COMPLEX ANALYSIS
:
: :.:.::
... ":
......
•<
; ...
.:
.
.f .. : ....~: ...
.
,,:".1
.:~:(..:.:::::,:. .. :.: ~ . .....
.
 ........ ...
..:...
,. ... :.....
l.
~ .. .~,. .
,... .. , ....
:1..
':.
:
...:~.... ~.
..: ..
..•i .~ , ~::...
·1· ...
~.~. !
.~;.~r::; . :..:•~.: .
r,
....
":
. ..·:1: ~ ....: :: . ..:._....
• • •• ":..:.;.::
::.
_.
. :.~ .:.;' +.. ~.~.~.:.:.~.: ~:.... : •. : . ~. .. ...
... .....
.:.! ..: ;.:: .
;:~:: i:;\:~~~:.,.~y: .: .. ,~.. :: .. .:.. .; ,
..
:
.
..
..
:~
.
International Series in Pure and Applied Mathematics
G. Spri.nJer and E. H.. SjMllier, Con..ulting Editors
COMPLEX ANALYSIS
An Introduction to the Theory of Analytic Functions of One Complex Variable Third Edition
L13i\~:~:. ...,~.:;"....A r '\·. .
Q~~~LR1)j
Complex Analysis Bentkr and Orszag: Advanced Mathematical Methods for Scientists and Engineers Buck: Advanced Calculus B'U8tJCker and Saaty: Finite Graphs and Networks
Cheney: Chester:
AhlJ1YI8:
Introduction to Approximation Theory Techniques in Partial Differential Equations Coddington and Levin30n: Theory of Ordinary Differential Equations Ctmte and de Boor: Elementary Numerical Analysi~: An Algorithmic Approach Dennemeyer: Introduction to Partial Differential Equations and Boundary Value Problems Dettman: Mathematical Methods in Physics and Engineering Golomb and Shank3: Elements of Ordinary Differential Equations Hamming: Numerical Methods for Scientists and Engineers Hildebrand.: Introduction to Numerical Analysis HfJ'USWkoider: The Numerical Treatment of a Single Nonlinear Equation Kalman, FalbJ and Arbib: Topics in Mathematieal Systems Theory Lass: . Vector and Tensor Analysis McCarty: Topology: An Introduction with Applications to Topological Groups M onJc: Introduction to Set Theory Moore: Elements of Linear Algebra and Matrix Theory Mour8Ufld and Duri3: Elementary Theory and Application of Numerical Analysis Pearl: Matrix Theory and Finite Mathematics PiPM and Harvill: Applied Mathematics for Engineers and Physicists Ralsttm and Rabinowitz: A First Course in Numerical Analysis RitgeT and Rose: Differential Equations with Applications Rudin: Principles of Mathematical Analysis ShapirQ: Introduction to Abstract Algebra Simmons: DiJlerentiai Equations with Applications and Historical Notes Simm0n8: Introduction to Topology and Modern Analysis Struble: Nonlinea.r Differential Equations
..
ito
Professor of Mat ematiCS, Emeritus Harvard University
< ~.. . .j
Lars V. Ahl'!'S
"":.
.....
.,
•.
,
. . ,'i". .~
~
::~·:McGrawRlD,Inc. ~. .:New York St .. Louis
Lisbon
<~ ...
San Francisco London Madrid.
\,.y
: ..
New Delhi San Juan Tokyo Toronto
.
slniaPore
•• I' •
Auckland Bogota ~cO City .Milan
" ::I
..
• " r
.
v
. v.'
, :
.....
•
.. . :~ .
.(
'
.'
•
. :_.:.. . .,: :.}:~~ ~~'/..
.
~ ',' '.
'.
~. ":', ~ :'
. ..
:...
•
~(_,
. . ~.
:.;S!
••
~~·~?"I'· ~.:~.
'
.... ~..~ ',;.::., .. ,. .:
J _
~.:.:
•
'::
", '1 ~
••
~.:::.
• ._:
'.' ". I
:
'.~ ~
"'1
•• '
.' ..'.
....
••
.. . '_".'
•
.....
.:.
•• •• _.'
.'
~:.' '.
.:
I
.'.
...
'.:
• ::~.,•
.'
:.  ;.
'.
_.'
•
•
•

.'
:

._~.
..
.......
. ': ~ ::.~?:./.:.: ,:':,:~~~~: :~.:~, .. ~ .:.: ;:.:: . '.~::~.': .:... , ..: .: ".,.~.~::,.~: :
..
.
.
.::~~::=~~..' ::..:....~.·;.'i ..',.....:.: .
To Ems
.~ . .~
(
COMPLEX ANAL VIIS
Copyrip.t @ 19711,1966 by McGrawHili, Inc. All richu f'Oklved. Copyript 1963 by NeGrawHiII, Ine, An rich" reaerved .. Printed in 'United 8~tee of AInerica. No pari of this publication may be reproduced. stored in a retrieval system" or transmit.ted, in any form or by any IIIe&Il8, elecVonic. mechanical. pbotooopyiDa, recordiD& or oiberwile, without die prior written permiasion of the publilher.
u.e
16 17 18 19 20 21 BRBBRB ThiI book wu"
96 9 8 7 6 5 4 3 2 1 0
The editors were Carol Napier and Stephen W&«Iey; the production auperviaor wu Joe Campanella.
in Modem SA by Monotype Comptlllition Company, Inc.
u.....,. 01CoaP' •• Cat.......
Ablfon, Lan Valerian, date Complex analyfJiL
in PuhUoatJon Data
Ineluct. index.
(International aeries in pure and applied matheJnatics)
&
1. AnaIytie functiolll.. I Title. QA331.A4S 1879 515'.93 7817078 ISBN Q.074J006671
".
~: : . : ,t,» .•..
..:.'.... . J' _~: "".::.
~.
: .!
!
.....
Xlll
.
The Binomial Equation Analytic Geometry .. ... Conjugation..
ER 1
COMPLEX NUMBERS
1
1 3 4 6
9
~
The Algebra oj Compla: Numbe.
:i
I
I
t
\. ·1~ Justification
1
.:The Spherical Representation
Representation oj Cornples Numbers Geometric Addition and Multiplication
12 12
17 18
15
'I
2
COMPLEX FUNcnONS
21 21
22 24
28
30 33
_ .
ii .
33
.
." .
".
• ~
v:
. ..I. ~
. Absolute Value 4 ..I
2C
•
I
.:.•.":1:
"
••
.
' ••••
I..
. 1.
.
. IIUII.n.
.13 SquareRoots i.I
i
Contents
.~.
..
.
I
.
!
..e Geometric
!
. .
. ..~ Inequalities lJ....
..1 Arithmetic Operations :.
~... ~ ~

.
.:"~
I
I
I
.
.~..1.
.
..
2 The Argument Principle 5..''.t.:i~".4 Length and Area. Zeros and Poles The LocaI Mapping The Maximum Principle
.:.
":".¥II
CONTENTS
CONTENTS
2.rtla.6 Locally Exact Differentials 4. {.3 Connectedness 1.
...3 Uniform Convergence 2.1 The Index of a Point with Respect to a Closed Curve 2.
::
..3 3.:.3 3..
.i :
:
I
3
Linear Transformations
3.3 Line Integrals as Functions of Arcs 1.'~a..5 Abel'B Limit Theorem
.2 The Integral Formula 2~ Higher Derivatives
114
3
The Exponential ond Trigonometric Functions
3.:" _ _'
" •
.. .!
.
.
I
1 Fundamental Theorems
.4
of Analytical Functions
Taylorfs Theorem
46
49 50 50
CHAPTER 3
~ .>?:.":::":::.' ...:.''::~!.3 Homology 4.3 EValuation of Definite Integrala
The Calculus oj Residues 5..1 Chains and Cycles 4.
.'i:..:£<.1 The Exponential 3.1 The Residue Theorem
148
152 154
162
148
.
... ...5
3.
I
I
97
.3 The Periodicity ·
3...GHAPTER 5
.
.'.\.'i' )/·~:.
144 146
142
5
69 73 75 76
5.. POUJer Serle.2 3. 1 3.
.:.4 3.> ':
...1 Line Integrals 1.2 Metric Spaces 1.4 The Logarithm
120
124 124 126 130 133
137
114 118
'I
I
I
43
44
3 Local Properties
·3 . 1.\.1 Bets and Elements 1.2 6..4 Power Series 2.
ANALYTIC FUNCTIONS AS MAPPINGS
Removable Singularities.'..:.
!
35
38
41 42
42
. <\ :
..:.5 Cauchy's Theorem in a Diak
104 105 109
112
101
Partial~ .1 Weieratrass's Theorem .2 The Trigonometric Functions 3.:· ~:_::...5 Definition and Basic Properties The Meanvalue Property Poisson's Formula Sehwarz's Theorem The Reflection Principle
The Cross Ratio Symmetry Oriented Circles Families of Circles
78
83
76
80
84
89
165 166 168 172
175 175 175 179
162
4
Elementary Conformal Mappings
4....
Cauchy's Integral Formula
2.
.:.6 Topological Spaces
./. ..1 The Use of Level Curves 4.2 Analytic Functions in Regions 2.4 Cauchy's Theorem for a Rectangle 1. ./.
I I
i
1 Elementary Point Set Topology 1.
The General Form oj CauChY~8Theorem
4.2 A Survey of Elementary Mappings 4.. Espansioru
r . Formula
193 198 201
..2 Rectifiable Arcs 1.3 Elementary Riemann Surfaces
89
93
. Multiply Connected Regions 7
I·
I
51
54 59 63 66 67 67
141 141
138
137
2 !..~.2 Simple Connectivity 4.1 Arcs and Closed Curves 2..
I
.3 The Laurent Series
I . ~ Infinite PfOduets
Canonieal Products
The Gamma Function Btiriing'.5 Continuous Functions 1. ... ~~.4 Compactness 1.
. I: ~..1 The Linear Group
i"
Harmonic Functions
6.. ..3 Conformal Mapping 2.
!
Conformolity
2.~"..2 3. Froctiona and Factorisation
~
184
187
187
1..I
I
:
CHAPTER 4
COMPLEX INTEGRATION
101
101
~ "rt~
1.4 The General Statement of Cauchy's Theorem 4.
191
. 12 The Taylor Series :.".3 6~4 6.5 Proof of Cauchy's Theorem 4.
. 1 6..r • :
SERIES AND PRODUCT DEVELOPMENTS
_ .
3 The DiJJerential Equation 3.
~..::.
~..
t:~=:. 2 Algebraic Functioru 2.2 Solution of Dirichlet's Problem
245 24li 248
.
. 1
265 265 266 268
270 272 272
273
5.tial Equations
I:_'
~.
...::..
.:.4 The Functional Equation The Zeros of the Zeta Function
:!
.1 Representation by Exponentials 1.~
".:~
7...... :.s
~
_:.' ..
.:/"'
..
4
The Riemann Zeta Function 4. : .
. DIRICHLET'S PROBLEM
229 229 229 232 233
CHAPTER 8
GLOBAL ANAL rnc FUNCTIONS
The Riemann Mapping Theorem
1..
'..
i:
I...
I: i
I
CONTENTS
CONTENTS
"...... .
!
214
216 218 219 219
220
2 Doubly Periodic Functions 2.:.
283
283
<
!
I
1
I:
i
1
. ~.. ' ••
.::"1'
:1····:.s Point of View
Harmonie Measures
Green's Function Para.
...2
206
CHAPTER 7
ELLIPTIC FUNCTIONS
263
263 263 264 264
Jensen' a Formula Hadamard's Theorem
207 208 212
213
1 Simply Periodic Functions
1...
I
3 A Closer Look ot Harmonie Functions
atl 3.
I.·.4 The Canonical Basis General Properties of Elliptic Functions
5 Normal Families
5~1 Equicontinuity
.
307
308 309 311 313 315 318 323
'
.4 1~
:1. The Monodromy Theorem 6 1.4 Families of Analytic Functions 5..Christoffel Formula
The Behavior at an Angle
Germs and Sheaves Sections and Riemann Surfaces Analytic Continuations along Arcs Homotopic Curves 1."'..Jr '..
2.'
..
5.
.. .3 Behavior at the Critical Points
301
306
304
.
:
~~.··:~·.3 Solutions at Infinity
II::
Regular Singular Points
'The Hypergeometric Differential Equation Riemann.3 2.1 The Resultant of Two Polynomials 2.'
".4
.
'
~
.~(t.':....1
251
252 257 259
. 
..
_'. r..
I
..
I..
:.':.Fourier Development 1. .::.5 The Classical Definition
223 225
.".1
.2 Extension of r(B) to the Whole Plane
4.. _.1
of Polygons
235 235 236 238
1.
~":'1'
'::'... .2 Normality and Compactness
Arzela's Theorem 5.
.. r~: .'
':'
.. _.1 The Period Module 2.
5 Canonical Mappings oj Multiply Connected Regions
5.
. .4.
.
.
I:: .2 The Schwarz.. .
. ... .7 Branch Points
284
287
289
291
295
WI
2.
2..(1")
5..~
..' ..
: Ii
. ..
....:
r"_.~~..
..~ ..
• ::'.2 Functions with the Meanvalue Property Hamaek's Principle
241
242
243
..1 The Product Development 4.1
I:
I'i!: .~ ~ineGT Differen...3
222
3
The Weie.3 4.
!
5.. .. ..:"
'.
Picortre Theorem
3..
~":
:.: .
.....':.
... :~
' '
••

'
•••
:.
:
_. .
".
. T..
'.
••
'
••
•
I

•
_.5 The Conformal Mapping by >.2 Definition and Properties of Algebraic Functions
3CKJ
300
2~4 The Triangle Functions of Schwarz
241
.. ... :'. :: :
• • •• ..:
.
275
277
279 283
CHAPTER 6
1
CONFORMAL MAPPING.3 Mapping on a Rectangle
.::. '.
.:. .
.
•• .
3 Entire Functions
3..
.4 Analytie Arcs
1 Analytic Continuation
11 The Weierstrass Theory ..':::"
'
: .2 Unimodular Transformations
2.'.3
:I .
.. '1 Ordinary Points
1
I.
• .1
Lacunary Values
4
The Dirichlet Problem
4.
".2 The.: :.
'._(T) 3.1 Subharmonic Functions 4.2 1~ 1...
.3 Use of the Reflection Principle 1.
1
.: J.2 Boundary Behavior 1.
.
I~:"r" ..
..>. ..
:::':
.
"
.
.
:..1 Statement and Proof 1.:'~~~~:'. • •
4:.3 Functions of Finite Order
I
.
•
.'::.i
284
I'
i
Z Conformal Mapping
2.. '.. .2
..
. :.4 The Modular Function ).
4. . .. •
I.
'.
•••
.
!
.strass Theory
it 1 The Weierstrass pfunction 3.
..2
. •• I
..' '..i I"..".1 3..2 The Functions r(z) and q(z) 3..
..
:.llel Slit Regions
I
.
.. 1..
Complex Analy8i8 has successfully maintained its place as the standard elementary text on functions of one complex variable." . There is. :!
..._". who has kindly permitted ..... The author still ~believes strongly in a geometric approach to the basics. ~ mapping has been added.. To some degree this infringes on the selfcontained exposition.
e.....:/.. In Chapter 2 a brief section on the change of length and area under ..~ I
~

.. ...."::
. It complements but does not replace the old proof..ae to reproduce it. has been retained and improved.:..theless. .third editions can be summarised as follows: . In Cbapter4 there is a new and simpler proof of the general form of ~~~. partly because of differences in student preparedness
I
.matical terminology. >/. .. never~. !: ... but it did not ?Eem necessary to change the style in any significant way.. It is due to A...1 ..' .I
.. There are no radical innovations in the new edition.. In a few places. . Notations and terminology have been modernized.
. ':: >'
..'.
. A short section on the Riemann zeta function has been included ...".
(~.: ... ~..I
II
I
Preface
I
1
I
!I
I·' ...a...and aims ..
.. . .
.
!.
Ii
i
!
. ~ ...II
I' I
'1 ~
..~
I • •
•
. .••. need for a new edition.. .::
. c·
'. ·I:'\.uy's theorem..~. The tage is minor. F. .
.
I·
...':':'>0
. Otherwise.. Bearden. 3. . the main differences between the second . ':_:'
'.
". for it forces the reader to fall back on for the definition and manipulation of double integrals.~·. 4.._ :'...... ·:··.•......
.>.throughout the book.~
»
.. .'
.. partly because of changes in current mathe\... .
. 2...
• •
i
I
I
j'
. .......::.. .: ..
. it was desirable to clarify certain points that exhas shown to have been a source of possible misunderstanding or · " Misprints and minor errors that have come to my attention ve been corrected.. and for this reason the introductory chapters are virtually unchanged..:)....<:
...
COMPLEX ANALYSIS
Lare V. AhlfoT8
:. weaknesses. who kindly let me share student reaction to a recent course based on my book.. 5. The author has successfully resisted the temptation to include Riemann surfaces as onedimensional complex manifolds. and the proof of the functional equation illustrates the use of residues in a less trivial situation than the mere computation of definite integrals . and errors in the second edition. The main purpose W88 to introduce the reader to the terminology of germs and sheaves while emphasizing all the classical concepts. It goes without saying that nothing beyond the basic notions of sheaf theory would have been compatible with the elementary nature of the book. The book would lose much of its usefulness if it went beyond its purpose of being no more than an introduction to the basic methods and results of complex function theory in the plane.:"
I
. 6.PREFACE
This always fascinates students. I am particularly grateful to my colleague Lynn Loomis. Large parts of Chapter 8 have been completely rewritten. It is my pleasant duty to thank the many who have helped me by pointing out misprints.
.::··::~. . if ~ s: 0.
""7"'(
_:~:.:
.relation is 1 .··:•. We begin our study of complex function theory by stressing and implementing this analogy.
In the seeond identity we have made use of the .
..~#{.
.:"':'<'.. ::. .::·~:)~·h. Assuming that the ordinary rules of arithmetic apply to complex numbers we find indeed
(1) (a
1. .
.' . h1. + ia)
(a.2:~:&~...
From elementary
algebra the
+ ~) + ("I + ia)
!S
= (a + y)
+ i(ft + &)
and
j'
I.:: .. Addition and multiplication do not lead out from the system of complex numbers.
.~g. to be purely imaginary..
(2)
(0:
+ iIl)(_.
. Two complex numbers are equal if and only if they have the same real part and the same imaginary part.:i~::i.1
COMPLEX NUMBERS
:L THE ALCEBRA
I
OF COMPLEX
NUMBERS
I:
It is fundamental that real and complex numbers obey the same basic laws of arithmetic.~~~~~~:~k
..'.. Zero is the only number which is at once real and purely imaginary.' .
::II
i
...:: . Arithmetic
Operotioru.'
I
:
!
i
'1
.
reader is acquainted with the imaginary unit i with the property s~ :: 1..·· :.1.. .
..":~>
I :. It is leu obviousihat di_on ia also possible..1.
~....
.l.
.. ..... the number is said. ~:.:fi~)~~ . a and J3 are the real and imaginary part of the complex number.. .i
. If the imaginary unit is combined with two real numbers aJ /l by the processes of addition and multiplication. it is of course real. We wish to
1

...
:
.~·i~\h1c. we obtain a complex number a + ill. If a == 0. .·.~.110)
+ i(a6 + fJ~) ....
Y for we know that ""It
(3)
p.* .a&)_
~ft·n~ (4)
" ..
+ (Ja + 02 +
0:8
aI'

This system of simultaneous linear equations has the unique solution
x=
.....oR _ ± (
(1
3 +41'
+ sj*' + (I
...:
. a'Y
. ...\iI
....
.nnot be combined arbitrarily.a + V at + fj!).. '.
~ ctthe·vaIues are
± v.2 2
.
.
Together with the first equa. = "'(2
This is equivalent to the system of equations
(4) thus the result
«8.
x2 yl
=
As a special case the reciprocal of
a
8.
EXERCISES
L Find
cr.
IDlution
the values of
5
"..I.~. we find at once
a il3 1+ i~
·~Beneewe must have
X'
+
=
+ i(3)(.y' From these equations we obtain (Xl
=a
+ at is not sero... provided that 'Y iii F O."
·~ere the square
we find
+ y"
=
V«2 + fJl...it!. has only four possible values: 1.~~~ equations (5) yield.
.2
COMPLEX
ANALVSI.2
1.. .
+ i~)(x + iy) ..
&y
(1 ±2iy1
for all combinations
r.be found in a simpler way.i6.
root is positive or zero. If the given number is a + ifj.
I. i 1.. in general... But ·these values ca...'
+ is'
..y 10 that their product has the sign of fJ~ This leads to the
1I. If numerator and denominator are multiplied with 'Y .:. two opposite values for :t and two ~he . L.
+ if>

. (.(o: + V. 1
(x
and
and we obtain the two equations
a
of signs..
i(a
0:
+ i~ =
1
a~
at
+ (jl
that these quantities are positive or zero regardless of the sign
We note that itt. we must have
a
+
+
+
2.~1~"4.. . Iii .
== 7X fJ == 3x
+ "'Iy.t + a
+ y2)t
= (Xl 
Nt)'
+ 4x2yi
~ as
+ fJlfo
Once the existence of the quotient has been proved.::
x
+ iy
(x and y real).
± i .
·thlt ~ jII. We have a + ilJ _ O:'Y + fjll + i fl'Y 'Y
2xy :::: fj.and . for the second ~ ru (4) is not a consequence of (5)~ We must therefore be careful 3: . Hz. We shall now show that the square root of a % complex number can be found explicitly..
COIIPLEX
NU"B~RS
3
ahow that (0: ifJ)/(r ill) is a complex ~umber.
.i8) _ (cry
. }
.. + i3)("
(a
. . complex
number ~ 0 is given by
+ v' 2 + (jl) = l(. ·0> . They eorrespond to values of n which divided by 4 leave the remainders 0. If the quotient is denoted by x + iY. Square Roots.r
t: + Val +. find the real and imaginary parts of
~
1
z
+ iP
= (.i)·...~
j...
...ia) 
+ ~a) + i(~ 7 + a'
1
.. 3.: + ltl! pi) .. . its value c~n.: Falll
. .
2. .~v..i.
l
x' . we are looking for a number x + iy such that
+ iy)'
=a
+ ifl.
. i.. Show that
z'
z
+
1 l'
1
By (2) this condition can be written
a
+ i~ ==
(1'X 
try)
+ i(h + 'YY).. if a
~ 0.y=... . 1.
genarally to indicate a
.:.(11 . a :i! o.nd ~
w9M
fields is a onetoone aor.ff)2 J and this is . are all different.<~:. .. ~ots.r'.
..~]>. Solve the quadratic equation
~or
Z2
+ (a + i~)z + ~+ i8 = O.. The subset C is & subfield of F.
an?
18
th
o. imporiaDt
.a. and lower align in (6). Hence R contains the natural numbers. .xl + 1 = 0.:.•ubfield..
:.::. Compute v . for at + 1 is &Iways ·:~·positive. So far our approach to complex numbers has been completely uncritical.. the equation of subtraction f$ x =:' a has always a 801utionJ~d the equation of diviBion {jx = a has a solution
+
+ +
'Whenever·(1 ~
One showe by elementary reuooing that the neutral elements results of subtraction and division are unique. ~ . Denote & ·aolution by i.. Also.. but this distinction is e.. We begin by recalling the characteristic properties of the realnumbr::r system which we denote by R.an d V J .. Thi~ means that addition and multiplication are defined. The equation Xl + 1 = 0 has no solution in R. a ~ 0 and purely imaginary if fJ = 0. this is exACtly what was shown Sec..tm
t We
aaaume that the reader baa a working knowledge of elementary algebra. only if a = ol.. .. Suppose now that a field F can be found which contains R as a ·.hkdl:'Ja:~ .. Let C be the subset of .rtifici~ and should ~ avoided...0 either a or a is positive.
. < a. Compute
These properties are common to all fields.is another field containing R and a root i' of the equation . Let ttl < (12 < (II < · . In addition.
+
t
+
~"lI!III
.1.uesof _II:: ~4r: I. and· ~. . hence (a . textbook in which a full axiomatic treatment of real numbers is given.i'. Since both square roots are in .. It 18 understood that all square roots of positive numbers are taken with the positive sign... ~'.8 < all < A for all n > HI. t
. and since it is a field it must contain the subfield formed by all rational numbers . 1 1 + 1. only positive numbers have real square roots and only ~egative numbers have purely imaginary square.. ..·. .
Although the above charaeterization of a field ill eomplete.)(x . " .. 0 can be solved. JusiiJication.
2. t The student who is not .. < · · · JI and assume the existence of a real number B such that ~ <: B for all R. They are real if fJ =.0.
... .
Find the four v&l. For if . (J = If. Our diseussion of the realnumber system is incomp1ete inasmuch &8 we have not proved the existence and uniqueness (up to isomorphisms) of ."' "_ . _. every field integral domaift: et/J .B jiefil. They coincide only if a: + ifj = O.respondence which pre..eneral complex.a E R+.
t. for a ifJ = cI + i/3' · a . In the first place. The set R+ is characterized by the following properties: (1) 0 is not a positive number: (2) if a. In other words except for sero.
is uaed quite. .
bet_ two
~~
. 0 if and only if (If = 0 or 13 =
o. 1O one.. satiBfymg the (l880C1alive commutative and diBlributive law.. · We have found that the square root of any complex number exists and bas two opposite values.. respectively: (If 0 = a..:: II.~ "'~i.Fconsisting of all elements which can be expressed in the form « + ifj real a and Ii.:.. therefore 1 = I'
Vi.lJot~
ooe. a system R with the postulated properties. The numhem 0 and 1 are neutral :Uements und~ addition and multiplication. the field R has an urder relation a < (J (or (j > a)...
e an
is a positive number.
EXERCISES
1. and in which the equation Z2 + 1 ::. 1.S.. F.tion Xl 1 = 0 has a solution while all the roles of arithmetic remain in force. In fact... except for trivial verifica)Vhich the reader is asked to carry out. Moreover. What is more. and the equation + 1 = 0 has exactly two roots in F.. it 18Dot possible to distinguish between the positive and negative square root of a complex number. By virtue of the order relation the sums 1. We have not questioned the existence of a number system in which the equa.obvi~us1y does not convey much to a student who ie not already at leut vqueIy faDllbar "Wlth the concept. We could of course distinguish between the upper. d'NlatioBetMt are 00nBidered.a.
sum... Finally J R satisfies the following completen. 1 + 1..COMPLEX
ANALYSIS
COMPLEX
NUMBERS
5
if a < O.::... Then 2:' + 1 = (z + .. thoroughly familiar with one of the constructive processes by which real numbers can be introduced should not fail to fill this gap by consulting any . the structure of C is independent of F..
I..
.a')' == ."oOIl.. (If • 1 _ cx for all cx. The correct way is to treat both square roots In a symmetric meaner...~).~ . From these conditions one derives all the usual rules for manipulation of ineqUalities.
00l'f'&0
. In particular one finds that every square al is either positive or zero.a! == i(ft . (3) the sum and the product of two positive numbers are positive. .
·Ii". Then the completeness condition requires the existence of a number A = litna~ a" with the following property: given any g > 0 there exists a natural number no such that A .. It is most easily defined in terms of the set R+ of positive real numbers: a < fJ if and only if (j .tJ. i and i.
: _. ~ is . This representation is unique.e88 crmdition: every increasing and bounded sequence of real numbers has a limit.
Since i has the same property.4. . Direct verification shows th~t .a..
. and Q: .
.rc_ WIth real coefficients occur in pairs ofoonjugate roots. The simplest and most direct method is the following: Consider all expressions of the form a + i~ where a. It is thus demonstrated that C and C1 are isomorphic. . .. The ~~rresponding property for quotients is a consequence: if ax = b. Conjugation.. )
:=
R(4.
particular. if the coefficients are real. There is a onetoone correspondence between C and C' which B880ciatea a + ifJ and a + i'(j. .~J... 18 everywhere replaced by i. ~." + {JI is always positive or zero. to make free use of both notations. all rules must remain valid if I.. · . By systematic use of the notations a and a it is hence possible to dispense with the use of separate letters for the real and i~ry part.. Show that the system of all matrices of the special form
~bined by matrix addition and JDB.. 2.. The fundamental property of conjugation is the one already referred to.... ' The tarminOloir &ndnot&tioliare jUtified hT
0"
. A complex number can be denoted either by a single letter 4. y.stand for any rational operation applied to the complex numbers abc ' .xZ + eft..#1\~ii
. it is identical with field)"
+
?Xpress.
1. r == ~ + i'l1. and i is not an element of a.Uix multiplication.. Show that the complexnumber system ean be thought of 88 the : ~d of all .!. In deriving the rules for complex addition and multiplication we used only the f~ct t~t .i{J is the conjugate of Q: #t The conjugate of a IS denoted by tl. symbols (+ does not indicate addition.. 1. 'U. t..»
.
'.!:'·':·~~i~1. 11 are real numbers. let R(a.. In order to give our definition s meaning it remains to exhibit a field F which contains R (or a subfield isomorphic with R) and in which the equation x2 + 1 0 has 8 root. It is more convenient. !}
R(atb. and this correspondence is evidently & field isomorphism. Other standard notations are z == z + i1l.. root of this equation.~ynomial x"2 + 1 ...
(~.. ==
==
':If f is a.
O. .
:: . Absolute Yalue..c.. {j are real numbers while the signs + and i are pure.b.) :' .
The existence of the complexnumber _field is now proved...
:": " • e. We now define the field of complex number8 to be the subfield C of an arbitrarily given F. A number is real if and only if it is equal to its conjugate. • ..
. moreover.c.. We have just seen that the choice of F makes no difference.:.Z + C. ·
4>
. The formulas
tac:
+
+
Re a
=
a+·4 2'
Ima =
aa
2i
These expressions are elements of & field F in which addition and multiplication are defined by (1) and (2) (observe the two different meani~ of the sign +) ~ The elements of the particular form a + iO are seen to constitute a subfield isomorphic to R.+ c.
.polynomials with real coefficients modulo the irreducible . or in the form a + ifJ with real a an·d /l...
o..' . ' then ax = b. we obtain in fact {O il)2 = (I + iO). though.· + c.i". is isomorphic to ·.t tid "'" a...
:::I
it1y understood that x. . Ita nonnegava aq~ JIOOtra'hIIecJ the mod~ua ab. .
'"

.'
.. .8
COMPLEX
ANALYSIS
COIIPLEX NUMBERS
7
the corresponding subset C' is formed by all elements a + £'/3. There are many ways in which such a field can be constructed.
EXERCISES
(For students with a background in algebra)
a+5=4+6 ab == a . representing an element of the field C.the field of complex numbers .:::
.
As an application. f and t are roots of the same aqua~ ~ we have the familiar theorem that the nonreal roots of an equa~ .b. · Th en
j .. ~ tr~ormation which replaces Q: ifJ by Q:.c... consider the equation
coZ"
+ Clzta1 + · .. and hence (b!a) = 6/4. More generally.. that
die corresponding subfield C for we can write
j
a
+ ifJ = (a + iO) + /3(0 + i1). The·field F has thus the required properties.this is indeed so.. ~s The real and Imagmary ?~t of a complex number a will a180 be denoted by Re a. namely.:. and the element 0 + it satisfies the equation x2 + 1 = 0.p IS calle~ complex c~J'Uflation. w = U + WJI and when used in this connection it
..' The produo.
. It lIJdeootedbyltlf •. The conjugation is an i'IUJol'Utory transformation: this means that d == 6. thtn f is a root of the equation
cor + ClzA1 + · · . Im a.~ 1. } .. but we have not yet shown that there exists a field F with the required properties.b.. the real and imaginary part in terms of the complex number and Its conjugate.tol~ ~ue of the C9Plplex num~.. and we can go baek to the simpler notation· a + ill where the + indicates addition in C aWl· i is & root of the equation Xl + 1 = 0..
~".:: abab .
2i(3
+ ...
. .
.~>.::.: ..\ '~.:..~i...la\1
==
ad
+
(ab
+ btl) + bb
1
The equality Re Q =: lal holds if and onJy if a is real and ~ If (9) is applied to (7). .
.4::\)~~""..~~.'.\::J..li ~:::i.bl
== 2(lal2
+
. We repeat the definition
.1~ :'.~:. "".
e
1~1·
important Inequalities to point out that there and hence all inequali
The quotient a/b) b F.':..~.. and··it·js benae ~WIIent·'to alb .
.:".:...'
.~~(:': ~.. ad
L Prove that
== lal'. From the definition of the absolute value we (9)
since both are ~ o~ In words: The absolute value of a product is equallo the product of the absolute
values of 'IJw !ackrr8~ . satisfies bee/b) ==
Ibl · la/bl == \al.~~ f:: :".. /. O·~ condition can be written in the ..~:}. and compute that solution.. Let us determine all cases of equality in (11)~ In (10) the equality .~ .
.
.".t. JI!.~ ....~}. 2.A:. and observe that la:l == product we obtain
lal.%~9i~4~'{r.i.: :}.iy are conjugate. lnequolitiu. The reader is well aware of the importance of the estimate (II) in the '..:.'.::~:_i':. ..:~'.. ::"j. If' b. '.~. ties must be between real numbers..
~b\:).
la + b1t la la
btl
+ lb.~.<f: !'~':~~:'~~'..". • J ~:~.:_' (Il) The abaolute value of a BUM is at moat equal to the BUm of the ab30lute Nluea oJ the terms.?:~.t == 'all · I~' · ·
G. "L~·".\}...' + 2 Re ab..::.::_ :~ "..\"'1'l~'\·it':'~·(.~~i.5... '.':.:v'::... ~:.<" .'\. ' . :':O:..:...'~.:.. ( .~.
We find
la + btl ==
or
(7)
(a
+ b)(d + b)
=. /:...~.form Ihlt(ajb) i: 0....t..'
:·.~S.:
:.·. holds if and only if a6 ~ 0 (it is convenient to let c > 0 indicate that . \
:~<~
~·~f:·..:'.'»'·. Find the conditions under which the equation az + bi + c == 0 in one complex unknown has exactly one solution.. I..'
.~~":~~ .1 ....: '::i.'. L Prove Lagrange's identity in the complex form
lOOl == 'al · Ibl
1..::.:.. Find the absolute values of
'·bumbers ..bt)!
The corresponding formula.~. ~ 0."~~~~:(~:~ ~':':. we obtain and hence
o.poajtiue)...'.• It is clear that this property extends to arbitrary fimte products .
~.
.:.~i".s.:. or
and hence we have also
deduce the inequalities
lal ~ Be a· s la. for the difference is (7')
(8)
EXERCISES
=
..'~. .....hi. o.~~·~·7·::\~'
.. .. = 11' Ibl .
.:.' .all
2 He
ab. ~.is real and .
For the absolute value of a
labll
and hence
= ab ~ 00
. ~ :' ." .
+ Ibt
l l 
ia + bl!
~
<Ial + .:~~..
..:~ ·:..\ ~.)
and
.!..:: adhb ~
'a'I'b. ~~: :~:....:.. ~..
'
..
if either la1 == 1or lb. In the· general
'. i~:~' itt· .:"?t.'....Ij
z
for z = x + iy and a = % .0.. .~ case..0. (10) : This is called the triangle i'lU?AfUlJlity for reasons which will emerge later .:.':..~.•="'.':.: .~ ...·~.t. =: 1~ What exception must be made if lar ...·~it<j. .~ .~ .. and we aball find it no leas imporlant in the theory of complex
L Verify by calculation that the values of
z..
. ~ 0'. Ial ::ii 1m a ~ lalA
The formula for the absolute value of a sum is not as simple.:~~...:..0
. .8
COMPLEX ANALYSIS
COIIPLEX
NUMBERS
•
the fact that the modulus of a. We shall now prove some which will be of constant use..<.t) (3 . real number coincides with its numerical value taken with the positive sign . Jly induction it can be extended to arbitrary sums:
and by addition we obtain the identity
+ btl + la .
t
lalc2 ~ • •
a.:" ._.:·...:.. .0•: .
+1
(3'+ 41)(1 + 2t) ..f·o.\~i .. It is perhaps well is no order relation in the complexnumber system... ~+~:if..
I
1db
a
b
I
:::11
where la. .·:·...)(2 + ~)(1 + ...
b Idb
I
<1
Prove Cauchy's inequality by induction.3)
+ bl.+ lant~ la1 + a2\ == lat' + 'a:l.10
COMPLEX
ANALYSIS
COMPLEX NUMBERS
11
case we proceed as follows: Suppose that equality holds in (11). . '
or For the same
combined to (12)
la\
= I(a 
b)
+ bl ~ la . o.b\. ...l = \(al + at) + a$ + · · . + a2 + · · · + a.::. used as a subscriptt cannot be confUBed :with the bnaginary unit. .0 we obtain
+ .
2.
•• ~
are ~mplex
u.....o. 4) . By (10) we have also
if
and ooly
any
for if the denominator should vanish there is nothing to prove. Assuming that at '#.1 ..
:1
•~
'~. • • • • • • • • 
."I:_~''''': .
t.
To Bum UP..1 (1 + 1::1 + · · · +~)
laJI
+ \a21 + · .
la
+ · . ..4.XIRCISES
ia + il3l ~ ial + ItJ'
which expresses that the absolute value of a complex number is at most equal to the sum of the absolute values of the real and imaginary part. then
1all +
Hence
1a21
But the numbering of the terms is arbitrary. Jet).. ..
:
. . Cauchy's inequality can a1so be proved by means of Lagrange/a :. . This choice is not arbitrary. ~ tI
d.pnbersz satisfying
II ".
• I. =
(AlGI
+ AtlIs + · .. Foremost is Cauc"vs inequality which states that
...1 = la.
+ •
To prove it.
'. 0 fori = I..:.. •. Suppose conversely that this condition is fulfilled. At: .~· + Ia.~+
(lall!
+ · · · + lan(
2)
(\b I
l2
+ .bl. Substituting in (15) we find after
simplifications.. identity (Sec.. but it is dictated by the desire to make the expression (15) WI small as possible. . ':': ". thus the ratio of any two nonzero terms must be positive.~ ~.~:
.. +:..: the 8ign of equality hold« in (11) tWo nonzero terms i8 positive .J < 1.~.Ibl s la .1. /.. ... ... _..~ ~:. and these inequalities
which is equivalent .. denote an arbitrary complex number.':i .~~albl a b....
1a 
b1
~
IIal .
I
a.
..(I1... by (7) (15) This expression is ~ 0 for all X4' We can choose
We obtain
la.
.tbere
.lal s: 1a .. reason Ibl . L If 1401 < 1.:~~..
t·~.. . Many other inequalities whose proof is less immediate are also of fre
L Prove that
·"nt
use.
.n and >"1 + ).. and if a2 'F..' ••
t i is a eonvenient summation index and.. i . + Ia' ~ lall + latl + · · .
. It seems pointless to b"n its 1J8et
4.11 + ~ + · · · +:: = lall
=
(1 + :: + · · · + ::) = ja.. I.· + la
if 1M ratio oj
ft}.bl + Ibl
can be
lat .
. 1.. I. Show ~t... '.
From (~5) we conclude further that the sign of equality holds in (14) if and only If the a.t : show that
}
+ · ·· +~.
..IbII·
Of course the same estimate can be applied to la A special case of (10) is the inequality . are proportional to the 0". ." + x. . ~Icl·· .]2 ~ in shorter notation..+ ~r ~ lal + 421 + las~+ ... ' ..
to (14).0 we conclude that aI/a) 5=.n~
.l
l)
: if
101 < 1 and Ibl < 1..· · + Ib.• . Ex.
. this formula can be extended to arbitrary products.ti.4~. and we express it through the equation
+
+
(17)
It is clear that.. lL
Vector addition.~ numbers.rp).:.
. which we have ani"" at~ the·:formula (17) .. The number the point and the vector will all be denoted by the same letter a. __ '.'. ..aJ 4 are the vertices of a rectangle which is symmetric with respect to both axes.¥i:. and the second coordinate axis (yaxis) is called the imaginary axis. we can therefore state: ..':. In this trigonometric form of & complex number r is always ~ 0 and equal to the modulus [c]. Observe that a b and a . what are the smallest and Jargest values of izl?
2.l~. and we shall often speak of the point a 88 8 synonym of the number a.. and we denote it by arg a.
'.'~~'::~)h:_.')}~x. Their product can be written in the from ala. however. As ~ual we identify all vectors which can be obtained from each other by parallel displacements.'~'::. For this reason we shall use geometry only for descriptive purposes and not for valid proof.
2.~j.<\\0'~1~:~fi:i::j}fY~~~i:.'Violatesour principles. and not from the axioms of geometry.'. Consider two complex numbers a1 =. unless the language is so thinly veiled that the an~ytic interpretation is selfevident.
.sin 'P18in 9'.. The latter result is new. The symmetric point of a with respect to the i~8ginary axis is
Hence we can write a == a + ill == r(cDS fP + i sin 'P). The orgumem of a producl ill equal to the BUm oj the arguments of the jocttwB . and the argument tpl 'P2. ..n~ b is Ja .': •• ~.a points from the end point of a to the end point of b.
+
PIG.:{:c/:. The addition of complex numbers can be visualized as vector addition.
:
• • '. . An additional advantage of the vector_ representation is that the length of the vector a is equal to [e]..bll == 2{lal' + IbP') become familiar geometric theorems. that all conclusions in analysis should be derived from the p~pertieB of real numbers. . ' '~. With this interpretation b the triangle inequality l4 + bl . ./:. but also by a vector pointing from the origin to the point.
. however. If the polar coordinates of the point (a. If this condition is fulfilled.tJ). """. We take the point of view. Place a second vector b 80 that its initial point coincides with the end point of 4. Then a + b iB represented by the vector from the Initial point of II to the end point of b..' . This is fundamental. This representation is constantly used..tpl COS CPI+ cos 'PlSin 'P2)].
We recognize that the product has the modulus rlf. 11).a we draw both vectors a and b from the same initial point. '1: _ ".. deep and unexpected justification of the geometric representation of com~ ~. To this end we let a co~~ number be represented not only by a point. then b . The polar angle f{J is called the argument or amplittule of the complex number. We must be fully aware.a.
. In the
.1~Geometrie Addition and Multiplication. "..
. The geometric representation derives its usefulness from the vivid mental pictures aa80ciated with a geometric language. By means of the addition theorems of the cosine and the sine this expression can be simplified to
(l6)
alOI
== rtr2(cos
(" 1
+ IPS) + i sin (lP 1 + rps) 1.::. . In order to derive a geometric interpretation of the product of two complex numbers we introduce polar coordinates. Hence the distance between the points a u..~. l.~~. The rule that we have jU8t formulated gives 8 . . .12
COMPLEX
ANALYSIS
COIIPLEX
NUMBERS
13
if and only if lal :S lei.
':'
.r1(COB 4'1 + i sin rpl) and 62 == T2(COS VJ'l + i sin ifP2)..«.~) are (r. = rl'I[(COB fJl 008 tpl .) + i(sin. we know that
a=rcos9'
.:3! JaJ + Ibl and the identity la + bpt + la . This attitude relieves us from the exigencies of rigor in connection with geometric considerations. the complex number a :::::: ifJ can be represented by the point with coordia nates (a.
.b are the diagonals in a perallelogram with the sides a and b (Fig.
fJ == T sin e.~.
':...
The four points a. The point a and its conjugate a lie symmetrically with respect to the real axis. To construct the difference b ."
... THE CEOMETRIC
COMPLEX
REPRESENTATION
OF
NUMBERS
ab
With respect to 8 given rectangular coordinate system in 8 plane.t the manner in . The first coordinate axis (saxis) takes the name of real oai«. t)u. and . The plane itself is referred to 88 the complex plane.. ..
Secondly.= arg 4z 
opposed to real analysis. The points 0..(20) . . 5.. aJ. 1.:. the polar angle is determined only up to multiples of 360°.In the case of division (17) is replaced by·
(18)
O:J arg .ark: A perfectly acceptable way to define angles and arguments would be to apply the familiar methods of calculus which permit us to
express the length of a circular arc &8 a definite integral. 1.~fJ:j}:+:·:. to be derived in Chap.14
COMPLEX ANALYSIS
COMPLEX
NUMBERS
11
first place the equation (17) is between angles rather than between numbers. 1. :. a1 is similar to the triangle whose vertices are Of aI. Prove that the points al} at. The clue lies in a direct connection between the exponential function and the trigonometric functions..analysis. Express the result in symmetric form.
t. For this reason.
This formula is trivially valid for n
aI = ..ngle with vertices at.<:. 88
=
rl[COS (rp) +isin(fJ)J
it holds also when n is a negative integer. :
. and at being. Suppose that a and b are two vertices of a square. if we want to interpret (17) numerically. given. . this similarity determines the point 61"2 (Fig.
= COB
n~
+ i sin n"
. except that the similar triangles are now 0.. 12. at and 0._.isiD.pproach.. & correct definition of the trigonometric functions.0.. :. at. all are vertices of an equilateral triangle if and only if a~ a: = B1U:2 + G2aa + aaal.
. The Binomial Equation.)
=
0..p
Bilppoaing that a "# 0 we write a = r{cos
z
+ i sin (/') and
==
p(cos B
+ i sin tI} . It follows indeed that the triangle with the vertices 0. Until we reach this point the reader is asked to subdue his quest for complete rigor. and secondly its proof rested on the use of trigonometry. 41a2. This leads to .. Sec.
t!'
·'1' . . and hence (17) has & meaning only if al snd ~ are #...
vector
multiplication.. we must agree that multiples of 3600 shall not count.. and sinee
Rem. . a2/al. i:
. By means of (17) 8 simple geometric construction of the product alOJ C&Il be obtained..
:Then (21) takes the fonn
o
1
PIG.. ..equation
t'L .
L Find the symmetric
points of
a
+ + a:
al
a.
. and to a computational proof of the addition theorems. which provides an extremely simple way to express cos 'nf' snd sin nIP in terms of cos tp and sin tp.
:
..::.. To find the nth root of a complex number a we have to solve the ... Thus it remains to define the argument in analytic terms and to prove (17) by purely analytic means.rg al. ...
.
.3. that the powers of a == r{co8 rp
(19)
+
From the preceding results we derive i sin rp) are given by
The geometric construction is the same. We remark first that the argument of 0 is not defined."::
(cos
IP
+ i sin ~). til.
. .
2"..1(008" . . 2.
== a.. EXERCISIS
with respect to the lines which bisect the angles between the coordinate axes. Find the center and the radius of the circle which cireumseribes the tria. ... 12). For r =: 1we obtain de MoivTe's formula
:. 02. offers a much more direct a..
.: :. The reason we do not follow t~ path is that complex· .. For the moment we postpone this proof and shall be content to disCUBS the consequences of (17) from a less critical standpoint . Find the two other vertices in all possible C&8eS. 3. 2..
..
" . ~. only the values k = 0.r :. If '" is given by (23)..':O: tp. + k~)} k = 0. The direetiot).expressed ~ radians the full angle is 2Ir.
. a directed line z = a + bt determines a right ha1f plane consisting of all points z withIm (z . and they are equally directed if b' is a positive multiple of b. The thing to remember is that a complex equation is ordinarily equivalent to two real equations...
..e] == r..3. •. Prove ~·aUy that the·midpoints of parallel chords to a circle
ur + (a)u
+ ' · .:<:.
. aXERC... 2.~.... COB fl..(0)_1... the equation of a circle is Iz . Problems of finding intersections between lines and circles..
EXERCISIS
L Express COB 3~.t~~:·: "~:. hyperbola. The angle between z = a + bt and z ~ a' + bit is arg b' [b..
Henee we
L What is
the value of 1 .. But this is not the only solution. .a)(i .a)/b < o and a left balfplane with Im (z . In algebraic form it can be rewritten as (e . in order to obtain a genuine locus these equations should be essentially the same ..
.al < r describes the inside of a circle.
They hove the mme
B
~metrically. ~ aD .. . In classical analytic geometry the equation of. ..+
W(w.
ft.t~a. in terms of cos .•..
s. observe that it depends on the order in which the lines are named. · · · . and the like usually become exceedingly simple when expressed in complex form .L ~ . Simplify 1 008 f{J cos 2". Hence the complete solution of the equation
z_ Vr[ C08(~ + k~) + iSin(. Express the fifth
and tenth roots of unity in algebraic form. Analytic Geometry. and we find th&t (22) is satisfied if and only if
Vr denotes
6=!+k'~'
n
"
where k is any integer. · • · J n :1 give different values of e. 2r Ism
n
all the mota can be expressed by 1.."
:
. . An inequality Iz .
L When does as bi C == 0 represent a line? 2.. _..Wi + Cc)2l

•
II
•
+ (l)1ilw(ftl)A?
==
Vr( coS.~~ ~ that> .
the nth roots are the vertices of
regular polygon
with '" sides.& locus is expressed as a relation between % and y..
+
+
+ ·· ·+
+
form .~'""1'::'
'"
_:'
. then all the nth roots can be expressed in the form wi · Va..t!
O. Prove that the diagonals of a parallelogram bisect each other and
$hat the diagonaJa of a rhombus are orthogonal. 1... Write the equation of an ellipse.
:. In fact.. k = OJ 1. . J n . ~~~. 2r
+
. The fact that this equation is invariant under complex conjugation is an indication that it represents a single real equation .. .
.E
(23)
w
= cosflo
. The lines are parallel whenever b' ~ a real multiple of b.
(21) is given by
However.l)A
~
0
for any integer h which is not a multiple of
:....tP' and sm e
1. It can just as well be expressed in terms of z and if sometimes to distinct advantage.. w". The lines are orthogonal to each other if b' [b is purely imaginary .. · The case a = 1 is particularly important. . on a (tiantetB. ~~i". . For instence..SEI ..
:
. It is also quite evident that if Va denotes any nth root of a. the parameter t runs through all real values.I .1..
and 1/4 intenJect the
. The roots of the equation z..a) == r2.. of B directed line can be identified with arg b.
Then are n nth roms of any complex number modulus and their arguments are equally trpacedt
.~:t.~..18
COMPLEX ANALYSIS
COMPLEX
NUM8ERS
17
This equation is certainly fulfilled if obtain the root
z
p'" =
r and
fI.n 1.. .a)/b > Ot An easy argument shows that this distinction is independent of the
. A straight line in the complex plane can be given by a parametric equation z ~ a + bt. parametric representation..... parabola in complex
+
+
sin 2"
I.a and b' are real multiples of ·b. tangente. · ... (oJ.• _ 1"'1'· l!·. Similarly.. If angl~ aTe. where a and b are complex numbers and b ~ 0.. and sin 5.. + isin~}
where
the positive nth root of the positi~e number r.. and ~ ".. ~
G·
. ~:: .· + sin rup. multiple of the full angle. (22) 1S also fulfilled if ~8 differs from i{J by s. COB 4". prove that
1+
+ .•. parallel or orthogonal Iines.
Pi..1 are called nth roots of unity. Two equations e = a + bt and z = 0' + bll represent the same line if and only if at ..: ':\ .. and if we set
.pefpendimtlar to the chords~ .
~.)plane with the XI. hat the stereographie projection transt forms every straight line in the zpJane into a circle on S which passes through the pole (0.. .
.)i'~i. .....
. :.OJ1) as shownin Fig.": ~'.: <:'" <!:.
.
Stereopaphic projection. no bsIf plane shall contain the ideal point . To prove this we observe that a eirele on the Sphere lies in a plane atXl + atX2 + a.. oo
= oo ..
.::' ·. CIO without violating the laws of arithmetic. from (24) we obtain
=1
+
+
+
and this correspondence is one to one. .0. It is Impossible.x..2atY
+ 40 + aa
=
o.. In terms of z and i this equation takes the form
++
(24)
Z=
Xl
1 . Conversely..:~>
.. the transformation (24) takes on a simple geometric meaning.' .~"/st~. number system..: ::.$.'.. For many purposes it is useful to extend the system C of complex numbers by introduction of a symboJ 00 to represent infinity..:. ."
.
. In the spherical representation there is no simple interpretation of addition and multiplication. the equation of any circle or straight line
Further computation yields
(26)
Xl
z+z = 1 + Izlt
zj
2:2
= i(l
+ Izl~)·
The correspondence can be completed by letting the point at infinity correspond to (O~O.'.. and (O}O~l) are in a straight line. We agree that every straight line shall pass through the point at infinity..
. The Spherical Representation. It is geometrically evident ..xa = ao.. ".':::_~'\:.. < 0 corresponds to the disk lz\ < 1 and the F..:. We· note that the hemisphere x.
• ". however. It is desirable to introduce a geometric model in which all points of the extended plane have a concrete representative. to define CIO + CIO and (..y.2alx . : . Its advantage lies in the fact that the point at infinity is no longer distinguished .
.18
COMPLEX
ANALYSIS
COMPLEX
NUMBERS
18
2... . If the complex plane is identified with the (zl.and X'raxis corresponding to the real and imaginary axis.
... Hence the correspondence is a central projection from the center (O. i.
(ao 
(Xl
+ y2)
.. . By contrast.
.1::::~:.Xa) '2
+ x~
1
+x
x. b=
a)
for all b ¢ 0 including b = 00.XI
+ ix. More generally._.aa this is the equation of a circle.4.. It is called a stereograpkic projection~ The context will make it clear whether the stereographie projection is regarded as a mapping from S to the extended complex plane. and the converse is also true.
3.
For Qo '#.~::. The points in the plane together with the point at infinity form the extended compkx plane.. where we can assume that at a~ ai = 1 and 0 ::::ao < 1./:. Writing z = x + iy we can verify that (27) and this means that the points (x. Its connection with the :finite numbers is established by setting a 00 == CoO + a = for all :finite OJ and
+
QO
b .":} .)
ati(z .<" .
~" •• ~"<'
i.
Indeed.. respectively. any circle on the sphere corresponds to a circle or straight line in the eplsne .. To this end we consider the unit sphere S whose equation in threedimensional space is xl x~ xl = 1.1) = ao(lzl:
1)
[zF' = xi
and hence (25)
(1 .l)~we can associate B complex number
IX)
hemisphere XI > 0 to its outside 1z1 > 1.XS).. . By special convention we shall nevertheless write a/O == 00 for a ~ 0 and bj Q) = 0 for b ¢ 00 . and we can thus regard the sphere as a representation of the extended plane or of the extended.i)
+ aa(lzll
.l). With every point on S~except (O.
11. In function theory the sphere S is referred to as the Riemann spMre. In the plane there is no room for s point corresponding to f but we can of COUl'Se introduce an "ideal" point which we call the point at infinity.
. '.O.
or
Ul(Z
+ i) a..f>. or vice versa.1).O) (X1JX2. 13.a. and for an = aa it represents a straight line.
.
function and f(.
.. ..) Howevert anaJyste are traditionally minded and
21
. INTRODUCTION TO THE CONCEPT OF
AI.
....: ...(z ..
"
..1)
2 ) 
From (35) and (36) we obtain alter a short computation Xlz.·:.
+ ~~ + xa~
(8
=
+ i)(z' + i')
. .::j. '. . Therefore. i:·_:.fanetioa 1(. Both differentiation and integration acquire new depth and significance... Let Z Z' denote the stereographic projections of 2. 2..:
:~~..._ "":. . we shall use the term "function" in its modern
meaning..t The notation y = J{x) will be used in 8 neutral manner with the understanding that % and II can be either real or complex.2'12
·
1.

_
.z)
(I
+ [zI2) (1 + Iz'l2) (1 + lzll)
~
+ <'zit (1 +
(1
. 5. .Z. and complex functions of & complex variable. ..:"'1.. By these agreements we
.. at the same time the range of applicability becomes radically restricted.
. As a practical matter we agree that the letters e and to shaH always denote complex variables...da for the
COIItinue _·lIP ttk 'eI·:· tv._ .~.2G
COMPLEX ANALYSIS
can be written in this form. . iI:._.. The correspondence is consequently one to one.)." . Find the radius of the spherical image of the circle in the plane whose center is a and radiue R.
•
•
.. real functions of a complex variable..
.x~)t
==
2 . ~"':"
.
(Xl 
xi)1
+ (Xt 
x02
+
(X3 .
..:
• I
. When we want to indicate that 8 variable is definitely restricted to real values. and use this to derive (28)..2(xtX~
+ x~ + xaX~).·i· r. r..'~.
00
ANALYTIC
FUNCTION

the corresponding formula is
d(z. we shall usually denote it by t.· .. . 4..·:of the fuaotion.%t.~
. .J. and let N be the north pole.) J (3:i. when stepping up to complex numbers we have to consider four different kinds of functions: real functions of a real variable. They are the only true "functions'} in the sense of the French "Theorie des fonctions' or the German "Funktionentheorie. only the analytic or holomorphic functions can be freely differentiated and integrated. It is easy to calculate the distance d(zJt) between the stereographic projections of z and H the points on the sphere a. Indeed. ~e problem for a regular tetrahedron in general position .' . " Nevertheless..I~~. Show that the triangles NZZ' and NzZ are similar.~)
=
V1 + Izis
2
EXERCISES
that z and z' correspond to diametrically opposite points on the Riemann sphere if and only if zi/ :::: 1.
L Show
The theory of functions of a complex variable aims at extending calculus to the complex domain..zit) (1 + lz' I')
+ \2'1
21z 
.
l)(lz'lt . '."
...x~) we ha ve :first J
2
COMPLEX FUNCTIONS
z..for a v. ·'. to indicate a complex function of a complex variable we use the notation tD == /(z) . complex functions of a real variable..".:z:~. Find the stereo graphic projections of the vertices .: .z)(z' ....re denoted by (Xl.~~!:!_~'. : J.
t Modern students are well aware that f 8tan.':
. thus. a
result we find that
(28) For ~ . A cube has its vertices on the sphere 8 and its edges parallel to the coordinate axes..
.
".
".
and
ftmdicm to underline that the function has only one value for eaob... :.~ ••
. :.. t It is not· necessary that a function be defined for all values of the independent variable.. 1
The following basic definition will be
adopted:
Definition 1..~\:'···...'~ ~·r~\i:. '. a product or 8 i quotient are all valid. let .. ·'··Thereare some familiar variants of the definition which correspond '~.....
.
a
X
definition makes decisive use of the absolute value.~"
f(a
+ ill)
..CoO..... ':' ..
'. ~.:.is
.only if the /oUOUJing i8 true: For BVerY E > 0 there exists a number 3· >
'1(%)
f . quotients..' .
=
Im
Ail
(1) is a consequence of (3)..
•
•
t.
. It is essential that the law by which a function is defined be formulated in clear and unambiguous terms... . :. we can use the same definition regardless of whether the variable ~·the function J(x) are real or complex.wall sometimes use the pleonastic term ai~alud
. by which we mean that if f(a) is defined. . :.: :qliotient continue to hold in the complex ease.. The formal definition...:tive UUCB ftAt q?0...: ~·:£if·:' . . . ~·_. until furtber notice.~/~~:.. Since the ~ 'of absolute value has a meaning for complex &S well as for real numbers..::.' . J u:aDYit ~ u. We trust the reader to formulate correct definitions td·Cover all the possibilities.~·lL.
r • ::... In other words. 1m f(z)~ and lJ(x)l. then f(x) is defined for all x sufficiently close to c. an alternative simpler notation we sometimes write: !(z) __..
=
A....~·:k:. .
. for it is the limit of the quotients
+
I(a
+ Il)
..r:.
0 with the property that ai < 0 and x . . the proofs depehd only on the properties of the absolute value expressed by
.:r:""'""+G lim 1m jex)
~(I
lim !(x) =
A..
5::'i"
.:.
t We . a product.J_. ...:.. <
5
for all valuea of x BUCh that
1$ 
.~ and a complex independent variable. ..
Condition (1) is evidently equivalent to (2) From (1) and (2) we obtain lim Re fez) ~ Re A (3) Conversely. Thus a .. Therefore we make merely an informal agreement that every function be defined on an open set.... and k &..:."'~ .::~::..
.':
·
: . =t:'1~..
The function f(x)
= f(a). 80 are'" Re !(x). There is nevertheless a fundamental difference between the cases of a ::.
is said to 'be continu0U8 at a if and only if A continuous function.J_.
. singlevaltl.·~.:::..':::.r.'
. Indeed.is continuous. . : :.
I
A... Ii.·1J!nit of the ..:. .. . for A .. :infinite limit.
. The jundion !(x) is said to have the limit A as x tend« to
(1) lim f(x)
a. ase where a·or A is infinite..:.. .. .
. without further qualification .::.:1.
. For the moment we shall deliberately underemphasise the role of point set theory.ed..rI: a. .~.tends to zero through real values.
•
•
•
• I. The derivative of a function is defined as 8 particular limit and can be considered regardless of whether the variables are real or complex.!(a) On the other side it is also the
h
..e.
. mined by the chain rule.. :".
:~.
...I:_! ". "_"
I···.r.." .f(a)
ih. '. all functions must be weU defined and consequently..
.. but in the complex case there is only •. If !(x) ... :. The formal treatment of point set topology' is deferred until the next chapter. The sum/(x) + g(x) and the productf(x)g(x) of two continuousfunetions are continuous.
'.Ihe.
= . (4)
f'(a}
= lim f(z) ~G
lea) .. The usual rules for forming the derivative of a BUm. . lim f(x) .. '::::TbiB
'·~.'
. .. Then I'(a) is onone side real. In the real.
if". .'_
. .::. :. To illustrate our point... nown results coneerning the limit of a·Sum...
. '"P: ..~!. the quotientj(x)/O(x) is defined and continuous at a if and only if g(a) ¢ 0.. case we can distinguish c W1t8en the limits ee and ._~u
I..'.22
COIIPLEX
ANALYSIS
COMPLEX
FUNCT'ONS
23
do not wish to cancel the earlier convention whereby a notation z = x + iy automatically implies that x and y are real.. The derivative of 8 composite 'function is deter~ . Limits and Continuity.
.••••
:.:..~) be a real function of a complex variable whose derivative exists at ..'! ·:The well..
1.. ::~ : . value of the variable.: .~. .
.. t
is one which is continuous at all points where it is defined.function of a complex varlabie:either baa the derivative zero or eIae . ..~ as such purely b&ginaryil· ThCreiooo f(o) must be zero."Y..: +.
.•
24
COMPLEX
ANALYSIS
COMPLEX
FUNCTIONS
The case of a complex function of a real variable can be reduced to the real ease..then v is said to be the conjugate harmonic futu>f
. .
(5)
lim fez
hO
+
i~lf. a:e ax
iJx
If (z) 11 =
(~)I + (au)' (au):! + (~)2== ~ax av ax ay ax
=
'1'(z)12
we haveJ for instance...ann differential equations which must be satisfied by the real and imaginary part of any analytic function.
ay
+ h)
h
.
fez)
II!:
i af
ay
=
i ~
ay
+ ~.""~. if we substitute purely imaginary values ik for h.f(z) .~~<~::~:.. ."':" .. .. t remark th~t the existence of the four partial derivatives in (6) is implied by the existence of /'(2).. . .
' •••
. In contrast. and v satisfy the CauchyRiemann equatIOns _(6)J . I(z + It) .. and the derivative becomes a partial derivative with respect to x. (f(z + h) .:~.(.... 2
These are the OamhyRi61l'f. Usmg this Information we·obtain from (6)
e:" .~~_... The definition of the derivative can be rewritten in the form
f{z) "'" Iim f(z
.n<o.
. The same is true of the quotient J(z)Jg(z... ..::.
.. Analytic Functioru../(z»/h we obtain
:I
Indeed. Strictly . By this fact u and v will have continuous partial derivati~ of orders. The complex notation has nevertheless certain formal advantages which it would be unwise to give up.aj .. s.
Au = a'u
&
H we write J{z) = u{z) + iv(:) it follows.". ":.
.
'..~.. We have thus 1(2) . ~Di:· ~:. !Ie
f(z) For the quantity
=~
+ i8u.::~
:.' ... .~. For the purpose of this preliminary investigation the reader may think primarily of functions which are defined in the whole plane .. H ~woharmonic functions 11.
. that u(z} and v(z) are both continuous._.~"~. In the general case it is necessary to exclude the points at which g(z) == O. The class of analytic fUnctions is formed by the complex functions of & complex vsriable which possess a derivative wherever the function is defined.8 complex variable has farreaching consequences for the structural properties of the function.Aartl
RMttum.
m
. ~~:
C!fIIII~
·0..f:ij"t~~:'(.
. B:0d in particular the mixed derivatives will be equal..)
== 0.. ••
. ....~!"~~~.:.speeking.~'~'~i:1. The term holtnnorphic function is used with identical meaning.. The sum and the product of two analytic functions are again analytic. then the imaginary part y is kept constant..:"':~?~~~ . az a~ a..
~_: .n.~p. from
The last expression shows that IJ'(z)J! is the Jacobian of u and v with respect to x and y.0
It follows that fez) must satisfy the partial differential equation
~
ax 

i aj
ay
&
which resolves into the real equations
w)
h
ax
=
ay'
av
ay = 
h
ax"
1....
:
1 '.(1~l857)
aDd.~.. (18261868) are reprded
'IfOI'k9Dlpbui_ the I8O. The investigation of these consequences is the central theme in complexfunction theory .f(z»
==
0 · J'(z.
fgu~of.
i I
A.:
::
..
As a first consequence J(z) is necessarily eontinuous. The limit of the difference quotient must be the same regardless of the way in which A approaches zero.
•
. the simplest is
.. H we write z(t) ::: z(t) iy(t) we find indeed
+
Similarly... H we choose real values for h.':..
tb!l"... we obtain
f(2)
=
z'(t)
==
x/{l)
+ iy'(t). Using (6) we can write down four formally different expressions for j'(z). .::
+ i~.. provided that g(z) does not vanish.. this very typical ease will thus not be included in our considerations.. .)..
.
A function u which satisfies Laplace' 8 equation Au =: 0 is said to be harmon~.."': " .) of two analytic functions. moreover. .... _....
ay
and the existence of z' (t) is equivalent to the simultaneous existence of x'{t) and 'II(t). We shall prove later that the derivative of an analytic function is i~lf analytic. ~
= axl
b' a2v
+ a'u
ayt
a'v
m!E
0
+ ayt = o.0
lim (J(z
I
+ h)
.... the existence of the derivative of 8 complex function of . but it will be clear that the results remain valid except for obvious modifications .::::..~~f!l~~(s:. The reel and imaginary part of an analytic function are thus ~mc.:':.fez) h .
For instance.ive!'(z). as complex functions of two real variables .at·.nd the most general conjugate function of z:' .yl ii' hence z' + ie.ion f(z) can thus be computed by means of the formula
1(1)
==
21£(z/2.
Since J(z) _ only determined up to a purely imaginary constant. however. tend to zero more rapidly than h ik in the sense that al/(h ik) 4 0 and E2/(h ik) 4 0 for 11. be &dded. The funet. U is the conjugate harmonic function of v.v(x. If we substitute x =: 2/2.u(x..
.y) mul .. is analytic .COMPLEX H COMPLEX ANALYSIS
FUNCTIONS
titm of u.. and for this purpose we make the explicit assumption that 14 and v have continuousfirstorder partifl.i~z .0). If u(x. where rp(y) is 8.. isdeflniteiy limited functions u(X". Observe that Zl .y) as a function of z and z which we will treat as independent van abIes (forgetting that they are in fact conjugate to each other). This is not the place to diSCUBS the weakest oonditions of regularity which can be imposed on harmonic functions. With this change of variable we can eonslder. By similar formal arguments we can derive 8 very simple method WblC~ allows us to compute. u = Xl . a. The conju:pte function must therefore satisfy
_e
fun f(z
A+iJ.y' is 2xy e where c is a constant. :1.
".y) of two real variables. We remark first that the conjugate function f(zj has the derivative zero with. is harmonic 'and au/iJx = 2x. + au k av . ~i. under exactly these regularity conI
..fez) =
Gi +
i ~)(h
+ ik) + '1 + iE. Actually...
+ at
+!t.
fP{y) is 8. ~ature ~f analytic functions.
It is reasonable to expect that this is a formal identity. In the same sense. and converaely.t:.... and then even w~n z and y are complex.y) =
vex
+ h. y we obtain . we can wnte x =:= i(z z).y) =
iii 11. With O.
+ i~'
ax
conclude that /(2). an. respect . In this form the in8tIi. we ~y is ~ well assume that 1(0) is real.!I' + 2i:ttI . There is an interesting formal procedure which throws considerable li~t on the.y + k)
. be considered as a function of J. z/2t.tO
+ 11. and in simple cases the computation can be made explicit. a14/ily = 2y. y = j. This formal re~ning supports the point of view that analytic funetlO~ are true functions of a complex variable as opposed to functions whi. . we denote this function by}(z).. We are thus tempted to say that an anal!tlc function IS Independent of it and a. without use of integration. Introducing complex variable z = x + iy and its conjugate z = :c . v is deterrn. .N o.(z/2) z/2$) = IU(z) + J(O)]. With this notation we can write down the identity·
u(x. e.nalytic with oontin1U)U8 derivaf.(fv k
au
aU.iy)]. we would obtain
tID:
+
ditions.ined only up to an additive oonstant. so that the use of the definite article.1'(0):0). It is proved in calculus.i derivatives. If the rules of calculus were applicable. the notation j(z) = u{.to z and may. then I(z) = 1&(z) + iv(z) .::
.
~ !(~ ~ ax·ay _ (Jz 2
4· ~).. We present this procedure with an explicit warnmg to the reader that it is purely formal and does not possess any power of proof .. z'l.. Consider a c?mplex fnnctionf(x. + iJe) h + ik
. the analytic function fez) whose real part is a given harmonic function u(x.. therefore. .
at Ctf) ai=2 1(Of ax+iay'
ax h + ay

These expressions have no convenient definition as limits but we can nevertheless introduce them 88 symbolic derivatives with' respect to z
where the remainders el..y)Whieh
to
_. function of y alone. that we can write
u(x
+ h. + ik)
. although traditional. which implies J(O) = 1'(0.f(z)
=~
ax
and z~ By comparison with (5) we find that analytic functions are characterized ~Y the ~on~t~on aI/iii = O.. The analytic function with the real part z'l.z).(x.. We wish to prove. Substitution in the second equation yields tp'(y) = O.ch are more adequately described.
av
2x.y + k)
.y7..
+
A: pqrely
..od..8 always analytic.
it holds === z/2i
1
ay a3: . The conjugate of a harmonic function can be found by integration..lI) we obtain by virtue of the rela
+
+
+
+
tions (6)
f(z and hence
+ 11. From the first equation v == 2xy + rp(y)..sfy the CauchyRimumn differential equations.y) + iv(X.wiU : c . function of z alone. that the function u + iv detennined by a pair of conjugate harmonic functions .y). is not quite accurate.Y) have CO'Jllin:UO'U3 jiraturder partial clef'ivative$ which .. ik . constant. Henee
:s
:=
av 
2y.ig.f(x.
imaginary ~ . .y)
= j[f(x
+ iy) + l(x
.) .
.. a polynomial of degree n has exactly n zeros.. The order of a zero a can also be determined by consideration of the successive derivatives of P(z) for z == a.. .. If this is true for all k we conclude from (9) that
1m bP'(z)
P(z)
=
11.al) (z' . P'(a) '" o. an are not necessarily distinct. Successive derivation yields P(a) = P'(a) = · .. This IS the socalled fundamental theorem of algebra which we shall prove later. it is shown in elementary algebra that (z : (ll)Pl(Z) where Pl(z) is a polynomial of degree n ....S
1..a)lP1(z) with P1(a) ¢ 0.«&:)1 < 0. 7..
But the Imaginary parts of reciprocal numbers have opposite sign.::' .. 1m b(z .(I.. We find that the sum of the orders of the zeros of a polynomial is equal to its degree.· · + ~
~l<t
is an analytic function.a}/b < 0 (see Chap. the order of a zero equals the order of the first nonvanisbing derivative.28
COMPLEX
ANALYSIS
COMPLEX
FUNCTIONS
21
. Find the most general harmonic polynomial of the form ax' + bx1y + cry2 + dy3.
.. Every constant is an ana1~c fun~tio~ with the derivative o. "..
. I S 2. considered as a pol~omlaI! IS In many respects exceptional and will be excluded from our cOnSlder~t~ons.
From (8) we obtain
(9) P(z)
a'u
1.
"f.~
.
!1
L.:':<':<':::':.):..:''':. Show that lL harmonic function satisfies the formal differential equation
lJz lJi = 0. · • · J a. S.
i
. is ~h~n said to be of degree n. "#..=I b
mb
za_1
al:a>O m b
·
t
P~z?.. • · .
P(z)
P'(z)
=
at
ao +
alZ
+ ..3.It Repetition of thlS process finally leads to a complete factorization
Im ea.. is in H and z is not.. The simplest nonconstant analytIC function 18 Z whose derivative is 1..· +
The notation (7) shall imply that a... ~\~~
.(z ..a.t a is a zero of order h.) • .. 1/ all zero« of a polytWmialP(z) lie in a half plane.)
t For
formal reasons. we have then
==
+ 2atZ + · .«no
1. known &8 Lucas's theorem: Theorem 1. If a. In a sharper formu1ation the theorem teUs us that the smallest convex polygon that contains the .
"1m
Z
bal"
<0
JI
(8)
equal to 
PCz) :::: a. .. Show that an analytic function cannot have a constant absolute value without reducing to a constant. L Prove that the functions u(z) and' u(!) are simultaneously harmonic. end the polynomial..
analytic. Polynomials.
where the (Zlt a2.. 2.' S. 'I
". From the factorization we conclude that P(z) does not vanish for any value of 2 different from aI. Suppose tha.: . The constant 0.0.. 4.. Sec.aeroa.: ':..} also contains the zeros of P'(z). If g(w) and f{z) are analytic functions.3). Therefore) under the same assumption... · (z .. for the function must have a meaning for complex values of the argument. Since the sum and product of two analytic functions are ' again analytic) it follows that every polynomial
(7)
P' (z) =
Z
1
at
+.:. . Prove rigorously that the funCtiODSJ(Z) andf(j) are simultaneously analytic. . More simply.. As an application we shall prove the following theorem. if each zero is counted as many times as its order indicates.
Its derivative is
Suppose that the half plane H is defined as the part of the plane where Im (z .
. Moreover. of P(.:.. A zero of order 1 is called 8. show
that gU(z)) is also
. If exactly h of the aj coincide.. In other words. :. their common value is called a zero of P(z) of the order h.
are rational in x and y. .:
.. If Pial) = 0. at.. Verify CauchyRiemann's equations for the functions z and z .
~
.
as a PO~
its degree is set
and consequently P'(z) ~ 0. then all zeroe oj the derivative pi (a) lie in the same half plane. 1. +
z .::>''. simple zero and is characterized by the conditions Pea} = 0./.. if the constant 0 is ~
00....
.
EXERCISE. Determine the conjugate harmonic function and the corresponding analytic function by integration and by the formal method. = P(4l)(a) = 0 while P(lI}(a) '" O. the factorization is uniquely determined except for the order of the factors. Then we can write P(z) = (z . For n > 0 the equation P(z) == 0 has at least one root. Suffice it to say that the method ean be extended to the general case and that 8 complete justification can be • given.
..0...~:'"
"I'{~
. We may define R( 00) as the limit of R{z) as z + co.tional.+ · · · + a....by definition equal to the order of the corresponding zero of Q(z). .~ I'~~~'... . zero or pole at CIC. ..~.
. However...+~b.
.. .'
. balf" + ba. belongs either to the numerator or to the denominator. 18 equal to the .mJ and if m == R( 00) :: a.n at c().. which we can rewrite as a rational function R1(z).. Aceordingly... if m > n R(z) has IL zero of order m ..e) at co Let the distinct finite poles of R(z) be denoted by Pl.by one.
I~:'_'. 4
(to)
We turn to the case of
8
rational function
We can now count the total number of zeros and polesin the extended plane.
. R(z) will be given· the value 00 at the zeros of Q(z)t It must therefore be considered as a function with values in the extended plane. :
:.
If R1{O) = 0 or ac. greater of the numbers m and n." .a are roots of the equation R(z) = a./b... The linear transfonnation 2 + a is called a parallel trcnBlaticm and liz is an inversion.. it should be noticed that the expression (11) does not appear in reduced
Such fractions."I'"
_ . .zao
I
==
G(z)
+ H(z)
we obtain
R1{a) ==
!lli'
+ atZ. the function R(z) _.. and we find indeed
with a8 . and this is essential. the order of the zero or pole at order of the ·zero or pole of Rl{Z) at the origin. and if the roots are counted as many times as the order of the zero indicates. ". or linear transformations will be studied at length in Chap. Greater unity is sehieved if we let the variable z as well as the values R(z) lange over the extended plane.
. . We ca..zn.. The former has a fixed point at GO. including those at DO.. This common number of zeros and poles is called the ord6r of
given as the quotient of two polynomials. de~ of G(z) is the order of the pole at co J and the polynomial G(z) 18 called the singular part of R(.': ~~.t! 0. and as such it is continuous..: :'" I... and H(z) is finite at co..
TID
aw 
+a
fJ •
8 and 81 are inverse to each other..
'.~:' .. the order
of each pole being increased. I ·.. and the order of 8 pole is. and consequently the same order... r
.
. In case Q(z) has multiple zeros. :..R(6~'+:f)~_GIG}+Hi(r)lI':. :.. The zeros of Q(z) are called poles of R(z).. The derivative
(11)
the rational function.. Rational Functions. In order to derive this representation we assume first that R{z) has a pole at co..
of the decomposition
< 11.+ b.rry out the division of P(z) by Q(z) until the degree of the remainder is at most equal to that of the denominator..:j:
.. 3.~ : ·~_:t.. and set
R( ~) = Rl(O). The result can be written in the form
(12) R(z)
With the notation
R{z)
+ atZ + · · + ~ == ~bo~+~b""').fJ~ :F
o.
..fJ ••
if
where the power Z..... if
m
The function R By
UBe
("J + i) is
IL
n.L~ . F."
.
. The· number of poles is the same.
. 3.
R
")
\Z
_ P'(z)Q(z) . R'(z) has the same poles as R(z).+ · · . . . but this definition would not determine the order of a. as arational function defined by the righthand member of (11).~
.
at¢'
1 I
where G(z} is a polynomial without constant term. The zeros of R(z} .QI(Z)P(Z)

Q{Z)2
+ fJ
+
3
exists only when Q(z) :. • ..Z+~"'. a has the same poles as R(z). 00
9
1'l
._. fJl..
. The count shows that the number of zeros. the point
at
00
is a pole of ordern . that P(z) and Q(z) have no common factors and hence no common sercs. the latter interchanges 0 and
CD. We assume..:.
:.
C()
is defined as the
Every rational function has a representation by partialfradiorut.::_ :. For the moment we noie merely that the equation 10 = S(z) has exactly one root.If·. ~. . I t is therefore preferable to consider the function R(l/z). function of
(12) we can write
r with
IL
pole at
r=
QO. we can state the following result = A rational function R(z) oj ortkr p haa p zeros and p poles and every equation R(z) = a has e:tadly P TOOts... J A rational function of order 1 is 8 linear fraction
S(z) ~ az
~t...
. Sec..
:':...
COMPLEX
ANALY81S
COMPLEX
FUNCTIONS
31
1. If a is any constant.:.
z
The transformations
=
81(UI} =
form.
32
COMPLEX
ANALYSIS
COMPLEX
FUNCTIONS
33
or with a cbange of variable
R(z)
Here
=
Gj~
~
(3) + Hi(z ~ fJ}
in z ~ ~ without eonstant term.a.:I:~('. 2.
..1 < 114+ law. .1~~~~:. _
:."!"\:i:~:~ ~~ . 3). . ......
'11:1'
. A sequence will be called Jufl4amental. .." . the sum of a convergent series is such & limit.
.. and the same
This is a rational function which cannot have other poles than ~~._.. What is the general form of a rational function which has absolute value 1 on the circle lzl == 11 In particular.._. . 80 is the sum..
~
s: ..
•
I
1.. ...
"p p
. In this section we study only the most eJeme?tary properties of power series.. how are the zeros and poles
Gj(~ ~ (3) is a polynomial
Consider now the expression
related to each other? 5....~~~. .. T~e test that serves this purpose bears the name of Cauchy. and if the terms are analytic funetlOD.~ . chances are good that the sum will also be analytic....
. O! all series wi~h analytic tenns the. spd. Sequences. (jtJ .
~
e. ..~.. ..:
'. . .1. At z = ~j we find that the two terms which become infinite have a difference Hi (z ~
(3) with afinite
is true at a).. functions are very special analytic functions.. However} it is only with the introduction of complex numbers that it becomes completely successful..
I
"I
. !he easiest way to achieve greater variety is to form limits.'
.... .. + A we can find nG such that Ia.( •
"
• .. fj'l and 00. . .._
.a.o~toxie . . If a. .£ for n ~ nG.
{... .~:
i
I
I .AI < e/2 for n i. one ..
• 
..
.8.
. a._
.m.. Only in rare eases can the convergence be proved by exhibiting the limit so it is extremely important to make use of a method that permits proof of the existence of a limit even when it cannot be determined explicitly.7. power series with complex eoeffleiente are the simplest..the 8UfIi~y. ... '_... . . Use the formul& in the preceding ·exercise to prove that there exists a unique polynomial P of degree < 1l with given values CI: at the points «l (Lagrange's interpolation polynomial) .
I
.
~:. if it satisfies the following condition: given any e > 0 there exists an no such that Ia..11 ~ no it follows by the tria....~ .'
... how large and how small
for
z = fJj~
(13)
THEORY
OF POWER SERIES
R(z) .G(z) 
it!
OJ
(z ~ fJi)
limit. no.
... ..
I .. and if this constant is absorbed in G(z) we obtain
(14) R(z)
= G(z} +
itt
Polyno~a.'' ._
r.. The test reads: A~ i8 convergent if and rmlll if it ill a Cauchy sequence. < E+ The sufficiency is closely connected with the definition of real num.(+ 2)·
.~.
"..
.. .1Iowever we wish to use 1 the ..ngle ineqllality that la. 
.. J ... or a Cauchy ~. finite limit is said to be corwergent} and any sequence which does not converge is divergent If li~.. ...!\ ."ay in which real numbers ~8D be introduced is indeed to ~~~. For instance..'
":=
s... .
_ .
"
• I
"
"j
:"
<. liIiVIoonce :of real numbeD haI~ JimitA ..bers. ~ :I'_f". .. .
This representation is well known from the calculus where it is used as a technical device in integration theory. A strong motivation for taking up this....1 . Use
the method of the text to develop
zt
%3 _
1
and
z(z
+
1 1)'(.t
J
...·~ . : .1 < I whenever n ~ 110 and m ~ nO.
.. eyery bouaded ..
Gj
(z ~ . called
The function
. == 00 7 the sequence may be said to diverge to infinity... .' I .
~. .I8 and rational. . ~ Oeuchy's eondition.. ...::. . 1 ~ 'i '_
. If the terms are f?nctioDs of a variable.
EXERCISES
l. .~.. show that
•
oil
ill
and if P is a
A'
A.._
•
. If a rational function is real on IzI situated?
can the order of R'(z) be?
:L ELEMENTARY
=
1.:: ..).:.... .. .. (z ~ fJ) is finite
6. The sequence {4t. .
'
A....: ('...
~:/". ..:. The necessity is immediate. .~ .. how are the zeros and poles
the singular part of R(z} at fJi'
s.: .
~. If Q is a polynomial with distinct roots at.
in partial fractions. .. .. study when we are not yet equipped to prove the most general properties (those that depend on integration) is that we need power series to construct the exponential function (Sec .. CIa.li has the limit A if to every E > 0 there exists an no such that Ja. .~: . :.... For tn.. .
:. 2.• ~ ~ I ~ '. i.
. If R(z) is 8 rational function of order n.. A sequence with a. A rational function without poles must reduce to 8 constant...:~j. polynomial of degree < n.A\ < . Therefore (13) has neither any finite poles nor a pole at ao."..l.
. If A is finite and e: > 0 there exists an no such that A. no. ~ the sequenee of values If...'
:. . > A . .e. The series (15) converges if and only if to every e > 0 there exists an no such that la.+l + · · · 4n+p 1 < e.. .....+
Itis easy to characterize the limes superior by its properties. contrary to the choi~ .
A
2.(%)} converges~orleverY. . I t is quite clear that the limes inferior and limes superior will be equal if and only if the sequence" converges to a finite limit or diverges to + 00 or to . convergence of (16) implies that the original series (15) is convergent .. a. and we denote its limit by A..co • In all cases there cannot be more than one number A with
(. then the limitj(x) is spin a function tiit'R~~~<~ BY·:definition. A very simple application of Cauchy's condition permits us to deduce the convergence of one sequence from that of another. . + tt. . In the opposite direction. hence it has a limit At which is finite or equal to + 00 • The number Al is known as the leas: upper bound or supremum (l.. a..00. It is clear that {At} is 8 nonincreasing sequence. and if this is the case the limit of the sequence is the BUm of
+
Ji•

(16)
lall + 1a21 + · · ... of s... and if they converge. Under this condition.1 for all pairs of subscripts. .. In the case of convergenee.u... indeed.. It may be finite. beIonp to E there . :
_. For p == 0 we find in particular that lanl < E.. Consider a sequence of funetions I.1~ Ill. on z..· ·
~+1
formed by the absolute values of the terms.i(2i)1 < ~for n ~ At.... ~ . The notations are frequently simplified to Rm and lim. then Alia ~ A .tl .} implies convergence of {b. tIw+.a == (A .fjnedon the 88me set E... there are arbitrarily large n for which aD > A .+ll + · ~· + l~'... if tlA :::: A .I + ..ence. is the greatest of the numbers CIt..o such that '1.} Hence convergence of {a. so does the original sequence. . It follows that A .·~dt.
+
R)
+
2.. . The reader Mould prove the following relations:
the series.2. < A + g for n S.. .
..
~
(A . Then the sum of the whole series is 8 = B" + R The series (15) can be compared with the series
The series is said to converge if and only if the corresponding sequence is convergent. the sequence {b~l may be termed a c~~ of the sequence Ia.
Ia. and it follows that A == lim an and a = lim a. We use the opportunity to recall the notions of lime« superior and lime8 in/eriar.. .~+ a.. . .} i is nondecreasing. UnJjatm eo".} i we shal] set a == max lai. for sll 11.8...b...
....
these properties . < A". .am) (a.a) < 3e.:. or .. If A = + there are arbitrarily large a and A = . that is.· · + G.ee if and only if a.. . Construct in the same way the least upper bound AI: of the sequence {anI: obtained from the original sequence by deleting· a1... or wp) of the numbers ~. a..(x). +
+ ~. :"
.. the new series converges or diverges together with (15).a. Given a real sequence {a. The sequence {a.exlsts an·n. are both finite.C A..
I' :.. ~ no and p ~ O. ) a
and determine
corresponding no. I. Series.
4>
••
. + 00.
+ as + .. From I~ ....b. ' If a finite number of the terms of the series (15) are omitted. which is impossible. The sequence of partial ·sums of (15) is 8 contraction of the sequence corresponding to (16). In other words. Hence the general term of a convergent series tends to zero.. A series with the property that the series formed by the absolute values of the terms converges is said to be absolutely convergent .
A == lim sup ".
34
COMPLEX
ANALYSIS
COMPLEX
FUNCTIONS
35
The real and imaginary parts of a Cauchy sequence are again Cauchy sequences.O'fatl < E we obtain tanl < laaoJ + e for n ~ no. tends to ..
Now we return to the sufficiency of Cauchy's condition.
8
3 By definition of a and A there exists
.}.. but no is allowed to depend.
. This condition is necessary but of course not sufficient.+ la. for
. so is {b..(z) . If it is true that lb.
Associated with this series is the sequence of its partial sums
8.
.... . it is the least number which is s:: all a...e with m.+lt'
~
14tt1 + ia. An infinite series is a formal infinite sum
t
(15)
a1
+ at + · . be the sum of the series which begins with the term. Applied to & series Cauchy's convergence test yields the following condition. for 'fir ~ no. If a ¢ A choose
e=~
fore. Hence a :::s A J and the sequence converges. +
==
at
•••
.a (ata. .. The limes inferior can be defined in the same manner with ioequalitiesreversed. .a.z that bilongs to E.· + a..._ For this reason we need to prove the sufficiency only for real sequences. if {an} is a Cauchy sequence.n ~ nQ.. .1..} (this is not a standard term).00 ~ In any case we write
(t. if e > 0 and i. an a" < a + e and an a. let R. < A + Et and it follows that aft.
jfl(X} \ < a we can keep n fixed and let m tend to 00 • It follows that 'f{x) . However.:....:..
.: . We say that a series with variable terms
Junctions is itself continuous. in most eases it will be found that the convergence is uniform only on a part of the set on which the functions are originally defined. If U = Ul u. The necessity is again trivial. .
+
+
+
+
. 'I:'he.. Show that the sum of an absolutely
convergent
for real values of x.
for 8 majorant if it is true that I!.(x) I ~ I: for n ~ no and 'all x in~.
<.(:1. conversely.. Try to arrange the proof so economically that the absolute convergence of the second series is not needed . ~ Suppose that the functions f.0+.
.+l + · · · + 4.1(x) I < E for aU n ~ no and all x in E.
..·
al
the series with positive terms
+ a2 + .....
: . and the. . Utili) (U1Va + u. We say in this situation that the sequence converges pointwise.
. prove that UV =: UtVl (UIV..f. For any E > 0 we are able to find an n such that
for all z.
i·
.../(x) 1 < e/3 for all x in E~ Let Xo be a point in E.)..t~~'~~. . in a somewhat weaker form .(x)) converges uniformly to I(x) on the set E if to every I > 0 there exists an no BUCk thal If. V = Vi + Vr·+ • .!Jt(z)' s 1a.[.Cauchy's necessary and sufficient condition has a counterpart for uniform convergence. but in order to have 1(1 + l/n)x ......• ::.
:rr. prove that
:: If(x)
.)1 +
.z:
elusion follows immediately by Cauchy's condition.. ~ no and aU zin E.v"2 11:VI) + · · . ).(x) I ~ M ~ for some constant M and for all sufficiently large n.{x) . : " :
.!(Xc) .
EXERCISES L Prove that a convergent sequence is bounded ..a + ·
t
. s.Xo I < 8.
+ a.38
COIIPLEX
ANALYSIS
COMPLEX
FUNCTIONS
37
For instance.(x)
+ /"+I(x) + · · · +/..1.on E i. It has the slight weakness that it applies only to series which are also absolutely convergent. &.In the case of series this criterion.f. it is true that
lim
11"""'+_
(1 + !).
series does not of
change if the terms are rearranged . > 0 there exists an no such that 11.(z)l
+
IJ...1 . . .(%) . + a. becomes . For the sufficiency we remark that the limit function f(x) exists by the ordinary form of Cauchy's test.
_
_
~
.
.(x) .. In the theory of analytic functions we shall find uniform convergence much more important than pointwise convergence.:
.(x) I is uniformly convergent. :: .(2:) .. This condition is frequently referred to as the Weierstrass M test. then the sequence (f..
.(xo)
.
... 2.·?~:. partlcularly simple. (It is easy if both series are absolutely convergent. We assert: The sequence II..._.
1/. Discuss the uniform convergence of the series
3....1_ " . Under the same condition on x it follows that
If(x) ... In the inequality If.. Because JII (x) is continuous at XI) we can find a > 0 such that 1/.. In Positive formulation: The sequence {f. Discuss completely the convergence and uniform convergence the sequence Inzn}r..l on E. .(x) . The most important consequence of uniform convergence is the following: The limit function of a uniformly convergent sequence of continu~
flex)
has
+ J't(x) + · · · + fn{X) + · .:. If lim Zn = A.
.. 4.: _..(x) 1 is a contract_ion of a convergent sequence of constants (a.......f1..
<
2.!m(X)' < afar alltn.4. .. Such an no exists for every fixed x.. For practical use the following test is the most applicable: If a sequence of functions tf.!ft(xo) I < E/3 for all x in E with Ix . _ .(x) I eonverges uniformly.(x) ._.(x) . the minorant converges uniformly..xi == lxl/n < e for n ~ na it is necessary that no > lxi/I!.. hypothesis means ~hat \I.f(xl')l
Therefore} if the majorant converges. M(a . .f and only if to every F. The general principle of contraction is more complicated. provided that at least one of the series is absolutely convergent. Hence the convergence is uniform.z
n
::::::=.. the first series is a minorant of the second.
.'J:
..
and we have proved that J(z) is continuous at Xu.. )
+ + . but the requirement cannot be met simultaneously for all x. but not uniformly. 'IIjII
lim n (21 1
+ 21 + · · · + ZII)
=
A. are convergent series. con. In these circumstances we have
If..(x) are continuous and tend uniformly to f(x) on the aet E.(xo)~
+ 1f.
.· · .{ . but has a wider range of applicability.
"
Proof.anl
110
1
== lim
R41a
8~(Z)~
(iii) In
lzl < R
. or that it converges for all values of e.z) for lz' < 1... Ia..8ntly divergent. I > 1/p~..· ':_r"\::. we shall find that every power series converges inside 8 circle and diverges outside the same circle. Set Vn of the binomial theorem n = (1 + cl! < 2/n. .. and by use In(n . 0 ~ R ~ co.
A little more
where the coefficients a.
We have to show that J'(z) Consider the identity
(19) l(lI! .zI < (lzl/p)·forJargen.·.+ · · · + ~z~ + .(. we shall prove the following theorem due to _~bel: 2...
I
+ · · . =L
and also /1(%)
.(z)
where BIt(Z)
=
ao +
41Z
+ . radius of convergencet with the following propertie«: (i) ... We ~h811show that the assertions in the theorem are true if R is chosen according to the formula
(18)

Z
1l~(tD»).
atZ " . p < R the convergence is uniform [or lzl ~ p.+
lz=
lz
Since z/' + 0 for \zj < 1 and IZ"I ~ 1 for It 1 ~ 1 we conclude that the geometric series converges to 1/(1 ... Thus 1aazlt I > (Izl/p)ft for infinitely many n..
Zttl
This is known as Hadamard's formula for the radius of convergence. (ii) If Izi > R the terms of the serie« are unbounded.:.
.. Then IIp> l/RJ and by the definition of limes superior there exists an no such that la..R is called the circle of cmwugenc6.
~:.~(Zfl) /1('0) •
= il(Z) ..:_.r
. manner . called the.. because
n«lnZftl has the same radius of convergence.... For every power series (17) there exi8t8 a number R.etric aeries
l+z+z'+···
whosepartial 1+ z
+~+ ..The series converge8 absolutely for every z with lzl < R~ If 0 .
'"
:
'"
""
. Indeed.=.
The last tenn can be
lIB
=
.. P01. It turns out that the behavior of the geometric series is typical . .
CIO
sums can be written in the form
The derived series.
.. This gives
L anZ"
Q
•
=
8~(Z)
+ Rn(z)
R.
L na..I .. .
..11/it < 1/p.(16" Series. H Izi > R we choose p so that R < p < ~zl.....'
'"
.
=
. and the terms are unbounded...
an).8~(Zfl») + (8~(Zo) .
1
lim sup
vra:l. ['ol < p < R..+ 4.(%) and
z
:2'0
8.4._lzttl.l < l/p"forn ~ no.:.~~. AB an almost trivial example we consider the geom.5. diverges for Izl ~ 1.a and thevariable z are complex. For lzl < R we shall write
J(z)
=
Vn 4 1.. (Xc)
. Ia...
""
..
=
(8.· ~+:Jo * + z!t) '.. and the aeries is
Theorem
cO'n8et]tt. but the difference is so slight that we need not do so in a formal.
""
.. Since lip < l/R there are arbitrarily large n such that 1a. nothing is claimed about the convergence on the circle. Itfollowsthatla.11
..
the sum of the serie8 is an analytic fumtion..
'"
. and is consequently convergent..'·:.
~IZO
+ .
.. More precisely..ll/fl > 1/ p.../1(")
+ (Rtl(Z)
where we 888UJl1e that reWritten 88

The circle lz' ==.1·+
i. except that it may happen that the aeries converges only for z = 0... To prove the uniform convergence for Izl ~ p < R we choose a p' with p < pi < R and find la. so that the power series (17) has a convergent geometric series as a maiorent....31
COMPLEX
ANALYSIS
COMPLEX
FUNCTtONS
31
2.i~%·. Since the maj orant is convergent and has constant terms we conclude by Weierstrass's M test that the power series is uniformly convergent .26
z~
%0
·Itl..:. The derivative can be obtained by termwiBe differentiation. If Izi < R we can find p so that lzl < p < R.. and hence 6" + o..l)lJ!. generally we may consider series
which are power series with respect to the center Zo.
1+
Then 0" > 0. and the derived series has the same radius of convergence. .':
:_.>
=
1
+ 8".
(17)
au
A power series is of the form
+ a:tZ + aza2.znl ~ (P!p')'" for n ~ no...
Same question for
il(.:..!l(ZO)\ < c/3 for n ~ nl.2'" is at least Rill"....f 1 Za ) ( '~_..
. :.
•
..::. Limit Theorem.:~(... Find the radius of convergence
of the following power series:
q*..:
m a positive integer.iWt~~:····~ .
•
.
1:
zn(1ql
<
1).1) of the real axis. if we look at the last line we see that air the power series becomes
f(It)(O)/k!.~ z .
~::
.....J.l
R".. 111(z) == 2«2 6alZ 12aa' + · . but the main part is still missing.
•
._ ~:_ ..s.. What
is the radius
The expression on the right is the remainder term in a convergent series. By the definition of derivative we can find o > 0 such that 0 < Iz < a implies
zo'
I
when 0
s...z)..': .. namely that every analytic function has a Taylor development.: ...5 ..
2.
.. but we have proved it only under the assumption that !(z) has a power series development. .·
+
....Zo
e coDvergent? ...
..00 ....
. ...a+ l~ = R....~z..\
I<
<
3
. There is also an nl such that '8~(Zo) ....(zo) ~~...z.
L For what values of
z is
f(z) ~ f(z. n]. There is a second theorem of Abel's which refers to the case where a power series converges at a point of the circle of convergence... symmetrically to the part (..
(18
Z
11 
zl/(1 
1z1) remains
n!
+ .) .
Geometrically....(zo) 1
z%0
< _!
3
L nl'z".zal
<
8...~·
Remark~
This is the familiar TaylorMaclawin development. and they are given explicitly by
I(z) == an + atZ + 4zZ' + ..
+
.. .
•
•
•
.
of
1 in powers of
3
1.'
I tit.. what is the radius of convergence of ~a.. 1:
~1
~
n!Z"... 1a I /\a.
. have radii of convergence ·R1 and R:2t show that the radius of convergence of I~b.
. We do know that the development is uniquely determined.
..
We have proved that !(zo) exists and equals
Since the reasoning can be repeated we have in reality proved much more: A power series with positive radius of convergence has derivatives of all orders. Expand ~ .. I! 2!·
:=
2.
t . We lose no generality by assuming that R == 1 and that the convergence takes place at z = 1. if it exists...~
Z
Zo

8
I(
"
ZO)
...has radius of conGO R
When all these inequalities
are combined it follows by (19) that
vergence
R.az "ttl? of l..R...: 7:··:..
II
•
•
•
•
.. ~ o·
•
OJ f~ this ~~ be &~t&inedby adding
..:~~.....
.z"
0
•
tend8 to f(l) bounded . (z) .:.
•
j(I)(Z)
=
k1al:
·
+ (k + 1) I aoHlZ + (k + 2)! aoH2Z2 + . It is customary to say that the approach takes place in a Stolz angle. Choose 8 fixed n ::: nfl. prove that 2':a.
In particular.~:...
If
J(z) == 1(0)
+ J'(O)z +
!"{O)
2!
Z2
+ .(z) .
~ If Zasz" has radius of convergence R.
2: zl
for n ~ nO...
<
[z . '..40
COIIPLEX
ANALYSIS
COMPLEX
FUNCTIONS
41
and we conclude that
EXERCISES
L Expand (1 .
ProoJta
We may
aSsume~~~.· +
4
/(·)(0)
approaches 1in BUCk a way that
z~
L~ o
•
convergeBj then /(s)
= ~ o.
.
•
•
•
•
. Abel'.. If limn .: ~. in
2
powers of
z....
Hence we can find no such that
convergence? 3.
.. what is Zn'a_z"? L If l:a~ and ~bgZn.....~· f(z) == al + 2U2Z + 3032' + · ~...
and
TheoreDl
3.
.
. ..
.. ..
•
. the condition means ~hat z stays in an angle < 180° with vertex 1. 7.a~z!L? " If j(z) = ::z... .
( . We write SA = Go the identity (summation by parts)
8~(Z) == ao + atZ + = 80(1 .
.
. The value of the constant is found by setting z == 0..yt.i1+ ~ (proof by the reader) .
.":
.). and the real version of the theorem shows that f is constant on every horizontal and every vertieal line. these functions seem to be derived from completely different sources and with different purposea in mind.
.z$ + . This is certainly 80 iff is defined in the whole plane.z) + 81(Z = (1 . and since ~ and ~ are reciprocals..+ na~l + · · ·
But 8nt" + 0.. and if willing to use imaginary arguments he will be able to derive Euler's formula eiz = cos x i sin x as a formal identity.." .
+
depth.
. :
.~. . ~ With the preparation given in the preceding section it will be easy to define e&'.' .
.
I • .
'
.
"

"
I
.
v.
all'
If (20) is to be satisfied.... is then dominated by the geometric series
E
e'
= 1 + I! + 21 + · · · + n! + · · ·
Z
Zl
2ft
L Izln = elzl/(1 CID
Izl)
< a/(l ~
:i
. and
make use of
Sn_l)Ztl
(20)
JI (z) == f(z)
== 1. and to derive the relations between 'these functions.i8 never zero.
v
~...
converges. and (22) foI.
I..
3. For real x the series development (21) shows that e#t > 1 for x > 0.the logarithm leads in tum to the correct definition of the argument of a complex number. and .:~:..a:. and hence to the nongeometrie definition of angle...
t
The person who approaches calculus exclusively from the point of view of real numbers will not expect any relationship between the exponential function e and the trigonometric functions cos x sud sin z.... We solve it by setting
f(z) f'(z)
= Bo =
+
· · · + «nzn = 80 + (81 . no doubt.oJ
. At the same time we can define theIogsrithm as the inverse function of the exponential. For iff = u iv. . for .~1+21
~.._. r.... Hence leiflll = eli' ~rill = 1J and \eriiwl = e
S•
..· · + 8 _lz»l) + 8$'.~
.· .
. we must have «n1 = nan.:~.22:
z .
"I •
.
D
with the initial value 1(0)
at
+ at! + · · · + ~zt' + ~.
+
8. so we obtain the representation
.
fez)
=
(1 . : \ ~
~.'. .
(.z)
! 8nZ"~ o.lzl).... The Trigonometric Functions.... + 2~ + ..'~'
.
4!
.. We may begin by defining the exponential fU1lCtion as the solution of the differential equation
!Jill. and that 8" ~ o. and the initial condition gives ao = 1" It follows by induction that ~ = lInt The solution is denoted by ~ or exp z} depending on purely typographical considerations.Z2) + .zt') + 8".a) (80 8IZ
+ a1 + · .· + (S3 ...~~ .'. But it took the genius of a Gauss to analyze its full
S
Remark .. Indeed. ..z. h. ~ a particular C88e of the addition theorem e· r' ::::: This shows 1..··
.zl ~ K(l . · + 8'Q~1(Z"l . from n = m on..1
It follows that
8~i
lJ(z)l
11 
z\
1+ Ke.
~
~ .
we obtain ~ AX iJy = ax iJy .' ~
..~
.....::
•
. R similarity between the Taylor developments of these functions.'.. He will notice. We must show of course that the series
(21)
We are assuming that \1 . Exponential.
.....2. It is a consequence of the differential equation that e~ satisfies the addition theorem
(22)
The first term on the right can be made arbitrarily small by choosing z sufficiently close to It and we conclude thatf(z) ~ 0 when z + 1 subject to the stated restriction.'_ ••• • ~:.
"":
. defined by
(23) Substitution
The trigonometric
functions are
in (21) shows that they have the series developments
co8.:
.
The... It does so in the whole plane. c = a + b. that e..BO}Z + .. say. Hence e ~ ·e= is a constant. au
au =
av
av = 0 = .
.'
_.42
8
COMPLEX
ANALYSIS
COMPLEX
FUNCTIONS
43
constant to ao. ec~ == «.. The remainder of the series ~B~..~:~~":'
~
~. The fact that the series has real coefficients shows that exp j is the complex conjugate of exp 2.... cos Z and sin z for complex z. e*) == O.lows for z == a. .l · er= + e· ... ~..
IL o
"..' .
....
.l < s for n ~ m. Choose m so large that \stt.
.. THE EXPONENTIAL AND TRIGONOMETRIC FUNCTIONS
we find that D(e= rr~) = e. We conclude that et.~.
. We have used the fact thatf(z) is constant iff(z) is identically zero. Indeed.~~.. 0 < et: < 1 for x < 0..+ an.
. namely from lover i to 1 and back over i to 1.~. I • •
of·~.sati sfies e+~ == 6..~~~if.iodicity. and formulas for cosh 2z......
efZ
1=
008 Z
+ i sin
z 1 .. every period must be an integral multiple of Wo. Use the addition formulas to separate cos (x + iy).sin y > y and cos 0 == 1 gives cos y > 1 y'J/2. When 11 increases from 0 to 2r. betweenfhe additive group of real numbers and the multiplicative'~~group .
+

r.). Express them through cos iz.
as well
88
the identity cost
Z
+ sinS! z =
for sJI z.1.
We _ythat/(z) has. :F 1.. . We denote it by "'0.value 1. = 1.ift(.
en
=
1 .eY
e
U
.nWa
would be a positive period < 6)0.Qf"lJIt'Ii8BJ. ·From ··an algebraic point of view the mapping w·= eiv establishes 8 t. the point 10 = ei..'
.ez)/2. and hence sin y < sin 1JI = 1. Therefore eW". tan (1 + a.th~p6riDd c.
'~
L_
". There exists an integer n such.. Thus 8 period of e.
Observe that all the trigonometric
EXERCISES
.. cos i. Actually) it is the smallest positive period... • •
For real z they reduce to the familiar Taylor developments of cos x and sin x._ 1_" . sinh 2z. TheP.. then w .sin z.
The addition formulas cos (a sin (a
+ b) + b)
= cos a cos b .+ cos oX = cosh'lI
ll
....
functions are rational functions of eH ... eU
we have sin yo == + 1.. We find for instance
tan z == . period we choose the following: From D sin Y = cos y :s: 1 and sin 0 == 0 we obtain sin y < y for y > 0.. The double inequality 0 < sin y < 1 guarantees that ei~ is neither ± 1 nor ±i". Since this is not possible. 2. sin iZt Derive the addition formulas.
•· . In the same way D cos y = .
• • •
mnltipea. which in turn leads to sin y > y .COMPLEX
ANALYSIS
COMPLEX
FUNCTIONS
41
Sln z
~
==
z
3!
21
+ 51 
Z~
.
!
(cosh
21/ 
These equations demonstrate the intimate relationship between the numbers e and .~
.
D sin z = cos
2..y'/6 'and finally to cos Y < 1 y'J/2 + U'/24. sec z... = coshZ 11 
cos!
X
=
. Consider now an arbitrary period w.. r. Because cos! yo
It follows likewise that D
C08 Z
+ sin'! yo ==
1
= . either by integration or by use of the meanvalue theorem..WI"'rI. between 0 and 7:/2...·~
.
The smallest positive period of fill is denoted by 2r~ In the course of the proof we have shown that
e'riJ2
L Find the values of sin i. with the significant di1ferenee that we have now redefined these functions without use of geometry. +. From (23) we obtain fwther Euler's formula
"... or et:: ::: 1.. and 4yo is indeed the smallest positive period. :. and therefore there is a yo between 0 and v'3 with cos Yo = O... It follows that c = iw with real 6Jj we prefer to say that til is a period of eU~ We shall show that there are periods.. The other trigonometric functions tan z. Show that
[cos zl2 = sinh! 11
=
i.. This inequality shows that cos v'3 < 0. which shows that cos y is strictly decreasing. 3..J~:::~. and that they are all integral multiples of & positive period Wo.. They are defined in terms of cos z and sin z in the customary manner.::.:
.~?i~~
":~:'+. Of the many ways to prove the existence of a.y2j6) > y/2 > 0.. and hence e4iw. .
•
and
[sin zIt
= sinh111
+ Binl!:r. sin (x + iy) in real and imaginary parts.sin'll oX
= ~ (cosh 2y
+ cos 2z)
cos 23:).3..~.." that is... Because sin y is positive and cos' Y + sin! 'Y == 1 it follows that sin y is strictly increasing. cosec z are of secondary importance.t. 6''' ::::± i. that is..
.~''.·:bQJb~. To see this.'c) ~ J~) '. take 0 < y < YfJ~ Then sin y > y(l ..... descri bes the unit circle Iwl == 1 in the positive sense.'~:.:j_~)~:'f~~~~i..sin a sin b = cos a sin b + sin a C03 b
are direct consequences of (23) and the addition theorem for the exponential function.f~wd·
by all"integral
8. For every w with Iwl = 1 there is one and only one y from the halfopen interval 0 S Y < 21r such that w ::= 8'''4 All this follows readily from the established fact that cos y is strictly decreasing in the "first quadrant. The hyperbolic cosine and sine are defined by cosh z == (e' + e·)/2t sinh z = (et . that nwo ~ w < (n + l)wG" If w were not equal to ~G. The
:.!\. ~~ . cot z.
. We have shown that 4yo is a period.eGinples numbers with absolute. : .
~~.of·.."•..
. since ~ is always ~O.~..== w.I:'. If we want to compare & value on the left with a value on the right.
We can also write these values in the form arc cos w .
.".i:~
~
~
... .J ~.:
.
.. .:III:
.. By definition. We see that every complex number other than 0 has injinif£ly many logarithms which differ from each other by multiples of 2ri. 1 + 2i? s.: \ . between the positive real axis and the haIf line from 0 through the point w. :'.t~y.::: i log (w +
+ V' WI

1)~
for w + v'w2 . The righthand member of the second equation (24) is a complex number of absolute value 1.".... Therefore. ii..) = log at arg (:1%2) = arg %1
+ log z. If a is restricted to positive numbers. .. then ~ has exactly q values and can be represented as W.. it has one and only one solution in the interval 0 S Y < 2.
11'"
With a change of notation. 7. ..
.~ '/'.
. :
. as we have just seen.. The imaginary part of log 10 is also called the argument of w.1 and w .:. This notation is so convenient that it is used constantly. _:
. 1. The infinitely many values of are cos w reflect the evenness and periodicity of eos z. First of all. By convention the logarithm...:. The addition theorem of the exponential function . the logarithm. 1.i. ~~..* sides and rmll~)~
. 3' n. 2. and
_
==
1D
+ .
EXERCISES 1.
+ arg z"
but only in the sense that both 'sides represent the same infinite set of complex numbers..'2' 4 Tt. if \z' = r and arg z == 9J then z ::= rei'..
z = arc cos 1D = i
±
vw
2
1).i.it is also satisfied by all y that differ from this solution by an integral multiple of 2r. ~. The Logorithm. log a shall be real. and it is interpreted geometrically as the angle.are cos 1D.. Sec.) . i/2.COIIP~X
ANALYSIS
COMPLEX
FUNCTIONS
47
3... 3. show that every remainder in the series for cos y and
sin y has the same sign 1:\8 the leading term (this generalizes the inequalities used in the periodicity proof. :.. and then a" can be interpreted as a power of a or at If b is a rational number with the reduced form p/q..'V w2 .'. 1L Show that the lOOts of the binomial equation" a are the vertices of & regular poiypn (eq.. == 2 IF . of a.. . Show how to define the "angles" in a triangle.. t
va. then we can merely assert that they differ by a multiple of 2ri (or 2. Z == log w is a root of the equation 6. L Determine all values of 2'.
== 2' . . Determine the real and imaginary parts of r: L Expre88 are tan to in terms of the logarithm . there is essentially only one elementary transcendental function.. positive number shall always mean the real logsrithm. ~5 f(_~ ill
. . bearing in mind that they should lie between 0 and. measured 'in radians.~~_<~..
log (ZlZ..
by solving
cos z
==
to
/1101. With this definition. For real y. The inverse sine is most easily defined by are
SIn W
•
log to
== log Iwl
+ i arg w.
.. and aPl has in general infinitely many values which differ by factors e2riRb• There will be a single value if and only if b is an integer n. even when the exponential function is not otherwise involved.) + r' = w. : j'
. Together with the exponential function we must also study its inverse function.l) . Find the real and imaginary parts of exp (e.0 the equation e'*'" == 10 is equivalent to
(24) eil
the equation
Finally we discuss the inverse cosine which is obtained
1( ../w2 v
log (w

1 . The symbol atJ where a and b are arbitrary complex numbers except for the condition a ~ OJ is always interpreted as an equivalent of exp {b log a). t..: ":i. Otherwise log a is the complex logarithm.. '
. For 10 '#. < 2 ~ ri3 ..: .1 are reciprocal numbers. Ute number 0 has no logarithm. In other words. Prove. unless the contrary is stated. the real logarithm of the positive number liD'. According to this definition the argument has infinitely many values whicb difier by multiples of 2r. and all has 8 single value.:.:. .
This is a quadratic equation in 6u with the roots and consequently
aU
The first equation has a unique solution x == log lwl. srg w.4. (1)2'1...'.1. prove that the sum of the angles is . that a < . . )1 3 2. ~. See.. ". In addition .. 1 .2.. ~
~:
..). L Findh e v aIue 0f e' lor z . for instance.cI~ly implies
4'
It is worth emphasizing that in the theory of complex analytic funetions all elementary transcendental functions can thus be expressed through ~ and its inverse log z. Forwhatva1uesofzis~equa1to2. (Compare with the remarks in Chap.
I .
. . only loosely connected with the flrst. It is perhaps more appropriate to sound a warning that greatest possible generality should not become a purpose. are expressly designed to develop geometric feeling by way of detailed study of elementary mappings. The second and third sections. At the same time we try to stre8B rigor in geometric thinking.3
ANALYTIC FUNCTIONS
AS MAPPINGS
A function to :: J(z) may be viewed as 8 mapping which represents ~ point z by its image w.<
.. In order to carry out this program it is desirable to develop the underlying concepts with sufficient generality. in a preliminary waYt the special properties of mappings defined by analytic functions. .L..
. The purpose of this chapter is to study. There is no need. to the point where the geometric image beeomes the guide but not the foundation of re~~ : . for our main concern is with the properties that are essential for the study of analytic functions. ·"::". no apologies are needed..
. If the student feels that he is already thoroughly familiar with this material.. ....
. The author believes that proficiency in the study of analytic functions requires a mixture of geometric feeling and oomputationel skill... In the :first section we develop the fundamentals of point set topology and metric spaces. .. for otherwise we would soon be forced to introduce 8 great number of ad hoc definitions whose mutual relationship would be far from clear." : .. .
~
. he should read it only for terminology . to go very far.
. : . Since presentday students are exposed to abstraction and generality at quite an early stage..
including those which are elements of both.:..0(%) I.... ". Two sets are equal if and only if they have the same elements.... This lends a geometric flavor to the language.." :
..
. The empty set is denoted by 9J.: ~
:. YES. For all considerations of limits and continuity it is essential to give B precise meaning to the terms "sufficiently near" and "arbitr~iJy near..." """:.. the triangle 'inequality is obviously fulfilled."..... Sec.z') .
"
"
. .
•
.. we shall have occasion to consider spaces whose elements are functions.y).0. the set all continuous functions defined on the interval a :!i x =! b.~:¥j:...
~.50
COMPLEX
ANALYSIS
ANALYTIC
FUNCTIONS
AI MAPPINGS
51
L ELEMENTARY
POINT
SET TOPOLOGY
and the De Morgan laws V Y) = ~X f'l ~ Y .." In the spaces Rand C of real and complex numbers.V(l
2lz . .yl . to say that. and an element as a point.cl + lzl")(l + 1z'11)
..
. ( .'
..: t.. denoted by X (\ Y.enlly near to y means that there exists an e > 0 80$·· that x EX whenever Ix :_. Sets and Elelnents.. respectively. An example of a function space is given by G[a. its elements.y) d(y.':. R and C are metric spaces with d(z. the foliowing terminology: For any. a set of real numbers has one complement with respect to the real line and another with respect to the complex plane.
'
."
"I' ""
1"'.
. It becomes a metric space if we define distance by d(J.y)! = ~ (Xi .
. whether finite or infinite in number.!'~i~t::':~f~..y) between any two points. In terma of dlstanee.. The n.
'
::..~{:~~~.& 0 an. The logical foundations of set theory belong to another discipline.
. 2.
. and this relationship is indicated by X C Y or Y::> X (we do not exclude the possibility that X = Y).the complement depends on the totality of points under consideration+ For instance. d{y. Topological considerations are extremely important for the foundation of the study of analytic functions.:'~:::.::"1:::.. For instance ..
.
.y) in 8U~h a way that the following conditions are fulfilled: 1.::::.ce d(x.
1.z).. The reader is familiar with the notation x E X which expresses that x is an element of X (as a rule we denote sets by capital letters and elements by small letters)."
.y . in that case a "point" is a function.y~)t. •• I. if X C Y we can consider the relative complement Y X which consists of all points that are in Y but Dot in X (we find it clearer to use this notation only when X C Y) . but should not be taken too literally. A set can also be referred to as 8 space.:' ..
. 3...
...
1.~Y·iI e ~Jthe. .. and they have obvious generalizations to arbitrary collections of sets. The last condition is the triangle inequality. d{x. One can of course form the intersection and union of arbitrary collections of sets.:c:._..: .. 2. ~ ..troduce.
' ".
..
... a set ~ contai7UI all z wji.etric space if there is defined.is formed by all points which are elements of both X and Y .
:. More generally..
..~:::.1 .':..ci. we iD.b]. Subsets of a given space are usually called point seta. In our language a set will be 8 collection of identifiable objects. We say that a set S is a m. . The complement of & set X consists of all points which are not in X.
d... 1.(X rl y) = ~X U .\.< Ii is called
of
>. .
e ".. such nearness can be expressed by a quantitative condition Ix . The term is traditionally used in a very wide sense and without strict limits.
.. For instance. .
••
"'.y) = 0 if and only if x ::::y.:.is the set of real ntuplea
+
with
B
" distance defined by d(x..y) = l~. and the first systematic study of topology
was motivated by this need . X conlainli points· arbitrarily near to y if to every e > 0 there exists an x E X such that lx ~ 91 < 5.dimensional euclidean space R.
". :.
•.d.::
. since this represents the euclidean distance between the stereographle images on the Riemann sphere.:r~~
.~'i..
..4). We recall that we
1
have defined a distance in the extended complex plane by
~x. What we need to describe nearness in quantitative terms is obviously a di8tan. in keeping with the fact that all our applications will be to very familiar objects+ In this limited framework no logical paradoxes can occur. The intersection of two sets X and Y.Dutive
laW8
X U (Y
X
r.
These are purely logical identities..' ~ . yl < £..~l:ot:~ 2: E·~ w1t~d(x."..~t B(~.yl < L For instance.2.
:
.:'.
:.'%::~ k~..."
..=". d(x. Similarly.g) == max I/(x) ...
~J}~~:f.x) = d(x.
'...s which are elements of either X or Y...z) s: d(x......"
. (Y
r.
:
'_. it will be denoted by We note that."
'
""""
" .y) . a nonnegative real number d(x.J
I t is helpful to keep in mind the di8trt.
~(X
The branch of mathematics which goes under the name of topology is concerned with all questions directly or indirectly related to continuity.
. for every pair x f 8. The union X U Y con· sists of all point.
'_ '.
:' .. Our approach will be quite naive.
. •
. Z) ==
(X U y)
n (X
V Z) = (X (\ Y) U (X
rv Z)
V Z)
(see Chap. X is a subset of Y if every element of X is also an element of Y. Metric Spaces.
The definition is interpreted to mean that the empty set is open (the condition is fulfilled because the set has no elements). We denote it by Jnt X.
+
exterior..~
:.. Show that a diacrete.z) < 6' gives d{x.
The proofs are so obvious that they can be left to the reader.
•
•
:
.. 'I
'1
•
'.tter space is bounded in the sense that all distances lie under a fixed bound.6) is an open di&k with center Zt) and radius a.and that X is open if Int X = X.y)~ show that 8 with the distance function a(x.
••••
.:
.. A point belongs to the boundary if and only if aJ1 its neighborhoods intersect both X and t"'JX.
. aX? L It is sometimes typographically simpler to write X' for .z) > 0. 7.. The closure is usually denoted by X.y) = ~(xty)/[l d(x. . (ii) The closure of X is the smallest closed set which contains X. A complete list' would be more confusing than helpful. We say that x E X is an isolated point of X if :t has a neighborhood whose intersection with X reduces to the point x.i). and since z was any point in B(y. to distinguish it from the closed baa formed by all x E 8 with d(xty) ~ o. X~ C Y.. For greater emphasis a ball is sometimes referred to as an open ball.:.X. ~ Observe that Int X C X C x. . Verify that this condition is fulfilled in the preceding exercise....which is not an isolated point . It exists..
center y and radius 8. As sueh it can be denoted by .. It is also referred to as the ooeigbborThe general definition of neighborhood is as follows:
Definition I.... .52
COMPLEX
ANALYSIS
ANALYTIC
FUNCTIONS
AS
MAPPINGS
53
the baU with
hood of y. and vice versa. is urged to interpret the proof in geometric tenns .:'
j...IX. If X is the set of complex numbers whose real and imaginary parts are rational. Suppose that there are given two distance functions d(x..a)... Notation: Bd X or ax.
:'
• __
:
..== X. A set is said to be discrete ij' all its points are isoleted.
.. In any metric space the empty set and the whole space are at the same time open and closed. Siio~"that .
In other words. In the complex plane B(zo.
EXERCISES
L If S is a metric space with distance function d(x..·aet·in·R or C is countable .
... The la..8) we conclude that B(y.y) and dt(xty) on the same space S. if Z E B(Yto).. or the intersection of all closed sets ~X. 3. .
space. ..\ .
~.'
..y)] is also a metric
The union of any collectitm 01 open seis i8 open~ The union of a jinif£ number of closed sets is closed . Also.) C B(y.6) is a neighborhood of z. It can also be described as the set of all points of which X is a neighborhood .The triangle inequality shows that B(z.. The inleTsectioo of any collection of closed ssts is closed.. .. :v.
.. With this notation. It is clear that x is an secumulation point of X if and only if every neighborhood of x contains infinitely many points from X. and we shall limit ourselves to the following: i1) terior.:~ ~ .d(y.. closed if X. It should be noted that the last two statements follow from the first two by
use of the De Morgan laws. X C Y implies Int X Clot Y. '. .The following is an immediate consequence of· the triangle inequality: Every ball is an open set.. . for it may be characterized as the union of all open sets ex. They are said to be equivalent if they determine the same open sets. Hence B(y.
.:'" ..: .
:
+
.X_. Show that d and al are equivalent if to every is > 0 there exists a a > 0 such that d(xty) < a implies dl(%~Y) < e. A point belongs to the closure of X if and only if all its neighborhoods intersect X.
. The complement of an open set is said to be closed...~ form a closed set. For added convenience we shall also introduce the notions of isolated point and accumulation point. Show by strict application of the definition that the closure of lz .zol < a.....
.
.. infrequently by CI X. _.:.. ..6.... (i) The interior of a set X is the largest open set contained in X. Indeed... a neighborhood of y is a set which contains all points sufficiently near to y.z) = 8._..the :accumUlation points of anY. for d(x.y) < ~' d(y. (iii) The boundary of X is the closure minus the interior.. it consists of all complex numbers z which satisfy the strict inequality I: ..
.. There are many terms in common usage which are direct1y related to the idea of open sets.
. and there may be other sets with the same property.. The following properties of open and closed sets are fundamental: The intersection of a finite number of open 8Ct8 is open.
:.. and the reader. An accumulation point is a point of X. what is Int Xt X>. how is X1I related to X? Show that X'I_.I == X'_I.Z 6~~ 3. Z.. . then 1/ = 1. It is also the complement of the closure. (iv) The exterior of X is the interior of r.. We use the notion of neighborhood to define
apen
set:
A Bet i8 open if it is a neighborhood of each of its eiement«:
Definition 2..i) is an open set..
boundary. We have just proved that it is an open set.Z0' < 0 j 8 lz .. closure. A
set N C 8 is called a neighborhood of YES
if if contains

a ball B(y.
If A is empty.:.. At the risk of being pedantic we repeat: Definition 3.
. ~. . A ltUbset 0/ a metric &paCe is connected if it cannot be represented as the union of two disjoint relatively open 8ets none of which is emptYt If E is open.I
'11 ". but not open in R. none of the sets is allowed to be empty . then the semielosed interval 0 ..
t The s~um
... . if we regard the closed interval 0 ~ ..~ x ~ b. a space is connected if it consists of a single piece. A contains a largest number...bi) with a".. This is meaningness unless we define the statement in terms of nearness. I ••
. Henceforth. one of them must fail to meet E. a < x < b.
"
.h. We denote this interval by (a.ctJ. . The largest number ·a of A. based on the fact that any monotone sequence has a finite or infinite limit.. The opposite assumption would lead to a similar eontradietion. Intuitively speaking. but an open seton E need not be open when regarded as a subset of S.
... a < x :s: bt a.. An interval is defined by an inequality of one of the four types: a < x < b.1. It is understood that A and B are dUjoim and nonempty. . ...
". All points of E lie between a and b. is called the grea~Bt loue« bound of E.l.
.. relatively closed means the same 88 closed. As an example. Since A and B are closed e would have to be a common point of A and B. we msy conclude that either A or B is empty. and we conclude that t must belong to E. and if A is the whole line we se·t a == + With this convention every set of real numbers has a uniquely determined greatest .
... a subset of E is relatively open if and only if it is open.
=~...
{"
"
:.. :~.. The nonempty connected
IJ'UbsetB of the real line are the
intervals..__. Again. Suppose that a point ~ from the open interval (a.. t Returning to the proof. . Connectedneu." . elosed.
Another common practice is ..
:~
". apace is often used in the following manner: Suppose that we are able to construct two complementary open subsets A and B of S.b2) and continue the process indefinitely. In this way we obtain B sequence of nested intervals (~. AB to A itself it is easily seen that A is open whenever it does not contain any largest number.00 or b = + co this includes the semiinfinite intervals and the whole line .
II
COMPLEX
ANALVSIS
ANALYTIC
FUNCTIONS
AS MAPPINGS
55
1. Suppose..
9f •. and its subspace topology is called the relative topology..'~':. The easiest ~8y is to give & negative ehsrscterisation: S is not connected if there exists a partition S == A V B into Open 8'Ub8et8 A and B.8. we agree to set a· == ..x or sup x for x E E.
We reproduce one of the classical proofs... Then the open sets defined by x <·E and x > E cover E... if it exists. we shall always mean that it has this property &8 s subspace. in contradiction with the fact that" a is the greatest lower bound.} have a common limit c.only if the set E is empty. if 8 is connected.b]... It is evident that the . Because the line is connected. x or inf % for x E E. See.
I':
I
I
. .. Before proving the converse we make an important remark. and hence R~is connected.' The most important result is that the whole line is connected.. Suppose that the real line R is represented as the union R == A V B of two disjoint closed sets. Cor instance. Then ~ would be a lower bound.bl or ta~b(. A· and its complement cannot both be open unless one of them is empty.. or A is the whole line.. lower bound. This contradiction shows that either A 'or B is empty..~ x < 1 is relatively open. we assume that E is a connected set with the greatest lower bound a and the least upper bound b.
(0..·:+.y) a8 on all of S.x < 1 as
Theorem 1. ~ ~':~...+s:. A subset E C S is said to be connected if it is connected in the relative· topology. . Consider the set A of all lower bounds. t For· a = .to denote open intervals by ]a. Trivial examples of connected seta are the empty set and any set that consists of IL single point..~are to be ·~tnCtllae(t .J and {b". If E is any nonempty subset of 8 metric space S We may consider E as a metric space in its own right with the same distance function d(:t. it is clear that a =' + 00 if and .
~.b .r·~I:I~:·· . The connectedness of a. 2. There are thus three possibilities: either A is empty. Let E be an arbitrary subset of R and call a a lower bound of E if a ~ % for all x E E. In the case of the real line it is possib1e to name all connected sets. that no point of E lies to the left of ~.. we may assume that at < b1• We bisect the interval (al. The sequences {a.
~"t••
.
:
.b) did not belong to E. and because ~ is connected. and this is indeed one of the fundamental properties of the realnumber system. Similarly. ~.b[ and semiclosed intervals by ]a..
a subspace of R.b) and closed intervals by [a. E B..... seq~ce
wu ~~~
~
i~ Chap. . it is commonly denoted 88 g.
. if E is closed. '. and a dosed set is connected if it cannot be deccnnposed into two closed sets.bt) and note that one of the two halves has its left end Point in A and its right end point in B. If neither is empty we can find at E A and hi E B. To avoid confusion neighborhoods and open sets on·E·are often referred to as relative neighborhoods and relatively open sets. Neighborhoods and open sets on E are defined 88 on any metric space.
• . the cases a arid b ··.. or semielosed interval with the end points a and b. when we say that a subset E has some specific topological property. complement of A is open. . With minor modifications the same proof applies to any interval . EAt b. I t follows that E ~ an open. It is always understood that a < b~
t We denote
open intervals by (a. limits included. We can therefore state: An open Bet i~connected if it cannot be decomp088d into two open sets. 2. is defined in a u corresponding manner. The least upper bound~ denoted as l._..
By Theore~ 3 the whole plane. in A40 All points in this neighborhood can be joined to al by & line segment. Suppose that C(a) were not connected.
The structure of connected sets in the plane is not nearly so simple as in the case of the line. ·We have
One of the sides of the polygon must then join a point in A 1 to a point in A2. The representation B•. The same is true of any 6neighborhood in Rtl. Then C(a) is sure to contain a. Eo () A. . that any two components are either disjoint or identical. . . The proof is the same..
A nonempty connected open 8et is called a reqion:
A nonempty open Bet in the plane i8 connected if and only if any two oj ita points can be joined by a polygon which liu in the set...: . and & half plane are regions. First. are evidently open..
.
". a component of a set is B.
Theorem 3. so that in fact C(a) == C(c)... an open disk 12. part would be empty. and for this reason it is sufficient to consider the case where at and Bt ~ joined by a.all < ! contained . if C E C(a) (\ C(b). If we . and by the same reasoning O(c) C C(4). and from there to a. moreover.
I
. disjoint. This is contrary to the definition of A!. We may aseume that a E A while B contains 8 point b.COMPLEX ANALYSIS
ANALYTIC
FUNCTIONS
AS MAPPINGS
17
proved:
In the course of the proof we have introduced the notions of greatest lower bound and least upper bound.J beE.. Theorem 4. Choose at E Al) a: E At and suppose that these points can be joined by & ~polygon in A. We prove first that the condition is necessary. then at: could be joined to this point by a line segment.~~~ ':

. Let us prove that At and Ai are both
regions may have the same closure.r
If E is the given set.. The closure of 8: region is called a closed region.atl where t runs through the interval 0 :S t ~ 1 The subsets of the interval 0 < t < 1 which correspond to points i~ A · and AI! respectively.. It would follow. Let A be an open connected set. if at E At.'
. We denote by AI the subset of A whose points can be joined to a by polygons in A~ and by AI the subset whose points eannot be so joined..~I < p. . Since bE 0(0) there is a eonneeted set Eo C E 'W~ich contains 0 and b. but the following characterization of open connected sets contains essentially all the information we shall need. This is a·contradiction. As the name Indicates. in which ease they are ealled the minimum and the maximum. the set must lie in a finite interval.
: .
':
.. For future use we remark that it is eve~ possible to join any two points by a polygon whose sides are parallel to the coordinate axes..moredimensional analogue of an open Interval... The notion of & [oining polygon is so simple that we need not give a formal definition . let 112 <I
be a neighborhood contained in A~ If a point in this neighborhood could be joined to a by a polygon. The theorem generalizes easily to R ~ and en~
Definition 4. for instance in the course of a proof . and from there to a by a polygon. 0:.
open. which is precisely what we
by way of 0. consider a point a E E and let C(a) denote the union of all connected subsets of E that contain a.can show that C(a) is connected . Finally.~~'J:'. ~Rt . then C(a) C C{c) by the definition of C(c) and the connectedness of O(a). for the set consisting of the single point a is connected. Then we could find relatively open sets A.. But Al contains the point a..'.d: JJ such that C(a) s: A U B. and choose a point a E A. in other words a component.
'
'. ·Any doeed and bounded nonempty set of real numbeT81ias a minimum and a maximum. A region is the. Indeed. • I'
want to prove.. It should be observed that different
Every set has a unique decomp08Uion into components. In order to be sure that the bounds are finite we must know that the set is not empty and that there is some finite lower bound and some finite upper bound. It happens frequently that we have to analyze the structure of sets which are defined very implicitly. and nonvoid. connected subset which is not contained in any larger connected subset . Hence a E C(c). In such cases the first step is to decompose the set into its maximal connected eomponenie..· () B. if 01 E A 1there exists B neighborhood Iz . and we have proved that the condition is necessary . and we conclude that Az is open.In order to prove the aufliciency we assume that A has a representation A z:: At U AI as the union of two disjoint open sets..· and we conclude t'h&t C(a) is connected. In other words. This segment has a parametric representation z == at t(a2 . and ~C8: . Since A was connected either At or A:z must be empty..
contradicts the connectedness of the interval. T~
+
Theorem 2. '. such a set is said to be bounded. A fl B == JJ. and we have proved that the condition of the theorem is sufficient. B . any two points in A can be joined
o.. and consequently C(a) = C(b)~ We call C(a) the component of a.
'. hence AI is empty. : ::...
". If the set is closed and if the bounds are finite. .
• ....1. and all points can be joined to d... Similarly C(b) == C(c).. then it is a maximal connected set. 'r'. Hence the whole neighborhood is contained in AIJ and At is open. they must belong to the set. Secondly.. line segment.= (Eon A) U (Eo n B) would be a decomposition into relatively open subsets.
in Rft.8) is connected B(a. re thinking of X. it is possible to analyze the proof ~d then:by ~nve at a more general result. Compactness. We are led to the following result: In a locally connected separable space every open set 18 a cQ'Untable union oj diBjoim regiOfl. Consider a E C(a) C E.
.y) E R2 with x == 0.
~
.. that the number of components is countable. ..= 8.' ._}. ~
.: .
t: t" :... To different components correspond diff~nt k..1).t are !he co~ponents of E? Are they all closed? Are they relatively open..
~:
. set E_ IS denae ~n S if E.~ .
"7
!
'I~
". furthermore... s. The definition of compactness reads. . whole generation of mathematicians to agree on the best approach...
I..4. In view of its importance the property deserves a.. The reader wiU~·findno difficulty in proving that a complete subset of a metric space is closed. Show that 8r
discrete set in a separable metric space is countable ..~.
if every Cauchy
sequence
countable.
"
a metric
.l·is a. and we would say that {xtt.are connected.x) + 0. I
. namely that every Cauchy sequence is convergent (Chap . Sec.. . It is clear that every convergent sequence is a Cauchy sequence. In the case of R" we can conclude. 2.CISES
show that the relatively.. given set can be covered by open sets. 1 and either y = 0 or y = lin for some positive ~teger ft. 1# d· · ·m For instance.:'.~: . We shall now introduce the stronger concept of compactness.
t
1. for he will not be able to discern the purpose of the definition. .
.
A subset is complete if it is complete when regarded as a subspace . Let A be the set of points (X. Whatever we do the uninitiated reader will feel somewhat bewildered..&t0 ~ x ~.::
.
r
.. Again.
. 'I :_''''::~.
Definition 5.~) • ~.
: :: . L Prove 'that the closure of a connected set is connected. ~a. .
containa a finite BUbcovering. Indeed. Show t hat the union of two regions 18 a region if and only If they have a common point .
58
COMPLEX ANALYSIS
ANALYTIC
FUNCTIONS
AS
IIAPPINGS
51
Theorem 5..'... A subcovering is a subcollection with the same property.: " .
ic"J 
I
.2. and let B
L If
xes
be the set with x > 0. :.:
'.
'. . and it is not hard to see that this property carries over to any R1I...& subset of the natural numbers. A set is said to be discrete if all its Points are isolated.. .y) E R2 such th..6) C C(a).r: " ... We shall sa.': i
·'r ~.
• .
".
EXE.open ~t must contain a point with rational coordinates..:
' '
"
.. IS.
~·..) .:. In view of this result it would be possible to dispense with the notion of compactness" at least for the purposes of this book.) + 0 as n and m tend to 00..
This is &.. . and a finite covering is one that consists of a finite number of sets.'. Cauchy sequence if d(xa. and that a closed 8'Ubsetof a complete space is complete..
A set X it.'
.. The proof is left to the reader.. open (closed) subsets of X are precisely those ~t8 that ean be expressed as the Interaection of X witb an open (closed) subset of S.8
~: . A little more generally the assertion is true for any space S w~ch 18 l~y con:netted.8 ... IS ~~en. For each component C(a). Let E be the set of points (x.. We. as a subset of a t. 2. Hence C(a) .. If E 18 open It con~ns B(a.... This is not surprising. but not conversely. Defin~tion 6. To see this we observe that every .. The consensus of present opinion is that it is best to focus the attention on the different ways in which 8. Verify that E is not locally connected . and it is a matter of taste which one to choose as definition..
I
:
.. 7. but this would be unwise} for it would mean shutting our eyes to the most striking property of bounded and closed sets of real or complex numbers.'
•
1_
'~••
:==
• : ... and consequently the set of components 1.:
. .y that 8.. 3). The set of points With rational coordinates is countable..
". special name..o) and because B(a. determine the ~ma.Oneighb~rhoods . .. y = sin 1/X'.. of a point a contains a connected neighborhood of a. There are several equivalent characterizations of compaetness. every open subset oj R is a countable union OJ t8J0I open interval8.l·::···
.. '. compact if and only if every open covering 01 X
In 'this context we. For R and C we have proved the converse... consequence of the fact that the . It is stronger than completeness in the sense that every compact space or set is complete. and may thus be expressed as a sequence {Pi}.' ~ J~. Is A U B connected? . By this we mean that any neighborhood.. Prove that the components of a closed set are closed (use Ex... for it took a.
rt'
. A nMlrit space i8 aaid to be compli!U
is convergent . The outcome would be that we would have to repeat essentially the same proof in many different connections.

.... "to
yo
. ..llest k such that P* E C(a)..: ~I':. As a matter of fact it will tum out that the compact subsets of Rand C are the closed bounded seta.'
.. we would say that x" ~ s: if d(x". conclude that the components are in onetoone correspondence with . : .
In Kit the components of any open set are
op6nt
i
: ..x. and we shall say that a metric space is separable If ther~ exists a countable subset which is dense in S. The notions of convergent sequences and Cauchy sequences are obviously meaningful in any metric space. Let us say that a collection of open sets is an open covering of a set X if X is contained in the union of the open sets. ~
~
space 8.
~ I:
'.. 'yl ~ 1.
"
... ~ Ex. For the opposite conclusion we need to show that every bounded set in R or C is totally bounded.~. 3 For this reason it makes no difference whether we formulate the definition for a full space or for a subset. . and let {x. choose itt point Xo and consider all balls B(xo. A set.. relatively open set). for it is not clear whether the . for d(x. In fact. If it does contain a point from X.
".Y) < 8/2 and Eft < &/2... Therefore it must contain a finite subeovering.. Determine n. for if X C B(Xl. The square can be subdivided into a finite number of squares with arbitrarily small side.:r). then we can replace it by 82aneighborhood around that point. ed is valid. (The preceding proof that any compact set is bounded becomes redundant. that is.
11).tXn+l) < en and hence d(x".ll B(y.. .. .~ totally bou .
. It converges to a limit y..y) :s.} is not convergent. Therefore B(xnr £:ft) admits a finite subeovering. such that d(x.xn) < E for m..=0·. formed by Us. neighborhoods wit~ centers on X. If {x.y) > 2e. It happens that this is of no avail.
... is compact if and only if it is complete and totally
bounded. X C B(xo. When Definition 7 is applied to a subset X C 8 it is slightly ambiguous.neighborhoods should be with respect to X or with respect to 8. We choose a fixed n ~ no for which d(xft. E B(Xl. it is not clear whether we require their centers to lie on X.
.) We have already proved one part of the following theorem:
Theorem 6..
We have already mentioned this particular consequence. it follows by the preceding reasoning that this collection is an open covering of X. it wou1d follow that the given sequence is finite.. then any two pointe of X have a distance <2£ + max d(Xi. Because B(xl. For the collection of all balls of radius e is an open covering.lies in the fact that many proofs become particularly simple when formulated in terms of open coverings... I}..8).xo) + d(y.. . X can be
This is certainly true of any compact set.) < tit implies d(x. which means the same as X C B(xa. and the squares can in turn be covered by disks with arbitrarily small radius. (T1t • . £ 1) which does not admit a finite subcovering.r:tr:ir.. If such 8 neighborhood does not meet X it is superfluous. We prove first that every compact space is complete. contains only finitely many Xli.Y E X it follows that d(x..
A subset of R or C
is compact if and wdy if it ie closed
and
Definition 7.
'.. except for a small point that should not be glossed over.. .. eft) C Bf:y. contains x" only for finitely many n). Suppose that there exists an open covering which does not contain any finite subcovering.H
COMPLEX ANALYSIS
ANALYTIC
FUNCTIONS AS MAPPINGS
11
and the covering is by open sets of S..:. a compact set is necessarily bounded (8 metric space is bounded if all distances lie under a finite bound).. in other words. and conversely every open subset of X can be expressed in this form (Sec.
ev
R.. and this y belongs to one of the open sets U in the given covering.::.8).. it contains a finite subeovering. and hence in a square. d(x.txn} > E for all m ~ no. We know that S can be covered by finitely many B(xt E. hence there exists a B(Xl. and our proof that bounded..). T For any X.. and complete subsets of a complete Space are closed. d
bounded . If each had a finite subcovering. If X is bounded it is contained in a disk. In one direction the conclusion is immediate: We know that a compact set is bounded and complete. Its importance.} there exists an e > 0 such that d(x.
To prove the other part. .xfJ) < 2r... and can be dropped.
"r •
.. But that is clearly impossible. E"n+l €n+pl < 2n+1• It follows that Xn is a Cauchy sequence. with r = max. .:~~ of:C.' rr.+l E B(x 1 The second property implies d(Xty. Consider now the collection of all open sets U which contain only finitely many XII.
bounded .. Secondly. But boundedness is not all we can prove. n ~ no. Write til = 211..
&
totally bounded set is totally
: .Y) < 8.
+
+ ·· ·+
ft)
+
Corollary.) has no finite sub cove ring. Suppose that X is compact. Because U is open. Forthis reason the ambiguity is only apparent.. it contains a ba.. for since each U.. .€:) U · · ·U B(x. Cauchy sequence in X. To see this. assume that the metric space S is complete and totally bounded. Then B(xft.:. The property in the definition is frequently referred to as the HeineBorel property.••••
t Here we
are
usinI the
fact that any subset of ~ this . suppose that we have covered X by eneighborhoods whose centers do not necessarily lie on X. It is convenient to define a stronger property called total boundedness:
trI)..tely many balls of radius t·
. e) contains only finitely many x. and the compactness implies that we can select finitely many that cover X. 1. and we have proved that X is bounded.ty) > 2e for infinitely many n..x d(xtf. t UN.r0 U · · · U B(xo. ~.But if U is an open set in 8} then U (\ X is an open subset of X (8. and we conclude that 8 has the HeineBorel property..y) ~ d(x. We observe that a totally bounded set is necessarily bounded. Choose n so large that d(xjf.
t
>
0.. I) . ."'.
.d(x. The reader ~
. If y is not the limit of {x. and if X is compact. This proves that X is totally bounded.: . is itself totally bounded we can find an x.
.. and we obtain a finite covering by 2..e). Then d(x""y) ~ d(x$tY) .":. namely by the single set U. '.} be a.r) . Let us take the case of C. This is a contradiction.. They form an open covering of X.xtt+'p) < eta. the same would be true of S.r.
A Bet X ie totally bounded ift lor every covered by ftnf..: .x. but R and C are complete. It follows that the eneighborhood B(y.Xj).€I) such that B(x'be.: ~s. (better.~~:. t It is clear how to continue the construction: we obtain a sequence Xft with the property that Bex"" E:R) has no finite subcovering and X..
same point...·.~~ '::.. The converse is also true. We are mainly concerned with functions ~hat'at'econtinUOU8< at an PO~ts in the domain of definition. but we preferred to emphasize the importance of total boundedness as early as possible. except that a sequence permits repetitions of the.:~~ :. and when Xl.. occasionally the latter are allowed .e. The previous time we used this reasoning was to prove that a compact space is complete.~.... ::~ :~ .RB...::.
L Give an alternate
Theorem 7a
A metric apace is compact if and only if every infinite sequence has a lima point.. It should be clear that results of this kind are particularly valuable for the purpose of presenting proofs BBconcisely as possible. The first part of the proof is a repetition of an earlier argument.18 regarded as a subspace.I
·~)ii·:. We conclude that the BolzanoWeierstrass property implies total bound ed ness. is not
decreBBing sequence of nonempty
empty (Cantor1s lemma).}..or complexvalued functions..~.~.··~~~~. ordinary distance being replaced by distance on the Riemann sphere ..td. the function.} that converges to y. .. .m~~~~ L . Use compactness to prove that a closed bounded set of real num
:> Et.uoWl. Observe that fUl(X~)) C X'.. ..e) V·· . end then we observed that a Cauchy sequence With a limit point is necessarily convergent.. L . are ¢ o.'~_. Show that the HeineBorel property·can also be expressed in the f?llowmg ~anner:.Weierstrass theorem.:. limit point must be convergent.
Show by example that this need not be true if the sets are merely closed S.~L~. in which ease the d?~a.. in such a way that X"'i E B(y. ....functions f whose domain is only a subset of S. A limit point is almost the same as an accumulation point of the set fonned by the points x..:> · · · is . ••• t X" have been selected we choose %tI+1 so that it does not lie in B(x}.. we just pointed out that a Cauchy sequence with a.. We construct a sequence I x" 1 &8 follows: Xl is arbitrary.
) ~... I t remains to prove the converse.~. for d(x.~ : ..
. ... .
.. . and it is denoted by f(X).~_~~.
i:.. Indeed. ..rj·..·. FunctIOns ~ also refe~ to as mappings: we say that! maps S into S'. This is always poaaible because these neighborhoods do not cover the whole space..5. In view of Theorem 6 that is what we had to prove.
.:.~. Functw..~'
. limit point of {x_l it has a neighborhood which contains only finitely many Xn (abbreviated version of the correct phrase). ')._::~>~ . We shall consider functions! which are de6n~ on a metric space S and have values in another metric space S'.:. V B(x".. : .
bounded. every neighborhood of y contains infinitely many x.• .
.. This theorem is usually referred to as the Balzano. Is the space complete? Show that the aneighborh~ are not "totally·
.~':iZ1:~~"i~:1=:i~~~!:t~~~... ~..."x > ! for all m and n.m. Indeed. Let 8.a
compact sets.. If y is not 8.:
: :: •••
.·~·_::· _~.e). Contin.E. '. But it is clear that {x"l has no convergent subsequence. .00 the set of aU sequences x = {x. . and f1U(X) ::> x.:
. 2. .
.~.
EXERCISES
proof of the fact that every bounded sequence of ~ompI~ n~mbers has a convergent subsequence (for instance by use of the Iimes inferior).
...
~~
~~
. suppose that £1. Defined(x y} = max Ix _ y I . In most cases it is safe to slur over the distinction ~ a function on S and its restriction to a subset are usually ~enoted by t~e same symbol.
. and it would follow that the sequence is finite..
.~~:::.::. :::.~~~·:.8 called the 'mage of X under I.: ~.
.Every collection of closed sets with an empty intersection contains a finite 8ubcollection with empty intersection . It deals with the notion of limit point (sometimes called cluster value) : We say that y is B limit point of the sequence {Xn} if there exists a subsequence {x . In the first place it is clear that the BolzanoWeierstrass property implies completeness... The inverse 1 1 11IUIfer (X') of X' C 8 consists of all z f 8 such thatf(z) f X'. If every B(y. .:/~i.
ft
The ~~er should reftect on the fact that we have exhibited three c~a:actenzations of compactness whose logical equivalence is not at all triviel.. : . would form an open covering..
1..:~_·(k~.
.:. NaturallYl we shall be mainly concerned with rea.~. In the compact case we could select a finite subcovering.. ~. £' 1) and nitl > nl:L It is clear that {~} converges to y.1 of real numbers such thatonlyafimtenumberofthex .'
:. and we wnte J :8~ S.~ {~: ..
'.12
COMPLEX
ANALYSIS
ANALYTIC
FUNCTIONS AS MAPPINGS
13
There is a third characterization of compact sets. ~~.) contains infinitely many X we can choose subscripts nk. For strict economy of thought it would thus have been better to prove Theorem 7 before Theorem 6.. It came to be recognized as an important theorem precisely because of the role it plays in the theory of analytic functions. If X C 8 the set of all values J(x} for x E 8 1. If y is 8 limit point. then the intersection
r:...' ~ ~
•
. .
. We are of course free to cOIlSl~er. ~t.+ O.
_.'. Th_edefi~tion of a continuous function needs practically no modification.:_.
.
. by induction. ..
I .. Suppose now that the space is not totally bounded+ Then there exists an ! > 0 such that the space cannot be coveredby finitely mimy eneighborhoods.:'. We showed in essence that every sequence has a limit point. :::::> B. f UI continuous at a if to every It > 0 there exists 0 > 0 such that d(xJa~ < 0 Implies d'U(x}J!(a)} < I... If there were no limit points the open sets containing only finitely many x..
i. ~~~~::.to take value~ in the extendedoomplex plane.If Hi
..·.
A)
bers has a maximum .l.'.·:i~.The spa~ 8 is the domain of. The original formulation was that every bounded sequence of complex numbers has a convergent subsequence.
.lingui8ticaJly clu.:...). For mstanoe..
~.)) '< 1/2 80 that d'(J(xJ.!ifor. We conclude that SI is 0
The proof is typical of the way the HeineBorel property can be used. the words "open" and "closed. Suppose that f is continuous on a compact set X.o! u.!(x'JJ) < e all P8Irs (X1tX. Under a continUOUB mapping the image of every compact set is cmnpact. 8. The emphasis is on the fact that a is not allowed to depend on Xl. :1 .
10. and we have proved thatf(x) is compact . For instance the mapping I(x) = x2/(1 + x2) of R into R has the image feR) = {y.p).
. and that f(S) is all of SI. " . For Inatanee the function z is ~njformly continuous on the whole complex plane. Then B == jl(A) U fl(B) is a representation of S as a union of disjoint open sets. ". continudy will be in constant UBe. ihe following is
bers also contains zero."""*~""".
Under a continuoUB mapping the i~age of any connected Be'

We may assume that I is defined and continuous on the whole space S.. and consequently closed.there exists a 0 > 0 such that d'(!(xl).Pw/2). It 18 said to be onto if f(S) = 8' .~. "" The image is a closed bounded subset of R.B(y.·on).
connected. In fact.p) . . There 18 a Yl With d<Xt.. A typical application is the assertion that a realvalued function which is continuous and never zero on 8 connected set is either always positive or always negative. 1 ...~ .: ./2 & !S:. There exists a firute subcovering: X C B(YhPI/2) V··· V B(Y.)rf(y".. 0 ::: y < I} which is neither open nor closed.1I. Accordingly. a function f is said to be uniformly continUOUB on X if to ev?ry e > 0... In this connection it is perhaps ~ful to point out that the property of being an open subset is not topological.. but the function z~
18 not . .i/ormly
The existence of a
is connected. not for the direct image.X2) < 6.lea is aaIled ~
be re1B_. For every y e X there is a ball B(y. defined on 8'. y.. here p may depend 00.: ."~. it "satisfiesf~I(f(x» == x and f(Jl(X') = X'." when referring to the inverse image.. TheoreIll 9.auch that d'(J(x)J(y» < E/2 for x e. but this is a deep theorem that we shall not need ./2.may ternuJ can (It "" to\ A"'" ".. If f is not defined on all of S. A continuou8 realvalued /unct£on on a compact Bet has a max~ mum and a minimum...» < 42 and d'(j(x. "" "~"
~oron.. In twa example I(R) fails to be closed because R is not compact._ o. If 8 is connected either jI(A) ::::. Let 8 be the sm~lest of ~e numbers pt/2.
Ig~
t~
..".. we have proved that compactness and connectedness are topological properties (Theorems 8 and 9). Suppose that f is defined and continuous on the compact set Consider a covering of !(X) by open sets U.X!) < o.Yk) < Pk/2. If Xes and Y C 8' and if X is homeomorphic to Y there is no re~n why X and Y should be simultaneously open.
Ibappmc WIth both propart.: or 0 Jl(B) = Of and hence A :::: or "B. Quite ~~era11y.. COrollary.VjI(U.• ay. .
. a condition 18 said to hold uniformly with respect to fir parameter if It c~ be expressed by inequalities which do not involve the parameter. On a compact set every contt'1I.84
COMPLEX
ANALYSIS"
ANALYTIC
FUNCTIONS
AS MAPPINGS
85
The following characterizations are immediate consequences of the def nition : A function is continuous if and only if tM inverse image oj every open set is open. maximum and a minimum follows by Theorem 2. .y.
==
A ~a?pi~ f:8 + 8' is said to be one to one if f(x) =+ J(y) only for Y. .. ~ ..:
I. and hence an interval.:hich 18 shared by all topological images is called a lop<llogical properl.
hr
+
On sets which are not compact some continuous functions are uniformly continuous and others are not.f(x2)) < f 88 desired..I~ •••
and ~..
I:· .: \.num. ~
<I
•
(~
~OI' one
to one)
~. But an interval which contains positive and negative ..
x
true:
Theorem..is a topologual mapping or a homeomorphism.t A mapping with both these
fOr'
Th~rem
conHnU0'U8. . Co~lder the covering of X the smaller balls B(y.UOUB function
is 'U.:..n. 0. . should of course be interpreted relatively to the domain of f~ It is very important to observe that these properties hold only for the inverse image..:. It follows that f(X) C U1 V·· · U Us. the image is connected. In this situation.. In fact.. It happens to be true if S = 8' == Btl (invariance oj the regt. Suppose that 8' = A V B where A and B are open and disjoint.t ~
1.. and suppose that d(Xt.p/2). .J with d(Xl.).: :. A fUnclion i8 ccmtinUO'US if and only if the inverse image of every closed set is closed.!(y.. if f and iI are both continuous we say th~t f. The inverse images II(U) are open and form a covering of X. Because X is compact we can select a finite subeovering: X Cf1(U1) U · .

x~
properties has an inverse 11. . and we obtain d(X2 Yk) < pt.. The notion . A property of a set . Hence d'(f(x.
+ x and x" j. However.
.

"
.continuous on the whole real line: sin z. As a matter of fact. The HeineBorel property is also one that deals only with o~n sets. .6.. but very essentially also a succession of points.
"
.:
}~
. If a Dond~functioii t . Which of the following functions are uniformly .. ~ .. Therefore it. and so on.
. ·
A topological space ill called a Hausdorff epeee if any two distinct points are crmtained in disjoint open sets.
.: z{t). makes perfect sense to speak of a compact topological space. In the presence of this condition it is obvious that the limit of a convergent sequence is unique.. (ii) The intersection of any two '?Pen sets is an open set . boundary.
:.. Neighborhoods could be avoided.
. if x#<1I we require the existence of open sets U..:_ "...
. The definition is clear: We say that x" j.. Topological SPOCeIJ.. (Consider the effect of removing a point. Connectedness was defined purely by means of open sets .·:~. . 01 points as z :.) . The point is that it may be unnatural to futroduce a distance in situations when one is not real1r needed. ...:s 1".. Prove that there exist x EXt Y E Y such that d(x..
.)·:..) . properties (ii) and (iii) were strongly emphasized. .:. the first serious difficulty we encounter ~.p6 tW interval «". "(1") m8.• '.. z if every neighborhood of x contains all but a finite number of the XII. Theorem 6 becomes meaningless...
.
Definition 9. However.... Indeed.."~. We shall never in this book have occasion to consider a space that is not a Hausdorff
n
space.
1r CONFORMALITY
Definition 8.
+
'.~_X. an are is not merely a set of points. ~"
. =. (Show that closed sets are mapped "on closed sets.
2.~.
.$ pi onto a ::. and it is immediately clear how to define Interior._. x sin x. y we are not able to prove that x = 1/... .
.
. Prove that a subset of the real line which 18 topologically equrvalent to an open interval is an open interval.
~.. = 2(. This awkward situation is remedied by introducing a new axiom which characterizes the topological space 88 a Hausdorff space .. The equation of an arc 'Y in the plane is most conveniently given in parametric form IE = x(t)J '!I = yet) where t runs through an interval a ~ t ~ fJ and x(t)J yet) are continuous functions. The JollQlUJ'ingconditions have to be fulfilled: (i) The emp·t:y set r& and the whole space T are open aetB.
'
..y) is a minimum.·· A topologicalsptJCe is a set T together with a collection o. . ..
.:.sJ. We can also use the complex notation z = z(t) ~ x(t) iy(O which baa several advantages. then..» define8~ 8NIle ~D.::
'. .
in
This is not the place to give examples of topologies that cannot be derived from a distance function. :. compact space is topological.. x sin (x2)J \x1i sin x? It is not necessary.II
COMPLEX
ANALYStS
ANALYTIC
FUNCTIONS
AS MAPPINGS
17
EXERCISES
1 onto the whole plane. Wlth
We now return to our original setting where aU functions and variables are restricted to real or complex numbers. As such it is eompaet and eonneeted. The reason for including this section has been to alert.. and the theorems remain true.
L Const~ct
a. It is also customary to identify the arc ")' with the continuous mapping of [a. called open sets.. we used distances to define open sets. and Theorem 7 becomes false: ~. closure. . ordered by inCI"9a8ing values of the" parameter. but there is really no strong reason to do this.
We recognize at once that this termin?logy is consistent with 04'llr earlier definition of an open subset of a metric space. Considered as a point set an arc is the image of a closed finite interVal under a eontinuous mapping. Closed sets are the complements of open sets.~ .(. The role of metric spaces will seem disproportionately small: all we actually need are some simple applications of connectedness and compactness..
" ...
'
. P. (iii) The union of an arbitrary collection of open sets S8 an open Bet. the reader that distances are dispensable.. Let X and Y be compact sets in a complete metnc space. t ::.
:.~ . s. . In other words.1.. But if x.
..~:::
.~ Hence the definition carries over to topological spaces...'.f its sub8ets... V such that x E U. It centers on the geometric consequences of the existence of a derivative. True enough.
'
_'
. L Prove that every continuous onetoone mapping of a. The whole section is mainly descriptive. :
. 2. topological mapping of the open disk
lzl <
convergent sequences. If we deci~e to consider the open sets as the primary objects we must postulate asaoms that they have to satisfy. and (i) is trivial. The observant reader will have noticed that most results the preceding sections were formulated in terms of open sets... _. but they are rather convenient: N is a neighborhood of x if there exists an open set U such that x E U and U C N. The following axioms lead to the commonly accepted definition of a topological space:
1. When following this custom it is preferable to denote the mapping by z == "r(t). : .
.
. 11 E V and U V = /2J. and not always convenient to express nearness in termsof distance.::... Arcs and Closed CurtJelf.. Such examples would necessarily be very complicated and would not further the purposes of tills book. ...
The change is reversible if and only if v. we speak of a differentiable or . A commonly used synonym is holomorphic. An arc is piecewise differentiable or piecewise regular if the same conditions hold except for a finite number of values t.ve~ 8Ome·. The opposite arc of z = Z(t). 2
88
a definite
+
When we consider the derivative
f'(z) = lim fez
/t900
+ h)
h
.~. 1 2. 0 ~ t .>'(r) is continuous and. it is customary to denote the restriction by the same letter [.:. it is important to show that certain properties of arcs are invariant under a change of parameter. For greater flexibility of the language it is desirable to introduce the following complement to Definition 10:
a
11. the equation z = t'! it4. in addition. we choose a point til from the interval (a.fez)
+
of a complexvalued function. 0 ~ t ~ 21..! t ~ (.. or a Jordan art. The differentiable or regular character of an arc is invariant under the change of parameter t = tp(r) provided that v. but the rule can be broken if it is clear from the context what the set is. it is no loss o~ generality to consider only the case where 0 is connected. For instance.. This convention saves us from writing down the equa. in the case of a piecewise
+
needed. differentiable or regular closed curve and of a simple closed curve (or Jordan curve) are obvious . A circle 0.enninal point of an arc remain the same after a change of parameter .
I
+
. . Analytic Functions in Regions. sometimes by . depending on the connection. A complexval1. The J!urpose of the shift is to get rid of the distinguished position of the initial point.! t . A junction /(z) is a'lUllytic on an arbitrary p~nt set A if it is the· restriction to A of a function which is analytic in some open sef containing A. .~ified. and . and hence singlevalued. new closed curve whose equation is z = z(t) for to . As a rule one should avoid speaking of an analytic function f(z) without referring to a specific open set n on which it is defined.
. A constant function z(t) defines a point cunJ6.:.ii! il.Ii a) for {l ~ t ~ to {l . .. If the derivative t(t) := x'(t) iy'(t) exists and is ..:open neipboriaood of Zo. z'(t) #.=:: tt. regular change of parameter. a ~ t ~ (I.(. In following this course. the simplest course is to consider two arcs as different as soon as they are given by different equations.
__0·· .It . . The existence of the derivative will therefore have a different meaning depending on whether z is an interior point or a boundary point of A.. The correct definitions of a. the initial and t. ~. for the specific choice of n is usually immaterial as long as it contains A.I. For closed curves a shift of the parameter is defined as follows: If the original equation is z == z(t). and this is its most important purpose it serves rule to distinguish between C and C. since the components of an open set are open. It is important to stress that the open set (} is part of the definition.. is said to be analytic in n if it has a deri~Ve at each point of Q. regardless of whether one equation may arise from the other by &: change of parameter. We give a formal statement of the definition: Definition 10.&.fJ) and define 8..0 the arc iR said to be regular. The way to avoid this is to insist that all analytic functions be defined on open sets.) is defined and has a derivati.z( t). ¢O. A change of the parametrle interval (a.:'i 1. defined on a set ~ in the complex plane. We shall say that the arc is differentiable if z'(t) exists and is continuous (the term continuously dUJerentiable is too unwieldy) i if. This is a ease in which the set Q need not be explicitly mentioned. Definition The last definition is merely an agreement to use a convenient tenni~l~gy. Opposite ares are sometimes denoted by 'Y and . Sometimes one says more explicitly that fez) is complex analytic. Logically. that is to say
regular arc. at these points z(t) shall still be continuous with left and right derivatives which are equal to the left and right limits of z' (t) and.l£d junction f(z). When this is the case. and if fez) is analytic in OJ then the restriction of f to D' is analytic in ll'. can be considen:d as a closed curve with the equation z = a re". if z(tJ = Z(t2) only for h '. which is one of the fonn t = GT + b.1. Another instance in whieh the mention of Il can be auppressed is the phrase: "Let I{z) be analytic at %0" It means that a function /(z. 0 ~ t ~ 1 arises by a reversible change of parameter from the equation z = t it2. as we will.01'0.. a . it is of course understood that Z E A and that the Iimit is with respect to values h such that z + h E A.(:J) can always be brought about by a linear change of parameter.>(") is strictly increasing. defined on an open set 0.II
COMPLEX
ANALYSIS
ANALYTIC
FUNCTIONS
AS MAPPINGS
II
We say that the first equation arises from the second by a change of parameter. Observe that f must first of all be a function. is the arc z .a.tio~ each time it is
+
+
regz~ 00. t ~ Ii and z = z(t . In particular. An are is a closed cuTtle if the end points coincide: z{a) = z(P). also. We will use this standard pfll'ametrization whenever a circle is introduced. . a > o. . originally defined as a locus Iz ~ e] = r. An arc is simple.. If 0' is an open subset of 0.!Ot the arc ~ has So tangent whose direction is determined by arg z'(0.. For instance. for regularity.
~.. consider the set A in the wplane which is defined by the inequa.1 <:: . ~ if It.e"'J.. lv.zsl Wi \ < Ul
Hence
and it follows that w ==' is continuous at Zl.. An anelytLc fwction . or Me cos z. Denote the principal branch by W =U i" = log a. For a given point w.. minimum p on A (Theorem 8.L ~l ..l

+
t:ve
=0
+
+ w!
IZ1 and
IW1
t21
=
Iwi .h + W1.. This minimum is positive. Indeed. has to be specified It .1 has consequently 8..le"1).
Therefore.e. let us prove that it is continuousChoose two points 21. .. . an al"l7f·i f .ii~
av
+
'Zl . . it follows that all values eh are on the same SIde of the real axis.:..l. :~...\ ... ._ the continuity must be proved. . v. for 2 + .
. Therefore.we can define VI .'.ANALYTIC
FUNCTIONS AS MAPPINGS
71
70
COMPLEX ANALYSIS
Although our definition requires all analytie functions to be singlevalued.1C raae of log Z in certain · be proved in the chapter on integration regions.:
if u an. There 18 nothing unique about the · · singlevalued branches have been h w~y ~ which the region and the each time we consider a f . Corollary).. provided that they are restricted to a definite region in which it is possible to select a singlevalued and analytie branch of the function.'
.
. and assume that
where ~ has a positive real part.d v sa tiaf y these equations in 0 and if th · envatives are . 1". In the case of log z we can use the same region fir obtained by excluding the negative real axis.IV .unc~on such 88 log z the choice of the branch define s sing1eval~ an: anaaJunyt~ambentalhfact that ~t is impossible to This will' . _ 1 are both . = l. must lie function of z Fro1m th'e c .. and s given e > 0..~pellB only when z lies on the excluded ot z in ~ eeause n' 18 co~ooted..n o an an ytlC function in 0 satisfy the Cauch y. %2 e fi and denote the corresponding values of w by WI = 'Ul iv._.I are reciprocals and hence real only if z parts of the real axi ~a.... Once the continuity is established the analyticity follows by derivation of the inverse function z = 1O~.· su a value they are all in the u r half smee l IS an analytic branch of log (z can the~fore define o snd e..r..0• S~ 0 mee 1 .. on IDOl ¥ we concl d bo derivative exists and equals l/z._. All the results of Chap II C Se 1 2 .
. the cun 1 ~ i6 = ti.1 11 .d. For instance. .1m !J...'.J
arc eos a == ilog {z +~}
in 0' whose derivative is
o...·red~ to a·eO~~~ l"{Ma
. In this wav we obtai m ai 81 evalu ed analytic function between singl agL ~ part lies
+~
+~
+ ~)
w:~ W.'. &.. . with the notations used in calculus Az ? 0 implies
l!w
+
The infinitely many values of are 008 z h i log (2 +~) In this ~ t e same 88 the values of the h If lin · case we restrict z to the complement 0' of a ~ ~. · real and ~ es x 0' 1. . u e as a ve that the Neither
Th~n
Wt
eU. M oreover z t P e an · en V.w~1 .. .et'lo(g.. m.R·. we may choose for n the complement of the negative real axis z ~ 0. c \lWl ~P: m O~ .. and for sufficiently smell a it is not empty....
"
:'"
'"
.fIJUl...t&ve proved that w is a continuous 4.'Wli ~ s. or U2 etta > I Ytlog 2 ind> Ul > 2. log z. With this choice 10 becomes a singlevalued function in 0. ..• IS never real ill n'. lremwn valid for functions which f alvti . and si . The continuOUS function Ie'" . This set is closed and bounded. an 10 th e latter case Ie"" _ ew.:: e'" _ lIt > in the disk ltD _ w I < e ~d he e"1 > o. this set is indeed open and connected.u.
."
au
OIl
au
8ii ~ . and V 2'1.
~. w. Itt :.21ri.min (P. Then
cannot lie in A.continuous then U.
.1 ~ "11". but this time we have no algebraic identity at our disposal.
z
'. the real and Imaai n. Again. for this would k is it possible that u < 1 2 rna e Ie. = 'Ul with 'U1. ~~nary iemann equations parte
i
+
ax = ay'
Conversely deri 4'. In 0: one and only one of the values of Vi has a positive real part. Hence w.'1 in the first exam I d tb set v z.0'" ' .101 
=
IW1 .
+ ~1
I
(1 + ~
)=
1 Vlz!
with the same branch of Vi. /1w 11m :=:: ~o M
I· _..YZt=T .w:l1 · IWI + 'W:ll
+ 102\ ~
1h
+ Uz > ttl·
v'Z
tZl ..
"
... ·e partial + 1S an . > O..lities ltD .. :.. .._.1 '
'. n . C osen In t ese examples.z.1D Aw40 M
D arc cos z = i z
and we obtain . . z. p is. it is possible to consider such multiplevalued functions as 0.e"'tl 5. oJ ..
.eVil < B"
.zs'
= leI . and define the principal branch of the logarithm by the condition [Irn log zl < 1r.. for A does not contain any point w. Choose & ". and we are forced to use a more general reasoning..Z2 8.i! log 2.1 . _ 1 = i~ ~_ . n p reursr. former case we would obtfti ' . are analytic on an open set' art".:.yzt=T and 2 _. . this ha.. u.~' ti . . ...2' ]8 never _I m . y == 0 and x .
the modulus..
~
We will investigate the meaning of this equation at a point z& == 2(to) with z'(ttJ) '# 0 and f'(zo) '# . Suppose that j(z) analytic and satisfies the condition If(z):ll . ax ax
av
A related property of the mapping is derived by consideration of the modulus \f'(z&)I. Hence /(z) is in any case a constant. that any two points in a. Give a. we can set u = kv with constant k (unless v is identioally zero).11 < 1 in a. . 1.3 we remarked. In general this change of scale will vary from point to point ..:£
X'
(to)
+
where
the partial deriritiris' ~':'taken
i£
veto)
at .i~
ax
ay
=~
ay
+ i ax = 0 ' av
we obtain
This relation ~rts t~at the angle between the directed tangents to "I at Za and to ~ at Wo .. or the argument i8 ctmstant..
u
au + v (tv
ay
ay
== u ~
au + v ~ = 0.3... 2. . we find that w' (t) also exists and is determined by
(1)
w'(t) ~ f'(z{t»)z'(t).
ax ax
and
WIth f(z)
¢
O.
2.
The vanishing of the derivative implies that au/axf au/aYt ilv/ax. if arg f(~) is constant. It is hence independent of the curve 7.~ ~ In' 'terms of
r(t. i~ is clear that both kinds of conforma1ity together imply :he existenee of f (zo)4> It is less obvious that each kind will separately Imply the same result.. In sense as well as in size. 3. But if u' + v2 ::= 0 at a single point it is constantly zero and J(z) vanishes identically. Under this condition the derivative of wet) = f(z(. ~ ~z.) can be expressed in the form
ttl'(to) :
is
. and we conclude again that f must reduce to a constant. and let fez) be defined and ~o.let us a~e that thepartial derivatives ai/ax and aljay are continuous. at least under additional regularity assumptions . For this reason curves through Zo which are tangent to
each other are mapped onto curves with a common tangent at Wo. We have
These equations permit the conclusion au/ax = avjiJx = 0 unless the determinant u2 + v2 vanishes. We conclude that u iv is constant. It follows that 'U and v are constant on any line segment in n which is parallel to one of the coordinate axes.IS equal to arg f(zo). To be more prec~se. In other words. region U. The same is trueiJ either the real part. If not. t ~ (J. tangent at Wo == !(ZI). C~nversely. If z'(t) exists. avj8y are all zero. effected by the transformation w = f(z).
+
(2)
argw'(to)
=
argf(zo)
+ argz(to). f'(z)
= ~ . Same problem for log log z. ~n the equation 10 = wet) = f(z(t» defines an arc "I In the wplane which may be called the image of r. Note that for this theorem it is essential that n is a region. But 11.0' (to) ~ 0" Hence y' has a. ~h~ linear change of sc:Ue a~ Zo. Finally..
If u2
+ v! is constant}
av u+v=o
au. In Sec.. Consider the case of an fez) which is analytic in n.
+
~nd t~ ~eans that any small line segment with one end point at %0 is.
Theorem II.ANALYTIC
FUNCTIONS
AS MAPPINGS
73
72
COMPLEX
ANALYSIS
the following theorem we shall list some simple conditions which have this consequence :
An analytic fU'Mlion in a region 0 whose derivative vanishes identicaUy must reduce to a constant. region can be joined within the region by a polygon whose sides are parallel to the axes. precise definition of a singlevalued branch of ~
in a suitable region. Show that either Re fez) > 0 or Re J(z) < 0 throughout 0. IS Independent of the direction. contracted or expanded in the ratio If(zo)t. and prove that it is analytic.. . ConjorlJUJl Mapping~ Suppose that an are 'Y with the equation z = z(t)J a =::.. In view of this p~perty the mapping by w = f{z) is said to be conformal at all points
and hence f(z) must be constant. we can only assert that fez) is constant on each component of nO. Mo~vert two curves whic:h form an angle at to are mapped upon curves forming the same angle. The first conclusion is that 1. the imaginary part.:
::<::"
. in connection with Theorem 3. '.ZGI
= If(zo)t
'
v'l="Z
1.kv is the real part of (1 + ik)!.ntinuollil in D.
EXERCISES
1im \J(z) .Y ·tftiS can 
~. and its direction is detennined by
o.
If u or v is constant. is contained in a region g.!(zo) I .. In t~ limit..
._.::~.
..:.
T
_:
~
1:.
I ~.O.• 1 . ·1.. . •
I.'I· iJy} e I()to
1( aj at) J( + 2 iJx + t·ay ~ tD. This is the simple and fruitful idea used by Riemann when he introduced the generaJ."' ~':::""': :u. It is helpful to think of the image of the whole region as a transparent film which partly covers itself....•• .
.. ~~ x>. is given by
+
L('Y)
=
J" V'x'(I)
s
2
+ y'(t)2
=
dt =
w(t)
J"
4
Iz'(t) I dt....ill~te
2~4.'_..... If the mapping of n by w = J(z) is topological..~:...•
: .
•
. .& 0 is sufficient to conclude that the ma~ ping is topological if it is restricted to a sufficiently small neighborhood of Zo.' .
and in the second case
at · at ax == . 31.r~: . but the images overlap. :...
. ".... Quite similarly.':":'":" ..:.~" .ls for integration with respect to ..Lt
:1 .
••
••
. _..."'": :. infinitesimal areas will clearly be multiplied by tf(z) 12.
..
••
.:.1 (t) ....
•
. 1+ I: . '."... On a circle the modulus is constant only if the radius vanishes or if the center Iies at the origin.. r"'IIJ
\i
~tz.. the condition that the change .. .._.
in
It is customary to use the shorter notations
(5) L(oy) ""
f ..
r..: : . ':<.~~~~::.
~_'t::..r ~
.L ..
.
..
......:.. We have found that under a conformal mapping fez) the length of an infinitesimal line segment at the point z is multiplied by the factor If(z) I. simpler proof of this important theorem. . We know from calculus that the length of a differentiable arc 'Y with the equation z = z(t) = x{t) iy(t).As a....y) at the point ZQ is 11'(zo)12 and hence 'F. ~~.~
•••
••¥
•••
: . :':i
:..J1.OJfor then the derivative of the inverse function must be equal to 1//1(%) at the point· z ~ jl(lO).. ~\ ::..: .. ' . . ~r' s"" .~.
..
The image curve 'Y' is determined by w ur'(t) = f (z( t) )1.
L( "t')
=
f
If'(z)
~
Ildzl·
Observe that in complex notation the calculus symbol r.
.. n.1 "Jl.::.

r
'I
:
:::::"
. . arg t/(tfJ)~ The expression (3)
srg [w'(to)/z'(to)1
must be independent
of
must therefore have a constant argument."""'. Later we shall present a.O .
~:~:...=
..:
r ..··rv...
."
··1.~~~: ~. •
' .. 'J' . we cannot assert th&t the mapping of the whole region is necessarily topological. It evidently preserves the size but reverses the sense of angles.of scale shall be the same in all directions implies that the expression (3) has a constant modulus.1 . A mapping by the conjugate of an analytic function with a nonvanishing derivative is said to be indireetly conformal. I~ · . ..': :::
."J ~~ ~ .... ~ =: . '"~.This follows easily if J'(2) F."'/:'"
I ••••
.. . The knowledge thatJ'(zo) =.
..
··1
•
. We shall prove later that jl(Z) can ne~er vanish the case of a topological mapping by an analytic function .
~
O:.::
.
.)
If angles are preserved.y).•
'A
__
.rg z'(to) is allowed to vary. ~: ..!: ~.~.~..~ .') =
= f(z(t») with the derivative
f:
If(z(t» Hz'(t) I til. Idzl. The argument cannot be constant on this circle unless its radius vanishes.r.:~~_. :1
~. I ._.~ *"~7'1=:.' ..:'''. .. v := v(x.· ~ .••• • _ •• .''I=:: .
dy .. for the Jacobian of the functions u = u(x.' ~r. .. .. ~
. : 1..:J~... .. ~ /i·· '<. To ..
:
'.:~~:.:
• •
.~.. r.•.. .
..•.....ized regions now known as Riemann surfoce«. ..::.. .(....~.:
• ••
_
.. This follows by the theorem on implicit functions known from the calcuIus. In the first case we obtain (4). ~. 1
.~_
..
JIrlz
•
...:.1
. • ~:~
••
.••
••.
"'.... But even if /'(z) ¢ 0 throughout the region 0.
••••
..
•
.
(4)
at ~= ax
t
· aj
ay
which is the complex form of the CauchyRiemann equations .. :. :
~_..:
..1.... Let us put this on a rigorous basis.
•...~
I.
.~...:. .. r "_ . Now let E be 8 point set in the plane whose srea
A(E) .".. .
.:."... the point represented by (3) describes a circle having the radius !l(aflax) icaJ/iJy)\.
Doubly covered region. J •• ..
•
..
regions 01 and O2 are one to one.arc length is replaced by Jdz I.
_ ·~"I jr I..:.. ·t\~. Here the mappings of the sub
=
2
1(
aJ al\ ax ..::~.... then the inverse function z = jI(W) is also analytic.1 . ..._ i.
: :.. 1 ~~':I.ay
·The last equation expresses the faCt that fez) is analytic. Because the distortion is the same in all directions. ·1· •.: : : .\ .
. I •• :~I·::.r....·. i' :a.1
• ...r ·1··
'... ~.'G ". .:. Its length is thus
L(..~ y. ~...... • 1..~
:
:~. .: ···.
.
. a ~ t ~ b..• :
••
•• _..
Ii c..
_.. " ~.!·.~. Length and Area.
..:
: ..:_:.
. •••
. •••
•• ..· .: ~.r"r_' r ··::1"· :· "'1' 1 :. .1
~..~..i~~~~:"!. ·5
~.74
COMPLEX
ANALYSIS
ANALYTIC
FUNCTIONS AS MAPPINGS
75
be rewritten as I() w to
what may happen we refer to Fig. ~. and hence we must have
+
F'IG.:.". ~#:i"t~· ...:c
"I:
~.. ."+.:·. . ••
..:.. ».
If we use only normalized representations. is called an inversion.
All linear transformations form Or group.J
The formulas (5) and (6) have important applications in the part of complex analysis that is frequently referred to as geometric function theory. and for that reason their importance goes far beyond serving as simple examples of conformal mappings.. For linear transformations we shall usually replace the notation 8(z)
a parallel aamlaiion. I~t_:
. The third transformation. and (8) identifies the group of linear transformations with the onedimensional projective group over the complex numbers" usually denoted by P(l.
• ••
1 .
.. if Be
(8)
(7) is said to be normalized if ad ..
transformation has two normalized represenother by changing the signs of the coefficients./~~ . for it will equip him with simple but very valuable techniques .C).:
•
•
••
. except that there are two opposite matrices corresponding to the same linear transformation. bijective differentiable mapping.
10 =: Z
+ «. is a Totatioo if Ikl = 1 and a homotheti« transformation if k > o. then by virtue of the CauchyRiemann equations. With the latter interpretation it becomes obvious that S is a topological mapping of the extended plane onto itself. LINEAR TRANSFORMATIONS
(a
1
bl)
dl
Cl
(at
C..y) is a. the topology being defined by distances on the Riemann sphere. W = 'lOt/WI we find that
WI W2
But if f(z)
UJJ.
by 8z. The reader will do well to pay particular attention to this geometric aspect.4 that a linear fractional transformation
(7)
w
= B(z) = ~
az
+b ce + d
dw .
. If fez) = u(x.. C)).. 2.0 we can set k = lkl · k/lkJ. We have already remarked in Chap. the associative Jaw (8182)83 = Sl(SttBi) holds for arbitrary transformations} the identity w == z is a linear transformation.t'%
'f'(z)
we obtain
(6)
r!
mapping of an open set containing E. C lbz dId!
+
Of all analytic functions the firstorder rational functions have the simplest mapping properties} for they define mappings of the extended plane onto itself which are at the same time conformal and topological. .. The ratios %1 :Z2 ¢ 0:0 are the points of the complex projective Iine.. :. Indeed. and
or} in matrix notation.. is called
az
+ b be ~ ad .. then by the rule for changing integration variables the area of the image Et = feE) is given by
A(E'}
=
fJ
1'U"v~ 
'UJ':c! dx dy.
... For arbitrary complex k #. See.b
.
=
Cla!
+ dle:
a1b2" + bid!). If we use subscripts to distinguish between the matrices that correspond to 81 82 it is immediate that 8182 belongs to the matrix product
.. obtained from each A convenient way to homogeneous coordinates.
:.
= ==
azl
CZl
+ dz2
+ bZ
I
A (E') =
ff
E
Ir(z)
JI dx dy.. we can also identify it with the group of twobytwo matrices with determinant 1 (denoted 8L(2. w :::::. The Linear Group. : .y) + iv(x.71
COMPLEX
ANALYSIS
ANALYTIC
FUNCTIONS
AS MAPPINGS
77
can be evaluated as Or double Riemann integral._
:
.. The linear transformations have also very remarkable geometric properties.
...:_:
:. except for remarking that the simplest linear transformations belong to matrices of the form
with ad ..
d2
b2) (ala! + b1c. H c ¢ 0 we can write
The first of these. ~
is a conformal
u. The representation It is clear that every linear tations. :··0 C2 + d == c!(z + d/t)·.
The main advantage of this notation is that it leads to a simple determination of a composite transformation w = StS2Z.1. w = kz.be = 1. express a linear transformation is by use of If we write z == Zl/Z2.~. The second.. 1.
3..~Io. We shall make no further use of the matrix notation. w := l/z.be "# 0 has an inverse
z = Sl(W) =
cw+a
The special values S( co) = ale and 8{ dIe) = co can be introduced either by convention or as limits for z ~ ao and z + dIe. and the inverse of a linear transformation is linear. and hence w == kz can be represented as the result of a hom 0the tie transformation followed by a rotation.
3.
cd '#. Tz = (Z. Tz" into 1. The proof is immediate.
In general it is of course necessary to solve this equation with respect to w. 
. :"
...) '. ==
00
the transformation
Z
reduces to
ZZ3
Za
z . there exists SMear transformation S which carries them into 1.(.ZJ.Z.Z" )
= arg
Zl Zt _
Z3
z_(
.. an inversion.
The er088 ratio (Z1. but this is unimportant . The correspondence must indeed be given by
(W.
.3
with the same property.obtain
(ae ·ca)l1Dt~
+ (ad~~
:i :.Z:2.
.Tzs1 TzJ == (Zt.:I:.~.~. .Zt. a rotation.
88
are distinct...Z2..Z4) is real on the image of the real axis under the
=
Tw.Z2.. Za to prescribed positions W].
":
... A convenany three of the points are
0 this is the equation of 8 straight line.
.Z..Z"
ralio (ZI.2.
Theorem 13. Show that any linear transformation which transforms the real axis into itself can be written with real coefficients.)~
7l1Tse. .&) . and a homothetic transformation followed by another translation.
or z.
.... in the extended plane.Zt) is real
if
and only
if the four
This is evident by elementary geometry.z"
respectively.}.
EXERCISES
L Prove that the reflection
The cross ratio is invariant under linear transformations. then. 0.. ':
cb)w
+ (bc·_
J. ZI.< .
.::.. i is not a linear transformation.. 
If
ac _ ca ==
.. Indeed. Tzi) .. By definition we have hence
(TZIJ Tz'l. T2Ttz and
=Z
z+2
+ 3'
Tzg == z
+ l'
==
(ZlJZ2.
In more
z" are diatincl points in the extended plane and T any linear trans/ormation.'._:.
Definition 12...~. w.
df2)w
+ bd
_ db ::: o~
The definition is meaningful only ~ tional value can be introduced as soon distinct..Wa)
fixed and preserves all distances is either 8 rotation or a rotation followed by reflexion in the real axis.
.• ..a.ZI.:::: STl(Tzt)
::: 8z1
z . : .). for if Sz :."
.71
COMPLEX
ANALYSIS
ANALYTIC
FUNCTIONS
AS MAPPINGS
71
and this decomposition shows that the most genersllinear transformsion is composed by 8. this condition is of the form .. If at ... and we would have 8 = Til We conclude that S is uniquely determined.. W2.0 we can
':. The values of 10 == TIz for real z satisfy the equation Tw Explicitly. TZ2.. the inversion falls out and the last translation is not needed . Tza. r .Z...':_~~. then (Tz]..%a
=
(ZJZLZ2~Z3) .
aw+b .. 3. 021.. 
Z.ZI. The
CT088
If T were another linear tl.) ill the imtlfJ6 of Zl under the
00. aw+b cw+d
and if the points lie on a circle this difference of angles is either 0 or +1rt depending on the relative location.
2:1.i '. . Tza. 0. Given three distinct points 2.ZS. .8S. If none of the points is co. OJ co invariant. Z.: ~': s .Z3.ZS.. • .. then STI would leave 1. If c == 0.%I)Z.:. translation.arg
z! Z2 
Za
Z41
.~.. 00 in this order.Z. for under this conditioIl: the coefficient ad _ c~ cannot also vanish.. 0.TrUs Z'p..%f into 1.:::: (Z.msformation
transformation TI and nowhere else.. precise formulation: Theoeem 12.
L Prove that the most general transformation which leaves the origin
With the help of this property we can immediately write down the linear transformation which carries three given points %1.
cW+a
linear transformation
which c.. 28..Z. The Cross Ratio.· 1/
ZIJ Z2.ZI.. 00.~
.WltW'2. for we obtain
arg
(Zl. B will be given by
(9) If
Z2.
2
T1z
find T1T2Z. For an analytic proof we need only show that the image of the real axis under any linear transformation is either a circle or B straight line....8 7'1 carries Tz1. Direct calculation shows that this is true only for the identity transformation. .
point8 lie on a circle or on a straight line.
2.
By cross multiplication we .
10
COMPLEX
ANALYSIS
ANALYTIC
FUNCTIONS
AS MAPPINGS
11
divide by this coefficient and complete the square, putation we obtain
After a simple com
The po,m, z and. a* are said to be 81Immetric tJJith respect to the circle C through z 1, Z:,. Z 3 if and ooly if (z *,z 1ft i,Z,) :::;;(ZtZ) ,z1~ZI)..
Definition 13. The points on C, and only those, are symmetric to themselves. The mapping which carries z into z* is 8 onetoone correspondence and is called reflection with respect to C. Two reflections will evidently result in 8 linear transformation .. • We wish to investigate the geometric significance of symmetry~ Suppose first that C is a straight line. Then we can choose %3 = co and the condition for symmetry becomes
(10)
which is the equation of a circle. The last result makes it clear that we should not, in the linear transformations, distinguish between circles and straight further justification was found in the fact that both correspond on the Riemann sphere. Accordingly we shallagree to use circle in this wider sense. t
theory of lin~~ A to circles the word
z* 
it!
Z
is
The following is an immediate corollary of Theorems 12 and 13:
Tbeorem 14. A linear tramfornt.ation carrie. circles inlo circles.
EXERCISES
Taking absolute values we obtain 1:*  ZIJ ~ any finite point on C, and we conclude that from all points on C. By (10) we have further
1.1' 8tl. Here
%
can be and z· are equidistant
Zt
which carries 01 i, i into 1, 1, O~ 2. Express the cross ratios corresponding to the 24 permutations of four points in terms of ).. == (Zl,Z2,ZS,Zt_}. I.. If the consecutive vertices ZI, %1, ZI, z, of a quadrilateral lie on a circle, prove that
1. Find the linear transformation
1Zl

Za\ .. \ Z2  z41 = IZI  Ztl · 1Z3

z41 + Iz!

zal • Iz1  z41
and interpret the result geometrically. '" Show that any four distinct points can be carried by a linear . transformation. to positions 1, 1~ k, where the value of k depends on k, the points.. How many solutions are there, and how are they related? 3.3. Symmetry. The points z and z are symmetric with respect to the real axis, A linear transformation with real coefficients carries the real axis into itself and z,. z into points which' are again symmetric. More generally, if a linear transformation T carries the real axis into & circle C,. we shall say that the points w = Ta and w* == Ti are symmetric with respect to C. This is a relation between w, w* and (J which does not depend on T. For if 8 is another transformation which carries the real axis into C, then SIT is a real transformation, and hence S'w = SlTz and 8lw* = SIT! are also conjugate, Symmetry can thus be defined in the following terms:
and hence z and z* are in different balf planes determined by c.. We t leave to the reader to prove that C is the bisecting normal of the segment between z and ' ... Consider now the ease of a finite circle C of center a and radius R. Systematic use of the invariance of the cross ratio allows us to conclude 88 follows;
(ZrZl,Zt,Za)
==
(J 
=
(z  0,21  a ,Zt a, R't J RZ,
Z1 

4,,%:11 
a)
.
a Zs  a z,  a
R2)::::;:
(
.I 
RI
a
,Zl
_
a,Sl 
a J'I
'
a)
+a
"'" (.
~!
a + a,zl,iI'IJSI)
This equation shows that the symmetric point of z is z* = RI I (I  a) or that a: and 2* satisfy the relation
(11)
(,.  4)(1  ti)
== Rt.
The product 1:*  01 Iz  at of the distances to the center is hence R2. Further, the ratio (z·  a}/(z  a) is positive, which means that z and z* ~ situated on the same half line from B.. There is a simple geometric construction for the symmetric point of z (Figt 32) .. We note that the 8ymmetric point of a is DO.
t This agreement
will be in force only when dealing with linear transfonnations.

t Unl_ .,. ~cdde &iid He 0. C. ..':
I
.. .
',:
",
:~~.:...
.
. ....
.
. ~:.':'.:,' '.~.
12
COIIPLEX
ANALYSIS
ANALYTIC
FUNCTIONS
AS MAPPINGS
13
Fie..12.
Reflection in
8
circle.
in the circle lz  21 == 1.. L Carry out the reflections in the preceding exercise by geometric construction .. ... Find the linear transformation which carries the circle ,z\ = 2 into 18 + l' ~ 1J the point 2 into the origin, and the origin into i. . ~ Find the most general linear transfonnation of the circle lzl = R Into It8elf.. 6.. SupJXI8e that a linear transfonnation carries one pair of concentric circles into another pair of concentric circles, Prove that the ratios of the radii must be the same.
2. Reflect the imaginary
axis, the line x
=
y, and the circle
1zt ==
1
(The symmetry principle .. II a linear transformation ) carries a circle 91 into a circle C" then it transforms any pair of symmetric points 1Dith respect to (J1 into a pair of symm.6tric points with respect to C2•
Theorem
15.
Find ~ lin~a.r transform~tion which carries lzl == 1and iz Into concentric CIrcles. What IS the ratio of the radii? .. Same problem for Izl = 1 and x = 2~
_. 7.
11 == }
1.4. Oriented Circle..
Briefly, linear transformations preserve SYIIlmetry.. If C1 or O2 is the real axis, the principle follows from the definition of symmetrY.. In the general case the assertion follows by use of an intermediate transformation which carries C1 into the real axis .. There are two ways in which the principle of symmetry can be used. If the images of Z And C under a certain linear transformation are known, then the principle allows us to find the image of z*. On the other hand, if the images of z and z* are known, we conclude that the image of C must be a line of symmetry of these images. While this is not enough to determine the image of C,. the information we gain is nevertheless ~ valuable. The principle of symmetry is put to practical use in the problem of finding the linear transformations which carry a circle C into a circle C' .. We can always determine the transformation by requiring that three points %1, $2,. %3 on C go over into three points Wt, WI, Ws on C'; the transformation is then (W,Wl,W2,Wa) = (Z,ZhZJJZ1}. But the transformation is . also determined if we prescribe that a point %1 on C shall correspond to 8r point WI on C' and that 8 point z, not on C shall be carried Into a point Wt not on G1• We know then that (the symme~ric point :of z, with respect to C) must correspond to (the symmetric point of WI with respect to C').. Hence the transformation will be obtained from the relation (10, WI, W2, w:) = (z,zl,zt,zi) ..
Because B(z) is analytic and
S (z)
,
=
(12
adbe
+ d)*
;¢
0
the mapping
== 8(z) is conformal for z # die and 00; It follows that a. pair of intersecting circles are mapped on circles that include the same angle. In addition, the sense of an angle is preserved. From an intuitive: point of view this means that right and left are preserved, but a more
to
precise formulation is desirable .. . An orientation of· a circle (J is determined by an ordered triple of points 21,''2,.1. on C.. With respect to this orientation a point z not on C is said to lie to the right of a if 1m (Z,'I,'1,81) > 0 and to the left of (J is ~ (Z,ZlJZI'~*) < 0 (this cheeks with· everyday use because (i,l,O, QO) = ",. It 18 essential to show that there are only two different orientationa, By this e mean that the distinction between left and right is the same for all triples, while the meaning may be reversed. Since the cross ratio is invariant, it is sufficient to consider the case where C is the real axis, Then
w:
w:
zt
(z,ZljzJ,%.)
+6 = cz + d
az
B
can be written with real coefficients, and
simple calculation gives
Im (2,%I,Z2,Z.)
= Icz + dl' 1m s.
adbe
EXERCISES
L Prove
that every reflection carries circles into circles ..
L
t,.!ii.~f~;~~P1.~~;:~.L~~~.~ :.!l~~:~~~ .._
....
:.. 11:" '. ~.. .. :.. ~ I.. • I•
We recognize that the distinction between right 8Dd left is the same as ~n.~; ~JMl~ ~~; ~w:er.~alfp1&1U'. Which is whieh
~~. .~
.,·~_i;;:
.. :
.",
,.
_
..:
...
:. >:i~<\~_:·':..
""',
.,
:'.<.:.
. "'::.".::: ...:.. : •.. 1
.. .:.
: ;.::.'... ~.·:~:·»;iYi»:,~:.<':~..: /~.. '~.,~ .
• .... .'... '; 'I' .
.
..
,
,',:'
.
'.'
.
,'I:
.
:.. :~: '. ... I.~=.":.. ~~
.....:
':""\
.'iI.~.~,i.'::;',~ .:~. ~·~d: :·~·t~·~r~~..:f~:ntXf~
..= .' ~. _.• :.... : .. :_ ......
..
., .:1':.
..
·COMPLEX
ANALYSIS
ANALYTIC
FUNCTIONS AS IIAPPINGS
85
A linear transformation 8 carries the oriented circle C into a. circle which we orient through the triple 8z1, SZ'l, 8z3.. From the invarianee of the cross ratio it follows that the left and right of C will be mapped on the left and right of the image circle. If two circles are tangent to each other, their orientations can be compared, Indeed, we can use a linear transformation which throws their common point to 00. The circles become parallel straight lines, and we know how to compare the directions of parallel lines .. In the geometric representation the orientation 21, Z2, z. can be indicated by an arrow which points from %1 over Zt to. z.. With the usual choice of the coordinate system left and right will have their customary meaning with respect to this arrow .. When the finite plane is considered as part of the extended plane, the point at infinity is distinguished. We can therefore define an absolute positive orientation of all finite circles by the requirement that 00 should lie to the right of the oriented circles. The points to the left are said to form the inside of the circle and the points to the right form its outside.
EXERCISES
,.
.;.
..
..{: ..
... I:
I~ .
circles through tl and b. On the other hand, the concentric circles about the origin, ltoJ = P, correspond to circles with the equation
I
zz
al b
:: p/lkl·
These are the circle8 oj·Apallonius with limit points a and b. By their equation they are the loci of points whose distances from a and b have a constant ratio, Denote by 0. the circles through a, b and by Cz the circles of ApolIonius with these limit points. The configuration (Fig. 33) formed by all the circles (J. and C2 will be referred to as the t:ircul4r net or the Steiner eirde« determined by a and b. It has many Intereeting Properties of which we shall list a few; 1. There is exactly one O. and one C2 through each point in the plane with the exception of the limit points. 2.. Every C1 meets every Ct under right angles .. 3. Reflection in a C1 transforms every C2 into itself and every O. into another C.. Reflection in a Ot transforms every (1. into itself and every CI into another Ct... 4. The limit points are symmetric with respect to each C! but not with respect to any other circle, )
are points on & circle, show that ZlJ Z3,. "4 and Z2, ZI, 2:, determine the 8811le orientation if and only if (Zt,.Zt"Z" z.) > o. 2. Prove that a tangent to 8 circle is perpendicular to the radius through the point of contact (in this connection B tangent should be defined 88 a straight line with only one point in common with the circle). S. Verify that the inside of the circle l2  e] == R is formed by all points z with Iz  al < R. ~ ... . ... The angle between two oriented. circles at a. point of Intersection 18 defined 88 the angle between the tangents at that point, equipped with the same orientation. Prove by analytic reasoning, rather than geometric inspection, that the angles at the two points of intersection are opposite to each other.
1. If
%1, %2, Za, Z,
3.5. Fa.milies oj Circles. A great deal can be done toward the visual . ization of linear transformations by the introduction of certain families of circles which may be thought of as coordinate lineaIn a circular coordinate system. Consider 8 linear transformation of the form
w
=
k.. za ., zb
..... ..... Steiner circle& ..
Here. z == a corresponds to .w == 0 and z = b to w :: GO. ~t fO~OW8 that the straight lines through the origin of the tDplane are rmagee of the
.
..
.
.
 ..... ,. . : .....
..
'
. . . ...
:' '..
. . •''.. •
. ·....' ..' ,,: (, .:\;' ,';,_',._, ·:·.::>;\;'i"~\:>:: ,:i,~";':;_" :;.;4:"';,.~,....,;;,i£:i~::;;1,;;;;;:":> :(i~ :,;_;_;:;"_, ;
:
:'.
::
.:.,
'J
..,"'~.: ,. I' • . ._._II ••••_.. I: ..... ,..•• I',.. , 0'·  >. . '
.,. =:.~ :··~··T·
:,. I
','_ :.'" ..''.':~'''.''
I': ",
••
.,. I ·1···· ,. _." .....:r<:,. , '.•• ~,.. . .... .. ....,... _. • ••• ".  ~....... . . .:!, <~:. .. ;or'.. _...!I;': ..
: •• ,' .. ~.: •
.
'.
_.
'
:.:.::.:..~")
::'.:::'11.':
~..:.::;7:~~:~.~::
.. ':{;":_":'::':i.: ::.~.::~~~.:~ .. r"::' .: ", " ,:
',1 •
: .,.,.~
I
..•.
•
:11''
!!:'=t?~:... i ..··.,, :.it ~.. l .....:.:.:
',', :
f:.:,.;.._(,._ ~:.T..... ." _ 
~:<:o..
r.~ ~_"~"ir. :.11: ., ... ::':1
~~~ .:...y.
I
•
COMPLEX
ANAL V... J. h' = b. .. if w = u iv the lines u . The fixed .. The corresponding... In particular.. b.'. Cleerly. oints of a linear transformation are found by solving the p equation
'. ~ _
.1)2 = 4fJ. We shall denote the images of the lines v = constant by Cit the circles of the other fa. ~
'. .
.~'.. Since the properties are invariant under linear transformations.. ' '. If a.... The condition for this is (a . b are said to be fixed points of T.
<
We can then use the Steiner circles determined by G..'
'PI "
. . b into a' JI b' it can be written in the form
(12)
.'" .... It is important to note...
.I
.... The transformation is then said to be hyperbolic~ When k increases the points Tz. If the equation (13) is found to have two distinct roots a and bJ the
transformation
can be written in the form
•
t ~.nlily by Ct.
.~.. wb'zb
.
)' )
..llellines correspond to mutually tangent circles .: . IL
. ~_. We can write any linear transformation in the form (12) with arbitrary at b and use the two circular nets to great advantage._1
.. The case C~ = 01 occurs when lk\ = 1. ~:' ... moreover....=k
1Da
wb
za · zb
.:.: constant and v constant correspond to two families of mutually tangent circles which intersect at right angles (Fig.. transformation to == Tz carries G.
:
.
.
..
~r ..' . . This configuration can be considered as a degenerate set of Steiner circles. the points Tz move along the circles a2...:. Under these circumstenees the whole circular net will be mapped upon itself.
... It is determined by the point a and the tangent to one of the families of circles.
. .. z # a.
. The situation is particularly simple if a' == a. b'.e.. .
It is evident that straight lines in the wplane correspond to circles through a... The special cases in which all C 1or all C'2are mapped upon themselves are particularly important.=. :.. its direction is given by arg w.. will flow along the circles a1 toward h.IS
ANALYTIC
FUNCTIONS
AS MAPPINGS
17
These properties are all trivial when the limit points are 0 and (tJ.
• c •
: .' • ~ ..a.
_w~_a_____ _ k . b to discuss the nature of the transformation.
". that the method is by no means restricted to this case.. "
. The value of k serves to identify the image circles C~ and C~.. i. It may happen.al/lz . but the orientation is reversed).
. para. .. Transformations with this property are called elliptu. Consider the transformation
W=
za 
w
+c . The consideration of this How provides a very clear picture of a hyperbolic transformation.' 
I
. A linear transformation with coinciding fixed points is said to be parabolic. 34).. however.
Degenerate Steiner Cll'cles. when the C1 are lines through the origin and the 01 concentric circles. they must continue to hold in the general case.For the discussion of parabolic transformations it is desirable to introduce still another type of circular net.
.. Z .
+
z..
.
(13)
z==
az+~ • 'YZ + a
In general this is a quadratic equation 'with two roots.. dow circulates about a and b in dift'erent directions. Then a. if 'Y = 0 one of the fixed points is co..'
It is clear that T transforms the circles C1 and CI into circles C~ and C~ with the limit points a'. with appropriate orientations Cl forms the . however. The general linear transformation with two fixed points is the product of & hyperbolic and an elliptic transformation with the same fixed points.. and it is convenient to represent z and Tz in the..':
~.same plane.. angle arg k with its image C~J and the quotient of the constant' ratios Iz .... ". We have C~ = C1 for all 01 if k > 0 (if k < 0 the circles are still the same.bl on C~ and Ot is 'kl. When arg "k varies..
.
:Jt {.... Indeed.. " . the line v = 1m e corresponds to the tangent of the circles el. that the roots coincide.
. and we can hence suppose that e is real.of Uiaad t'_
.:::: for z
ft
. In more general cases it is advisable to begin with 8 study of the image curves of the lines % :::. "::...'
:~~:. In (14) 8 multiplicative factor is arbitrary.IQ
+ c. One of the most fruitful ways is to study the correspondence of curves induced by the point transformation. . the other the repellent fix:d porn..nsformations.ttl ? What happens in the parabolic case.!iii
_'"
. In this section we shall consider those mappings which can be defined by elementary functions. hyperbolic..
COMPLEX
ANALYSIS
.. (The limit is the aliract'ive. .
L Show
which satisfies 8 z :.: XI and y ~ '110.... or parabolic? 2.
(14)
of transformations depends on a certain number of real parameters. In addition." . it can well be compared with the visualization of a real function by its graph... .nalytic function to = fez).. fixed point as n :+ CD.J
. One of the most important problems is to determine the conformal mappings of one region onto another..
4.. ..orthogonal. 11 acts as a parameter and can be eliminated or retained !~rding to convenience . which we shall not try to make precise. '.. we naturally wish to gain information about the specific geometric properties of the mapping.
z == 2z _ l'
to
== 3z _ l'
2z
to
=
az4
z _ 1'
w==
z
22
. . It is therefore natural that all questions connected with conformal mapping have received a great deal of attention. determined by a' and w'. In an obvious way. II ~ v(. into curves of a family of wellknown character.. in ·the . If we writeJ(a) ~ u(z.. special case.. We suppose now that a == a' is the only fixed point. A linear transformation that is neither hyperbolic.
It is clear that the circles 01 and C2 are carried into the circles C~ and C.'
".and are called. elliptic.
.tI} .._.Y).== ±2.
By this transformation the configuration consisting of the circles eland O2 is mapped upon itseH. Any such information will strengthen our visual conception of the mapping... the image of z == Xo is given by the parametric equations u ~ 1£(xo.tion
Se=
lIZ
+ b' cz + d
oj Level Curves. lrt__ ... Find all linear transformations which represent
rotations of the
Riemann sphere.y). The special properties of the function fez) may express themselves in the fact that certain simple curves are transformed.
:~t Wf~
.u. 8 family
. ELEMENTARY CONFORMAL
MAPPINGS
. They.
Such was the case for mappings by linear tra.. What happens when n + . Topther. provided that straight lines are included as 8.y) + iv(z.'...~ ~." .
"
.. toa
61
== za
w
+c. Then every CJ is mapped upon it8eIf and the ~bolic transfonnation can be considered as a 80w along the circles Ct.. parabolic if a + d . .
Is any of these transformations elliptic. we . to tz\ = 1 and
I
lz 
1\= 4. 3 that a linear tranSformation carries circles into circles. :... . If S is hyperbolic or loxodromic.... progress in this direction has increased our knowledge of analytic functions considerably. the curves form an orthogonal net in the tDplane.zo.plane..1. L Find all circles which are orthogonal. . nor parabolic is said to be loxodrorntea
EXERCISES :L Find the fixed points of the linear transformations
W
The conformal mapping associated with an analytic function affords an excellent visualization of the properties of the latter. The
USB
are normalized by ad.. Then w :: fIJ' and we can write . ~ilarly... show that S"Z converges to a.
.. conformal mapping enters naturally in many branches of mathematical physics and in this way accounts for the immediate usefulness of complexfunction theory.
7. the same for all z. hyperbolic if a + d < 2 or >2" that a Iinear transformation some integer n is necessarily elliptic.. We proved in See....may eonsider the CUrves' u(~~lI)' . except! when z coincides with the ot~er fixed point..== toa
OIJ
c.' The image of 71 1/1 is determined in the .. How many real parameters are there in the family of all Iinear transformations? How many in the families of hyperbolic..)
S. parabolic transformations? How many linear transformations leave a given circle C invariant?
._.~I'~
~"
~:
.. By consideration of the Steiner circles it was possible to obtain 8 complete picture of the correspondence. 0 way. elliptic..
ANALYTIC
FUNCTIONS
AS IIAPPINGS
II
Any transformation
which carries a into a' can be written in the form
. When a conformal mapping is defined by an explicit a.be = 1~ Show that 8 is elliptic if and only if 2 < a + d < 2.1I).·U. and v(z.. Suppose that the coefficients of the transforma.'
.
. In general we can therefore only consider the mapping of an anguler sector onto another. There is no point in deriving their equations.rpl ~ 2r. i :\ ..qJJ).tive
Dealo.
. .rg z satisfies the inequality
(15)
It is easy to show that S(~l.
.. and in this case S(fPl.((J.. For a ¢ 1 the transform. shaH be formed by all points Z .yf) k2 .:I! 0 such that one value of .
w z
For B different family of image curves consider the cireleslw ..~\.
.....
. The mapping is conformal at all points z .\
~:
. line x == Zo > 0 ."1 + n 2r} is geometrically identical with S«(/J1.
.
":.::= Vo are equilateral hyperbolas with the diagonals and the coordinate axes for a.) but may determine a different branch of Z". On the other hand} the image of x == Xo is Vi 4x5<xI ... and then we may as well suppose that a is positive.d OJ but an angle 8 at the origin is transformed into an angle a9.. For instance. and if a is fractional ~ is noteven singlevalued.qJ2J is a region. Both families represent parabolas with the focus at the origin whose sxes are pointed in the negative and positive direction of the uaxis.. family.. .
.
+
and represents a family of Iemniseates with the focal points ± 1 The orthogonal family is represented. if we follow the change of arg w when z traces the. ..&tion of the whole plane is not one to one. : >.. the image curve must have a loop (Fig~ ~).
.
where arg z stands for the value of the argument singled out by the condition (15).
. by
t
<I
Xl 
y' == 2hxy
+1
=
and consists of all equilateral hyperbolas with center at the origin which pass through the points ± 1..y! and v == 2xy. In the case of the third power 10 = Z3 the level curves in both planes are cubic ~es.. 35 and 36~. This function is analytic with the nonvanishing deriva. consider only the case of real a.
arg w ==
a
arg z
PIG.symptotes.. Their orthogonality is wellknown from analytic geometry .. . J5. we find that ... .
. Since
We
arg
'to'
= W=
a arg
'zJa
z
concentric circles about the origin are transformed into circles of the same .arpl) in the wplane. Let us consider the mapping w = z' in detail.> 0 must have the form indicated in Fig..11 = k in the wpl&ne. It should be observed that S(f{Jl n · 211". .¥'l) is mapped onto the sector B(a~l. and half lines from the origin correspond toother balf lines..~l) S 2. Since 11 == %1 .10
COMPLEX
ANALYSIS
ANALYTIC
FUNCTIONS
AS MAPPINGS
11
In other cases it may be more convenient to use polar coordinates and study the images of concentric circles and straight lines through the origin. where 0 < 9'1 .. of w =: z« is defined by the condition
In this region a unique value
eplane .. we recognise that the level curves u == Uo and 11 .. The families of level curves are shown in Figs.. Among the simplest mappings are those by a power 10 ::: zz.u) and the image of y ."
':
'_
.. for their
general shape is clear ~thout ealeulation.' . 37..
"

.
==
a _.a..::. It is a folium of Descartes.1
+
The mapping is one to one only if a(9! .. The sector S(~l.
. They are of course orthogonal to each other.l. The equation of the inverse image can be written in the form (x' + y2) 2 = 2(x' .::::yo is v2 == 4yl(tfa + u).~ .' . the curves ~ = flo . Similarly.
21
< r/2
can even transform zero angles into straight angle 8. The left half of the strip'.. and if Y2 . Firat. we map 01 onto & circular region. All these transformations have the characteristic property that they map a family of straight linea or circles on & similar family.YI = 11'" the image is a half plane. The function t == ~ + 1'1 == e~ maps the strip r/2 < y < 1t/2 onto the half plane ~ > Ot On the other hand. their use is essentially limited to regions whose boundary is made up of circular arcs and line segments. The power serves the particular purpose of straightening angles.
J. corresponds to a half circle. By these means we can for instance find a standard mapping of any
FIG...+ 1 = tanh
el
2
z
on the unit disk {wi < 1. When faced with the problem of mapping a region Dl oonformally onto another region O2... it is usually advisable to proceed in two steps. tools at our disposal are linear transformations and transformations by a power.
.. cut off by the im_aginary axla.
. ..
.. the general problem of conformal mapping can be reduced to the problem of mapping & region onto a disk or a. Hence
tD
=
e. that this mapping problem has a solution for every region whose boundary consists of 8 simple closed curve. The main. and by the logarithm . For this reason.z = %0 and y == Yo are mapped onto circles about the origin and rays of constant argument . 6.2. It is useful to write down some explicit formulas for the mapping . We are thus able to map 8 parallel strip onto & half plane. In other words. .
. A SurtJey oj Elementary Mapping8.: . :: :. Y2 S 2'F is mapped onto an angular sector...12
COMPLEX
ANALYSIS
ANALYTIC
FUNCTIONS
AS MAPPINGS
maps ~ > 0 onto
lwl <
1. 4.. and then we map the circular region onto Ot. We shell prove. In particular) a horizontal strip Yt < Y < Y2.~.>. and hence onto any circular region....
by 10 == e% is very simple.
wpbme . and with the aid of the exponential function we maps the strip IIm.. by the exponential function. The linea. The mapping is one to one in any region which does not contain two points whose difference is a multiple of 2ri. in Chap. half plane...
The mapping
'I
w=r+l
tl
.. Any other straight line in the zplane·is mapped on a logarithmic spiral.
39.. The fina1 transformation'
map onto
by a cubic polynomial w = aoZl alzt + 6tZ + 610 The familiar transformation z = 2'1 . the transformation (16) allows us to include in our list of elementary conformal mappings the mapping of the outside of an ellipse or the region between the branches of a hyperbola onto a circular region. . Next we define
Hence the image of a circle Iwl = p < 1is an ellipse with the major axis p pl and the minor axis pl . In this connection we shall treat explicitly a special case which occurs frequently. if the third angle has the ve~x a and if the sides from a meet again at b..Mapping by z .b) w?ieh transforms the given region into an angular sector. 1(_
w~z~..
The sign of the square root is uniquely determined by the condition \w\ < 1..:.e
For a mOre detailed study of the mapping obtain ... The image of a radius is half a branch of a hyperbola.:. without loss of generality we may assume the end points of the segment are + 1.. we begin with the preliminary mapping Zl :: (z ..

... ::
::... In fact.
"' .
As a final and less trivial example we shall study the mapping defined
+
w==zi+l
Zl
1
maps the half plane onto ~wl < 1...~:. or its complement.
. The region is a wedge W1~ the angle 2..p..
. . the half circle is a wedgeshaped region which in turn can be mapped onto &.
.a) will map the region between them onto ~ parallel strip.a}j(z ..b) maps the triangle onto a circular sector.. ~ ~:...
...a)J(z . A little more generally.
.~) (z +~) := 1.aL/3ao allows us to get rid of the quadratic term...nsformation Zl = 1/(% . By an appropriate power w = z~ this sector can be mapped onto a.
~":.. . the same method applIes to a cireular tnangle with two right angles. Let it be required to map the complement of a line ~gment onto the inside or outside of a circle. r:. half plane.. ~f plane. It does not.. allow us to map the inside of an ellipse or the inside of a hyperbolic branch. Clearly.1
z+1
maps the wedge on the full angle obtained by exclusion of the negative real axis. and a suitable exponential transformation maps the strip onto a half plane."
.. the tra.
.COMPLEX
ANALYSIS ANALYTIC
FUNCTIONS
AS IIAPPINGS
region whose boundary consists of two circular arcs with common end points..
Elimination of
(17)
(I
yields
and elimination of (18)
p
t!mt
Zl =
z....
(16) \Ve set 19 "" pei' and
x=~
y
:=:
2
1( 1). however. Such a region is either a circular w~et whose angle may be greater than 21'". If the end points of the arcs a and b. the linear transformatIon z: = (z .
+
z. ...~)
. . :.
.. :.. By m~ans of a power this sector can be transformed into 8. If the circles are tangent to each other at the point 4.....
=
vZt
8
as the square root whose real part is positive and obtain the right half plane. . The correspondence is illustrated in Fig. If the sign is changed) we obtain & mapping onto \10\ > 11
+ urI). : .:~ . half circle. The preliminary transformation
ar. The ellipses (17) and the hyperbolas (18) are confocal. ::. for (z .
p
(p + !) cos 8
p
SIn
fI.. .. Elimination of the intermediate variables leads to the correspondence
(16)
Z
=
~(tD +.
/n) < arg z < k(2r/n). (Lindelof .. this cut has an upperand & l~wer "edge . The coefficient for z is chosen 80 as to make the derivative vanish for z c= ± 1. k .e except for the positive real axis.y2 = 3 and the xaxis divide the splane are mapped onto half planes. 4. very similar to the one in the case of the simpler mapping w == :1.
.3. For orientation the reader may again lean on Fig. Map the outside of the parabola yl == 2px on the disk Iwl < 1 80 that z 0 and z .:. for instance a tag or a color.mspping by a power 'W == Z".
.
." Corresponding to the " angles in the zplane we consider n identical copies of the wplane with the cut. Map the region between 'z\ . and one with the semiaxes p~ + pI in the tDpiane. but then the segment (.
'
·:
.. This idea leads to the notion of a Riemann aurJau.. .. The simplest Riemann surface is connected with the .ne the corresponding angle is 36..1:· I·.I.. ' .... We know that there is a onetoone correspondence between each angle (k .. and never in logical proofs.2. Map the outside of the ellipse (x/a)! + (ylb)' == 1 onto lwl < 1 with preservation of sYlllmetrieS4L L Map the part of the eplane to the left of the righthand branch of the hyperbola x2 .) s. Similarly. y ~ 0 on the outside of the unit circle so that the points at co correspond to each other . It is now easy to visualize the correspondence between the z. If this is not the case. the COrreB.. three of them onto the upper half plane and three onto the lower.a21 == pl(p > a) on the disk lwl < 180 that symmetries are preserved. nJ and the whole wpIan..and seplenes... the one in the aplane has an asymptote which makes the angle ..·11
COMPLEX
ANALYSIS
ANAL YTIC FUNCTIONS
AS MAPPINGS
17
and by obvious normalizations we can reduce the polynomial to the form tD ~ Zl .. .11 < 1 on t~e inside of the unit circle. Choose the mapping so that the two symmetries are preserved. . Elementary Riemann SurJaees. For our purposes it is sufficient to introduce Riemann surfaces in a purely descriptive manner.
W = 2' 
3z. We are free to do so as Iong ss we use them merely for purposes of illustration... (Lindelof...3:. but when z describes an ellipse 10 will trace the corresponding larger ellipse three times. ~.
. Map the complement of the arc
lzl :::
excluded from the splene.. p~e ..(J with the positive real axis.
'.) Hint: Consider on one side the mapping of the upper half of the region by w = Z2 J on the other side the mapping of a quadrant by
I::
J. .y' == 1 on a half plane.. The visualization of a function by means of the corresponding mapping is completely clear only when the mapping is one to one.
.
. ..
r .
.:::
EXERCISES
All mappings are to be conformal L Map the common part of the disks le] < 1 and I' ..2) must be
1. It is not our intention to give) in this connection.
I'
4.~..:::. p/2 correspond to 1D == 1 and w == O. I .) L Map the inside of the lemniscate lZI . (Lindelof.
. _:
.J. but they are reciprocal and yield the same value of w.
_. unique r we may impose the condition ~rl< 1. ) 7. . We note in particular that the six regions into which the hyperbola ax' . we can still give our imagination the necessary support by the introduction of generalized regions in which distinct points may have the same coordinates.
•
.1.. In order to do this it is necessary to suppose that points which occupy the same place can be distinguished by other characteristics. The whole pattern of confocal ellipses and hyperbolas remains invariant... 39.. and in the wpla. 0 and the vertex to w == 1...1)(2. a rigorous definition of this notion.
. They will be the "sheets" of the Riemann surface.at on the disk Iw1 < 1 80 that the focus corresponds to w . To the circle Itt = p < 1 corresponds an ellipse with the semiaxes p~l ± p in the splane.. ~ they are distinguished by & tag k which serves to identify the eerreeponding ~Ie~ When z moves ~ ia... Map the inside of the righthand branch of the hyperbola Xl . The image of each angle is thus obtained by performing a "cut" along the positive axis. where n > 1 is an integer.yt ::::. The inside of the righthand branch of the hyperbola corresponds to the whole wplane with an incision along the negative real axis up to the po~t 2. Making use of the transformation (16) we introduce an auxiliary variable r defined by
2==r+~1 1 r
Our cubic polynomial takes then the simple form
1D
==
r' + •.
_
.and wplanes. For the region between two hyperbolic branches whose asymptotes make an angle ~ 2r /3 the mapping is one to one. In order to obtain a.RC?nding
• 1 • • ••• • •• .. a radius arg 9 corresponds to hyperbolic branches in the z... : I: .1 and 'Z == i o~ a half plane.. Points with the eame tag are considered to lie in the same sheet or layer... (Lindelof .. . The situation is thus . r
1
We note that each z determines two values {.
1
•••
. 2.
.
. except for one or more cuts. the half planes must be joined across one of the three intervals (. attacbed . The result of the construction is a Riemann surface whose points are in onetoone correspondence with the points of the zplane.p~bIem f~r
1D 
1D 
(z2 ... _ if it connects h sheets. Each region corresponds to a half plane on which we mark the boundary points 1 and 1.I) . one white and one shaded.
EXERCISES
L Describe
the Riemann
surface
_ 1( w==2 z+z·
FIG. 311..e. the clearest picture of the Riemann surface is obtained by direct consideration of the fundamental regions in the eplsne. . The Riemann surface corresponding to w = e. The origin Will not be & point of the Riemann surface.80D we must attach the lower edge of the first sheet to the upper edge of the second sheet.Sa. The shaded regions are those in which C08 Z has a positive imaginary part. It represents a cross section of the surface in the case that the cuts are chosen parallel to each other.1.'. we introduce the subregions which correspond to the upper and lower half plane..QO to 1 and from 1to co...is of similar nature. ~ . completing the cycle ."
. The choice of the correct junction is automatic from a glance at the corresponding situation in the eplane. 1).
" "
. A region which is mapped in. 310. (1. but the idealized model shall be free from this discrepancy. the lower edge of the second sheet to the upper edge of the third.
1)
associated
with
the function
'11.1)1. but . The line :t Tar eorresponds to both edges' of the positive cut if k is even.
. . For fundamental regions of w == c~ z we may Ch~OBe the 8t~ps (Ie . A point of this kind is called a branch point H o~ Riemann surf&C~ is considered over the extended plane..
Fundamental
regions of
COS ... For any two adjacent regions.. It connects a.
. We will illustrate the procedure by consideration of the Riemann surface defined by 10 == cos z.. Each strip is mapped onto the whole wplane ~th cuts along the real axis from . An attempt to illustrate the connection between the sheets is made in Fig. The resulting surface has infinitely many simple branch points over 10 = 1 and 1D = 1 which aiternatingly connect the odd and even sheets .. and a curve must wind n times around the origin . (1. to each other 80 that they form an endless screw. In the last step the lower edge of the nth sheet is attached to the upper edge of the first sheet. corresponding to the fact that ~ is never zero. The reader should bear in mind that any two points on the same level can be joined by an arc which does not intersect any of the cuts. The reader will find it easyto construct other Riemann surfaces.DO. For this re_8. The interpretation is even simpler if. the point at QO IS also a branch point. In more general cases a branch point need not connect all the sheets. The point 1D = 0 is in a special position.
to the edges of the negative cut if k is odd. Same problem for L Sam.
.ll the sheets. ~ The cut along the positive axis could be replaced by a cut along any simple arc from 0 to IX).. the cuts are in no way distinguished lines on the surface. Whatever the advantage of such representations may be.
FIG .. this conespondence is continuous if continuity is defined in the sense suggested by the construction. & onetoone manner onto the whole plane.1)22" < y < k ~2r onto a sheet with a cut along the positive axis..
2. the Riemann surface obtained in this way should be considered as identical with the one originally constructed...
_. as in Fig. JU. The sheets are.. < z < Ie.
..
"
. the introduction of specific cuts is necessary for descriptive purposes. we find that the edges of the corresponding cuts must be joined crosswise so as to generate a simple branch point at 'W = ± 1. ·and
5:..
COMPLEX
ANALYSIS
ANALYTIC
FUNCTIONS
AS MAPPINGS
It
point to should be free to move on the Riemann surface. What is more.co).. In other words .1).before it closes. If we consider the two strips which are adjacent along the line x = kr. it is said to he of order h .
The Riemann surface of eos z. In 1\ physical sense this is not possible without selfintersection. is called & lunda~tal region. and 80 on. In this case the function maps each parallel strip (k ...
MGut.1(t) == u(f) + iv(t) is a continuous function. that the derivative of an analytic function is continuous.
. E.. without resorting to complex integrals or equivalent tools.. in many elementary cases indefinite integrals can be found by inversion of known derivation formulas.4
COMPLEX
INTEGRATION
L FUNDAMENTAL
THEOREMS
Many important properties of analytic functions are very difficult to prove without use of complex integration.tiv.. &c. 87t 1961).
'. Am.
t Witllout use of integration R.. At present the integrationfree proofs are. much more difficult than the classical proofs. to say the least. If . T_ Whyburn. Both proof8lean on & topological theorem due to G. They can be defined by a limit process which mimics the definition of a real definite Integral. it is only recently that it became possible to prove. Soc.
'81
1. Actually.1. L. For instance. Plunkett proVed the continuity of the
.As in the real ease we distinguish between definite and indefinite integrals. Connell and P. .. t . or that the higher derivatives exist.sJl prefer to define complex definite integrals in terms of real integrals+ This will save us from repeating existence proofs which are essentially the same as in the real case. the reader must be thorougbly familiar with the theory of definite integrals of real continuous functions+ The most immediate generalisation of & real integral is to the definite integral of a complex function over & real interval. we sh. (Bttll. Am.~ 1900). N aturally. An indefinite integral is a function whose derivstive equals a given analytic function in a region. The definite integrals are taken over differentiable or piecewise differentiable arcs and are not limited to analytic functions.. Porcelli provedthe existeiloo·of·alfdeii. Math.
derivative (Bull. H. Une Intepals.
Whenever a line integral over an arc 'Y is considered.. In the righthand member 0(·(4). real (J in (2) and find
== e" with a
The integral (4) has also a very obvious additive property.~ . the interval of integration 'has to be subdivided in the obvious manner.h{iii)
".'Y by the equation z ::..
.i fJ onto Q ~ t < b.
.b u(t)
G
dt
+ i fL v(t)
G
for changing the variable of integration we have
dt. '. . . In Chap..
B
···+
IT'/ th. the integral over
+
Using this notation.2.) /. line integrals with respect to 3: or 1/ can be introd.»)tl(T) is the derivative of Z(t(T» with respect to T.1th+ /.. It is quite clear what is meant by subdividing an arc 'Y into a finite number of subarea. we assume that t(T') is piecewise differentiable .. lar.: :. 3.. .' ._
. let it be tacitly understood that 7 is piecewise differentiable./tk·~ ~(1.b f(z(t))z'(t}
In particu
dt
=
f:!(Z(t(T») Z'{t(T»)t'(r) d1'. The old and the new initial point determine two subarcs TIt '12. if c = a + iP is a complex constant we obtain
(2) for both members are equal to
~ut z'(t(.
a
We conclude that
(5) To see this we choose e
holds for arbitrary complex f(t)..
This is our definition of the complex line integral of J(z) extended over the arc 'Y.'
••• I'.
. A . the fundamental inequality
h f(z(t} )z' (t) tit.' .: '. if z'{l) is not continuous throughout.""
4>
value of the integral..change of parameter is determined by an increasing function t = teT) which maps an interval a ~ T . but then there is nothing
to prove.
:: . "" 0.flv) dt
+ i fa·
J__ /(z)
(av
dz
=
J_~af(z(t»)(z'(t»
dt. then f(z(t» is also continuous and we can set
(4)
and the corresponding integrals satisfy the relation
(6)
1"1
J • • • +"'"111f dz = 1.
'
.. Sec . H the functionJ(z). t We consider now a piecewise differentiable arc ~ with the equation z = z(t)~ a ~ t ~ b. and the invariance follows from the fact that the integral over '11 + 72 is equal to the integral over ""12 ""1..uced by
t II is not defined
if
fa" f dt... A subdivision can be indicated by a symbolic equation
~ = 'Y1
For
(:I
= arg Ja f(t) dtthe expression on the left reduces
lt
to the absolute
+ Yi + .102
COMPLEX
ANALYSIS
COMPLEX
INTEGRATION
103
defined in an interval (a..
J.: ::.
./di! ~ (J.· + ..f tk ..
and by a change of variable the last integral can be brought to the form
When a ~ b. By the rule
closed curve is also invariant under a shift of parameter.
. z(t)... b ~ t ~ a.
+ tlu) dt.1..b)J we set by definition
(1)
J{~ J(t)
(J
dt =
J.
This integral has most of the properties of the real integral. mtegr&l (4) has the same value whether 'Y be represented by the equation z = z(t) or by the equation z = Z(t(T». and (3) results. we defined the opposite arc ./ dz + J'I'*f dz + +~+ .. In addition to integrals of the form (4) we can also consider line integrals with respect to it The most convenient definition is by double conjugation Finally.' '.
/. and hence the . ·We have thus
JJb (au
(3)
.1.is defined and continuous on "Y.
"
. The most important property of the integral (4) is its invariance under a change of parameter...
Two notations are in common use.'
. If "f is rectifiable and fez) continuous on "Y it is possible to define integrals of type (8) as a limit
/.f(z(t»l:l(t)1
dt...z("'_I)I_
1:1
This integral is again independent of the choice of parameter. From the parametric equation z = z(t) == a + pei'. = b. (u dy + II dx)
'Z{tl) ...2.
+
which separates the real and imaginary part. Idz\ which is by definition

J. It is quite easy to show that piecewise differentiable arcs are rectifiable...)
.y(tI:1). In other wordS.x(ti 'y(t..2(t1. . and if the integral
1.... p th + q dy are often studied
B8
functions (or fWlCtionala) of
the
length of 7.. The length of an arc can also be defined as the least upper bound of all sums
(10)
depends only on the end points.. 1'1 J To say that an integral depends only on the end points is equivalent to saying that the integral over any closed curve is sero. Rectifiable Arcs.t the functions x(t) and yet) are of bounded varitltibn..) .Conversely..~ < t..
:.y(t. H this least upper bound is finite we sa. 0 ~ t ~ 2'1'"...1 ds = f. III ·
The ineq~jty
Ilk]
~
Here the limit is of the same kind as that encountered in the definition of a definite integral.
/.} + · ~· + IY(t.J dz I~ f.Az1 an example we compute the length of 8 circle..Iz'(t) 1de ==
as expected ..
L "'".ys tha. An essentially different line integral is obtained by integration with respect to arc length. trast to (5) we have now
In con
while (6) remains valid in the same form.104
COMPLEX ANALYSIS
COMPLEX
INTEGRATION
101
With (7)
f=u
+ iv we find that
the integral (4) can be written in the form
j.z(~I)i
1l. ~ . and hence
102.t}l
IT P d3.flrlzl
=
/.y that the arc is rectifiable.
in which case formula (7) would serve as definition of the integral (4). General line integrals of the fonn
For f = 1 the integral (8) reduces to
I....x{ t'l).z{ t..3. and if the integral over any closed
fol1ows that
/'f1 ~ I.) 
z(t. of a full circle we obtain z'(t) = ipei. if 71 and 1'1 have the 88Jlle initial point snd the same end point.»Iz(t..y(tiI) I ~ and lz(t.... but not piecewise differentiable.. if O
'Yt
'Vi
and .

•
.
I/. Indeed.1)' s Ix(tl) .. 1.... + q dy..2(10)'
+
lz(tt) . have the same end
is
&
closed curve.
•I
• ~.y(to)1
+ · · .
f
')'I
closed curve. one sa. Of course we could just 88 well have started by defining integrals of the form
lx(tl)
:. An important class of integrals is characterized by the property that the integral over an arc depends only on its end points.
.! da = lim l J(z(t.. When the latter Burne are bounded. and the definition is (8)
are bounded at the same time.
!o2r
p
dt = 21rp
It is then BSSUIned that p and q are defined and continuous in a region 0 and that 'Y is free to vary in o. then "I sod T have the same end points.
. it is clear that the BUms (10) and the corresponding sums
Because l!(t~) . Line Integrals GlI Functioru 0/ Are«... we obtain
lz(tv .. we require that p dz + q dy = ( p d:e + q dy .) . (u dx 
v dy)
+ i J.s(tlVl J Iy(ti) .) .t}
ft
..J1
and consequently points. if "f is 8
the arc 'Y. and that the two definitions of length coincide . However. In the elementary theory of analytic functions it is seldom necessary to consider ares which are rectifiable.z(to)l
iY(~l) . .(t) is rectifiable if and only if the real and imaginary parts of z(t) are oj
bounded variation. It is a matter of taste which one prefers..
(9) is a consequence of (3)..+ Ix(!. An arc z == . the notion of rectifiable arc is one that every mathematician should know.z(taI)I
.z(t1)1
+ · · ~+
\:(l. then "II curve v~j~~ it
where a = to < t1 < · .
q and U can be either real or complex. tJZ4'
p tk
+ q dy.
in
Q
aF
8z
. It is customary to write dU == (au/ax) ch + (8f!Jily) dy and to say that an expression p dx + q dy which can be written in this form is an exact differential.. The function U.
""
. Sec. the function is well defined.. Consider ~'/~~'""C ".y) = /''' P(X. _whose sides are parallel to the coordinate axes (Fig .y)
If this is so..
From
O~.y(b)
.
and the value 'of this difference depends only on the end points.. FCz) fulfills the CauchyRiema.
. for if the condition is fulfilled we can write.·i.J dz would
Dn
not be
defined) F(z) is analytic with the derivative /(z) (Chap.
"

"
"
"
. the aame result holds for all closed curves which do not pass through 4J for in the cOmplementary region of the point a.S'"~t:. contained in 0.y(t)
tit "
The integral /. for if two functions have the same partial derivatives their difference must be constant.. (11) does not always hold. Moreover. To prove the necessity we choose a fixed point (xo..COIIPLEX
ANALYSIS
COMPLEX
INTEGRATtON
101
FIC..the Indefinite integral is still analytic ~ ~singlevalued. 1.AB an immediate application of the above result we find that (ll)
" if and only if f is the derivative of an analytic functicm in o.
"ee~~" 2r~" : We: obtain"
"" "
""
+
the lower limit of the integral being irrelevant.
fr P ax + q dy
=
feb (~~
x'(t)
+~
y'{t}) dt
=
/a" ~
U(x(O... Thus an integral depends only on the end points if and only if the integrand is an exact differential. if it exists. (z . B function which is analytic in the whole plane. by choosing the last segment vertical} we can show that aU jily ==...fez)
/7
.
th~~"~exPre_on it
'fc. rep~'ted by the""eqUation z == a pe".a)'" is the derivative of (2 .. If n is negative.aF . if we choose the last segment of r horizontal.41) and define a function by
U(x.
L (z 
a)" tk = 0
/7
P dx
+ q dy..
}
"
+ eonst. .y) ==
Under these circumstances we shell prove later that fez) is itself analytic.U(x(a).==:
t
. 2.ayt
The sufficiency follows at once.a)tt+l/(n + 1)."th:~ 0.
aF{z) _ ~~() ay
defintNl in OJ depends only
U(x..t "aulax == p. depends
Dnly
the end points
of
== U(x(b). we can keep y constant and let x vary without changing the other segments. . TheoreID I. but ~ I..) EO. " .q. with the usual notations.. mined up to an additive constant.y(a»). au lay == q. In fact.'~~/.
Since the integral depends only on the end points. On the last segment we can choose x" for parameter and obtain U(X.y... When isj{z) dz ::: f(z) €!x + if(z) dy an exact differential? According to the definition there must exist lL function F(z) in 0 with the partial derivatives
The following theorem gives a necessary and sufficient condition under which a line integral depends only on the end points.
J dz. The line integral
of(z)
ax 
.. _
ainee fez) is by sssumption continuous (otherwise
/.. In the same way..41
follows at once tha. Observe that p. is uniquely deter . join it to (x"y) by a polygon v..nD equation

on the end points of 'Y if und only if there exists a function with the partial derivatives au lax = p.y) tb:
for all closed curves 'Y J provided that the integer n is ~ O.. For ~n:= 1.2). tuith coMntWUa J.
(The continuity of f'(z) is taken for granted..
_
J
:
J
. 4. hence.. we find
that
(13) '1(R) = 'l(R{I)
f
is the value of
f'(z) T j(II)·rb = 0
I
+ '1(R(U) + '1(R<') + '1(R(t).101
COMPLEX ANALVSIS
COMPLEX
INTEGRATION
101
This result shows that it is impossible to define a singlevalued branch of log (z . Compute
to 1 + i..
...61 = R. . On the other hand) if the closed curve y is contained in 8 ha. • ". 1._...d).
for the positive sense of the circle..:
... 1.:.. Its perimeter can be considered as 8 simple closed curve consisting of four line segments whose direction we choose 80 that R lies to the left of the directed segments.J{Z)f'(z) dz
is purely imaginary.... It is natural to begin with a case in which the topological considerations are trivial ..
'.
". <. RC. for in such a half plane & singlevalued and analytic branch of log (z .1
.
.. R{Il.
P(II} di?
A1i8We. but they diJIer in their topological rather than in their analytical content.e.
I:
•
..
(12)
where 'Y is the directed line segment from 0 .: . a rectangle R defined by Inequalities a ~ x ~ bJ C ::: y ~ d. specifically..
. is not 8 region. a reference to Fig..: . In the theorem that follows we consider a.
. (The continuity off(z) is takenfor granted. (a.. It is importaut to note that this fact can be verified explicitly and does not make illicit use of geometric intuition.·.
8 parameter. .
k... Show that
The proof is based on the method of bisection.a) in an annulus PI < 1z .) L Assume that f(z) is analytic and satisfies the inequality lJ(z) < 1 in a region n...
t This is Itanda. Show that J. .: :::··:f~:'i1. If R is divided into four congruent rectangles R(l) J R(I)..c) .2).. We consider... by use of and second.. z..4. Cauchy'S TheoremJor IJ Rectangle.. then ( J(z) dz == JaR
o.. the integral vanishes.
EXERCISES L Compute
L Describe a set of eircumstaaees
under which the formula "'" 0
/.. and we denote it by oR.loglldz
is meaningful and true.a}.
for every closed curve in o. We recall to the reader that such a funetion is by definition defined and analytic in an open set which contains R. by observingthat
.'L
.
Let us introduce the
s.L Compute
x
== ~
(z
+!) ""~(z + ~) on the circle. If P(z) is a polynomial and C denotes the circle I: .c) . (b. ~ . 42 is helpful..
for the positive sense of the circle.) 7.:. 3 Sec.I'
X
dz.. in two ways: first. Compute
If
the function J(z) is analytic on R.
. Suppose that f(z) is analytic on' a closed eurve v (i.. function which is analytic on the rectangle R.1f plane which does not contain a. (b._
. Note that by earlier oonvention aR ill alao the boundary of B M a point set (Chap . . Nevertheless.'··\':>·'.a) can be defined ... t We emphasize that R is chosen 88 a closed point set and.. We refer to this closed curve as the boundary turve: or contour of R. what
Ie
for the integrals oyer the common sides cancel each other..
"
_"
.f is analytic in 8 region that contains "().. The order of the vertices is thus (a . notation
'1(R)
== JaB J(z) dz (
11
which we will also use for any rectangle contained in the given one._'!.:2?riR!P'(a).rd notation. There are several forms of Cauchrs theorem.d). .. The following is a preliminary version of CauchyJ8 theorem:
Theorem 2.
:..\. and we shall WI8 it repeatedly. I..e] < PJ.
.
.dL E. We make now the observation that
) 'R".. we can choose 8 so that or
(15)
Since . 2..j'(z*)
I<
important use ...(z .
(bid)
•
i
(~ £')
FIG.. the choice sba11 be made according to some definite rule .owa~in·Eig...
.::.... " :. Lit J(z) be analytic on 1M set R' obtained frmn a rectangle R by mnitting a finite number of interior points ri. 1~1.. .2"*}
for ...z ·1·ldz I·
Bisection of rectangle. for evidently R can be divided into smaller rectangles which contain at most
f aR~
.. We· dhWe~B.

. This process can be repeated indefinitely) and we obtain a sequence of nested rectangles R ::> RI :) Ht:J ..z*l < B.~H.
. But if d and L are the corresponding quantities for the original rectangle R....
Iq(Rft) I :s: e ( J 'R... 44. will be contained in a prescribed neighborhood Iz . Secondly) if B > 0 is given.= apply
(
dz
=
0
It is sufficient to consider the case of 8 single exceptional point r.
a
If(z} ....(8 . is contained in Iz ... 2."
.~
~z ....tJ._ :.z*l < 6 as soon &8 n is sufficiently large. 3
J
1'l(R(l»)l ~ i111(R) I.')~. . We 88SUme that a satisfies both conditions and that R.
f
R(2}
~
.z*1 < 0.(z . By (16) we have hence l. At the same time the proof is simpler than the earlier proofs inasmuch ~ it leans neither on double integration nor on differentia.1
J(z) dz == .
Z
dz = 04
as~·ab. Ie: ...
.iDm ~ine ~
1.. By virtue of these equations we are able to write 7J(Rn) =
B(3)
. It follows from (13) that at least one of the rectangles Rei).
~
1 f
(b. of the diagonal of Rr.:l .
.
.::. < 8..
.z *)f'(z *)] dz.
tl
R(4)
and it follows by (15) that
(16)
J'R"
r
[f(z)
. :'.z·...
1'J(R.
_ . k 4... We shall prove at once the following stronger theorem which will find very
I
fez) .. .110
COMPLEX
ANALYSIS
COMPLEX
INTEGRATION
111
(1'4 d)
..
...(R) = 0. This beautiful proof. · ~ R.z·1 is at most equal to the length d. First of all.:. 4..(RIl) l :s. If it i8 true that
Theorem 3. must satisfy the condition
==
1.
....
o....)l ~ il'1(R"~l)
I
The rectangles RA converge to 8 point z* ERin the sense that Rn. · with the property and hence
(14)
In the last integral Iz .. e)
These trivial special cases of our theorem have already been proved in Sec. If L. denotes the length of the perimeter of RnJ the integral is hence ~ d'ff>Ln.ri)J(Z) == 0
J
~·. is due to €4 Goursat who discovered that the classical hypothesis of a continuous f(z) is redundant.e: is arbitrary..J(z*) . we can only have . then
lim •• (.. tion under the integral sign.
lor aU i. 4.. and the theorem is proved ..!*)!'(z*)l
< liz .(::..'~. The hypothesis in Theorem 2 can be weakened considerably..
ft
and comparison with (14) yields l'l(R) 1 ~ dL e..
R(l}
.f(z*)
Z
z*
.fez *) .. which could hardly be simpler... it is clear that tin = 2"d and L~ = ZL. respectively.
':': . We recall that the proof depended on the fact that 1 and 2 are the derivatives of z and z2/2.. we choose 6 so small that J(z) is defined and analytic in 12... We denote this rectangle by RI. if several R(1:) have this property.
." '..." .
.
.. after cancellations.~~: I_:..
where u consists of the horizontal line segment from the center (xo.. .· :·:f.
I
I 1
I
I I
lcz 
(
dz
2ri
Q.i..'
_
"....~~'"..zol < p to be de noted by 6.. it is necessary to make a special assumption concerning the region D in which J(z) is known to be analytic and to which the curve ~ is restricted..:
~. For the applications it is very important that the conclusion of Theorem 4 remains valid under the weaker condition of Theorem 3..
\_::{~~L. .\~~~.~.·~:. If j(z) is analytic in an open disk A. We are not yet in a position to formulate this condition.
COMPLEX
INTEGRATION
113
I
I I
I
·~
•r
•
Indeed.Yo) to (x.
"I'
112
COMPLEX
ANALYSIS
.. On the other hand..
~•. It is not true that the integral of an analytic function over a closed curve is always zero.. of the theorem i8 certainly fulfilled if j(2) is an81ytic and bounded on R'. :.Yo) to (x.J? _
. and hence Theorem 4 holds for any of these regions.
s
/..
·.'. A rectangle.'·..:
:~. in A.~¥Jti~
:
~~\~J~:f{:(.....
.
seen that aF/ay = if(z).s..
:: ~>...
.1
Theorem 2 to all but the rectangle R« in the center.
. the same proof would go through for any region which contains the rectangle with the opposite vertices %0 and z as soon as it contains z.
:.. and for this reason we must restrict attention to a very special case.ri)/(') 0 for all j. This choice defines "the same function F(z).·...:.

1.':>' '. reach full generality./: :::.i:r. _ .
u
.(a .:.~_'.: ..lz I~lrl'
that Ro is a square of center
r. We observe that the hypothesis. . 
r
I I I f
when C is 8 circle about a..• Theorem 4. elementary
esti
Thus we obtain
and since e is arbitrary the theorem follows.. Uien (18) hold3 lor any closed
58
curve ""t in 4.A~.0:~1M~:~2:::.\:>::. however.. : ...
(17)
!/(z)
dz = 0
If e > 0 we can choose the rectangle Ro so small that
The proof is a repetition of the argument used in proving the second half of Theorem 1..~< . we obtain.. In what follows we assume that {1 is an open disk tz . .··~~. j
.·:'~·:~ ~. We state this 88 a separate theorem ..·1\8
mates show that
we may.k. :. Cauchy's Theorem in a Di. . or the inside of an ellipse all have this property. and we obtain aF lax == J(z}... '.:...5.
.~~ .<>·:. In order to make sure that the integral vanishes..:s~ '_u ''..Yo) and the vertical segment from (x..'.. .A with the derivative f(z).. If the corresponding equations (12) are added. We define a function F(z) by
(19) F(z) =
1/(2)1 ~ Is on aRo..' obtainMl by omiltintl a finite number of pointB fJ from an open disk A.. 5. By this method we cannot.y). ~
:=:
).. :.. we have found that
I
~~~~~~~~~~~I~.~~f:. . '.:.. :. Clearly.'~:..~ ::~ . :: .'. .. . by Theorem 2 a can be replaced by a path consisting of a vertical segment followed by & horizontal segment.. Hence F(z) is analytic in. then
. 1. tl..4. Let fez) be analytic in the region . If f(z) mti8fita the CORdititm lim.:::. and /(z) dz is an exact differential.
So·
"K'
!kr. "~... a half plane.
By (17) we have thus
g
rl
Lfdz
! /IJS f dz 1 ~
If we assume...(~~iv:·. it is immediately
Theorem...>..
. (18) for every closed curoe .
.a) vanishes except perhaps at a finite number of points. nition can be based on the following lemma:
2.
:
."_.:::". In case there are exceptional points on the lines x = Xo and y :: Yo the reader will easily convince himself that a similar proof eanbe carried out J provided that we use four line segments in the place of three. Assume first that no fi lies on the lines x = Xa and y = Yt).
.. We can now define the i'l'l.. of Theorem 4:
".a). and since this function is continuous it must reduce to a constant. the defi. moreover..
=:
Indes oj a Point unth Respect to a Closed Curve.~
lTz dz a 
a
I
(%O· . has numerous important applications. Through
whenever Zl(t) is continuous.a
Since z(JJ)
by the equation
z{a) we obtain elOO = 1. 44). log Iz ....
When z describes a closed curve.
. known as Cawhyt8 integral formula.a) .
..a)
=
f
y
d log Iz 
al + i
Jd
"1
arg (z ..a). . It is clear:that n(. Above all. :.
.a
It is defined and continuous on the closed interval [a. ~=I
'.=.
.' :..
.th respeot to G.
t..liri~pe.. We conclude as before that F(z) is an indefinite integral of f(z) .8'1.tnmediBte 00ilBeq~.c) (z(t) .4(. /:: . .a) in 8.
. and therefore h(/J) must be Dr multiple of:2ri~ ... be either vertical or horizontal. The simplest proof is computational. We have thus
~(I)
= z{t) . and it has the derivative
h'(t)
=
z{t) .1.. If the precewise differentiable clo8td curve "I does not ptlN through the pmnt aJ then the value of the integral
r
.al returns to its initial value and arg (z .
"~
'
: . Z
£ I
I
_1
.
2.de% oj the point a with respect to the curv.
".. This would seem to imply the lemma.COIIPLEX
ANALYSIS
COIiPlEX
. ..·~I .:
::
. The.a
z'(t)
INTEGRAL
FORMULA
a very simple application of Cauchy's theorem it becomes possible to represent an analytic function f(z) as a line integral in which the variable z enters as & parameter. 'Y. it enables us to study the local properties of an analytic function in great detail.a.
: . for we cannot let a pass through the exceptional points.tIo. The
n(oy a)
J
1 = :
f."~.. but more careful thought shows that the reasoning ia of no value unless we define arg (z .za dz 2ri
•
With a suggestive termiuology the index is also called the 1IJi. CAUCHY".RdiRg number of ~ wi.. a ~ t S (J... ....
The proof must be modified.rty ·"···ani. for we can write
r
=
f
T
d log (s .··. . unique way . z(t) . From this equation it follows that the derivative of .:. and the theorem follows.....)'0)
This lemma may seem trivial. By an obvious application of Theorem 3 we find that the value of F(z) in (18) is independent of the choice of the middle segment.... let us consider the function
h(t)
==
j'
a
z' (t) cU.. z(a) . .. . If the equation of l' is z = a(t)...a) increases or decreases by a multiple of 21"...

INTEGRATION
111
~l .
:..
"10.~
~ . It is then possible to avoid the exceptional points by letting f1 consist of three segments (Fig.r
is a multiple oJ 2ri. This representation.. s= n(')'.. the last segment can. 1~
Lemma 1.. As 8 preliminary to the derivation of Cauchy's formula we must define a notion which in a precise way indicates how many times a closed curve winds around a fixed point not on the curvet> If the curve is piecewise differentiable t as we shall assume without serious lOBS of generality. This proves the Iemma.
:..a)/(z .(e ...O) = n(C.
COMPLEX
'NTEGRATtON
117
tire 8t1me circle.b) .:' I.. ..
'~:JL
.....
~ :
. . Let d"j. (6) if "(is pieeewise difFerentiable the new definition is equivalent to the old.. and it is desirable to formulate.u) for any continuous closed curve 7 that does not pass through 6J whether piecewise differentiable or not.. We shall say. and set tI == til + · . . polygon which does not meet y.. Suppose that %1 lies in the lower half plane and Z2 in tk 'Upper half plane. ..a) = O. we find that n(~~O) = n{C. the components of the eomplement.j
.
.:. They serve to illustrate the topological use of winding numbers ... fl be the points in which L1.I
· ·
.a.j}
~
(i) 1f'Y lies inside of a t4rcle.'
~
.
'.
·EXERCISES
~
For the proof we draw the half lines Ll and L" from the origin through ZI and %2 (Fig .O) = I. Give an alternate proof of Lemma 1 by dividing "f into a finite number of subarea such that there exists a singlevalued branch of arg (z .. hF.
. This proves that n(~. It is possible to define n(l"..a)1 . Denote the subarc fram 21 to Zt in the direcl. ~ .
\ .a) on each subare.!o
• I
¥. Introducing the closed curves Ut :::::71 82 :" 01 .:
• J .. be the directed line segment from the initial to the terminal point of "Yl. Any two points the same region determined by 'Y can be joined by a. Hence the origin belongs to the
J
+
I:iiiji:
+
These are not routine exercises. there is exactly one which contains the point at infinity.b) is never real and ~ o.a)
== 0 for all points a otdBide of
t. For this pwpose "t is divided into subarea "(I. i:: i : ..
.. To justify the definition. and if iy does not meet the segment we must have
in
(1 J. We shall find. 8J be two poinl8 071 a closed·curve 'Y which does not pa88 through the origin... :1·. a geometric condition which leads to this eonsequence.. L" intersect & circle C about the origin.· •• :.4) :::::: "t... If 1 is sufficiently large. . Pay particular attention to the compactness argument that is needed to prove the existence of such B subdivision... n( Outside of this segment the function (. .
.. ~ . for short. Consequently. 7..«) == n ( 'Y . . 1. and zero in the unbounded region.. . If C is described in the positive sense....
'
...lh..0) ::: 1 .J. .. the case n('Y. 2.
41
+

. • I . ~ is contained in a al disk 1z1 < P < la\ and we conclude by (i) that n(T.
.b)] is analytic in the complement of the segment. If "'{I does rwt meet the negative real axis and ')'2 doe8 not meet the positive real a:tis:t then n{ 'Y .2 ~ a)/(z . .. . For simplicity we take a == o.b)l. Denote the directed line segments from %1 to f1 and from %2 to 12 by 61. then n(~..01 ... . 02..
. (ii) As a Junction of a the index n(~. As 8 point set ~ is closed and bounded. Let tl. But cr1 does not meet the negative axis. For this reason it is sufficient to prove th ~t n( I)' ..O) + n(cr"O) becanse of cancellations. Let 21..za

zb
1 ) dz
==
o· '
Pic.ion of the curve by 1'lf and the 8"UbarcJrmn %2 to %1 by '1. '5
hence n( 'Y ..
. and we obtain n(al.J
111
COMPLEX
ANALYSIS
.... (a) the result is independent of the subdivision. :: . . Its complement is open and can be represented as a union of disjoint regions..:
~
. prove the following: .:·· .
••
. tT2 'Y~ al . :
• ~
0..
:..O) + n(ul.a) :: 1 particularly important... Lemma 2. Its derivative is equal to (z . T determines one and only one unbounded region . For this reason the principal branch of log [(2 .~.a) i8 constant in each of the regions determined by "f. each contained in a. the arc C1 from tl to i2 does not intersect the negative axis..
. :.a) = 0 in the unbounded region. that y determinu these regions . If the complementary regions are considered in the extended plane. 45).
.:<.
unbounded region determined by tTl..: ".:disk that does not include B. cr•• We define ·n(T}a) to be the value of n(a.O) = 0" For a similar reason n(a"O) = OJ and we conclude that n(y.. (c) the properties (i) Gd~(Ji)0( the 't ·continue to hold .a)..b) if 7 does not meet the line segment from a to b. .:. and the arc C1 from f2 to rl does not intersect the positive axis..·
~.
. (d) n(cr"Xl) ::: I. ..x ) = 1or 1..
r: .f(a) dz
"f
Z
a
= O.:. and it is convenient to change the notation and rewrite (21) in the form
. and lei 'Y be a closed curve in A.. hence n(O"I...
For.:.) . and let 0'2 be constructed in the same way with '12 in the place of ""Ii.\. We apply Caueby's theorem to the function
F(z)
In this statement we have suppressed the requirement that a be B point in A. We have then
(21) J(a)
= ~ r !(z) dz.'
.a) • fCa) .'
.
~ ..
. z! E ~ with Im 2.
'
. We .. (e) the first intersection Xl of the positive real axis with "Y lies on 1'1.. provided that the order of a with.::
.
: . The most common a...1 < 0.
2ri.
.. The Integral Formula.2. but it
=:
..
': ~'::.za dz. hence n(O'"l~z) == 0 for a E 'Y!.'. The notion of winding number leads to a quick proof of one part of the theorem....
• ~.. These points may be chosen so that there are no other points of 'Y on the line segments from 0 to %1 and from 0 to %1.2t>(J.
"..:
.
r ......J.. ..
fez) . Choose the notation so that the intersection %2 farthest to the right is with ')'1 (Fig. '. f.. The presence of exceptional points fi is permitted...J(a)..JT
.may thus t~t a as avariable..4.s: .
where n(~.a) == lim (f(z)
~a
... '
I
:. (excluding the end
and we observe that the integral in the righthand member is by definition 2ri ...0. provided none of them coincides with a..!.a) ¢ o.z) == n(u2}x) = 1 for small x > 0 (Lemma 2). Let 0'"1 be the closed curve that consists of the line segment from 0 to 21 followed by ')'1 and the segment from %2 to 0.:.a) and the integral in the righthand member are both zero. f(z) 2ri
7. Consider a closed curve 'Y in '.
.
I . ' ..~~~
~~'tlSlrra~~k~~~~~~~~
~~~~~
..:~~i .. Let 1'1 and 'rl be the arcs of 'Y from %1 to Zt. The Jordan cun6 theorem asserts that every Jordan curve in the
plane determines exactly two regions. For any point a not on 'Y (20) 1&{.z) == 1 for Z E 1'1. '".
. We have done 80 in view of the obvious interpretation of the formula (20) for the case that a is not in .. in this case n(1'.
:
I
•
.J(a»
0
which is the condition of Theorem 5. :'.a) is the index of a with respect to
points). Let J(z) be analytic in an open disk 4. the other on 'Yt. and no other points on.
'
1(.. Suppose that /(z) is analytic in an open disk 4.
.. '.... :.
'
= J(z)
za
..
. Z..
=
.":. I..:::
a it is not defined.J.. and a point a E A which does not lie on 'Y.
1 ..
(22)
..
1
It is clear that Theorem 6 remains valid for any region 0 to which Theorem 5 can be applied. Indeed. 46)4 Prove the following: (a) n(C1trX2) = 0. (b) n(ut.. 1m Z2 > 0.. respect to "f remains equal to 1.
We conclude that
f
Fie.
I' ".:.~¥:.a).dl(r) s. Indeed" it permits us to compute J(a) as soon as the values of f(%) on 'Y are given. Then fTl ~ 0"2 = 'Y or 1"4 The positive real axis intersects both '11 and "'II (why?). This will be so if there exists & point a with n(~.
"
.:
.. together with the fact that f(z) is analytic in A.
..
.
2ri J~ 2

a
and this we interpret as a rtpresenta#on formula.'
.
This equation can be written in the form
f
f(z) tk = !(4) dz... :".. We may assume that Re z > 0 on y.
_. and that there are points Zl. (~) there exists 8 segment of the positive real axis with one end point on "1.
Part of the Jordan curve theorem...J .. ".:r:1~. 'Yza 1"za
f
J...
. In (21) we may let a take different va1ue~.pplication is to the case where n("Y..a) = 1.. ... n(T.111
COMPLEX
ANALYSIS
COMPLEX
INTEGRATION
111
This function is analytic for z satisfies the condition
lim F(z)(z
. '
· r . We have thus proved: Theorem 6.
".yiP The points x between the end pointe satisfy n('Y ..
¢
e.. :"
i' _.~\'~>#/~..
..
.. namely that the complement of a Jordan curve 'Y has at least two components..
I
•
'..
:
. applied to the function .. Compute
2... or we could prove a general theorem concerning line integrals whose integrand depends...r!¥~'~:·:~~::~2·::f~~~~:~ .*.I:. .)(.:{'_·:~::·..!ldrI
J
= ~ ( fer) dr. Let 20 be B point not on 'Y and choose the neighborhood !z .o)
(24:)
( dr !(..ed by 'Y..
L 1. and that we have proved it only when f(z) is analytic in & disk. ':'I:':>'I. . Actually.B) = 1 we have n(C..0
H the differentiations can be justified. Compute
Isl
Izf .{z) . }a (rJ(f) Z)R+l . .
.6~
...F.
EXERCISES
1. .H:
..
f.2)(r 
%0)
n»
= _!_
1.ZD! · :...tioos. Theorem 6 can be applied to f(z) in l!J.. Since n(C..(r..'. At the same time we shall have obtained a convenient representation formula for the derivatives ..~.:_~..(z) =
2Z:t ~
1
by decomposition of the integrand in partial fractions.~ . For such z we obtain by (22)
fez)
z.)~ ~ . Compute
IUACtion
Then the
J 113
idz. We must remember that it is valid only when n(Ytz) = 1.. To & point a if 0 we determine 8 aneighborhood !J" contained in 11.J. . a circle C about 4..F 1(20)
.~. : The general case is proved by induction..df fer) 2rt
G'
2)1
tenda to the limit F1(Zft) as z + 10..The representation formula (22) gives us an ideal tool for the study of the local properties of analytic functions.
1.:": ~ . We consider a functionf(z) which is analytic in an arbitrary region O...•
. <f . we shall prove only the following lemma which is all we need in the present case: Lemma 3.120
COIIPLEX
ANALYSIS
COMPLEX
INTEGAATtON
121
It is this formula which is usually referred to as Cauchy)s integral formula.... 3..e) =. For the justification we could either refer to corresponding theorems in the real case.""_..~ :.(t) dr (r ..:.>!
< 12 . and its derivalive F~(z) ::: nF »1(Z) . .. Since every point in 0 lies inside of some such circle. In particular we can now show that an analytic function has derivatives of all orders.zol < 6/2 we attain that zl > B/2 for all r I! y.
. a ~~...zo)
f
'Y
fr

ql~~ Z
rdf
to
)
we obtain at once
IF1(2)
.
z
=
Provided that the integral can be differentiated integration we find
(23) and
under the sign of
z
Zo
f
7
..(f)/(r we conclude that the difference quotient F t{z) ..
2ri l c
r
and this inequality proves the continuity of F1(z) at Zo. F.(z).I<. we shall have proved the existence of sll derivatives at the points inside of C. the existence will be proved in the whole region U.[I. .II).. and in .. .
"
".:.(z.··zo) .!:!~~~
I·''··~··~~~
~~:~t.t:.
:
.l)F.
We prove first that F1(z) is continuous..~....<r 
'P(t)
dr
z)"
all
ill analytic in each 01 the regiona determin.'.f ~~)t"]+ '·)f~ ~).
(z .!!!.(r) is continuous on the arc 'Y.i1::'' f~:~::~~::~'. analytically on a parameter.F. F19m this part of the lemma.... 2ri "
•••
.
is
under the condition [e] ~ p~ Hint: make use of the equations zJ ::: p' and
!ck! =
ip liz..z) = 1 for all points z inside of C. From
'1 
Ft(z)

F1(zo) == (z .
..ZD! < & so that it does not meet 'Y~ By _ restricting z to the smaller neighborhood Iz . From the identity
·F. . .. ·
.~ !.a.
"
"
... Suppose that ...(z)..
.
z
. Suppose we have shown that F~_l(Z) == (n .
:.:. which are then also analytic . Hence it is proved that F~(z) = F.20)...:. Higher Deri".~. .6.
._. by the induction hypothesis.ends to zero for z + Zo..alc ldal (Ial
. 1. that 3 /(z) is the derivative of an analytic function F(z).
f 1'1
Iz . ~:'.I.. ..
••• _. and by Liouville's theorem l/P(z) would be constant..~ .
lit:
. neighborhood of ZG~ Now.=.
S ::
'I'
••
'::
_:
"
•
.. then /(z) is
lilt 2 2. The inequality (25) is known 88 Ca~hY'8 _mat6.
.t) is analytic as 8 function of ZEn for any fixed t.
.
:
..t) dt
88
For Liouville's theorem we need only the esse n == 1. applied to 'P(r)/(r . It is clear that Lemma 3 is just what is needed in order to deduce (23) and (24) in a rigorous way..:.8 theorem..+l(zo).
~
.. find an upper bound for tf(tI)(z)I in \t' ~ p < R. .
'.Z)M dz.Zo is bounded in 8.
.. We know that P(z) + co for z+ co. the object being to minimize the function M(r).. In order to make the best use of the inequality it is important that r be judiciously chosen. We know now that j(z) is then itself analytic.for some n and all sufficiently large Izl reduces to a polynomial.n be stated as follows: If fez) is defined. Then
5.. Hence F~(zo)exists and equals nFD+l{zO).. Suppose further that rp(z. Indeed. S. Since this is not so. .!!
p).. ~..
I
_
_ .122
COIIPLEX
ANALYSIS
COMPLEX
INTEGRATION
123
we can conclude that FA(:) is continuous.~.
'
". :. I. We conclude that the function is constant. If P(z) were never zero. The first is known as M arma'.t Cauchy's inequality will yield .
. It shows above
(X) .
.0 must have a root.t) df
Fill in the necessary details to obtain
I
and use Lemma 3 to prove (26) .
.: :.t) be continuous as 9.
point can never satisfy lJ<")(z)I > nln.:
_. v. find the best estimate of 1/<·)(0)1 tha... The hypothesis means that If(t)l S M on all eireles. function of both variables when z lies in a region 11 and a ::S: t < fj. A more general form of Lemma 3 reads as follows = Let the function rp(z.
··1. the same kind . the first term t.\:. The remaining factor in the second term is continuous.~ ~:
.':.
. the quotient in the first term tends to a derivative which by the induction hypothesis equals (n . . and assume that If(r) l ~ M on C.ouvilWs theorem: A junction which is analytic and bounded in the whole plane muat reduce to a constam.. and therefore l/P(z) tends to zero. We have thus proved that an analytic function has derivatives of all orders which are analytic and can be
JI
the successive derivatives of an analytic function cannot be arbitrary. 0 fur
The hypothesis implies. . A second classical result goes under the Dame of Li.
To prove this represent fP(ztt)
a Ca.where M(r) is the maximum of 111 on It . This implies boundedness (the absolute value is continuous on the Riemann sphere and has thus a finite maximum). and in the second term the factor of z ..... Prove
f
z'"(l . '
that EXERCISES
:I. as we have already remarked in Sec. Hence we can let r tend to and (25) leads to f/{a) = 0 for all a.Zn).
r.
...
..
r.. and it ca.~ ••:'" :'. and if
all closed curves y in 0.
.Zo and let z tend to Zo.. by what we have already proved. If fez) is analytic for 'zi < 1 and 'f(z)l ~ 1/(1 . If fez) is analytic and If(z)l ::: M for 'z! ::: R..
I..
. ..
2ri}crz·
(
'P(r.zl).al = Y. and has the limit F +1(zo). Show that the successive derivatives
of an analytic function at a Formulate 8 sharper theorem of
(25)
F(z) == is analytic in z and
(26)
J: 'P(z. :
~.:..
..
.
. polynomial of degree> O. Suppose that Pea) is 8. the function l/P(z) would be analytic in the whole plane . If we apply (24) with z = a. if we divide the identity by z ..:.
that a function which is analytic in the whole plane and satisfies all inequality 11(z) 1 < Izl.p
IT f dz ... the equation P(z) =.l)F. there must always exist an M and an r so that (25) is fulfilled. :.. we obtain at once
t
analytic in
o.. For the proof we make use of a simple estimate derived from (24) Let the radius of C be T. and conHnuom in a region 0.uchy integral
'P(s t) """ _!_
.
=.. Compute
an
represented by the formula (24)~ Among the consequences of this result _welike to single out two classical theorems....
.'~ '. Liouville's theorem leads to an almost trivial proof of the Jundamental theorem of algebra... _.
in Theorem 5.
f(a) + (z . Repeating the process tion !2(Z) which equals (Jl(Z) . the function which is equal to 1(2) for z ~ a and which has the value
(27) 2ri}cfa.11(a»/(z z == a.J(a»)/(z . as z tends to a is I'(a).a)2J.(a).
a we find
The necessity and the uniqueness are trivial since the extended function must be continuous at B..
..
an analytic function which is for z == 4. . . We shall prove the following precise theorem:
.a) for z ¢ a and fi(a) for
is defined can be written in the
j(z) fl(Z) .. In this section we will make a closer study of the local properties. function bY!l(z).. that Cauchy's theorem in a circular disk remains true under these weaker conditions. . It is natural to denote the extended function by f(z) and the value (27) by f(a) .
.a)f'J(Z) . We showed moreover.a)P(z) == O.. for we were required to apply Cauchy's theorem to a function of the form (f(1.. .
== J(a)
+J'i~)(z 
a)
+f'i~) (z 
a)i
+ . This remark is more important than it may seem on the surface. SUpp086 that fez) ie aft!1lytic in the region 11' obtained by omiUing a point a from a region it A necessary and sufficient condition that there exist .
used in the proof of Cauchy's formula. and we obtain the foUowing form
fez)

= ~ ( fer) dt 2ft lo r . provided that none of them coincides with e..an analytic function in {)which coincides with j(z) in rt is that lim (e .(a)
Differentiating
n times and setting z
==
+ ~· ~ + (z .• ' _(
. representation of fez} through an integral which in its dependence on e has the same character at the exceptional points as everywhere else. It is not defined for z == a but it satisfies the condition lim (s . . It follows that the exceptional points are such only by lack of information. all satisfying the fundamental condition of Theorem 3. . .·. The recursive scheme by which f~(z)
form
~.
'. To prove the sufficiency we draw & circle C about a 80 that C and its inside are contained in o..
'.a and equal to J'(a) .eDl
(28) fez)
8. and so on. . This was an essential point in our .
" "
.z
. Points with this character are called removable singularitie8. is valid] and we can write .derivation of Cauchy's integral formula. Cauchy'e formula.f(a)
za
3. Finally.a)f1{a)
+
(z .. Taylor's Theorem. .a)'lJ~l(a) + (z .
~. It will include & classification of the isolated singularities of analytic functions.. .... r.
=
. If J(z) ill analytic in a region UJ comaining a.. .. . it was pointed out that Cauchy's integral formula remains valid in the presence of 8 finite number of exceptional points.. and not by their intrinsic nature..1.. it is p08Bibl.+a
fez)
=:
/(a)
+
(z ...(a)~
This determines
8 theorem:
the coefficients i. ¢ a inside of Ct But the integral in the righthand member represents an analytic function of 8 throughout the inside of C.t24
COMPLEX
ANALYSIS
COMPLEX
INTEGRATION
125
L LOCAL PROPERTIES OF ANALYTIC
FUNCTIONS
We have already proved that an analytic function has derivatives of all orders.a). Consequently. ....lorities.
. . The limit of F{z) . The extended funcUon is uniquely determined~
.(z)." :..
_!_ ( fer)
dr
:..
I'" (a) ==
of Taylor'
nlf . In Theorem 3 we introduced a weaker condition which could be substituted for analyticity at a finite number of points without affecting the end result.
for z = a is analytic in D. .. Cauchy's formula provides us with a... Hence there exists equal to F(z) for z '#.. . .a)'1. Let us denote this we can define an analytic func.a)/l(2) = fl(a) + (z .V. . We apply this result to the function
17(z) == fez) ._ . Indeed.
== a we obtain
From these equations which are trivially valid also for z
Theorem 7. Removable Singr..
.a)J{z) = o.
for all .) ....
1.
COMPLEX
ANALYSIS
COMPLEX
INTEGRATION
127
.. We show now thatj(z) is identically zero in all of O. Let El be the set on which J(z) and all derivatives vanish and E2 the set on which the function or one of the derivatives is different from zero, El is open by the above reasoning, and E, is open because the function and all deriva. tives are continuous. Therefore either EI or Ht must be empty. If E. is empty, the function is identically zero, If El is empty, /(z) can never vanish together with all its derivatives. Assume that j(z) is not Identically zero.. Then, if /(a) = 0, there exists a first derivative J(h)(a) which is different from zero. We say then that a is a zero of order h, and the result that we have just proved expresses that there are no zeros of infinite order. In this respect an analytic function has the same local behavior 88 a polynomial, and just as in the case of polynomials we find that it is possible to write fez) .... (.I  o)"I ... z) ( where fl(Z) is analytic and fl(a) ¢ O. In the same situation, since/A(z) is cont.inuous~jl(Z) F 0 in a neighborhood of a and z == a is the only zero of J(z) in this neighborhood. In other words, the zeros of an analytic function which does not vanish identically are isolated.. This property can also be formulated as a uniqueness theorem: If f(z) and g(z) are analll~ in D, and if I(z) == g(z} on a ut which has an accumulation poiRl in D, then l(z) ia identScaUy eqWJl to g(z).. The conclusion follows by consideration of the difference J{z) .;._g(z}. Particular instances of this result which deserve to be quoted are the following: If f(z) is identically zero in s subregion of 0, then it is identically zero in UJ and the same is true if J(z) vanishes on an arc which does not reduce to a point. We can also sa.y that an analytic function is uniquely determined by its values on any set with an accumulation point in the region of analyticity. This does not mean that we know of any  way in which the values of the function can be computed, We consider now & function f(z) which is analytic in 8 neighborhood of at except perhaps at a itself, In other words, fez) shall be analytic in 8 region o· < aJ < 0.. The point a is called an isolated riflgularitg of f(z). We have already treated the case of a removable singularity. Since we can then define f(l1) so that fez) becomes analytic in the disk < &J it needs no further consideration. t ~ Jim J(z,) = 00; the point a is said to be & pole of 1(2), and we set
This finite development must be well distinguished from the infinite Taylor series which we will study later. It is, however, the finite dev:elopment (28) which is the most useful for the study of the locBl. propertie~ :f J(z). Its usefulness is enhanced by the fact that f.. z) has a simple explicit ( expression as ft. line integral. Using the same circle C as before we have first
s ( ) = _!_
II
e
2ri [o
( f,,(r) dr.
r
z
For j.(r) we substitute the expression obtain~~ from (28). There will be one main term containing fer). The remarmng terms are, except for constant factors, of the form
F,(a) =
lo (r  a)p(r  z)'
(
df
l' ~
1~
But Ft(a)
_ 1 ~ z  a lo
r ( r  z _ r 1 a) 1
dr = 0,
identically for all a inside of C. By Leroma3 we have F.+1(a) = Fi·)(a)/r.! andthusF,(a) = o for all II ~ 1. Hencetheexpressionforj,,(z)reducesto (29) The representation
fB(Z) =
2n
1(
f(r) tit lc (r  a)~(r  z)·
is valid inside of C. U f(a) and all deriva.tives J(z) ~ J,.(z)(z  a)A
f(p)
3.2. Zeros and Poles.
write by (28) (30)
(a) vanish, we can
for any n. An estimate for !R(Z) can be obtained by (~9). ~he d~sk with the circumference C has to be contained in the reglon 0: III ~hich I(z) is defined and analytic. The absolute value If(z) 1 has a maximum M on C; if the radius of C is denoted by R, we find
'Z 
lfR(Z), ~
for 1J: 
R"l(R 
M
Iz 
a1)
'Z ._ a'
a\ < R..
BIIB)"
By (ao) we have thus
\f(z) 1 ~
(
\Z 
a\)" ..R  MR  ar R Iz
Iz  I'll < R.
Hence f(z) == 0
I(a)
There exists a 8' .S 6 such that f(z) ~ 0 for 0 < I.  01 < a'" In this region the function ,(.) = l/f(z) is defined and analytic. But the singularity of g(.) a is removable, a:nd g(z) _ an analytic exten=.
00.
.......
But
(1.1 
+
0 for n + co, since
inside of C.
. t If a is :~__ ~:. ~Y term is uae.d iIa a ~
•
;,..
.'
I•
.:
.....
~".~
1ODl~.
j..;
at
. ~
.. ',.
~!~ ..... ~ ..._,.
.. .:. •
.
.
~.
,...
(o·A:,:; .o:·~: ,:r. ,~ ....... :
• .. • .'
, L.
..
.".
.. f(~) is frequently .id ~ h6:replOr at A;·ta for aDalytic~ . _. , .. '
L
.
.
•
.~: .. ,

: ~ •.
'
I
~,.
:.~ ~:
:.. ,
I....
•
0
..
....
_
•
'
•
_~ •••
••
'
:.
••
...: • '::1
;lo,.
o
•
I~
.....~
.Jo.i:
~
,. ,
•
..
'"
',..
.~ .
..• '".:
or'
. ".::,
::
..
,"\:~:I:' ,.
121
COMPLEX ANALYSIS COMPLEX INTEGRATION
sion with g(a) == o. Since g(z) does not vanish identically, the zero at a has a finite order, and we can write g(z) == (z  a)lg1t{z) with g..(a) ~ The number h is the order of the pole, and !(z) has the representation fez) = (z  a)lf,(z) where f.(z) = 1/g1(z) is analytic and different from zero in a neighborhood of a. The nature of a pole is thus exactly the same 88 in the case of a rational function. A functionf(z) which is analytic in a region 0, except for poles, is said to be meramorphic in o. More precisely, to every a E n there shall exist a neighborhood lz  al < B, contained in 0, such that either fez) is analytic in the whole neighborhood, or else fez) is analytic for 0 < Iz  al < ai and the isolated singularity is s pole. Observe that the poles of 8 meromorphic function are isolated by tkjinititm. The quotient J(z}/g(z) of two analytic functions in (] is a meromorphic function in 0, provided that g(z) is not identically zero. The only possible poles are the zeros of g(z), but a common zero of J(z) and g(z) can also be 8 removable singularity.. H this is the case) the value of the quotient must be determined by continuity. More generally, the sum, the product, and the quotient of two meromorphic functions are meromorphie. The case of an identically vanishing denominator must be excluded, unless we wish to CO~aider the constant co as a meromorphic function .. For 8 more detailed discussion of isolated singularities, we consider the conditions (1) lim lz  ol«'f(z)1 == OJ (2) lim lz  a\G\f(z)I·= 00, for
...... 0 ........
121
o.
where 9'(z) is analytic at z = a. and find
1(1)
F
or z '# a we can divide by (z  a)1.
= BIo{,

a).I.
+ Bl1{z
 a)H.
+ · · · + B1(z
_ a)1
+ tp(z}.
The part of this development which precedes tp{z) is called the Iri l :n~!(~)guIat z = a. A pole ~ thus not only an order, but also ~eI~ 81n ar part. The difference of two f cti ..h singular part is a.nalytic at a. un ODS Wlt the same In case (iii) the point a is an essential '8Olated. . the neighborhood of an essential singularity I(~) is at ~fagularilY •. In unbounded and comes arbitrarily close to ze e sam~ tI~e 0.£ th~ ccmplicated behavior of a function in ther:eig:o~h:~a~tenzatlon tIal 81ngulanty, we prove the follOwing classical theorem of ~eo~em. 9. ~n analytic function tomes arbitrarily clOlJe 10 any complex v ue In every nesghborhood of an essential singularity.
;e~~~::~
real values of a.. If (1) holds for & certain a..then it holds for all larger a~ and hence for' some integer m.. Then (z  a)"1(z) has 8 removable singularity and vanishes for z == G.. Either J(z) is identically zero, in which case (1) holds for all a, or (z  a)j(I) has a zero of finite order k: In the latter case it follows at once that (1) holds for all a > h = m  k, while (2) holds for all a < h: ~ume now that (2) holds for some a; then it holds for all smaller a, and hence for some integer n. The function (a  a)t(z) has a pole finite order l, and setting h == n + l we find again that (1) holds for a > h and (2) (or a < h, The discussion shoW8 that there are three possibilities: (i) condition (1) holds for all at and J(z) vanishes identica.Ily; (ii) there exists an integer h such that (1) holds for a > hand (2) for a· < b; (iii) neither (1) nor (2) holds for any tl. Case (i) is uninteresting, In case (ii) h may be called the algebraic order of f(z) at 4" It is positive in case of a pole, negative in case of zero.. and zero if J(z) is analytic but #: 0 at a.. The remarkable thing is that the order is always an integer; there is no singlevalued analytic function which tends to 0 or 00 like a. fractional power of \z  al. In the case of 8 pole of order h let 11Sapply Theorem 8 to the analytic function (,.  a)"'/(z).. We obtain a development of the form
0;ch !Ience
to choose fj
If the assertion were not true we uld find a~ ~ that I/(z)  AI >'8 in neigh~r=:I: z a, or any a < 0 we have then lim 12  al'"11(.2')
ex:
a would not be an essential
ingly, there exists a p. with ~
>
O. Sincein
= 0,
that ~
Iz  allll/(z)1
J.2' «1'lf(z}  AI ... 0, and ;e arec~~~ that case lim Iz  alllJAI c 0 it would f 11 ~a 0 ow and a would not be an essential singularity of
sin.ity
n:(=~:::;a~~ _I
A
to.
of I(z) _ A
Ac
d
or
a.
fez). The ,,?ntradiction proves the theorem.  T~ ,notion. of isolated singularity applies also to functions which 1c m a n~ghborhood 121 > R of cO, Since I( co) is not defined : c a to as an Isolated singularity. smd by convention it has the same ~ of frert;vable singularity, pole, or essential singularity as the an y 0 g e :f(l/z) at z = O. If the singularity is nonessential j() z has an algebreie order h such that lim ____..,( ).  .th· , Z ""I Z 18 net er zero nor infinit d~ · ~. y, a:n zor 8 pole the smgular part is a polynomial in If · an essential cnn~ ..l it he Z.. 1X) 18 · Y. t function has Thearem 9 in ~UJ:&rl neighborhood of infinity.the property expressed by every
:aIr
EXERCISES
j
(z  a)'1{z)
=:i
B..
+ Bltl(Z
 a)
+ · · ..+ 81(% . .
a)l~l
+ tp(z)(z .
a)'
that; f!!(Z)
~ .. ~;:v, .~~~
..
I'
JII
,
•
~4
i.". ~'::._:.',. '.
.:.
4=.:......
and fI(z) have the algebraic orders 1& and _k at _ ....,~, :'Jiow z _ the onler1a. + k,JI,the orderli  h, ~41 +:~&nOta.iJlich "~::.L!)d_.mPix ~I..) . . . ... .. .:.... ... .~', : ~:..... r;· .,. L ~j1_ffr~;:,:,~:t' ~ ~. . '.,," • 5.. '5 :yi t
.',.
~.. '!';_, .
oj!
 ~ : j.. ,J ~ • '" . ~., .......... ~
:
,
~,.
••
1 ... .... : ':~
'oj: ."!;~~ '. :; ,: <, .
:'
':'
~. ~:~.:. . : . ..: ... ,'

.. . ,',.
. .::.. :~·Ii~~;t>:.t"!;t;. I,,:': Ji.: _I..: ...· .. :·/:'::;;'_'~::::";;'il1:'::".; .:,.... .: ··:.~i.:~:':~'.·4:···~.1.~ ; .:~;k;t::: .~, . ..... , '¥1"'l1.~.;"::::....' :~..:. .• . '~. ,. .:~:.:,t., 
.
 :::.,. .~,. '!.:,~,:~. ~ :~::.:.:~.:'::.'.;../ .:.:.~' •..;,~'. ._ : :
~ :.,~:.:~'>;':.:.:.}.~': .....J5·:·~/"'~.~~~·"!P, r: ... .. 1 :: .."
f#.; ~
.~:.>:>:: <., :.:'. :~::::::.~~.::>..:".':..~:..:.<'~ :..,>. ~.::.. " ..: ~ . ' .
:·,,1 • •
.,
of'.:
"';0:.
...1
.
I
.~
.~: ..... : .~ ..~"!"
~:"tl.
:,..
~
1
...
.."
.:".II~ i~..:~';::'. l
..... ' ... ,....:,
.:

i;'.!..~~.:.. ~..;. ~ : _.~

t ~ .,.~ ~'L ~
••••
•
..
,
~.
••
•
.,."
~·:;':d}'"_'·· ..:..L"Y~7:·'':· .:..
~,.,....~:
•
:0.0;., ••
~(;1.\~~
130
COMPLEX
ANALYSIS
COMPLEX
INTEGRATION
131
2. Show that a function which is analytic in the whole plane and has
8
at
nonessential singularity at 3. Show that the functions
DO •
a)
reduces to a polynomial. e, sin z and cos z have essential singuJarities i.
pass through a zero
10. Let Zj be ~ ZBr08 of a fUndum fez) whiCh 18 antilytic in a . dISk 4 .and does not fJantM identicaUy 'r each zero beina lUta:",~ ..... as many .I ~ _..........A ttm.es as its order indicates. For every closed curve 'Y in 4 which does not
Th~rem ~
.... Show that any function which is meromorphie in the extended plane is rational. s. Prove that an isolated singularity of /(z) is removable as soon as either Re J{z) or Im f(z) is bounded above or below. Hint: Apply a fraetionallinear transformation. L Show that an isolated singularity of f(z) cannot be a pole of exp f(z) Hint: land ef cannot have a common pole (why?)+ Now apply Theorem 9.
+
(32)
,
nh ,Sj) = ~
L~(~)
dzJ
where the sum has only a finite numher of terms # 0..
The function w an d we find
=
J(z) maps y onto a closed curve
r in the
wp Iane,
'
3.3. The Local Mapping. We begin with the proof of a general formula which enables us to determine the number of zeros of an analytic
function. We are considering a. function fez) which is analytic and not identically zero in an open disk e, Let"'Y be a closed curve in a such that f(z) ~ 0 on '1. For the sake of simplicity we suppose first that fez) has only a finite number of zeros in .6., and we agree to denote them by z], Z" ••• , ZR where each zero is repeated 88 many times as its order indicates. By repeated applications of Theorem 8,. or rather its consequence (30), it is clear that we can write /(z) = (z  Zl)(Z  Z2) .. • ', (e  z.)g(z) where g(z) is analytic and ¢ 0 in.6. Forming the logarithmic derivative we obtain
Jr
(33)
( dw
to
=
f jf(Z) dz · /(z)
7"
The formula (32) has thus the following interpretation: n(r,O) =
X n(y,2;).
J
•
f'(z) [(z)
=
z  z.
1
+
z
1
Z2
+... +
z
1
Zt.
+ g'(zt,
g(z)
The simplest and moat useful application is to the case where it is ~wn beforehand that each n(Y,zj) must be either 0 or 1. Then (32) YI~ds a formula for the total nuntber of z.er08 enclosed by "'i. This is evidently the CSBewhen "J is a circle" Let a be an arbitrary complex value, and apply Theorem 10 to fez) _ a. The zeros of J(z)  a are the roots of the equation fez) = a and we denote them by zj(a).. In the place of (32) we obtain the formula '" . 1 t' £., n(Y,2;(a»  2ri ,.fez) (z) a dz
:1
for z ¢ z;, and .. articularly p yields
on 'Y.
Since g(s) '# 0 in a, Cauchy's theorem
f
f
(31)
nh',21)
"Y
tt(z) dz ::: g(z)
o.
1 2wi
and (33) takes the form n(r,a)
=X
J
n(y,zi(a».
Recalling the definition.of ~(""',%i) we find
+ nh,za) + · · · + n(oy,z.) =
..
f I'(z) dz. f(z)
"t
It is necessary to assume that fez) '# a on "'Y" If a and b are in the same region determined by r, we know that
n(r,a)
This is still true if J(z) has infinitely many seros in 4.. It is clear that " is contained. in a concentric disk Ii' smaller than 4. Unless J(z) is identically aero, a case which must obviously be excluded, it has only a. finite number of zeros in A'.. This is an obvious consequence of the BolzanoWeierstrass theorem, for if there were infinitely many zeros they would have an accumulation point in the closure of Il', and this is impossible. We can now apply (31) to the disk Il.'. The zeros outside of fl.' satisfy n('YJzj) = 0 and hence do not contribute to the sum in (31)~ We have thus proved:
.
.'
= n(r,b),
.
and hence we ha.ve also
l: n('Y,zi(a»
J'
"'"
In(y,zi(b».
I~ .., is. a ~cle, it follows that /(z) takes the values a and b e~ua.Ily many times inside of ')'. The following' theorem on local correspondence is an immediate consequence of·tbis result.
,
.!~..
,I
... .
.
'.
'.
. ....
...
':~, ';.';
..
'"*/(·).~ ~~.~
'..: '.:
TheonmI 11.. 8t1p'p088 that fez) it analytic tU '0, J(se} == 1Do, aM t}~:~~ f(~) ~ tDo 01 order ft at ~ :1/. >O,·ie.~tI f:mOll".,~'fJ ~ a.~ .•. > O,NCA lAta/or ~ CI ~ 141~ tD.I < 8 U&e ~
"'=.0;'"
,,"~::¥~;~i_~}gt;::~:L::i:;rbj~i~_;:, ~ "::r'~.::;:.~n,:.~·:lor:¢~ ... :
.. ~ "

~·~.H <
'',,' ,'',
y
••••••••••

:
,"<,'
'<'
"'_','
'L,
.
 ':
. . .:.
.
.:'
.'
•
•
. ~: . .:
".':
.
•. _.: : ..::.....;.~_;:".:,{: .. ... ... ,~
.:.:,'
:.~.: .: ~.. ~:. ,; , .·~L~~ > .
·:a~~!::.l$]"t;;:<·.i.,~·~::.:.;._.·:.,:::;.~!1l¥"o:'~~'....._~mt_~'~~~~~ ~~
•. :..,; :....
..!",
.
.
.
This is merely another way of saying that the image of every sufficiently small disk Is .': ::.
•
. On the . In this neighborhood it is possible to define a singlevalued analytic branch of Vfi(Z).
w+prane
J
'..
Corollary 1. .~~~~·~"~~.'.
..1~.
:. ..
._ ::..
h·
.'
.
•
.. and hence every value a . prove the existence of an 8..i.
If f(z) ie analytic at
to . :. S~ce o~n set~ in the splane correspond to open sets in the wpiane the mve~_ f~ction of J{z) is continuous. f. .
':"..t.wol < a which does not intersect I' (Fig..· .Qalytie ..
~
"o'V .~... v.:
..: .....• J:1
:
.'
..' _.
1m
==
Wo
+ t(z)
to
008
z with
·'0 ...
·...
:
. : .
...:: ~:.a..%0' < e...4~X
~PO~~
. fez) .' : . and hence the inverse mapping is
mentary character and determines n equally spaced values r for each value of 10. i It is understood that multiple roots are counted according to their multiplicity) hut if e.._~\:~..
~
J.
fez)·
r(z) = (..z....'.
<:: =::::::~. :{"... _ ~ .: ..:~> __ .<~~::..~ '. ~'··~::~~~·:~··~·/'::'·~·_'.: e." .: .i.~"'.zo}h. ~ 0:'.. . <'...:·~·~·:~::" ..
"
. {' .. CoJollary 1of Tb~rem aQ~~q.tplane
=
_Iplane
Branch point: n
3..Wo in this disk..' /:..stant analytic fundion
maps open 8et&
onto open
seta. ~. " .. • L
Local eorreepondenee.'~\..
::. :.". s taken ~ tim~.... :
.~:.
'< .. if the local mapping is one to one...._..: :
<:. Same problem for to == e: ("..
. ~
... Under the assumption of Theorem 11 we can write
likewise conformal.:·~. there exists 8 neighborhood Iw ..:' :" '..:
':....
:.( '.. ~::." .
0"
zpane
F'IQ. . By performing the mapping in two steps we obtain B very illuminating picture of the local correspondence. A
w·plane
FIe.'
.:.' .
"..~'. .. the mapping 10 = Wo is of an ele
+ r
r
2... Figure 48 shows the inverse image of & small disk and the n arcs which are mapped onto the positive radius.~~~.. Choose £ > 0 80 that Ig(z) . Let y be the circle Iz ....
•....t := t(z) is topological in a neigh ..~:.':' .g(zo) I < Ig(zo) I for lz ...q .•. . The mapping can be restricted to 8 neighborhood of %0 contained in A..
_~ •••_
~..~: .
noncon.: z . :!
.. I and 80 that Zn is the only zero of J{z) .:·r.. It IS sufficient to choose e so thatf'(z) does not vanish for 0 < Iz .
r .:..I. ...:.20' ...:.
.J. :. borhood of to.
~ • . Theorem 11 can hold only with n = 1.: :~/.::~:~~'::··i~: ~' :.)g(z)
'10 
is analytic at %0 and g(zo) # o.
.::\
..
+
Determine f(8) explicitly . Indeed. For n > 1 the local correspondence can still be described in very precise terms.'.". ~~
..·:.: .
. IfJ(z) is analytic at the origin andf(O) P£ 0.
state:
<:Orollary 2.~..':..: . and the mapping is topological... ...:..~~.~. and hence !'(zo) must be different from zero .
. ~.zol < !."..zo' < a....
.
• • •
11 has a
J
~ery
.
.: : L Apply the representation j(2)
L Determine explicitly the largest disk about the origin whose image under the mapping tD . .other hand. contains a neighborhood 1w ...i". It follows immediately that all values a in this neighborhood are taken the same number of times inside of~._.
•
.~lJoXimwn
. In the C&Se n = 1 there is onetoone correspondence between the disk woi < i and an open subset A of Iz ..1001<.: :.it . .~.!Do == (z .....Prineipl.:
~ .::..\ ••
:.. :
::.
.._~.~
..
IXERCISES
Since r'(zo) =: h(zo) ~ 0 the mapping . . I.with '(20) ¢
Ot it mapa a neighbor
hood of
to conformally
and topologically onto a region~
From the continuity of the inverse function it follows in the usual way that the inverse function is analytic.zo1 <!. .(z)..:·:.. ..·.
.. .. The equation f(z) == Wo bas exactly n coinciding roots inside of 'Y.~~~/)~i~~:. 47).
. .~..
.. which we denote by h(z). ·.. of the equation fez) == a are simple for a :P!: 100.zol ... .
.
~:·~·~.:= z' z is one to one.::~<:~:·:.(z) 'such that !(zA) 1(0) + g(z)a in & neighborhood of O...
principle<~~~
(!.l"~..
...:~·.
_).
'
.
~
. 41. 132
COIIPLEX ANALYSIS
COIIPLEX
INTEGRATION
133
We can choose I so that J(z) is defined and analytic for Iz ..:
:...~ : _..~..
....
.. We have thus where g(z)
fez) !Do :... .
. and we are able to
Conversely. and r its image under the mapping 10 = l(z)4' Since tDo belongs to the complement of the closed set I'.
.:.
~ .»: . . is sufficiently small we can assert t~t ~ roots.. _:. 47.J~~~~'..
.~(. positive formulation essentially the same theorem can be stated in the form:
Wo
=:
The proof is clear.
.:{.:"":~.:~::.
_~* '..
...:i .~~.
~:.l It onto lr~ 1 with ...·{.:~:~:. 8tiU. Therefore if it is known that fez) takes some value of modulus < M.
f(z&) is any value taken in 0. Lett .134
COMPLEX
ANALYSIS
COMPLEX
INTEGRATION
131
analytic functions.. the original form of the theorem can be applied.:.:'. there is nothing to prove. in the more general case. l/r.. or if If'(O) I = 1.~.1
I'_
'20'
maps . It follows easily that f(z) must reduce to a constant. ' < M.. . and hence If(zo) I II! not the maximum of If(z)\..
If
Theorem 12'.:~~L:.·: . The rather specialised assumptions of Theorem 13 are not essential.~ijQn 8f(~1f) . ~ .'
". Letting r tend to 1 we find that 1{Z) I s 1 for Jill z. If Zo is an interior point. which can be rewritten as 1/(')' :! izl/R~ $jmilarly. _~. IS. . and (35) would lead to
Suppose that ~f(zo)t were
8 maximum. dr = ire"" d9 on "I and obtain for a = 20
. If the formula (22) is specialized to the case where ~ is a circle of center 020 and radius r..oon4i~/( . .
Then we would have I/(zo + rei·)t ~ If(zo)l. we can write t = Zo + rei'. if the upper bound of the modulus is M· instead of I. I/(z) . : ~ .
the contradiction I!(zo)l < If{zn)t. .. then II(z) I ~ M in the whole disk. If 11(z) l = lzi for some z ¢ 0...w~ < R and] . In 8..
. by continuity. If(z) I :.::. : ~..zol .. but should be looked upon as the result of & normalization. consequence of Cauchy's integral formula. Circl. ~~~~~w..!. 18 contained III the region of analyticity. and if the strict inequality held for a single value of 8 it would hold.. . on 9r whole arc...) If J(z) is analytic and nonconstant in a region D.
. If it is known that If(z) f ~ M on Iz I = R.If(zo + re") Id8..::.. Thus ~f(z)l must be constantly equal to I!(zo) \ on all sufficiently small circles fz .. .~ . if J(z) is known to s&tisfy the conditions of the theorem in a disk of radius R.wol < E contained in the image of o. then I/(zo) I is also the maximum of If(z) i in a disk lz . and trus is the assertion of _the theorem. has a maximum on E~ Suppose that it is assumed at z.wiu..::. then if(z) I ::: Iz' and "(0) I S 1. in a neighborhood of Zo. then its ab80lute value \/(:) 1 ha« no maximum in o. gOing < <:
~.: ~~~..DU)~·Pnerallyt We may·. then f(z) = ~ with a constant e of absolute value I.... that 11(z) must reduce to a constant.0·~~:h~"1:·~D~·~·'~ ~..hieh.replace the condition 1(0) == 0 by an ~~~. If f{z) is defined and continuous on a closed bounded set E and analytic on the· interiM of EJ flten tM maximum of If{z) lonE is assu1118d on the boundary of E. to the function J(Rz)...~~. and preference is quite justly given to proofs of this kind. As a rule all proofs based on the maximum principle are very straightforward. < M..iff ~ 1.3) is a constant of absolute value M.ji~ ...'
_". ~. The equality can hold only if 1(.v.~:·8.. But then the mean value of If(zc + rei.:.zDI < 6 contained in E... Consider now the case of a function J(z) which is analytio in the open disk 1z1 < R and continuous on the closed disk Izi ~ R.. l~~ar ~9~~ & ·~bich •• PI ~.
'J
(34)
This formula shows that the value of an ana. = tD... .lytic function at the center of a circle is equal to the arithmetic mean of its values." :'. We have given preference to the first proof because it shows that the maximum principle is 8r consequence of the topological properties of the mapping by an analytic function.. .. it meens that 1/1(:)l attains its maximum and. l. . it may be expected that 8 better estimate can be given.~~:> ~. .~~~~ •• '.
:".
. But this is not possible unlessJ(z) is constant in the component of the interior of E which contains Zo· It follows by continuity that tf(z) I is equal to its maximum on the whole boundary of that component. 10
r. .... I~ this neighborhood there are points of modulus > Iwol.~::. r. :~ .:#. From (34) we derive the ineqnality
(35)
If(zo)! ~
1 2. ..:. This boundary is opt empty and it is contained in the boundary of E..~"'_'_.:~"..r:'
~~:~~
!.. The resultioa inequality is.· ~~ .
The maximum principle can also be proved analytically.~_'.. T. there exists a neighborhood \10 .. This reasoning provides a second proof of the maximum principle. If the equality holds at a single point. Theorem 12.'s.)1 would be strictly less than If(zo)l. Thus the maximum is always assumed at a boundary
Since E is compact. As a result we obtain If(Rz) I :i Iz1..:::. For instance.
. to f(Rz)/M. On the circle Izl :c: r < 1 it is of absolute value :s..e.. : ::~
'>:.. . Because of its simple and expJimt form~lation it is one of the most useful general theorems in the theory of functIons.\~< It. flO) = 0. as a. on the. . (The maximum principle.. M"lIR. hence..>.. We apply the maximum principle to the function Il(Z) which is equal to J(z)/z for z ¢ 0 and to j'(O) for B == O.' . we apply the theorem to J~)/M or.
.
. Theorems to this effect are very useful.:s.. . hence.. If J(z) is analytic for Jz r < 1 and satisfies the conditions If(z) I S 1..... the . The following particular result is known as the lemma of
Schwarz:
Theorem.:~ : :
. If Zo is on the boundary.%01 == r and... subject to the condition that the closed disk Iz . and hence '/1(Z)1 ~ l/r for lal ~ r.' ::.
.. .' ..tion· of oppoeite arcs~· ...»: '.'.'. :
. 2 1 _] Zl . THE"· GENERAL FORM
r
'I'(z)\
<
=
1
Izl2
. or 181(z)} :S: lTtl.. The gener&lization ean proceed.z. ..
. _.. Deduce that a linear transformation of the unit disk onto itself preserves noneuclidean lengths. For one thing we can seek to characterize the regions in which Cauchy's theorem has universal validity. . .. that (1 .. .If(z).t the Me of smallest noneuclidean length that joins two given points in the unit disk is a circular arc which is orthogonal to the unit circle... L Prove by use of Schwarz"s lemma that every onetoone conformal mapping of IL disk onto another (or a...131
COMPLEX
ANALYSIS
COMPLEX
INTEGRATION
137
of the original theorem..
.<...::~ .. In the first place we must generalize the notion of line integral.anu
'z'
which is valid when "VI..~.. show that
If(z)
.
~z.
.
OF CAUCHY'S
THEOREM
is analytic and Im J(z) ::: 0 for Im
1/(2) ...:. .'. (2) subdivision of ~~an (3)· fusiOn . The· guiding principle is that two chains should be considered icUm.'. {.: :. it is clear that different fonna! 8UMS can represent the same eh~n. form a subdivision of the arc ~..ZlZt
1+
Zt 
Zt %IZ. and we define the corresponding.:(~).. Since the righthand member of (37) has a meaning for any finite collection. ~ Derive corresponding ·inequalities if J(z) maps 1z\ < 1 into the upper half planet ". . Explioitly.. Just ~ there is nothing unique about the way in which an arc can be subdivlded.
2. '1'1 + 'Y! + · · · + ~1tt which need not be an arc.og .led by considering line integrals over arbitrary chains... .lz (z =
S...)
R'l ... . ~
1
lzt ~
1 implies
. To this end we examine the equation
(37)
1"I
transformation. " • ...:. Show that Boll analytic function J(z) with l/(z)l < 1 for < 1 maps every ~ on an arc with smaller or equal noneuclidean length.' • _ • :. we find that the following operations do not change the identity·' of· a chain = (1) permutation of two arcs...wof(z}
.I.
.ub&rcl to a single arc.1. vt..
j
If(z).Jdz
. Chairu and Cycles. I'...:
' •• : •• "
.' .wo)
j'
s
R(z .
J
fdz
=>
J.. half plane) is given by a linear transfonnation. . : . (4) ·on~·ofan lie. in two directions..•.. _ . Hence we obtain lSf(T1f)1 ~ lr\. For the purpose of studying the local properties of analytic functions this was quite adequate. Show by use of (36).::'1 .
' '. 1 and 2.. Prove tha. . or directly.
. If 'Y is a piecewise differentiable arc contained in 1z1 < 1 the integral
L• •
+. nothing prevents us from considering an arbitrary formal sum .'. Such formal sums of arcs are called chainB~ It is clear that nothing is lost and much JJl&y be PJ. and check the result by explicit eomputation.integral by means of equation (37).ioZ
j.
. _." . . .~:.:.
_'. we can consider an arbitrary region and look for the curves ~ for which the assertion of Cauchy's theorem is true.. _..Jdz + fTJdz + · · · + /.Rzo) 1
I ..
implies that f(z)

1 In EX.!(Zo) . .zal
101
and
Imf(z)
11'(%) 1 s
~y
x
+ iy) .) The shortest noneuclidean length is called the noneuclidean di8t..
• ••••
.~+'"·· +. (Make use of a linear transformation that carries one end point to the origin. How should noneuclidean length in the upper half plane be defined
4.. .. the other to s point on the positive real axis... this inequality can be written in the form (36)
EXERCISES 1.
1 for
11 1. If this prinoiple is Analyzed..tieaI if they yield t~ Mme line integrals for all functions J.!~:I" '_••• :'" •••• ~~.
is called the llonfUClidean length (or hyperbolic length) of~.z! 1 ..
: .
I· •
. .2)
2.. .
z > 0..
. but from a more general point of view we cannot be satisfied with a result which is so obviously incomplete.On thi8 bUia it would be e8sy to _ a
~j
":"":
.. Secondly.. Derive a formula for the noneuclidean distance between %1 and Zit Answer:
M(f(z)
M'l .... prove that equality
is a linear
In our preliminary treatment of Cauchy's theorem and the integral formula we considered only the case of 8 circular region..
. If fez)
satisfies the hypothesis
between the end points.nce1ii. . . "(t. 4..
however. If the complement of 11 is connected. Very simple combinatorial considerations show that a chain is a cycle if and only if in any representation the initial and end points of the individual arcs are identical in pairs.. With this notation every chain can be written in the form (38) where the Gj are positive integers and the rts are all different.
" _ ':~ . .. and 8 parallel strip are simply connected. One of these sets contains co.The coefficients will be arbitrary integers.'
.. :.. and it is for simply connected regions that Osuchy's theorem is universally valid. According to this convention the outside of a circle is not simply connected.
fM'
TheoreDl 14. The suggestive language we have used cannot take the place of 8 mathematical ~finition. region.'
. let A be the bounded set. Inasmuch as the situation does not involve any logical pitfalls. .
I'"
'
. . Thus it is immediately possible to tell whether a chain is 8 cycle or not.a} = nhl. the integral of an exact differential over any cycle is zero.: ~: : ~>~'
. In the course of our work we shall find. Our choice has the advantage of leading very quickly to the essential results in complex integration theory...
Definition I..2. It is clear that all theorems which we have heretofore formulated only for closed curves in a region are in fact valid for arbitrary cycles in a.
.. The sum of two chains is defined in the obvious way by juxtaposition . . The last device enables us to express any two chains in terms of the same arcs. There is little doubt that all readers will know what we mean if we speak about a region without holes. Such regions are said to be simply connected. and their sum is obtained by adding corresponding coefflcienta.: ). we shall dispense with this formalization. The definition can be applied to regions on the Riemann sphere. In the applic&tions we shall consider chains which are contained in a given open set rt By this we mean that the chains have a representation by arcs in n and that only such representations will be considered.. but fortunately very little is needed to make the term precISe.E A. Let ~ be any cycle in 0. for its complement consists of a closed disk and the point at infinity.
'"th
'. The last example shows the importance of taking the complement with respect to the extended plane.:
. The necessity of the condition is almost trivial..~. _. a haIf plane. it is convenient to denote the sum as a multiple. Consequently n("(. single closed curve.y and continue the reduction of 'Y) (38) until no two '1.?f:". It is clear that the additive property (37) of line integrals remains valid for arbitrary chains.fJtl~~ . .t is commonly accepted. Indeed 8.. A region is simply connected to the extended plane is connectedt
if
its complement toith reqect
At this point we warn the reader that this definition is not the one tha. The zero chain is either an empty sum or a Bum with all coefficients equal to zero. and terms with zero coefficients can be added at will.~. = .. The index of a point with respect to a cycle is defined in exactly the same way as in the case of a..iJ... liesat ~ center:..
..~~\ ... and this condition is usually chosen as definition..region without holes is obviously one whose complement consists a single piece.. more or less equaIly important.a) = 0 all cycle8 'Y in II and aU points a tohich do not belong to D. A chain is a cycle if it can be represented as a BUm of closed curves.~_....
".. In particular. It states tb.
'.
• I'
I.
'. The seta A and B have a shortest distance 0 > 04' Cover the whole plane & .. For opposite arcs we are allowed to write a( .. . We ~ free to chocse the net ~'·~~.(.&t & closed curve in a simply connected region cannot wind around any point which does not belong to the region. We are thus led to the following definition:
0/
ThiB alternative condition is also very suggestive.
A region [) is simply connected if and only if n('Y. and this is evidently the most symmetric situation.
.h~ boundary I. it must be contained in one of the regions determined by 7. for the complement of the strip in the finite plane is evidently not connected.'
~'{
.131
COIIPLEX
ANALYSIS
COMPLEX INTEGRATION
131
formulate a logical equivalence relation which defines the identity of chains in a formal manner.
.
.a) == 0 for all finite points in the complement.. and inasmuch as Q) belongs to the complement this must be the unbounded region. For our purposes it is nevertheless better to agree that all regions lie in the finite plane unless the contrary is explicitly stated.a} nhIJ4). SiJnple Conneetivity. It seems quite evident that this condition is not fulfilled in the case of 8 region with 8 hole..
+
+
4.a.··· ..'.. and the otheris consequently bounded.. union A U B of two disjoint closed sets. It has the same properties and in addition we can formulate the obvious but important additive law' expressed by the equation n(Yl "V1.
••• ':I"~' . We assume that the complement of D can be represented as the ..~ . It is easy to Bee that & disk.:'. One of these states that any closed curve can be contracted to B point.
• .=. 1.. are opposite.~~~t. .. .:
:. the main reason being that our definition cannot be used in more than two real dimensions. that the property expressed by Definition 1 is equivalent to a number of other properties.lle~~ @CI~ Q of side < 8/0.3:.~y. Jr.: ·~~0. When identical chains are added. For the precise proof of the sufficiency an explicit construction is needed.~'tffi~~~
...~k:(t.')
.
.
4.~ .'v. .::"':.'::.f~ Before praying.
.'
.
1
J. then (40) . .
''':..140
COMPLEX
ANALYSIS
COMPLEX
INTEGRATiON
141
[] while its integral
f. ..: ...
Curve with index I. 19.
(40)
The definitive
1/ f(z) is analytic
in 0.{ '
. it is: evident that n(y. .
'.
curve of Q is denoted by 8Qi we assume explicitly that the squares Q are closed and that the interior of Q .
.
. I' ••
J
.:r:'p... In combination with Theorem 14 we have the following corollary. .:~q.. any side which meets A is a common side of two squares included in the Bum (39)....'1'2 o...a) ~ O.. . A cycle 'Y in an open set 0 is said to be homologous to zero with respect to n if n(1'.:
'.a) = 1..
.' .:.._
I. . Consider now the cycle (39) where the sum ranges over all squares Qi in the net which have a point in common with A (Fig.~..3..
I
I'
. . Homology... .
".}~. ~::. z dz a "'" 2nnh. and since the directions are opposite the Bide does not appear in the reduced expression of 'Y.. 4~9)4 Because a is contained in one and only one of these squares.... our terminology does not quite agree with standard usage..the
• ". .. 1..J
f""oo.
Definition 211 In symbols we write.
._:._ •. It was Emil Artin who discovered that the characterization of homology by vanishing winding numbers ties in precisely with what is needed for the general version of Cauchy's theorem...a) F 0 for some a outside of ll... Indeed.:. .:: .a) = 0 for aU PQ'ints a in the complement of n.
.:...
4.J
I.
r.. then 1/(% .::. In fact. The notation 1'1 ")'2 shall be equivalent to"'Yl .. ·ti~ up ~th·tha _·~·:in··h f4qU i·i3.I
FIG. ~
.•• :.:.. but which a cycle in an arbitrary region or open set may or may not have. The characterization of simple connectivity by Theorem 14 singles out a property that is common to all cycles in a simply connected region... ·. This property plays an important role in topology and therefore has a special name. form of Cauchy's theorem is now very easy to stateTheorem 15. then it holds for all analytic functions in 0.:.
In a different formulation. namely those of the form 1/(z .As:~ibted· . '
~.
. ..·.I:'

... I
o. 0 (mod 0)..J~I!i. .{. Furthermore..ll.. .lies to the left of the directed line segments which make up aQ. Corollary 1 If I(z) is analytic in a 8imply connected region 0.&. This idea has led to a remarkable simplifica.outlifl·Wt·~~on. '... then
jor every cycle "Y which i8 homologous to zero in
o.. and 7 0 (mod 0) implies 7 0 (mod rt) for all 0' ::> {l Again. .
..
~.C~ _..~. ~..
.. :.
"
.
. . .~:
. and our theorem is proved.... When it is clear to wha.
. .I
I.. it is clear that 7 does not meet B. Wl . we make an observation 'W~Cb.
.
.: .."~ The General Statement oj Cauchy's Theorem.
:_ ... Homologies can be added and subtracted.: •. ~ _ ~~:. ~ " .
•..tion of earlier proofs.._. we are claiming that if 7 ~ such that (40) holds for 8 certain collection of analytic functions. .
.. it is equally clear that 7 does not meet A.. Hence 'Y is contained in 0..~'..~ ..
.
. But if the cancellations are carried out.'I
_)!
:.
"
.
:
.a) is analytic in
...1.J
. 0 need not be mentioned.. We remark now that Cauchy's theorem is certainly not valid for regions which are not simply eonneeted.
theorem.. if there is a cycle y in n such that n{toy...
.a) with a not in 0.
I·
.t open set we are referring.
.. ...I: .
holds for all cycles 'Y in
.
then
f is analytic
in D.. I'a is a stun of oriented line segments which are sides of exactly one Qj.:'. and if 8 is sufficiently small it is not empty. The union of the squares Q/..
= F(z) 
+
4. we find that
eJ1(. '
:'.'.. who has kindly permi~ted itB use in this connection
r
t This
.'
:..:. : ..)~' (.
:'~~~"!/~.5.. considered as a point set.. .
=..
:11
..
. Proof of Cauchy's Theorem" t We begin with a construction that parallels the one in the proof of Theorem 14.COMPLEX ANALYSIS
COIIPLEX
INTEGRATION
143
the validity of (40) for all closed curves "Y in a region means that the line integral of J rk is independent of the path.. or that f rk is an exact differential.
(41)
if}
=
i«
if j #. because n is bounded the set J is finite.r) = n(.' :~: . ': .
.. . j E J. In a simply connected region every analytic function is thus a derivative..jo
fez) = ~ ( f(r) dt.
.. we cover the plane by B net of squares of side 0.
In fact.I) has the derivative zero and is therefore a constant.

'. is contained in 1b it follows that n(~.To we merely write it in the form exp «lin) log fez))..'..
and consequently we can set log fez)
define ~
F(zo) log 1(20).
'.::':~ . Choosing a point Zo E n and one of the infinitely many values log f(zo). A particular application of this fact occurs very frequently: Corollary 2~ If f(z) is analytic and :# 0 in a simply connected region 0.zJalued analytic lwancheB oj log /(1.. the elosedsquares in the net which are contained in 0.'" . There is a point to E Q which is not in o.:... ..'
. It is possible to join r and r0 by a line segment which lies in Q and therefore does not meet Oa. . then it is p08sible to defi~ Bingle. In particular.) and

~inn. consists of closed regions whose oriented boundaries make up the
FIC" .11
Suppose now that My...fo) = O. .... we know that there exists an analytic function F(z) in n such that F'(z) = f'(z)/f(z)~ The function !(z)rF(. Consider a point E (} . r ~ z
r
proof follows a suggestion by A ~F ~ Bearden. n('Y.
If z lies in the interior of
Qjot
cycle
'and hence Clearly. We denote by Oa the interior of the union V Qj (Fig. Sinee v. 2n Jr. Given 3 > 0.':..0. but otherwise arbitrary. and we denote by Qir j E J.. '
. Let ~ be a cycle whieh is homologous to zero in 0. By Theorem 1 there is then a singlevalued analytic function F(z) such that F'(z) = f(z) (the pleonastic term "singlevalued' is used for emphasis only). ol + tot:
1(·0> ~
f(z) . It belongs to at least oneQ which is not aQj. Assume first that' n is bounded.r) = 0 for all points t on rIJ.
.)
. we choose a 80 small that 'Y is contained in U.:':. . 410).
::101
a.u" we
+ q dy
+ q dy
=
!.a) = n(~. . For the proof of Theorem 16 we show first that l' can be replaced by Or polygon t1 with horizontal and vertical sides such that .. They divide the plane into BOrne finite rectangles Hi and some unbounded regions R~ which may be regarded as infinite rectangles. so that·~ ~ 0 (mod 0').o~. ~t n(iJRi 4. we sha.(I~)."G. It is clear.t 'If [1 is unbounded.Choose a point a.i.. To continue the proof we extend all segments that make up (t to infinite lines (Fig .
J
.. the . p dx
=
There does not seem to be direct way of modifying the proof of Theorem 15 so that it would cover this more general situation... for no ailies on a. ..
:
~.) 1" if k == i and 0 if k . If 'Y is given by z = z(t). Locally Esact Differentia"'. The 'separate proof of Cauchy's theorem has been included because of its special appeal. It will thus be sufficient to prove the theorem for polygons WIth sides parallel to the axes. With: ·this·. The corresponding aubares 'Y i of 'Y have the property that each is contained in a disk of radius p which lies entirely in 11.a) = 0. It is also true that A(cr.) . Hence the integral (42) is zero and we have proved the theorem for bounded f. we replace it by its intersection g' with a disk IzI < R w hieh is large enough to contain 1'.
j
In other words.ll follow Artin's proof of Theorem 16. . In either ease n("Y. 0..) and ft{•• a~} =.ln. p. If P dx
+ q dy
o. Let the distance from "Y to the complement of l1 be p.
is locaJly exact in 0.
(~ f f
T
2T'l
rs
fer)
dr) SZ
dz =
J r (~f 2. the coefficients n(a. This condition is certainly fulfilled if pax q dy = f{z) dz with j analytic in fi.

1.oJ
that
for all 'Y = aR where R is a rectangle contained in o..every loc~y ~t differential has the same integral over a as over 7. . The proof is applicable to {t.lJi) aRi
where the sum ranges over all finite rectangles. 148) that this is so if and only if (43)
FIG. and we conclude that the theorem is valid for arbitrary o.. . We detemune 8 >·0 so
.r) = O. The end points of 'Y' can be joined within that disk by a polygon at consisting of a horizontal and a vertical segment.a) for a in the complement of 0. 1..6. . Any ~int a in ~e comple~ent of D' is either in the complement of n or lies outside the disk. ThIS property implies. We construct fT 88 an approximation of 'Y. f~ction z(t) is _ lmiformly continuous on the closed interval [arb].ai) are well determined..
divide [o"bJ into subintervals of length < &. . from the interior of each Rit and form the cycle {44}
aD
=
L
n{a.In the discussion that follows we shall also make use of points chosen from the interior of each R~."#.411). (43) is then true for any cycle y 0 (mod 0).
loy p d» + q dYJ
/.TION
145
order of integration can be reversed." .. We shall therefore end up by presenting two different proofs of Cauchy's general theorem. p dx
+ q dy
=0
obtain
for every cycle 'Y ~ 0 in
/ . Since the differential is exact in the disk. .at
from (~).1.
. similarly.. then
lUi pax
and if u
= l. in particular. As in the earlier editions of this book.
• I • • •
any
as desired.~.
fz(t) .
. . p dx + q dy = 0
+
.. n(u. nt~R"a~) ~ '··0· ot :iill j.144
COIIPLEX
ANA. n<. A differential p d:t + q dy is said to be locally exact in 0 if it is exact in some neighborhood of each point in rt It is not difficuit to see (Ex. mind it follows f ~t n(""a.4l1
J.LYSIS
COMPLEX INTEGRA. .
'Y
r
dz
2
)f(f)
dr~
On the right the inside integral is n(7.i P d:t + q dy.z(t')
I < plor it ~ til < o and
Theorem 16. and ~ce (1 ~ O. and by Theorem 15.
. respectively.I said to ~orm a h.. 11 :i.":::
: .y ~ nO wMre n == n(y. • . If thiS circle.~.
jdz
+ · ~ + c.aDcontains the multiple CUii. but by an elementary theorem in linear algebra.. • • . . A region which is not simply connected is called multiply connected.
.a) is constant when a varies over anyone of the components Ai and that n(y. let Ail At.
+
r
i
n(a. Suppose that a point a in the closed rectangle R. and not on "(.
t>
. .. On the other hand. pdx
and Theorem 16 is proved. and assume that co belongs to A If . But it is clear that the cycle CI~l CITt + . and the number of basis elements is one less than the connectivity. and a = ai· From this property of a ~ 0'0 we wish to conclude that a(l is identical with (T up to segments that cancel against each other... Moreover._::
:.c (JR. I I. the periods of the indefinite Integral.. a region of_ connectivity n arises by punching n holes in the Riemann sphere . for if two linear combinations were homologous to the same cycle their difference would be a linear oombination which is homologous to zero . hence it cannot be homologous to .l'Yal. times around the ·points in Ai.O)~ The integral of an ~aJytie ~~tion over a cycl~ is a multiple of the single period '_
. any cycle m the annulus satisfies ... whose corresponding coefficient n(rt.
jdz
=
elf
.
.tla .
:
.
JIk
+ C!J
")1'. . is an arbitrary cycle in 0..t (45)
a=
f
For a given cycle T in D.be the constant value of n(.. and hence n(ct.. we conclude that c = o... i == 1~.
.
. we can prove.
4.
.~:
. TtJ .a. and the integral along an arc from Zo to z is determined up to additive multiplee of the periods. that n('Y. and a homololY 'basis is formed by ~y circle Izl ... In other words. let us consider the extJ emely simple case of an ann~ defined by "I < 121 < rl~ The complement has the components .Suppose that the reduced expression of a ...C. :'. It is not the only homology basis. be the com .
+ qdy = 0.'1. .0 and rl = 00. _.7.· + c..:) iJR •..Ct'Yt . On the other band. More precisely. for any analytic function J(z) in 0. The vanishing of the periods is a ~ and sufficient condition for the existence of a sin21evalued indefinite integral . such that n(riJa) == 1 for a cf Ai and n('Y~Ja) = 0 for all other points outside of D.zero unless all the Ci vanish.141
COIIPLEX
ANALYSIS
COIIPLEX
INTEGRATION
147
the interior of u. In the case of finite connectivity.
"f ~
el'Yt
+ Ct1'1 + .. just as in Theorem 14. We find that any point outside of (I has the index zero with respect to the cycle".l" ·
.• • • .
are
+
j
. Consequently..tli) is different from zero are actually contained in 0. · + e. and it follows that ai and ak must have the same index with respect to this cycle.. .. r.1"
. let Ci. They are ca.Bi) = n(a"a) = 0. the line segment between a and ~ does not intersect a..1
fda. The same reasoning applies if t1 ij is the common side of a finite rectangle Hi and an infinite rectangle R~~ Thus every side of a finite rectangle occurs with coefficient zero in a . In a less precise but more suggestive language.
I
depend only on the function.uo..1. lies to the left of «o. By Theorem 18 we obtain.
. .. ponents of the complement of fl.Every cycle is thus homologous to a linear combination of the cycles ~1.omology bari8 for the region n. ·i:.... "rIn order to illustrate.: .
We prove now that all the R.~.. is denoted by 0. We conclude by the local exactness that the integral of p dx q dy over any aRi which occurs effectively in (45) is zero."Y. ~1 < r < rt.l winds e. I We have thus shown that n(tt .. we include the degenerate cases'l . 'Yt... were not in o. we choose the orientation so that R. does not contain flil:.t"...a) == 0 in A.
. duplicating the construction used in the proof of the same theorem we can find cycles ~i.1f
t
1'".. proving tha.
The numbers
i.. .
~
JI. we may conclude that every homology basis has the same number of elements.a) = 0 because 11 ~ 0 (mod n).. Let ail: be the common side of two adjacent rectangles Ri.. ..... or.. Multiply Conneeted Regioru. Then the cycle f1 .Iled modulu oj periodicity of the differential f dz.. (} is said to have the finite connectivity n if the complement _of{}has exactly n components and infinite connectivity if the complement has infinitely many eomponents.cycle is a linear combination of the periods with integers as coefficients.rl and Jzl· 1= 'I.. The annulus is doubly eonneeted.Ct'l'1 .::.n 1. with less accuracy.
In view of these circumstances the cycles '1...4) for a E A. We have found that the integral of f(z) over any .a) = 0 for all a = a.
~~. .
. Then n(a. these indices are c and 0.. .0'0. .R. J A..
.t:.. We find that every region with a finite homology baeis bas finite connectivity. This linear combination is uniquely determined.. belongs to the unbounded region determined by (1.·· .
We repeat the definition in the following form: Definition 3..
.
q dy is locally exact
to sero.. The constant Ri which produces this result is called the reBidue of fez) at the point OJ.
. For a first orientation. What are the possible values of
f~
over a closed curve in the region?
5.. vanishing period. was introduced.~~ ~.
lri. To each a. In order to make this method more systematic a simple formalism.. Show that the bounded regions determined by a closed curve are simply connected..Djl < {. We tum now to the discussion of a function f(z) which is analytic in lit region n except for isolated singularities. 4.."a)!(a)
== ~
f7za dz /(%) 2ri
J'll
2. 4 under particularly simple circumstances .: ::.~. that the region obtained from a simply connected region by removing m points has the connectivity m + 1..:1 .j is contained in 0'..
"
.. . ... Indeed.o. the founder of complex integration theory.
EXERCISES
L Prove without use of Theorem 16 that p dx
in Q if and only if
+
For instance.. .al < at
It is helpful to use such selfexplanatory notations as R == ~f(z). . known 88 the calculus of residues. Cauchy's integral formula can now be expressed in the following form: If fez) i8 analytic in a region n~ then. Let 7 bea cycle in 0' which is homologous to zero with respect to Then 'Y satisfies thM homology
book.weIt
.~ . . ". The proof is a repetition of the proof of Theorem 6. • J (1. our diseussion of the local behavior has already shown that all removable singularities can simply be ignored. ~_ . S..:: .. _. The r~1..yerify~ that 1M pointadj· __ '.~ can be defined in any region such that the points + 1are in the same component of the complement. ~ ..
..d use of residues amounts essentially to an application of the results proved in Bee. tt.~
. all results which were derived as consequences of Cauchy's· theorem for a disk remain valid in arbitrary regions for all cycles which are ho~us
.. In this connection we point out that there is of course no longer any need to give 8.ve shown that the determination of line integrals of analytic functions over closed curves can be reduced to the determination of periods..
aj
has a. .
'"
..aj)Cj . let us assume that there.
":. THE CALCULUS OF RESIDUES
which i8 hotnologous to zero in o.. :'.
:.1.. while the unbounded region is doubly connected. and find 9..W of fez) at an isolated singularity a is the unique complex nuff1rb6r R whU!h makes J(z} . We are thus in possession o~ a method which in many cases permits us to evaluate integrals without resorting to explicit calculation. are only 8 finite number of singular points.. The Residue Theorem.
.::. . ....~.
n(.can be defined in any simply connected region which does not contain the origin..
. Lwi7f.RJ{z . za and z. if we set R.a) the derivativ6 of·a singlevalued analytic Junction in an annultt8 0 < Iz .
n. Show that a singlevalued analytic branch of. Draw a circle OJ about tlj of radius <&iJ and let
for every cycle
'Y
(46)
be the corresponding period of fez)..~~. .·~·li"'. 4.: ..: From the point of view adopted in this
R. . ..~(:. Our first. separate proof of Theorem 15 in the presence of removable singularities. . 3. . denoted by a1..\?. The region obtained by excluding the points tIj will be denoted by a'.... :'.t..'. ...
i
w8". Cl~ly. by Cauchy.n·.. :"
'.
. Show that singlevalued analytic branches of log z. Therefore.' Under certain circumstances it turns out that the periods can be found without or with very·little computation..
5. B2.. homology basis.""1: n{')'. Prove
(
pd:t+qay=o
for every rectangle Reg with sides parallel to the axes.OJ) has the period 2...: I¥.:. ..
. = Pj/Zri. and thjs application is typical.141
COMPLEX ANALYSIS
COMPLEX
INTEGRATION
141
whose value is of course independent of the radius r.tla8il7. the combination fez) _
z
The results of the preceding section ha. .i.

. This is of great value for practical purposes as well as for the further development of the theory.. . there exists a ij > 0 such that the doubly connected region o < Iz . ~'. : .task is to review earlier results in the light of the more general theorems of Sec. The particular function l/(z ....
In rigorous texts considerable effort is spent on proving that these intuitive notions have .a)..dr ·r •
I. )' at z = a is e.'
I.~e~
"
•• ' •••
.) is either o or 1.) ..b)..a)f(z) for z ~ B..a)l
+ $'(z)
• ••_. for then the function (z .'
. number of the isolated points BJ.. In presentations of Cauchy's theorem.. then .a.....a)2.
/. then it i~ clear that r is homologous to zero with respect to 0'.. bounds 0..
:'"
.
. a R8'lnaTk.a)(z . If..'
.~ . For essential singularities there is no such procedure of any practical value. and indeed quite easy. ·~h~. and we have proved:
Theorem 17.a
J
J)
Ress_. We need... Then we have simply
~ J't J(z) dz = l Res...
to recognize that the residue equals the coefficient Be: Indeed. is even more immediate.b)
This is the T~ue ~OTem~ except for the restrictive assumption that there are only a finite number of singularities.150
COM'LEX
ANALYSIS
COMPLEX
INTEGRATION
151
By virtue of the homology we obtain. subregion of Dt and the orientation is chosen 80 that the subregion lies to the left of o.'
.0 isolate the classical case. :""'!" ..~. for then the same proof can be applied. The following atatements are therefore trivial consequences of Theorems 15 and 17: If 'Y ~b~ n and fez) is analytic on the set 0 + "i..a precise meaning.
In the applications it is frequently the case that each n(""a. and if n + 'Y is contained in a larger region 0'. prising that the residue theorem is comparatively seldom used in the presence of essential singularities .(z)(z .OJ)P"
and since P. the term Bl(.
..
I
.a)~ is omitted. with the notation (46). The residue at a is obviously ea/(a . 1.
. The set of all points a with n('Y. Rj finally
Since the principal part at 8 pole found.a) + !. All that is needed is to accept the following definition: ..
A cycle r ia 8aid to bound the region n if and only if n(r..0 for all points a not in
n. the remain~r· is e\1i~t1y 8:.j!(z) _
Jor any cycle 'Y which i8 homologous to zero in 0 and doe« not pas8 throUgh any of the points tlj. The main objection to this procedure is the necessity to allot time and attention to rather delicate questions which are peripheral in comparison with the main iBBUes.4
+·
t
•
+ B (z
1
.a)1 we find that the residue of e/(z ."
. •
. and the residue at b is tJ/(b . In the general case we need only prove that n('Yraj) = 0 except for & finite number of points Q.
a
~
fTf(z)
dz
=
l n("f.a. the curve 'Y to which the theorems are applied is supposed to form the complete boundary of 8.ve at our disposal " simple procedure to determine the residues.
/. The residue theorem is of little value unless we ha. there is no mention of homology) nor is the notion of index used explicitly. ~ . J1'f(r) .. The complement is consequently a compact set. Instead. and as such it cannot contain more than a finite.
. and thus it is not sur .a) iB defined and equal to 1fur all points a E 0 and either undefined or equal to
Definition 4..rf: 0 only for a finite number of the singularities.8 .a) = 0 is open and contains all points outside of s large circle._.(z . The assertion follows by routine reasoning. .
.
.~..+ e(z .. Dividing by (z .
_ •••••
. H b == a. the situation is slightly more complicated.a). With respect to poles the situation is · entirely different. f dz = !n('Y.. For the residues of the function
~ J. Therefore n( 'Y..
•
B.o..1(z) tU "'" 0
"'..
where the sum is extended over all singularities enclosed by "y..': . 2wi
!.'
. I(:!)
i .. Let J(z) be analytic except Jor isolated singularitia region Q Then (47)
oJ in
at the poles a and b ¢ 4..
= 2ri.'
'.
./ dz
==
r
J
n(y. the integral formula and the residue theorem which follow more classical lines. only look at the expansion f(z}
=
.
'
:. we have thus a very simple For simple poles the method residue equals the value of the instance let it be required to find
I
is always either given or can be easily method for finding the residues. We must then expand e' by Taylor's theorem in the form e' == e..
1
(z . With the general point of view that we have adopted it is still possible.:
"
• I

f(z) ..)R.
..
.If r lies in 8 disk which does not contain ·the origin} then n(r. ..
o
Corollary.:~""r. We observe that a cycle ~ which bounds n must contain the set theoretic boundary of o..
.
. Indeed. We can now generalize Theorem 10 to the case of a meromorphic function.":.(z).. the sums in (48) are finite.(z)l < IPttI(z)i on _ ... '''..
". since "y is closed.
Let
I
The assumption implies that J(z) and g(z) are zerofree on over..a)'1A(z). A typical application of Rouche's theorem would be the following..
.(z)
where P tj~l is & polynomial of degree n . we have thus n(r. then every neighborhood of ZG contains points from D and points not in ll.a.O) == O.(a) ~ 0 and obtain j'(z) = h(z .1
:f"..
(49) ~
(48)
. Theorem 18 is usually referred to as the argument principu.)g(Cj)
l
._:' "i . then
If I(z) is meromorphic
in
n with the sero«
a... This observation is the basis for the following corollary..
. n(~. replacing h.i R 88 P . and when we apply (47)J the integral formula..g{z)' < If(z) on '1.O) where r is the image cycle of v. results.a) has a simple pole at z = a with the residue !(a)."fDir' <' 1" hu" lUdt4I . known as Rouche' 8 theorem:
l' be homologous to zero in 0 and such Uta/.~ .. For a zero of order J:l we can write j(z) = (z.. .:~ ~~"A~.~.
. we find by the same calculation as above. nomiaI equation PRl(Z) == 0" . ."
:
.I(z)
in 0.O) = 0.. If such a neighborhood were free from points of "V.:
:.a) + f~(z)/f. For a suitably chosen n it may happen that we can prove the inequality R"lf.':. ThiS result
is iniportant for the study of the inverse function. and hence every neighborhood of Zo must meet .:."_<""~::""If)~~p1Y ~ h. 1%' = R. Normally.
'Y
paB8
through
the notations: of Thib. The name refers to the interpretation of the lefthand member of (48) as n(r. and the pow
and the residue ·hg(a) at a pole.1"
:.. meet D.
=~
1
tl'here the sum ranges over the singularities a. .
.
has the residue hg(a) at a zero c of order iii We obtain thus the formula

Theorem 18. Then f(a:} has the ssme number of zeros in Izl . Consequently (z)/J(z) = h. if Zo lies on the boundary of n. Indeed..
bl.2.J. .
... Another application ofthe residue theorem occurred in the proof of Theorem 10 which served to determine the number of zeros of ananelytic function....:
"
.'
.. When Theorem 18 is applied to F(z).z) i8 either (W' 1for an'll point a not on . This contradicts the definition. but for Cauchy's theorem and related considerations this assumption is not needed. If a(z) iI analytic in
n. I
•
nC"I. Theorem 18 can be generalized in the following manner. n('Y.. E [1.
:. Cauchy's integral formula can be considered as a. the function J(z)/(z ."..b.
. special case of the residue theorem. and the assertion follows. then
gee)
~(~1
_.. ~
i.. and we see that f'lf has a simple pole with the residue h.. Using Taylor'S theorem we can write J(z) = Pal(Z)
_
i
+ Z"/. Suppose that we wish to find the number of zeros of a function J(z) in the disk 12'1 !S: R.
lor every cycle
whith i8 homologous to UTO in (].
.:
I. with f..1.
g(z)
~(~i !
dz J
nh. The converse of the preceding statement is not true.._':.the equatien fez) to~ "~' ".z) would be defined and constant in the neighborhood.. . one would try to choose 'Y so that it is identical with the boundary of 0.1(z .)
~4'
More
11 < 1
5.. Sa must lie on 1". and this number can be determined by approximate solution of the poly .
on "(.rem 11 we know ·that .. .and does not any of the zeros or ptjle8.
. If f has a pole of order h.. they satisfy the inequality g(z) 1 1(1.
.. The values of F(z) = g(z)/J(z) on " are thus contained in the open disk of center 1 and radius 1.
:
r. Then !(z) and g(z) have the 8ame number of zeros enclosed by 'Y.
.'
. t.. The Argu~nt Principle.
'~:..1'
' .i("'?.' In the formula (32) this residue is accounted for by a corresponding repetition of terms.
.)g(bt}. with It. that f'// has the residue h~ The following theorem results:
It is understood that multiple zeros and poles have to be repeated as many times as their order indicates. then
~ I.a)'i~(z). Suppose that J(z) and g(z) are analytic in n and satisfy the inequality 'f{z) .l
"'.r :.
WitlJ
j
*
:=e
~~. '.a)Alfl(Z) + (z . for a point on ~ may well have a neighborhood which does not.
..152
COMPLEX
ANALYSIS
COMPLEX
tNTEGRATION
153
If fez) is analytic on
n + 'Y except
J(z) dz
for isolated singularitie8
R~s_.. :~.I(Z).
and we find that the 2. di ealculus of residues yields. ~ ::: ~.
.~·:~·::·~~~~/!.' ~ .:.: ·":'·~~:.1 and apply Rouch~'s theorem.~.._:
.~. All.ttly a laborsaving device. the integrand must be closely connected with some analytic function....
/'(z) !(Il) =.. Look for the bIggest term. ..wol ~ 6 Thus the power sums of the roots 2j(W) are singlevalued analyt~c f~etioDB of w.. while a real integral is always extended over an interval. very efficient tool for t~ ~valuat:ion of definite ~tegr~. .. and t~ ~due . ":·:'I7. ::".I(W)Z + a~(w) = 0
functions of
tD
whose coefficients are analytic
ill.e' + ~ = ~. curve.2..0 . equation
(50) is replaced by f'(:) 1(2) ...e
W
z liz....:"~.~)!O The value f ~.:.:.. . :.'~t·. .i . .. but they all apply under rather special circumstances The technique can be learned at the hand of typical examples.. &1 3. Taking this into account ·we find that the integral equals
•• "1:
.:~..
!\:
o·
.fl RD(z+~}~(z .'
...)P... let us compute Jo a
..... complex rather than integr~s IS?O at ~ advantaget for clearly the ~ua~on of a complex mtegrallS equrv en to the evsJuation of two definite Integrals.Hz + 3 = 0 have their modu
Z7
..
"
.J. · + a.'·.
+
+...~"]~.~ ·:'·:~~~~~~~·~:··.~~··. Ill! > 1.
. in the right half_plane? Bint: Sketch the un~ of the unagmary and apply the argument principle to a large half disk. "
.::. and the method is far from infallible..COMPLEX 114 COMPLEX ANALYSIS
INTEGRATION
155
(49) with g(z) ~ e...:.::}":..:..~.. complete mastery does not guarantee success.:().:.~ :. ' . I •
~ ~
..
•
'..
~:
.' .~.'..t a is l/(a .·.: :·:·.. _ 2."
~<~.'.: :.
If (49) is applied with g(z)
=
=
Z"".w z.
5.. •• .3. for usually we are only required to integrate elementary functions.~'1'i~ .!
!..C. :~:.).> > .'.. Evoluation of Definite IntegraLt.
'.:><·'~·.. ·····' ~ ._t Ii ·tly hut even if the ordinary methods of calculus can be app I grM exp 01 .~.... . ': . integrals of the form
t
The right. It is much more serious that the technique of complex integration applies only to closed curves.. e B.'. .': a~
..'
'.(3) with
.
'. '.
'
~:' The denominator ..'·:..
//:. It is very natural to make the substitution z == eitl which immediately transforms (51) into the line integral
+ al(w)zl + . Hence they are also analytic. but eveo . . How many roots of the squataon r ...!)]~..:<.5~. . polynomial equation· z"
(51)
/02.. There are a number of ways in which this can be accomplished.::. 1.
EXERCISES 1. As an example.
.(W) are the roots of a. '~..
.i
1
J 1. .:.
can be factored into (z . particularly important when it IS impossible to find the mdefinlte nlied .:~..{~ ...< 1... :.
.
i1
2: zlw)ft.is onehalf of the integral from 0 to 2r. In the first place.1»0\ < 6.
1w .
.0i:.:.:'..2Z&
+6z
l

2
t=
+
dO cos
e
J
a> 1. when Izi . :' .!.
+1_
It remains only to determine the residues which correspond to the poles of the integrand inside the unit circle.
For
fi
= 1 the in verse function rl(1») can thus be represented explicitly by
f1(W)
=<
1 2..
~. J ..!:: .
. and they can all be extended to the complex domain. :.. however.
iJ.
.~ integral is not extended over (0. but this technique is usually very laborious. • Th f t that the the use of residues is frequet..
Ius between 1 and 2? ' 3 0 Ii 3.
..01
J. The calc~ua ?f residues p": vides 8.hand member represents an analytic function of w for 11.
.
_..: :. :... we obtain
(50)
i ~1
L Zj(w)
1l
=
2. but since cos 6 takes the same ·~ues in the intervals (0. But it is well known that the elementary symmetric functions can be expressed as polynomials in the power sums...::!
."..'.~.. How many roots of the eqUAtion Zl +. _~. ::~t·~""\'·~:~~~. .:?\:.:
.·~ently 'al.
. .
..a) (z .
. A special device must be used in order to reduce the problem to one which concerns integration over a closed. .
rea:I
Q."':~ . R( cos OJ sin
0)
dtJ
where the integrand is a rational function of cos B and sin 8 can be easily evaluated by means of residues.) and (w.
:: :. Of course these integrals can also be computed by explicit integration.'/ •••• '. some serious limitations. How many roots does the equation
0 have in the disk Izi < 11 Hifit:.~:·f···..1"
.
2ri
1
J
1S::lIl·
There are. {j == a~~ .: ...: ~.:. ~ ..~·
h '''. .. This is not very serious.dz.
• :.
~>~~Y~~.r·:~:
.i
I&zol'_'~
Iv) 
f'{z)
'W Z
dz.Zw) is is clear that the integral from o to .
._::. For the proof we integrate over the perimeter of a rectangle with the vertices X2~ Xt iY...u. For one thing.~ .l(8)tfit·_..~ .. Hence the integral over the right vertical side is...'~ .
.
Since leUl . tH .rDI?cf•• · ~':....meirC1e ia~.. For fixed Xl. Xl where Y > 0. encloses the poles ~P8t~b&lf plane. provided that the rational function R(z) has a zero of at least order two at infinity. but we use a path which avoids ~~.
R(z)..
R(x) rk
converges if and only if in the rational function R(x) the degree of the
denominator is at least two units higher than the degree of the numerator.~~ (53) has a meaning and calls for evaluation.: :: :_.:. . In the discussion we have 8BSUJD. in the lower half ~ane (~412) . ~.>0
L
Res R(z)e"
r<A
(}
1
+ _!_)
Jr.2ri
.. and the corresponding integral is equal to 2ri times the sum of the residues in the upper half plane. ~..:. As soon as XI. ...
f _"IQ
R(x) dx == 2ri
:1'>0
L Res RCz)·
3. XI iY.it~. Hence the integral over the right vertical side is less than a constant times lIX!..!tJttriliil~~ ~..
f~".
f
Xt
Xl
R(x)eiz dx
.
over a symmetric interval hss the desired limit for we are of course required to prove that :
In reality
... 2.ed. 'and noothers aIJ BOOn as "'l)~~y 'are8l1fti~~t1i J&rge~and'is rndBcien~lJID8ll. We obtain
where . An integral of the form.c.. o~ tb~ j~tegr~8 (53) may well exist... The same method can be applied to an integral of the form
(52)
The las~ integral can be evaluated explicitly and is found to be < 1.1..iDWJ~a1_~~. if R(z) has simple ~~ w~ch eoineide with zeros of Bin x or C08 x.. I
R(x)e<s dz . Let us suppose. . except for a constant factor. .OJ~18B. As p ) ao obvious estimates show that the integral over the semicircle tends to 0.) whue·
~.. :
• •I • • •
<.~ ...R(x)e~ fix
whose real and imaginary parts determine the important (53)
J:. for .p
This inequality proves that dz = 2ri
r:.
..:lI'. . this rectangle contains all the poles in the upper half plane.
integrals
f ~. tacitly.p) and the semicircle from p to p in the upper half plane.. ~. We use the same method as before.80·that·~:·ean ·lftite l.
+
+
less thaD
f
y
o
e» dy < 1 rzl XI
JY
G
ell dy. Then the second inte._.. that R(z) has no poles on ~he real axis since otherwise the integral (52) has no meaning... O.under the sole condition that B( ao) . this it is not convenient to use semicircles..·.:. even if we were successful we would only have proved that the integral
In
case
f
~II
R(z)e"" dz
p + 00. ev is bounded in the upper half planet we can again con· elude th&t the integral over the semicircle tends to zero.: :. and we obtain
has a limit when Xl and XI tend independently to CD In the earlier examples this question did not arise because the convergence of the integral was assured beforehand.:. .. oricin by following & small semicircle of radius .
..r . .~ . It is easy_to see t~t this closed curve....irlBt&ncej·'that R(z) has 8 simple pole at z = 0. X"2 and Y are sufficiently large.A denotes s constant. :~. and if no pole lies on the real axis. The integral over the upper horizontal side is evidently less than IrY(X1 + Xz)/Y multiplied with a constant."...~lt_"'''.:\t$~.. If p is large enough this curve encloses all poles in the upper half plane. .R. ~ .
. and a corresponding result is found for the left vertical side.
1'>0
L Res R(z)e'"
.
L Res R(z)eb.. : ":'<. .
R(x) cos x dxt
f_ R(x) sin x th..~. and we conclude that
f :.. XI it tends to 0 as Y + ee . it is not so easy to estimate the integral over the semicirCle. . + &(.•··c~
..f~.~:~iio__ ~~Of _jiiit·:·1&.
11>0
It is less obvious that the same result holds when R(z) b8s only a simple zero at infinity.)i8~~_lftlc'..: . &ppoae that .L..
J~. R(x). thepole at the origin. Under the hypothesis IzR(z)\ is bounded. namely.. . : ~:.. The standard procedure is to integrate the complex function R(z) over a closed curve consisting of a line segment (p.:::. How~J one. and secondly..COMPLEX INTEGRATION
151
COMPLEX
ANALYSIS
t
157
of the integral is found to be v/v' at . ~ ::.'" th = 2ri
.
.v4'
f.
':. This.' .
. ier··it~?daea.
I
:""
'I
".~ :..
"~.. f'~.. and es before we m assume that H( «l) = o.. COMPLEX ANALYSIS
INTEGRATION
158
.
•.
.
'rrr J.. ~.. or ain~2: evaluated by the same technique....: :~::. . .
".v
com
ID. _The
. Indeed...i..~!
..... Slmple. As the Simplest example we.>.::. ~.'!is ..0
~
..e2ri.L~~··· .' .'
~u::'~ '+':. ".::. :: '.:..~~
o
Fie.' t· ·al by the fact that the Integrand IS odd.. '.. For practical purposes it may be preferable not to use any preliminary 8Ubstituti~ and integrate the function ziR(z) over the closed curve shown in F~ 414~ We have then to use the branch of . j'_
~ ..."
'!. we can apply the residue theorem to the function ZIe+1R(Z2). .'.
pr..0 < a < 1. . The new feature is the fact that R(z)z« is not singlevalued...~~::
:.. The simplest procedure is to start with the substitution % = it which transforma the integral into
2
The limit on the left is called the Ca1J.. .. and the eorrespo
f ana bie" ·
.pothes.::~ ~_:t ...:
'......
. " ~ .
.2ri ~ Res R(z)e'
.' •• r I ~ ... as linear
ill
c~:
m. • • . ..
&
+ f· R(x)e""
.d.nge o.
+
r " sin x dz
X
== !. The next category of integrals have the form
. '.. ~..1tOt:~··to:tbe·:hy.. one very large and one very small (Fig.
~'
.1
'!
. .. these factors ean. n. For convergence R(z) must have & zero of at least order two at co and at most a simple pole at the origin .owever.. whose of:tbe residue
:...' ~.~' .:..n: · In the general case where several poles lie on the real aXIS we obtain
a... 1'63.. :"
integrals can be reduced to the form (52) by a oba. The evaluation c8.
:
:
'
.
.i:¥..tIR(zt) dz
= fo"
(ZI . *I. e ng .
2
We remark that integrals containing a factor ooS"'. 1
+
(:Il)Ja+l)R(zl) liz. This·method. h·· at alue and since imaginBJ'y part it is not necessary to take t e prmClp v. r .:d~' .. We use a closed curve consisting of two line segments along the positive and negative axis and two semicircles in the upper half plane.~j
.+lR(zi)
dz. '. As long as we avoid the negative imaginary axis. In the the equatlon IS fIYl .. however t is just the circumstance which makes it possible to find the integrai from 0 to CC.~~.. needs some justifie&tion.
10" z..'.. (z)"" == e1riazl«.1) in the· upper half plane.
theoIem... '.lIsfor determination of the ._. .:~' . :.: ....a. . ei.CAy priflCi~ val~ ~hth~ it exists although the integral itseH has no meanmg. : . These are the same as the residues of z!fR(z) in the whole plane. It is an essen YPO:: that all the poles on the real axia be..
:' . and the integral equals
(1 .
.g
n
.....'.lent'·Iiee· brtrween' 0 'and 21r«. have
L Res R(z)e" .'
fY~...' .:.dx x
H.~! ~._
. .
dx
=
2ri
[L
Res R(z)e
u
00
+ IB1
the exponent a is real and may be supposed to lie in the interval . ' ._":: ~ . .
0
in front itI ¢ OJ we are finally able to determine the value :ijf.1:0
+ Ti
tial h' th" where the notations are se1fexpl~atory. side we observe that onehalf of the residue 0 ~ been mcluded j is as if onehalf of the pole were counted as lymg In the uppe: half PI. the integrand is even we find
f~ee' the factor
JJJ
b· .:..
Se arating the real and imaginary part we o~rve that~he real part Qf p .
.~'iJ.
. : ~>"~'l' : ~.~ the desired integral.. •• '. . 4(13)4' Under our assumptions it is easy to show that the integrals over the semicircles tend to zero.. ·.!2.\ . f
WheR
R(x)e'fN# dx
.residues of zk+1R(.....
~'f'.
. it is well defined and analytic in the region obtained by omitting the negative imaginary axis.~ ~
LI..alJIOns
of terms
C08
mx and sin fRX.
":..'~.
1 = m
j
m
R
(z) m

ei~dx~
4 .COMPLEX 1..' ~ {~... ..
..
we may choose the branch whose argument lies between Ta and awa...
f"
~
R(x)eio: dx . ~enee the residue theorem yields the value of the integral
Zk
..
. ./0'" X"'R(x)
dx
~~.. 412
It is clear that we are led to the result
lim
~o
J
.t
i:~:r:J
tbi:
!O" t%«HR(t
l)
tit.
...
For the function
f_ ......"
" • '.'..)
:::c
n •.j.":"..v.tri.
.
pr
. :
~~
'~.....
:.
(}_
I
.
for Iz I
<
As an illustration.x
(j
x!
sin x
+ at
a real. ..2itf...1..integration by parts.iY. 00. p e s..al
la'
# p.. : ..)
J0
t"
log (1
< a < 2)..and '1'".2ie"" sin :e) th "'"0
=1
If we choose log eU == ix. :: .e!ul. Show
z. n positive integers)...)s' a real. the points 0 and T have to be avoided.sin e...
+x
2 )
(h) xH·.
..becomes infinite like log 8..e2U ~ 1 .. We apply Cauchy's theorem to the rectangle whose vertices are OJ 1f'. As a final example we compute the special integral
that in Sec 5 3 Exam 13th .S y _. we find that this function is real and negative only for z == nT. however.elia = .. we must choose log ($) == ri/2.. In the region obtained by omitting these half lines the principal branch of log (1 . The equation can now be written in the form
• • ••
.
"
'.. 1 and if JrJ < 1.
dx
J0 1
r:
+
a'
2
dx a real
1
J
(~..... the imaginary part lies between 0 and 71'4' Therefore. then
f(t)
_
log (. (1 .. extended over a rightangled isosceles triangle.
'
..
e)
sin t: z'
2.e2w(cos 2x i sin 2x)..
Consider the function 1 .
1
FIC .
(b) (d) (f)
X2
X
ce
(e) (g) (i)
~ ~ o x'
cos x
+
+2
J0
(~x'
X2 (Xl
+
Zl dx
.
function...
(Try... (Suggested by a student) L ~/:aluate the following integrals by the method of resid~e8:
JOT log
sin 8 dB. ·
+
Hint: Use zZ 

p
2
to conve rt the mtegr al to a Iine integral of a rational i · e
can sometimes be used to evaluate showthat if fez) is analytic and boun:
*5. From the representa.
xl/3
+ Xl dx.tion 1 . and we do this by following small circular quadrants of radius It Because of the periodicity the integrals over the vertical sides cancel against each other. The same proof applies near the vertex 1r.... The integral over the upper horizontal side tends to 8...
COMPLEX ANALYSIS
COMPLEX
INTEGRATION
1'1
11"
log 2 
(~2) + lor i
log sin x dz
+ (.etu) is hence singlevalued and analytic... (0 dx
J0
(~
(1
+ ZZ)l log x dx.(l.
and we find
..
dx.11
FIG •• 14
0.: . in order to obtain the principal branch with an imaginary part between . (f)
•• +
(b\
'J
(Z2 _ 1)2'
1
(c)
1 sin e'
(d)
cot z. 1_1<1
/f
f(z) d:e_ ~. will tend to zero.
. 11'" + iY.') i = 0. Indeed.Ip
J
Irk I l' fz . From the fact that 11 . z)" (m. and we obtain
+
() f
C
(a)
/0
eo
a
x'
+~in2 z ' lal
+
>
d:c
1.:. since the imaginary part of the logaritlun ~ bounded we need only consider the real part.5x1
+ 6'
lOx:!
+ 9'
J
(~
(J
+ a:. Complex integration
t:
mtegrals. Y ~ 0.. . e mtegral may be..' .:
. the integrals over the quadrants can also be seen to approach sero 88 & + o.
:
"
'. Compute
I.tt)
.
o
..
. Finally.e!ul/ Iz 1~ 2 for $ + 0 we see that log 11.(
EX·EReISES
L Fin~ the poles and residues of the following functions:
(a)
(1
Z2
+ 5z + 6'
.. and. and since a log a )0 0 the integral over the quadrant near the origin.
lIZ
COMPLEX
ANALYSIS
COIIPLEX
INTEGRATION
113
I. HARMONIC
FUNCTIONS
The real and imaginary parts of an analytic function are conjugate harmonic functions, Therefore, all theorems on analytic functions are also theorems on pairs of conj ugate harmonic functions. However) harmonic functions are important in their own right, and their treatment is not always simplified by the use of complex methods, This is particularly true when the conjugate harmonic function is not singlevalued.
This, it. should be ~emembered, is the most natural way of passing from harmoruc to analytic functions. From (55) we pasa to the differential (56)
f dz = (:: d:t
+ ~ dy ) + i ( du
=
:; th
+ ::
dy )
We assemble in this section some facts about harmonic functions that are intimately connected with Cauchy's theorem. The more delicate properties of harmonic functions are postponed to a later chapter ..
6.1. Dejinition and Basic PropertielJ. A realvalued function u(z) or u(z,Y)t defined and singlevalued in a region 0, is said to 'be harmonic in Ilt or a potential Junction, if it is continuous together with its partial derivatives of the first two orders and satisfies Laplacf/B equation. (54) We shall see later that the regularity conditions can be weakened~ but this is a point of relatively minor importance. The sum of two harmonic functions and a constant multiple of a harmonic function are again harmonic j this is due to the linear character of Laplace's equation, The simplest harmonic functions are the linear functions ax by.. In polar coordinates (r,6) equation (54) takestbe form
In this expression the real part is the differential of u , dx +......____".d'iJ
au
ax
au·
By
II·
If ~ has a conjugate harmonic function v, then the imaginary written &8 dv_rJv av au au  a:z: dx + ay ay = d:t + iJx ay.
pari can be
au
In general, however, there is no singlevalued conjugate function and in ~: CIrcumstances it is better not to use the notation av, Instead we
•
and call
(57)
au =
J dz

au au ay" + axil1l
au.
We have by (56) du
*au the
conjugate differmtiol of
=
+
+ i .du.
r
!(r:)
+ :~ = O.t
+
. B~ Cauchy's theorem th~ integral of f dz vanishes along any cycle which IS h?molo~lliI to zero In f.l. On the other hand, the integral oE the exact diflerential au vanishes along all cycles. It follows by (57)tbat
(58)
This shows that log r is a harmonic function and that any harmonic function which depends only on T must be of the form a log r b.. The argument 9 is harmonic whenever it can be uniquely defined. If u is harmonic in 0, then
(55)
IS
.. _1· anwytic, for
f J7
·du = J,. 
~ tk + iJu ay
iJy ax
=
0
f(z)
·t+
= au _ ax
i au
ay
1.._.
Wl'1
mg U == au V ax'
=  au we ay
IUt.V8
for all Cf?les 'Y which are homologous to zero in n. lei The_m~gra1 in (58) has an important interpretation which cannot be . t ~e?tioned. If., is a regular curve with the equation z = z(t) the direction o~ the tangent is determined by the angle a = srg Zl (t) J .~. we can ~te dx = Idzl cos a, dy = Ilk I sin a. The normal which pointe to the right of the tangent has the direction (:J... /2 d thus cos a ==  sin fJ sin a  C08 R. The .. a...,. an
.' v
expresBlon
au· au an_ 8i
C08
fJ
au + ay sin fJ
..
t This form cannot
be used for
T
C.
..
•• ' .',
is a directional deriva.tive of u, ,the: righthand· normal tkrivGtive with l'e8peCt,~ t~ euryre 'Y. .l!~~~I~, ·d.;~;(~/4n) 1.11 and (58) can be . .~ .. .
.,. ,
:
..
"
...
.
'1:' :. I
.
..
. . .: :.: :;./~~.:~/:: ..:..:
'i~'::':\):<:···.:::~·::··/:;::~.
.....
'M
COMPLEX
ANALYSIS
COMPLEX
INTEGRATION
115
written in the form
(59)
for every cycle
r which is homologous
to zero in
D.
This is the classical notation. Its main advantage is that au/an actually represents a rate of change in the direction perpendicular to y. For Instance, if 'Y is the circle lzl == r, described in the positive sense, au/an can be replaced by the partial derivative lJu/ar. It has the disadvantage that (59) is not expressed as an ordinary line integral, but as an integral with respect to arc length. For this reason the classical notation is less natural in connection with homology theory" and we prefer to use the notation ·du. In 9. simply connected region the integral of * du vanishes over all cycles, and u has a singlevalued conjugate function v which is determined up to an additive constant. ' In the multiply connected ease the conjugate function has periods
For Ul == 1, u, :; U the formula reduces to (58). notation (60) would be written as
In the classical
6.Z. The Meanmlue Property. Let us .apply Theorem 19 with Ul == log rand 1£1 equal to & function U, harmonic in Izl < p.. For 0 we choose the punctured disk 0 < Iz1 < p, and for 'Y we take the cycle C 1  C 2 where C; is a circle Izi = r, < p described in the positive sense. On a circle JzJ = r we have *du = r(au!iJr) de and hence (60) yields
log rl f
JC
Jll
1
aau dB
r

rc, u dB = log Jr J
Tt
Ct
"1 au dB 
ar
J CI
(
u dfJ
•
an f,..l *du = f,.... ~ Idzl
corresponding to the cycles in a homology basis. There is an important generalization of (58) which deals with a pair of harmonic functions. If VI and Ut are harmonic in 0, we elaim that
(60)
In other words, the expression
is constant, and this is true even if u is only known to be harmonic in an annulus. By (58) we find in the same way that
for every cycle 'Y which is homoJogous to zero in R. According to Theorem 16, Sec. 4.6, it is sufficient to prove (60) for 7 = iJR, where R is 8. rectangle contained in 11. In R, Ul and 'Ut have singlevalued conjugate functions Vl, V2 and we can write
U,1 *dul
1;11.,
f
T~d8
is constant in the case of an annulus and zero if u is harmonic in the whole disk. Combining these results we obtain:
Theorem 20a The arithmetie m.ean of 'a'harmonic function over concentric circles lzl = r i8 a linear'function of log T,


U2:
*dul :;:::; dv2 ttl

UI tWl ;:::: 1l.1 dvs
+ Vl du,
 d(UlfJl) ..
Here d(utVl) is an exact diJJerentiw, and u1dvt + vldut is the imaginary part of (UI + wl)(dul + "dv,). The laSt J2:(z) are theorem, conclude
Theorem
(60)
(61)
;..
]·1 ....
J "dB""
a=
a: Jog r
+ p,
and changing to
&
differential can be written in the form FJ2 dz where FI(a) and analytic on R.. The integral of FJ2 dz vanishes by Cauchy'S and so does therefore the integral of its imaginary part, We that (60) holds for 'Y = aR,. and we have proved:
19. If ut and
Ut
,
and if u is
lUrmonic
in a disk
l1li
0 and tIu! arilhmetic mean is constant
.. .. n the latter case fJ 'I onlPn we find
(62)
u{O), by continuity,
new
are harmonic in a region OJ then
*dU2 'UI
. u(zo} "'" ~ fo"'Uj{Zo
+ rei') dB.
....
.~
f
~
Ul
*dUt ;::::0
,
" It is 'clear ihat~'l(62)'aotik{;&Iio',haw"beeD derived' fitjin the eorre.....
•• ' • :• _ ••••• : .... "
't' .; ...!.J. :,:: .
". :·'I.:~I.
"
'
t.'L/'~:. ·.,0;J~,~~~t~f.'',o'".!!:.'.:"I'~ ,~;.:
' .. ,..J" • ~::_..: ':~ ~ ":,':. • .. :....
.
.....
..
.
'
.
. ';' ..... :.: ./'~ .: : :.::
 : ..
'
.
"
..,'
.,'...
s,»
' ••
:.:.
1 ...
COMPLEX ANALYSIS
COMPLEX
INTEGRATION
117
sponding formula for analytic functions, Sec. 3.4,. (34).. It leads directly to the maximum principle for harmonic functions: A nonconatant harmonic junclion has neither a ~maximum nor a minimum in its region of definition. ConsequenilYJ the maximum and the minimum on a closed bounded Bet E ate taken on the boundary of E .. Theorem 21.
anypoint to the center. the closed disk IzJ ~ R, z
maps
To be explicit, suppose that u(z) is harmonic in The linear transformation
=
Set)
=
ReEf + a) R+af
to
The proof is the same as for the maximum principle of analytic functions and will not be repeated. It applies also to the minimum for the reason that u is harmonic together with u.. In the case of analytic functions the corresponding procedure would have been to apply the maximum principle to l//(z) which is illegitimate unless J(z) iJI!:. o. Observe that the maximum principle for analytic functions follows from the maximum principle for harmonic functions by applying the latter to log '/(z)l which is harmonic when I(z) F 0..
EXERCISES
function u(S(t» is harmonic in
'r1·~ 1 onto
Izl :S:
R with r
= 0 corresponding Irl ~ 1, and by (62) we
z
= a4'
The
obtain
u{a) "'" 2~
From
ItJ ... l
J
u(B{r»
d arg
r.
r = R(z
we compute
R'
 a)  i12
d arg
r ==
i ~ = i (
r
Z
1
+ R! 4 a
a _
az
_) liz = (
Z
a
+ RI liz az _) a 
d8~
is harmonic and bounded in 0 is B removable singularity in the sense that when 1£(0) is properly defined.
1. If
1J
< Izi < p, show
that the origin 'U becomes harmonic in 1z1 < p
On ~bstituting rewritten as
R2
= zj
the coefficient of dB in the last expression can be
that f(z) is analytic in the annulus Tl < [s] < T2 and continuous on the closed annulus. If M(T) denotes the maximum of If(z) I for [z I = TJ show that
2. Suppose
z za
·or, equivalently;
88
+ za  
Ri  lalt Iza 12
M(r)
s M(rl)aM(r,)la
where a = log (T2/r): log (rtJrJ (Hadamard's threecircle theorem) Discuss cases ·of equality. Hint: 4pply the maximum principle to & linear combination of log If(z) l and log I: I··
4
(zza a + za ~) + ~+ 2
!
=
Re z
ZQ
+ a.
We obtain the two forms (6) 3 of (1 11 a)  2;
J R" 
6.3. Poisson's Formula. The maximum principle has the following important consequence: If u(z) is continuous on a closed bounded set E and harmonic on the interior of E,. then it is uniquely determined by its values on the boundary of E~ Indeed, if Ul and Ut are two such functions with the same boundary values, then UI  U2 is harmonic with the boundary values 04 By the maximum and minimum principle the difference Ul  U2 must then be identically zero on E .. There arises the problem of finding u when its boundary values are given. At this point we shall solve the problem only in the simplest case, namely for a closed disk .. Formula (62) determines the value orvat the center of the disk. But this is all we need, for there exists a linear transformation which carries
Iz _
larl u(z) al:!l
'el B
1 d9 "'" 2.
J
tsJ 'I!IIIR
Re z
+ a u(z)
a
dB
Z
Poi88lJ'11,'8
formula.
In polar coordinates,
1 (h R'rt u t't" = 2;: )0 R2 _ 2,R cos (8 _ rp) (.)
+ r2 v(Re") d9.
ohlwn
In the derivation we have assumed that u(z) is harmonic in the closed ~~sk.. However, the result remains true under the weaker condition that 'U(z:~ is 'harmonic in the open disk and continuous in the closed disk. Ind~, if 0 < r < 1, then .(rz) is harmonic in the closed disk, and we
.. ';'.. '.
,': :'.(ro)
"*" ~~r!~'T;;~ w(n) "jt
.1
..
iV·~·J
.
~I
_.
.
di.:'
.~,
..
. ::. ,'. • • ", '.;. :: •• : .:). I • ( .: I
.
•
.
.
. ..
'
: _.•.•. , .•:....... .• '~"'~.'.'.~.~.: ....·,I~.·:. . • , . ): •.~.;'.~:.:
. .. ',
;r
.::»:
.
"
.: •••••••• ::.~ •• ::~
. .. .
;~
' ••
.:
.
it follows that any inequality m ~ U S M implies m :::: Pu ~ M. formula (63) gives
(65) .ei'.. Choosing R = 1we define. by U1 the corresponding function for C.. As in (65) the integral can again be written as the real part of an analytic function.
.
formula.... (. and hence IPlI.... and we conclude that (63) remains valid.lIR
la\' It: _ al2
u(z) d8
and
Pu+v = Pu
+ r.. Under this condition..s at Bo.. ."+_
: U(O) dO
lzl < R..
~. The functional is linear inasmuch as
u(a)
=
2.. ·:8inee 'U 1 is ·~ntinUOU8 and vanishes ·
'.
contintwu8 Jor
and c.. 'on· the o$hEq'. provided it is sufficiently regular..::::. _....~~ . Schwarz: Theorem 23.:. :
...' ..: 
..~\
.. Moreover. ...1"
COMPLEX
ANA. . Suppose that u(z) is harmonic for Iz\ ~ R.'
.
provided that U is continuou. Clearly...
. It follows that PU1 is zero on the open arc Cit and since it is continuous PU1(z) + 0 as z + eill E C2• In proving (68) we may suppose that U(Bo) = 0. for if this is not the case we need only replace U by U . .
1
/ R2 1. Schwarz's .!
.. We ~eduee from (67) that P. does it have the boundary values u(z) on Iz1 = R? There is reason to clarify the notations. Indeed. Because u(z) is uniformly continuous on \z\ ~ R it is true that u(rz) + u(z) uniformly for It I = R. A. by the same reasoning as before. The question of boundary values is settled by the following fundamental theorem that was first proved by H. ' . harmonic everywhere except on the closed arc Cl~ The expression Re e~' + z == 1 ~.
:
. ...
I •
~~.4.TheoreHl_ Theorem 22 serves to express 8...
:.
:
.
' ."
!'.
'.. . :. '.'. Since PU1 can be regarded as a line integral over Cl it is. note that u = 1 yields (67)
for
all
1al < R.Re~.. • '.
".. . for any piecewise continuous function U(IJ) in 0 ~ 6 ~ 2r... u(~)
== Re
[~i / ~~ : 1]"
1.' and denote by UI the function which coincides with U on C1 and vanishes on C!.z le'f 
6.. and consequently it is Or harmonic function. We have already remarked that Pu is harmonic.:_ ~
..
lz1 < R.
Jzl'
'z'
zit
vanishes on 1%1 == 1 for z :#..LYSIS
COMPLEX
INTEGRATION
111
Now all we need to do is to let r tend to 1... But the righthand side of formula (64) has a meaning as soon as u is defined on =: R.aJ < J ~ _
lim PuC')
== U(6a)
This formula is known as Schwarz's
As a special case of 1(64). = c...:..... > 0 we can find 01 and C" sueh that ei.

. JU1(B)! < 1/2.

.. together ~t~ th? linear and positive character of the functional..(z)I < for all < 1.."
"
''::
. as such it is called 8 functional.
Pt.1
:... Given .. for all I. . :.U(9o).. ::... . Observe that Pu(z) is not only a funct~on of z. let C« and C" be complementary arcs of the unit circle.. :.
It
The junction Pu(z) is harmonic lor
. . U ~ 0 implies Pu(z) ~ o· because of this property Pn is said to be a positive linear functional.tl =R
u(t)
for constant c. Pu = PU1 + Pu". given harmonic function through its values on a circle. I'.::~ :. From this property. and IU(IJ) I < e/2 for ei'E C .U
=
cPu
for all
lal <
R.. A
.
'.
The theorem leads at once to an explicit expression for the conjugate function of u. The question is..":
1/.. ."+:':'.. .. is an interior point of C.band.all this the Poisson integral of U. :..:. 10
1
t.. ~ ':"
.
(68)
The bracketed expression is RIl analytic function of z for follows that u(z} is the real part of
(66) where C is an arbitrary real constant.
. : ... Then
(64)
Pu(z)
= 2. theorem: Theorem 22. We shall formulate the result as a. ~ .. for instance piecewise continuous.. but also a function of the function U. To study the boundary behavior..'..
...
Gen
'0
mIlO
10·log sin
I dl.·= log tl + e].. CIO.
I .
. Show that the reasoning can be carried out successfully by application to u .the length of the arc.... Ii 1(2) is analytic in a neighborhood of co and if zI Re f(z) _. .e.. 1).
..Izt2 =
z)(c1'· .t~ll~.P S4' A good try would be to apply the maximum principle to u . replace each value of U(6) by the value at the point opposite to z. 3.. Given a fixed z inside the unit circle we determine for each eif the point ei'* which is such tha.'Y·.1 arg
We regard 6* as a function of 6 and differentiate..P.. .
...ei"l < Ot which is precisely what we had to prove . !by·. for we cannot immediately apply the maximum and minimum principle to 11 . .
. or by simple calculation.P.~\. If eland Ct are complementary area on the unit circlet set U ::.~
~t ~ is· renlq~JIhle.
'.
U{ 0)
:0::
1 . .> 0.
1z)/(ei'. The point at co presents an added difficulty..z ... . and use this fact to compute
.
:DO
1.
. cut off by the straight linea through z and the end points of C 1.
•
.. .. for instance U( +0) 0..lim /(1. : _'r:.1711
COMPLEX ANALYSIS
COIIPLEX
INTEGRATION
171
In other words.I).~t·1
. 88SUme that U has a jump at 0. continuous on the real axis.. can be represented 88 l\ Poisson integral (Ex. 0 when
7. Prove that B function which is harmonic and bounded in the upper half plane.' ~
...qn~.e].. to find Pu(z).~(.CI) for y _.
....S
at ei'o there exists a 8 such that .
•• <:»
. so we must have

Remark.:1 on Cl. the isolated
.. Bw:.~e. on taking absolute values.
.. In Ex. .i
and. ...~c~.i'..au for . 2.!..
~~~Y
+ /1 ~~t~~.
Zl
e..l
tends to 0 as z + O. d8· Te'" .~i7:. Im <ViZ).) exists.
(~' 
z) is negative.. .
obtain
ei~d8 ei'. There is an interesting geometric interpretation of Poisson's formula. H inl: U (66)...'· . It
represents a hsrmonic function in the upper half plane with bound&rJ" values U(~) at points of continuity (Poisson'S integral for the half plane). ~

Y t)~
+y
t
U(O df
FIG.t e"t z and e"" are in a straight line (Fig...'.'
Re fez) _..• • .fo.Q~. . show that . This almost workaJ for the function tends to 0 for y + 0 and to .zl! = d8
1and hence
If fez) is analytic in the whole plane and if cc. Assume that U(~) is piecewise continuous and bounded for all real Show that Pu(z)
1 = . Be
::
. == 0.
.. d 9* ei' ..
. Since z is constant we
1..zll"." '.~~. (70) It follows by (69) and (70) that

erslise to arbitrary jumps and to the case of the circle" .PUI(z) I < a/2 for Iz::. opposite to Ct. Show that the mean...
\z\'

=
Ie" .
~.)
. .
. value formula (62) remains valid for u.. 0 when z to 00. s. also due to Schwarz..
2~
1:\2 .J" (x .e'''1
follows that IPu(z)l < 'Pull + IPu1t < g as soon as Izl < 1 and Iz . (In other worda..... 1.Ga~. that (69) But the ratio (ei"

< o../~
. 415). r 1.
EXERCISES
~+
1.• ~~
.: ... • :.~. but we lack control when 'xi + ao.41. '.. Find Pu(z) explicitly and show that 2rPu(z) equals .... ". and take the average over the circle..~~.. It is clear geometrically. with the idea of letting I tend to o.:~ . Show that Pu(z) . show that f is a constant. U = 0 on C..
...
:

:. Consider the case of a symmetric region: 0* = o. <. and the intersection of n with the real axis by a. The proofs of these statements consist in trivial verifications.:. .
:. v(z}. &how that Jfs) is odd.:::~. The domain U can be taken to be symmetric with respect to 8 circle C rather than with respect to a straight line.fCz) is analytic and vanishes on an interval it must be identically zero. Therefore U« is a constant. Consider
Uo(z)
==
uo(z) .
• •
•• 1:
. and hence identica. It vanishes on the half circle.cipk.+
. and v . part of our theorem asserts that J(z) has an analytic con. Hint: Choose a as in (61).>'''.
. real on (1'.. The construction can be repeated for arbitrary disks. that ie.. Under these conditions we want to show that f(z) is the restriction to 0+ of a function which is analytic in all of g.:. We conclude that V is harmonic in the whole disk. ..
+
. because V tends to zero by definition and P» vanishes by obvious symmetry..that the Uo coincide in overlapping disks.:
.. Suppose that v(x) is continuous in n+ V a. Undersuch conditionaj'(e) has an analytic continuation which maps symmetric points with respe ct to C onto symmetric points with respect to C'. and aero on a.~'
: ~.5.. In other words.for we are assuming that fez) is already known to· be analytic in all of o. hut it is a rather weak result. An elementary aspect of the gymmetry principle. The maximum and minimum principle implies that V = P» in the upper half disk. Indeed. There are many more general variants first formulated by H" A. . and nothing at all need to be assumed about the real part.·~ f~C#~ J which is &ll&lytic in a SYII1metria region o can .
..Pv is harmonic in the upper half of the disk.~:~..u. let us again consider a disk with center on a. Then show that Uo is the real part of an analytic function fo{z} and use the preceding exercise to conclude that the singularity at 0 is removable. v(z) = v(z) . . and let fT be the part of the real_axi8 in ll..172
COIIPLEX
A. Assume now that fez) is analytic in 0 and real on at least one interval of the real axis. With the notation f = u + iv we have thus u(z) = u(i).. Then IJ has a haTmonic extension. 2. The Reflection Principle..•
'
<...::~:::~. S~~ :tha~.. and if fez) is an analytic function} then f(Z) is also analytic. and we conclude that j(z) == ~ in 0. and when z tends to C it may be assumed that /(z) approeehee another circle C'.i auo/ay
2
Let 0+ be Ik part in the upper half plane of a 81J1nmetric region 0.lly.. if u(z) is harmonic and j(z) analytic in 8 region then u(J) is harmonic and l(J) analytic as functions of z in the region n* obtained by reflecting n in the real axis...'.I~ . We have to show that V is harmonic on fT. if v is the imagtnary part of an analytic function I(z) in n+} then fez) has an analytic extension WAlth 8atisfie8 J(z) == j(z). to 0 which satisfie8 the symmetry relation .
.~:::~':<:..
EXERCISEI
av = 0 ax· auo/ax .
L If f(2)
.tinualion to Q.J. For a point Xo E a consider & disk with
6.. For the remaining part of the theorem. analytic in 0+.NALYSIS COMPLEX INTEGRATION 173
*a..'·/i~~1t··analytic in 0 and . Even in this formulation the assumptions are too strong.center Xo cont&ined in n. has been discussed already in connection with linear transformations (Chap. .. and equal to v(z) in the mirror image of n+.(i)t
On the real diameter it is clear that
aUo/ax
=_
= 0 and also
auo
fly
=
2 auo
iJy
It follows that the analytic function vanishes on the real axis.·<j:. .<. 0 on 6. Let us denote the intersection of ~ with the upper half plane by 0+...
..ginary part v(z) vanishes on a.'~. Because 11 is connected it must intersect the real axis along at least one open interval. This is important. continuous and. 3.. the main thing is that the ima.. Sec.3).
For the proof we construct the function V(z) which is equal to v(z) in 0+. and the same proof can he repeated for the lower half....~~'.: . v(J) =. Z E 0* if and only if. and also on the diameter... ·lftitrtea··in ··the tanif'll ~. The principle of reflection is based on the observation that if u{z) is a harmonic function. and in particular at Xo.
If u(z) is harmonic
for 0
<
IzI
<
p
and lim zu(z) = Or prove that u
can be written in the form u(z) = a log 121 uu(z) where a is a constant and Uo is harmonic in [z! < p. We have proved that u.. It is clear .e:O
In tM same Bituation. In the definitive statement of the reflection principle the emphasis should therefore be on harmonic functions. 3. hamionic in 0+. Since J(z) . has a conjugate harmonic function Ut) in the same disk which we may normalize so that Uo = Re J(z) in the upper half. More preelsely. We have already extended v to the whole disk. Schwarz.:. and this constant is evidently zero.
Theorem 24. The difference V .. The definition can be extended to all of OJ and the theorem follows..L··:·'. .i)..::.
<•
is analytic in the whole plane and real on the real axis purely imaginary on the imasinary axis..~. .<. ..(z) = ue(. by Theorem 23. ...~' ~..
'
'
. and satisfies the symmetry condition /(z) == f(z).'.i II! (1. ~~ .'. and let Pv denote the Poisson integral with respect to this disk formed with the boundary values V.. The theorem has obvious generalizations. or reflection pri11. . ... then u(z) is likewise harmonic.::..>: . Suppose that J(z) is defined on n+ U a..!'( ~~~~~..
2.174
COMPLEX
ANALYSIS
5
Izl = 1. .~. J. central theorem
.'/.. mainly for the purpose of defining the exponential and trigonometric functions... TI&eiDN.
. infinite products. ~'..>.2.
..... Without use of integration we were. but it must also be kept in mind that the whole theory originated from 8 desire to be able to manipulate explicit analytic expressions. f • .i~":·:... and other limits. ·The first subsection deals with more general properties of sequences· of analytic functions .::. partly with quite explicit representations of elementary transcendental functions and other specific functiooa.
L POWER IIERIEI EXPANSIONS
In a preliminary way we have considered power series in Chap. This question will nOw be resolved in the affirmative.
.. analytic ~ctionf(Z) = u + iv. Use (66) to derive & formula for f (z) In terms of u(z)t s.: . complete to an. :. show
SERIES AND ·PRODUCT DEVELOPMENTS
real on the real axis. [Refteet over the real axis. u(z) ~ Ky for'll> 0. ·r . not able to prove that every ADalytic function has a power series expansion. :. Such expressions take the form of infinite series... 4 to show thatf(z) is bounded.:. If J(z) is an.. essentially 88 an application of Cauchy'8 theorem . .alytic in Izi ~ 1 and satisfies 111= 1 on
that J(z) is rational.
. and use Ex.: The.)
Very general theorema have their natural place in the theory of analytic functions..:. In this chapter we deal partly with the rules that govern such limits._".. prove that u = ~ ·th 0 :S k ~ K.. enu~..1•. 1. 1£ u(z) is harmonic and 0 .
f.r) dr. In these circumstances no single function Ift(z) is defined in all of n..'
. uniformly on every compact..(z) = z/(2z" 1) and let n.b! Cauchy's theorem. The theorem is proved. J~(z) converges uniformly to J/(Z) on every compact 3Ub.t:. 2. . Suppose that .~rem 1 is due to Weierstrass.z
<r
•
Letting n tend to
II.
'... subset oj n.
Because of the uniform convergence
/.... H E < 1 we have then 2121+1 < &/2 and 11 + 2:rl > i so that (2) . It is practically evident that lim !tt(z) = z in the disk
on
for n
"y
we 0 tam.oQ n.and n is the union of the n..JJ. 0 provmg uniform eoDverge~ on a compact point set A can ~ f~ta~ by use of the maximum Prlncipie. analytu terms.
where C is the circle It . With another formulation. The theorem ean then be expressed 88· follows: If a Mn6B with...(z)
+
= ~ ( f. and the IJI!rietJ can be differ61JJ:iated term J~ I_ Th task f·" UN 1JC. :
. Moreooer...
For any given ·value of z we can make lrl < e.)flog (l/lzl). in the preceding example the sequence (!w(z)} tends to the limit function f(z) uniformly on every compact subset of the region fl... Ia
Starting from
!.
'1. be the disk lzl < 2~1/ft...:
"
'"
..:'
•. this is the natural condition in the theorem that we are going to prove. e . on 8 compact set Iz1 has a maximum r < 1 and the set is thus contained in the closed disk 1z' ~ r.: I'~ ~ ~ • . for n greater th&n a certain no~ In the general ease no will not be the same for all points of 0.(z) 
Z
= _2z~1/(2zn
+ 1).al S r. The analyticity of fez) follows most easily by use of Morera's theorem (Chap . in the most typical case the regions f4s form an increasing sequence..al = r and 11'1 al we obtain by uniform convergence J(z) = ~ ( fer) dr. The limit function J{z) must also be considered in some region D..Z.3). It follows that the convergence is uniform in any closed disk lzl :S: r < 1. 2ri l c f . Sec. the assumption implies that this disk lies in ~ for all n greater than a certain no.
and by Morera's t~eorem it follows that fez) is analytic in Iz _ e] < r Consequently f(~) IS analytic in the whole region n.
. In order to study the uniformity of the convergence we form the difference .
I.(z) dz = 0
regi. As a very simple example takeJ. then the sum.set of n. We shall find that we can frequently prove uniform convergence on every compact subset of 0.J(z) 18 UO'JU6ylsc In 0.. C·· · ..ai < r.p < r. In fact._.'
. the other hand. and by repeated applications It follows that f. · An alternative and more explicit proof is based on the integral formula
f. and therefore the convergence is uniform on every ~omplLCt subset.. we have
~d simple estimates show that the convergence is uniform for Iz ~ e] ..(z) dz = 0.f(z)dz
=
!n.(z) is defined and analytic in Dr region~. b
>n
o..(.
Theorem 1. . 01C O2 c·· .....Z)I the same reasoning yields lim f!(z) =
..
. t If ~ is any closed curve contained in \z ... Its ~pplication to series whose terms are analytic functions is partieularly nnportant.
. ( 2ri}c
(t 
f(f)
dr z)!
= 1'(z)
...'U~/1il"orml 11 on.tida' tna:xi~ ita
't
cmwer
. ~".
. The disk iB compact and hence has a finite suboovering. 4. _ . the
BUb8et oj (I.tfoo ·~'. (
2ri}crz
lz1 < 1 which we choose as our region 0.c u. region 0.!>(z) converges uniformly to fCi) (e) on every compact subset of O. and for this reason it would not make sense to require that the convergence be uniform in D.
and this formula shows that f{z) is analytic in the disk. This me&DAthat it is oon~ed in a fixed 0. Any ~ompact subset of n can be covered by a finite number of such closed disks. . This is the typical situation.. 6V8rIJ
. ~em.)
I.
t In fact.171
COMPLEX
ANALYSIS
SERIES AND PRODUCT
DEVELOPMENTS
177
We are considering a sequence tf.... (z) is analytic in the refrion D."./4 by taking n > log (4/e.(z) + · · ·
~pad
J
J.. the formula.
fez)
= !t(z)
+b(z) + · · · + f.at =s.
!~(z) = ~ ( faCt) df 2ri l c cr .
.(z)} where each/.
. form an open covering of ~z . Then /(z) is analytic in o..(z)} converges to a limit junction fez) in a region 0..! l/. in an equivalent formulation. r be a closed disk contained in D.rn1r. In fact with the AOti.
•.l. Let ~z . _. and clearly._f!68 .. or on any subset of such a closed disk. yet the limit f(z) may exist at all points of OJ although the convergence cannot be uniform..
.ente tf. on. In fact.~I. and thal the seqtt. if J(z) is to be defined in UJ each point of 0 must belong to all n. < I.tiaNf Of.
.zeros. for they give the number of roots of the equation f.) i8 analytsc in the repreaentGtiOn
f'egitm 0. . In particular.H(') ..
In the Iast formula C is any circle Iz . Tbeorem 2..:. if J(z) is ana.). The Taylor Series.....zo1 ~ p such that the closed disk Iz .. positive minimum on the circle \z .(z)
=~
ro !'(z) az. 2ri l o !. Prove that
4.
•
oil
!(A)(Z) + n! •
(s .:
'.'01 :::. p can be chosen arbitrarily T close to the shortest distance from Zo to the boundary of a.'
•
'..
containing
'fit
tlaM •
(1 .
Theol'elll3~ 1/ f(..:.%01 'tJ which we denote by C. where lim r. thenf(z) is either identically zero or has at moat m.... We have proved:
converges for Re z L Prove that
> IJ
and represent its derivative in series form. On the other hand.....>~:F:~:: . We show now that every analytic function can be developed in a convergent Taylor series.'
.Z 1n+1 p"(p _ 201>
Iz ~
=
e
uniformly on all compact sets ..sol ~ p is contained in o...
.'
.1
.(z) have at most m zeros in 0.I
:::
'
...
I
••
:
(
. The integral on the right is there0 fore zero..:.:. .
...
EXERCISES. ... For inetanoe.%01 < r. uniformly on C.
summation.
: ·~IJ:.NALYSIS SERIES .Zo)
+J+l(Z)(Z
.
'.'
. . .2..'a} "HI
". Since Zo was arbitrary. ..~.?<~i{ft..
S. 4J Sec. As a generalization
elude that the limit function is analytic . 
'D)
+
.
.: ... If(z) 1 has a. The following theorem is due to A. then Weierstrass's theorem applies and we can COD
of Theorem 2J prove that if the /. together with the corresponding representation of the remainder term..
:
:. < p.2z
+ 3·
.1.21.. If M denotes the maximum of 1/(2)l on C. we can write fez) = f('e) with _ 1( f.
.lytic in a region Q containing %0..(z) converge8 to !(z)t uniformly lin every compact subset oj O~ then J(z) i8 either identically zero or neva equal to 260 in 0.z)'
J{r) dr
prove that
. ..)
It I < 1. .:..z
.
and that the latter series represents an analytic function for Re z
> O.. .
that the series
r(z)
• =! . For this reason uniform convergence on the boundary of A implies uniform convergence on A.
...:. Theorem 8.
. According to this theorem.t)f(z)
== 1' .. we obtain at once the estimate I/'*l(z) (z .. ·
... L Using Taylors theorem applied to a branch of log (1

.. 2ri l fez)
+ I'
i~o)(. the theorem follows.ZO)+1
But the integrals on the left are s1I zero. Since it is also true that !~(a) + f'(.. For any point Zo E 0 there is therefore a number r > 0 such that !(z) is defined and ¢O for 0 < Iz . Hurwitz:
If the junctions /A(S) are analytic and F 0 in a region 0.
(1 + ~)'"
n
~
MI2 .
•
I.
(Develop
in a double aeries and reverse the order of
case
1.
. This is an almost immediate consequence of the finite Taylor development given in Chap. := 1.. .. and consequently I(zo) ¢ 0 by the same interpretation of the integral...
.. We conclude that the remainder term tends uniformly to zero in every diRk I.171
COIIPLEX
A.{z) ::::: inside of C.AND PRODUCT DEVELOPMENTS
171
mum in A on the boundary of A. Show
+ z/n).... 3.o)"H(r .... if the functions 18(2) are analytic in the disk lal < It and if it can be shown that the sequence converges uniformly on each circle lzl == r..
'.·.:":
.
•
.
"
. Suppose that f(z) is not identically zero.:.lim
2.. . and if f". = It follows that l/!ft(z) converges uniformly to 1//(z) on C. The zeros of J{z) are in any isolated. we may conclude that for
.
.2rt) c <r
lim ~ ( !!(z) dz ~.
we must first of all choose a welldefined branch.j "
.4 67
t
1 Zl
1· 3 z 15
1~ · 5 z 7 3
+
• ••
+ z)1l = 1 + I'Z + (~)z! + · · · + (:) r + .
where the .
H the logarithmic series had a radius of convergence greater t..+1]. There are simple techniques which allow us to compute.2 + 3 .~_:.21 + 41 .
where the binomial coefficients are defined by
(n
PO) =
jJ(p.: I:. and hence it is exactly 1. and secondly we have to choose a center Zo #. If ~ is the branch with a positive real part..4 + "5 .>:~:.eoeflleients are uniquely determined and equal to the Taylor coeffioients of fez). .. We gave earlier a direct proof that power series can be differentiated term by term.
:'. every convergent power series is its own Taylor series.:it . 
1) · · ~ ~ 
1 ·2 . between r/2 and Tr/2. log (1 + z) = z .ban 1. all the eoeffioientethat we are likely to need. Since this is not the ease.o. in order to find the first n ooefficienis of .. 6Z1
+ .. but if it is there is no guarantee that the series still representsj'(e) at all points which are simultaneously in II and in the circle of convergence. . We recall that the developments
The series developments of the cyclometrie functions are tan e and arc sin z are most easily obtained by consideration of the derived series. we have 1
~ =1
+
+
+ 2 + 2 '4 "
2
1 zt
1·3
1·3·5 + 2. . The radius of convergence cannot be greater than that of the derived series. and through integration we obtain
.. and we obtain
t= dz
. caleulate the suceessive derivatives.t .· · ·
(1
. Of course. choosing the branch which is respectively equal to 1 or 0 at the origin..(8) sueh
. For combinations of elementary functions it is mostly not possible to find a general law for the coefficients. I t may well be larger.110
COMPLEX
A. It is elementary to compute the coefficients.:.. .z?i
+ z" zl
ZOi
+ ··
II
~ +1i1+ .. Similarly. It is convenient to introduce the notation [z"] for any function which is analytic and has a zero of at least order n at the origin... · · n
n + 1).22
for any path inside the unit circle. ~
. From the expansion
~
1
e=l+·z+21+··
cos
•
z'
Z4
+ z!
1
= 1 . part. Unless po is a positive integer.
III
. one of the derivatives will be a negative power of 1 + 2.. however. H we want to represent a fractions! power of z or log z through 8 power series.61 +
Z
z"!
Zl
···
arctsnz=za+5"'7+··
where the branch is uniquely determined as arc tan z = }o 1 +
Z1
Z7
·
smz =
31+ 5! ·7!
Zl
Z5
Z7
+ ···
served as definitions of the functions they represent. For justification we can either rely on uniform convergence or apply Theorem 1. . the radius of convergence must be exactly 1. . and all its derivatives would be bounded in Iz1 < 1.• sulitrieiJ. if the binomial series were convergent in a circle of radius > 1.
ZR
we obtain by integration
Z=
1 . the function (1 + s). Since this branch is singlevalued and analytic in 1z1 < 1J the radius of convergence is at least 1. less precisely. In order to find the first few coefficients we need not. It amounts to the same thing if we develop the function (1 z)~ or log (1 e) about the origin.~ uetermfu&. and hence unbounded._·:·poJjnomial P ..~.
•
for Izi < 1.'J With this notation any function which is analytic at the origin can be written in the form
J{z} = ao
+ alz + · · · + a. :. Zl ZB Z4 Z.. Thus the radius of convergence is precisely 1 except in the trivial case in which the binomial series reduces to a polynomial. 4.ztt + [. This is also a direct consequence of Weierstrass's theorem. with a reasonable mount of labor. arcsmz
=
z
+ 23 + 2"45 + 2..the Taylor expansio~ ...'. {z] denotes a function which "contains the factor z.NALYSIS
SERIES
AND PRODUCT
DEVELOPMENTS
181
The radius of convergence of the Taylor series is thus at least equal to the shortest distance from Za to the boundary of O.. then log (1 + z) would be bounded for lz' < 1.
.
The series represents the principal branch of arc sin z with a real. Thus.. as we have remarked before..
we can detemune a polynomial Rtt. it follows by Theorem 3 that the radius of convergence is at least equal to the smaller of the radii of convergence of the given series J(.RtIJ =:= [Z"+11. are the Taylor coefficients of /(z) . 1 5 (w
1
same notaPA(w) + [10.. For the existence of the inverse function it is neceSStVy and sufficient that g'{O) # 0. The coe:fficlen~ of R..z. f(z)g(z)
c::
Explicitly we obtain aubo
+ (aobl + B1ba)z + + (Gob. .~::ht.+ a~ + .· .(Q.{w) such that P..g(zo)t To Simplify..~UI~~ ~
.(Q.: ..51D'
1
+ 3 + 3 toY
10'
1
+ [". we set P.· + a. . Using the
jew) .. and from s = P . let us sssume that 2:0 == 0 and g(O) = o~ We can then set
with at ~ 0.+ [Ztt+l] ...R.
.
2
With an abbreviated notation we write fez) =..~:. provided that g(O) = bo ¢ o.
:
...ea + ~.(to) determines the coefficients of g1(1D)~ ...... such that P Il ~ Q. In fact.o'.112
COIIPLEX
ANALYSIS
SERIES
AND
PRODUCT
DEVELOPMENTS
113
that /(z) ._ltll~J~"~.~. we must be able to expand the inverse function of an analytic function tD "'" g(z). the expansion of I(w) must be in powers of 10 . ~:.& quotient J(z)/g(z) the same method can be applied.~'!~'{~ ..(z)
+ [ZR+l]. Then f . ~ .(10) + [WitH].Go
41W ~1D.. In the practical computation of P ilQ.::. under the assumption a1 ¢ 0 the notations [za+1J and (Wtoll] are interchangeable. a For practical purposes the development of the inverse function is found by successive substitutions..
and tha.P Al(Qa(S» + b3aiZ" + [zt*l] == P__ l(QRl(Z) aJ') + b.t any expression of the form [wlt+l] becomes a [%. and we have only to take b...
.5 + [z7J
With appro
Zl
Z5
and fI(a) tions as before we writel(w) with QIt(O) O..t1~\:. The degree of P.(Q1I(z» for powers ~ n..~~: t. + bIZ + btZ + .. [zA+11..(Q.. For instance. Hence P... ~.
In the case of .· ... ..zlt [Z"H].(z» = z + [Z. .t.. we can takePl(tD) = 'tIJ/(h If P1 is given..
to include fifth powers....". aro~nd til.bo).[w. m. : . Finally. if g(z) is developed... + [Wi])' + [tD1] .(Q.+1] and g(z) == Q. we proceed by induction. and the coefficients of the terms of degree :S. = c.f>~... Our problem is to determine a polynomial p.
+ [ztt+
1
+ [~+ll
"
.(s) does not matter.e) and g(z). It is also important that we know how to form the development of a composite fun~tion J(g(z». Here we may suppose that g(O) = 0 and we are looking for the branch of the inverse function z = r1(w) ~hich is analytie in a neighborhood of the origin and vanishes for w = O.P . it is of course not necessary to determine the terms of degree higher than n.
. == PAl + b.. .P.
.· · + + ··· == btS + btll + · · · + b.~I(Z» + P~l(QKI(Z»aJ' + b. + a16nt + · .{z) [z .4izt..+1]) + [zro+l] = P .n in P nQII are the Taylor coefficients of the prod
g{z) == atZ
+ B2Z1 + .{z» [z~11... In the last member the first two terms form a known polynomial of the form z c.z" + ..:.. I~ order to prove the existence of a polynomial P .~:~ ~~::: . .+1].:::I
priate rem. hence we assume that
+
It is then clear that f(z)g(z) == P" (z) Qtt(z) + [z+1]. to
+~~. but the expressions are too complicated to be of essential help .. and the Taylor coefficients of J(g{z» are the coefficients of P. Substituting
::I
+ + .
..· ==
Q. (g(z) [.(z) has a zero of at least order n + 1 at the origin. Thus we obtain /(g(z)) = P.1...
. .{z) + [zl\+ll 10 = g(z) we have to observe that 1) = Pa(Q.+1] we obtain z = P.
:.+ ..arz1' + [zA+l] == P~I(Q.. In this case..+lJ... continued until the remainder contains the factor Za+l. + b.~. suppose that J(z) :: 0:0 g(z) =a bo
+ alZ + B2Z2 + ..W" and obtain
+
+
+
+
.. and since g(O) ¢ 0 we find that Jla = R.::.1f!t./.. ~ ft.(z)
+ [rFl]
uct f(i)f(z). By use of ordinary long division.. it is true in any case that the coefficients of %"'.· . They can be determined explicitly in determinant form.· ~
it • I
In deriving this expansion we have not even mentioned the question of convergence..(z})
Z
1( zI +3 10 + 3" + [waJ)' . are the Taylor coefficients of f(z)/g(z). we write
z
and substitute
=W
+ 3 ..... To illustrate the method we determine the expansion of tan to from the series
p (QII(z»
II
+
+
+
1D=arotanz=z3+5~
If we want the development
Zl
zI
' •• .
. [Z~+l]. Clearly..~!D~. but since the development is identical with t~e Taylor development of f(z)g(z). ~.
". to
...'1
'
P.sinden we obtain
10
this expression in the terms to the right.....