1

COMPLEX ANALYSIS

M.A. (Previous)

Directorate of Distance Education

Maharshi Dayanand University
ROHTAK – 124 001
2

Copyright © 2003, Maharshi Dayanand University, ROHTAK

All Rights Reserved. No part of this publication may be reproduced or stored in a retrieval system or trans-
mitted in any form or by any means; electronic, mechanical, photocopying, recording or otherwise, without the
written permission of the copyright holder.

Maharshi Dayanand University

ROHTAK – 124 001

Developed & Produced by EXCEL BOOKS PVT. LTD., A-45 Naraina, Phase 1, New Delhi-110028
 3

Contents
Unit-I
1. Anaytic Functions 5
2. Complex Integration 16
Unit-II
1. Zeros of Analytic Function 41
2. Laurent's Series 42
3. Isolated Singularities 49
4. Maximum Modulus Principle 53
5. Meromorphic Functions 57
6. Calculus of Residues 63
7. Evaluation of Integrals 68
8. Multivalued Function and its Branches 76
Unit-III
1. Transformations 82
2. Conformal Mappings 91
3. Space of Analytic Functions 96
4. Factorization of Integral Function 103
5. The Gamma Function 106
6. Runge's Theorem 118
7. Mittag Leffler's Theorem 121
Unit-IV
1. Analytic Continuation 123
2. Schwarz's Reflection Principle 132
3. Monodromy Theorem and its Consequences 133
4. Harmonic Functions on a Disk 138
5. The Dirichlet's Problem 141
6. Green's Function 149
7. Canonical Product 151
Unit-V
1. Growth and Order of Entire Function 156
2. The Range of an Analytic Function 168
3. Univalent Functions 179
4

Complex Analysis

M.Marks: 100
Time: 3 Hours

Note: Question paper will consist of three sections. Section I consisting of one question with ten parts of
2 marks each covering whole of the syllabus shall be compulsory. From Section-II, 10 questions to be set
selecting two questions from each unit. The candidate will be required to attempt any seven questions
each of five marks. Section-III, five questions to be set, one from each unit. The candidate will be required
to attempt any three questions each of fifteen marks.
Unit I: Analysis functions, Cauchy-Riemann equation in cartesian and polar coordinates . Complex integration.
Cauchy-Goursat Theorem. Cauchy's integral formula. Higher order derivatives. Morera's Theorem. Cauchy's
inequality and Liouville's theorem. The fundamental theorem of algebra. Taylor's theorem.
Unit-II: Isolated singularities. Meromorphic functions. Maximum modulus principle. Schwarz lemma. Laurent's
series. The argument principle. Rouche's theorem. Inverse function theorem.
Residues. Cauchy's residue theorem. Evaluation of integrals. Branches of many valued functions with special
reference to arg z, log z and za.
Unit-III: Bilinear transformations, their properties and classifications. Definitions and examples of Conformal
mappings.
Space of analytic functions. Hurwitz's theorem. Montel's theorem. Riemann mapping theorem.
Weierstrass' factorisation theorem. Gammar function and its properties. Riemann Zeta function. Riemann's
functional equation. Runge's theorem. Mittag-Leffler's theorem.
Unit IV: Analytic Continuation. Uniqueness of direct analytic continuation. Uniqueness of analytic continuation
along a curve. Power series method of analytic continuation. Schwarz Reflection principle. Monodromy theorem
and its consequences. Harmonic functions on a disk. Harnack's inequality and theorem. Dirichlet problem.
Green's function.
Canonical products. jensen's formula. Poisson-Jensen formula. Hadamard's three circles theorem.
Unit V: Order of an entire function. Exponent of Convergence. Borel's theorem. Hadamard's factorization
theorem.
The range of an analytic function. Bloch's theorem. The Little Picard theorem. Schottky's theorem. Montel
Caratheodory and the Great picard theorem.
Univalent functions. Bieberbach's conjecture (Statement only) and the "1/4 theorem.
 COMPLEX ANALYSIS 5

UNIT – I

1. Analytic Functions
We denote the set of complex numbers by . Unless stated to the contrary, all functions will be
assumed to take their values in . It has been observed that the definitions of limit and continuity
of functions in  are analogous to those in real analysis. Continuous functions play only an
ancillary and technical role in the subject of complex analysis. Much more important are the
analytic functions which we discuss here. Loosely, analytic means differentiable.
Differentiation in  is set against the background of limits, continuity etc. To some extent the
rules for differentiation of a function of complex variable are similar to those of differentiation of
a function of real variable. Since  is merely R2 with the additional structure of addition and
multiplication of complex numbers, we can immediately transform most of the concepts of R 2
into those for the complex field .
Let us consider the complex function w = f(z) of a complex variable z. If z and w be separated
into their real and imaginary parts and written as z = x + iy, w = u + iv, then the relation w = f(z)
becomes
u + iv = f(x + iy)
From here, it is clear that u and v, in general, depend upon x and y in a certain definite manner so
that the function w = f(z) is nothing but the ordered pair of two real functions u and v of two real
variables x and y so that we may write
w = u (x, y) + i v(x, y)
If we use the polar form, then f can be written as
w = f(z) = u (r, ) + iv (r, )
1.1. Definition. A function f defined on an open set G of  is differentiable at an interior point z0
of G if the limit
f (z)  f (z 0 )
lim (1)
zz0 z  z0
exists. If z = z0 + h, h being complex, then (1) is equivalently written as
f (z 0  h )  f (z 0 )
lim (2)
h 0 h
When the limit (1) (or (2)) exists, it must be the same regardless of the way in which z
approaches z0 (or h approaches zero). The value of the limit, denoted by f (z0), is called the
derivative of f at z0. In    language, the above definition of derivative is the statement that
for every positive number  there exist a positive number  such that
f (z)  f (z 0 )
 f ' (z 0 ) <  (3)
z  z0
whenever 0 < |z  z0| < .
For the general point z, we have
f ( z  h )  f (z)
f (z) = lim
h 0 h
which may also be expressed as
6

f (z  Δz)  f (z)
f (z) = lim
Δz  0 Δz
Suppose w = f(z). We sometimes define
w = f(z + z)  f(z)
and write the derivative as
dw Δw
= lim
dz Δz0 Δz
If f is differentiable at each point of G, we say that f is differentiable on G. We observe that if f is
differentiable on G, then f (z) defines a function f  : G. If f  is continuous, then we say that f
is continuously differentiable. If f  is differentiable, then f is said to be twice differentiable.
Continuing in this manner, a differentiable function such that each successive derivative is again
differentiable, is called infinitely differentiable. It is immediate that the derivative of a constant
function is zero.
If f is differentiable at a point z0 in G, then f is continuous at z0, since
 f ( z)  f ( z 0 ) 
lim [f(z)  f(z0)] = lim   zlim (zz0)
zz0 zz0
 z  z0  z0
= f (z0). 0 = 0
i.e. lim f(z) = f(z0)
zz0

A continuous function is not necessarily differentiable. In fact differentiable functions possess

many special properties. For example, f(z) =z is obviously continuous but does not possess
derivative, since, by definition
zhz h
f (z) = lim = lim
h 0 h h  0 h
If we write h = rei, then
f (z) = lim e2i
h 0

So, if h0 along the positive real axis ( = 0), then f (z) = 1 and if h0 along the positive
imaginary axis ( = /2), then f (z) = 1. Hence f (z) is not unique and it depends on how h
approaches zero. Thus, we find the surprising result that the function f(z) =z is not
differentiable anywhere, even though it is continuous everywhere. In fact, this situation will be
seen for general complex functions unless the real and imaginary parts satisfy certain
compatibility conditions.
Similarly, f(z) = |z|2 is continuous everywhere but is differentiable only at z = 0 and the functions
|z|, Re z, Im z are all nowhere differentiable in  .
1.2. Definition. Let G be an open set in . A function f : G  is analytic (holomorphic) in G if
f(z) is differentiable at each point of G. Here, it is important to stress that the open set G is a part
of the definition.
Equivalently, a function f(z) is said to be analytic at z = z0 if f(z) is differentiable at every point
of some neighbourhood of z0. We observe that f(z) = |z a|2 is differentiable at z = a but it is not
analytic at z = a because there does not exist a neighbourhood of a in which |za|2 is
differentiable at each point of the neighbourhood.
If in a domain D of the complex plane, f(z) is analytic throughout, we sometimes say that f(z) is
regular in D to emphasize that every point of D is a point at which f(z) is analytic. Further, if f(z)
 COMPLEX ANALYSIS 7

is analytic at each point of the entire finite plane, then f(z) is called an entire function. A point
where the function fails to be analytic, is called a singular point or singularity of the function.
The set (class) of functions holomorphic in G is denoted by H(G). The usual differentiation rules
apply for analytic functions. Thus, if f, gH(G), then f + gH(G) and fgH(G), so that H(G) is a
ring. Further, superpositions of analytic functions are analytic, chain rule of differentiation
applies. Thus, if f and g are analytic on G and G1 respectively and f(G)  G1, then gof is analytic
on G and
(gof) (z) = g(f(z)) f (z) for all z in G.
1.3. Remark. The theory of analytic functions cannot be considered as a simple generalization of
calculus. To point out how vastly different the two subjects are, we shall show that every
analytic function is infinitely differentiable and also has a power series expansion about each
point of its domain. These results have no analogue in the theory of functions of real variables.
Further, in the complex variable case, there are an infinity of directions in which a variable z can
approach a point z0, at which differentiability is considered. In the real case, however, there are
only two avenues of approach (e.g. continuity of a function in real case, can be discussed in
terms of left and right continuity).
Thus, we notice that the statement that a function of a complex variable has a derivative is
stronger than the same statement about a function of a real variable.
1.4. Cauchy-Riemann Equations. Now we come to the earlier mentioned compatibility
relationship between the real and imaginary parts of a complex function which are necessarily
satisfied if the function is differentiable. These relations are known as Cauchy-Riemann
equations (CR equations). We have seen that every complex function can be expressed as
f(z) = u (x, y) + iv (x, y), where u(x, y)  u and v(x, y)  v
are real functions of two real variables x and y. We shall denote the partial derivatives
u u  2 u  2 u  2 u
, , , , by ux, uy, uxx, uyy, uxy respectively.
x y x 2 y 2 xy
1.5. Theorem. (Necessary condition for f(z) to be analytic). Let f(z) = u(x, y) + iv (x, y) be
defined on an open set G and be differentiable at z = x + iy  G, then the four first order partial
derivatives ux, uy, vx, vy exist and satisfy the Cauchy-Riemann Equations ux = vy, uy = vx .

Proof. By definition, we have

f ( z  h )  f ( z)
f (z) = lim
h 0 h
We evaluate this limit in two different ways. Let h = h1 + i h2  0, h1, h2  R First let h0
through real values of h i.e. h = h1, h2 = 0. Thus, we get
f (z  h )  f (z) f ( x  iy  h 1 )  f ( x  iy)
=
h h1
u ( x  h 1 , y)  u ( x , y) v( x  h 1 , y )  v( x , y)
= i
h1 h1
Letting h10, we obtain
u v
f (z) = ( x, y)  i (x, y) (1)
x x
Secondly, let h0 through purely imaginary h = ih2, h1 = 0. Then, we have
8

f (z  h )  f (z) f ( x  iy  ih 2 )  f ( x  iy)
=
h h2
u ( x , y  h 2 )  u ( x , y) v( x , y  h 2 )  v( x , y)
=i 
h2 h2
Letting h20, we obtain
u v
f (z) = i (x, y)  i (x, y) (2)
y y
Since f (z) exists, i.e., f(z) has unique derivative, so from (1) and (2), equating real and
imaginary parts, we get the Cauchy-Riemann equations
u v u v
 and  (3)
x y y x
i.e. ux = vy and uy = vx
1.6. Remarks. (i) We have f(z) = u + iv which gives
f u v f u v
 i ,  i
x x x y y y
From these two results, CR equations, in complex form, can be put as
f 1 f
 .
x i y
(ii) We note that unless the differential equations (3) i.e. CR equations are satisfied,
f(z) = u + iv cannot be differentiable at any point even if the four first order partial derivatives
exist.
For example, let us take
f(z) = Re z = x, z = x + iy
u u v v
Then  1,  0,  0, 0
x y x y
Thus, although the partial derivatives exist every where, CR equations are not satisfied at any
point of the complex plane. Hence the function f(z) = Re z is not differentiable at any point.
(iii) The condition of the above theorem is not sufficient. Actually, CR equations are
useful for proving non-differentiability. They are not, on their own, a sufficient condition for
differentiability. For this, as an example, we consider the function
( z ) 2 / z, z  0
f(z) =  , z = x + iy
0 , z  0
and show that f(z) is not differentiable at the origin, although CR equations are satisfied at that
point. By definition, we have
f (z)  f (0) (z) 2
f (0) = lim  lim 2
z 0 z z 0 z
2
 x  iy 
= lim  
( x , y )( 0, 0 )  x  iy 
 
 COMPLEX ANALYSIS 9

1 if z  0 along real axis


= 1 if z  0 along imaginary axis
 1 if z  0 along the line y  x

Thus f (0) is not unique and hence f(z) is not differentiable at the origin.
Now, to verify CR equations, we have
f(0) = 0  u(0, 0) = 0, v(0, 0) = 0
( z ) 2 ( z ) 3 ( x  iy) 3
Also f(z) =   2
z zz x  y2
x 3  3xy 2 y 3  3x 2 y
From here, u(x, y) = , v( x , y)  2
x y
2 2
x  y2
Therefore, at (0, 0)
ux = 1, uy = 0, vx = 0, vy = 1
Thus CR equations are satisfied at the origin.
To make CR equations as sufficient an additional condition of continuity on partial derivatives
is imposed.
1.7. Theorem. (Sufficient condition for f(z) to be analytic). Suppose that f(z) = u(x, y)
+ i v(x, y) for z = x + iy in an open set G, where u and v have continuous first order partial
derivatives and satisfy CauchyRiemann equations in G. Then f is analytic in G.

Proof. For z = x + iy, let h = h1 + ih2 G. We have

u = u(x, y)
u + u = u (x + h1, y + h2)
so that
u = u(x + h1, y + h2)  u (x, y) (1)
Similarly,
v = v(x + h1, y + h2)  v(x, y) (2)
Since ux, uy, vx, vy are continuous at the point (x, y), applying mean value theorem for functions
of two real variables, we get
u = (ux + 1) h1 + (uy + 2) h2
= (ux h1 + uy h2) + (1 h1 + 2 h2) (3)
v = (vx h1 + vy h2) + (1 h1 + 2 h2) (4)
where 1 = 1 (h1, h2), 2 = 2 (h1, h2), 1 = 1 (h1, h2)
and 2 = 2 (h1, h2)  0 as h1, h20
Thus u + i v = (ux + ivx) h1 + (uy + ivy)h2
+ (1 + i1) h1 + (2 + i2) h2
Making use of CR equations, we obtain
u + iv = (ux + ivx) h1 + (vx + iux) h2
+ (1 + i1) h1 + (2 + i2) h2
= (ux + ivx) (h1 + ih2) + (1 + i1) h1 + (2 + i2) h2
= (ux + ivx) h + (1 + i1) h1 + (2 + i2) h2.
Therefore,
f (z  h )  f (z) Δu  iΔv

h h
10

 ( iη1 )h1  (2 iη 2 )h 2 

= ux + ivx +  1  (5)
 h 
Now, we note that |h1|  |h|, |h2|  |h| so that
(1 iη1 )h1  (2 iη 2 )h 2 h h
 |1 iη1 | 1  |2 iη 2 | 2
h h h
 |1 iη1 |  |2 iη 2 |
Therefore, as h = h1 + ih20, the expression in the square bracket of (5) approaches zero.
Consequently, taking limit as h0 in (5), we obtain
f (z  h )  f (z)
f (z) = lim = ux + ivx
h 0 h
which shows that f(z) is analytic at every point of G.
The two results, those of necessary and sufficient conditions for f(z) to be analytic, can be
combined in the form of the following theorem.
1.8. Theorem. Let u and v be real-valued functions defined on a region G and suppose that u and
v have continuous first order partial derivatives. Then f : G  defined by
f(z) = u(x, y) + iv (x, y) is analytic iff u and v satisfy the Cauchy-Riemann equations.
1.9. CR Equations in Polar Co-ordinates : We know that in polar co-ords. (r, ),
x = r cos, y = r sin
y
r= x 2  y 2 , θ  tan 1  
x
u u r u θ
Now,  
x r x θ x
u  x  u 
   y 
= 
r   θ  x  y 2 
 2
 x y 
2 2

u 1 u
= cosθ  sin θ (1)
r r θ
u u r u θ
 
y r y θ y
u  y  u  x 
 
= 
r  2
 2  θ  x 2  y 2 
 x y 
u 1 u
= sin θ  cosθ (2)
r r θ
Similarly,
v v 1 v
 cosθ  sin  (3)
x r r θ
v v 1 v
and  sin θ  cos  (4)
y r r θ
u v u v
Using CR equations  with (1) and (4),   with (2) and (3), we get
x y y x
 COMPLEX ANALYSIS 11

 u 1 v   v 1 u 
   cosθ     sin  = 0 (5)
 r r θ   r r θ 
 u 1 v   v 1 u 
   sin θ     cos  = 0 (6)
 r r θ   r r θ 
Multiplying (5) by cos  and (6) by sin  and then adding, we find
u 1 v
 0
r r θ
u 1 v
i.e.  (7)
r r θ
Again, multiplying (5) by sin  and (6) by cos  and then subtracting, we have
v 1 u
 0
r r 
v 1 u
i.e.  (8)
r r θ
Equations (7) and (8) are the required CR equations in polar co-ordinates.
1.10. Remark. We can express f (z) in polar co-ords. as
u v
f (z) = i
x x
u 1 u v
= cosθ  sin θ  i cosθ
r r θ r
1 v
i sin 
r θ
u v v u
= cosθ  sin θ  i cosθ  i sin θ
r r r r
 u v   u v 
=   i  cosθ    i  i sin 
 r r   r r 
 u v 
= (cos   i sin )   i 
 r r 
 u v 
= ei   i 
 r r 
dw iθ w
i.e., e
dz r
dw i iθ w
Similarly, we get  e
dz r θ
1.11. Theorem. A real function of a complex variables either has derivative zero or the
derivative does not exist.
Proof. Suppose that f(z) is a real function of complex variable whose derivative exists at z0.
Then, by definitions
f (z 0  h )  f (z 0 )
f (z0) = lim
h 0 h
let h = h1 + ih2.
12

If we take the limit h0 along the real axis, h = h10, then f (z0) is real (since f is real). If we
take the limit h0 along the imaginary axis, h = ih20, then f (z0) becomes purely imaginary
number, where f is real. So we must have f (z0) = 0.
Further, in this case we also observe that if f(z) is analytic then, using CR equations, we
conclude that f(z) is a constant function.
1.12. Example. Show that the function f(0) = 0,
x 3 (1  i)  y 3 (1  i)
f(z) = u + iv =
x 2  y2
x 3  y3 x 3  y3
=  i
x 2  y2 x 2  y2
is continuous and that the CR equations are satisfied at the origin, yet f (0) does not exist
Solution. We have
x 3  y3 x 3  y3
u= , v 2
x y
2 2
x  y2
When z  0, u and v are rational functions of x and y with non zero denominators. It follows that
they are continuous when z  0. To test them for continuity at z = 0, we change to polars and get
u = r (cos3  sin3), u = r(cos3 + sin3)
each of which tends to zero as r0, whatever value  may have. Now, the actual values of u and
v at origin are zero since f(0) = 0. So the actual and the limiting values of u and v at the origin
are equal, they are continuous there. Hence f(z) is a continuous function for all values of z. Now,
at the origin
u u ( x,0)  (0,0)
 lim
x x0 x
x3 / x2
= lim 1
x 0 x
u v v
Similarly,  1,  1,  1
y x y
Hence CR equations are satisfied at the origin.
f (z)  f (0)
Again, f (0) = lim
z0 z
( x 3  y 3 )  i( x 3  y 3 ) 1
= lim .
z 0 x 2  y2 x  iy
i
If we let z0 along real axis (y = 0), then f (0) = 1 + i and if z0 along y = x, then f (0) =
1 i
Thus f (0) is not unique and hence f(z) is not differentiable at the origin.
Similar conclusion (as for example 1.12) holds for the following two functions
 I m (z 2 )
 , z0
(i) f(z) = u + iv =  | z |2
0 , z0

 COMPLEX ANALYSIS 13

 z5
 , z0
(ii) f(z) = u + iv = | z |4
0, z0

1.13. Example. Real and imaginary parts of an analytic function satisfy Laplace equation.
Solution. Let f(z) = u + iv be an analytic function so that CR equations ux = vy, uy = vx are
satisfied. Differentiating first equation w.r.t. x and second w.r.t. y and adding, we get
 2u  2u  2v 2v
  
x 2 y 2 xy yx
where continuity of partial derivatives implies that the mixed derivatives are equal i.e. vxy = vyx.
Hence, we get
2u = 0
Similarly, differentiating first equation w.r.t y and second w.r.t x and then subtracting, we find
 2u  2u  2v  2v
  
yx xy x 2 y 2
i.e. 2v = 0
1.14. Definition. A real valued function (x, y) of real variables x and y is said to be harmonic
on a domain D  , if for all points (x, y) in D, it satisfies the Laplace equation in two variables.
Thus, from the above example 1.13, we observe that u and v are harmonic functions. In such a
case, u and v are called conjugate harmonic functions i.e. u is referred to as the harmonic
conjugate of v and vice-versa where f(z) = u + iv is analytic. Harmonic functions play a part in
both physics and mathematics.
1.15. Remarks. (i) CR equations, in polar form, are
1 1
ur = v, vr =  u
r r
Differentiating first equation w.r.t r and second w.r.t , we get
1
vr = ur + r ur r , vr =  u
r
Thus, using the continuity of second order partial derivatives, we get
1
ur + r urr =  u
r
1 1
i.e. urr + ur + 2 u = 0 which is the polar form of Laplace equation.
r r
(ii) The function u (or v) can be obtained from v (or u) via CR equations. Thus, we can
obtain an analytic function f(z) = u + iv if either u or v is given. For this we use
zz zz
x= , y where z = x + iy
2 2i
Suppose u is given.
u u
We denote by (x, y), by (x, y)
x y
u v u u
Therefore, f (z) = i  i
x x x y
= (x, y)  i(x, y)
14

= (z, 0)  i (z, 0)
Where we have set x = z, y = 0
Then, f(z) =  [(z, 0)  i(z, 0)] dz + c, c being a constant.
Similarly, if v is given, we can find f(z).
1.16. Power series. An infinite series of the form
 
(i)  an zn or (2)  an (zz0)n
n 0 n 0

where an, z, z0 are in general complex, is called a power series. Since the series (2) can be
transformed into the series (1) by means of change of origin, it is sufficient to consider only the
series of type (1).

The circle |z| = R which includes all the values of z for which the power series  an zn
n 0
converges, is called the circle of convergence and the radius R of this circle is called the radius
of convergence of the series. Thus, the series converges for |z| < R and diverges for |z| > R,
nothing is claimed about the convergence on the circle.
The radius of convergence R of a power series, using ratio test or Cauchy‟s root test, is given by
the formula
1
an 
R = lim = lim | a n | n
n  a n 1 n 

1
1 a
or = lim n 1 = lim | a n | n
R n  a n n 

The number R is unique and R =  is allowed, in that case the series converges for arbitrarily
large |z|.
 
The given power series  an zn and the derived series  n an zn1 (obtained by differentiating
n 0 n 1
1
the given series) have the same radius of convergence due to the fact that lim n n  1 .
n 

1.17. Remark. Our interest in power series is in their behaviour as functions. The power series

can be used to give examples of analytic functions. A power series  an zn with non-zero radius
n 0

of convergence R, converges for |z|< R, and so we can define a function f by f(z) =  an zn
n 0
(|z| < R). The function f(z) is called the sum function of the power series.
1.18. Theorem. A power series represents an analytic function inside its circle of convergence.

Proof. Let the radius of convergence of the power series  an zn be R and let
n 0
 
f(z) =  an zn, (z) =  n an zn1.
n 0 n 1

The radius of convergence of the second series is also R. Suppose that z is any point within the
circle of convergence so that |z| < R. Then there exists a positive number r such that |z| < r < R.
For convenience, we write |z| =  , |h| = . Then  < R. Also h may be so chosen that  +  < r.
 COMPLEX ANALYSIS 15

Since  an zn is convergent in |z| < R,  an rn is bounded for 0 < r < R so that |an rn| < M where M
is finite positive constant. Thus we have
f (z  h )  f ( z)
 (z)
h
  (z  h ) n  z n 
=  an   n z n 1 
n 0  h 
  n (n  1) n 2 
=  an  z h  ...  h n 1 
n 0  2 
  n (n  1) n 2 
  | an |  | z | | h | ... | h |n 1 
n 0  2 
 M  n ( n  1) 
<  n ρ n 2  ... n 1 
n 0 r  2 
 M 1  n ( n  1) 
n 2 2
=  n  ρ  ... n 
n 0 r   2 
 M

=  n  (ρ ) n  ρ n  nρ n 1 
n 0 r

M   ρ     ρ    ρ  
n n n
=        n   (2)
 n 0  r   r  ρ  r  
n 2
  ρ   ρ   ρ  
Now,     1    …..
n 0  r  r  r 
1
=
ρ 
1
r
r
=
r   
n 2
 ρ ρ ρ
and     1      ......
n 0  r  r r
1 r
= 
ρ r ρ
1
r
n 2 3
   
Let us write S =  n    1.  2.    3   ...
r r r r
2 3
  
Then S     2   .......
r r r
Subtracting, we get
2
 ρ ρ ρ
S 1        ...
 r r r
/ r 
= 
1 / r r  
16

ρr
or S=
( r  ρ) 2
Using the values of these sums, (2) becomes
f ( z  h )  f ( z) M r r r 
 (z)     
h   r    r   (r  ) 2 
M r
=
(r  ρ )(r  ρ) 2
which tends to zero as 0
f ( z  h )  f ( z)
Hence lim = (z)
h 0 h
It follows that f(z) has the derivative (z). Thus f(z) is differentiable so that f(z) is analytic for
|z| < R.
Again, since the radius of convergence of the derived series is also R, so (z) is also analytic in
|z|<R. Successively differentiating and applying the theorem, we see that the sum function f(z) of
a power series possesses derivatives of all orders within its circle of convergence and all these
derivatives are obtained by term by term differentiation of the series. In other words, a power
series represents an analytic function inside its circle of convergence.
2. Complex Integration
Let [a, b] be a closed interval, where a, b are real numbers. Divide [a, b] into subintervals
[a = t0, t1], [t1, t2],…, [tn1, tn = b] (1)
by inserting n1 points t1, t2,…, tn1 satisfying the inequalities
a = t0 < t1 < t2<…< tn1 < tn = b
Then the set P = {t0, t1,…, tn} is called the partition of the interval [a, b] and the greatest of the
numbers t1 t0, t2  t1,…, tn  tn1 is called the norm of the partition P. Thus the norm of the
partition P is the maximum length of the subintervals in (1).
2.1. Arcs and Curves in the Complex Plane. An arc (path) L in a region G   is a
continuous function z(t) : [a, b]G for t  [a, b] in R. The arc L, given by z(t) = x(t) + iy(t),
t  [a, b], where x(t) and y(t) are continuous functions of t, is therefore a set of all image points of
a closed interval under a continuous mapping. The arc L is said to be differentiable if z(t) exists
for all t in [a, b]. In addition to the existence of z(t), if z(t) : [a, b]  is continuous, then z(t)
is a smooth arc. In such case, we may say that L is regular and smooth. Thus a regular arc is
characterized by the property that x (t ) and y (t ) exist and are continuous over the whole range of
values of t.
We say that an arc is simple or Jordan arc if z(t1) = z(t2) only when t1 = t2 i.e. the arc does not
intersect itself. If the points corresponding to the values a and b coincide, the arc is said to be a
closed arc (closed curve). An arc is said to be piecewise continuous in [a,b] if it is continuous in
every subinterval of [a, b].
2.2. Rectifiable Arcs. Let z = x(t) + iy(t) be the equation of the Jordan arc L, the range for the
parameter t being t0  t  T.
Let z0, z1,…, zn be the points of this arc corresponding to the values t0, t1,…, tn of t, where t0 < t1
< t2 <…< tn = T. Evidently, the length of the polygonal arc obtained by joining successively z0
and z1, z1 and z2 etc by st. line segments is given by
 COMPLEX ANALYSIS 17

n
n =  |zr  zr1|
r 1
n
=  ( x r  iy r )  ( x r 1  iy r 1 )
r 1
n
=  |(xr  xr1) + i (yr  yr1)|
r 1
n
=  [(xr  xr1)2 +(yryr1)2]1/2
r 1

z2
z1

z0

If this sum n tends to a unique limit l<, as n and the maximum of the differences tr  tr1
tends to zero, we say that the arc L defined by z = x(t) + iy(t) is rectifiable and that its length is l.
In this connection, we have the following result.
“A regular arc z = x(t) + iy(t), t0  t  T is rectifiable and its length is
T 2 2 1/2
 t 0 [( x (t)) + ( y (t)) ] dt”.

2.3. Contours. Let PQ and QR to be two rectifiable arcs with only Q as common point, then the
arc PR is evidently rectifiable and its length is the sum of lengths of PQ and QR. Thus it follows
that Jordan arc which consists of a finite number of regular arcs is rectifiable, its length being the
sum of lengths of regular arcs of which it is composed. Such an arc is called contour. Thus a
contour C is continuous chain of finite number of regular arcs. i.e. a contour is a piecewise
smooth arc.
By a closed contour we shall mean a simple closed Jordan arc consisting of a finite number of
regular arcs. Clearly, every closed contour is rectifiable. Circle rectangle, ellipse etc. are
examples of closed contour.
2.4. Simply Connected Region A region D is said to be simply connected if every simple closed
contour within it encloses only points of D. In such a region every closed curve can be shrunk
(contracted) to a point without passing out of the region(Fig.1). If the region is not simply
connected, then it is called multiply connected(Fig. 2).

Simply connected region Multiply connected regions

Fig. 1 Fig. 2
18

2.5. Riemann’s Definition of Complex Integration

First, we define the integral as the limit of a sum and later on, deduce it as the operation inverse
to that of differentiation.
Let us consider a function f(z) of the complex variable z. We assume that f(z) has a definite
value at each point of a rectifiable arc L having equation
z(t) = x(t) + iy(t), t0  t  T.
We divide this arc into n smaller arcs by points z 0, z1, z2,…, zn1, zn ( = Z, say) which correspond
to the values
t0 < t1 < t2,…, <tn1 < tn (= T) of the parameter t and then form the sum
n
 =  f(r) (zr  zr1)
r 1

where r is a point of L between zr1 and zr. If this sum  tends to a unique limit I as n and
the maximum of the differences tr  tr1 tends to zero, we say that f(z) is integrable from z0 to Z
along the arc L, and we write
I =  f(z) dz
L

The direction of integration is from z0 to Z, since the points on x(t) + iy(t) describe the arc L in
this sense when t increases.
2.6. Remarks. (i) Some of the most obvious properties of real integrals extend at once to
complex integrals, for example,
 [f(z) + g(z)] dz =  f(z) dz +  g(z) dz,
L L L
 K f(z) dz = K  f(z) dz, K being constant
L L
and  f(z) dz =   f(z) dz,
L' L
where L denotes the arc L described in opposite direction.
(ii) In the above definition of the complex integral, although z0, Z play much the same
parts as the lower and upper limits in the definite integral of a function of a real variable, we do
not write
I =  zZ0 f(z) dz

This is dictated essentially by the fact that the value of I depends, in general, not only on the
initial and final points of the arc L but also on its actual form.
In special circumstances, the integral may be independent of path from z0 to Z as shown in the
following example.
2.7. Example. Using the definition of an integral as the limit of a sum, evaluate the integrals
(i)  dz (ii)  |dz| (iii)  z dz
L L L
where L is a rectifiable arc joining the points z =  and z = .
Solution. We first observe that the integrals exist since the integrand is continuous on L in each
case.
(i) By definition we have.
 COMPLEX ANALYSIS 19

n
 dz = lim  (zr  zr1) 1
L n   r 1

= lim [z1  z0 + z2  z1 +…+ zn  zn1]

n 
= lim (zn  z0) = 
n 
n
(ii)  |dz| = lim  |zr  zr1|
L n  r 1
= lim [|z1  z0| + |z2  z1| +…+ |zn  zn1|]
n 

= Arc length of L
= l (say)
n
(iii) Let I =  z dz = lim  (zr  zr1) r (1)
L n   r 1

where r is any point on the sub arc joining zr1 and zr.
Since r is arbitrary, we set r = zr and r1 = zr1 successively in (1) to find
n
I = lim  zr (zr  zr1)
n   r 1
n
I = lim  zr1 (zr  zr1)
n   r 1
Adding these two results, we get
n
2I = lim  (zr + zr1) (zr  zr1)
n   r 1
n
= lim  (z 2r  z 2r 1 ) = lim (z 2n  z 02 ) = 2  2
n   r 1 n 

1
 I = (2  2)
2
In particular, if L is closed, then  =  and thus
 dz = 0,  z dz = 0.
L L

2.8. Theorem (Integration along a regular arc). Let f(z) be continuous on the regular arc L
whose equation is z(t) = x(t) + iy(t), t0  t  T. Prove that f(z) is integrable along L and that

 f(z) dz =  t 0 F(t) [ x (t)  i y (t)]dt ,

T

L
where F(t) denotes the value of f(z) at the point of L corresponding to the parametric value t.
Proof. Let us consider the sum
n
 =  f(r) (zr  zr1)
r 1

where r is a point of L between zr1 and zr. If r is the value of the parameter t corresponding to
r, then r lies between tr1 and tr. Writing F(t) = (t) + i(t), where  and  are real, we find that
n
 =  [ (r) + i(r)] [xr  xr1) + i(yr  yr1)]
r 1
n n
=  (r) (xr  xr1) + i  (r) (yr  yr1)
r 1 r 1
20

n n
+ i   (r) (xr  xr1)   (r) (yr  yr1)
r 1 r 1
= 1 + i 2 + i 3  4 (say)
= 1  4 + i (2 + 3)

We consider these four sums separately.

By the mean value theorem of differential calculus, the first sum is
n
1 =  (r) (xr  xr1)
r 1
n
=  (r) x (r) (tr  tr1)
r 1
(f(a + h) f(a) = hf (a + h), 0    1
xrxr1 = x(tr) x(tr1)
= (trtr1) x (r))
where r his between tr1 and tr.
We first show that 1 can be made to differ by less than an arbitrary positive number, however
small, from the sum
n
1 =  (tr) x (tr) (tr  tr1)
r 1
by making the maximum of the differences tr  tr1sufficiently small.
Now, by hypothesis, the functions (t) and x (t) are continuous. As continuous functions are
necessarily bounded, there exist a positive number K such that the inequalities
|(t)|  K, | x (t)|  K
hold for t0  t  T.
Moreover, the functions are also uniformly continuous, we can, therefore, preassign an arbitrary
positive number , as small as we please, and then choose a positive number , depending on ,
such that
|(t)  (t)| < , | x (t)  x (t)| < ,
whenever |t  t| < 
Hence if the maximum of the differences tr  tr1 is less than , we have
|(r) x (r)  (tr) x (tr)|
= | (r) { x (r)  x (tr)}+ x (tr) {(r)  (tr)}|
 |(r)|.| x (r)  x (tr) |+| x (tr)|.|(r)  (tr)|
< 2K
and therefore
|1  1| < 2K (T  t0)
By the definition of the integral of a continuous function of a real variable, 1 tends to the limit
n
 t 0 (t) x (t)dt  a f ( x )dx  lim  f(xi) xi
T b
n  i 1

as n and the maximum of the differences tr  tr1 tends to zero. Since |1  1| can be made
as small as we please by taking  small enough, 1 must also tend to the same limit.
 COMPLEX ANALYSIS 21

Similarly the other terms of  tend to limits. Combining these results we find that  tends to the
limit
 t [(t) x (t)  (t) y (t)] dt
T
0

+ i  tT0 [(t) x (t) + (t) y (t)] dt

=  tT0 F(t)[ x (t) + i y (t)] dt
and so f(z) is integrable along the regular arc L.
2.9. Remark. The result of the above theorem is not merely of theoretical importance as an
existence theorem. It is also of practical use since it reduces the problem of evaluating a
complex integral to the integration of two real functions of a real variable.
More generally, it can be shown that if f(z) is continuous on a contour C, it is integrable along C,
the value of its integral being the sum of the integrals of f(z) along the regular arcs of which C is
composed.
2.10. Theorem. (Absolute value of a complex integral). If f(z) is continuous on a contour C of
length l, where it satisfies the inequality

|f(z)|  M, then  f (z)dz  M l

C
Proof. Without loss of generality, we assume that C is a regular arc.
Now, if g(t) is any complex continuous function of the real variable t, we have.
n n
 g( t r )(t r  t r 1 )   |g(tr)| (tr  tr1)
r 1 r 1
and so, on proceeding to the limit, we get
|  tT0 g( t ) dt |   tT0 | g( t ) | dt
Hence, using the result of the previous theorem, we have

 f (z)dz   F( t )[x ( t )  i y ( t )]dt

C C

  tT0 | F( t ) | | x ( t )  i y ( t ) | dt
 M  tT0 | x ( t )  i y ( t ) | dt
(f(z) = F(t) on C  |F(t)|  M)
dz
T
= M t | | dt
0 dt

= M  tT0 |dz| = M l.
2.11. Remarks. (i) The result of the above theorem (2.10) is also called estimate of the integral.
(ii) So for we had assumed that f(z) is only continuous on the regular arc L along which
we take its integral. We now impose the restriction that f(z) is analytic and suppose further that
L lies entirely within the simply connected domain D within which f(z) is regular. Then
 f(z) dz certainly exists, since f(z) is necessarily continuous on L. But we are now in a position
L
to infer much more about this integral i.e. the integral is independent of path of integration. An
equivalent form of this result is Cauchy theorem - the keystone in the theory of analytic
functions.
22

2.12. Cauchy Theorem (Elementary Form). First we consider the elementary form of Cauchy
theorem which requires the additional assumption that the derivative of f(z) is continuous. This
form of Cauchy theorem is also known as Cauchy fundamental theorem, which has the following
statement.
If f(z) is analytic function whose derivative f (z) exists and is continuous at each point within
and on a closed contour C, then
 f(z) dz = 0
C
Proof. Let D denotes the closed region which consists of all points within and on C. If we write
z = x + iy, f(z) = u + iv, then we have
 f(z)dz =  (u + iv) (dx + idy)
C C
=  (u dx vdy) + i  (v dx + udy) (1)
C C
P Q
Now, we use the Green‟s theorem for a plane which states that if P(x, y), Q(x, y), , are
y x
continuous functions within a domain D and if C is any closed contour in D, then
 Q P 
 (P dx + Qdy) =     dx dy (2)
C D  x y 
By hypothesis f (z) exists and is continuous in D, so u and v and their partial derivatives ux, vx,
uy, vy are continuous functions of x and y in D. Thus the conditions of Green‟s theorem are
satisfied. Hence applying this theorem in (1), we obtain
 v u   u v 
 f(z)dz =      dx dy + i     dx dy
C D  x y  D  x y 
 u u   u u 
=    dx dy  i    dx dy
D  y y  D  x x 
(using CR equations)
=0+i0=0
Hence the result.
2.13. The General Form of Cauchy’s Theorem (Cauchy-Goursat Theorem). An important
step was pointed out by Goursat who showed that it is unnecessary to assume the continuity of
f (z), and that Cauchy‟s theorem is true if it is only assumed that f (z) exists at each point within
and on C. Actually, the continuity of the derivative f (z) and its differentiability are
consequences of Cauchy‟s theorem. The theorem states as follows:
If a function f(z) is analytic and one-valued within and on a simple closed contour C, then
 f(z) dz = 0
C
Proof. First of all, we observe that the integral certainly exists, since a function which is analytic
is continuous and a continuous function is integrable. For the proof of the theorem, we divide up
the region inside the closed contour C into a large number of sub-regions by a network of lines
parallel to the real and imaginary axes. Suppose that this divides the inside of C into a number of
 COMPLEX ANALYSIS 23

squares C1, C2,… CM say, and a number of irregular regions D1, D2,…, DN say, parts of whose
boundaries are parts of C (Fig. 1) .
F E

D C

Fig. 1 Fig. 2
C
A B
Then B
M N
 f(z) dz =   f(z) dz +
m 1 Cm
  f(zdz (1)
C n 1 D n

where each contour is described in positive (anti-clockwise) direction.

Consider, for example, any two adjacent squares ABCD and DCEF with common side CD
(Fig.2). The side CD is described from C to D in the first square and from D to C in the second.
Hence the two integrals along CD cancel. So all the integrals cancel except those which form
part of C itself, since these are described once only. Moreover within the integrals of R.H.S. of
(1), there are contained integral along all the parts of the contour C into which C is divided on
account of the subdivision. Thus the result (1) is true.
We now use the fact that f(z) is analytic at every point. This means that, if z0 is any point inside
or on C, then
f ( z)  f ( z 0 )
 f ' (z 0 ) < 
z  z0
provided that 0 < |z  z0| <  = (z0)
i.e. if |zz0| < , then
|f(z)  f(z0)  (zz0) f (z0)|   |z  z0| (2)
If we consider any particular region Cm or Dn in the above construction, it is evident that we can
choose its side so small that (2) is satisfied if z0 is a given point of the region, and z is any other
point. It is not, however, immediately obvious that we can choose the whole network so that the
conditions are satisfied in all the partial regions at the same time. We shall prove that this is
actually possible. i.e. “having given , we can choose the network in such a way that, in every
Cm or Dn, there is a point z0 such that (2) holds for every z in this region”. This actually, means
that the function is uniformly differentiable throughout the interior of C. We prove it by well
known process of subdivision.
Suppose that we start with a network of parallel lies at constant distance l between every
consecutive pair of lines. Some of the squares formed by these lines may each contain a point z 0
of the desired type. We leave these squares as they are. The rest we subdivide by lines midway
between the previous lines. If there still remain any parts which do not have the required
property, we subdivide them again in the same way.
Obviously, there are two distinct possibilities. The process may terminate after a finite number
of steps and then the result is obtained, or it may go on indefinitely.
In the second case, there is atleast one region which we can subdivide indefinitely without
obtaining the required result. We call this region, including its boundaries, R1. After the first
24

subdivision, we obtain a part R2 contained in R1 with the same property. Proceeding in this way,
we have an infinity of regions R1, R2,…, Rn each contained in the previous one, and in each of
which inequality (2) is impossible.
Since R1  R2  R3,…, there must be a point z0 common to all the regions Rn (n = 1, 2,…) and
since the dimensions of Rn decrease indefinitely, we can have |zz0| <  for sufficiently large n,
say n > n0 and for every z in Rn. But f(z) is analytic at z0. Hence (2) holds for this z0 in Rn if
n > n0. This contradicts the statement that in no Rn, there exists a point z0 satisfying inequality
(2). Thus the second possibility is ruled out and (2) is satisfied for every point in the region C.
Now, let us consider one of the squares Cm of side lm. In Cm, by inequality (2), we have
f(z) = f(z0) + (z  z0) f (z0) + (z),
where |(z)|   |z  z0|
Hence,  Cm f(z) dz =  Cm [f(z0) + (zz0) f (z0)] dz +  Cm (z)dz (3)
The first integral in (3) simplifies to
[f(z0)  z0 f (z0)]  Cm dz + f (z0)  Cm zdz
and therefore vanishes, since  Cm dz = 0,  Cm z dz = 0 (By definition). Also, by virtue of the
result regarding absolute value of a complex integral, we obtain.
|  Cm (z)dz| <   Cm |zz0| |dz|
<  2 lm. 4lm,
since |zz0|  2 lm for z0 inside Cm and z on Cm and the length of Cm is 4lm.
lm
z0

lm

z
In the case of any one of the irregular region Dn, the length of the contour is not greater than
uln + n, where n is the length of the curved part of the boundary. Hence
| Dn (z)dz| <  2 ln (4ln + n).
Adding all the parts, we obtain
M N
|  C f(z)dz|   | C f (z)dz |   | D f(z)dz|
m 1 m
n 1
n

M N
=  | Cm (z) dz | +  | Dn (z)dz|
m 1 n 1
<   2 4lm2 +   2 ln (4ln + n)
< 4 2  (lm2 + ln2) +  2 l  n (4)
where l denotes some constant greater than every one of the lns. Now (lm2 + ln2) is the area of a
region which just includes C and is therefore bounded. Also  n is the length of the contour C.
Hence the R.H.S. of (4) is less than a constant multiple of . But the L.H.S. is independent of ,
and  is arbitrarily small, it follows therefore that
 C f(z) dz = 0
which proves the theorem.
 COMPLEX ANALYSIS 25

2.14. Cor. Suppose f(z) is analytic in a simply connected domain D, then the integral along any
rectifiable curve in D joining any two points of D is the same i.e. it does not depend on the curve
joining the two points i.e. integral is independent of path.
Proof. Suppose the two points A(z1) and B(z2) of the simply connected domain D are joined by
the curves C1 and C2 as shown in the figure.
Then, by Cauchy‟s theorem.
 ALBMA f(z) dz = 0 D B(z2)
i.e.  ALB f (z)dz   BMA f (z)dz  0
L M
i.e.  ALB f (z)dz   AMB f (z)dz  0
i.e. C1 f (z)dz  C2 f (z)dz . A(z1)

2.15. Extension of Cauchy’s Theorem to Contours Defining Multiply Connected Regions.

By adopting a suitable convention as to the sense of integration, Cauchy‟s theorem can be
extended to the case of contours which are made up of several distinct closed contours.
Consider, for example, a function f(z) which is analytic in the multiply connected region R
bounded by the closed contour C and the two interior contours C1, C2 as well as on these
contours themselves. The complete contour C* which is the boundary of the region R is made
up of the three contours C, C1 and C2 and we adopt the convention that C* is described in the
positive sense if the region R is on the L.H.S. w.r.t. this sense of describing it. Then by Cauchy‟s
theorem
C* f(z) dz = 0
where the integral is taken round the complete contour C* in the positive sense.

a b 


c d


C

Practically, we deal with this case by drawing transversals like ab, cd and by applying Cauchy‟s
theorem for a simple closed contour abbadccda. It is found convenient in applications to
express the same result in the form
 C f (z)dz   C1 f (z)dz   C2 f (z)dz
where all the three integrals are now taken in the same (positive) sense.
An exactly similar result holds in case there are any finite number of closed contours C1, C2,…,
Cm inside a closed contour C and f(z) is analytic in the multiply connected region bounded by
them as well as on them. We then have
 C f (z)dz   C1 f (z)dz   C2 f (z)dz  ...   Cm f (z)dz.
where all the contours are described in positive sense.
26

2.16. Theorem. (Cauchy’s Integral Formula). Let f(z) be analytic inside and on a closed
contour C and let z0 be any point inside C. Then
1 f ( z)
f(z0) =  dz
2πi C z  z 0
f ( z)
Proof. We consider the function This function is analytic throughout the region bounded
z-z 0
by C except at z = z0.
Then, by 2.15, we have
f (z) f ( z)
C z  z dz   z  z dz
0 0
where  is any closed contour inside C including the point z0 as an interior point.

z0



Let us choose  to be the circle with centre z0 and radius . Since f(z) is continuous, we can take
 so small that on ,
| f(z)  f(z0) | < 
where  is any preassigned positive number.
Now,
f (z) [ f (z)  f (z 0 )]  f (z 0 )
γ dz   γ dz
z  z0 z  z0
dz f (z)  f (z 0 )
= f(z0)  dz    dz (1)
z  z0 z  z0
For any point z on ,
z  z0 =  ei  dz =  i ei d
i
dz 2   e d 2
  z  z  0  ei  0 id  2i
0
and
f (z)  f (z 0 ) [ f (z)  f (z 0 )] iθ
γ dz   02 π ρ e idθ
z  z0 ρ e iθ
=  02 π [ f (z)  f (z 0 )]idθ
<  20 π d = 2
Hence from (1), we get
f (z)
C dz  2πi f (z 0 ) < 2
z  z0
 COMPLEX ANALYSIS 27

Since  is arbitrarily small and L.H.S. is independent of , it follows that

f ( z)
C dz  2πif (z 0 )  0
z  z0
1 f (z)
i.e. f(z0) =  dz.
2πi C z  z 0
which proves the result.
2.17. Cor. (Extension of Cauchy’s Integral Formula to Multiply Connected Regions): If f(z)
is analytic in a ring shaped region bounded by two closed contours C 1 and C2 and z0 is a point in
the region between C1 and C2, then
1 f ( z) 1 f (z)
f(z0) = C 2 dz   C1 dz .
2πi z  z0 2πi z  z0
where C2 is the outer contour.
Proof. Describe a circle  of radius  about the point z0 such that the circle lies in the ring shaped
f (z)
region. The function is analytic in the region bounded by three close contours C1, C2
z  z0
and .

z0



C1

C2

Thus by 2.15, we have.

f (z) f (z) f (z)
C 2z  z0
dz  C
1 z  z0
dz  
z  z0
dz

where the integral along each contour is taken in positive sense. Now, using Cauchy‟s integral
formula, we find.
f (z) f (z)
C 2 z  z dz  C1 z  z dz  2i f (z 0 )
0 0
or
1 f (z) 1 f (z)
f (z 0 )   dz   dz .
2πi C 2 z  z 0 2πi C1 z  z 0

2.18. Poisson’s Integral Formula. Let f (z) be analytic in the region |z|  R, then for 0 < r < R,
we have
1 2 (R 2  r 2 )f (Re i )
2 0 R 2  2Rr cos(  )  r 2
F(r ei) = d

where  is the value of  on the circle |z| = R.

Proof. Let C denote the circle |z| = R.
Let z0 = r ei ,  < r < R by any point inside C, then by Cauchy‟s integral formula,
28

1 f (z)
f(z0) = 
2i C z  z 0
dz (1)

R2
The inverse of z0 w.r.t. the circle |z| = R is and lies outside the circle, so by Cauchy‟s
z0
theorem, we have
1 f (z)
2i C
0= dz (2)
R2
z
z0
Subtracting (2) from (1), we get
R2
(z 0  )f (z)dz
1 z0
2i C
f(z0) =
R2
(z  z 0 )(z  )
z0
1 (R 2  z 0 z 0 )f (z)dz
2i C (z  z 0 )(R 2  zz 0 )
= (3)

Now, any point on the circle C is expressible as z = R ei. Also z0 = r ei , so z0 = r ei
Therefore,
R2 z0 z 0 = R2r2 (4)
(zz0) (R z z 0 ) = z R2  z2 z 0  z0 R2 + z0 z 0 z
2

= R3 ei  R2 ei2 r ei  r eiR2 + r2R ei

= R ei [R22r R cos () + r2] (5)
and
dz = R i ei d
Thus, (3) becomes
1 2 (R 2  r 2 )f (Rei )d
2 0 R 2  2Rr cos(  )  r 2
f(r ei) = (6)

which is the required result.

Formula (6) can be separated into real and imaginary parts to get (f(z) = u + iv)
1 2 (R 2  r 2 )u (R, )d
 0 R 2  2Rr cos(  )  r 2
u(r,) =

1 2 (R 2  r 2 ) v(R, )d
 0 R 2  2Rr cos(  )  r 2
v(r,) = .

2.19. Theorem (The derivative of an analytic function). Let f(z) be analytic within and on a
closed contour C and let z0 be any point inside C, then
1 f (z)
f ' (z 0 ) 
2i  C
(z  z 0 ) 2
dz

Proof. Let z0 + h be a point in the nighbourhood of z0 and inside C, (z = h). Then Cauchy‟s
Integral formula at these two points, gives
 COMPLEX ANALYSIS 29

1 f (z)
f (z 0 )   dz.
2πi C z  z 0
1 f ( z)
and f (z 0  h )  C dz
2πi z  z0  h
Subtracting the first result from second, we get
f (z 0  h )  f (z 0 ) 1 f (z)
  dz (1)
h 2 πi C (z  z 0 )(z  z 0  h )

We observe in (1) that as h0, the required result follows. We have thus only to show that we
can proceed to the limit under the integral sign. We consider the difference
f (z 0  h )  f (z 0 ) 1 f ( z)
  dz
h 2πi C (z  z 0 ) 2
1 f (z) 1 f (z)
=    dz
2πi C (z  z 0 )(z  z 0  h ) 2πi C (z  z 0 ) 2
h f (z)dz
= C (2)
2πi (z  z 0 ) 2 (z  z 0  h )
Since f(z) is analytic on C so f(z) is bounded on C. Thus |f(z)|  M on C, M being an absolute
positive constant. Let us denote the distance of z 0 from the points nearest to it on C by  and the
length of C by l. Then if |h|<,
f (z)dz Ml | h |
h C  2 (3)
(z  z 0 ) (z  z 0  h ) δ (δ | h |)
2

which is bounded and tends to zero as |h|0. Thus, taking limit as |h|0, it follows from (2)
that
f (z 0  h )  f (z 0 ) 1 f (z)
lim   dz
h 0 h 2πi C (z  z 0 ) 2
Hence f(z) is differentiable at z0 and
1 f (z)
f (z0) =  dz
2 πi C (z  z 0 ) 2
which is Cauchy‟s integral formula for f (z) at points within C.
2.20. Generalization. This result (2.19) has a very significant consequence in the fact that f (z)
is itself analytic within C. i.e. derivative of an analytic function is also analytic. To prove this, it
is enough to show that f (z) has derivative at any point z0 inside C.
Using Cauchy‟s integral formula for f (z0) and f (z0+h) with the same restriction on h as before,
we get
1  1 1 
f (z0+h)  f (z0) =  C f (z)   2
dz
2 πi  (z  z 0  h ) (z  z 0 ) 
2

f ' (z 0  h )  f ' (z 0 ) 1 f (z)(2z  2z 0  h )dz

i.e.  
h 2πi C (z  z 0 ) 2 (z  z 0  h ) 2
by means of arguments parallel to those used in the proof of Cauchy‟s formula for f (z0), we can
easily show that as |h|0, the integral on R.H.S. tends to the limit
30

2 f (z)
 dz
2 πi C (z-z 0 ) 3
Thus f (z) has a differential co-efficient at z0, given by the formula
|2 f (z)
f (z0) =  dz.
2πi C (z  z 0 ) 3
The arguments can obviously be repeated and we get the following result as a generalization. If
f(z) is analytic inside and on a closed contour C, it possesses derivatives of all orders which are
all analytic inside C. The nth derivative f n(z0) at any point z0 inside C being given by the
formula
|n f (z)dz
f n(z0) =  .
2πi C (z  z 0 ) n 1

2.21. Remark. From Cauchy integral formula, we observe a remarkable fact about an analytic
function. Its values everywhere inside a closed contour are completely determined by its values
on the boundary. In fact the values of each derivative of an analytic function are determined just
by the values of the function on the boundary.
dz
4.22. Example. Evaluate  , m  1, 2,...M where C is a single closed contour.
C (z  z 0 ) m
1
Solution. The function is analytic except at z = z0. Hence if C does not enclose z0,
(z  z 0 ) m
then by Cauchy‟s theorem, the integral is zero. If C encloses z0, then we choose a circle  of
small radius  with centre z0.

 z0

C

Thus, we get
dz dz
I=  
C (z  z 0 ) (z  z 0 ) m
m
γ

On , z  z0 =  ei, dz = e id
i

ρe iθ idθ
 I =  02 π m imθ  02 π i m+1 ei(m1) d
ρ e
2πi, if m  1
=
0 , if m  1
0, if z  z 0 is outside C

Thus I = 0, if z  z 0 is inside C, m  1
2i if z  z is inside C, m  1
 0
4.23. Example. Evaluate
 COMPLEX ANALYSIS 31

ez
 dz
C z ( z  16)
2

where C is a closed contour between the circles of radius 1 and 3, centred at origin.
Solution. The integrand is analytic except at z = 0, z = + 4 which are not points of the given
region. Therefore, by Cauchy‟s theorem, the integral vanishes.
2.24. Example. Using Cauchy‟s Integral Formula show that
e 2z 8πe 2
 dz  i
C ( z  1)
4
3
where C is the circle |z| = 3
Solution. By Cauchy‟s integral formula for derivatives, we have
|n f (z)dz
f n(z0) =  (1)
2πi C (z-z0 ) n 1
where f(z) is analytic inside and on C.
In the present case, C is |z| = 3, f(z) = e2z, z0 = 1, n = 3 and f(z) is analytic inside and on the
circle |z| = 3.
Also, f 3(1) = 8 e2
 (1) becomes
|3 e 2z
8e2 =  dz
2πi C (z  1) 4
e 2z 8πe 2
  dz  i
C (z  1) 4 3
Hence the result.
2.25. Exercise. Using Cauchy‟s integral formula, prove that
sin π z 2  cos π z 2
(i)  dz  4πi
C (z  1)(z  2)
where C is the circle |z| = 3
sin 6 z 21
(ii)  dz  πi
C ( z  π / 6)
3
16
where C is the circle |z| = 1
cos z  πi
(iii)  dz 
C z(z  4) 2
where C is the circle |z| = 1
e zt 1
(iv)  dz  (sin t  t cos t )
C ( z  1)
2 2
2
where t > 0 and C is the circle |z| = 3.
2.26. A Complex Integral as a function of its upper limit. Let f(z) be analytic in a region D
and let
F(z) =  zz f(w) dw
0
32

where z0 is any fixed point in D and the path of integration is any contour from z 0 to z lying
entirely in D. It follows from Cor. (2.14) to Cauchy‟s theorem that the value of F(z) depends on z
only and not on the particular path of integration from z 0 to z. F(z) is called the indefinite
integral of f(z). We prove below the analogue, in the theory of functions of a complex variable,
of the well known „fundamental theorem of integral calculus.‟ It asserts that the operations of
integration and differentiation are inverse operations.
2.27. Theorem. The function F(z) is analytic in D and its derivative is f(z)
Proof. Since F(z) =  zz f(w) dw
0
 F(z+h) =  zzh f(w) dw
0
Thus
F(z + h) F(z) =  zzh f ( w )dw  zz f ( w )dw
0 0

=  zz 0 f ( w )dw  zzh f ( w )dw

0
=  zzh f ( w )dw
F(z  h )  F(z) 1 zh
  z f ( w )dw
h h
where by Cauchy‟s theorem, we may suppose that integral is taken along the straight line from z
to z + h . Thus
F(z  h )  F(z) 1 1
 f (z)   zzh f ( w )dw   zzh f (z)dw
h h h
1 zh
= z [ f ( w )  f (z)]dw
h
Since f(z) is analytic so it is continuous, given  > 0, there exists a  > 0 such that
|f(w)  f(z)| <  whenever |w  z| < 
Therefore, if 0 < |h| < , we have
F(z  h )  F(z) 1 zh
 f ( z)   | f(w)  f(z)| |dw|
h |h| z
1 zh 1
<  | dw | | h |
|h| z |h|
Hence
F(z  h )  F(z)
lim  f ( z)  0
h 0 h
or F(z) = f(z)
which proves that F(z) is analytic and that its derivative is f(z).
2.28. Morera’s Theorem. (Converse of Cauchy’s theorem). If f(z) is continuous in a region D
and if the integral  f(z)dz taken round any closed contour in D vanishes, then f(z) is analytic in
D.

Proof. When the integral round a closed contour vanishes, then we know that the value of the
integral
F(z) =  zz f(w)dw
0
 COMPLEX ANALYSIS 33

is independent of path of integration joining z0 and z. Also, we have

F(z  h )  F(z) 1 z  h
 z f(w) dw
h h
and further
F(z  h )  F(z) 1
 f (z)   zzh [f(w)  f(z)] dw
z h
where we are free to assume that the path of integration is the straight line joining the points z
and z + h. Since f(z) is continuous in D, we find that (previous theorem 2.27)
F(z) = f(z)
i.e. F(z) is analytic with derivative f(z) But we have the result that derivative of an analytic
function is analytic. Thus we finally conclude that F(z) i.e. f(z) is analytic in D.
2.29. Cauchy’s Inequality (Cauchy’s Estimate). If f(z) is analytic within and on a circle C
given by |z  z0| = R and if |f(z)|  M for every z on C, then
M| n
|f n(z0)| 
Rn
Proof. Since f(z) is analytic inside C, we have by Cauchy‟s integral formula for nth derivative of
an analytic function
|n f (z)
f n(z0) =  dz.
2πi C (z  z 0 ) n 1
Since on the circle |z  z0| = R,
z  z0 = Rei, dz = Rei id
and the length of the circle is 2R, therefore
|n f ( z )dz
| f n(z0) | = 
2 π C (z-z 0 ) n 1
| n | f (z) || dz |
 
2π C | z-z 0 |n 1
| n 2 π M | Re iθ idθ | | n 2 π M
    dθ
2π 0 | Re iθ |n 1 2π 0 R n
|n M M| n
= 2π  n
2π R n
R
M| n
Hence |f n(z0)| 
Rn
2.30. Liouville’s Theorem. A function which is analytic in all finite regions of the complex
plane, and is bounded, is identically equal to a constant.
or
If an integral function f(z) is bounded for all values of z, then it is constant
or
The only bounded entire functions are the constant functions.
Proof. Let z1, z2 be arbitrary distinct points in z-plane and let C be a large circle with centre at
origin and radius R such that C encloses z1 and z2 i.e. |z1| < R, |z2| < R.
Since f(z) is bounded, there exists a positive number M such that |f(z)|  M  z.
By Cauchy‟s integral formula,
34

1 f (z)dz
f(z1) = 
2πi C z  z1
1 f (z)dz
f(z2) = 
2 πi C z  z 2
1 f (z)(z 2  z1 )
 f(z2)  f(z1) =  dz
2πi C (z  z 2 )(z  z1 )
Thus
| z 2  z1 | | f (z) || dz |
|f(z2)  f(z1)|  
2π C |z  z1| | z  z 2 |
M | z 2  z1 | | dz |
 
2π C | z  z1 || z  z 2 |
M | z 2  z1 | | dz |
  | z  z1 | | z |  | z1 |
2π C (| z |  | z1 |)(| z |  | z 2 |)
Now, on the circle C, z = R ei, |z| = R,
dz = Rei id
Therefore,
M | z 2  z1 | 2  | Re i id |
|f(z2)  f(z1)| 
2  0 (R  | z1 |)(R  | z 2 |)
M | z 2  z1 | R
= 2π
2π (R  | z1 | (R  | z 2 |)
M | z 2  z1 | 1
= .
 | z |  | z |  R
1  1  1  2 
 R  R 
which tends to zero as R.
Hence f(z2)  f(z1) = 0 i.e. f(z1) = f(z2)
But z1, z2 are arbitrary, this holds for all couples of points z 1, z2 in the z-plane, therefore
f(z) = constant.
2.31. The Fundamental Theorem of Algebra. Any polynomial
P(z) = a0 + a1 z +…+ zn zn, an  0, n  1 has at least one point z = z0 such
that P(z0) = 0 i.e. P(z) has at least one zero.
Proof. We establish the proof by contradiction.
1
If P(z) does not vanish, then the function f(z) = is analytic in the finite z-plane. Also when
P( z )
|z|, P(z) and hence f(z) is bounded in entire complex plane, including infinity.
Liouville‟s theorem then implies that f(z) and hence P(z) is a constant which violates n  1 and
thus contradicts the assumption that P(z) does not vanish. Hence it is concluded that P(z)
vanishes at some point z = z0
2.32. Remark. The above form of fundamental theorem of algebra does not tell about the
number of zeros of P(z). Another form which tells that P(z) has exactly n zeros, will be
discussed later on. Of course, here we can prove this result by using the process of algebra as
follows :
 COMPLEX ANALYSIS 35

By the fundamental theorem of algebra, proved above, P(z) has at least one zero say z = z 0 such
that P(z0) = 0
Then,
P(z)  P(z0) = a0 + a1z + a2 z2 +…+ an zn
 (a0 + a1 z0 + a2 z02 +… an z0n)
= a1 (z  z0) + a2(z2  z02) +…+ an (zn  z0n)
= (z  z0) Q(z)
where Q(z) is a polynomial of degree (n1). Applying the fundamental theorem of algebra
again, we note that Q(z) has at least one zero, say z 1 (which may be equal to z0) and so
P(z) = (zz0) (zz1) R(z), where R(z) is a polynomial of degree (n2). Continuing in this manner,
we see that P(z) has exactly n zeros.
2.33. Taylor’s Series. We have observed that a convergent complex power series defines an
analytic (holomorphic) function. Here, we discuss its converse i.e. we proceed to prove that if
f(z) is an analytic function, regular in a neighbourhood of the point z = a, it can be expanded in a
series of powers of (z a). These two results combine to demonstrate that a function is analytic in
a region iff it is locally representable by power series. The following theorem extends Taylor‟s
classical theorem in real analysis to analytic functions of a complex variable.
2.34. Taylor’s Theorem. Suppose that f(z) is analytic inside and on a closed contour C and let a
be a point inside C. Then
f ' ' (a )
f(z) = f(a) + f (a) (za) + (z  a)2 +…
|2
f n (a )
…..+ (za)n
|n
 f n (a )
= f(a) +  (za)n
n 1 | n
The infinite series is convergent if |z a| <  where  is the distance from a to the nearest point of
C. In the region |za|  1 where 1 < , the series is uniformly convergent.
δ  δ1
Proof. Let 2 = so that 0 < 1 < 2 < . Then, by hypothesis, f(z) is analytic within and on
2
the circle  defined by the equation |z  a| = 2. Let a + h be any point of the region defined by
|z a|  1.


2

1 a
a

C
36

Since a + h lies within the circle , using Cauchy‟s integral formula

1 f (z)
f(a + h) =  dz
2πi γ z  a  h
1 1
=  f ( z) dz
2πi γ  h 
(z  a )1  
 za 
 
1 f (z)  1 
  h 
= dz
2 πi γ z  a 
1
 za 
1 f (z)  h h2 hn
=  1    ...
2 πi γ z  a  z  a (z  a ) 2 (z  a ) n
h n 1   1 b n 1 
  
 1  b  1  b  b  ...  b  1  b 
2 n
 dz
(z  a ) (z  a  h ) 
n
 
1 f (z) h f (z) h2 f (z)
=  dz   dz   dz
2πi γ z  a 2πi γ (z  a ) 2
2πi γ (z  a ) 3
hn f (z) h n 1 f (z)dz
+….+  n 1
dz   n 1
2πi γ (z  a ) 2πi γ (z  a ) (z  a  h )
Using Cauchy‟s integral formulae for the derivatives of an analytic function, we get
h2
f(a + h) = f(a) + h f (a) + f (a) +…..
|2
hn n
+ f (a) + n
|n
h n 1 f (z)dz
where n = 
2πi γ (z  a) n 1 (z  a  h )
Thus
n hr
f(a + h) = f(a) +  f r (a ) + n
r 1 |r
But on account of continuity, f(z) is bounded on the circle . Thus there exists a positive constant
M such that |f(z)|  M on . Also, when |z a| = 2,
| z  a h|  |za|  |h| > 2  1
where a + h lies in the circle and |za|  1 implies | a + h a|  1 i.e. |h|  1.
Now, applying the result regarding the absolute value of a complex integral we have the
inequality
1 | f (z) || h |n 1| dz |
|n|  
| 2πi | γ | z  a |n 1| z  a  h |
M | h |n 1 | dz |
2   2 n 1 ( 2  1 )

 COMPLEX ANALYSIS 37

M | h |n 1
= 2 2
22 n 1 ( 2  1 )
n
M | h | | h |
=  
δ 2  δ1  δ 2 
Since |h|  1 < 2, it follows that as n, n0
so that we have the identity
 f n (a ) n
f(a+h) = f(a) +  h
n 1 |n
Changing over from a + h to z, we thus have the so called Taylor‟s series (expansion)
 f n (a )
f(z) = f(a) +  (z a)n
n 1 |n
So far, we have proved only the convergence of this series for all values of z such that |z a|  1.
It is however possible to prove more i.e. the uniform convergence as follows.
Since |h|  1, we have.
n
M δ1  δ1 
|n|   
δ 2  δ1  δ 2 
and we observe that the expression on the right is independent of h. Therefore, given >0, there
exists an integer N = N(), independent of h, such that |n| <  for n  N. This proves the
uniform convergence of the Taylor‟s series of f(z) in the region |z  z0|  1 < 
2.35. Remarks. (i) The above theorem is sometimes known as the Cauchy-Taylor theorem
(ii) By putting a = 0, Taylor‟s expansion reduces to
 f n (0) n
f(z) = f(0) +  z
n 1 |n
which is known as Maclaurin‟s series.
(iii) Taylor‟s series can be put as

f(z) =  an (z a)n
n 1
where
f n (a ) 1 | n f ( z)
an =  .  dz
| n | n 2 πi ( z  a ) n 1
1 f (z)
=  dz .
2πi (z  a ) n1
f n (a )
(iv) Using an = , the result of Cauchy‟s inequality (2.29) can be put as
|n
f n (a ) M| n M
|an| =  n
 n
| n | n R R
M
i.e. |an|  n
R
38

2. 36. Theorem. On the circumference of the circle of convergence of a power series, there must
be at least one singular point of the function represented by the series.
Proof. Suppose that there is no singularity on the circumference |z  a| = R of the radius of
convergence of the power series.

f(z) =  an (za)n
n 0
Then, the function f(z) will be regular in a disc |za| < R + , where  is sufficiently small

positive number. But from this it follows that the series  an(za)n must converge in the disc
n 0
|z a| < R + and this contradicts the assumption that |z a| < R is the circle of convergence.
Hence there is at least one singular point of the function


f(z) =  an (z a)n
n 0

on the circle of convergence of the power series  an (za)n .
n 0
2.37. Example. Expand log (1 + z) in a Taylor‟s series about the point z = 0 and determine the
region of convergence for the resulting series.
Solution. Let f(z) = log (1 + z)
1 1
Then f (z) = , f ' ' ( z)  
1 z (1  z) 2
………………………………
………………………………
(1) n 1| n  1
f n(z) =
(1  z) n
Hence f(0) = 0, f (0) = 1, f (0) = 1
………………………………
………………………………
f n(0) = (1)n1 |n1
 By Taylor‟s theorem,
z2
f(z) = log (1+z) = f(0) + zf (0) + f (0) +
|2
zn n
+ f (0) +…
|n
z2 zn
=0+z+ (1) +…+ (1)n1 |n1+……
| 2 | n
z 2 z3 zn
=z  .....  (1) n 1  .....
2 3 n
Now, if un denotes the nth term of the series, then
(1) n 1 z n (1) n z n 1
un = , u n 1 
n n 1
 COMPLEX ANALYSIS 39

un 1
 lim 
n  u n 1 | z |
1
Hence by D‟Alembert‟s ratio test, the series converges for  1 i.e. |z| < 1
|z|
2.38. Example. If the function f(z) is analytic when |z| < R and has the Taylor‟s expansion

 an z , show that if r < R,
n
n 0
1 2π 
 |f(rei)|2 d =  |an|2 r2n
2π 0 n 0
Hence prove that if |f(z)|  M when |z| < R,

 |an| r  M
2 2n 2
n 0

Solution. Since f(z) is analytic for |z| < R, so f(z) is analytic within and on a closed contour C
defined by |z| = r, r < R. Thus f(z) can be expanded in a Taylor‟s series within |z| = r so that

f (z) =  an zn
0

=  an rn ein, z = rei
0
 
 |f(z)| = f(z) f (z)  an rn ein  a m rm eim
2
n 0 m 0
 
=   an a m r e
m+n i(nm)
n 0 m 0

The two series for f(z) and f (z) are absolutely convergent and hence their product is uniformly
convergent for the range 0  0  2. Thus, the term by term integration is justified. So, we get
 
2π m+n 2 π
|f(z)|2 d =   an a m r  0 e
i(nm)
0 d
n 0 m 0
  0, n  m
=  an a n rn+n. 2, 0 e i ( n m) d  
n 0 2 , n  m
1 2π 
 0 |f(re )| d =  |an| r
i 2 2 2n
or (1)
2π n 0
Now, from (1), we get
 1 2π
 |an| r =
2 2n
 |f(rei)|2 d
0 2π 0
1 2π
  M2 d
2π 0
1 2
= M 2 = M2
2π
which proves the required result.
2.39. Example. If a function f(z) is analytic for all finite values of z and as |z|, |f(z)| = A|z|K,
then f(z) is a polynomial of degree  K.
Solution. Here, f(z) is analytic in the finite part of z-plane. Also, it is given that
40

lim |f(z)| = A|z|K (1)

|z|
We can assume that f(z) is analytic inside a circle C defined by |z| = R, where R is large but
finite. Hence f(z) can be expanded in a Taylor‟s series as

f(z) =  an zn (2)
0
where
f n (0) |n f ( z)
an =   dz
| n | n 2 π i C ( z  0) n 1
1 f ( z)
Thus =  dz
2πi C z n 1
1 | f (z) || dz | 1
|an|     |f(z)| |dz|
| 2πi | C | z | n 1 2 π R n 1 C
M
  | dz | , M = max. |f(z)| on C.
2 π R n 1 C
M M A | z |K
= n 1
2πR  n  (using (1))
2π R R Rn
AR K A
=  n K
Rn R
A
Thus |an|  A RKn =
R n K
which tends to zero as R, if n  K > 0
i.e. an = 0  n such that n > K .
Now, from (2), we conclude that f(z) is a polynomial of degree  K. Hence the result.
 COMPLEX ANALYSIS 41

UNIT-II


1. Zeros of Analytic function

A zero of an analytic function f(z) is the value of z such that f(z) = 0.
Suppose f(z) is analytic in a domain D and a is any point in D. Then, by Taylor‟s theorem,

f(z) can be expanded about z = a in the form

 f n (a )
f(z) =  an (z a)n an = (1)
n 0 |n
Suppose a0 = a1 = a2 =…….= am1 = 0, am  0 (2)
so that f(a) = f (a) =……= f m1(a) = 0, f m(a)  0
In this case, we say that f(z) has a zero of order m at z = a and thus (1) takes the form

f(z) =  an (z a)n
n m

=  an+m(z a)n+m
n 0

= (z a)m  an+m(z a)n
n 0

Taking  an+m (z a)n = (z) (3)
n 0
we get
f(z) = (z a)m (z) (4)
 
Now (a) =   a n  m (z  a ) n 
 n 0  z a
  
= a m   a n  m ( z  a ) n   a m
 n 1  z a
Since am  0, so (a)  0
Thus, an analytic function f(z) is said to have a zero of order m at z = a if f(z) is expressible as
f(z) = (z a)m (z)
where (z) is analytic and (a)  0.
Also, f(z) is said to have a simple zero at z = a if z = a is a zero of order one.
1.1. Theorem. Zeros are isolated points.
Proof. Let us take the analytic function f(z) which has a zero of order m at z = a. Then, by
definition, f (z) can be expressed as
f(z) = (z a)m (z), where (z) is analytic and (a)  0.
Let (a) = 2K. Since (z) is analytic in sufficiently small neighbourhood of a, if follows from
the continuity of (z) in this neighbourhood that we can choose  so small that, for |za| < ,
| (z)  (a)| < | K |
42

Hence | (z) | = |(a) + (z)  (a)|

 | (a)|  |(z)  (a) |
> | 2K |  |K|
= |K|, for |z  a | < 
and thus, since K  0, (z) does not vanish in the region | z  a | < .
Since f(z) = (z  a)m (z), it follows at once that f(z) has no zero other than a in the same region.
Thus we conclude that there exists a nbd of a in which the only zero of f(z) is the point a itself
i.e. a is an isolated zero.
The above theorem can also be stated as “Let f(z) be analytic in a domain D, then unless f(z) is
identically zero, there exists a neighbourhood of each point in D throughout which the function
has no zero except possibly at the point itself.”
From the isolated nature of zeros of an analytic function, we are able to deduce the following
remarkable result.
1.2. Theorem. If f (z) is an analytic function, regular in a domain D and if z1, z2,…,zn,… is a
sequence of zeros of f (z), having a limiting point in the interior of D, then f (z) vanishes
identically in D.
Proof. Let a be the limiting point of the sequence of zeros z1, z2,…, zn,… of f(z). Then virtue of
continuity of f (z), f(a) = 0. Again, since f(z) is regular in the domain D and a is an interior point
of D, we can expand f(z) as a power series in powers of z  a as

f(z) =  an (za)n (1)
n 1
which converges in the neighbourhood of a. Now, either f(z) is identically zero in this region on
account of the vanishing of all co-efficient an, or else there exists a first co-efficient am (say)
which does not vanish. But if the latter is the case, we have already seen that there is a
neighbourhood of which does not contain any zero of f(z) other than a itself. This contradicts the
hypothesis that a is the limiting point of the sequence of zeros z 1, z2,…,zn. We are thus led to the
conclusion that f(z) is identically zero in the circle of convergence of the series (1) .
We are now free to repeat the same reasoning, starting with any point inside this circle, as the
hypothesis now holds for any such point. In this manner by repeated employment of the same
reasoning, it can be shown that f(z) is identically zero throughout the interior of D.
1.3. Remarks. The following two results are direct consequences of the above theorem
(i) If a function is regular in a region and vanishes at all points of a subregion of the given
region, or along any arc of a continuous curve in the region, then it must be identically zero
throughout the interior of the given region.
(ii) If two functions are regular in a region, and have identical values at an infinite
number of points which have a limiting point in the region, they must be equal to each other
throughout the interior of the given region.
i.e. If two functions, which are analytic in a domain, coincide in a part of that domain, then they
coincide in the whole domain.
For this, we take f(z) = f1(z)  f2(z).
2. Laurent’s Series. Now, we discuss the functions which are analytic in a punctured disc i.e. an
open disc with centre removed. We have seen that a function f(z) which is regular in a of a point
z = a, can be expanded in a Taylor‟s series in powers of (z a) and that this power series is
convergent in any circular region with centre a, contained within the given neighbourhood
 COMPLEX ANALYSIS 43

In case, however, the function is not analytic in the neighbourhood of a point a including it, but
analytic only in a ring shaped region (sometimes called annulus) surrounding a, the expansion of
f(z) in a Taylor‟s series in powers of (z a) ceases to be valid. The question naturally arises as to
whether f(z), for values of z in the above said ring shaped region, can be expanded in powers of
(z a) at all. The following theorem answers this question.
2.1. Laurent’s Theorem. Let f(z) be analytic in the ring shaped region between two concentric
circles C and C of radii R and R (R < R) and centre a, and on the circles themselves, then
 
f(z) =  an (z a)n +  bn (z  a)n
n 0 n 1
z being any point of the annulus and
1 f ( w )dw
an =  ,
2πi C ( w  a ) n 1
1 f ( w )dw
bn = 
2πi C' ( w  a ) n 1
Proof. Since f(z) is analytic on the circles and within the annulus between the two circles, by
Cauchy‟s integral formula

R
R
R a C
r1 

1 f (w )dw 1 f (w)dw
f(z) =    (1)
2πi C w  z 2πi C' w  z
consider the identity
1 1 1
  (2)
w  z ( w  a )  (z  a )  za 
( w  a ) 1 
 w  a 
Applying the result
1 bn
= 1 + b + b2 +…+ bn1 +
1 b 1 b
n 1 n
b
=  br 
r 0 1 b
on R.H.S. of (2), we obtain
 
1 n 1  z  a   z  a  
r n
1 1
     
w  z w  a  r 0  w  a   w  a  z  a 
1
 w a 
n 1 (z  a ) r (z  a ) n 1
=  r 1
 (3)
r 0 ( w  a ) (w  a) n w  z
Interchanging z and w, we get
44

n 1
1 (w  a) r (w  a) n 1
  (4)
z  w r 0 (z  a ) r 1 (z  a ) n z  w
Equations (3) and (4) can be written as
n
f ( w ) n 1 (z  a ) r f ( w )  z  a  f ( w )
    w on C (5)
w  z r 0 ( w  a ) r 1  w a  w z
n
f ( w )  f ( w ) n 1 ( w  a ) r f ( w )  w  a  f ( w )
     w on C (6)
zw w  z r 0 (z  a ) r 1  za  zw
Let M and M be the maximum values of |f(w)| on C and C respectively. Also let |z  a| = r1.
Equations of circles C and C are |w  a| = R and |w  a| = R respectively.
From the figure, it is clear that
w a R' 
  1 if w lies on C' 
za r1 
 (7)
za
  1 if w lies on C 
r1
w a R 
The absolute value |un(z)| of general term of the series in (5) is
(z  a ) n
|un(z)| = f (w )
( w  a ) n 1

n
r1n Mr 
 n 1
M  1
R R R

similarly, the absolute value |un(z)| of general term of the series (6) is
n
(R ' ) n M'  R ' 
|un(z)|  M'   
r1
n 1
r1  r1 
Hence the series of positive terms
n
M'  R ' 
n
M  r1 
   and Σ  
R R r1  r1 
r1 R'
are both convergent as  1,  1 .
R r1
Consequently by Weierstrass M-test, both the series in (5) and (6) are uniformly convergent.
Hence term by term integration is valid. Integrating (5) and (6), we obtain
1 f ( w )dw n 1 (z  a ) r f ( w )dw
  
2πi C w  z r 0 2πi C ( w  a ) r 1
(z  a ) n f ( w )dw
+ 
2πi C ( w  a ) n ( w  z)
and
1 f ( w )dw n 1 (z  a )  r 1
    f ( w )(w  a ) dw
r
2πi C' w  z r 0 2πi C'
 COMPLEX ANALYSIS 45

1 (w  a n ) f (w )
+  dw
(z  a ) n 2πi C' zw
1 f ( w )dw
Taking ar =  ,
2πi C ( w  a ) r 1
1 r
br+1 =  (wa) f(w) dw
2πi C
and
n
1  z  a  f (w )
2i C  w  a  w  z
Un =   dw ,

n
1  w  a  f (w)
2i C'  z  a  z  w
Vn    dw

We get
1 f ( w )dw n 1
   (za)r ar + Un (8)
2πi C w  z r 0
1 f ( w )dw n 1 b r 1
   r 1
 Vn (9)
2πi C' w  z r 0 ( z  a )
Adding (8) and (9) and using (1), we get
n 1 n
f(z) =  a r (z  a ) r   b r (z  a ) r  U n  Vn (10)
r 0 r 1
n
1  z  a  f (w )
Now, |Un| =    dw
2πi C  w  a  w  z
n
1  r1  M | dw |
    ( |w z| = |(wa)  (za)|
2π C  R  R  r1
 |wa|  |za| = Rr1)
n
1  r1  M
=   2R
2  R  R  r1
n
M  r1 
  
1 1  
r R
R
r1
which tends to zero as n, since  1.
R
Thus lim Un = 0
n 
Similarly, we can get lim Vn = 0
n 
Making n in (10), we obtain
 
f(z) =  a r (z  a ) r   br (z a)r
r 0 r 1
or
 
f(z) =  a n (z  a ) n   bn (z a)n (11)
n 0 n 1
46

where
1 f ( w )dw
an = 
2i C ( w  a ) n 1
,

1 f ( w )dw
2i C' ( w  a ) n 1
bn 

which proves the theorem.

2.2. Remarks. (i) The result (ii) can be put in a more compact form as

f(z) =  an (za)n
n  
where the co-efficients are given by the single formula
1 f (w )
an =  dw
2πi γ ( w  a ) n 1
where  denotes C when n  0 and C when n < 0 since however the integrand is analytic in the
annulus R < |za| < R, we may take  to be any closed contour which passes round the ring.
(ii) The function f(z) which is expanded in Laurent‟s series is one-valued. Laurent‟s
theorem will not provide an expansion for multi-valued function.
(iii) In the particular case when f(z) is analytic inside C, all the coefficients bn are zero,
by Cauchy‟s theorem, and the series reduces to Taylor‟s series.
(iv) The series of positive powers of z  a converges, not merely in the ring, but
everywhere inside the circle C. Similarly the series of negative powers of z  a converges
everywhere outside C.

(v) The series of negative powers of z  a i.e.,  b n (z  a ) n is called the principal part
n 1

of Laurent‟s expansion, which the series of positive powers i.e.  a n (z  a ) n is called the
n 0
regular part.
(vi) There is no handy method, like that for Taylor‟s series, for finding the Laurent co-
efficients. But if we can find them by any method (generally by direct expansion), their validity
is justified due to the fact that Laurent‟s co-efficients are unique.
1
2.3. Example. Expand f(z) = in a Laurent‟s series valid for the regions.
(z  1)(z  3)
(i) | z |< 1, (ii) 1 < |z| < 3, (iii) |z| > 3, (iv) 0 < |z +1| < 2
1 1 1
Solution. f(z) = = 
(z  1)(z  3) 2(z  1) 2(z  3)
(i) for |z| < |, we have
1 1
f(z) = (z +1)1  (z + 3)1
2 2
1
1 1  z
= (z +1)1  1  
2 6  3
1 z z z 
 
2 3
1
= 1  z  z 2  z 3  ...  1         ...
2 6  3  3   3  
 COMPLEX ANALYSIS 47

1 4 13
=  z  z2
3 9 27
(ii) for |z| > 1, we have
1
1 1  1 1  1 1 1 
 1    1    ...
2(z  1) 2z  z  2z  z z 2 z 3 
1 1 1 1
=  2  3  4  ...
2z 2z 2z 2z
and for |z| < 3, we have.

1
1 1 1 1 z
=  1  
2(z  3) 6  z  6  3 
1  
 3
1 z z 2 z3
=     ...
6 18 54 162
Hence the Laurent‟s series for f(z), valid for the annulus 1 < |z| < 3, is
1 1 1 1 1 3 z 2 z3
f(z) =…….. 4  3  2     
2z 2z 2z 2z 6 18 54 162
(iii) for |z| > 3
1 1
f(z) = 
2(z  1) 2(z  3)
1 1
1  1 1  3
 1    1  
2z  z  2z  z 
1 4 13
= 2  3  4 .....
z z z
(iv) We put z + 1 = u, then 0 < | u | < 2 and we have
1 1
f(z) = 
(z  1)(z  3) u (u  2)
1
1 1  u
=  1  
 u 2u  2 
2 u 1  
 2
1 1 u u2
=     .........
2u 4 8 16
1 1 z  1 (z  1) 2
=     .....
2(z  1) 4 8 16
c 1 
 z  
2 z 
2.4. Example. Show that e   a nzn

1 2π
where an =  cos (n  c sin ) d
2π 0
48

c 1 
 z 
2 z 
Solution. The function f(z) = e is analytic except at z = 0 and z = . Hence f(z) is analytic
in the annulus r1  |z|  r2, where r1 is small and r2 is large. Therefore, f(z) can be expanded in
the Laurent‟s series in the form
 
f(z) =  a n z n   b n z n (1)
n 0 n 1
where
1 f (z)
2i C' z n 1
an = dz ,

1 f (z)
2i C' z n 1
bn  dz ,

C being any circle with centre at the origin for the shake of convenience, let us take C to be the
unit circle |z| = 1 which gives z = ei
Now,
C
1 e 2 (z  z 1 )
an = 
2i C' z n 1
dz ,

1 2  e Ci sin  ie i d
2i  0
=
e i ( n 1) 
1 2 i (C sin n)
2  0
= e d
1 2π i 2π
=  cos(c sin θ  nθ)dθ   F(θ)dθ (2)
2π 0 2π 0
where F() = sin (C sin   n).
Since F(2 ) = sin [C sin (2 ) n (2  )]
=  sin (C sin   n + 2n)
=  sin (C sin  n) = F()
2π
  0 F() d = 0
Thus, from (2), we have
1 2π
an =  cos ( C sin  n) d
2π 0
1 2π
=  cos ( n   C sin ) d
2π 0
We note that if z is replaced by z1, the function f(z) remains unaltered so that
bn = (1)n an
Hence from (1), we get
 
f(z) =  a n z n   (1) n a n z n
0 1

=  a n zn

1 2π
where an =  cos (n  C sin ) d
2π 0
2.5. Example. Prove that the function f(z) = cosh (z + z1) can be expanded in a series of the type
 COMPLEX ANALYSIS 49

 
n
 a n z   bn z
n
0 1
1 2π
in which the co-efficients of z and zn, both are given by
n
 cos n cosh (2 cos) d .
2π 0
Solution. The function f(z) = cosh (z + z1) is analytic except at z = 0 and z = . Hence f(z) is
analytic in the annulus r1  |z|  r2, where r1 is small and r2 is large. Therefore, f(z) can be
expanded in the Laurent‟s series as
 
f(z) =  a n z n   b n z n (1)
n 0 n 1
where
1 f (z)dz
2i C z n 1
an = ,

1 f (z)
bn  
2i C z n 1
dz

C being any circle with centre at the origin. We take C to be the unit circle |z| = 1 which gives z
= ei
Now,
1 2  cosh(z  z )
2i 0
an = dz
z n 1
1 2 π cosh(2 cos )e i id
= 
2π 0 e i ( n 1) 
1 2π
=  cosh (2 cos ) ein d
2π 0
1 2 i 2
2  0 2 0
= cosh(2cos)cos n d  F ()d (2)

where F() = cosh (2 cos ) sin n

We note that F(2  ) =  F()
2π
  0 F() d = 0
Thus (2) becomes
1 2π
an =  cosh (2 cos ) cos n d (3)
2π 0
It is clear that
1 2π
bn = an =  cosh (2 cos ) cos (n) d = an
2π 0
Thus, from (1), we find
  
cosh (z + z1) =  a n z n   a n z n  a 0   a n (z n  z n )
0 1 1
where an is give by (3)
3. Isolated Singularities
The point where the function ceases to be analytic is called the singularity of the function.
Suppose that a function f(z) is analytic throughout the neighbourhood of a point z = a, say for
50

| z  a| < , except at the point a itself. Then the point a is called an isolated singularity of the
function f(z). In other words, the point z = a is said to be isolated singularity of f(z) if
(i) f(z) is not analytic at z = a
(ii) there exists a deleted neighbourhood of z = a containing no other singularity.
z 1
For example, the function f(z) = has three isolated singularities at z = 1, 2, 3
(z  1)(z  2)(z  3)
respectively.
It is not difficult to construct examples of singularities which are not isolated. For instance, the
function
1
f(z) = has such a singularity at z = 0
1
sin
z
1 1
In fact, sin = 0 when = n
z z
1 1
or z= . Thus, f(z) ceases to be analytic when z = . When n, z0. Therefore in
nπ nπ
1
every neighbourhood of origin there lies an infinite number of points of the form and hence
nπ
z = 0 is not an isolated singularity.
In case of isolated singularity at z = a f(z) can be expanded in a Laurent‟s series in positive and
negative powers of z  a in the region defined by r < |z  a| < R and r may be taken as small as
we please. Thus, with suitable definitions of an and bn in this region, we have
 
f(z) =  a n (z  a ) n   bn(z a)n, 0 < |z a| < R
n 0 n 1

n
where  bn (za) is the principal part of the expansion of f(z) at the singular point z = a.
n 1
There are now three possible cases, discussed as follows.
3.1. Removable Singularity. If the principal part of f(z) at z = a contains n term i.e. bn = 0  n,
then the singularity z = a is called a removable singularity of f(z). In such a case we can make
f(z) regular when |za | < R by suitably defining its value at z = a. For example, the function
sin z
f(z) = is undefined at z = 0. Also we have
z
sin z 1  z3 z5 
 z   ....... 
z z  | 3 | 5 

2 4
z z
= 1   ....
|3 |5
sin z
Thus contains no negative powers of z.
z
If it were the case f(0)  1, then z = 0 is a removable singularity which can be removed by simply
redefining f(0) = 1. Singularities of this type are of little importance.
3.2. Pole. If the principal part of f(z) at z = a contains a finite number of terms, say m, i.e. bn = 0
 n such that n > m, then the singularity is called a pole of order m. Poles of order 1,2,3… are
 COMPLEX ANALYSIS 51

called simple, double, triple poles. The coefficient b1 is called the residue of f(z) at the pole a.
Thus, if z = a is a pole of order m of the function f(z), then f(z) has the expansion of the form
 m
f(z) =  a n (z  a ) n   b n (z  a ) n
n 0 n 1
 b1 b2 bm
=  a n (z  a ) n    ... 
n 0 z  a (z  a ) 2 (z  a ) m
1  m 1
 
= b m  b m1 (z  a )  ...  b1 (z  a )   a n (z  a ) mn 
(z  a ) m n 0 
(z)
=
(z  a ) m
where (z) is analytic for |z a| < R and (a) = bm  0
1
We can therefore find a neighbourhood |z a| <  of the pole in which |f(z)|  |bm| |za|m.
2
Hence if f(z) has a pole of order m at z = a, then | f(z)|  as za in any manner, i.e., an
analytic function cannot be bounded in the neighbourhood of a pole.
For example, the function
z2
f(z) =
(z  5) 2 (z  4) 3
has z = 5, z =  4 as poles of order two and three respectively. Moreover, if f(z) has a pole of
1
order m at a, then is regular and has a zero of order m there, since
f ( z)
1 (z  a ) m
= , where (z) is regular and does not vanish when |z  a| < .
f ( z) (z)
1
Similarly, we note that the converse is also true, i.e., if f(z) has a zero of order m at z = a,
f ( z)
has a pole of order m there.
Further, note that poles are isolated, since zeros are isolated.
3.3. Isolated Essential Singularity. If the principal part of f(z) at z = a has an infinite number of
terms, i.e., bn  0 for infinitely many values of n, then the singularity a is called isolated essential
1
singularity or essential singularity. In this case, a is evidently also a singularity of
f ( z)
1
1 1 1
For example, e z = 1 +    ...
z | 2 z 2 | 3z3

has z = 0 as an isolated essential singularity.

3.4. Example. Find the singularities of the function

c
e z a
f(z) = z
e a 1
indicating the character of each singularity.
52

c
e z a exp( c / z)  a
Solution. f(z) = 
z
 za 
ea 1 exp1   1
 a 
1
e c / z a c / z a
 z a 
=   e  1  e.e. a 
e. e z a / a  1  
1
  z  a (z  a ) 2 
=e 1  e1 
c/za
  ....... 
  
2
a | 2a
 c  c  1
2 
=  1     ...
 z  a  z  a  | 2 
  za   za 
2 
 1  e1   ...  e 2 1   ...  ........ 
  a   a  
Clearly, this expansion contains positive and negative powers of (z  a). Moreover, terms
containing negative powers of (z a) are infinite in number.
Hence by definition, z = a is an isolated essential singularity.
e c / z a
Again, f(z) = z / a
e 1
Evidently, denominator has zero of order one at
ez/a = 1 = e2ni
i.e., z = 2nia
Thus, f(z) has a pole of order one at each point z = 2nia, where n = 0, + 1, + 2….
3.5 Behaviour of an Analytic Function near an isolated Essential Singularity. As we know
that if z = a is a pole of an analytic function f(z), then | f(z)| as za in any manner. The
behaviour of an analytic function near an isolated essential singularity is of a much complicated
character. The following theorem is a precise statement of this complicated nature of f(z) near an
isolated essential singularity and this theorem is called Weierstress theorem.
3.6. Theorem. If a is an isolated essential singularity of f(z), then given positive numbers l, ,
however small, and any number K, however large, there exists a point z in the circle | z  a | < l
at which | f(z)  K < .
or
In any neighbourhood of an isolated essential singularity, an analytic function approaches any
given value arbitrarily closely.
Proof. We first observe that if l and M are any positive numbers, then there are values of z in the
circle |z  a| < l at which | f(z) | > M (1)
For, if this were not true, then we would have | f(z) |  M for | z  a| < l. If the principal part in
the Laurent expansion of f(z) about a is

 b n (z  a ) n ,
n 1
 COMPLEX ANALYSIS 53

1 f ( w )dw
where bn = 
2πi γ (w-a) -n1
and  is the circle | w  a | = r, r being sufficiently small, then
1 f (w)
| bn | =  dw
2πi γ ( w  a ) n 1
1 n 1
=  ( w  a ) f ( w )dw
2πi γ
M n 1
 r  | dw |
2
M n 1
 r 2r  Mr n
2
By the result of the absolute value of a complex integral, this holds for all n  1 and r, so that,
making r0, we find that bn = 0 for n  1.
This implies that there is no isolated essential singularity at z = a. But this contradicts the
hypothesis that a is an isolated essential singularity of f(z). Thus, the observed result (1) is true,
i.e., “in the neighbourhood of an isolated essential singularity, f(z) cannot be bounded.”
Now, let us take any finite, but arbitrary positive number K. There are now two distinct
possibilities, either f(z) K has zeros inside every circle |z  a| = l or else we can find a
sufficiently small  such that f(z)  K has no zero for | z  a | < l. In the first case, the result
follows immediately. In the second case, choosing a sufficiently small l, we have | f(z)  K |  0
in | z  a | < l, so that
1
(z) = is regular for |z  a | < l, except at a, where as we shall just see,
f (z)  K
(z) has an essential singularity. We have.
1
f(z) = K
(z)
If (z) were analytic at a, f(z) would either be analytic or have a pole at a. On the other hand if
(z) has a pole at a, f(z) would be obviously analytic there. Thus we reach at the contradiction
and therefore, (z) has an essential singularity at a. So, due to (1), given  > 0, there exists a
point z in the circle |z  a | < l such that
1
|(z) > , i.e. | f(z)  K| < 

and hence the theorem is proved.
3.7. Remark. The above theorem helps us to understand clearly the distinction between poles
and isolated essential singularities. While | f(z)| , as z tends to a pole in any manner, at an
isolated essential singularity f(z) has no unique limiting value, and it comes arbitrarily close to
any arbitrarily pre assigned value at infinity of points in every neighbourhood of the isolated
essential singularity.

4. Maximum Modulus Principle. Here, we continue the study of properties of analytic

functions. Contrary to the case of real functions, we cannot speak of maxima and minima of a
complex function f(z), since  is not an ordered field. However, it is meaningful to consider
maximum and minimum values of the modulus | f(z)| of the complex function f(z), real part of
54

f(z) and imaginary part of f(z). The following theorem known as maximum modulus principle, is
also true if f(z) is not one-valued, provided | f(z)| is one-valued.
4.1. Theorem. Let f(z) be analytic within and on a simple closed contour C. If | f(z)|  M on
C, then the inequality | f(z)| < M holds every where within C. Moreover | f(z)| = M at a point
within C if and only if f(z) is constant.
In other words, | f(z)| attains the max. value on the boundary C and not at any interior point of the
region D bounded by C.
Proof. We prove the theorem by contradiction. If possible, let | f(z)| attains the maximum value
at an interior point z = z0 of the region D enclosed by C. Since f(z) is analytic inside C, we can
expand f(z) by Taylor‟s theorem in the nbd. of zo as

f(z) =  a n (z  z 0 ) n
n 0
1 f (z)
where an =  dz
2 πi γ (z  z 0 ) n
and  is the circle |z  z0| = r, r being small.
We have z  z0 = r ei i.e. z = z0 + r ei, 0    2.

Also | f(z)|2 = f(z) f (z)

  imθ
=  a n r n e inθ  a m r m e
n 0 m 0
 
=   a n a m r n  m e i ( n m )θ
0 0
Integrating both sides from 0 to 2, we get
 
0
2π
| f (z) |2 dθ    a n a m r mn  02 π e i ( n m )θ dθ
0 0

=  anan r2n 2, n = m
n 0

=  |an|2 r2n 2 (1)
0

2 π i ( n m )θ 0 if n  m
where 0 e dθ  
2π if n  m
From (1), we have for n = 0,
2π 2 2
 0 | f(z)| d = |a0| 2
and putting z = z0 in this, we find
2π 2 2
 0 | f(z0)| d = |a0| 2
or
| f(z0)|2  02 π d = |a0|2 2
or
| f(z0)|2 2 = |a0|2 2  | f(z0)|2 = |a0|2 (2)
Also, since f(z) has max. value at z = z0, so
| f(z)|2  | f(z0)|2 = |a0|2
Hence from (1), we get
 COMPLEX ANALYSIS 55

 1 2π
 | an | r 
2 2n
 | f (z) |2 dθ
n 0 2π 0
1 2π
  | f (z 0 ) |2 dθ
2π 0
1
= |a0|2 2 = |a0|2
2π
Thus,
|a0|2 + |a1|2 r2 + |a2|2 r4 +… |a0|2 for positive values of r.
Hence |a1| = |a2| = |a3| =….= 0
i.e. a1 = a2 = a3 = 0
which implies
f(z) = a0 = constant.
Hence | f(z)| cannot attain a max. value at an interior point of D which is a contradiction to our
supposition.
Also | f(z)| attains a max. value at an interior point of D if it is constant and in that case. |
f(z)| = M throughout D.
4.2. Theorem. Let f(z) be analytic within and on a simple closed contour C and let f(z)  0
inside C. Further suppose that f(z) is not constant, then | f(z)| cannot attain a minimum value
inside C.
1
Proof. Since f(z) is analytic within and on C and also f(z)  0 inside C, so is also analytic
f ( z)
within and on C.
1
Therefore, by maximum modulus principle, cannot attain a maximum value inside C
f (z)
which implies that | f(z)| cannot have a minimum value inside C.
4.3. Theorem. Let f(z) be an analytic function, regular for |z| < R and let M(r) denote the
maximum of | f(z)| on |z| = r, then M(r) is a steadily increasing function of r for r < R.
Proof. By maximum modulus principle, for two circles
|z| = r1 and |z| = r2, we have
| f(z)|  M(r), where r1 < r2
which implies M(r1)  M (r2), r1 < r2
and M(r1) = M(r2) if f(z) is constant.
Also M(r) cannot be bounded because if it were so, then f(z) is a constant (by Lioureille‟s
theorem). Hence M(r) is a steadily increasing function of r.
4.4. Schwarz’s Lemma. Let f(z) be analytic in a domain D defined by |z| < R and let
M
| f(z)|  M for all z in D and f(0) = 0, then | f(z)|  |z|.
R
M i
Also, if the equality holds for any one z, then f(z) = ze , where  is real constant.
R
Proof. Let C be the circle |z| = r < R.
Since f(z) is analytic within and on C, therefore by Taylor‟s theorem

f(z) =  an zn at any point z within C.
n 0
56

i.e. f(z) = a0 + a1 z + a2 z2 +
Since f(0) = 0, we get a0 = 0
 f(z) = a1z + a2 z2 + a3 z3 (1)
f ( z)
Let g(z) = (2)
z
then we have
g(z) = a1 + a2z a3z2 (3)
The function g(z) in (2) has a singularity at z = 0 which can be removed by defining g(0) = a 1 .
Now, g(z) is analytic within and on C and so by maximum modulus principle, |g(z)| attains
maximum value on C, say at z = z0 and not within C.
Thus |z0| = r < R and
f (z 0 ) f (z) M
|g(z0)| = = max. value of |g(z)| =  (4)
z0 z r
and thus for any z inside C, we have.
M
|g(z)| <
r
f (z) M M
i.e. <  |f(z)| < |z| (5)
z r r
This inequality holds for all r. s. t. r < R.
Now, L. H. S. is free from r, making rR in (5), we find
M
|f(z)| < |z|  z s. t |z| < R.
R
Also, from (4), we note that for the point z0 on C,
M
|f(z0)| = |z0|
r
Making rR, we get
M
|f(z0)| = |z0|
R
M i
i.e., f(z) = ze for z lying on |z| = R.
R
which proves the result.
4.5. Remarks. (i) If we take M = 1, R = 1, then Schwarz‟s lemma takes the form as follows.
“If f(z) is analytic in a domain D defined by |z|< 1 and | f(z)|  1 for all z in D and
f(0) = 0, then | f(z)|  |z|. Also if the equality holds for any one z, then f(z) = z ei, where  is a
real constant.”
(ii) In view of the power series expansion,
z2
f(z) = f(0) + z f (0) + f (0) +….
|2
we get
f (z) z2
= f (0) + f (0) +….
z |2
 COMPLEX ANALYSIS 57

where we have assumed that f(z) satisfies the conditions of Schwarz‟s lemma so that
f( z) M

z R
This implies that
z2 M
| f (0) + f (0) +……..| 
| 2 R
By setting z = 0, we obtain
M
f (0) 
R
(iii) Let f(z) be analytic inside and on the unit circle, |f(z)|  M on the circle and f(0) = a
where 0 < a < m . Then
M | z | a
| f(z)|  M
a | z | M
inside the circle.
For its proof, we consider
f (z)  a
(z) = M
a f (z)  M 2
f (0)  a a a
Then (0) = M M 2 0
a f (0)  M 2
a  M2
Also, (z) is regular at every point on the unit circle.
f (z)  a Ma
Also, |(z)| = M M 1
a f (z)  M 2 aM  M 2

Thus, (z) satisfies all the conditions of Schwarz‟s lemma.

Therefore,
f ( z)  a
|(z)| = M  |z|
af (z)  M 2
M | z | a
which gives |f(z)|  M .
a | z | M
5. Meromorphic Function. A function f(z) is said to be meromorphic in a region D if it is
analytic in D except at a finite number of poles. In other words, a function f(z) whose only
singularities in the entire complex plane are poles, is called a meromorphic function. The word
meromorphic is used for the phrase “analytic except for poles”. The concept of meromorphic is
used in contrast to holomorphic. A meromorphic function is a ratio of entire functions.

Rational functions are meromorphic functions.

z2 1
e.g. f(z) = 5
z  2z 3  z
(z  1)(z  1) (z  1)(z  1)
= 
z(z  2z  1)
4 2
z(z 2  1) 2
58

(z  1)(z  1)
=
z ( z  i) 2 ( z  i) 2
has poles at z = 0 (simple), at z = + i (both double) and zeros at z = + 1 (both simple)
Since only singularities of f(z) are poles, therefore f(z) is a meromorphic function.

Similarly, tan z, cot z, sec z are all meromorphic functions.

A meromorphic function does not have essential singularity. The following theorem tells about
the number of zeros and poles of a meromorphic function.
5.1. Theorem. Let f(z) be analytic inside and on a simple closed contour C except for a finite
1 f ' ( z)
number of poles inside C and let f(z)  0 on C, then  dz = N  P
2πi C f (z)
where N and P are respectively the total number of zeros and poles of f(z) inside C, a zero (pole)
of order m being counted m times.
Proof. Suppose that f(z) is analytic within and on a simple closed contour C except at a pole
z = a of order p inside C and also suppose that f(z) has a zero of order n at z = b inside C.
Then, we have to prove that
1 f ' ( z)
C dz = np
2πi f ( z)
Let 1 and T1 be the circles inside C with centre at z = a and z = b respectively.

a 1
b
T1 C

Then, by cor. to Cauchy‟s theorem, we have

1 f ' ( z) 1 f ' (z) 1 f ' (z)
 dz =  dz +  dz (1)
2πi C f (z) 2πi γ1 f (z) 2πi T1 f (z)
Now, f(z) has pole of order p at z = a, so
g (z)
f(z) = (2)
(z  a ) p
where g(z) is analytic and non-zero within and on 1. Taking logarithm of (2) and differentiating,
we get
log f(z) = log g(z)  p log (z  a)
f ' ( z) g ' ( z) p
i.e.,  
f ( z) g ( z) z  a
Therefore,
f ' ( z) g' (z) dz
 dz   dz  p  (3)
γ1 f ( z ) γ1 g ( z ) γ1 z a
 COMPLEX ANALYSIS 59

g ' (z)
Since is analytic within and on 1, by Cauchy theorem,
g(z)
g' (z)
 dz  0
γ1 g(z)

f ' (z)
Thus (3) gives  dz =  2ip (4)
γ1 f (z)
Again, f(z) has a zero of order n at z = b, so we can write
f(z) = (z  b)n (z) (5)
where (z) is analytic and non-zero within and on T1
Taking logarithm, then differentiating, we get
f ' ( z) n ' (z)
 
f (z) z  b (z)
or
f ' ( z) dz ' (z)
 f (z) dz  n  z  b   (z) dz (6)
T 1 T1 T1

' (z)
Since is analytic within and on T1, therefore
(z)
' (z)
 (z) dz = 0 and thus (6) becomes
T1
f ' (z)
 dz = 2in (7)
T1 f (z)
Writing (1) with the help of (4) and (7), we get
1 f ' ( z)
 dz = p + n = np (8)
2πi C f (z)
Now, suppose that f(z) has poles of order pm at z = am for m = 1, 2,…,r and zeros of order nm at
z = bm for m = 1, 2,…,s within C. We enclose each pole and zero by circles 1, 2,…, r and T1,
T2,…, Ts. Thus (8) becomes
1 f ' ( z) s r
 dz =  n m   p m
2πi C f (z) m1 m1
s r
Taking  n m  N,  p m  P, we obtain
m1 m1
1 f ' ( z)
 dz = N  P which proves the theorem. This theorem is also known
2πi C f (z)
as the argument principle which can be put in a more explicit manner as follows :
5.2. Theorem (The Argument Principle). Let f(z) be meromorphic inside a closed contour C
and analytic on C where f(z)  0. When f(z) describes C, the argument of f(z) increases by a
multiple of 2, namely
C arg f(z) = 2 (NP)
where N and P are respectively the total number of zeros and poles of f(z) inside C, a zero (pole)
of order m being counted m times.
Proof. Let arg f(z) = 
60

So, we can write

f(z) = | f(z)| ei
i.e. log f(z) = log | f(z)| + i (1)
Then as proved in the above theorem 4.1,

1 f ' ( z)
NP=  dz
2πi C f (z)
1
=  d (log f(z))
2πi C
1
=  d (log | f(z)| + i )
2πi C
1 1
=  d (log | f(z)| +  d (2)
2πi C 2π C
The first integral in (2) vanishes, since log | f(z)| is single valued, i.e., it returns to its original
value at z goes round C. Now,  d is the variation in the argument of f(z) in describing the
C
contour C,
Therefore d = C arg f(z)
Thus, (2) becomes
C arg f(z) = 2 (N  P)
This formula makes it possible to compute the number NP from the variation of the argument
of f(z) along the boundary of the closed contour C and is known as argument principle.
In particular, if f(z) is analytic inside and on C, then P = 0
1
and N= C arg f(z).
2π
5.3. Rouche’s Theorem. If f(z) and g(z) are analytic inside and on a closed contour C and
|g(z)| < | f(z)| on C, then f(z) and f(z) + g(z) have the same number of zeros inside C.
Proof. First we prove that neither f(z) nor f(z) + g(z) has a zero on C.
If f(z) has a zero at z = a on C, then f(a) = 0
Thus |g(z) < | f(z)|  |g(a)| < f(a) = 0
 g(a) = 0  | f(a)| = |g(a)|
i.e. | f(z)| = |g(z)| at z = a
which is contrary to the assumption that
|g(z)| < |f(z)| on C.
Again, if f(z) + g(z) has a zero at z = b on C,
then f(b) + g(b) = 0  f(b) = g(b)
i.e. | f(b)| = |g(b)|
again a contradiction.
Thus, neither f(z) nor f(z) + g(z) has a zero on C.
Now, let N and N be the number of zeros of f(z) and f(z) + g(z) respectively inside C. We are to
prove that N = N.
 COMPLEX ANALYSIS 61

Since f(z) and f(z) + g(z) both are analytic within and on C and have no pole inside C, therefore
the argument principle
1 f'
 dz  N  P, with P  0 , gives
2πi C f
1 f' 1 f ' g '
 dz  N, 2i  f  g dz  N'
2πi C f C
Subtracting these two results, we get
1  f ' g' f ' 
   dz = N N
2πi C  f  g f 
g ( z)
Let us take (z) = so that g = f 
f ( z)
Now, |g| < |f|  |g/f| < 1 i.e. ||< |
Therefore,
f 'g' f 'f '   f' f ' (1  )  f'
 
f g f  f f (1  )
f' '
= 
f 1 
f 'g' f ' '
i.e.   (2)
f  g f 1 
Using (2) in (1), we get
1 ' 1
 (1 + )1 dz
2i C 1   2i C
N  N = dz  (3)

Since we have observed that || < 1, so binomial expansion of (1 + )1 is possible and this
expansion in powers of  is uniformly convergent and hence term by term integration is possible.
Thus,   (1+)1dz =  (1+2  3 +…) dz
C C
=   dz    dz +  2  dz (4)
C C C
Now, the functions f and g both are analytic within and on C and f  0 g  0 for any point on C,
therefore  = g/f is analytic and non-zero for any point on C. Thus  and its all derivatives are
analytic and so by Cauchy‟s theorem, each integral on R.H.S. of (4) vanishes. Thus
1
  (1 + ) dz = 0
C
and therefore from (3), we conclude N  N = 0
i.e. N = N
5.4. Theorem (Fundamental Theorem of Algebra). Every polynomial of degree n has exactly
n zeros.
Proof. Let us consider the polynomial
a0 + a1 z + a2 z2 + …+ an zn, an  0
We take f(z) = an zn, g(z) = a0 + a1 z + a2 z2 +…+ an1 zn1

Let C be a circle |z| = r, where r > 1.

62

Now, | f(z)| = |an zn| = |an| rn

|g(z)|  |a0| + |a1| r + |a2| r2+…+ |an1| rn1
 (|a0| + |a1| +…+ |an1| rn1
Therefore,
g( z) (| a 0 |  | a1 | ... | a n 1 |) r n 1

f ( z) | an | rn
| a |  | a 1 | ... | a n 1 |
= 0
| an | r
Hence |g(z) < |f(z)|, provided that
| a 0 |  | a 1 | ... | a n 1 |
<1
| an | r
| a |  | a 1 | ... | a n 1 |
i.e. r> 0 (1)
| an |
Since r is arbitrary, therefore we can choose r large enough so that (1) is satisfied. Now,
applying Rouche‟s theorem, we find that the given polynomial f(z + g(z) has the same number of
zeros as f(z). But f(z) has exactly n zeros all located at z = 0. Hence the given polynomial has
exactly n zeros.
5.5. Example. Determine the number of roots of the equation
z8  4z5 + z2  1 = 0
that lie inside the circle |z| = 1
Solution. Let C be the circle defined by |z| = 1
Let us take f(z) = z8  4z5, g(z) = z21.
On the circle C,
g(z) z2 1 | z | 2 1
 8 
f (z) z  4z 5
| z |5 | 4  z 3 |
11 2 2
   1
4 | z |3 4  1 3
Thus |g(z)| <|f(z)| and both f(z) and g(z) are analytic within and on C, Rouche‟s theorem implies
that the required number of roots is the same as the number of roots of the equation z8  4z5 = 0
in the region |z| < 1. Since z3  4  0 for |z| < 1, therefore the required number of roots is found to
be 5.
5.6. Inverse Function. If f(z) = w has a solution z = F(w), then we may write
f{F(w) } = w, F{ f(z)} = z. The function F defined in this way, is called inverse
function of f.
5.7. Theorem. (Inverse Function Theorem). Let a function w = f(z) be analytic at a point z = z0
where f (z0)  0 and w0 = f(z0).
Then there exists a neighbourhood of w0 in the w-plane in which the function w = f(z) has a
unique inverse z = F(w) in the sense that the function F is single-valued and analytic in that
neighbourhood such that F(w0) = z0 and
1
F(w) = .
f ' (z)
 COMPLEX ANALYSIS 63

Proof. Consider the function f(z)  w0. By hypothesis, f(z0)  w0 = 0. Since f (z0)  0, f is not a
constant function and therefore, neither f(z)  w0 not f (z) is identically zero. Also f(z)  w0 is
analytic at z = z0 and so it is analytic in some neighbourhood of z0. Again, since zeros are
isolated, neither f(z)  w0 nor f (z) has any zero in some deleted neighbourhood of z0. Hence
there exists  > 0 such that f(z)  w0 is analytic for |z  z0|   and f(z)  w0  0, f (z)  0 for
0 < |zz0|   . Let D denote the open disc
{z : |z z0| < }
and C denotes its boundary
{z : |zz0| = }.
Since f(z)  w0 for |z  z0|  , we conclude that | f(z) w0| has a positive minimum on the circle
C. Let
min | f(z) w0| = m
zC

and choose  such that 0 <  < m.

We now show that the function f(z) assumes exactly once in D every value w1 in the open disc
T = { w : |w w0| < }. We apply Rouche‟s theorem to the functions w0  w1 and
f(z)w0. The condition of the theorem are satisfied, since
|w0w1| <  < m = min |f(z) w0|  |f(z)  w0| on C.
zC

Thus we conclude that the functions.

f(z)w0 and (f(z) w0) + (w0 w1) = f(z) w1
have the same number of zeros in D. But the function f(z)  w0 has only one zero in D i.e. a
simple zeros at z0, since (f(z) w0) = f(z)  0 at z0.
Hence f(z)w1 must also have only one zero, say z1 in D. This means that the function f(z)
assumes the value w, exactly once in D. It follows that the function w = f(z) has a unique
inverse, say z = F(w) in D such that F is single-valued and w =f {F(w)}. We now show that the
function F is analytic in D. For fix w1 in D, we have f(z) = w1 for a unique z1 in D. If w is in T
and F(w) = z, then
F( w )  F( w 1 ) z  z1
 (1)
w  w1 f ( z )  f ( z1 )
It is noted that T is continuous. Hence z  z1 whenever ww1. Since z1 D, as shown above f
(z1) exists and is zero. If we let ww, then (1) shows that
1
F(w1) = .
f ' ( z1 )
Thus F(w) exists in the neighbourhood T of w0 so that the function F is analytic there.
6. Calculus of Residues
The main result to be discussed here is Cauchy’s residue theorem which does for
meromorphic functions what Cauchy’s theorem does for holomorphic functions. This
theorem is extremely important theoretically and for practical applications.
6.1. The Residue at a Singularity. We know that in the neighbourhood of an isolated singularity
z = a, a one valued analytic function f(z) may be expanded in a Laurent‟s series as
 
f(z) =  a n (z  a ) n   b n (z  a ) n
n 0 n 1
The co-efficient b1 is called the residue of f(z) at z = a and is given by the formula
64

1 1 f (z)dz
Res (z = a) = b1=  f(z) dz | bn 
2πi γ 2πi γ (z  a ) n 1
Where  is any circle with centre z = a, which excludes all other singularities of f(z). In case,
z = a is a simple pole, then we have
 b
Res (z = a) = b1 = lim (za) f(z) | f(z) =  an (za)n + 1
z a 0 za
A more general definition of the residue of a function f(z) at a point z = a is as follows.
If the point z = a is the only singularity of an analytic function f(z) inside a closed contour C,
1
then the value  f(z) dz is called the residue of f(z) at a.
2πi C
6.2. Residue at Infinity. If f(z) is analytic or has an isolated singularity at infinity and if C is a
circle enclosing all its singularities in the finite parts of the z-plane, the residue of f(z) at infinity
is defined by
1 1
Res (z = ) =  f(z) dz, | or Res (z =)   f(z) dz
2πi C 2πi C
Integration taken in positive sense
the integration being taken round C in the negative sense w.r.t. the origin, provided that this
integral has a definite value. By means of the substitution z = w1, the integral defining the
residue at infinity takes the form
1 dw
 [f(w1)] 2 ,
2πi w
taken in positive sense round a sufficiently small circle with centre at the origin.
Thus, we also say if
lim [f(w1) w1] or lim [ z f(z)]
w 0 z 
has a definite value, that value is the residue of f(z) at infinity.
6.3. Remarks. (i) The function may be regular at infinity, yet has a residue there.
b
For example, consider the function f(z) = for this function
za
1
Res (z = ) =   f(z) dz
2πi C
1 b
=  dz
2πi C z  a
b 2 π reiθ idθ
=  , C being the circle |z a| = r
2πi 0 reiθ
b 2π
=  dθ  b
2π 0
 Res ( z = ) = b
1
Also, z = a is a simple pole of f(z) and its residue there is  f(z) dz = b
2πi C
| or lim (za) f(z) = b
z a
Thus Res (z = a) = b  Res (z = )
 COMPLEX ANALYSIS 65

(ii) If the function is analytic at a point z = a, then its residue at z = a is zero but not so at
infinity.
(iii) In the definition of residue at infinity, C may be any closed contour enclosing all the
singularities in the finite parts of the z-plane.
6.4. Calculation of Residues. Now, we discuss the method of calculation of residue in some
special cases.
(i) If the function f(z) has a simple pole at z = a, then, Res (z = a) = lim (za) f(z).
z a
(z)
(ii) If f(z) has a simple pole at z = a and f(z) is of the form f(z) = i.e. a rational
( z)
function, then
(z)
Res (z = a) = lim (za) f(z) = lim (za)
z a z a ( z)
(z)
= lim
z a  ( z )   (a )
za
(a )
= ,
 ' (a )
where (a) = 0, (a)  0, since (z) has a simple zero at z = a
(iii) If f(z) has a pole of order m at z = a then we can write
(z)
f(z) = (1)
(z  a ) m
where (z) is analytic and (a)  0.
1 1 (z)
Now, Res (z = a) = b1 =  f(z) dz =  dz
2πi C 2πi C (z  a ) m
1 | m 1 (z)
=
| m  1 2i C (z  a )
 m 11
dz

1
= m1(a) [By Cauchy‟s integral formula for derivatives] (2)
| m 1
Using (1), formula (2) take the form
1 d m 1
Res (z = a) = m 1
[(za)m f(z)] as za
| m  1 dz
1 d m 1
i.e. Res (z = a) = lim [(za)m f(z)] (3)
z a | m  1 dz m 1

Thus, for a pole of order m, we can use either formula (2) or (3).
(iv) If z = a is a pole of any order for f(z), then the residue of f(z) at z = a is the
1
co-efficient of in Laurent‟s expansion of f(z)
za
1
(v) Res (z = ) = Negative of the co-efficient of in the expansion of f(z) in the
z
neighbourhood of z = .
66

z4
6.5. Examples. (a) Find the residue of at z = ia
z2  a 2
z4
Solution. Let f(z) = .
z  a2
2

Poles of f(z) are z = + ia

Thus z = ia is a simple pole, so
Res (z = ia) = lim (z + ia) f(z)
z   ia

z4
= lim (z  ia )
z ia (z  ia )(z  ia )
z4
4 4 or (z) 
= lim
z

a z  ia
z ia z  ia  2ia z 4 /(z  ia )
as f (z) 
(z  ia )
a 4i ia 3
= 
2a 2
iz 4
(b) Find the residues of e z at its poles.
e iz
Solution. Let f(z) = 4
z
f(z) has pole of order 4 at z = 0, so
1  d3  i
Res (z = 0) =  3 (e iz )   | (z) = eiz
| 3  dz  z 0 6
Alternatively, by the Laurent‟s expansion
e iz 1 i 1 i
     ...
z 4 z 4 z3 | 2 z 2 | 3 z
we find that
1
Res (z = 0) = co-efficient of
z
i
= .
6
z3
(c) Find the residue of at z = .
z 1
2

1
z3 z3  1 
Solution. Let f(z) =  = z 1  2 
z 2 1  1   z 
z 2 1  
 z2 
1 1
= z ( 1  2  4  ....)
z z
1 1
= z +  3  ...
z z
Therefore,
 COMPLEX ANALYSIS 67

1
Res (z = ) = (co-efficient of )  1
z
z3
(d) Find the residues of at its poles.
(z  1) 4 (z  2)(z  3)
z3
Solution. Let f(z) =
(z  1) 4 (z  2)(z  3)
Poles of f(z) are z = 1 (order four) and z = 2, 3 (simple)
Therefore,
z3
Res (z = 2) = lim (z2) f(z) = lim = 8
z2 z2 (z  1) 4 (z  3)
27
Res (z = 3) = lim (z  3) f(z) =
z 3 16
z3
For z = 1, we take (z) = ,
(z  2)(z  3)
(z) 3 (1)
where f(z) = and thus Res (z = 1) =
(z  1) 4
|3
8 27
Now, (z) = z +5  
z 2 z 3
48 162
3(z) = 
(z  2) 4
(z  3) 4
303
3(1) =
8
Thus,
303 101
Res (z = 1) = 
8 | 3 16
6.6. Theorem. (Cauchy Residue Theorem). Let f(z) be one-valued and analytic inside and on a
simple closed contour C, except for a finite number of poles within C. Then
 f(z) dz = 2i [Sum of residues of f(z) at its poles within C]
C
Proof. Let a1, a2,…., an be the poles of f(z) inside C. Draw a set of circles r of radii  and centre
ar (r = 1, 2,…, n) which do not overlap and all lie within C. Then f(z) is regular in the domain
bounded externally by C and internally by the circles r.

a5
a 1 1 5

C
2 a2 a3 a4 4
3
68

Then by cor. to Cauchy‟s Theorem, we have

n
 f(z) dz =   f(z) dz (1)
C r 1 γr

Now, if ar is a pole of order m, then by Laurent‟s theorem, f(z) can be expressed as

m bs
f(z) = (z) + 
s 1 (z  a r )
s

where (z) is regular within and on r .

Then
m bs
 f(z) dz =   dz (2)
γ r (z  a r )
s
γr s 1

where  f(z) dz = 0, by Cauchy‟s theorem

γr

Now, on r |zar| =  i.e. z = ar + ei

 dz =  iei d
where  varies from 0 to 2 as the point z moves once round r.
m
1s 2 π (1s )iθ
Thus,  f(z) dz =  bs   0 e idθ
γr s 1

= 2i b1
= 2i [Residue of f(z) at ar]
2π (1s )iθ 0, if s  1
where 0 e dθ = 
2π if s  1
Hence, from (1), we find
n
 f (z)dz   2i [Residue of f(z) at ar]
C r 1

n 
= 2i   Re sidue of f (z) at a r 
r 1 
= 2i [sum of Residues of f(z) at its poles inside C.]
which proves the theorem.
(z)
6.7. Remark. If f(z) can be expressed in the form f(z) = where (z) is analytic and
(z  a ) m
(a)  0, then the pole z = a is a pole of type I or overt.
(z)
If f(z) is of the form f(z) = , where (z) and (z) are analytic and (a)  0 and (z) has a
(z)
zero of order m at z = a, then z = a is a pole of type II or covert. Actually, whether a pole of f(z)
is overt or covert, is a matter of how f(z) is written
7. Evaluation of Integrals
Cauchy‟s residue theorem provides the natural way to deal with a contour integral  f(z) dz
C
where f(z) has poles inside C. For an integral round just one type I pole Cauchy‟s integral
formula or Cauchy‟s formula for derivatives should be used. Of course, when f(z) has no poles
inside or on C, Cauchy‟s theorem applies. In Cauchy‟s residue theorem‟s applications to
 COMPLEX ANALYSIS 69

evaluation of definite integrals, we frequently need to find the limiting value of an integral along
a path as that path shrinks or expands indefinitely.
For evaluation of integrals of the type  02 π f( cos , sin ) where f(cos , sin ) is a rational
function of cos  and sin  and bounded for the entire range of integration, we use the
z  z 1 z  z 1
transformation z = ei i.e. cos  = , sin θ  This will reduce the integral to the form
2 2i
 g(z) dz where g(z) is a rational function of z, with no singularity on the unit circle |z| = 1,
C
denoted by C.
Another type of integrals, we shall consider, are  f(x) dx, which are obtained by using
complex function f(z).
a dθ
7.1. Example. Evaluate  0π 2 , where a > 0
a  sin 2 θ
a dθ
Solution. Let I =  0π 2
a  sin 2 θ
2a dθ 2a dθ
=  0π   0π
2a 2  2 sin 2 θ 2a 2  1  cos 2θ
a dt
=  02 π , 2 = t
2a 2  1  cos t
a dt
=  02 π
 e it  e it 
2a 2  1   

 2 
it it
Putting z = e so that dz = e idt, we get
2a dz
I=  1
C 2( 2a  1)  ( z  z ) iz
2

2a dz
= 
i C  z  2(2a 2  1)z  1
2

dz
= 2ai  2
C z  2( 2a  1) z  1
2

or
I =2ai  f(z) dz (1)
C
1
where f(z) = and C is the unit circle |z| = 1
z 2  2(2a 2  1)z  1
Now, poles of f(z) are given by
z2  2 (2a2 + 1) z + 1 = 0
i.e. z = (2a2 + 1) + 2a a 2 1
we take  = (2a +1) + 2a a 2  1
2

 = (2a2 + 1)  2a a 2  1
Thus, poles of f(z) are z = 1
Clearly, || > 1 and since || = 1  || < 1
70

Thus, f(z) has only one simple pole z =  lying within C.

Res (z = ) = lim (z) f(z)
zβ
1 1
=
   4a a 2  1
Hence by Cauchy‟s Residue theorem
 f(z) dz = 2i (sum of residues of f(z) at its poles within C)
C
 1 
= 2i  
 
 4 a a 2
 1 
Thus, from (1), we get
  2i  
I = 2ai  
 
 4a a  1  a 1
2 2

π adθ π
Similarly, we have  0 2 
a  cos2 θ a 1
2

2π
7.2. Example. Prove that  02 π e cos θ cos(sinθ  nθ)dθ  , n is a +ve integer.
|n
Solution. Let I =  02 π e cosθ cos (sin  n) d
= R.P.  02 π e cosθ e(sin n)i d
= R.P.  02 π e cosθ + i sin  in d
2 i in
= R.P. 0 ee d
iθ
= R.P.  02 π e e (e iθ ) n dθ
putting z = ei, we get
dz 1 ez
I = R.P.  e z z n  R.P.  n 1 dz
C iz iC z
1
= R.P.  f(z)dz (1)
iC
ez
where f(z) = , C is the unit circle |z| = 1
z n 1
Evidently, z = 0 is a pole of order n + 1 for f(z), lying within C.
1  dn z  1
 Res (z = 0) =  ne  
|  dz
n  z 0 | n
Thus, using Cauchy‟s residue theorem, we get
1 1 2π
I = R.P. 2πi 
i |n |n
Note that  02 π e cosθ sin (sin   n) d = 0
Similarly, we have
 COMPLEX ANALYSIS 71

2π cos θ 2π(1) n
0 e cos(sinθ  nθ)dθ  .
|n
7.3. Theorem. Let f(z) be a function of the complex variable z satisfying the conditions
(i) f(z) is meromorphic in the upper half of the complex plane i.e. Im. z  0.
(ii) f(z) has no pole on the real axis
(iii) z f(z)0 uniformly as |z| for 0  arg z  .
(iv)  0 f(x) dx and   f(x) dx both converge.
Then   f(x) dx = 2i  Res;
where  Res. denotes the sum of residues of f(z) at its poles in the upper half of the z-plane.
Proof. Let us consider the integral  f(z)dz, where C is the contour consisting of the segment of
C
the real axis from R to R and the semi-circle in the upper half plane on it as diameter.
Let the semi-circular part of the contour C be denoted by T and let R be chosen so large
that C includes all the poles of f(z).

R R
O

Then by Cauchy‟s residue theorem,

R
 f(z) dz =  R f(x) dx +  f(z) dz = 2i  Res. (1)
C T
By hypothesis (iii), |z f(z)| <  for all points z on T. If R is chosen sufficiently large, however
small be the +ve number , then for such R,
|  f(z) dz| = |  0π f(Rei) Rei id |
T

= |  0π zf(z) d |
<   0π d = 
It follows that as |z| = R,  f(z) dz0
T
Now, since hypothesis (iv) holds,

  f ( x )dx  lim  R f(x) dx
R
R 
Taking limit as R in (1), we get

  f(x) dx = 2i Res.
Hence the result
7.4. Example. By method of contour integration, prove that
72

 dx
0
(1  x 2 ) 2
 /4

Solution. Consider the integral

1
 f(z) dz, where f(z) =
C (1  z2 )2
C being the closed contour consisting of T, the upper half of the large circle |z| = R and the real
axis from R to R.
Poles of f(z) are z = + i(each of order two) f(z) has only one pole of order two at z = i within C.
We can write
(z) 1
f(z) = , where (z) =
( z  i) 2 ( z  i) 2
1 1
 Res (z = i) = (i) =
| 1 4 i
Hence, by Cauchy‟s residue theorem
1 π
 f(z) dz = 2i =
C 4i 2
R π
or  f(z) dz +  R f(x) dx =
T 2
dz R dx 
i.e.   R  (1)
T ( z  1) (1  x 2 ) 2 2
2 2

Now, using the inequality,

1 1
|z1 + z2|  |z1|  |z2|,  , we get
| z1  z 2 | z1 |  | z 2 |
dz | dz | | dz |
  
T (1  z 2 2
) T ( z 2
 1) 2
T (| z |2 1) 2
1 πR
=  | dz | 2 0 as |z| = R
(R 2  1) 2 T (R  1) 2
dz
so lim  0
R  T (1  z 2 ) 2

Making R in (1), we obtain

 dx
   π/2
(1  x 2 ) 2
 dx
or 0 = /4.
(1  x 2 ) 2
dx π ( a  2b )
7.5. Example. Prove that    a > 0, b > 0
( x  a )(x  b )
2 2 2 2 2
2ab 3 (a  b) 2
Solution. Consider the integral  f(z)dz,
C
1
where f(z) =
(z 2  a 2 )(z 2  b 2 ) 2
 COMPLEX ANALYSIS 73

and c is the closed contour consisting of T, the upper half of the large circle |z| = R and the real
axis from R to R.
Poles of f(z) are z = + ia (simple) and z = + ib (double)
Only poles of f(z) lying within C are z = ia (simple) and z = ib (double)
(z)
f (z) 
1 (z  ib) 2
Res (z = ia) =
2ia (a 2  b 2 ) 2 1
(z)  2
(z  a 2 )(z  ib) 2
(3b 2  a 2 )i
Res (z = ib) =
4b 3 (a 2  b 2 ) 2
i  2 3b 2  a 2 
Thus, sum of residues = 2 2 
  
4(a  b )  a
2
b3 

=

i  2b 3  a (3b 2  a 2 ) 
4ab 3 (a 2  b 2 ) 2
i[( b 3  a 3 )  3b 2 (b  a )]
=
4ab 3 (a 2  b 2 ) 2
i(b  a )[b 2  a 2  ab  3b 2 ]
=
4ab 3 (a 2  b 2 ) 2
i(b  a )(a  b)(a  2b)  i ( a  2b )
= 
4ab (a  b )
3 2 2 2
4ab 3 (a  b) 2
So, by Cauchy residue theorem,
 f(z) dz =  f (z)   R f(x) dx
R

C T

  i ( a  2b )  π (a  2b )
= 2i  2
= (1)
 4ab (a  b)  2ab (a  b)
3 3 2

dz
Now, |  f(z)dz| = 
T (z  a )(z  b )
2 2 2 2 2
T

| dz |
= 
T(| z |2 a 2 )(| z |2 b 2 ) 2
πR
= 0 as |z| = R.
(R  a )(R 2  b 2 )
2 2

Making R in (1), we obtain

 dx π ( a  2b )
   .
( x 2  z 2 )(x 2  b 2 ) 2 2ab 3 (a  b) 2
Hence the result.
Deductions. (i) putting a = 1, b = 2, we get
 dx 5π dx 5π
   or  0 
( x 2  1)(x 2  4) 2 144 ( x 2  1)(x 2  4) 2 288
74

(ii) Putting a = 2, b = 1, we get

 dx π dx π
   or  0  .
( x  2)(x  1)
2 2 2
9 ( x  2)(x  1)
2 2 2
18
(iii) Putting a = 3, b = 2, we get
 dx 7π dx 7π
   or  0 
( x 2  9)(x 2  4) 2 1200 ( x 2  9)(x 2  4) 2 2400
2θ
7.6 Jordan’s Inequality. If 0    /2, the  sin   
π
This inequality is called Jordan inequality. We know that as  increases from 0 to /2, cos
decreases steadily and consequently, the mean ordinate of the graph of y = cos x over the range 0
 x   also decreases steadily. But this mean ordinate is given by
1 θ sin θ
 cos x dx 
θ 0 θ
It follows that when 0    /2,
2 sin θ
 1
π θ
7.7. Jordan’s Lemma. If f(z) is analytic except at a finite number of singularities and if f(z)0
uniformly as z, then
lim  e imz f(z) dz = 0 , m > 0
R  T

where T denotes the semi-circle |z| = R, Im. z  0, R being taken so large that all the singularities
of f(z) lie within T.
Proof. Since f(z)0 uniformly as |z|, there exists  > 0 such that | f(z)| <   z on T.
Also |z| = R  z = Rei  dz = Rei id |dz| = Rd
iθ
|eimz| = | e im Re | = |eimR cos  emR sin |
= emR sin
Hence, using Jordan inequality,
|  eimz f(z) dz|   |eimz f(z)| |dz|
T T

<  0π e  mR sin a R d
= 2  R  0π / 2 e  mR sin θ d
2a
/ 2
  sin 
= 2  R 0 e  2 mR / 
d 
 2
i.e.  sin  

(1  e  mR )
= 2R
2mR / π
π π
= (1  e mR ) 
m m
Hence lim  e imz f(z) dz = 0
R  T

7.8. Example. By method of contour integration prove that

 COMPLEX ANALYSIS 75

 cos mx π ma
0 dx  e , where m  0, a > 0
x2  a2 2a
Solution. We consider the integral
e imz
 f(z)dz, where f(z) = 2
C z  a2
and C is the closed contour consisting of T, the upper half of the large circle |z| = R and real axis
from R to R.
1
Now, 2 0 as |z| = R
z  a2
Hence by Jordan lemma,
e im z
lim  2 dz  0
R  T z  a 2

i.e. lim  f (z)dz  0 (1)

R  T

Now, poles of f(z) are given by z = + ia (simple), out of which z = ia lies within C.
e  ma
 Res (z = ia) =
2ia
Hence by Cauchy‟s residue theorem,
e  ma π ma
 f(z)dz = 2i = e
C 2ia a
or
R π ma
 f(z)dz +  R f(x)dx = e
T a
Making R and using (1), we get
 e imx π
  2 dx  e ma
x  a2 a
Equating real parts, we get
 cos mx π
  2 dx  e ma
x  a2 a
 cos mx π ma
or 0 dx  e
x2  a2 2a
Hence the result.
Deduction. (i) Replacing m by a and a by 1 in the above example, we get
 cos ax π
0 dx  e a
x 1
2
2
Putting a = 1, we get
 cos x π π
0  e 1 
x 1
2
2 2 e
(ii) Taking m = 1, a = 2, we get
 cos x π
0 dx  2 .
x2  4 4e
76

x 3 sin mx π ma /  ma 
7.9. Example. Prove that    e 2
cos  m > 0, a > 0
x4  a4 2  2
Solution. Consider the integral  f(z)dz, where
C

z 3e imz
f(z) = 4
z  a4
and C is the closed contour….
z3
Since 4 0 as |z| = R, so by
z  a4
Jordan lemma,
z 3e imz
lim  4 dz = 0 (1)
R  T z  a 4

Poles of f(z) are given by

z4 + a4 = 0
or z4 =  a4 = e2ni ei a4
or z = a e(2n+1)i/4, n = 0, 1, 2, 3.
Out of these four simple poles, only
z = aei/4, a ei3/4 lie within C.
φ( z ) φ( z )
If f(z) = , then Res ( z = ) = lim ,  being simple pole.
ψ( z ) z α ψ' ( z )
 For the present case,
z 3 e imz e imz
Res (z = ) = lim 3
= lim .
z α 4 z z α 4
i/4 iz/4
Thus, Res (z = ae ) + Res (z = a e )
1
= [exp (ima ei/4) + exp (ima eiz/4)]
4
1   i  1     1  i  
= expima   expima  
4   2    2  
1   ma   ima    ima 
exp
= exp   exp 
4  2   2   2 
1   ma   ma 
= exp  cos 
2  2   2
Hence by Cauchy‟s residue theorem,
  ma   ma 
 f(z) dz =  f(z) dz +  R f(x) dx = i exp   cos 
R

C T  2   2
Taking limit as R and using (1), we get
 x 3e imx   ma   ma 
  4 dx  πi exp  cos 
x  a4  2   2
Equating imaginary parts, we obtain
 x 3 sin mx   ma   ma 
  dx  π exp  cos 
x a
4 4
 2   2
 COMPLEX ANALYSIS 77

x 3 sin mx
 π   ma   ma 
or 0 dx  exp  cos 
x a
4 4
2  2   2
8. Multivalued Function and its Branches
The familiar fact that sin  and cos  are periodic functions with period 2, is responsible for the
non-uniqueness of  in the representation z = |z| ei i.e. z = rei. Here, we shall discuss non-
uniqueness problems with reference to the function arg z, log z and za. We know that a function
w = f(z) is multivalued when for given z, we may find more than one value of w. Thus, a
function f(z) is said to be single-valued if it satisfies
f(z) = f(z(r, )) = f(z(r,  + 2))
otherwise it is classified as multivalued function.
For analytic properties of a multivalued function, we consider domains in which these functions
are single valued. This leads to the concept of branches of such functions. Before discussing
branches of a many valued function, we give a brief account of the three functions arg z, log z
and za.
8.1. Argument Function. For each z , z  0, we define the argument of z to be
arg z = [arg z] = {  R : z = |z|ei}
the square bracket notation emphasizes that arg z is a set of numbers and not a single number.
i.e. [arg z] is multivalued. Infact, it is an infinite set of the form {  + 2n : n  I}, where  is
z
any fixed number such that ei = .
|z|
For example, arg i = {(4n +1) /2 : n  I}
1
Also, arg   = { :   arg z}
z
Thus, for z1, z2  0, we have
arg (z1 z2) = {1 + 2 : 1  arg z1, 2  arg z2}
= arg z1 + arg z2
 z1 
and arg   = arg z1  arg z2
 z2 
For principal value determination, we can use Arg z = , where z = |z| ei,  <   
(or 0   < 2). When z performs a complete anticlockwise circuit round the unit circle, 
increases by 2 and a jump discontinuity in Arg z is inevitable. Thus, we cannot impose a
restriction which determines  uniquely and therefore for general purpose, we use more
complicated notation arg z or [arg z] which allows z to move freely about the origin with 
varying continuously. We observe that
arg z = [arg z] = Arg z + 2n, nI.
8.2. Logarithmic Function. We observe that the exponential function ez is a periodic function
with a purely imaginary period of 2i, since
ez+2i = ez. e2i = ez, e2i = 1.
i.e. exp (z + 2i) = exp z for all z.
If w is any given non-zero point in the w-plane then there is an infinite number of points in the
z-plane such that the equation
w = ez (1)
is satisfied. For this, we note that when z and w are written as z = x + iy and w =  ei ( < 
 ), equation (1) can be put as
78

ez = ex+iy = ex eiy =  ei (2)

From here, e =  and y =  + 2n, n  I.
x

Since the equation ex =  is the same as x = loge  = log  (base e understood), it follows that
when w =  ei( <   ), equation (1) is satisfied if and only if z has one of the values
z = log  + i ( + 2n), n  I (3)
Thus, if we write
log w = log  + i ( + 2n), nI (4)
we see that exp (log w) = w, this motivates the following definition of the (multivalued)
logarithmic function of a complex variable.
The logarithmic function is defined at non-zero points z = rei ( <   ) in the z-plane as
log z = log r + i ( + 2n), n  I (5)
The principal value of log z is the value obtained from (5) when n = 0 and is denoted by Log z.
Thus
Log z = log r + i i.e. Log z = log |z| + i Arg z (6)
Also, from (5) & (6), we note that
log z = Log z + 2ni, n  I (7)
The function Log z is evidently well defined and single-valued when z  0.
Equation (5) can also be put as
log z = {log |z| + i :  arg z}
or [log z] = {log |z| + i :  [arg z]} (8)
or log z = log |z| + i = log |z| + i arg z (9)
where  =  + 2n,  = Arg z.
From (8), we find that
log 1 = {2ni, n  I}, log (1) = {(2n+1) i, nI}
In particular, Log 1 = 0, Log (1) = i . Similarly log, log i = {(un+1) i/2, nI}, log (i) =
{un1) i/2, nI} In particular, Log i = i/2, Log (i) = i/2.
Thus, we conclude that complex logarithm is not a bona fide function, but a multifunction. We
have assigned to each z  0 infinitely many values of the logarithm.
8.3. Complex Exponents. When z  0 and the exponent a is any complex number, the function
za is defined by the equation.
a
w = za = e log z = ea log z = exp ( a log z) (1)
where log z denotes the multivalued logarithmic function. Equation (1) can also be expressed as
w = za = {ea(log |z| + i) :  arg z}
or [za] = {ea(log |z| + i) :  [arg z]}
Thus the manivalued nature of the function log z will generally result in the many-valuedness of
za. Only when a is an integer, za does not produce multiple values. In this case za contains a
1
single point zn. When a = (n = 2, 3,…), then
n
w = z1/n = (r ei)1/n = r1/n ei( + 2m )/n, mI
We note that in particular, the complex nth roots of + 1 are obtained as
wn = 1  w = e2mi/n, wn = 1  w = e(2m+1)i/n, m = 0,1,…, n1.
2i
For example, i = exp (2 i log i) = exp [2i (4n+1) i/2]
 COMPLEX ANALYSIS 79

= exp [(4n+1) ], n  I

In should be observed that the formula
xa xb = xa+b, x, a, b,  R
can be shown to have a complex analogue (in which values of the multi-functions involved
have to be appropriately selected) but the formula
x12 x2a = (x1 x2)a, x1, x2, a  R
has no universally complex generalization.
8.4. Branches, Branch Points and Branch Cuts. We recall that a multifunction w defined on a
set S   is an assignment to each z S of a set [w(z)] of complex numbers. Our main aim is
that given a multifunction w defined on S, can we select, for each z S, a point f(z) in [w(z)] so
that f(z) is analytic in an open subset G of S, where G is to be chosen as large as possible? If we
are to do this, then f(z) must vary continuously with z in G, since an analytic function is
necessarily continuous.
Suppose w is defined in some punctured disc D having centre a and radius R i.e. 0 < |z  a| < R
and that f(z)  [w(z)] is chosen so that f(z) is at least continuous on the circle  with centre a and
radius r (0 < r < R). As z traces out the circle  starting from, say z0, f(z) varies continuously, but
must be restored to its original value f(z0) when z completes its circuit, since f(z) is, by
hypothesis, single valued. Notice also that if z  a = r ei(z), where (z) is chosen to vary
continuously with z, then (z) increases by 2 as z performs its circuit, so that (z) is not
restored to its original value. The same phenomenon does not occur if z moves round a circle in
the punctured disc D not containing a, in this case (z) does return to its original value. More
generally, our discussion suggests that if we are to extract an analytic function from a multi-
function w, we shall meet to restrict to a set in which it is impossible to encircle, one at a time,
points a such that the definition of [w(z)] involves the argument of (za). In some cases,
encircling several of these „bad‟ points simultaneously may be allowable.
A branch of a multiple-valued function f(z) defined on S   is any single-valued function F(z)
which is analytic in some domain D S at each point of which the value F(z) is one of the values
of f(z). The requirement of analyticity, of course, prevents F(z) from taking on a random
selection of the values of f(z).
A branch cut is a portion of a line or curve that is introduced in order to define a branch F(z) of a
multiple-valued function f(z).
A multivalued function f(z) defined on S   is said to have a branch point at z0 when z
describes an arbitrary small circle about z0, then for every branch F(z) of f(z), F(z) does not
return to its original value. Points on the branch cut for F(z) are singular points of F(z) and any
point that is common to all branch cuts of f(z) is called a branch point. For example, let us
consider the logarithmic function
log z = log r + i = log |z| + i arg z (1)
If we let  denote any real number and restrict the values of  in (1) to the interval  <  <  +
2, then the function
log z = log r + i (r > 0,  <  <  + 2) (2)
with component functions
u(r, ) = log r and v(r, ) =  (3)
is single-valued, continuous and analytic function. Thus for each fixed , the function (2) is a
branch of the function (1). We note that if the function (2) were to be defined on the ray  = , it
80

would not be continuous there. For, if z is any point on that ray, there are points arbitrarily close
to z at which the values of v are near to  and also points such that the values of v are near to  +
2. The origin and the ray  =  make up the branch cut for the branch (2) of the logarithmic
function. The function

Log z = log r + i (r > 0,  <  < ) (4)

is called the principal branch of the logarithmic function in which the branch cut consists of the
origin and the ray  = . The origin is evidently a branch point of the logarithmic function.


O
x

For analyticity of (2), we observe that the first order partial derivatives of u and v are continuous
and satisfy the polar form
1 1
ur = v  , vr   u 
r r
of the CR equations. Further
d
(log z) = ei (ur + ivr)
dz
1  1
= ei   i 0   iθ
r  re
d 1
Thus (log z) = (|z| = r > 0,  < arg z <  + 2)
dz z
In particular
d 1
(log z) = (|z| > 0,  < Arg z < ).
dz z
1
Further, since log = log z,  is also a branch point of log z. Thus a cut along any half-line
z
from 0 to  will serve as a branch cut.
Now, let us consider the function w = za in which a is an arbitrary complex number. We can
write
 COMPLEX ANALYSIS 81

a
w = za = e log z = ea log z (5)
where many-valued nature of log z results is many-valuedness of za. If Log z denotes a definite
branch, say the principal value of log z, then the various values of z a will be of the form
za = ea(Log z + 2ni) = ea Log z e2 i a n (6)
where log z = Log z + 2ni, nI.
The function in (6) has infinitely many different values. But the number of different values of za
will be finite in the cases in which only a finite number of the values e2 ian, nI, are different
from one another. In such a case, there must exist two integers m and m (m = m) such that e2
iam
= e2 iam
or e2 ia(m m) = 1. Since ez = 1 only if z = 2in, thus we get a (mm) = n and
therefore it follows that a is a rational number. Thus z a has a finite set of values iff a is a rational
number. If a is not rational, za has infinity of values.
We have observed that if z = rei and  is any real number, then the branch
log z = log r + i (r > 0,  <  <  + 2) (7)
of the logarithmic function is single-valued and analytic in the indicated domain. When this
branch is used, it follows that the function (5) is single valued and analytic in the said domain.
The derivative of such a branch is obtained as
d a d a
(z )  [exp (a log z)] = exp ( a log z)
dz dz z
exp (a log z)
=a = a exp [(a 1) log z]
exp (log z)
= aza1.
As a particular case, we consider the multivalued function f(z) = z1/2 and we define
z1/2 = r ei/2, r > 0,  <  <  + 2 (8)
where the component functions
u(r, ) = r cos /2, v(r, ) = r sin /2 (9)
are single valued and continuous in the indicated domain. The function is not continuous on the
line  =  as there are points arbitrarily close to z at which the values of v (r, ) are nearer to
r sin /2 and also points such that the values of v(r, ) are nearer to  r sin /2. The function
(8) is differentiable as C-R equations in polar form are satisfied by the functions in (9) and
d 1/ 2  1 1 
(z )  e iθ (u r  iv r )  e iθ  cosθ / 2  i sin θ / 2 
dz 2 r 2 r 
1 iθ / 2 1
= e  1/ 2
2 r 2z
Thus (8) is a branch of the function f(z) = z1/2 where the origin and the line  =  form branch
cut. When moving from any point z = rei about the origin, one complete circuit to reach again,
at z, we have changed arg z by 2. For original position z = rei, we have w = r ei/2, and after
one complete circuit, w = r ei(+2)/2 =  r ei/2. Thus w has not returned to its original value
and hence change in branch has occurred. Since a complete circuit about z = 0 changed the
branch of the function, z = 0 is a branch point for the function z 1/2.
82

UNIT – III

1. Transformations
Here, we shall study how various curves and regions are mapped by elementary analytic
function. We shall work in  i.e. the extended complex plane. We start with the linear
function.
w = Az …(1)
where A is non zero complex constant and z  0. We write A and z in exponential form as
A = aei, z = rei
i( + )
Then w = (ar) e …(2)
Thus we observe from (2) that transformation (1) expands (or contracts) the radius vector
representing z by the factor a = | A | and rotates it through an angle  = arg A about the origin.
The image of a given region is, therefore, geometrically similar to that region. The general linear
transformation
w = Az + B …(3)
is evidently an expansion or contraction and a rotation, followed by a translation. The image
region mapped by (3) is geometrically congruent to the original one.
Now we consider the function
1
w= …(4)
z
which establishes a one to one correspondence between the non zero points of the z-plane and
the w-plane. Since z z = | z |2, the mapping can be described by means of the successive
transformations
1
Z= z, w= Z …(5)
| z |2
Geometrically, we know that if P and Q are inverse points w.r.t. a circle of radius r with centre
A, then
(AP) (AQ) = r2
Thus  and  are inverse points w.r.t. the circle | z  a | = r if
(  a) (  a ) = r2


 Q
a
P
r A

where the pair  = a,  =  is also included. We note that , , a are collinear. Also points 
and  are inverse w.r.t a straight line l if  is the reflection of  in l and conversely. Thus the
first of the transformation in (5) is an inversion w.r.t the unit circle | z | = 1 i.e. the image of a
non zero point z is the point Z with the properties
1
z Z = 1, | Z | = and arg Z = arg z
|z|
 COMPLEX ANALYSIS 83

Thus the point exterior to the circle | z | = 1 are mapped onto the non zero points interior to it and
conversely. Any point on the circle is mapped onto itself. The second of the transformation in
(5) is simply a reflection in the real axis.

y z

Z
x
O
w
|z| =
1
1 1
Since lim =  and lim = 0,
z0 z z  z
it is natural to define a one-one transformation w = T(z) from the extended z plane onto the
extended w plane by writing
T(0) = , T() = 0
1
and T(z) =
z
for the remaining values of z. It is observed that T is continuous throughout the extended z
plane.
When a point w = u + iv is the image of a non zero point z = x + iy under the transformation
1 z
w = , writing w = results in
z | z |2
x y
u= , v= 2 …(6)
x 2  y2 x  y2
1 w
Also, since z=  , we get
w | w |2
u v
x= , y= …(7)
u 2  v2 u 2  v2
The following argument, based on these relations (6) and (7) between co-ordinates shows the
1
important result that the mapping w = transforms circles and lines into circles and lines.
z
When a, b, c, d are real numbers satisfying the condition b2 + c2 > 4ad, then the equation
a(x2 + y2) + bx + cy + d = 0 …(8)
represents an arbitrary circle or line, where a  0 for a circle and a = 0 for a line.
If x and y satisfy equation (8), we can use relations (7) to substitute for these variables.
Thus using (7) in (8), we obtain
d(u2 + v2) + bu  cv + a = 0 …(9)
which also represents a circle or a line. Conversely, if u and v satisfy (9), it follows from (6) that
x and y satisfy (8). From (8) and (9), it is clear that
(i) a circle (a  0) not passing through the origin (d  0) in the z plane is transformed into a
circle not passing through the origin in the w plane.
84

(ii) a circle (a  0) through the origin (d = 0) in the z plane is transformed into a line which does
not pass through the origin in the w plane.
(iii) a line (a = 0) not passing through the origin (d  0) in the z plane is transformed into a circle
through the origin in the w plane.
(iv) a line (a = 0) through the origin (d = 0) in the z plane is transformed into a line through the
origin in the w plane.
1
Hence we conclude that w = transforms circles and lines into circles and lines.
z
Remark : In the extended complex plane, a line may be treated as a circle with infinite radius.
1.1. Bilinear Transformation. The transformation
az  b
w= , ad  bc  0 …(1)
cz  d
where a, b, c, d are complex constants, is called bilinear transformation or a linear fractional
transformation or Möbius transformation. We observe that the condition ad  bc  0 is necessary
for (1) to be a bilinear transformation, since if
b d
ad bc = 0, then  and we get
a c

a (z  b / a ) a
w=  i.e. we get a constant function which is not
c(z  d / c) c
linear.
Equation (1) can be written in the form
cwz + dw  az  b = 0 …(2)
Since (2) is linear in z and linear in w or bilinear in z and w, therefore (1) is termed as
bilinear transformation.
When c = 0, the condition ad  bc  0 becomes ad  0 and we see that the transformation
reduces to general linear transformation. When c  0, equation (1) can be written as
a (z  b / a ) a  b / a  d / c 
w=  1
c (z  d / c) c  z  d / c 
a bc  ad 1
=   …(3)
c c 2
z  d/c
We note that (3) is a composition of the mappings
d 1 bc  ad
z1 = z + , z2 = , z3 = z2
c z1 c2
a
and thus we get w = + z3 .
c
The above three auxiliary transformations are of the form
1
w = z + , w= , w = z …(4)
z
Hence every bilinear transformation is the resultant of the transformations in (4).
But we have already discussed these transformations and thus we conclude that a bilinear
transformation always transforms circles and lines into circles and lines because the
transformations in (4) do so.
 COMPLEX ANALYSIS 85

From (1), we observe that if c = 0, a, d  0, each point in the w plane is the image of one and
d
only one point in the z-plane. The same is true if c  0, except when z =  which makes the
c
d
denominator zero. Since we work in extended complex plane, so in case z =  , w =  and
c
d
thus we may regard the point at infinity in the w-plane as corresponding to the point z =  in
c
the z-plane.
Thus if we write
az  b
T(z) = w = , ad  bc  0 …(5)
cz  d
Then T() = , if c = 0
T() = , T    = , if c  0
a d
and
c  c
Thus T is continuos on the extended z-plane. When the domain of definition is enlarged in this
way, the bilinear transformation (5) is one-one mapping of the extended z-plane onto the
extended w-plane.
Hence, associated with the transformation T, there is an inverse transformation T 1 which is
defined on the extended w-plane as
T1(w) = z if and only if T(z) = w.
Thus, when we solve equation (1) for z, then
 dw  b
z= , ad  bc  0 …(6)
cw  a
and thus
 dw  b
T1(w) = z = , ad  bc  0
cw  a
1
Evidently T is itself a bilinear transformation, where
T1() =  if c = 0
T1   = , T1 () =  , if c  0
a d
and
c c
From the above discussion, we conclude that inverse of a bilinear transformation is bilinear. The
d a
points z =  (w = ) and z = (w = ) are called critical points.
c c
1.2. Theorem. Composition (or resultant or product) of two bilinear transformations is a
bilinear transformation.
Proof. We consider the bilinear transformations
az  b
w= , ad  bc  0 …(1)
cz  d
a w  b1
and w1 = 1 , a1d1  b1c1  0 …(2)
c1 w  d 1
Putting the value of w from (1) in (2), we get
86

az  b 
a 1    b 1 (a a  b c) z  ( b d  a b )
 cz  d   1 1 1 1
w1 =
 az  b  (c 1 a  d 1 c) z  (d 1 d  c 1 b )
c1    d1
 cz  d 
Taking A = a1a + b1c, B = b1d + a1b,
C = c1a + d1c, D = d1d + c1b, we get
Az  B
w1 =
Cz  D
Also AD  BC = (a1a + b1c) (d1d + c1b)  (b1d + a1b) (c1a + d1c)
= (a1ad1d + a1ac1b + b1cd1d + b1cc1b)
 (b1dc1a + b1d d1c + a1bc1a + a1bd1c)
= a1ad1d + b1bc1c  b1dc1a  a1bd1c
= ad(a1d1  b1c1)  bc(a1d1  b1c1)
= (ad  bc) (a1d1  b1c1)  0
Az  B
Thus w1 = , AD  BC  0
Cz  D
Is a bilinear transformation.
This bilinear transformation is called the resultant (or product or composition) of the
bilinear transformations (1) and (2).
The above property is also expressed by saying that bilinear transformations form a
group.
1.3. Definitions. (i) The points which coincide with their transforms under bilinear
transformation are called its fixed points. For the bilinear transformation w = az  b , fixed
cz  d
points are given by w = z i.e. z = az  b …(1)
cz  d
Since (1) is a quadratic in z and has in general two different roots, therefore there are
generally two invariant points for a bilinear transformation.
(ii) If z1, z2, z3, z4 are any distinct points in the z-plane, then the ratio
(z  z 2 )(z 3  z 4 )
(z1, z2, z3, z4) = 1
(z 2  z 3 )(z 4  z1 )
is called cross ratio of the four points z1, z2, z3, z4. This ratio is invariant under a bilinear
transformation i.e.
(w1, w2, w3, w4) = (z1, z2, z3, z4)
1.4. Transformation of a Circle. First we show that if p and q are two given points and K is a
constant, then the equation
zp
= K, (1)
zq
represents a circle. For this, we have
|z  p|2 = K2|z  q|2
 (z  p) ( z  p ) = K2(z  q) ( z  q )
 (z  p) ( z  p ) = K2(z  q) ( z  q )
 zz  pz  p z  pp = K2(z z  qz  qz  qq )
 COMPLEX ANALYSIS 87

 (1  K2) z z (p  qK2) z  ( p  q K2)z = K2q q  pp

 p  qK 2   p  qK 2  | p |2  K 2 | q |2
 z z   2 
z  
 2 
z  =0 …(2)
 1 K   1 K  1 K2
Equation (2) is of the form
zz  bz  bz  c = 0 (c is being a real constant)
which always represents a circle.
Thus equation (2) represents a circle if K  1.
If K = 1, then it represents a straight line
| z  p |= | z  q |
Further, we observe that in the form (1), p and q are inverse points w.r.t. the circle. For this, if
the circle is |z  z0| =  and p and q are inverse points w.r.t. it, then
z  z0 = ei, p  z0 = qei,
 2 i
q  z0 = e
a
Therefore,

zp e i  ae i a e i  ae i
= 
zq  2
 ae i  e i
e i  e i
a

(cos   i sin )  a (cos   i sin ) a

=K , K
a (cos   i sin )  (cos   i sin ) 
( cos  a cos)  i( sin   a sin )
=K
(a cos   cos)  i(a sin    sin )
1/ 2
 ( cos   a cos ) 2  ( sin   a sin ) 2 
=K  2 
 (a cos    cos )  (a sin    sin ) 
2

= K, where K  1, since a  
Thus, if p and q are inverse points w.r.t. a circle, then its equation can be written as
zp
= K, K  1, K being a real constant.
zq
1.5 Theorem. In a bilinear transformation, a circle transforms into a circle and inverse points
transform into inverse points. In the particular case in which the circle becomes a straight line,
inverse points become points symmetric about the line.
zp
Proof : We know that = K represents a circle in the z-plane with p and q as inverse
zq
points, where K  1. Let the bilinear transformation be
az  b dw  b
w= so that z=
cz  d  cw  a
Then under this bilinear transformation, the circle transforms into
88

dw  b
p dw  b  p(q  cw)
 cw  a =K  =K
dw  b dw  b  q(a  cw)
q
 cw  a
ap  b
w
w (d  cp)  (ap  b) cp  d cq  d
 =K  =K …(1)
w (d  cq)  (aq  b) aq  b cp  d
w
cq  d
The form of equation (1) shows that it represents a circle in the w-plane whose inverse points are
ap  b aq  b
and . Thus, a circle in the z-plane transforms into a circle in the w-plane and the
cp  d cq  d
inverse points transform into the inverse points.
cq  d
Also if K = 1, then equation (1) represents a straight line bisecting at right angle the join
cp  d
ap  b aq  b
of the points and so that these points are symmetric about this line. Thus in a
cp  d cq  d
particular case, a circle in the z-plane transforms into a straight line in the w-plane and the
inverse points transform into points symmetrical about the line.
1.6. Example. Find all bilinear transformations of the half plane Im z  0 into the unit circle
| w |  1.
Solution. We know that two points z, z , symmetrical about the real z-axis(Im z = 0) correspond
1 1 Comment [a1]:
to points w, , inverse w.r.t. the unit w-circle. (|w| = 1). In particular, the origin and the
w |w|
point at infinity in the w-plane correspond to conjugate values of z.
Let
az  b a( z  b / a)
w= = …(1)
cz  d c (z  d / c)
be the required transformation.
Clearly c  0, otherwise points at  in the two planes would correspond.
Also, w = 0 and w =  are the inverse points w.r.t. | w | = 1. Since in (1), w = 0, w = 
b d
correspond respectively to z =  , z =  , therefore these two values of z-plane must be
a c
conjugate to each other. Hence we may write
b d
 = ,  =  so that
a c
a z
w = (2)
c z
The point z = 0 on the boundary of the half plane Im z  0 must correspond to a point on
the boundary of the circle | w | = 1, so that
a 0 a
1=|w|= 
c 0 c
 COMPLEX ANALYSIS 89

a
 = ei  a = cei, where  is real.
c
Thus, we get
z
w = ei   …(3)
z
Since z =  gives w = 0,  must be a point of the upper half plane i.e. Im  > 0. With this
condition, (3) gives the required transformation. In (3), if z is real, obviously | w | = 1 and if
Im z > 0, then z is nearer to  than to  and so | w | < 1. Hence the general linear
transformation of the half plane Im z  0 on the circle | w |  1 is
z
w = ei   , Im  > 0.
z
1.7. Example. Find all bilinear transformations of the unit | z |  1 into the unit circle | w |  1.
OR
Find the general homographic transformations which leaves the unit circle invariant.
Solution. Let the required transformation be
az  b a (z  b / a )
w=  …(1)
cz  d c (z  d / c)
Here, w = 0 and w = , correspond to inverse points
b d
z= , z= , so we may write
a c
b d 1
 = ,  = such that |  | < 1.
a c 
a  z    a  z   
So, w=     …(2)
c  z  1/   c  z  1 
The point z = 1 on the boundary of the unit circle in z-plane must correspond to a point on the
boundary of the unit circle in w-plane so that
a 1   a
1=|w|= 
c  1 c
or a  = c ei, where  is real.
Hence (2) becomes,
z 
w = ei  , ||<1 …(3)
 z 1

This is the required transformation, for if z = e ,  = be , then
i i

e i  be i
| w |= = 1.
be i ( )  1
If z = rei, where r < 1, then
|z  |2  |  z  1|2
= r2 2rb cos(  ) + b2  {b2r2  2br cos(  ) + 1}
= (r2  1)(1  b2) < 0
and so
| z   |2
|z  |2 < |  z  1|2  <1
|  z  1 |2
i.e. | w |< 1
Hence the result.
90

1.8. Example. Show that the general transformation of the circle | z |   into the circle
| w |   is
 z 
w =  ei  ,
2 
|  | < .
 z   
Solution. Let the transformation be
az  b a  z  b / a 
w=    …(1)
cz  d c  z  d / c 
The points w = 0 and w = , inverse points of | w | =  correspond to inverse point z = b/a,
z = d/c respectively of | z | = , so we may write
d 
2
b
 = ,   , ||<
a c 
Thus, from (1), we get
 
 
a  z    a  z   
w=    …(2)
c  2  c   z   2 
 z  
  
Equation (2) satisfied the condition | z |   and | w |  . Hence for | z | = , we must have
| w | =  so that (2) becomes
a z  
 = | w | = , z z = 2
c  z  zz
a 1 z   a 1 z  
= 
c z z c z z
a 1
= , | z   | = | z|
c 
a a
  =  =  ei,  being real.
c c
Thus, the required transformation becomes
 z 
w =  ei   , |  | < .
2 
 z   
1.9. Example. Find the bilinear transformation which maps the point 2, i, 2 onto the points 1,
i, 1.
Solution. Under the concept of cross-ratio, the required transformation is given by
( w  w 1 )(w 2  w 3 ) (z  z1 )(z 2  z 3 )

( w 1  w 2 )(w 3  w ) (z1  z 2 )(z 3  z)
Using the values of zi and wi , we get
( w  1)(i  1) (z  2)(i  2)
=
(1  i)(1  w ) (2  i)(2  z)
w  1  z  2  2  i  1  i 
or =   
w  1  z  2  2  i  1  i 
w  1 4  3i z  2
or =
w 1 5 z2
 COMPLEX ANALYSIS 91

w 1 w 1 (4  3i)(z  2)  5(z  2)

or =
w  1  ( w  1) (4  3i)(z  2)  5(z  2)
3z(3  i)  2i(3  i) 3z  2i
or w = 
 iz(z  i)  6(3  i)  (iz  6)
3z  2i
or w=
iz  6
which is the required transformation.
1.10. Exercise
Find the bilinear transformation which maps
(i) 1, i, 2 onto 0, 2, i respectively.
(ii) 1, i, 0 onto 1, i, 1 respectively.
(iii) 0, 1,  onto , i, 1 respectively.
(iv) 1, , i into 0, , 1 respectively.
(v) , i, 0 onto 0, i,  respectively.
(vi) 1, 0, 1 onto i, , 1 respectively.
(vii) 1, i, 1 onto i, 0, i respectively.
i
Hint : For the terms like etc, we use
w
i in 
= lim | form
  w n  n  w 
0 1
= lim
n  1  0

= 1
Answers :
2(z  1) (i  2 ) z  1 z  (1  i)
(i) w = (ii) w = (iii) w =
(1  i)z  2 iz  1 z
z 1 1 (1  i)z  (i  1)
(iv) w = (v) w =  (vi) w =
1 i z 2z
zi
(vii) w =
zi
2. Conformal Mappings
Let S be a domain in a plane in which x and y are taken as rectangular Cartesian co-ordinates.
Let us suppose that the functions u(x, y) and v(x, y) are continuous and possess continuous
partial derivatives of the first order at each point of the domain S. The equations
u = u(x, y), v = v(x, y)
set up a correspondence between the points of S and the points of a set T in the (u, v) plane. The
set T is evidently a domain and is called a map of S. Moreover, since the first order partial
derivatives of u and v are continuous, a curve in S which has a continuously turning tangent is
mapped on a curve with the same property in T. The correspondence between the two domains
is not, however, necessarily a one-one correspondence.
For example, if we take u = x2, v = y2, then the domain x2 + y2 < 1 is mapped on the triangle
bounded by u = 0, v = 0, u + v = 1, but there are four points of the circle corresponding to each
point of the triangle.
92

2.1 Definition : A mapping from S to T is said to be isogonal if it has a one-one transformation

which maps any two intersecting curves of S into two curves of T which cut at the same angle.
Thus in an isogonal mapping, only the magnitude of angle is preserved.
An isogonal transformation which also conserves the sense of rotation is called conformal
mapping. Thus in a conformal transformation, the sense of rotation as well as the
magnitude of the angle is preserved.
The following theorem provides the necessary condition of conformality which briefly
states that if f(z) is analytic, mapping is conformal.
2.2. Theorem : Prove that at each point z of a domain D where f(z) is analytic and f (z)  0, the
mapping w = f(z) is conformal.
Proof. Let w = f(z) be an analytic function of z, regular and one valued in a region D of the
z-plane. Let z0 be an interior point of D and let C1 and C2 be two continuous curves passing
through z0 and having definite tangents at this point, making angles 1, 2, say, with the real
axis.
We have to discover what is the representation of this figure in the w-plane. Let z1 and z2 be
points on the curves C1 and C2 near to z0. We shall suppose that they are at the same distance r
from z0, so we can write
z1  z0 = rei1, z2  z0 = rei2.
Then as r  0, 1  1, 2  2. The point z0 corresponds to a point w0 in the w-plane and z1
and z2 correspond to point w1 and w2 which describe curves C1 and C2, making angles 1 and 2
with the real axis.

C2
y
C1' C '2
v
C1 z w1
1
z0 w0
w2
z2
tangent
2 1 2 1
x u

Let w1  w0 = 1 ei1 , w2  w0 = 2 e i2 ,

where 1, 2  0  1, 2  1, 2
respectively.
Now, by the definition of an analytic function,
w  w0
lim 1 = f (z0)
z z 0 z  z
1 0

Since f (z0)  0, we may write it in the form Rei and thus

1e i1 
lim = Rei i.e. lim 1 e i ( 1 1 ) = Rei
rei1 r
1
 lim = R = |f (z0) |
r
 COMPLEX ANALYSIS 93

and lim (1  1) = 

i.e. lim 1  lim 1 = 
i.e. 1  1 =   1 = 1 + 
Similarly, 2 = 2 +  .
Hence the curves C1 and C2 have definite tangents at w0 making angles 1 +  and 2 + 
respectively with the real axis. The angle between C1 and C2 is
1  2 = (1 + )  (2  ) = 1  2
which is the same as the angle between C1 and C2. Hence the curve C1 and C2 intersect at the
same angle as the curves C1 and C2. Also the angle between the curves has the same sense in the
two figures. So the mapping is conformal.
Special Case : When f (z0) = 0, we suppose that f (z) has a zero of order n at the point z0.
Then in the neighbourhood of this point (by Taylor‟s theorem)
f(z) = f(z0) + a(z  z0)n+1 + …, where a  0
Hence w1  w0 = a(z  z0)n+1 + ….
i.e. 1 ei1 = | a | rn+1 ei[ + (n +1)1] + …
where  = arg a
Hence lim 1 = lim [ + (n + 1) 1] =  + (n + 1) 1 |  is constant
Similarly lim 2 =  + (n + 1) 2
Thus the curves C1 and C2 still have definite tangent at w0, but the angel between the tangents is
lim(2  1) = (n + 1) (2  1)
Thus, the angle is magnified by (n + 1).
 
Also the linear magnification, R = lim 1 = 0 | lim 1 = R = |f (z0)| = 0
r r
Therefore, the conformal property does not hold at such points where f (z) = 0
A point z0 at which f (z0) = 0 is called a critical point of the mapping. The following theorem is
the converse of the above theorem and is sufficient condition for the mapping to be conformal.
2.3. Theorem : If the mapping w = f(z) is conformal then show that f(z) is an analytic function
of z.
Proof. Let w = f(z) = u(x, y) + iv(x, y)
Here, u = u(x, y) and v = v(x, y) are continuously differentiable equations defining conformal
transformation from z-plane to w-plane. Let ds and d be the length elements in z-plane and
w-plane respectively so that
ds2 = dx2 + dy2, d2 = du2 + dv2 …(1)
Since u, v are functions of x and y, therefore
u u v v
du = dx + dy, dv = dx  dy
x y x y

2 2
 u u   v v 
 du2 + dv2 =  dx  dy    dx  dy 
 x y   x y 
 u 2
v 
2
  u   v  
2 2

i.e. d2 =       dx 2        dy 2

 x   x   y y
    
94

 u u v v 
+ 2  dxdy …(2)
 x y x y 
Since the mapping is given to be conformal, therefore the ratio d2 : ds2 is independent of
direction, so that from (1) and (2), comparing the coefficients, we get
2 2
 u    v 
2 2
 u   v  u u v v
        
 x   x    y   y   x y x y
1 1 0
2 2
 u    v  =  u    v 
2 2
         …(3)
 x   x   y   y 
u u v v
and  =0 …(4)
x y x y
Equations (3) and (4) are satisfied if
u v v u
 ,  …(5)
x y x y
u v v u
or  ,  …(6)
x y x y
Equation (6) reduces to (5) if we replace v by v i.e. by taking as image figure obtained by the
reflection in the real axis of the w-plane.
Thus the four partial derivatives ux, uy, vx, vy exist, are continuous and they satisfy C-R equations
(5). Hence f(z) is analytic.
2.4. Remarks
(i) The mapping w = f(z) is conformal in a domain D if it is conformal at each point of the
domain.
(ii) The conformal mappings play an important role in the study of various physical
phenomena defined on domains and curves of arbitrary shapes. Smaller portions of these
domains and curves are conformally mapped by analytic function to well-known domains
and curves.
2.5. Example : Discuss the mapping w = z .
Solution. We observe that the given mapping replaces every point by its reflection in the real
axis. Hence angles are conserved but their signs are changed and thus the mapping is isogonal
but not conformal. If the mapping w = z is followed by a conformal transformation, then
resulting transformation of the form w = f( z ) is also isogonal but not conformal, where f(z) is
analytic function of z.
2.6. Example : Discuss the nature of the mapping w = z2 at the point z = 1 + i and examine its
effect on the lines Im z = Re z and Re z = 1 passing through that point.
Solution. We note that the argument of the derivative of f(z) = z2 at z = 1 + i is
[arg 2z]z = 1+ i = arg(2 + 2i) = /4
Hence the tangent to each curve through z = 1 + i will be turned by the angle /4. The
co-efficient of linear magnification is |f (z)| at z = 1 + i, i..e |2 + 2i| = 2 2 . The mapping is
w = z2 = x2  y2 + 2ixy = u(x, y) + iv(x, y)
We observe that mapping is conformal at the point z = 1 + i, where the half lines y = x(y  0)
and x = 1(y  0) intersect. We denote these half lines by C1 and C2, with positive sense upwards
and observe that the angle from C1 to C2 is /4 at their point of intersection. We have
 COMPLEX ANALYSIS 95

u = x2  y2, v = 2xy
The half line C1 is transformed into the curve C1 given by
u = 0, v = 2y2 (y  0)
Thus C1 is the upper half v  0 of the v-axis.
The half line C2 is transformed into the curve C2 represented by
u = 1  y2, v = 2y (y  0)
Hence C2 is the upper half of the parabola v2 = 4(u  1). We note that, in each case, the
positive sense of the image curve is upward.
For the image curve C2 ,
dv dv / dy 2 2
  
du du / dy  2 y v
dv
In particular = 1 when v = 2. Consequently, the angle from the image curve C1 to the
du

image curve C2 at the point w = f(1 + i) = 2i is , as required by the conformality of the
4
mapping there.

'
y C2 v C1
C1
/
'
/4 C2 4 2i

1+i

/2 C3 /2 C 3'

O 1 x O 1 u

Note. The angle of rotation and the scalar factor (linear magnification) can change from point to
point. We note that they are 0 and 2 respectively, at the point z = 1, since f (1) = 2, where the
curves C2 and C2 are the same as above and the non-negative x-axis (C3) is transformed into the
non-negative u-axis (C3).
2.7. Example. Discuss the mapping w = za, where a is a positive real number.
Solution. Denoting z and w in polar as
z = rei, w = ei, the mapping gives  = ra,  = a.
Thus the radii vectors are raised to the power a and the angles with vertices at the origin are
multiplied by the factor a. If a > 1, distinct lines through the origin in the z-plane are not mapped
onto distinct lines through the origin in the w-plane, since, e.g. the straight line through the origin
2
at an angle to the real axis of the z-plane is mapped onto a line through the origin in the
a
dw
w-plane at an angle 2 to the real axis i.e. the positive real axis itself. Further = aza1,
dz
which vanishes at the origin if a > 1 and has a singularity at the origin if a < 1. Hence the
mapping is conformal and the angles are therefore preserved, excepting at the origin. Similarly
the mapping w = ez is conformal.
96


2.8. Example. Prove that the quadrant | z | < 1, 0 < arg z < is mapped conformally onto a
2
4
domain in the w-plane by the transformation w = .
(z  1) 2
4
Solution. If w = f(z) = , then f  (z) is finite and does not vanish in the given quadrant.
(z  1) 2
Hence the mapping w = f(z) is conformal and the quadrant is mapped onto a domain in the w-
plane provided w does not assume any value twice i.e. distinct points of the quadrant are mapped
to distinct points of the w-plane. We show that this indeed is true. If possible, let
4 4
= , where z1  z2 and both z1 and z2 belong to the quadrant in the z-plane.
(z1  1) 2
(z 2  1) 2
Then, since z1  z2, we have (z1  z2) (z1 + z2 + 2) = 0
 z1 + z2 + 2 = 0 i.e. z1 = z2  2. But since z2 belongs to the quadrant, z2  2 does not, which
contradicts the assumption that z1 belongs to the quadrant. Hence w does not assume any value
twice.
3. Space of Analytic Functions
We start with the following definition
3.1. Definition. A metric space is a pair (X, d) where X is a set and d is a function from X 
X into R, called the distance function or metric, which satisfy the following conditions for x,
y, zX
(i) d(x, y)  0
(ii) d(x, y) = 0 if x = y
(iii) d(x, y) = d(y, x)
(iv) d(x, z)  d(x, y) +d (y, z)
Conditions (iii) and (iv) are called ‘symmetry’ and ‘triangle inequality’ respectively. A
metric space (X, d) is said to be bounded if there exists a positive number K such that
d(x, y)  K for all x, y  X .
The metric space (X, d), in short, is also denoted by X, the metric being understood. If x
and r > 0 are fixed then let us define
B(x; r) = {x  X : d(x, y) < r}
B(x; r) = {y  X : d(x, y)  r}
B(x; r) and B(x; r) are called open and closed balls (spheres) respectively, with centre x
and radius r. B(x; ) is also referred to as the -neighbourhood of x.
Let X = R or  and define d(z, w) = |zw| . This makes both (R,d) and (, d) metric
spaces. (, d) is the case of principal interest for us. In ( , d), open and closed balls are
termed as open and closed discs respectively.
A metric space (X, d) is said to be complete if every sequence in X converges to a point of X,
R and  are examples of complete metric spaces.
If G is an open set in  and (X, d) is complete metric space then the set of all continuous
functions from G to X is denoted by C(G, X).
The set C(G, X) is always non empty as it contains the constant functions. However it is
possible that C(G, X) contains only the constant functions. For example, suppose that G is
connected and X = N = {1, 2, 3, 4,..}. If f  C(G, X) then f(G) must be connected in X and
hence, must be singleton as the only connected subsets of N are singleton sets.
 COMPLEX ANALYSIS 97

In this section we shall be mainly concern with the case when X is either  or .
To put a metric on C(G, X), we need the following result.
3.2. Theorem : If G is open in  then there is a sequence {Kn} of compact subsets of G such

that G = Kn . Moreover, the sets Kn can be chosen to satisfy the following conditions :
n 1
(a) Kn  int Kn+1 ;
(b) K  G and K compact implied K  Kn for some n.
Now we define a metric on C(G, X).

Since G is open set in , we have G = K n where each Kn is compact and Kn  int Kn+1. For
n 1
n  N, we define
n (f, g) = sup {d(f(z), g(z)) : z  Kn}
for all functions f and g in C(G, X).
Also if we define
 n
 n (f , g )
  2 
1
(f, g) = for all f, g  C(G, X)
n 1 1   n (f , g )
then (C(G, X), ) is a metric space. In fact (C(G, X), ) is a complete metric space.
3.3. Definitions : A set   C(G, X) is normal if each sequence in  has a subsequence which
converges to a function f in C(G, X).
A set   C(G, X) is normal iff its closure is compact.
A set   C(G, X) is called equicontinuous at a point z0 in G iff for every  > 0 there is a  > 0
such that for |z  z0| < ,
d(f(z), f(z0)) < 
for every f in .
 is said to be equicontinuous over a set E  G if for every  > 0 there is a  > 0 such that for z
and z in E and |z  z| < , we have
d(f(z), f(z)) <  for all f in 
Notice that if  consists of a single function f then the statement that  is equicontinuous at z0 is
only the statement that f is continuous at z0. The important thing about equicontinuity is that the
same  will work for all the functions in . Also for  = {f} to be equicontinuous over E is
equivalent to the uniform continuity of f on E.
Further, suppose   C(G, X) is equicontinuous at each point of G then  is
equicontinuous over each compact subset of G.
3.4 Arzela-Ascoli Theorem : A set   C(G, X) is normal iff the following two conditions are
satisfied :
(a) For each z in G, {f(z) ; f  F} has compact closure in X.
(b) F is equicontinuous at each point of G.
Let G be an open subset of the complex plane and H(G) be the collection of holomorphic
(analytic) functions on G.
The following theorem shows that H(G) is a closed subset of C(G, ).
98

3.5. Theorem. If {fn} is a sequence in H(G) and f belongs to C(G, ) such that fn  f then f is
analytic and fn(k)  f(k) for each integer k  1.
Proof. To show f is analytic on G, we shall use the following form of Morera‟s theorem which
states
“Let G be a region and let f : G   be a continuous function such that  f = 0 for every
T
triangular path T in G ; then f is analytic in G”.
Let T be a triangle contained inside a disk D  G. Since T is compact, {fn} converges to f
uniformly over T. Hence  f = lim  f n = 0, since each fn is analytic. Thus f must be analytic in
T T
every disk D  G. This gives that f is analytic in G.
To show that fn(k)  f(k), let D denote the closure of B(a, r) contained in G. Then there is a
number R > r such that B (a ; R)  G.
If  is the circle |z  a| = R then by Cauchy‟s integral formula,
k! f n ( w )  f ( w )
2i  ( w  z) k 1
fn(k)(z)  f(k)(z) = dw for z in D.

Let Mn = sup { |fn(w)  f(w)| : |w  a| = R}. Then by Cauchy‟s estimate, we have

k!M n R
|fn(k)z  f(k)(z)|  for | z  a|  r …(1)
(R  r ) k 1
Since fn  f, lim Mn = 0
Thus, it follows from (1) that fn(k)  f(k) uniformly on B (a ; R). Now let K be an arbitrary
compact subset of G and 0 < r < d(K, G) then there are a1, a2,…,an in K such that
n
K  B(a j ; r)
j1
Since fn(k)  f(k) uniformly on each B(aj ; r), it follows that fn(k)  f(k) uniformly on K, which
completes the proof of the theorem.
Cor. 1 : H(G) is a complete metric space.
Proof. Since C(G, ) is a complete metric space and H(G) is a closed subset of C(G, ), we get
that H(G) is also complete using “Let (X, d) be a complete metric space and Y  X. Then
(Y, d) is complete iff Y is closed in X”.

Cor. 2 : If fn : G   is analytic and fn 1
n (z) converges uniformly on compact sets to f(z) then

f(k)z = f
n 1
(k)
n (z)

3.6. Hurwitz’s Theorem. Let G be a region and suppose the sequence {fn} in H(G) converges
to f. If f  0, B (a ; R)  G and f(z)  0 for |z  a| = R then there is an integer N such that
for n  N, f and fn have the same number of zeros in B(a ; R).

Proof : Let  = inf {|f(z)| : |z  a| = R}

Since f(z)  0 for |z  a| = R, we have  > 0.
 COMPLEX ANALYSIS 99

Now fn  f uniformly on {z : |z  a| = R} so there is an integer N such that if n  N and

|z  a|= R then

|f(z) = fn(z)| < < |f(z)|
2
Hence by Rouche‟s theorem, f and fn have the same number of zeros in B(a; R).
Cor : If {fn}  H(G) converges to f in H(G) and each fn never vanishes on G then either f  0 or
f never vanishes.
3.7. Remark : Another form of Hurwitz‟s theorem is “Let {fn(z)} be a sequence of functions,
each analytic in a region D bounded by a simple closed contour and let f n(z)  f(z) uniformly in
D. Suppose that f(z) is not identically zero. Let z 0 be an interior point of D. Then z0 is a zero of
f(z) if and only if it is a limit point of the set of zeros of the functions f n(z), points which are
zeros of fn(z) for an infinity of values of n being counted as limit points.”
Proof. Let z0 be any point of D and let  be a circle with centre z0 and radius  so small that 
lies entirely in D. Suppose  neither contains nor has on it any zero of f(z) except possibly for
the point z0 itself. Then |f(z)| has a strictly positive lower bound on the circle |z  z0| = , say
|f(z)|  K > 0 …(1)
Having fixed  and K, we can choose N so large that, on the circle,
|fn(z)  f(z)| < K for all n > N …(2)
From (1) and (2), we get
|fn(z)  f(z)| < |f(z)|
Thus, if we set g(z) = fn(z)  f(z), then on the circle |z  z0| = , |g(z)| < |f(z)|
Hence by Rouche‟s theorem, for n > N, g(z) + f(z) i.e. fn(z) has the same number of zeros as f(z)
inside the circle  . Thus if f(z) = 0, then fn(z) has exactly one zero inside  for n > N.
Therefore, z0 is the limit point of the zeros of fn(z). If f(z0)  0, then fn(z0)  0 inside  for n > N
which completes the proof.
3.8. Definition : A set   H(G) is called locally bounded if for each point a in G there are
constants M and r > 0 such that for all f in  ,
|f(z)|  M for |z  a| < r.
Alternately,  is locally bounded if there is an r > 0 such that
sup {|f(z)| : |z  a| < r, f  } < .
That is,  is locally bounded if about each point a in G there is a disk on which  is uniformly
bounded.
3.9. Lemma : A set  in H(G) is locally bounded iff for each compact set K  G there is a
constant M such that
|f(z)|  M
for all f in  and z in K.
3.10. Montel’s Theorem : A family  in H(G) is normal iff  is locally bounded.
Proof : Suppose  is normal. We have to show  is locally bounded. Let, if possible, suppose
that  is not locally bounded. Then there is a compact set K  G such that
sup {|f(z)| : z  K, f  } = 
That is, there is a sequence {fn} in  such that
100

sup{|fn(z) : z  K}  n.
Since  is normal there is a function f in H(G) and a subsequence { f n k } such that f n k  f.
This gives lim sup( | f n k (z)  f(z) : z  K} = 0
k 

Let |f(z)|  M for z in K. Then

nk  sup{ | f n k (z)| : z  K}
= sup{ | f n k (z)  f(z) + f(z) | : z  K}
 sup{ | f n k (z)  f(z)| : z  K} + sup{ | f(z) | : z  K}
 nk  sup{| f n k (z)  f(z)| : z  K} + M
 lim nk  M .
k 
A contradiction since <nk > is strictly monotonically increasing sequence. Hence our supposition
is wrong. So  must be locally bounded.
Conversely suppose that  is locally bounded. Then for each z in G, {f(z) : f  } has compact
closure. We now show that  is equicontinuous at each point of G. Let a be any fixed point of
G and  > 0. By hypothesis, there is an r > 0 and M > 0 such that B (a ; r)  G and
|f(z)|  M for all z in B (a ; r) and for all f in  .
1
Let |z  a| < r and f  . Then by Cauchy‟s formula, with r(t) = a + reit, 0  t  2,
2
1 f ( w )(a  z)
|f(a)  f(z)| 
2  (w  a )(w  z)dw


4M
 |a  z|
r
Choose  straight line 0 <  , min  ,
r r 

 2 4M 
Then |a  z| <  gives
|f(a)  f(z)| <  for all f in  .
 F is equicontinuous at a  G.
Hence by Ascoli-Arzela theorem.  is normal.
Cor. A set   H(G) is compact iff it is closed and locally bounded.
3.11. Definition : A region G1 is called conformally equivalent to G2 if there is an analytic
function f : G1   such that f is one one and f(G1) = G2.
It is immediate that  is not equivalent to any bounded region by Liouville‟s theorem. Also it
follows from the definition that if G1 is simply connected and G1 is equivalent to G2 then G2
must be simply connected.
We now prove Riemann mapping theorem which states that every simply connected region
G in the plane (other than the plane itself) is conformally equivalent to the open unit disc
D. We shall use the following results.
Theorem (1) : Let G be an open connected subset of  . Then the following are equivalent.
(a) G is simply connected.
(b) For any f in H(G) such that f(z)  0 for all z in G, there is a function g in H(G) such that
f(z) = [g(z)]2.
 COMPLEX ANALYSIS101

Theorem (2) : Let f : D  D be a one one analytic function of D onto itself and suppose
f(a) = 0. Then there is a complex number c with | c | = 1 such that
za
f = ca where a(z) = with | a | < 1.
1  az
Theorem (3) (Open mapping theorem) : Let G be a region and suppose that f is a non-constant
analytic function on G. Then for any open set U in G, f(U) is open.
3.12. Riemann mapping theorem. Let G be a simply connected region which is not the whole
plane and let a  G. Then there is a unique analytic function f : G   having the properties :
(a) f(a) = 0 and f (a) > 0
(b) f is one one
(c) f(G) = {z : | z | < 1}
Proof : First we show f is unique.
Let g be another analytic function on  such that g(a) = 0, g(a) > 0 g is one one and
g(G) = {z : | z | < 1} = D.
Then f0g1 : D  D is analytic, one one and onto
Also f0g1(0) = f(a) = 0 . So there is a constant c with | c | = 1
and f0g1(z) = cz for all z. [Applying theorem (2) with a = 0]
But then f(z) = c g(z) gives that 0 < f (a) = cg(a).
Since g(a) > 0, it follows that c = 1. Hence f = g and so f is unique.
Now let  = {f  H(G) : f is one one, f(a) = 0, f (a) > 0, f(G)  D}
We first show    .
Since G   so there exists b   such that b  G
Also G is simply connected so there exists an analytic function g on G such that [g(z)]2 = z  b.
Then g is one-one
For this let z1, z2  G such that g(z1) = g(z2)
Then [g(z1)]2 = [g(z2)]2
 z1  b = z2  b
 z 1 = z2
 g is one-one.
So by open mapping theorem, there is a positive number r such that
B(g(a) ; r)  g(G) …(1)
Let z be a point in G such that g(z)  B(g(a) ; r)
Then | g(z) + g(a) | < r
 | g(z)  g(a) |< r
 g(z)  B(g(a) ; r)
 g(z)  g(G) [using (1)]
So  some w  G such that
g(z) = g(w)
 [g(z)]2 = [g(w)]2
 zb =wb
 z=w
Thus, we get,
g(z) = g(z)
 g(z) = 0
102

But then z  b = [g(z)]2 = 0 implies b = z  G, a contradiction.

Hence g(G)  B(g(a) ; r) = 
Let U = B(g(a) ; r). There is a Mobius transformation T such that
T(C  U ) = D
Let g1 = T0g then g1 is analytic and g1(G)  D.
g (z)  
Consider g2(z) = 1 where  = g1(a).
1   g 1 (z)
Then g2 is analytic, g2(G)  D and g2(a) = 0
Choose a complex number c, | c | = 1, such that
g3(z) = c g2(z) and g3(a) > 0
Now g3   hence   .
Next we assume that  =   {0} …(2)
Since f(G)  D, sup {| f(z) : z  G}  1 for f in . So by Montel’s theorem,  is normal.
This gives  is compact.
Consider the function  : H(G)  C
as (f) = f (a)
Then  is continuous function. Since  is compact, there is an f in  such that f (a)  g (a)
for all g   .
As   , (2) implies that f  . We show that f(G) = D. Suppose w  D such that w  f(G).
Then the function
f (z)  w
1  wf (z)
is analytic in G and never vanishes. Since G is simply connected, there is an analytic function
h : G   such that
f (z)  w
[h(z)]2 = …(3)
1  wf (z)
w
Since the Mobius transformation T  = maps D onto D,
1  w
we have h(G)  D.
Define g : G   as
| h (a ) | h (z)  h (a )
g(z) = .
h (a ) 1  h (a )h (z)
Then g(G)  D, g(a) = 0 and g is one-one.
| h (a ) | h (a )[1 | h (a ) |2 ] | h (a ) |
Also g (a) = . =

h (a ) [1 | h (a ) | ]
2 2
1 | h (a ) | 2
f (a )  w
But | h(a) |2 = = | w | = | w | [ f(a) = 0]
1  wf (a )
Differentiating (3), we get
2 h(a) h(a) = f (a) [1  | w |2]
f (a )(1 | w | 2 ) f (a )(1 | w | 2 )
 h(a) = 
2h (a ) 2 |w|
 COMPLEX ANALYSIS103

f (a )(1 | w |2 ) 1 f (a )(1 | w |)

 g(a) = .  > f  (a)
2 |w| (1 | w |) 2 |w|
Thus g  . A contradiction to the choice of f .
Hence we must have f(G) = D.
Next we prove  =   {0}.
Suppose {fn} is a sequence in  and fn  f in H(G).
Then f(a) = lim fn(a) = 0 Also fn(a)  f (a) so f  (a)  0
n 

Let z1 be an arbitrary element of G and let w = f(z1). Let wn = fn(z1). Let z2  G, z2  z1 and K
be a closed disk centred at z2 such that z1  K.
Then fn(z)  wn never vanishes on K since f is one one But fn(z)  wn converges uniformly to f(z)
 w on K as K is compact. So Hurwitz‟s theorem gives that f(z)  w never vanishes on K or
f(z) = w.
If f(z)  w on K then f is constant function throughout G and since f(a) = 0, we have
f(z)  0. Otherwise we have f is one. So f  can never vanish. This gives
f (a) > 0 [ f  a)  0]
and so f .
Hence  =   {0} which completes the proof of the theorem.
4. Factorization of an Integral Function
We know that a function which is regular in every finite region of the z-plane is called an
integral function or entire function. In other words, integral function is an analytic function
which has no singularity except at infinity.
z2
e.g. ez = 1 + z + +…
2
The simplest integral functions are polynomials. We know that a polynomial can be uniquely
expressed as the product of linear factors in the form :
 z  z   z 
f(z) = f(0) 1  1  ...1  
 z 1  z 2   z n 

where z1, z2,…, zn are the zeros of the polynomial.

An integral function which is not a polynomial may have an infinity of zeros zn and the product
 z 
 1   taken over these zeros may be divergent.
 zn 
So, a integral function cannot be always factorized in the same way as a polynomial and thus we
 z 
have to consider less simple factors than 1   . We observe that
 z n 

(a) An integral function may have no zero e.g. ez.

(b) An integral function may have finite number of zeroes e.g. polynomials of finite degree.
(c) An integral function may have infinite number of zeroes. e.g. sin z, cos z.
4.1 Theorem : The most general integral function with no zero is the form eg(z), where g(z) is
itself an integral function.
104

Proof : Let f(z) be an integral function with no zero, then f (z) is also an integral function and
f (z)
so is .
f (z)

f ( w )dw
z
Let F(z) = 
z0
f (w )
, where the integral is taken along any path from fixed point z 0 to a point z.

Thus f(z) = [log f(w) ] zz o = log f(z)  log f(z0)

 log f(z) = F(z) + log f(z0)
 f(z) = exp [log f(z0) + F(z)]
= eg(z), where g(z) = log f(z0) + F(z) is itself an integral function.
Hence the result.
4.2. Construction of an Integral Function with Given Zeros. If f(z) is an integral function
with only a finite number of zeros, say z1, z1,..,zn, then the function
f ( z)
(z  z1 )(z  z 2 )...( z  z n )
is an integral function with no zeros. Also we know that the most general form of an integral
function is eg(z), where g(z) is an integral function. Thus, we put
f ( z)
= eg(z)
(z  z1 )(z  z 2 )...( z  z n )
 f(z) = (z  z1) (z  z2)…(z  zn) eg(z)
If, however, f(z) is an integral function with an infinite number of zeros, then the only limit point
of the sequence of zeros, z1, z2, …,zn,… is the point at infinity. To determine an integral
function f(z) with an infinity of zeros, we have an important theorem due to Weierstrass.
4.3. Weierstrass Primary Factors. The expressions
E0(z) = 1  z
 z2 zp 
Ep(z) = (1  z) exp  z   ...   , p  1
 2 p
are called Weierstrass primary factors. Each primary factor is an integral function which has
only a simple zero at z = 1. Thus, Ep(z/a) has a simple zero at z = a and no other zero.
The behaviour of Ep(z) as z  0, depends upon p, since for | z | < 1, we have
  z2 z p 
Ep(z) = exp log(1  z)   z   ...  
  2 p 
 z2 z p z p1 z p 2   z2 z p 
= exp   z   ..    ..    z   ..  
 2 p p 1 p  2   2 p 
 z p 1
z p  2    z  n
= exp   ...  exp   

 p 1 p  2   n p1 n 
z p1 z p 2
 log Ep(z) =   …..
p 1 p  2
1
Hence if K > 1 and | z |  , then
K
| log Ep(z) |  | z |p+1 + | z |p+2 + …
 COMPLEX ANALYSIS105

= | z |p+1(1 + | z | + | z |2 + ….)
= | z |p+1 1   2  ... 
1 1
 K K 
K p+1
= |z|
K 1
This inequality helps in determining the convergence of a product of primary factors. In
1
particular, when | z |  , then
2
|log Ep(z) |  2 | z |p+1 …(1)
4.4. Theorem : If z1, z2,…,zn,… be any sequence of numbers whose only limit point is the
point at infinity, then it is possible to construct an integral function which vanishes at each of the
points zn and no where else.
Proof : Let the given zeros z1, z2,…,zn,… be arranged in order of non-decreasing modulus i.e.
| z1 |  | z2 |  ….  | zn |  …
Let | z n | = rn , n = 1, 2, …
 pn
r
and let p1, p2,… be a sequence of positive integers such that the series
n 1
  r
 is convergent
n 

for all values of r. It is always possible to find such a sequence since rn = | zn | is increasing
n
r  1
definitely with n and we can take pn = n, since    n for rn > 2r and hence the series is
 n
r 2
convergent.
Now, let
  z 
f(z) =  E pn  
n 1  zn 
This integral function is found to have the required property according to the specifications of
the theorem. To prove this, we observe that if | zn | > 2 | z |, then by the inequality (1) for
|log Ep(z) |, we have
p n 1 p n 1
 z  z r 
log E p n    2  2 
 zn  zn  rn 

 z 
and hence by Weierstrass‟s test, the series
n 1
 log E
  is uniformly convergent for
 zn 
pn

|zn| > 2R, |z|  R and also by Weierstrass‟s test for the uniform convergence of an infinite product
to be convergent, so is the product

 z 

n 1
E pn   , |zn| > 2R,
 zn 
|z|  R

 z 
Hence f(z) is regular for |z|  R and its only zeros in this region are those of  E pn   ,
n 1  zn 
|zn| > 2R i.e. the points z1, z2, …, zn, … Since R can be taken as large as we please, therefore we
conclude that f(z) is the integral function which vanishes only at z1, z2,..zn,... and nowhere else.
Hence the result.
Remarks : (i) The function f(z) is not uniquely determined, since we have a wide choice of the
sequence of positive integers p1, p2, …, pn,…
106

(ii) The most general integral function with the points z1, z2,…zn as its only zeros is f(z) eg(z),
where f(z) is the integral function constructed above and g(z) is an arbitrary integral
function.
4.5. Weierstrass Factorization Theorem. Let f(z) be an entire function and let {zn} be the
non-zero. Zeros of f(z) repeated according to multiplicity, suppose f(z) has a zero at z = 0 of
order m  0 (a zero of order m = 0 at z = 0 means f(0)  0). Then there is an entire function g(z)
and a sequence of integers {pn} such that

 z 
f(z) = zm eg(z)  E pn  
n 1  zn 
Proof : According to the preceding theorem {pn} can be chosen such that

 z 
h(z) = zm  E pn  
n 1  zn 
f (z)
has the same zeros as f(z) with the same multiplicity. It follows that has removable
h (z)
singularity at z = 0, z1, z2,… Thus f/h is an entire function and, furthermore, has no zeros. Since
 is simply connected, there is an entire function g(z) such that
f (z)
= eg(z)
h (z)

 z 
i.e. f(z) = zm eg(z) E
n 1
 .
pn
 zn 
4.6. Theorem : If f(z) is an integral function and f(0)  0, then f(z) = f(0) P(z) eg(z), where P(z)
is the product of primary factors and g(z) is an integral function.
Proof : We form P(z) from the zeros of f(z). Let
f (z) P (z)
(z) = 
f ( z) P( z)
Then (z) is an integral function, since the poles of one term are cancelled by those of the other.
Hence
z
g(z) = 
0
(t)dt = [log f(t)  log P( t )]0z

= log f(z)  log f(0)  log P(z) + log P(0)

 log f(z) = g(z) + log f(0) + log P(z)
 f(z) = f(0) P(z) eg(z)
Hence the result.
Remark : This factorization is not unique.
5. The Gamma Function
Here, we shall construct a function, called gamma function or Euler‟s gamma function which is
meromorphic with pole at non-positive integers i.e. z = 0, 1, 2,….
There are two natural approaches to construct the gamma function. One is via the
Weierstrass product and the other is via a Mellin integral. Certain properties are clear
from one definition but not from the other, although the two definitions give the same
function. We start with the former approach which involves more algebraic properties of
 COMPLEX ANALYSIS107

the gamma function. For this, we introduce functions which have only negative zeros. The
simplest function of this kind is

 z
G(z) =  1  e z / n …(1)
n 1  n
Obviously, the function G(z) has only positive zeros and moreover, from (1), we note that

 z 2  sin z
zG(z) G(z) = z  1  2   …(2)
n 1  n  
Now, zeros of

 z  1  ( z1) / n
G(z  1) =  1  e …(3)
n 1  n 
are given by z = 1  n, n = 1, 2, …
Thus G(z  1) has the same zeros as G(z), and in addition, a simple zero at z = 0. By Weierstrass
factorization theorem, we can write
G(z  1) = zg(z) G(z) …(4)
where g(z) is an integral function.
Using (1) and (2) in (3), we get
 
 z 1 n  zn
  n e (z1) / n = z eg(z)   n e z / n …(5)
n 1   n 1  
For determining g(z), we take logarithmic derivative of both sides of (5). This gives the equation

 

  z  1  n  n   z  g(z)    z  n  n 

1 1 1 1 1
…(6)
n 1 n 1
Replacing n by n + 1, the series on L.H.S. of (6) can be written as
 
 1  1   1 1  1  1 
 
n 1  z  n

n  1 z  
n 1  z  n

n  1
 
 1         
1 1 1 1 1 
= 
z n 1  z  n n  n 1  n n  1 
1   1 
  , where   
1 
 
1 1
=  =1
z n 1  z  n n  n 1  n n  1
Using this in (6), we conclude that g (z) = 0 and thus g(z) is constant which we denote by .
Hence (4) reduces to
G(z  1) = z e G(z) …(7)
To determine , we put z = 1 in (7) to have
G(0) = eG(1)

G(0) = 1, G(1) =  1  e 1 / n
1
But from (1),
n 1 
n

e =  1  e 1 / n
1
Hence, …(8)
n 1  n 
The nth partial product of this infinite product can be written in the form.
 1  ... 
1 1 1

(n + 1) e  2 3 n 
and we therefore deduce from (8) that
108

 = lim 1    ...   log n 

1 1 1
n   2 3 n 
This constant  is called Euler’s constant, and its approximate value is 0.577216.
5.1. Definition : The gamma function is defined by
1
(z) = z …(9)
ze G (z)
 1
e  z 1  z  e z / n
z 
=   …(10)
n 1 
n
We observe that (z) is well-defined in the whole complex plane except for z = 0, 1, 2,…
which are simple poles of the function. Hence (z) is meromorphic with these poles, but no
zeros.
5.2. Properties of Gamma Function. The following are the simple properties of gamma
function.
(i) Using (7) in (9), we get
e  ( z 1)  e  ( z 1) 
(z  1) = 
(z  1)G (z  1) (z  1)ze  G (z)
1  e  z  (z)
=  
z  1  zG (z)  z  1
i.e. (z) = (z  1) (z  1)
or (z + 1) = z (z) …(11)
(ii) Using (2), (9) and (11), we get
(1  z) (z) = z(z) (z)
1 1 1
= .
z e  z G(z) e z G(z)
1 
=  …(12)
zG (z)G(z) sin z
in (12), we get    =  . Then putting z =  in (11), we find
1 1 1
(iii) Putting z =
2 2 2

   =           = 2  .
1 1 1 1
2 2  2  2
(iv) For z = 1, (8) and (10) result in
 1
(1) = e  1   e1 / n = 1.
1
n 1 
n
Further, if n is a positive integer, then using (11) repeatedly, we get
(2) = 1. (1) = 1, (3) = 2 (2) = 2 =1. 2
(4) = 3(3) = 3.2 = 1.2.3 and so on.
Thus, finally, we get
(n + 1) = n
Thus gamma function can be considered as a generalization of the factorial function.
(v) Taking logarithm of both sides of (10) and then differentiating, we get
 COMPLEX ANALYSIS109

 (z) 1 
=        
1 1
(z) z n 1  z  n n 
Differentiating again, we get
d   (z) 
 

 
1 1 1
=  
dz  (z)  z 2
n 1 ( z  n ) 2
n 0 ( z  n) 2
Similarly, replacing z by 2z in (10), we find that
d  (2z)  
=2 
1
dz  (2z)  n  0 ( 2z  n )
2

Thus, we have
  1 
  z   
d   (z)  d   2
+  
dz  (z)  dz   1
 z  
  2  
 

 (z  n ) 
1 1
= 2 2
n 0 n 0 z  n  1 
 
 2
 

= 4  
1 1
2
 n 0 ( 2 z  2 n ) 2
n 0 ( 2 z  2 n  1) 

1
=4  |Even (2n) + odd (2n + 1) |
m 0 ( 2z  m)
2

d  (2z) 
=2
dz  (2z) 
  1 
  z   
d  (z) d  (2z) 
 
2
 2 
dz   2z) 
Thus +z
dz  (z)  1
   z  
  2  
which on integration gives
  z  
1
 (z)  2  (2z)
 2 +a
(z)  1  ( 2z )
 z  
 2
Integrating again, we get
log (z)   z   = log (2z) + az + b
1
 2
where a and b are arbitrary constants.
(z)   z   = eaz + b (2z)
1
Thus …(13)
 2
1
To determine a and b, we substitute z = and z = 1 so that we make use of the results
2
     , (1) = 1 to have
1
2
110

   (1) = e(a/2) + b (1), (1)    = ea+b (2)

1 3
2 2

1
or e(a/2)+b =  , ea + b =  [ (z + 1) = z(z)
2
    =  1  
a 1 3 1
 + b = log ,
2 2 2  2
=   
1 1 1 1
a+b = log   log 2  ]
2 2 2 2
Solving, these equations, we obtain
1
a = 2 log 2, b= log  + log 2
2
Using the values of a and b in (13), we get
 (2z) = 22z  1(z)   z  
1
…(14)
 2
Formula (14) is known as Lengendre‟s duplication formula.
(1) n
(vi) Residue of the gamma function at z = n, n = 0, 1, .. is .
n
To prove this, we know that the function (z) has the simple poles at z = 0, 1, 2, …, n,…
Also, we have (z + 1) = z (z)
(z  1)
i.e. (z) =
z
By repeated application of this formula, we can write
(z  1) (z  2) (z  3) (z  n  1)
(z) =    ... 
z z(z  1) z(z  1)(z  2) z(z  1)..(z  n )
Hence, the residue at z = n is given by
Res(z = n) = lim (z + n) (z)
z  n

(z  n  1)
= lim (z + n)
z  n z(z  1)(z  2)...( z  n )
(z  n  1)
= lim
z  n z ( z  1)(z  2)...( z  n  1)

(1)
=
 n (n  1)(n  2)...( 2)(1)
1 (1) n
= = .
(1) n(n  1)(n  2)...2.1
n
n

5.3. Integral Representation of (z). For Re z > 0, we define


(z) =  e  t t z 1dt …(1)
0
Integral (1) is also called Mellin integral or Mellin transform. We shall show that this function is
well defined for Re z = > 0 and the integral (1) is convergent.
For this, let 0 < C < . If the principle value of tz  1 is taken, then
 COMPLEX ANALYSIS111

C 
(z) =  e  t t z 1dt   e  t t z 1dt …(2)
0 C
We note that
t

t z1 e 2 t z 1 t z 1
|e t |= t
0 t
as t  

2
e e2
Hence, for large values of C, the second integral in (2) converges, since
 

 e t dt   e dt = 2e
 t z 1 t / 2 C/2

C C

For the first integral, we find that

C C C

e dt   | t z 1 | dt   t Re z 1 dt < , Re z > 0.
 t z 1
t
0 0 0

Thus we conclude that (z) is well defined for Re z > 0.

Moreover (z) is analytic in this domain.
We have another integral form of gamma function which is obtained by substituting t = x2 in (1).
Thus

(z) = 2  e  x x 2 z 1 dx,
2
Re z > 0 …(3)
0

Remark : The integral (1) diverges for Re z  0 which causes limitation of the gamma function
as an analytic function in the domain Re z > 0. However, we may extend this domain of
definition to the entire complex plane except the point z = 0, 1, 2, …
Using integration by parts, we get from (1)
 
(z + 1) = 
0
et tz dt = z  et tz1 dt = z (z)
0

(z  1)
or (z) = , Re(z + 1) > 0
z
The repeated application of this formula gives
(z  n  1)
(z) = , n = 1, 2, 3, .. …(4)
z(z  1)(z  2)...( z  n )
If n is chosen such that Re (z + n + 1) > 0, then (4) represents an analytic function for all z
except 0, 1, 2,…
5.4. Analytic Continuation of (z). By definition, we have

(z) = 
0
et tz1 dt (Re z > 0) …(1)

From this we are not able to infer anything about (z) on the imaginary axis or to the left
on it. Now, let us consider
F(z) =  e  w (w)z1 dw …(2)
C

where C is the contour consisting of the real axis from  to  ( > 0), the circle | w | = 
described in the positive direction and the real axis from  to  again. The many-valued
function ([w)z1 = e(z1)log(w) is made definite by taking log (w) to be real at w = . The
112

contour integral (2) is then uniformly convergent in any finite region of the z-plane, since the
question of convergence now need to be settled only at infinity. But since |wz1| = wx1, where
z = x + iy, the integral is uniformly convergent for z in any finite region throughout which
Re z  a > 0. Thus F(z) is regular for all finite values of z.
Now, if w =  ei, then
log([w) = log  = i(  ), for w  C i.e. on the contour. The integral
along the real axis therefore gives


 {e+(z  1)(log   i) + e(z 1)(log  + i)} d



= 2i sin z  e z1 d

On the circle of radius ,
|(w)z1| = |e(z1)log  + i(  ) | | O(1) means a function which  0
= e(x  1)log   y(  ) | O(1) means a bounded function
= O(x1) | O(x) means a function of order x.
The integral round the circle of radius  is thus O(x) = O(1) as   0 if x > 0. Hence, letting 
 0, we get

F(z) = 2i sin z 
0
e z1 d = 2i sin z (z), Re z > 0

F(z) 1
i.e. (z) =  i F(z) cosec  z, Re z > 0.
 2i sin z 2
1
Now, i F(z) cosec z is an analytic function of z except possibly at the poles of cosec z i.e.
2
z = 0, 1,  2,… is equal to (z) for Re z >0. Thus it is the analytic continuation of (z) for the
domain defined by the points on the imaginary axis and points to the left of it i.e. this function
can be taken as analytic continuation of (z) over the whole z-plane. But we already know that
(z) is regular at z = 1, 2,… Hence the only possible poles are z = 0, 1, 2,…
5.5. Stirling Formula. In most cases where the gamma function can be applied, it is essential to
have some information on the behaviour of (z) for very large values of z. Such asymptotic
character of (z) is given by Stirling formula. There are many proofs of this formula. We use
the well known method of comparing a sum (n) with the corresponding integral  (t) dt.
By definition, we have
1
e  z   z  z/n
z 
(z) = 1   e
n 1 
n

z z 
 log (z) =   n  log1  n   z  log z
n 1
…(1)

where we take the principal value of each logarithm. If [u] denotes the greatest integer not
greater than u, then we have
N1 n 1
N
[u ]  u  1 / 2 n  u  (1 / 2)
 u  z du    uz
du
0 n 0 n
 COMPLEX ANALYSIS113

 1 
N 1 n 1  n z  N 1 n 1
 1  
=   

2
uz
 1du    n   z  log(u  z)  u 

n 0 n n 0  2  n
 
 
N 1 n
 
  n  z  z  log(n  1  z)   n  2  z  log(n  z)  1
1 1
=
n 0
N 1 N 1

  n  2  z  log(n  1  z)    n  2  z  log(n  z)  N

1 1
=
n 0 n 0
N 1
N
 1   1 
=   n  2  z  log(n  z)    n  2  z  log(n  z)  N
n 1 
n 0  
(Replacing n + 1 by n in 1st series)
N 1
=   n   z  log(n  z)   N   z  log(N  z)
1 1
n 1  2   2 
N 1
  z   log z    n   z  log(n  z)  N
1 1
 2 n 1  2 
N 1
 1  1 
=   log(n  z)   z   log z   N   z  log(N  z)  N
n 1  2  2 
…(2)
N 1 N 1 N 1
 z 
Now,   log(n  z) =   log n  log1   =   log1    log N  1
z
n 1 n 1   n  n 1  n
N 1
z z 
N  1  z 1   ... 
1 
  n  log1  n   log
1
= 
n 1  2 N 1
Thus (2) becomes
1
N [u ]  u  N 1
z z 
 uz
2 du =
  n  log1  n   log
n 1
N 1
0

 z 1   ... 
1 1   1
   z   log z
 2 N 1  2
 1 
+  N   z  log (N + z)  N …(3)
 2 
Now, we shall show that if n is a positive integer, then
log (n) = log n  1 =  n   log n  n + C + O(1)
1
…(4)
 2
where C is a constant.
To prove this, we observe that
n 1
log ( n  1) =  log v
v 1
1
v
2 1/ 2
Now  log t dt = 
1
{log(v + t) + log (v  t)} dt
v 0
2
114

1/ 2
  t 2 
 log v  log1  2  dt = log v + Cv
2
=
0   v 
1
v
2
i.e. log v =  log t dt
1
 Cv, where Cv = O(1/v2).
v
2
1
n
2 n 1
Hence log (n) = log( n  1) =  log t dt   C
1/ 2 v 1
v

=  n   log  n     n    log
1 1 1 1 1
 2  2  2 2 2
1 
+   C v  0(1) (  log x dx = x log x  x}
2 v 1

=  n   log n  n + C + 0(1), where C is constant.

1
 2
 1  1 
log  n   = log n 1   = log n + kig 1   
1
 2  2n   2n  
Now, using (4) and the relations
1 1
1 +  ...  = log N +  + 0(1)
2 N 1
 O 2 
z 1
and log (N + z) = log N +
N N 
in equation (3) and then making N  , we obtain from (1) that
 [u ]  u 
1
 1 1
log (z) =  z   log z  z + log 2 +  2 du …(5)
 2 2 0
uz
where the integral is known as the error term. Writing
 1
u
(u) =  [ w ]  w  dw ,
0 2
we find that (u) is bounded since
(n + 1) = (n) for integer values of n.
Thus,
 [u ]  u 
1  
2 du = (u ) du  (u )
0 u  z 0 u  z 0 (u  z) 2 du (Integ. by parts)

 du  1
= O  2   O r  ,
0 u  r 2
 2 ur cos    
where r = | z | .
The usefulness of the error term is that it tends to zero uniformly in every sector of complex
numbers z = rei such that  +       , 0 <  < .
Hence (5) assumes the familiar form of Stirling formula for complex values of z, as
 COMPLEX ANALYSIS115

 1 
log (z) =  z   log z  z + log 2 + O 
1 1
 , as | z |  
 2 2 | z |
which can also be expressed in the form
(z)  zz(1/2) ez 2  , | z |  .

5.6. Riemann’s Zeta-function. Riemann‟s Zeta-function (z) is defined by series n
n 1
z
i.e.

 (z) = 1z + 2z + 3z + … + nz + …

1 1 1 1
= z  z  z  ...  z + …
1 2 3 n
which is known to converge uniformly and absolutely in any bounded closed domain to the right
of the line Re z = 1. Hence the function (z) is an analytic function, regular when Re z > 1.
We proceed to show that analyticity of  (z) can be extended to the region where Re z > 0,
by the function
(1  21z)  (z) = 1z  2z + 3z  4z + …
From this we shall be led to the result that the only singularity of  (z) in the right half plane
Re z > 0 is a simple pole at the point z = 1 and residue of  (z) at this pole is 1.
For Re z > 1, we have
  
(1  21z)  (z) = n
n 1
z
(1  21 z ) = n
n 1
z
 2 (2n ) z
n 1

= 1z  2z + 3z  4z + …

the re-ordering of the terms being justified by absolute convergence.
It can be shown that this last mentioned series in uniformly convergent in any bounded closed
domain D in which Re z   > 0. For this we use the following criterion (Hardy‟s Test) for the
uniform convergence of a series of terms which are functions of a complex variable. According
to this criterion the series an(z) un(z) is uniformly convergent in a bounded closed domain D, if
in D
(i) an(z) has uniformly bounded partial sums
(ii) |un(z)  un+1(z)| is uniformly convergent
(iii) un(z)  0 uniformly, as n  .
Now, let us take
an(z) = (1)n, un(z) = (n +1)z
Then, the partial sums of an(z) are alternately 1 and 0, so that condition (i) of Hardy‟s test is
satisfied. Condition (iii) is also satisfied, since
|un(z)| = (n + 1)Re z  (n + 1),
so that un(z)  0 uniformly in the bounded closed domain D in which Re z   > 0.
For the condition (ii), we observe that
un(z)  un+1(z) = (n + 1)z  (n + 2)z
n2

t
 z 1
=z dt
n 1
Hence
n2

| t
 z 1
|un(z)  un+1(z)|  | z | | dt
n 1
116

n 2

t
  1
|z| dt
n 1
< | z | (n + 1)  1
Since z is bounded in D, by Weierstrass‟s M-test, it follows that |un(z)  un+1(z)| converges
uniformly. Thus it follows finally that the series
1z  2z + 3z  4z + …
converges uniformly in D. Hence its sum function n(z) is a regular function in the region defined
by Re z > 0. But for Re z > 1, we have
(z) = (1  21z)  (z) …(1)
Equation (1) provides the analytic continuation of  (z) into the region 0 < Re z  1. Now
1  21z has simple zeros at the points z given by (1  z) log 2 = 2Ki, where K is any integer or
zero.
(z)
From (1), (z) =
1  21z
and thus it follows that a point of the above mentioned set is a simple pole of (z), provided (z)
does not vanish there and that (z) has no singularity in the right half plane Re z > 0. Since (1)
= log 2  0, the point z = 1 is a simple pole of (z). The residue at this pole is
Res(z = 1) = lim (z  1)  (z)
z 1

(z  1)(z)
= lim
z 1 1  21z
(z) (1)
= lim 1z  =1
z 1 2  1 log 2
1 z
It is observed that no other zero if 1  21z is a pole of (z). Hence we conclude that the only
singularity of (z) in Re z > 0 is a simple pole at z = 1 and residue at this pole is 1.
5.7. Analytic Continuation of  (z). We observe that
 
w z 1
0 e w  1 dw = 
0
{wz1 enw} dw

 
=   wz1 enw dw =  nz  vz1 ev dv
0 0

= nz (z) = (z) (z)

the operation being justified for Re z > 1.
Hence, we get

1 w z 1
 (z) = 0 e w  1 dw, Re z > 1
( z )
…(1)

In the same manner, it can be shown that if Re z > 1,

1 ( w ) z 1
 (z) = 
2i sin z(z) C e w  1
dw, …(2)

where the contour C starts at infinity on the positive real axis, encircles the origin once in the
1
positive direction, but excludes all the poles of other than w = 0 i.e. the points
(e w  1)
w =  2i,  4i,….
 COMPLEX ANALYSIS117

Now, using the functional equation of the gamma function, i.e.


(z) (1  z) =
sin z
we may write (2) in the form
i(1  z) ( w ) z 1
2 C e w  1
(z) = dw …(3)

As in case of the gamma function, this contour integral is an integral function of z. The formula
(3) therefore provides the analytic continuation of (z) over the whole plane. The only possible
singularities of (z) are the poles of (1  z) i.e. z = 1, 2, … But we know that (z) is regular at
z = 2, 3, … Hence the only possible pole is at z = 1. The residue at this simple pole is 1, since
for z = 1, the contour integral in (3) is
dw
 e w  1 = 2i, by theorem of residues, and (1  z) has a simple pole at z =1,
C
with residue 1.
5.8. Riemann’s Functional Equation. The function (z) satisfied the functional equation
1
(1  z) = 21z z cos z (z) (z)
2
To prove this, we consider the formula
1 ( w ) z 1
(z) = 
2i sin z(z) C e w  1
dw …(1)

and let z have any value and we deform C into the contour Cn consisting of the square with
centre at the origin and sides parallel to the axis, of length (4n + 2), together with the positive
real axis from (2n + 1) to . In this process, we pass over the poles of the integrand at the
points w = 2i, 4i, …,2ni and 2i, 4i, …, 2ni. The residue at 2Ki(K > 0) is
1
 i z
e(z  1)(log 2K  i/2) = (2K)z1 i e 2

and the residue at 2Ki (K > 0) is

1
i z
e(z  1)(log 2K + i/2) =  (2K)z1 i e 2
The sum of these two residues is
1
(2K)z12 sin z.
2
Hence formula (1) gives
z  1
1 ( w ) n
 2 sin z (2K) z 1
1
sin z(z) g(z) =   w …(2)
2i C e  1 2 K 1
n

(         2i  Re s )
C Cn poles Cn

Now, let Re z < 0. On the square

|(w)z1| = e(x1) log |w | y arg (w) = O(nx1)
1
and |(e  1) |  K, a positive finite constant.
w

Moreover, the length of the square is O(n), as n  . Hence, this part of the integral is O(nx),
and hence it tends to zero, since x < 0. The remaining part of the integral obviously tends to
zero. Hence letting n  , we have from (2)
118


z(2)z1  K z 1
1
sin  z(z) (z) = 2 sin
2 K 1

1 1 1
 2 sin  z cos z (z) (z) = 2 sin z (2)z1 (1  z)
2 2 2
1
 cos z (z) (z) = (2) (1  z)
z1
2
1
 (1  z) = 21z z cos z (z) (z) …(3)
2
which is the required function equation and provides a relationship between (z) and
(1  z).
We have proved the functional equation for Re z < 0. By analytic continuation, it holds for
all z.
Remark : An equivalent form of the functional equation (3) is obtained by using the identity of

the gamma function i.e. (z) (1  z) = in (3) to get
sin z
1
(z) = 2z z1 sin z (1  z) (1  z)
2
6. Runge’s Theorem
Before proving the theorem we state some basic concepts which will be used in sequel.
6.1. Definition : A metric space (X, d) is called connected if the only subsets of X which are
both open and closed are  and X.
A maximal connected subset of (X, d) is called a component of X. We note that
(i) If A and B are connected subsets of (X, d) such that A  B   then A  B is also
connected.
(ii) If A is a component of X then A is closed in X.
(iii) If G is open in  then components of G are open.
(iv) A subset A of a metric space is open iff int (A) = A.
Definition : If A is subset of a metric space and x  X then the distance from x to the set A,
denoted by d(x, A) is defined as
d(x, A) = inf{d(x, a) : a  A}
6.2. Definition : Let A be a subset of a metric space (X, d). Then the boundary of A denoted
by A, is defined as the set of all points of X which are neither in the interior of A nor in the
interior of (X  A). We note that
(i) A = A  X  A
(ii) int(A) = A  A
6.3. The extended plane. The set   {}   is called extended plane. If we define
2 | z  z |
d(z, z) = for z, z  
[(1 | z |2 )(1 | z |2 )]1/ 2
2
and d(z, ) = for z  
(1 | z | 2 )1 / 2
then (, d) is a metric space.
 COMPLEX ANALYSIS119

6.4. Definition : A path in a region G   is a continuous function r : [a, b]  G for some

interval [a, b] in R. If r : [a, b]   is a path then the set {r(t) : a  t  b} is called the trace of 
and is denoted by {}.
Note that trace of a path is always a compact set. We have the following two useful results :
Lemma (a). Let K be a compact subset of the region G ; then there are straight line segments 1,
2,…,n in GK such that for every function f in H(G),
n
f (w )
f(z) = 
1
k 1 2 i 

w z
dw
k

for all z in K.
The line segments form a finite number of closed polygons.
Lemma (b). Let  be a rectifiable curve and let K be a compact set such that K  {} =  . If f
is a continuous function on {} and  > 0 then there is a rational function R(z) having all its
poles on {} and such that
f (x)
 w  z  R (z) < 

for all z in K.

6.5. Theorem (Runge Theorem). Let K be a compact subset of  and let E be a subset of
  K that meets each component of   K. If f is analytic in an open set containing K and
 > 0 then there is a rational function R(z) whose only poles lie in E and such that
|f(z)  R(z)| <  for all z in K.
Proof : For f, g  C(K, ), define
(f, g) = sup{|f(z)  g(z)| : z  K }
Then  is a metric on C(K, ). Also (fn, f)  0 iff <fn> converges uniformly to f on K. So
Runge‟s theorem says that if f is analytic on a neighbourhood of K and  > 0 then there is a
1
rational function R(z) with poles in E such that (f, R) < . By taking  = , it is seen that we
n
want to find a sequence of rational functions {Rn(z)} with poles in E such that (f, Rn)  0, that
is, Rn converge uniformly to f on K.
Let B(E) be the collection of all functions f in C(K, ) such that there is a sequence {Rn} of
rational functions with poles in E such that {Rn} converges uniformly to f on K. Then B(E) is a
closed subalgebra of C(K, ) that contains every rational function with a pole in E. To say B(E)
is an algebra we mean that if f and g are in B(E) and    then f, f + g, fg  B(E). We now
prove two lemmas.
Lemma 1. Let V and U be open subsets of  with V  U and V  U =  . If H is a component
of U and H  V   then H  V.
Proof. Let a  H  V and let G be the component of V such that a  G. Then H  G is
connected since H  G   and H  G  U. Since H is a component of U so H  G  H. This
gives G  H . But G  V as G  V. So G  H =  as G  H  G  U  V  U =  .
Now V is open in  and G is component of V so G is open
 G = G
Also G = G  G
 G = G  G
120

Taking complements,
  G =   ( G  G)
=   [( G  ( G)]
= (  G )  G.
 H  G = H  (  G)
= H  [(  G )  G]
= H  [(  G ) [H  G =  )
As   G is open in  so H  (  G ) is open in H that is, H  G is open in H.
Also G is open in  so   G is closed in .
 H  (  G) is closed in H
 H  G is closed in H.
Now H  G is open as well closed in H. Since H is connected and G  , we must have
H  G =  . That is, H = G as G  V, we have H  V.
Lemma (2). If a    K then (z  a)1  B(E).
Proof. Case 1. When   E.
Let U =   K and let V = {a   : (z  a)1  B(E)}
Then E  V  U.
We first show “if a  V and |b  a| < d(a, K) then b  V”. …(1)
Since |b  a| < d(a, K) and d(a, K) = mf{|a  k| : k  K|}, there exists a number r, 0 <r < 1, such
that
|b  a| < r |z  a| for all z in K i.e. |b  a| |z  a|1 < r  z  K.
But
1
b a
(z  b)1 = [(z  a)  (b  a)]1 = (z  a)1 1   …(2)
 za
Hence |b  a| |z  a|1 < r < 1 for all z  K gives that
1 
1  b  a    b  a 
n


 za
   
n 0  z  a 
…(3)

converges uniformly on K by Weierstrass M-test.

ba
k
Let Qn(z) =  
n
1
 then (z  a) Qn(z)  B(E) since G  V and B(E) is an algebra. Since
k 0  z  a 

B(E) is closed, equation (1) and the uniform convergence of (3) imply that (z  b)1  B(E). So
b  V.
Now (1) implies B(a ; )  V where  = d(a, K).
 a is interior point of V. But a is any arbitrary point of V So V is open.
We claim that V  U = .
Let b  V, then there exists a sequence {an} in V such that b = lim an. Since V is open so
V V =  . Thus b  V. So (1) implies
|b  an|  d(an, K)
Letting n  , we have 0  d(b, K) i.e. d(b, K) = 0 or b  K. Thus V  U =  .
[ K =   U]
Let H be a component of U =   K. Then by hypothesis,
HE
 COMPLEX ANALYSIS121

So HV [ E V ]
BY Lemma (1), H  V. But H is arbitrary so U  V.
 V =U
Case 2.   E.

Let d be the metric on . Choose a0 in the unbounded component of   K such that
1
d(a0, )  d(, K) and
2
|a0| > 2 max {| z | : z  K}
Let E0 = (E  {})  {a0}.
Then E0 meets each component of   K. If a    K. Case 1 gives that (z  a)1  B(E0). If
we show that (z  a0)1  B(E) then it will follow that B(E0)  B(E) and so (z  a)1  B(E) for
z 1
each a in K. Now  for all z in K so
a0 2
 n
 z 

1 1 1
=   
z  a0  z  a0 n 0  a 0 
a 0 1  
 a 0 
k
 z 
n
Converges uniformly on K. So Qn(z) = a01
k 0
  a
 is a polynomial and Qn converges
0 

uniformly to (z  a0)1 on K. Since Qn has its only pole at , Qn  B(E). Thus (z  a0)1  B(E).
Proof of main theorem. If f is analytic on an open set G and K  G then for each  > 0, there
exists a rational function R(z) with poles in K such that
|f(z)  R(z)| <  for all z in K, by lemma (a) and (b).
Since B(E) is an algebra, lemma (2) implies that R  B(E).
Hence the result.
Corollary. Let G be an open subset of the plane and let E be a subset of   G such that E
meets every component of   G. Let R(G, E) be the set of rational functions with poles in E
and consider R(G, E) as a subspace of H(G). If f  H(G) then there is a sequence {Rn} in
R(G, E) such that f = lim Rn. That is, R(G, E) is dense in H(G).
7. Mittag Leffler’s Theorem
Let G be an open subset of  and let {ak} be a sequence of distinct points in G such that {ak} has
no limit point in G. Let {Sk(z)} be the sequence of rational functions given by
mk A jk
Sk(z) = 
j1 ( z  a k )
j

where mk is some positive integer and A1k, A2k,…, A mk k are arbitrary complex coefficients.
Then there is a meromorphic function f on G whose poles are exactly the points {a k} and such
that the singular part of f at ak is Sk(z).
Proof : Since G is open in C there is a sequence {Kn} of compact subsets of G such that

G = K n , Kn  int (Kn+1) and each component of Kn contains a component of G.
n 1
Since each Kn is compact and {ak} has no limit point in G, there are only a finite number of
points ak in each Kn.
122

Define the sets of integers In as follows :

I1 = {k : ak  K1}
In = {k : ak  Kn  Kn1} for n  2.
Define functions fn by

fn(z) = S
kI n
k (z) for n  1.

Then fn is rational and its poles are the points

{ak : k  In}  Kn  Kn1
Note that if In is empty then let fn = 0.
Since fn has no poles in Kn1 (for n  2), it is analytic in a neighbourhood of Kn1. By Runge‟s
theorem, there is a rational function Rn(z) with its poles in G such that
n
|fn(z)  Rn(z)| <  
1
for all z in Kn1.
2

Let f(z) = f1(z) +  [f
n 2
n (z)  R n (z)] …(1)

We claim that f is required meromorphic function.

Let K be a compact subset of G  {ak : k  1}. Then K is a compact subset of G. So there is an

integer N such that K  KN as G = K n .
n 1

If n  N then
n
|fn(z)  Rn(z)| <  
1
for all z in K.
2
So by Weierstrass M-test, the series (1) converges uniformly to f on K. Thus f is analytic on
G{ak : k  1}.
It remains to show that each ak is a pole of f with singular part Sk(z). For this consider a fixed
integer k  1. Then there is a number R > 0 such that
|aj  ak| > R for j  k.
Thus f(z) = Sk(z) + g(z) for 0 < |z  ak| < R
where g is analytic in B(ak ; R). Hence z = ak is a pole of f and Sk(z) is its singular part.
 COMPLEX ANALYSIS123

UNIT-IV


1. Analytic Continuation
From the results regarding zeros of an analytic function, it follows that if two functions are
regular in a domain D and if they coincide in a neighbourhood, however small, of any point
a of D, or only along a path-segment, however small, terminating in a point a of D, or only
at an infinite number of distinct points with a limit-point a in D, then the two functions are
identically the same in D. Thus it emerges that a regular function defined in a domain D is
completely determined by its values over any such sets of points. This is a very great
restraint in the behaviour of analytic functions. One of the remarkable consequences of this
feature of analytic functions, which is extremely helpful in studying them, is know as
analytic continuation. Analytic continuation is a process of extending the definition of a
domain of an analytic function in which it is originally defined i.e. it is a concept which is
utilized for making the domain of definition of an analytic function as large as possible.
Let us suppose that two functions f1(z) and f2(z) are given, such that f1(z) is analytic in the
domain D1 and f2(z) in a domain D2 We further assume that D1 and D2 have a common part D12
(D1  D2). If f1(z) = f2(z) in the common part D12, then we say that f2(z) is the direct analytic
continuation of f1(z) from D1 into D2 via D12. Conversely, f1(z) is the direct analytic continuation
of f2(z) from D2 into D1 via D12. Indeed f1(z) and f2(z) are analytic continuations of each other.
Both f1(z) and f2(z) may be regarded as partial representations or elements of one and the same
function F(z) which is analytic in the domain D1  D2, and is defined as
 f (z) for all z  D1
F(z) =  1
 f 2 (z) for all z  D 2
under the condition that f1(z) = f2(z) at an infinite set of points with a limit-point in D12

D1 D2

D12

It is observed that for the purpose of analytic continuation, it is sufficient that the domains D1
and D2 have only a small arc in common.
1.1. Definition. An analytic function f(z) with its domain of definition D is called a function
element and is denoted by (f, D). If zD, then (f, D) is called a function element of z. Using this
notation, we may say that (f1, D1) and (f2, D2) are direct analytic continuations of each other iff
D1  D2   and f1(z) = f2(z) for all zD1  D2.
Remark. We use the word „direct‟ because later on we shall deal with continuation along a
curve. i.e. just to distinguish between the two.
124

1.2. Analytic continuation along a chain of Domain. Suppose we have a chain of function
elements (f1, D1), (f2, D2),…, (fk, DK),…,(fn, Dn) such that D1 and D2 have the part D12 in
common, D2 and D3 have the part D23 is common and so on. If f1(z)= f2(z) in D12, f2(z) = f3(z) in
D23 and so on, then we say that (fk, DK) is direct analytic continuation of (fK1, DK1). In this
way, (fn, Dn) is analytic continuation of (f1, D1) along a chain of domains D1, D2,…, Dn. Without
loss of generality, we may take these domains as open circular discs. Since (fK1, DK1) and (fK,
DK) are direct analytic continuations of each other, thus we have defined an equivalence relation
and the equivalence classes are called global analytic functions.

1.3. Complete Analytic Function

Suppose that f(z) is analytic in a domain D. Let us form all possible analytic continuations of (f,
D) and then all possible analytic continuations (f1, D1), (f2, D2),…, (fn, Dn) of these continuations
such that
 f1 (z ) if z  D1
 f (z ) if z  D
 2 2
.......... .......... ......
F(z) = 
.......... .......... ......
.......... .......... .......

 f n (z ) if z  D n

Such a function F(z) is called complete analytic function. In this process of continuation, we may
arrive at a closed curve beyond which it is not possible to take analytic continuation. Such a
closed curve is known as the natural boundary of the complete analytic function. A point lying
outside the natural boundary is known as the singularity of the complete analytic function. If no
analytic continuation of f(z) is possible to a point z0, then z0 is a singularity of f(z). Obviously,
the singularity of f(z) is also a singularity of the corresponding complete analytic function F(z).
1.4. Theorem (Uniqueness of Direct Analytic Continuation). There cannot be two different
direct analytic continuations of a function.
Proof. Let f1(z) be an analytic function regular in the domain D1 and let f2(z) and g2(z) be two
direct analytic continuations of f1(z) from D1 into the domain D2 via D12 which is the domain
common to both D1 and D2. Then by definition of analytic continuation, f2(z) and g2(z) are two
functions analytic in D2 such that
f1(z) = f2(z) and f1(z) = g2(z)
at all points z in D12 i.e. f2(z) = g2(z) in D12. Thus f2(z) and g2(z) are two functions analytic in the
domain D2 such that they coincide in a part D12 of D2. It follows from the well known result
that they coincide throughout D2. i.e. f2(z) = g2(z) throughout D2. Hence the result.
1.5. Example. Given the identity sin2z + cos2z = 1 holds for real values of z, prove that it also
holds for all complex values of z.
Solution. Let f(z) = sin2z + cos2z1 and let D be a region of the z-plane containing a portion of
x-axis (real axis). Since sin z and cos z are analytic in D so f(z) is also analytic in D. Also
f(z) = 0 on the x-axis. Hence by the well known result, it follows that f(z) = 0 identically in D,
which shows that sin2 z + cos2 z = 1 for all z in D. Since D is arbitrary, the result holds for all
values of z.
 COMPLEX ANALYSIS125

Remark. This method is useful in proving for complex values many of the results true for real
values.
1.6. Analytic continuation along a curve Let  be a curve in the complex pane having equation
z = z(t) = x(t) + iy(t), a  t  b.
We take the path along  to be continuous. Let a = t0  t1  … tn = b be the portion of the
interval. If there is a chain (f1, D1), (f2, D2),…,(fn, Dn) of function elements such that (fK+1, DK+1)
is a direct analytic continuation of (fK, DK) for K = 1, 2…., n1 and z(t)  DK for tK1  t  tK, K
= 1, 2,…, n then (fn, Dn) is said to be analytic continuation of (f1, D1) along the curve .
Thus we shall obtain a well defined analytic function in a nbd. of the end point of the path, which
is called the analytic continuation of (f1, D1) along the path . Here, DK may be taken as discs
containing z(tK1) as shown in the figure.

Z(t2)
Z(tn-1)
Z(t0)

D1
D2 Z(tn)
D3 Dn1 Dn
Z(t1)
Dn

Further, we say that the sequence {D1, D2,…, Dn} is connected by the curve  along the partition
if the image z([tK1, tK]) is contained in DK.
1.7. Theorem (Uniqueness of Analytic Continuation along a Curve). Analytic continuation
of a given function element along a given curve is unique. In other words, if (fn, Dn) and
(gm, Em) are two analytic continuations of (f1, D1) along the same curve  defined by
z = z(t) = x(t) + iy(t), a  t  b.
Then fn = gm on Dn  Em
Proof. Suppose there are two analytic continuations of (f1, D1) along the curve , namely
(f1, D1), (f2, D2),…, (fn, Dn)
and (g1, 1), (g2, 2),…, (gm, m)
where g1 = f1 and E1 = D1
Then there exist partitions
a = t0  t1  ………..  tn = b
and a = s0  s1  ………. < sm = b
such that z(t) Di for ti1  t  ti, i = 1, 2,…,n and z(t) Ej for sj1  t sj, j = 1, 2, …,m. We
claim that if 1  i  n, 1  j  m and
[ti1, ti)  [sj1, sj]  
then (fi, Di) and (gi, Ej) are direct analytic continuations of each other. This is certainly true when
i = j = 1, since g1 = f1 and E1 = D1. If it is not true for all i and j, then we may pick from all (i, j),
for which the statement is false, a pair such that i + j is minimal. Suppose that ti1  sj1, where i
 2. Since [ti1, ti]  [sj1, sj]   and sj1  ti1 , we must have ti1  sj. Thus sj1  ti1  sj. It
follows that z(ti1)  Di1  Ei  Ej. In particular, this intersection is non-empty. None (fi, D1) is
a direct analytic continuation of (fi1, Di1). Moreover, (fi1, Di1) is a direct analytic continuation
126

of (gj, Ej) since i + j is minimal, where we observe that t i1  [ti2, ti1]  [sj1, sj] so that the
hypothesis of the claim is satisfied. Since Di1  Di  Ej  , (fi, D1) must be direct analytic
continuation of (gj, Ej) which is contradiction. Hence our claim holds for all i and j. In particular,
it holds for i = n, j = m which proves the theorem.

1.8. Power series Method of Analytic continuation. Here we consider the problem of
continuing analytically a function f(z) defined initially as the sum function of a power series

 an (zz0) whose circle of convergence C0 has a finite
n
n 0
non-zero radius. Thus, we shall use only circular domain and Taylor‟s expansion in such domain.
The first thing to observe here is that, when the continuation has been carried out, there must be
at least one singularity of the complete analytic function on the circle of convergence C0. For if
there were not, we would construct, by analytic continuation, an analytic function which is equal
to f(z) within C0 but is regular in a larger concentric circle C0. The expansion of this function as
a Taylor‟s series in powers of z z0 would then converge everywhere within C0, which is,
however, impossible since the series would necessarily be the original series, whose circle of
convergence is C0.
To carry out the analytic continuation, we take any fixed point z 1 within C0, and calculate the
values of f(z) and its successive derivatives at that point from the given power series by repeated
term-by-term differentiation. We then form the Taylor‟s series
 f n (z1 )
 (z  z1)n (1)
0 | n
whose circle of convergence is C1, say. Let 1 denote the circle with centre z1 which touches C0
internally. By Taylor‟s theorem, this new power series is certainly convergent within 1 and has
the sum f(z) there. Hence the radius of C1 cannot be less than that of 1. There are now three
possibilities
(i) C1 may have a larger radius than 1. In this case C1 lies partially outside C0 and the
new power series (1) provides an analytic continuation of f(z). We can then take a point z2 within
C1 and outside C0 and repeat the process as far as possible.
(ii) C0 may be a natural boundary of f(z). In this case, we cannot continue f(z) outside C0
and the circle C1 touches C0 internally, no matter what point z1 within C0 was chosen.
(iii) C1 may touch C0 internally even if C0 is not a natural boundary of f(z). The point of
contact of C0 and C1 is then a singularity of the complete analytic function obtained by the
analytic continuation of the sum function of the original power series, since there is necessarily
one singularity of the complete analytic function on C1 and this cannot be within C0.

Thus, if C0 is not a natural boundary for the function f(z) =  an (zz0)n, this process of forming
n 0
the new power series of the type (1) provides a simple method for the analytic continuation of
f(z), know as power series method.
Remark. Power series method is also called standard method of analytic continuation.
1.9. Example. Explain how it is possible to continue analytically the function
f(z) = 1 + z + z2 +…+ zn +…
outside the circle of convergence of the power series.
 COMPLEX ANALYSIS127

Solution. The circle of convergence of the given power series is |z| = 1. Denoting it by C0, we
observe that within C0 the sum function f(z) = (1z)1 is regular. Further, this function is regular
in any domain which does not contain the point z = 1. We carry out the analytic continuation by
means of power series. If a is any point inside C0 such that a is not real and positive, then
|1 a| > 1  |a| (1)
Now, the Taylor‟s expansion of f(z) about the point z = a is given by
 f n (a )
 (z  a)n (2)
n 0 |n
f n (a ) 1
But  and thus (2) becomes
| n (1  a ) n 1
n
 (z  a ) n 1  za 
 n 1
    (3)
n 0 (1  a ) 1  a n 0  1 a 
za
Clearly, (3) converges for 1
1 a
and its circle of convergence C1 is given by |z  a| = |1  a|. It follows from the inequality (1) that
C1 goes beyond C0 and hence (3) provides an analytic continuation of f(z) outside C0, since we
note that the sum function of (3) is also (1z)1.
On the other hand, if a be a real point inside C0 such that 0 < a < 1, C1 touches C0 at z = 1, which
is, therefore, a singularity of the complete analytic function obtained by analytic continuation of
f(z).
1.10. Example. Show that the function
1 z z2
f(z) =  2  3  ...
a a a
can be continued analytically.
Solution. We have
1 z z2

f(z) =   ... (1)
a a2 a3
This series converges within the circle C0 defined by |z| = |a| and has the sum
1/ a 1
f(z) = 
1 z / a a  z
The only singularity of f(z) on C0 is at z = a. Hence the analytic continuation of f(z) beyond C0 is
possible. For this purpose we take a point z = b not lying on the line segment joining z = 0 and
z = a. We draw a circle C1 with centre b and radius |a  b| i.e. C1 is |z  b| = |ab| This new circle
C1 clearly extends beyond C0 as shown in the figure

a b

O
128

Now we reconstruct the series (1) in powers of z  b in the form

 (z  b) n f n (b) 1
 n 1
, where  (2)
n 0 (a  b) | n ( a  b) n 1
1
This power series has circle of convergence C1 and has the sum function . Thus the power
az
series (1) and (2) represent the same function in the region common to the interior of C0 and C1
Hence the series (2) represents an analytic continuation of series (1).
1.11. Example. Show that the circle of convergence of the power series
f(z) = 1 + z + z2 + z4 + z8 +……
is a natural boundary of its sum function
Solution. We have
 n
f(z) = 1 +  z 2
n 0
Evidently, |z| = 1 is the circle of convergence of the power series. We write
q n  n
f(z) = 1 +  z 2   z 2 = f1(z) + f2(z), say.
0 q 1
Let P be a point at z = r e2ip/2q lying outside the circle of convergence, where p and q are
integers and r > 1.
We examine the behaviour of f(z) as P approaches the circle of convergence through
radius vector.
P

n n n / 2q n n q 1
Now, z 2  r 2 e 2ip 2  r 2 e ip 2
q n n 1q
 f1(z) = 1 +  r 2 e πip 2
n 0
which is a polynomial of degree 2q and tends to a finite limit as r1
 n n 1q
Also f2(z) =  r 2 e πip 2
q 1
Here, n > q so 2n+1q is an even integer and thus
n 1q
e πip 2 1
 n
 f2(z) =  r 2  as r1
q 1
q
Thus f(z) = f1(z) + f2(z), when z = e 2 πip / 2 .
q
Hence the point z = e 2 πip / 2 is a singularity of f(z). This point lies on the boundary of the circle
q
|z| = 1. But any arc of |z| = 1, however small, contains a point of the form e 2 πip / 2 , where p and
q are integers. Thus the singularities of f(z) are everywhere dense on |z| = 1and consequently
|z| = 1 constitutes the natural boundary for the sum function of the given power series.
 COMPLEX ANALYSIS129


1.12. Example. Show that the power series  z3n cannot be continued analytically beyond the
n 0
circle |z| = 1
Solution. Here |un(z)|1/n = |z3n|1/n = |z3| = |z|3
So the series is convergent if |z| < 1
 Circle of convergence is |z| = 1 Now take the point P at z = r e 2 πip / 3q , r > 1 and then
proceeds as in the above two examples.
1.13. Example. Show that the power series
z 2 z3
z  .......
2 3
may be continued analytically to a wider region by means of the series
(1  z) (1  z) 2 (1  z) 3
log 2  
2 2  22 3  23
Solution. The first series converges within the circle C1 given by |z| = 1 and has the sum function
f1(z) = log(1 + z). The second series has the sum function
  1  z  1  1  z  2 1  1  z 3 
f2(z) = log 2 +         ....
  2  2  2  3  2  
  1  z  1 z
= log 2 + log 1    1
  2  2
1 z 
= log 2 + log   = log (1 +z)
 2 
1 z
and thus is convergent within the circle C2 given by = 1 i.e. |z1| = 2 thus we observe that
2
(i) f1(z) is analytic within C1
(ii) f2(z) is analytic within C2
(iii) f1(z) = f2(z) in the region common to C1 and C2.
Hence the second series is an analytic
continuation of the first series to circle C2
which evidently extends beyond the circle
C1, as shown in the figure.

z =1
z = 1 C2
C1

Remark. The circles C1 and C2 touch internally at z = 1 which is a singularity for both f1(z) and
f2(z) i.e. z = 1 is a singularity of the complete analytic function whose two representations
(members) are f1(z) & f2(z).

1.14. Example. Show that the functions defined by the series

130

1 + az + a2 z2 +…..
1 (1  a )z (1  a ) 2 z 2
and  
1  z (1  z) 2 (1  z) 3
are analytic continuations of each other.
1
Solution. The first power series represents the function f1(z) = and has the circle of
1  az
1 1
convergence C1 given by |az| = 1 i.e. |z| = The only singularity is at the point z = (a  0) on
|a| a
the boundary of the circle. The second series has the sum function
1 (1  a )z (1  a ) 2 z 2
f2(z) =   .....
1  z (1  z) 2 (1  z) 3
1  1 a  1 a  2 
2
= 1   z    z ......... 
1  z   1  z   1  z  
1 1 1 a 
=  z  1
1 z 1 a  1 z 
1  z
1 z 
1 1 z 1
= . 
1  z 1  az 1  az
and has the circle of convergence C2 given by
z(1  a )
 1 i.e. |z(1a)| = |1z|
1 z
i.e. |z(1a)|2 = |1z|2
i.e. z z (1a)2 = (1z) (1  z) , where a is assumed to be real and a > 0
i.e. z z (1a)2 = 1 (z + z ) + z z
i.e. (x2 + y2) (1+a2  2a) = 1 2x + x2 + y2, z = x + iy
i.e. (x2 + y2) a (a2) = 1 2x
2x 1
i.e. x2 + y2   0
a (2  a ) a (2  a )
2 2
 1   1 a 
i.e.  x    ( y  0) 2   
 a (2  a )   a (2  a ) 
 1  1 a
Thus the circle C2 has the centre  ,0  and radius .
 a (2  a )  a (2  a )

Since the two circles depend upon a, where we shall assume that a > 0, we have the
following cases
Case I. Let 0 < a < 1. In this case, the two circles C1 and C2 touch internally, since the distance
between their centres is equal to the difference of their radii. Thus the first series represent the
analytic continuation of the second from C2 to C1 y
C1

x
0
C2
 COMPLEX ANALYSIS131

1
Case II. If a = 1, then the second series reduces to and the first series is 1 + z + z2 +….
1 z
1
which has the sum function .
1 z
Case III. If 1 < a < 2. In this case, the two
1
circles touch externally at z = so that the
a
two series have no common region of O
convergence. Nevertheless they are analytic
1 C2
continuations of the same function C1
1  az
1
Case IV. If a = 2, then the first series represents the function within C1 given by
1  2z
1 1 z
|z| = and the second series defines the sum function in the region 1
2 1  2z 1 z
1
i.e. z z  (1  z)(1  z), i. e.x  .
2
Thus the second series represents the function y 1
1 1 x
in the half plane x < . We note that the line 2
1  2z 2
1 1
x= touches the circle |z| = as shown in the x
2 2 O
figure. Hence in this case, the second series
represents the analytic continuation of the first
1 C1
series from the region |z| < to the half plane x
2
1
< .
2
Case V. Let a > 2. In this case C1 and C2 touch
internally, where C1 being the inner circle, as shown
in the figure. Hence the second series represents an 0
analytic continuation of the first series from C1 to C 1
C2 C2
1.15. Example. Show that the function defined by C2
f1(z) =  0 t3 ezt dt
is analytic at all points z for which Re(z) > 0. Find also a function which is analytic continuation
of f1(z).
Solution. f1(z) =  0 t3 ezt dt
132


 e zt e zt e zt e zt 
= t 3  3t 2 2  6t 3  6 4  (Integ. by parts)
 z z z z 0
 1  6
= 6  0  4   4 , if Re(z) > 0
 z  z
6
Let f2(z) =
z4
Then f1(z) = f2(z) for Re(z) > 0
The function f2(z) is analytic throughout the complex plane except at z = 0 and f1(z) = f2(z)  z
s. t. Re(z) > 0. Hence f2(z) is the required analytic continuation of f1(z).
2. Schwarz’s Reflection Principle
We observe that some elementary functions f(z) possess the property that f(z)  f (z) for all
points z in some domain. In other words, if w = f(z), then it may happen that w  f (z) i.e. the
reflection of z in the real axis corresponds to one reflection of w in the real axis. For example,
the functions
z, z2 + 1, ez, sin z etc
have the above said property, since, when z is replaced by its conjugate, the value of each
function changes to the conjugate of its original value. On the other hand, the functions
iz, z2 + i, eiz, (1 + i) sin z etc
do not have the said property.
2.1. Definition. Let G be a region and G* = {z : zG} then G is called symmetric region if
G = G*
If G is a symmetric region then let G+ = {z  G : Im z > 0} G = {z G : Im z < 0} and
G0 = {z  G : Im z = 0}.
2.2. Theorem (Schwarz’s Reflection Principle). Let G be a region such that G = G* if
f : G+  G0  is a continuous function which is analytic on G+ and f(x) is real for x in G0 then
there is an analytic function g : G  s.t. g(z) = f(z) for all z in G+  G0.

Proof. For z in G, define g(z) = f (z) and for z in G+  G0, define g(z) = f(z).
Then g : G  is continuous. We will show that g is analytic. Clearly g is analytic on G +  G.
To show g is analytic on G0, let x0 be a fixed point in G0 and let R > 0 be such that
B(x0 ; R)  G.
It is sufficient to show that g is analytic on B(x0 ; R) We shall apply Morera‟s theorem.
Let T = [a, b, c, a] be a triangle in B(x0 ; R). Assume that T  G+  G0 and [a, b]  G0 Let 
represent T together with its inside. Then g(z) = f(z) for all z in . [ T  G+  G0] By
hypothesis f is continuous on G+  G0, so f is uniformly continuous on . So given  > 0, there
is a  > 0 s.t. z, z  implies
|f(z)  f(z)| <  whenever |z  z| < .
Choose  and  on the line segments [c, a] and [b, c] respectively so that |  a| <  and
|  b| < . Let T1 = [, , c, ] and Q = [a, b, , , a]. Then  f   f   f
T T1 Q
 COMPLEX ANALYSIS133

G+

 
a b G0

But T1 and its inside are contained in G+ and f is analytic there.

So  f 0
T1

  f  f
T Q
By if 0  t  1, then
| [t  + (1  t) ]  [tb + (1t) a]| < 
so that
| f (t  + (1  t) )  f(t b + (1  t) a) | < .
Let M = max. {1 f(z) | : z  } and l be the perimeter of T then
|  f   f | = |(ba)  01 f(tb + (1 t)a)dt  ()  01 f(t +(1 t)) dt |
[ a ,b ] [ β ,α ]

 | b a|  01 [f(t b +(1t)a)  f(t + (1t))] dt |

+ |ba)  ()| |  01 f(t  + (1t)) dt |
 |b  a| + M |(b) + (  a)|
 l + 2M.
Also |  f |  M |a  |  M 
[ α ,a ]

and |  f |  M.
[ b,β ]

 |  f | |  f  f  f  f||  f  f ||  f ||  f|

T [ a ,b ] [ , ] [  ,a ] [ b , ] [ a ,b ] [, ] [  ,a ] [ b , ]

 l + 4M
Choosing  > 0 s. t.  < . Then
|  f | < (l + 4M). Since  is arbitrary it follows that  f  0 . Hence f
T T
must be analytic.
3. Monodromy Theorem and its Consequences
We first give some definitions.
3.1. Definition. Let 0, 1 : [0, 1]G be two closed rectifiable curves in a region G then 0 is
homotopic to 1 in G if there is a continuous function
F : [0, 1]  [0, 1]  G
134

such that
F(s, 0) = 0(s)
F(s, 1) = 1(s) (0  s < 1)
F(0, t) = F(1, t) (0  t  1)
3.2. Definition. 0, 1 : [0, 1]G are two rectifiable curves in G such that 0(0) = 1(0) = a and
0(1) = 1(1) = b then 0 and 1 are fixed-end-point homotopic (FEP homotopic) if there is a
continuous map F : [0, 1]  [0, 1]G s.t.
F(s, 0) = 0(s), F(s, 1) = 1(s)
F(0, t) = a, F(1, t) = b for 0  s, t  1.
We note that the relation of FEP homotopic is an equivalence relation on the curves from
one given point to another.
3.3. Definition. An open set G is called simply connected if G is connected and every closed
curve in G is homotopic to zero.
3.4. Definition. A function element is a pair (f, G) where G is a region and f is an analytic
function on G.
For a given function element (f ,G) define the germ of f at a to be the collection of all function
elements (g, D) such that aD and f(z) = g(z) for all z in a neighbourhood of a. The germ of f at
„a‟ is denoted by [f]a.
Notice that [f]a is a collection of function elements.
3.5. Definition. Let  : [0, 1]  C be a path and suppose that for each t in [0, 1] there is a
function element (ft. Dt) such that
(i) (t)  Dt ;
(ii) for each t in [0, 1] there is a  > 0 such that | s t| <  implies (s)  Dt and
[fs](s) = [ft](s)
Then (f1, D1) is called analytic continuation of (f0, D0) along the path .
Remark. Since  is a continuous function and (t) is in the open set Dt so there is a  > 0 such
that (s)  Dt for | st| < 
So part (ii) of above definition implies
fs(z) = ft(z) for all z  Ds  Dt,
whenever |s  t| < 
3.6. Theorem. Let  : [0, 1]  be a path from a to b and let {(ft, Dt) : 0  t  1} and
{(gt, Bt) : 0  t  1} be analytic continuations along  such that [f0]a = [g0]a. Then [f1]b = [g1]b
Proof. Consider the set
T = {t  [0, 1] : [ft](t) = [gt](t)}
Since [f0]a = [g0]a so 0  T. Thus T  .
We shall show that T is both open and closed. To show T is open, let t be a fixed point of
T st t  0. By definition of analytic continuation, there is a  > 0 such that for |s  t| < ,
(s)  Dt  Bt and
[fs](s) = [ft](s)
[gs](s) = [gt](s)
But t  T implies
ft(z) = gt(z) for all z in Dt  Bt.
 COMPLEX ANALYSIS135

Hence [ft](s) = [gt](s) for all (s) in Dt  Bt.

So [fs](s) = [gs](s) whenever |st| < .
That is, s  T whenever |s  t| < 
or (t  , t + )  T.
If t = 0 then the above argument shows that [a, a + )  T for some  > 0. Hence T is open.
To show T is closed let t be a limit point of T. Again by definition of analytic
continuation there is a  > 0 s. t. for | s  t| < , (s)  Dt  Bt and
[fs](s) = (ft](s)
[gs](s) = [gt](s) (1)
Since t is a limit point of T there is a point s in T s.t. |s  t| < . Let G = Dt  Bt  Ds  Bs.
Then (s)  G . So G is non-empty open set thus by definition of T,
fs(z) = gs(z) for all z in G.
But (1) implies
fs(z) = ft(z) and gs(z) = gt(z) for all z in G.
 ft(z) = gt(z) for all z in G.
Since G has a limit point in Dt  Bt , this gives [ft](t) = [gt](t)
Thus t  T and so T is closed.
Now T is a non-empty subset of [0, 1] s.t. T is both open and closed. So connectedness of [0,
1] implies T = [0, 1].
Thus 1  T and hence [f1](1) = [g1](1) i.e. [f1]b = [g1]b as (1) = b.
3.7. Definition. If  : [0, 1]   is a path from a to b and {(ft, Dt) : a  t  1} is an analytic
continuation along  then the germ [f1]b is the analytic continuation of [f0]a along .
Remark. Suppose a and b are two complex numbers and let  and  be two paths from a to b.
Suppose {(ft, Dt)} and {(gt, Dt)} are analytic continuations along  and  respectively s.t.
[f0]a = [g0]a. Now the question is “Does it follow that [f1]b = [g1]b”?. If  and  are the same path
then above result gives an affirmative answer. However if  and  are distinct then the answer
can be no.
3.8. Lemma. Let  : [0 1]  be a path and let {ft, Dt) : 0  t  1} be an analytic continuation
along . For 0  t  1 let R(t) be the radius of convergence of the power series expansion of
ft about z = (t). Then either R(t)   or R : [0,1]  (0, ) is continuous.
Proof. Suppose R(t) =  for some value of t.
Then ft can be extended to an entire function.
It follows that fs(z) = ft(z) for all z in Ds so that R(s) =  for all s in [0, 1] ; that is R(s)  . Now
suppose that R(t) <  for all t. Let t be a fixed number in [0, 1] and let a = (t).

Let ft(z) =  an (z a)n
n 0
be the power series expansion of ft about a.
Now let 1 > 0 be such that |st| < 1 implies that
(s)  Dt  B(a ; R(t)) and [fs](s) = [ft](s Fix s with |st| < 1 and let b = (s).
Now ft can be extended to an analytic function on B(a; R(t)) Since fs agrees with ft on a
neighbourhood of fs can be extended so that it is also analytic on B(a ; R(t))  Ds. If fs has
power series expansion
136


fs(z) =  bn (z b)n about z = b.
n 0
Then the radius of convergence R(s) must be at least as big as the distance from b to the
circle |z  a| = R(t) ; that is,

R(s)  d(b, {z : |z a| = R(t)})

 R(t)  |a b|
This implies R(t)  R(s)  |a b|
R(t)
i.e. R(t)  R(s)  |(t)  (s)| a
Similarly, we can show b
R(s)  R(t)  |(t)  (s)|
 |R(s)  R(t)  |(t)  (s)| for |s  t| < 1.
Since  : [0, 1]  is continuous so given  > 0,  2 > 0 s.t.
|(t)  (s)| <  for |s  t| < 2.
Let  = min {1, 2}. Then  > 0 and
|R(s)  R(t)| ,  for |st| < .
Hence R is continuous at t.
3.9. Lemma. Let  :[0, 1]  be a path from a to b and let {(ft, Dt) : 0  t  1} be an analytic
continuation along . There is a number  > 0 s.t. if  : [0, 1]   is any path from a to b with
|(t)  (t)| <  for all t and if {(gt, Bt) : 0  t  1} is any continuation along  with [g0]a = [f0]a ;
then [g1]b = [f1]b.
Proof. For 0  t  1, let R(t) be the radius of convergence of the power series expansion of
ft about z = (t).
If R(t)   then any value of  will be sufficient.
So suppose R(t) <  for all t.
Since R is a continuous function and R(t) > 0 for all t, R has a positive minimum value.
1
Let 0 <  < min {R(t) : 0  t  1}
2
Suppose  : [0, 1]  is any path from a to b with |(t)  (t)| <  for all t and {(gt, Bt) :
0  t  1} is any continuation along  with [g0]a = [f0]a.
Suppose Dt is a disk of radius R(t) about (t).
Since |(t)  (t)| <  < R(t), (t)  Bt  Dt for all t.
Define the set T = {t  [0, 1] : ft(z) = gt(z) for all z in Bt  Dt}
Then 0  T since [g0]a = [f0]a. So T  .
We will show 1  T. For this it is sufficient to show that T is both open and closed subset of
[0, 1].
To show T is open, let t be any fixed point of T.
Choose  > 0 such that
| γ(s)  γ( t ) |, [ f s ] γ (s )  [ f t ] γ (s ) ,
 (1)
| σ(s)  σ( t ) |, [g s ]σ (s )  [g t ]σ (s ) 
and (s)  Bt whenever |st| < .
We now show that Bs  Bt  Ds  Dt   for |s t| < . For this we will show (s)  Bs  Bt 
Ds Dt for |st| < . If |st| <  then
 COMPLEX ANALYSIS137

|(s)  (s)| <  < R(s)

so that (s)  Ds.
Also |(s)  (t)| = | (s)  (s) + (s)  (t)|
 |(s)  (s)| + |(s)  (t)| < 2 < R(t)
 (s)  Dt.
Since we already have (s)  Bs  Bt by (1), so
(s)  Bs Bt  Ds  Dt = G.
Since t  T, it follows that ft(z) = gt(z) for all z in G.
Also (1) implies fs(z) = ft(z) and gs(z) = gt(z) for all z in G.
Thus fs(z) = gs(z) for all z in G.
But since G has a limit point in Bs  Ds, we must have s  T.
That is, (t  , t + )  T
 T is open.
Similarly we can show that T is closed.
 T is non-empty open and closed subset of [0, 1]. As [0, 1] is connected, we have
[0, 1] = T.
Thus 1  T and the result is proved.
3.10. Definition. Let (f, D) be a function element and let G be a region which contains D; then
(f, D) admits unrestricted analytic continuation in G if for any path  in G with initial point in D
there is an analytic continuation of (f, D) along .
3.11. Theorem (Monodromy Theorem). Let (f, D) be a function element and let G be a region
containing D such that (f, D) admits unrestricted continuation in G. Let a  D, b  G and let 0
and 1 be paths in G from a to b; let {(ft, Dt) : 0 < t  1} and {(gt, Dt) : 0  t  1} be analytic
continuations of (f, D) along 0 and 1 respectively. If 0 and 1 are FEP homotopic in G then
[f1]b = [g1]b.
Proof. Since 0 and 1 are fixed end point homotopic in G there is a continuous function F : [0, 1]
 [0, 1]G such that
F(t, 0) = 0(t), F(t, 1) = 1(t)
F(0, u) = a, F(1, u) = b
For all t and u in [0, 1]
Let u be a fixed point of [0, 1]. Consider the path u, defined by
u(t) = F(t, u) for all t  [0, 1].
Then u(0) = F(0, u) = a, u(1) = F(1, u) = b
 u is a path from a to b.
By hypothesis there is an analytic continuation
{(ht,u, Dt,u) : 0  t  1}
of (f, D) along u.
Now {(ht,0, Dt, 0) : 0  t  1} and {(ft, Dt) : 0  t  1} are analytic continuations along 0 so by
theorem 3.6, we have
[f1]b = [h1,0]b
Similarly, [g1]b = [h1,1]b
To prove the theorem, it is sufficient to show
[h1,0]b = [h1,1]b
Consider the set
U = {u  [0, 1] : [h1,u]b = [h1,0]b. We will show 1  U.
138

Now, 0  U. So U  .
We claim U is both open and closed subset of [0, 1].
Let u be any point in [0, 1].
We assert that there is  > 0 such that if |u v| < 
then [h1,u]b = [h1,v]b (2)
By lemma 3.9, there an  > 0 such that if  is any path from a to b with |u(t)  (t)| <  for all t
and if {(kt, Et)} is any continuation of (f, D) along , then
[h1,u]b = [k1]b (3)
Now F is uniformly continuous function so there is  > 0 s.t.
|F(t, u)  F(t, v)| <  whenever |u v| < 
i.e. |u(t)  v(t)| <  whenever |u v| < .
So for |u v| < , v is a path from a to b with
|u(t)  v(t)| <  for all t and {(ht,v, Dt,v)} is continuation of (f, D) along v,
so by (3),
[h1,u]b = [h1,v]b
Suppose u U sp that [h1,u]b =[h1,0]b. Then as proved above, there is a  > 0 s.t. |u v| < 
implies [h1,u]b = [h1,v]b
i.e. v  (u  , u + ) implies [h1,v]b = [h1,0]b
i.e. v (u  , u + ) implies v  U
i.e. (u  , u + )  U.
Hence U is open.
To show U is closed, we show U = U. Let u  U and  be the +ve number satisfying (2). Then
there is a v  U such that
|u v| < 
So by (2), [h1,u]b = [h1,v]b.
Since v  U, [h1,v]b = [h1,0]b Therefore [h1,u]b = [h1,0]b. so that u  U.
Thus U is closed as U = U.
Now U is a non-empty open and closed subset of [0, 1] and [0, 1] is connected.
 [0, 1] = U
So 1 U and the result is proved.
The following corollary is the main consequence of the Monodromy theorem.
3.12. Corollary. Let (f, D) be a function element which admits unrestricted continuation in the
simply connected region G. Then there is an analytic function F : G C such that F(z) = f(z) for
all z in D.
Proof. Let a be a fixed point in D and z is any point in G. If  is a path in G from a to z and
{(ft, Dt) : 0  t  1} is an analytic continuation of (f, D) along  then let F(z, ) = f1(z) since G is
simply connected,
F(z, ) = F(z, ) for any two paths  and  in G from a to z. Thus F(z) = F(z, ) is a well
defined function from G to C. To show that F is analytic let z  G. Let  be a path in G from a to
z and {(ft, Dt)} be the analytic continuation of (f, D) along . Then F() = f1() for all  in a
neighbourhood of z. Hence F must be analytic.
4. Harmonic Functions on a Disk
If G is an open subset of  then a function u : GR is called harmonic if u has continuous
second partial derivatives and
 COMPLEX ANALYSIS139

 2u  2u
 0
x 2 y 2
We recall the following facts about harmonic functions.
(i) A function f on a region G is analytic iff Ref = u and Imf = v are harmonic functions
which satisfy the Cauchy-Riemann equations.
(ii) A region G is simply connected iff for each harmonic function u on G there is a
harmonic function v on G such that f = u + iv is analytic on G.
If f : G  is an analytic function then u = Re f and v = Im f are called harmonic
conjugates.

With this terminology, (ii) implies that every harmonic function on a simply connected
region has a harmonic conjugate. If u is a harmonic function on G and D is a disk s.t. D 
G then there is a harmonic function v on D s.t. u + iv is analytic on D.
4.1. Theorem. If u : G  is harmonic then u is infinitely differentiable.
Proof. Let z0 = x0 + iy0 be a fixed point in G and  > 0 be choosen s.t.
B(z0 ; )  G.
Then u has a harmonic conjugate v on B(z0 ; ), that is, f = u + iv is analytic and hence infinitely
differentiable on B(z0 ; ). Thus it follows that u is infinitely differentiable.
The next result gives a property that harmonic functions share with analytic functions.
4.2. Mean Value Theorem. Let u : GR be a harmonic function and let B (a : r) be a closed
disk contained in G. If  is the circle |z  a| = r
1 2π
then u(a) =  u(a + rei) d.
2π 0
Proof. Let D be a disk such that B(a ; r)  D  G and let f be an analytic function on D such
that u = Re f. Then by Cauchy‟s Integral formula,
1 f ( z)
f(a) =  dz.
2πi γ z  a
1 2π
=  f(a + rei) d where z  a = rei
2π 0
1 2π
 u(a) + i v(a) =  [u (a + rei) + i v(a + rei)] d where v = Im f.
2π 0
Equating real parts on both sides, we get
1 2π
u(a) =  u(a + rei) d
2π 0
4.3. Definition. A continuous function u : GR has the Mean Value Property (MVP) if
whenever B (a ; r)  G,
1 2π
u(a) =  u(a + rei) d.
2π 0
4.4. Maximum Principle. Let G be a region and suppose that u is a continuous real valued
function on G with the MVP. If there is a point a in G such that u(a)  u(z) for all z in G then u
is a constant function.
140

4.5. Definition. The function


Pr() =  r|n| ein
n 
for 0  r < 1 and  <  < , is called the Poisson kernel.
Let z = rei where 0  r < 1.
1  r e iθ 1  z
Then 
1  r e iθ 1  z
= (1 + z) ( 1z)1
= (1+z) (1 + z + z2 + z3 + …)

= 1 + 2  zn
n 1
 
= 1 + 2  rn ein = 1 + 2  rn (cos n + i sin n)
n 1 n 1
 1  r e iθ  
 Re  iθ
 = 1 + 2  rn cos n

1 r e  n 1

= 1  rn (ein + ein)
n 1
= Pr()
1 r e i  1  r e i  1  r e i  1  r e i  r e  i   r 2
Also i
  
i   i

 1  r e i  r e  i  r 2
1 r e  1  r e  1  r e 
(1  r 2 )  2ir sin θ
=
1  2r cosθ  r 2
 1  r e iθ  1 r2
so that Pr() = Re  iθ

 1  2r cosθ  r 2
1 r e 
4.6. Proposition. The Poisson kernel satisfies the following:
1 π
(i)  Pr() d = 1 ;
2π  π
(ii) Pr() > 0 for all ,
Pr () = Pr()
and Pr is periodic in  with period 2
(iii) Pr() < Pr() if 0 <  < ||   ;
(iv) for each  > 0, lim Pr() = 0 uniformly in  for   |  |  .
r 1


Proof. (i) For a fixed value of r, 0  r < 1, the series  r |n| ein converges uniformly in  t0
n 
Pr(). So applying term by term integration, we have

π π
 r e
|n| in
  π Pr()d =   π
n 
 
=  r |n|  ππ ein d = 2 + 2  rn  ππ cos n d = 2.
n  n 1
1 π
  Pr() d = 1
2π  π
 COMPLEX ANALYSIS141

1 r2
(ii) We have Pr() =  0 since r < 1.
| 1  r e iθ | 2
1 r2
Also Pr() =
1  2r cos θ  r 2
1 r2 1 r2
 Pr() =   Pr (θ)
1  2r cos(θ)  r 2 1  2r cosθ  r 2
Further since cos  is periodic with period 2, it follows from
1 r2
Pr() =
1  2r cos θ  r 2
that Pr() is also periodic in  with period 2.
(iii) Let 0 <  <   . Define f : [, ]R
by f(t) = Pr(t) for all t in [, ].
 2(1  r 2 )r sin θ
Then f (t) = Pr(t) = 0
(1  2r cosθ  r 2 ) 2
 f() > f() and so Pr() < Pr()
We have 0 < Pr()  Pr() if   ||   (1)
1 r2
Also Pr() =
1  2r cos δ  r 2
1 r2
 lim Pr() = lim 0
r 1 r 1 1  2r cos δ  r 2

Thus (1) implies lim Pr() = 0

r 1
Hence lim Pr() = 0 uniformly in  for   ||  .
r 1

5. The Dirichlet’s Problem

The Dirichlet‟s Problem consists in determining all regions G such that for any continuous
function f : G  R there is a continuous function u : GR such that
u(z) = f(z) for z in G
and u is harmonic in G.
The next theorem states that the Dirichlet‟s Problem can be solved for the unit disk.
5.1. Theorem. Let D = {z : |z| < 1} and suppose that f : DR is a continuous function. Then
there is a continuous function
u : DR
such that
(a) u(z) = f(z) for z in D ;
(b) u is harmonic in D.
Moreover u is unique and is defined by the formula
1 π
u(r ei) =  Pr ( t) f(eit) dt
2π  π
for 0  r < 1, 0    2.
Proof. Define u : D  R
142

1 π
  Pr (θ  t ) f (e )dt for 0  r  1
it
as u(r ei) =  2π  π
 f ( r e iθ ) for r  1

Then u(ei) = f(ei)
 u(z) = f(z) for z in D.
It remains to show that u is continuous on D and harmonic in D.
(i) u is harmonic in D.
If 0  r < 1 then
1 π
u(r ei) =  Pr(t) f(eit)dt
2π  π
1 π 1  r e i (θ t ) 
=  π Re  i (θ t ) 
f(eit)dt
2π 1  r e 
 1 1  r e i (θt )  
= Re   ππ f (e it )  i (θ t )  
dt
 2π 1  r e  
 1  e it  r e iθ  
= Re   ππ f (e it )  it iθ  
dt
 2π  e  r e  
Define g : D  by
it  e  z 
it
1 π
g(z) =   π f (e )  it  dt
2π e  z 
Since u = Re g, then g is analytic it follows that u is harmonic in D.
(ii) u is continuous on D.
Since u is harmonic on D, it only remains to show that u is continuous at each point of the
boundary of D. For this we prove the following.
Given  in [, ] and  > 0 there is a , 0 <  < 1 and an arc A of D about ei such that for
 < r < 1 and ei in A,
|u (r ei)  f(ei)| < 
We prove the result by taking  = 0.
Since f is continuous at z = 1, there is a  > 0 such that
1
|f(ei)  f(1)| <  if || < . (1)
3
Let M = max. {1 f(ei) : ||  }
Since for each  > 0,
lim Pr() = 0 uniformly in , for   ||  ,
r 1
there is a number , 0 <  < 1, such that

Pr() < (2)
3M
δ
for  < r < 1 and || 
2
1
Let A be the arc {e : || < }. Then if ei  A and  < r < 1,
i
2
 COMPLEX ANALYSIS143

1 π
u(r ei)  f(1) =  Pr(t) f(eit)dt  f(1)
2π  π
1 π
=  Pr( t) [f(eit)  f(1)]dt
2π  π
1 1
 Pr(t) [f(e )  f(1)]dt +  Pr(t)[f(e ) f(1)]dt
it it
=
2π |t|δ 2π |t|δ
1 δ δ
If |t|   and ||   then |t  |  |t|  ||    
2 2 2
So from (1) & (2), it follows that
1   
|u (r ei)  f(1)|   2M  
3  3M 
i.e. |u(r ei)  f(ei)| <  for  = 0
Since f is continuous function, it follows that u is continuous at e i.
Finally to show u is unique, suppose that v is a continuous function an D such that v is
harmonic on D and v(ei) = f(ei) for all . Then u  v is harmonic in D and (u  v) (z) = 0 for all
z in D.
So u  v  0 is u  v .
Cor. (a) : If u : D  R is a continuous function that is harmonic in D
1 π
then u(r ei) =  Pr( t) u (eit) dt
2π  π
for 0  r < 1 and all .
Moreover, u is the real part of the analytic function
1 π e it  z
f(z) =  u(eit)dt
2π π e it  z
Cor. (b): Let a   ,  > 0 and suppose h is a continuous real valued function on {z : |za| = };
then there is a unique continuous function w : B(a; ) R such that w is harmonic on B(a : )
and w(z) = h(z) for |z a| = .
Proof. Consider f(ei) = h(a + ei)
Then f is continuous on D.
So  a continuous function u :D  R such that
u is harmonic in D and u(ei) = f(ei).
Define w : B(a ; )  R
za 
as w(z) = u   for z B (a ; ).
 ρ 
Then w is harmonic on B(a ; ) since u is harmonic on D.
Also for z = a + ei,
w(z) = u(ei) = f(ei) = h(a + ei) = h(z).
Hence w(z) = h(z) for |za| = .
We shall use the following result in proving next theorem.
5.2. Lemma. Let G be a bounded region and suppose that w :G  R is a continuous function
that satisfies the MVP on G. If w(z) = 0 for all z in G then w(z) = 0 for all z in G.
5.3. Theorem. If u : G  R is a continuous function which has the MVP then u is harmonic.
144

Proof. Let a  G and choose  s.t. B(a ; )  G.

It is sufficient to show u is harmonic on B(a ; )
By last cor., there is a continuous function
w :B(a ; )  R
which is harmonic in B(a ; ) and w (a + ei) = u(a + ei) for all . Since uw satisfies the
MVP and (u  w) z = 0 for |za| = , it follows that u  w in B(a, ). Since w is harmonic on
B(a, ); we have u must be harmonic.
5.4 Harnack’s Inequality. If u :B( a ; R)  R is continuous harmonic in B(a ; R) and u  0,
then for 0  r < R and all ,
R r Rr
u(a)  u(a + r ei)  u(a).
Rr Rr
Proof. Define w :D  R as
w ( ei) = u(a + Rei) for 0    1
Then w is continuous function onD s.t. w is harmonic in D. So by cor. (a) to theorem 5.1,
1 π
w( ei) =  P( t) w(eit) dt for 0   < 1 and all .
2π  π
1 π 1 ρ2
=  π w(eit)dt (1)
2π 1  2ρ cos(θ  t )  ρ 2
1 r2
Since 0  r < R so 0 
r
 1.  Pr (θ) 
R 1  2r cosθ  r 2
r
Replacing  by in (1), we get
R
r2
1
r  1 π R2
w  e iθ  =  π w (e it )dt
 R  2π 2r r2
1  cos(θ  t )  2
R R
1 π R 2
 r 2
or u(a + r ei) =  .u (a + R eit) dt (2)
2π π R 2  2rR cos(θ  t )  r 2
Now R  r  | Reit  r ei|  R + r
 (Rr)  |R eit  r ei|2  (R + r)2
2

1 1 1
  iθ 2

(R  r ) 2
| Re  r e |
it
(R  r ) 2
Multiplying by (R  r ), we get
2 2

R r R2  r2 Rr
 
R  r | R e it  r e iθ |2 R  r
Rr R2  r2 Rr
  2 
R  r R  2rR cos(θ  t )  r 2 R  r
u (a  R e it )
Multiplying by and integrating w.r.t t between the limits  to , we get
2π
1 R r π 1 Rr π
   u(a+Re )dt  u(a +re )     u(a + Re )dt
it i it
2π  R  r  π 2π  R  r  π
 COMPLEX ANALYSIS145

[using(2)]
Rr Rr 1 π
 u(a)  u(a + r e )  
i it
that is   u(a) since u(a) =  π u(a +R e )dt
Rr R r 2π
Hence the result.
5.5. Definition. If G is an open subset of  then Har(G) is the space of all harmonic functions
on G.
5.6. Remark. Since Har(G)  C(G, R) , Har(G) is given the metric that it inherits from C(G, R)
We now prove Harnack‟s theorem. The following results will be used.
(i) Let (X, d) be a complete metric space and Y  X. Then Y is complete iff Y is closed.
(ii) Let (X, d) be a metric space. Then a set A  X is closed iff for every sequence <xn>
in A with xnx, we have xA.
(iii) A metric space is connected iff it is not the union of two non-empty disjoint open
sets.
5.7. Harnack’s Theorem. Let G be a region
(a) The metric space Har(G) is complete
(b) If {un} is a sequence in Har(G) such that u1  u2 … then either un(z) uniformly
on compact subsets of G or {un} converges in Har(G) to a harmonic function.
Proof. (a) We know C(G, R) is complete metric space and Har(G)  C(G, R) so to show Har(G)
is complete, it is sufficient to show that Har(G) is a closed subspace of C(G, R).
Let {un} be a sequence in Har(G) such that unu in C(G, R). Then {un} converges uniformly to
u in C(G, R).
  u = lim  un (1)
n 
Let B (a ;r) be a closed disk contained in G. Then
1 2π
un(a) =  un(a + r ei) d (2)
2π 0
as un is a harmonic function.
Now, {un(a)} converges to u(a) so using (1) and we have
1 1 2π
u(a) = lim un(a) = lim  2 π un(a + ri) d = i
 u (a+re ) d.
n  2π n  0 2π 0
This shows that u has the MVP. Now u : G R is a continuous function having MVP. So u is
harmonic i.e. u Har(G). Thus Har(G) is closed and so complete.
(b) Assume that u1  0 because, otherwise, we may replace un by un  u1. Let
u(z) = sup {un(z) : n  1} for each z in G. Then for each z in G, we have either u(z)   on
u(z)  R and un(z)  u(z).
Define A = {z G : u(z) = }
B = {z G : u(z) < }
Then G = AB and A  B =  .
We will show that both A and B are open. Let a G and R be chosen such that B(a ; R)  G.
Then by Harnack‟s inequality
R | z  a | R | z  a |
u n (a )  u n (z)  u n (a ) (3)
R | z  a | R | z  a |
for all z in B(a ; R) and all n  1.
146

R | z  a |
If aA then un(a) so that u n (a )  un(z) implies un(z) for all z in B(a ; R) that
R | z  a |
is,
B(a : R)  A.
So, ‘a’ is interior point of A. But ‘a’ is arbitrary point of A. So every point of A is its
interior point and hence A is open.
If aB then u(a) < . Using right half of (3), we have
u(z) <  for |za| < R
i.e. u(z) <  for all z in B(a ;R)
 B(a; R)  B and so B is open.
Since G is connected, we have either A =  or B =  that is, either B = G or A = G.
Suppose A = G. Then u  
R ρ
Also ifB(a ; R)  G and 0 <  < R hen M = > 0 and so (3) implies.
R ρ
M un(a)  un(z) for |za|  .
Hence un(z) uniformly for z inB (a ;). Thus we have shown that for each a in G there is a
 > 0 such that un(z) uniformly for |za|  . So un(z)  uniformly for z in any compact
set.
Now suppose B = G. Then u(z) <  for all z in G. If  < R then there is a constant N, which
depends only on a and  such that
M un(a)  un(z)  N un(a) for |za|   and all n.
So if m  n, we have
0  un(z)  um(z)
 N un(a)  M un(a)
 c [un(a)  um(a)]
for some constant c .
Thus {un(z)} is uniformly Cauchy sequence onB (a ; ). It follows that {un} is a Cauchy
sequence in Har(G). Har(G) is complete, so {un} must converge to a harmonic function. Since
un(z)  u(z), we have {un} converges to u is Har(G). This completes the proof of the theorem.
5.8. Subharmonic and Superharmonic Functions. Let G be a region and let  : G  R be a
continuous function. Then  is called subharmonic function if wheneverB(a ; r)  G,
1 2π
(a)   (a + r ei)d
2π 0
Also  is called a superharmonic function if wheneverB (a ; r)  G,
1 2π
(a)   (a + r ei)d
2π 0
Clearly every harmonic function is subharmonic as well as superharmonic. In fact, u is harmonic
iff u is both subharmonic and superharmonic. Also observe that  is superharmonic iff  is
subharmonic.
5.9. Definition. If G   then the boundary of G in  is called extended bounded of G and is
denoted by G.
Clearly G = G if G is bounded and
 COMPLEX ANALYSIS147

G = G  {} if G is unbounded.

5.10. Maximum Principle: Let G be a region and let  and  be real valued functions defined
on G such that  is subharmonic and  is superharmonic. If for each point a in G
lim sup (z)  lim inf (z),
z a z a
then either (z) < (z) for all z in G or  =  and  is harmonic.
5.11. Definition. If G is a region and f : G  R is a continuous function then the Perron
Family P(f, G), consists of all subharmonic functions
:GR
such that lim sup (z)  f(a) for all a in G.
z a
Since f is continuous, there is a constant M such that
|f(a)|  M for all a in G
So the constant function M is in P(f, G) and the Perron Family is never empty.
5.12. Theorem. Let G be a region and f : G  R be a continuous function :
then
u(z) = sup{(z) :  P(f, G)} defines a harmonic function u on G.
5.13. Definition. Let G be a region and f : G R be a continuous function then the harmonic
function u defined by
u(z) = sup {(z) :   P(f, G)}
is called the Perron Function associated with f.
5.14. Definition. A region G is called a Dirichlet’s Region if the Dirichlet‟s Problem can be
solved for G. That is, G is a Dirichlet‟s Region if for each continuous function f : GR there
is a continuous function u :G  R such that u is harmonic in G and
u(z) = f(z) for all z in G.
By theorem 5.1, it follows that a disk is a Dirichlet‟s Region, but the punctured disk is not, as
shown below.
Let G = {z : 0 < |z| < 1}, T = {z : |z| = 1} so that G = T  {0}
Define f : G  R by
f(z) = 0 if z  T
and f(0) = 1.
log | z |
For 0 <  < 1, let u(z) =
log 
Then u is harmonic in G, u(z) > 0 for z in G,
u(z) = 0 for z in T and u(z) = 1 if |z| = 
Suppose that v  P(f, G)
Since |f|  1, |v(z)|  1 for all z in G.
If R = {z :  < |z|< 1} then lim sup v(z)  u(a) for all a in R. By the maximum principle,
z a
v(z)  u(z) for all z in R.
Since  is arbitrary, this gives that for each z in G,
v(z)  lim u(z) = 0
 0

Hence the Perron function associated with f is identically zero function and the Dirichlet‟s
Problem cannot be solved for the punctured disk.
148

In this section we will see conditions that are sufficient for a region to be a Dirichlet
Region.
Notation. For a set G and a point a in G, let G(a ; r) = G  B(a ; r) for all r > 0.
5.15. Definition. Let G be a region and let a G. A barrier for G at „a‟ is a family {r : r > 0}
of functions such that
(a) r is defined and superharmonic on G(a ; r) with 0  r(z)  1;
(b) lim r(z) = 0
z a
(c) lim r(z) = 1 for w in G  {w : |w  a| = r}.
z w

5.16. Theorem. If G is a Dirichlet Region then there is a barrier for G at each point of G.
Proof. Suppose a  G s.t. a  .
|za |
Let f(z) = for z  
1 | z  a |
with f() = 1. Then f is continuous on G.
So there is a continuous function u :G  R such that u is harmonic on G and u(z) = f(z) for z in
G. In particular, u(a) = 0 and a is the only zero of u inG.
Let cr = inf {u(z) : |za| = r, z  G}
= min {u(z) : |za| = r, z G} > 0
Define r : G(a ; r)  R by
1
r(z) = min {u(z), cr}.
cr
Then {r} is a barrier for G at a.
The next result provides a converse of above theorem.
5.17. Theorem. Let G be a region and let a  G such that there is a barrier for G at a. If f :
G R is continuous and u is the Perron Function associated with f then
lim u(z) = f(a).
z a

Proof. Let {r : r > 0} be a barrier for G at a.

For convenience assume that a  .
Also by replacing f by f  f(a), if necessary, we can suppose that f(a) = 0.
Let  > 0 and choose  > 0 such that |f(w)| <  whenever w  G and |wa| < 2. Let  = .
Let ψ̂ : G  R be defined by
ψ̂ (z) = (z) for z in G(a ; )
and ψ̂ (z) = 1 for z in G  B(a ; ).
Then ψ̂ is superharmonic.
If |f(w)|  M for all w in G, then M ψ̂   is subharmonic.
We claim that M ψ̂  .  P(f, G).
If w  G  B(a ; ) then
lim sup [M ψ̂ (z)  ] = M   < f(w).
z w
Because ψ̂ (z)  0, it follows that
 COMPLEX ANALYSIS149

lim sup [M ψ̂ (z)  ]    for all w in G.

z w
In particular, if w  G  B(a; ) then
lim sup [M ψ̂ (z)  ]   < f(w) by the choice of .
z w
Hence M ψ̂ (z)    u(z) for all z in G. (1)
Similarly
lim inf [M ψ̂ (z) + ]  lim sup (z)
z w z w
for all  in P(f, G) and w in G. By Maximum Principle,
(z)  M ψ̂ (z) +  for  in P(f, G) and z in G
Hence u(z)  M ψ̂ (z) +  for all z in G (2)
From (1) & (2),
M ψ̂ (z)    u(z) < M ψ̂ (z) +  for all z in G.
But lim ψ̂ (z) = lim (z) = 0 so
z a z a
  lim u(z)   for all z in G.
z a
As  is arbitrary +ve number, we get
lim u(z) = 0 = f(a).
z a
This completes the proof.
Cor. A region G is a Dirichlet Region iff there is a barrier for G at each point of G.
6. Green’s Function
Here we introduce Green‟s function and discuss its existence. Such function plays an important
role in differential equations and other fields of analysis.
6.1. Definition Let G be a region in the plane and let a  G. A Green‟s Function of G with
singularity at a is a function ga : G  R with the properties.
(a) ga is harmonic in G {a} ;
(b) g(z) = ga(z) + log |za| is harmonic in a disk about a;
(c) lim ga(z) = 0 for each w in G.
z w

6.2. Remarks. (1) For a given region G and a point a in G, ga need not exist. However, if it
exists, it is unique.
To prove this suppose ha is another Green‟s Function for G with singularity at a. Then haga is
harmonic in G. But (c) implies lim [ha(z)  ga(z)] = 0 for every w in G. So by Maximum
z w
Principle, we have ha = ga.
(2) A Green‟s Function is positive In fct, ga is harmonic in G {a} and lim g(z) = + 
z a

since g (az ) log |za| is harmonic at z = a. So by Maximum Principle.

ga(z) > 0 for all z in G{a}.

(3)  has no Green‟s Function with a singularity at zero.

Suppose that g0 is the Green‟s Function with singularity at zero. Let g = g0 so g(z) < 0 for all z
as g0 is positive.
150

We will show that g must be constant function. For this let z1, z2 be two complex numbers s.t.
0  z1  z2  0. Let  > 0 be given. Then there is a  > 0 such that
|g(z)  g(z1)| <  if |z1  z| < .
So g(z) < g(z1) +  if |z z1| < .
Let r > |z1  z2| > . Then
g ( z1 )   z  z1
hr(z) = log
δ r
log 
 r
is harmonic in   {z1}.
Also g(z)  hr(z) for z on the boundary of the annulus
A = {z :  < |z  z1| < r}. By the Maximum Principle,
g(z)  hr(z) for z in A.
In particular, g(z2)  hr(z2).
Letting r  , we get
g(z2)  lim hr(z2) = g(z1) + 
r
Since  was arbitrary chosen positive number,
g(z2)  g(z1)
Interchanging the role of z1 and z2, we have
g(z1)  g(z2)
 g(z1) = g(z2)
Hence g must be a constant function which is a contradiction. Thus  has no Green‟s function
with a singularity at zero
The next theorem shows when do Green‟s Functions exist.
6.3. Theorem. If G is a bounded Dirichlet Region then for each a in G there is a Green‟s
Function on G with singularity at a.
Proof. Define f : G  R by
f(z) = log |z  a|
and let u :G  R be the unique continuous function which is harmonic on G such that
u(z) = f(z) for z in G.
Then ga(z) = u(z)  log |z  a| is the required Green‟s Function on G with singularity at a.
The next result shows that Green’s Functions are conformal invariants.

6.4. Theorem. Let G and  be regions such that there is a one-one analytic function f of G onto
 ; let a  G and  = f(a).
If ga and  are the Green‟s Function for G and  with singularities a and  respectively, then
ga(z) =  (f(z))
Proof. Let  : GR be defined by  =  of.
We shall show  = ga.
For this it is sufficient to show that  has the properties of the Green’s Function with
singularity at z = a.
Clearly  is harmonic in G{a}.
 COMPLEX ANALYSIS151

Let w  G and {zn} be a sequence in G with znw.

Then {f(zn)} is a sequence in . So there is a subsequence {znk} such that f(znk)  f(w) in .
So (f(znk))  0
Since this happens for any convergent subsequence of {f(zn)}, it follows that
lim (zn) = lim (f(zn)) = 0.
n  n 
Hence lim (z) = 0 for every w in G.
z w
By power series expansion of f about z = a, we have
f(z) = f(a) + A1 (za) + A2 (za)2 +…..
or f(z)   = (z a) [A1 + A2 (za) +….]
Hence log |f(z)  | = log |za| + h(z) (1)
where h(z) = log |A1 + A2 (z a) +….| is harmonic near z = a since A1  0
Suppose (w) = (w)  log |w| where  is a harmonic function on . Since f : G is onto
and w  so
w = f(z) for some z  G.
Thus  (f(z)) = (f(z))  log |f(z)  |
i.e. (z) = [(f(z))  h(z)]  log |z a| [using (1)]
Since 0f  h is harmonic near z = a, (z) + log |za| is harmonic near z = a. Therefore  is a
Green‟s Function of G with singularity at a. Hence it follows by uniqueness of Green‟s Function.
ga = 
i.e. ga(z) = (z)
i.e. ga(z) = (f(z))
Hence the result.
7. Canonical Product
We recall the Weierstrass factorization theorem for entire functions. Let f(z) be an entire
function with a zero of multiplicity m  0 at z = 0. Let {zn} be the non-zero zeros of f(z),
arranged so that a zero of multiplicity K is repeated in this sequence K times. Also suppose that
|z1|  |z2| ….. . If {pn} is a sequence of integers such that
p n 1
  R 
   < , for every R > 0, then
n 1  | z n | 

P(z) =  E p (z / z n ) (1)
n 1 n

converges uniformly on compact subsets of the plane, where by definition of primary factors, we
have
 z2 zp 
Ep(z) = (1z)exp  z   ...   (2)
 2 p 
for p  1 and E0(z) = 1z
Then the Weierstrass theorem says that
f(z) = zm eg(z) P(z) (3)
where g(z) is an entire function.
152

We are interested in the case in which g(z) and P(z) have certain characteristics which result in
properties of f(z) and conversely. A convenient assumption for P(z) is that all the integers pn are
equal. Then we see that this is to assume that there is an integer p  1 such that

 | z n |( p1) <  (4)
n 1

i.e. it is an assumption on the growth rate of the zeros of f(z). Further, if we assume that p is the
smallest integer for which the series (4) converges, then the product

P(z) =  E p (z / z n ) (5)
n 1

is called the canonical product associated with the sequence {zn} of zeros of f(z) and the integer
p is called the genus of the canonical product. The restriction on g(z), we impose, is that it is a
polynomial. Such an assumption must impose a growth condition on eg(z). When g(z) is a
polynomial, then we say that f(z) is of finite genus and we define the genus of f(z) to be the
degree of this polynomial or to be the genus of the canonical product whichever is greater.
Now we drive Jensen‟s formula which says that there is a relation between the growth
rate of the zeros of f(z) and the growth of M(r) = sup {|f(rei)| : 0    2} as r increases. For
this, we shall use Gauss-Mean Value Theorem which states that if f(z) is analytic in a domain D
which contains the disc |z  z0|  , then
1 2π
f(z0) =  f(z0 +ei) d
2π 0
If u is the real part of f(z), the above result gives Gauss-mean value theorem for harmonic
function, as
1 2π
u(z0) =  u(z0 +ei) d
2π 0
7.1. Jensen’s Formula. Let f(z) be analytic in the closed disc |z|  R and let f(0)  0, f(z)  0 on
|z| = R. If z1, z2,…, zn are zeros of f(z) in the open disc |z| < R repeated according to their
multiplicity, then

n  R  1 2π
log |f(0)| =   log   i
 0 log | f(Re )d.
i 1  i 
| z | 2 π
Proof. Consider the function
n R 2  zi z
F(z) = f(z)  (1)
i 1 R (z  z i )
We observe that F(z) is analytic in any domain in which f(z) is analytic and further F(z)  0 for
|z|  R. Hence F(z) is analytic and never vanish on an open disc |z| <  for some  > R.
Also |F(z)| = |f(z)| (2)
on |z| = R, since
n
R2  z z n
R 2  z i Re i
 R (z  zi ) 
R 2 e i  Rz i
, z = Rei
i 1 i i 1

n
R ( R  z i e i )
=
i 1 Re i (R  z i e i )
 COMPLEX ANALYSIS153

n R  z i e i
=  , |ei| = 1
i 1 R  z i e i
=1
Since F(z) is analytic and non-zero in |z| < , log F(z) is analytic in z| <  and consequently its
real part log |F(z)| is harmonic there. Hence using Gauss-Mean value theorem for log |F(z)|, we
get
1 2π
log |F(0)| =  log |F(Rei)|d (3)
2π 0
Now, from (1),
n R
F(0) = f(0)   
i 1  z i 
n R
so that |F(0)| = |f(0)| 
i 1 | z i |
and thus
n R
log |F(0)| = log |f(0)| +  log
i 1 | z i |
i i
Also by (2), |F(Re )| = |f(Re )| on |z| = R.
Therefore (3) becomes
n R 1 2
 0
log |f(0)| +  log = log |f(Rei)|d
i 1 | z i | 2
or
n R 1 2π
log |f(0)| =   log +  log |f(Rei)|d
i 1 | z i | 2π 0
7.2. Poisson-Jensen Formula. Let f(z) be analytic in the closed disc |z|  R and let f(z)  0 on |z|
= R. If z1, z2,…, zn are the zeros of f(z) in the open disc |z| < R repeated according to their
multiplicity and z = rei, 0  r < R, then
n R 2  zi z
log |f(z)| =   log
i 1 R (z  z i )
1 2  (R 2  r 2 ) log | f (Re i ) |
2 0 R 2  2Rr cos(  )  r 2
+ d.

Proof. Consider the function

n R 2  zi z
F(z) = f(z)  (1)
i 1 R (z  z i )
Clearly F(z) is analytic in any domain in which f(z) is analytic and F(z)  0 for |z|  R. Hence
F(z) is analytic and never vanish on an open disc |z| <  for some  > R. Also
|F(z)| = |f(z)| on |z| = R
Since F(z) is analytic and non-zero in |z| < , log F(z) is analytic in |z| <  and consequently
its real part log |F(z)| is harmonic there. Hence using Poisson integral formula (unit-I) for
log |F(z)|, we get
154

1 2  (R 2  r 2 ) log | F(Re i ) |
2 0 R 2  2Rr cos(  )  r 2
log |F(z)| = d (2)

Now, log |F(Rei)| = log |f(Rei)| on |z| = R.

n R 2  zi z
Also log |F(z)| = log |f(z)| 
i 1 R (z  z i )
n R 2  zi z
= log |f(z)| +  log
i 1 R (z  z i )
Therefore (2) becomes
n R 2  zi z
log |f(z)| =   log
i 1 R (z  z i )
1 2  (R 2  r 2 ) log | f (Re i ) |
2 0 R 2  2Rr cos(  )  r 2
+ d

7.3. The Maximum Modulus of an Entire Function. Let f(z) be a non-constant entire function.
Define
M(r) = Max {|f(z)| : |z|  r}
Since f(z) is entire, by maximum modulus principle, |f(z)| reaches its maximum value M(r) on
the circle |z| = r so that
M(r) = max {|f(z)| : |z| = r}
In fact, M(r) is a steadily increasing unbounded function of r. For this, by maximum
modulus principle, we have.
|f(r1 ei)| < M(r2)
whenever r1 < r2 and consequently M(r1) < M(r2) and if M (r) were bounded, then by Liouville‟s
theorem, f(z) would be constant.
7.4. Hadmard’s Three Circle Theorem. Let f(z) be analytic in r1  |z|  r3 and let r1 < r2 < r3.
Let Mi be the maximum values of |f(z)| on the circles |z| = ri (i = 1, 2, 3), then

 M1
log( r 3 / r1) log( r3 / r2 ) log( r2 / r1)
M2 .M3
Proof. Let F(z) = zf(z), where  is a real constant to be determined later. Since f(z) is analytic
and also z is analytic, therefore F(z) is analytic in the annulus r1  |z|  r3.
The function F(z) is not, in general, single-valued. But if we cut the annulus along negative real
axis, we obtain a domain in which the principal branch of this function is analytic. By the
maximum modulus principle, |F(z)| attains maximum value on the boundary of the cut annulus.
If we consider a branch of this function which is analytic in the part of the annulus for which
/2  arg z < 3/2, we see that the principal value cannot attain its maximum modulus on the cut
and so must attain it on one of the boundary circles of the annulus. Thus it is shown that when
r1  z|  r3,
|z f(z)|  max. {r1α M1 , r3α M 3 }
Hence if r1 < r2 < r3, we must have
r2α M 2  Max.{r1α M1 , r3α M 3 } (1)
We choose  so that
 COMPLEX ANALYSIS155

r1α M1  r3α M 3
which gives
(r1/r3) = M3/M1
log(M1 / M 3 )
i.e. = (2)
log(r3 / r1 )
With this value of , (1) gives
r2α M 2  r1α M1
or M2  (r2/r1) M1
log( M1 / M3 )

log( r3 / r1 )
= (r2/r1) .M1
Hence,
M 2 log( r3 / r1)  (r2 / r1 )  log( M1 / M3 ) . M1log( r3 / r1)
= (M1 / M3 )  log( r2 / r1) . M1log( r3 / r1)
= M1log( r3 / r2 ) . M3log( r2 / r1) (3)
where we have used the result
a log b  (e log a ) log b  (e log b ) log a  b log a
7.5. Remark. We say that a function f(x) of a real variable x is convex downwards (or simply
convex) if the arc
y = f(x), x1 < x < x2
lies below the chord joining the points (x1, f(x1)) and (x2, f(x2)). Equivalently, the condition may
be stated as
x x x  x1
f(x)  2 f ( x1 )  f ( x 2 ), x1 < x < x2 (4)
x 2  x1 x 2  x1
where the chord has the equation
f ( x 2 )  f ( x1 )
y  f(x1) = (x  x1) (Two point form)
x 2  x1
 x x  x  x1
i.e. y =  2  f ( x1 )  f (x 2 )
 x 2  x1  x 2  x1
Hadmard‟s three circle theorem may no be expressed in convexity form by saying that M(r) is a
convex function of log r since the inequality (3) may be written as (taking logarithm or both
sides)
log r3  log r2 log r2  log r1
log M(r2)  log M(r1) + log M(r3) .
log r3  log r1 log r3  log r1
156

UNIT - V


1. Growth and Order of an Entire Function

We recall that a polynomial Pn(z) of degree n has exactly n zeros. Further the rate of growth as
z of |Pn(z)| increases as its degree n increases. Hence there exists a relationship, via the
degree, between the number of zeros and the growth of the polynomial
Again, let f(z) be a non-constant entire function. We define
M(r) = max. {| f(z)| : |z|  r}
We have already proved that M(r) is a steadily increasing unbounded function of r and
thus M(r) steadily approaches  as r.
It is the growth rate of M(r) which is most easily related to the distribution of zeros of f(z). In
fact, considerable information about entire functions is gained by studying how fast M(r)
approaches infinity. The technique to be used is to compare M(r) for large r with exp(r) for
various , where  is a real constant.
An entire function f(z) is said to be of finite order if there exists a real  such that
M(r)  exp(r) for all sufficiently large r (1)
We also then define the order  of f(z) as
 = inf {  0 : M(r)  exp (r) for sufficiently large r} (2)
i.e. the lower bound  of numbers  for which (1) is true is called order of the entire function
f(z). We write  = + if f(z) is not of finite order i.e. the set in (2) is empty. If f(z) is of order ,
then
M(r)  exp (r+)
for every positive value of  but not for negative values, provided r is sufficiently large.
Functions of finite order are, after polynomials, the simplest integral functions.
1.1. Theorem. Let  be the order of an integral function f(z), then
log log M(r )
 = lim sup
r log r
where M(r) = max |f(z)| on |z| = r.
Proof. Let 1 = inf. {  0 : M(r)  exp (r) for sufficiently large r} (1)
Then by definition, 1 is the order of the function f(z). To prove the result, we are to prove that
1 = , where
log log M(r )
 = lim sup (2)
r log r
ρ1 
Let  > 0 be arbitrary, then (1) suggests that M(r)  exp (r ) for sufficiently large r .
Taking logarithm of both sides, we get
ρ 
log M(r) < r 1
Again, taking logarithm, we find
log log M(r)  (1 + ) log r
 COMPLEX ANALYSIS157

log log M(r )

or < 1 + 
log r
log log M(r )
 lim sup  1 + 
r log r
   1 + 
Since  is arbitrary, so   1 (3)
On the other hand if  > 0, then (2) shows that
log log M(r )
<+
log r
or log log M(r) < ( + ) log r = log r(+)
or log M(r) < r( +)
or M(r) < exp (r+)
It follows that
inf. {  0 : M(r)  exp (r) for sufficiently large r} <  +
i.e. 1 <  + 
Since  is arbitrary, we obtain
1   (4)
From (3) and (4), we conclude that
1 = 
Remark. In view of the above theorem, the order  of an entire function f(z) is also given by
log log M(r )
 = lim sup (5)
r log r
log log M(r )
It should be noted that if lim exists whether finite or infinite, then this limit gives
r log r
the order of f(z). It is only when this limit does not exist that we find the limit (5) to obtain the
order of f(z).
1.2. Example. Find the order of the following functions
(i) f(z) = a0 + a1z + a2z2 +…+ anzn, an  0
(ii) eaz, a  0, (iii) cos z, (iv) sin z, (v) cos z ,
λ
(vi) e z , which  is a positive integer
z
(vii) e e
Solution. (i) Here, M(r) = |an| rn for large |z| = r
log log M(r ) log log (| a n | r n ) 
 lim = lim form
r log r r log r 
1 1
| a n | n r n 1
log(| a n | r ) | a n | r n
n
= lim
r 1/ r
n
= lim 0
r   log(| a | r n )
n
Hence the order of a polynomial is zero.
(ii) Here, M(r) = e|a| r
158

log log M(r ) log (| a | r ) 

 lim = lim form
r log r r log r 
1
|a|
| a |r
= lim 1
r 1/ r
Hence order of eaz is 1
(iii) Since
z2 z4 z6
cos z = 1    .... ,
|2 |4 |6
we find that
| z |2 | z |4
|cos z|  1 + 
|2 |4
r2 r4
1+   ...
|2 |4
e r  e r
= , in the disc |z|  r
2
e r  e r
Thus |cos z|  if |z|  r
2
e r  e r  1  e 2 r 
Hence M(r) = = er  

2  2 
 1  e 2 r 
 log M(r) = r + log  

 2 
 1  1  e 2 r 
= r 1  log 

 r  2 
 1  1  e 2 r 
log r  log 1   

log log M(r )  r  2 
 lim = lim
r log r r log r
  1  1  e 2 r  
 log 1    

  r  2  
= lim 1  
r log r
 
 
 
=1
Thus it follows that order of cos z is 1
(iv) Proceeding as above, we find that order of sin z is also 1
(v) Here, we observe that
e r
 e r
M(r) =
2
 COMPLEX ANALYSIS159

1
and thus as in (iii), the order of cos z comes out to be .
2
λ λ
(vi) Here, M(r) = e r and so by definition, order of e z is 
log log M(r ) λ log r
or lim = lim λ
r log r r   log r

(vii) In this case,

r
M(r) = e e
so that
log log M(r ) r 
lim = lim form
r log r r   log r 
1
= lim = lim r = 
r 1 / r r
z
Hence e e is of infinite order.
For further discussion, we shall need the following theorem.
1.3. Theorem. If the real part of an entire function g(z) satisfies the inequality Re g(z) < r+ for
every  > 0 and all sufficiently large r, then g(z) is a polynomial of degree not exceeding  .
Proof. Since g(z) is entire function, so by Taylor‟s expansion, we have

g(z) = a0 + a1 z + a2z2 +……=  a n z n
n 0
1 g( z)
where an =  n 1 dz (1)
2πi C z
C being the circle |z| = r.
Now when n > 0,
1 g(z) 1   m  dz

   a m z  n 1
2i C
n 1
dz 
2i C z  m 0 z
1  2π m  miθ reiθ
= 0 a m r e n 1 ( n 1) iθ
dθ | z = rei
2 π m 0 r e
where term by term integration being justified in view of the uniform convergence of the
series.
1 g (z) 1  2
So , 
2i C z n 1
dz   a m r mn e ( mn )i d
2  m  0 0
=0 (2)
Thus, from (1) and (2), we get
1 g(z) 1 g(z)
an =  dz   dz
2πi C z n 1 2πi C z n 1
1 g(z)  g(z)
=  dz
2 πi C z n 1
1 2 Re g(z) 1 dθ
=  dz   02 π Re g(reiθ ) n inθ
2πi C z n 1 π r e
Thus, it follows that
160

1 2π
|an|  i
 0 |Re g(re )| d | |ein| = 1
π rn
1
or |an| rn   02 π |Re g(rei)| d (3)
π
On the other hand
1 g ( z) 1 2π
a0 =  dz   g(rei) d
2πi C z 2π 0
1 2π
so Re(a0) =  Re g(rei) d 4)
2π 0
Hence from (3) and (4), we conclude that
1 2π
2 Re (a0) + |an| rn   {|Re g(z)| + Re g(z)} d
π 0
But the integrand is equal to 2 Re g(z) or 0 according as Re g(z) > 0 or  0. Since, by hypothesis
Re g(z) < r+, it follows that
1
2 Re a0 + |an| rn   02 π 2 Re g(z) d
π
1 2π
<  0 2 r+ d
π
2
= r+ 2 = 4r+
π
which holds for >0 and all sufficiently large r.
If we write this inequality in the form
|an| < 4 r+n + (2 Re a0)rn
and then make r, we see that an = 0 when n >  and so g(z) is a polynomial of degree not
exceeding  and hence the proof of the theorem.
1.4. An Estimation of Number of Zeros. We shall denote by N(r) the number of zeros of an
entire function f(z) in the closed disc |z|  r.

1.5. Theorem. If f(z) is an entire function of order , then for every  > 0, the inequality
N(r)  r+ holds for all sufficiently large r.
Proof. Without loss of generality, we may suppose that f(0) = 1. For if f(z) has a zero of order m
cf ( z )
at the origin, we may consider g(z) = , where c is chosen so that g(0) = 1 and since the
zm
functions f(z) and g(z) have the same order, for our consideration it will be unimportant that the
number of zeros of f(z) and g(z) differ by m.
We also assume at first that f(z) has no zero on |z| = 2r and we suppose that the zeros zi of f(z) are
arranged in non-decreasing order of their moduli so that |zi|  |zi+1|.
We apply Jensen‟s formula (Unit-IV) with R replaced by 2r and n = N(2r). We thus have
1 2π N( 2r )  2r 
 0 log |f(2re )| d = log |f(0)| +  log 
i
2π i 1  | zi | 
N( 2r )  2r  1 2 π
or  log   i
 0 log |f(2re )| d (1)
i 1  i 
| z | 2 π
 COMPLEX ANALYSIS161

N( 2r )  2r  N ( r )  2r 
Now,  log    log 
i 1  | z i |  i1  | zi | 
 N(r) log 2 (2)
 2r 
since for large r, log    log 2
 | zi | 
1 2π 1 2π
 log |f(2re )| d 
i
And  log M(2r) d = log M(2r) (3)
2π 0 2π 0
Also, since  is the order (assumed finite) of the function f(z), then by definition, for every
 > 0, we have
log M(2r)  (2r)+/2 (4)
Thus, we conclude from (1), (2), (3) and (4) that
N(r) log 2  (2r) +/2
2ρ / 2 r ρ 
or N(r)   r+ , (5)
(log 2) r/ 2

2ρ / 2
since r is large and  > 0 implies that  1. The inequality (5) continuous to hold if
(log 2)r/ 2
we remove the restriction that there may be no zeros on |z| = 2r for we may apply the inequality
on slightly larger circles (since zeros are isolated i.e. cannot cluster) and use the fact that N(r) is
right continuous. We note that if N(r) = m on |z| = r, then N(r) = m on [r, s] for some s > r,
otherwise f(z) would have a limit point of zeros. But zeros cannot cluster and hence N(r) is right
continuous.
1.6. Exponent of Convergence. Let {z1, z2,…} be a sequence of non-zero complex numbers
with |zn|. The exponent of convergence  of the sequence is defined by

 = inf {t > 0 :  | z i |t  }
i 1
If the sequence is finite, we define  = 0. It should be noted that 0     and  =  iff
 
t t
 | z i |   for all t > 0. Also  = 0 iff  | z i |   for all t > 0.
i1 i1


e.g. if zi = 2i, then  = 0, since  (2i ) t converges for all t > 0. We shall mainly use this concept
i 1
in the case in which zi are the zeros (counted according to multiplicity) of an entire function f(z)
and we shall always assume that |z i|  |zi+1|. We sometimes call  as the convergence exponent
of f(z). or exponent of convergence of zeros of f(z).
Thus, we observe that
 = 0  existence of finite number of zeros of f(z)
 > 0  existence of infinite number of zeros of f(z)
1.7. Remark. N(r) and  are both measures of the growth of the numbers |z i|. If the zeros are
densely distributed, N(r) increases rapidly with r and since |zi| slowly,  is large. We may
compare the definition of  with that of the genus p already defined. It is evident that p, which is
an integer, exists if  <  and in this case p    p +1. The definition of p shows that  = p +1

implies  | zi| < .
i 1
162

1.8. Theorem. If f(z) is an entire function order  and convergence exponent , then   .
Proof. If  is infinite, the inequality,    is trivial. Again, if the number of zeros is finite, then
 = 0 and    holds. We may therefore suppose that  is finite and that there are infinitely
many zeros which we arrange as a sequence {zn} such that
|zn|  |zn+1| and |zn| as n.
By definition of N(r), it is observed that
N(|zn|)  n (1)
The strict inequality N(|zn|) > n will hold if
|zn| = |zn+1|
Also, we have proved that N(r)  r+, for large r.
Thus N(|zn|)  |zn|+ (2)
if  > 0 and n is sufficiently large.
Thus from (1) and (2), we get
|zn|+  n (3)
for sufficiently large n.
Since |zn| , we may assume that |zn|  1. Given t > , we may take  < t so that t/(+)>1
Then for sufficiently large n, we have from (3),
|zn|t  nt/(+)
 
Consequently,  | z n |t   nt/(+) (4)
n 1 n 1
Since t/(+) > 1, the series on R. H. S. of (4) converges

t
and so  |zn| <  (5)
n 1
for every t > .
Since by definition,

 = inf. { t > 0 :  |zn|t < }
n 1
and since (5) holds for every t > , we must have   
1.9. Borel’s Theorem. The order of a canonical product is equal to the exponent of convergence
of its zeros.
Proof. If  and  be respectively the order and convergence exponent of the conical product,
then we have proved in the previous theorem that    . So we only need to prove here that
 < .
Let us recall the notation
 1 1 
EK(w) = (1w) exp  w  w 2  ......  w K  , K  0
 2 K 
 K wn 
= (1w) exp   

 n 1 n 
K  | w |n 
 |EK(w)|  (1+|w|) exp   
 (1)
n 1  n 

Since exp |w|  1+|w| and for large |w|,

| w |K | w |K 1
  … |w|
K K 1
 COMPLEX ANALYSIS163

 | w |K  K | w |n
 K  

 K  n 1 n
K
| w |n
 | w |K  
n 1 n

Therefore (1) becomes

|EK(w)|  (exp |w|) (exp |w|K)
= exp (|w| + |w|K)
 exp (2|w|K)
 exp (c|w|) (2)
where c  2,   K and |w|  1.
1
On the other hand if |w|  and K  0, then we have (Unit – III)
2
|EK(w)|  exp (2|w|K+1)
and so
1
|EK(w)|  exp [ (2|w|) for   K + 1 and |w|  (3)
2
Again, if 1/2  |w|  1 and K > 0, it can easily be shown that for some constant c,
|EK(w)|  exp (c|w|K+1)
Since |w|  1, it follows that
|EK(w)|  exp (c|w|) (4)
1
for   K +1 and  |w|  1
2
From (2), (3) and (4), we conclude that for K    K + 1, there exists c such that for all w
|EK(w)|  exp (c|w|) (5)
  z 
Now, let P(z) =  E p   be the canonical product with zeros at zn, n = 1, 2,… Then by
n 1  zn 
definition, p is the genus of the canonical product and we know that p satisfies the inequality
p    p + 1. If  = p +1, let  = p+1 while if  < p+1, let  satisfy  <  < p +1.

Since  = inf {t > 0 :  |zi|t < }
i 1
we conclude that
 
 
 |zi| < . Let  |zi| = a (say)
i 1 i 1
Then, we get
  z 
|P(z)|   exp  c n 
n 1  z 
 
  
= exp  c | z |  | z n |λ 
λ
 n 1 
= exp (ac|z|) (6)
which holds for every  >  and all z.
Since (6) is of the form
M(r)  exp(r+)
164

and by definition.
 = inf. {  0 : M(r)  exp(r)}
therefore, we conclude that

which completes the proof.
1.10. Theorem. Let P(z) be a canonical product of finite order  and q > 0 and  > 0. Then for
all sufficiently large |z|
|z  zi| > |zi|q implies log |P(z)| > |z|+.
Proof. Let |z| = r, |zi| = ri. We have
 1 1 
EK(w) = (1w) exp  w  w 2  ...  w K  K  0
 2 K 
which gives
  wn 
EK(w) = exp    
 (1)
 n K 1 n 
1
Also, for |w|  , we have
2
|EK(w)|  exp (2|w|K+1) (2)
Now, from (1), we get
  wn 
EK(w) = 1 exp  

 n K 1 n 
  | w |n 
 |EK(w)|  1 exp  

 n K 1 n 
 1/exp (2|w|K+1) | using (2)
= exp (2 |w|K+1)
 exp (2|w|) if   K + 1
Thus log |EK(w)|  2|w| (3)
1
where |w|  and   K +1
2
1
Also, for sufficiently large A > 0 and for all w satisfying |w|  , we have
2
K | w |n
  A |w|K
n 1 n
Further, if K    K +1, then |2w|K  |2w| and consequently
|w|K  2K |w|  2|w|
 K w n  K | w |n 
Hence log exp   
 n 1 n  A |w|  2A|w|
K
(4)
 n 1 n 
1
provided K <   K + 1 and |w| 
2
Now, we consider a fixed z with |z| = r > 1. We shall estimate separately the factors of
 z
P(z) =  E p  
i 1  zi 
 COMPLEX ANALYSIS165

1 1
for which |z/zi| > i.e. ri < 2r and those for which |z/zi|  i.e. ri  2r.
2 2
We note that p is the genus of P(z). We thus write
z z
|P(z)| =  E p    E p   (5)
ri  2 r  i  ri 2 r
z  zi 
Now, we estimate log |P(z)| in (5) by using (4) to write the logarithm of each factor
corresponding to ri < 2r as
z  z  p 1  z n 
log E p    log 1   exp    
 zi   zi   n 1 n  z i  
 
n
z p 1 z 
= log 1   log exp   
zi n 1 n  z i 
λ
z z
 log 1   2A (6)
zi zi
provided p    p + 1.
Again by applying (3) to the factors in (5) for which ri  2r, we conclude for each of them that
λ
z  z
log E p    2 (7)
 zi  z i

provided   p + 1.
Now, the order  of P(z) is also the exponent of convergence of the sequence {zi} . If  < p+1,
let  satisfy  <  < p + 1 and if  = p +1, let  = p +1. Thus in either case, the definition of
convergence exponent gives

λ
 | zi | <  with p    p + 1.
i 1

Let  | z i |λ = B, where B is a finite constant.
i 1
Using the estimates (6) and (7), we obtain from (5) that
 
z 1 1
log |P(z)|   log 1 
zi
 2A | z |   zi
 2 | z |  zi
ri  2 r ri  2 r ri  2 r


z 1
  log 1
zi
 2(A  1) | z |  zi
ri 2 r i 1

z
=  log 1
zi
 2(A  1)B | z |
ri 2 r
Thus
z
log |P(z)|   log 1
zi
 D | z | (8)
ri 2 r
where D is a suitable constant.
166

Since by our assumption |z|  1, we see that the inequality (8) must hold for every 
> . By hypothesis
z
|zzi| > |zi|q i.e. 1  |zi|q1
zi
so we get
z q 1
 log 1    log z i
ri 2 r zi ri 2 r

  (q+1) N(2r) log (2r) (9)

Also, we know the result
N(2r)  (2r)+/2 (10)
which holds for all sufficiently large r and arbitrary  > 0. Using (9) and (10), we may
write (8) as
log |P(z)|  (q+1) (2r)+/2 log (2r)  D r (11)
We choose  <  +  which is always possible for any prescribed . Then since r is large,
we may have
1 1
(q+1) (2r)+/2 log (2r)   r+ and D r   r+
2 2
Substituting these estimates in (11), we obtain
log |P(z)|  r+
which proves the required result.
1.11. Hadmard’s Factorization Theorem. If f(z) is an entire function of finite order , then
f(z) = zm eg(z) P(z)
where m is the order of zeros of z at z = 0, g(z) is a polynomial of degree not exceeding  and
P(z) is the canonical product associated with the sequence of non-zero zeros of f(z)
Proof. We have already shown in Weierstrass‟s factorization theorem that an entire function f(z)
can be expressed as
f(z) = zm eg(z) P(z) (1)
where g(z) is itself an entire function. Here, we shall use the addition hypothesis, that f(z) is of
finite order , to show that g(z) is a polynomial of degree not exceeding . It is clear hat the
division of f(z) by czm does not affect either the hypothesis or the conclusion of the theorem and
so it is sufficient to consider the representation f(z) = eg(z) P(z), so that
f (z) f (z)
eg(z) =  | eg(z) | 
P( z ) P( z )
f (z)
i.e. eRe g(z) =
P( z)
Taking logarithm, we get
Re g(z) = log |f(z)|  log |P(z)| (2)
By the definition of order, it follows that
|f(z)|  exp (r+)
for sufficiently large |z| = r and all  > 0.
Thus log |f(z)|  r+ (3)
 COMPLEX ANALYSIS167

If  is the convergence exponent of the non-zero zeros of f(z), then we have proved the result
that   . Also, by Borel‟s theorem,  is the order of the canonical product P(z) and so it
follows from theorem 1.10 that
log |P(z)| > r+ for large |z| = r.
Thus
log |P(z)| < r+  r+ (4)
where    and r is large.
From (3) and (4), we get
log | f(z)|  log |P(z)|  2r+
and thus (2) gives
Re g(z)  2r+
Since r is large, we conclude from theorem 1.3 that g(z) is a polynomial of degree not
exceeding .

1.12. Example. Using Hadmard‟s factorization theorem, prove that

  z2 
sin z = z  1  2 
n 1  n 

Solution. The zeros of sin z are at z = 0, + 1 + 2…, i.e. non-zero zeros of sin z are + 1, + 2… .
 1  1
Since the series  diverges and  2 converges, so that p = 1 is the least integer such that
n 1 n n 1 n
p 1
1
 (n0) converges. Thus the genus of the canonical product is 1 and thus the canonical
n
product associated with non-zero zeros of sin z is of the form

 z
P(z) =  1   e z / n
n    n
n 0
  z   z
=  1   e z / n  1   e  z / n
n 1  n n 1  n
  z2 
=  1  2 
n 1  n 

Now, order of sin z is 1. Since z = 0 is a simple zero of sin z, Hadmard’s factorization of
sin z may be written as
  z2 
sin z = zeg(z)  1  2 
n 1  n 
where g(z) is a polynomial of degree not exceeding 1 i.e. order of sin z. let g(z) = a0 + a1z.
  z2 
 sin z  ze a 0a1z  1  2 
n 1  n 
We are to find a0 and a1. For this we write
sin πz   z2 
 e a 0 a1z  1  2  (1)
z n 1  n 
168

sin πz
Since   as z 0, so making z0 in (1) we get
z
 = ea0
sin πz   z2 
  π e a1z  1  2  (2)
z n 1  n 
Again, replacing z by z in (2), we get
sin πz   z2 
 π e a1z  1  2  (3)
z n 1  n 
Equations (2) and (3) give e a1z  e a1z  a1 = 0
  z2 
Hence sin z = z  1  2  .
n 1  n 

2. The Range of an Analytic Function

Here, the range of an analytic function is investigated. A generic problem of this type is :
Let  be a family of analytic functions on a region G which satisfy some property P. What can
be said about f (G) for each f in  ? Are the sets f (G) uniformly big in some sense ? Does there
exists a ball B(a; r) such that f (G)  B(a ; r) for each f in  ? The answers to these questions
depend on the property P that is used to define .
We start with the following lemma.
2.1. Lemma. Let f be analytic in D = {z : |z| < 1} such that f(0) = 0, f (0) = 1 and |f(z)|  M for
all z in D. Then M  1 and
 1 
f(D)  B  0 ; 
 6 M
Proof. Let f(z) = z + a2 z2 + a3 z3 +…
Since f(z) is analytic in D = B(0 ; 1) so by Cauchy‟s Estimate,
|an|  M for n  1
 |a1|  M
 M1 [ a1 = 1]
1
Let z D s.t. |z| = . Then
4M

| f(z)|  |z|   |an zn|
n 2
n
1   1 
   M.  
4M n 2  4M 
1 1  1 1 
=  1    ..... 
4M 16M  4M 16M 2 
1 1 1 1 1 1  12M  4  1
=  .     
4M 16M  1  4M 16M  4 4M  16M  4  6M
1  
 4M 
 12M  4 2 
 Min value of 16M  4 is 3 when M  1
 COMPLEX ANALYSIS169

 1  1
Suppose w B  0 ;  Then | w |
 6 M  6 M
Consider the function g(z) = f(z)  w.
1 1
For |z| = , |f(z)  g(z)| = |w| <  |f(z)|
4M 6M
 1 
So by Rouche‟s theorem, f and g have the same number of zeros in B  0 ;  . Since f(0) = 0
 4 M
 1 
so g(z0) = 0 for some z0  B  0 ; 
 4M 
  1  
 f(z0) w = 0 for some z0 D  B 0 ; 4M   D
   
i.e. w = f(z0) for some z0D
i.e. w f(D)
 1 
Hence B  0;   f(D)
 6M 
2.2. Lemma. Suppose g(z) is analytic on B(0; R), g(0) = 0, |g(0)| =  > 0 and |g(z)|  M for all z,
then
 R 2μ 2 
g(B(0; R))  B  0 ; 
 6M 

g (Rz )
Proof. Let f(z) = for z D where D = {z : |z| < 1}.
R g' (0)
Then f is analytic on D, f(0) = 0, f (0) = 1 and
g(Rz ) | g(Rz ) | M
|f(z)| =   for all z in D.
R g' (0) Rμ Rμ
So by lemma 2.1,
 μR 
B  0;   f(D)
 6M 
 R 2μ 2   R 2μ 2 
To show B  0 ;   g(B(0; R)), let w B  0 ;
 

 6M   6M 
R 2μ 2
Then |w| < B
6M
w Rμ
 
Rμ 6M
w  μR 
  B  0;   f(D)
Rμ  6M 
w
 = f(z) for some zD.
Rμ
w g(Rz )
 = where |z| < 1
Rμ Rg ' (0)
170

 w = g(Rz) where |Rz| < R

 wg(B(0 ; R)) [  Rz  B(0 ; R)]
Hence the result.
2.3. Lemma. Let f be an analytic function on the disk B(0 ; r) such that
|f (z)  f (a)| < |f (a)| for all z in B(a ; r), z  a ; then f is one one.
Proof. Suppose z1 and z2 are points in B(a ; r) s. t. z1  z2. Let  be the line segment [z1, z2] then

|f(z1)  f(z2)| =  f ' (z)dz

γ

  f ' (a )dz   [ f ' (z)  f ' (a )]dz

γ γ

 
   f ' (a )dz   [ f ' (a )  f ' (z)dz   f ' (z)dz 
 γ γ γ 
 |f (a)| |z1  z2|   |f (z)  f (a)| |dz|
γ
> |f (a)| |z1  z2|  |f (a)| |z1  z2| = 0
 f(z1)  f(z2)
Hence f is one one.
2.4. Bloch’s Theorem. Let f be an analytic function on a region containing the closure of the
disk D = { z : |z| < 1} and satisfying f(0) = 0, f (0) = 1. Then there is a disk S  D on which f is
1
one-one and such that f(S) contains a disk of radius .
72
Proof. Let K(r) = max {| f (z)| : |z| = r} and let h(r) = (1r) K(r).
Then h : [0, 1]  R is continuous such that h(0) = K(0) = 1 and h(1) = 0 .
Let r0 = sup {r : h(r) = 1}.
Then h(r0) = 1, r0 < 1 as h(1) = 0 and h(r) < 1 if r > r0. Choose a s.t. |a| = r0 and |f (a)| = K(r0).
1
Then |f (a)| = [  (1r0) K (r0) = h(r0) = 1]
1  r0
1
Let z  B(a ;0) where 0 = (1-r0).
2
Then |za| < 0.
1 1
So |z| = |za+a| < |za| + |a| < (1r0) + r0 = (1+r0)
2 2
 1 
h  (1  r0 ) 
1   2  1 1  1  
 |f (z)|  K  (1  r0 )     h 2 (1  r0 )   1
2  1  1 (1  r ) 1 (1  r )  0    
0 0
2 2
1
 |f (z)| < for |za| < 0.
ρ0
So |f (z)f (0)|  |f (z)| + |f (a)|
1 1 3  1 1 
<   | f ' (a )   
0 20 20  1  r0 20 
 COMPLEX ANALYSIS171

Using Schwarz‟s Lemma, this gives

3
|f (z) f (a)| < |za| for z  B(a ; 0)
2ρ 02
 1 
Let S = B  a ; ρ 0  Then for zS,
 3 
1  ρ0 
|f (z)  f (a)| <   | z  a | 3 
2ρ 0  
 1 1 
i.e. |f (z)  f (a)| < |f (a)}   | f ' (a ) |  
 1  r0 2ρ 0
So by Lemma 2.3, f is one-one on S.
1
Now we will show that f(S) contains a disk of radius .
72
 1 
Define a function g : B  0 ; ρ 0   C
 3 
as g(z) = f(z +a) f(a)
Then g(0) = f(a) f(a) = 0
1
and |g(0) = |f (0)| =
2ρ 0
 1 
Let zB  0 ; ρ 0  . Then the line segment  = [a, z+a] lies in S  B(a ; 0)
 3 

So |g(z)| =  f ' ( w )dw

γ

  | f ' ( w ) | | dw |
γ
1 1
< | z |
ρ0 3
  1 
By Lemma 2.2, we get g  B 0 ;  0    B(0 ; )
  3 
2
 1   1 
2
 ρ0  
 3   2ρ 0  1
where = 
1 72
6 
3
 1    1 
 B  0 ;   g B 0 ; ρ 0  
 72    3 
1
Now we show B(f(a) ; )  f(S).
72
1 1
Let w B(f(a) ; ) . Then |wf(a)| <
72 72
 1 
 w f(a)  B  0 ; 
 72 
172

 1 
 w f(a) = g(z) for some zB  0 ; ρ 0 
 3 
1
 wf(a) = f(z+a)  f(a) where |z| < 0
3
1
 w = f(z+a) where |z+aa| < 0
3
 w f(S) as z + a S.
 1 
Hence B  f (a ) ;   f(S).
 72 
2.5. Corollary. Let f be analytic function on a region containing the closure of B(0 ; R); then
1
f(B(0; R)) contains a disk of radius R|f (0)|
72
Proof. If f (0) = 0 then the result is trivial.
So assume that f (0)  0.
f (Rz )  f (0)
Consider the function g(z) =
Rf ' (0)
Then g is analytic on a region containing the closure of B(0; 1). Also g(0) = 0, g(0) = 1.
So by Bloch‟s theorem,
 1 
B  g(a ) ;   g(B(0 ; 1)) for some a .
 72 
1
We claim that B(f(aR) ; R|f (0)|)  f(B(0; R))
72
1
Let wB(f(a R) ; R|f (0)|)
72
1
Then |wf(a R)| < R |f (0)|
72
1
 |w(g(a) Rf (0) + f(0)| < R |f (0)|
72
w f (0) 1
   g (a ) 
Rf ' (0) Rf '(0) 72
w f (0)
  = g(z) where z  B(0 ; 1)
Rf ' (0) Rf ' (0)
 w f(0) = R f (0) g(z)
 wf(0) = f(Rz) f(0) where z  B(0; 1)
 w = f(Rz) where |Rz| < R
 wf (B(0; R)) as Rz  B(0; R)
1
Hence B(f(aR) ; R|f (0)|)  f(B(0; R))
72
2.6. Definition. Let  be the set of all functions f analytic on a region containing the closure of
the disk D = {z : |z| < 1} and satisfying f(0) = 0, f (0) = 1. For each f in , let (f) be the
 COMPLEX ANALYSIS173

supremum of all numbers r such that there is a disk S in D an which f is one-one and such that
1
f(S) contains a disk of radius r. Then (f) 
72
Bloch‟s constant B is defined as
B = inf {(f) : f }
1
By Bloch‟s theorem, B 
72
1
If we take f(z) = z then B  1. So  B  1.
72
However, better estimates than these are known. In fact, it is known that 0.43  B  0.47.
Although the exact value of B remains unknown, but it has been conjectured (guessed) that
(1 / 3(11 / 12)
B=
1  3 (1 / 4
2.7. Definition. Let  be the set of all function f analytic on a region containing the closure of
the disk D = {z : |z| < 1} and satisfying f(0) = 0, f (0) = 1.
For each f in , define
(f) = sup {r : f(D) contains a disk of radius r}
Landau‟s constant L is defined by
L = inf {(f) : f }
Clearly L  B and L  1. Although exact value of L is unknown but it can be proved that
0.50  L  0.56
In particular, L > B.
2.8. Proposition. If f is analytic on a region containing the closure of the disk D = {z : |z| < 1}
and f(0) = 0 f (0) = 1; then f(D) contains a disk of radius L, where L is Landau‟s constant.
Proof. We shall show that f(D) contains a disk of radius  where
 = (f)
Since  = sup {r : f(D) contains a disk of radius r} for each nN, there is a point n in f(D) such
that
 1
B  α n ; λ    f(D)
 n
Now n  f(D)  f(D) and f(D) is compact. Since every compact metric space is sequentially
compact, f(D) is sequentially compact. So every sequence of points in f(D) contains a
convergent subsequence. In particular, <n> contains a subsequence <  n k > such that
 n k    f(D).
We may assume that  = lim n. We show that B( ;)  f(D). Let wB( ;). Then |w| < .
1
Choose n0 s.t. |w| < 
n0
Since n so there exists an integer n1 > n0 s.t.
1
|n | <   |w| for n  n1.
n0
 |wn| = |w + n|
 |w| + | n|
174

1
<
n0
1
<   if n  n1
n
 1
 w  B  α n ; λ    f(D)
 n
Hence B(; )  f(D).
2.9. Corollary. Let f be analytic on a region that contains the closure of B(0 ; R); then f(B(0 ;R))
contains a disk of radius R|f (0)| L.
Proof. If f (0) = 0 then the result is trivial. So assume that f (0)  0. Applying the above theorem
f (Rz )  f (0)
to the function g(z) = , we get the required result.
Rf '(0)
2.10. Definition. If G is an open connected set in  and f : G  is a continuous function such
that z = exp f(z) for all z in G then f is a branch of the logarithm.
2.11. Lemma. Let G be a simply connected region and suppose that f is an analytic function on
G that does not assume the values 0 or 1. Then there is an analytic function g on G such that
f(z) = exp (i cosh [2g(z)] for z in G.
Proof. Since f never vanishes, there is a branch l of log f(z) defined on G; that is el = f.
1
Let F(z) = l(z)
2πi
We claim that F does not assume any integer value. If F(a) = n for some integer n, then
f(a) = exp (2 i F(a)) = exp (2 in) = 1 which is not possible.
Since F cannot assume the values 0 and 1, it is possible to define
H(z) = F(z)  F(z)  1
Now H(z)  0 for any z so that it is possible to define a branch g of log H on G, that is, e g = H.
e 2g  e 2g
 cosh (2g) + 1 = 1
2
2
(e g  e  g )
=
2
2
1 1
= H  
2 H
 H(z)  F(z)  F(z)  1 
= 2F  1 
and  F(z)  F(z)  1
 H(z) 
1
= l
πi
This gives
f = exp (l) = exp [ i +  i cosh (2g)]
= exp ( i). exp ( i cosh (2g))
=  exp ( i cosh (2g))
Hence the result.
 COMPLEX ANALYSIS175

2.12. Lemma. Let G be a simply connected region and suppose that f is an analytic function on
G that does not assume the values 0 or 1. Let g be an analytic function on G such that
f(z) = exp (i  cosh [2g(z)]) for z in G.
Then g(G) contains no disk of radius 1.
Proof. Let n be a positive integer and m any integer.
We claim that g cannot assume any of the values
1
{+ log ( n  n  1)  im : n1, m = 0, + 1, + 2,….}
2
If there is a point a in G such that
1
g(a) = + log n  n  1)  im
2
then 2 cosh [2g(a)] = e2g(a) + e2g(a)
= eim ( n  n  1)  2  e imπ ( n  n  1)  2
= (1)m[ ( n  n  1) 2  ( n  n  1) 2 ] [ ei = 1]
= (1)m [2(2n1)]
 cosh (2g(a)) = (1)m (2n1)
So f(a) =  exp [(1)m (2n1) i]
=1 [ (2n1) is odd]
which is not true. Hence g cannot assume any of the values
1
{+ log ( n  n  1)  imπ : n  1, m = 0, + 1, + 2,….}
2
These points form the vertices of a grid of rectangles in the plane. The height of an
arbitrary rectangle is
1 1 π
imπ  i(m  1) π   3
2 2 2
The width is log( n  1  n )  log( n  n  1)  0.
Now, (x) = log ( x  1  x )  log( x  x  1) is a decreasing function so that the width of
any rectangle  (1) = log (1+ 2 ) < log e = 1 . So the diagonal of the rectangle < 2. Hence
g(G) contains no disk of radius 1.
2.13. Little Picard Theorem. If f is an entire function that omits two values then f is a constant.

Proof. Let a and be two values omitted by f. so that

f(z)  a and f(z)  b for all z.
f a
Then the function omits the values 0 and 1
ba
So assume that f(z)  0 and f(z)  1 for all z. By lemma 2.12, there is an entire function g such
that g() contains no disk of radius 1.
We want to prove f is a constant function. Let, if possible, suppose that f is not a constant
function. Then g is also not a constant so there is a point z0 with g(z0)  0 by considering
g(z + z0) if necessary, we may suppose that
g(0)  0
176

 g(B(0 ; R)) contains a disk of radius LR |g(0)|. Choosing R sufficiently large we

get that g() contains a disk of radius 1. A contradiction. Hence f must be constant.
2.14. Schottky’s Theorem. For each  and , 0 <  <  and 0    1, there is a constant
C(, ) such that if f is an analytic function on some simply connected region containing the
closure of B(0; 1) that omits the values 0 and 1 and such that |f(0)|   ; then |f(z)|  C(, ) for
|z|  .
Proof. We consider two cases
1
Case I. Suppose  |f(0)|  .
2
Since f never vanishes there is a branch l of log f(z) s.t. 0  Im l(0) < 2.
1
Let F(z) = l(z)
2πi
1
Then |F(0)| = |log |f(0)| + i Im l(0)|
2π
1 1
< (log  + 2) = log  + 1 = C0() (say)
2π 2π
 |F(0)| < C0() (1)
Also | F(0)  F(0)  1 |  | F(0) |  | F(0)  1 |
= |F(0)|1/2 + |F(0)1|1/2
 [C0()]1/2 + [C0() +1]1/2 = C1() (say)
Now F cannot assume the values 0 and 1, it is possible to define
H(z) = F(z)  F(z)  1
Also H(z)  0 for any z so define a branch g of log H s.t.
0  Im g(0)  2.
If H(0)  1, then
|g(0)| = | log| H(0)| + i Im g(0)|
 log |H(0)| + 2
 log C1() + 2
If |H(0)| < 1 then
|g(0)|   log |H(0)| + 2
 1 
= log   + 2
 | H(0) | 
= log | F(0)  F(0)  1 | + 2
 log C1() + 2
 |g(0)|  C2() where C2() = Log C1() + 2.
If |a| < 1 then g(B(a ; 1|a|)) contains a disk of radius L(1|a|) |g(a)| [By cor. 2.9]
On the other hand g(B(0 ; 1)) contains no disk of radius 1. So we must have
L(1|a|) |g(a)| < 1 for |a| < 1
1
i.e. |g(a)| < for |a| < 1 (2)
(1 | a |) L
If |a| < 1, let  be the line segment [0, a]; then
|g(a)| = |g(0) + g(a)  g(0)|
 COMPLEX ANALYSIS177

 |g(0)| + |g(a)  g(0)|

 C2() +  g' (z)dz

γ

 C2() + |a| max {|g(z)| : z  [0, a]}

|a|
 C2() + (using (2)) (3)
L(1 | a |)
β
Let C3(, ) = C2() +
L(1  β)
Then (3) gives
|g(z)|  C3(, ) if |z|  .
Consequently, if |z|  ,
|f(z)| = |exp [ i cosh 2g(z)]|
 exp [ |cosh 2g(z)|]
 exp [ e2|g(z)|]
 exp [ e 2C3 (α,β) ] = C4 (, ) (say)
1
Case II. Suppose 0 < |f(0)| <
2
In this case (1f) satisfies the condition of case (I) so that
|1f(z)|  C4(2, ) if |z|  .
Hence |f(z)| = |f(z)| = |1f(z)+1|  |1f(z)| + 1  1 + C4(2, ) if |z|  
If we define C(, ) = max. {C4(, ), 1 +C4(2, )}, we have
|f(z)|  C(, ) if |z|  
2.15. Corollary. Let f be analytic on a simply connected region containing B(0 ;R) and suppose
that f omits the values 0 and 1. If C(, ) is the constant obtained in Schottky‟s Theorem and
|f(0)|   then
|f(z)|  C(, ) for |z|  R
Proof. It follows by considering the function f(Rz) for |z|  1.
2.16. Montel-Coratheodory Theorem. If  is the family of all analytic functions on a region G
that do not assume the values 0 and 1, then  is normal in C(G, ).
Proof. Fix a point z0 in G and define the families G and H by
G = {f  : |f(z0)  1},
H = {f  : |f(z0)|  1}
So that  = G  H
We shall show that G is normal in H(G) and H is normal in C(G, ). To show G is normal in
H(G), it is sufficient to show that G is locally bounded in view of Montel‟s theorem.
Let a be any point in G and  be a curve in G from z0 to a. Let D0, D1, D2,…, Dn be disks in G
with centres z0, z1, z2,…, zn = a on {} and such that zk1 and zk are in Dk1 Dk for 1  k  n .
Also assume that Dk  G for 0  k  n . By Schottky‟s theorem for D0, there is a constant C0
such that
|f(z) |  C0 for z in D0
In particular
|f(z1)|  C0
178

So by Schottky‟s theorem G is uniformly bounded by C1 on D1 . Continuing, we have G is

uniformly bounded on Dn. Since a was arbitrary, this gives that G is locally bounded. Hence by
Montel‟s theorem, G is normal in H(G).
Now consider H = {f   : |f(z0|  1 }
If f  H then 1/f is analytic on G because f never vanishes. Also 1/f omits the values 0 and 1
1
and | ( ) (z0) |  1.
f
~ 1
Hence H  { : f  H }  G
f
~
and H is normal in H(G).
So if {fn} is a sequence in H there is a subsequence { f n k } and an analytic function h on G such
 1 
that   converges in H(G) to h. So either h  0 or h never vanishes (By Cor. to Hurwitz‟s
 f n k 
theorem, Unit III). If h  0 then fnk(z)   uniformly on compact subsets of G. If h never
1 1
vanishes then is analytic and it follows that f n k (z)  uniformly on compact subsets
h h (z)
of G.
2.17. Great Picard Theorem. Suppose an analytic function f has an essential singularity at
z = a. Then in each neighbourhood of a, f assumes each complex number, with one possible
exception, an infinite number of times.
Proof. We prove the theorem by taking a = 0 .
Suppose that there is an R such that there are two numbers not in {f(z) : 0 < |z| < R} . Also
suppose that f(z)  0 and f(z)  1 for 0 < |z| < R. Let G = B(0; R)  {0}
Define fn : G   by
z
fn(z) = f ( )
n
Then each fn is analytic and no fn assumes the value 0 or 1. So by Montel-Caratheodory
Theorem, {fn} is a normal family in C(G, ).
1
Let { f n k } be a subsequence of {fn} such that f n k   uniformly on {z : |z| = R} where  is
2
either analytic on G or    .
1
If  is analytic, let M = max{ |(z) | : |z| = R}
2
z 1
Then |f( )| = | f n k (z)|  | f n k (z)  (z)| + |(z)|  2M for nk sufficiently large and |z| = R .
nk 2
R
Thus |f(z)|  2M for |z| = and for sufficiently large nk.
2n k
By Maximum Modulus Principle, f is uniformly bounded on concentric annuli about zero.
This gives that f is bounded by 2M on a deleted neighbourhood of zero and so z = 0 must
be a removable singularity. Therefore  cannot be analytic. So   . In this case f has a
pole at zero. A contradiction. So at most one complex number is never assumed.
If there is a complex number w which is assumed only a finite number of times then by
taking a sufficiently small disk, we again arrive at a punctured disk in which f fails to
assume two values.
2.18 Remark. An alternate framing of the above theorem is as follows :
 COMPLEX ANALYSIS179

If f has an isolated singularity at z = a and if there are two complex numbers that are not
assumed infinitely often by f then z = a is either a pole or a removable singularity.
2.19 Corollary. If f is an entire function that is not a polynomial then f assumes every complex
number, with one exception, an infinite number of times.
1
Proof. Consider the function g(z) = f ( ) .
z
Since f is not a polynomial, g has an essential singularity at z = 0. So result follows from
Great Picard Theorem. We observe that Cor. 2.19 is an improvement of the Little Picard
Theorem.
3. Univalent Function
The theory of conformal mappings on simply connected regions is of special significance
from the point of view of geometric function theory. Essentially there are three types of
simply connected spaces in the extended complex plane
(i) The open unit disc
(ii) The entire finite complex plane
(iii) The extended complex plane or the Riemann sphere.
The cases (i), (ii) and (iii) are referred to as hyperbolic, parabolic and elliptic cases
respectively.
A simple geometric restriction namely the injectivity imposed on functions defined on the unit
disc throws a wealth of information on the geometric and analytic properties of such functions.
In general, we say that a function f(z) is univalent (simple, schlicht) in a region D if it is analytic,
one-valued, and does not take any value more than once in D. The function w = f(z) then
represents the region D of the z-plane on a region D of the w-plane, in such a way that there is a
one-one correspondence between the points of the two regions. In other words, a univalent
function assumes each value in its range precisely once.
3.1. Theorem. If f(z) is univalent in D, then f (z)  0 in D.
Proof. On the contrary, suppose that f (z0) = 0. Then f(z)  f(z0) has a zero of order n (n  2) at
z0. Since f(z) is not constant, we can find a circle |zz0| =  on which f(z)  f(z0) does not vanish,
and inside which f (z) has no zeros except z0. Let m be the lower bound of | f(z) f(z0)| on this
circle. Then by Rouche‟s theorem, if 0 < |a| < m, f(z)  f(z0) a has n zeros in the circle, it cannot
have a double zero, since f (z) has no other zeros in the circle. This is contrary to the hypothesis
that f(z) does not take any value more than once.
3.2. Remarks. (i) If f(z) is univalent in D, then the mapping w = f(z) is conformal at every point
of D.
(ii) A univalent function of a univalent function is univalent
Proof. If f(z) is univalent in D and F(w) in D, then F{ f(z)} is univalent in D, since F{ f(z1)} =
F{f(z2)} implies f(z1) = f(z2) as F is univalent and this further implies z1 = z2, since f is univalent.
(iii) In the above relationship, to every point of D corresponds just one point of D. We
can therefore consider z as a function of w, say z = (w). This is called the inverse function of
w = f(z).
(iv) The inverse function is univalent in D, since it is one-valued and analytic.
3.3. Theorem. A univalent function w = f(z) which represents a unit circle on itself, so that the
centre and a given direction through it remain unaltered, is the identical transformation w = z.
Proof. We have | f(z)| = 1 for |z| = 1 and f(0) = 0.
Hence by Schwarz‟s lemma (unit –II)
180

|w| = | f(z)|  |z|

But, applying Schwarz‟s lemma to the inverse function, we have |z|  |w|
Hence, we get
|w| = |z|, i.e. | f(z)/z| = 1, |z|  1.
Since a function of constant modulus is constant, it follows that
f(z)/z = a, i.e., f(z) = az
where |a| = 1. The remaining conditions then show that a = 1 and thus we obtain w = z.
3.4. Remark. The class of functions f(z) which are univalent for the open disc |z| < 1 and such
that f(0) = 0, f (0) = 1, has been studied in great detail. The function w = z belongs to this class
and represents the unit circle on itself. We denote this class by and the unit open disc by U. As
an immediate consequence of Taylor‟s series development, to every f(z) in F we have a power
series expansion

f(z)= z +  a n z n
n 2

which is valid throughout U. We further observe that the class F is not closed under either
addition or multiplication.
as (f +g) (0) = 2 and (fg)(0) = 0 for f, g  F.
3.5. Bieberbach’s Conjecture. The study of geometric function theory has been given a solid
foundation by Riemann with his fundamental mapping theorem. Later on, Koebe and others
formulated the theory of univalent functions which are normalized by f(0) = 0 f (0) = 1 (the class
F) Normal families of analytic functions was introduced by Montel and Coratheodary established
his kernel theorem on sequences of univalent functions.
Ludwig Bieberbach proposed his famous conjecture in 1916 which says : If

f(z) = z +  a n z n
n 2

is univalent in the open unit disc |z| < 1 then |an|  n (n = 2, 3,….) with equality if and only if f(z)
is a rotation of the Koebe function
z
K(z) = = z + 2z2 + 3z3 +….
(1  z) 2
Bieberbach proved his conjecture for n = 2 using the area principle which was established by T.
H. Gronwall. Several sub-classes of univalent functions for which the conjecture can be easily
verified by other geometric means have been introduced and studied.
 1 
3.6. Theorem.  The " Theorem" : For any function of the class F, no boundary point of the
 4 
1
map of the unit circle is nearer to the origin than the point .
4
The 1/4 theorem may be stated in another form as follows.

If f(z) = z +  a n z n  F then |a2|  2 with equality if and only if
n 2
z 1 1
f(z) = with || = 1. Further, f(U) contains all w with |w| < and the constant
(1  αz) 2 4 4
cannot be improved for all f  F where U denotes the units open disc.