­ mally the map f is an amplification by lf(zo)I together with a twist (a counterclockwise rotation) through Arg(f'(z0)). 29 30 A Friendly Approach to Complex Analysis The central result in this chapter is the necessity and (under mild condi­ tions) sufficiency of the Cauchy-Riemann equations for the complex differ­ entiability of a function in an open set. 2.1 Complex differentiability Definition 2.1. (1) Let Ube an open subset of (C, f: U --t (C and z0 E U. Then f is said to be complex differentiable at z0 if there exists a complex number L such that z - zo lim f( ) f( ) = L z - zo that is, for every E > 0, there is a J > 0 such that whenever z EU and 0 < Jz - zol < J, we have z---+zo I f(z) - f(zo)_ z-zo LI < . E We denote this L ( which can be shown to be unique) by (2) A f'(zo)or dz(zo). function f : U --t (C which is complex differentiable at all points of the open set U is called holomorphic 1 in U. (3) A function that is holomorphic in (C is called entire, that is, the domain of f is understood to be the whole of (C and moreover, f is holomorphic in (C. Let us look at a simple example of an entire function. Example 2.1. Consider the function f : (C --t (C defined by f(z) (z E q. We show that f is entire. Note that 2 - z5 f(z)- f(zo) z ---=z -zo z -zo = z + zo :::::: 2zo z2 for z near z0, and so we guess that f'(zo)= 2z0. Let us show this now. For z -/- zo, we have f(z)- f(zo) l _ _ _ _ _ _ _ _ - 2zol z - z0 = z2 - z l- _ _ _ 5 - 2zol = Jz + zo - 2zol = Jz - zol- z-zo 1The word "holomorphic" is derived from the Greek "holos" meaning "entire", and "morphe" meaning "form" or "appearance". Complex differentiability 31 So we see that the left hand side can be made as small as we please when z is close enough to z0. Let E > 0. Set 15 := E satisfies O < lz-zol < 15, we have I f(z)-f(zo) z-zo > 0. Then whenever z EC -2zol = lz-zol < 15 = E. Hence f'(zo) = 2z0. As z0 EC was arbitrary, f is holomorphic in C, that is, f is entire, and from the above it follows that d zz dz = 2z, z EC. ◊ On the other hand, here is an example of a natural mapping which is not complex differentiable. Example 2.2. Consider the function g : C ➔ C defined by g(z) = z (z E q. We show that g is differentiable nowhere. Suppose that g is differentiable at z0 EC. Let E := ½ > 0. Then there exists 15 > 0 such that whenever z satisfies O < lz- zol < 15, we have l g(z)-g(zo) z-zo -zo -g'( Zo) - -- -g'( Zo < E. z )I 1 1:Z-zo See the picture on the left of Figure 2.2. The above says that whenever z is in the punctured disc of radius 15 with center z0, we are guaranteed that this inequality holds. We will now make special choices of the z as the blue and the red point shown in the figure, and show that the special cases of the inequality above yield that g'(zo) must lie in discs of radius 1/2 with centers at -1 and 1. See the picture on the right of Figure 2.2. But these discs have no intersection, and this will be our contradiction. We give the details below. (9 • Takmg z Fig. 2.2 15 Non complex differentiability of complex conjugation = zo + 2, we have O < lz-zol < 15, and so 1: = :: - g'(zo)I =I:;� - g'(zo)I = II - g'(zo)I < E. (2. 1) 32 A Friendly Approach to Complex Analysis Also, taking z = zo + i�, we have 0 < lz - zol < J, and so l z-zo - g'(zo z _ zo )I = ,-iiJJ/22 -g'(zo )I = I 1 + g'(zo)I < 1 It follows from (2.1) and (2.2) that E. 2 = 11-g'(zo)+l+g'(zo)I::; 11-g'(zo)l+ll+g'(zo)I < E+E = 2E = 2-� a contradiction. So g is not differentiable at z0. (2.2) = 1, ◊ 2 Exercise 2.1. Show that f : (C ➔ (C defined by f(z) = lzl for z E C, is complex differentiable at 0 and that f' (0) = 0. We will see later (in Exercise 2.9) that f is not complex differentiable at any nonzero complex number. Exercise 2.2. Let D be a domain, and f : D ➔ (C be holomorphic in D. Set D* := {z E (C: z ED}, and define f*: D* ➔ (C by f*(z) = f(z) (z ED*). Prove that f* is holomorphic in D*. The following reformulation of complex differentiability is useful to prove el­ ementary facts about complex differentiation. Roughly speaking, the result says that for a complex differentiable function f with complex derivative L at z 0, f(z) -f(zo) - L · (z -zo) goes to 0 "faster than z - zo". Lemma 2.1. Let U be an open set in C, z 0 EU, and f : U ➔ C. Then the following are equivalent: (1) f is complex differentiable at zo with f'(zo) = L. (2) There exists an r > 0, and a function h : D(zo,r) ➔ C, where D(zo,r) := {z EC: lz - zol < r} , such that (a) f(z) = f(zo) + (L + h(z))(z -zo) for lz - zol < r, and (b) lim h(z) = 0. z➔zo Proof. (2)::::}(1): For z E D(z 0, r) \ {zo}, we have, upon rearranging, that f(z) - f(zo) Z - Zo _ L = h(z) z.2+° 0 and so f is complex differentiable at zo, and f'(zo) = L. (1)*(2): Now suppose that f is complex differentiable at zo. Then there is a J1 > 0 such that whenever 0 < lz - zol < J1, z EU and I f(z) - f(zo) Z - Zo _ f'(zo)I < 1. 33 Complex differentiability Set r := 81, and define h: D(zo, r) ---+ C by h(z)= { f(z) - f(zo) _ f'(zo) if Z-=/=- Zo, z-zo if z = zo, 0 Then f(z) = f(zo) + (f'(zo) + h(z))(z - zo) holds whenever lz - zol < r. If E > 0, then there is a 8 > 0 (which can be chosen smaller than r) such that whenever O < lz - zol < 8, we have I f(z) - f(zo) - f'(zo)I (= lh(z) - OI) < z - zo This completes the proof. E. □ For example, using this lemma, we see that holomorphic functions must be continuous. Exercise 2.3. Let D be a domain in C. Show that if f : D ➔ (C is complex differentiable at zo E D, then f is continuous at zo. Later on, we will see that if f is holomorphic in D, then in fact it is infinitely many times differentiable in D! Using Lemma 2.1, it is also easy to show the following. Proposition 2.1. Let U be an open subset of C. Let f, g : U ---+ C be complex differentiable functions at z 0 EU. Then: (1) f + g is complex differentiable at zo and (f + g)'(zo) = f'(zo) + g'(zo). (Here f + g: U---+ C is defined by (f + g)(z) = f(z) + g(z) for z EU.) (2) If a EC, then a· f is complex differentiable and (a· f)'(zo) = af'(zo). (Here a· f: U---+ C is defined by (a· f)(z) = af(z) for z EU.) (3) fg is complex differentiable at zo and moreover, there holds that (fg)'(zo) = f'(zo)g(zo) + f(zo)g'(zo). (Here Jg: U---+ C is defined by (fg)(z) = f(z)g(z) for z EU.) Remark 2.1. Let U be an open subset of C, and let Hol(U) denote the set of all holomorphic functions in U. Then it follows from the above that Hol(U) is a complex vector space with pointwise operations. On the other hand, the third statement above shows that the pointwise product of two holomorphic functions is again holomorphic, and so Hol(U) also has the structure of a ring with pointwise addition and multiplication. Example 2.3. It is easy to see that if f(z) := z (z Eq, then f'(z) = 1. Using the rule for complex differentiation of a pointwise product of two A Friendly Approach to Complex Analysis 34 holomorphic functions, it follows by induction that for all n E N, z f-t z n is entire, and d n Z dz = nz n -1 _ ◊ In particular all polynomials are entire. Exercise 2.4. Prove Proposition 2.1. Exercise 2.5. Let D = {z E (C : lzl < 1} and Hol(D) denote the complex vector space of all holomorphic functions in D with pointwise operations. Is Hol(D) finite dimensional? Exercise 2.6. Let U be an open subset of C, and let f : U ➔ (C be such that f (z) =I= 0 for z E U and f is holomorphic in U. Prove that the function J: U ➔ C, defined by is holomorphic, and that (y)' Exercise 2.7. Show that in (C \ (y) (z) = ftz for all z E U, ) (z) = - :��)�2 ( (z E U). {O}, for each m E Z, !zm = mzm -l Just like one has the chain rule for the derivative of the composition of maps in the real setting, there is an analogous chain rule in the context of composition of holomorphic functions. Proposition 2.2. (Chain rule) Let (1) (2) (3) (4) DJ, D9 be domains, f: DJ➔ C be holomorphic in DJ, g: D9➔ C be holomorphic in D9 , and f(DJ) C D9 • Then their compos ition go f : DJ ➔ C, defined by (go f)(z) = g(f(z)), z E Dt, is holomorphic in Dt and (gof)'(z) = g'(f(z))f'(z) for all z E Dt. f g Fig. 2.3 The chain rule: (go f)'(z) = g'(f(z))f'(z), z E Dt · Complex differentiability 35 Proof. Let z0 E DJ. Then f(z0) E D9. From the complex differentiabil­ ity off at z0 and that of g at f (z0), we know that there are functions h f and h9, defined in the discs D(z0, TJ) C DJ and D(f(z0), r9) C D9 such that f(z) - f(zo) = (f'(zo) + h1(z))(z - zo), g(w) - g(f(zo)) = (g'(f(zo)) + h9(w))(w - f(zo)), and lim h (w) = 0. lim h z = 0, z-+zo 1( ) w➔ f(zo) 9 But it follows from the continuity of f at z0 that when z is close to z0, w := f(z) is close to f(zo), and so if z-=/= zo, but close to zo, we have (g O f)(z) - (g O f)(zo) = (g'(f(zo)) + h9(f(z)))(f'(zo) + h1(z)), Z - Zo □ and so the claim follows. Example 2.4. From Exercise 2.7, we know that (�) = - 2 z EC \{0}, :, :z but this is also easy to see from the definition, because 1 1 Zo-Z -1 z-+zo 1 Z Zo = zzo(z - zo) zzo ----+ - z3 Z-Zo for zo EC \{0}. 1 f := 1 + z 2 g := z i .,,,.-------.._ -i DJ :=C \{-i,i} Fig. 2.4 D9 :=C \{0} Application of the chain rule. Now consider the functions f := l+z 2 on DJ :=C\{-i,i} and g := 1/ z on D9 :=C \{0}. Clearly, f (DJ) c D9, and so, by the Chain rule, 1 2z d 1 • 2z = ( ) =(1 + z2)2 (1 + z2)2 dz 1 + z2 inC \{-i,i}. ◊ 36 A Friendly Approach to Complex Analysis Exercise 2.8. Assuming that expz is entire and that exp' z this later), show that 1 +z) z r-+ exp (1-z is holomorphic in the unit disc I[)):= 2.2 {z E (C: [z[ < = expz (we will prove 1}, and find its derivative. Cauchy-Riemann equations We now prove the main result in this chapter, which says roughly that a = u + iv is holomorphic if and only if its real and imaginary u, v (viewed as real valued functions living in an open subset of IR.2 ) function f parts satisfy a pair of partial differential equations, called the Cauchy-Riemann equations. Cauchy-Riemann Let U be an open subset of <C, and let f : U ➔ <C be a function. Then taking any point (x,y) EU, we have f(x +iy) E <C, and we can look at the real part u(x,y) of f(x + iy), and the imaginary part v(x,y) of f(x + iy). See Figure 2.5. � v(x,y) - -,f(x, y) u(x,y) Fig. 2.5 The real and imaginary parts u, v off. 37 Complex differentiability If one changes the point (x,y), then f(x + iy) changes, and so do u(x,y) and v(x,y). In this manner, associated with f , we obtain two real-valued functions u: U---+ IR., U 3 (x,y) f-t Re(f(x + iy)) =: u(x,y), v: U---+ IR., U 3 (x,y) f-t Im(f(x + iy)) =: v(x,y). Our first result in this section is the necessity of the Cauchy-Riemann equations for complex differentiability, and we will prove this result in Theorem 2.1 below. The result says that if f is complex differentiable at (xo,Yo) E U, then OU OX av o y and OU o y av at (xo,Yo), OX and these two equations are called the Cauchy-Riemann equations. So for a function to be complex differentiable, it has got to satisfy these equations. In o,ther words, if these equations aren't satisfied by the real and imaginary part of some complex valued function at some point, then at that point we know that the function can't be complex differentiable. Here is an example. We had seen earlier (by "brute force", using the E-8 definition of complex differentiability), that z f---t z is not complex differentiable any where in the complex plane. Let us revisit Example 2.2. Example 2.5. For the function g: C---+ C defined by g(z) = z (z EC), we have that u(x,y) = Re(g(x + iy)) = Re(x - iy) = x, Thus v(x,y) = Im(g(x + iy)) = Im(x - iy) = -y. OU o (x,y) = 1 i- -1 x av =o (x,y). y This shows that the Cauchy-Riemann equations can't hold at any point. So we recover our previous observation that g is complex differentiable nowhere. ◊ Before proving the necessity of the Cauchy-Riemann equations for complex differentiability, let us also mention the second important result we will show in this section, namely the sufficiency of the Cauchy-Riemann equa­ tions for holomorphicity in an open set. More precisely, we will show in 38 A Friendly Approach to Complex Analysis ThE:Jorem 2.2 below that if u,v : U ➔ IR are two real-valued functions in an open set U that are continuously differentiable in U (as functions of two real variables), and the Cauchy-Riemann equations are satisfied by u and v everywhere in U, then the new complex-valued function f: U ➔ C, created from u,v by setting f(x+iy) := u(x,y)+iv(x,y), (x,y) EU, (so that u,v turn out to be the real and imaginary parts, respectively, of the f we have constructed), is holomorphic in U. This important result will enable us to establish the holomorphicity of important functions without having to go through the rigmarole of verifying the E-8 definition. Here is an example. Example 2.6. Consider the functions u,v defined in the punctured plane IR2 \ {(0, 0)} as follows: X u(x,y) := 2 v(x,y):= 2-y 2' (x, y) =/- (0, 0). X +y 2' X +y Then we have y 2 -x 2 1 • (x 2 + y 2 ) - x • 2 x &u 2 x 2 + y 2 x ) &x ( ( 2 + y2 )2 ' &u -x-2y -2xy 2 y 2 x 2 x 2 &y ( + ) ( + y 2 )2 ' y2 - x2 &v 2xy &v and . = &x (X 2 +y 2 )2 (x 2 + y 2 ) 2 ' y 8 Clearly (x,y) M (x 2 +y 2 ) 2 , y 2 -x 2 , ±2xy are continuous in IR2 and (x 2 +y 2 ) 2 is nonzero in IR2 \ {(0, 0)}. So it follows that each of these partial derivatives is continuous in IR2 \ {(0, 0)}. So u,v are continuously differentiable in IR2 \ {(0, 0)}. Also the Cauchy-Riemann equations hold. Thus f := u + iv is holomorphic in (C \ {0}. In fact, the f above is just the reciprocation map z M 1/ z: X -y X iy z z I z =/- 0. f = u + iv = x 2 y 2 + i x2 y 2 ) = x2 + y 2 = = = + + zz -;; ' • •( - W ◊ Theorem 2.1. Let U be an open subset of (C and let f: U ➔ (C be complex differentiable at zo = xo + iyo E U. Then the functions (x,y) M u(x,y) := Re( f(x + iy)) : U ➔ IR and (x,y) M v(x,y) := Im(f(x + iy)) : U ➔ IR are differentiable at (xo, Yo) and &u &v (xo, Yo) = (xo, Yo) and &x &y (2.3) 39 Complex differentiability Proof. (The idea of the proof is easy, we just let (x,y ) tend to (xo,Yo) by first keeping y fixed at y0, and then by keeping x fixed at x 0, and look at what this gives us. See Figure 2.6.) (xo,Y) (xo,Yo) (x,yo) Fig. 2.6 Proof of the necessity of the Cauchy-Riemann (CR) equations for complex differentiability. Let z0 = (x0, y0) EU. Let E > 0. Then there is /j > 0 such that whenever 0 < [z - z0[ < 8, we have z EU and I f(z ) - f(zo) Z - Zo - J'(zo)I < E. (2.4) Step 1. We will show that �: (xo,Yo) exists and equals Re(f'(z0)). Let z := x + iy0, where x E � is such that O < [x - x0[ < 8. Then z - zo = x - xo, and so O < [z - zo[ = [x - xo[ < 8. Thus l u(x,yo) - u(xo,Yo) _ Re(f'(zo))I x-xo f(x + iyo = I Re ( l z = = ;� xo + iyo) ) - Re(f'(zo))I = IRe ( f( ; �?o)) - Re(f'(zo))I ::::; I f(z ) - f(zo) - ' z f ( o)I < E, z-zo using (2.4). Thus the partial derivative ou . u(x,Yo) - u(xo,Yo) - (xo,Yo) = 11m OX x➔xa = Re(f'(zo)). X - Xo 40 A Friendly Approach to Complex Analysis Step 2. We show that i:(xo,Yo) =Im(f'(zo)). Proceeding in a manner similar to Step 1 above, we also have with the same notation that v(x,yo)-v(xo,Yo) _ Im(f'(zo))I l X-Xo (x + iyo xo + iyo) = IIm ( f ; �� ) - Im(f'(zo))I (z = IIm ( f ; :s; I = o) :? ) - Im(f'(zo))I f(z) - f(zo) Z - Zo = - f'(zo)I < E, ov . v(x,yo)-v(xo,Yo) =Im(!'(zo )). Thus and so -;:,- (Xo,Yo ) = 11m uX x----+xo X - Xo .av OU I f (zo) = ox (xo,Yo)+ i o (xo,Yo)x (2.5) Step 3. We show that :: (xo,Yo) = -Im(f'(zo)). Now let z := xo + iy, where y E JR is such that O < IY - Yol < 8. Then z - zo = i(y - Yo) and so O < lz - zol = IY - Yol < 8. Thus using the fact that for a, b E JR there holds that Re( a+ ib) =Im(i(a+ ib)), we obtain that Im( (f u(xo, y) - u(xo,Yo) + Im(!'(zo))I = I i (z) - f(zo)) + Im(f'(zo))I I y-� y-� (z = Im (-f ; :?o) + f'(zo)) :s; I If = (z) - f(zo) Z - Zo J'(zo) I < E. I Thus the partial derivative ou _ . u(xo,y) - u(xo,Yo) _ -Im(!'( )) . zo -;:,- (xo,Yo ) - 11m y----+yo uy y-yo Recall that in Step 2, we had obtained i:(xo,Yo) =Im(f'(zo)), and so, we have established one of the two Cauchy-Riemann equations, namely that Complex differentiability 41 Step 4. We show that ::(xo,Yo) = Re(f'(zo)). Proceeding as in Step 3 above, with the same notation , and using the fact that for a, b E JR there holds that Im(a+ib) = -Re(i(a+ib)), we have that I v(xo,y)- v(xo,Yo) R ! (z) - f(zo) - e( '(zo))I = I-Re (J ) - Re(f'(zo) I y-yo y-yo z) :S 1-J( - f(zo) - f'(zo)I y-yo f(z)- f(zo) = I - f'(zo)I < E. Thus the partial derivative Hence Z- Zo av _ Re(!'(zo )). _ 1.1m v(xo,y) - v(xo,Yo) � (xo,Yo ) uy Y-+Yo y-yo J'(zo) = :: (xo,Yo) - i :: (xo , Yo). (2.6) From (2.5) and (2.6), it follows that av av au au = (xo,Yo) = - (xo,Yo)(xo,Yo) and ax ax (xo,Yo) ay ay So we have got both Cauchy-Riemann equations. Finally, we show that u, v are differentiable (as real valued functions of two real variables) at (xo,Yo). For z = (x,y) satisfying O < l z - zol < 8, lu(x,y)-u(xo,Yo)-(�(xo,Yo)) (x-xo)-(�(xo,Yo)) (y-yo)I ll(x,y)- (xo,Yo)ll2 _ - lu(x,y) - u(xo,Yo)- (�(xo, Yo)) (x - xo) + (�(xo, Yo)) (y -Yo)I ll(x,y)- (xo,Yo)ll2 IRe(f(z) - f(zo) - f'(zo)(z - zo))I =���-��--��-�<€ lz- zol Thus u is differentiable at (x0, Yo). Similarly, v is also differentiable at (xo,Yo). □ Remark 2.2. We will see later on that in fact the real and imaginary parts of a holomorphic function are infinitely many times differentiable. 42 A Friendly Approach to Complex Analysis Exercise 2.9. Consider Exercise 2.1 again. Show that J is not differentiable at any point of the open set (C \ {0}. One also has the following converse to Theorem 2.1. This is a very useful result to check the holomorphicity of functions. Theorem 2.2. Let (1) U be an open subset of C, (2) u,v : U ➔ IR. be continuously differentiable, and (3) u,v satisfy the Cauchy-Riemann equations: for all (x,y) EU, au av au av (x,y) = (x,y) and (x,y) = - (x,y). ay ax ax ay Then f := u + iv : U ➔ C is holomorphic in U and J' (X + iy) = ;: (X' y) + i ;: (X' y) for X + iy E u. Proof. Let z0 = x 0 + iy0 E U. Let E > 0. Let o > 0 be such that whenever z = x +iy belongs to the disc D(z0, o) := { w EC: lw- z0 1 < o}, we have z EU , l au au (x,y )- (xo,Yo) I ax ax av av < E, and 1ax(x,y )- ax(xo,Yo) I < E. (2.7) (That this is possible follows from the fact that u,v are continuously dif­ ferentiable.) Let z = x + iy be a fixed point in the punctured disc D(z0, o) \ { z0}, and consider the part of the line in D (z0, o) joining z0 to z. A point on this line can be represented by p(t) = (1-t)zo + tz = ((1- t)xo + tx, (1- t)yo + ty )­ In particular, p(O) = z0 and p(l) = z. See Figure 2.7. z = (x,y) p(t) zo = (xo,Yo) Fig. 2.7 = (1 - t)z0 + tz = ((1 - t)xo + tx, (1 - t)yo + ty) Points p(t), on the line joining zo to z. 43 Complex differentiability Define r.p1,r.p2 : (-1, 1) ➔ JR.by [r.p1(t)] · = u(p(t)) ] [ v(p(t)) r.p2(t) · Then using the chain rule we obtain · - xo)+ �; (p(t)) · (y - Yo) ] · - xo)+ -y (aavp(t)) · (y - Yo) �: (p(t)) · (x ['Pi(t)] ·[ . av 'P2I (t) -a(p(t)) (x x Let 0 u a av (p(t)) = (p(t)), ya xa au av B(t) := - (p(t)) = (p(t)), ya xa where we have used the Cauchy-Riemann equations to obtain the rightmost equalities. Thus with this notation, we have ['Pi(t)]= A(t)(x-xo) B(t)(y-yo)]=[Re((A(t)+iB(t))(z zo)) [ ] r.p�(t) B(t)(x x0) + A(t)(y-y0) lm((A(t) + iB(t))(z zo)) A(t) := - - So f(z) - · - f(zo) = u(x,y) - u(xo,Yo)+i(v(x,y) - v(xo,Yo)) = r.p1(l) - r.p1(0) + i(r.p2(l) - r.p2(0)) = = Thus 11 (11 11 1 1 + 'Pi(t)dt+i A(t)dt r.p;(t)dt i · - zo)­ B(t)dt) (z u a( ) . av( )) - f(zo) - ( -a Xo,Yo + i-a Xo,Yo u a u a ) . [ av ((p(t)) (p(O)) dt+ l ((p(t)) - f(z) Z-Zo X X ) 1 av [ (p(O)) dt. i l = o a a a a x x x x o By (2.7), it follows that v a u a f(z) f(zo) ((xo,Yo)+i(xo,Yo) <E+€=2€. a a z zo x x This holds for all z satisfying O < lz - zol < 8, and so f is complex differ­ entiable at z0 and 1 I - - . ) I f'(zo) = �: (xo, Yo)+i �: (xo,Yo). This completes the proof. □ 44 A Friendly Approach to Complex Analysis Let us revisit Example 2 .1 again, but now instead of using the E-8 def­ inition of complex differentiability, we will use the above result to check holomorphicity of the squaring map. Example 2. 7. For the function f : (C -+ C defined by f(z) =z2 (z E q we have that u(x,y)= Re(f(x+ iy))=Re(x 2 -y v(x,y)=Im(f(x + iy))=Im(x 2 -y 2 + 2xyi)=x 2 - y2, 2 + 2xyi)=2xy. Thus au 2 x= (x,y) ax au (x,y) = -2y = ay which shows that the Cauchy-Riemann equations hold in C. So we recover our previous observation that f is entire, and since J'(z)= �:(x,y)+i::(x,y)=2 x +2yi =2z, ◊ we also obtain that f'(z)=2z for z EC. Example 2.8. (exp, sin, cos are entire.) For the function g: (C-+ C defined by g(z)=expz (z E q we have that u(x,y)= Re(g(x+ iy))=Re(e x (cosy+i siny))=e x cosy, Thus v(x,y)= Im(g(x+ iy))=Im(e x (cosy+i sin y))=e x sin y. au av e "' cosy= (x,y), (x,y) ay ax au av x (x,y) = -e siny = - (x, y), ax ay which shows that the Cauchy-Riemann equations hold in C. So we arrive at the important result that exp is entire, and since "' x g'(z)= �:(x,y)+i;:(x,y) =e cosy+ie sin y=expz, we also obtain that : exp z=exp z for z EC. Hence from Proposition 2 .2 , z also the trigonometric functions exp(iz) - exp(-iz) . smz= 2i and cosz= exp(iz)+exp(-iz) 2 Complex differentiability 45 are entire functions, and moreover, d l. S nz � p( iz)- (-i) exp(-iz) -----iz) i ex-------- = exp( iz) +exp(-= cosz ' and =- d = - COS Z � � exp(-iz) i exp( iz) +(-i) --------2 = 2 exp( iz)- exp(-iz) - ------� = - Sl. n Z . ◊ Example 2.9. (Holomorphicity of Log.) We will show that the principal logarithm is holomorphic in the open set C\ (-oo, O]. Note that the principal logarithm is defined in the bigger set (C \ { 0}, but we had seen earlier that it is not continuous in this bigger set ( because at each negative real number, it is discontinuous). We had also seen that in the smaller set (C \ (-oo, O], the principal logarithm is continuous. We will now use this continuity to show that Log is in fact holomorphic in (C \ (-oo, O], and that d l - Log(z) = - for z E (C \ (-oo, O]. dz z First note that when z , zo EC\ (-oo, O] are distinct, then Log(z) -/- Log(zo)­ (Why?) Let E > 0. Set Since expw is differentiable at wo := Log(zo), there is a 81 > 0 such that whenever w satisfies O < lw -wol = lw- Log(zo)I < 81, there holds that I I 1-1 expw-zo expw - exp Wo -zo < - expwo w- Log(zo) w-wo E1- But by the continuity and injectivity of Log in (C \ (-oo, O], there exists a o > 0 such that whenever O < lz -zol < o, we have 0 < I Log z - Log zo I < 81. Thus with w := Log z, and O < lz -zol < o, we have O < lw -wol < 81, and so I I z -zo ----- --zo < Log z - Log zo E1. I 46 But then I we have A Friendly Approach to Complex Analysis I z - zo lzol whenever 0 <lz-zol <J, �lzol-E1 �-.So Log z - Log zo 2 Log z-Log z0 z-z0 1 - zo I ( I = zo - z-zo Log z-Log z0 ) = zo Log z-zo z-Log z0 1 1 <El· -- • lzol/2 lzol I · 1 1 z -z o · Log z - Log zo 1 z-zo Log z-Log zo 2E1 = --2 <E. 1 Zo 1 ·� 1 lzol Thus Log is holomorphic in (C \ (-oo, 0] and moreover, d Log z = �- ◊ z dz We now consider an example illustrating the fact that the assumption of differentiability of u, v in Theorem 2.2 (as opposed to mere existence of partial derivatives of u, v satisfying the Cauchy-Riemann equations), is not superfluous. Example 2.10. Consider the function f: (C--+ (C given by xy(x+iy) . f( X +iy )X 2 +y 2 if x + iy-/- 0, and f (0) = 0. We have that for nonzero (x, y) E IR2, x2 y u(x, y) = Re(f(x + iy)) = 2 X +y 2, xy2 v(x, y) = Im(f(x + iy)) = 2 X +y 2, and u(0, 0) = v(0, 0) = 0. Thus av au au av (0, 0) = 0 = - (0, 0), (0, 0), and (0, 0) = 0 = ax ay ay ax which shows that the Cauchy-Riemann equations hold at the point (0, 0). However, the function is not complex differentiable at 0, since if it were, we would have au .av I f (0) = ax (0, 0) + i ax (0, 0) = 0 + iO = 0, and for E = 1/4, there would exist a corresponding J such that whenever 0 <lz- 0I = Ix+ iyl <J, we would have xy f(z) - f(O) = 2 2 <E, x +y z-0 I j'(O)I I I 47 Complex differentiability but taking x + iy= 2 + i 2, we arrive at the contradiction that J J 1 I I I xy < f. = 4· x 2 + y2 This shows that f is not complex differentiable at 0. We note that there is no contradiction to Theorem 2.2, since for example u is not differentiable at (0,0). It it were, its derivative at (0,0) would have to be the linear transformation 2= [:] � [ :: (0,0) :� (0,0)] [:] = [ 0 0] [:] = 0. But then with f. := 1/3 > 0, there must exist a J > 0 such that whenever 0 < ll(x,y)-(0,0)11 2 < J, we would have 1 x2 lu(x,y)-u(0,0)-0((x,y)-(0,0))1 -------------= 2 y 2 3 < f.=3 ll(x,y)-(0,0)11 2 (x +y )2 . Then with (x, y) = ( �'�), we have ll(x,y) -(0,0)11 2= � < J, and so J2 J 4 · 2 -=x y 1 1 ----��� <E=-1 == ' 2 2 2 (x +y ) ! J ! (�+ :) J8 a contradiction. So u is not differentiable at (0,0). 3 v'§ ◊ The Cauchy-Riemann equations can also be used to prove some interest­ ing facts, for example the following one, which highlights the "rigidity" of holomorphic functions alluded to earlier. See also Exercise 2.12 below. Example 2.11. (Holomorphic function with constant modulus on a disc is a constant.) Consider the disc D= {z EC: lz-zol < r}. We will show using the Cauchy-Riemann equations that if f : D -+ C is holomorphic in D, with the property that there is a c E JR such that IJ(z)I= c for all z E D, then f is constant. (We will use this fact in later when we learn about a result called the "Maximum Modulus Theorem".) See the picture below. f ,-------------. 48 A Friendly Approach to Complex Analysis Let u, v denote the real and imaginary parts of f. By assumption, we have c2= 1/1 2= u2 + v2, and so by differentiation, au u ax v + ay v + au u va ax va ay = o, = 0. va u. . va . = - a Ill thefirst equatIOn and = Vsmg ax ay ay au x a . . Ill the second equat10n, au au u- -v- =0, ax au u To remove To remove au ay au ax a y (2.8) a y + v au ax (2.9) = 0. au . , multiply (2.8) by u, (2.9) by v, and add: (u 2 + v 2) = 0. ax au . , multiply (2.8by ) -v, (2.9by ) u, and add: (u 2 + v 2) = 0. ay If c= 0 so that u + v = c2= 0, then u= v= 0 and so f= 0 in D. If C-/=- 0, then 2 2 au ax = au ay va va . . = = 0. = 0, and the CR equat10ns give also ax By the Fundamental Theorem of Integral Calculus, it follows that x ay 1 u (�, Yo)d�= 0, Y u ) - u( , Yo)= 1 � ( , rt)drt= 0. y u(x, Yo) - u(xo, Yo)= u(x, y xo a aX x Yo x So the value of u at any (x, y) is the same as the value of u at zo= (xo, Yo). Thus u is constant in D. (x y) (xo, Yo ) J (x, Yo) Similarly, v is constant in D. Consequently, f= u + iv is constant in D. ◊ Exercise 2.10. Show that z i---+ z3 is entire using the Cauchy-Riemann equations. 49 Complex differentiability Exercise 2.11. Show that z I-+ Re(z) is complex differentiable nowhere. Exercise 2.12. Let DC (C be a domain. Show, using the Cauchy-Riemann equa­ tions, that if f : D ➔ (C is holomorphic in D, with the property that f(z) E R for all z E D, then f is constant in D. Exercise 2.13. Let D C (C be a domain. Show that if f : D ➔ (C is holomorphic in D, with the property that J' (z) = 0 for all z E D, then f is constant in D. Exercise 2.14. Suppose that f : (C ➔ (C is an entire function, and let u := Re(!), v := Im(!). Suppose, moreover, that there exists a differentiable h: R ➔ R such that u =ho v. Prove that f must be a constant. Exercise 2.15. Let k ER be a fixed, and let f be defined by f(z) = (x 2 -y 2 )+kxyi for z = x + iy, x, y ER. Show that f is entire if and only if k = 2. 2.3 Geometric meaning of the complex derivative In ordinary calculus we have learnt the geometric meaning of the derivative of a real valued function f:�➔�at a point xo E�: f'(x0) is the slope of the tangent to the graph off at x0• See Figure 2.8. f slope= f'(xa)-------..._ Xo Fig. 2.8 Meaning off' (xo)lim :z:-+:z:o J(x) - f(xo) J(x) - J(xo) =f'(x0) implies R::-j'(xo) for x near xo, i.e. X - Xo X - Xo f(x) - J(xo) R::- J'(xo)(x - xo)- This means that locally, around x0, J(x) - J(x0) looks like the action of the linear map h rt f'(x0 )h: �➔�on x - Xo- Visually this means that near x 0, there is very little difference between the (tangent) line with slope f'(x0) passing through (x0, f(x0)) and the graph off. That is, if we zoom into the graph of the function around the point (x0, f(x0)), then the graph looks like a straight line. 50 A Friendly Approach to Complex Analysis In this section, we pose an analogous question for a complex valued function map f : U ➔ C defined on an open set U, that happens to be complex differentiable at a point z0: What is the geometric meaning of the complex number f'(z0)? We can't draw a graph off, because z as well as f(z) belong to C = �2, and so (z, f(z)) would be a point in �2 x �2 = �4 ! But we can draw a copy of U in the plane on the left hand side, and on the right hand side, we can imagine a copy of C, with f mapping points from U on the left to points on the right, as shown below. f � C ·f (z) We will now show that the complex number f'(z0) describes the action of the complex differentiable function locally infinitesimally around z0 by a anticlockwise rotation through the angle Arg(f'(zo)) together with a scal­ ing/ magnification by lf'(zo)lf(z) - f(zo) = (z) f(zo) �f'(zo) for z near zo, i.e. f'(z0), implies f lim z➔zo Z - Zo Z - Zo f(z) - f(zo) � f'(zo)(z - zo). But from the geometric meaning of complex multiplication, we know that when we multiply z-zo by f'(z0), z- zo gets rotated anticlockwise through the angle Arg(f'(zo)), and the length of z - z0 gets multiplied by the length of f'(z0), namely we get a magnification by the factor lf'(zo)I. In order to further understand this geometrically, look at Figure 2.9. z zo Fig. 2.9 f � f(zo) Geometric local meaning of the complex derivative. Complex differentiability v'3 51 Suppose that f'(zo) = + i, so that lf'(zo)I = 2 and Arg(f'(zo)) = ?T/6. First look at z - z0 shown in the domain U as the solid line segment between z and z0. On the right hand side, we have shown a translated version of this line segment as a dashed line, emanating from f (z0). In order to find out where f(z) is, we just use the fact that f(z) - f(z0) is approximately equal to f'(zo) multiplied by z - z0• So the solid line denoting f(z) f(z0) on the right hand side is obtained by rotating the rightmost dashed line anticlockwise through an angle of Arg(f'(zo)) ( assumed to be 30 ° in this picture), and magnifying the length of the dashed line by lf'(zo)I = 2. Note that if want to find the image of another point z which is near z0, we have to repeat the same procedure. Namely we first look at the line segment joining z0 to z, which is the solid line on the left. We have shown a translated version of this as a dashed line, emanating from f (z0), in the picture on right hand side. In order to find the position of f(z), we first rotate the leftmost dashed line anticlockwise through an angle of f' ( z0), that is 30 ° , and magnifying the length of the leftmost dashed line by lf'(zo)I = 2. In this manner, we obtain the solid line on the right hand side representing f(z) - f(z0). Placing one end at f(z0), the other end is the point f(z) (approximately!). So we see that locally the action off is as follows. Imagine the domain as a rubber sheet, and look at a point z0 on this rubber sheet. Tear out a small portion of this rubber sheet around z0. The function f takes the point z0 on this rubber sheet to a point f(zo) somewhere in the complex plane. If we want to know how the rest of the points on our little torn rubber sheet are mapped by f, one follows this procedure. We place our rubber sheet such that zo on our rubber sheet is lying over the point f(zo) in the complex plane. (Imagine pinning it on the plane with the pin passing through the point marked z0 on our little torn rubber sheet.) Then we stretch out our rubber sheet about the point z0 by a factor of lf'(zo)I, and then rotate this stretched rubber sheet anticlockwise by an angle of Arg(f'(z0)) around the point z0 on the rubber sheet. In order to stress this geometric interpretation, let us revisit Example 2.1 yet again, where we considered the squaring map z i-+ z 2. See Figure 2.10 and Example 2.12. 52 A Friendly Approach to Complex Analysis . 2,5 ; ' 1 : ; . 2·····L ....l±J.------1 ..... � LL_I , !_J.. L I l' : .. J! ... � .. � ...... ) .. l Q.8 Q.6 o.4 0.2- 0 Q,2 Q,4 0,6. 0,8 -1 -1 Fig. 2.10 -(),5 0 �.5 1 The image of a square and a square grid under the mapping z 2 • Example 2.12. Suppose that we assume complex differentiability of the squaring map f, z H z 2 , at a point z0 E C. Let us then try to prove that the complex derivative of the squaring map at z0 must be 2z0 by figuring out geometrically the local amplification at z0 and the local rotation at z0 produced by the squaring map. We first ask: what is the local rotation produced? In order to find this out, consider a point z close to z0 along the ray joining z0 to z. Look at Figure 2.11, which shows the effect of the squaring map: the angle is doubled, and the distance to O is squared. Thus z 2 lies on the ray joining 0 to z5, which makes an angle 2Arg(z0) with the positive real axis. Hence the line segment joining z5 to z 2 is obtained by rotating the line segment joining z0 to z anticlockwise through the angle Arg(z0). Consequently, Arg(f'(zo)) = Arg(zo)- 53 Complex differentiability Z H z2 z �o) 0 Fig. 2.11 Calculation of the amount of local rotation produced by the squaring map. Next we ask: what is the local amplification produced? In order to find this out, consider a point z close to z0 which is at the same distance from O as z0, but it makes an slightly bigger angle 0 + d0 with the positive real axis. Look at Figure 2.12, which shows the effect of the squaring map. Since d0 is tiny, the length lz-zol is approximately lzol ·d0, while the length lz 2 - z5I is approximately lzol 2 • 2d0. Consequently the magnification factor lf'(zo)I must be equal to (lzol 2 · 2d0)/(lzol · d0) = 2lzol- Z H z2 z 0 Fig. 2.12 0 Finding the amount of local magnification produced by the squaring map. A Friendly Approach to Complex Analysis 54 Putting it all together, we have J'(zo) = lf'(zo)I · ( cos(Arg(J'(zo))) +isin(Arg(J'(zo)))) = 2lzol · ( cos(Arg(zo)) +isin(Arg(zo))) = 2z0 .. So by investigating the local behaviour of the squaring map f around the point z0, we could find out the complex derivative f'(z0). ◊ Example 2.13. (Complex conjugation is complex differentiable nowhere.) See Figure 2.13. w w Fig. 2.13 z t-+ z is not holomorphic. Suppose that z r-+ z is complex differentiable at z0. Then the local be­ haviour of the map around zo should be a rotation followed by an am­ plification. Consider the point z near z0 obtained by a tiny horizontal translation. From the picture, by looking at the images z0 and z, we see that since z - z0 = z - z0, no rotation is produced. On the other hand, if we look at w which is near z0 obtained by a tiny vertical displacement, then from the picture, we see that w - z0 = -(z - z0), and so there is a rotation through 180 ° . But this means that locally the map is not a rota­ tion (because if it were, all infinitesimal vectors emanating at z0 would be rotated by f by a same fixed amount!). ◊ Exercise 2.16. We know that the power function z r-+ z n , n EN, is entire. Find its complex derivative by investigating its local behaviour. Hint: Since the rotation and magnification is the same for all tiny arrows ema­ nating from zo, one can simply the argument given in Example 2.12 by looking at what happens to a tiny vector perpendicular to the ray joining Oto zo. 55 Complex differentiability Exercise 2.17. We know that the exponential function z >---+ e z is entire. Find its complex derivative by investigating its local behaviour. Hint: Take a point zo, and move it vertically up through a distance of o. Look at the image to determine the amplification. Similarly, by moving zo horizontally through o, determine the amount of rotation produced locally. Exercise 2.18. Give a visual argument to show that the map z >---+ Re(z) is not complex differentiable any where in <C. Conformality. Look at Figure 1.16 on page 20 again, which shows the action of the entire mapping exp. In this picture, we see that just like in the domain, also the images of the vertical and horizontal lines are mutually perpendicular. We had mentioned that this is the property of conformality, that is, of the preservation of angles between curves in the domain together with their "orientation" is something which is possessed by all complex differentiable functions in domains. Moreover, we had mentioned that this property is possessed by every holomorphic function in its domain. Let us now see why holomorphic functions possess this property, based on what we have learnt about the local action of complex differentiable functions. f f(p) Fig. 2.14 Let f : U ---+ Conformality of a holomorphic (C be holomorphic in ing at a point p E U. U. f. Here cp = Arg(f'(p)). Imagine two smooth curves intersect­ Since the curves are smooth, they have tangents at p, say T1 and T2. Look at Figure 2.14. Near the point p, there is very little difference between the curve and its tangent line at p, so we may assume that the curves are replaced by their tangent lines. These tangent lines make a certain angle. Now let us look at what f does to these lines. Each of the curves is mapped to new curves in (C by f, and these intersect at f of p, that is f(p). They are smooth too, and they possess new tangent lines. But since the local action of f around p is rotation clockwise by Arg(f' (p)) followed by magnification, the new tangent lines are obtained by just rotating clockwise the old tangent lines, and magnifying the image. So it is obvious that the angle will be the same, and so will the orientation. So the conformality of holomorphic maps isn't a mystery anymore! 56 A Friendly Approach to Complex Analysis Relation to real differentiability, and the Cauchy-Riemann equa­ tions revisited. Consider a map f : U -+ C which is complex differentiable at z0 = (x 0, y0) in the open set U. If u, v denote the real and imaginary part of the function f, then we know that u, v : U -+ IR are differentiable at (xo,Yo)- Thus f, that is, the map (x,y) H (u(x,y),v(x,y)): U-+ IR2 , is differentiable (in the real variable sense), and we know that its derivative is the linear transformation l �:(xo,Yo) �� (xo,Yo) ·- [ A. av av (xo,Yo) (xo,Yo) ay ax which describes the local action. But since f is complex differentiable, we know that its local action is a anticlockwise rotation by 0 := Arg(f'(zo)), followed by a magnification by r := lf'(zo)I, and so it is the linear trans­ formation described by the matrix cos0 -sin0 r[ sm · 0 cos 0] · Since A must equal this, we obtain that au av ( Xo,Yo) = r cos0 = ( Xo,Yo), ax ay au av (Xo,Yo) = r sin0 = - ay ( Xo,Yo). ax Moreover, f'(zo) = r(cos0+isin0) = �:(xo,Yo) +i��(xo,Yo)- Summarizing, if f is complex differentiable at a point zo, then it is real differentiable (as a mapping from U c IR2 to IR2 ), but what distin­ guishes complex differentiability from mere real differentiability is that the real derivative for a complex differentiable mapping is not just any old lin­ ear transformation, but a very special one: it is an anticlockwise rotation through an angle 0 followed by a scaling by r. 2.4 The d-bar operator The two Cauchy-Riemann equations can be written as a single equation by introducing what is called the "d-bar operator" {) az· 57 Complex differentiability Let us define the differential operators � := ! az 2 ( � � -i� := ) and ax ! az ay 2 ( � +i � ax ay ). By a differential operator, we mean something which can act on functions and produce new functions. For example, the above two differential opera­ tors can act on smooth functions cp , : U -+ JR, where U is an open subset of JR2 , and by the smoothness of cp we mean that it possesses at least its first partial derivatives with respect to x and y everywhere in U. Then acp := ! az 2 Also, for u, v ( acpx a /'P) ay and acp ! := az 2 a a ( acpx +i acpy ). smooth real-valued functions in U, we define . . a a av (u+iv) := aauz +i. az (u+iv) := aauz+i. aavz and az . az With this notation, we have for a holomorphic f = UC <C that a az u + iv in the open set a . =au . av +i­ -(u+iv) az az az .1 1 av . av au . au =- (-+i-) +i- (-+i- ) 2 ax ay 2 ax ay = ! ( au_ av) +i ( au+ av) ax 2 a 2 ax a -f = ! y y = O+iO = 0, using the Cauchy-Riemann equations for the real and imaginary parts off. Also, we have �f fu = �(u+iv)= au+i av fu fu fu av _ av ( a x i ay ) = ! ( au+ av) + i (- au+ av) 2 ax ay 2 ay ax 1 au .1 av au . av = - · 2- + i- · 2- = - + i2 ax 2 ax ax ax au au ( _ i ay ) +i!2 2 ax = ! ! = J'. Summarizing, for f holomorphic in U that : f = 0, z and :zf = J'. u, v 58 A F'riendly Approach to Complex Analysis So philosophically, we ought to think of holomorphic functions as "func­ tions of z, z which are independent of z". Example 2.14. z is not holomorphic because 8 _ {)( ") OZZ = OZ X - zy = 1 2 8 OX+z ay 1 . 1 . = - -z · - • z =1 =/- 2 . (8 ) . . 1 (8 X- 2 2 8 OX+z Oy ) y o. 2 ◊ Example 2.15. lzl = zz is not holomorphic in (C \ {O} because 2 8 - 8 1 --=(zz) = --=(x +y ) =2 oz 2 oz 2 . (8 8 -+zax ay ) (x +y ) 2 2 =½(2x+i2y) =x+iy =z =f. 0 in (C \ {O}. ◊ Example 2.16. z 2 is entire because ( !(z 2 ) = :z 1 =2 ( x2 - 8 2 y +2xyi) . 8 -+zax ay )( 2 2) .1 ( X -y +z2 8 . 8 )( 8 . 8 ) -+zax ay ) 2xy =½(2x -i2y)+i½(2y+i2x) =0. Moreover, :z (z 2 ) 2 2 = !(x - y +2xyi) 1 =2 1 ( 8 . 8 --iax ay ) 2 2) .1 ( (x -y +z2 1 --iax ay (2xy) = (2x+i2y)+i (2y -i2x) =2(x+iy) =2z. 2 Exercise 2.19. Show that 4 2.5 2 !! = .6., where .6. := ::2 ◊ + ::2 is the Laplacian. Notes The section on the geometric meaning of the complex derivative follows closely the exposition in [Needham (1997)]. Exercises 2.16 and 2.17 are taken from [Needham (1997)]. Chapter 3 Cauchy Integral Theorem and consequences Having become familiar with complex differentiation, we now turn to inte­ gration. In this chapter we will learn a very important theorem in complex analysis, called I The Cauchy Integral Theorem, I We will begin by defining "contour integration". And then we will show the Cauchy Integral Theorem. One might ask: Why is this contour inte­ gration and the Cauchy Integral Theorem so important? The importance of integration in the complex plane stems from the fact that it will lead to a greater understanding of holomorphic functions, for example, the funda­ mental fact that holomorphic functions are infinitely many times complex differentiable! In this chapter we will learn the following main topics: (1) (2) (3) (4) Contour integration and its properties. The Fundamental Theorem of Contour Integration. The Cauchy Integral Theorem. Consequences of the Cauchy Integral Theorem: (a) (b) (c) (d) 3.1 Existence of a primitive. Infinite differentiability of holomorphic functions. Liouville's Theorem and the Fundamental Theorem of Algebra. Morera's Theorem. Definition of the contour integral In ordinary calculus, given a continuous function f : [a, b] ➔ JR., l b f(x)dx 59 (3.1) 60 A Friendly Approach to Complex Analysis has a clear meaning. Now suppose we wish to generalize this in the complex setting: given z, w complex numbers, want to give meaning to something like 1 w f(() d(. Then a first question is: How do we get from z to w? In JR, if a< b, then there is just one way of going from the real number a to the real number b, and so our data in the real case is just: (1) a< b, and (2) a continuous function f: [a, b] -t R But now z and w are points in the complex plane, and so there are many possible connecting paths along which we could integrate. See Figure 3.1. - - - - o----e:::,,.---00 a b Fig. 3.1 JR - - - - z c,;r Which path to go from z to w? So in the complex setting, besides specifying the end points z and w, we will also specify the path 'Y taken to go from z to w, and we will replace the above expression (3.1) in the real case by an expression which looks like this in the complex setting: 1 f(z)dz. We call such an expression a "contour" integral, for the computation of which we need the following data: (1) A domain D (c IC), and z,w ED. (2) A continuous function f : D-+ C. (3) A smooth path 'Y : [a, b] -t D joining z to w. Cauchy Integral Theorem 61 We note that we need not merely a path, but a smooth path joining z to w. What do we mean by "smooth" here? Recall that a path,: [a, b]-+ D is a continuous function. We can decompose, into its real and imaginary parts: ,(t) = x(t) + iy(t), where x, y : [a, b] -+ R The path, is called differentiable. Here are some examples. t E [a, b], smooth if x, y are continuously re be given by,(t) = t(l+i), t E [O, l]. Then Example 3.1. Let,: [O, 1]-+ the real and imaginary parts x, y : [O, 1] -+ IR of, are given by x(t) = t, y(t) = t, t E [O, l]. Since x, y are continuously differentiable on [O, 1],, is a smooth path. See Figure 3.2. l+i 'Y 0 Fig. 3.2 The smooth path 1. Similarly consider the two paths,1, , 2 : [O, 21r] -+ IR, given by where t ,1 (t) = exp(it) and ,2(t) = exp(2it), E [O, 21r]. Then since the real and imaginary parts of these paths are cos t, sin t, cos(2t), sin(2t), all of which are continuously differentiable, it follows that,1, , 2 are smooth paths. See Figure 3.3. Fig. 3.3 The smooth path 1. So the set of image points (the ranges of,1 and ,2) is the same, that is: {,1 (t) : t E [O, 21r]} = {,2(t) : t E [O, 21r]} = {z circle with center 0). However ,1 and ,2 are functions are not the same: for example, ,1(1r) E re : lzl = 1}, (the unit different paths, because the = -1-/=- 1 = , 2(1r). ◊ 62 A Friendly Approach to Complex Analysis Remark 3.1. It is very common and convenient to refer to the t E [a, bl} path/curve itself. range {'y(t) : of a path 'Y : [a, b] --+ C as the W ith this usage, a path becomes a concrete geometric object (as opposed to being a mapping), such as a circle or a straight line segment in the complex plane and hence can be easily visualized. The difficulty with this abuse of terminology is that several different paths can have the same image, and so it causes ambiguity. The precise definition of the contour integral is given below. Definition 3.1. Given (1) a domain D, : (2) a continuous function f D --+ C (with real and imaginary parts de­ noted by u, v: D ➔ �), and (3) a smooth path 'Y : [a, b] --+ by x, y : [a, b] --+ �), we define i f(z)dz := := := D (with real and imaginary parts denoted b 1 f('Y(t))'Y'(t)dt b 1 ·(x'(t)+iy'(t))dt b 1 (u('Y(t)) x'(t) -v('Y(t)) y'(t))dt b 1 (v('Y(t)) x'(t) u('Y(t)) y'(t))dt. (3.2) (u('Y(t))+ iv('Y(t))) · · • +i • + We note that the two integrals in the last line above are just the usual Riemann integrals of real-valued continuous functions. The contour integral can be interpreted geometrically as follows. The term 'Y'(t)dt = x'(t)dt + iy'(t)dt can be viewed as an infinitesimal incremental piece of the contour. multiply this by the (almost constant) value f('Y(t)) We off on this incremental piece. Finally, we add up all these contributions along the contour to get the total as the integral b 1 f('Y(t))'Y'(t)dt. See Figure 3.4. 63 Cauchy Integral Theorem D f('Y(t)) f 'Y(b) � iy'(t)dt \ 'Y(a) x'(t)dt Fig. 3.4 Geometric meaning of the contour integral. Here is an example. Example 3.2. Let (l) D = C, (2) 'Y be the smooth path given by (3) f = (z Hz). l Then f(z)dz = = 11 11 t(l + i) • (1 + i)dt t(l - i) · (1 + i)dt = = 2 r1 tdt = 2 . Jo 'Y(t) = t(l + i), t E t2 1 l 2o =2 . 11 t(l 2 - [O, i2 )dt = 1], and 11 t(l + l)dt 1 2 = 1. ◊ Exercise 3.1. Consider the three paths 11,,2,13: [0,27r]--+ <C defined by = exp(it), 12(t) = exp(2it), 13(t) = exp(-it), 11(t) for t E [O,27r]. integrals are all different. Show that their images are the same, but the three contour 64 A Friendly Approach to Complex Analysis Exercise 3.2. Let f be holomorphic in a domain and let 'Y: [O, 1] ➔ D be a smooth path. Show that :/('Y(t)) = J'('y(t)) · "/(t) for all t E [O, l]. In the sequel, we will often assume that our smooth paths are parameterized by [O, 1] (rather than some more general interval [a, bl). Let us explain why we may assume this. Suppose that 'Y : [a, b] ➔ and -;:;; : [c, d] ---+ are two smooth paths, such that there is a continuously differentiable function cp : [c, d] ➔ [a, b] such that a= cp(c), b = cp(d), and -;:;;(t) = "f(cp(t)) for t E [a, b]. We call such smooth paths "equivalent". (Imagine going from 'Y(a) = -;::;(c) to 'Y(b) = -;::;(d) along the same route, but with possibly different speeds.) See Figure 3.5. We now show the following. re b cp(t) 1 '' : cp a C d t re, '(b)�',(a) � J '\') � ,(,p(t)) ) � Fig. 3.5 'Y(a) = -;::;(c) Equivalent paths. Equivalent paths give the same integral. By the chain rule it follows that b 1 J(-;::;(t))-;::;'(t)dt = 1 f('Y(cp(t))h'(cp(t))cp'(t)dt b hf(z)dz = (-r=cp(t)) id J('Y(T)h'(T)dT = c In particular, given any 'Y: [a, b] ➔ cp(t) = 1 "Y J(z)dz. re, we can define cp : [O, 1] ---+ [a, b] by t)a + tb, t E [a, b]. Then cp is continuously differentiable, and cp(O) = a, cp(l) = b. So with c := 0, d := 1 in the above, and with -;:;; : [O, 1] ➔ re defined by -;:;; = 'Y o cp, (1- we see that h J(z)dz = i f(z)dz, Cauchy Integral Theorem 65 and so there is no loss of generality (when it comes to statements about contour integrals) in assuming that the smooth path is parameterized by [O, l]. Contour integrals along piecewise smooth paths. We extend the definition above to paths with "corners". A path , : [a, b] -+ C is called a piecewise smooth path/curve if there exist points a < c1 < · · · < and, is continuously differentiable on For such a path, we define 1 7 f(z)dz := l a + Here is an example. ci 1 <b Cn [a, c1l, [c1, c2l, ..., [cn -1, cn l, [en, b]. f(,(t)),'(t)dt + 1:: J c2 q f(,(t)),'(t)dt + Example 3.3. Let 7y be the path from O to 1 _ ,(t)= { See Figure 3.6. t c1, ..., Cn such that f(,(t)),'(t)dt + · · · 1: f(,(t)),'(t)dt. + i given by ift E [O, 1] l+(t-l)i iftE(l,2]. l+i ' Fig. 3.6 Then we have J2 12 { zdz = f\1dt+ 1� lo = = fo 1 tdt + 1 0 The piecewise smooth path (1 + (t- l)i)idt= ( (i lo + (t- l))dt -1 = 1 +i. !2 +i+ 4-l 2 1 'i- tdt+ 12 1 (1 - (t- l)i)idt 66 A Friendly Approach to Complex Analysis Let us look at the answers obtained in Examples 3.2 and 3.3. We see that z f---+ z), the end + i), and the paths joining the integrand is the same ( the nonholomorphic function points of the paths are the same (namely O and 1 these two points are different ('-y and ::Y). See Figure 3.7. l+i ' Fig. 3.7 Iz 0 ' The two paths "Y and-;;;. hz The answer we obtain in each case is different too: Thus the integral z f---+ z. depends dz = l -=/- 1 +i= dz. on the path for the nonholomorphic integrand This is not strange, because from the definition of the contour integral of course we expect the value of the contour integral to depend on the route chosen. The main goal in this chapter is to show that the contour integration of a holomorphic function along two paths joining the same points from z to w is the same provided that the map is holomorphic everywhere in the region between the two paths! It turns out that this is a fundamental result (called the Cauchy Integral Theorem) in complex analysis because many further results follow from this. Let us check this in an example. Example 3.4. Consider the same two contours 1, ::Y considered in Exam­ ples 3.2 and 3.3 above. But instead of taking the nonholomorphic map z f---+ z, I 1 + i)t(l + i)dt = 1 1 2itdt = i, zdz = 1 1 t ldt + 12 + (t l)i)idt consider now the = · 1 [ tdt + lo function (l h 1 zdz = entire (1 j\1 z: Then we have and - (t - l))dt = !2 - !2 + i = i. We note that this time the answer is the same for I and for :::;;. ◊ Cauchy Integral Theorem Exercise 3.3. Integrate the following functions over the circle anticlockwise: 67 lzl = 2, oriented (1) z+z. (2) z 2 - 2z + 3. (3) xy, where z = x + iy, x, y E JR. Exercise 3.4. Evaluate i Re(z)dz, where 'Y is: (1) The straight line segment from Oto 1 + i. (2) The short circular arc with center i and radius 1 joining Oto 1 + i. (3) The part of the parabola y 3.1.1 = x 2 from x = 0 to x = l joining Oto 1 + i. An important integral Let us now calculate a very important simple contour integral, which will recur again and again in the course. We fix a useful convention: through­ out the book, unless otherwise stated, a circular path with center z0 and radius r > 0 traversed in the anticlockwise direction will mean the path C: [O, 21r] -+ <C given by C(t) = z0 + rexp(it), t E [O, 21r]. (Thus the circle is traversed once.) See Figure 3.8. C zo r Fig. 3.8 The circular path C with center zo and radius r traversed in the anticlockwise direction. We will now calculate the integrals L (z - zotdz, n E Z. This calculation will prove to be very useful later on. Theorem 3.1. Let C be a circular path with center z0 and radius r traversed in the anticlockwise direction. Then { ( _ )nd = { 21ri if n = -1, Jc z zo z O if n -/- -1. >0 68 A Friendly Approach to Complex Analysis We note that the answer is independent of r. Proof. We have C(t) = zo+ rexp(it) = zo+ rcost+ irsint, t E [0,27!"], and so C'(t) = -rsint+ircost= ir(cost+i sint) = irexp(it) , t E [0,27!"]. We now consider the two cases: 1 ° When n = -l, we have { (z - zotd z = { (z - zo)-1d z = le le = fo 21r f 21r lo 1 · irexp(it)dt it rexp(") idt= 27l"i. 2° When n =/- -l, we have fc (z - zo) d z= fo21r r exp(nit) • irexp(it)dt 2 = fo ir +l exp(i (n+ l)t)dt 2 2 = -r fo :in((n+ l)t)dt+ ir fo :os((n+ l)t)dt n n n 1r n+ 1 n+1 = 0+0= 0. □ This completes the proof. We will see later that this has significant consequences. For instance, suppose that we have an f that has a "expansion in terms of integral powers of z" (whatever that means) , in an annulus A. := { z E <C : r < lz - zol < R} with center z0, inner radius r, and outer radius R like this: f(z) = L an (z - zot , z EA.. nEZ We will give precise meaning to the (infinite) "sum" above later on, but for now, one may just imagine a finite sum (so that all but finitely many an s are zeros). Then by multiplying both sides by (z - z0 )-(m +l) for some m E Z, we obtain __f_(_z)_...,.. = " an (z - zo) n-m-1 , L..i (z - zo)m +l and so l -. 1 nEZ 1 f(z) " an (z - zo) n-m-ld z = a . d z = L..i m 27ft e (Z - Zo )m +l e ,,, nEa.. Cauchy Integral Theorem 69 In the above, we assumed that the sum passes through integration over C, which for finite sums, follows from the definition of the integral, and we will see this in the next section. When the sum is not finite, we will make precise the details later, but things work out essentially as suggested by this calculation. The upshot of it all is that the coefficients are expressible in terms of a contour integral, and we will see later that any function holomorphic in an annulus will have such an expansion. Exercise 3.5. Let C be the circular path with center O and radius 1 traversed in the anticlockwise direction. Show that for O :S k :S n, (n) k 3.2 z) n = � { (lz+ dz. k +1 27ri }0 Properties of contour integration In this section we will show some useful properties of contour integration. The following result follows in a straightforward manner from the definition of contour integration. Proposition 3.1. Let D be a domain in <C and "I: [a, b] ---+ D be a piecewise smooth path. Then the following hold: (1) For all continuous f, g : D---+ <C, l (f + g)(z)dz= l f(z)dz+ l g(z)dz. (2) For all continuous f : D ---+ <C and all a E <C, l (af)(z)dz =al f(z)dz. Let C(D; <C) denote the vector space over <C of all complex-valued continu­ ous functions on D with pointwise operations. Then the above result says that each piecewise smooth path "/ in D induces a linear transformation from C(D;<C) to <C, namely f f--t l f(z)dz: C(D;<C)---+ <C. Exercise 3.6. Prove Proposition 3.1. 70 A Friendly Approach to Complex Analysis Opposite paths. Given a smooth path 'Y: [a, b] ---+Din a domain D, its oppos ite path,-"(: [a, b] ---+ D, is defined by (-'Y)(t) = 'Y(a+b-t),t E [a, b]. Then (-'Y)(a) = 'Y(b) and (-'Y)(b) = 'Y(a), and so-"( starts where 'Y ends, and ends at the starting point of 'Y while traversing the same path of 'Y but in the opposite direction. See Figure 3.9. Fig. 3.9 The opposite path --y to the path 'Y· But why do we denote the opposite path by -"(? Here's why. Proposition 3.2. Let 'Y : [a, b]---+ D be a smooth path in a doma in D an d f : D ---+ C be a continuous function . Then 1 -"! Proof. ['Y We have f(z)dz = -1 'Y f(z)dz. b f(z)dz 1 f( (-'Y)(t)) · (-'Y)'(t)dt b 1 f('Y(a+b- t)) · ('Y'(a + b- t)) · (-l)dt (r= a� b -t) l a -1 f('Y(T)). 'Y'(T)dT = f(z)dz. -1 b f('Y(T)). 'Y'(T)dT □ Exercise 3.7. Show that-(-')')= 'Y, where 'Y: [a,b]-+ Dis a smooth path in a domain D. Concatenation of paths. Let D be a domain and let "(1 : [a1, b1] ---+ D and 'Y2 : [a2, b2]---+ D be two paths such that 71 Cauchy Integral Theorem (so that 'Y2 starts where 'Yl their "concatenation" by: ('Yi+'Y2)(t) = { ends). Define to be 'Yl (t) for a 1 ::; t ::; b 1 , b 'Y2(t - 1 +a2 ) for b 1 ::; t::; b 1 +b2 - a2 . 'Yl 'Yl +'Y2 Concatenation ')'1 Fig. 3.10 'Yl + 'Y2 : [a 1 ,b 1 +b2 - a2 ] 'Y2 + ')'2 of two paths ')'1, ')'2. Proposition 3.3. Let D be a domain in <C and let 'Yl : [a 1 ,b 1 ] ➔ D and 'Y2: [a2,b2] ➔ D be two paths such that 'Y 1 (b1) ="f2 (a2 ). Then 1 Proof. ')'1 +"12 1 "/1 +"12 We have f(z)dz =1 f(z)dz = bl+bra 1�1 1 1�1 1 1 1 2 a1 = + = = = + "/1 "/1 1 f(z) dz + "/1 1 "/2 f(z)dz. f(('Y 1 +'Y2)(t))('Y 1 +'Yd(t)dt f(('Y 1 +'Y2)(t))('Y 1 +'Y2)'(t)dt b 1 +b 2- a 2 f(('Y 1 +'Y2)(t))('Y 1 +'Y2 )'(t)dt b1 f('Y 1 (t))'Y�(t)dt b 1 +b 2- a 2 f('Y2 (r - b 1 +a2)h�(r - b 1 +a2 )dr b1 f(z) dz+ f(z) dz+ j 1 b2 a2 "/2 f('Y2 (s))'Y�(s)dt f(z) dz. (s = r - b 1 + a2 ) □ Exercise 3.8. If')': [a, b] ➔ Dis a smooth path in a domain D, and f: D ➔ IC is continuous, then show that f Ir+<--r) f(z)dz = 0. A useful estimate. We now prove an inequality for the size of the contour integral in terms of the size of lfl along the contour, and the length of the contour. This will prove to be indispensable in the sequel. 72 A Friendly Approach to Complex Analysis Proposition 3.4. Let be a domain in <C, 1 : [a, b] ---+ D be a piecewise smooth path and (3) f : D ---+ <C be a continuous function. (l) (2) D 11 Then 'Y f(z)dzl::; ( max lf('Y(t))I) · (length of 1). (3.3) tE[a,b] Recall that the length of I is given by 1 b J(x'(t))2 + (y'(t))2, where x, y : [a, b] ---+ IR are the real and imaginary parts of 1. See Fig­ ure 3.11. 1(b) y'(t)dt 1(a) Fig. 3.11 The arc length of the path 'Y is the sum of the incremental arc lengths ds, where ds = (x'(t)dt)2 + (y'(t)dt)2 = (x'(t))2 + (y'(t)) 2 dt. ✓ ✓ Proof. Consider first a curve To see this, let 11 cp(t)dtl b b 1 <p(t)dt = r = · exp(i0), where r = exp(-i0) • r • exp(i0) = exp(-i0) · = <p : [a, b] ---+ <C, for which we prove 1 b b 1 cp(t)dt = 1 Re (exp(-i0) · b r?: 0 and 0 E (-1r, 1r]. T hen exp(-i0) • <p(t)) dt + i 1 b <p(t)dt Im (exp(-i0) · cp(t)) dt. 73 Cauchy Integral Theorem But the left hand side is real, and so the integral of the imaginary part on the right hand side must be zero. Consequently, 11 cp(t)dtl b =1 b Re (exp(-i0) • cp(t)) dt :::; 1 [Re (exp(-i0)·cp(t))[ dt b :::; 1 [exp(-i0)·cp(t)[ dt b = 1 [cp(t)[ dt. b The proposition now follows, since with cp(t):= f('Y(t))·'Y'(t), t E II If 'Y(t) = f(z) dzl x(t) 11 f('Y(t))'Y'(t)dtl ::::: 1 lf('Y(t))'Y'(t)[dt b + =1 b [f('Y(t))ll'Y'(t)[dt :::; ( max [f('Y(t))I) J ['Y'(t)[dt. tE[a,b] b iy(t), where x, y are real-valued, then 1 ['Y'(t)[dt b = [a, b], b =1 b This completes the proof. a J(x'(t))2 + (y'(t))2 dt = length of 1'· □ Exercise 3.9. Calculate the upper bound given by (3.3) on the absolute value of the integral i z2 dz, where 'Y is the straight line path from O to 1 find its absolute value. + i. Also, compute the integral and Exercise 3.10. Using the calculation done in Exercise 3.5, deduce that (:) ::; 4 . n 3.3 Fundamental Theorem of Contour Integration Let us recall the Fundamental Theorem of Calculus in the real setting: Theorem 3.2. (Fundamental Theorem of Calculus) If F: [a, b]--+ JR. is continuously differentiable and F' =:f on [a, bl, then 1 f(x)dx b = F(b) - F(a). 74 A Friendly Approach to Complex Analysis This is an important result, because it facilitates the computation of the Riemann integral. Indeed, if we know that a function is the derivative of something, then it is easy to calculate its integral. For example, b3 _a3 d because x 2 = x 2dx = ( 3) . dx : 3 ab Analogously, we will now see that if f is the derivative of a holomorphic function, then the calculation of the contour integral l I fz ( )dz is easy, since we have (similar to the Fundamental Theorem of Calculus in the real setting): Theorem 3.3. (Fundamental 1 Theorem of Contour Integration) Let (l) Dbe a d omain in C, (2) 'Y : [a, b] ➔ Dbe a piecewise smooth path, (3) f: D ➔ C be a continuous function in D, (4) F: D ➔ C be a holomorphic function such that F' =fin D. Then I fz ( (a)). ( b)) - F, ( )dz = F,( How does this theorem help? One can now calculate some contour integrals very easily (just like in ordinary calculus). Here is an example. Example 3.5. Since 1 zdz = 'Y d z ( :) dz =z z( E q, for (l + i)2 _ 0 2 2 2 1 + 2i + i2 any 'Y joining O to 1 2 1 + i, + 2i - 1 = i, 2 and so in particular, we recover the answer obtained in Example 3.4. ◊ I In particular, as we have also seen in the previous example, fz ( )dz = Fw ( ) - Fz ( ) is independent of the path 'Y joining the points z tow, when f possesses an "antiderivative" or "primitive" Fin D. 1The naming of this result is done just to highlight the similarity with the real analysis analogue. However, in complex analysis, this isn't all that "fundamental". We will soon learn about Cauchy's Integral Theorem, which is certainly more fundamental! 75 Cauchy Integral Theorem = z for all z E C. Indeed, the calculation in Examples 3.2 and 3.3 shows that the contour integral along paths joining O to 1 + i does depend on the path chosen. ◊ Example 3.6. There is no function F : (C ➔ (C such that F'(z) Proof. (of Theorem 3.3.) For z = x + iy ED, where x,y are real, define the real-valued functions U,V,u,v by F(x + iy) = U(x,y) + iV(x,y), f(x + iy) = u(x,y) + iv(x,y). Also, set 'Y(t) = x(t) + iy(t) (t E [a, bl), where x,y are real-valued. Then by the Cauchy-Riemann equations, we have u(x,y) + iv(x,y) = f(x + iy) = F'(x + iy) au av .a v .a u = a (x,y) + i a (x,y) = a (x,y) - i a (x,y). x x y y By the chain rule and the above, we have ! U(x(t),y(t)) = !� (x(t),y(t)) · x'(t) + !� (x(t),y(t)) · y'(t) = u(x(t),y(t)) · x'(t) - v(x(t),y(t)) · y'(t). ! Similarly, i V(x(t),y(t)) = !: (x(t),y(t)) • x'(t) + !; (x(t),y(t)) • y'(t) = v(x(t),y(t)) · x'(t) + u(x(t),y(t)) · y'(t). Thus J(z)dz = l b J('Y(t))'Y'(t)dt ( b = l u(x(t),y(t)) + iv(x(t),y(t))) (x' (t) + iy'(t) )dt =lb ! U(x(t),y(t))dt + i lb ! V(x(t),y(t))dt = U(x(b),y(b))-U(x(a),y(a)) +i(V(x(b),y(b))-V(x(a),y(a))) = F('Y(b)) - F('Y(a)). This completes the proof. □ 76 A Friendly Approach to Complex Analysis Exercise 3.11. Show, using the Cauchy-Riemann equations, that z 1--t primitive in <C. z has no Exercise 3.12. (Integration by Parts Formula.) Let f, g be holomorphic functions defined in a domain D, such that J', g' are continuous in D, and let 'Y be a piecewise smooth path in D from w E D to z E D. Show that i f(()g'(()d( = f(z)g(z) - J(w)g(w) - Exercise 3.13. Evaluate i i J'(()g(()d(. cos zdz, where 'Y is any path joining -i to i. Definition 3.2. A path 'Y : [a, b] ---+ C is said to be closed if 'Y(a) = "f(b). Corollary 3.1. Let (1) (2) (3) (4) D be a domain in C, 'Y : [a, b] ---+ D be a closed piecewise smooth path, f: D ---+ C is a continuous function in D, F: D---+ C be a holomorphic function such that F' = f in D. Then 1 Proof. f(z) dz = 0. 1 f(z)dz = F('Y(b)) - F('Y(a)) = 0 since "f(b) = 'Y(a). 1 Example 3.7. Form E Z \ {0}, z ED:= C \ {0}, □ � ( ::) = zm -1; so d zm -ldz = 0 for any closed path 'Y in D. What if m = 0? Note that Log' z = I/ z for z E jj := C \ (-oo, 0), and so for any path 7 in D, we do have f !dz= 0. }-:y z However, in D, !z doesn't have a primitive; see Exercise 3.16. ◊ Cauchy Integral Theorem 77 i Exercise 3.14. Use the Fundamental Theorem of Contour Integration to write down the value of expzdz where I is a path joining O and a + ib. Equate the answer obtained with the parametric evaluation along the straight line from O to a+ ib, and deduce that 1 1 _ a( e a cosb-l)+be a sinb e a x cos (bx)d x• a 2 + b2 o Exercise 3.15. Applying the Fundamental Theorem of Contour Integration to expz and integrating round a circular path, show that for all r > 0, [211" J o e rcosO cos(rsin0 + 0)d0 = 0. Exercise 3.16. Show that 1/z has no primitive in the punctured plane <C \ {O}. The Cauchy Integral Theorem 3.4 We will now show one of the main results in complex analysis, called the Cauchy Integral Theorem. Theorem 3.4. (The Cauchy Integral Theorem) Let (1) (2) (3) D be a domain in <C, f: D --+ <C be holomorphic in D, and 'Yo, 'Yi : [O, 1] --+ D be two closed, piecewise smooth, D-homotopic paths. Then 1 ')'O f(z)dz = 1 f(z)dz. ')'1 Before we go further, let us try to understand the statement. Notice, first of all, that the two paths in D are closed. Secondly, what do we mean by saying that the two closed paths are "D-homotopic"? Intuitively, this means the following. Look at Figure 3.12, where we have depicted the two paths in the domain. Imagine you have placed a rubber band along 'YO· For 'Yo to be D-homotopic to 'Yi, we should be able to deform this rubber band so as to get 'Yl, with the condition that each intermediate position of the rubber band lies in D. Clearly this is not possible sometimes, for example if the domain has holes. See for example the picture on the right of Figure 3.12, we expect the two paths in the domain D taken as punctured complex plane D = <C \ {O} to be not D-homotopic. 78 A Friendly Approach to Complex Analysis intermediate position of the rubber band obstruction Fig. 3.12 ')'O, ')'1 in D :=CareC-homotopic in the picture on the left, but in the picture on the right, the two paths ')'O, ')'l in D = C\ {O} are not C\ {O}-homotopic. We have the following precise definition. Definition 3.3. Let D be a domain in C and ')'o, ')'1 : [0, 1] -+ D be closed paths. Then 'Yo is said to be D-homotopic to 'Yl if there is a continuous function H: [0, 1] x [0, 1] -+ D such that the following hold: (Hl) For all t E [0, 1], H(t,0) = 1'o(t). (H2) For all t E [0, 1], H(t, 1) = 'Yi(t). (H3) For alls E [0, 1], H(O,s)= H(l,s). We can think of the Has a family of closed paths from [0, 1] to D, param­ eterized by "time", the s-variable. You may think of the closed path at time s as the position of the rubber band at time s, Initially, when s = 0, H(·, 0) is the path 'Yo, while finally, when the times= 1, we end up with H(·, 1), which is the path 'Yl· So far, this is what (Hl) and (H2) say. The requirement (H3) just says that at each point of time s, the intermediate path 'Ys := H(·,s) is closed too. Continuity of H means that the rubber band never breaks, and the deformation takes place smoothly. The picture below illustrates this. H(·,s) H(-,0) = 'Yo 79 Cauchy Integral Theorem Example 3.8. Let D = C, and 'Yo, 11 : [0, 1] --+ C be the two circular paths given by 'Yo= 4exp(27rit), and 11 = 2i + exp(27rit), fort E [0, 1]. Then 'Yo is C-homotopic to 'Yl· Indeed, we can define H by just taking a "convex combination" of the points 1o(t) and 11(t). See Figure 3.8. Set H: [0, 1] x [0, 1]--+ C by H(t,s)= (1- s) · 1o(t) + s · 11(t)= (4 - 3s) · exp(27rit) + s • 2i , 0 ::::; t, s ::::; 1. Then His clearly continuous, and moreover, (Hl) for all t E [0, 1], H(t,0) = (4- 0) · exp(27rit) + 0 · 2i = 10 (t), (H2) for all t E [0, 1], H(t, 1) = (4- 3) · exp(27rit) + 1- 2i = 11(t), and (H3) for each s E [0, 1], H(0,s) = (4- 3s) · 1 + s · 2i = H(l,s). Hence (Hl), (H2), (H3) are satisfied and so 10 is C-homotopic to 11. On the other hand, the same two paths are not C \ {0}-homotopic. Why is that? If they were C \ {0}-homotopic, then by the Cauchy Integral Theorem, the contour integral of the holomorphic function 1/z in C \ {0} along the two paths would be the same. But we have 1! dz= 27l"i =I= 0= 1 !dz, ')'1 Z 'Yo Z where the last equality follows from the Fundamental Theorem of Contour Integration, because 1/ z has the primitive Logz in C \ (-oo, 0]. ◊ Exercise 3.17. Let D be a domain in <C. Show that D-homotopy is an equivalence relation on the set of all closed paths in D. In particular, we can say ",o, 11 are D-homotopic" instead of saying that ",o is D-homotopic to 11". Proof. ( of Theorem 3.4.) We will make the simplifying assumption that the homotopy H is twice continuously differentiable. This smoothness con­ dition can be omitted (see for example [Conway (1978)]), but then the proof becomes technical. Moreover, the assumption of twice continuous differen­ tiability is mild, and we will invoke this below when we will exchange the order of partial differentiation: fP H 8s8t 82 H 8t8s · 80 A Friendly Approach to Complex Analysis Idea of the proof: Let 'Ys := H(·, s) be the intermediate curve at times. Define I( s ) := 1 f(z)dz, s E [O, l]. /S (We will use differentiation under the integral sign with respect tos to show that d /(s) 0, d showing thats H I( s) is constant, and in particular = 1 f(z)dz = I(0) = J(l) = 10 1 1 f(z)dz, 'Yl which is the desired conclusion.) We have 8H dI d d [1 dz s z ( )= f( ) = l f(H(t, s)) 8t (t, s)dt ds ds o ds ,. = 1( 1 1 d s s :s (f(H(t, ))�� (t, )) t 82 H 8H 8H ( s s s s )) [J (t, s )) dt ) + f(H(t, )) [J f'(H(t, (t, (t, ) l = o s t 88 8t (1 =l o = 1 ( f'(H(t, s)) 8H (t, s) 8H (t, s) + f(H(t, s)) [J82[JH (t, s)) dt 1! 8t 88 t s (f(H(t, s )) �� (t, s)) dt, and so by the Fundamental Theorem of Integral Calculus, (1 d 8H dI (t, s)) dt f(H(t, s)) (s) = l ( 88 ds o dt 8H 8H = f(H(l, s)) 88 (l, s) - f(H(0, s )) 88 (o, s) H(l, a) -H(l, s ) -f H s))lim H(O, a)- H(O, s) ( (l, o---+s a - s a - s a a s ) , ) , H( , H( H( l l )-H(l, s ) -f(H(l,s))lim - l =f(H(l,s))lim o---+s o---+s a - s a - s =f(H(l,s))lim o---+s = 0. 1 1 Hence the maps H I( s) : [0, 1] -+ (C is constant. In particular, 11 f(z) dz = J(l) = J(0) = 12 f(z) dz. 81 Cauchy Integral Theorem This completes the proof. But where did we use the holomorphicity of f? We used this in order to get the equality in the third line above. Indeed, we claim that :/f (H (t, Let f = s))) = f' (H (t, s)) · �� (t, s ). = X u + iv, and H + iY, where u, v, X, Y are real valued. We suppress the arguments (t, s) below. Then we have :/f (H (t, s ))) = = :/u (X, Y ) + iv (X, Y )) u a , (X Y ax +i ( = u a ( = f' v ax (X ax +i ( = a u a ax (X ax aX as (X, Y ) · Y ' av ). ). aX as (X, Y ) · Y (X ' )+ + 8s + aX _ v (X as v (X, Y ). y a u aY as (X, Y ) · ax )). ( +iY ) 8s ) aY a Y (X ' aY a (X ' Y ) . 8s + i ax s a ax aX . av +iY ) · : u . (X, Y ) ay a = f' a x as ( 8s ) aY . aY +ia H (t , s s ) )) · �� (t,s). The careful reader would have also noticed that we also assumed that f' is continuous when we differentiated under the integral sign. Again, the result holds without this assumption, but we will not do this here. The interested student can find the complete proof of the Cauchy Integral Theorem for example in [Conway (1978)]. D Exercise 3.18. We have seen that if C is the circular path with center O and radius 1 traversed in the anticlockwise direction, then { !dz = 21ri. la z Now consider the path S, comprising the four line segments which are the sides of the square with vertices ±1 ± i, traversed anticlockwise. Draw a picture to convince yourself that Sis (C \ {O}-homotopic to C. Evaluate parametrically the integral { !dz, ls z and confirm that the answer is indeed 21ri. 82 A Friendly Approach to Complex Analysis Exercise 3.19. Let a> 0, b > 0, and E: [O,27r] ➔ <C be the elliptic path By considering 3.4.1 = acost+ibsint, jEz-1 dz, show that 12 E(t) 71" O t E [O,27r]. 1 271" • 0 2 d0 = a2(cos 0)2 + b2( sm ab • ) Special case: simply connected domains Consider a "degenerate" closed path, which is constant. That is, if D is a domain and p E D, then consider "/p : [0,1] ➔ D given by "/p (t) = p, t E [0, l]. Then "/p is closed, because "/p(O) = p = "/p (l). For any continuous f: D ➔ C, what is 1 IP f(z)dz? 0, because 'Y�(t) = 0 for each t E [a, b] and 1 1p 1 f(z)dz = [ f('Yp (t)) · "f�(t)dt = 0. lo In light of this, we see that an important special case of the Cauchy Integral Theorem is obtained when we know that a closed path 'Y is D-homotopic to a point (that is, the constant path "/p (t) = p for all t E [0,1]). In this case we say that 'Y is D-contractible. Imagine placing a rubber band along 'Y, and then shrinking it to a point such that each intermediate position of the rubber band is in D. Indeed for a D-contractible path 'Y (which is D-homotopic to a point p E D), and for a holomorphic function f: D ➔ C, we have by the Cauchy Integral Theorem that 1 1 J(z)dz = 1 IP J(z)dz = 0. A domain in which every closed path is D-contractible is called simply connected. For example, the domains C, II)):= {z EC: lzl < 1}, C \ (-oo,0], are all simply connected. For example if we take any p in the domains in first two cases, then the homotopy given by H(t, s):= (1- s)'Y(t) + sp, t, s E [0,1] does the job. In the case when D = C\ (-oo,0], given any 'Y: [0,1] ➔ D, we first choose any real p > 0 for example p = 1, and then use the same Has above. Note that none of the above domains have any "holes" in them. On Cauchy Integral Theorem 83 the other hand, domains with holes are not simply connected. For example, the punctured complex plane (C \ { 0} is not simply connected. For example consider the circular path with center O and any positive radius r traversed once in the anticlockwise direction. We have seen that 1 !dz= 27ri. cZ But if C were (C \ { 0}-contractible to a point in the punctured plane, we should have had 1 !dz= 0 cZ by Cauchy's Integral Formula. So this means that C is not (C \ {O}­ contractible to a point in (C \ {O} and so (C \ {O} is not simply connected. Similarly, one can show that the annulus { z E (C : 1 < lzl < 2} is not simply connected. See Figure 3.13. The obstruction of the hole can be thought of as a nail or a pillar emanating from the plane, which prevents a rubber band encircling it from being shrunk to a point in the domain while always staying in the plane. simply connected not simply connected Fig. 3.13 The domains IC, ll)) := {z E IC : JzJ < 1} and IC\ (-oo, OJ are simply connected, while the annulus A:= {z E IC: 1 < JzJ < 2} and the punctured plane IC\ {O} aren't. 84 A Friendly Approach to Complex Analysis We have the following corollary of the Cauchy Integral Theorem. Corollary 3.2. Let (1) D be a simply connected domain, (2) 'Y be a closed piecewise smooth path in D and (3) f: D ➔ C be holomorphic in D. Then i f(z)dz = 0. This corollary itself is also sometimes called the Cauchy Integral Theorem. Example 3.9. For any closed path 'Y, since exp is entire, and (C is simply i connected, we have In fact for any entire function It can happen for a f: D ➔ function non (C that expzdz = 0. f := 1/z • Then f(z)dz = 0, for any closed path 'Y· simply connected domain i for every closed smooth path 'Y in 2 i f, f(z)dz = 0 1 'Y D. D ◊ and a holomorphic Here is an example: take D= C\ {0}, �dz=O z for every closed path 'Y in the punctured plane because 1/z 2 possesses a primitive in (C \ {0}: Exercise 3.20. Integrate the following functions over the circular path given by Jzl = 3 traversed in the anticlockwise direction: (1) Log(z - 4i). (2) 1 z-1 (3) Principal value of iz -3. 85 Cauchy Integml Theorem Exercise 3.21. (Winding number of a curve.) Suppose that 1 : [0, 1] ➔ C is a smooth closed path that does not pass through 0. We define the winding number of 1 (about 0) to be 1 1 1 w (1) .·- - / -dz - - 21ri 'Y z 21ri 11 O '(t) ,(t) ']__dt. (1) Using the observation that exp(21ria) = 1 if and only if a E Z, show that w(,) E Z by proceeding as follows. Define cp: [0, 1] ➔ C by To show that w(,) E Z, it suffices to show that cp(l) = 1. To this end, calculate cp' (t), and use this expression to show that cph is constant in [0, 1]. Use this fact to conclude that cp(l) = 1. (2) Calculate the winding number of r1: [0, 1] ➔ C given by r1(t) (t E [0, l]). = exp(21rit) (3) Prove that if ,1, 12: [0, 1] ➔ C are smooth closed paths not passing through 0, and 11 · 12 is their pointwise product, then w(,1 · 12) = w(,1) + w(,2). (4) Let m E N. Calculate the winding number of the curve rm: [0, 1] ➔ C given by rm(t) = exp(21rimt) (t E [0, 1]). (5) Show that the winding number function I M w(,) is "locally constant", by which we mean that if ,o: [0, 1] ➔ C \ {0} is a smooth closed path, then there is a 8 > 0 such that for every smooth closed path 1 : [0, 1] ➔ C \ {0} such that II,- ,olloo := max{l,(t)- ,o(t)I : t E [0, 1]} < 8, we have w(,) = w(,o)­ (In other words, if we equip the set of curves with the uniform topology, and equip Z with the discrete topology, then IM w(,) is continuous.) 3.4.2 What happens with nonholomorphic functions? We now highlight the fact that the Cauchy Integral Theorem may fail if one drops the assumption of holomorphicity of f. Let us see what happens when we consider our favourite nonholomorphic function, the complex con­ jugation map z H z. We will show that rather than the integral around the closed loop 'Y being 0, the contour integral of z around 'Y yields the area enclosed by 'Y, which is of course very much dependent on 'Y, and two <C-homotopic paths can enclose widely different areas (just imagine two concentric circles with different radii). We will only give a plausibility argument by resorting to a specific picture, as shown in the picture below. 86 A Friendly Approach to Complex Analysis (x(t), y(t)) (x(a), y(a)) II (x(b), y(b)) For the smooth path 'Y : [a, b] -+ <C, we have j zdz 'Y = 1 b ab =1 = (x(t) - iy(t))(x'(t) + iy'(t))dt x(t)x'(t) + y(t)y'(t) + i(x(t)y'(t)-y(t)x'(t))dt (x(b))2 -(x(a))2 b + (y(b))2 -(y(a))2 + i 2 = O+i 1 (x(t)y'(t)-y(t)x'(t))dt. _, ' ' x'(t)dt Fig. 3.14 l b f (x(t)y'(t)-y(t)x'(t))dt la ' ' �I b x(t)y'(t)dt and b l l� -x'(t)dt x'(t)y(t)dt. 87 Cauchy Integml Theorem Look at Figure 3.14. From the two pictures on the top, we see that b l x(t)y'(t)dt = (Area enclosed by "f). From the two pictures on the bottom, we see that Thus l b l x'(t)y(t)dt = -(Area enclosed by "f). b z dz = i l (x(t)y'(t) - y(t)x'(t))dt = 2i • (Area enclosed by "f). So in contrast to the Cauchy Integral Theorem for holomorphic functions, we see that for this nonholomorphic function, the integral along a closed contour is not zero, but yields the area of the contour! Exercise 3.22. Suppose a coin of radius r rolls around a fixed bigger coin of radius R. Then the path traced by a point on the rim of the rolling coin is called an epicycloid, and it is a closed curve if R nr, for some n EN. See Figure 3.15. = Fig. 3.15 The epicycloid with n = 6. (1) With the center of the fixed coin at the origin, show that the epicycloid can be represented parametrically as z(t) r((n 1) exp(it) - exp(i(n l)t)), t E [O, 21r]. = + + (2) By evaluating the integral of z along the epicycloid, show that the area en2 closed by the epicycloid is equal to 1rr (n l)(n 2). + + In the rest of this chapter we will learn about several consequences of the Cauchy Integral Theorem. In particular, we will learn (1) that in simply connected domains every holomorphic function possesses a primitive; (2) · that holomorphic functions are infinitely many times differentiable; (3) that bounded entire functions are constants (Liouville's Theorem) (and also use this to prove the Fundamental Theorem of Algebra); (4) a result called Morera's theorem, which is a sort of a converse to the Cauchy Integral Theorem. A Friendly Approach to Complex Analysis 88 3.5 Existence of a primitive We will show that on a simply connected domain, every holomorphic func­ tion is the derivative of some holomorphic function. Theorem 3.5. If (1) D is a simply connected domain and (2) f: D-+ C is holomorphic, then there is a holomorphic function F : D -+ C such that for all z E D, F'(z) = f(z). Proof. Fix a point p E D. Define F : D -+ C by F(z)=i f(()d(, zED, where 'Y is any smooth path in D joining p to z. Why is this F well-defined, that is, why does the F(z) not depend on which path we take joining p to z? If:=:; is another smooth path in D that joins p to z, then 'Y -:=:; is a closed smooth path in the simply connected domain D. See Figure 3.16. 7z z 'Y p Fig. 3.16 'Y - ::Y forms a closed path. 1_ Cauchy Integral Theorem gives so that 1 7z 0= f ( () d( 7-7 f(z)dz = 1 b f(()d( - � f(()d(, h = � f ( () d(, and F is well-defined. Ir Next we show the holomorphicity of F and that F' = f in D. Since f is holomorphic in D, it is also continuous there, and so given a z E D and an E > 0, there is a 8 > 0 such that whenever lw - zl < 8, we have lf(w) - f(z)I < f.. Thus if we take aw such that O < lw - zl < 8, then F( (z) = W�Z w2=: f(()d(). f(()d(- (Jw Jz 89 Cauchy Integral Theorem If 'Yzw is a straight line path joining z to w, then the concatenation of 'Yz with the concatenation of 'Yzw with -,w is a closed path, and so by the Cauchy Integral Theorem, we obtain 0 = 1 'Y.z+'Yzw-")'w 1 f(()d( = f(()d( + 'Yz 1 'Yzw f(()d(-1 f(()d(. 'Yw /1;::\) 'Y - , ,' ', p 1 1 1 The Fundamental Theorem of Contour Integration gives and so F(w I and so F( w 2 = 2 = Id(= 'Yzw : 'Yzw (z) - z = f( ) w � z z :( ) - f(z)I = ('d( = w - z, 'Yzw f(()d( - w � z = w � 1 (!(() -f(z))d(, z 'Yzw I i.w W � z = w� zl l 1 1 'Yzw f(z)d( (!(() -f(z))d(I li.w (!(() -f(z))d(I max If(() - f(z)I) · (length of 'Yzw ) :::; -1 -- (1 (E"/zw W - Z :::; ---1 1:lw-zl = I w-z Thus F'(z) = f(z), and Fis holomorphic. 1 €. □ Remark 3.2. A primitive for a holomorphic function f in a simply con­ nected domain is unique up to a constant. Indeed, if F, F are both primi­ tives for F, then F' = f = F' in D, and so d (F - F) = F' - F' = f - f = O in D. dz By Exercise 2.13, it follows that there is a constant C such that F-F = C in D. So F = F + C in D. 90 A Friendly Approach to Complex Analysis Example 3.10. exp(-z 2 ) is entire. So there exists an F, which is also entire, such that for all z EC, F'(z) = exp(-z 2 ). (But one cannot express F in terms of elementary functions. One primitive is given by F(z) = 1 2 e-C d( 'Yz for z E C, where 'Yz is the straight line path joining O to z. Then in particular, for real x, and it turns out that this ( and so any other primitive too) can't be expressed in terms of elementary functions.) ◊ Exercise 3.23. Suppose that Dis a domain. If f is holomorphic in D, and there is no F holomorphic in D such that F' = f in D, then we know that D cannot be simply connected. Give a concrete example of such a Dand f. Corollary 3.3. If ( 1) D is a simply connected domain, (2) f : D ---+ C is holomorphic, (3) 'Y : [a, b] ---+ D and -::Y : [c, d] ---+ D are two smooth paths such that they have the same start and end points, that is, 'Y(a) = -::Y(c) and "f(b) = -::Y(d), then 1 Proof. 1 f(z) dz= /4 f(z) dz. f has a primitive F and so f(z)dz = F('Y(l))- F('Y(O)) = F(-::Y(l)) - F(-::Y(O)) = /4J(z). □ 91 Cauchy Integral Theorem 3.6 The Cauchy Integral Formula We will now learn about a result, called the Cauchy Integral Formula, which says, roughly speaking that if we have a closed path I without self­ intersections, and f is a function which is holomorphic inside 1, then the value of f at any point inside I is determined by the values of the function on 1! This illustrates the "rigidity" of holomorphic functions. Later on, in the next chapter, we will study a more general Cauchy Integral Formula, which will allow us to even express all the derivatives of f at any point inside I in terms of the values of the function on 1. So we can consider the basic result in this section as the "n = 0 case" of the more general result to follow. Theorem 3.6. (The Cauchy Integral Formula for circular paths) Let (1) D be a domain, (2) f: D--+ C be holomorphic in D, (3) r > 0, zo ED and the disc�:= {z EC: lz-zol � r} CD. Then f(w) =� 21ri r f(z) dz, zw lcr lw-zol < r, where Cr is the circular path Cr(t) = zo + r exp(it), t E [O, 21r], with center zo and radius r > 0 traversed in the anticlockwise direction. • In order to prove this result, we will first prove the following technical fact, which will also prove to be useful later on. 92 A Friendly Approach to Complex Analysis Proposition 3.5. Let (1) D be a domain, zo ED, (2) f: D---+ <C be holomorphic in D \ {zo}, and continuous on D, (3) r > 0 and the disc Ll := {z E <C: lz - zol::::; r} be contained in D. Then f f(z) dz, f(zo)= � 2m lc r z-zo where Cr is the circular path Cr (t) = zo +rexp(it), t E [0,21r], with center zo and radius r > 0 traversed in the anticlockwise direction. Proof. Let E > 0. Then there is a 8 > 0 (which we can arrange to be smaller than r) such that whenever O < !z - zol ::::; 8, lf(z) - f(zo)I < E. Consider the circular path C0 , with center zo and radius 8 traversed in the anticlockwise direction. But C0 and Cr are easily seen to be D \ {zo}­ homotopic. /� ('•·----Qqo ·. i .•' _____\cJ r : '// "�- \_o' Indeed the homotopy H can be obtained by just taking the convex combi­ nation of the points on Cr and C0: H(·, s) := (l-s)Cr(·)+sC0(·), s E [O, l]. Thus by the Cauchy Integral Theorem, we have f f(z) dz= f f(z) dz. lc 0 z-zo Jc r z-zo Hence, I� r 1� r I� / r _l dz JJi)_dz-f(zo)� JJi)_dz-f(zo)I = 21rilcrz-zo 21rilc 8 Z-Zo 21rilc 8 Z-Zo l f(z) - J(zo) = dzl 21rilc 0 z-zo (z) (zo) < max lf - f I). 27r8 - zECo 21rlz - zol ( f < -·27rD=E. 21r8 Since E > 0 was arbitrary, the claim follows. □ 93 Cauchy Integral Theorem The following is an immediate corollary. Corollary 3.4. Let (1) D be a domain, (2) f: D----+ C be holomorphic in D, (3) r > 0, zo ED and the disc Ll := {z EC: lz - zol::; r} c D. 1 Then f(z) 1 . --dz, f(zo) = 2 7fZ Cr Z - Zo where Cr is the circular path Cr (t ) = zo +rexp(it), t E [0,21r], with center zo and radius r > 0 traversed in the anticlockwise direction. We now prove the basic version of the Cauchy Integral Formula, namely Theorem 3.6. Note that as opposed to the previous Corollary 3.4, now the w can be any point inside the circle with center z0 and radius r , and not necessarily the center zo as in the corollary above. Proof. (of Theorem 3.6.) Let w be such that lw - zo I < r . Choose a o > 0 small enough so that the circular path C0 with center w and radius o is contained in the interior of Cr . But now Cr and C0 are D \ { w}-homotopic, and this can be seen in the same manner as in Example 3.8. Since !(·) · -W is holomorphic in D \ {w}, it follows from the Cauchy Integral Theorem that the second equality holds below: 1 1 f(z) dz. f(z) dz=� 21rz c8 Z - W 21rz Cr Z - W This completes the proof. f(w) = � D 94 A Friendly Approach to Complex Analysis Exercise 3.24. Let O < a < l, and let 'Y be the unit circle with center O traversed anticlockwise. Show that { J, Y 21r 1 i { dz= dt. (z-a)(az-l) }0 l+a2-2acost Use Cauchy's Integral Formula to deduce that 1 2 ,r 0 1 dt 1 +a2-2acost 21r = -1-a -- • 2 Exercise 3.25. Fill in the blanks. (l) (2) { exp zdz=---, where 'Y is the circle lzl },y z-1 wise direction. r z +1 : dz J, z -1 = --- ) = 2 traversed in the anticlock­ where 'Y is the circle lz -11 1 traversed in the where 'Y IS · 1e I z -i·1 · the circ . 1 traversed m the anticlockwise direction. (3 ) (4 ) (5 ) z 2 +l _ dz = ___ 1 z2 1 anticlockwise direction. jf , 1 z 2 +l · the circ · 1e I z + 11 = 1 traversed m . the -dz = ___ , where 'Y IS -2 , z -1 anticlockwise direction. z: +l dz = ---1 clockwise direction. { }, z Exercise 3.26. Does z 1--t , where 'Y is the circle lzl 1 2) zl-z ( = 3 traversed in the anti­ have a primitive in {z E (C: 0 < lzl < 1}? Corollary 3.5. {Cauchy's Integral Formula for general paths) Let (1) D be a domain, (2) f: D-+ C be holomorphic in D, (3) zo ED, and (4) 'Y be a closed path in D which is D \ {zo}-homotopic to a circular path C centered at z0, such that C and its interior is contained in D. Then we have f(zo) Proof. =� 27rZ 1 f(z) 'Y Z - Zo dz. 1 By the Cauchy Integral Formula for circular paths, it follows that f(zo) 1 =. 2 11"Z f(z) --dz. CZ - Zo 1 95 Cauchy Integral Theorem 1 But since 'Y is D \ {z0 }-homotopic to C, by the Cauchy Integral Theorem, f(z) dz = _1_ f(z ) dz. _1_ 21ri c z - Zo 21ri 'Y z - Zo This completes the proof. □ This result highlights the "rigidity" associated with holomorphic functions mentioned earlier. By this we mean that their highly structured nature (everywhere locally infinitesimally a rotation followed by a magnification) enables one to pin down their precise behaviour from very limited infor­ mation. That is, even if we know the effect of a holomorphic function in a small portion of the plane (for example the values along a closed path), its values can be inferred at other far away points in a unique manner. The picture below illustrates this in the case of the Cauchy Integral Formula, where knowing the values of f on the curve 'Y enables one to determine the values at all points in the shaded region! exp(iz) . 3.27. Let F be defined by F(z) = Exercise -- and let R > 1. z2 + 1 (1) Let u be the closed semicircular path formed by the segment S of the real axis from - R to R, followed by the circular arc T of radius R in the upper half plane from R to - R. Show that { F(z)dz = �e }" (2) Prove that I exp(iz)I � 1 for z in the upper half plane, and conclude that for large enough lzl, IF(z)I � 2/Jzl2• (3) Show that lim { F(z)dz = 0, and so lim { F(z)dz = �- �ooh �ooh e (4) Conclude, by parameterizing the integral over Sin terms of x, that r cosx dx := lim { cosx dx = �R ➔ oo }_ R l+x 2 J_ oo l+x 2 e oo Exercise 3.28. Evaluate R 1 2 O ,,- ecos 9 cos(sin 0)d0. Hint: Consider exp(exp(i0)). 96 A Friendly Approach to Complex Analysis 3. 7 Holomorphic functions are infinitely differentiable In this section we prove the fundamental property of holomorphic func­ tions in a domain, namely that they are infinitely many times complex differentiable. Let us contrast this with the situation in Real Analysis. We have already seen in Example 0.1 that the derivative may fail to be differentiable at isolated points. There are even more extreme examples of this phenomenon, and there exist functions f : � ➔ � which are differentiable everywhere, but f' is differentiable nowhere! We refer the interested reader to §3.8, [Gelbaum and Olmsted (1964)], where one can find an example of a function g : � ➔ � that is continuous everywhere, but differentiable nowhere; the integral f(x) = fo x g(�)d�, x E �, of this g, then gives our sought for f : � ➔ � that is differentiable every­ where, but whose derivative (g!) is differentiable nowhere. Corollary 3.6. Let (1) D be a domain, and (2) f: D ➔ <C be holomorphic in D. Then f' is holomorphic in D. Note that the above gives the following chain of implications: I f E Hol(D) I⇒I f' E Hol(D) I⇒I f" E Hol(D) I⇒ ··· So whenever f is holomorphic in a domain D (that is, f E Hol(D)), it is infinitely many times complex differentiable. Here is a plan of how we will show this. Formula, we know that where f(z) =� 2m 1 From the Cauchy Integral f(() d , Cr (- Z ( Cr is a circle centered at z with radius r. If we were to formally differentiate under the integral sign, we would get an expression for the derivative of f: f'(z) = _1 21ri 1 f(() d(, Cr ((-z)2 Cauchy Integral Theorem 97 Having shown this formula, we will show that . f'(w)- f'(z) 1Im----w➔z Z-W exists by using the above expression for the derivative at z and w. Let zo E D. Let g be defined by f(z)- f(zo) I·r z ---1r zo, g(z)= { z-zo if z= zo. f'(zo) Clearly g is holomorphic in D \ {zo} and continuous in D. We will now apply the technical fact we had shown in Proposition 3.5 tog. Choose an r > 0 small enough so that the disc {z EC: lz-zol � r} is contained in D, and let Cr denote the circular path with center zo and radius r traversed anticlockwise. Then 1 f g(z) dz . J'(zo) = g(zo) = 2 7rZ lcr z- Zo = _1 f f(z)- f(zo) dz ( 3 .4 ) 21ri lcr (z-zo) 2 1 f(z) dz_ f(zo) { = _1_ f dz 21ri lcr (z-zo) 2 21ri lcr (z - zo) 2 1 f f(z) dz- O · =( 3·5 ) 21ri lcr (z-zo) 2 Thus for w inside Cr, but with w f z0, we have that f(z) dz f(z) = _1 f _1 f J'(w) = 21ri } z-w 2 dz 21ri J z -w ) )2 ' cr ( 00 ( where C0 is a small circular path with center w and radius 8 that lies inside Cr. The second equality above follows from the Cauchy Integral Theorem, because J(-) (· - w)2 is holomorphic in D \ {w} and the paths Cr, C0 are D \ {w}-homotopic. Proof. Cr,___-� 98 A Friendly Approach to Complex Analysis So we have for w -/ zo inside Cr that f'(w) - f'(zo) = _1 _ (-1 f f(z) dz) f(z) dz __ l f 2 21ri lcr (z - zo)2 w - Zo w - zo 21ri lcr (z - w) __1_ f f(z)(2z - Zo - w) dz - 21ri Jcr (z - w)2 (z - zo)2 What does this look like when w � zo? The numerator looks like f(z)(2z - zo - zo) = 2f(z)(z - zo), while the denominator looks like (z - zo)2 (z - zo)2 . So we guess that (w) - f'(zo) = -2_ f f(z) dz lim f' w-tzo w - Zo 21ri lcr (z - zo) 3 ' and we prove this claim below. So let us calculate f'(w) -f'(zo) _ -2_ f f(z) dz= (w _ zo _l_ f (3z -zo -2w)f(z) dz ) w -zo 21ri Jc/z -zo)3 21ri Jcr (z - w) 2 (z -zo)3 Of course if we manage to show that the integral is bounded by some constant for all w close to zo, then we see that since this is being multiplied by w - zo, as w ➔ zo, we can make the overall expression as small as we please. To this end, let us consider a disc with center z0 and having radius smaller than r , say r/2, and we will confine w (which is anyway supposed to be near z0) to lie within this disc. So from now on, w will lie in the (compact set) { w E <C: lw - zo I Consider the continuous map <p, ( ) 'P. (3z - zo - 2w)f(z) z, w H (z - w)2 (z - zo)3 I I : Cr x { w E <C: lw - zol:::; :::; i} . i} (=:KC <C 2 = JR4 ) ➔ R But K is a compact set in JR4 because it is closed and bounded. Thus the continuous function <p: K ➔ JR has some maximum value M � 0 on the compact set K. Hence, f (�z - z) ( 2w)fj:) dzl :::; _!_M(length of Cr )= _!__M21rr = Mr, 21f 21f 21ri jCr Z - W 2 Z - Zo and so f'(w) - f'(zo) - � f(z) dzl :::; lw - zolM r W�o 0. w - zo 21ri Jcr (z - zo) 3 This shows that f' is differentiable at zo. As the choice of zo was arbitrary, f' is holomorphic in D. I� I r □ Exercise 3.29. Suppose f is holomorphic in a domain D. Is it clear that if n EN, then J<n) has a continuous complex derivative? Cauchy Integral Theorem 3.8 99 Liouville's Theorem; Fundamental Theorem of Algebra Here is one more instance of the rigidity associated with holomorphicity. Theorem 3.7. (Liouville's Theorem) Every bounded entire function is constant. Let us again contrast this with the situation in Real Analysis. There we know that for example x H sin x is differentiable everywhere on �, and it is bounded too: I sin xi :::; 1 for all x ER But sin is not a constant function. On the other hand, in light of the above Liouville's Theorem, the entire function z H sin z, being nonconstant, must necessarily be unbounded on (C! We had checked this by brute force using the definition earlier on page 23, and proved that I sin(iy)I ➔ oo as y ➔ ±oo. Proof. Let M;:::: 0 be such that for all z EC, IJ(z)I :::; M. Suppose that w E C, and let I be the circular path with center w and radius R, where R is any positive number. Then (from the proof of Corollary 3.6, see in particular (3.5)) f(z) , 1 2 dz, f (w) = -. 27ri '"Y (z - w ) and so 1 f(z) IJ'(w)I = 1dz < _..!__. M . 21rR = M_ R 21ri '"Y (z-w)2 l - 21r R2 1 1 But since R > 0 was arbitrary, it follows that f'(w) = 0. So f' (w) = 0 for all w EC, and hence f is constant. We had seen this in Exercise 2.13, but here is another way to see this. If z E C, by considering the straight line path 'Yz joining Oto z, we have f(z) - f(0) = This completes the proof. 1 '"Y z J'(()d( = 0. □ This result can be used to give a short proof of the Fundamental Theorem of Algebra2 • Corollary 3.7. (Fundamental Theorem of Algebra) Every polynomial of degree ;:::: 1 has a root in C. 2Despite its name, there is no purely algebraic proof of the theorem, since any proof must use the completeness of the reals, which is not an algebraic concept. Additionally, it is not really "fundamental" for modern algebra; its name was given at a time when the study of algebra was mainly concerned with the solutions of polynomial equations with real or complex coefficients. A Friendly Approach to Complex Analysis 100 For a polynomial p: (C--+ C, given by p(z) =co+ c1z + · · · + cd zd , z EC, where co, c1, · · · , cd EC, Cd -/ 0, the number dis called the degree of p. A number z0 EC such that p(zo) = 0 is called a zero/root of p. Proof. Suppose p(z) =co+ c1z + · · · + cd z d is a polynomial with d � 1, and such that it has no root in C. That is, for all z E C, p(z) -I 0. But since p is entire and nonzero, its reciprocal, namely the function f given by f(z) = l/p(z) (z EC), is entire; see Exercise 2.6. In Exercise 1.24, we had shown an estimate on the growth of a polynomial: there exist M, R > 0 such that jp(z)I � Mlzld whenever lzl > R. In the compact set {z EC : lzl ::::; R}, the continuous function z r-+ jp(z)I has a positive (because pis never 0) minimum m. Thus lf(z)I::::; min{ �d ' M !} , z EC. By Liouville's Theorem, f must be constant, and sop must be a constant, a contradiction to the fact that d � l. For example, the polynomial p given by p(z) = z - 3z guaranteed to have a root somewhere in the complex plane. 1976 28 □ + v'399 is Exercise 3.30. Let f be an entire function such that f is bounded away from 0, that is, there is a 8 > 0 such that for all z E <C, lf(z)I 2: 8. Show that f is a constant. Exercise 3.31. Show that an entire function whose range of values avoids a disc {w E <C: lw - wol < r} must be a constant. Exercise 3.32. Asstime that f is an entire function that is periodic in both the real and in the imaginary direction, that is, there exist Ti, T2 in JR such that f(z) = f(z +Ti)= f(z + iT2) for all z E <C. Prove that f is constant. Exercise 3.33. A classical theme in the theory of entire functions is to try to characterize the entire function f based on the way Iii grows for large lzl. Here is one instance of this. ( 1) Show that if f is entire and lf(z)I � I expzl, for all z E <C, then in fact f is equal to c • expz for some complex constant c with lei � 1. (Thus if a nonconstant entire function "grows" no faster than the exponential function, it is an exponential function.) ( 2) One may be tempted to argue that this can't be right on the grounds that "polynomials grow more slowly than the exponential function", but surely p =f. exp z. Find the flaw in this reasoning by showing that if p is a polynomial satisfying IP( z) I � I exp zI for all z E <C, then p 0. Hint: Look at z = x < 0. = 101 Cauchy Integral Theorem Exercise 3.34. Let f : (C -+ (C be an entire function. Suppose that a1, a2 are complex numbers such that a1 =/ a2, and such that a1, a2 are contained in the interior of the circular path C with radius R > 0 and center 0, traversed once in the counterclockwise direction. Ii z f( ) c ( z - a1)( z - a2) (1) Prove that dz 21rR max IJ(z)lI :S (R - I a1 l)(R - I a2 I) zEC a (2) F ind a, /3 E (C such that for all z E C, -,------,--,---....,.. - -z - a1 (z - a1)(z - a2) (3) Express 1 /3 + -z - a2 · z z z dz in terms of { f( ) dz and { f( ) dz. { f( ) }0 (z-a1)(z-a2) }0 z-a1 }0 z-a2 Use the Cauchy Integral Formula to simplify these latter expressions. (4) Deduce Liouville's Theorem. 3.9 Morera's Theorem: converse to Cauchy's Integral Theorem Recall the Cauchy Integral Theorem, where we have learnt that if (1) (2) (3) (4) Dis a domain, f : D-+ <C is holomorphic, b. is any disc such that b. C D, 'Y is a closed piecewise smooth path in b., then i f(z)dz = 0. Now we will see that the following converse to the above result holds. Theorem 3.8. (Morera's Theorem) If (1) D is a domain, (2) f : D-+ <C is a continuous function such that (3) for every closed rectangular path 'Y in every disc contained in D, i f(z)dz = 0, then f is holomorphic in D. In other words if the contour integral is zero for some special paths, then we are allowed to conclude that f is holomorphic! A Friendly Approach to Complex Analysis 102 Proof. Let zo E D, and let b. be a disc with center z0 such that b. C D, and 'Yzo ,z is the path joining zo to z by first moving horizontally and then moving vertically. Zo 'Yzo, z Define F: b.---+ <C by F(z) = 1 f(()d(, z Eb.. We will show that F is holomorphic in b., and its derivative is f. This shows that f is holomorphic in b.! Why? f (being the derivative of a holomorphic function) is itselfholomorphic in b.. Let z E b.. Suppose E > 0. Since f is continuous, there exists o > 0 such that whenever lw-zl < o and w Eb., it follows that lf(w)-f(z)I < E. We have 1'zo,z F(w) -F(z) = 1 1'zo,w f(()d(-1 '"t'zo,z f(()d(. Using the fact that the integral off on closed rectangular paths is zero, F(w) - F(z) = 1 "Yz,w f(()d(, where 'Yz,w is the path joining z to w by again first moving horizontally and then moving vertically. See the picture below which shows one particular case, and we have (suppressing the integrand): F(w) - F(z) = lz !(()d(- l. .!(()d( = o, 0 = 1 1 1 1 1 B + C + -"'(z,w =0 + -D + l l fc-(l 1J "Yz,w + + f(()d(. + 103 Cauchy Integral Theorem ..... - - ... jj\n � Zo Thus for 0 w F( 2 < lw - zl < 8, = : (z) - f(z) = w �z = w �z 1 1 'Yz,w "Yz,w ':'/z,w D f(()d( - f(z) w � z (!(() - f(z))d(, 1 'Yz,w ld( where we have used the Fundamental Theorem of Contour Integration for the holomorphic function 1 to obtain 1 "Yz,w ld( = w-z. Consequently, from the above, and using the fact that the length of ':'fz,w is we have IRe(w - z)I + IIm(w - z)I 21w - zl, :::; E This completes the proof. 3.10 :::; -1w-z --1 (IRe(w - z)I + IIm(w - z)I) < 2E. □ Notes The proof of Theorem 3.4 follows closely the exposition in [Beck, Marchesi, Pixton, Sabalka (2008)]. Exercises 3.3, 3.4, 3.12, 3.20, 3.27 are taken from [Beck, Marchesi, Pixton, Sabalka (2008)]. Exercises 3.10, 3.14, 3.18, 3.19, 3.22, 3.24 are taken from [Needham (1997)]. Exercise 3.21 is taken from [Rudin (1987)]. Exercise 3.26, 3.33, 3.34 is taken from [Flanigan (1972)]. Exercise 3.28 is taken from [Howie (2003)]. Chapter 4 Taylor and Laurent series In this chapter we will first learn about the fundamental result which says that a holomorphic function f has a power series expansion around any point in the domain D where it lives. See the picture on the left below. L c (z - zo) 00 Taylor series: n n =O n Laurent series: L c (z - zot nEZ n That is, for each z0 E D, there exists an R > 0 such that J(z) = L=O c (z - zot, 00 n lz - zol < R. n Vice versa, every power series n =O that converges for at least two points converges in lz - zol < R for some R and is holomorphic there. En route we will also prove further fundamental results on holomorphic functions: (1) The (general) Cauchy Integral Formula and the Cauchy inequality. (2) The classification of zeros and the Identity Theorem. (3) The Maximum Modulus Theorem. 105 106 A Friendly Approach to Complex Analysis In the second part of the chapter, we will learn about Laurent series, which are like power series, except that negative integer powers of the terms z - zo also occur in the expansion. This will be useful to study functions that are holomorphic in annuli (and in particular punctured discs). See the picture on the right above. They are also useful to classify "singularities" , and to evaluate some real integrals, as we will see at the end of this chapter. 4.1 Series Just like with real series, given a sequence (an)nE J\I of complex numbers, one can form a new sequence (sn)nE J\I of its partial sums: s2 := a1 s3 := a1 + a 2, + a2 + a3, Definition 4.1. = = a := lim S n . (1) The series" a converges if (s n)nE J\I converges, and" L..,,; n L..,,; n n-+ oo n=l n=l = (2) The series Lan diverges if (sn)nE J\I diverges. n=l = = L (3) Lan converges absolutely if the real series Jan i converges. n=l n=l From the result in Exercise 1.25, which says that a complex sequence con­ verges if and only if the sequences of its real and imaginary parts converge, = Lan converges ¢? n=l the real series = = L Re(a ) and L Im(a ) converge. n n n=l n=l Thus the results from real analysis lend themselves for use in testing the convergence of complex series. For example, it is easy to prove the following two facts, which we leave as exercises. 00 Exercise 4.1. If Lan n=l converges, then lim n--+oo an = 0. 00 Exercise 4.2. If Lan n=l 00 converges absolutely, then Lan n=l converges. 107 Taylor and Laurent series Exercise 4.3. Show that if lzl < 1, then f: Exercise 4.5. Show that the series 1-• satisfying Re(s) > 1. Thus s 1-t t n=O Exercise 4.4. Show that if Jzl < 1, then z n converges and that nz -l n + 2-• + f: n=O z n = 1� z. = l � 2• z) ( 3-• +··· converges for all s EC 00 1 ((s) := " �n• n=l is a well-defined map in the half-plane given by Re(s) > 1, and is called the Riemann zeta function. The link of the Riemann zeta function with the number theoretic world of primes is brought out by the Euler Product Formula, which says that if p1 := 2 < p2 := 3 < p3 := 5 < · · · is the infinite list of primes in increasing order, then K 1 IT----., K --+oo k=l 1 - Pk ((s) = lim Re(s) > 1. Bernhard Riemann (1826-1866) showed that the function ( can be extended holomorphically to C \ {1 }. It can be shown that the function ( has zeros at -2, - 4, -6, ..., called "trivial zeros", but it also has other zeros. All the non­ trivial zeros Riemann computed turned out to lie on the line Re(s) = 1/2. This led him to formulate the following conjecture, which is a famous unsolved problem in Mathematics. Conjecture 4.1. (Riemann Hypothesis) zeta function lie on the line Re(s) = 1/2. 4.2 4.2.1 All non-trivial zeros of the Riemann Power series Power series and their region of convergence Let (cn)nE N be a complex sequence (thought of as a sequence of "coeffi­ cients"). An expression of the type 00 LCn Z n n =O is called a power series in the complex variable z. Thus we imagine putting in specific values of z in the above series. Then for some z E C, the power series will converge, while for other values of z it may diverge. 108 A Friendly Approach to Complex Analysis Example 4.1. All polynomial expressions are power series, with only finitely many nonzero coefficients. Polynomials converge for all z E C. The power series 00 n =O converges whenever I z I < 1. It diverges if I z I � 1 since , ( lim z n n--+oo = o). ◊ A fundamental question is: L 00 For what values of z E (C does the power series n =O Cn Z n converge? The following result gives the answer to this question. L 00 Theorem 4.1. For n =O Cn Z n , exactly one of the following hold: (1) Either it is absolutely convergent for all z EC. (2) Or there is a unique nonnegative real number R such that L L 00 (a) n =O Cn Z n is absolutely convergent for all z EC with lzl Cn Z n is divergent for all z EC with lzl 00 (b) n =O < R, and > R. (The unique R > 0 in the above theorem is called the radius of convergence of the power series, and if the power series converges for all z E C, we say that the power series has an infinite radius of convergence, and "'.rite "R = oo".) diverge R zo converge Fig. 4.1 Convergence region of a power series in C. 109 Taylor and Laurent series What happens on the circle lzl = R? Complex power series may diverge at every point on the boundary (given by lzl = R), or diverge on some points of the boundary and converge at other points of the boundary, or converge at all the points on the boundary. There is no general result answering what happens at each point on the circle, and one just has to look at the specific power series at hand to find out the behaviour. Proof. (of Theorem 4.1.) Let S := { y E [0, oo) : 3z EC such that y = lzl and � Cn Z n converges} . Clearly OE S. Only two cases are possible: l S is not bounded above. In this case we will show that the radius of convergence is infinite. Given z EC, there exists a y ES such that lzl < y. But as y ES, there exists a zo EC such that y = lzol and O converges. It follows that its terms tend to 0 as n ➔ oo, and in particular, they are bounded: lcnzol :=::; M. Then with r := lzl/lzol ( < 1), we have lcn zn l L Mr = lcnzol 00 But n =O n converges (r < c��,) n :=:; Mrn (n EN). 1!), and so by the Comparison Test, 00 LCn Z n n =O is absolutely convergent. Since z was arbitrary, the claim follows. 2_0 S is bounded above. In this case, we will show that the radius of con­ vergence is sup S, that is, (a) if lzl (b) if lzl < sup S, L enz 00 then > supS, then n =O n converges absolutely, and 00 Lcnzn diverges. n =O 110 A Friendly Approach to Comple$ Analysis If z E <C and lzl < sup S, then by the definition of supremum, it follows that there exists a y E S such that lzl < y. Then we repeat the proof in 1 ° as follows. Since y E S, there exists a zo E <C such that y = lzol and converges. It follows that its terms tend to O as n ➔ ao, and in particular, they are bounded: lcn zol::; M. Then with r := lzl/lzol (< 1), we have lcn zn l = lcn zol C�'i) n ::; Mrn 00 00 n=O n=O (n EN). But L Mrn converges (r < 1!), and so L Cn Z n is absolutely convergent. Finally, if z E <C and I z I > sup S, then setting y .- I z I, we see that y ¢ S, and by the definition of S, diverges (otherwise y E S). • z The uniqueness of R can be seen as follows. If R, R have the property described in the theorem and R < R, then R+R R<r:=--<R. 2 00 00 □ As r < R, Lcn rn converges. As R < r, Lcn rn diverges, a contradiction. =l � The calculation of the radius of convergence is facilitated in some cases by the following result. 111 Taylor and Laurent series Theorem 4.2. Consider the power series l . Cn + I exists, . If L := 11m I -then n--+oo Cn (1) the radius of convergence is 1/L if L-/=- 0. (2) the radius of convergence is infinite if L = 0. Proof. Let L -/=- 0. We have that for all nonzero z such that lzl that there exists a q < 1 and an N large enough such that lc for all n > N. 1 z: n+ lcnZ +l l I (This is because I c 1 ;: = I Cn + jlzl '-5, q l Cn < l < 1/ L zl n� Llzl < 1. So we may take q = (Llzl 1)/2 < 1.) Thus by the Ratio Test, the power series converges absolutely for such z. + If L = 0, then for any nonzero z E <C, we can guarantee that there exists a q < 1 such that lc +1 z: + 1 1 n = Cn +l jlzl '-5, q < l for all n lcnZ > N. I I (This is because Cn I I n--+oo Cn+l --z --+ 0IzI Cn = 0 < 1. So we may take q = 1/2 < 1.) Thus again by the Ratio Test, the power series converges absolutely for such z. On the other hand, if L -/=- 0 and lzl enough such that lc l + z: l l t CnZ c 1 as I : zl n� Llzl > = > 1/ L, then there exists an N large c l n+l jlzl > 1 for all n > N, Cn 1. By the Ratio Test, the power series diverges. □ 112 A Friendly Approach to Complex Analysis Example 4.2. Consider the power series Lz oo n=l 1 lim (n+ l) 2 n2 • We have = 1' 1 n--+oo n n2 and so the power series converges for lzl < 1 and diverges for lzl > 1. Note that if lzl = 1, then and since L n1 00 n=l 2 converges, it follows that I: oo n=l n :2 converges absolutely. Thus at every point of the circle lzl = 1, the power series converges. We see that this situation is in contrast to the case of the geometric series 00 where we had convergence at no point of the circle Exercise 4.6. Consider the power series izl = 1. L enxn. If L := 00 n=O lim n--+oo ◊ vTcJ exists, then (1) the radius of convergence is 1/ L if L-:/ 0. (2) the radius of convergence is infinite if L = 0. Exercise 4.7. Show that L nnzn converges only when z = 0. 00 n=l zn Exercise 4.8. Show that � 0nn converges for all z E C. n=l Exercise 4.9. Find the radius of convergence of the following complex power series: oo (-l)n n L-n-z' n=l 00 ""' 2012 wn n=O n z ' 113 Taylor and Laurent series 4.2.2 Power series are holomorphic We have seen that polynomials are power series with an infinite radius of convergence, that is, they converge in the whole of <C. They are of course also holomorphic there. This is not a coincidence. Now we will see, more generally, that a power series f(z) := that converges for lzl there holds that L CnZn 00 n=O < R is actually holomorphic there, and for lzl J'(z) = d: (co+ c1z + c 2 z2 + · · ·) = c1 + 2c 2 z + 3c3z 2 + · · · = f n =l < R, Cnnzn-l, (as expected, if one imagines differentiating the series termwise, as we do in the case of finite sums, that is, polynomials). Theorem 4.3. Let R > 0 and f(z) := f'(z) = Proof. L ncnzn-l 00 n =l for lzl L CnZn converge for lzl < R. Then 00 n=O < R. Step 1. First we show that the power series g(z) := L nCnZn-l = C1 + 2c z + · · ·+ nCnZn-l + ... 00 n =l is absolutely convergent for lzl By hypothesis 2 < R. Fix z and let r satisfy lzl < r < R. converges, and so there is some positive number M such that lcn rn l for all n. Let p := lzl/r. Then O ::=; p < 1, and 1 z n -1 Mnpn-1 lncn zn -ll = lcn rn l · - · nl-l :C:::: --r • r r <M L npn-I converges (to 1/(1-p) Exercise 4.4). By the Comparison Test, =l L ncnzn-I converges absolutely. 00 2 n 00 n =l ; A Friendly Approach to Complex Analysis 114 Step 2. Now we show that f'(zo) = g(zo) for lzol < R, that is, lim (f(z) - f(zo) - g(zo)) = 0. z--+zo Z - Zo As before, let r be such that lzol < r < R and since z ---+ zo, we may also restrict z so that lzl < r. Let E > 0. As L nc r -l converges absolutely, there is an N such that 00 n n =l n Keep N fixed. We have f(z) - f(zo)= f f LC ( 00 n =l n z n - zo), and so for z -=I- zo, n f(z) - f(zo) = Cn Z - Zo = Cn (zn -1 + zn -2 ZO z - zo z - zo n =l n =l + ... + Zo- 1). f Thus f(z) - j(zo) Cn (zn -1 + zn -2 ZQ + ... + Zo- 1 -nzo- 1). - g(zo) = Z - ZQ n =l We let S1 be the sum of the first N - 1 terms of this series (that is, from n= l ton= N -1) and S2 be the sum of the remaining terms. Then since izl, izol < r, it follows that !S2 1 s; L ic l ( r -1 + r -1 + ... + r -1 +nr -1) = L 2nlc lr 00 00 n =N Also, S1 = n n n n terms LC N n =l n l n (z - n n n =N n n 1 + zn-2 zo + · · · + zz0- 2 + z0- 1 -nz0-1) < �- 115 Taylor and Laurent series is a polynomial in z and by the algebra of limits, . � Cn (z0n- 1 + z0n-2 zo + · · · + zoz0n-2 + z0n- 1 11m S1 = � N z➔zo - n=l = LC N n=l n (nz�-l - nz�- 1) = 0. nz0n- 1 ) So there is a 8 > 0 such that whenever lz - zol < 8, we have 1S11 < E/2. Thus for lzl < r and 0 < lz - zol < 8, we have I f(z) - f(zo) z-� - g(zo)/ � IS1 I + IS2 I < :_ 2 This means that f'(zo) = g(zo), as claimed. + :_ = E. 2 D Remark 4.1. If (cn )nE N is a sequence of real numbers, then consider the real power series From Real Analysis, we know that such a power series converges in an in­ terval of the form (- R, R) and diverges in JR \ [- R, R] for some R ;:::: 0. The two results in Theorems 4.1 and 4.3 imply that if we replace the real variable x by a complex variable z, then we can "extend/continue" the real power series to a holomorphic function in the disc given by I z I < R in the complex plane. So we can view real analytic functions (namely functions of a real variable having a local power series expansion) as restrictions of holo­ morphic functions. This again highlights the interplay between the worlds of real analysis and complex analysis. (We have seen a previous instance of this interaction when we studied the Cauchy-Riemann equations.) By a repeated application of the previous result, we have the following. Corollary 4.1. Let R > 0 and let f(z) := Then fork;:::: 1, LC Z 00 n n=O L n(n - l)(n - 2) · · · (n -k+ l)c z -k 00 j{k)(z) = . n n=k 1 In particular, for n;:::: 0, Cn = JC n)(0). 1 n. n n converge for lzl for lzl < R. < R. (4.1) 116 A Friendly Approach to Complex Analysis Proof. This is straightforward, and the last claim follows by setting z = 0 in (4.1): f n(n-1) · · · (n - k + l)c n zn -k-ll = k!ck. n=k+l z=0 jCkl(O)=k(k-1) · · · lck + z Also, f (O) = co. □ There is nothing special about taking power series centered at 0. One can also consider Lc n (z-zot, n=0 00 where zo is a fixed complex number. The following results follow immediately from Theorems 4.1 and 4.3. Corollary 4.2. For L C (z - zo) , exactly one of the following hold: 00 n=0 n n (1) Either it is absolutely convergent for all z E C. (2) Or there is a unique nonnegative real number R such that (a) L c (z - zot is absolutely convergent for lz - zol < R, and =0 L c (z - zot is divergent for lz - zol > R. 00 n n 00 (b) n=0 n Corollary 4.3. Let zo E <C, R > 0 and f(z) := for lz-zol jCkl(z) = < R. Then L c (z - zot converge 00 n=0 n L n(n -1) · · · (n - k + l)c (Z - zor-k for lz-zol < R, 00 n=k n n In particular, for n � 0, Cn = .!,f( l(zo). n. k � 1. Remark 4.2. (Uniqueness of coefficients.) Suppose that L C (Z - zot and I:cn(Z - zot 00 00 n=0 n=0 are two power series which both converge to the same function f in an open disk with center z0 and radius R > 0. Then from the above, for n � 0, we have f( n )(zo) = Cn · Cn = n.I n 117 Taylor and Laurent series Exercise 4.10. For lzl < 1, what is 12 + 22 z + 32 z2 + 42z3 +. • •? LC Z . 00 Exercise 4.11. True or false? All statements refer to power series n n n =O (1) The set of points z for which the power series converges equals either the singleton set {O} or some open disc of finite positive radius or the entire complex plane, but no other type of set. (2) If the power series converges for z = 1, then it converges for all z with lzl < 1. (3) If the power series converges for z = 1, then it converges for all z with lzl = 1. (4) If the power series converges for z = 1, then it converges for z = -1. (5) Some power series converge at all points of an open disc with center O of some positive radius, and also at certain points on the boundary of the disc (that is the circle bounding the disc), and at no other points. (6) There are power series that converge on a set of points which is exactly equal to the closed disc given by lzl :S 1. (7) If the power series diverges at z 4.3 = i, then it diverges at z = 1 + i as well. Taylor series We have seen in the last section that complex power series 00 Lcn(z-zot n=O are holomorphic in their region of convergence lz - zol < R, where R is the radius of convergence. In this section, we will show that conversely, if f is holomorphic in the disc lz - zo I < R, then f(z) = L cn(z - zot whenever lz - zol < R, 00 n=O where the coefficients can be determined from the f. Thus every holomorphic function f defined in a domain D possesses a power series expansion in a: disc around any point zo E D. Theorem 4.4. If f is holomorphic in D(zo,R) := {z E (C: lz - zol < R}, then f(z) = co +c1(z - zo) + c2(z - zo)2 + c3(z - zo)3 + · · · for z E D(zo,R), where for n � 0, d( - _1 f f(() Cn - 21ri Jc (( - zo)n+l ' and C is the circular path with center z0 and radius r, where O < r < R traversed in the anticlockwise direction. , -Q _ A Friendly Approach to Complex Analysis 118 ' ' ' ', , : :, ' ' c---,,, '' ' r zo R - -- ............ ' --� ', ,' ' ' ' ' , : ,: ' ,,,." Proof. Let z E D(zo, R). Initially, let r be such that lz-zol < r < R. Then by Cauchy's Integral Formula, f(() d( = _1 f(() d( f(z) = _1 2ni c (-z 2ni c (-zo+ zo-z f(() d( = 2ni c l z-zo ((-zo) ( (-zo) 1 1 �1 z-zo lz-zol Then lwl = -'------'- < . Thus Set w := ---. 1 1 .,, -zo r Wn 1 1 2 3 n 1 ----=--= l+w+w +w +···+w +-­ l z-zo -w -w 1 1 (- Zo z-zo (z-zot (z-zor-1 = +--+···+----+-----1 (-zo ((-zo)n-1(( - z)' ((-zo)n-l and so plugging this in the above, we obtain 1 (z-zot-1 (z-zot d( (() (-1- + ...+ + f(z) = _1 f ((-zo)n((-z)) ((-zo)n (-zo 2ni c where =co+c1(z-zo)+ · · ·+ Cn-1(z-zo)n-l + Rn(z), 1 f(()(z-zor d(. Rn(z) := _1 2ni c ((-zo)n((-z) So we would be done ifwe manage to show that Rn(z) goes to O as n--+ oo. We note that IJI is bounded on the circle, since it is a continuous real valued function on the compact set C, that is, there is an M > 0 such that for all (EC, lf(()I < M. (Here with a slight abuse, we think ofthe path C, and the set ofpoints C(t), t E [O, 2n], as being the same.) Also, for (EC, lz-zol nn (z-zot l ((- nl = ( r ) �O. zo) Taylor and Laurent series 119 But what about 1/1( - zl for ( E C? Is this bounded by something? The picture below shows that indeed this term is bounded by the reciprocal of the "distance between the circle C and z" . We have 1(-zl = 1(-zo-(z-zo)I 2: 1(-zol-lz-zol = r- lz-zol- Thus n M lz - zol n �O. ) IRn (z)l:::;( r -lz - zol r Thus the series co+ c1 (z - zo)+ c2 (z - zo)2 + c3 (z - zo)3 + · · · converges to f(z). Note that we have only shown the expression Cn = _l 1 ( f( ) d 21ri c ((- zo)n +l ( where r is such that Iz-zoI < r < R. But by the Cauchy Integral Theorem, we see that this integral is independent of r, and any value of r E (0, R) can be chosen here: ( ! -� +l is holomorphic in the punctured disc D*(z0, R) given by (1) (• - ZQ n 0 < I z - zo I <__R; (2) besides C, if C is another circular path with center zo and some other radius E (0, R) then C, C are D*(zo, R)-homotopic. r D This completes the proof. But we had learnt earlier in Theorem 4.3 that whenever we have a J(z) = L c (z - zot for lz - zol < R, n=O n we know that JC n )(zo) n! for n 2'. 0. And in the above result, we found different expressions for the coefficients en , given in terms of integrals. But the coefficients of the power Cn = 120 A Friendly Approach to Complex Analysis series expansion are unique in any disc, and so these have to be the same. With this observation in mind, we obtain the following result. Corollary 4.4. (Taylor 1 Series) If (1) D be a domain, (2) f: D -+ (C is holomorphic, and (3) zo ED, then f(z) = f'(zo) f"(zo) - zo) + --, -(z - zo) 2 +... , f(zo) + --(z 1 1. 2. lz - zol < R, where R is the radius of the largest open disk with center z0 contained in D. Also, J< n)(zo) = _2:!_ 1 f(z) 21ri c (z - z0) n +l dz ' (4.2) where C is the circular path with center zo and radius r , where O traversed in the anticlockwise direction. <r<R (4.2) is called the (general) Cauchy Integral formula. Earlier we had just seen the cases when n = 0 (Theorem 3.4) and n = 1 (Proof of Theorem 3.6). Also, we can obtain from here that J< n )(w) = _2:!_ 21ri 1 f(z) dz ' c (z - w) n +l for any point w Eb.:= {z E (C: lz-zol < R}. This follows from the Cauchy Integral Theorem, since we can first consider a small circle C& centered at w for which the above formula holds, and then note that the paths C, C& are b. \ {w}-homotopic, while the function f(-)/(· - w) n +i is holomorphic in b. \ {w}. Proof. f(z) From Theorem 4.4, we have = co+c1(z-zo)+c2(z-zo) 2+c3(z-zo) 3+... , z E D(zo, R), (4.3) where D(zo, R) := {z E (C: lz - zol < R} and R is the radius of the largest open disk with center zo contained in D. Also, for n :::: 0, 1 f(z) _ 1 d Cn - 21ri C (z - zo) n +l z, ------------- 1 Named after Brook Taylor (1685-1731) who, among others, studied this expansion in the context of real analytic functions. 121 Taylor and Laurent series and C is the circular path with center z0 and radius r, where 0 < r < R traversed in the anticlockwise direction. But from Corollary 4.3, the power series above is infinitely many times differentiable, and also, for n � 0, 1 n) JC (zo) = Cn. 1 n. Thus the result follows. □ Summarizing, if f is holomorphic in iz - zol < R, then f(z) = 00 00 1 f(() f( )(zo) n ((z - zo)n = L L +l d( · . _ n 1 (z - zo) , ) 2 ( ni C zo n. ) 1 ( n=0 n=0 n for lz-zol < R, where C(t) = zo+r exp(it), t E [0, 2n], and r is any number such that 0 < r < R. Example 4.3. The exponential function f, z 1---t f(z) := expz, is entire. So we know that we can write exp z = 00 CnZn for z EC, L n=0 for some coefficients Cns. What are these coefficients? They are given by f(n)(zo) , n � 0. Cn = n.1 Since ! exp z = exp z, it follows that j(nl(o) = 1, and so f(z) = f(n)( 0) '°'-(z -Ot = '°' �zn, oo oo � n=0 n! � n! n=0 ◊ for all z EC. Example 4.4. The function f defined by f(z) = Log(z) is holomorphic in C \ (-oo, 0]. The largest open disc with center zo = 1 in this cut plane is D = {z EC: lz - 11 < 1}. Since JCnl(zo) = we have Log(l + w) = w - (-lt (n - 1)! z[j w2 2 +···+ = (-lt(n - 1)!, (-ltwn n + · · · for lwl < 1. Exercise 4.12. Show that for z EC: 3 z sin z = z - 3! 5 - ··· +5! + z z2 and cos z = 1 - 2! 4 z -+··· +4! . ◊ 122 A Friendly Approach to Complex Analysis Exercise 4.13. Find the Taylor series of the polynomial z 6- z4 + z2 - 1 with the point zo =1 taken as the center. Exercise 4.14. Find en 's in the Taylor series (1) f(z) = 1 1'0z LC Z n=O n n around Ooff given by exp((2 )d(, z EC, where ')'oz is the straight line path from Oto z. Hint: J'(z) =exp( z2 ). z2 (2) J(z)= ( z l)2, + z EC\{-1}. Hint: Fori z l<l, 1 z +l =1- z+z2 -+···. Here is a consequence of the general Cauchy Integral Formula. Corollary 4.5. (Cauchy's inequality) If (1) f is holomorphic in D(zo,R) := {z EC: lz - zol < R} and (2) 1/(z)I :SM for all z E D(zo, R), n1M then for n 2'. 0, lf( n ) (zo)I :S �n • Proof. 1� r Let C be the circle with center zo and radius r < R. Then 1/Cnl(zo)I = 21ri f(z) Jc (z- zo)n+l dzl I I n! M n!M n! f(z) <-max----- -21rr= ---21rr= --. - 21r zEC (z- zo) n +l 21r r n +l rn The claim now follows by passing the limit r /' R. □ Exercise 4.15. Suppose that f is an entire function for which there is an M > 0 and an integer n 2: 0 such that for all z E C, IJ( z)I :S Ml z ln - Use Cauchy's inequality to prove that J<n+i) (z) = 0 for all z and show that f is a polynomial of degree at most n. What happens when n =O? Exercise 4.16. Evaluate { s ;�1: dz, where C is the circular path with center 0 la z and radius 1 traversed in the anticlockwise direction. 4.4 Classification of zeros Suppose f : D-+ C is holomorphic in a domain D. We ask the question: What do the zeros of a nonzero f look like? (A point z0 ED is a zero off if f(zo) = 0.) We will learn in this section that the answer to this question is, that the zeros are "isolated". Such a thing doesn't happen with continuous 123 Taylor and Laurent series functions. Zeros of nonzero continuous functions, which aren't as rigid as holomorphic functions needn't be isolated. Example 4.5. (1) expz has no zeros in <C. Indeed, I exp(z)I = e Re(z) > 0 for all z E <C. (2) cos z-3 has infinitely many zeros in <Cat 2nn±ilog(3+2v'2), n E Z, all of which lie on a horizontal line, and they are isolated, with a distance of 2n between any two isolated adjacent zeros. We had seen this in Exercise 1.38. (3) The polynomial p, p(z) = (z + l)3 z 9 (z - 1)9, has zeros at -1, 0, 1. ◊ If p is a nonzero polynomial such that p(zo) = 0, then by the Division Algorithm there exists a polynomial q (the quotient) such that p(z) = (z - zo)q(z) (that is, the remainder is 0). Now we have two possible cases: 1° q(zo)-/= 0. Then zo is an isolated zero of p. 2 ° q(z0) = 0. Then we repeat the above procedure with p replaced by q. Eventually, we obtain p(z) = (z - z0 )m q(z) for some m 2: 1 and q(zo)-/= 0. Then we we call m the multiplicity/order of zo (as a zero of p). We will now see in Theorem 4.1 below that the same sort of a thing holds for holomorphic functions f ( replacing the polynomial p), except that we end up with another holomorphic function (g instead of the polynomial q). This is not completely surprising, since we know that power series are analogues of polynomials, and every holomorphic function has a local power series expansion. But let us first give the following definitions. Definition 4.2. Let D be a domain and f: D--+ <C be holomorphic in D. A point zo ED is called a zero off if f(zo) = 0. If there is a smallest m E N such that (1) f(m )(zo)-/= 0 and (2) f(zo) = · · · = f(m-l)(zo) = 0, then z0 is said to be a zero of f of order m. that f(o) := f.) (We adopt the convention We have the following result on the classification of zeros of a holomorphic function. 124 A Friendly Approach to Complex Analysis Proposition 4. 1. ( Classification of zeros) Let (1) D be a domain, (2) f: D-+ <C be holomorphic in D and (3) zo ED be a zero of f. Then there are exactly two possibilities: 1° There is a positive R such that f(z)=0 for all z satisfying lz-zol < R. 2_0 There exists an m E N such that zo is a zero off oforder m, and there exists a holomorphic function g: D-+ <C such that g(zo)-/= 0 and f(z)=(z - zo)m g(z) for all z ED. We note that ifwe are in case 2 ° , then since g is continuous and as g(zo)-/= 0, gis not zero in a small disc b. centered at z0, and so f is nonzero in b.\ {0} using the fact that for z Eb.\ {zo}, f(z)=(z - zorg(z)-/= 0. Thus zo is the only zero of f in b., that is, z0 is isolated. Also, we will soon learn about something called the Identity Theorem, and from that result, it will follow that in fact in case 1 °, we can conclude that f 0 in the whole of the domain D. Proof. = We have a power series expansion for f in a disc with some radius R > 0 and center zo: f(z)=co+ c1(z - zo)+ c2 (z - zo) 2 + · · · for lz - zol < R. Since f(zo)=0, we know that co=0. Now there are exactly two possibil­ ities: 1° All the Cn are zero. Then f(z)=0 whenever lz - zol < R. 2_0 There is a smallest m ;:::: 1 such that Cm -/= 0. Then we have that co=c1 = · · · =Cm -1=0, and so using the fact that f( n l(zo) , n;:::: 0, Cn = n.1 it follows that zo is a zero of order m. Moreover, from the power series expansion, we have + Cm +k(z-zot f(z)=Cm(z-zor+cm +l(z-zo)m l+. · · =(z-zor I: (X) k=O for lz - zol < R. Thus, if we define g: D-+ <C by f(z) for z -/= zo, z ( - zo ) m g(z)= { k � Cm +k(z - zo) for lz - zol < R, (X) (4.4) 125 Taylor and Laurent series From (4.4), the two definitions give the same value whenever both are applicable. So g is well-defined. Moreover, we have: is holomorphic in D. For z =/- z0, this follows by observing that both f and 1/(· -zor are holomorphic in D\ {O}. For lz -zol < R, this follows since g is given by a power series! (2) g(zo) = Cm =/- 0. (Definition of m!) (3) f(z) = (z -zo) m g(z) for z =/- D\ {zo} for z ED\ {z0} follows from (4.4). On the other hand if z = z0, then both sides are zero. Thus for all z E D, f(z) = (z -zo)m g(z). (4) z0 is a zero of order m. Indeed, Cn = fn (z0 )/n! for all n's, and so the claim follows using Cm =/- 0, while co = c1 = · · · = Cm-1 = 0. (1) g □ This completes the proof. Example 4.6. (1) mr is a zero of sin z for each n E Z. We know that sin z is not identically zero in any neighbourhood of mr by just looking at the restriction of sin z to the real axis. We ask: what is the order of mr as a zero of sin z? Since sin' z = cos z and cos zl z =mr = (-l) n =/- 0, it follows that mr is a zero of sin z of order 1. (2) exp(z2 ) -1 has a zero at O since exp(02 ) - 1 = 1 - 1 = 0. What is its order? We have exp(z 2 ) = 1 + z2 z4 - + - + · · · ' z E <C, 1! 2! 1 z2 and so exp(z 2 ) - 1 = z 2 g (z) (z E q, where g(z) := + 1 + · · · 1 1. 2. g is given by a power series that converges in <C, and so g is entire. Also, g(O) = 1 =/- 0. Thus the order of O as a zero of exp(z 2 ) - 1 is 2. Alternately, we could have observed that 2 2 d� (exp(z ) -1{= = (exp(z )) · 2zl =O 0 z = 0, ::2 (exp(z 2 ) -1{= = (exp(z 2 )) · 2 + (exp(z 2 )) · (2z) 2 1 =O 0 z = 2 =/- 0, and so the order of O as a zero of exp(z 2 ) - 1 is 2. ◊ Exercise 4.17. Let D be a domain, m E N, R > 0 and zo E D. Let f, g : D ➔ <C be m holomorphic such that g(zo) -/= 0 and whenever lz-zol < R, J(z) = (z-zo) g(zo). Prove that zo is a zero of f of order m. 126 A Friendly Approach to Complex Analysis Exercise 4.18. Find the order of the zero zo for the function fin each case: (1) zo = i and f(z) = (1 + z2 ) 4 • (2) zo = 2mri, where n is an integer, and f(z) = expz -1. (3) zo = 0 and f(z) = cosz -1 + �(sinz) 2 . Exercise 4.19. Let fbe holomorphic in a disc that contains a circle, in its interior. Suppose there is exactly one zero zo of order 1 off, which lies in the interior of ,. Prove that 1 zj'(z) zo = dz. 27ri , J(z) 1 Exercise 4.20. Let D be a domain and f be holomorphic in D such that f has a zero of order m > 1 at zo ED. Prove that the function z >-+ (f(z))2 has a zero of order 2m at zo, and that J' has a zero of order m - 1 at zo. Exercise 4.21. A complex valued continuous function on a domain needn't exhibit the dichotomy of behaviour of its zeros as in Theorem 4.1 for holomorphic func­ tions. Give an example of a function f: <C ➔ <C, such that O is neither an isolated zero nor is it the case that fis identically zero in a small disc around O centered at 0. 4.5 The Identity Theorem In this section, we will learn the Identity Theorem, which once again high­ lights the rigidity of holomorphic functions. It says roughly that a nonzero holomorphic function cannot have an accumulation of zeros in its domain. Theorem 4.5. Let (1) D be a domain, (2) f: D ➔ <C be a holomorphic function in D, (3) (zn) nE N be a sequence of distinct zeros off which converges to z* ED. Thenf is identically zero in D. Proof. First let us note that z* is itself a zero of f because by the con­ tinuity off, we have f(z*) = = f ( n➔oo lim Zn ) = lim f(zn) = lim 0 = 0. n➔oo n➔oo We claim thatf 0 in some disc� with center z* and radius r > 0. If not, then z* is a zero of some order m, and f(z) = (z -z*) m g(z) for z in�, and Taylor and Laurent series 127 g(z) -=f. 0 in �- But then for all large n, 0 = f(zn ) = (zn - z*) m g(zn ) -=f. 0, a contradiction. Next we will show that f 0 in all of D. Suppose that f(w) -=f. 0 for some w E D. Then there is a path, : [O, 1] ---+ D that joins z* tow: 1(0) = z* and 1(1) = w. Let S := {t E [0, 1] : f(,(T)) = 0 for 0:::; T:::; t}, and T := supS. We note that supS exists since S is nonempty (0 belongs to S!) and S is bounded above (by 1). W hat is T? If we think of t as time as we travel from z* (time t = 0) tow (time t = 1) along,, then T is the largest time such that f has been O along the path covered so far. If T = 1, then we are done (since then by continuity, f(w) = f(T) = 0). So let us suppose that T < 1. Then f(,(T)) = 0, by continuity. But then f(z) = 0 for z's in a disc around,(T) of radius say 8, since,(T) can't be an isolated zero! This implies that f(,(t)) = 0 for t's that are bigger than T (because for all t's close enough to T, l,(t) -,(T)I < r). = \ ,(T) / '' ,,' This contradicts the definition of T. Thus T can't be less than 1. This D completes the proof. Example 4.7. We had shown that for all z E re, (cosz)2 + (sinz)2 = 1 using the definitions of cosz and sinz. Here is a proof using the above result. Consider f: re➔ re, f(z) = (cos z) 2 + (sinz)2 -1, z Ere. Then f is entire. Also, for all x E �, f( x) = (cosx)2 +(sinx)2 -1 = 0 by Pythagoras's Theorem. So by the result above, f 0 in re! ◊ = The following is an immediate consequence of this result. Corollary 4.6. (Identity Theorem) Let (1) D be a domain, (2) f, g: D ➔ re be holomorphic in D, (3) (zn ) n EN be a sequence of distinct points in D which converges to z* E D, and such that for all n EN, f(zn ) = g(zn )Then f(z) = g(z) for all z ED. Proof. Define h : D ---+ re by h(z) = f(z) -g(z), z E D, and note that the ZnS are zeros of the holomorphic function h. By the result above, h 128 A Friendly Approach to Complex Analysis □ must be identically zero in D, and so the claim follows. Example 4.8. We know that exp: <C---+ <C defined by expz = exp(x + iy):= ex(cosy + i siny), z=x + iy E <C, is an entire function such that exp x = ex for x E IR. In other words, exp is an entire extension of the usual real exponential function. Is there any other entire extension possible? We show that the answer is no! Suppose that g: <C---+ <C is entire and g(x) = ex for real x. But then expx = g(x) for all x ER In particular, and 1/n ➔ 0 E <C. So by the Identity Theorem, expz = g(z) for all z E <C. So there is only one entire function whose restriction to IR is e x . This explains the naturalness of the definition of the complex exponential in Section 1.4.1. ◊ Exercise 4.22. Show, using the Identity Theorem, that for all z1, z2 EC, cos(z1 + z2) = (cosz1)(cosz2) - (sinzi)(sinz2), by appealing to the corresponding identity when z1, z2 are real numbers. Exercise 4.23. Let D be a domain and let Hol(D) be the set of all functions holomorphic in D. Then it is easy to check that Hol(D) is a commutative ring with the pointwise operations (f + g)(z) (f · g)(z) = f(z) + g(z), = f(z)g(z), for z E D and f,g E Hol(D). (By a commutative ring R, we mean a set R with two laws of composition + and· such that (R, +) is an Abelian group, · is associative, commutative and has an identity, and the distributive law holds: for a,b,c E R, (a+ b)· c = a· c + b· c.) Check that Hol(D) is an integral domain, that is, a nonzero ring having no zero divisors. In other words, if f • g =0 for f,g E Hol(D), then either f =0 or g =0. If instead of Hol(D), we consider the set C(D) of all complex-valued con­ tinuous functions on D, then C(D) is again a commutative ring with pointwise operations. Is C(D) an integral domain? (This shows that continuous functions are not as "rigid" as holomorphic functions.) 129 Taylor and Laurent series Exercise 4.24. Let f, g be holomorphic functions in a domain D. Which of the following conditions imply f = g identically in D? (1) There is a sequence (zn ) nE N of distinct points in D such that J(zn ) = g(zn) for all n EN. (2) There is a convergent sequence (zn ) nE N of distinct points in D with its limit in D such that f(zn ) = g(zn) for all n EN. (3) 'Y is a smooth path in D joining distinct points a, b E D and f (4) w ED is such that J <nl(w) = g <n)(w) = g on 'Y­ for all n 2: 0. Exercise 4.25. Suppose that f is an entire function, and that in every power series (that is, for every zo E q f(z) Prove that f is a polynomial. 4.6 = L c (z - zot, at least one coefficient is 0. n=O n The Maximum Modulus Theorem In this section, we prove an important result, known as the Maximum Modulus Theorem which says that for a nonconstant holomorphic function f : D ➔ C, Ill can't have a maximizer in the domain D. Theorem 4.6. (Maximum Modulus Theorem) Let (l) D be a domain, (2) f: D ➔ (C be holomorphic in D, (3) z0 ED be such that for a ll z ED, lf(zo)I 2:: lf(z)IThen f is constant on D. Proof. Let r > 0 be such that the disc with center z0 and radius 2r is contained in D. Let Cr be the circular path Cr(t) = z0 + rexp(it), t E [O, 21r]. Then by the Cauchy Integral Formula, 1 !(zo ) = 21ri 1 =27r 1 f( l 1 -dz = z) z - zo 21ri 271' ---'-----'---'-'-irexp ' f(zo +rexp(it)). (it )dt 1271' f(zo + rexp(it))dt, 7 0 O rexp(it) 130 A Friendly Approach to Complex Analysis where the last expression can be viewed as the "average" of the values of f on Cr. Since lf(zo + rexp(it))I::; lf(zo)I for all t, the above yields lf(zo)I = 1 � 2 1 f(zo + rexp(it))dtl ::; � 2 211: 211: 1211:lf(zo + rexp(it))ldt = fz 1 ::; � l ( o)ldt lf(zo)I2 So in the above, all the inequalities ::; have to be equalities, and by rear­ ranging, we get 1211: ( lf(zo)l - lf(zo + rexp(it))I)dt = 0. ;:,:o But the integrand is pointwise nonnegative, and so in light of the above, we can conclude that lf(zo + rexp(it))I = lf(zo)I for all t. But by replacing r by any smaller number, the same conclusion would hold. Thus f maps the disc�:= {z E <C: lz-zol ::; r} into the circle {w E <C: lwl = lf(zo)I}- This implies by Example 2.11 that f is constant on �- The Identity Theorem now implies that f must be constant on the whole of D. 2� □ Example 4.9. Let lHI := {z E <C : Re(z) :2 O} denote the right half plane, and consider f : lHI --+ <C given by e ) f(z) = ;l-/ , z E lHI. Then it can be shown that llfll= := max lf(z)I zEIH[ exists. Without worrying about the existence of this maximum, let us instead see how, assuming its existence, the Maximum Modulus Theorem enables us to calculate its value. Suppose that zo E lHI is a maximizer. Then this maximizer z0 can't have a real part which is positive, by the Maximum Modulus Theorem. So zo E iffi?., that is, z0 = iy0 for some Yo ER But 1 e p( iy) , y ER lf(iy)I = l �i - I= Jy + 1 y+1 1 = 1. ◊ Thus llfll= = max lf(z)I = maxlf(iy)I = max zEIH[ yEIR yEIR l J0 +1 + y 2 H+i 2 2 Exercise 4.26. Let D be a domain and let f: D ➔ <C be a nonconstant holomor­ phic map. Prove that there is no maximizer for the map z >-+ IJ(z)J on D. Exercise 4.27. (Minimum Modulus Theorem) Let D be a domain and let f: D ➔ <C be holomorphic in D. Suppose that there is a zo ED such that for all z ED, IJ(zo)I :S: Jf(z)J. Then prove that either f(zo) = 0 or f is constant on D. Exercise 4.28. Consider the function f defined by mum and minimum value of Jf(z)J on {z E <C: JzJ f(z) = z 2 :S: 1}. - 2. Find the maxi­ 131 Taylor and Laurent series 4. 7 Laurent series Laurent series generalize Taylor series Indeed, while a Taylor series . I:cn (z-zor n=O has nongenative powers of the term z-z0, and converges in a disc, a Laurent series is an expression of the type Lcn (Z-zor nEZ = · · · +C- 1 (z-Zo)-l +Co+C1 (z-zo) 1 + · · · , which has negative powers of z - z0 too . We will see that (1) Laurent series "converge" in an annulus {z E <C : r < lz - zol < R} with center z0 and gives a holomorphic function there, and (2) conversely, if we have a holomorphic function in an annulus with center z0 and it has singularities that lie in the "hole" inside the annulus, then the function has a Laurent series expansion in the annulus For . example, we know that for all z E <C, z z 2 z3 + + +· , l! 2! 3! ·· and so for z =/- 0, we have the "Laurent series expansion" exp z = 1+ exp 1 1 1 1 1 1 = 1+ + + + � 2! z2 3! 3! ... . � Note that exp(l/z) is holomorphic in <C \ {O}, which is a degenerate annulus centered at O with inner radius r = 0 and outer radius R = +oo! Let us first define what we mean by the convergence of L c (z-z ) . nEZ n 0 n 132 A Friendly Approach to Complex Analysis Definition 4.3. The Laurent series L c (z - z0) nEZ L C- (z - zo)00 If n=l n L e (z - zo) nEZ n n n n n converges (for z) if converges and L e (z - zor converges. 00 n=O n converges, then we write nEZ 00 00 n=l n=O and call this the sum of the Laurent series. Example 4.10. For what z E <C does the Laurent series 1 4z 1 8z ... + _ _3 + _ _2 + _!_ + 1 + z + z 2 + z 3 + ... converge? We have: 2z (1) 1 + z + z 2 + z 3 + · · · converges for zl < 1, and it diverges for lzl > 1. 1 1 < 1 and diverges for 1 � I > 1, + :3 + :5 + · · · converges for (2) 2 2zI 2z 4 8 that is, it converges for lzl > 1/2 and diverges for lzl < 1/2. Hence the given Laurent series converges when lzl < 1 and lzl > 1/2, that is, it converges inside the annulus {z E <C: 1/2 < lzl < 1}, and it diverges when lzl > 1 or when lzl < 1/2. ◊ For what z does L C (Z - zor converge? nEZ n (1) From Theorem 4.1, for L C (Z - zor, there is some R such that it 00 n =O n converges for lz - zol <Rand diverges for lz - zol > R. (2) What about the series L c_ 00 n (z - zo)- n ? The power series n=l also converges for lwl <Rand diverges for lwl > Then L C- (z - zo)- R. L C- W 00 n n=l Set w := (z-zo)-1. n converges when 1/lz - zol < R, that is for n=l lz - zol > 1/.R =: r, and diverges for lz - zol < r. 00 n n 133 Taylor and Laurent series Hence the Laurent series converges in the annulus {z Ere: r and diverges if either lz - zol < r or lz - zol > R. < lz-zol < R} Is it holomorphic in the annulus where it converges? L c (z - zot is holomorphic in {z Ere: lz - zol < R} and so in 00 (1) z I-+ n n=O particular, also in {z Ere: r < lz - zol < R}. (2) The map 00 n � 9 W i-----=--t �C-nW n=l is holomorphic in {w E re : lwl < .R}. Also we see that the mapping z � (z - z0)-1 : re\ {zo} -+ re is holomorphic. So their composition go f is holomorphic in {z Ere: lz - zol > r}, that is, 00 )-n g f � zi-=---:+ �C-n (z - zo o n=l is holomorphic in {z E re : r < lz - zol}, and so also in particular in the annulus {z Ere: r < lz - zol < R}. L + L c_ 00 00 Hence the sum z I-+ en (z - zot n (z - zo)- n=l is holomorphic in the annulus {z Ere: r < lz - zol n=O n = < R}. L c (z - zot nEZ n Summarizing, we have learnt that any Laurent series L C (z - zo) nEZ n 2 converges in an annulus {z Ere: r map z I-+ n < lz - zol < R} for some r, R, and the L C (z - zot n nEZ is holomorphic in {z Ere: r < lz - zol < R}. That conversely, a function holomorphic in an annulus has a Laurent series expansion is the content of the following theorem. 2which may be empty! 134 A Friendly Approach to Complex Analysis Theorem 4.7. If f is holomorphic in A:= {z E (C: r then f(z) = where (1) Cn l . =2 1f'I, 1 C L C (z - zo) nEZ n f(() , (( - zo)n+l d( (2) C is the circular path given by C(t) (3) p is any number such that r n < lz -z0/ for z EA, = z0 + pexp(it), < p < R. < R}, (4.5) t E [O, 21r], Moreover, the coefficients are unique in (4.5). Example 4.11. Define f : (C \ {O} --+ (C by f (z) = z3 exp(l/z), z =/- 0. Then f is holomorphic in A:= {z E (C: 0 < /zl < +oo}. So by the above result, f must have a Laurent expansion LCn Z n . nEZ In this case we can find the coefficients just by inspection: since 1 1 1 exp; = 1 + ; + + · · · , z =/- 0, 21z2 2 z l 1 3 1 3 we have f(z) = z exp; z + z + ! + ! + ! + · · ·, for z =/- 0. 4z 2 3 Consequently, 1 1 1 · · · , c_1 = , co = , c1 = , c2 = 1, c3 = 1, 1 1 2. 41. 3. and Cn = 0 for n 2: 4. ◊ Proof. (of Theorem 4.7.) (Existence.) Fix z EA. Chooser and R such that r <r < lz -zol < R < R. Let ,1 and ,2 be the circular paths + rexp(it), ,2(t) = zo + Rexp(it), and 0:= Arg(z) + 1r/2. Let 13 ')'1(t) = fort E [0, 21r + 0], ZQ . z -zo . ,3 (t) = ti I z -zo I : [r, R] --+A be the path (This is just a straight line path joining 11 and ,2, and the peculiar mul­ tiplication by i produces a rotation by 90 ° , ensuring that this path avoids z!) See Figure 4.2. 135 Taylor and Laurent series Fig. 4.2 Laurent series. Clearly the path 'Y := "/2 -"/3 - "(1 + "/3 is A \ {z}-homotopic to a small () circle C8 centered at z. Also, f - is holomorphic in A\ {z}, and so . -z 1 ( ( f( ) d( = f f( ) d( = (z) · 21ri, f z z ( ( J c 0 7 where the first equality follows from the Cauchy Integral Theorem, and the second equality follows from the Cauchy Integral Formula. Thus since the contour integral along "/3 cancels with that along --y3, we obtain 1 ( !( ) d( = � (z) = � f 27ri "( ( - Z 27ri = 1 - fl - 1 72 /73 ( ( f( ) d(-� f( ) d(. 2m 171 ( - z 2m 172 (-z � "fl + f/ f(() _ /73 (-Z d( (II) (I) We will see below that the integral (I) gives the term (X) Lcn(z-zot, n=O while the integral (II) gives the term n L C-n(z-zo)- , (X) n=l and these put together we will yield the desired Laurent series expansion of . f 136 A Friendly Approach to Complex Analysis 1 Step 1. In this step we will show that - . 21ri We have for ( E "/2 that !(() ( -z !(() ( -zo + zo-z 1 f�) f "/2 <, Z d( = n=O cn(z-z0 )n. !(() !(() ((-zo)( 1-w) ' z-zo ((-zo) (1---) (- Zo z-zo lz-zol lz-zol where w := = -- < 1, and so --· We have lwl = r -zo -zo I( ., R I n W l 2 3 n-1 --=l + w + w + w + ···+ w --. + l-w l -w Using this, we obtain f(() = f(() (1 + w + ···+ wn+ _:!!:_) l-w ( -zo ( z-zo = !(() + ...+ !(() n(z _ zor-1 + f ( )( n t . ((-zo) (-zo ((-zo) ( ( -z) ( -z �1 Thus f(() d( 21ri "12 (-z where =�1 �1 f(() d( - (z-zot-1 f(() d( + ...+ 21ri "12 ((-zo) n 21ri "12 (-zo l n f(() . + 21ri "12 (( - zo)n(( -z) d( (z-zo) =Co+ c1(z-zo) + ··· + Cn-1(z-zor-1 + Rn(z), 1 1 Rn(z) := _ 21ri 1 f( ( )(z-zot d(. zo)n( ( -z) Here we have used the fact that "/2 is A-homotopic to any circle C with center z0 and radius p where r < p < R, and so by Theorem 3.4, 1 "12 ((- 1 !(() d !(() d ( ( zo)k = c ((-zo)k fork=l, ..., n -l. We would be done and obtain "12 ((- �1 21ri f(() "12 ( - Z d( = f n=O = Ck Cn(z-zot if we show that lim Rn = 0. There is an M > 0 such that for all ( E "/2, n-+oo lf(()I < M. Moreover, I(-zol =Rand I(-zl = 1(-zo-(z-zo)I 2: I(-zol - lz-zol = R - lz-zol, 137 Taylor and Laurent series and so n M IR (z)I::; lz - zol � O. n ) ( R-lz - zol R n 1 3 2 Thus co+ c 1(z - zo)+ c2 (z - zo) + c3 (z -zo) +· ·· = -. 1ri2 1 1 �1 l. Step 2. We will show that-21ri We have __l 21ri 1 f(() d( z '°Yl (- = ( -zo Set w :=--. Then lwl = z -zo 1 1 L (() d ', 12 z (. f(() oo C- n (z - zo)-n. -r- d( = z n=l '°Yl ', - = _1 1f f(() d( z - Zo) - (( - Zo) 21ri '°Yl ( 21ri '°Y1 f(() (z - zo) (1- r < .1 Thus z zo I I ( -zo Z - Zo) d( W n ----=--=l+w+w2 +w3 +···+wn-1 +--, l and so l 27ri _ ( - zo z - zo 1 f(() d( = Z 1 �1 1 11 ( - = l-w 1- w 1 ( (-zo)n ..+ ((-zor- + (-1 · + ) d( f(() z -zo (z-zo)n (z-() (z-zo)n 11 1 l d(· f(() d ·-- +... + 21ri 11 ((-zo)-n+l 21ri 11 f(() ( z-zo (z-zo)n _1 21ri -l + 21ri 11 /(()((- zo) d ( (z- zo)n (z - () �1 n = C-1(z - zo)-1+···+ C-( n -l)(z - zo)-(n -l)+ Rn (z), 1 where !(()(( - zor d . Rn (z) := _1 ( 21ri 11 (z - zo)n (z - () Here we have used the fact that 'Yl is A-homotopic to any circle C with center z0 and radius p where r < p < R, and so by Theorem 3.4, 1 1 f(() d !(() d = C k ( = c ((- zo)k ( zo)k '°Y1 ((- A Friendly Approach to Complex Analysis 138 for k =-l, ..., -(n- 1). There is an M > 0 such that for all ( E 'Yi, lf(()I < M. Moreover, 1(- zol =rand iz-(I= l(z-zo )- ((-zo)I 2:: lz-zol-I(-zol= lz-zol - _ and so ( IRn(z)I::; r n Iz-zo I) M --+<X> n --+ Iz-zo I-r o. Thus C-1(z-zo)-1 + · · · + C-(n-I)(z-zo )-( n-l) + · · ·=-� 211'i 1 r, f(() d . ( z "/1 ( - This completes the proof of the part on the existence of the Laurent ex­ pansion. Uniqueness of coefficients. Cauchy's Integral Formula allows us to show that the Laurent expansion is unique, that is, if f (z)= 2:cn(Z-zo r for r < lz-zol < R, nEZ d (z-zo )n+l then for all n, Cn = Cn , If n-/- l, then (z-zo )n= - ( ----). So n+ 1 dz l (z-zo rdz = 0 (n-/- 1), where C is given by C(t)= zo +pexp(it), t E [O, 211']. By a direct calculation, f _l_ dz= f lo lc z-zo 21r l ' irexp(it)dt= 211'i. rexp(it) Hence, if term-by-term integration is justified in the annulus, we would have {e: (z-zo ) lC n -m-1 �� Cn(z-Zo ) dz nEZ � Cn l{e: (z-Zo )n-m-l dz = 211'iC·-m , = �nEZ C and the claim about the uniqueness of coefficients would be proved. This term-by-term integration can be justified as follows. We have ( � Cn(z-zo )n-m-l = .. · �- nEZ + Cm-2 - (z-zo) 3 + Cm-1 - (z-zo )2 ) + (cm+l + Cm+2(z-zo ) + · · ·) +� z-z o Cm = fi(z) +-+ fz(z). Z-Zo 139 Taylor and Laurent series We need only show that Ji, h have a primitive in the annulus and then l { Lcn(Z -zor-m- dz jC nEZ = r j C (1i(z) + Cm Z -Zo + h(z)) dz = 0 + 21ricm + 0 = 21ricm, 00 as required. But h(z) = LCm+n (z -zor-1 for lz -zol < R, and so if n=l L Cm+ (z -zot 00 F2(z) := - n n=l n for lz -zol < R, then : F2(z) = h(z), and so His a primitive of /2. For Ji, we have z 00 - 00 � -Cm nWn+l, _ � _Cm-nn+l _ -� f1 (z) � (z zo) where w 1 z -zo = - -, and this is valid for R > lz -zol > r. So 1 converges for lwl < -. If we set r then ! G(w) =- f 1 for lwl < -, r n l Cm-n w - . Hence if define Fi by n=l f 1 Cm-n (z -zo)-n for z EA, F1 (z) = G (--) = z -zo n n=l then 00 = (z -zo)-2 LCm n (z -zo)-n+l = fi(z). n=l □ 140 A Friendly Approach to Complex Analysis Note that the uniqueness of coefficients is valid only if we consider a par­ ticular fixed annulus. It can happen that the same function has different Laurent expansions, but valid in different annuli, as shown in the following example. Example 4.12. Consider Fig. 4.3 f defined IC\ {O, 1}, A:= {z E IC: 0 by f(z) < lzl < = 1 zz-l ) ( 1} and A:= {z E IC: 1 f is holomorphic in the annulus A:= {z EC : 0 < Laurent expansion there? Since lzl < 1, we have 1 f(z) = =- 1 +z + z2 + z3 + · · · ) = - !( z(z-l) z and so the Laurent series coefficients are given by C-2 =C_ 3 C-1 =Co = ··· = C1 , z EC\ {0, 1}. = lzl < < lzl < +oo}. 1}. What is its ! - 1-z- z -z 2 z 3 -••• , 0, = · · · = -1. But f is also holomorphic in the annulus A:= {z EC: 1 < lzl}, with inner radius 1, and outer radius infinite. So f has a Laurent series expansion in A too! Since lzl > 1, we have l = ! l = 2_ 1 + + 2_ z z2 z2 ( z2 (1-1/z) 1 1 = 2 + 3 + z z .... Thus now the Laurent series coefficients are given by f (z) = z(z-l) + 2_3 + · · ·) z c_z = c_ 3 = ... = 1, C-1 =Co = C1 = · · · = 0. So we notice that the coefficients are different, but this is not surprising, since the annuli we considered for the Laurent expansions were different too. ◊ Exercise 4.29. Reconsider Example 4.12, and now find Laurent expansions for f also in the annuli A1 := {z E IC: 0 < lz-11 < 1} and A1 := {z E IC: 1 < lz-11}. 141 Taylor and Laurent series 4.8 Classification of singularities If we look at the three functions sinz z 1 z3 ' 1 exp-, z then we notice that each of them is not defined at 0, and refer to O as a "sin­ gularity" of these functions, because the function is not defined there. But we will see that each of these functions behave very differently near their common singularity. In other words, the "nature of the singularity" differs in each case. We will explain precisely how the behaviour is different in each case, and this is what we mean by classification of singularities. More­ over, we will learn two results, which will allow us to find out the type of singularity at hand. Of these two characterization results for singularities, one result will be in terms of limits, while the other will be in terms of what happens with Laurent coefficients. We first give the following definition. Definition 4.4. Let f be a complex valued function which is not defined at a point z0, and suppose that it is holomorphic in some punctured disc {z E <C: 0 < lz - zol < R} centered at zo with some radius R > 0. Then we call zo an isolated singularity of f. Example 4.13. For example, each of the functions sinz z 1 ' z3 1 exp-, z has an isolated singularity at 0. On the other hand, f given by f(z) := 1 sin (l/z) has a singularity at 0, but it is not an isolated singularity. (At z =l/mr, n E Z, the function f is not defined.) ◊ Definition 4.5. An isolated singularity z0 off is called (1) a removable singularity off if there is a function F, holomorphic in the disc {z E <C: lz - zol < R} such that F =fin the punctured disc {z E <C: 0 < lz - zol < R}. ( )I = +oo, that is, 2 ( ) a pole offif lim lfz z➔zo for all M > 0 there is a 8 > 0 such that whenever O < lz - zol ( )I > M. lfz 3 ( ) an essential singularity off if z0 is neither removable nor a pole. < 8, 142 A Friendly Approach to Complex Analysis Example 4.14. (1) The function f given by f(z) since for z =/=- 0, we have = sin z has a removable singularity at 0, z l sin z z3 z 5 -- =; ( z - ! + 3 z 5! - + · · · ) =]; 00 t ( -1 2n z 2n ' ( + l )! and the right hand side, being a power series with an infinite radius of convergence (why?) defines an entire function F. Since this entire function F coincides with the given function f in the punctured plane (C \ {0}, it follows that f has a removable singularity at 0. 1 (2) The function \ has a pole at 0, since lim - 3 = +oo. z--+0 Z1 1 Z ! (3) The function exp has an essential singularity at 0. Indeed, z (a) 0 is not a removable singularity, because for example lime½ = +oo. X',,0 1 x (b) 0 is also not a pole, since lime = 0, and so it can't be that x/'O lim lf(z)I = +oo. ◊ z--+0 We will now learn our first characterization result for singularities, in terms of limiting behaviour. Theorem 4.8. ( Classification of singularities via limits) Suppose z0 is an isolated singularity of f. Then z0 is removable {cc} zo is a pole {cc} lim (z - zo)f(z) = 0. z➔zo (a) ,( lim (z - zo)f(z) = 0) and z➔zo (b) =in EN such that lim(z -zot + 1 f(z) = 0. a z--+z zo is essential Proof. I {cc} ( The smallest such n is called the order of the pole zo of f.) VnEN,( lim(z- zorf(z)=O). (1) z0 removable==> z➔zo z �� a (z - zo)f(z) = 0 I· Let z0 be removable, and let F be holomorphic in D(zo, R) := {z EC: lz - zol < R} Taylor and Laurent series 143 such that F = f for O < lz - zol < R. Then using the fact that F is continuous at z 0, we obtain lim (z-zo)F(z) = lim (z - zo)F(z) = 0· F(zo) = 0. z➔ zo z➔zo Hence for everyE > 0there exists a 8 > 0such that for O < lz - zol < 8, there holds that and so l(z -zo)f(z) - OI =l(z -zo)F(z)-OI <E, lim (z-zo)f(z) = 0. z➔zo Next we will show that lim (z - zo)f (z) = 0 =} zo is removable . z➔zo Suppose that lim (z-zo)f(z) z--+zo = 0, and that f is holomorphic in the punctured disc {z EC: 0< lz-zol< R}. Then f has a Laurent expansion there: f(z) = L C (z-zo) , n n nEZ 0< lz -zol< R, where the coefficients Cn 's are given by contour integrals along any circle Cr with center z 0 in the punctured disc and r < R: 1 f(z) dz, n Cr (z-zo) +l We show that C-n = 0for n EN. Given E > 0, take r small enough so that l(z - zo)f(z)I <Eon Cr. Then we have for n ENthat Cn 1 I C- n I = 121ri = 2_l_ 1ri 1 1-1 1 (z - zo)f(z) d f(z) d zl n+l zl = 21ri (z - zo)- n +2 (z -z o )Cr Cr 1 E n 1 n < ---2 1r r = Er - < 1:R -l_ 21r r-n +2 Since the choice ofE> 0was arbitrary, it follows that c_n = 0for all n EN. Consequently, with (X) n=O we see that Fis holomorphic in {z E C: lz - zol < R}, and F=fin the punctured disc {z EC: 0< lz - zol< R}. 144 A Friendly Approach to Complex Analysis (2) zo is a pole ::::} { ,( lim (z - zo)f(z) z--+zo = 0) and :3n EN such that lim (z - zor+l f(z) z➔za = 0. Suppose that z0 is a pole of.f. Then z0 is not removable (why?), and so by the first part, it follows that ,( lim (z - zo)f(z) z➔zo = o). (Here ---, is the symbol for negation, to be read as "it is not the case that".) There is some R > 0 such that lf(z)I > 1 in the punctured disc D := {z E (C: 0 < lz - zol < R}. Define g in this punctured disc D by g(z) = 1 f(z) · Since z0 is a pole off, it follows that lim g(z) z➔zo lim (z - zo)g(z) z➔�o = 0. In particular, also = 0, and so by the first part above, g has a holomorphic extension G to {z E (C: lz - zol < R}. Also, G(zo) = lim G(z) = lim g(z) = 0. z--+zo z--+zo So z0 is a zero ofG, and since G is not identically zero in a neighbourhood ofz0, it follows from the result on the classification ofzeros that z0 has some order n EN, and there is a holomorphic function H defined for lz- zol < R such that H(zo) =/- 0 and G(z) = (z - zo) n H(z). In particular, for 0 < lz - zol < R, we have 1 1 1 f(z) = g(z) = G(z) = (z - zo) n H(z) · Hence 0 - 1·Im (z - zo) 1Im . (z - z0) n +lj (z) 0 . z--+zo H( Z) H(zo) z--+zo Done! 145 Taylor and Laurent series o) • lim (z - zo)f(z) = and z--+zo . ( 3n EN such that lim (z - z0 )n+l f(z) = 0. z--+zo ==> z0 is a pole. } Choose the smallest such n, call it n*: thus lim (z - zo)n .+l f(z) = 0, z➔zo o). · lim (z - zo)n * f(z) = z--+zo ( So (z - zor• f(z) has a removable singularity at z = zo. Thus there exists an F, holomorphic in {z EC: lz - zol < R} such that for 0 < lz - zol < R, (z - zo)n • f(z) = F(z). Note that F(zo) = lim F(z) = lim (z - zor• f(z)-=/- 0 z➔zo z➔zo (owing to our choice of n*.) From F(z) fz ( ) (z - zo)n• for 0 < lz - zol < R, we obtain lim IJ(z)I = lim z--+zo z--+zo So zo is a pole off. IF(z)I 1 = +oo. = IF(zo)I · lim n z--+zo z ZoI n . lz - zol · '--.,.--' I ,60 (3) This follows easily from the first two parts. Indeed, if z0 is an essential singularity of f, then it is not removable, and so o), • lim (z - zo)f(z) = z--+zo ( and so we get the desired claim at least for n = 1. But since z0 isn't a pole, by (2) we also get the desired conclusion for all other n's as well. Conversely, if o) , for all n EN, · lim (z - zorf(z) = z--+zo ( then in particular, from the n = I case, we get by (1) that zo can't be a removable singularity. And from the cases for all other n ?: 2, we can conclude, in light of (2), that zo can't be a pole either. This zo must be an essential singularity of f. □ 146 A Friendly Approach to Complex Analysis Let us reconsider Example 4.14. Example 4.15. sin z = z--+0 lim sin z = sin 0 = 0, and so 0 is a removable z--+0 z sinz . . smgulanty of -- . z 1 and lim z 4 • \ = lim z = 0, 0 is a pole of 3, and lim z · \ = (2) As , ( z--+0 Z z--+0 Z z--+0 Z 1 moreover the order is 3 since , ( lim z2 • 3 = and , ( lim z3 • \ = z--+0 z--+0 z z 1 1 1 1 1 + + -+ -For x > 0 ex = 1 + • • • > -Thus for all (3) 3!x3 n!xn ' l!x 2!x 2 1 1 n EN, xn e;;; > 1 and n. (1) We have lim z • o) o) o). So for all n EN, , ( lim Z n exp! Z z--+0 = o). 1 Thus 0 is an essential singularity of exp -. z ◊ We will now learn our second characterization result for singularities, in terms of the Laurent series coefficients. Theorem 4.9. (Classification via Laurent coefficients) Let I: c (z - zot for 0 < lz -zol < R, for some R > 0. (1) zo be an isolated singularity of f, and (2) f(z) Then = nEZ n zo is removable zo is a pole z0 is essential {=} {=} {=} For all n < 0, Cn = 0 There exists an m EN such that (a) C-m i- 0 and (b) for all n < -m, Cn = 0 Then the order of the pole zo ism. There are infinitely many negative indices n such that Cn i- 0. 147 Taylor and Laurent series removable pole -m essential Proof. 0 infinitely many nonzero I (1) zo is removable =? (for all n < 0, Cn = 0) I- Suppose that z0 is a removable singularity. Then f has a holomorphic extension F defined for lz - zol < R. But then this holomorphic F has a Taylor series expansion 00 F(z) In particular, for O = Lcn(Z - zo) n n =O < lz - zol < R, for lz-zol < R. we obtain L Cn(z-zo) = L C (z-zo) , 00 f(z) = n n =O nEZ n n and by the uniqueness of the Laurent series expansion in an annulus, it follows that Cn = Cn for n � 0, and Cn = 0 for all n < 0. I (For all n < 0, Cn Suppose that Cn = 0) =? z0 is removable = 0 for all n < 0. 1· Then for O < lz - zol < R, 00 nEZ and so defining F for I z - zo I F(z) n =O < R by = L C (z - zo) , 00 n =O n n we see that Fis holomorphic in {z E C : lz - zol < R} (because it is a power series!) and moreover F = f for O < lz - zol < R. (2) We will show that if z0 is a pole of order m, then f(z ) = C-m (Z - Zo ) m 00 C-1 � ( )n +···+--+co+ �en z-zo , Z - Zo n =O 148 A Friendly Approach to Complex Analysis with C- m =/- 0. Suppose zo is a pole of order m. Using Theorem 4.8, + lim (z - Zo)((z - zo)m f(z)) = lim (z - zor l f(z) = 0, z-+zo z-+zo the function (z - zorf(z) has a removable singularity at zo. We have + (z - zorf(z) = (z - zo)m Cn(z - zo)n = Cn(Z - zo)n m , L L nEZ nEZ and by the previous part, this last series has all coefficients of negative powers of z - zo equal to 0, that is, 0 = c-(m+ ) = c-(m+2) = · · ·. Hence l for O < I z - zo I < R, (z - zo)m f(z) = C- m + C-m+1(z - zo) + C-m +2(z - zo)2 +··· (4.6) and so c 1 C- m +co +c1(z - zo) +c2(z - zo)2 +···. +···+ _ (z - zo) (z - zo)m Moreover, the function C-m + C-m+1(z - zo) + C- m +2(z - zo)2 + ··· is holomorphic, and so it has the limit C- m as z-+ zo. Thus f(z)= lim(z-zorf(z) = lim(c_ m +C-m +1(z-zo)+c_ m +2(z-zo)2 +- ··) = C- m · z-tzo z-+zo On the other hand, since z0 is a pole of order m, we know that •( lim (z - zo)m f(z) = z-+zo o). Consequently, (4.6) implies that C- m =/- 0. Suppose now that f(z) = L c m ( Z -_ Zo +···+�+co + Cn(z - zo)n, Z - Zo n=O with c_ m =/- 0. We will show that z0 pole of order m. Suppose that there is some m E N such that C- m =/- 0 and Cn = 0 for all n < -m. Then (z-zorf(z) = C- m +C- m +1(z-zo)+c_ m +2(z-zo)2 +···, and since the right hand side defines a holomorphic function, say h, in {z E <C: lz - zol < R}, it follows that )m lim (z - zo)m f(z) = lim h(z) = h(zo) = C- m =/- 0, z-+zo m+ lim (z - zo) l f(z) = 0 · C- = 0. z-+zo z--+zo and m Thus zo is a pole of order m off, by Theorem 4.8. (3) This is immediate from the previous two parts and the fact that an D essential singularity is neither a removable singularity nor a pole. 149 Taylor and Laurent series Let us reconsider Example 4.14. Example 4.16. (1) We have for z i- 0, si z : = � (z _ ;� + �: _ ;� + _ ...) = 1 _ ;� + �: _ ;� + _ ... . Since there are no negative powers of z appearing in the Laurent series sinz . . . . . expansion, 1t follows that 0 1s a removable smgulanty of --. (2) For z z i- 0, 1 1 - =···+0+-+0+··· . z3 z3 Thus c_3 = 1 order 3. (3) For z 0, i- i- 0, while 0 = c_4 = c_5 =··· . So 0 is a pole of z\ of 1 1 1 1 exp-;; = 1 + l! + 2! 2 + ! +··· · z z 3 z3 For infinitely many negative indices n (in fact for all of them), C-n 1 =I-/- 0. n. Thus 0 is an essential singularity of exp .!_. z ◊ Exercise 4.30. Let D be a domain and f be holomorphic in D such that f has only one zero zo in D, and the order of zo is m > 1. Show that the function z f---t 1/f(z) has a pole of order matzo. Exercise 4.31. Let D be a disc with center zo. Suppose that f is nonzero and holomorphic in D \ {zo}, and that f has a pole of order mat zo. Show that the function z f---t 1/ f (z) has a holomorphic extension g to D, and that g has a zero of order m at zo. Exercise 4.32. Let D be a domain and zo ED. Suppose that f has a pole of order m at zo and that f has the Laurent series expansion f(z) = L:Cn(z nEZ - zof for O < lz - zol < R, where R is some positive number. Show that C-1 1 dm-1 lim - _1 ((z - zo) m f(z)) . = (m- l '. z--+zo z d ) m A F'riendly Approach to Complex Analysis 150 Exercise 4.33. True or false? (1) If f has a Laurent expansion z-1 +eo+c1z+- ••,convergent in some punctured disc about the origin, then f has a pole at 0. (2) A function may have different Laurent series centered at zo, depending on the annulus of convergence selected. (3) If f has an isolated singularity at zo, then it has a Laurent series centered at zo and is convergent in some punctured disc O < [ z - zo[ < R. (4) If a Laurent series for f convergent in some annulus R1 < [z - zo[ < R2 is actually a Taylor series (no negative powers of z - zo), then this series actually converges in the full disc given by [z - zo[ < R2 (at least). (5) If the last conclusion holds, then f has at worst removable singularities in the full disc given by [z - zo[ < R2 and may be considered holomorphic throughout this disc. Exercise 4.34. Decide the nature of the singularity, if any, at O for the following functions. If the function is holomorphic or the singularity is isolated, expand the function in appropriate powers of z convergent in a punctured disc given by 0 < [z[ < R. . 1 sinz, sm ;;, 1 sin l' z sinz z . 1 z sm­ z Exercise 4.35. True or false? I !I= (1) lim exp +oo. z➔O Z (2) If f has a pole of order m at zo, then there exists a polynomial p such that fhas a holomorphic extension to a disc around zo. (z -pzo)m (3) If f is holomorphic in a neighbourhood of 0, then there is an integer m such that 1._ has a pole at 0 whenever n zn > m. (4) If f, g have poles of order m f,mg respectively at zo, then their pointwise product f · g has a pole of order m1 + mg at zo. Exercise 4.36. Give an example of a function holomorphic in all of (C except for essential singularities at the two points 0 and 1. Exercise 4.37. The function f given by f(z) = l/(z - 1) clearly does not have a singularity at 0. As it has the Laurent series z-1 + z-2 + z-3 + • • • for [z[ > 1, one might then say that this series has infinitely many negative powers of z, and fallaciously conclude that the point 0 is an essential singularity of f. Point out the flaw in this argument. Exercise 4.38. Prove or disprove: If f and g have a pole and an essential singu­ larity respectively at the point zo, then fg has an essential singularity at zo. Taylor and Laurent series 151 Wild behaviour near essential singularities 4.8.1 We now show a result which illustrates the "wild" behaviour of a function f at its essential singularity any E > zo. It says that given any complex number w, z 0, there w. So the 0, and any arbitrary small punctured disc� with center is a point z in � such that lies within a distance E from f(z) image of any punctured disc centered at the essential singularity is dense in C. Or in even more descriptive terms, f comes arbitrarily close to any complex value in every neighbourhood of zo. Theorem 4.10. ("Casorati-Weierstrass" 3) Suppose z 0 is an essential singularity of f. Then (1) for every complex number w, (2) for every 8 > 0, and (3) for every E > 0, there exists a z EC such that lz - zol 8 \ . zo <>----------; 2'.,/ Proof. and 8 8 and lf(z) - wl • f � < E. f(z) wa----� E ' Suppose the statement is false. Then there exist w E C, E > 0 such that whenever z ED:= {z EC: 0 < lz - zol < 8}, lf(z) - wl for < z ED 2: E. Then g defined by g(z) 1 = f(z) - w is holomorphic there, and lim (z z➔zo l z O:; g( )I = 1 lf(z) - wl 1 zo)g(z) :; �) > 0 we have = 0, since zED. 3This result was published by Weierstrass in 1876 (in German) and by the Sokhotski in 1873 (in Russian), So it was called Sokhotski's theorem in the Russian literature and Weierstrass's theorem in the Western literature. The same theorem was published by Casorati in 1868, and by Briot and Bouquet in the first edition of their book (1859), called Theorie des fonctions doublement periodiques, et en particulier, des fonctions elliptiques. However, Briot and Bouquet removed this theorem from the second edition (1875). 152 A Friendly Approach to Complex Analysis g has a removable singularity at zo. Let its extension be denoted again g. Let m be the order of the zero of g at zo. (Set m = 0 if g(zo) =/- 0.) Then g(z) = (z - z0) m h(z), for some function h holomorphic in D and such that h(zo) =/- 0. Then for 0 < lz - zol < 8 So by (z - zo)m + 1 (J(z) - w + w) (z - zo) m +l f(z) 1 (z - zo)m +l __ + (z - zo) m +l · w g(z) z - zo + (z - zo)m +l . w h(z) z-tzo 0 + 0 . w = 0. --t h(zo) zo (when m = 0) or a pole z0 ( when m E N). Hence zo can't be an essential singularity of f, a Thus either f has a removable singularity at at □ contradiction. Example 4.17. The function exp(l/ z) has an essential singularity at 0. We show that it takes on any given nonzero w (= p exp(i0) E <C, p,0 E JR), in any arbitrarily small neighbourhood of 0. Setting z = rexp(it), we need to solve 1 cost sint exp - = exp (-- - i--) = r z r - pe'0 . By equating the absolute values, we obtain cost -- = 1ogp. r On the other hand, by looking at arguments, we see that a solution is given when _ Using (cost)2 + (sint)2 = 1, sint r = 0_ we have 1 r=--;=====� J(logp)2 + 02 Moreover tant = ---. logp But we are allowed to increase 0 by integral multiplies of 21r, without chang­ ing w. Bearing this in mind, it is clear from the above expression for r that we can make r as small as we please. ◊ 153 Taylor and Laurent series The above example illustrates a much stronger theorem than the Casorati­ Weierstrass Theorem, due to Picard, which says that the image of any punctured disc centered at an essential singularity misses at most one point of IC! In our example above, the exceptional value is w = 0. A proof of Picard's Theorem is beyond the scope of these notes, but can be found in the book [Conway (1978)]. Exercise 4.39. Prove, using the Casorati-Weierstrass Theorem, that if f has an essential singularity at zo, and if w is any complex value whatever, then there exists a sequence z1, z2, Z3, · · · such that lim f (zn) = w. lim Zn = zo and n--+cx:i n--tcx:i 4.9 Residue Theorem Suppose that D is a domain and that a holomorphic f : D \ { z0} ---+ IC has an isolated singularity at zo. Let J(z) = L Cn(z - zo) nEZ n for O < lz - zol < R. Then we call the coefficient C- 1 the residue off at zo, and denote it by res(!, zo)- Why the name? In everyday language, "residue" means something which is "left over". In the above too, one can think of c_ 1 as something which is left over in the following manner. We know that ( f(z)dz = 21ric_ 1 = Cn (Z - zo)n )dz, � O�� 1 + 1 dz= (z Cr nEZ Cr lcr where Cr is given by Cr(t)= z0 +rexp(it), E [0,21r], and r < R. Note that we have 0 for n-=/=- -1, Cn Z-Zo n )dz= { 2. 1ric_ 1 £or n _ - - 1, Cr 1 1 (L r 1( ( ) and so if we imagine integrating termwise formally in the Laurent series expansion off, then 21ric_ 1 is what is left over. What is the big fuss about this residue? The equation 21ric_ 1 = 1 f(z)dz Cr gives a way of computing contour integrals via calculating the residue of f at z0 (which amounts to finding the value of the coefficient c_1 in the 154 A Friendly Approach to Complex Analysis Laurent expansion off). So if there is way of calculating c_1 easily, then we can compute 1 'Y f(z)dz ( = f f(z)dz) = 21ric_1 lcr (4.7) for any closed path I which is D\ {zo}-homotopic to Cr. And we have seen in Exercise 4.32 that there is a way of calculating c_l via the formula dm-1 1 c_l = ' lim - m _1 ((z - zo)m f(z)), (m-1.) z ➔ zo dz when z0 is a pole of order m off. It turns out that some awkward real integrals can be computed by first relating them to a contour integral for an appropriate holomorphic f and some path 1, and then using this route via the residue to evaluate the contour integral, and eventually also the awkward real integral. Here is an example. 1 Example 4.18. Consider the real integral 1 3cos 5 0 0 + We view this as the contour integral along a circular path as follows. First we write 1 z+ � exp(i0) + exp(-i0) z2 + 1 cosB = = = 2 2z ' 2 271" ---d0. where z := exp(i0). So if, is the circular path with radius 1 and center at 0 defined by 1(0) = exp(i0), then ,'(0)d0 = iexp(i0)d0 f 21r }0 1 d0 = 5+3cos0 1 'Y 0 E [O, 21r], = izd0, and so l - z2 5+3 ( 1 :1) iz dz = 2 Let f be the function defined by 2i f(z) = - (3z + l)(z + 3 · ) 1'Y 2i (3z+l)(z+3) dz · 155 Taylor and Laurent series Then f has two poles, one at -1/3 and the other at -3, both of order 1. Of these, only the one at -1/3 lies inside '"Y· Thus, 1271" 0 1 d0 = 5+ 3 cos 0 1 'Y f(z)dz = 21ri · res (1, _!3 ) • r What is c_1 ? We have in a punctured disc with center -1/3 that f ( Z) = C l � ! Z + Co + C1 (Z 2 + �) + C 3 (Z + � ... , + and so (z + co (z + .... Thus �) f(z) = c_ 1 + �) + C -1 Consequently, . 1 Im 1271" = z➔-1/3 o 1) - !( Z) = (Z + 3 1 1·Im z➔-1/3 7r d0 = -. 5 +cos 3 2 0 2i i = --. 3(z + 3) 4 ---- ◊ More generally, if f has a finite number of poles in D, a result similar to (4.7) holds, and this is the content of the following. Theorem 4.11. (Residue Theorem) Let (1) D be a domain, (2) (3) (4) (5) f be holomorphic in D\ {P1, ... ,PK}, f have poles at pl,···,PK oforder m1,••·,mK, respectively, '"Y be a closed path in D\ {p1, ... ,PK} and "( be such that for each k = l, ..., K, "( is D \ {Pk}-homotopic to a circle Ck centered at Pk such that the interior ofCk is contained in D and contains only the pole Pk. Then 1 'Y K f(z)dz = 21riLres( f,pk)k=l 156 Proof. A Friendly Approach to Complex Analysis For each k = l, ..., K, we can write f(z) = = L C-n,k(z - Pk)-n + L Cn,k(z - Pkt= fk(z) + hk(z), n=l n=O mk 0 < lz - Pkl < Rk, for some Rk > 0, where the sum with negative powers of z - Pk is denoted by fk, and the sum with nonnegative powers of z - Pk is denoted by hk. Note that hk is holomorphic for lz - Pk! < Rk, and that fk is a rational function, defined in <C \ {pk}, having only one singularity, namely a pole at Pk· Thus f - !k is holomorphic in a small disc around Pk· Set g := f - (!1 +···+ fK)- Fix a k E {1, ..., K}. Write g = (f - fk) k L j# Observe that both f - fk and each Ji for j -/- k is holomorphic in a small disc around Pk· Thus g is holomorphic in a small disc around Pk· This happens with each k E {1, ..., K}. Thus, g is holomorphic in D. We note that as 'Y is D \ {pi}-homotopic to a circle 01 centered at p1, by the Cauchy Integral Theorem, that is, So j "I i i g(z)dz = 0, (!-(Ji+···+ fK))dz = 0. By the Cauchy Integral Theorem, 1 fk(z)dz = "/ f(z)dz = 1 f(z)dz = 2nic-1,k, k = l, · · · , K. j fk(z)dz = Lk=l 2nic_1,k = 2ni Lk=l res(f,Pk)L k=l Ck K K K "I o Exercise 4.40. Evaluate/ L g(z) dz along the path I shown below. "/ 1 + expz -5+10i 10+10 v 5+5i -5+5i -5-2i , , , I -5-5i 5-5i � 10-2i □ 157 Taylor and Laurent series Integral of a real rational function using the Residue Theorem. As we mentioned earlier, the Residue Theorem can be used to calculate contour integrals, and sometimes gives an easy way to calculate some real integrals. Let us see how it can be used to calculate the improper integrals of rational functions. Consider a real integral of the type 1_: f(x)dx. Such an integral, for which the interval of integration is not finite, is called an improper integral, and it is defined as 100 -oo 1 a--+-oo l f(x)dx = lim ° f(x)dx + lim b--++oo a b l J(x)dx, lo (4.8) when both the limits on the right hand side exist. In this case, also 1_: f(x)dx = rE/_f!- 1:r f(x)dx. 00 (4.9) We call the right hand side in (4.9), if it exists, the Cauchy principal value of the integral. (However, it can happen that the Cauchy principal value exists, for example, lim r--++oo J r -r r xdx = lim ( r--++oo 2 - 2 r2 ) 2 = 0, 2 b but the improper integral doesn't exist: l xdx = b .) 2 lo We assume that the function fin (4.8) is a real rational function whose denominator is different from O for all real x, and the denominator has degree at least two more than the degree of the numerator. Then it can be seen that the limits on the right hand side in (4.8) exist, and so we can start from (4.9). Consider l f(z)dz, around a path, shown in Figure 4.4. (j -r 0 r Fig. 4.4 'Y consists of the semicircular arc u and the straight line joining -r tor. 158 A Friendly Approach to Complex Analysis Since f is rational, 1 f(z)dz = H z f (z) has finitely many poles in the upper half plane, and if we choose r large enough, 'Y encloses all of these poles in its 1 interior. By the Residue Theorem, we then obtain 'Y CT f(z)dz + r J -r f(x)dx = 21ri L k: lm(pk)>0 res(f,Pk), where the sum consists of terms for all the poles that lie in the upper half-plane. From this, we obtain r J f(x)dx = 21ri -r L k: Im(pk)>O res(f,Pk) - 1 CT f(z)dz. We show that as r increases, the value of the integral over the corresponding semicircular arc u approaches 0. Indeed, from the fact that the degree of the denominator off is at least two more than the degree of the numerator, it follows that there are Hence for r > Consequently, ro, lim 1= r-++= 1 -= a M, ro large enough such that M lf(z)I < ll (lzl > ro). W f(z)dzl � � 1rr = � . 1r f(z)dz = 0, and so f(x)dx = 21ri L k: lm(pk)>O res(f,pk)- Let us see an example of this method in action. Example 4.19. We will show that The function i p1 = exp (7r ) , 4 f given by f (z) 1 1 -- dx O (X) 1 + X 4 =� l+z = 7r1n· 2y2 has four poles of order 1: 31ri 51ri P2 = exp ( ) , p3 = exp ( ) , 4 4 1 p4 = exp ( 71ri 4) . 159 Taylor and Laurent series The first two of these poles lie in the upper half plane. We have res(f,Pl ) = = z➔ r1mp1 z1+- Pi Z 4 1 res(f,p2 ) = � z �2 =4 z-p2 = z��2 1+ z4 4 1: As 1 f is 21ri ( 4 even -i (f(x) = f(-x) / 00 1 exp ( i :)+ for all x 1 E pf 41 = - exp (1+ z4 ) 4� = !(1 +z t=p2 : x dx = z -pi 1 1 - (7ri) 4 i iexp ( - exp ( - : )) JR), 00 � = i we have ' (1+ p�) Z -p2 ___1_�_ Thus 1 (1+ Z 4) - (1+ P4i ) r 1m z➔ p1 i : ). = 7r sin i = �. 7r 1 dx = 1 2 _ 00 1 + x4 dx = 2v12· lo 1 + x4 ◊ Here is an example of the computation of an exotic integral where the integrand is not a rational function. Example 4.20. (Fresnel Integrals4 ) We will show that We consider 1 f lo 00 cos(x 2 )dx = exp(iz2 )dz, f lo 00 sin(x )dx 2 = �2v2 where, is shown below. lzl =R 4These integrals arise in optics, in the description of diffraction phenomena. 160 A Friendly Approach to Complex Analysis i Since exp(iz 2 ) is entire, by the Cauchy Integral Theorem we have 0= = 1 exp(iz 2 )dz R exp(ix 2 )dx + -1 R 14 1!: exp(iR2 exp(2i0))iR exp(i0)d0 exp (it2 exp (ii)) exp (i�)dt. We will show that the middle integral goes to 0 as R -+ oo. First note that I jexp(iR2 exp(2i0))iRexp(i0) = jRexp(R2 (i cos(20) - sin(20))) = Re -R 2 sin(20) I . But Figure 4.5 shows that whenever the angle t is such that 0 < t < i, 2 sin t -<-. 1f t Indeed, the length of the arc PAQ, which is 2t, is clearly less than the length of the semicircular arc PBQ, which is 1r sin t, and so the inequality above follows. (Alternately, we could note that t H sin tis concave in [0, 1r], because its second derivative is - sin t, which is nonpositive in [0, 1r], and so the graph of sin t is lies above that of the straight line graph of 2t/1r joining the two points (0, 0) and (1r/2, 1).) p B Q Fig. 4.5 Here P is any point on the circle with center O and radius 1 such that OP makes an angle t with the positive real axis. We reflect P in the real axis to get the point Q, and let M be the intersection of PQ with the real axis. With Mas center and radius PM, we draw a circle, meeting the real axis on the right of A at the point B. Taylor and Laurent series = 20 yields exp(2i0) )iR exp(i0) I = Re-R Applying this inequality with t !exp(iR 2 and so I 1 ,,. 4 I 2 exp(iR exp(2i0))iRexp(i0)d0 s;R which tends to O as lim R-H oo f R lo R➔ 1 .. 4 2 161 2 sin( 2 6) s;Re-R �, 20 7r 2 e-4R nd0= (l-e-R ), 4R +oo. Hence we obtain exp(ix )dx 2 = f lim R lo R-H oo i = 1+ f v2 lo oo 2 exp (it exp (i�)) exp (i�)dt 2 4 e-t Here we have used the known 5 fact that f e_"' oo lo dx 2 = 2 dt = (1 + i)-Jrr_ 2v2 J1r. 2 So by equating real and imaginary parts, f 00 lo Exercise 4.41. Evaluate 1 = 2 cos(x )dx 1 o f 00 lo sin(x )dx 2 = �2v2 ◊ 2 " cos0 5 + 4 cos0 d0. Exercise 4.42. Evaluate the following integrals: (1) (2) () 3 (4) 00 o 1 - -2 dx. 1 +x roo roo (1 1 1 (a2 + x2 )(b2 + x2) dx, lo lo 00 o 1 + x2)2 dx where a> b > 0. . 1 +x 2 . 1 +x 4 dx Exercise 4.43. If n E N and C is the path C(t) = exp(i0), 2 { e:�: dz. Deduce that f " ecose cos(n0 - sin0)d0 = 2�. kz 5With J h := o l [ 00 2 e-x dx, 1 2 = = ( f 00 l/l e-x dx 121 0 2 00 0 2 ) ( � f 00 lo 2 e-Y dy 7l' e-r rdrd0 = -. 4 ) 0 E [O, 271'], find 2 f 00 2 = [o lo e-(x +Y ldxdy l 00 162 A Friendly Approach to Complex Analysis Exercise 4.44. Let f have a zero of order 1 at zo, so that 1/f has a pole of order 1 at zo. Prove that res (y, zo) = '� . f ( o) 1 Exercise 4.45. Prove that res (-.- -,k1r) smz = (-l)k . Exercise 4.46. The nth Fibonacci number fn, where n � 0, is defined by the following recurrence relation: Jo = 1, Ji = 1, fn = fn-1 + fn- 2 for n � 2. Let F(z) := L fnz . (X) n n=O (1) Prove by induction that fn S:: 2n for all n EN. ( 2) Using the estimate fn S:: 2n , deduce that the radius of convergence of Fis at least 1/2. _: _ 2• 1 z Hint: Write down the Taylor series for zF(z) and z F(z) and add. (3) Show that the recurrence among the fn implies that F(z) = 2 (4) Verify that res ( n+l l � _ 2 , 0) z ( z z ) (5) Using the Residue Theorem, prove fn = Js ( (1 = fn • +2�r +l - c �) 2 n +l ) · 1 around a circle with center O and radius R zn+l(l-z-z2 ) and show that this integral vanishes as R-+ +oo. Hint: Integrate 4.10 Notes §4.8 follows closely [Beck, Marchesi, Pixton, Sabalka (2008)]. Exercise 4.11, 4.24, 4.33, 4.35, 4.36, 4.37, 4.38, 4.39 are taken from [Flanigan (1972)]. Exercise 4.14 is taken from [Volkovyski'i'., Lunts, Aramanovich (1991)]. Ex­ ercise 4.15, 4.25 are taken from [Rudin (1987)]. Exercise 4.28 is taken from [Beck, Marchesi, Pixton, Sabalka (2008)]. Exercise 4.40 is taken from [Ash and Novinger (2007)]. Chapter 5 Harmonic functions In this last chapter, we study: (1) harmonic functions, which are real-valued functions that solve a certain partial differential equation, called the Laplace equation, (2) that real and imaginary parts of holomorphic functions are harmonic, and that the converse holds locally and globally on simply connected domains, (3) some consequences of the above interplay between harmonic and holo­ morphic functions, in particular in a certain "boundary value problem", called the Dirichlet problem. 5.1 What is a harmonic function? Definition 5.1. Let Ube an open subset of JR.2 . A function u: U-+ JR. is called harmonic if uhas continuous partial derivatives of order 2 ( abbrevi­ ated by writing uE C2 ), and it satisfies the Laplace equation: {J2u {J2u (�u)(x, y) := a 2 (x, y) + 2 (x, y) = 0 for all (x, y) EU. x ay Example 5.1. Let U = JR.2 • Consider the function u : U -+ JR. given by u(x, y) = x2 - y2 for (x, y) E JR.2 • Then a2 u a au 2x ax = au 2 ay = - y a2 u 2, ax2 = a2 u ay2 = -2. 2u y u E C 2 and �u = O in JR.2 , x xy 2 2 + ) , ( ax2 ay2 ( , ) = - = 0. Since u is harmonic in JR.2 . ◊ Thus 163 164 A Friendly Approach to Complex Analysis Of course, not all functions are harmonic. Example 5.2. Consider the function u given by u(x,y) (x,y) E JR2. Then for all (x,y) E JR2, cP u a2 u ox 2 (x,y) + oy2 (x,y) = 2 + 2 = 4 # 0. Since x2 + y2 for 6..u is never O in JR2 , u is not harmonic in any open subset of JR2 • ◊ Exercise 5.1. Show that the following functions u are harmonic in the correspond­ ing open set U. 2 2 2 (1) u(x,y) = log(x + y ), U = ffi. \ {(O, O)}. (2) u(x,y)=e°' siny,U=ffi.2• Exercise 5.2. Show that the set Har(U) of all harmonic functions on an open set U forms a real vector space with pointwise operations. Exercise 5.3. Is the pointwise product of two harmonic functions also necessarily harmonic? Why bother about harmonic functions? Harmonic functions are important because they satisfy the Laplace equation, which is important among other things for two primary reasons: (1) The Laplace equation is the prototype of an important class of PDEs, namely "elliptic equations", which is one of the three main classes of PDEs. I Class of PDE I Main example Parabolic Hyperbolic Diffusion equation Wave equation 82 u 8 2 u =0 8x 2 + 8y2 82 u 82 u ---=0 8x 2 at 2 8 u 82 u ---=0 8t 2 8x 2 (2) The Laplace equation arises in many applications, for example in physics in the following scenarios. In hydrodynamics, the "velocity potential" of the fluid flow satisfies the Laplace equation, while in elec­ trostatics, the electrostatic potential satisfies the Laplace equation. The Laplace equation also has an important link with stochastic processes. We describe this very roughly below. Harmonic functions Imagine the open unit disc lIJ) := {z 165 E <C: lzl < 1}. Consider a particle starting at a point z ED and undergoing Brownian motion (for exam­ ple think of a pollen grain in water, bombarded by many tiny water molecules producing "random" motion). Intuitively one feels that since the motion is random, eventually the particle will leave the boundary := {z 1[' E <C : lzl = 1} of the disc lill. Let us denote by (z the point on 1[' where the particle first exits the unit circle 1[', having started at z. So we get a random variable (z which lives on the unit circle. See Figure 5.1. lIJ) f lives on 1[' Fig. 5.1 Brownian motion and the Dirichlet problem. Now let f : 1[' -+ ffi. be a given continuous function. Then we can think of f((z) as being a real-valued random variable on T. Let us denote its expectation by IE(f((z )). This depends on where one starts initially, that is, it depends on z. Let u: lIJ)-+ ffi. be given by u(z) = IE(f((z )), z E lill. It turns out that u is then harmonic, and in fact it is a solution to the "Dirichlet problem", which is the boundary value problem, where given f : 1[' -+ ffi. on the boundary 1[', we have to find a function u, solving the Laplace equation in the interior lIJ) of 1[', such that it has a continuous extension to 1[', matching with the given data f: {Au= 0 in ID), uhr = 5.2 f. What is the link between harmonic functions and holo­ morphic functions? It might appear that harmonic functions should belong just to the realm of Real Analysis. In this section, we will now learn about two results, which will amply justify their study in Complex Analysis. Roughly a function is 166 A Friendly Approach to Complex Analysis harmonic in an open set if and only if locally, it is the real part of some holomorphic function. Theorem 5.1. Let (l) U be an open subset of C and (2) f: U-+ C be holomorphic in U. Then : :: ;::���,} are harmonic functions in U. Vice versa, we will also learn the following converse to this. Theorem 5.2. Let (l) U be simply connected and (2) u: U-+ ffi. be harmonic in U. Then there exists a function v : U -+ f := u + iv is holomorphic in U. ffi., such that v is harmonic in U and Note that for the fin the conclusion of the above result, we have Re(!)= and Im(!) = v. u, So what we are saying is that every harmonic function in a simply connected domain is the real part of some holomorphic function defined there. Since a unit disc is simply connected, in particular it follows that every harmonic function is locally the real part of a holomorphic func­ tion defined (at least in that disc). We will see later on in Exercise 5.5 that the assumption of simply connectedness is not superfluous, and that given a harmonic function in a non-simply connected domain, there may fail to exist a globally defined holomorphic function in the whole domain whose real part is the given harmonic function. But we are jumping ahead. Let us first focus on the first result mentioned above. Before proving Theorem 5.1, let us revisit Example 5.1. U Example 5.3. In our previous example, when we have u = Re(z 2 )= Re(x2 -y2 = ffi.2 and + 2xyi) and z 2 is entire. u = x2 - y2 , So Theorem 5.1 delivers again our old observation that u is harmonic in ffi.2 • In fact from our calculation, Theorem 5.1 also gives us that v (We can of course check this by brute force: o and so o2u ox2 + 2u OU = 2Y' = ox2 ox o 2v = 0 + 0 = 0.) oy2 O' := ov = 2x ox 2xy = ' o2 v oy2 Im(z 2 ) is harmonic. = 0, 167 Harmonic functions On the other hand, it follows from Example 5.2 and Theorem 5.1 that for any open subset U of C, u holomorphic function defined in := U. x2 + y2 is not the real part of any ◊ Proof. (of Theorem 5.1.) We have f(x + iy) = u(x, y) + iv(x, y) for (x, y) E U. Since f is infinitely many times differentiable, we know that u, v have partial derivatives of all orders, and so by the Cauchy-Riemann equations, we have {J2 u =� au ( ) ax ax ax2 and so av ( ) ax ay (C,;,R) � 2 (u� ) u is harmonic. Similarly, a v =� av ( ) ax ax ax2 2 (C,;,R) � (- au ax ay ) 2 (v� ) � ay av ( ) ax (C,;,R) � (- au ay ay 2 au - ay2 ' �ay (- axau) )· (C,;,R) � (- av ay a2 v ay ) ay2 ' and so v is harmonic as well. (Alternately, we could have also noted that v □ = Re(-if).) Now we show Theorem 5.2, which gives a converse to the above result when the open set in question is a simply connected domain. As mentioned earlier, for more general domains, it can happen that there are harmonic functions which aren't globally the real part of a holomorphic function; see Exercise 5.5, where we take (1) (2) U := JR.2 \ {(O, 0)} (which is not simply connected) and u := log(x 2 + y 2 ) (which is harmonic in U). Then there is no holomorphic f in (C \ {O} such that u = Re(!) in U. Proof. (of Theorem 5.2.) We will construct a holomorphic f with real part u, and then v First set := Im(!) will serve as the required harmonic function. au .au g=--i. ax ay We will prove that g is holomorphic, and then to construct a primitive of g, necessarily holomorphic, which will be the f we seek. To show that g 168 A Friendly Approach to Complex Analysis is holomorphic, we will use Theorem 2.2. First we note that since u is harmonic, the functions Re(g) OU = ox and Im(g) = - OU oy have continuous partial derivatives. Moreover, using the fact that u satisfies the Laplace equation, we can now see that the real and imaginary parts of g satisfy the Cauchy-Riemann equations: 2 2 = (Re(g)) = O = = = (Im(g)), OX x X x2 y y2 y y o (ou) o u o u o ( ou) o o o o -o o ou ou ou = !_ ( ) = !_ ( ) = _!_ (- ) = _!_(Im(g)). o!_(Re(g)) oy ox ox oy ox oy ox y o o Hence g is holomorphic in U, and by Theorem 3.5, since U is simply con­ nected, it has a primitive G in U. Decompose G = + into its real and imaginary parts Then u iv u, v. i ov ox + ox' OU - i OU = ' OU g=G = OX y o and so (5.1) o(u-u) =O, OX and so it follows from the Fundamental Theorem of Integral Calculus, that u - u is locally constant along horizontal lines. Also, (5.1) gives OU OV (C_:_R) oy ox OU oy ' where have used the Cauchy-Riemann equations to obtain the last equality. Hence o(u-u) =0 oy showing that u - u is locally constant along vertical lines as well. Since U is a domain, any two points in U can be joined by a stepwise path, and so it follows from here that u - u must be a constant, say C ( E ffi.), in U. Consequently, f == a - c = (u - c) + iv= u + iv, and so we see that f is holomorphic in U, u = Re(!), v := v = Im(G) is □ harmonic in U and f = u + iv. 169 Harmonic functions Definition 5.2. Let U be an open set and u : U-+ IR be harmonic in U. v : U -+ IR such harmonic conjugate of u. Any harmonic function U is called a Example 5.4. u v := 2xy that f := u + iv is holomorphic in is a harmonic conjugate of the harmonic function := x2 -y2 considered in Example 5.1 because f where := u +iv= x2 -y2 + 2xyi = (x + iy)2 = z2, z = x + iy, ◊ is entire. Harmonic conjugates are obviously not unique since we can just add a con­ stant to a harmonic conjugate and get a new harmonic conjugate. In a simply connected domain, for each harmonic function, Theorem 5.2 guar­ antees the existence of a harmonic conjugate, but it is not particularly useful for finding a harmonic conjugate (from the above proof, we see that it relies on the construction of the primitive G). A more direct way is to look at the Cauchy-Riemann equations, as shown in the following example. Example 5.5. Let u = -(sinx)(sinhy). Then �:� = +(sinx)(sinh y), �:� = -(sinx)(sinh y), and so u is harmonic in IR2 . As IR2 is simply connected, we know that there is a harmonic conjugate. Can we find one? If then we know that u v is a harmonic conjugate, + iv will be holomorphic, and so the pair u, v satisfy the Cauchy-Riemann equations. So we need av such that av OU = - = +(smx)(coshy), ay ax av au . = -(cosx)(smh y). = ax ay Integrating (5.2) with respect to v= for some constant C(y) is C(y) ( )( . h sm (5.3) while keeping y fixed, we see that depending on y (5.2) -(cosx)(coshy) + C(y), differentiable, we have - cos X x ) (5.3) = must y. (5.4) So assuming for the moment that ov (5.4) ( )( . h ) dC ( ) = - cos X sm y + d y . ay y 170 A Friendly Approach to Complex Analysis Consequently dC (y) = 0, dy and so C(y) = C for all y. So we could take any constant C, and in particular we may choose C = 0. So based on the above rough reasoning, we take v = -(cosx)(coshy) as a candidate for a harmonic conjugate of u. In order to verify this guess, we can quickly note that the Cauchy-Riemann equations are satisfied by the pair (u, v), so that f := u + iv is holomorphic. But instead, let us use an old calculation we performed in Exercise 1.37 to check directly that f := u + iv is holomorphic by actually finding the f: f = u +iv= -(sinx)(sinhy) - i(cosx)(coshy) = -i ((cosx)(cosh y) - i(sinx)(sinh y)) = -i cos z, ◊ which is entire. Exercise 5.4. Find harmonic conjugates for the following harmonic functions in ll�.2 : 2 3 x e siny, x - 3xy - 2y, x(l + 2y). Exercise 5.5. Show that there is no holomorphic function f defined in C \ {O} whose real part is the harmonic function u defined by u(x,y) = log(x2 + y 2), (x, y) E JE. 2 \ {(O, O)}. Hint: If v is a harmonic conjugate of u, then also h(z) := z 2 exp( -(u + iv)) is holomorphic. Find lhl, and conclude that h' = 0. Show that h' = 0 implies that 2/z has a primitive in C \ {O}, which is impossible. Exercise 5.6. Is it possible to find a v : JE.2 -+ lE. so that f defined by J(x + iy) = x3 +y 3 +iv(x,y), (x,y) E JE. 2 , is holomorphic in C? 5.3 Consequences of the two way traffic: holomorphic ++ harmonic In the previous section, the two results given in Theorems 5.1 and 5.2 show that there is a two way traffic between the real analysis world of harmonic functions and the complex analysis world ofholomorphic functions, allowing a fruitful interaction between the two worlds. 171 Harmonic functions Holomorphic functions In the previous three chapters we have learnt many pleasant properties possessed by holomorphic functions. Let us now use some of these to derive some important properties of harmonic functions. In particular, we will show the following results in this section: (1) If u is harmonic, then it is c= . (2) The Mean Value Property for harmonic functions. (3) The Maximum Principle for harmonic functions. (4) Uniqueness of solutions to the Dirichlet Problem. 5.3.1 Harmonic functions are smooth Corollary 5.1. Harmonic functions are infinitely many times differen­ tiable. (Note that the definition of a harmonic function demands only twice con­ tinuous differentiability. The remarkable result here say s that thanks to the fact that the Laplace equation is satisfied, in fact the function has got to be infinitely differentiable. A result of this ty pe is called a regularity result in PDE theory.) Proof. zo = U. Let r > 0 such that the disc D with center Suppose that u is a harmonic function in an open set (xo, Yo) E U. Then there is a zo and radius r is contained in U. But ulv is harmonic in D and D is simply connected. So there is a holomorphic function f defined in that Re(!) in = u in D. D, such But f is infinitely many times complex differentiable D. Consequently, u is infinitely many times differentiable in D, and in particular at zo E D. As the choice of zo E D was arbitrary, the result follows. □ Exercise 5.7. Show that all partial derivatives of a harmonic function are har­ monic. 172 5.3.2 A Friendly Approach to Complex Analysis Mean value property Using the Cauchy Integral Formula, we immediately obtain the following "mean value property" of harmonic functions, which says that the value of a harmonic function is the average (or mean) of the values on a circle with that point as the center. Theorem 5.3. (Mean-value property of harmonic functions) Let (1) U be an open set, (2) u: U-+ ffi. be harmonic in U, (3) Zo EU, (4) R > 0 be such that the disc {z EC: lz - zol < R} CU. 271" 1 Then u(zo) = - 1 u(zo + rexp(it))dt for all r such that O < r < R. 21r 0 Proof. The disc D := {z EC : lz - zol < R} is simply connected, and so there is a holomorphic function f defined in D, whose real part is u. But now by the Cauchy Integral Formula, if C is the circular path given by C(t) = zo + rexp(it) (t E [O, 21r]), then 2 f(zo + rexp(it)) . . f(z) =1 1 71" -dz 1 ------irexp(it)dt f(zo ) =r exp(it) 21ri c z - zo 21ri O 271" = _!_ f(zo+rexp(it))dt. 21r lo 1 f Equating real parts, the claim is proved. 5.3.3 D Maximum Principle From the Maximum Modulus Theorem (see page 129), we also obtain the following. Theorem 5.4. (Maximum Principle) Suppose that (1) U is a simply connected domain, (2) u: U-+ ffi. is harmonic in U, (3) z0 EU is a point such that u(z0) 2: u(z) for all z ED. Then u is constant in U. There is a holomorphic function f defined in U whose real part equals u. But then the function g : U -+ C defined by g(z) = exp(f(z)) Proof. 173 Harmonic functions (z EU) is holomorphic too. We have lg(zo)I = I exp(f(zo))I = e Re(zo) = eu(zo) � e u(z) = lg(z)I for all z EU. By the Maximum Modulus Theorem applied tog, it follows that g must be constant in U. Thus lgl is also constant in U, that is, /gl = e Re(f) = e u is a constant in U. Taking the (real) logarithm, it follows that u is constant in U. □ 5.4 Uniqueness of solution for the Dirichlet problem The Maximum Principle has an important consequence about the unique­ ness of solutions to the Dirichlet problem, as explained below. Let := 11' := ][J) {z EC: /z/ < 1}, {z EC: lzl = l}. r.p lives here Then the Dirichlet problem is the following: Given r.p : 11' -+ IR, continuous, find a u : ][J) U 11' -+ JR such that ( 1) u is continuous on ][J) U 11', (2) uhr = r.p, (3) u has continuous partial derivatives of order (4) -6.u = 0 in ][J). 2 in ][J), The given function r.p is called the boundary data. The reason one is interested in solving the Dirichlet problem is that the need arises in ap­ plications, for example in heat conduction, electrostatics, and fluid flow. Using the Maximum Principle, we can show the following. Proposition 5.1. The solution to the Dirichlet problem is unique. 174 A Friendly Approach to Complex Analysis Proof. Indeed, let u1, u2 be two distinct solutions corresponding to the boundary data cp. In particular, u1 = u2 on 1l'. So u1 must differ from u2 somewhere inside lDl. Without loss of generality suppose that there is a point w E lDl where u1(w) > Then u := u1 - u2 is such that u2 (w). (Otherwise exchange their labels.) (1) u=0on'll', (2) u(w) > 0, (3) u is harmonic in lDl. Let zo E [)) U 1l' be the maximizer for the real-valued continuous function u on the compact set [)) U 1l'. From (1),(2), zo e/. 1l'. So zo E IDl. We have u(z0) 2: u(z) for all z E IDl, and so by the Maximum Principle, u must be constant in IDl. But u is continuous on[)) U 1l', and u is 0on 1l'. So it follows that the constant value of u must be 0everywhere in IDlU'll'. Hence u1 = u2, a contradiction. D Remark 5.1. It can be shown that the following expression, called the Poisson Integral Formula, boundary data cp: u(rexp(it))= - 1 21r 1 o 2 -rr gives the solution to the Dirichlet problem with 1 1- r2 - 2rcos (O - t) + r2 cp(exp(i0))d0 (( E 1l'). This can be derived using the Cauchy Integral Formula, but there are some technical subtleties, and so we will not prove this here. Exercise 5.8. (Some half-plane Dirichlet problems.) Given the "boundary data" b : JR ➔ JR, we consider the problem of finding a continuous, real-valued function h defined in the closed upper half-plane y 2: 0, such that h is harmonic in the open upper half-plane y > 0 and moreover, h(x, 0) = b(x). (1) If b is just a polynomial p, then show that we can simply take h given by h(x,y) = Re(p(x+iy)). (2) Prove that if b(x) = 1 -1+x -2, then (x,y) I-+ Re(b(x + iy)) is not a solution (because of the pole at z Show that 1 2 y = 2 h(x,y) := Re t z+i X + y+ 1) (--i .) gives a solution to the Dirichlet problem. Exercise 5.9. Let u: JR2 ➔ JR be a harmonic function such that (x,y) E JR2 . Prove that u is constant. u(x, y) = i). > 0 for all Hint: Let f be an entire function whose real part is u. Consider exp(-!). Harmonic functions 175 Exercise 5.10. The regularity of functions satisfying the Laplace equation is not completely for free. Here is an example to show that a discontinuous function may satisfy the Laplace equation! Consider the function u : IB.2 ➔ ffi. defined to 4 be the real part of e-l/z when z #- 0 and O at the origin. (1) Verify that u is discontinuous at 0. (2) Check that u(x, 0) = e-l/x 4 , u(0, y) = e-l/y . 4 (3) Being the real part of a holomorphic function in C\ {O}, we know already that u satisfies the Laplace equation everywhere in IB.2 \ { (0, O)}. Show that also 2 {J u 8x2 exist, and that 82 u ax 2 (0, 0) + (0, 0) and 82 u 8y2 (0, 0) 82 u ay2 (0, 0) = 0. Exercise 5.11. Let D1, D2 be domains in C. Let rp : D1 ➔ D2 be holomorphic. Show that if h : D2 ➔ ffi. is harmonic, then h o rp: D1 ➔ ffi. is harmonic as well. Now suppose that rp : D1 ➔ D2 is holomorphic, a bijection, and also that rp- 1 : D2 ➔ D 1 is holomorphic. We call such a map rp a biholomorphism. Con­ clude that a function h : D2 ➔ ffi. is harmonic if and only if h o rp : D1 ➔ ffi. is harmonic. Thus the existence of a biholomorphism between two domains allows one to transplant harmonic (or even holomorphic) functions from one domain to the other. This mobility has the advantage that if D1 is "nice" (like a half plane or a disc), while D2 is complicated, then problems (like the Dirichlet Problem) in D2 can be solved by first moving over to D1, solving it there, and then transplanting the solution to D2. A first natural question is then the following: Given two domains D1 and D2, is there a biholomorphism between them? An answer is provided by the Riemann Mapping Theorem, a proof of which is beyond the scope of this book, but can be found for example in [Conway (1978)]. Theorem 5.5. (Riemann Mapping Theorem) Let D be a proper (that is, D #- q simply connected domain in C. Then there exists a biholomorphism rp: D ➔ lill := {z EC: lzl < l}. Thus the above result guarantees a biholomorphism between any two proper simply connected domains (by a passage through llll). Unfortunately, the proof does not give a practical algorithm for finding the biholomorphism. Show that the "Mobius transformation" rp: lHI ➔ llll, where lHI := {s EC: Re (s) > O}, given by s-1 , s E lHI, rp(s) = s+ l is a biholomorphism between the right half plane lHI and the disc llll. 176 5.5 A Friendly Approach to Complex Analysis Notes Exercises 5.5, 5.6 and the proof of Theorem 5.2 is taken from [Beck, March­ esi, Pixton, Sabalka (2008)]. Exercise 5.8 is based on [Flanigan (1973)]. Solutions Solutions to the exercises from the Introduction Solution to Exercise 0.1 Suppose the derivative of f' at O exists, and is equal to L. Then taking E := 1 >0, there exists a 8 >0 such that whenever0 < Ix -0I < 8, we have I f'(x) - f'(O) x-0 - LI < E. In particular, with x := 8/2, we have0 < Ix -0I = 8/2 < 8, and so there must hold that f'(x) - f'(O) 2(8/2)-0 = 1 - LI = 12 - LI < E. (5.5) x-0 (8/2) -0 I LI On the other hand, with x := -8/2, we have0 < Ix -0I = 8/2 < 8, and so again there must hold that I f'(x) - f'(O) x-0 LI = ,-2(-(-8/82/2)-)-00 - LI = 12+ LI < E. (5.6) From (5.5) and (5.6) it follows, using the triangle inequality for the real absolute value, that 4 = l2+L+2-LI::::; l2+LI + 12- LI < E+ E = 2E = 2, a contradiction. Hence f' cannot be differentiable at0. 177 178 A Friendly Approach to Complex Analysis Solutions to the exercises from Chapter 1 Solution to Exercise 1.1 Since (x,y) =/- 0, at least one among x,y is nonzero, and so x2+ y2 =/- 0. Thus ( ( X -y ) E JR2 x2+y2 ' x2+y2 Moreover, (x,Y) · - = X ( x2+y2 -y ' x2+y2 ) ) ( X X -y -) X• (--y -)+y -X•--- y• -· x2 yx2+y2 x2+y2 ' x2+y2 + 2 x2+y2 -xy+xy (x2 y2 ' x2 y2 ) = (1,0). + + Hence for (x,y) =/- (0,0), we have (x,y)-1 = ( x X2+y2 , -y X2+y2 ) in C. Solution to Exercise 1.2 Since 0 E (-i, i), tan0 ER We have __1__ --,------,-1---,--,tan 0 +i ( 2 = 2 2) 2 +(tan0) 1 1- itan0 1 +(tan0) sin 0 _ _ _ . (cos0)2 2 (cos0) c 0 =----) +i ( _ �os�---(cos0)2+(sin0)2 (cos0)2+(sin0)2 = . (sin0)(cos0) (cos0)2 • +i = (cos 0 )2+i"(sm 0 )(cos 0 ). 1 1 Hence .. 1+itan0 . = (l+itan0)((cos0)2+i(sm0)(cos0)) . 1- itan 0 sin0 . = (cos0)2 - - · (sm0)(cos0) cos 0 +i ((sin0)(cos0)+ :�:: • (cos0)2 ) = (cos0)2 - (sin0)2+i2 (sin0)(cos0) = cos(20)+isin(20). 179 Solutions Solution to Exercise 1.3 Let P C C be a set of positive elements of C. Then since i -/=- 0, by (P3), either i E P or (if/. P and -i E P). By (P2), we have -1 = i · i = ( -i) · ( -i) E P. (5.7) 1 = (-1) · (-1) E P. (5.8) Again by (P2), But 1-/=- 0, and (5.7), (5.8) contradict (P3) for x = l. Solution to Exercise 1.4- See Figure 5.2. cos j i 0 + isin j 1 -v'2i Fig. 5.2 Location of the complex numbers 0, 1, -3/2, i, -v'2i, cos�+ isin�- 3 3 180 A Friendly Approach to Complex Analysis Solution to Exercise 1.5 We have for 0 E � that (cos 0+ isin 0) 3 (cos0+ isin0)3 = cos(30)+ isin(30). But = (cos0+ isin0)((cos0)2 - (sin0)2 + i2(cos0)(sin0) = (cos0)((cos0)2 - (sin0)2 ) - (sin0)2(cos0)(sin0) +i(· .. ) . Hence equating the real parts on both sides, we obtain cos(30) = Re((cos0+ isin0)3 ) = (cos0)((cos0)2 - (sin0)2 ) - 2(cos0)(sin0) 2 = (cos0)((cos0)2 -1+ (cos0)2 )-2(cos0)(1-(cos0)2 ) = (cos0)3 - cos0 + (cos0)3 - 2 cos0 + 2(cos0)3 = 4(cos0)3 - 3cos0. Alternatively, the Binomial Formula (a + b) n = to hold for a, b E C, n E N and so t (�) k=O n k ak b - continues cos(30) = Re((cos0+ isin0)3) = Re((cos0)3 + 3 (cos0)2 (isin0)+ 3 (cos0)(isin0)2 + (isin0)3 ) = (cos0)3 - 3 (cos0)(sin0)2 = 4(cos0)3 - 3cos0. Solution to Exercise 1. 6 We have 1+ i = J2 ( � + i�) (1+ i) 10 = J2 (cos�+ isin�). Hence 10 = (0) 10 (cos�+ isin�) = 25 (cos (10 • �)+ isin (10 • �)) = 32 (cos (21r+ i) + isin (21r+ i)) = 32 (cos (i) + isin (i)) = 32(0+ i -1) = 32i. Solution to Exercise 1. 7 The angle made by 2+ i with the positive real axis is tan- 1 (1/2), and the angle made by 3+i with the positive real axis is tan- 1 (1/ 3 ). Thus the angle Solutions 181 made by (2+i)(3+i) with the positive real axis is tan-1(1/2)+tan-1(1/3). On the other hand, since (2 + i)(3 + i) = 6 - 1 + i(2 + 3) = 5 + 5i, the angle made by (2 + i)(3 + i) with the positive real axis is tan- 1 (5/5) = tan- 1 1 = n:/4. Consequently, 7r: 11 11 4 = tan- 2 + tan- 3 Solution to Exercise 1. 8 Suppose that the vertices A, B, C of the equilateral triangle are at the complex numbers ZA, ZB, zc, and that they are labelled in the anticlockwise fashion. Since t'(AC) = t'(AB) and L.CAB = n:/3, we have zc - ZA = (cos i + i sin i) (zB- ZA)- (5.9) We argue by contradiction and let p, q, m, n E Z be such that zc - zA = p + iq, and ZB - ZA = m + in. Then (5.9) becomes p + iq = ( � + � i) (m + in), that is, p= q = m J3 n, and 2 2 m-v'3 n -2-+2• (5.10) (5.11) Thus (by multiplying (5.10) by -n and (5.11) by m and adding), we obtain qm - pn = v'3 (m2 + n2 ). 2 But m 2 + n2 =/. 0 (since ZB =/. ZA!), and so we obtain V3 = a contradiction. 2(qm - pn) E (Ql m2 +n2 182 A Friendly Approach to Complex Analysis Solution to Exercise 1. 9 We write -1 = 1 • (cos1r + i sin1r),and we seek w = p(coso: + i sino:) such that w4 = p4 (cos(4o:) + i sin(4o:)) = 1 • (cos1r + i sin1r). Hence p4 = 1 and sop= l. Also,we have 4o: E {1r,1r ± 21r,1r ± 41r,•••},and so o:E{i, �±i, ±1r, � ···}· So we get w = p(cos a + i sin a) = 1 (cos a + i sin a) belongs to 31r . . 31r 51r . . 51r 71r . . 71r cos +ism 4, COS 4 4 +iSlll 4, COS 4 +iSlll 4, COS 4 +iSlll 4} { 7r • • 7r - l+i -l+i -1-i 1-i { }. - J2' J2' J2'J2 The four fourth roots of unity are depicted in the complex plane in Fig­ ure 5.3. 1-i v'2 Fig. 5.3 Location of the complex numbers w that satisfy w 4 = - 1. Solution to Exercise 1.10 We have 0 = z 6 - z 3 - 2 = (z 3 ) 2 - 2 z 3 + z 3 - 2 = (z 3 - 2) (z3 + 1), 183 Solutions and so z 3 = 2 or z 3 = -1. The equation z 3 = 2 holds if and only if . . 21r . . 41r 3tn (cos 21r +ism 3tn (cos 41r +ism 3tn} , z E { v2 ) , v2 ) , v2 3 3 3 3 that is, J3) 2 , 3tn (- 1 +i. zE { v2 2 3tn (- 1 - i. v2 2 J3) 2 , 3tn} . v2 On the other hand, z3 = -1 holds if and only if 51r . . 51r . . 1r . . 1r +ism } z E { cos +ism , cos 1r +ism 1r, cos that is, 3 3 3 vf3 3 , vf3 zE{!+i -1 !_i }· 2 2' ' 2 2 So z6 - z 3 - 2 = 0 if and only if [(z 3 if that is, = 2) or (z 3 = -1)], that is, if and only J3) J3) 1 . vf3 . vf3 . . 3tn (--1 +i3tn (--1 -i3tn v2 +i- -1 -1 -i- } . v2 z E { v2 2 2 ' 2 2 ' '2 2' '2 2 Solution to Exercise 1.11 Suppose w E (C \JR is such that w 3 = 1. Then (w- l)(w 2 + w +1) = 0, and since w I- 1, we have w 2 + w + l = 0. Hence ((b - a)w+(b - c))((b - a)w 2 + b - c) = (b - a) w + (b - a)(b - c)(-1) + (b - c)2 = (b- a)2 · 1 + (b - a)(b - c)(-1) + (b- c)2 = (b - a)(b - a - b + c) + (b - c)2 = (b - a)(c - a)+ (b - c)2 = (be - ca - ab+ a2 + b2 - 2bc + c2 2 = a2 3 + b2 + c2 - ab - be - ca = 0. 184 A Friendly Approach to Complex Analysis Hence either (b - a)w = c - b or (b - a)w2 = c - b. But the latter case is the same as (b- a)w3 =(c-b)w, that is, (c-b)w =b- a. So we have that lb- al =le-bl, and the angle between the line segments joining a to b and a to c is 7f /3; see Figure 5.4. a ' J_, ' � a or 2,r ' 3 a, b, c Fig. 5.4 ' 2,r 3 form an equilateral triangle. In either case, we obtain that the triangle formed by a, b, c is equilateral. If a, b, c are all real, then the equilateral triangle must degenerate to a point r E IR, and so a =b =c ( = r). Thus we recover the real case result. Solution to Exercise 1.12 Let w E C\IR be such that w3 = 1. Since (1-w)(l+w+w2) = ,0 and w =/- 1, we have l+w+w2 = 0 . Also, l+w2 +w4 = l+w2 +w-w3 =l+w2 +w = .0 We have 3 n (l + l) n + (l + w) n + (l + w ) n =� 3 2 3 3 ( ) But 1 +1 +1 if k (1 + wk + w2k ) = { 1 + w + w2 if k 1 + w 2 + w 4 if k _ 3 if k { 0 if k 0 if k =0 =1 =2 Thus (l + l)3n + (1 + w)3n + (1 + w2 )3n = 3 . ( 3 ; (l + wk + w2k ). =0 =1 =2 mod 3, mod 3, mod 3 mod ,3 mod ,3 mod .3 (3;) (3;) + + ...+ G:)). Solutions 185 But also (1 + 1)3n + (1 + w)3n + (1 + w2)3n = = 23n + (-w2)3n + (-w)3n 23n + (-lt + (-l)n = 23n + 2(-l)n, and so the claim follows. Solution to Exercise 1.13 Q ' P B' Fig. 5.5 RP and SQ have equal lengths and meet at right angles. See Figure 5.5. Let the points A, B, C, D in the plane correspond to the complex numbers a, b, c, d, respectively. Since AB' is obtained from AB by rotating AB about A in an anticlockwise fashion by 90 ° , we have that B' corresponds to the complex number a - i(b - a). Since Pis the midpoint of BB', it follows that P corresponds to a+ b - i(b - a) 2 Similarly, Q, R, S correspond to c + d - i(d - c) d + a - i(a - d) b + c - i(c - b) 2 2 2 186 A Friendly Approach to Complex Analysis respectively . If we denote the complex numbers corresponding to P, Q, R, S, by p, q, r, s, respectively, then · b+c-i(c-b) - d+ a-i(a-d) . iq-s ( ) =i ( ) 2 2 -b+c- a+d+ib ( + c-d- a) 2 b-----'a) + c+d-i( d-c) = -------'= -a-b+i( ---'------'-p + r. 2 2 Hence lq - sl =IP - rl, showing that f(QS) =f(PR). Also, since mul­ tiplication by i produces a rotation about the origin by 90 °, we see that PR J_ QS. Solution to Exercise 1.14 Let z1 =x1 + iy1, z2 =x2 + iy2, where x1, x2, Y1, Y2 belong to JR. Then z1z2 =X1X2 - Y1Y2 +i(X1Y2 + Y1X2 ), and lz1z2 1 2 =(x1x2 - Y1Y2 )2 + (x1Y2 + Y1X2 )2 =XiX�- �+ YiY�+ XiY�+ �+ YiX� =Xi (X�+ y�)+ Yi (Y�+ x�)=(xi + yi )(x�+ y�) =lz1 l 2 lz2 1 2 Since lz1I, lz2I, lz1z2I are all nonnegative, it follows that lz1z2I =lz1I lz2ISolution to Exercise 1.15 Let z =x+iy, where x, y E R Then (z)=X-iy=X-i(- y)=X+iy=Z. Also, zz=(x+iy)(x-iy)=x2 + y2 +i(-xy+ xy)=x2 + y2 =lzl 2Finally, z+ z x+ jef+ x- jef 2 x = =x=Re (z) and = 2 2 2 z- z ;i+iy-;i+iy 2iy =y=Im (z). = = 2i 2i 2i 187 Solutions Solution to Exercise 1.16 Let z = x+iy, where x, y E R Then lzl =Ix+iyl = Jx 2 +y2 = Jx 2 +(-y) 2 =Ix- iyl = lzl, IRe(z)I = lxl = H::; Jx 2 +y2 =Ix+iyl = lzl, Jx 2 +y2 = lx+iyl =lzl. IIm(z)I =IYI = #::; Since z is obtained by reflection of z in the real axis and since O E �, the distance ofz to O is equal to the distance ofz to 0. In other words, lzl = lzl. The inequalities IRe(z)I ::; lzl and IIm(z)I ::; lzl just say that the length of any side in a right angled triangle is at most the length of the hypotenuse. z Solution to Exercise 1.17 First we note that lazl = lal lzl = lal lzl < 1-1 = 1, and so az-=/- 1. We have z-a z-a z-a ( z -a ) zz-az-az + aa . . = = 1-az 1-az 1-az-az + aazz 1-az 1-az lzl2 -az-az + lal2 1-az-az + lal2 lzl2 1-az-az + lal2 1zl2 + lzl2 + lal2 -1-lal2 lzl2 1-az-az + lal2 lzl2 2 2 2 lzl +lal -1-lal2 lzl = 1 + -------,---,,.,.,-cc:--1-az-az+lal2 lzl2 2 lzl + lal2 -1-lal2 lzl2 =1+ ll-azl2 _ l _ (1-lzl2 )(1-lal2 ) · 11-azl2 188 A Friendly Approach to Complex Analysis Thus I 2 (1- lzl2 )(1- lal2 ) < z-a 1 = 1- 1- 0 = 1. 1- az 11- azl2 2:0 as [z[::;1, [a[<l Solution to Exercise 1.18 Let w E C be such that p(w) = 0, that is, eo + c1 w + • • • + cd wd Co+ C1 W + • · · + Cd W d = 0 = 0, and so, using the fact that the Ck = 0. Hence are all real for O � k � d, we obtain 0 =Co+ C1 w + ... + Cd W d = Co+ C1 w + ... + Cd W d =Co+ ciw + · · · + Cd W d = Co+ C1W + · · · + cd (w)d where the last equality follows from the observation that w )k w k = -- = - - = (1 � k � d. � � k times k times Consequently, 0 = c0 + c1w + · · · + cd (w)d =p(w). Solution to Exercise 1.19 Let a= lal(cosa+ i sina), and b = lbl(cos,B + i sin,B), where a, ,B E [O, 21r). Then ab= lal(cosa+ i sina) · lbl(cos,B - i sin,B) = !al lbl(cosa + i sina)(cos,B - i sin,B), and so Im(ab) = !al lbl(-(cosa)(sin,B) + (sina)(cos,B)) = !al lbl sin(a -,B). lal A a Bb Fig. 5.6 The area of .6.OAB formed by the triangle with vertices at 0, a, b. 189 Solutions = The area of the triangle OAB formed by 0, a, b (where O B b) is 1 1 . . 2 f(OA)-f(OB)-smL'.AOB = 21al·lbl·lsm(a-,B )I But for a square matrix M . l For z1, z2, Z3 EC, we have = I I 1 Im(ab) = 1Im(ab)I = 2 2 Solution to Exercise 1.20 1 Z1 Z1 w := i · det [1 z2 z2] 1 Z3 Z3 = 0, A= a and 1 Z1 Z1 -i · det [1 z2 z2 . 1 Z3 Z3 = [mi1], detM = L (sign o-) · mia(i), aESn where Sn is the set of all permutations on the set { 1, • • • , n}. Hence detM = L (sign o-) · mia(i) = detM, aESn where M is the matrix obtained from M by taking entry wise complex con­ jugates. Hence 1 Z1 Z1 det [ 1 z2 z2 1 Z3 Z3 l 1 Z1 Z1 = det [1 z2 z2 1 Z3 Z3 l [ l 1 Z1 Z1 = - det 1 z2 z2 , 1 Z3 Z3 where the last equality is obtained by interchanging the second and the third columns. Consequently, 1 Z1 Z1 i · det [ 1 z2 z2 1 Z3 Z3 l [ l 1 Z1 Z1 1 z1 z1 1 Z3 Z3 1 Z3 Z3 = -i · det 1 z2 z2 = -i · ( - det [1 z2 z2 1 Z1 Z1 = i · det [1 z2 z2] 1 Z3 Z3 Thus w is its own complex conjugate, and hence it is real. l) 190 A Friendly Approach to Complex Analysis Solution to Exercise 1.21 We have 2 2 lz1 + z21 + lz1 - z21 = (z1 + z2)(z1 + z2) + (z1 - z2)(z1 - z2) = Zl • Zl + Zl · Z2 + Z2 • Z1 + Z2 • Z2 +z1 · z1 + z1 · (-z2) + (-z2) · z1 + (-z2) · (-z2) 2 2 2 2 = lz1l + � +� + lz2l + lz1l - � - � + lz2l 2 2 = 2(lz1l + lz2l ). Consider the paralellogram P with vertices at 0, z1, z2, z1 +z2 in the complex plane. Then lz1 +z2I denotes the length of one diagonal of P, while lz1 -z21 is the length of the other diagonal of P. Also, lz1I, lz21 are the lengths of the sides on P. So the above equality says "In a parallelogram, the sum of the squares of the lengths of the diagonals equals twice the sum of the squares of the sides." Solution to Exercise 1.22 For z1, z2 EC, we have lz1I = lz1 - z2 + z2I � lz1 - z2I + lz2I, and so lz1I - hi� lz1 - z2ISince this holds for all z1, z2 EC, by swapping z1 and (5.12) z2 in (5.12), lz2I - lz1I � lz2 - z1I = I - (z1 - z2)I = I - 11 lz1 - z2I = lz1 - z2I- (5.13) (5.12) and (5.13) give llz1I - lz2II � lz1 - z2ISolution to Exercise 1.23 (1),(2),(3): Fig. 5.7 Left to right: The set of points described by lz - (1- i)I = 2, lz - (1- i)I < 2 and 1 < lz - (1- i)I < 2, respectively. Solutions 191 (4) Let z = x + iy, where x, y ER Then Re(z - (1- i)) = 3 if and only if x - 1 = 3, that is x = 4. 4 0 (5) Let z = x + iy, where x, y ER Then IIm(z - (1- i))I < 3 if and only if IY + 11 < 3, that is -4 < y < 2. (6) {z EC: lz - (1- i)I = lz - (1 + i)I} is the set of all complex numbers z whose distance to 1-i is equal to its distance to 1 + i. So it is the set of all points which lies on the perpendicular bisector of the line segment joining 1- i to 1 + i. So the set is the real line �l+i 1-i Fig. 5.8 The set of points z satisfying [z - (1 - i)[ (7) The equation lz - (1 - i)I + lz - (1 + i)I = [z - (1 + i)[ is R = 2 says that the sum of the 192 A Friendly Approach to Complex Analysis distances of z to 1 + i and to 1-i is 2. But the distance between 1-i and 1 + i is 2. So z lies on the line segment joining 1-i to 1-i. This conclusion can also be arrived at analytically. If z = x + iy, where x,y E �, then 2 = J(x -1) 2 + (y + 1) 2 + J(x -1) 2 + (y -1) 2 ?:': IY + 11 + IY-11 ?:': 1 +JI+ 1-JI= 2. Thus IY + 11 + IY-11 = 2 and x = 1. l+i 1-i Fig. 5.9 The set of points joining 1 - i to 1 + i. z satisfying lz - (1 - i) I+ iz - (1 + i)I = 2 is the line segment (8) The set of z such that lz -(1-i)I + lz -(1 + i)I = 3 lies on an ellipse E with foci at 1 + i and 1-i, and so {z E (C: lz - (1-i)I + lz - (1 + i)I < 3} is the region in the interior of the ellipse E. Fig. 5.10 The set of points the ellipse E. z satisfying lz - (1- i)I + lz - (1 + i)I < 3 is the interior of 193 Solutions Solution to Exercise 1.24- Co s· C1 ···+ zd-l + zd) . mce lcd-1 1 +...+ lc �I + �)= O, lim n ➔ oo ( n nd 1 nd there exists an N large enough so that lcoI l iI lcdl lcd- 1 I +···+ ed-1 < 2· +Nd N � Hence for lzl> N =: R, we have c cd- 1 d lp(z)I= l z lled + + od I +...+ � z zd-1 z cd l d � lzl (lcdl- ; +···+ �� 1 + :� z l ed- i i + + +� � l z ld ( lcdl- ( · ·· zd-1 z zd )) For z r 0, we have p()z = z _j_ C d- 1 d( Cd + -+ z I hl I) � lzld (1 cdl- c � +...+ )�� 1 + c il lc I lc I • � lzld (lcdl- ; ) = ; lz ld �J)) .._.,..., =:M Solution to Exercise 1.25 ("If' part) Suppose that the two real sequences (Re(Zn)) nE fil and (Im(zn)) nEN are convergent respectively to Re(L) and Im(L). Then given E > 0, lzn - LI = J(Re(zn) - Re(L)) 2 +(Im(zn) - m I (L))2 2 2 =E < (J2) + (J2) for n > N, where N is large enough to ensure that for all n > N, and IIm(zn) - m I (L)I< IRe(zn) - Re(L)I < J2 So (zn) nEN converges to L. J2" ("Only if' part) Suppose (zn) nEN converges to L. Given E > 0, let N be such that for all n > N, l z - LI< E. But then for n > N, IRe(zn) - Re(L)I= IRe(zn - L)I� lzn - LI< E, and I (zn - L)I� lzn - LI< E. I (L)I= Im IIm(zn) - m Hence the two real sequences (Re(zn)) nEN and (Im(zn)) nEN are convergent respectively to Re(L) and Im(L). 194 A Friendly Approach to Complex Analysis Solution to Exercise 1.26 ("Only if' part) Suppose (zn )nE N converges to L. Then (Re(zn ))nE N and (Im(zn ))nE N converge to Re(L) and Im(L), respectively. Hence (Re(zn ))nE N and (-Im(zn ))nE N converge to Re(L) and -Im(L), respectively, that is, (Re(zn ))nE N and (Im(zn))nE N converge to Re(£) and Im(£), respectively. Consequently (zn)nE N converges to L. ("If' part) Suppose (zn)nE N converges to L. By the previous part, ((zn ))nE N converges to (£), that is, (zn )nE N converges to L. Solution to Exercise 1.27 Let (zn )nE N be a Cauchy sequence in C. The inequalities IRe(zn ) - Re(zm)I = IRe(zn - Zm)I S lzn - Zml and IIm(zn ) - Im(zm)I = IIm(zn - Zm)I S lzn - Zml show that the two real sequences (Re(zn ))nE N and (Im(zn ))nE N are then also Cauchy sequences. Since � is complete, they are convergent to, say, a, bE�, respectively. But then (zn )nE N converges in (C to a+ ib. Hence (C is complete. Solution to Exercise 1.28 Let zoEC and E > 0. Set 8 = E > 0. Then whenever lz - zol < 8, we have IRe(z) - Re(zo)I = IRe(z - zo)I S lz - zol < 8 = E. So z f---t Re(z) is continuous at zo. Since the choice of zoEC was arbitrary, it follows that z f---t Re(z) is continuous (in C). Solution to Exercise 1.29 Let U:= {zEC: Re(z) • Im(z) > 1}, and set F:= CU (the complement of U). If (zn )nE N is a sequence in F such that (zn )nE N converges to L in C, then we have Re(zn ) · Im(zn ) S 1 for all nEN, (5.14) and (Re(zn ))nE N, (Im(zn ))nE N converge respectively to Re(L) and Im(L). Thus (Re(zn ) · Im(zn ))nE N is also convergent, with limit Re(L) · Im(L). (5.14) then gives Re(L) • Im(L) s 1. So LE F. Hence Fis closed, and so CF= U is open. Solutions 195 Next we show that U is not a domain. Suppose that it is. Then there is a (stepwise) path 1: [a, b] ➔ U that joins 1(a)=2+2i EU to the point 1(b) = -2 - 2i E U. Since the map z f-t Re(z) : <C ➔ JR is continuous, it follows that t 8 Re(,(t)) : [a, b] ➔ IR is continuous too. We have <p(a)= Re(,(a))= Re(2+2i)= 2, and <p(b)= Re(,(b))= Re(-2 - 2i)=-2. Since <p(a) = 2 > 0 > -2 = <p(b), it follows by the Intermediate Value Theorem that there exists a t* E [a, b] such that O = <p(t*) = Re(,(t*)). But then Re(,(t*)) • Im(,(t*)) = 0 • Im(,(t*)) = 0 -:f 1, showing that 1(t*) (/. U, a contradiction. So U is not path-connected, and hence not a domain. Solution to Exercise 1.30 Since D is open, it is clear that its reflection in the real axis, D*, is also open. Let w1, w2 E D*. Then w1, w2 E D. As D is a domain, there exists a stepwise path 1 : [a, b] ➔ <C such that 1(a) = w 1, 1(b) = w2 and for all t E [a, bl, 1(t) E D. Now define [a, b] ➔ <C by ,*(t) = 1(t), t E [a, bj. Then ,*(a) = w1 = w1, 1*(b) = w2 = w2, and for all t E [a, bl, 1*(t)=1(t) E D*. As 1* is the composition of the continuous functions 1 and z f-t z, 1* is continuous. Also, since I is a stepwise path, there exist points to =a < t1 < · · · < tn < tn+l =b such that for each k = 0, 1, · · · , n, the restriction ,l[t k ,t k +il has either con­ stant real part or constant imaginary part. Consequently 1*l[t k ,t k +il (which has the same real part as ,l[t k .f k +il and has imaginary part which is mi­ nus the imaginary part of ,l[t k ,t k +il) also has either a constant real part of a constant imaginary part. So 1* is a stepwise path too. Hence D* is path-connected. As D* is open and path-connected, it is a domain. ,* : Solution to Exercise 1.31 We have 9 exp(i ;) =exp(i(41r+i))=e 0 (cosi+i sini)=l(O+i-l)=i, and exp(3 +1ri)=e3 (cos1r+i sin1r)=e3 (-1+i • 0)=-e3 . 196 A Friendly Approach to Complex Analysis Solution to Exercise 1.32 Let z = x + iy, where x,y E IR. Then ex (cosy + i siny) = 1ri. Taking absolute values on both sides, we obtain ex = 1r, and so x = log1r. Hence cosy + i siny = i, which means that siny = 1 and cosy = 0. Thus y = � +21rk, k E Z. See Figure 5.11. 1 Fig. 5.11 Possible values of y when cosy+ i siny = i are given by y = Hence if expz = 1ri, then z E { log1r +i Ci � + 21l'k, k E Z. +21rk) , k E Z} . Vice versa, if z E log1r +i ( � +21rk) for some k E Z, then expz = e log-,r (cos (i +21rk) +i sin (i +21rk)) =1r(0 +i · 1) = 1ri. Consequently, exp z = 1ri if and only if z E { log1r +i ( i +21rk) , k E Z} . Solution to Exercise 1.33 Let ,(t) :=exp( it), t E [0, 21r]. Then ,(t) =exp( it) = e0 ( cos t +i sint) =cost+ i sin t. The point ( cost, sint) lies on a circle of radius 1 and center (0,0), and with increasing t, this point moves anticlockwise. Hence the curve t r-+ 1(t) is the circle traversed in the anticlockwise direction, as shown in Figure 5.12 . 197 Solutions 1(t) = cost+isint 1(0) = 1( 21r) = 1 ---+------+----!--- Fig. 5.12 The curve t >-+ ')'(t) := exp(it), t E [0,27r]. Solution to Exercise 1.34 We have exp(t+it) = e t(cost+isint). So the image curve is given by ti---+ (et cost,et sint). We have sketched this (not to scale!) in Figure 5.13. Thus the curve is a spiral , and as t \. - oo, et(cost+isint) converges to 0, while the curve spirals outwards as t / +oo. -- t/'+oo t=O Fig. 5.13 The image of the line y = x under the map z = x + iy >-+ expz. Solution to Exercise 1.35 We have 2 exp(z 2 ) =exp((x+iy)2 ) =exp(x2 -y2+2xyi) =ex -Y\cos(2xy)+isin(2xy)), 198 A Friendly Approach to Complex Analysis and so I exp(z 2 )1 = ex -Y , Re( exp(z 2 )) = ex -Y cos(2xy), 2 2 Im( exp(z 2 )) = ex -Y sin(2xy). Also, we have for z-=/- 0 that 2 1 - iy ) exp ! = exp (- -.-) = exp ( : 2 Z X = e '" 2 �Y 2 X +iy +y ) + i sin ( 2 cos ( 2 x �\2 x ( and so it follows from here that lexp 2 2 2 �I= e 7 y2 and )) , :c2�y2 , Re ( exp �) = e :c 2 2 2 2 Im ( exp �) = e :c �Y �Y 7 7 cos ( 2 x sin ( 2 x y2 y2 ), ). Solution to Exercise 1.36 For z1, z2 EC, we have (sin z1)(cos z2 )+ (cos z1)(sin z2 ) exp(iz1) � exp(-iz1) exp(iz2 ) + exp(-iz2 ) ) ) ( =( i 2 exp(iz1) + exp(-iz2 ) exp(iz2 ) � exp(-iz2 ) ) ( ) + ( i 2 2 exp(i(z1 + z2 )) - 2 exp(-i(z1 + z2 )) . ( + ) =�� = �. � Solution to Exercise 1.37 Let z = x + iy, where x,y ER Then cos z = cos(x + iy) = (cos x)(cos( iy)) - (sin x)(sin(iy)) = (cosx) ( e- Y + eY 2 e- Y _ eY ) ) - (sinx) ( 2i sinhy = (cosx)(coshy) - (sinx) (-- -) i = (cosx)(cos hy) - i(sinx)(sinhy). 199 Solutions Thus I coszl 2 = (cosx)2 (coshy)2 +(sinx)2 (sinhy)2 = (1- (sinx)2 )(coshy)2 +(sinx)2 = ( e2Y - 2 + e-2Y · ) 4 e2Y +2+e- 2Y (coshy)2 - (sinx)2 (coshy)2 +(sinx)2 ( - 1) 4 = (coshy)2 - (sinx)2 (coshy)2 +(sinx)2 ((coshy)2 - 1) = (coshy)2 -�+�- (sinx)2 = (coshy)2 - (sinx)2 . Solution to Exercise 1.38 Let z = x+iy, where x, y E JR. Then cos z = 3 gives = 3, (sinx)(sinhy) = 0. (cosx)(cosh y) (5.15) (5.16) We note that sinhy = 0 if and only if y = 0. But y = 0 is impossible, since this would mean z (= x + iy = x) would be real, and there are no real solutions x to cosx = 3! Thus we must have sinhy cf. 0, and so (5.16) implies that sinx = 0. Hence x E {mr: n E Z}. But then cosx = ±1. As eY + e- Y > 0 for all y E IR, 2 we see from (5.15) that cosx can't be -1. So x E {21rn : n E Z} and cos x = l. Then cosh y = 3 yields coshy = eY + e- Y ---=3 ' that is, (eY )2 - 6eY +1 eY = 0. Thus 2 6 "9=1 = ± � = 3 ± y� -l = 3 ± 2V2 ' 2 and soy= log(3+2J2°) or y = log(3- 2V2) = log 9� 3+2 2 = log J2 = - log(3+2V2). 1 3+2 2 200 A Friendly Approach to Complex Analysis Consequently, z E {21rn ± i log(3+2J2), n E Z}. Vice versa, if z =21rn ± i log(3+2J2) for some n E Z, then cosz = (cos(21rn))(cosh(±(3 ± 2V2))) - i(sin(21rn))(sinh• • • )) "-v---' '-..--' =1 =0 =cosh(±(3 ± 2V2)) = log(3+2v'2) e + e- log(3+2v'2) 2 3+2J2+3 - 2J2 3+2J2+(3+2J2)-l =3_ = = 2 2 Consequently, cos z =3 if and only if z E {27rn ± i log(3+2J2), n E Z} Solution to Exercise 1.39 0 Fig. 5.14 The set {z E IC: z =f. 0, ¾ < IArg(z)I < n- Solution to Exercise 1.40 We have Log(l+i) =Log ( V2 ( � +i �)) =Log(V2 (cos�+i sin�)) =Log(V2 exp(i�)) =log V2+ii Solution to Exercise 1.41 We have Log(-1) =Log(l · exp(i1r)) =log l+i1r =0+i1r =i1r, Log(l) =Log(l · exp(iO)) =log 1+iO =0+iO =0. Solutions 201 With z = -1, we have Log(z 2 ) = Log((-1) 2 ) = Log(l) = 0, while 2 • Log(z) = 2 · Log(-1) = 2 • i1r. So we see that when z = -1, Log(z 2 ) = 0 =/- 2 · i1r = 2 · Log(z). Solution to Exercise 1.42 Let A:= {z EC: 1 < z < e}. Then z EA if and only if z = rexp(iArg(z)), where 1 < r < e and Arg(z) E (-1r,1r]. For such a z, Log(z) = Log(rexp(iArg(z))) = log r +iArg(z) and O = log 1 < log r < log e = 1. So the image lies in the rectangle ]I:= {x +iy: 0 < X < 1, -7r < y < 1r}. Log � Vice versa, if x + iy E ll, then z := exp(x + iy) = ex exp(iy) E A since lzl = ex E (1, e), and log(z) = Log(e x exp(iy)) = log ex + iy = x + iy. Hence the image of A under Log is precisely ll. Solution to Exercise 1.43 The principal value of (1 +i) 1 -i is exp((l - i)Log(l +i)). We have Log(l +i) = Log ( v'2 exp (i�)) = log v'2 +ii. Thus the principal value of (1 +i) 1 -i is exp( (1 - i)Log(l +i)) = exp ( (1 - i) (log v'2 +ii)) = e10g v'2+¾ exp (i (i - log h)) In ... (1 +i) exp(-i log vIn2) v'2 = v � e4 = e l (1 +i) ( cos(log v'2) - i sin(log v'2)). 202 A Friendly Approach to Complex Analysis Solutions to the exercises from Chapter 2 Solution to Exercise 2.1 For z -=/- 0, we have f(z) - f( 0) _ = lzl 2 - 0 = ff . 0 z z- 0 z- 0 Thus given€> 0, we setb =€> 0, and then whenever 0 < lz-01 = lzl <b, f(z)-f(O) 01 = lffl = ff = lzl < b = z-o z lz l Hence f is complex differentiable at 0, and f'(0) = 0. l €. Solution to Exercise 2.2 Let w0 E D*. Then wo E D. Since f is holomorphic in D, given € > 0, there exists ab> 0 such that whenever 0 < lz- w0 1 <b, z ED and I f(z) - !J_Wo) J' - (wo)I < z-wo €. Now let w be such that 0 < lw-wol < b. But then 0 < lw-wol = lw-wol = lw-wol <b, 1j(W)-f(Wo) and so w ED. Thus w E D*. Moreover, l f*(w) -f*(wo) -f'(wo)I = w-wo (5.17) w-wo -f'(wo)I = I f(wj_ - !J_Wo) - f'(wo)I w-wo = I f(w_}_ !J_Wo) - J'(wo)I < € (using (5.17)). W-Wo So f* is complex differentiable at w0 and (f*)'(wo) = f'(wo)- As wo ED* was arbitrary, J* is holomorphic in D*. Solution to Exercise 2.3 Since f is complex differentiable at z0, there exists an r > 0 and a function h: D(zo,r) ➔ <C, where D(zo,r) := {z E (C: lz-zol < r} CD such that f(z) = f(zo) + (f'(zo) + h(z))(z - zo) for lz-zol < r, 203 Solutions and lim h(z) =0. We can choose an r' < r so that in D*(zo,r'):= {z E <C: 0 < lz - zol < r'} (c D(zo,r) CD), we have lh(z)I < 1. Now given E > 0, set z➔zo , r' . { lf'(zoE)I + l } Then whenever 0 < lz - zol < 8, we have z ED(zo,r') and so 8 =min IJ(z) - f(zo)I =lf'(zo) + h(z)llz - zol :::; (IJ'(zo)I + lh(z)I) lf'(zoE)I + 1 < (lf'(zo)I + 1) =. lf'(zo)E I + 1 E Hence f is continuous at zo. Solution to Exercise 2.4 Using the fact that f, g : U ➔ <C are complex differentiable functions at z0 EU, it follows from Lemma 2.1 that there exists an r > 0, and functions hf , hg : D(zo,r) ➔ <C, where D(zo,r) := {z E <C: lz - zol < r}, such that for lz - zol < r f(z) =f(zo) + (f'(zo) + h 1 (z))(z - zo), g(z) = g(zo) + (g'(zo) + hg (z))(z - zo), and lim h 1 (z) =0 = lim hg (z). z➔zo (5.18) (5.19) z➔zo (1) Adding (5.18) and (5.19), we obtain for lz - zol < r that (f + g)(z) =(f + g)(zo) + (J'(zo) + g'(zo) + hf + g (z))(z - zo), where hf + g (z):=h 1 (z) + hg (z) in D(zo,r). Moreover, lim h J + g (z) = lim (h1 (z)+hg (z)) = lim h 1 (z)+ lim hg (z) =0+0 =0. So by Lemma 2.1, it follows that f + g is complex differentiable and (f + g)'(zo) =f'(zo) + g'(zo)(2) Multiplying (5.18) by a, we obtain for lz - zol < r that z➔zo z➔zo z➔zo z➔zo (a· f)(z) =(a· f)(zo) + (a· J'(zo) + h0 - J (z))(z - zo), where h0 . 1 (z):=a· h 1 (z) in D(zo,r). Moreover, lim h0 . 1 (z) = lim (a· h 1 (z)) =a· lim h 1 (z) =a· 0 =0. z➔zo z➔zo So by Lemma 2.1, it follows that a · f is complex differentiable and (a· f)'(zo) =a· f'(zo)z➔zo 204 A Friendly Approach to Complex Analysis (3) Multiplying (5.18) and (5.19), we obtain for lz - zol < r that (fg)(z) = (fg)(zo) + (J'(zo)g(zo) + f(zo)g'(zo) + h19 (z)) (z - zo), where h19 (z):= f(zo)hg (z)+g(zo)h1(z)+(z-zo)(J'(zo)+h1(z))(g'(zo)+h9 (z)) in D(z0,r). Moreover, lim h19 (z) z➔zo = J(zo) · 0 + g(zo) · 0 + 0 · (J'(zo) + 0) · (g'(zo) + 0) = 0. So by Lemma 2.1, it follows that fg is complex differentiable and (fg)'(zo) = J'(zo)g(zo) + J(zo)g'(zo)Solution to Exercise 2. 5 Suppose that Hol(l[])) is finite dimensional with dimension d. Then the d+ 1 vectors 1, z, z2, • • • , zd E Hol(l[])) must be linearly dependent. So there exist scalars ao, · · · ,ad, not all zeros, such that ao · 1 + a1 · z +···+ad · zd = 0 (z E ][])). Let k E {0, 1, · · · , d} be the smallest index such that ak -/- 0. Then by differentiating k times and evaluating at 0 E ][]), we get 0 + ak · k! + 0 = 0, and so ak = 0, a contradiction. Solution to Exercise 2. 6 Let z0 E U. Since f is holomorphic at z0, there exists a r > 0 and a complex-valued h defined on D(zo,r) := {z EC: lz - zol < r} CU such that f(z) = J(zo) + (J'(zo) + h(z))(z - zo), z E D(zo,r), and lim h(z) z➔zo = 0. Thus with g := 1/ f, we have -1 g( z) = -(1 g zo) + (J'(zo) + h(z))(z - zo), (5.20) 205 Solutions and so g(zo) = g(z) + (f'(zo) + h(z))g(zo)g(z) · (z - zo)- Rearranging, we obtain g(z) = g(zo) + (-f'(zo)g(zo)g(z) - h(z)g(zo)g(z)) · (z - zo) h(z) f'(zo) f'(zo) f'(zo) ) (z zo) = g(zo) + ( - (f(zo))2 + (f(zo))2 f(zo)f(z) f(zo)f(z) f'(zo) = g(zo) + ( - (f(zo))2 + <p(z)) · (z - z0), where h�) f'(�) f'�o) <p(z) ·- ---2 - --z E D(zo,r) . - -----,--------,-------,--,( f(zo)) f(zo)f(z) f(zo)f(z)' . Using the continuity off at zo and (5.20), we obtain Hence g is differentiable at z0 and f'(zo) ' g (zo) = - (f(zo))2. Solution to Exercise 2. 7 We have already shown this for m 2: 0. Suppose now that m = -n for some n E N. Then consider the map z f---7 zm = z -n = 1 zn = 1 f(z)' where f(z) := zn , z E C \ {O}. Since f is holomorphic and pointwise nonzero in C \ {O}, it follows that 1/ f is holomorphic too, and 1 ' 1 1 m-1 f'(z) nzn -l ( ) (z) = = -n. n +l = m. z-m+l = mz = 2 (f(z)) f z (zn )2 This completes the proof. Solution to Exercise 2.8 Let f : lIJ) -t C be defined by z f(z)=_l+ , 1-z ZE lIJ) ' 206 A Friendly Approach to Complex Analysis and g: C--+ C be define d by g(z) = expz, z EC. go f is holomorp hic in ][}), and more ove r, and so (go f)'(z) = g'(f(z) ) • 1+ z 1+ z • .!}_ ) ) = exp ((1-z dz 1-z f'(z) = exp (-��;)· (-(l+z): = exp (=- 2 z C� )- � d�(l+z)) l z z l+z (l+z) ) (-z · ( 1 -z)2 1 (1-z)2 exp z)) d 1+ Thus - (exp (--1- z dz The n f(][})) c C = D9, (- 1 + z) 1-z 2 =---( 1- z)2 exp 1 1 -z ) z) z 1+ (--1 -z , E][})_ Solution to Exercise 2. 9 Le t z = x + iy, whe re x, y E R We have lzl 2 = x2 + y2 . So if u, v de note the re al p art and the imaginary p art of lzl 2 , the n we have u = x2 + y2 , v = 0. So Since au 2x, ax= au = 2y, ay z f=. 0, it follows that x av =0, ay av O ax= or y is nonze ro. Cauchy-Rie mann equations is violate d, since e ithe r So at le ast one of the So lzl 2 is not diffe re ntiable at any nonze ro comp lex numb e r. Solution to Exercise 2.10 Le t z = x + iy, z3 whe re x,y ER The n 2 3 3 = (x +iy) = x + 3x (iy) + 3x(iy) + (iy) 2 = x3 - 3xy2 + i(3x2 y - y 3 ). 3 207 Solutions So if u, v denote the real and imaginary parts of z3, then we have = x3 - 3xy2 and v(x, y) = 3x2 y - y3. u(x, y) We have u, v are continuously differentiable, and av au 3 2 3 2 = x - y ax ay' av au -=-6xy= --. ax ay and So the Cauchy-Riemann equations are satisfied everywhere in IR 2 • Hence z r-+ z3 is entire. Solution to Exercise 2.11 Let z = x+iy, where x, y Re(z) = Re(x+iy) = x. Re(z), then we have ER We have denote the real and imaginary parts of So if u, v U=X, V So we have = 0. au av =I-:/- 0 = for all (x, y) E ax ay 2 IR . Thus the Cauchy-Riemann equations are satisfied nowhere in IR2 • Hence Re(z) is not complex differentiable at any point in C. Solution to Exercise 2.12 Let u, v be the real and imaginary parts of au ax = av ay = 0 and Hence u(x, Yo) - u(xo, Yo) = au ay 1. x xo f. = - Then _ av ax v= 0. Thus = 0. au (�, Yo)d� aX = 0 whenever the straight line segment joining ( x, Yo) E D to ( xo, in D. Similarly, u(xo, y) - u(xo, Yo) = 1, Y au Yo 8 y (xo, 17)d17 = 0 Yo) E D lies 208 A Friendly Approach to Complex Analysis whenever the straight line segment joining (xo, Yo) E D to (x0, y) E D lies in D. So the value of u along horizontal and vertical line segments lying inside D is constant. As D is path-connected, it now follows that u is constant in D (because any two points in D can be joined by a stepwise path). Hence f = u + iO = u is constant in D. Solution to Exercise 2.13 OU .av +i Let u, v be the real and imaginary parts off. Then f'(z) = ox ox gives =0 8u_ov_ . 0 D ox-ox- m ' and using the Cauchy-Riemann equations, also that Hence u(x, Yo) - u(xo, Yo)= 1. ou !l(t, Yo)dt = 0 XO UX x whenever the straight line segment joining (x, Yo) E D to (xo, Yo) E D lies in D. Similarly, u(xo, y) - u(xo, Yo) = 1 y ou (xo, 'TJ)d'TJ = 0 Yo 8 y whenever the straight line segment joining (xo, Yo) ED to (xo, y) E D lies in D. So the value of u along horizontal and vertical line segments lying inside D is constant. As D is path-connected , it now follows that u is constant in D (because any two points in D can be joined by a stepwise path). Similarly v is also constant in D. Consequently, f = u + iv is constant in D. Solution to Exercise 2.14 We have using the chain rule that av OU (x, y) = hI (v(x, y)) (x, y), ox ox ov OU (x, y) = hI (v(x, y)) (x, y). oy oy 209 Solutions Using these and the Gauch-Riemann equations, we obtain �:(x,y) = h'(v(x,y)) · �:(x,y) = h'(v(x,y)) · �:(x,y) = h'(v(x,y)) • (h'(v(x,y))�:(x,y)) = (h'(v(x,y)))2 • ;:(x,y) = -(h'(v(x,y)))2 • �: (x,y), au (x,y) = 0. As (1 + (h'(v(x,y)))2 ) � 1 ay au (x,y) = 0. ay Using the Cauchy-Riemann equations, we also obtain av au = . (x,y) = - ay(x,y) 0 ax From this, it follows that and so (1 + (h'(v(x,y)))2 ) > 0, �:(x,y) = h'(v(x,y));: (x,y) = h'(v(x,y)) • 0 = 0, and using the Cauchy-Riemann equations again, we have av au (x,y) = 0. (x,y) = ay ax But now it follows that u is constant locally along horizontal and along vertical lines. Since D is a domain, it is path-connected, and so we know that any two points in D can be connected by a stepwise path. This implies that u is constant in D. Similarly v is constant in D. So f = u + iv is constant in D as well. Solution to Exercise 2.15 ("If'' part) Suppose k = 2. Then f(z) = x 2 -y2 + 2xyi = x 2 + (iy) 2 + 2x( iy) = (x + iy) 2 = z 2, which we have seen is entire in Example 2.1. ("Only if"' part) Now suppose that f is entire. Then the Cauchy-Riemann equations must be satisfied everywhere in IR2 • So in particular, av au for all x,y E IR. = 2x = kx = ax ay If we take x = 1, then the above yields that k = 2. A Friendly Approach to Complex Analysis 210 Solution to Exercise 2.16 z 0 0 We have that the length of z-zo is lzol tan(d0) � lzold0, while the length of zn-z0 is \z0 \ n tan(nd0) � \zol n nd0, and so the magnification produced locally by z c-+ zn is given by lzn-zlJ'I lz-zol � \zo\ n nd0 lzo\d0 = n lzol n -1. Also, from the picture, we see that the anticlockwise rotation produced locally is equal to (n-1)0. Hence f'(zo) Consequently, d = n \zol n -I(cos((n-1)0) + isin((n-1)0) ) = n (lzol (cos0 + isin0n n -1 = nz;-1. zn dz = nzn -I, z EC. Solution to Exercise 2.17 i( y + 8)1---------<>z Fig. 5.15 t5 iy1----------<>Zo expz exp � expzo Calculation of the amount of local magnification produced by exp. From Figure 5.15, we see that the magnification produced is ex . t5 -8 = ex . 211 Solutions expz exp Zo Z iy1-------<)--�o-<> y expzo � y From the picture, we see that the anticlockwise rotation produced is y. Thus the complex derivative at z0 must be e "' (cosy+isiny) = exp(x+iy). Consequently, exp' z = exp z. Solution to Exercise 2.18 Let zo E C. Move the point z0 along the line with slope 1 through a distance of o to get the point z. Similarly move z0 horizontally to the left by a distance o to get the point z. Re(•) is complex differentiable at Suppose that the real part mapping z0• In Figure 5.16, by looking at the images of z, z under the mapping Re(·), we see get conflicting values of the local rotation produced as being 45 ° 0 ° , which cannot happen. at z0• As the choice of zo E and map cannot be complex differentiable arbitrary, the map is complex differentiable nowhere. z 0 0 Zo z Re(·) � C was Clockwise 45 ° rotation No rotation � Re(z) Fig. 5.16 So the Non complex differentiability of Re(·). ,) Re(zo) Re(z) 212 A Friendly Approach to Complex Analysis Solution to Exercise 2.19 For f = u + iv with smooth u, v E C2, that is, twice continuously differen­ tiable, we have 213 Solutions Solutions to the exercises from Chapter 3 Solution to Exercise 3.1 We have ,-1 =cos t+ i sin t, ,-2 =cos(2t)+ i sin(2t), and ')'3 =cos t - i sin t, and in each case, (Re(1'k(t)))2 + (In1(1'k))2 = 1, k = 1,2,3, and so the image of ')'k is contained in the circle 11' with center O and radius 1. Also if z =exp(i0) with 0 E [O, 21r), then z = ,-1(0) = ,-2(0/2) = ')'3(21r - 0), and so every point on 11' belongs to the image of each of the curves ,-1, ,-2, 1'3 · We have 1 1 1 1 1 1'i -dz= z 1'2 !dz = f z lo 1' !dz= f z lo 3 O 2 1r --1 .- • i exp(it)dt =21ri, exp(it) 21r . • 2i exp(2it)dt =41ri, exp 2it) ! � 21r . • (-i) exp(-it)dt =-21ri. exp -it) Solution to Exercise 3.2 Let ,-(t) = x(t) + iy(t), t E [O, 1], where x, y are real-valued. Also, let u, v be the real and imaginary parts of f, respectively. Then f' (,'(t)) · ,-' (t) = ( �: (x(t), y(t)) + i :: (x(t), y(t))) (x'(t)+ iy'(t)) au av = a (x(t), y(t)). x'(t) - a (x(t), y(t)) · y'(t) x x +i ( �: (x(t), y(t)) · y'(t)+ �: (x(t), y(t)) · x'(t)) au au = o (x(t), y(t)) · x'(t) + a (x(t), y(t)). y'(t) y x +i ( �� (x(t), y(t)) · y'(t)+ :: (x(t), y(t)) · x'(t)) (using Cauchy-Riemann equations) d d = u(x(t), y(t)) + i v(x(t), y(t)) (using the Chain Rule) dt dt = !(u(x(t), y(t)) + iv(x(t), y(t))) = !f(,-(t)). 214 A Friendly Approach to Complex Analysis Solution to Exercise 3.3 Let, be the c ir cular path ,(t)=2exp(it), t E [0,27r]. I, (1) We have (z+z)dz= 121r 1 (2exp(it)+2exp(-it)) • 2i • exp(it)dt =4i I, (2) We have (z 2 - 2 O 7' 121r 121r 2z+3)dz= = (exp(2it)+l)dt=4i • 0+4i • 271"=87ri. (4exp(2it) - 4exp(it) +3) • 2i • exp(it)dt i(8exp(3it)- 8exp(2it)+6exp(it))dt =0+0+0=0. I, (3) We have xydz= 121' 1 1-'Ir.._,._., 1' 2 cos t-2 sint-2i·(cost+i s int)dt =4i 2 O 7' (sin(2t))(cos t+i s int)dt (s in(2t)) cos t dt- 2 =4i odd function 121' ( 0 cos t- cos(3t))dt =0- 2(0- 0)=0. Solution to Exercise 3.,4. (1) ,(t)= (1+i)t, t E [0,1), and so 1 'Y 1 Re(z)dz= O 1 l+i t(l+i)dt= -. [ 7r/2,0], and so (2) ,(t)=i+exp(it), t E o o Re(z)dz=l (cos t)iexp(it)dt =l i(cost)2 1 'Y / o 2 -7r =l -1r/ 2 (i cos(2t)+1 2 / 2 _ s in(2t) )dt -7r 2 . 1 7r 1 1 .71" =0+i · - · - + - = - +i-. 2 2 2 2 4 2 - (cost)(sint)dt 215 Solutions (3) 1(t) = t + it2, t E [O, 1], and so 1 7 Re(z)dz = 1 o1 1 t · (1 + 2it)dt = 2 1 1 2 + 2i • -3 = -2 + -3 i. Solution to Exercise 3. 5 By the Binomial Theorem, Thus for O :S k :S n, + zr = � (n) l-k-1 ' � f, Z zk+l l=O (1 and so 21ri _1 1 C (1 + zr dz= _1 z k+l 21ri 1t C (n)zl-k-ldz l=O £ Solution to Exercise 3.6 Let u,, VJ, U9, Vy : [a, b] --+ IB. be such that f('Y(t))'Y'(t) g('Y(t))'Y'(t) = U9(t) + iVg(t), t E [a, b]. Then 1 (f + g)(z)dz = = U1 (t) + iV1 (t), b l (f + g)('Y(t)) • "f'(t)dt b = l (!b(t)) • "f'(t) + g('Y(t)) • "f'(t))dt b b = l (U,(t) + Ug(t))dt + i l (V,(t) + Vg(t))dt b b b b = l u1 (t)dt + l U9(t)dt + i l v1 (t)dt + l Vg(t)dt = 1 f(z)dz + 1 g(z)dz. 216 A Friendly Approach to Complex Analysis (2) Let a = p + iq where p, q E IB., and let U, V : [a, b] ➔ IB. be such that f('y(t)) · 'Y'(t) = U(t) + iV(t), t E [a, b]. Then 1 b (a· f)(z)dz = 1 (p + iq)(U(t) + iV(t))dt b b = 1 (pU(t) - qV(t))dt + i 1 (pV(t) + qU(t))dt b b b b = p(1 U(t)dt + i1 V(t)dt) + iq(1 U(t)dt + i1 V(t)dt) b b = (p+iq)(1 U(t)dt+i 1 V(t)dt) =a· 1 f(z)dz. Solution to Exercise 3. 7 -(-'Y): [a, b] ➔ <C is given by (-(-'Y))(t) = (-'Y)(a + b - t) = 'Y(a + b - (a+ b - t)) = 'Y(t), t E [a, bl, and so -(-'Y) = 'Y· (This is obvious visually.) Solution to Exercise 3.8 We do have that "f(b) = (-'Y)(a), and so the two paths 'Y and -'Y can be concatenated, and we have 1 -r+(--y) f(z)dz = 1 ')' f(z)dz + f f(z)dz = J_')' 1')' f(z)dz -1')' f(z)dz = 0. Solution to Exercise 3.9 Let 'Y: [O, 1] ➔ <C be defined by 'Y(t) = (1 + i)t, t E [O, 1]. By Pythagoras's Theorem, the length of 'Y is Jl 2 + 1 2 = v'2. l+i 1 0 1 Also, l('y(t))2 1 We have 1 and so 7 217 Solutions = It+itl2 = 2t2 , and so max l('y(t))2 1 = 2 · 1 2 = 2. Thus z2dzl 11"I tE[0,1] :S ( max j('y(t)) 1) · (length of 'Y) = 2V2. 2 tE[0,1] 1 1 ( +i)3 , z2 dz = ( (t+it)2 · (1+i)dt = ( (1 +i)3 t2 dt = l lo lo 2v12 )3 J2z2 dz = = - -. 3 3 117 I ( 3 Solution to Exercise 3.10 We have (2:) I (2:)1 I la = = 2�i ( n l ;:t dzl + l n ( +_ )_ _n I) . <- 2_ 271" . 1 = max l_I _ _z__ (max - l_ z 1 271" izl=l 1 z izl=l :S (1+1) n = 2 n = 4 _ 2 2 n+l 2 2 n Solution to Exercise 3.11 Suppose that F = U+iV is a primitive of z in C. Then au . av -+iax ax . . lT])2 ln JN.. • = - - i- = F' = Z = X - iy av . au ay ay _ Fix x0 ER Then for (x,y) E JR2, we have V(x,y) - V(xo,y) = 1x 8V (�,y)d� = 1x -yd�= -xy + xoy. a xo X xo So V(x,y) = -xy+cp(y), where cp(y) := V(xo,y) +xoy. Hence X = aV ay = -X+'{)'(y), that is, cp'(y) = 2x for all x E JR, which is clearly impossible, since in particular, we would obtain 2 · 1 = 2 = cp'(y) = 2 • 0 = O! 218 A Friendly Approach to Complex Analysis Solution to Exercise 3.12 The map ( i---+ f(()g'(() + f'(()g(() has a primitive since (Jg)'= f g' + f'g. So by the Fundamental Theorem of Contour Integration, 1 (!(()g'(() + f'(()g(()) d( = f(z)g(z) - f(w)g(w), and a rearrangement proves the claim. Solution to Exercise 3.13 We have sin ' z = cos z, and so cos z has a primitive in IC. Thus by the Fundamental Theorem of Contour Integration · i") e -1 -e1 exp (.i·i") - exp (-i· .. = COS Z dZ ... = Sln i- Sln (-i.) = 2Sln i= 2 . . 2i i 1'Y =(e-¾)i. Solution to Exercise 3.14 Since exp' z = exp z in C, we have 1 1 exp zdz = exp(a+ib)-exp O = ea (cos b+i sin b)-1 = ea cos b-1+iea sin b, for a path 'Y joining O to a + ib.If 'Y(x) =(a + ib)x, x E [O, 1], then exp zdz = fo\xp(a+ib)·(a+ib)dx = Hence (a -ib) (a2 + b2) 1 exp zdz= fo 1 e x(cos(bx)+i sin(bx))(a+ib)dx. a fo 1 e x(cos(bx) + i sin(bx))(a + b )dx.Thus a fo 1 e x cos(bx)dx = Re((a -ib) 1 exp zdz) 2 2 a = Re((a -ib)(ea cos b - 1 + iea sin b)) Hence 11 =a(ea cos b - 1) + bea sin b. a a _ a(e cos b - 1) + be sin b · e a x cos(bx)dx 2 2 a +b o 219 Solutions Solution to Exercise 3.15 Consider the closed circular path C with center at O and radius r > 0 traversed in the anticlockwise direction: C(0) = rexp(i0), 0 E [0,21r]. By the Fundamental Theorem of Contour Integration, we have 0 = = la exp zdz = 121r 121r e r cos 0+ir sin 0 . ri exp(i0)d0 e r cosO • r • i • exp(i(rsin0 + 0))d0. So we obtain in particular (by equating real parts) that {21' e r cosO cos(rsin0 + 0)d0 = 0. Jo Solution to Exercise 3.16 Suppose Fis holomorphic in C\ {0} and that F' = 1/z in C\ {0}. Consider a circular path C with a positive radius centered at O traversed in the anticlockwise direction. Then by the Fundamental Theorem of Contour Integration, since C is closed, we have la 1 F'(z)dz = 0. On the other hand, we know that C F'(z)dz = 1 �dz= 21ri, cZ a contradiction. Solution to Exercise 3.17 (ERl) Let 'Y : [0, 1] -+ D be a path which is closed. Define the map H: [0, 1] x [0, 1]-+ D by H(t,s) = 'Y(t), t, s E [0, l]. Then His continuous, and H(t, 0) = 'Y(t), for all t E [0, 1), H(t, 1) = 'Y(t), for all t E [0, 1), H(0,s) = "f(0) = "f(l) = H(l,s), for alls E [0, l]. Hence 'Y is D-homotopic to itself. So the relation is reflexive. 220 A Friendly Approach to Complex Analysis (ER2 ) Let 'Yo,'Yi: 0 [ ,1] ➔ D be closed paths such that 'Yo is D-homotopic to'Yl· Then there exists a continuous H: 0 [ ,1] x 0 [ ,1] ➔ D such that H(t,0)='Yo(t), for all t E 0 [ ,1], H(t,1) ='Y1(t), for all t E 0 [ ,1], H(0,s) = H(l,s), for alls E 0 [ ,l]. Let H: 0 [ ,1] x 0 [ ,1] ➔ D be defined by H(t,s) = H(t,1 -s) for all t, s E 0 [ ,l]. Then ii is continuous and [ ,1], H(t,0) = H(t,1) = '}'1 (t), for all t E 0 H(t,1) = H(t,0) ='Yo(t), for all t E [0,1], H(O,s) = H(0,1- s) = H(l,1- s) = H(l,s), for alls E 0 [ ,1]. Thus'Yl is D-homotopic to'YO· Hence the relation is symmetric. (ER3) Let 'Yo,'Y1, '}'2 be closed paths such that 'Yo is D-homotopic to '}'1 and 'Yi is D-homotopic to '}'2. So there exist two continuous functions H,K:[0,1] x 0 [ ,1] ➔ D such that H(t,0) = 'Yo(t), for all t E 0 [ ,1], K(t,0)='Y1(t), for all t E [0,1], K(t,1) ='Y2(t), for all t E 0 [ ,1], H(t,1) = 'Yi (t), for all t E 0 [ ,1], H(0,s) = H(l,s), for alls E [0,1], K(0,s) = K(l,s), for alls E [0,1]. Let L: [0,1] x 0 [ ,1] ➔ D be defined by sE O H(t,2s) [ , ½L L(t,s) = { K(t,2(s- 21 )) s E (21 ,1] Then L(t,0) = H(t,0) = 'Yo(t), for all t E 0 [ ,1], [ ,1]. L(t,1) =K(t,1) = ','2 (t), for all t E 0 Also, for 0 ::; s ::; ½, L(0,s) = H(0,2s) = H(l,2s) = L(l,s), and for ½ < s ::; 1, L(O,s)=K(0,2(s-�)) =K(l,2(s-�)) =L(l,s). Moreover, if ((tn ,sn ))nE N is a sequence in 0 [ ,1] x (to,½), then (½,1] that converges to lim L(tn , S n ) = lim K(tn ,2(sn - � )) = K(to,0) = '}'1 (to) n-too 2 n-too = H(to,1) = L (to, �) =L ( }�� (tn , S n )) . Hence it follows that L is continuous. Consequently 'Yo is D-homotopic to ,.,,2. So the relation is transitive. As the relation of D-homotopy is reflexive, symmetric and transitive, it is an equivalence relation. 221 Solutions Solution to Exercise 3.18 From the picture , we see that C is C \ {O}-homotopic to S. -1 +i "/2 s "/3 "/1 -1-i Let for t "/4 1-i 'Y1 ( t ) := (1- t )(l-i)+ t (l+i) = 1+i(2 t - 1), 'Y2( t ) := (1- t )(l+i)+ t (-1+i) = (1- 2 t )+i, "f3( t ) := (1- t )(-1+i)+ t (-1-i) = -1 +i(l- 2 t ), "f4( t ) := (1- t )(-1-i)+ t (l-i) = 2 t - 1-i, E [O, l]. Then We have 1 1 1 1 f !dz = !dz. !dz+ !dz+ !dz+ �z �z nz �z kz 1 2i _ 1 2i(l-i(2 t - 1)) / dt / dt }0 l+i(2 t - 1) - }0 1+ (2 t - 1)2 1 1 2 t- 1 1 / 2 dt + 2i / dt 2 }0 l+ (2 t - 1) }0 l+ (2 t - 1)2 1 U d 1 d (u=2t-1) i· jl -t+ 1 --2 t = 2 -1 1 +u -1 1 +u i(tan-11- tan- 1 (-1))+ 0 = i(�- ( - �)) =ii. Similarly, using the facts that 1 1 / 1r 2 2 t- 1 / dt = and dt = O, 2 2 }0 l+ (2 t - 1)2 lo l+ (2 t - 1) 222 A Friendly Approach to Complex Analysis we obtain 1 1 1 -2 dt= f -2·(-(2t- 1) - i)dt ! dz = ( 1+(2t- 1)2 lo 1 - 2t+i lo z 72 1 73 . . =0+(-1)(-i) 2 =i2, 7f 7f 1 1 -2i -2i-(-l+i(2t- l))dt ! dz = ( dt= f lo 1+(2t- 1)2 lo -1+i(l - 2t) z = -i•(-1)-�+0=i� and 2' 2 l 1 2 2·((2t- l)+i)dt ! dz = f dt= f lo 2t- 1 - i lo 1+(2t- 1)2 74Z 1 . . = 0 +i=i-. 2 2 7f Thus Is� dz 7f =4 · (ii) =21ri, as expected. Solution to Exercise 3.19 For a circular path C with center O and a positive radius, traversed in the anticlockwise direction, we have 1 - dz =21ri. cZ1 But the elliptic path Eis IC\ {O}-homotopic to C, as shown in Figure 5.17. C E Fig. 5.17 E, C are lC \ {O}-homotopic. 223 Solutions So by the Cauchy Integral Theorem, / 2" f ! dz= f ! dz= 21ri, and so lE z le z I 21ri= f !dz= . . -(-asin0+ib cos0)d0 +i 0 b sm 0 z cos a lE lo 2 / " (-a sin0 +ib cos0)(a cos0 - ibsin0) B d = a 2 (cos0)2 +b2 (sin0)2 lo / 2" (b2 - a 2 )(cos0)(sin0) +iab((cos0)2 +(sin0)2 ) B d = a 2 (cos0)2 +b2 (sin0)2 lo / 2" (b2 - a 2 )(cos0)(sin0) +iab · 1 B _ d · a 2 (cos0)2 +b2 (sin0)2 lo /2,r 21r I d0 Equating the imaginary parts, 2 (cos0)2 +b2 (sin0)2 o ab· a l - Solution to Exercise 3.20 See Figure 5.18. Fig. 5.18 (1) z H Log(z - 4i) is holomorphic in IC\ {r + 4i : r :::; O}. So by the Cauchy Integral Theorem i Log(z - 4i)dz= 0. (2) If C denotes the circle with center 1 and any positive radius, say, r, then we know that ( _]_ dz= 21ri. I le z- A Friendly Approach to Complex Analysis 224 Since 1 /(•-1) is holomorphic in re\ { 1}, and since the two circular paths C and C are re \ {1 }-homotopic, it follows from the Cauchy Integral Theorem that 1 1 f --dz = f --dz = 27ri. 1 l z z le e -1 (3) We have ( i z - 3 = exp((z -3)Logi)= exp ( (z - 3) log 1 +ii)) = exp ( (z -3)( 0 +ii)) = exp (ii· (z -3)), i and so z t-+ i z - 3 is entire. By the Cauchy Integral Theorem, i z - 3 dz = 0. Solution to Exercise 3.21 (1) We have <p (t)= exp I s) ( rt ,'(s) ds) . d ( lot ,'( ) 'Y ds = <p(t) lo ,(s) 1 and so cp'1 - <p')' = 0. Hence dt �i�j �gj. ,2 ( s) !__('!'_' ) = cp',- cp,' = _Q_ = 0 dt and so = ,2 ,'(t) , . ,(t) , But ')'(0)= 1(1) (since I is closed). Hence cp(l) = cp(0)= exp ( lo Consequently, w(1) E Z. (2) We have o �(�j ds) = exp(0)= 1. 1 r 1 r�(t) dt w(I'i) = 27ri lo r 1 (t) _1_ [ 1 27ri exp(27rit) dt = 27ri lo exp(27rit) 1 = - · 27ri = 1. 27ri 225 Solutions 1 (3) We have (11· ,2)'(t) =,Ht)· ,2(t) + ,1(t),�(t), t E [0, 1], and so 1 1 (,1 . ,2)'(t) d w(,1· ,2) = - . t 2 7r2 o (11· 12)(t) 1 1 (t) + ,1(t),�(t) d , ) t = _ _ r W · ,2 . 21ri lo ,1(t) ,2(t) 1 / yi{th�(t) 1 / 'Y� (t)·yA() dt dt + _ =_ 2ni}0 �-,2(t) 2ni}0 ,1(t)·yA(J 1 1 = w(,1) + w(,2). (4) We have rm =r1····· r1 (m times), and so w(rm) = w(r1) + · · · + w(r1) (m times) = m· w(r1) =m· 1 =m. (5) Consider the map <p: [0, 1] --+JR.given by <p(t) = l,o(t)/, t E [0, 1], which measures the distance of 1(t) from 0. Being a continuous function, it has a minimum value d0, and do > 0 since 10 does not pass through 0. Take i5 =do/2 > 0. Let I be a smooth closed path such that max l,(t) - ,ol < 8. II, - ,olloo :=tE[0,1] We will show that I is (C \ {0}-homotopic to ,o. Define the function H: [0, 1] x [0, 1]--+ C\ {0} by H(t, s) =(1-s),o(t) +s,(t), t, s E [0, l]. Then H is continuous, H(t, 0) =,o(t) for all t E [0, 1], H(t, 1) =1(t) for all t E [0, 1], and H(O, s) =(1 - s),o(0) +s1(0) = (1 -s),0 (1) +s1(1) = H(l, s) for alls E [0, l]. Also we note that H(t, s) is never 0, because it is a convex combination of ,o(t) and 1(t), and if (1 - s),o(t) + s1(t) =0 for some t, s, then we arrive at a contradiction. See Figure 5.19. 1(t) -----<>--------<>: ' o--: ' Fig. 5.19 0 < do/2 ' ,o(t) ' That H(t, s) is never 0. 226 A Friendly Approach to Complex Analysis Indeed, 1· do 2 = > sl,,o(t) - 'Y(t)! = l'Yo(t)I bo(t) - ((1 - s)'Yo(t) + S')'(t))I = bo(t) - OI � do. Thus by the Cauchy Integral Theorem, w('Y) l = -. 21ri J J l 1 l -dz= -. -dz= w('Yo)21ri 7 z 7z Solution to Exercise 3.22 (1) 1r-nt t 0 Q nr The length of the arc on the big circle subtended by LQ'0Q is t · nr. As the small coin rolls without slipping, the angle made by 0' P with 00' is (t • nr)/r = n • t. From the picture, we see that 0' (nr + r) exp(it), and P = (nr = + r) exp(it) + exp(-i(1r - nt)) · rexp(it) -------- '--v--' produces a O'P' clockwise rotation = (n + l)rexp(it) + (-1) • r • exp((n + l)it). 227 Solutions (2) The area enclosed by the epicycloid 'Y is 1 { 27r 1 dz= }o r((n+l) exp(it) - exp((n+l)it)). 1 2i 2i / = ; fo i r((n+l)i exp(it) - (n+l)i exp((n+l)it))dt 2 1r r2 ((n+1) exp(-it) - exp(-(n+l)it)) • (n+l)i ( exp(it) - exp((n+l)it))dt 2 1r (n+l)r = --1 ((n+1) - (n+1) exp(int) - exp(-int)+ l)dt 2 2 0 2 = (n+l)r ((n+l) · 27r+0+0+27r) 2 = (n+l)r2 ?T(n+2) = ?Tr2 (n+l)(n+2). Solution to Exercise 3.23 Take for example f given by f(z) = 1/z for z E D := (C \ {O}. Then we have seen that f does not have a primitive in D. (See Example 3.7 and Exercise 3.16.) Solution to Exercise 3.24 Let 'Y(t) 1 7 = exp(it), t E . i (z - a)(az - 1) [O, 1] and so dz = 1 = 27r 0 2 1r 0 2,r 0 2 ,r 1 i (exp(it) - a)(a exp(it) -1) . . ) dt iexp (it - exp(it) ----,----,-----.....,..,.......,......,..dt (exp(it) - a)(a - exp(-it)) exp(it) 1 -...,..dt = 1 -----,--,------,--,------,-----,- =1 0 2 ,r (exp(it) - a)(exp(-it) - a) 1 --�dt I exp(it) - al2 1 = 1 ------2 2 dt 0 2 ,r ((cos t) - a) +(sin t) 1 = 1 ----dt • 1 - 2a cost + a2 0 228 A Friendly Approach to Complex Analysis Since the mapping z f-+ i/ (az -l) is holomorphic in a disc containing the unit circle 'Y (because 0 < a < l), by the Cauchy Integral Formula, 1 _l_ So 1 21ri 'Y 2 0 7r azi_ l z -a 1 ------2 dt= + 1- 2a cost a dz = 1 -I _i_· a z -l z=a = _i_· -. a2 - 1 i. 2 7f . az-1 d z=2 1r i---=--. -2 a -l 1- a2 'Y z i a Solution to Exercise 3.25 -1 -1 -1 ( 2) (3) ( 4) ( 5) (2) j 1 1 1 1 'Y expzd -l Z z2 + : l dz z 'Y 1 z= 21riexpzl + 1d = z 'Y z - 1 2 z -l 'Y 3 (5) (1) (1) 1 1 (3) (4) = 21riexp l= 21r ie. 2 z::l d z = 21riz + z = 0. z=l 1 -l z + 12 + 1 11 = 21ri. = 21r i l z=l 1+ 1 2 (-1) 2 + 1 z2 + 1d � d z=2 1ri z + 11 =-21r i. =21ri z=j 2 'Yz - 1 'Yz-(-1) z-l z=-1 -1- 1 2 z2 + 1 d z = z + 1(-1- - _1_) d z 2 z -l z + l 'Y 'Yz2 - 1 1 1 1 z +1 z +1 z2 + 1 z2 + 1 2 d 2 d z = 21ri-1 - z-21ri-- 1 'Y z-l 'Yz-(-1) 2 z=l 2 z=-1 = 21ri(l) - 21r i(l) =0. = 2 2 Solutions 229 Solution to Exercise 3.26 Suppose F is a primitive. Consider the closed path I given by lz - =I0 ½ traversed anticlockwise. By the Fundamental Theorem of Contour Integra­ tion, 1 1 z(z}_ l) dz= 1 F'(z)dz = 0, . I since I is closed. On the other hand, by the Cauchy Integral Formula, 1 1 d z 2 Z ( Z - l) 1 1 . z -1 d -z = 211'i� 1 = 211'i2 -0 l 0 Z - 1 z=O So we arrive at a contradiction. Hence 1 z(z2 - 1) does not p ossess a primitive in {z EC: <0 lzl < 1}. 'Y = 2 'Y Z - = -211'i. Solution to Exercise 3.27 iR T s i -R -i R Fig. 5.20 The path u = S + T. (1) By the Cauchy Integral Formula, we have 1 1 1 exp(iz) ex iz exp (.i�) z+i ; dz= . dz= 211'i z + z+i z-i (. 1) 1 e ex i) (i p _ . • _ . -� 2 2 - 11'i i +i - 11'i 2i - e · (2) Let z = z + iy, with x, y real and y 2: Then 0. O' F(z)dz = O' O' I . z=i I exp(iz)I = I exp(i(x + iy))I = I exp(-y + ix)I = e-Y � 1 (since y 2: )0 Thus . IF(z)I = 1_ I exp(iz)I < _ 2 2 lz +ll 1z +11· 230 A Friendly Approach to Complex Analysis But lz2 I- I - ll::; lz2 - (-1)1 = < IF(z)I - 2 lz 1 1 lz 2 + 11, and so if lzl < + 11 - 1 lzl - 1 2 < 1 l ---3_ \!'2. F(z)dzl::; 21rR • 1_!1 � IF(z)I ::; 21rR• �2 E \!'2, - lzl 2 since lzl 2::; 2lzl 2 - 2 for lz2 :2: 2, that is, lzl 2: (3) We have 2: (for R 2: v'2) R _ 47f --=:_:f O . - R / F(z)dz So lim R➔ oolT = 0. Since f F(z)dz = f F(z)dz - f F(z)dz = '!!_ - f F(z)dz, ls la it now follows that lim (4) Let S(x) 1 = x, x E e lr f F(z)dz = '!!__ e R➔ ools [-R,R]. Then lim lr f F(z)dz = '!!__O = !!__ R ➔ oolr e e R · R sinx exp(ix) l R cosx • dx= J -- dx+i J -dx F(z)dz= J -- R x2 + 1 - R x2 + 1 - R x2 + 1 s R =J cosx --x d +O -R x2 +1 where we have used the fact that last equality. Hence R lim J R➔ oo cosx - R x2 +1 sin x x2 +l is an odd function to get the r F(z)dz = !!__ dx = lim R➔ ools e Solution to Exercise 3.28 Consider the entire function exp z and let C : [O, 21r] -+ (C be the circular path with center O and radius 1 given by C(0) = exp(i0), 0 E [O, 21r]. By the Cauchy Integral Formula, 1 I exp z l -- dz = expz -. 21ri CZ - 0 z=O = expO = 1, 231 Solutions 1 But 211" 211" exp z exp(exp(i0)) . . · )d0 =i· 1 exp(expi dz=1 iexpi (0 ( ·0))d0 . ( 0) expi cz- 0 0 0 2 11" =i Jo{ exp(cos0+i sin0)d0 =i Jo{ Hence {o J 211" =- { }0 211" e cos 0 (cos(sin0)+i sin(sin0))d0 2 11" ec089 sin(sin0)d0+i { }0 211" ecos0 cos(sin0)d0 ecos e cos(sin0)d0 = 21r. Solution to Exercise 3.29 If f is holomorphic, then f(n) is holomorphic too, and so is its derivative j(n+l). But j(n+l), being complex differentiable, is in particular continu­ ous. So f(n) has a continuous complex derivative. Solution to Exercise 3. 30 Since for all z EC, lf(z)I � 8 > 0, in particular f(z) =/- 0 for all z EC, and so 1/ f is entire. But for all z EC I I :::; �' ftz) and so by Liouville's Theorem, 1/ f is constant. Thus f is constant as well. Solution to Exercise 3.31 Let g be defined by g(z) := f(z) - wo for z E C. Then g is entire and lg(z)I � r for all z E C, and so g is bounded away from 0. Hence by Exercise 3.30, g is constant, and so f = g + w0 must be constant too. Solution to Exercise 3.32 Consider the compact set K := {(x, y) : 0 :::; x:::; T1 , 0 :::; y :::; T2}. The continuous function (x, y) 1----r lf(x +iy)I assumes a maximum value Mon 232 A Friendly Approach to Complex Analysis K. But for mEZ nEZ there are integers n, m such that x + iy =xo + nT1 + i (yo + mT2) for some xo E [O, T1) and Yo E [O, T2). Owing to the periodicity off, f(x + iy)=f(xo + nT1 + i (yo + mT2))=f(xo + iyo) Ef(K), and so for all x, y E JR, lf(xo + iyo)I SM. Hence f is bounded on C, and by Liouville's Theorem, must be constant. Solution to Exercise 3.33 (1) Let g be defined by g(z) = exp(-z) • f(z) for z EC. Then g is entire. Moreover, since lf(z)I SI expzl , it follows by rearranging that lg(z)j = I exp(-z) · J(z)I S 1 for all z EC. So by Liouville's Theorem, g is constant with value say, c. Since jg(z)j S 1, we obtain jcj S1. Consequently, g(z) = exp(-z) · f(z) =c, and so f(z) = c · expz for all z E C, where c is a constant such that leis 1. (2) We know that if p is a polynomial of degree d � 1, then there exist M, R > 0 such that jp(z)I � Mjzj d for jzj > R. Thus with z = x < -R < 0, we have jzj > Rand so Mlxl d Slp(z)I Sle x l =ex S 1 ( because x < 0). Hence lxl d S 1/M for all x < -R, a contradiction. Sop is a constant, say equal to co. But then lp(z)I S I expzl again gives with z = x < 0 that leol S lex l = e x for all x < 0, and so leol = 0. Consequently, p=co= 0. Solution to Exercise 3.3,4. (1) We have for z E C that lz - ail � lzl - la1I= R - la1I, lz - a2I � lzl - la2I= R - la2I- 233 Solutions Thus 11 lf(z)I f(z) < max dzl • (length of C) zEC lz 1 z a2) - a2I a1llz c (z a )( M < · 21rR, - -,------,----,,.....,---,--...,,.. (R-la1l)(R - la21) where M := maxlf(z)lzEC (2) Since a1 =/. a2, we have and so 1 z - a1 1 z - a2 (3) We have 1 z - a2 - (z - ai) (z - a1)(z - a2) -1 1 f(z) _ f(z) dz f(z) dz ( c (z - a1)(z - a2) - c a1 - a2 z - a1 z - a2 ) 1 f(z) dz) f(z) dz a1 - a2 c z - a2 c z - a1 If we consider a small disc .6.1 with center a1 and radius r1 > 0 with boundary C1, then we see that C and C1 are C\ {a1}-homotopic, and the function (z) , z EC\ {ai} g(z) := f z - a1 is holomorphic. Thus by the Cauchy Integral Theorem, f(z) dz. f(z) dz= c z - a1 c1 z - a1 f(z) dz= (a1). Thus But by the Cauchy Integral Formula, 1. f 21ri c1 Z - a1 f(z) dz= 21rif(a1). c z-ai Similarly, f(z) dz= 21rif(a2)c z-a2 21ri(f(a1) - f(a2)) f(z) -----. dz = ------Consequent1y, a1 - a2 c (z - a1)(z - a2) 1 1 1 1 -1 (1 1 1 234 A Friendly Approach to Complex Analysis (4) Suppose that f is a bounded entire function with a bound Mon lfl, that is, lf(z)I ::::; M for all z E C. Suppose that a1, a2 are any two distinct points in C. Let C be a circular path with radius R > 0 and center 0, traversed once in the counterclockwise direction that contains a1, a2 in its interior. Then using the results in parts (i) and (iii), I la1- a2I · 21ri(f(a1)- f(a2)) lf(ai) _ f(a2)I = 21r a1- a2 f(z) = la1- a2I dz 21r · c (z- a1)(z- a2) I 21rRM < la1- a2I . 21r (R- la1l)(R - la2I) As R can be made as large as we please and since 21rRM --------+ 0 (R- la1l)(R - la2I) 11 I as R --+ oo, it follows that lf(a1)- f(a2)I = 0. This implies that J(a1) = J(a2). Hence f is constant. 235 Solutions Solutions to the exercises from Chapter 4 Solution to Exercise 4.1 = = = L L lm(an )Re(an ) and Since Lan converges, so do the real series n= l n= l n= l Hence lim Re(an ) = 0 and lim lm(an ) = 0. Thus lim an = 0 too. n➔oo n➔oo n➔oo Solution to Exercise 4.2 As = L la l converges, and for all n EN, Re(a ) :S la l, lm(a ) :S la l, n= l n n = = n= l n= l n n n = converge by the Comparison Test. Hence Lan converges. n= l Solution to Exercise 4.3 With S n := l+z+• · ·+zn - 1 +zn , we have that ZS n = z+z2 +• • •+zn +zn+l, and so (1-z)s n = 1-zn+l_ Since lzl < 1, z-/:- 1 and so 1-z-/:- 0. So 1-zn+l S n = 1 + Z + · · · + Zn -l + Zn = --1-z (5.21) Thus = lim S n➔= n = lim n➔ = 1-zn+l 1-z = 1-0 1-z 1 1- z' 1 zn = lim S n = --. n➔ L.., L.., 1 Z n=O n=O (Note that we have used the fact that since lzl < 1, one has and so '°' zn converges, with '°' = = lim zn+l = 0, n➔ and this can be justified as follows: lzn+l -0I r:= lzl < 1.) lzJ n+l n ..::..f 0, since 236 A Friendly Approach to Complex Analysis Solution to Exercise 4.4 For n E N, let S n := 1 + 2z + 3z2 + • • · + (n - l)zn -2 + nzn -l_ Then zsn = z + 2z2 + · · · + (n - l)zn -l + nzn. Hence l-Z n (1 - z)Sn = l + z + z2 + · · · + zn-1 - nzn = -- nzn. l-z Thus l-zn nzn ---. Sn = 2 1-z (1-z) (We could have also obtained this expression from (5.21) by differentiating with respect to z.) If we set r := lzl, then O::; r < l and so 1 r=-l+h 1 where h := - - 1 > 0. We have r n. ( -l) . 2 . n h (1 + h) = l + (;) h + (;) h2 + ... + (:) hn � (;) h2 = � Hence n 2 n (n-1) h 2 (n-1) h =----2 ' 0<nrn=---<n·----n • 2 • • (l + h) and so by the Sandwich Theorem, lim nrn n➔oo lim S n n➔oo = 0. Consequently, n n _ 1 = lim ( 1 - Z 2 - nz ) = 1- 0 2 -_0 = 1-z (1-z)2 " n➔oo 1-z (1-z) (1-z) Solution to Exercise 4.5 I I I We have :s I I I = exp(s · �og(n)) exp(s�log n) 1 1 1 1 - eRe(s·logn) - e(logn)·(Re(s)) - (elogn)Re(s) - nRe(s)· = Recall that � _!._ converges if p > l. Hence if Re(s) > 1, then L., nP n= l 1 Re(s) n n=l converges. Thus 1 L 00 00 Lns n=l converges absolutely for Re(s) > 1, and in particular , it converges for Re(s) > 1. 237 Solutions Solution to Exercise 4. 6 Let L =/=- 0. We have that for all z such that lzl < 1/L that there exists a lzl :S: q < l for q < l and an N large enough such that ,Vlcn zn l = n all n > N. This is because � lzl � Llzl < l. (For example take q = (Llzl + 1)/2 < 1.) So by the Root Test, the power series converges absolutely for such z. If L = 0, then for any z E C, we can guarantee that there exists a q < l such that v'lcn zn l = lzl :S: q < l for all n > N. This is n because � lzl � 0lzl = 0 < 1. (So we may arrange for example that q = 1/2 < 1.) So again by the Root Test, the power series converges absolutely for such z. On the other hand, if L =/=- 0 and lzl > 1/L, then there exists an N large enough such that v'lcn zn l = lzl > 1 for all n > N. This is n because � lzl � Llzl > l. So again by the Root Test, the power series diverges. vfcJ vfcJ vfcJ Solution to Exercise 4. 7 When z = 0, the series clearly converges with sum 0. Suppose z =/=- 0. Then lzl =/=- 0. Choose N E N such that N > 1/lzl. Then for n > N, lnzl > Nlzl > 1, and so lnn zn - 0I = lnzln > ln = 1, showing that Ln z 00 Thus if z =/=- 0, then n n diverges. n=l Solution to Exercise 4. 8 V� We have lim nf"l = lim � = 0. So the radius of convergence of n-+oo n-+oo n oo n I::n n=l is infinite, and the power series converges for all z E C. 238 A Friendly Approach to Complex Analysis Solution to Exercise ,4.9 (1) We have (-1r+1 n . l lim n + = 1lffi --=1 ' n--+oo n--+oo n + 1 (-l)n n (-1r n and so the radius of convergence of - - z is 1. n n=l (2) We have .!. 2012 _ (n +1)2012 . . hm l hm (1 + ) - 1, I 2012 n--+oo n--+oo n n L 00 and so the radius of convergence of (3) We have lim n--+oo 1 (n + 1)! 1 n! Ln2012 00 z n n=O is 1. 1 = lim _ _ = 0 ' n--+oo n and so the radius of convergence of f� n. n=O z +1 n is infinite. Solution to Exercise 4.10 For l z l < 1, we know that and so f (z ):=1 +2z +3z2 +4z3 z f (z )=g(z ):= z +· · ·= (l 1 _ +2z2 +3z3 +4z4 +•··= z )2 , z 2 (1- z ) for l z l < 1. So g(z ):= z +2z2 +3z3 +4z4 +··· converges for l z l < 1, g is holomorphic in the disc l z l < 1, with g' (z ) = 1 +22z +32z 2 +42z3 + ... for l z l < 1. On the other hand, g(z)= zf (z )= for l z l < 1, (1- z )2 z and so _ z 1 , _ d 2 _ 1- z +2z _ 1 + z g (z )- ( )-1 +z 2 2 3 dz (1- z ) •(l- z ) (1- z )3 -(1- z )3" (l- z ) 0 239 Solutions Solution to Exercise 4.11 L :2 converges}= {z EC: lzl � 1}. n oo (1) False. For example, { z EC: n=l (2) True. (- 1 r L --z n 00 (3) False. For example z = -1. n n=l converges for z = 1, but diverges for (4) False. See the example in (3). (5) True. Same example as in (3). L zn2 n oo (6) True. For example, consider . n=l (7) True. The radius of convergence is � 1 and 1 1 +ii= v'2 > 1. Solution to Exercise 4.12 Since d2n d 2n+l sinz = (-ltsinz and d 2n+l sinz = (-ltcosz, 2n dz z and sinO = 0 and cosO = 1, we have 1 dn z3 z5 + ••• +- zn = z- sinz = '"' - ( . sinz) n z=O 3! 5! L.., n! dz n =O I 00 z z - - •· •. Alternately, + 1 2. 41. + oo oo 1 · 1 exp(iz) + exp(-iz) 1 cosz = -----'-----'-----'-� = - ( L -i nz n + L - ( -1) ni·nz n ) . 2 2 n =O n! n! n =O 2 n Since i = (-lt, we have 2 4 Similarly, cos z = 1- z2 iz 3 z 4 + iz 5 z6 1 + iz + + 4! ( 6! · · · 2! 2 4 2 5 6 3 iz z z z iz ... 1 - iz + 4! 2! + 6! + ) 1 4 1 1 = 1- -z 2 + -z - -z 6 + - · · · . 2! 4! 6! cos z = 1 3! 5! - 3! 5! - 240 A Friendly Approach to Complex Analysis Solution to Exercise ,4.13 Let p(z) = z6 - z4 + z2 - 1, z E C. Then p'(z)=6z5 - 4z3 + 2z, p"(z) = 30z4 - 12z2 + 2, p"' (z)=120z3 - 24z, p (4l (z)=360z2 - 24, p (5l (z)= 720z, p (6l (z) = 720, p (7l (z) = p (8l (z)= · · · = 0. Hence p(l)=1- 1 + 1- 1 = 0, p'(l) = 6 - 4 + 2 = 4, 1! p"(l) 30- 12+2 = = 10 ' 2! 2 p"' (l) 120- 24 --=---=16 ' 3! 6 4 ( p ) (1) = 360- 24 = 14, � 24 p (5 l (l) 720 =6 ' = 5! 120 p (6)(1) = 720 = 1. -6!- 720 Thus for all z E <C, z6 -z4 + z2 - 1 p'(l) p (6)(1) = p(l) + �(z - 1) + · · · + - -(z- 1)6 +0 61 = 4(z -1) + lO(z-1) 2 + 16(z-1)3 + 14(z-1)4 +6(z-1)5 +(z-1)6 . Solution to Exercise 4.14 (1) The function z H exp(z2 ) possesses a primitive, say g, in the simply connected domain <C. Thus f(z)= 1 'YDz exp((2 )d( = 1 'YOz g'(()d( = g(z) - g(0). 241 Solutions So f'(z)= g'(z)= exp(z 2 )= I L 1n1 00 2n z n=O . . Consequently, I d2n+l 1 d2n 1 1 f'(z) = and (z) = 0. 2 (2n)! dz n (2n+1)! dz2n +J' z=O z=O n! (2n)' Hence j( 2n+ll(O) = -,-· and J( 2n+ 2 l(O) = 0. Also, f(O) = 0. Thus n. oo j(n)(0) oo j( 2n+l)(O) oo l ---z 2n+l = --Zn = z 2n+l. f(z)= n! + 1)! + l)(n!) (2n (2n n =O n =O n =O (2) For lzl < 1, 1 -- = l -z+z2 -z3 +z4 -+··· z+l and since power series are holomorphic in the region of convergence with complex derivative given by termwise differentiation, we obtain for lzl < 1 that l d l ---2 = - --= -1+2z - 3z2 +4z3 -+ · · • . (z+1) dzz+l Multiplying both sides by -z 2 gives L z L L L 2 oo = z 2 - 2z3 - 3z4 +- ...= (-1 r . ( n - l) .Zn (z + 1) 2 n=2 for lzl < 1. So we have co = c1 = 0 and en = (-lt · (n - 1) for n � 2. Solution to Exercise 4.15 Let z EC. Let R > lzl. Then (n+l)! · max lf(z)I Rn +l lzl:":'.R . max M-lzln= (n+l)! -M-Rn = (n+l)!M_ < (n+l)! Rn +l lzl:','.R Rn +l R But the choice of lzl > R was arbitrary, and so f(n +l)(z)= 0. Since z EC was arbitrary, we obtain that j(n +l) 0 in C. By Taylor's Theorem, for all z EC, n oo j( k )( ) j( k )( ) f(z) = �zk , �(z - ol = k =O k =O since j(n+l)(0) = f(n + 2 )(0) = f(n +3)(0) = · · · = 0. So we see that f is a polynomial of degree at most n. If n = 0, then f is a bounded entire function, and our conclusion ob­ tained above says that f is constant. So the special case when n = 0 is Liouville's Theorem. lln+l)(z)I � = L L A Friendly Approach to Complex Analysis 242 Solution to Exercise 4.16 1 By the Cauchy Integral Formula, 2013! -27ft . Hence 1 c d 2012 . sin z dz = s nz I d Z 2012 r Z2013 z=0 = (-1) 2012/2 sinzl z=O =0. sinz dz 2013 =0. cZ Solution to Exercise 4.17 z0, there exists a 8 > 0, which can be chosen R, such that g(z) =/-0 for lz -zol < 8. We have f(zo) =0, but f(z) = (z-zorg(z) (lz-zol < R) shows that f(z) =/-0 for O < lz-zol < 8. By the continuity of g at smaller than So by the Theorem on Classification of Zeros, there exists an m E N, which is the order of zo as a zero off and a gholomorphic in Then for lz - zol that this implies 0 =/- while if m 0 =/- < R, D such that g(z0) =/-0. (z - zo)m g(z) = (z - zorg(z). We show m = m. For if m > m, then we get the contradiction that g(zo) = > m, we have lim z➔zo g(z) = lim z➔zo (z - zo)m - m g(z) =0 · g(zo) =0, then we get the contradiction that lim g(z) = lim (z - zo)m - mg(z) =0 · g(zo) =0. g(zo) = z➔zo z➔zo Consequently, m = m, and so Zo is a zero of order m = m. Solution to Exercise 4 .18 (1) We have f(z) = (1 + z2 ) 4 = ((z - i)(z +i)) 4 = (z - i)4 (z +i) 4 , g(z) := (z +i) 4 , g is entire, g(i) = f(z) = (z - i)4g(z). Soi is a zero off of order 4. and so with (2) We have f(2n1ri) = 1 -1 J'(2mri)= expzl So 2n1ri is a zero of f of order 1. =0, and (2i) 4 = 16 =/- 0 and . = 1 =f. O. z=2n1ri Solutions 243 1 (3) f(O) = 1 - 1 + (0)2 = 0, and we have 2 1 (1-cos(2z)) 1 . f(z) = cosz -1 + (smz)2 = cosz-1 + · 2 2 2 3 1 = cosz ---- cos(2z) 4 4 = (1- �� + :; - :� +-...) - � _ ! (l _ 4z2 + 16z4 _ 26 z6 + _ ... ) 4 2! 4! 6! 1 + ! . )z 2 + ( .!_ _ ! . 6 )z4 + ... = (1 _ � _ ! ) + 4 4 2! 4 2! 4! 4 4! (-.!. ..___....., 0 ± ________, 0 '-----v------" #0 and so zo is a zero of order 4. Solution to Exercise ,4..19 First we note that if z is a point that is distinct from z0 in the disc, then f(z)-=/- 0. By the result on the classification of zeros, f(z) = (z-zo)g(z), where g is holomorphic in the disc, and g(zo)-=/- 0. Thus _1_ 21ri 1 zf'(z) d z "I f(z) l + 1 = _ _ z( · g(z) (z zo) · g'(z)) dz 21ri "I (z-zo)g(z) z(g(z) + (z - zo) · g'(z)) 1 g�) � = 21ri "I (z -zo) + = z(g(z) (z zo) · g'(z)) J (Cauchy Integral Formula) z=zo g(z) zo(g(zo) + 0 · g'(zo)) g(zo) = zo. 1 1 A Friendly Approach to Complex Analysis 244 Solution to Exercise 4-.20 By the result on the classification of zeros, f(z) = (z - z0)mg(z), where g is holomorphic in D and g(z0) =/=- 0. Thus (f(z))2 = (z - zo)2m ( g(z))2 . ....__., =:G z ( ) Clearly (f(z0)) 2 = 0, and G is holomorphic in D with G(zo) = (g(z0)) 2 =/=- 0. Hence z0 is a zero ofz r-+ (f(z))2 of order 2m. Also, J'(z) = m(z - zor- g(z) + (z - zor9'(z) m 1 = (z - zo) - (mg(z) + (z - zo)g'(z)), 1 and so f'(zo) = (zo-zo)m- 1g1(zo) (m� ) O· g1(zo) = 0. As 91 is holomorphic and l g1 (zo) = mg(zo) + 0 · g'(zo) = mg(zo) + 0 = mg(zo) =/=- 0, it follows that zo is a zero off' oforder m - l. Solution to Exercise 4-.21 Consider the function f : <C ➔ <C given by f(x,y) = x sin ! ifx =/=- 0, X and f(0, *) = 0. Then f is obviously continuous at any point (x o, Yo) where x0 =/=- 0. Moreover, since lf(x, Yo) - f(O, Yo)I = Ix sin� - 01 = lxl I sin� I ::':'. lxl · 1 = Ix - 0I for all x =/=- 0. Thus f is continuous also at the points (0, *). Hence f is continuous everywhere in <C. By the definition off, 0 is a zero off. 0 is clearly not an isolated zero off, since f (_!_, nn sin(mr) = 0, o) = _!_ nn n E N. Also, f is not identically zero in any disc centered at 0 because for n E N, f C2n ! O = 2 l)n' ) ( n ! l)n sin ((2n + l)i) = (2n ! ( l =/=l) n - r o. 245 Solutions Solution to Exercise 4.22 We know that for all x1, x2 EJR, cos(x1 + x2) = (cosx1)(cosx2)- (sinx1)(sinx2). Fix x EJR and consider the entire function f given by f(z) := cos(z+x)- ((cosz)(cosx)- (sinz)(sinx)), (5.22) z EC. We have f(y) = 0 for ally E JR, thanks to (5.22), and so by the Identity Theorem, f(z) = 0 for all z EC, that is, cos(z + x) = (cosz)(cosx)- (sinz)(sinx), z EC. (5.23) But the choice of x E JR was arbitrary, and so (5.23) holds for all x E R Next, fix a z EC. Consider the entire function g given by g(w) := cos(z + w)- ((cosz)(cosw)- (sinz)(sinw)), w EC. Then we have g(x) = 0 for all x EJR by (5.23). Another application of the Identity Theorem yields g(w) = 0 for all w EC. Hence cos(z + w) = (cosz)(cosw)- (sinz)(sin w), w EC. (5.24) But the choice of z E C was arbitrary. Consequently, (5.24) holds for all z EC (and for all w EC). Solution to Exercise 4.23 Suppose f, g EHol(D) are such that (5.25) (f · g)(z) = f(z) · g(z) = 0, z ED. Suppose there exists zo E D such that f(zo) -:/- 0. By the continuity off, there exists a 8 > 0 such that f(z)-:/- 0 whenever lz-zol < 8. (5.25) then implies g(z) = 0 for lz-zol < 8. By the Identity Theorem, g 0 in D. So Hol(D) has no zero divisors. On the other hand, C(D) is not an integral domain, and we show this below. Let zo ED and let 8 > 0 be such that the disc Li:= {z ED: lz-zol < 8} CD. Consider the continuous function r.p: JR---+ JR defined by 0 if t :S 0, r.p(t) = { t if t > 0. Define f, g by f(z) := r.p(Re(z-zo)), = for z ED. g(z) := r.p(-Re(z-zo)), 246 A Friendly Approach to Comp lex Analysis g > 0 here Fig. 5.21 f > 0 here Construction off, g E C(D) using cp. Being the composition of continuous functions, f, g E C(D). Also f(z) > 0 for all z in the right half of�, and so f-:/- 0 in C(D). Similarly g(z) > 0 for all z belonging to the left half of�, and so g -:/- 0 in C(D). Nevertheless, f ·g=O. Solution to Exercise 4.24 (1) False. Take D = C, f = exp, g = l. Then for all n E N, we have f(21rin) = exp(21rin) = 1 = g(21rin), but f -:/- g (for example, because f(i1r) = -1-:/- 1 = g(i1r)). (2) 'Irue. (3) 'Irue. Let 'Y(t) = x(t) + iy(t), t E [a, b]. Consider a point to where either x'(to) or y'(to) is nonzero. (If they are both always 0, then a = b, a contradiction.) Suppose x'(to) > 0 (the other cases are handled similarly). Then x(t) > 0 in a neighbourhood of t0. So x is increasing there. Take tn = t0 + n ?: N, where N is large enough so that to+ if E [a, b]. Set Zn = 'Y(tn)- Then (zn)n2: N is a sequence of distinct points (since their real parts ate distinct), which converges to 'Y(t0). By the Identity Theorem, f = g in D. (4) 'Irue. By the Taylor expansion around w, f =gin a disc with center w, and so by the Identity Theorem, f =gin D. ¼, Solution to Exercise 4.25 Let K = {z E C : lzl :::; l}. For each z E K, there exists a smallest n(z) E {0, 1, 2, 3, · · •} such that for all w near z, f(w) = L C (z)(w - zr 00 n=O n and Cn(z)(z) = 0. Hence u< n(z))(z))/((n(z))!) = 0, and so f( n(z))(z) = 0. Let rp : K-+ NU {0} be defined by cp(z) = n(z). Since K is uncountable, 247 Solutions while NU {0} is countable, there exists an N such that rp-1(N) is infinite. Let (zn )nE N be a sequence of distinct points in rp-1(N). As K is compact, this has a convergent subsequence (znk hE N with limit, say z* E K. As j(N)(znk ) = 0 for all k, by the Identity Theorem (applied to j(N)), we have j(N) = 0 in K, and so also in C. By Taylor's Theorem, oo j(n)(O) n N-1 j(n)(0) n _ f(z) z _ z , I:-, n.1n.- n=O for all z E <C, and so f is a polynomial. L- n=O Solution to Exercise 4.26 Let zo E D be such that Jf(zo)I 2: Jf(z)J for all z E D. By the Maximum Modulus Theorem, f is constant in D, a contradiction. Solution to Exercise 4.27 Let J(zo) =/- 0. Then Jf(zo)I > 0 and so for all z ED, Jf(z)J 2: Jf(zo)I > 0, implying that for all z E D, f(z) =/- 0. Now consider the holomorphic function g := 1/ f in D. We have 1 1 Jg(zo)I = IJ = Jg(z)J, z ED, (zo)I 2: IJ(z)J and so by the Maximum Modulus Theorem, g is constant. But then f is constant too. Solution to Exercise 4.28 Let zo be a maximizer, which exists since z f-+ Jf(z)J is continuous and K := {z E (C : JzJ :::; 1} is compact. But z0 can't be in the interior of L: indeed, if JzoJ < 1, then by the Maximum Modulus Theorem (applied to f on lill := { z E (C: JzJ < 1}), f would be constant in lill, which it clearly isn't. Hence z E 'll' := {z E (C: JzJ = 1}. So max Jf(z)J = max Jf(z)J = max I exp(2it) - 2J = 1- 1 - 2J = 3. zEK tE[0,21r) izl=l Similarly, if z1 is a minimizer, z1 can't be in the interior of K. Indeed, zf 2 =/- 0, and so by the Minimum Modulus Theorem, f would be a constant, a contradiction. So z1 E 'll'. Hence min Jf(z)J = min Jf(z)J = min I exp(2it) - 2J = JI - 2J = 1. zEK tE[0,21r) izl=l See Figure 5.22. - 248 A Friendly Approach to Complex Analysis '][' exp(2it) minimizer / maximizer +-----.-----+=====at -1 2 1 0 Maximizer and minimizer for Jz 2 - 21 in the unit disc. Fig. 5.22 Solution to Exercise 4.29 For z E A1 := {z E (C: 0 < lz-11 < 1}, we have 1 z(z-l) 1 (z-l+l)(z-1) = = z z � (1 - (z-1) + (z-1) 2 - (z -1) 3 +- · • ·) 1 � -1 (z-1) - (z-1) 2 + (z-1) 3 + 1 - + .. • . On the other hand, for z E A1 := {z E (C: 1 < lz -11} 1 z(z-1) 1 (z-1+l)(z -1) = = 1 (z-1) 2 1 (z _ l) 2 (l + 1 1 (1- z-l + (z -1) 2 1 1 (z-1) 2 - (z -1) 3 + 1 - _1 _) z-l 1 (z-1) 3 1 (z-1)4 - (z-1) 5 - + .. ·) +- .. · · Solution to Exercise 4.30 By the result on classification of zeros, f(z) = (z-zorg(z), z ED, where g is holomorphic in D and g(zo) =I- 0. Also, since zo is the only zero off in D, g( z) =f=. 0 for all z E D. So 1/g is holomorphic, and it has a Taylor expansion in a disc with center zo: there exists an R > 0 such that 1 g (z) = L en(z - zot for lz-zol < R, 00 n=O 249 Solutions and co-/- 0 (because g( zo)-/- 0). Thus for 0 1 J( z ) = = 1 ( z- zo)mg( z ) Co + ( z- zO )m 1 < lz - zol < R, 00 cn ( z- zot ( z - zo)m � = 00 Cm-1 � n + ···+ -- + � Cm+n (Z-Zo) • m -1 ( z - zo) z - zo n =O C1 Hence 1/ f has a pole of order m at zo. Solution to Exercise 4.31 z t---+ ( z - z0 )m f( z ) has a holomorphic extension, say h, to ·( lim ( z - zor f( z ) z-tzo = o), D. Also, and so h( zo)-/- 0. Moreover, since f( z )-/- 0 for zED, also h( z )-/- 0 for all zED. Hence 1 f( z ) = ( z - zor h( z ) for all zED\ {zo} and g defined by g( z ) is holomorphic in D. Since = : ( z - zor 1 h( zo) h( z ) -/- 0, z0 ' zED is a zero of g of order m. Solution to Exercise 4.32 We must have Cn = 0 for all n < -m. Thus L 00 C-m C-m r1 + ...+ � + Cn (Z - zot+ m z - zo ( z - zo m-1 ( z - zo) n =O -1 So ( z- zorf( z ) = C-m +c-m+1(z-zo)+·· •+c-1(z-zor +···. Hence m ( z - z0 ) f( z ) has a holomorphic extension, say g, to f( z ) = �:={zEC: lz-zol <R}. So for lz-zol < R, g( z) = c_m +c-m+1(z-zo)+···+c-1(z-zo)m-l +···. By Taylor's Theorem, c_ 1 = dm-l g 1 (zo). m ! ( - 1) dzm-l 250 A Friendly Approach to Complex Analysis But gCm-l) is holomorphic in b. and in particular, continuous at z0. So Also, for 0 < lz- zol < R, g( z ) = ( z - z or f( z ), and so for z -/- Zo in b., m 1 d g(m-l)( z ) = --(( z - z orf( z )). dz m- 1 Hence c_1 = 1 1 im-l)(zo) = --- lim g(m-ll(z) (m - l)! (m -1)! z-tzo m -1 1 d lim m _ 1 (( z - z orf( z )). ( m- 1)1. z ➔zo dz Solution to Exercise 4.33 (1) (2) (3) (4) (5) True, since c_1 = 1-/- 0, and c_2 = c_3 = · · · = 0. True. True. True. True. Solution to Exercise 4.34 (1) sin z does not have a singularity at 0, and for sin z z E (C, 5 z3 z = z- +- +• • • . 3! 5! (2) sin ! has an essential singularity at 0, since for z-/- 0, z . 1 sin-=••• z 1 1 1 +-- --3 + -. 5 5!z 3!z z sinz (3) -- has a removable singularity at 0, since we have that z sinz lim z • -- = lim sinz = 0. sinz Also, -z z➔O z z➔O 1 1 = 1- 1 , 0. -z2 + z 4 - z 6 + - · · · £or z 4 3! 5! 7! 251 Solutions sinz (4) -2- has a pole of order 1 at 0,since for z =f. 0, z sinz 1 z z3 z 5 = + + - ... . 7 � 3! 5! 7! (5) 1/( sin(l/z)) does not have an isolated singularity at 0 because with Zn = 1/(mr),n EN,we have sin _!_ =sin(mr) =0 1 Zn n--+oo =- --+ 0. ( We h ad also observed this . m . Example 4.13.) n1r ( 6) z sin - has an essential singularity at 0 since for z =f. 0 z 1 1 . 1 z sm - = · · · - +-4 - -2 + 1. 3!z 5!z z and Zn Solution to Exercise 4.35 ( 1) False. limje ½j =lime ½ =O,and so,(limlexp !l=+oo). x,?'O x,?'O z--+0 z (2) '!rue. There exists an R > 0 such that C-m C-1 C-m +l """' f( Z ) = + · · · + -- + � Cn ( Z - zo)n, + (z - z0 )m (z - zo)m -l z - zo n =O for 0 < !z - zol < R, and so with P :=C-m + C-m +1(z - zo) + · · · + C-1(z - zo)m -l, we have for 0 < !z - zo[ < R, f(z) - (Z p(z) - Zo )m =� � cn (z - zo)n . n =O ( 3) 'Irue. Let the order of0 as a zero off be m. (Take m =0 iff(0) =/- 0.) Then there exists a holomorphic function g such that f(z) =zm g (z) and g(0) =/- 0. Hence for n > m,and z =f. 0, m f(z) g (z) -- z g(z) --zn - m Thus [g(z)[ f(z) 1 =lim lim =+oo =[g(zo)[ · lim - -z--+0 I zn I z--+0 [z[n-m z--+0 [z[n m since g (zo) =f. 0 and n > m. 252 A Friendly Approach to Complex Analysis (4) True. In some punctured disc D= {z E C : 0 < [z - zo[ < R}, f,g are nonzero, and there exist holomorphic functions h f , h9 such that h 1 (zo ) =/- 0,h 9 (zo ) =/- 0,and for all zED, ™ t =(z - zo )™ 1 h 1 (z ) , tz =(z - z0 ) 9 hg(z ) . g ) f z) Thus h 1 (zo ) h 9 (zo ) =/- 0 and for all zED l =(z - zo )™ 1 +m 9 h 1 (z ) h 9 (z) . f(z ) g(z ) Consequently,fg has a pole of order m1 +m9 at z0• Solution to Exercise 4.36 Consider f given by f(z) = (exp�)+exp( � )' zEC\{0,1 }. l z Then f is holomorphic in C\{0,1 }. The function exp( 1 /( 1 - z)) is holo­ morphic in a neighbourhood of z= 0, while the function exp(l/z ) has an essential singularity at 0. Thus,their sum,namely f, has an essential sin­ gularity at 0. (Why?) On the other hand exp(l/z) is holomorphic in a neighbourhood of 1 ,while exp(l/( 1 - z)) has an essential singularity there. So f has an essential singularity at z=l. Solution to Exercise 4.37 We have seen that if zo is an isolated singularity of a function g with the Laurent series expansion g(z )= Cn(z - zot L nEZ for O < [z - z0[ < R and for some R > 0, and there are infinitely many indices n < 0 such that Cn =/- 0,then z0 is an essential singularity of g. However,for the given f, the annulus for the Laurent expansion z-1 +z-2 +z-3 +... is given by [z[ > l. The correct annulus to consider for deciding the nature of the singularity at z = 0 is of the form O < [z[ < R for some R > 0. In fact,for [zl < 1 we have 1 f(z )=- _ =-(l+z+z2 +z3 +···), 1 z showing that f is holomorphic for lz[ < 1 ,and f does not have a singularity at z=0. 253 Solutions Solution to Exercise 4.38 It is clear that oz is an isolated singularity of Jg . Indeed, since J and g both have an isolated singularity atzo, we have that J is holomorphic in a punctured disc 0 < lz-zo <l Rf for some Rf > 0, and g is holomorphic in a punctured disc 0 < -z lz o< l R9 for some R9 > 0. ThusJg is holomorphic in the punctured disc 0 < zI zo I < min{ RJ, R9}. Suppose thatJg has a removable singularity or a pole atz 0• Then there exists an m � 1 such that = lim (z-zo) J(z)g(z) = 0. z➔zo SinceJ has a pole atoz , say of order m1, J C-m1 (z-zo )mf is nonzero nearzo and C-m1+1 C-1 n Z Zo ) , + + z -+� + -z L..., Cn (o (z o )m _ 1 · · · -z n =O for 0 < lz zo< l Rand for some R > 0. Here C-m-:/, 0. So forz-:/- zo, J (Z ) = f but nearoz , we have = (z-zo) g(z ) = 1 Z Zo ) m1J()Z (- 1 -zo) =J(z)g(z) (z • -zo) =(z ...______...., __________ -tO -tO 1 0. 0- 0. --+--· C-m1 So g must have a pole at z0 or a removable singularity at z0, a contradiction. Consequently Jg has an essential singularity at z 0. z-tzo Solution to Exercise 4.39 Set E := 1/n =: o (> 0). By the Casorati-Weierstrass Theorem, there exists a Z n in the punctured disc aroundzo with radius o such that ( JI z )-wl < E. n That is, zl o z < l l/n and ( J I z ) wl < E. Hence (z N converges to ) n n n nE oz , and (J(z )n) nE N converges tow. Solution to Exercise 4.40 1 + exp z = 0 if and only if z E {'1ri + 21rni: n E Z}. So Log( z ) J(z) := 1 + expz is holomorphic in C ( \( oo,0]) \ {1ri + 21rni: n E Z}. J has poles of order 1 at the points {1ri + 21rni: n E Z}, of which two lie inside the given path --y:-1r and 31ri. See Figure 5.23 . A Friendly Approach to Complex Analysis 254 3ni � ✓ ,1 � ::c ' -?ri , ,, ✓ ,2 Fig. 5.23 1 We have J(z)dz = 'Y 1 J(z)dz + 1 The curves ,1 and 12. J(z)dz = 2ni (res(J, 3ni) - res(!, -ni)). 12 11 So we need to calculate res(!, 3ni) and res(!, -ni). We can write Log(z) = C-1,31ri h --� . + 31ri, 1 +expz z - 3ni h w here 31ri is holornorphic in a neighbour hood of 3ni. Thus z - 3ni (z - 3ni)Log(z) _ . _ . - 1llll . · L Og (Z ) C-1 ' 3,ri. - 1llll z--+31ri z--+37ri expz - exp(3ni) 1 +expz 1 · Log(3ni) = -1 ( logl3nil + i '!!..2) = expz 1 z=31ri = -log3 - logn - i'i. We can write Log(z) C-1,-1ri " + h-,ri, 1 +expz z - (-ni) w here h-1ri is holornorphic in a neighbour hood of -ni. Thus z - (-ni) (z - (-ni))Log(z) . . · Log(Z ) = 1llll C-1 ' -1ri· = 1llll z--+-1ri expz -exp(-ni) z--+-1ri 1 +expz 1 -Log(-ni) = -l ( logl - nil + i( - '!!..)) = 2 expzlz=-1ri 1 . 7f = -logn+ i . 2 Log(z) dz=2ni(-log3-logn- i'!!..+logn- i'!!..) = 2n2 - (2nlog3) i. So 2 2 l+expz 1 255 Solutions Solution to Exercise 4.41 Let 'Y be the circular path given by "!(0) { 2n }0 cos 0 d0 5+4cos 0 = = 1 1 = exp(i0) for 0 E [O,27r). Then 1 z+ !. z __ z2 +l 2 . _!_dz = dz 2 +5z+2) !. "' "'2iz(2z '5+4z+ ' --z iz 2 z2 +l dz. 2iz(2z+l)(z+2) 7 Let f be defined by z 2 +1 f (z) : = 2iz(2z+l)(z+2)· Then f has three poles, at 0, -1/2, -2, and each is of order 1. Of these, the poles at O and -1/2 lie inside 'Y· So by the Residue Theorem, 1 2n 0 cos 0 dB 5+4cos 0 = 27ri(res(f, 0) +res(!, -1/2)) z·(z +1) ·(i·Im ----'-------'-+ 2 = 27ri (z+l/2)·(z2 +1) 1.Im ) z ➔ -I/ 2 2iz(2z+l)(z+2) z ➔ O 2iz(2z+l)(z+2) 1 . .§.4 1 = 27fi· (--- + ) 2i·1·2 2i.•(-½)·2·! 7f 3 = 1 5 27ri· (- - -) 4i 12i Solution to Exercise 4.42 (1) Let Ji be defined by 1 fi(z) = l+z2· Then Ji has poles at i and -i, both of order 1. Hence 1 1 . . . . z-i -- dx = - •2m •res(JI,i) = 7ri· hm-2 z ➔ i 1+Z 2 2 1+ X 0 1 1 7r = 7ri· lim -- = 7ri· - = -. 2i 2 z ➔ i Z +i (2) Leth be defined by 1 h(z) = (a2 + z2 (b2 + z2 ) . ) 1 00 256 A Friendly Approach to Complex Analysis Then h has poles atai, -ai, bi, -bi, all of order 1. As h is even, rlo = (a2 +x2)(b1 2 +x2)dx = 1 2· 7f = i = 27ri (res(h,ai) +res(h,bi)) 1__ + __1__ ) (-a (b (a )2a b )2b 2 - 2(a2�b2) (3) Leth be defined by 2 i (i-¾) 2 - = 2 i 2ab(:+b)" 1 h(z) = ( 1+z2 )2 Then h has poles at i and -i, both of order 2 . We have ro = ( 1+x1 ) dx = !2 · 27r · res(h, ) l i 2 2 i 1 7ri m · li ..!!:_ ((z- i )2 • . ) . 2 1! z---+idz (z-i) (z+i)2 -2 . -2 . . 7r = 7rz • hm --- = 7rz • - = -. z---+i(z+i)3 -8i 4 = (4) Let f4 be defined by Then f4 has poles at i Pl = exp (: ), P2 = exp 1= Ct), p3 = exp c;i ), p4 = exp c;i ), all of order 1. Hence o 1+x 2 --4 dx 1+x 1 = 2 · 27ri (res(f4,p1)+res(f4,p 2 )) +p12 l+p 22 7ri ( l pf + ) 4 4P� + 1 + _ _ + � + _l_) = 7ri (-Pl P2 + _l_ + _l_) = 7ri 4pf 4p1 4p� 4p2 4 4p1 4p2 i i e e (i ) e (( )-e xp 7r/4)- xp 7r /4 + xp 7r/4 _ . xp( i7r/4) -ITT () 4 4 -i sin(7r/4) i sin(7r/4) = 7ri (+ = �) = 7ri · (-i ) . _2.._ 2 y'2 y'2 2 (_E!_ = Solutions 257 Solution to Exercise 4.43 1 By the Residue Theorem, expz expz z 21ri 1· dn exp --· d ( z n+l ·-1m-z1ri-res -_ 2 . n ( z n +l, c zn +l n! z--+0 dz zn +l) n 21ri . d 21ri . 21ri = - · hm - expz = - · hm expz = - · expO n! z--+0 dz n n! z--+0 n! 21ri . 21ri = l= . n! n! Hence exp(cos0+isin0) 21ri . - = · i.(cos0+ism . O)dO . n! cos((n+1)0)+ism((n+1)0) 0 o) _ 121r 121r 121r 1 =i exp(cos0+isin0) • ( cos(n0) - isin(n0))d0 =i exp(cos0)( cos(n0 - sin0) - isin(n0 - sin0))d0 . Equating the imaginary parts, we obtain exp(cos0) • cos(n0 - sin0)d0 ok Solution � = 1. n. to Exercise 4.44 For z in a small punctured disc D centered at zo, f(z) =I= 0 and J(z) = (z - zo)h(z) (5.26) for some holomorphic function h such that h(zo) =I= 0. From (5.26) we have f'(z) = h(z)+(z - zo)h'(z) and in particular, f'(zo) = h(zo)- Now 1 1 for zED\ {zo}, J(z) (z - zo)h(z) ¾ and since where do = is holomorphic in D, 1 h(zo) 1 h(z) = = do+d1(z-zo)+··· for zED, 1 f'(zo)" Hence for zED\ {z0}, 1 1 do -·(do+d1(z - zo)+·· ·) = --+d1+d2(z - zo)+··· = f(z) z - zo z - zo and so res (y, zo) = do = f'(�o) . 258 A Friendly Approach to Complex Analysis Solution to Exercise 4.45 Let J be given by J(z) = sinz. J has zeros of order 1 at kn, k E Z. So by the previous exercise, 1 1 1 res (--, kn) = , 1 = -l k = (-1) k k'Tf sinz sin zlz=k1r ( ) ( ) Solution to Exercise 4.46 (1) We have Jo = l ::=; 2° = 1, Ji = 1 ::=; 21 = 2, and if Jm ::=; 2m for all m ::=; n (for some n � l) , then m m m m m Jm+I = Jm + Jm- 1 ::=; 2 + 2 -l = 2 -l · 3 < 2 -l · 4 = 2 +l. (2) If lzl < 1/2, then v'lcn zn l = lzl ::=; ffn · lzl = 2lzl < 1 for all n EN. So by the Root Test, vfcJ · 00 n Llenz l n=O converges when lzl < 1/2. Hence the radius of convergence of F is � 1/2. (3) We have for lzl < 1/2: zF(z) = Joz+ fiz2 + hz3 + • • • , Joz2 + fiz3 + · · · . z2 F(z) = Adding these, zF(z)+ z2 F(z) = 1 · z +(Ji+ Jo)z2 + (h + fi )z3 + · · · = fiz+ hz2 + hz3 + · · · = (Jo+ fiz+ hz2 + hz3 + · · ·) - Jo = F(z) - l. Hence 1 = F(z) - zF(z) - z2 F(z) = (1 - z - z2 )F(z). 1 1 So for lzl < - , we have F(z) = 2 l-z-z2 (4) We have F(z) 1 Jo+···+ Jn -IZ n -I + Jn Z n + Jn+1zn+I + · · · zn+I Jn Ji Jo = n+I + n + · · ·+ + Jn+I + Jn+2 z+ · · · , --; z z (5.27) 259 Solutions and so 1 , res ( n+1 (l- z - z2 ) z o) = coefficient of! in (5.27) =f n· z (5) We have for Jzl = R > 2: 11- z - z2 1 2::: Jz2 + z l - 1 =l z l · lz+ 11- 1 =R · lz+ ll - 1 2: R· (lz l - 1)- 1 = R · (R- 1)- 1 = R2 - R- l > 0 (since R > 2). Hence if CR : [0,27r] -+ (C is the circular path given by CR(t) = Rexp(it), t E [0,271"], then I }{0R I l l < _l_ . · 271" R dz n n 2 2 + + R l R - R- l z l(l- z - z ) 1 1 R ➔ oo . Rn R2 - R- l -+ O 1 Define G by G(z) := n+ . Then G has z l(l- z - z 2) (a) a pole at 0 of order n + l, -1+v'5 of order 1, 2 -1- v'5 (c) a pole at of order 1. 2 Thus (b) a pole at 1 -l+v'5 -1-v'5 1 res(G,0)+res (G,--- ) +res (G,--- ) =-. G(z )dz 2 2 271"Z CR for all R > 2. Hence res(G,0)+res ( -l+v'5 ,a, 2 ) +res (G, -1-v'5 2 ) \ = Rlim � { G(z )dz =0, ➔ O 27ri JcR . that 1s, fn = -res ( G, -l+v'5 -1- v'5 ) - res ( G, ). 2 2 260 A Friendly Approach to Complex Analysis We have res 1 + v'5) = (G , ---2 1· 1m z➔ -1!v's ( z- -1 + v'5) 2 1 Also, Hence ) f =-1 . ( l+v'5 )n+l _ _!__ ( 1-v'5 n+l n y'5 y'5 2 2 . ( ( 1 +2 - � Js) •+' - ( Js) •+') . 1-2 1 z n+l (l - z - z 2 ) 261 Solutions Solutions to the exercises from Chapter 5 Solution to Exercise 5.1 (1) We have for (x, y) E JR2 \ {(O, 0)} that 2x 1 --2x=--, 2 2 2 + x +y 2 x y 2x 2 2y 2 + 2x 2 -4x 2 (2x) - --,-,--....,........, (x 2 +y 2)2 x 2 +y 2 (x 2 +y 2)2 · 2(y 2 -x 2) (x 2 +y 2)2 · Similarly noticing the symmetry in the roles of x and y, we have 2y a 2u 2(x 2 -y 2) and 2 = 2 + 2 2 . y ) (x xz + yz' ay au ay Consequently a a 2 u 2(y 2 -x 2) 2(x 2 -y 2) 2u -+= ---'---------'---+---=0 . 2 a y2 (x 2 +y 2)2 (x 2 +y 2)2 ax Since u is C2 and 6-u=0 in JR2 \ {(0, 0)}, u is harmonic there. (2) We have for (x, y) E JR2 that aU a zu ax ax 2 aU a 2u X = e cos y, ay a y2 • = e smy, So X x = e sm y, and, . . = ex( -smy ). a2 u + a2 u= e x siny + ex(-siny)=0 . As u is C2 and 6-u=0 ax 2 a yz IR , u is harmonic in IR2 . in 2 Solution to Exercise 5.2 Consider the vector space V of all real-valued functions defined on U, with pointwise operations. Then we know that V is a real vector space. We will show that Har(U) is a subspace of this vector space V, and hence a vector space with pointwise operations. We have (81) The constant function O assuming value 0 everywhere on U belongs to Har(U). Indeed 262 A Friendly Approach to Complex Analysis (S2) Let u , v E Har(U). Then 82 (u +v) 82 (u +v) [J2 u 82 v 8 2 u 8 2 v ----+----=-+-+-+8x2 8y 2 8x2 8x2 8y 2 8y 2 _ 82 u 02 u 82 v 02 v ) + (-+) - (-+2 2 8x 8y 8x2 8y 2 =o+o=o. (S3) Let a: E JR and u E Har(U). Then 82 u 82 u c.92 (o: · u) c.92 (o: · u) = + + 0: 0: 8x2 8y 2 • 8x 2 • 8y 2 = a: ( ::� + ::�) =a: · 0=0. Hence Har(U) is a real vector space with pointwise operations. Solution to Exercise 5.3 u :=x=Re(z) and u :=x +y=Re(z - iz) are harmonic in JR2, and their pointwise product is u • u=x • (x +y)=x2 +xy. We have a a a2 (u . u) a2 (u . u) 8x2 + oy 2 = 8x (2x +y)+ oy (x)=2+0=2#0. So the pointwise product of two harmonic functions need not be harmonic. Solution to Exercise 5.4 (1) Let u =ex sin y. We seek a v such that u + iv is holomorphic. So the Cauchy-Riemann equations must be satisfied. Hence av OU - = -- = -eX cosy. ax &y So ifwe keep y fixed, we obtain by integrating that v = -e x cosy+C(y), for a constant C(y), which depends on y. Thus :: =ex siny+ C'(y)= :: =ex siny, and so C'(y)=0, giving C(y)=K. So we try v :=-e x cosy. Then u + iv = ex siny+ i(-ex cosy)= ex (siny - icosy) = -iex (cosy+isiny)= -iexp(x+iy)= -iexp(z), where z=x+iy. Hence u+iv = -iexp z, which is indeed holomorphic. Hence v = -ex cosy is a harmonic conjugate for u :=ex sin y. 263 Solutions (2) Let u = x 3 - 3xy2 - 2y. We seek av such that u + iv is holomorphic. So the Cauchy-Riemann equations must be satisfied. Hence av au -=--= 6xy+2. ax ay So if we keep y fixed, we obtain by integrating that x2 2 v = 6 y + 2x+C(y) = 3x y + 2x+C(y), 2 for a constant C(y), which depends on y. Thus av au - = 3x2 + C' (y) = -= 3x2 - 3y 2 ay ax and so C'(y) = -3y2 , which gives 3 C(y) =-3 y + C = -y3 + C, 3 and so we try v = 3x 2 y + 2x - y3• With this v, we have u +iv= x 3 - 3xy2 - 2y + i(3x 2 y+2x - y3) = x 3 +3x(iy) 2 +3x2 (iy)+(iy) 3 - 2y + i2x = (x + iy)3+2i(x+iy) = z2 + 2iz for z = x + iy. Thus u +iv= z2 + 2iz, which is indeed holomorphic. So v := 3x 2 y+2x -y3 is a harmonic conjugate of u := x 3 - 3xy2 - 2y. (3) Let u := x(l + 2y). We seek a v such that u + iv is holomorphic. So the Cauchy-Riemann equations must be satisfied. Hence av au -=--=-2x. ax ay So if we keep y fixed, we obtain by integrating that x2 v =-2 + C(y) = -x2 + C(y), 2 for a constant C(y), which depends on y. Thus OV OU - = CI (y) = -= 1 +2y. ay ax Thus y2 + C = y + y2 + C. 2 So we try v :=-x2 + y + y2 . With this v, we have u +iv= x(l + 2y)+i(-x 2 +y+y 2 ) = x+iy + 2xy+i(y2 C(y) = y+2 · - x2 ) = x + iy - i((x2 - y2 ) + i2xy) = x +iy - i(x + iy) 2 = z - iz2 for z = x + iy. Hence u + iv = z - iz2 is indeed holomorphic, and v :=-x2 +y+y2 is a harmonic conjugate of u := x(l + 2y). 264 A Friendly Approach to Complex Analysis Solution to Exercise 5. 5 Let v be a harmonic conjugate of u. Then f := u + iv is holomorphic in C \ {0}. Hence h := z 2 exp(-f(z)) is holomorphic in C \ {0}. We have 2 lhl = lzl 2 1exp(-f(z))I = lzl 2 e-Re(f(z)) = lzl 2 e-u = lzl 2 e-loglzl w 2 1 = 1. = lzl · But then h must be constant in each disc contained in C\ {0}. Consequently h must be constant in C \ {0}. Thus h' = 0. But 2 h' = 2z exp(-f(z)) + z exp(-f(z)) · (-J'(z)), and so f'(z) = �So 1/ z would have a primitive in C\ {0}. Now if, is the path 'Y(t) = exp(it), 0 ::; t ::; 21r, we have the contradiction that 1 1 �dz= f'(z)dz = 0. z 'Y Hence u has no harmonic conjugate in C \ {0}. 2 · 21ri = 'Y Solution to Exercise 5.6 Set u := x 3 + y 3• If f were holomorphic, then u would be harmonic. But fP u fP u & & = (3x2 ) + (3y2 ) &y &xZ &yZ &x for x =/- -y. Hence the answer is "no". + = 6x + 6y = 6(x + y) =/- 0 Solution to Exercise 5. 7 It suffices to show that if u is harmonic, then so are �: and �:. We know that u is infinitely many times differentiable. We have 265 Solutions Similarly Solution to Exercise 5.8 (1) Let b(x) = p(x) =co+ c1x + • · · + cdxd . Then p(z) := p(x + iy) =Co+ C1Z + · · · + Cd Z d is entire, and so h := Re(p(x+iy)) is harmonic. Moreover for all x E �, h(x, 0) = Re(p(x + iO)) = Re(p(x)) = Re(b(x)) = b(x). (2) We have b(z) := b(x + iy) := 1 is not defined at z = i. But 1 + z2 i z+i is holomorphic in the upper half-plane, and so its real part is harmonic there. Moreover, 0+1 h(x, 0) = x2 + (0 + 1)2 1 = x2 + 1 = b(x) for all x ER Solution to Exercise 5. 9 Let f be an entire function whose real part is u. (Note that C is simply connected.) Then exp(-f) is entire too. We have Iexp(- !)I= e-Re(f) = e-u:::; 1, since u(x, y) > 0 for all x, y E �- By Liouville's Theorem, exp(-f) is a constant. Hence I exp(-f) I is constant too, that is, e-u is constant. Consequently the real logarithm log(e-u) = -u is constant, and so u is constant as well. A Friendly Approach to Complex Analysis 266 Solution to Exercise 5.10 (1) For z =r exp(i0), we have exp (- :4 ) = exp (-r14 exp(-i40)) . l/n and 40 = -1r, that is, with 1 = : Xn + iyn , Zn : = :;:; exp ( -i. 4 4 we have u(xn ,Yn ) = exp(-n4 exp(i1r)) = exp(-n4 • (-1)) = en • So we have that (xn ,Yn ) -+ (0,0), but it is not the case that u(xn ,Yn ) -+ 0, showing that u is not continuous at (0,0). So if we taker (2) We have u(x, 0) = exp = (- 7r) \ ) (x + i)4 = u(0,y) exp (- (0 +\i)4 ) (3) We have ,::iu _u (0 ax ' 0) = lim x---+0 u(x = = 4 e-l/x , and exp (- i4�4 ) ' 0) - u(0 ' 0) X-0 = = and so 0 ::; I e- x � 4 I 4 4 4 . e-l/x - 0 hm --x---+0 X (The last equality follows from the fact that e 1 /x (- 1 \ ) = e- /y . exp = 4 . e- 1 /x hm -x---+0 X 1 4 = 0. 1 1 1 1 l+-+(-) +··•>2! x4 x4 x4 2 = ::; lxl3 .) Similarly, :: (0,0) = 0. Thus 4 au u d -(x, 0) - a-e- i/x - 0 (0, 0) a2u ax x ____ -(0 0) = lim a x = lim -=d= x---+0 x---+0 X a x2 ' x-0 4 -4 4 1 . e - 1/ x . � -4e- 1 /x = 1m ---- = lim ---- = 0. x---+0 x---+0 x6 x 4 2 1 1 1 1 e-l/x 4 _ + · · · > -, __ ::; 2lxl2 -) 0::; (As e 1/x = 1 +-+ ) ( 2x8 x6 2! x4 x4 Similarly, a2u (0,0) = 0. ay 2 a2u a2u Hence (0,0) + 2 (0,0) = 0 + 0 = 0. 2 ay ax I I 267 Solutions Solution to Exercise 5.11 (1) Let zo E D1. Then cp(zo) E D2. Let A be a disc with center cp(z0) and radius E > 0 small enough so that A C D 2 and cp- 1 (A) C D 1. Since A is simply connected, there is a holomorphic function G defined in A such that g = Re(G) in A. The composition of the holomorphic maps cpl,r•(�) : cp- 1(A) ---+ A and G : A ---+ <C is holomorphic, and so Re(G o cpl,r•(�)) is harmonic in cp- 1 (A). But for z E cp- 1(A), 'Pl cp -•(�)(z) = cp(z) EA, and so (Go 'Pl,r•(�))(z) = G(cp(z)) = g(cp(z)) =(go cp)(z). So (gocp)l cp -•(�) is harmonic in cp- 1 (A). As z0 Ecp- 1 (A), in particular, we have A(g o cp)(zo) = 0. Since the choice of z0 E D1 was arbitrary, g o <p is harmonic in D1. (2) If h : D2 ---+ JR is harmonic, then by the first part, it follows that h o <p : D1 ---+ JR is harmonic. Now suppose hocp: D1 ---+ JR is harmonic. Then since cp-1 : D 2 ---+ D1 is holomorphic, by the first part, (hocp) ocp- 1 : D 2 ---+ JR is harmonic. But (ho cp) o cp- 1 =ho (cp o cp- 1) =ho (idvJ = h, where idv2 : D2 ---+ D2 is the identity map z Hz (z E D2). So h : D 2 ---+ JR is harmonic. p (= s) A (= -1) Fig. 5.24 0 B (= 1) Triangle inequality in t:..PO' B (3) By the triangle inequality in APO'B shown in Figure 5.24, for s (= P) in IHI, we have Is+ 11 = £(PA) =£(PO')+ £(0'A) =£(PO')+ £(0'B) > f(PB) = Is- 11, where we have used the fact that O' is on the perpendicular bisector of AB to get the third equality. So cp(s) ElI)) for all s EIHI. The function <p is clearly holomorphic: for s EIHI, 1 s = cp'(s) = l· ! +(s-l)· (- : 2 = ) �s\:;); (s 1 ) (s: 1 ) 2" s l 268 A Friendly Approach to Complex Analysis Now consider 'lj; : lI)) ---+ IHI given by 'lj;(s)=��;, zElDl. (This expression for 'lj;, which is a candidate for cp- 1, is obtained by 8 -1 .) solving for s in the equation z = cp(s)= s+l I -l N' ' ' I A 0 1 +3_ 1 B Triangle inequality in b.PO' B Fig. 5.25 = In the Figure 5.25, we know that the angle subtended by the diameter AB at any point of the circle is go o, and so for any point P ( z) in lDl, LAPE> go 0 • So Re('lj;(z))=Re(��;) = l'I/J(z)I cos0 = l'I/J(z)I cos(?T-LAPE)> 0. Thus 'lj;(z) E IHI for all z E lDl. The map 'lj; is holomorphic in lDl: for z ElDl, , 1 1 l-z+l+z = 2 + (l + z) ( ) = 'lj; (z) = l· · 2 2 (1 z) l-z (l-z)2· (1 z) Finally, for s EIHI, we have s-1 1 + -+ + ; s +1 = s 1 s - 1 = 2s = , ('lj; 0 cp)(s)= s 1 s + l-s+l 2 1_ s+l and for z ElDl, we have l+z ---1 1 + z-l + z 2z --------z (cpo'lj;)(z) = l+z+l-z - 2 --+1 l-z So cp is a bijection and cp- 1 ='lj;. �+: Bibliography Beck, M., Marchesi, G., Pixton, D., and Sabalka, L. (2008). A First Course in Complex Analysis, http://math. sf su. edu/beck/papers/complex.pdf Conway, J. (1978). Functions of One Complex Variable I, 2nd Edition. (Springer). Fisher, S. (1999). Complex Variables, 2nd Edition. (Dover). Flanigan, F. (1972) Complex Variables (Dover). Flanigan, F. (1973). Classroom Notes: Some Half-Plane Dirichlet Problems: A Bare Hands Approach. American Mathematical Monthly 80, 1, pp. 59-61. Gelbaum, B. and Olmsted, J. (1964). Counterexamples in Analysis (Dover). Gilman, J., Kra, I. and Rodriguez, R. (2007). Complex Analysis. In the Spirit of Lipman Bers (Springer). Howie, J. (2003). Complex Analysis (Springer). Needham, T. (1997). Visual Complex Analysis (Oxford University Press). Ash, R. and Novinger, W. (2007). Complex Analysis, 2nd Edition. (Dover). Remmert, R. (1991) Theory of Complex Functions (Springer). Rudin, W. (1987). Real and Complex Analysis, 3rd edition. (McGraw-Hill). Shastri, A. (2000). An Introduction to Complex Analysis (Macmillan Publishers India). Shaw, W. (2006). Complex Analysis with MATHEMATICA (Cambridge Univer­ sity Press). Shurman, J. (2012). Course Materials for Mathematics 311: Complex Analysis, http://people.reed.edu/-jerry/311/mats.html. Tall, D. (1970) Functions of a Complex Variable (Dover). Volkovyskff, L., Lunts, G. and Aramanovich, I. (1991). A Collection of Problems on Complex Analysis (Dover). 269 Index concatenation of paths, 71 conjugate, 11 contour integral, 62 convergent series, 106 curve, 15 fl-contractible curve, 82 C, 1 absolute value, 10 absolutely convergent series, 106 Argand plane, 5 d-bar operator, 57 de Moivre's formula, 8 degree of a polynomial, 100 Dirichlet problem, 173 divergent series, 106 domain, 15 biholomorphism, 175 binomial coefficients, 10 Binomial Formula, 180 Binomial Theorem, 10 Bombelli, 4 boundary data, 173 elliptic PDE, 164 entire function, 30 epicycloid, 87 essential singularity, 141 Euler Product Formula, 107 exponential, 18 Cardano, 3 Casorati-Weierstrass Theorem, 151 Cauchy Integral Formula, 91 Cauchy Integral formula, 120 Cauchy Integral Theorem, 77 Cauchy principal value, 157 Cauchy sequence, 14 Cauchy's inequality, 122 Cauchy-Riemann equations, 37 chain rule, 34 classification of zeros, 124 closed path, 76 commutative ring, 128 complete, 14 complex differentiable function, 30 complex exponential, 18 complex number, 1 complex plane, 5 Fibonacci number, 162 field, 2 Fresnel integrals, 159 Fundamental Theorem of Algebra, 99 Fundamental Theorem of Contour Integration, 74 Gaussian integer, 8 harmonic conjugate, 169 harmonic function, 163 holomorphic function, 30 271 272 A Friendly Approach to Complex Analysis homotopic paths, 78 Identity Theorem, 127 imaginary part, 1 improper integral, 157 integral domain, 128 Integration by Parts Formula, 76 isolated singularity, 141 Laplace equation, 163 Laurent series, 131 Liouville's Theorem, 99 Log, 24 logarithm function, 24 Prime Number Theorem, x primitive, 88 principal argument, 24 principal value, 26 radius of convergence, 108 real analytic function, 115 real part, 1 removable singularity, 141 Residue Theorem, 155 Riemann zeta function, 107 Riemann Hypothesis, x, 107 Riemann Mapping Theorem, 175 Riemann zeta function, x root of a polynomial, 100 roots of a complex number, 8 Mobius transformation, 175 Maximum Modulus Theorem, 129 Mean Value Property, 172 Minimum Modulus Theorem, 130 Morera's Theorem, 101 simply connected domain, 82 smooth path, 61 stepwise path, 16 order of a pole, 142 order of a zero of a function, 123 ordered field, 4 Taylor series, 120 triangle inequality, 12 trigonometric functions, 22 partial sums, 106 path, 15 path-connected, 15 Picard's Theorem, 153 Poisson Integral Formula, 174 pole, 141 power series, 107 winding number, 85 zero of a function, 123 zero divisor, 128 zero of a polynomial, 100 zero of a certain order of a function, 123 A FRIENDLY APPROACH TO COMPLEX ANALYSIS The book constitutes a basic, concise, yet rigorous course in complex analysis, for students who have studied calculus in one and several variables, but have not previously been exposed to complex analysis. The textbook should be particularly useful and relevant for undergraduate students in joint programmes with mathematics, as well as engineering students. The aim of the book is to cover the bare bones of the subject with minimal prerequisites. The core content of the book is the three main pillars of complex analysis: the Cauchy-Riemann equations, the Cauchy Integral Theorem, and Taylor and Laurent series expansions. Each section contains several problems, which are not purely drill exercises, but are rather meant to reinforce the fundamental concepts. Detailed solutions to all the exercises appear at the end of the book, making the book ideal also for self-study. There are many figures illustrating the text. World Scientific www.worldscientific.com 9047 SC 1 1111 1 11 1 1 111 1 1111 1 1 ISBN 978-981-4578-99-B(pbk) 9 789814 578998