Introduction to COMPLEX ANALYSIS ROLF NEVANLINNA Academy of Finland Vv. PAATERO University of Helsinki Translated by v. KGvARI and G. Ss. GOODMAN Imperial College of Science and Technology, London A vv ADDISON-WESLEY PUBLISHING COMPANY Reading, Massachusetts - Menlo Park, California - London - Don Mills, OntarioThis volumeis the only authorized English translation of Einfihrung in die Funktionentheorie by Rolf Nevanlinna and V. Paatero published by Birkhauser Verlag, whose copyright notice on the original edition reads © Birkhtuser Verlag Basel, 1964. Copyright © i969 by Rolf Nevanlinna, All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission of the Copyright holder. Printed in Great Britain. Library of Congress Catalog Card No. 69-18119. {9@N 0:201-05265-2 DEFGHIUKLM-CO-79876563FOREWORD The present textbook is based upon lectures given by the authors at Helsinki University and at the University of Ziirich, and is a translation of the German edition, Einfiihrung in die Funktionentheorie, published by Birkhduser Verlag, Basel, in 1964, It is assumed that the reader is acquainted with analytic geometry and the calculus, so that this introduction to the theory of functions may be begun in the third or fourth year of undergraduate study in college. As the Table of Contents indicates, the present volume is limited to the presentation of the elements of the theory of functions, and the authors have attempted to make the material both comprehensible and precise. Among the sections in which this volume deviates more or less from other presentations we must mention the following: the introduction of the complex numbers, the concept of homotopy and its application, the integral theorems, the theory and application of harmonic functions, in particular harmonic measure, and the correspondence of boundaries under conformal mapping. Exercises have been placed at the end of each chapter, and all 320 of these exercises should be solved by the student for better insight into the subject matter, whether he learns the subject through lectures or by self-study. In introducing the elementary functions (Chapters 2-7) we have followed in many places the presentation given by Ernst Lindelof in his Finnish text- book, Johdatus funktioteoriaan (introduction to the theory of functions). This is particularly true for a considerable number of the exercises of these chapters. We have received assistance in our work from various sources. First we owe thanks to Dr. G. 8. Goodman and Dr. T. Kévari for the effort and interest which they have put into the translation of the book. We also express our appreciation to Addison-Wesley Publishing Company and, in particular, to Professor A. J. Lohwater for the valuable advice and generous assistance which he has given in the editing of this edition, Helsinki, September, 1968 Rolf Nevanlinna V. PaateroChapter 2 §L §3 Chapter 2 gl §2 §3 Chapter 3 §1 3 Chapter 4 Chapter 5 $1 2 33 Chapter 6 $1 §2 3 §5 Chapter 7 st §2 CONTENTS The Concept of an Analytic Function The complex numbers Point sets in the complex plane . Functions of a complex variable General Properties of Rational Functions Themthpowr 2. 2... ee ee Polynomials . 2... ee ee Rational functions Linear Transformations Basic properties of linear transformations . Mapping problems ee 8am 8 Stereographic projection. . . . 2. 2 1. 1 1 we Mapping by Rational Functions of Second Order... The Exponential Function and its Inverse. The General Power Definition and basic properties of the exponential function . Mapping by means of the exponential function. The ‘ogarithm 7 The general power ® ey . The Trigonometric Functions Thesineandcosine . . a ‘The tangent and the cotangent oe & 4; . ¥ The mappings given by the functions tanz and cotz. Their inverse functions ‘The mappings given by the functions sinz and couzs The functions aresinz and arccosz . . Survey of the Riemann surfaces of the elementary functions. * Infinite Series with Complex Terms General theorems . a ee Powerseries © 1 1. ee ee ee eee 37 45 52 65 70 B 7B 82 85 87 91 95 100viii CONTENTS Chapter 8 sl §2 8 4 Chapter 9 sl §2 8 $4 §5 86 97 Chapter 10 §l § 8 $5 Chapter 11 st 2 3 $5 86 Chapter 12 a §2 8 Chapter 13 $l §3 Integration in the Complex Domain. Cauchy’s Theorem. ‘Complex line integrals rn The primitivefunction . 2 6. 1 1 ew ee Cauchy’stheorem . . so The general formulation of Caucty’s 's mewren. ‘Cauchy’s Integral Formula and its Applications Cauchy’s formula . "The Taylor expansion of an analytic function - Consequences of Cauchy’s integral formula ‘The Laurent expansion Isolated singularities of an analytic function The inverse of an analytic function . Mapping by a rational function . The Residue Theorem and its Applications The residue theorem. . . ApSTRAtION of the sesldue theotemn to the evaluation of definite integrals oF a ge @ ‘The partial-fraction expansion ofcot7z | 1 1. eC The argument principle . s © 8 Applications of the argument principle. uf Harmonic Functions Preliminary considerations . Lee Gauss's mean-value theorem. The maximum and minimum prin= Ole) se ke tw eee we em eS Poisson'sformula. . . . . 2 ee ee Theharmonicmeasure . . . . 1. The Dirichlet problem Harnack’s principle Analytic Continuation ‘The principle of analytic continuation. . . . The monodromy theorem ee ‘The inverse of a rationalfunction . . ee ee Harmonic continuation. The reflection principle . Entire Functions Infinite products . . Product representation of the function 4 w=singz. The Weierstrass factorization theorem. . . Jensen’s formula. The growth of entire functions . 108 114 17 121 131 135 138 146 149 155 162, 167 169 173 175 77 184 191 195 197 206 213 216 217 219 225 228 231 236Chapter 14 §. a2 93 4 §s §6 §7 §8 9 g10 sll 2 Chapter 15 $1 §2 3 Chapter 16 1 §2 §3 4 §5 Chapter 17 $l §2 $3 4 §5 §6 §7 CONTENTS Periodic Functions Definitions of simply and doubly periodic functions . Reduction of simply periodic functions to the exponential function The basic properties of doubly periodic functions . ‘The Weierstrass p-function. . . . . . ‘The Weierstrass 2- and o-functions . Representation of doubly periodic functions by means of the o-function . # tp sae ‘The differential equation of the function ez). Representation of doubly periodic functions as rational functions of pand p’. « Addition theorem for doubly petiodic functions . Determination of a doubly Beviosic function with prescribed principal parts ms ms Mapping bya doubly periodic ic function of order 2 Ellipticintegrals . 6. 1 1 ew ee ‘The Euler T-Function Definition of the J-function . Stirling’sformula. . . ‘The product representation of the rh function | The Riemann ¢-Function Definition and the Euler product formula. . 2... Integral representation of the {-function . Analytic continuation of the ¢-function Riemann’s functional equation . The zeros of the {-function and the distribution of rime numbers The Theory of Conformal Mapping The Riemann mappingtheorem . . 1. we Construction of the solution . . Boundary correspondence under conformal mapping. . The connection between conformal mapping and the Dirichlet problem * “sw we The conformal mapping of polygons Triangle functions ‘The Picard theorem ix 242 244 246 249 256 259 261 262 264 266 272 278 280 285 289 291 292 295 298 305 308 315 324 327 334 337 343CHAPTER 1 THE CONCEPT OF AN ANALYTIC FUNCTION The theory of functions is concerned with complex-valued functions of a complex variable. Our study is confined to those functions which are differentiable in a sense which will be made precise later on; such functions are known as analytic functions. In order to create a basis for the theory, we begin by introducing the complex numbers in a manner which will lead us naturally to their interpretation as vectors in the plane. §1. THE COMPLEX NUMBERS 1.1. Two-dimensional Vector Spaces We begin by stating the axioms for a two-dimensional vector space over the real numbers. Let there be given a set R, whose elements a, b,...,x, y, ... shall be called points or vectors, satisfying the following conditions. I. To every two elements a, 6 € R there corresponds an element c € R, known as their sum and written c = @ + 5, obeying the following rules: Ll. a+6=b +a (the commutative law). 12. a+ (6+) = (a+b) +c (the associative law). 1.3. There is a zero in R, denoted x = 0, with the property that a+0=aforeveryae R. 14. Theequation a + x=bhas oneand only onesolution,x=b—ae R. Il. To every vector a and every real number A there corresponds a vector b = Age R, known as their product, and obeying the following rules. ILL. A(ua) = Apa @, » real numbers). 112. (+ wa = Ag + ya, Aa + b) = Aa + Ab (the distributive law). 13. la =a. 14. The product Ag vanishes if and only if A = 0, or a = 0, or both A= and a =0.¢ IL.5. The axiom of dimension: there exist two vectors aand bin R which are linearly independent, that is, for which the equation + We shall use the symbol 0 for the number zero as well as for the vector zero without, we trust, provoking any confusion. 12. THE CONCEPT OF AN ANALYTIC FUNCTION §1 da + ub =0 has only the solution A = 2 = 0, but every three vectors a, b, cin Rare linearly dependent, that is, the equation Aa + wb + ve = 0 always has a solution such that at least one of the three numbers A, 2, v does not vanish. This axiom asserts that the dimension of the vector space R is equal to two. In the resulting “affine plane” every vector x admits a representation in terms of its coordinates in a two-dimensional reference system. Such a system is given by a basis for R, that is, by two linearly independent vectors e,, @,€R. From J1.5 it follows that every vector x € R has two numbers ¢, and é, associated with it (its coordinates in this reference system) such that x= Ger + S202. 1.2, Plane Euclidean Geometry Axioms I and II define a two-dimensional vector space whose geometry is the geometry of the affine plane. It becomes a Euclidean geometry once we introduce a (Euclidean) measure of length and angle. We can arrive at such a measure by defining, for any two vectors x and y in R, a scalar product (x, y) with the following properties. TIL.1. (x,y) is a real, symmetric function of its arguments x and y: @&y) =»). 11.2. (x, y) is linear in each argument. IIL3. (x, y) is positive definite, that is, (x, x) 2 0, and equality holds only for x = 0. The dength, norm, or modulus |x| of a vector x is defined by Ix] = +VG, 3). It is easily proved (Exercises 1 and 2){ that the following inequalities hold: 1) Schwarz’s inequality (x, y?? S [x|? |p|? 2) The triangle inequality |x + y| S |x| + |yI- The angle [x, y] between two vectors x, (0) is defined by Gy, x] Ty ‘Two vectors are therefore orthogonal if (x, y) = 0. cos [x, y] = +A function f(x) is said to be linear, if f(Ax) = Af) and f(x: + x2) F(x) +f). The linearity of the scalar product (x, y) asserts, therefore, that this product obeys the distributive law. + Unless there are indications to the contrary, the numbers will always refer to the exercises at the end of the chapter.§1 THE COMPLEX NUMBERS 3 Ife), e2 is a basis for R and if the vectors x, y have the representations x= be, + ben, y= me, + Me» in terms of this basis, then 2 ev) = Eien + Soe, me, + 92€2) = Be Sabie where the g,, denote the real constants Bu = (Cre) (812 = gr). The square of the norm of x is the quadratic form bel? = Gx) = 3B gabiba = gibt + 2ei2bibs + 82083- Tt reduces to the Pythagorean form Ix? = of + if and only if the coordinate system is orthonormal; that is, (1,¢2)= 0, fer] = Jes} = 1 (the Cartesian coordinate system), 1.3. Extension of the Set R to a Vector Algebra In what follows, we shall not introduce a metric into the plane R for the time being, so that we shall be dealing with an affine geometry on R defined by the postulates in groups I and IJ. The problem before us is to see whether it is possible to extend Land II so as to give R the structure of a field (or algebra), and, if this is possible, to discover in how many different ways it can be done. The vector space R becomes an algebra once we are able to define, for any two elements x, y € R, a “product” z=xyeR which satisfies the following axioms. IV.1. The product is commutative: xy = yx. IV.2. The product is bilinear, that is, linear in each factor. IV.3, The product is associative: x(yz) = (xy)z. IV.4. The product xy vanishes, xy = 0, if and only if at least one factor vanishes. Ld. Our task, then, is 10 find all bilinear forms xy © R which satisfy these axioms IV. In order to arrive at the general solution to this problem, we shall assume, at first, that we already have a product xy defined on Rin accordance with axioms TV and see what this tells us.4 "TRE CONCEPT OF AN ANALYTIC FUNCTION §1 If we fix the vector y # 0 in the product z = xy, we obtain a linear transformation in x which maps the plane R into itself. This mapping is one-to-one, for if z, = x,y, 2) = x2), then 2 — 22 = (1 — 2). Since y was assumed to be different from 0, z, — z2 will vanish if and only if X, — X, = 0. Different vectors x therefore have (foreach fixed y # 0) different image vectors z = xy. On the other hand, the range of the mapping z = xy is the whole plane R. For, if x, and x, are two vectors in R, and A, and A, are two arbitrary real numbers, then Quay + Aexady = Amy + Aaxay = Az + Aaza, where z, = x,y, Z,; = x.y. From this we see that the image vectors z,, 2, are linearly independent if and only if the vectors x,, x2 are linearly independent. Hence, if x,, x2 is a basis for R, then z,, z, will also be a basis. If the vector x-has the coordinates 4,, A in the system (x), x), then its image vector has the same coordinates in the system (z,, 22), for z= xy = A,z, + Azz). Hence, the set of image vectors z = xy covers the plane R exactly once if x runs through all values in R (for y fixed). Thus, for any given vector y ¥ 0, there is precisely one vector x which makes the product xy take a prescribed value z; that element is the “quotient” x=2/y. 1.5. Definition of the Unit Vector e If, in particular, we take z = y(#0), then there is a definite vector e = e, € R having the property that e,y = y. We shall show that e, is independent of the choice of y. Let y, and y, be two non-zero vectors. Ife), =¥1, €2¥2=I2 then ya D2 V2 =Y2 = = (ey) => 2Y2= Ye = Ivy ( wy, and this last expression is, by axiom IV.3 (the associative law), equal to €:(¥i¥2/Vi) = e1¥2- Hence, e2¥2 = €1Y2, oF (@; — &2)¥2 = 0, from which it follows that e, = e2, since y, # 0. The element e (#0) defined uniquely by the equation ey=ye=y dl) is called the unit vector, or unit, in R. 1.6, Definition of the Vector i Let a be an arbitrary vector in R and consider the equation waa.8 THE COMPLEX NUMBERS 5 If this equation has a solution x = x,, x} = a, then, for every vector x € R, we have x —a=x?— xP = (x — xe t+ x), so that the equation x? — a =0 has, in addition to x = x,, one further solution x = —x,. Let us choose @ = —e and solve the equation x4e=0. (1.2) The existence of a solution will be shown in an exercise (Exercise 3). We denote the solutions by x = +/ (#0). The vector jis linearly independent of the vector e, for, if i = Ae (A real), then we would have —e = i? = (Ae? = Ae? = Me, or (1 +*)e = 0, which is impossible, since both 1 + A? 4 Oand e#0. The vectors x = e and x = i span the entire plane R. An arbitrary vector x € Rhas the coordinate representation x = e+ ni. This representation has been found under the assumption that there is a product, defined for pairs of vectors x,, x2 € R, which satisfies the axioms IV. If x, and x, are written in terms of coordinates, x= het mi x= he + mi, it follows from IV and the definition of the basis vectors e and i via (1.1) and (1.2) that the product x,x, must have the form 1X2 = (Ere + mi)(Ere + mai) = EE — aemde + Erne + mb) (1.3) The quotient x,/x2 (x. # 0) is defined to be that vector x = ge + yi which, when multiplied by the vector x, = £,¢ + oi, yields the vector x; = Ee + i. With the aid of (1.3), we can obtain the coordinates €, 7 of x from the equations HE-mi=o mb + Sn= 1 Therefore the quotient x,/x2 is given by the expression my Set mt fib: + 2 % hetmi +i 1.7, The Solution of the Extension Problem We now turn all this around and choose any two linearly independent vectors in R, label them e¢ and i, and define the product x,x, of two vectors x, = e + 9 (J =1, 2) by means of Eq. (1.3). We shall then have ex = xe = x and i? + e = 0, and all the axioms IV will be satisfied. The verification of axioms IV.1-3 we leave to the reader. To prove IV.4, we mds — &im , 4 + aad (1.4)6 THE CONCEPT OF AN ANALYTIC FUNCTION $l observe that the equation x,x2 = 0 is equivalent, by (1.3), to the coordinate equations f8— mm = £m + mb, = 0. Squaring and then adding, we obtain Gi + WG + nD = 0. Consequently, ¢; = 7, =0 or €; = 42 = 0, that is, x, = 0 or x, =0 (or X = X2 = 0), as required by axiom IV.4. We have, therefore, completely solved the problem before us: Af the vectors e and i are any two arbitrarily chosen linearly independent vectors, then (1.3) furnishes a definition for the product of two vectors in R which makes R into a field (or algebra, that is, a vector space which satisfies the axioms IV), and this definition of the product is the only one that is compatible with all the axioms. 1.8. Notation for Complex Numbers. Absolute Value and Argument Having made R into a field in which every vector, or complex number, can be written as £,e + £2, we want to say something about notation. Vectors ge (& real) along the e-axis we shall denote, for brevity, by ¢ alone, by dropping the e. In view of the property xe = ex = x which defines the unit, this can hardly lead to confusion. Furthermore, in keeping with a long-standing custom we shall denote the coordinates of a complex number z = ¢,e + €5f = &, + 21 by &, = x, &: = y, and write zoxtiy. The real number x is called the real part of z, and the real number y is called the imaginary part of z. These terms can be abbreviated to x=Rez, y=Imz. We now introduce a Euclidean metric into the “complex plane” R by defining the scalar product (z,, z2) of two complex numbers z, = x, + iy, Zq = X2 + iy2 as (21, 22) = 1X2 + Ya This means that the basis vectors e and / are orthogonal to one another, and that their Iengths are one: [e| = |i] = 1. The modulus or absolute value of a complex number z = x + iy is then given by 2] = +V(, 2) = +Vx? + y?. If we go over to polar coordinates, we get z = r(cos ¢ + isin gd),$1 . ‘THE COMPLEX NUMBERS = 7 where r = |z!, ¢ = arc tan y/x. The quantity ¢ is called the argument of z: argz= ¢ = arc tan y/x. As long as z # 0, ¢ is defined up to a multiple of 27 (we say “modulo 27,” and write “mod 27”). In this notation, the product of two complex numbers % = ry{cos $y + isin dy) (k = 1, 2) 2422 = rra{cos (b, + $2) + isin ($1 + $2)}. From this it follows that The absolute value of the product of two complex numbers is equal to the product of their absolute values, while the argument of the product is equal 10 the sum (mod 27) of the arguments of the factors: lzizal = lal lz], arg 122) = arg 2, + arg z; (mod 27). The latter rule presupposes that the factors are different from zero, since the argument of the number z = 0 is indeterminate. From the product rule it follows that fal teal? arg oT arg z, — arg z, (mod 27). 23 23 If all » factors, z = r(cos ¢ + isin ¢), of a product are equal we obtain [r(cos $ + isin ¢)" = r*(cos nd + isin nd). This yields as a special case, for r = 1, de Moivre’s formula (cos f + isin )" = cos nd + isin nd. (1.5) The numbers x + iy and x ~ iy are said to be complex conjugates. The complex conjugate of the complex number z is denoted by 2; obviously, Rez=3(4+9, Imz= 1-9. Geometrically speaking, the addition of complex numbers corresponds to vector addition (according to the parallelogram rule). The difference 2, — 2, corresponds to a vector whose initial point is at z, and whose end- point is at z;. The modulus of the difference |z, — z,| gives the distance between the points z, and z3. Since the complex numbers form an algebra (axioms I-IV), the rational operations of arithmetic (addition, subtraction, multiplication, and division) obey the same rules as in the real case. Over and beyond this, the defining equation i? = —1 must be taken into account.8 THE CONCEPT OF AN ANALYTIC FUNCTION §2 §2. POINT SETS IN THE COMPLEX PLANE 1,9. Convergent Sequences A sequence of complex numbers 24s Zap eee Zee (1.6) tends to a limit, lim z, = 2, (7) if, to any arbitrarily prescribed number « > 0, a number ne > 0 can be~ Sound such that In-zl r. A set of points {z} in the extended plane |z| < © is said to be open if each of its points is the center of some circular neighborhood which belongs entirely to the set. . An open set of points {z} in the extended plane |z| S « forms a domain if it is possible to join any two points in {z} by a polygonal path which lies entirely in {z}. (This condition makes the open set connected.) Any domain containing the point zp is called a neighborhood of Zp. A point a is called a cluster point (or sometimes, a limit point or accumula- tion point) of a set of complex numbers {z} if every circular neighborhood of a contains at least one point z 4 a of {z}. From this it follows that every neighborhood of a cluster point @ must contain infinitely many points of the set. If a set contains all of its cluster points, the set is said to be closed. The set of points |z| < is open. The extended plane [z| S$ « is both open and closed (Exercise 18). Open sets and closed sets are important particular classes of sets, but an arbitrary set of points is, in general, neither open nor closed. A closed set which cannot be split into two disjoint closed subsets is called a continuum. A set of points {z} is said to be compact if any infinite subset of it has a cluster point belonging to {z}. (A compact set is therefore closed.) The closed plane |z| $ « is compact. The set of points in the plane which do not belong to a given set {z} forms what is called the complement of {z}. The complement of an open set is closed, and the complement of a closed set is open (Exercise 20). + The finite plane can also be extended in other ways. For example, in projective geometry, there is the so-called line at infinity with its infinitely many, infinitely distant points.10 ‘THE CONCEPT OF AN ANALYTIC FUNCTION §2 Let G be a domain. If G does not contain every point of |z| S «, then the points of the complement of G fall into two classes: a) Boundary points of G. These do not belong to G, but are cluster points of G. The set of boundary points forms the boundary of G. b) Exterior points of G. These are points which belong neither to G nor to the boundary of G. This set can be empty. If Fis the boundary of the domain G, then the union G U I (that is, the set of all points which belong either to G or I’) is a closed set (Exercise 21). It is called the closure of the domain G. ‘The union of a domain and its boundary is also called a closed domain... §3. FUNCTIONS OF A COMPLEX VARIABLE 1.11. Definition of 2 Function. Continuity ¢ Functions of a complex variable are defined in the same manner as functions of a real variable: Af to every value z in a domain G there correspondsa definite complex value w, then the mapping f: z —> w is said to be a function defined in the domain G. The number w = f(z) is called the value of the function at the point z. In what follows we shall consider first only those functions which assume finite values in a finite domain, that is, a domain belonging to the finite plane |z| < ©. The real and imaginary parts, u and », of the function f(z) are real functions of the real variables x and y (z = x + iy): =ux,y), 0 = ox, y). Conversely, any two such functions always define one complex function {@a=ut+ wofzx+y. Continuity is defined in the same way as in the real case: A function w = f (2) is continuous at the point z if, to every positive number «, there corresponds a positive number p,, such that [f@+42)-/@|<« whenever [Az] < pe. Geometrically speaking, the continuity of a function w = f(z) at z = zo means this: to an arbitrarily small disk K, centered at wo = f (zo) there corresponds a disk K, about zp with the property that w = f(z) lies in K,, whenever z lies in K,. The limit of a function is defined in the same way as the limit of a sequence in Section 1.9. Everything that was said there about the limit of a sequence of complex numbers applies here as well. Combining the definitions of§3. FUNCTIONS OF A COMPLEX VARIABLE ll continuity and of a limit we can say that a function f(z) is continuous at a point z if lim f(z + 4z) = f(z). 40 The real and imaginary parts of a continuous function are obviously continuous functions themselves, and conversely. The same holds for the absolute value and the argument of a continuous function, provided that the function does not vanish { f(z) # 0). 1.12. Differentiable Functions The derivative of a function of a complex variable is defined in the same way as in the real case. Let f(z) be a function defined in a neighborhood of the point z. If the difference quotient Af f@+ 4) -f@ a: az tends to a finite limit A whenever Az tends to zero, then the number A is called the derivative of f at the point z and is denoted by “)=Ax= lim 2 f'@)= A= lim 5 (1.10) The difference « = Af/dz — A therefore tends to zero with 4z. Hence, Af has the expansion. Af = Adz + Az e = Adz + Ax(42), (1.10) where (dz) denotes a number which tends to zero as Az + 0. If, on the other hand, there exists a constant A such that (1.10)’ is satisfied, then Af/4z = A + (42). Accordingly, the difference quotient 4f/4z tends to the limit 4 as 4z + 0. Hence, A is the derivative f’(z) of the function f@)- Equations (1.10) and (1.10)' are therefore equivalent. The expansion (1.10)’ expresses the fact that the function fis differentiable at the point z. The first term is the differential Y= f'@Az corresponding to the increment 4z. The total increment df of the function J is obtained by adding to the differential df the re- mainder term 4z(4z). AL If f'(z) # 0, this remainder term is negligible ate) in comparison with df as 4z > 0. Therefore, for small values of |z|, the differential df isa good #2) approximation to the increment 4f: the ratio of Figure 112 ‘THE CONCEPT OF AN ANALYTIC FUNCTION §3 the length of the vector p(p) (where p = |Az|) to |df| tends to zero with Az. If, in particular, we choose f(z) = z, then dz = Az. Accordingly, we can write df= f'(z)dz. The derivative f’ is therefore the quotientt of two differentials df and dz: af zB f'®).- (This notation is due to Leibniz.) 1.13. The Cauchy-Riemann Differential Equations We shall now investigate what form the equations that define the derivative (1.10, 1.10’) assume when we separate the variables into their real and imaginary parts: z=x+tiy, f(z) = u(x, y) + to(x, y). Accordingly, dz = Ax + idy and u(x + dx, y + Ay) — u(x, 9) = du, v(x + Ax, y + Ay) — (x, y) = do. We also write A=a+iB, — A2(dz) = p((p), + i(p)a), where (p), and (p); denote real numbers which tend to zero with p= [dz] = Vida + Gy. If the function fis differentiable, then, by (1.10)’, Au + idv = (a + if)(Ax + idy) + p((e), + i(p)2)- Comparison of the real and imaginary parts yields Au = «Ax — PAy + ple), 4v = BAx + ady + ple). These formulas say that u(x, y) and v(x, y), considered as functions of the real variables x and y, are differentiable at the point (x, y). If we set dy =0 (so that now p = [4x|) and then divide by Ax, the first equation yields Au [Ax| Tn 7% + Fy (49 > as Ax >0. The function u(x, y) therefore possesses the partial derivative u, = a. + The differentials dz = 4z and df = f’ dz are finite (and not “‘infinitesimal’””) com- Plex quantities. On the other hand, as we have already seen, the differential df approximates 4f better and better, the smaller the modulus of dz = dz is.§3 FUNCTIONS OF A COMPLEX VARIABLE 13 In a similar way, we can prove that wu, thus have the following result. If the function f (z) = u + iv is differentiable, then the functions u(x, y) and v(x, y) are also differentiable. The partial derivatives are related by the equations B, vy = Band vy =. We WH ty =e (1) These partial differential equations, which relate the real and imaginary parts of a differentiable function, are known as the Cauchy-Riemann differential equations. The converse of this result is also true. If u(x, y) and u(x, y) are differentiable functions of x and y, and if their partial derivatives uz, Uy, Ux, vy satisfy the Cauchy-Riemann differential equations, then the complex function F@) = ux, y) + iol, y) is differentiable with respect to the variable z = x +- iy. Its derivative, f(z), is S'@) = Uz + ty = v, — in. Proof. The assumption that u and v are differentiable at the point (x, y) means that they have the expansions Au = u,dx + u,Ay + plp),, dv v,Ax + v,dy + pl) (1-12) Here, the quantities (p), and (p), vanish when p = V(4x)? + (dy)? tends to 0. If we multiply the second equation in (1.12) by /and add it to the first equation, we get Af = Au + idv = (uz + ivJAx + (uy + JAY + lp), where (p) = (p), + i(p)2 > 0 as p > 0. Since, by hypothesis, the Cauchy- Riemann equations (1.11) are satisfied, we can write Af = (uz + iv,)Ax + (vz + iu,Ay + plp) = (uz + in,)Ax + iy) + plp). But, according to (1.10)’, this means that f(z) is differentiable and has the derivative f(z) = tz + iv, = vy — iy. Remark, If a function u(x, y) of two real variables is differentiable at a point &y): Au = aAx + BAy + plp) ((p) +0 as p= V(dx)? + (dy)? > 0), then, as we pointed out above, it possesses partial derivatives u, = « and u, = B at this point. We know from the theory of real functions of several variables that, conversely, the mere existence of the derivatives u, and u, at the point (x, y) in no way implies the differentiability of the function u(x, y) at this point. If, however, we make the additional hypothesis that the partial14 THE CONCEPT OF AN ANALYTIC FUNCTION 8 derivatives wu, and u, also exist in a neighborhood of the point (x, y) and are continuous at this point, then one can conclude that w is differentiable at the point. (We use the mean-value theorem; cf. Exercises 27 and 28.) 1.14, The Definition of an Analytic Function Iff (2) is defined ina finite domain G, and is differentiable in z at each point of G, then f (2) is said to be an analytic function in G. A function is said to be analytic at a point if it is analytic in a neighborhood of the point. The definition of an analytic function in the domain G says that at every point of G this function possesses a finite derivative (2) = tim 2 FO= jim ie" . This definition of an analytic function was given by the founder of the theory of complex functions, Augustin Cauchy (1789-1857). From what we have just discussed, it is equivalent to the following definition, given by Bernhard Riemann (1826-1866). A function f(z) = u(x, y) + ive, y) is analytic in a domain G if the Functions u(x, y) and v(x, y) are differentiable throughout G and the Cauchy- Riemann differential equations Uy = Dy My = Dy are satisfied. Later on, we shall show that still more follows from these assumptions, namely, that an analytic function is continuously differentiable, (that is, it possesses a continuous derivative), and the same holds for its real and imaginary parts. In their original definition of an analytic function, the founders of complex function theory, Cauchy and Riemann, required the continuity of the derivatives. That this property already follows from the differentiability assumption was proved only at a later date (Edouard Goursat, 1900). We shall see, in fact, that the existence of ai! the higher derivatives follows from the analyticity of a function. Despite its apparent simplicity, then, the definition of an analytic function constitutes an enormous restric- tion compared with the general definition of a complex function. The real and imaginary parts u and v of an analytic function have con- tinuous partial derivatives of all orders, as we shall prove Jater on. From the Cauchy-Riemann differential equations (1.11) it follows that u and v satisfy Laplace’s equation —@U eu 4U= ae t ye 0 (cf. Exercise 32). Functions of this sort are called harmonic functions.§3 FUNCTIONS OF A COMPLEX VARIABLE 15 Two harmonic functions which are related by the Cauchy-Riemann equations are said to be harmonic conjugates. The real part and the imaginary part of an analytic function are therefore conjugate harmonic functions, 1.15, The Rules of Differentiation The derivative of a complex function has been defined in precisely the same way as the corresponding notion in the case of a function of a single real variable. The definition is based upon the rational operations of arithmetic and the notion of limit, both of which remain unchanged when we go over from the real domain to the complex domain. Accordingly, all the rules of differential calculus (for example, for the derivative of a sum, or a product, or a quotient) remain intact and can be applied, without further ado, to complex functions. The same also holds for the composition of functions. Suppose that the function w = w(z) is analytic in the domain G, of the z-plane and that its values lie in a domain G,, of the w-plane. If, further, £ = {(w) is an analytic function of w in the domain G,,, then the composite function z + w > $= U(w@)) =f is analytic in G,. To prove this, it is only necessary to show that fis a differentiable function of z. We leave this to the reader as an exercise (Exercise 29), At the same time, we note that the chain rule for the derivative of a composite function holds in the complex domain: a dl dw a lO" eae 1.16. Conformal Mapping by Analytic Functions Suppose that the function w = w(z) is continuous in a neighborhood of the point zp and is differentiable at z. Suppose, further, that w'(z9) # 0. alf the point z moves along the ray arg (z — 29) = ¢ (© const.), then its image w(z) = w(zp) + Jw moves along a well-defined curve yy which starts at Wo = w(Z) and is such that Aw n , Fig = W Go) + (Az) = w'@o)[l + 2), where <(Az) = (Az)/w’(zo) > Oas Az -> 0. Therefore, as 4z > 0, we have |Aw| ' Aw , re > |wGdb are Ge = arg. dw ~ p> argw'G). (1.13) From the second formula in (1.13) it follows that yg possesses a tangent at the point wo which makes an angle f= $ + arg w'(Z) with the real axis.16 THE CONCEPT OF AN ANALYTIC FUNCTION §3 If we take two rays, arg (z — zo) = 4, (v = 1, 2), then the angle between the tangents to their image curves at the point w, turns out to be t2—th=b2- dr Under the mapping from the z-plane into the w-plane this angle therefore remains fixed. If the angle of inclination of the ray in the z-plane increases, then the angle of inclination of the tangent to the image curve also increases: the orientation is preserved. From the first formula in (1.13), we conclude that the ratio of corre- sponding distances |4w| and |4z| tends to the same limit |w’(zo)| indepen- dently of the direction chosen for the vector Az. Geometrically speaking, this means that the mapping of the z-plane into the w-plane resembles a similarity mapping in a neighborhood centered at zo, and the resemblance becomes stronger as the size of the neighborhood shrinks. Ls [> On the basis of these considerations we make the following definition: If the function w = w(z) has a non-vanishing derivative at the point zo, then the mapping z > w is called conformal at the point zo. This definition is equivalent to the existence of the following (finite) limits: jim, aga = (a) [Aw 2 Taz] #0, (b) where Aw = w(zp + Az) — w(Zp). 1.17. In order to enlarge upon the foregoing considerations, we shall separate the complex function w(z) into its real and imaginary parts. We shall assume that the real and imaginary parts of the function w = w(z) = ux, y) + Fol, »)§3 FUNCTIONS OF A COMPLEX VARIABLE 17 are differentiable at the point zp, and that the functional determinant is positive, Gu, v) a, y) = tb, — U0, > 0 (1.14) at the point zo. If, in addition, one of the two limits (a) and (b) mentioned in Section 1.16 exists, then w(z) is differentiable at the point zp and w’(zp) # 0. To prove this, we shall first assume the existence of the limit (a). If we write Az = dz = dx + idy # Oand Aw = Au + ido, then lim are 4 dv dx — dudy . A240 Az du dx + dv dy since, by (1.14), du and dv do not both vanish, This limit and therefore also the expression dv dx — dudy _v, dx? + (v, — u,) dx dy — u, dy’ du dx do dy uz dx? + (Uy + 0,) dx dy + v, dy? = arctan is independent of dx and dy, From this it follows that the coefficients of the quadratic forms occurring in the numerator and denominator are proportional, By My ty =A = AY, + 0D, where A depends only upon zp. Consequently, (1 + WY, — 4.) = 0 and so Uy = Vy Uy = “Dy. The Cauchy-Riemann equations therefore hold, and the existence of w'(Zo) # 0 follows from (1.14) and Section 1.13. We shall now prove the same result from the existence of the limit (b) instead of (a). (The hypotheses made at the beginning of this section are still assumed to be in force.) The expression im (Al? _ diet + do? ar+0 [dz ax? + dp = HEF 2D dx? + Atty + v5v,) de dy + (U5 + 05) dy? Get a (#0) is independent of dx and dy. From this it follows, as in the first case, that tity H0x0,=0, += 2+ v2>0, (1.15)18 ‘THE CONCEPT OF AN ANALYTIC FUNCTION 3 and, consequently, & ome ort, 2, 2, t, 3 wy oH bz = [Dp ae flys From the second equation in (1.15), we obtain 2 = +]. Because of assumption (1.14), 4 = 1, and the Cauchy-Riemann equations hold. From this we conclude, as above, that the derivative w'(z9) 4 0 exists. Remark. According to the foregoing proof, it follows from the existence of the limit (b) that » = +1 or » = —I, depending upon whether the functional determinant u,v, ~ uyr, is positive or negative. In the latter case, the mapping z — wis indirectly conformal at the point 29 (Exercise 39). EXERCISES ON CHAPTER 1 1, Prove Schwarz’s inequality (x, y)? $ |x|? |y|? (Section 1.2). Under what conditions does equality hold? Hint. |dx + py|? = Qoe + py, Ax + py) = A(x, x) + 2An%, y) + Uy, 9) is positive definite with respect to A and p (or it vanishes identically). 2. Prove the triangle inequality (Section 1.2): Weal — [eal] S ls + 22) S lai] + l22]- Under what conditions does equality hold? Generalize the inequality on the right to n complex numbers. 3. Prove, with the aid of axioms IV, Section 1.3, that the equation wte=0 (1.2) possesses a solution. Here e is that vector for which ex = xe = x holds for all values of x. Hint. Take as a basis for the x-plane the vector e and a vector /, linearly independent of e. Let xamet+of, y=Pethf LP=netynt be the coordinate representation of two arbitrary vectors x and y and of the vector f?. Then xy = aPre + (iB + eB) S + 22Bof? = (Br + yieaBade + (4182 + 2B, + yroPo) f Since xy = 0 only if x = 0 or y = 0, the system of equations ayBy + yie2B2 = 0, m8 + (oy + y2%2)82 = 0EXERCISES ON CHAPTER | 19 has only the solutions «, = «, = 0 or 8, = B, = 0. Such will be the case if and only if the determinant of the coefficients of 8, and f,, a(t y20%2) — yied = of + yoo cee — 7103 AO for all a? + «3 > 0. This quadratic form in «, and a, is therefore definite, and the discriminant of the form is positive: Gy, + v7) > 0. Now solve Eq. (1.2), setting x = ae + @,f The equation Stes (al t yyod + le + Qayay + y208) f= 0 is equivalent to the system of equations a+ yet '=0, ae + 7202) = 0, which possesses two pairs of real roots: Ya 2 e; oe 0) = Fee ‘ ay, + 7) a Vn FAD 4, Show that the product x,x2 defined in Section 1.7 satisfies the axioms IV. 1-3. 5. What is the geometric interpretation of the multiplication of complex numbers? 6. Find the real and imaginary parts of the quotient of the complex numbers 2, =X, + I, 22 = Xz t iy. by multiplying top and bottom by the complex conjugate 2, of z2. 7. Prove that the determinant of the real and imaginary parts of the complex numbers a and b is (ab — ab)/2i. 8. The value of a rational expression goes over into its complex conjugate when every complex number occurring in the expression is replaced by its complex conjugate. (Show this.) Deduce from this the theorem: If the coefficients of the algebraic equation agz" + az""! + +++ +a,=0 are real, then the complex roots of the equation are conjugate in pairs. 9, Prove that (z = x + iy) lel + byl 5 va Si! When does equality hold? 10, Prove that S |x| + [>1- if a and 6 are complex numbers with |a| = 1 or |6| = 1.20. THE CONCEPT OF AN ANALYTIC FUNCTION 11. Prove that if [a] < 1 and |b] <1. 12, Prove that § + = 8) + Ss for |z| = t. 13, Derive the formulas ley + za)? = |z,[? + [za]? + 2 Re iz), le, ~ zal? = [zuP? + [zp]? — 2 Re (2,22). The second one includes the Law of Cosines. 14, Separate the following expressions into real and imaginary parts (z is complex): 1 z-i iz? > = z-i zi in}, 15. Explain the geometrical significance of the absolute value and the argu- ment of the function (z —- i)/(z + i) and investigate how they vary as z varies. 16. Prove the following theorem: lim z, =z #0 holds if and only if the conditions Jim |z,| = |z] and lim arg z, = arg z (mod 27) are fulfilled, 17, Prove the Bolzano-Weierstrass Theorem: Every bounded infinite set of points has a cluster point. 18. Show that the set of points |z| S © is both open and closed. 19. Give an example of a set of points in the plane which is neither open nor closed. 20. Show that in the complex plane (a) the complement of an open set is closed; (b) the complement of a closed set is open. 21. Prove that (a) the boundary of a domain is a closed set; (b) the union GU Fof a domain G with its boundary Tis a closed set. 22. Prove that the union A U B of two closed sets 4 and Bis closed.EXERCISES ON CHAPTER | 21 23. Prove that the intersection 4 © B of two closed sets A and B (that is, the set of points common to A and 8) is closed. 24. Prove that (a) the union A U B and (b) the intersection 4. B of two open sets A and B are open. 25. Prove that the union of arbitrarily many open sets is open. 26. Prove the Heine-Borel Theorem: If a bounded, closed set A is covered by a set C of open disks (that is, A is contained in the union of the disks in C), then there is a finite number of disks in C which cover A, 27. Prove that if a real function u(x, y) of two real variables possesses partial derivatives which are continuous at a point, then the function is differentiable at this point. 28. Show by means of the example u = x?y/(x? + y) that the continuity of a function and the existence of partial derivatives is not sufficient for the differentiability of the function. 29. Prove the chain rule for the differentiation of a composite function. 30. Verify that the real and imaginary parts of the functions of z given in Exercise 14 satisfy the Cauchy-Riemann equations. 31. By use of polar coordinates, investigate the real and imaginary parts of z" and 1/z*, with particular rerefence to their sign. 32. Prove that the real and imaginary parts of an analytic function w(z) = u(x, y) + i o(x, y) satisfy Laplace's equation provided u and » possess continuous partial derivatives of the second order. 33. Prove that, in a disk on which the derivative of a complex function vanishes identically, the function is constant. 34, Show that every polynomial of the first degree a + bx + cy with real coefficients is the real part of an analytic function of z (= x + iy), and construct this function. 35. What are the most general polynomials, with real coefficients, of second degree U(x, y) =a + bx + cy + dx? + exy + fy, VQ, y) =a + bx + cy +d'x* + e'xy + fy, for which U(x, y) + i V(x, y) is an analytic function of z = x + iy? Show that the function in question is a polynomial of second degree in z. 36. What is the most general polynomial, with real coefficients, of the form ax + 3bx2y + 3exy? + dy* which is the real part of an analytic function? Construct this function.22. THE CONCEPT OF AN ANALYTIC FUNCTION 37. Show that the form of the Cauchy-Riemann equations and of Laplace’s equation is preserved when the coordinate system x, y is replaced by another rectangular coordinate system whose axes are in the same position relative to one another as were the original coordinate axes. 38. Show that in polar coordinates the Cauchy-Riemann equations have the form a or rap’ or rag and Laplace’s equation has the form eu 120 1 @U wT ort age 39, The real and imaginary parts of a function w(z) = u(x, y) + i o(x, y) are differentiable at a certain point zp and x, v)/0(x, y) = u,v, — u,v, < 0. Prove that if the limit (b) in Section 1.16 exists, the mapping z > w = u— iv is conformal at the point zo and the given mapping z — w is indirectly conformal (conformal with the sense of the angles reversed). 40. Suppose that the function w(z) is analytic in a domain G which is symmetric with respect to the real axis. Show that f(z) = w(Z) is then an analytic function ofzin G.CHAPTER 2 GENERAL PROPERTIES OF RATIONAL FUNCTIONS §1. THE 2-TH POWER 2.1, Continuity and Differentiability The function w = z" (na positive integer) is a continuous function of z. For every value of z it possesses the derivative dwidz=nz""". Proof. The increment 4w in the function w when z changes by dz is, by the binomial formula, Aw = (2 + Azl = 2 = ned + G ) 2242) 4 +++ + (dz), from which it follows that 4] < mlzlt]4z] + 6 Te Prada)? +e [zi We see that |Aw| < ¢ whenever |4z| is sufficiently small. The n-th power is therefore a continuous function of z. For the difference quotient of the function, we obtain Aw _ nz () 2A ye 4 (Az! Az and, therefore, () 2 This yields, for the derivative, the formula dw dz dz 2.2. The Binomial Equation We want to find out what values of z satisfy the equation Z=a#0. (2.1) ae sane 2 Az] +--+ + [Az nz! Introducing the notation z=r(cosd+ising), a= p(cosy + isin py), 2324 GENERAL PROPERTIES OF RATIONAL FUNCTIONS §1 we can write (2.1) in the form r'(cos nd + isin nd) = p(cos % + isin #). This equation is fulfilled if and only if rap, mpap+k2n (k=0,41,42,.... (2.2) From Eq. (2.2) it follows that Those values of ¢ which differ by multiples of 27, so that the difference of the corresponding values of k is divisible by n, give rise to the same value of z. The function z” therefore assumes every value a # 0 at n different points, namely, - Vp [coe (¢ + 0) + isin (é+«%)| (k=0,1,..42—1). (23) As a > 0, the values of z given by (2.3) all tend to zero, and for z = 0 they all coincide. At this point, wOO)=-0 W=1,2,...2-1, wMO) 40, zi We say that the function w = z" possesses a zero of n-th order (or n-tuple zero) at the origin. 2.3. The Mapping by Means of the Power Function The a-th power, w=2 (n22), (2.4) defines a function which is analytic in the entire z-plane. Its derivative, w' = nz™1, does not vanish for z # 0, so that by Chapter 1, (2.4) gives a mapping of the z-plane into the w-plane which is conformal whenever z # 0. From (2.4) we have |w| = lz", argw = margz. (2.5) Thus circles |z| = const. in the z-plane correspond to circles |w| = const. in the w-plane. Every angle with vertex at the origin is multiplied by n. The mapping is therefore not conformal at the origin z = 0, where w’ = 0. If we let z make one circuit around the circle |z| = r, its image point w will go around the circle [w| = r" n times. Thus a different values of z, of the form (2.3), correspond to each value of w = p(cos ¥ + isin #) # 0. These values of z may also be written as nome =01,.-.2-0, 26)§l THE n-TH POWER 25 where 3 4 isin? 20 = Vp (cos © + isin and 2a, an €, = cos — + isin — n n is a root of the equation z” = 1. To every ray y emanating from the origin in the z-plane there corres- ponds a ray from the origin in the w-plane which makes an angle with the positive real axis that is # times as large as the corresponding angle made by y. When y is rotated through the angle 27/n, its image ray turns through an angle of 27. Thus, to the sector Fo: 0S angzs there corresponds the entire w-plane. To every point on the positive real axis in the w-plane there corresponds one point on each of the two rays that bound the sector Fy; otherwise, the correspondence is one-to-one. We can obtain a correspondence which is everywhere one-to-one by slitting the w-plane along the positive real axis from the origin to infinity and associating the “upper edge” (arg w = 0) of the slit with the positive real axis in the z-plane and the “lower edge” (arg w = 27) with the ray arg z = 2x/n. If the ray y is rotated once more through the angle 27/n, its image ray turns through the whole angle 2z all over again. As in the previous case, the sector F: 2 < are2 g 2 n n is therefore mapped one-to-one onto the entire slit plane. Two points zp and z, in the sectors Fo and F;, respectively, are mapped onto the same point in the w-plane whenever z, = zp¢,. This procedure shows that the function (2.4) maps every sector Fy 7 sags sk+ = (& =0,1,. -1) in a one-to-one manner onto the whole plane slit along the positive real axis. To every point in the z-plane there corresponds a definite point in the w-plane. Conversely, to every point w # 0 there correspond precisely 2 different points in the z-plane. The inverse of the function (2.4), z= Vw awl, 2.2 has therefore » different values or branches: it is an n-valued function.26 GENERAL PROPERTIES OF RATIONAL FUNCTIONS §l 2.4. The Riemann Surface for the a-th Power We can extend the mapping given by the n-th power w =z" by defining w= 0 = lim z* (z -> «) as the image of z= 0. In order to make the mapping between the extended z-plane and the extended w-plane one-to-one, we now slit the w-plane along the positive real axis, make x replicas of it, and associate with each replica of the slit plane a different sector F, (k = 0, 1,...,2— 1). We then bind these planes together along the edges of the slits and form a connected surface R,, in such a way that R,, will be completely zplane woplane Figure 3 swept out by the image of a ray argz = const. when this ray makes one complete turn about the origin. To achieve this, we join the “lower” edge of the first sheet to the “upper” edge of the second sheet, the lower edge of the second sheet to the upper edge of the third, etc., until, finally, we join the lower edge of the -th sheet to the upper edge of the first sheet. Thus there arises a closed, n-sheeted surface R,,, which is known as a Riemann surface (cf. Fig. 3). The function (2.4) maps the extended z-plane in a one-to-one manner onto the m-sheeted Riemann surface R,,. Its inverse function (2.7) is therefore single-valued on this surface. Each point w # 0 appears on every sheet of the surface, and therefore 7 times in all, but the point w = 0 appears only once on the surface because all the sheets are joined together there. If we go once around the point w = 0, we always proceed to a new sheet. Only after n revolutions do we return to our starting point. If we make k revolutions about the origin, where Aargw=k-2z, many-sheeted surface comes from Riemann.§1 ‘THE 7TH POWER ar then the image point in the z-plane moves, according to (2.5), in such a way that Qa Aare =k. The point z therefore returns to its starting point only when k is divisible by m5 otherwise z ends up in some other sector Fo, . . ., F,-; of the z-plane, The n sheets of the surface are joined together at the points w = 0 and w = «, Such points are called branch points of (n — 1)-st order of the Riemann surface Ry. The Riemann surface R,, is also said to be a covering surface of the plane |w| S$ ©, This important topological notion is defined in the following way. Let G, and G,, be two domains in the complex planes |z| < © and |w| S ©, respectively. G, is called an unbranched covering surface of G, if the following conditions are fulfilled. 1) There is a mapping which associates to each point z € G, a unique point w € G,,. 2) Whenever this mapping takes a point z =a into a point w = 6, it maps a certain neighborhood U, of z = a topologically (one-to-one and continuously in both directions) onto a certain neighborhood U, of w = 6. 3) Every point w ¢ G,, is the image of at least one point z € G,. The mapping furnished by the power w = 2” satisfies these properties for the domains G,: 0 < |z| < «, G,: 0 <|w| <. Under the mapping the latter domain is covered precisely n times. The covering surface G, is un- branched relative to the “underlying surface” G,,. If we add the branch points z=w=0 and z= w= «, G, is called a branched covering surface of the w-plane with branch points at 0 and «. The many-sheeted surface R,, defined above is therefore a relatively branched covering surface of |w| < 0. R, can be made to a topological map of the plane G, (|z| S ~) by taking the n-sheeted circular domains |w| R® that corre- spond under the mapping w = z" to the circular neighborhoods |z| < r and |z| > R, respectively, as the “‘circular neighborhoods” of the branch points w=0, 0. R, as well as G, are therefore both coverings of Gy (|w| S «). The surface R,, has been introduced ‘‘in space” to aid our geometrical intui- tion. But it is just as easy to grasp the structure of this surface by means of the domain G, when, as above, it is divided up into “sheets” F, (k = 0, ..., a — 1) each of which corresponds to a complete replica of the underlying surface G,. This fact should be borne in mind when, later on, we construct Riemann surfaces which make the inverses of other single-valued mappings z —> w single-valued.28 GENERAL PROPERTIES OF RATIONAL FUNCTIONS $1 2.5. The Cyclic Group of the z-th Roots The points at which the function (2.4) assumes the same value w can be found, according to Eq. (2.6), by taking any one of them, say z, and making the substitution S=zk (k= 0,41, +2, These substitutions form what is called a group. A group is defined in the following way. (2.8) A set R of elements a, 5, c, . . . forms a group if the following postulates hold: 1) To every ordered pair of elements (a, 4) in R there is associated an element c in R, written c = ab. 2) This operation is associative: a(bc) = (abe. 3) There exists an element ¢ in R, called the identity, such that, for every a € R, the equation ae=ea=a holds. 4) Every element a € R possesses an inverse, a~!, in R, satisfying aga'=a'a=e, In the set (2.8) we take as elements the substitutions S themselves, and we define as the group product $\5; of Sj=ze and = S) = ze the result of composing them: SySp = zeBelt = zelth, which is again a substitution. The conditions (1) and (2} are obviously fulfilled. If we set e=zl=z, then (3) is also satisfied. Finally, the substitution inverse to S = ze is So = ze* = ze, for SS = SS = zeqtet =z. Since et = ¢# for every integer k, the number of different substitutions is finite, in fact equal to 2, Such a group is said to be finite and of order n.§2 POLYNOMIALS 29 By repeated application of the substitution S = ze, we obtain all the substitutions of the group (2.8). If we set S? = SS, etc., the different sub- stitutions of the group can be written as S° =e, S,S%,..., S57". The elements of the group are thus powers of a single element 5 in the group. Such a group is said to be cyclic. The group (2.8) (more generally, any cyclic group) is commutative, or abelian, that is, S\S; = S25, is valid for all substitutions S,, S, in the group. §2. POLYNOMIALS 2.6. Zeros A polynomial of degree 7 is an expression of the form P,(Z) = a2” + az"! 4+ +++ + ay (2.9) The coefficients are given complex numbers ap, a, ..., Gn, With ay # 0. It follows from Section 2.1 that P,(z) is continuous and, at every finite value of Z, possesses a well-defined derivative which can be computed according to the tule familiar from the real case. Replacing z by x + iy in (2.9) and expanding (x + iyy-* for each value of k (k = 0,1,...,— 1) according to the binomial formula, we obtain a decomposition of P,(z) into real and imaginary parts: PZ) = Uylx, y) + VAC y)- The functions U,(x, y) and V,(x, y) are polynomials in x and y of degree n. They are, according to Chapter 1 (Section 1.14), harmonic functions and satisfy the Cauchy-Riemann differential equations. We want to find out what values a polynomial will assume when z varies. This question can be resolved with the help of the Fundamental Theorem of Algebra: A polynomial P,(2) (n & 1) vanishes for at least one value of z. We shall defer the proof of this theorem until later (Section 9.13), The following theorem also holds: If the polynomial P,{2) vanishes for z = z,, then P,@) = @ — 21) Pai@), (2.10) where P,_,(z) is a polynomial of degree n — 1. Indeed, if we divide P,(z) by (z — z,) until the remainder is independent of z, we obtain the identity Pilz) = (@ — 21)Pas@) + R. Setting z = z, yields R = 0, since P,(z,) = 0.30 GENERAL PROPERTIES OF RATIONAL FUNCTIONS §2 If we apply successively the Fundamental Theorem of Algebra and the theorem just proved, we find that Pz) = aglz — 2)(z — 22)... (@ — 2), where the z,, Z2,..., Z, are, apart from the ordering, uniquely determined complex numbers. The polynomial P,(z) vanishes for the values z = z,, 2, .. ., Zn, and only for these, The values z, need not all be distinct. If v, values of the z, are equal tO 2, v2 equal to z,.. ., m% equal to z (z;, .. ., 2, all distinct), then P,(Z) = ag(z — 24)"(z — 2)... (2 — 2)", 3 y= me If P{z) has the representation P@)=(-2O@2) 2), where Q(z) is a polynomial and Q(zo) # 0, then z = 2g is said to be a zero of Eth order of the polynomial P(z). The polynomial P,(z) above therefore has a zero of »,-th order at each point z, (i = 1, 2,.. ., ). From the result just obtained, it follows that an algebraic equation of the a-th degree possesses exactly 2 roots (when each root is counted according to its multiplicity or order). We speak of an n-th order c-point zo of a polynomial P(z) when zp is an n-th order zero of the polynomial P(z) — c. From this it follows that zhe polynomial P,(z) assumes every finite value precisely n times. 2.7, The Behavior at Infinity We now want to investigate the behavior of a polynomial P,(z) when |z| > . If we write ry Pr P,(z) (a8 +25 +3 then ol Z| |P.Z)| = +. QA1 lag + 4 By io Zz 2 Since the second factor tends to the limit |ao| # 0 when |z{ > =, we have lim [PG =». [z|r00 Moreover, it follows from (2.11) that§3 RATIONAL FUNCTIONS 31 We therefore say that the polynomial P,(z) possesses a pole of n-th order at infinity. Another way of expressing this is to say that the polynomial assumes the value «© exactly times, §3. RATIONAL FUNCTIONS 2.8. Zeros and Poles A rational function is obtained whenever we apply the rational operations of arithmetic finitely many times to the variable z and to certain given complex numbers, We can write the result as P@) RE) GG)’ (2.12) where P(z) and Q(z) are polynomials P(2) = age” + a2") ++ + Q(z) = bo" + bz") 4 ++ + by (a, 69 + 0) without common factors. The real and imaginary parts of the function R(z) are rational functions of x and y. From Section 1.15 it follows that a rational function R(z) is differentiable whenever Q(z) # 0; the rule for differentiation is the same as in the real case. 2.9, If we factor P(z) and Q(z) so that P(Z) = aoe — ay)P(Z — a2)? «(2 — a), Q@) = bolz ~ Bz — fo)"*... (2 — Bi)" where a # 0, bp # O and a # 8, for all i, j, we obtain for R(z) the repre- sentation (2 = aE — a)... — oye RG) = KSA Ma Oa 2.13 OK GIR By = B" en) with K = do/bo. The normal form (2.13) of the rational function R(z) places in evidence all of its finite zeros and poles. From (2.13), RE) = ~ «)"*Ri), where R,(z) is a rational function which, at the point z = «, assumes a finite, non-zero value. We thus designate the point «, as a zero of order y, of the function R(z). Similarly, a2, a3, ..., a, are zeros of orders p22, [235 ++ +s Pay respectively, of the function R(z).32. GENERAL PROPERTIES OF RATIONAL FUNCTIONS, §3 Now let us write (2.13) in the form where R,(z) is again a rational function which, at the point z = §,, has a finite; non-zero value. Therefore |R(z)| increases beyond all bounds when z -> f, but in such a way that (z — 8,)"'R(z) tends to a finite, non-zero limit. For this reason we say that R(z) possesses a pole of order v, at the point f,. Similarly, 8 is a pole of order v2, etc. At every point z which is different from ay, a2)... ap, By, Boy - function R(z) assumes a finite, non-zero value. -» Br the 2.10. The Order of the Function R(z) Let us consider the behavior of the function R(z) when |z| increases beyond all bounds, We write R(z) in the form There are three different cases to treat, depending upon. the relative magnitudes of m and a. 1) Form > an, lim R(z) = ~. Since bk R(2) ay jim oe by * 0, «, is a pole of order mi ~ n of R(z). 2) Form =n, lim R(2) = 2 z rn ’0 The function therefore has a finite, non-zero value at z= «. 3) Form n m = m a m—n m m=n m _ m=n n — nem m G ( ): (2.18; ©) ~ Gala) +B Gn (=, 2.18) The difference R(z) — F(z) (2.19) is a rational function, It is finite at every finite point z # 8, (A = 1,2,...,0). In order to study the difference at the point 8,, we write it as Ya Seley) By (1), the expression in brackets remains finite for z = B,; the same holds for the remaining terms as well. The expression (2.19) therefore has a finite value even at 8,. In a similar way we can conclude that the difference is finite at every point 8,. Therefore (2.19) is a polynomial in z. Since the difference can also be written in the form {rey — Gy, (; RG) — Fle) = (RE) ~ Gadd) — YG (=): ant 2 it vanishes at infinity. The polynomial (2.19) therefore has no pole at infinity either. It is therefore a constant, whose value, from the above, must be zero. Thus our assertion is proved. The rational function (2.12) can therefore be written in the form : 1 RG) = Gale) + Z Gy (4) , (2.20) and this representation in partial fractions is unique (see Exercise 6). EXERCISES ON CHAPTER 2 1. Split the function w = z? into real and imaginary parts and investigate: (a) how straight lines in the z-plane are mapped into the w-plane; (b) how straight lines in the w-plane are mapped into the z-plane.36 © GENERAL PROPERTIES OF RATIONAL FUNCTIONS Hint. Consider first the lines parallel to the coordinate axes. Ifa configuration in the z-plane is rotated about the origin, what happens to its image in the w-plane? 2. Which curves in the z-plane correspond to the circles |w — 6] = g under the mapping w = z?? Check the dependence of these curves on the value of [dl/p. Solution. The curves are lemniscates |z — a||z + a| = p witha = Vb. 3. What values does the function Vz possess at the point z =i? Write these values in trigonometric and in algebraic form. In what sequence do these values come up, when, starting at z = 1 with the function value 1, z describes the unit circle once in the negative direction (that is, in the direction in which arg z decreases)? 4. Derive the rule for differentiation of a rational function. 5. Decompose the functions L +i? ZT’ into partial fractions. 6. Show that the partial fraction decomposition (2.20) of a rational function is unique, 7. If the order of a rational function is p, then, by Section 2.11, the function assumes a given value c at p points, some of which may coincide for certain values of ¢, For which values of c does this happen? 1 xz + Iz + 2P andCHAPTER 3 LINEAR TRANSFORMATIONS §1. BASIC PROPERTIES OF LINEAR TRANSFORMATIONS 3.1. The Group of Linear Transformations The term linear (or bilinear) transformation is used to designate a rational function w = w(z) of first order. Its general form is b w= S@) = 2s. Gy where the determinant of the coefficients is assumed to be different from zero: ad — be #0. (3.2) Were this determinant zero, w would be either indeterminate (@ = b =e = d = 0) or constant (and therefore a rational function of order zero), The linear transformations satisfy the group axioms (1)-(4) given in Section 2.5. In particular, the identity transformation J (w = z) plays the role of the identity in the group, and the inverse S~' of the transformation (3.2) is the transformation ~—dw + o -1 Sy) The collection (3.1) therefore forms a group under composition. The linear transformation (3.1) furnishes a one-to-one mapping of the extended z-plane onto the extended w-plane (Section 1.9). In particular, it places in correspondence the points qa zea, wes and z=-%, woo, € c The mapping is conformal at all other points in the plane, for when z # —d/c, dw ad—be & (a+ap*” 3,2, The Invariance of the Cross-Ratio Let z, (#, —d/c) be an arbitrary point of the z-plane and let w, be its image: a+b cz +d" 37 ws38 LINEAR TRANSFORMATIONS §1 Then A similar formula holds for any other pair of associated points z,, W2 (z, # ~,—d/c). Forming their quotient, we get Yok “ (3.3) where A is the constant Equation (3.3) holds for all finite pairs z, w = (az + by/(ez +d), If z3, 3 is a third pair of corresponding points, then MoM Bo 7, Wy W223 Zp Elimination of \ between this equation and (3.3) converts the transformation 3.1) into the following form: Qw = w/w = w2) _ (2 — aE — 22) Gs — willows — wa)” 3 — 2s — 22)" Conversely, (3.4) defines a linear transformation which makes correspond to any three distinct given values z = 2,, Z2, Z; any three arbitrarily prescribed values w,, W2, W3. On each side of formula (3.4) appears what is known as the cross-ratio of four points «, B, y, 5: (3.4) _@= lle 8) (By = gg 5)" The equation (0%, Ws, Way Wa) = Gs 2222) (ay asserts that the cross-ratio of four points is invariant under a linear transforma- tion. Up to now we have assumed that the points z,, w, (v = 1, 2, 3) are different from «. However, formula (3.4)' also holds when any one of these values is infinite. For example, if w; = ©, then we have =; 0, 0, Wy, Wa) = (25 my Ws) = = It is easy to see that even in this case (3.4)' determines a linear transformation which carries the points z = z, into the points w = w, (v = 1, 2, 3).$l BASIC PROPERTIES OF LINEAR TRANSFORMATIONS 39 3.3. The Steiner Circles Let us choose two arbitrary points z = z,, z, (#4, —d/c) and denote their images under (3.1) by w = w,, w2, respectively. The transformation (3.1) can then be put into the form (3.3), where the constant A is non-zero. Con- sequently, w— wil yy eal wow jay l2— Zl, 3.3) hw] ~ Ml = 2 3) _— en . a8 pw, At te (mod 27). (3.3) Let p be a positive number. Then leral_ lz = 29] is the equation of a circle known from elementary geometry as the circle of Apollonius: it is the locus of all points z for which the ratio of the distances from the “limiting points” z, and z, is constant. As z moves around this circle, the point w, according to (3.3)’, must move on a circle with the limiting points w, and w2. The ratio of the distances from w to these limiting points is |Alp = const. The expression $ = arg z-% oe arg (z — 21) — arg (z — z2) gives the angle between the vectors z — z, and z — z, (Fig. 4). When z moves from z, to z. along a circular arc «, ¢ remains constant and equals the angle o between the tangent to this arc at z, and the vector z; — z,, According to Eq. (3.3)’, ww = b= are, ~ Pot ated is also constant. Therefore the image point w also moves along a circular arc 8 which makes the angle ¢o + arg A with the vector w, — w . If z now moves along the complementary arc of the arc «, 6 = ¢9 + 7 and, consequently, yb = bo + arg + 7. Hence, walso moves along the complementary arc of B. The circle through the points z, and z, therefore corresponds to a circle through the points w, and w,. We shall now construct all the “Steiner circles” which belong to the limiting points z, and z,. There are two kinds of circles to be distinguished. The first kind consists of the Apollonius circles (circles abour the limiting points) which correspond to different values of p; the.second kind consists of the circles through the limiting points which correspond to different values of ¢ (circles through the limiting points). From the foregoing, we see that a40 LINEAR TRANSFORMATIONS §h linear transformation (3.1) maps this system of Steiner circles onto a similar system in the w-plane having limiting points w, and w,. Under this trans- formation circles of the first kind are mapped onto circles of the first kind, and circles of the second kind are mapped onto circles of the second kind. From elementary geometry it is known (and, in any case, easy to prove) that through every point in the plane there passes exactly one circle of the first kind and one of the second kind in the Steiner system belonging to any two arbitrary limiting points. The two kinds of circles intersect orthogonally. Figure 4 This result continues to hold when one or the other of the limiting points (z; or 22, w; Or wz) moves off to infinity. If, for example, z, 4 ©, z, = ~, then the “circles” of the second kind become straight lines through the point z, and the circles of the first kind become concentric circles about this point. For this reason, straight lines and circles are not distinguished in the theory of linear transformations: by a “‘circle” is understood either a (finite) circle or a straight line. 3.4. The Mapping of Circles We now can prove easily: A linear transformation is a circular collineation: it maps circles onto circles.§1 BASIC PROPERTIES OF LINEAR TRANSFORMATIONS 41 Proof. Let C, be an arbitrary circle in the z-plane. We select two points z, and z, on it and denote their image points under the mapping (3.1) by », and w. C, can then be thought of as a circle of the second kind through the limiting points z, and z,. Accordingly, it will be mapped by (3.1) onto a circle of the second kind through the limiting points w, and w2, which proves our assertion. A variety of other important consequences can be drawn from the above- mentioned properties of the Steiner circles, Let C be a Steiner circle of the first kind with limiting points z, and z2. The straight line through z, and z, cuts the circle C in two diametrically Figure 5 opposite points, z= a, z= b. By the definition of the circle C, the cross- ratio (a — 2)(a - 22) (@= 2) = 2)" es Such a sequence of points a, b, z1, z2 is said to be harmonic. The points z, and z, are called symmetric points with respect to the circle C. It is easy to show that z, and z, are polar with respect to the circle C; this means that z, (or z,) is the mid-point of the chord which joins the points of contact with the circle of the tangents drawn from the point z, (z;, respectively). From (3.5) it follows that r,r, = p*, where r, and r, are the distances of the points z, and z, from the center of the circle C and p is the radius of C. In particular, if C is a straight line, then C is the perpendicular bisector of the segment 2,22, and the reflected points z, and z, are symmetric with respect to C. Now we shall prove: Ifa linear transformation maps a circle C, onto a circle Cy, and if z and z* are symmetric points with respect to C,, then their image points w and w* are symmetric with respect to C,,. Indeed, C, is a Steiner circle of the first kind with limiting points z and z*. By Section 3.3, C,, is then also a Steiner circle of the first kind with limiting (@, b, 2, 22) =42 LINEAR TRANSFORMATIONS §1 points w and w*. Therefore, w and w* are symmetric points with respect to Cy. 3.5. Fixed Points We consider now the fixed points of a linear transformation (3,1); that is, the points z which coincide with their image points under the mapping. Such points must satisfy the condition aft+b feed’ or cP +d-ati—-b=0. G6) If the discriminant 4 =(a— dy + 4be = (a + d)? — A(ad — be) (3.7) of the quadratic equation (3.6) does not vanish, then (3.6) has two distinct roots ¢ = £,, f,, which are finite when c # 0. If, however, c = 0, then one of the roots is = 2. In the following, we shall suppose that 4 # 0, and consider first the case c # 0, so that there are two distinct finite fixed points, {, and ¢,. From Section 3.3 it follows that the system of Steiner circles with limiting points z, = ¢, and z, = ¢, remains fixed under the transformation (3.1), with circles of the first kind going over into circles of the first kind, and circles of the second kind going over into circles of the second kind. Figure 6§1 BASIC PROPERTIES OF LINEAR TRANSFORMATIONS, 43 If 4d #0, but c=0, then by (3.2), 4 = (@— d)’, so that a # d and neither @ nor d vanishes. The fixed points are b Legate and =o, and the mapping of the Steiner net is the same as in the case ¢ # 0. The circles of the first kind are concentric circles about ¢,, while the circles of the second kind are straight lines through £,. The transformation (3.1) assumes the form w—% =A %), (3.8) with A= a/d. This is a similarity (or homothetic) transformation with the fixed point £, as center. 3.6. Elliptic and Hyperbolic Transformations If 4 #4 Oandc # 0, then, by Section 3.2, the transformation is of the form woh yeh, wl z-% where A # Ois a finite constant. If, in particular, |A] = 1, then it follows from (3.9) (cf. Section 3.3) that every circle of the first kind in the Steiner system with limiting points ¢, and {, is left invariant. Under the mapping, then, the points z “flow” along these invariant circles in such a way that circles of the second kind pass over into one another. Such a mapping is said to be elliptic. Similarly, one can speak of an elliptic transformation when 4 # 0, c=0. Now, the second fixed point is at infinity, and the transformation is of the form (3.8), with || = |a/d| = 1, that is, lal = |d]. The transformation (3.8) represents a rotation (through the angle arg A) about the fixed point £,. The value A = —1 gives rise to an elliptic transformation of a particular kind, If we set z; = w, = ©), 22 = wz = 2 in (3.3)”, then the circles of the second kind are also invariant: the arcs bounded by the limiting points are merely interchanged. This elliptic transformation is involutory; that is, if the point z is mapped into the point w, then w is mapped into z, as we can easily see from (3.9) (A = —1). Now suppose that A is real and positive. Comparing the arguments of the left and right sides of (3.9) shows that, in the Steiner system belonging to the limiting points ¢, and £,, the circles of first kind go over into one another, while the circles of second kind remain fixed. The transformation therefore represents a “flow” along the circle of second kind. A transformation of this kind is said to be hyperbolic. Incase d 4 0,c = 0, €, equals infinity ; the transformation is of the form (3.8). It is hyperbolic when arg \ = arg a — arg d = 0 (mod 2m), The map- ping is homothetic; ¢, is the center and A is the quotient or scale factor. G9)44 LINEAR TRANSFORMATIONS § The general mapping (3.9) can always be thought of as the result of composing an elliptic transformation with a hyperbolic transformation. Let atgA = $. We first effect the elliptic transformation in which A is replaced by A, =cos d + isin ¢ (|A,| = 1), and then the hyperbolic transformation with the A-value A, = |A]. The resulting general linear transformation has two different fixed points and is called Joxodromic. 3.7. Parabolic Transformations We now turn to the case d =0. The transformation (3.1) is then called parabolic. . First let us assume that c = 0. Then, by (3.7), a = d # 0, and the linear transformation (3.1) is thus of the form weztw (w=d/d¢ 0). (3.10) If w = 0 (that is, 6 = 0), this is the identity transformation. Every point is a fixed point. If w # 0, (3.10) is a simple translation (parallel displacement) with the “displacement vector” «w. The only fixed point is { = «0. The straight lines L, in the direction w are the streamlines; their perpendiculars, the lines L,, pass over into one another under the flow. It remains for us to investigate the case A = 0, c # 0. Equation (3.6) then yields one finite fixed point r-424 2" This is the only fixed point of the parabolic mapping, since the point z = goes over into the point w = a/c, so that « is not a fixed point. We now obtain +d clt+d cl+d cz+d’ and if we set cz + d= e(z — £) + cf + din the last expression, we obtain +d (ch+ay 1 ad—be* ad—be z—f Here at _(a+dp_4_(a+dy o+d=2S8, ad~be= SF ST, and so (hed | ad—be Thus, the transformation assumes the form§2 MAPPING PROBLEMS 45 where es Se Cag be ata (Oe). Tf we set ry. 1 1 Wee Bee 1D then (3.1) reduces to the form Wea Zl be, (3.12) This is a parallel displacement of the z’-plane. Under this mapping all straight lines L; in the direction w’ are left invariant. The lines L; which cut them orthogonally go over into one another, being displaced by an amount equal to the magnitude of w’. We can return to the original variables (z and w) by use of the inverse transformation of (3.11), z = ¢ + 1/z’. The lines L; thereupon go into circles L, which aré mutually tangent at the point z = w = @, but have no other points in common. Their common tangent is the line z = { + 1/(7w’), which is the image of the straight line z’ = vw’ (7 a real parameter). The perpen- diculars L; go over into circles I, which are also tangent to one another at z= Cand cut the circles L, orthogonally. Under a parabolic transformation the circles L, appear as invariant “streamlines”; the orthogonal family of circles L, remains unchanged as a whole, but individual circles are trans- formed into other circles of the family. The parabolic transformations arise as the limiting case of the elliptic and hyperbolic transformations when the fixed points coincide; the systems of circles L, and L, can thereupon be regarded as the limiting case of the Steiner system when the limiting points ¢, and ¢, coincide. From (3,11), it follows that for = 0 gas (3.13) This transformation is called an inversion. Since 1 z|=>- arg z’ = —arg z, Fl=a rg we can obtain (3.13) by a reflection in the real axis followed by a further reflection in the unit circle (cf. Section 3.4.). Each of these two transformations is itself indirectly conformal. §2. MAPPING PROBLEMS 3.8. The Conformal Mapping of Two Circular Domains onto one Another By a circular domain K we mean a domain whose boundary consists of a circle C, so that X is either the interior or the exterior of C. If, in particular, Cis a straight line, then X is one of the two half-planes bounded by C.46 LINEAR TRANSFORMATIONS §2 We pose the following problem: Map a given circular domain K, of the z-plane conformally onto a given circular domain K,, of the w-plane. Solution 1. We choose three arbitrary points z,, z2, Z; on the periphery of , and three arbitrary points w,, w2, w; on the periphery of K,, and ordered in the same sense as 2,, Z2,23.¢ The equation (Ww, W3, W1, W2) = (Z, Z3, 21, Za), or (w= ww = 2) _ (@ = AVE — a) (ws — wiles — 2) (@s — 21)/@s — 22)" defines a linear transformation of z into w which carries the points z, into the points w, (v = 1, 2, 3). Since a linear transformation maps circles onto circles, the circle C, which passes through the points z,, z,, , corresponds to the circle C, through w,, 2, w;, Each circular domain bounded by C, corresponds to a circular domain bounded by C,,; it is the ordering of the points z, and w, which forces K, to be mapped onto K,. From this we see that the mapping problem can be solved when three pairs of boundary points (z,, w,) (v = 1, 2, 3) are arbitrarily chosen, so long as they have the same sense. The mapping is then uniquely determined by the choice of these boundary points. Solution 2. We can also construct the mapping of K, onto K,, so that two arbitrary points 2) and wo in the interiors of K, and K,,, respectively, corre- spond. According to §1, if such a linear mapping exists at all, it carries the symmetric point zj of Zp relative to the circle C, onto the symmetric point wé of w relative to C,. It is therefore of the form W— Wo z—2 wwe z— zy (constant). (3.14) As we know, such a mapping transforms the Steiner circles about the limiting points zo, 2% into the Steiner circles about the limiting points wo, w%. The circles + If Kis the interior of C, and therefore a finite circular domain, the point z traverses the circle C in the positive sense with respect to K if the function arg (z — a), for a an arbitrary interior point of K, increases. It is easy to show that this property is independent of the choice of the point a. On the other hand, if K is the exterior of C, the definition of the positive sense is reversed, so that the boundary point z moves in the negative sense with respect to the interior of C. If, in particular, C is a straight line, so that K is a half-plane, the positive direction on C with respect to K is defined as above: it is the direction in which arg (z — a) increases when a is any interior point of K. Intuitively, this means the following: if we set up the coordinate axes in the z(= x + iy}planc in the usual way (that is, so that the positive y-axis when turned in a clockwise sense through 90° will coincide with the positive x-axis), then the domain Kremains to the left of the boundary when a point traverses the boundary in the positive sense (cf. Exercise 15).§2 MAPPING PROBLEMS 47 C, and C,, themselves belong to these pencils, however, and there exists one value of A which forces (3.14) to map C, onto C,,. We determine this value of A in the following way. Let z, and w, be arbitrary points on the circles C, and C,,, respectively. If \ is calculated from the equation Wi Wo 21 = 20 , wow AD oe (3.14) and substituted into (3.14), we can see that the transformation (3.14) maps the circle C, onto the circle C,. w, is thereby the image of z,, and the mapping problem is solved. This course of reasoning shows that we can find a conformal mapping of the domain K, onto the domain K,, in which we can choose two pairs of corre- sponding points: a pair of interior points (29, w)) and a pair of boundary points (2), ,). Once they are picked, the linear mapping (3.14) is uniquely determined. One may ask whether there exist analytic functions w = w(z) other than linear transformations which perform this conformal mapping, under the side conditions mentioned aboye. Later (Section 9.16) we shall show that there are no others. The linear transformation determined above is therefore the only solution of the mapping problem. 3.9. The Conformal Mapping of the Unit Disk onto Itself We shall now consider in more detail the mapping problem posed in Section 3.8, for the special case in which K, and K,, are the unit disks |z| S 1,|w| S 1. Using the second method of solution, we choose two interior points 2 (0 < |Z9| < 1) and wo (0 < |wo| < 1) which are to correspond to one another. The transformation we are seeking is then of the form (3.14), where the constant A is still to be determined. In the calculation of the symmetric points of zo and wo we observe that |zo| |zp| = 1, arg zg = arg Zp, so that (3.45) Hence, the circles |z| = 1 and || = 1 correspond if and only if the constant 7 is unimodular (that is, has modulus 1).48 LINEAR TRANSFORMATIONS §2. Formula (3.15) therefore embraces all the linear transformations which map the unit disk onto itself and make the interior points zo and wy correspond, 3.10. If, in particular, we choose wy = 0, then _. z-% we" T Tae defines a mapping of the disk |z/ < 1 onto the disk |w] < 1, which carries the point z, to the origin. We can also write this transformation in the form y= az+b_ bz+ a’ here a, 6 are arbitrary complex numbers with |5| < Ja]. Formula (3.16) arises when we set 7 = a/d and z, = ~b/a. It therefore embraces (under the condi- tion || < [a|) all (directly) conformal mappings of the unit disk onto itself. Let us compute the fixed points of this transformation. We obtain these whenz=w={: G16) be ~(a-@L-b=0. If we put a = « + iB, then a — 4 = 2i8 ( real), and the fixed points are therefore + Vib)? ~ 6 hia 5 For |b] = || (# 0) there is only one fixed point, £, = f, = ¢ = if/b, which lies on the circumference of the unit disk: |{] = 1. ‘The mapping is parabolic. For [4] > |8| there are two distinct fixed points, £ = £,, £2, which both lie on the unit circle: 2 2 _ ge ne =e-F +e The mapping is hyperbolic. For || < ||, finally, the fixed points are bia= Gs VB= TDD. Both points lie upon the same straight line through the origin and are symmetric with respect to the unit circle. The mapping is elliptic. In the parabolic mapping, the circles which go through the fixed point and which are tangent to the unit circle (the so-called oricycles) are invariant, while, in the hyperbolic case, the Steiner circles of second kind through the fixed points (the hypercycles) and, in the elliptic case, the circles of first kind