A Complete Solution Guide To Complex Analysis 3rd Bak

A Complete Solution Guide to
Complex Analysis
by Kit-Wing Yu, PhD
kitwing@hotmail.com
Copyright c 2020 by Kit-Wing Yu. All rights reserved. No part of this publication may be
reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic,
mechanical, photocopying, recording, or otherwise, without the prior written permission of the
author.
ISBN: 978-988-74155-0-3 (eBook)

ISBN: 978-988-74155-1-0 (Paperback)
ii
About the author
Dr. Kit-Wing Yu received his B.Sc. (1st Hons), M.Phil. and Ph.D. degrees in Math. at the
HKUST, PGDE (Mathematics) at the CUHK. After his graduation, he has joined United Chris-
tian College to serve as a mathematics teacher for at least seventeen years. He has also taken
the responsibility of the mathematics panel since 2002. Furthermore, he was appointed as a
part-time tutor (2002 – 2005) and then a part-time course coordinator (2006 – 2010) of the
Department of Mathematics at the OUHK.
Besides teaching, Dr. Yu has been appointed to be a marker of the HKAL Pure Mathematics
and HKDSE Mathematics (Core Part) for over thirteen years. Between 2012 and 2014, Dr. Yu
was invited to be a Judge Member by the World Olympic Mathematics Competition (China). In
the research aspect, he has published over twelve research papers in international mathematical
journals, including some well-known journals such as J. Reine Angew. Math., Proc. Roy. Soc.
Edinburgh Sect. A and Kodai Math. J.. His research interests are inequalities, special functions
and Nevanlinna’s value distribution theory. In the area of academic publication, he is the author
of five books:
• A Complete Solution Guide to Real and Complex Analysis I.
• Problems and Solutions for Undergraduate Real Analysis I.
• Problems and Solutions for Undergraduate Real Analysis II.
• Mock Tests for the ACT Mathematics.
• A Complete Solution Guide to Principles of Mathematical Analysis.
iii
iv
Preface
There are many books entitled Complex Analysis. For examples, Ahlfors [1], Bak and Newman
[4], Freitag and Rolf [9], Gamelin [10], Lang [14], and Stein & Shakarchi [24]. (I believe that
you can find more if you search your library website.) As an introductory textbook for complex
analysis, I would like to recommend the book by Bak and Newman [4]. The reasons are that this
book gives readers some important, insightful and interesting background to build up the theory
of complex analysis. Furthermore, their presentation is quite clear and concise, so I believe that
you can grasp the main concepts and skills in an easier way. Finally, the book contains a lot of
helpful examples and problems.
There are a total of 300 exercises in the third edition [4], but only 225 of them are provided
by solutions. (The exercises without solutions are marked an asterisk.) In my opinion, some
solutions are a bit “brief” and I guess that some students still have difficulties when reading
them. To provide assistance, I decide to write a solution manual for this book and I hope that
students / instructors can benefit from this solution book.
Before you read this book, I have a gentle reminder for you. As a mathematics instructor
at a college, I understand that the growth of a mathematics student depends largely on how
hard he/she does exercises. When your instructor asks you to do some exercises of Bak and
Newman’s book, you are not suggested to read my solutions unless you have tried your best to
prove them yourselves.
The features of this book are as follows:
• It covers all the 300 exercises with detailed and complete solutions. As a matter of fact,
my solutions show every detail, every step and every theorem that I applied.
• There are 34 illustrations for explaining the mathematical concepts or ideas used behind
the questions or theorems.
• Different colors are used in order to highlight or explain problems, lemmas, remarks, main
points/formulas involved, or show the steps of manipulation in some complicated proofs.
(ebook only)
• Necessary lemmas with proofs are provided.
• Useful or relevant references are provided to some questions for interested readers.
Finally, if you find such any typos or mistakes, please send your valuable comments or opinions
to
kitwing@hotmail.com
v
vi
so that I will post the updated errata on my website
https://sites.google.com/view/yukitwing/
from time to time.
Kit-Wing Yu
January 2020
List of Figures
1.1 The graph of z1 z2 and z1 z2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.2 The locations of the roots. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 The product of the 4 diagonals of a regular 5-gon. . . . . . . . . . . . . . . . . . 9
1.4 The graph of lemniscate of Bernoulli. . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5 The graph of S in the plane C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.6 S is a circle when T is a circle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.7 S is a circle minus (0, 0, 1) when T is a line. . . . . . . . . . . . . . . . . . . . . . 19
P
1.8 The effect of f (z) = 1z on . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
6.1 The function |f (z(t))| has a local maximum at z = 1 on the line 1 + t exp( π2 ). . . 76
6.2 The function |f (z(t))| has a local minimum at z = 1 on the line 1 + t exp( 33π
18 ). . 77
6.3 The function |f (z(t))| has an inflection point at z = 1 on the line 1 + t exp( 3π
4 ). . 77
7.1 The graph of the set S. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
8.1 The geometry of the circle C and the annulus B. . . . . . . . . . . . . . . . . . . 102

8.2 The “inside” of a simple closed polygonal path. . . . . . . . . . . . . . . . . . . . 105
8.3 A decomposition of a closed polygonal path. . . . . . . . . . . . . . . . . . . . . . 106
8.4 The circular arc C connecting −|z| and z clockwise. . . . . . . . . . . . . . . . . 107
10.1 The regular closed curve C surrounding ±1. . . . . . . . . . . . . . . . . . . . . . 135
π
11.1 The contour CR with central angle 4. . . . . . . . . . . . . . . . . . . . . . . . . 148
11.2 The square CN with vertices ±(N + 21 ) ± i(N + 21 ). . . . . . . . . . . . . . . . . . 152
11.3 The rectangle CN with vertices (N + 21 ) ± i(N + 12 ) and −(N − 21 ) ± i(N + 12 ). . 153
11.4 The rectangle ΓR with vertices at ±R and ±R + 2πi. . . . . . . . . . . . . . . . . 154
12.1 The formations of the contours C(0; R) and C(0; 2). . . . . . . . . . . . . . . . . 164
12.2 The closed contour formed by γ1 , γ2 , γ3 and γ4 . . . . . . . . . . . . . . . . . . . . 165
13.1 The sector S and the disc D(z0 ; δ). . . . . . . . . . . . . . . . . . . . . . . . . . . 173

13.2 The rectangles R and R′ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
13.3 The angles of ∠γ1 , γ2 and ∠f (γ1 ), f (γ2 ). . . . . . . . . . . . . . . . . . . . . . . . 177
13.4 The conformal mapping of the region between the two circles and the annulus. . 183
13.5 The conformal mapping of H onto a rectangle. . . . . . . . . . . . . . . . . . . . 184
vii
List of Figures viii
14.1 The shape of a Norman window N . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
16.1 The level curves of u(x, y) = k for k ∈ (0, 21 ) ∪ ( 12 , 1). . . . . . . . . . . . . . . . . 205

16.2 The temperature problem with the prescribed boundary values. . . . . . . . . . . 206
16.3 The angle Arg (ω 2 − 1). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
19.1 The intersections of y = x and y = tan x. . . . . . . . . . . . . . . . . . . . . . . 227

sin2 x
19.2 The intersections of y = x2 and y = tan x − x. . . . . . . . . . . . . . . . . . . 229
Contents
Preface v
List of Figures viii
1 The Complex Numbers 1
2 Functions of the Complex Variable z 23
3 Analytic Functions 37
4 Line Integrals and Entire Functions 47
5 Properties of Entire Functions 55
6 Properties of Analytic Functions 67
7 Further Properties of Analytic Functions 83
8 Simply Connected Domains 101
9 Isolated Singularities of an Analytic Function 109
10 The Residue Theorem 121
11 Applications of the Residue Theorem to the Evaluation ... 139
12 Further Contour Integral Techniques 163
13 Introduction to Conformal Mapping 173
14 The Riemann Mapping Theorem 187
15 Maximum-Modulus Theorems for Unbounded Domains 197
16 Harmonic Functions 201
17 Different Forms of Analytic Functions 211
18 Analytic Continuation; The Gamma and Zeta Functions 221
ix
Contents
19 Applications to Other Areas of Mathematics 227
Index 237
Bibliography 239
CHAPTER 1
The Complex Numbers
Problem 1.1
Bak and Newman Chapter 1 Exercise 1.
Proof.
(a) It is clear that

1 1 6 − 2i 6 − 2i 3 1
= × = = − i.
6 + 2i 6 + 2i 6 − 2i 36 + 4 20 20
(b) It is easy to see that
(2 + i)(3 + 2i) 4 + 7i 4 + 7i 1 + i −3 + 11i 3 11
= = × = = − + i.
1−i 1−i 1−i 1+i 2 2 2
√
(c) Since − 21 + 23 i = cis 2π
3 , we have
1 √3 4 2π 6π + 2π 2π 2π 1
√
3
− + i = cis 4 × = cis = cis 2π + = cis =− + i.
2 2 3 3 3 3 2 2
(d) We have 

 i, if n = 4m + 1 for some m ∈ Z;

−1, if n = 4m + 2 for some m ∈ Z;
in =

 −i, n = 4m + 3 for some m ∈ Z;

1, if n = 4m for some m ∈ Z.
We have completed the proof of the problem.
Problem 1.2
Proof. Suppose that (x + iy)2 = −8 + 6i. By the discussion on [4, p. 3], we see that
s s
√ √
−8 + 64 + 36 8 + 64 + 36
x=± = ±1 and y = ± · sgn (6) = ±3.
2 2
√
Hence the two values of −8 + 6i are ±(1 + 3i), completing the proof of the problem.
1
Chapter 1. The Complex Numbers 2
Problem 1.3
Proof. We have
√ q√
√ √
− 32i ± ( 32i)2 − 4 × (1) × (−6i) − 32i ± −32 + 24i √ √
z= = = −2 2i ± −8 + 6i.
2 2
By Problem 1.2, we see immediately that
√ √ √
z = −2 2i ± (1 + 3i) = 1 + (3 − 2 2)i or − 1 − (3 + 2 2)i.
This completes the proof of the problem.
Problem 1.4
Proof. Suppose that z1 = a + bi and z2 = c + di, where a, b, c and d are real. Furthermore, we
suppose that P (z) = a0 z n + a1 z n−1 + · · · + an−1 z + an , where a0 , a1 , . . . , an ∈ R and n ∈ N ∪ {0}.
(a) Then z1 + z2 = (a + c) + (b + d)i = (a + c) − (b + d)i = a − bi + c − di = z1 + z2 .

(b) We have z1 z2 = (a + c)(b + d)i = ac − bd + (ad + bc)i = ac − bd − (ad + bc)i = z1 · z2 .
(c) Since a0 , a1 , . . . , an ∈ R, we have ak = ak for all 0 ≤ k ≤ n. By repeated applications of
parts (a) and (b), we have
n
X n
X n
X n
X
P (z) = ak z n−k = ak z n−k = ak · z n−k = ak · z n−k = P (z).
k=0 k=0 k=0 k=0
(d) If z = A + Bi, then we have z = A − Bi = A + Bi = z.
We complete the proof of the problem.
Problem 1.5
Proof. Since P (z) = 0 if and only if P (z) = 0, it follows from Problem 1.4(c) that P (z) = 0 if
and only if P (z) = 0. This completes the proof of the problem.
Problem 1.6
Proof. Using rectangular coordinates, let z = a + bi. Then z 2 = (a2 − b2 ) + 2abi so that
p
|z 2 | = (a2 − b2 )2 + 4a2 b2 = a2 + b2 = |z|2 .
Using polar coordinates, if z = rcis θ, then we have z 2 = r 2 cis 2θ so that |z 2 | = r 2 = |z|2 . We
end the analysis of the problem.
3
Problem 1.7
Proof. Suppose that z = rcis θ.
(a) For every n ∈ Z, we have z n = r n cis nθ so that |z n | = r n = |z|n .
(b) Since z = rcis (−θ), we have z · z = r 2 cis (θ − θ) = r 2 = |z|2 .

√
(c) Let z = a + bi. Then we have Re z = a, Im z = b and |z| = a2 + b2 . Since a2 ≤ a2 + b2
and b2 ≤ a2 + b2 , we get |Re z| ≤ |z| and |Im z| ≤ |z|. The fact a2 + 2|a| · |b| + b2 ≥ a2 + b2
implies that (|a| + |b|)2 ≥ a2 + b2 which reduces to exactly
|Re z| + |Im z| ≥ |z|.
Now we study when an equality occurs. For example, |Re z| = |z| if and only if |a|2 = a2 +b2
if and only if b = 0. The case for |Im z| = |z| is similar. Next, |z| = |Re z| + |Im z| if and
only if a2 + b2 = |a|2 + 2|a| · |b| + |b|2 if and only if |a| · |b| = 0 if and only if
|a| = 0 or |b| = 0.
In conclusion, an equality occurs if and only if Re z = 0 or Im z = 0.
Problem 1.8
Proof.
(a) By Problem 1.7(b) and then Problem 1.4(a), we have
|z1 + z2 |2 = (z1 + z2 )(z1 + z2 ) = (z1 + z2 )(z1 + z2 ). (1.1)
By expanding the right-hand side of the equation (1.1) and then use Problems 1.7(a) and
(c), we get
|z1 + z2 |2 = |z1 |2 + |z2 |2 + 2Re (z1 z2 ) (1.2)

2 2
≤ |z1 | + |z2 | + 2|z1 z2 | (1.3)
= |z1 |2 + |z2 |2 + 2|z1 ||z2 |
= (|z1 | + |z2 |)2
which gives the triangle inequality.
(b) Note that we apply Problem 1.7(c) to get the inequality (1.3), so the equality (1.3) occurs
if and only if Im (z1 z2 ) = 0 if and only if z1 z2 ∈ R.
(c) Replace z1 by z1 − z2 in the triangle inequality, we know that |z1 − z2 + z2 | ≤ |z1 − z2 | + |z2 |
which is equivalent to
|z1 | − |z2 | ≤ |z1 − z2 |.

Problem 1.9
Bak and Newman Chapter 1 Exercise 9.∗
Proof.
(a) Let z1 = a + bi and z2 = c + di, where a, b, c, d ∈ Z. Thus z1 z2 = ac − bd + (ad + bc)i,

where ac − bd, ad + bc ∈ Z. Since |z1 z2 | = |z1 | · |z2 |, we have |z1 z2 |2 = |z1 |2 · |z2 |2 so that
(a2 + b2 )(c2 + d2 ) = (ac − bd)2 + (ad + bc)2 . (1.4)
If we define u = ac − bd and v = ad + bc, then we get the desired result.
(b) Since all a, b, c and d are nonzero, without loss of generality, we may assume they are all
positive. Besides the formula (1.4), we also note thata
(a2 + b2 )(c2 + d2 ) = (ac + bd)2 + (ad − bc)2 . (1.5)
Now ad + bc > 0 and ac + bd > 0, we follow from the two formulas (1.4) and (1.5) that it
suffices to check that at least one of ad − bc and ac − bd is nonzero. Assume that
ad = bc and ac = bd. (1.6)
Then the product of the same sides of the equations (1.6) gives a2 cd = b2 cd which implies
a2 = b2 . Similarly, the product of the opposite sides of the equations (1.6) gives abd2 = abc2
so that c2 = d2 . This is a contradiction to the hypotheses and hence we have ad − bc 6= 0
or ac − bd 6= 0.
(c) We label
u = ac − bd, v = ad + bc, s = ac + bd and t = ad − bc.
Direct computation gives
u2 − s2 = (u − s)(u + s) = (−2bd)(2ca) = −4abcd.
Since a, b, c and d are nonzero, we have u2 6= s2 . Similarly, we can show that
u2 − t2 = (u − t)(u + t) = (a2 − b2 )(c2 − d2 ).
By the hypotheses, we know that u2 6= t2 . Furthermore, we also have v 2 6= s2 and v 2 6= t2 .

Otherwise, u2 + v 2 = s2 + t2 will imply either u2 = s2 or u2 = t2 , a contradiction.
Consequently, the sets {(ac − bd)2 , (ad + bc)2 } and {(ac + bd)2 , (ad − bc)2 } are distinct.
(d) The assumption a, b, c, d 6= 0 means that the vectors from 0 to the complex numbers z1
and z2 are not parallel to the real axis or the imaginary axes. If we write
z1 = r1 cis θ1 and z2 = r2 cis θ2 , (1.7)
then both θ1 and θ2 are not a multiple of π2 . Next, notice that a2 6= b2 if and only if
a 6= ±b. Therefore, the vector from 0 to z1 is not parallel to the lines y = ±x. In other
words, θ1 is not a multiple of π4 . Similarly, c2 6= d2 if and only if θ2 is not a multiple of π4 .
a
This comes from the fact that |z1 z2 | = |z1 | · |z2 |.
5
– Geometric interpretation of part (b). We obtain directly from the representa-

tions (1.7) that
z1 z2 = r1 r2 cis (θ1 + θ2 ) and z1 z2 = r1 r2 cis (θ1 − θ2 ).
Assume that both the vectors from 0 to z1 z2 and z1 z2 were parallel to the real axis or
the imaginary axis. Then the previous analysis shows that both θ1 ± θ2 are multiples
of π2 , i.e.,
mπ nπ
θ1 + θ2 = and θ1 − θ2 =
2 2
but they imply that
(m + n)π
θ1 =
4
which is a contradiction. Thus at least one of the vectors from 0 to z1 z2 and z1 z2 is
not parallel to the axes. Consequently, we have
cis (θ1 + θ2 ) 6= 0 or cis (θ1 − θ2 ) 6= 0. (1.8)
Finally, we recall from the equations (1.4) and (1.5) that ac − bd and ad − bc are
exactly the real and imaginary parts of z1 z2 or z1 z2 respectively, so the results (1.8)
ensure that both u and v are nonzero.
– Geometric interpretation of part (c). To show that {u2 , v 2 } and {s2 , t2 } are
distinct, it is equivalent to showing that
(|u|, |v|) 6= (|s|, |t|) and (|u|, |v|) 6= (|t|, |s|). (1.9)
In Figure 1.1, we see easily that (|u|, |v|) = (|s|, |t|) if and only if the difference
between the arguments of z1 z2 and z1 z2 is a multiple of π.
Figure 1.1: The graph of z1 z2 and z1 z2 .

Assume that Arg (z1 z2 ) = Arg (z1 z2 ) + mπ for some m ∈ N. Then we have
2θ2 = (θ1 + θ2 ) − (θ1 − θ2 ) = Arg (z1 z2 ) − Arg (z1 z2 ) = mπ
so that θ2 = mπ 2 which is a contradiction by the description at the beginning of
part (d). Next, it is also true that (|u|, |v|) = (|t|, |s|) if and only if the difference
between the arguments of z1 z2 and z1 z2 is a multiple of π2 . In this case, we have
Arg (z1 z2 ) = Arg (z1 z2 ) + nπ
2 for some n ∈ Z and then similar argument can show
that
nπ
2θ1 = (θ1 + θ2 ) − (θ2 − θ1 ) = Arg (z1 z2 ) − Arg (z1 z2 ) = ,
2
a contradiction to the fact that θ1 6= nπ 4 . Hence we obtain the results (1.9) which
mean that {u2 , v 2 } and {s2 , t2 } are different.
Problem 1.10
Proof. We replace z2 by −z2 in the equation (1.2) to obtain

|z1 − z2 |2 = |z1 |2 + | − z2 |2 + 2Re (z1 (−z2 )) = |z1 |2 + |z2 |2 − 2Re (z1 z2 ). (1.10)
Then the sum of the equations (1.2) and (1.10) imply that
|z1 + z2 |2 + |z1 − z2 |2 = 2(|z1 |2 + |z2 |2 ).
Geometrically, the sum of the squares of the lengths of the diagonals (left-hand side) is double
to the sum of the squares of the lengths of the sides (right-hand side).b This completes the
proof of the problem.
Problem 1.11
Proof. Suppose that z = x + iy 6= 0. By the definition, we notice that

π y π
− ≤ tan−1 ≤ .
2 x 2
However, recall that Arg z is the angle which the vector (originating from 0) to z makes with
the positive x-axis (see the figure on [4, p. 6]), so it takes values [−π, π] (modulo 2π). Thus
their connection can be expressed as
 y


 tan−1 , if x > 0;

 x



 y

 tan−1 + π if x < 0 and y ≥ 0;



 x


 y
Arg z = Arg (x + iy) = tan−1 − π, if x < 0 and y < 0;

 x



 π



 , if x = 0 and y > 0;

 2




 −π,

if x = 0 and y < 0.
2
b
See also [26, Exercise 1.17, p. 9].
7
Problem 1.12
Proof. Let z = rcis θ. Here we mainly follow the steps in the example on [4, pp. 8, 9].c
(a) Since r 6 cis 6θ = 1cis 0, we have r = 1 and 6θ = 0 (modulo 2π). Then we have
6θ = 0, 2π, 4π, 6π, 8π, 10π.
Hence the six solutions are given by

π 2π 4π 5π
z1 = 1, z2 = cis , z3 = cis , z4 = cis π = −1, z5 = cis and z6 = cis .
3 3 3 3
(b) Since r 4 cis 4θ = 1cis π, we have r = 1 and 4θ = π (modulo 2π). Then we obtain
4θ = π, 3π, 5π, 7π.
Hence the four solutions are given by

π 3π 5π 7π
z1 = cis , z2 = cis , z3 = cis and z4 = cis .
4 4 4 4
√
(c) Since r 4 cis 4θ = 2cis 2π
3 , we have r =
4
2 and 4θ = 2π
3 (modulo 2π). Then we have
2π 8π 14π 20π
4θ = , , , .
3 3 3 3
Hence the four solutions are given by
√
4 π √
4 2π √
4 7π √
4 5π
z1 = 2cis , z2 = 2cis , z3 = 2cis and z4 = 2cis .
6 3 6 3
The locations of the roots in parts (a) to (c) have been shown in Figure 1.2 below. This ends
the proof of the problem.
Problem 1.13
Proof. Let ω be an nth root of unity other than 1. Since
0 = ω n − 1 = (ω − 1)(ω n−1 + ω n−2 + · · · + 1),
we conclude that ω satisfies the equation
z n + z n−1 + · · · + 1 = 0,
completing the proof of the problem.

c
In fact, there is a general formula of finding the nth roots of the equation z n = r, see [1, p. 16].
(a) The locations of the roots in part (a). (b) The locations of the roots in part (b).
(c) The locations of the roots in part (c).
Figure 1.2: The locations of the roots.
Problem 1.14
Proof. By the comments on [4, p. 9] or [14, pp. 13 - 16], the n-th roots of 1 are located at the
vertices of the regular n-gon inscribed in the unit circle. Let z1 , z2 , . . . , zn be the n-th roots of
zn = 1
with z1 = 1. Without loss of generality, we may assume that the vertices all be connected to
z1 . (For example, see Figure 1.3 for the regular 5-gon.) Then the lengths of the diagonals are
exactly |z1 − z2 |, |z1 − z3 |, . . . , |z1 − zn |. By Problem 1.13, we know that
z n−1 + z n−2 + · · · + 1 = (z − z2 )(z − z3 ) · · · (z − zn )

9
which certainly implies that
|1 − z2 | × |1 − z3 | × · · · × |1 − zn | = 1n−1 + 1n−2 + · · · + 1 = n.
Figure 1.3: The product of the 4 diagonals of a regular 5-gon.
This ends the proof of the problem.
Problem 1.15
Proof. In the following, we suppose that z = x + iy.
(a) This is a closed disc centred at i with radius 1. By Definition 1.6, it is not a region.
(b) The equation becomes |(x − 1) + iy| = |(x + 1) + iy|. By the definition of the modulus and
after squaring, we get
(x − 1)2 + y 2 = (x + 1)2 + y 2 . (1.11)
Thus we must have x = 0, i.e., the set of solutions of the equation is exactly the imaginary
axis. Since squaring an equation may create new solutions of the original equation, we
have to check that every solution of the equation (1.11) is also a solution of the original
equation.d In fact, if z = iy, then it is easy to see that
z−1 | − 1 + iy|
= = 1.
z+1 |1 + iy|
d
√ 2
For
√ example, if x = 2 − x, then x + x − 2 = 0 whose solutions are x = 1 or x = −2. However, note that
−2 = 2 + 2 = 2 which is impossible, so x = −2 is not a solution of the original solution.
In other words, all solutions of the equation (1.11) are also solutions of the original equa-
tion. By Definition 1.6 again, it is not a region.
(c) We have (x − 2)2 + y 2 > (x − 3)2 + y 2 which gives x > 25 . Similar to part (b), if x > 25 ,
then z = x + iy satisfies the original inequality. Hence the solution set of the inequality is
Re z > 25 . By Definition 1.6, it is a region.
(d) Note that |z| < 1 is the open unit disc centred at 0 and Im z > 0 is the upper half plane,
so the solution set is the upper half of the open unit disc centred at 0. By Definition 1.6,
it is a region.
(e) The equation is equivalent to z · z = 1 which is x2 + y 2 = 1. This is a unit circle centred

at 0, so it is not a region by Definition 1.6.
(f) We have x2 + y 2 = y which is x2 + (y − 21 )2 = 212 . Hence it is a circle centred at i

2 and the
radius is 21 . Similar to part (e), it is not a region.
(g) By the definition, we have |(x2 − y 2 − 1) + 2xyi| < 1 so that

p
(x2 − y 2 − 1)2 + 4x2 y 2 < 1
(x2 − y 2 )2 − 2(x2 − y 2 ) + 1 + 4x2 y 2 < 1.
After simplification, we obtain
(x2 + y 2 )2 − 2(x2 − y 2 ) < 0
which is known as a lemniscate of Bernoulli [15, pp. 121 - 123], see Figure 1.4, which
is generated by “Desmos” (https://www.desmos.com/), below:
Figure 1.4: The graph of lemniscate of Bernoulli.
We notice that this set is open but not connected. By Definition 1.6, it is not a region.
11
Problem 1.16
Proof.
(a) Let z = x + iy. Then the equation becomes

p
x2 + y 2 = x + 1 (1.12)
which, by squaring both sides, reduces to
y 2 = 2x + 1. (1.13)
Clearly, this is a parabola which opens rightward. Similar to Problem 1.15(b), we have to
check that every solution of the equation (1.13) is also a solution of the equation (1.12).
To see this, let (x, y) satisfy the equation (1.13). Since y 2 ≥ 0, we have x ≥ − 12 so that
x + 1 > 0. Next, we substitute the equation (1.13) into the left-hand side of the equation
(1.12) to get
p p p
x2 + y 2 = x2 + 2x + 1 = (x + 1)2 = |x + 1| = x + 1.
This shows that the solutions of the equations (1.12) and (1.13) are identical.
(b) Geometrically, it says that the sum of the distances from z to 1 and from z to −1 is always
4, so the locus of the points is the ellipse with foci (±1, 0) and the semi-major axis 2, i.e.,
x2 y 2
+ 2 = 1,
22 b
where b is the semi-minor axis. By the definition, we have b2 = 22 − 12 = 3 so that the set
of points satisfying the given equation is
x2 y 2
+ = 1.
4 3
(c) Notice that n ∈ Z. There are several cases for consideration.

1
– Case (i): n = 0. The equation becomes z = z which is Problem 1.15(e).
– Case (ii): n = 1. In this case, z = 1. If z = x + iy, then it implies that x − iy = 1
so that x = 1 and y = 0.
– Case (iii): n = 2. Now we have z = z. If z = x + iy, then x + iy = x − iy so that
y = 0. In other words, the set of solutions in this case is the real axis.
– Case (iv): n ≥ 3. Obviously, z = 0 is a solution of the equation. Suppose that
z 6= 0. Then |z n−1 | = |z| implies that |z| = 1. Next, z n−1 = z also gives
z n = z · z = |z|2 = 1,
i.e., z is an n-th root of unity. Hence the set of solutions in this case consists of the
n-th roots of unity and z = 0.
– Case (v): n ≤ −1. In this case, z 6= 0. Similar to Case (iv), |z n−1 | = |z| implies
that |z| = 1 and z n−1 = z gives
z n = 1, (1.14)
where n ≤ −1. If we rewrite the equation (1.14) as z −|n| = 1, then we get
z |n| = 1
so that z is an |n|-th root of unity.

Problem 1.17
Proof. Let z = x + iy, where x2 + y 2 = 1. Then it is easy to check that

z−1 x − 1 + iy
=
z+1 x + 1 + iy
x − 1 + iy x + 1 − iy
= ·
x + 1 + iy x + 1 − iy
x2 + y 2 − 1 2y
= 2 2
+i
(x + 1) + y (x + 1)2 + y 2
2y
=i .
(x + 1)2 + y 2
Note that if y > 0, then z−1 z−1

z+1 lies on the positive imaginary axis; if y < 0, then z+1 lies on the
negative imaginary axis. By the definition of Arg w given in the problem, we obtain immediately
that  π
 , if Im z > 0;
z − 1   2
Arg =
z+1 
 − π , if Im z < 0.

2
Problem 1.18
Proof.
√ By [4, Eqn.
√ (4), p. 10], the solutions of√x3 − 6x = 4 are the
√ three real-valued possibilities
for
p
3 3
2 + 2i + √2 − 2i. p Suppose that √ a = 2 + 2i and b = 2 − 2i. Then we have a2 =
3 3
3 2 3 2
(2 + 2i) = 2 i, b = (2 − 2i) = 2 3 −i and ab = 2. We note that
3 2
√ √
√3 3 i 3 i
i= + , − + , −i
2 2 2 2
and √ √
√
3 3 i 3 i
−i = − , − − , i.
2 2 2 2
Thus we have √ √
a2 − ab + b2 = 2 3 − 2, −2 3 − 2 or − 2.
13
√
3
√
3
By the hint, we conclude that the three real-valued possibilities for 2 + 2i + 2 − 2i are given
by
a 3 + b3 2 2 √ √
x =a+b= 2 2
=√ , √ or −2=1+ 3, 1− 3 or − 2,
a − ab + b 3−1 − 3−1
Remark 1.1
By direct substitution, we see that x = −2 is a root of the equation x3 − 6x − 4 = 0.
Therefore, we have
x3 − 6x − 4 = (x + 2)(x2 − 2x − 2)
√
so that the other real roots of the equation x3 − 6x − 4 = 0 are exactly 1 ± 3.
Problem 1.19
Proof. Suppose that α, β, γ are the roots of the equation
x3 + px − q = 0. (1.15)
Then we have 
 α + β + γ = 0,
αβ + βγ + γα = p, (1.16)

αβγ = q.
√
If γ = 0, then q = 0 and thus the equation x3 + px = 0 has three roots 0, ± −p. Now the roots
are real if and only if p < 0 if and only if 4p3 + 27q 2 = 4p3 < 0. Therefore, we may suppose that
all the roots of the equation (1.15) are nonzero. Thus it follows from the equations (1.16) that
2αβγ
(α − β)2 = α2 + β 2 + γ 2 − γ 2 −
γ
2αβγ
= (α + β + γ)2 − 2(αβ + βγ + γα) − γ 2 −
γ
2q
= −2p − γ 2 − . (1.17)
γ
Similarly, we also have
2q 2q
(β − γ)2 = −2p − α2 − and (γ − α)2 = −2p − β 2 − . (1.18)
α β
Suppose that (α − β)2 , (β − γ)2 and (γ − α)2 are roots of the new equation
y 3 + Ay 2 + By + C = 0,
where A, B and C will be determined very soon. By the formulas (1.17) and (1.18), it yields
that
2q
y = −2p − x2 −
x
xy = −2px − x3 − 2q
(x3 + px − q) + xy = −px − 3q
xy + px = −3q
3q
x=− . (1.19)
y+p
Substituting the expression (1.19) into the original equation (1.15), we arrive at
3q 3 3q
− +p − −q =0
y+p y+p
−27q 3 − 3pq(y + p)2 − q(y + p)3 = 0
−qy 3 − 6qpy 2 − 9qp2 y − 27q 3 − 4qp3 = 0. (1.20)
Recall that q 6= 0, so the equation (1.20) reduces to
y 3 + 6py 2 + 9p2 y + (27q 2 + 4p3 ) = 0
so that the product of its roots is given by
(α − β)2 (β − γ)2 (γ − α)2 = −(4p3 + 27q 2 ). (1.21)
Hence the formula (1.21) implies the following results:
• Case (i): The equation (1.15) has three distinct real roots if and only if 4p3 + 27q 2 < 0.
• Case (ii): The equation (1.15) has a repeated (real) root if and only if 4p3 + 27q 2 = 0.
• Case (iii): The equation (1.15) has two complex roots if and only if 4p3 + 27q 2 > 0.
Hence we complete the proof of the problem.
Remark 1.2
(a) By Case (ii) in the proof of Problem 1.19, it is believed that “three real roots” should
be replaced by “three distinct real roots”.
(b) If we assume that the equation (1.15) has a real root, then the algebra will become
much simpler. In fact, let α be a real zero of the function f (x) = x3 + px − q. Then
we can write
f (x) = (x − α)(x2 + ax + b),
where a, b ∈ R. Direct expansion gives f (x) = x3 + (a − α)x2 + (b − aα)x − bα, so we
have α = a and then
p = b − a2 and q = ba.
Now 4p3 + 27q 2 < 0 if and only if
4(b − a2 )3 + 27b2 a2 < 0

4b3 − 12b2 a2 + 12ba4 − 4a6 + 27b2 a2 < 0
4b3 + 15b2 a2 + 12ba4 − 4a6 < 0
(−a2 b2 − 4ba4 − 4a6 ) + (4b3 + 16a2 b2 + 16ba4 ) < 0
−a2 (2a2 + b)2 + 4b(b + 2a2 )2 < 0
(b + 2a2 )2 (4b − a2 ) < 0
if and only if a2 − 4b > 0 which is equivalent to saying that the equation x2 + ax + b = 0

has two (distinct) real roots. Hence, we have completed the proof of the problem.
15
Problem 1.20
Proof.
(a) It is clear that
(1 − z)P (z) = 1 + 2z + 3z 2 + · · · + nz n−1 − z − 2z 2 − · · · − (n − 1)z n−1 − nz n

= 1 + z + z 2 + · · · + z n−1 − nz n . (1.22)
Assume that z0 was a zero of P (z) with |z0 | = r > 1. Then the expression (1.22) implies
that
1 + z0 + z02 + · · · + z0n−1 = nz0n
and then
nr n ≤ 1 + r + · · · + r n−1 . (1.23)
However, since r > 1, it is easy to see that r k − 1 > 0 for every k ∈ N so that
nr n − (1 + r + · · · + r n−1 ) = (r n − 1) + (r n − r) + · · · + (r n − r n−1 )
= (r n − 1) + r(r n−1 − 1) + · · · + r n−1 (r − 1)
>0
which contradicts the inequality (1.23). Hence no such z0 exists and all the zeros of P (z)
lie in |z| ≤ 1.
(b) Let P (z) = a0 + a1 z + · · · + an z n . If a0 = a1 = · · · = an = 0, then P (z) ≡ 0 which is

meaningless. Therefore, we may assume that ak > 0 for some k ∈ {0, 1, 2, . . . , n}. We
consider
(1 − z)P (z) = a0 + a1 z + · · · + an z n − a0 z − a1 z 2 − · · · − an−1 z n − an z n+1

= a0 + (a1 − a0 )z + (a2 − a1 )z 2 + · · · + (an − an−1 )z n − an z n+1 . (1.24)
Assume that w was a zero of P (z) with |w| = r > 1. On the one hand, the expression
(1.24) implies that
a0 + (a1 − a0 )w + (a2 − a1 )w2 + · · · + (an − an−1 )wn = an wn+1
so that
an r n+1 ≤ a0 + (a1 − a0 )r + (a2 − a1 )r 2 + · · · + (an − an−1 )r n . (1.25)
On the other hand, we know that r k − 1 > 0 for every k ∈ N and since ak > 0 for some k,
we see that
an r n+1 − [a0 + (a1 − a0 )r + (a2 − a1 )r 2 + · · · + (an − an−1 )r n ]

= an r n (r − 1) + an−1 r n−1 (r − 1) + · · · + a1 r(r − 1) + a0 (r − 1)
= (r − 1)(an r n + an−1 r n−1 + · · · + a1 r + a0 )
> 0,
but this contradicts the inequality (1.25). Hence no such w exists and all the zeros of P (z)
lie in |z| ≤ 1.
Consequently, we complete the proof of the problem.
Remark 1.3
Problem 1.20(b) is the well-known Eneström and Kakeya Theorem. See, for examples, [17,
pp. 136, 137] and [18].
Problem 1.21
Proof.
(a) Let z0 be a fixed point such that |z0 | < 1. Then there exists a r > 0 such that |z0 | < r < 1.
It is clear that |kz k | ≤ kr k for every |z| ≤ r and k = 0, 1, 2, . . .. Let Mk = kr k . Since
√
k
√
k
α = lim sup kr k = lim kr = r < 1,
k→∞ k→∞
∞
X
the Root Test [27, Theorem 6.7, p. 76] ensures that the series kr k converges. By The-
k=0
∞
X
orem 1.9 (The Weierstrass M -test), the series kz k converges uniformly to a continuous
k=0
function f (z) on D(0; r). Since z0 is arbitrary, we conclude that f (z) is continuous on
D(0; 1).
(b) Let z = x + iy with x > 0. Obviously, we know that

1 1 1
=p ≤ 2.
k2 +z 2 2
(k + x) + y 2 k
∞
X 1
Since converges, Theorem 1.9 (The Weierstrass M -test) guarantees that the series
k2
k=1
∞
X 1
k2 + z
k=1
converges uniformly to a continuous function g(z) in the right half-plane Re z > 0.

Problem 1.22
Proof. Suppose that S is a polygonally connected set. Assume that S was disconnected. Then
there are disjoint open sets U and V of C such that
S ⊆ U ∪ V, S ∩ U 6= ∅ and S ∩ V 6= ∅. (1.26)
Select a ∈ S ∩ U and b ∈ S ∩ V . Since U and V are disjoint, a 6= b. Since S is polygonally

connected, there exists a polygonally line L : [0, 1] → S connecting a and b, i.e., L(0) = a and
L(1) = b. By the definition, we have
L = [a, z1 ] ∪ [z1 , z2 ] ∪ · · · ∪ [zn , b]

17
for some n ∈ N. Since each line segment is continuous, L is continuous and then L([0, 1]) is
connected by [27, Theorem 7.12, p. 100]. However, the set relations (1.26) imply that
L([0, 1]) ⊆ U ∩ V, L([0, 1]) ∩ U 6= ∅ and L([0, 1]) ∩ V 6= ∅.
In other words, L([0, 1]) is disconnected, a contradiction. Hence we end the proof of the problem.

Problem 1.23
Proof. First of all, the graph of S is shown in Figure 1.5.
Figure 1.5: The graph of S in the plane C.
Define
T = {x + iy | x > 0 and y = sin x1 } and I = {iy | y ∈ R}.
Then we have S = I ∪ T . We assume that S was disconnected, i.e., there are disjoint open sets
U and V such that
S ⊆ U ∪ V, S ∩ U 6= ∅ and S ∩ V 6= ∅. (1.27)
Since (0, 0) ∈ S, we may suppose that (0, 0) ∈ U so that I ∩ U 6= ∅. Since I is obviously

polygonally connected, it is connected by Problem 1.22. If I ∩ V 6= ∅, then the definition shows
that I is disconnected, a contradiction. Hence we know that
I ⊆ U. (1.28)
Next, the openness of U implies that there exists a δ > 0 such that
{z = a + ib | |z| < δ} ⊆ U.
1
We note that the points xn = 2nπ with n ∈ N satisfy sin x1n = sin 2nπ = 0, so we can select N
large enough such that the points zn = xn + i sin x1n satisfy
r 2
1
|zn | = (xn − 0)2 + sin − 0 = xn < δ
xn
for all n ≥ N .e In other words, it means that zn ∈ U for all n ≥ N or equivalently, T ∩ U 6= ∅.
Recall that sin x1 is continuous on R \ {0}, so the function f : (0, ∞) → C defined by
1
f (x) = x + i sin
x
is continuous on (0, ∞) and its image f ((0, ∞)) is connected by [27, Theorem 7.12, p. 100].
Since T = f ((0, ∞)), T is also connected and the argument in the previous paragraph shows
that
T ⊆ U. (1.29)
Combining the set relations (1.28) and (1.29), we conclude that S ∩ V = ∅ which contradicts
the assumption (1.27). Hence S is connected which completes the proof of the problem.
Remark 1.4
In Problem 1.23, the union of the set T with its limit point (0, 0) is called the Topologist’s
sine curve, see [19, Example 7, pp. 156, 157]. Furthermore, it is obvious that the point 2i
cannot be connected by any curve in S because | sin x1 | ≤ 1.
Problem 1.24
Proof. Recall from [4, Eqn. (3), p. 17] that
x y x2 + y 2
ξ= , η= and ζ =
x + y2 + 1
2 x + y2 + 1
2 x2 + y 2 + 1
so that
ζ
x2 + y 2 = . (1.30)
1−ζ
1 1
Since ζ ≥ ζ0 , we have 1−ζ ≥ 1−ζ0 . Hence it follows from the equation (1.30) that
ζ0
x2 + y 2 ≥ ,
1 − ζ0
ζ0
i.e., T is the exterior of the circle centred at 0 with radius 1−ζ0 . This completes the proof of the
problem.
Problem 1.25
Proof.
e
In fact, this argument also shows that (0, 0) is a limit point of S.
19
(a) Suppose that T is a circle on C. Then ζ 6= 1 and T has the form
x2 + y 2 + Dx + Ey + F = 0. (1.31)
Using the expression (1.30) and [4, Eqn. (2), p. 17], the equation (1.31) reduces to
ζ Dξ Eη
+ + +F =0
1−ζ 1−ζ 1−ζ
Dξ + Eη + (1 − F )ζ = −F. (1.32)
P P
Recall that a circle on is the intersection of with a plane Aξ + Bη + Cζ = G, where
A, B,
P C, G ∈ R. As the equation (1.32) is in this form, the corresponding set S is a circle
on that doesn’t contain (0, 0, 1) as required. See Figure 1.6 for an illustration.
Figure 1.6: S is a circle when T is a circle.
(b) If T is a line in the form ax + by + c = 0 for some a, b, c ∈ R, then we obtain

aξ bη
+ +c=0
1−ζ 1−ζ
aξ + bη − cζ = −c.
P
Thus S is again a circle on that minus the pole (0, 0, 1), see Figure 1.7.
Figure 1.7: S is a circle minus (0, 0, 1) when T is a line.

Problem 1.26
Proof. Let P (z) = an z n + an−1 z n−1 + · · · + a1 z + a0 , where an 6= 0. Then we have

an−1 a0 an−1 a0
|P (z)| = z n an + + · · · + n = |z|n × an + + ··· + n . (1.33)
z z z z
By [1, Eqn. (12), p.10], we know that |z1 | − |z2 | ≤ |z1 − z2 | holds for every complex z1 and z2 .
Apply this result to the right-hand side of the expression (1.33), we deduce that
an−1 a0
|P (z)| ≥ |z|n × |an | − + ··· + n . (1.34)
z z
Let M = max(|a0 |, |a1 |, . . . , |an−1 |). If |z| > 1, then it follows from the triangle inequality that
an−1 a0 |an−1 | |an−2 | |a0 | Mn
+ ··· + n ≤ + 2
+ ··· + n ≤ .
z z |z| |z| |z| |z|
|an |
Therefore, if |z| ≥ max(1, 2M n
|an | ), then
Mn
|z| ≤ 2 and the inequality (1.34) becomes
an−1 a0 |an |
|P (z)| ≥ |z|n × |an | − + · · · + n ≥ |z|n ×
z z 2
which tends to ∞ as |z| → ∞. Consequently, we have P (z) → ∞ as z → ∞ which completes
Problem 1.27
1 1 x y
Proof. Let z = x + iy. Then we have z = x+iy = x2 +y 2
− i x2 +y 2.
(a) Using [4, Eqn. (3), p. 17], we have

x y x2 + y 2
ξ= , η= and ζ = .
x + y2 + 1
2 x + y2 + 1
2 x2 + y 2 + 1
By [4, Eqn. (3), p. 17] again, we see that
x
x2 +y 2 x
ξ′ = x −y = =ξ
( x2 +y 2
2 ) + ( x2 +y 2 )
2 +1 x2 + y2 + 1
and
−y
′ x2 +y 2 −y
η = x −y = = −η.
( x2 +y 2
2 ) + ( x2 +y 2 )
2 +1 x2 + y2 + 1
Finally, we get
x 2 −y 2
′
( x2 +y 2 ) + ( x2 +y 2 ) x2 + y 2 1
ζ = = = 2 = 1 − ζ.
x
( x2 +y 2
2) + ( x2−y
+y 2
)2 +1 x2 + y 2 + (x2 + y 2 )2 x + y2 + 1
1 1 1 1 P
(b) Obviously, (−P 2 , 0, 2 ) and ( 2 , 0, 2 ) are on . By part (a), if the point
P z corresponds to
1
(ξ1 , η1 , ζ1 ) on , then the point z corresponds to (ξ1 , −η1 , 1 − ζ1 ) on . The form of the
corresponding points P show that they lie on the circle C which is the intersection of the
plane ξ = ξ1 and . It is clear that the centre of circle C is (ξ1 , 0, 21 ). See Figure 1.8 for
an illustration.
21
1 P
Figure 1.8: The effect of f (z) = z on .
By considering the projections of the points (ξ1 , η1 , ζ1 ), (ξ1 , −η1 , 1 − ζ1 ) and the circle
(ξ1 , 0, 21 ) on the η − ζ plane, we observe that (η1 , ζ1 ) and (−η1 , 1 − ζ1 ) are points on a circle
centred at (0, 12 ). Therefore, the equation of the straight line L passing through (0, 12 ) and
(η1 , ζ1 ) is
(2ζ1 − 1)η − 2η1 ζ + η1 = 0. (1.35)
By substituting the point (−η1 , 1 − ζ1 ) into the equation (1.35), it is easy to check that
it lies on L. In other words, the point (−η1 , 1 − ζ1 ) is the image of the 180◦ rotation of
the point (η1 , ζ1 ) about the centre (0, 21 ). Going back to the three dimensional situation,
it implies that the point (ξ1 , −η1 , 1 − ζ1 ) is the image of the 180◦ rotation of the point
(ξ1 , η1 , ζ1 ) about the diameter with endpoints (− 21 , 0, 12 ) and ( 21 , 0, 12 ).
Remark 1.5
Another way to intercept Problem 1.27(b) can be found in [20, p. 143].
Problem 1.28
P
Proof. Let T be a circle or a line in C. By Problem 1.25, its corresponding S ⊂ is also a
circle. By Definition 1.11, S has the form
Aξ + Bη + Cζ = D, (1.36)
where A, B, C, D ∈ R. Substituting the result of Problem 1.27(a) into the equation (1.36), it
leads
Aξ ′ + B(−η ′ ) + C(1 − ζ ′ ) = D
Aξ ′ − Bη ′ − Cζ ′ = D − C. (1.37)
P
By Definition 1.11 again, the set S ′ with the form (1.37) is also a circle on . By Proposition
1.12, its corresponding projection T ′ is a circle or a line in C. In conclusion, we have shown
that f (z) = 1z maps circles and lines in C onto other circles and lines. This ends the proof of
the problem.
Remark 1.6
The map in Problem 1.28 is sometimes called the inversion through the unit circle, see
[14, Exercise I.3.1, p. 17].
CHAPTER 2
Functions of the Complex Variable z
Problem 2.1
Proof. We consider the monomial P (x, y) = (x + iy)n first, where n ∈ N ∪ {0}. It is clear that
Px = n(x + iy)n−1 and Py = in(x + iy)n−1 which imply
Py = iPx .
Thus the necessity of Proposition 2.3 holds for a monomial. Since an analytic polynomial is a
finite linear combination of monomials, the necessity of Proposition 2.3 also holds for an analytic
polynomial. This completes the proof of the problem.
Problem 2.2
Proof.
(a) For every real z, it follows from Definition 2.4 that
f (z + h) − f (z)
f ′ (z) = lim (2.1)
h→0 h
h∈C
exists. In particular, we may take h to be real in the limit (2.1) and the resulting limit
still equals to f ′ (z), i.e.,
f (z + h) − f (z)
f ′ (z) = lim . (2.2)
h→0 h
h∈R
Since both z and h are real, f (z + h) and f (z) are also real so that the quotient in the
limit (2.2) is real-valued. Hence f ′ (z) is real-valued for real z.
(b) The limit (2.1) exists for all imaginary points z. If z = iy, where y ∈ R, then we have
f (iy + h) − f (iy) 1 f (iy + ir) − f (iy)

f ′ (iy) = lim = · lim . (2.3)
h→0 h i r→0 r
h∈C r∈R
23
Chapter 2. Functions of the Complex Variable z 24
By the hypothesis, both f (iy + ir) and f (iy) are real-valued so that the quotient in the
limit (2.3) is also real-valued. Hence f ′ (iy) is purely imaginary.
Problem 2.3
Proof.
(a) Since Px = 3x2 − 3y 2 − 1 + 6ixy and Py = −6xy + i(3x2 − 3y 2 − 1), we have Py = iPx . By
Proposition 2.3, P is analytic.
(b) Since Px = 2x and Py = 2iy, we have Py 6= iPx . By Proposition 2.3, P is not analytic.
(c) Since Px = 2y − 2ix and Py = 2x + 2iy, we have Py = iPx . By Proposition 2.3, P is

analytic.
Problem 2.4
Proof. Assume that P was a nonconstant analytic polynomial which takes imaginary values
only. Then Q = iP takes only real values. Since Qy = iPy = i(iPx ) = iQx , Q is a nonconstant
analytic polynomial. By the Example 1 on p. 24, no such Q exists. Hence no such P exists and
we complete the proof of the problem.
Problem 2.5
Proof. For Problem 2.3(a), we have P (z) = z 3 − z so that
P ′ (z) = 3z 2 − 1 = 3(x + iy)2 − 1 = 3x2 − 3y 2 − 1 + 6ixy = Px .
Similarly, for Problem 2.3(c), we have P (z) = −iz 2 so that
P ′ (z) = −2iz = −2i(x + iy) = 2y − 2ix = Px .
We end the proof of the problem.
Problem 2.6
Proof. By Definition 2.4, we have
h1 (z + t) − h1 (z) f (z + t) − f (z) g(z + t) − g(z)

h′1 (z) = lim = lim + lim = f ′ (z) + g ′ (z)
t→0 t t→0 t t→0 t
25
and
h2 (z + t) − h2 (z)
h′2 (z) = lim
t→0 t
f (z + t)g(z + t) − f (z)g(z)
= lim
t→0 t
f (z + t)g(z + t) − f (z)g(z + t) + f (z)g(z + t) − f (z)g(z)
= lim
t→0 t
f (z + t) − f (z) g(z + t) − g(z)
= lim · g(z + t) + lim · f (z)
t→0 t t→0 t
= f ′ (z)g(z) + f (z)g ′ (z). (2.4)
If g(z) 6= 0, then we have

1 1
g(z+t) − g(z)
[g−1 (z)]′ = lim
t→0 t
g(z) − g(z + t)
= lim
t→0 tg(z)g(z + t)
1 g(z + t) − g(z)
= − lim · lim
t→0 g(z)g(z + t) t→0 t
g′ (z)
=− 2 . (2.5)
g (z)
Using the results (2.4) and (2.5), we see that
h′3 (z) = [f (z)g −1 (z)]′

= f ′ (z)g−1 (z) + f (z)([−1 (z)]′
f ′ (z) f (z)g′ (z)
= −
g(z) g2 (z)
f ′ (z)g(z) − f (z)g ′ (z)
= ,
g2 (z)
Problem 2.7
Proof. By Definition 2.4 and the identity an − bn = (a − b)(an−1 + an−2 b + · · · + bn−1 ), we know
that
(z + h)n − z n
(z n )′ = lim = lim [(z + h)n−1 + (z + h)n−2 z + · · · + z n−1 ] = nz n−1 . (2.6)
h→0 h h→0
Thus Proposition 2.6 holds for a monomial. If P (z) = α0 + α1 z + · · · + αN z N , then the formula
(2.6) and repeated applications of Proposition 2.5 imply that
P ′ (z) = α1 + 2α2 z + · · · + N αN z N −1 .

Problem 2.8
1
Proof. Let Sn = n n , where n ≥ 2. Then log Sn = logn n > 0 which tends to 0 as n → ∞. Since
Sn = elog Sn and ex is continuous on R, it has led that

lim Sn = lim elog Sn = exp lim log Sn = e0 = 1.
n→∞ n→∞ n→∞
Problem 2.9
Proof.
1
(a) Now Cn! = 1 so that |Cn! | n! = 1. Therefore, we get
1 1
L = lim sup |Ck | k = lim |Cn! | n! = 1
k→∞ n→∞
and so R = 1 by Theorem 2.8.

1 1 1
n n n
(b) Now Cn = (n + 2n ) so that |Cn | n = (n + 2n ) n = 2 · (1 + 2n ) . Since 2n → 0 as n → ∞,
we have
n n1
lim sup 1 + n =1
n→∞ 2
and then
1
n n1
L = lim sup |Cn | n = lim sup 2 · 1 + n = 2.
n→∞ n→∞ 2
By Theorem 2.8, we conclude that R = 12 .
Problem 2.10
Proof. By the hypothesis, we have

1 1
L = lim sup |ck | k and R = ,
k→∞ L
where 0 < L < ∞.
(a) By Problem 2.8, we have

1 p 1 p 1
lim sup |kp ck | k = lim sup k k · |ck | k = lim sup k k · lim sup |ck | k = L.
k→∞ k→∞ k→∞ k→∞
Therefore, the radius of convergence of the power series is also R.

27
(b) Now it is clear that

1 1
lim sup ||ck || k = lim sup |ck | k = L,
k→∞ k→∞
so the radius of convergence of the power series is also R.
(c) Obviously, we have

1 1
lim sup |c2k | k = lim sup(|ck | k )2 = L2
k→∞ k→∞
so that the radius of convergence of the power series is R2 .
This has completed the proof of the problem.
Problem 2.11
Proof. Let R be the radius of convergence of the sum of the two power series. Now we claim
that
R ≥ min(R1 , R2 ). (2.7)
To this end, we suppose first that R1 < R2 . Thus for every z with |z| ≤ R1 , the hypotheses
imply that both
X∞ X∞
an z n and bn z n (2.8)
n=0 n=0
converge so that their sum

∞
X ∞
X ∞
X
n n
(an + bn )z = an z + bn z n (2.9)
n=0 n=0 n=0
converges for such z. In other words, R ≥ R1 . However, R1 < R < R2 is impossible. Otherwise,
if w satisfies R1 < |w| = R < R2 , then
∞
X ∞
X ∞
X
(an + bn )wn − bn w n = an w n
n=1 n=0 n=0
converges which means that the radius of convergence of the first power series (2.8) is greater
than R1 , a contradiction. Consequently, we obtain R = R1 which satisfies the inequality (2.7)
in this case. Next, if R1 = R2 , then the formula (2.9) is still valid for any z with |z| ≤ R1 = R2 .
By the definition, it implies the inequality (2.7) in this case.
To show that the strict inequality can hold in the estimate (2.7), we consider the power series
∞
X ∞
X
zn and (−z n )
n=0 n=0
both have radius of convergence 1, but their sum

∞
X
(1 − 1)z n = 0
n=0
has radius of convergence ∞. This has completed the proof of the problem.
Problem 2.12
Proof. Let |z| = 1. If z = 1, then the series
X∞ X∞
z 1
=
n=1
n n=1
n
is divergent. Without loss of generality, we may assume that z 6= 1. Now we want to apply
Dirichlet’s Test [27, Theorem 6.14, p. 78]:
Lemma 2.1 (Dirichlet’s Test)

P
Suppose that an is a series of complex numbers whose partial sums {An } is
bounded. If {bn } is a monotonically decreasing sequence and lim bn = 0, then the
P n→∞
series an bn converges.
Let an = z n and bn = n1 . Then the sequence {bn } satisfies the conditions of Lemma 2.1
(Dirichlet’s Test). Next, it follows from the triangle inequality that
n
X n
X zn − 1 2|z| 2
|An | = ak = zk = z · ≤ = ,
z−1 |z − 1| |z − 1|
k=1 k=1
so it is bounded for z 6= 1. Hence Lemma 2.1 (Dirichlet’s Test) asserts that the power series
∞
X zn
n=1
n
converges at all points on the unit circle except z = 1. This completes the proof of the problem.

Problem 2.13
Proof.
(a) Suppose that L is finite. For every n ≥ 1, we know that

a a a
2 3 n+1
an+1 = a1 × × × ··· × . (2.10)
a1 a2 an
By the definitiona , for every ǫ > 0, there is an N ∈ N such that k ≥ N implies
ak+1
−L <ǫ
ak
or equivalently,
ak+1
L−ǫ< <L+ǫ (2.11)
ak
a
For example, [27, p. 49].
29
for all k ≥ N . Rewrite the expression (2.10) as

a a
N +1 n+1
0 < an+1 = aN × × ··· ×
aN an
and then using the inequalities (2.11) to obtain
aN (L − ǫ)n−N < an < aN (L + ǫ)n−N

1 N 1 1 N
aN
n
(L − ǫ)1− n < ann < aN
n
(L + ǫ)1− n . (2.12)
Taking n → ∞ to each inequality (2.12) to deduce

1
L − ǫ ≤ lim ann ≤ L + ǫ.
n→∞
Since ǫ is arbitrary, we conclude that

1
lim ann = L
n→∞
as desired.
Next, if L = ∞, then for every M > 0, there exists an N ∈ N such that k ≥ N implies
ak+1
> M.
ak
Using similar argument as above, it gives
an+1 > aN · M n+1−N

1 1 N
n+1
an+1 n+1
> aN · M 1− n+1
which shows that 1

n+1
lim an+1 > M.
n→∞
Since M is arbitrary large, we conclude that
1
lim ann = ∞.
n→∞
1
(b) Let an = n! > 0. Then we have
an+1 1
lim = lim = 0.
n→∞ an n→∞ n + 1
Hence the required result follows immediately from part (a).

Problem 2.14
Proof.
(−1)n
(a) Take an = n! . Then we have
an+1 −1
lim = lim = 0.
n→∞ an n→∞ n+1
1
By Problem 2.13(a) and Theorem 2.8, it can be seen easily that R = L = ∞.
(b) Suppose that

∞
X ∞
X
ω n+1 z 2n+1
f (ω) = and g(z) = .
n=0
(2n + 1)! n=0
(2n + 1)!
1
Then it is obvious that g(z) = f (z 2 ). Take an = (2n+1)! . Since
an+1 (2n + 1)!

lim = lim = 0,
n→∞ an n→∞ (2n + 3)!
1
Problem 2.13(a) asserts that lim ann = 0 and then Theorem 2.8 ensures that
n→∞
∞
X ω n+1
(2n + 1)!
n=0
converges for |ω| < ∞. Therefore, the relation g(z) = f (z 2 ) implies that the radius of
convergence of the power series of g is also ∞.
n!
(c) Let an = nn . Using [12, §215, Eqn. (1)] or [27, Eqn. (5.24), p. 61], we find that
an+1 (n + 1)! nn n n
lim = lim × = lim = e−1 .
n→∞ an n→∞ (n + 1)n+1 n! n→∞ n + 1
By Problem 2.13(a) and Theorem 2.8, we conclude that R = e.

2n
(d) Let an = n! . Then
an+1 2n+1 n! 2
lim = lim × n = lim = 0.
n→∞ an n→∞ (n + 1)! 2 n→∞ n + 1
Thus Problem 2.13(a) and Theorem 2.8 yield that R = ∞.

Problem 2.15
Proof.
(a) Let x ∈ R be such that | sin x| ≤ 21 . We claim that | sin(x + 2)| > 1
2. To see this, the
condition | sin x| ≤ 12 implies that there is a k ∈ Z such that
π π
− ≤ x − kπ ≤ .
6 6
π 2π
Since 3 < π < 4, we have 2 <2< 3 so that
π π π π 2π 5π
= − + < x + 2 − kπ < + = .
3 6 2 6 3 6
Consequently, we have the claim that
1
| sin(x + 2)| > . (2.13)
2
31
Next, since 0 ≤ | sin n| ≤ 1 for every n ∈ N, we certainly have

1
| sin n| n ≤ 1 (2.14)
for every n ∈ N. Now we claim that given ǫ > 0, one can find a positive integer N such
that
1
| sin N | N > 1 − ǫ. (2.15)
1
To prove this claim, we first notice that lim 2− n = 1, so there exists an N ′ ∈ N such that
n→∞
n ≥ N ′ implies
1
2− n > 1 − ǫ. (2.16)
On the one hand, if | sin N ′ | > 21 , then we get from the inequality (2.16) that
1 1
| sin N ′ | N ′ > 2− N ′ > 1 − ǫ.
On the other hand, if | sin N ′ | ≤ 21 , then the previous claim (2.13) and the inequality (2.16)
imply that
1
− 1
| sin(N ′ + 2)| N ′ +2 > 2 N ′ +2 > 1 − ǫ.
Hence they mean that our claim (2.15) is true or equivalently, there exists a sequence {nk }
of positive integers such that
1
lim | sin nk | nk ≥ 1. (2.17)
k→∞
Finally, the estimates (2.14) and (2.17) guarantee that

1
lim sup | sin n| n = 1.
n→∞
By Theorem 2.8, it concludes that R = 1.
(b) By Theorem 2.8, we have

2 1
L = lim sup(e−n ) n = lim e−n = 0
n→∞ n→∞
so that R = ∞.
Problem 2.16
Proof. By the definition, we have

1 1 √ √
2k
lim c2k = lim (2k ) 2k = lim 2= 2
k→∞ k→∞ k→∞
and
1 h 1 k2 i 2k−1
1 h 1 k i 2k−1
k
√
2k−1
lim c2k−1 = lim 1+ = lim 1+ = e.
k→∞ k→∞ k k→∞ k
Since e > 2, we establish from Theorem 2.8 that
1
R= √ .
e
This completes the analysis of the problem.
Problem 2.17
Proof. The following proof is basically adopted from [22, Theorem 3.50, pp. 74, 75]. Suppose
that βn = Bn − B for all n = 0, 1, 2, . . .. It is clear that
n
X
Cn = ck
k=0
= a0 b0 + (a0 b1 + a1 b0 ) + · · · + (a0 bn + a1 bn−1 + · · · + an b0 )
= a0 Bn + a1 Bn−1 + · · · + an B0
= a0 (B + βn ) + a1 (B + βn−1 ) + · · · + an (B + β0 )
= (a0 + a1 + · · · + an )B + a0 βn + a1 βn−1 + · · · + an β0
= An B + a0 βn + a1 βn−1 + · · · + an β0 . (2.18)
Since An B → AB as n → ∞, it follows from the expression (2.18) that Cn → AB if and only if
lim (a0 βn + a1 βn−1 + · · · + an β0 ) = 0. (2.19)

n→∞
∞
X ∞
X
Since ak converges absolutely, |ak | is finite and we denote this number by α. Given ǫ > 0.
k=0 k=0
Since βn → 0 as n → ∞, there exists an N ∈ N such that n ≥ N implies |βn | < ǫ. Therefore,
the triangle inequality shows that
|a0 βn + a1 βn−1 + · · · + an β0 | ≤ |a0 βn + · · · + an−(N +1) βN +1 | + |an−N βN + · · · + an β0 |

≤ |an−N βN + · · · + an β0 | + (|a0 | · |βn | + · · · + |an−(N +1) | · |βN +1 |)
< |an−N βN + · · · + an β0 | + ǫ(|a0 | + |a1 | + · · · + |an−(N +1) |)
≤ |an−N βN + · · · + an β0 | + ǫα. (2.20)
Since an → 0 as n → ∞, we fix N in the inequality (2.20) and then take n → ∞ to establish
lim sup |a0 βn + a1 βn−1 + · · · + an β0 | ≤ ǫα.

n→∞
Since ǫ is arbitrary, this proves the desired result (2.19) and thus
lim Cn = AB.
n→∞
Problem 2.18
Proof. If the complex number z satisfies |z| < R1 , then the power series
∞
X
|an | · |z n |
n=0
33
converges. Similarly, if |z| < R2 , then the power series

∞
X
|bn | · |z n |
n=0
converges. Therefore, both series converges within |z| < min(R1 , R2 ).

Clearly, by replacing an and bn by an z n and bn z n respectively in Problem 2.17, we get
k
X k
X
(aj z j )(bk−j z k−j ) = aj bk−j · z k = ck z k
j=0 j=0
and then Problem 2.17 ensures that

∞
X
cn z n
n=0
converges within |z| < min(R1 , R2 ). We have completed the proof of the problem.
Problem 2.19
Proof.
(a) Suppose that |z| < 1. Apply the given identity, we have
lim (1 − z)(1 + z + z 2 + · · · + z N ) = lim (1 − z N +1 )

N →∞ N →∞
∞
X
(1 − z) zn = 1
n=0
X∞
1
zn = .
1−z
n=0
∞
X ∞
X
n
(b) Let cn z be the Cauchy product of z n with itself. By the definitionb , we have
n=0 n=0
n
X
cn = (1 × 1) = n + 1.
k=0
Combining this fact and part (a), we see that
1 X∞ X∞ X ∞ X∞
1
n n n
2
= z z = (n + 1)z = nz n +
(1 − z) n=0 n=0 n=0 n=0
1−z
which asserts that

∞
X 1 1 z
nz n = − = .
(1 − z)2 1 − z (1 − z)2
n=0

b
On p. 28.
Problem 2.20
Proof. Let the power series under consideration is

∞
X
Cn z n . (2.21)
n=0
Recall from [1, p. 53] that x is an accumulation point of the set S if and only if every neighbor-
hood of x contains infinitely many points of S. Particularly, since S has an accumulation point
at 0, each neighborhood D(0; n1 ) contains a point zn 6= 0 of S. Thus we have
1
|zn | <
n
which means that {zn } converges to 0. By Theorem 2.12 (The Uniqueness Theorem for Power
Series), we conclude that the power series (2.21) is identically zero. This completes the proof of
the problem.
Problem 2.21
∞
X
Proof. Consider the power series g(z) = Dn z n , where
n=0

 1, if n = 0;
Dn =

0, if n ∈ N.
Since f ( n1 ) = g( n1 ) for every n = 2, 3, . . . and n1 → 0 as n → ∞, we deduce from Corollary 2.14

that Cn = Dn for all n = 0, 1, 2, . . .. In other words, we have f (z) = 1 and then
f ′ (z) ≡ 0,
so f ′ (0) > 0 is impossible. Hence we have completed the proof of the problem.
Problem 2.22
Proof. We write
∞
X
g(z) = f (z + α) = Cn z n .
n=0
1
Since lim sup |Cn | n < ∞, Theorem 2.8 ensures that the radius of convergence of g is nonzero.
n→∞
Hence it follows from Corollary 2.11 that
g(n) (0) f (n) (α)
Cn = =
n! n!
for all n ∈ N ∪ {0}, completing the proof of the problem.
35
Problem 2.23
Proof.
1
(a) By Problem 2.8, we have lim sup n n = 1 so that the comment following Corollary 2.14 on
n→∞
p. 32 implies that the series converges throughout |z − 1| < 1.
(b) As we have shown in Problem 2.14(a) that the radius of convergence of the series there is
∞, so the comment following Corollary 2.14 on p. 32 implies that the radius of convergence
of
X∞
(−1)n
(z + 1)n
n!
n=0
is also ∞.
(c) Note that

∞
X ∞
X 1 n
2 n
n (2z − 1) = 2n n2 z − . (2.22)
2
n=0 n=0
By Problem 2.8, we see that

1 2
lim sup(2n n2 ) n = lim 2n n = 2.
n→∞ n→∞
Again the comment following Corollary 2.14 on p. 32 implies that the series (2.22) con-
verges throughout |z − 21 | < 12 .
CHAPTER 3
Analytic Functions
Problem 3.1
Proof. Notice that f (z) = f (x, y) and z = x + iy, so it is easy to see that
f (x + h, y) − f (x, y) f (z + h) − f (z)
fx = lim = lim
h→0 h h→0 h
h∈R h∈R
and
f (x, y + h) − f (x, y) f (z + ih) − f (z)
fy = lim = lim .
h→0 h h→0 h
h∈R h∈R
Problem 3.2
Proof.
(a) Obviously, fx = 2x and fy = 2iy exist in a neighborhood of any point on the line y = x
and furthermore, they are continuous at all points on the line y = x. Since fy = ifx if
and only if y = x, Proposition 3.2 asserts that f is differentiable at all points on the line
y = x.
(b) By Definition 3.1, if f was analytic at z = (x, y), then f is differentiable in a neighborhood
of z = (x, y) which is prohibited by part (a).
Hence we have completed the proof of the problem.
Problem 3.3
37
Chapter 3. Analytic Functions 38
Proof. If g is differentiable at w, then Definition 2.4 implies
g(s) − g(w) = (s − w)[g ′ (w) + ǫ(s)],
where ǫ(s) → 0 as s → w. Therefore, we get
g(f (z + h)) − g(f (z)) = [f (z + h) − f (z)] · [g ′ (f (z)) + ǫ(h)], (3.1)
where ǫ(h) → 0 as h → 0. By dividing both sides of the expression (3.1) by h, we see that
g(f (z + h)) − g(f (z)) f (z + h) − f (z)

= [g′ (f (z)) + ǫ(h)] ×
h h
and thus
g(f (z + h)) − g(f (z))
(g ◦ f )′ (z) = lim
h→0 h
f (z + h) − f (z)
= lim [g′ (f (z)) + ǫ(h)] × lim
h→0 h→0 h
= g′ (f (z)) · f ′ (z).
Remark 3.1
Problem 3.3 is the Chain Rule for the complex variable case.
Problem 3.4
Proof. Since g 2 (z) = z, if z0 ∈ C, then we have
g2 (z) − g 2 (z0 )
=1
z − z0
which gives
g(z) − g(z0 ) 1
= . (3.2)
z − z0 g(z) + g(z0 )
√
Since g is continuous at z0 , it deduces from the expression (3.2) that
g(z) − g(z0 ) 1 1 1 1
g′ (z0 ) = lim = lim = lim p = = √ .
z→z0 z − z0 z→z0 g(z) + g(z0 ) z→z0 [g(z)]2 + g(z0 ) 2g(z0 ) 2 z0
Since z0 is arbitrary, we obtain our expected result and we have completed the proof of the
problem.
Problem 3.5
39
Proof. Suppose that f is analytic in the region D ⊆ C. By Definition 3.3, f is differentiable in

f
D. Since f ′ ≡ 0 in D, Proposition 3.1 implies that iy = fx = f ′ ≡ 0 in D, i.e.,
fx ≡ fy ≡ 0 (3.3)
in D. If f = u + iv, then the identities (3.3) mean that
ux ≡ uy ≡ vx ≡ vy ≡ 0
in D. Therefore, we follow from Theorem 1.10 that both u and v are constant in D. In particular,
f is constant in D. This completes the proof of the problem.
Problem 3.6
Proof. Suppose that f is analytic in the region D ⊆ C. By Definition 3.3, f is differentiable in

D. By Problem 3.3, we establish that f 2 is also differentiable in D so that Definition 3.3 again
implies that f 2 is analytic in D. Furthermore, we get from Problem 3.3 that
[f 2 (z)]′ = 2f (z)f ′ (z) ≡ 0
in D. By Problem 3.5, [f (z)]2 is constant. Let [f (z)]2 = A. Then we take the modulus to both
sides to get
|f (z)|2 = |A|
in D. In other words, |f | is constant in D. By Proposition 3.7, f is constant in D, completing
Problem 3.7
Proof. Let f : D → C be a nonconstant analytic function in the region D. Let f = u + iv. If f

mapped D into a straight line y = ax + b with a, b ∈ R, then the point f (x + iy) as a point in
C (or in R2 ), we read
v(x, y) = au(x, y) + b
in D. Since f is analytic in D, it is differentiable there and Proposition 3.1 says that
ux = vy = auy and uy = −vx = −aux (3.4)
in D. These two equations (3.4) imply that
(1 + a2 )ux = 0
in D. Since a is real, we have ux = 0 and then the second equation (3.4) gives uy = 0 in D. By
Theorem 1.10, u is constant so that Proposition 3.6 implies that f is constant which contradicts
to the hypothesis.
Next, assume that f mapped D into a circular arc of the circle |w − w0 | = r centred at
w0 ∈ C and radius r > 0, i.e., |f (z) − w0 | = r. Consider the map g(z) = f (z) − w0 which is also
nonconstant analytic in D. Now
|g(z)| = |f (z) − w0 | = r
in D. Hence Proposition 3.7 ensures that g and then f are both constants, a contradiction again.
Problem 3.8
Proof. Suppose that f (x, y) = (x2 − y 2 ) + iv is analytic at z. Then it must be differentiable in

a neighborhood D(z; r) of z and the Cauchy-Riemann equations give ux = vy and uy = −vx in
D(z; r). Since ux = 2x and uy = −2y, we have vx = 2y and vy = 2x so that
v(x, y) = 2xy + g(y) and v(x, y) = 2xy + h(x) (3.5)
in D(z; r). Therefore, the equations (3.5) imply that h(x) = g(y) = C for some constant C in
D(z; r). In conclusion, f must take the form
f (z) = f (x, y) = (x2 − y 2 ) + i(2xy + C) = (x + iy)2 + iC = z 2 + iC
for some constant C. This completes the proof of the problem.
Problem 3.9
Proof. Assume that f was an analytic function satisfying the requirement. By Proposition 3.1,
it satisfies the Cauchy-Riemann equations, i.e., vy = ux = 2x and vx = −uy = −2y. Therefore,
we have
v(x, y) = 2xy + F (x) and v(x, y) = −2xy + G(y)
for some functions F (x) and G(y). Clearly, they imply that
F (x) − G(y) = −4xy. (3.6)
Put y = 0 in the equation (3.6) to get
F (x) = G(0)
for all x in the domain of F . In other words, F (x) is a constant function. Similarly, it can
be shown that G(y) is also a constant function. However, the equation (3.6) means that −4xy
is a constant for all x and y, a contradiction and then no such analytic function exits. This
completes the proof of the problem.
Problem 3.10
Proof. Since f is entire, Proposition 3.1 implies that

u x = vy . (3.7)
By the given form, ux and vy are functions of x and y respectively. These facts and the equation
(3.7) force that ux = vy = A for some real constant A. Then simple integration gives
u(x) = Ax + B and v(y) = Ay + C
for some B, C ∈ R. Consequently, we have
f (z) = Ax + B + i(Ay + C) = A(x + iy) + B + iC = Az + B + iC.
41
Problem 3.11
Proof. By the definition, we have ez = ex cos y + iex sin y so that u = ex cos y and v = ex sin y.
(a) Direct computation gives ux = ex cos y, uy = −ex sin y, vx = ex sin y and vy = ex cos y.
It is clear that all ux , uy , vx and vy exist and continuous everywhere. Furthermore, since
ux = vy and uy = −vx hold everywhere, Proposition 3.2 ensures that ez is differentiable
everywhere, i.e., ez is entire.
(b) If z1 = a + ib and z2 = c + id, then we see that
ez1 +z2 = e(a+c)+i(b+d)
= ea+c [cos(b + d) + i sin(b + d)]
= ea+c [cos b cos d − sin b sin d + i(sin b cos d + cos b sin d)]
= ea+c (cos b + i sin b) × (cos d + i sin d)
= ez1 ez2 .
Problem 3.12
Proof. Let z = x + iy. Then it follows from the definition of the modulus of z (see p. 5) that
p
|ez | = |ex+iy | = |ex | · |eiy | = ex · | cos y + i sin y| = ex · cos2 t + sin2 t = ex ,
Problem 3.13
Proof. When z = x + iy, we notice that

ez = ex eiy . (3.8)
Suppose that z is a ray lying in the right half-plane {z ∈ C | Re (z) = x > 0}, i.e., x → +∞ and
y → L for some finite L or L = ±∞. Then it follows from the expression (3.8) that
|ez | = ex → ∞.
Similarly, if z is a ray lying in the left half-plane {z ∈ C | Re (z) = x < 0}, i.e., x → −∞ and
y → L for some finite L or L = ±∞. Then the expression (3.8) implies that
|ez | = ex → 0.
Finally, if the ray lies on the imaginary axis, then x = 0 and z = iy so that
|ez | = |eiy | = 1.
Since y → +∞ or y → −∞, ez moves along the unit circle anticlockwise and clockwise infinitely
many times respectively. We have completed the proof of the problem.
Problem 3.14
Proof.
(a) Since e0 = 1, ez = e0 implies that z = 2kπi for all k ∈ Z.
π π
(b) Since e 2 i = i, ez = e 2 i implies that z = πi
2 + 2kπi = ( π2 + 2kπ)i for all k ∈ Z.
(c) If we write z = x + iy, then ez = −3 will give ex eiy = 3eπi . Then its modulus implies that
ex = 3 or equivalently, x = ln 3. In addition, eiy = eπi leads to us that y = π + 2kπ for all
k ∈ Z. Hence the solutions of the equation have the form
z = x + iy = ln 3 + (2k + 1)πi,
where k ∈ Z.
(d) Similar to part (c), if we write z = x + iy, then ez = 1 + i implies that
√ 1 i √ π
ex eiy = 2 · √ + √ = 2e 4 i .
2 2
√ √
Thus ex = 2 and then x = ln 2 = 12 ln 2. Besides, it is easy to see that y = π
4 + 2kπ for
all k ∈ Z. In conclusion, the solutions of the equation take the form
1 1
z = x + iy = ln 2 + 2k + πi,
2 4
where k ∈ Z.
Problem 3.15
Proof.
(a) Recall the formulas of sin z and cos z, we obtain immediately that
1 iz 1 1
2 sin z cos z = 2 × (e − e−iz ) × (eiz + e−iz ) = (e2iz − e−2iz ) = sin 2z.
2i 2 2i
(b) Direct computation implies

h1 i2 h 1 i2
cos2 z + sin2 z = (eiz + e−iz ) + (eiz − e−iz )
2 2i
1 2iz −2iz 1
= (e + 2 + e ) − (e2iz − 2 + e−2iz )
4 4
= 1.
(c) Using the property (ez )′ = ez and Problem 3.3, we have

d h 1 iz i 1 1
(sin z)′ = (e − e−iz ) = (ieiz + ie−z ) = (eiz + e−iz ) = cos z.
dz 2i 2i 2
43
Problem 3.16
Proof. Recall the definitions of hyperbolic sine and cosine as follows:
ex − e−x ex + e−x
sinh x = and cosh x = , (3.9)
2 2
where x ∈ R.
(a) By the definition of sin z, we know that

π 1 π π 1 π π 1
sin + iy = [ei( 2 +iy) − e−i( 2 +iy) ] = (e−y+i 2 − ey− 2 i ) = (ey + e−y ) = cosh y.
2 2i 2i 2
(b) Using the result of Problem 3.20 in advance, we have

p
| sin z| = (sin x cosh y)2 + (cos x sinh y)2
r
e2y + 2 + e−2y e2y − 2 + e−2y
= sin2 x · + cos2 x ·
r 4 4
e2y + 2 + e−2y − 4 cos2 x
=
r 4
ey + e−y 2
= − cos2 x
2
q
= cosh2 y − cos2 x. (3.10)
On the left side or the right side of the square, we have z = ±(N + 12 )π + iy, where
−(N + 21 )π ≤ y ≤ (N + 12 )π so that cos ±(N + 21 )π = 0 and then the expression (3.10)
implies
| sin z| = cosh y.
By the A.M. ≥ G.M., we know that cosh y ≥ 1. Hence we have | sin z| ≥ 1 in this case.
Next, on the top side or the bottom side of the square, we see that z = x ± i(N + 12 )π,
where −(N + 21 )π ≤ x ≤ (N + 12 )π. By the definition (3.9), cosh y is clearly an even
increasing function of y, so we have cosh ±(N + 12 )π = cosh(N + 21 )π ≥ cosh π2 ≥ 2 and
then
cosh2 y − cos2 x ≥ 4 − 1 = 3.
√
Hence we deduce from the expression (3.10) that | sin z| ≥ 3 ≥ 1.
(c) If y → ±∞, then it follows from the definition (3.9) that cosh y → ∞. Since 0 ≤ cos2 x ≤ 1
for all x ∈ R, we conclude from the expression (3.10) that | sin z| → ∞ as Im z = y → ±∞.
Problem 3.17
Proof. Since cos z = 21 (eiz + e−iz ) and sin z = 1 iz

2i (e − e−iz ), we obtain from the Chain Rule that
1 h d iz d −iz i
(cos z)′ = (e ) + (e )
2 dz dz
1h d d i
= i (eiz ) − i (e−iz )
2 d(iz) d(−iz)
−1 iz
= (e − e−iz )
2i
= − sin z,
Problem 3.18
Proof. By the definition of sin z, we have
eiz − e−iz
=2
2i
eiz − e−iz = 4i. (3.11)
Set w = eiz . Then the equation (3.11) becomes w2 − 4iw − 1 = 0 and then the quadratic formula
[4, Eqn. (1), p. 4] gives
p √
4i ± (4i)2 − 4 × (−1) 4i ± 2 3i √
w= = = (2 ± 3)i.
2 2
√
Thus eiz = (2 ± 3)i and then we get
√ π
iz = ln(2 ± 3) + + 2kπ i
π 2
√
z= + 2kπ − i ln(2 ± 3),
2
where k ∈ Z. This ends the proof of the problem.
Problem 3.19
z
Proof. Since ee = 1 = e0 , we have
ez = 2kπi, (3.12)
where k ∈ Z. If k = 0, then ez = 0 which is impossible. Thus we suppose that k 6= 0.
π
If k > 0, then 2kπi = eln(2kπ)+i 2 and the equation (3.12) gives
π
ez = eln(2kπ)+i 2 .
Consequently, we obtain
π π
z = ln(2kπ) + i + 2nπi = ln(2kπ) + i 2nπ + ,
2 2
where n ∈ Z.
45
3π
Next, if k < 0, then we have 2kπi = (−2kπi)(−i) = eln(−2kπ)+i 2 . Similarly, the equation
(3.12) becomes
3π
ez = eln(−2kπ)+i 2
which implies
3π 3π
z = ln(−2kπ) + i
+ 2nπi = ln(−2kπ) + i 2nπ + ,
2 2
where n ∈ Z. Thus we complete the proof of the problem.
Problem 3.20
Proof. By the definition of sin z, we have

ei(x+iy) − e−i(x+iy)
sin(x + iy) =
2i
e ix−y − e−ix+y
=
2
e (cos x + i sin x) − ey (cos x − i sin x)
−y
=
2i
y
e +e −y e−y − ey
= × (i sin x) + cos x
2i 2i
ey + e−y ey − e−y
= × (sin x) + i · cos x
2 2
= sin x cosh y + i cos x sinh y,
Problem 3.21
Proof. Using similar attack as in the proof of Problem 2.14(a), the radius of convergence of the
series of f (z) is ∞. Furthermore, it is clear that the series converges absolutely, so Problem 2.17
says that
∞ hX
wn−k i X 1 X n k
∞
X ∞ n ∞ n
z n X wn X zk
f (z)f (w) = = · = Ck z · wn−k . (3.13)
n=0
n! n=0
n! n=0
k! (n − k)! n=0
n!
k=0 k=0
(z+w)n
By the binomial theorem, the bracket in the formula (3.13) is exactly n! , so the formula
(3.13) becomes
∞
X (z + w)n
f (z)f (w) = = f (z + w). (3.14)
n!
n=0
Let x ∈ R. By the power series representation of ex [25, Definition 4.5.1, p. 90], it is easy to
see that
f (x) = ex . (3.15)
Recall that the series converges absolutely so that rearrangement is permitted. Thus we have
y2 y3 y2 y4 y3
f (iy) = 1 + iy − − i + ··· = 1 − + − ··· + i y − + ··· . (3.16)
2 3! 2 4 3!
By the power series representation of sin y and cos y [25, p. 102]), the equation (3.16) can be
written as
f (iy) = cos y + i sin y. (3.17)
Hence we deduce immediately from the formulas (3.14), (3.15) and (3.17) that
f (z) = f (x + iy) = f (x)f (iy) = ex (cos y + i sin y) = ex+iy = ez
as desired. This completes the proof of the problem.
Problem 3.22
Proof. By Problem 3.21, we have

∞
X
iz (iz)n
e = (3.18)
n!
n=0
and
∞
X (−iz)n
e−iz = . (3.19)
n!
n=0
The difference of the two expressions (3.18) and (3.19) establishes

z3 z5
eiz − e−iz = 2i z − + − · · · = 2ig(z).
3! 5!
Hence it follows from the definition of sin z that g(z) = sin z, completing the proof of the
problem.
Problem 3.23
Proof. Since sin z is entire, it follows from Theorem 2.9 and Problem 3.22 that
z2 z4
cos z = (sin z)′ = g′ (z) = 1 − + − ··· .
2! 4!
CHAPTER 4
Line Integrals and Entire Functions
Problem 4.1
Proof. We say that z ∼ ω if there exists an one-to-one C 1 mapping λ satisfying Definition 4.4.
Now we are going to check the definition of an equivalence relation directly.
• Reflexivity: The identity i : [a, b] → [a, b] is clearly an one-to-one C 1 mapping, i′ (t) > 0
and z(i(t)) = z(t) for all t ∈ [a, b]. Thus we have z ∼ z.
• Symmetry: Suppose that z ∼ ω. By the definition, there exists an one-to-one C 1 mapping

λ : [c, d] → [a, b] such that λ(c) = a, λ(d) = b, λ′ (t) > 0 and
ω(t) = z(λ(t)) (4.1)
for all t ∈ [c, d].

We need a modified exercise from Rudin as follows:a
Lemma 4.1
Suppose that f ′ (x) > 0 on [a, b]. Then f is strictly increasing in [a, b] and if g
denotes its inverse function, then g is differentiable on [f (a), f (b)] and
1
g′ (f (x)) =
f ′ (x)
on [a, b].
Since λ′ (t) > 0 on [c, d], Lemma 4.1 implies that its inverse function λ−1 : [a, b] → [c, d] is
differentiable and
1
(λ−1 )′ (λ(t)) = ′ >0 (4.2)
λ (t)
for all t ∈ [c, d]. Furthermore, the formula (4.2) ensures that λ−1 is also an one-to-one C 1
mapping on [a, b].
a
In fact, Lemma 4.1 is [22, Exercise 2, p. 114] with the open interval (a, b) replaced by closed interval [a, b].
A proof of this exercise can be found in [26, pp. 85, 86] and a proof of Lemma 4.1 can be modified from it once
one-sided limits are considered.
47
Chapter 4. Line Integrals and Entire Functions 48
To continue our proof, if s = λ(t), then t = λ−1 (s) and the expression (4.1) gives immedi-
ately that
z(s) = ω(λ−1 (s))
which means that ω ∼ z.
• Transitivity: Suppose that z ∼ ω and ω ∼ ζ. Then there are one-to-one C 1 mappings
λ1 : [c, d] → [a, b] and λ2 : [e, f ] → [c, d] such that λ1 (c) = a, λ1 (d) = b, λ2 (e) = c,
λ2 (f ) = d, λ′1 (t) > 0 for all t ∈ [c, d], λ′2 (s) > 0 for all s ∈ [e, f ] and
ω(t) = z(λ1 (t)) and ζ(s) = ω(λ2 (s)). (4.3)
Now the composition function λ = λ1 ◦ λ2 : [e, f ] → [a, b] satisfies λ(e) = a and λ(f ) = b.
Furthermore, Problem 3.3 guarantees that
λ′ (s) = (λ1 ◦ λ2 )′ (s) = λ′1 (λ2 (s)) · λ′2 (s) > 0
for all s ∈ [e, f ]. Finally, the two equations (4.3) show that
z(λ(s)) = z(λ1 (λ2 (s))) = ω(λ2 (s)) = ζ(s)
for all s ∈ [e, f ]. By Definition 4.4, z ∼ ζ.
Hence these prove that the concept of “smooth curves” is an equivalence relation and this
Problem 4.2
Proof. Note that ż(t) = 2t + 2it. By Definition 4.3, we see that

Z Z 1
f (z) dz = f (t2 + it2 )(2t + 2it) dt
C 0
Z 1
= (t4 + it4 )(2t + 2it) dt
0
Z 1
2
= 2(1 + i) t5 dt
0
2i
= ,
3
Problem 4.3
Proof. In this case, we have ż(t) = cos t − i sin t so that Definition 4.3 asserts
Z Z 2π Z 2π
f (z) dz = (sin t − i cos t)(cos t − i sin t) dt = −i dt = −2πi. (4.4)
C 0 0
Our result is different from that in Example 2 because our curve is in the opposite direction of
that in Example 2.b We have ended the proof of the problem.
b
We remark that the result (4.4) is an immediate consequence of Proposition 4.7.
49
Problem 4.4
Proof. We follow the given hint. Suppose that f = u1 + iv1 , g = u2 + iv2 and α = A + iB, where
u1 , u2 , v1 , v2 are real-valued functions and A, B ∈ R. Suppose, further, that C is a smooth curve
given by z(t), where a ≤ t ≤ b. Then we follow from Definitions 4.1 and 4.3 that
Z Z b
[f (z) + g(z)] dz = [f (z(t)) + g(z(t))] · ż(t) dt
C a
Z b

= [u1 (z(t)) + u2 (z(t))] · ż(t) + i[v1 (z(t)) + v2 (z(t))] · ż(t) dt
a
Z b Z b
= [u1 (z(t)) + u2 (z(t))] · ż(t) dt + i [v1 (z(t)) + v2 (z(t))] · ż(t) dt
a a
hZ b Z b i
= u1 (z(t)) · ż(t) dt + i v1 (z(t)) · ż(t) dt
a a
hZ b Z b i
+ u2 (z(t)) · ż(t) dt + i v2 (z(t)) · ż(t) dt
a a
Z b Z b
= [u1 (z(t)) + iv1 (z(t))] · ż(t) dt + [u2 (z(t)) + iv2 (z(t))] · ż(t) dt
Za Z a
= f (z) dz + g(z) dz.

C C
Similarly, we have
Z Z b
αf (z) dz = (A + iB)[u1 (z(t)) + iv1 (z(t))] · ż(t) dt
C a
Z b

= [Au1 (z(t)) − Bv1 (z(t))] + i[Av1 (z(t)) + Bu1 (z(t))] · ż(t) dt
a
Z b Z b
= [Au1 (z(t)) − Bv1 (z(t))] · ż(t) dt + i [Av1 (z(t)) + Bu1 (z(t))] · ż(t) dt
a a
hZ b Z b i
= (A + iB) u1 (z(t)) · ż(t) dt + i v1 (z(t)) · ż(t) dt
a a
Z b
= (A + iB) [u1 (z(t)) + iv1 (z(t))] · ż(t) dt
Z a
=α f (z) dz.
C
Problem 4.5
Proof. Fix a ∈ C. Let b ∈ C with a 6= b and C be a smooth curve connecting a and b.

Furthermore, we let z : [0, 1] → C be a parametrization such that z(0) = a and z(1) = b. Since
F is analytic on C, Proposition 4.12 implies that
Z
F (b) − F (a) = F (z(0)) − F (z(1)) = F ′ (z) dz = 0.
C
In other words, F (b) = F (a). Since b is arbitrary, F is a constant, completing the proof of the
problem.
Problem 4.6
Proof. It is clear that z(t) = eit is a parametrization of the smooth curve |z| = 1. By Definition
4.3, we know that Z Z 2π
f (z) dz = f (eit )ieit dt.
|z|=1 0
By the hypothesis, f (eit )ieit is a continuous complex-valued function of t so that the proof of
Lemma 4.9 gives Z
f (z) dz = Reiθ
|z|=1
for some R ≥ 0 and θ ∈ [0, 2π].c

Since R and f (z) are real for all |z| = 1, we have
Z
f (z) dz = R
|z|=1
Z
−iθ
=e f (z) dz
|z|=1
Z 2π
= f (eit )iei(t−θ) dt
0
n Z 2π o
= Re f (eit )[− sin(t − θ) + i cos(t − θ)] dt
0
Z 2π
=− f (eit ) sin(t − θ) dt. (4.5)
0
Since −f (eit ) sin(t − θ) ≤ |f (eit ) sin(t − θ)| for all t ∈ [0, 2π], the equation (4.5) reduces to
Z Z 2π Z 2π Z 2π
f (z) dz ≤ |f (eit ) sin(t − θ)| dt ≤ | sin(t − θ)| dt = | sin t| dt. (4.6)
|z|=1 0 0 0
Direct computation shows

Z 2π Z π Z 2π
| sin t| dt = sin t dt − sin t dt = 4.
0 0 π
We conclude from the inequality (4.6) that

Z
f (z) dz ≤ 4,
|z|=1
ending the proof of the problem.
Problem 4.7
c
Since the integral is a complex number, we can express it in the polar form.
51
Proof. Suppose that f (z) = α + βz, a ≤ b and c ≤ d. Let z1 = a + ic, z2 = b + ic, z3 = b + id

and z4 = a + id. Now the line segment Γ1 from z1 to z2 can be parameterized by
γ1 (t) = z1 + t(z2 − z1 )
for t ∈ [0, 1]. Then Definition 4.3 shows that
Z Z 1
dz = 1 · (z2 − z1 ) dt = z2 − z1 (4.7)
Γ1 0
and
Z Z 1
(z2 − z1 )2 z 2 − z12
z dz = [z1 + t(z2 − z1 )](z2 − z1 ) dt = z1 (z2 − z1 ) + = 2 . (4.8)
Γ1 0 2 2
Similarly, if Γ2 , Γ3 and Γ4 are the line segments from z2 to z3 , z3 to z4 and z4 to z1 respectively,
then we have
Z Z Z
dz = z3 − z2 , dz = z4 − z3 and dz = z1 − z4 (4.9)
Γ2 Γ3 Γ4
as well as
Z Z Z
z 2 − z22 z 2 − z32 z12 − z42
z dz = 3 , z dz = 4 and z dz = . (4.10)
Γ2 2 Γ3 2 Γ4 2
Since
Z 4 Z
X
f (z) dz = f (z) dz,
Γ k=1 Γk
the integrals (4.7), (4.8), (4.9), (4.10) and Proposition 4.8 together imply that
Z 4 h Z
X Z i 4 Z
X 4 Z
X
f (z) dz = α dz + β z dz = α dz + β z dz = 0
Γ k=1 Γk Γk k=1 Γk k=1 Γk
which completes the proof of the problem.
Problem 4.8
Proof.
(a) Clearly, ż(θ) = Rieiθ 6= 0 for all t ∈ [0, 2π], so C is a smooth curve by Definition 4.2. Since
k 6= −1, it must be true that
z k+1 ′
= zk .
k+1
z k+1
Since k+1 is analytic on the smooth curve C, Proposition 4.12 implies that
Z
(Re2πi )k+1 (Re0 )k+1
z k dz = − = 0.
C k+1 k+1
(b) By Definition 4.2, we get

Z Z 2π Z 2π
Rk+1 i(k+1)θ 2π
f (z) dz = (Reiθ )k · Rieiθ dθ = iRk+1 ei(k+1)θ dθ = e = 0.
C 0 0 k+1 0

Problem 4.9
Proof.
2
(a) Let f (z) = z − i and F (z) = z2 − iz. Since ż(t) = 1 + 2it 6= 0, C is a smooth curve. In
addition, F is analytic on C and F ′ (z) = f (z), so we apply Proposition 4.12 to conclude
that
Z h (1 + i)2 i h (−1 + i)2 i
(z − i) dz = F (z(1)) − F (z(−1)) = − i(1 + i) − − i(−1 + i) = 0.
C 2 2
(b) If Γ is the line segment from −1 + i to 1 + i which is parameterized by
γ(t) = (−1 + i) + t[1 + i − (−1 + i)] = (2t − 1) + i
for t ∈ [0, 1], then we have
Z Z 1
(z − i) dz = (2t − 1) · 2 dt = 0. (4.11)
Γ 0
Since the sum C + Γ is a smooth closed curve and f (z) = z − i is entire, Theorem 4.16
(The Closed Curve Theorem) implies that
Z Z Z
(z − i) dz + (z − i) dz = (z − i) dz = 0. (4.12)
C Γ C+Γ
Combining the two results (4.11) and (4.12), we conclude that

Z
(z − i) dz = 0.
C
Problem 4.10
Proof.
(a) Recall that (ez )′ = ez . If z(t) = it for t ∈ [0, 1] which is the parametrization of a smooth
curve connecting 0 and i, then it follows from Proposition 4.12 that
Z i
ez dz = ez(1) − ez(0) = ei − 1.
0
(b) Recall that ( 21 sin 2z)′ = cos 2z. If z(t) = π2 + it for t ∈ [0, 1] which is the parametrization
of a smooth curve connecting π2 and π2 + i, then it follows from Proposition 4.12 that
Z π +i
2 sin 2z(1) sin 2z(0) sin(π + 2i) sin π sin(π + 2i)
cos 2z dz = − = − = .
π 2 2 2 2 2
2
By Problem 3.20, we see that sin(π + 2i) = −i sinh 2, so we conclude that

Z π +i
2 −i sinh 2
cos 2z dz = .
π 2
2

53
Problem 4.11
Proof. Let a, b ∈ D. Define the curve C connecting a and b by z(t) = (1 − t)a + tb for t ∈ [0, 1].
Since D is convex, we see that C ⊆ D. Thus Proposition 4.12 is applicable to f ′ along the curve
C to get Z
f (b) − f (a) = f ′ (z) dz. (4.13)
C
Since |f ′ (z)| ≤ 1 on D and C is a line segment of length b− a, Theorem 4.10 (The M -L Formula)
reduces the expression (4.13) to
Z
|f (b) − f (a)| = f ′ (z) dz ≤ 1 · |b − a| = |b − a|
C
which is our desired result. This ends the proof of the problem.
Problem 4.12
Proof. If a = b, then we have ea − eb = 0 = a − b. Therefore, we suppose that a 6= b in the

following discussion. Let f (z) = ez and D = {z ∈ C | Re z < 0}. Then D is a convex region
containing a and b and f is analytic in D. Furthermore, if z = x + iy ∈ D, then x < 0 so that
|f ′ (z)| = |ez | = |ex · eiy | = ex ≤ 1. (4.14)
Hence we may apply Problem 4.11 to conclude that
|ea − eb | ≤ |a − b|. (4.15)
We note that the inequality in Problem 4.11 can be strict provided that we have |f ′ (z)| < 1
in D. Since x < 0, we know that the inequality (4.14) is in fact strict and this implies that the
inequality (4.15) there is strict. We complete the proof of the problem.
CHAPTER 5
Properties of Entire Functions
Problem 5.1
Proof. It is clear that f is entire. Since f ′ (z) = 2z, f ′′ (z) = 2 and f (k) (z) = 0 for all k ≥ 3, it
follows from Corollary 5.7 that
f (z) = 4 + 4(z − 2) + 2(z − 2)2 .
Problem 5.2
Proof. It is obvious that f is entire. Since f (k) (z) = ez for all k ≥ 1, it follows from Corollary
5.7 that
∞
X
a a ea 2 a (z − a)k
f (z) = e + e (z − a) + (z − a) + · · · = e ,
2! k!
k=0
Problem 5.3
Proof.
(a) Using Problem 3.3 repeatedly, we have
f (k) (z) = (−1)k+1 f (k)(−z)
for all k ∈ N. In other words, the derivative of an odd (resp. even) function is an even
(resp.) function. Furthermore, f (z) = −f (−z) implies that f (0) = 0. These two facts
combine to give
f (2n) (0) = 0
55
Chapter 5. Properties of Entire Functions 56
for all n = 0, 1, 2, . . .. Hence we obtain from Corollary 5.7 that
X f (2k−1) (0) ∞
′ f (3) (0) 3
f (z) = f (0)z + z + ··· = z 2k−1 .
3! (2k − 1)!
k=1
(b) By the analysis in part (a), we have f (2k−1) (0) = 0 for all k ∈ N, so we follow again from
Corollary 5.7 that
∞
X f (2k) (0)
f ′′ (0) 2
f (z) = f (0) + z + ··· = z 2k .
2! (2k)!
k=1
We have ended the proof of the problem.
Problem 5.4
Proof. Suppose that

∞
X
f (z) = Ck z k .
k=0
On the one hand, using [4, Eqn. (1), p. 63], we have

Z
1 f (ω)
Ck = dω, (5.1)
2πi C ω k+1
where k = 0, 1, 2, . . .. On the other hand, Corollary 2.11 implies that
f (k) (0)
Ck = (5.2)
k!
for all k = 0, 1, 2, . . .. Hence our desired formula follows immediately by comparing the different
expressions (5.1) and (5.2). We complete the proof of the problem.
Problem 5.5
Proof. Let g(z) = f (z + a). Then g is entire because f is entire. It is easy to check that
g(k) (z) = f (k) (z + a) so that g(k) (0) = f (k) (a) for all k = 0, 1, 2, . . .. Now it follows from Problem
5.4 that
Z Z
(k) (k) k! g(ζ) k! f (ζ + a)
f (a) = g (0) = k+1
dζ = dζ, (5.3)
2πi C ζ 2πi C ζ k+1
where C is any circle surrounding the origin. If ζ(t) = Reit for some R > 0 and 0 ≤ t ≤ 2π,
then Definition 4.3 shows that the integral in the expression (5.3) becomes
Z Z 2π
f (ζ + a) f (Reit + a)
dζ = · Rieit dt. (5.4)
C ζ k+1 0 (Reit )k+1
57
Let ω = ζ + a. Then the locus of ω is the circle centred at a with radius R. Besides, we have
Reit = ζ(t) = ω(t) − a and ω̇(t) = Rieit . Therefore, we observe from the integral (5.4) that
Z Z 2π
f (ζ + a) f (Reit + a)
dζ = · Rieit dt
C ζ k+1 0 (Reit )k+1
Z 2π
f (ω(t))
= ω̇(t) dt
[ω(t) − a]k+1
Z0
f (ω)
= k+1
dω, (5.5)
C ′ (ω − a)
where C ′ is the circle surrounding the point a. Combining the two formulas (5.3) and (5.5), we
conclude that Z
(k) k! f (ω)
f (a) = dω.
2πi C ′ (ω − a)k+1
Problem 5.6
Proof.
(a) By the expression (5.1), we have

Z
f (k) (0) 1 f (ω)
Ck = = dω,
k! 2πi C ω k+1
f (ω)
where C is the circle centred at the origin with radius R. Since f is entire, ω k+1
is
continuous on the smooth curve C. Now the length of C is 2πR and
f (ω) |f (ω)| M
k+1
= k+1 ≤ k+1
ω R R
on C, so Theorem 4.10 (The M -L Formula) implies that
Z
1 f (ω) 1 M M
|Ck | = k+1
dω ≤ · k+1 · (2π) = k+1 ,
2π C ω 2π R R
as desired.
(b) Apply part (a) with R = M = 1 to get |Ck | ≤ 1 for all k = 0, 1, 2, . . ..

Problem 5.7
Proof. Let R > 0. On the circle |z| = R, we have |f (z)| ≤ A + BRk . By Problem 5.6(a), we
know that
A + BRk
|Cj | ≤ , (5.6)
Rj
where j = 0, 1, 2, . . .. If j > k, then j − k > 0 and the inequality (5.6) implies that
A B
lim |Cj | ≤ lim + = 0.
R→+∞ R→+∞ Rj Rj−k
Hence Cj = 0 for all j = k + 1, k + 2, . . . and Theorem 5.11 (The Extended Liouville Theorem)
follows, completing the proof of the problem.
Problem 5.8
Proof. On the circle |z| = R > 0, the hypothesis implies that

3
|f (z)| ≤ A + BR 2 . (5.7)
Thus we deduce from the bound (5.7), Problem 5.6(a) and Theorem 5.5 (The Taylor Expansion
of an Entire Function) that
3
(k) k!(A + BR 2 )
|f (0)| = |k!Ck | ≤ . (5.8)
Rk
Notice that the inequality (5.8) holds for every R > 0, so if k ≥ 2, then we conclude that
3
(k) k!(A + BR 2 )
lim |f (0)| ≤ lim = 0.
R→+∞ R→+∞ Rk
Hence f (k) (0) = 0 for all k = 2, 3, . . . and then f is a linear polynomial. We have completed the
Problem 5.9
Proof. Since f is entire, f ′ is everywhere differentiable by Corollary 5.6. By Definition 3.3, f ′

is also entire. By Theorem 5.11 (The Extended Liouville Theorem), f ′ (z) is a polynomial of
degree at most 1 and then we apply Theorem 5.5 (The Taylor Expansion of an Entire Function)
to f ′ to get
f (k) (0) = 0 (5.9)
for k = 3, 4, . . .. Furthermore, we have f ′ (0) = 0 so that Theorem 5.5 (The Taylor Expansion of
an Entire Function) and the values (5.9) ensure that
f ′′ (0) 2
f (z) = f (0) + f ′ (0)z + z = a + bz 2 ,
2!
′′
where a = f (0) and b = f 2(0) .
Next, we recall from Problem 5.4 that
Z
′′ 1 f ′ (ω)
f (0) = dω, (5.10)
2πi C ω2
59
where C is a circle surrounding the origin. In particular, we take C to be the unit circle |ω| = 1
′
so that f ω(ω)
2 ≪ 1 on C. Therefore, we apply Theorem 4.10 (The M -L Formula) to the integral
(5.10) to yield
1
|f ′′ (0)| ≤ · 1 · 2π = 1
2π
and hence
|f ′′ (0)| 1
|b| = ≤ .
2 2
Problem 5.10
Proof. Assume that f was a nonconstant entire function satisfying the two equations. By re-
peated applications of the two equations, we can show that f satisfies the following equations
f (z) = f (z + n) and f (z) = f (z + in) (5.11)

for all n ∈ Z.
Consider the unit square D = {a + ib | a, b ∈ [0, 1]} which is compact. Since f is entire, it
modulus is continuous on C and thus |f (D)| is a compact subset of R. By the Heine-Borel
Theorem [27, p. 30], |f (z)| is bounded on D.
Let z = x + iy ∈ C. Recall that x can be written as the sum of [x] and {x}, where [x] and
{x} are the greatest integer function and the fractional part of the real number x respectively.
We note that [x] ∈ Z and 0 ≤ {x} < 1. Similarly, we have y = [y] + {y}. Therefore, we obtain
z = x + iy = ([x] + i[y]) + ω,
where ω = {x} + i{y}. It is easily seen that ω ∈ D. By the equations (5.11), we see that
f (z) = f ((ω + i[y]) + [x]) = f (ω + i[y]) = f (ω).
In other words, f is bounded on C and Theorem 5.10 (Liouville’s Theorem) implies that f is
constant, a contradiction. Hence no such function exists and we have completed the proof of
the problem.
Problem 5.11
Proof. Let P (z) be a real polynomial of odd degree. By Problem 1.5, we know that the complex
zeros of a real polynomial appear in conjugate pairs. Consequently, the number of complex
zeros of P (z) must be even. Since deg P (z) is odd, Theorem 5.12 (The Fundamental Theorem
of Algebra) asserts that the number of zeros (counting multiplicity) of P (z) must be odd. Hence
P (z) has at least one real zero. This ends the analysis of the problem.
Problem 5.12
Proof. Let P (z) be a real polynomial of degree n. By Theorem 5.12 (The Fundamental Theorem
of Algebra), we can express P (z) as
P (z) = an (z − z1 )(z − z2 ) · · · (z − zn )
for some an ∈ C and z1 , z2 , . . . , zn are the zeros of P (z). By Problem 1.5, zi is a zero of P (z) if
and only if zi is a zero of P (z). There are two cases:
• Case (i): zi is real. Then zi is also real and so P (z) has z − zi as its factor.
• Case (ii): zi is non-real. Then we have
(z − zi )(z − zi ) = z 2 − 2Re (zi )z + |zi |2
which is a real quadratic polynomial.
Hence P (z) is a product of real linear and quadratic polynomials, completing the proof of the
problem.
Problem 5.13
Proof. We prove the result in several steps.
• Step 1: v(x, y) ≡ 0 if and only if y = 0. If we suppose that P (x+iy) = u(x, y)+iv(x, y),
where u and v are real-valued functions, then the hypothesis says that v(x, y) = 0 if and
only if y = 0.
• Step 2: Either v(x, y) > 0 or v(x, y) < 0 in the upper half-plane. Since P is a
polynomial, it is differentiable everywhere and thus v is continuous on C. Fix a point
x first, if v(x, y1 ) > 0 and v(x, y2 ) < 0 with 0 < y1 < y2 , then the Intermediate Value
Theorem [27, p. 101] asserts that there exists a y0 ∈ (y1 , y2 ) such that
v(x, y0 ) = 0
which contradicts Step 1. This means that either v(x, y) > 0 or v(x, y) < 0 for all y > 0.
Next, we suppose that x1 is a number such that v(x1 , y) > 0 for all y > 0. Fix a y > 0. If
v(x2 , y) < 0, where x2 6= x1 , then we apply the Intermediate Value Theorem to conclude
that
v(x0 , y) = 0
for some x0 ∈ (x1 , x2 ) or x0 ∈ (x2 , x1 ) which is a contradiction again. Thus we have the
desired result.
• Step 3: Either vy (x, 0) ≥ 0 for all x ∈ R or vy (x, 0) ≤ 0 for all x ∈ R. Assume

that vy (x, 0) changed sign on R, i.e., there exist x1 < x2 such that vy (x1 , 0) > 0 and
vy (x2 , 0) < 0. Define G1 , G2 : R → R by
G1 (y) = v(x1 , y) and G2 (y) = v(x2 , y).
The assumption vy (x1 , 0) > 0 implies that G1 is strictly increasing in (−δ1 , δ1 ) for some
δ1 > 0. Similarly, G2 is strictly decreasing in (−δ2 , δ2 ) for some δ2 > 0. Therefore, by
using Step 1, one can find a ǫ > 0 such that
v(x1 , ǫ) > v(x1 , 0) = 0 and v(x2 , ǫ) < v(x2 , 0) = 0. (5.12)

61
Now if we define F (x) = v(x, ǫ) which is a continuous function with respect to x, then we
follow from the Intermediate Value Theorem and the inequalities (5.12) that
F (x3 ) = v(x3 , ǫ) = 0
for some x3 ∈ (x1 , x2 ), a contradiction to Step 1.
• Step 4: u(x, 0) is monotonic on R. Since P (z) is entire, Proposition 2.3 or 3.1 implies
that
ux (x, 0) = vy (x, 0)
so that either ux (x, 0) ≥ 0 or ux (x, 0) ≤ 0 for all x ∈ R by Step 3. Hence u(x, 0) is
monotonic on R.
• Step 5: P (z) = α has only one solution for all α ∈ R. Let deg P (z) = n. By
Theorem 5.12 (The Fundamental Theorem of Algebra), the polynomial P (z) has exactly n
roots z1 , z2 , . . . , zn ∈ C, counted with multiplicity. Let zk = xk +iyk , where k = 1, 2, . . . , n.
Then we get
P (z) = A(z − z1 )(z − z2 ) · · · (z − zn ) (5.13)
for some A ∈ C.
Recall that P = u + iv, so P (zk ) = 0 implies that u(xk , yk ) = v(xk , yk ) = 0. By Step 1,
we must have yk = 0 for all k = 1, 2, . . . , n. Hence we obtain from the expression (5.13)
that
P (z) = A(z − x1 )(z − x2 ) · · · (z − xn ). (5.14)
Without loss of generality, we may assume that x1 ≤ x2 ≤ · · · ≤ xn . Now our hypothesis
forces that A must be real. Since P (x, 0) = u(x, 0), we get from the expression (5.14) that
u(x, 0) = A(x − x1 )(x − x2 ) · · · (x − xn ).
Assume that P (x, 0) had at least two distinct zeros, say x1 < x2 . Since P (x, 0) is a
continuous function of x, P (x, 0) = u(x, 0) cannot be monotonic, but this contradicts
Step 4. Therefore, we conclude that x1 = x2 = · · · = xn and the representation (5.14)
reduces to
P (z) = A(z − x1 )n . (5.15)
πi
Next, if n > 1, then z1 = x1 + e n is a non-real complex number and
P (z1 ) = A(z1 − x1 )n = Aeπi = −A. (5.16)
Recall that A is real, so the result (5.16) contradicts to our hypothesis that P is real if
and only if z is real. Consequently, we have n = 1 and so the expression (5.15) gives
P (z) = Az − Az1 ,
as desired.
We complete the analysis of the problem.
Problem 5.14
Proof. If P (α) = P ′ (α) = · · · = P (k−1) (α) = 0 and P (k) (α) 6= 0, then Corollary 5.7 implies that
P (k) (α) P (k+1) (α)

P (z) = (z − α)k + (z − α)k+1 + · · · = (z − α)k Q(z),
k! (k + 1)!
where
P (k) (α) P (k+1) (α)
Q(z) = + (z − α) + · · ·
k! (k + 1)!
P (k) (α)
and Q(α) = k! 6= 0.
Conversely, suppose that P (z) = (z − α)k Q0 (z), where Q0 (α) 6= 0. Then we have
P ′ (z) = (z − α)k−1 Q1 (z),
where Q1 (z) = (z − α)Q′0 (z) + kQ0 (z). Therefore, it is easily seen that
P (n) (z) = (z − α)k−n Qn (z), (5.17)
where Qn (z) = (z − α)Q′n−1 (z)+ kQn−1 (z) which satisfies Qn (α) = kQn−1 (α) for n = 1, 2, . . . , k.
Thus the expression (5.17) implies immediately that

 0, if 1 ≤ n ≤ k − 1;
P (n) (α) =

k!Q0 (α), if k = n.
Hence we have completed the analysis of the problem.
Problem 5.15
Proof. Fix a point z ∈ C. Consider the line segment connecting 0 and z which is parameterized
by {tz | t ∈ [0, 1]}. Suppose that Kz = {t ∈ [0, 1] | |f (tz)| ≤ 1} which is a bounded set. Define

 sup Kz , if Kz 6= ∅;
d= (5.18)

0, otherwise.
Obviously, we have 0 ≤ d ≤ 1. Let zd = dz. We claim that
|f (zd )| ≤ max{1, |f (0)|}. (5.19)
On the one hand, if d = 0, then zd = 0 so that
|f (zd )| = |f (0)|. (5.20)
On the other hand, suppose that d > 0 and {tn } is a sequence of Kz with limit point t0 . Since
f is continuous and |f (tn z)| ≤ 1 for all n ∈ N, we have
|f (t0 z)| = lim |f (tn z)| ≤ 1.

n→∞
In other words, t0 ∈ Kz and thus Kz is closed in [0, 1]. By the Heine-Borel Theorem, Kz is
compact. Furthermore, we recall from [27, Problem 4.27, pp. 44, 45] that d ∈ Kz which implies
|f (zd )| = |f (dz)| ≤ 1. (5.21)

63
Hence our claim (5.19) follows immediately from the results (5.20) and (5.21).
By the definition (5.18), we have |f (tz)| > 1 for all t ∈ [d, 1] and the hypothesis implies
that |f ′ (tz)| ≤ 1 for all t ∈ [d, 1], i.e., |f ′ (ω)| ≤ 1 along the line segment connecting zd and z.
Combining this fact and Proposition 4.12, we havea
Z z Z z
′
|z| ≥ |z − zd | ≥ |f (ω)| dω ≥ f ′ (ω) dω = |f (z) − f (zd )| ≥ |f (z)| − |f (zd )|
zd zd
and so
|f (z)| ≤ |f (zd )| + |z| ≤ max{1, |f (0)|} + |z|,
where z ∈ C. Consequently, it follows from Theorem 5.11 (The Extended Liouville Theorem)
that f is a linear polynomial. This completes the proof of the problem.
Problem 5.16
Proof. Let P (z) = an z n + an−1 z n−1 + · · · + an be a polynomial having zeros z1 , z2 , . . . , zn , where

an 6= 0. By [4, Eqn. (4), p. 67], the centroid of zeros of P (z) is
z1 + z2 + · · · + zn an−1
=− . (5.22)
n nan
Now we let ω1 , ω2 , . . . , ωn−1 be the zeros of P ′ (z) = nan z n−1 + (n − 1)an−1 z n−2 + · · · + a2 z + a1 .
In other words, we have
nan z n−1 + (n − 1)an−1 z n−2 + · · · + a2 z + a1 = nan (z − ω1 )(z − ω2 ) · · · (z − ωn−1 ). (5.23)
By comparing coefficients of z n−2 on both sides of the expression (5.23), we see immediately
that
(n − 1)an−1 = −nan (ω1 + ω2 + · · · + ωn−1 )

ω1 + ω2 + · · · + ωn−1 an−1
=− . (5.24)
n−1 nan
Hence the result follows if we combine the two expressions (5.22) and (5.24). This completes
Problem 5.17
Proof. Let S be a convex set. The statement for n = 1 and n = 2 are trivial. Assume that the
statement is true for n = k ∈ N, i.e., if z1 , z2 , . . . , zk ∈ S, then
a1 z1 + a2 z2 + · · · + ak zk ∈ S, (5.25)
k
X
where aj ≥ 0 for all 1 ≤ j ≤ k and aj = 1. Now let z1 , z2 , . . . , zk , zk+1 ∈ S, aj ≥ 0 for all
j=1
k+1
X
1 ≤ j ≤ k + 1 and aj = 1.
j=1
a
In fact, we also apply Definition 4.3 and Lemma 4.9 to show the third inequality holds.
k
X
Define A = ai . Then the definition gives 1 − A = ak+1 which implies that
j=1
a a2 ak
1
(a1 z1 + a2 z2 + · · · + ak zk ) + ak+1 zk+1 = A z1 + z2 + · · · + zk + (1 − A)zk+1 . (5.26)
A A A
k
X aj
Since = 1, the induction assumption (5.25) ensures that the bracket on the right-hand
A
j=1
side of the expression (5.26) is an element of S. Therefore, the convexity of S further implies
that the right-hand side of the expression (5.26) belongs to S. Hence our required result follows
from induction and it ends the proof of the problem.
Problem 5.18
Proof.
(a) Let z1,k , z2,k , . . . , zk,k be the zeros of Pk (z), where k ∈ N. Applying [4, Eqn. (4), p. 67],
we have
1
z1,k + z2,k + · · · + zk,k 1 h (k−1)! i
= × − 1 = −1.
k k k!
(b) Without loss of generality, we may assume that zk,k is a zero of Pk (z) with maximal
possible modulus. We note that
′ zk
Pk+1 (z) = 1 + z + · · · + = Pk (z)
k!
for every k = 0, 1, 2, . . .. By the hypothesis, we know that
|zj,k+1 | ≤ |zk+1,k+1 |
for all j = 1, 2 . . . , k. In other words, we get
zj,k+1 ∈ D(0; |zk+1,k+1 |)
for all j = 1, 2 . . . , k + 1. Since D(0; |zk+1,k+1 |) is convex, it contains the convex hull of
z1,k+1 , z2,k+1 , . . . , zk+1,k+1 by Definition 5.13. Finally, Theorem 5.14 (The Gauss-Lucas
Theorem) implies that all the zeros of Pk+1 ′ (z) = Pk (z) lie within the convex hull of the
zeros of Pk+1 (z). Particularly, we obtain
|zk,k | ≤ |zk+1,k+1 |
and hence {|zk,k |} is an increasing sequence.

Problem 5.19
65
Proof. Let Q(z) = 1 + z + z 2 + · · · + z n . Then it is easy to check that Q′ (z) = P (z). Note that
z n+1 − 1
Q(z) = ,
z−1
so the zeros of Q(z) lie on the unit circle.
Let S = {z1 , z2 , . . . , zn+1 } be set of the zeros of Q(z) and Conv (S) the convex hull of S.
Since S ⊆ D(0; 1) and D(0; 1) is a convex set, Definition 5.13 ensures that Conv (S) ⊆ D(0; 1).
Hence we deduce from Theorem 5.14 (The Gauss-Lucas Theorem) that the zeros of P (z) lie
within Conv (S) ⊆ D(0; 1). This completes the proof of the problem.
Remark 5.1
In fact, the convex hull of a finite set S = {z1 , z2 , . . . , zn+1 } is given by
n X o
n+1
Conv (S) = a1 z1 + a2 z2 + · · · + an+1 zn+1 aj ≥ 0 for 1 ≤ j ≤ n + 1 and aj .
j=1
See, for example, [26, Eqn. (10.33), p. 268].
Problem 5.20
Proof. Using the method of Chapter 1, we see that the solutions of the polynomial equation
z 2 = i are
1
z = ± √ (1 + i).
2
√
Therefore, we have i = √12 (1 + i).
√
To find the estimates of i by Newton’s method, we first let
f (z) z2 − i
f (z) = z 2 − i and g(z) = z − = z − .
f ′ (z) 2z
Take z0 = 1. Then we have

1−i 1+i
z1 = g(z0 ) = 1 − = ,
2 2
1+i − 2i 1+i 1+i 3
z2 = g(z1 ) = − = + = (1 + i),
2 1+i 2 4 4
17
z3 = g(z2 ) = (1 + i),
24
865
z4 = g(z3 ) = (1 + i).
1224
CHAPTER 6
Properties of Analytic Functions
Problem 6.1
1
Proof. It is easily seen that f (z) = is analytic everywhere except at z = 0. Therefore, if
√ z
z ∈ D(1 + i; 2), then we obtain
1 1
=
z (1 + i) + [z − (1 − i)]
1 1
= · z−(1+i)
1+i 1+
1+i
1 h z − (1 + i) [z − (1 + i)]2 i
= · 1− + − · · ·
1+i 1+i (1 + i)2
X∞
(−1)k (z − 1 − i)k
= .
(1 + i)k+1
k=0
Problem 6.2
Proof. By partial fractions, we have

1 1 2 1 1 1
f (z) = = = · + · . (6.1)
1 − z − 2z 2 (1 + z)(1 − 2z) 3 1 + z 3 1 − 2z
We know that
X ∞ X ∞
1 1
= (−z)n and = (2z)n (6.2)
1 + z n=0 1 − 2z n=0
for |z| < 1 and for |z| < 12 respectively. Hence, for |z| < 21 , by putting the identities (6.2) into
the expression (6.1), we see that
∞ h
X 2(−1)n 2n i n
f (z) = + z ,
3 3
n=0
67
Chapter 6. Properties of Analytic Functions 68
Problem 6.3
∞
X
Proof. We note that the series z n converges for |z| < 1, so Theorem 2.9 implies that
n=0
X
∞ ′ ∞
X
n
z = nz n−1 (6.3)
n=0 n=0
throughout |z| < 1. Therefore, we observe from the formula (6.3) that
∞
X ∞
X X
∞ ′ 1 ′ z
nz n = z nz n−1 = z zn =z· = (6.4)
1−z (1 − z)2
n=0 n=0 n=0
throughout |z| < 1.

Similarly, further application of Theorem 2.9 gives
X
∞ ′′ ∞
X
zn = n(n − 1)z n−2 . (6.5)
n=0 n=0
throughout |z| < 1. Hence we deduce from the formulas (6.4) and (6.5) that
∞
X ∞
X ∞
X
n2 z n = n(n − 1)z n + nz n
n=0 n=0 n=0
∞
X
2 z
=z n(n − 1)z n−2 +
(1 − z)2
n=0
X
∞ ′′ z
= z2 zn +
(1 − z)2
n=0
1 ′′ z
2
=z +
1−z (1 − z)2
2z 2 z
= 3
+
(1 − z) (1 − z)2
z(1 + z)
= .
(1 − z)3
Problem 6.4
Proof. Assume that f ( n1 ) = 1

n+1 for all n ∈ N. Consider g(z) = f (z) − z
1+z . Then we have
1 1 1
1 1
n
g =f − 1 = − =0
n n n +1 n+1 n+1
for all n ∈ N. Since { n1 } is a sequence of distinct points and 1

n → 0 ∈ D(0; 1), Theorem 6.9 (The
Uniqueness Theorem) implies that
z
f (z) = (6.6)
1+z
69
in D(0; 1). Since f is analytic in D(0; 1), it must be continuous there. This means that
lim f (zn ) = f (−1) (6.7)

n→∞
holds for every sequence {zn } ⊆ D(0; 1) such that zn → −1 as n → ∞ and zn 6= −1 for all
n ∈ N. However, the representation (6.6) guarantees that the limit (6.7) is impossible. Hence
no such analytic function exists and we complete the proof of the problem.
Problem 6.5
Proof. We fix z2 ∈ R. Let f (z) = sin(z + z2 ) and g(z) = sin z cos z2 + cos z sin z2 . Then f and g
are entire with respect to z. Since
1 1
f =g
n n
for all n ∈ N, it follows from Corollary 6.10 that f ≡ g in C (with respect to z).
Next, we fix z1 ∈ C and consider F (ω) = sin(z1 + ω) and G(ω) = sin z1 cos ω + cos z1 sin ω.
Obviously, they are entire with respect to ω. The previous paragraph ensures that
1 1 1 1 1
F = sin z1 + = f (z1 ) = g(z1 ) = sin z1 cos + cos z1 sin =G
n n n n n
for all n ∈ N. Thus Corollary 6.10 shows that F ≡ G in C (with respect to ω) and hence the
formula
sin(z1 + z2 ) = sin z1 cos z2 + cos z1 sin z2
holds for all z1 , z2 ∈ C. We complete the proof of the problem.
Problem 6.6
Proof. Let D = C \ {(k + 21 )π | k ∈ Z}. Since π is irrational, we have n1 ∈ D for all n ∈ N.

Since f (z) agrees with the analytic function tan z on the set { n1 } and n1 → 0 in D as n → ∞,
Corollary 6.10 implies that
f (z) = tan z
in D. Thus f (z) = i if and only if tan z = i. Using the identities for sin z and cos z on [4, p.
41], we have
eiz − e−iz
= −1
eiz + e−iz
eiz − e−iz = −eiz − e−iz
eiz = 0
which is impossible.
Furthermore, if f is entire, then f is continuous in C. Thus using similar argument as in
the proof of Problem 6.4, it is impossible that f is not entire. This completes the proof of the
problem.
Problem 6.7
Proof. We notice that N ∈ N. Since f is entire and |f (z)| ≥ |z|N , f (z) → ∞ as z → ∞. By

Theorem 6.11, f is a polynomial. Let f (z) = an z n + an−1 z n−1 + · · · + a1 z + a0 , where an is a
nonzero constant. Then the triangle inequality implies that
|f (z)| ≤ |an | · |z|n + |an−1 | · |z|n−1 + · · · + |a1 | · |z| + |a0 |

|an−1 | |a0 |
≤ |z|n · |an | + + ··· + n
|z| |z|
n
≤ 2|an | · |z| (6.8)
for sufficiently large z. Assume that n < N . Now the hypothesis and the inequality (6.8) give
|z|N ≤ 2|an | · |z|n
for sufficiently large enough z, but this means that an = ∞, a contradiction. Hence deg f (z) ≥ N
and we complete the proof of the problem.
Problem 6.8
Proof. Suppose that f is constant. Since |f (z)| ≤ 2 for√all |z| = 1 with Im z ≥ 0, it is actually
true for all |z| ≤ 1. In particular, we have |f (0)| ≤ 2 < 6.
Without loss of generality, we may assume that f is nonconstant. Define F (z) = f (z)f (−z).
By the definition on [4, p. 86], F is analytic in D(0; 1) and continuous on the compact set
D(0; 1) so that max |F (z)| exits in D(0; 1) by the Extreme Value Theorem. Now Theorem 6.13
(The Maximum Modulus Theorem) implies that this maximum must lie on |z| = 1. Hence our
hypotheses assert that
|F (z)| ≤ max F (z) = max(|f (z)| · |f (−z)|) ≤ 2 × 3 = 6

|z|=1 |z|=1
for all |z| ≤ 1. Particularly, if z = 0, we have
|f (0)|2 = |F (0)| ≤ 6
√
which implies |f (0)| ≤ 6, completing the proof of the problem.
Problem 6.9
Proof. Let z = x + iy and K ⊆ C be compact. Since ez is continuous on K, the Extreme Value

Theorem ensures that the function |ez | attains its maximum and minimum in K. Let K ◦ be
the interior of K and z0 = x0 + iy0 ∈ K ◦ . Assume that z0 was an extreme of the |ez |. We
know that K ◦ is open, so there exists a δ > 0 such that D(z0 ; δ) ⊆ K ◦ . It is obvious that
z± = z0 ± 2δ ∈ D(z0 ; δ) so that
δ δ
|ez+ | = ex0 + 2 > ex0 = |ez0 | and |ez− | = ex0 − 2 < ex0 = |ez0 |
71
which imply contradictions. Hence the modulus of ez can only have an extreme on ∂K which
Problem 6.10
Proof. Let f (z) = z 2 − z = z(z − 1) which is clearly nonconstant and analytic in D(0; 1). On
the one hand, since f (0) = 0, we have
min |f (z)| = 0.
|z|≤1
On the other hand, the Extreme Value Theorem and Theorem 6.13 (The Maximum Modulus
Theorem) ensure that max |f (z)| exists on |z| = 1. To find this maximum, if |z| = 1, then
|z|≤1
z = reiθ so that
q √
|f (z)| = |z − 1| = (cos θ − 1)2 + sin2 θ = 2 − 2 cos θ.
Therefore, f takes its maximum when θ = π or equivalently at z = −1 and furthermore, we

have
max |f (z)| = 2.
|z|≤1
Problem 6.11
Proof. Let C = ∂D(α; r) for some α ∈ C and r > 0. Since f is analytic in the compact set
D(α; r), it is analytic in an open set D containing D(α; r) by Definition 3.3. Consequently, there
exists a R > r such that
D(α; r) ⊆ D(α; R) ⊆ D. (6.9)
Otherwise, for each n ∈ N, one can find a ωn ∈ D(α; r+ n1 ) but ωn ∈/ D. Since {ωn } ⊆ D(α; r+1)
and |ωn | ≤ |ωn − α| + |α|, {|ωn |} is a bounded sequence. Thus the Bolzano-Weierstrass Theorem
[27, Problem 5.25] ensures that there is a convergent subsequence {ωnk }. Let ωnk → ω0 as
k → ∞. Next, we note that {ωnk } ⊆ C \ D. Since D is open in C, C \ D is closed in C so that
ω0 ∈ C \ D. (6.10)
1
However, since |ωnk − α| < r + nk for all k ∈ N, we have
|ω0 − α| ≤ r,
i.e., ω0 ∈ D(α; r) which is contrary to the set relation (6.10). Hence the set relation (6.9) holds
for some R > r.
Let z0 ∈ D(α; r) and n ∈ N. Using Theorem 6.4 (The Cauchy Integral Formula) to the
analytic function f n to get
Z
n 1 f n (z)
f (z0 ) = dz. (6.11)
2πi C z − z0
Now we want to apply Theorem 4.10 (The M -L Formula) to the expression (6.11), so we check
n (z)
/ C, the function fz−z
the hypotheses. Since z0 ∈ 0
is continuous on C. Let ρC (z0 ) = inf |z − z0 |.
z∈C
Recall that ρC (z0 ) = 0 if and only if z0 ∈ C. Therefore, ρC (z0 ) > 0. Since ρC (z0 ) ≤ |z − z0 | for
all z ∈ C, it must be true that
f n (z) Mn
≤ . (6.12)
z − z0 ρC (z0 )
Consequently, it follows from Theorem 4.10 (The M -L Formula) and the estimate (6.12) that
Z
n 1 f n (z) 1 Mn M nr
|f (z0 )| = dz ≤ · · (2πr) =
2πi C z − z0 2π ρC (z0 ) ρC (z0 )
which gives
r n1
|f (z0 )| ≤ · M. (6.13)
ρC (z0 )
Take n → ∞, the inequality (6.13) implies that |f (z0 )| ≤ M for all z0 ∈ D(α; r). This proves
Theorem 6.13 (The Maximum-Modulus Theorem) and we complete the proof of the problem.
Remark 6.1
For Problem 6.11, please also see the book of Pólya and Szegő [21, p. 29].
Problem 6.12
Proof. If both f and g are constant, then there is nothing to prove. Thus, without loss of
generality, we may assume that at least one of f and g is nonconstant. Let z0 be a maximum
of |f (z)| + |g(z)| in D. Furthermore, we let f (z0 ) = R1 e−iα and g(z0 ) = R2 e−iβ for some
α, β ∈ [0, 2π]. We consider the function h : C → C defined by
h(z) = f (z)eiα + g(z)eiβ .
Since f and g are analytic in D and at least one of them is nonconstant, h is nonconstant
and analytic in D. Therefore, the Extreme Value Theorem and Theorem 6.13 (The Maximum
Modulus Theorem) show that the modulus |h(z)| attains its maximum on ∂D.
By the definition, we have
|h(z)| ≤ |eiα | · |f (z)| + |eiβ | · |g(z)| = |f (z)| + |g(z)| (6.14)
for all z ∈ D. Besides, we also have
|h(z0 )| = |f (z0 )eiα + g(z0 )eiβ | = R1 + R2 = |f (z0 )| + |g(z0 )|. (6.15)
Now these two facts (6.14) and (6.15) combine to say that the modulus |h(z)| attains its maxi-
mum, which is exactly |f (z0 )| + |g(z0 )|, at z0 . Hence we conclude from the previous paragraph
that |f (z)| + |g(z)| takes its maximum on ∂D. This completes the proof of the problem.
Problem 6.13
73
Proof. Suppose that P (z) is a nonconstant polynomial and R > 0. Assume that P (z) 6= 0 in the
region D(0; R). Now Theorem 6.14 (The Minimum Modulus Theorem) says that the minimum
modulus of P (z) in D(0; R) exists on |z| = R, i.e., there exists a z0 such that |z0 | = R and
|P (z0 )| < |P (z)|
holds for all |z| < R. Particularly, we have
|P (z0 )| < |P (0)| (6.16)
However, we recall from Problem 1.26 that |P (z)| ≥ M |z|n for large |z|, where M is a positive
constant and n = deg P . So if R is large enough, then we have
|P (z0 )| > 2|P (0)|
which definitely contradicts the inequality (6.16). Hence P (ω) = 0 for some ω ∈ C which proves
the Fundamental Theorem of Algebra. This ends the analysis of the problem.
Problem 6.14
Proof. If P (z) is constant, then P (z) = a0 with |a0 | ≤ 1 and thus there is nothing to prove. We
suppose that P (z) is nonconstant. By Problem 5.6(b), we have
|ak | ≤ 1 (6.17)
for all k = 0, 1, 2, . . . , n. Take R > 1 and consider
P (z) a0 a1
f (z) = n
= n + n−1 + · · · + an (6.18)
z z z
defined in the compact annulus D = {z ∈ C | 1 ≤ |z| ≤ R}. Since f is C-analytic in the region
{z ∈ C | 1 < |z| < R}, the Extreme Value Theorem and Theorem 6.13 (The Maximum Modulus
Theorem) guarantee that the maximum modulus of f in D occurs on ∂D. Thus if |z| = 1, since
|P (z)| ≤ 1 for all |z| ≤ 1, then |f (z)| ≤ 1. Next, if |z| = R, then we obtain from the inequality
(6.17) and the expression (6.18) that
|a0 | |a1 | 1 1 1 R
|f (z)| ≤ n
+ n−1
+ · · · + |a n | ≤ 1 + + · · · + n
= 1 − n+1
· →1
|z| |z| R R R R−1
as R → ∞. In other words, this means that |f (z)| ≤ 1 for all |z| ≥ 1 and hence
|P (z)| ≤ |z|n
for all |z| ≥ 1. This completes the proof of the problem.
Problem 6.15
Proof. We notice that any line passing through z = 2 can be parameterized as
z(t) = 2 + teiθ , (6.19)
where t ∈ R and θ ∈ [0, 2π] is the angle between the line and the positive real axis. Therefore,
we observe that
f (z(t)) = (1 + teiθ ) · (−2 + teiθ )2

= 4 − 3t2 e2iθ + t3 e3iθ
= (4 − 3t2 cos 2θ + t3 cos 3θ) + i(−3t2 sin 2θ + t3 sin 3θ).
Direct computation gives
g(t) = |f (z(t))|2
= (4 − 3t2 cos 2θ + t3 cos 3θ)2 + (−3t2 sin 2θ + t3 sin 3θ)2
= t6 + 9t4 + 16 − 6t5 cos θ − 24t2 cos 2θ + 8t3 cos 3θ
so that
g′ (t) = 6t5 + 36t3 − 30t4 cos θ − 48t cos 2θ + 24t2 cos 3θ

= 6t(t4 − 5t3 cos θ + 6t2 + 4t cos 3θ − 8 cos 2θ) (6.20)
and
g′′ (t) = 30t4 + 108t2 − 120t3 cos θ − 48 cos 2θ + 48t cos 3θ.
Since the problem requires that |f (z)| has a relative extremum at z = 2, it follows from
Fermat’s Theorem [27, Theorem 8.6, p. 129] that g′ (0) = 0 which is obvious by the expression
(6.20). It is evident that
g ′′ (0) = −48 cos 2θ,
so we have g′′ (0) > 0 if and only if π4 < θ < 3π 5π 7π ′′
4 and 4 < θ < 4 . Similarly, g (0) < 0 if and
only if 0 ≤ θ < π4 , 3π 5π 7π
4 < θ < 4 and 4 < θ ≤ 2π. In other words, it follows from the Second
Derivative Test [2, §4.17, pp. 188, 189] that |f (z)| has a local minimum at z = 2 on any line
(6.19) with θ ∈ ( π4 , 3π 5π 7π
4 ) ∪ ( 4 , 4 ) and |f (z)| has a relative maximum at z = 2 through any line
π
(6.19) with θ ∈ [0, 4 ) ∪ ( 4 , 4 ) ∪ ( 7π
3π 5π
4 , 2π].
Now it remains to test the behaviour of |f | along the line (6.19) for
π 3π 5π 7π
θ= , , and .
4 4 4 4
In fact, it follows easily from the First Derivative Test [2, Theorem 4.8, p. 188] that |f | attains
its local maximum at z = 2 when θ = π4 , 7π 4 and |f | has a local minimum at z = 2 when
θ = 3π4 , 5π
4 . This completes the proof of the problem.
Problem 6.16
Proof. It is clear that f is analytic in C \ {0}. Note that
z2 − 1 (z − 1)(z + 1)
f ′ (z) = 2
=
z z2
75
so that f ′ (z) = 0 if and only if z = ±1. Since f (1) 6= 0, Theorem 6.17 ensues that z = 1 is
a saddle point of f . Since f (−1) = 0, the modulus |f | definitely has the absolute minimum at
z = −1 and thus z = −1 is not a saddle point of f by Definition 6.16. In conclusion, z = 1 is
the only saddle point of f .
Suppose that
z(t) = 1 + teiθ (6.21)
where t ∈ R and θ ∈ [0, 2π] is the angle between the line and the positive real axis. Therefore,
we observe that
(2 + teiθ )2
f (z(t)) =
1 + teiθ
(4 + 4teiθ + t2 e2iθ )(1 + te−iθ )
=
1 + 2t cos θ + t2
4(1 + t2 ) + t(4 + t2 )eiθ + 4te−iθ + t2 e2iθ
=
1 + 2t cos θ + t2
4(1 + t2 ) + t(8 + t2 ) cos θ + t2 cos 2θ t3 sin θ + t2 sin 2θ
= 2
+i
1 + 2t cos θ + t 1 + 2t cos θ + t2
so that
g(t) = |f (z(t))|2
h 4(1 + t2 ) + t(8 + t2 ) cos θ + t2 cos 2θ i2 t3 sin θ + t2 sin 2θ 2
= +
1 + 2t cos θ + t2 1 + 2t cos θ + t2
h2 (t) + h2 (t)
= 1 2 2 , (6.22)
h3 (t)
where
h1 (t) = 4(1 + t2 ) + t(8 + t2 ) cos θ + t2 cos 2θ,

h2 (t) = t3 sin θ + t2 sin 2θ,
h3 (t) = 1 + 2t cos θ + t2 .
Direct computation shows
h1 (0) = 4, h′1 (0) = 8 cos θ, h′′1 (0) = 8 + 2 cos 2θ,

h2 (0) = 0, h′2 (0) = 0, h′′2 (0) = 2 sin 2θ, (6.23)
h3 (0) = 1, h′3 (0) = 2 cos θ, h′′3 (0) = 2.
Now it yields from the expression (6.22) that
h3 (h1 h′1 + h2 h′2 ) − (h21 + h22 )h′3

g′ = 2 ·
h33
and then the values (6.23) imply that

g′ (0) = 0
for every θ ∈ [0, 2π]. Next, further differentiation gives
h33 [h3 (h1 h′1 + h2 h′2 ) − (h21 + h22 )h′3 ]′ − [h3 (h1 h′1 + h2 h′2 ) − (h21 + h22 )h′3 ] · (3h23 h′3 )
g′′ = 2 ·
h63
h3 [h3 (h1 h′1 + h2 h′2 ) − (h21 + h22 )h′3 ]′ − [h3 (h1 h′1 + h2 h′2 ) − (h21 + h22 )h′3 ] · (3h′3 )
=2·
h43
and then
[h3 (h1 h′1 + h2 h′2 )]′ − [(h21 + h22 )h′3 ]′
g′′ (0) = 2 · . (6.24)
h33 t=0
On the one hand, [h3 (h1 h′1 + h2 h′2 )]′ = h′3 (h1 h′1 + h2 h′2 ) + h3 [h1 h′′1 + (h′1 )2 + h2 h′′2 + (h′2 )2 ], so the
values (6.23) give
d
[h3 (h1 h′1 + h2 h′2 )] = 2 cos θ(32 cos θ) + [4(8 + 2 cos 2θ) + 64 cos2 θ]
dt t=0
= 128 cos2 θ + 8(4 + cos 2θ). (6.25)
On the other hand, since [(h21 + h22 )h′3 ]′ = (2h1 h′1 + 2h2 h′2 )h′3 + (h21 + h22 )h′′3 , we get from the
values (6.23) that
d
[(h2 + h22 )h′3 ] = 32 + 128 cos 2 θ. (6.26)
dt 1 t=0
Putting the results (6.25) and (6.26) into the expression (6.24), we observe that
g′′ (0) = 2[128 cos 2 θ + 8(4 + cos 2θ) − 32 − 128 cos2 θ] = 16 cos 2θ
Hence the argument of the proof of Problem 6.15 can be applied here to yield that |f | has a
local minimum at z = 1 on any line (6.21) with θ ∈ [0, π4 ) ∪ ( 3π 5π 7π
4 , 4 ) ∪ ( 4 , 2π]. Similarly, |f |
has a local maximum at z = 1 through any line (6.21) with θ ∈ ( 4 , 4 ) ∪ ( 5π
π 3π 7π
4 , 4 ). Now the
remaining unclear points are
π 3π 5π 7π
θ= , , and . (6.27)
4 4 4 4
However, we can check from the graphs of |f | that the point z = 1 is in fact an inflection pointa
along the lines corresponding to these values (6.27).
Figure 6.1 to Figure 6.3 below are some typical examples for the above discussion. In fact,
the
p orange line, the ′light red ′′line and the blue line correspond to the graphs of the function
g(t) = |f (z(t))|, g (t) and g (t) respectively. They show us that |f | has a local maximum,
a local minimum and an inflection point at z = 1 on the line 1 + t exp( π2 ), 1 + t exp( 33π
18 ) and
1 + t exp( 3π
4 ) respectively.
Figure 6.1: The function |f (z(t))| has a local maximum at z = 1 on the line 1 + t exp( π2 ).
a
Points where the concavity changes direction, see [2, p. 191].
77
Figure 6.2: The function |f (z(t))| has a local minimum at z = 1 on the line 1 + t exp( 33π
18 ).
Figure 6.3: The function |f (z(t))| has an inflection point at z = 1 on the line 1 + t exp( 3π
4 ).
Problem 6.17
Proof.
(a) By direction computation, we have
z3 + z − 2 (z − 1)(z 2 + z + 2)
f ′ (z) = = .
z3 z3
√
So f ′ (z) = 0 if and only if z = 1, −1±i
2
7
. By Theorem 6.17, the saddle points of f (z) are
exactly √ √
−1 + i 7 −1 − i 7
z1 = and z2 = .
2 2
(b) We divide the proof into several steps:

√
Finding possible relative extrema of |f (z)|. Note that |z1 | = |z2 | = 2.
– Step 1: √
Let z = 2eiθ , where θ ∈ [0, 2π]. Then we have
√
|f (z)| = |f ( 2eiθ )|
p √ √
3 − 2 2 cos θ · 5 − 4 cos 2θ
= (6.28)
p 2
√ √
15 − 12 cos 2θ − 10 2 cos θ + 8 2 cos θ cos 2θ
= . (6.29)
2
Suppose that g(θ) is the function inside the square root of the expression (6.29). It
is clear from the expression (6.29) that it suffices to find the extrema of g(θ). Simple
differentiation shows that
√ √ √
g′ (θ) = 24 sin 2θ + 10 2 sin θ − 8 2 sin θ cos 2θ − 16 2 cos θ sin 2θ
√ √ √
= 24 sin 2θ + 10 2 sin θ − 8 2 sin θ(1 − 2 sin2 θ) − 32 2(1 − sin2 θ) sin θ
√ √
= 48 2 sin3 θ − 30 2 sin θ + 24 sin 2θ.
Therefore, g′ (θ) = 0 if and only if

√ √
48 2 sin3 θ − 30 2 sin θ + 24 sin 2θ = 0
√ √
sin θ(8 2 sin2 θ − 5 2 + 8 cos θ) = 0
√ √
sin θ(−8 2 cos2 θ + 8 cos θ + 3 2) = 0
if and only if √ √
3 2 2
sin θ = 0 or cos θ = ,− .
4 4
√
Note that cos θ = 3 4 2 is impossible because of the definition. Consequently, Fermat’s
Theorem ensues
√
that g(θ) takes its possible relative maxima or minima when sin θ = 0
2
or cos θ = − 4 .
– Step 2: Determination of the relative minima of |f (z)|. We know that
√ √
g′′ (θ) = 144 2 sin2 θ cos θ − 30 2 cos θ + 48 cos 2θ.
If sin θ = 0, then cos θ = ±1 and cos 2θ = 1 so that

√
g ′′ (θ) = 48 ∓ 30 2 > 0.
By the Second Derivative Test, g(θ) takes its relative minima when sin θ = 0 and we
follow from the expression (6.28) that
p √ p √
√ 3−2 2 √ 3+2 2
|f ( 2)| = or |f (− 2)| =
2 2
are relative minima of |f (z)|.
79
√
2
– Step 3: Determination of the relative maxima of |f (z)|. If cos θ = − 4 , then
√
14
sin θ = ± 4 and cos 2θ = − 34 . In this case, we have
g′′ (θ) = −75 < 0,
so the Second Derivative

√
Test implies that g(θ) and thus |f (z)| takes its relative
2
maxima when cos θ = − 4 .
√ √
2 14
Next, the pair of values cos θ = − 4 and sin θ = 4 corresponds to
√ √
√ iθ 1 28 −1 + i 7
z= 2e = − + i = = z1 .
2 4 2
√ √
2 14
Similarly, the pair of values cos θ = − 4 and sin θ = − 4 corresponds to
√ √
√ iθ 1 28 −1 − i 7
z= 2e = − − i = = z2 .
2 4 2
Hence we conclude immediately that
|f (zi )| = max |f (z)|

√
on |z| = 2, where i = 1, 2.
(c) Rewrite f as
(z − 1)2 (z + 1) z 2 z z 1−z
f (z) = 2
= 1− 2
(1+z) = z − 2
1− 2
= z −1+ . (6.30)
z |z| |z| |z| |z|2
We apply the idea of the proof of Problem 6.16.
– Case (i): Lines through z1 . Consider
z1 (t) = z1 + teiθ ,
where t ∈ R and θ ∈ [0, 2π] is √the angle between the line and the positive real axis.
Since z1 (t) = − 12 + t cos θ + i( 27 + t sin θ), we have
√
|z1 (t)|2 = 2 + ( 7 sin θ − cos θ)t + t2 .
√
7
Since z1 = − 21 + t cos θ − i( 2 + t sin θ), the expression (6.30) becomes
√
3 √7 3 − t cos θ + i( 7 + t sin θ)
f (z1 (t)) = − + t cos θ + i + t sin θ + 2 √ 2
2 2 2 + ( 7 sin θ − cos θ)t + t2
3
3 − t cos θ
= − + t cos θ + √ 2
2 2 + ( 7 sin θ − cos θ)t + t2
√ √
h 7 7
+ t sin θ i
i + t sin θ + √ 2
2 2 + ( 7 sin θ − cos θ)t + t2
3 −1 − (√7 sin θ − cos θ)t − t2
= − t cos θ · √
2 2 + ( 7 sin θ − cos θ)t + t2
√7 3 + (√7 sin θ − cos θ)t + t2
+i + t sin θ · √
2 2 + ( 7 sin θ − cos θ)t + t2
and then
h21 (t) + h22 (t)
g(t) = |f (z1 (t))|2 = ,
h23 (t)
where
3 √
h1 (t) = − t cos θ [1 + ( 7 sin θ − cos θ)t + t2 ],
2
√7 √
h2 (t) = + t sin θ [3 + ( 7 sin θ − cos θ)t + t2 ],
2 √
h3 (t) = 2 + ( 7 sin θ − cos θ)t + t2 .
Direct computation shows that

3 1 √
h1 (0) = , h′1 (0) = (3 7 sin θ − 5 cos θ),
2√ 2
3 7 1 √
h2 (0) = , h′2 (0) = (13 sin θ − 7 cos θ),
2 √ 2
h3 (0) = 2, h′3 (0) = 7 sin θ − cos θ.
Therefore, we obtain
h3 (h1 h′1 + h2 h′2 ) − (h21 + h22 )h′3

g ′ (0) = 2 ·
h33 t=0
√ √ √ √
3 3 7
2[ 4 (3 7 sin θ − 5 cos θ) + 4 (13 sin θ − 7 cos θ)] − ( 49 + 63
4 )( 7 sin θ − cos θ)
=
√ √ 4
6(4 7 sin θ − 3 cos θ) − 18( 7 sin θ − cos θ)
=
√ 4
3 7
= sin θ
2
so that g′ (0) = 0 if and only if θ = 0, π. By the First Derivative Test, it is easily seen
that |f | has local minima at z = z1 through the lines z1 ± t.
– Case (ii): Lines through z2 . Next, we consider
z2 (t) = z2 + teiθ .
√
7
Since z2 (t) = − 12 + t cos θ + i(− + t sin θ), we have
2
√
|z2 (t)|2 = 2 − ( 7 sin θ + cos θ)t + t2 .
√
7
Since z2 = − 21 + t cos θ − i(− 2 + t sin θ), the expression (6.30) gives
√
3 √7 3 − t cos θ + i( − 7 + t sin θ)
f (z2 (t)) = − + t cos θ + i − + t sin θ + 2 √ 2
2 2 2 − ( 7 sin θ + cos θ)t + t2
3
3 − t cos θ
= − + t cos θ + √ 2
2 2 − ( 7 sin θ + cos θ)t + t2
√ √
h 7 − 7
+ t sin θ i
i − + t sin θ + √ 2
2 2 − ( 7 sin θ + cos θ)t + t2
3 −1 + (√7 sin θ + cos θ)t − t2
= − t cos θ · √
2 2 − ( 7 sin θ + cos θ)t + t2
81
√7 3 − (√7 sin θ + cos θ)t + t2

+i − + t sin θ · √ .
2 2 − ( 7 sin θ + cos θ)t + t2
Thus we get
H12 (t) + H22 (t)

G(t) = |f (z2 (t))|2 = ,
H32 (t)
where
3 √
H1 (t) = − t cos θ [−1 + ( 7 sin θ + cos θ)t − t2 ],
2
√7 √
H2 (t) = − + t sin θ [3 − ( 7 sin θ + cos θ)t + t2 ],
2
√
H3 (t) = 2 − ( 7 sin θ + cos θ)t + t2 .
Simple differentiation yields that

3 1 √
H1 (0) = − , H1′ (0) = (3 7 sin θ + 5 cos θ),
2√ 2
3 7 1 √
H2 (0) = − , H2′ (0) = (13 sin θ + 7 cos θ),
2 √ 2
H3 (0) = 2, H3′ (0) = −( 7 sin θ + cos θ).
Clearly, these values imply that
H3 (H1 H1′ + H2 H2′ ) − (H12 + H22 )H3′

G′ (0) = 2 ·
H33 t=0
√ √ √ √
2[− 43 (3 7 sin θ + 5 cos θ) − 3 4 7 (13 sin θ + 7 cos θ)] + 18( 7 sin θ + cos θ)
=
√ √ 4
2(−12 7 sin θ − 9 cos θ) + 18( 7 sin θ + cos θ)
=
√ 4
3 7
=− sin θ.
2
Hence G′ (0) = 0 if and only if θ = 0, π. Similar to Case (i), |f | has local maxima at
z = z2 through the lines z2 ± t.
CHAPTER 7
Further Properties of Analytic Functions
Problem 7.1
Proof. Let K be the compact domain of f . Consider ef (z) and e−if (z) . Since f nonconstant
analytic on K, both ef (z) and e−if (z) are nonconstant analytic on K. Furthermore, we know
that ef (z) 6= 0 and e−if (z) 6= 0 on K. Thus the Extreme Value Theorem, Theorems 6.13 (The
Maximum Modulus Theorem) and 6.14 (The Minimum Modulus Theorem) together assert that
max |ef (z) |, max |eif (z) |, min |ef (z) | and min |e−if (z) | assume on the boundary ∂K. Recall that ex
is an increasing function on R, so the facts
|ef (z) | = eRe f (z) and |e−if (z) | = eIm f (z)
imply that maxima/minima of eRe f (z) correspond to maxima/minima of Re f (z) and similarly,
maxima/minima of eIm f (z) correspond to maxima/minima of Im f (z). Hence Re f and Im f
assume their maxima and minima on ∂K, completing the proof of the problem.

Problem 7.2
Proof. Let f be a nonconstant analytic function on the region D. By Theorem 7.1 (The Open
Mapping Theorem), f (D) is open in C. Since f is continuous and D is connected, f (D) is also
connected. By Definition 1.6, f (D) is also a region. This completes the proof of the problem.
Problem 7.3
Proof.
(a) Assume that z was an interior point of S. Then there exists a δ > 0 such that D(z; δ) ⊆ S.
By Theorem 7.1 (The Open Mapping Theorem), f (D(z; δ)) is an open set in C containing
f (z), i.e., f (z) is an interior point of f (S) = T which contradicts the hypothesis.
83
Chapter 7. Further Properties of Analytic Functions 84
(b) The graph of the set S is shown in Figure 7.1 below. By the definition of f (z) = z 2 ,
Figure 7.1: The graph of the set S.
we see that f (S) ⊆ D(0; 4). We claim that f (S) = D(0; 4). Let u + iv ∈ D(0; 4) and
f (x + iy) = u + iv. Then we have u = x2 − y 2 and v = 2xy. Solving these equations give
4x4 − 4ux2 − v 2 = 0
and it implies that √

2 u± u2 + v 2
x = .
2
If we take s √
u+ u2 + v 2
x=− < 0,
2
then we have
√ z = x + iy ∈ S1 ⊂ S and we have the claim. Next, the points z1 = 1 and √
z2 = 2 + i 23 are clearly boundary points on S. Since f (z1 ) = 1 and f (z2 ) = − 21 + i 23 ,
1
they are clearly interior points of D(0; 4).

Problem 7.4
Proof. Suppose that f is nonconstant and D = D(0; 1). We want to verify that
f (D) = D. (7.1)
If ζ ∈ D \ f (D), then the map g : D → D defined by

1
g(z) =
f (z) − ζ
85
is nonconstant C-analytic in D. Now the Extreme Value Theorem and Theorem 6.13 (The
Maximum Modulus Theorem) assert that |g(z)| ≤ max |g(w)| and then
w∈C(0;1)
1 1 1 1
≤ max ≤ max =
f (z) − ζ w∈C(0;1) f (w) − ζ w∈C(0;1) |f (w)| − |ζ| 1 − |ζ|
for all z ∈ D. Thus we have 0 < 1 − |ζ| ≤ |f (z) − ζ| which means that the distance between the
point ζ and the set f (D)a is at least 1 − |ζ|. Therefore, we have
1 − |ζ|
D ζ, ⊆ D \ f (D)
2
and so the set D \ f (D) is open in C. Next, we follow from Theorem 7.1 (The Open Mapping
Theorem) that f (D) is open.
Recall the set relation
D = (D \ f (D)) ∪ f (D). (7.2)
Since D \ f (D) and f (D) are disjoint open subsets of D, the relation (7.2) implies that D is
disconnected by the definition, but this is definitely a contradiction. In other words, we have
either f (D) = D or f (D) = ∅. However, the latter is impossible by the hypothesis and then we
have the result (7.1), completing the analysis of the problem.
Problem 7.5
Proof. Let D = D(0; 1). If f (z) = 1, then we are done, so we suppose that f is nonconstant.
We claim that the number of zeros of f inside D is finite. To this end, let {αn } be the sequence
of zeros (counted with their multiplicities) of f inside D. Since |f (z)| = 1 on |z| = 1, we
actually have {αn } ⊆ D. As a consequence of the Bolzano-Weierstrass Theorem, {αn } contains
a convergent subsequence {αnk }. Let αnk → α0 . Since f is continuous in D, we have
0 = lim f (αnk ) = f (α0 )
k→∞
which implies that α0 ∈

/ C(0; 1), i.e., α0 ∈ D. Since D is a region, Theorem 6.9 (The Uniqueness
Theorem) says that f ≡ 0 in D but it contradicts the continuity of f in D. This proves our
claim.
Next, we define the function
f (z)
g(z) = , (7.3)
Bα1 (z)Bα2 (z) · · · Bαn (z)
where Bα (z) is the Möbius transformation considered on [4, p. 95]. There are two cases.
• Case (i): g is constant. Then it means that

f (z) = cBα1 (z)Bα2 (z) · · · Bαn (z) (7.4)
for some c ∈ C. If αj 6= 0, then f is undefined at z = α1j which contradicts the hypothesis
that f is entire. Consequently, we have αj = 0 for all j = 1, 2, . . . , n so that the expression
(7.4) reduces to
f (z) = cz n
as required.
a
That is ρf (D) (ζ) = inf |f (z) − ζ|.
z∈D
• Case (ii): g is nonconstant. We claim that g is analytic in D. In fact, it is clear for

z ∈ D \ {α1 , α2 , . . . , αn }. In the case z = αj , we need the following lemma:b
Lemma 7.1
If f is analytic in a region G and f 6≡ 0 on G, then for each α ∈ G with f (α) = 0
there exists an integer n ≥ 1 and an analytic function g : G → C such that
f (z) = (z − α)n g(z) and g(α) 6= 0.
Apply Lemma 7.1 to the function f (z), the expression (7.3) can be rewritten as
Y n
(z − α1 )(z − α2 ) · · · (z − αn )h(z)
g(z) = n = h(z) (1 − αj z), (7.5)
Y z − αj
j=1
1 − αj z
j=1
where h is analytic in D and h(αj ) = 6 0 for all j = 1, 2, . . . , n. Thus the representation

(7.5) implies that g(z) is analytic at z = αj for all j = 1, 2, . . . , n.
Since g is nonconstant analytic in D, we apply the Extreme Value Theorem and Theorem
6.13 (The Maximum Modulus Theorem) to the expression (7.3) to get
|f (ζ)|
|g(z)| ≤ max |g(ζ)| = max n = max |f (ζ)| = 1 (7.6)
|ζ|=1 |ζ|=1 Y ζ − αj |ζ|=1
1 − αj ζ
j=1
for all z ∈ D. Recall that |αj | < 1 so that |αj |−1 > 1. By the representation (7.5) and the
property of h, we see that g(z) 6= 0 for all z ∈ D. Now it follows from Theorem 6.14 (The
Minimum Modulus Theorem) that
1 = min |f (ζ)| = min |g(ζ)| ≤ |g(z)| (7.7)

|ζ|=1 |ζ|=1
for all z ∈ D. Combining the inequalities (7.6) and (7.7), we have |g(z)| = 1 for all z ∈ D,
i.e., g(z) = eiθ for some θ so that f has the form (7.4) and then our result follows in this
case.
Problem 7.6
P (z)
Proof. Suppose that D = D(0; 1) and f (z) has poles at {α1 , α2 , . . . , αn } ⊆ D. Let f (z) = Q(z) ,
where P (z) and Q(z) are polynomials. Then we have
P (z)
f (z) = n .
Y
(z − αj )
j=1
b
Read [7, Corollary 3.9, p. 79] for a proof of it. The case for polynomials has been given on [4, p. 67].
87
Define
Yn
P (z) z − αj P (z)
g(z) = f (z)Bαj (z) = n · = n (7.8)
Y 1 − αj z Y
(z − αj ) j=1 (1 − αj z)
j=1 j=1
which is rational in D. Since |αj | < 1, we have |αj |−1 > 1 and then the expression (7.8) ensures
that g(z) has no poles in D. Furthermore, the representation (7.8) also implies that, on |z| = 1,
|g(z)| = |f (z)| · |Bαj (z)| = |f (z)|.
Problem 7.7
Proof.
(a) Let D = D(0; 1) and |a| < 1. Recall that Ba : D → D given by

ω−a
Ba (ω) = (7.9)
1−a·ω
α
is analytic and bounded in D by 1. Put a = R and ω = Rz into the representation (7.9) to
get
z z
−α R(z − α)
B Rα = R α Rz = 2
R 1− R · R R −α·z
which is analytic in D(0; R). Note that |α| < R. Following the idea on [4, p. 95], on
|z| = R, we have
z 2 R(z − α) R(z − α)
B Rα = 2 ·
R R − α · z R2 − α · z
R2 (|z|2 − αz − αz + |α|2 )
= 4
R − R2 αz − R2 αz + |α|2 · |z|2
R2 − αz − αz + |α|2
= 2
R − αz − αz + |α|2
= 1.
In other words, the function B Rα ( Rz ) maps the circle |z| = R into C(0; 1).
n
1 Y 2
(b) Let f (z) = (z − α1 )(z − α2 ) · · · (z − αn ) and g(z) = n (R − αk z). It is clear that both
R
k=1
f and g are analytic throughout D(0; R) and part (a) ensures that they have the same
magnitude on |z| = R, i.e.,
|f (z)| = |g(z)| (7.10)
on |z| = R. Since we have
1
g(0) = · R2n = Rn 6= 0
Rn
Theorem 6.13 (The Maximum Module Theorem) and Theorem 6.14 (The Minimum Mod-
ule Theorem) imply that there exist points |z1 | = |z2 | = R such that
|g(z1 )| > Rn and |g(z2 )| < Rn . (7.11)

Consequently, we obtain from the results (7.10) and (7.11) that

pn
p
|z1 − α1 | · |z1 − α2 | · · · |z1 − αn | = n |f (z1 )| > R
and p p
n n
|z2 − α1 | · |z2 − α2 | · · · |z2 − αn | = |f (z2 )| < R.
Problem 7.8
Proof. If f is constant, then there is nothing to prove. Thus we may assume that f is noncon-
stant. Since f is analytic in the annulus 1 ≤ |z| ≤ 2, if we define
f (z)
g(z) = ,
z2
then g is also nonconstant analytic in the annulus. By Theorem 6.13 (The Maximum Modulus
Theorem), max |g(ω)| occurs on |ω| = 1 or |ω| = 2 so that we obtain
|f (z)| |f (ω)|
= |g(z)| ≤ max |g(ω)| = max ≤1
|z|2 |ω|2
throughout the annulus. Hence we have |f (z)| ≤ |z|2 throughout the annulus which ends the
Problem 7.9
1
Proof. Define F (z) = 10 f (2z). Then F is analytic in D(0; 1). Furthermore, we have |F (z)| ≤ 1
1
in D(0; 1) and F ( 2 ) = 0. It is easily seen that our F is exactly the function f considered in [4,
Example 1, p. 95]. Therefore, we have
z − 21
|F (z)| ≤
1 − 12 z
or equivalently,
z − 21
|f (2z)| ≤ 10 . (7.12)
1 − 12 z
1
Put z = 4 into the inequality (7.12), we get
1 20
f ≤ ,
2 7
Problem 7.10
89
Proof. Let D = D(0; 1). We define g : D → C by
g = Bf (α) ◦ f
where Bα is the Möbius transformation

z−α
Bα (z) = .
1 − αz
Since f and Bf (α) are analytic in D, g is also analytic in D. Furthermore, since f and Bf (α) are
bounded by 1 in D, g is also bounded by 1 in D. By direct computation, we know that
1
Bα′ (α) = . (7.13)
1 − |α|2
Using Problem 3.3 and the fact (7.13), we obtain
f ′ (α)
g′ (α) = Bf′ (α) (f (α)) × f ′ (α) = (7.14)
1 − |f (α)|2
Since f (α) 6= 0, we have 1 − |f (α)|2 < 1. Furthermore, it is impossible that |f (α)| = 1 by

Theorem 6.13 (The Maximum Modulus Theorem). Hence the expression (7.14) implies that
|f ′ (α)| < |g ′ (α)|.
Problem 7.11
Proof. By Problem 7.10, we may assume that f (α) = 0 so that
f (z) − f (α) f (z)

f ′ (α) = lim = lim . (7.15)
z→α z−α z→α z−α
Again the assumption and [4, Example 1, p. 95] give
|f (z)| ≤ |Bα (z)| (7.16)
throughout D(0; 1). Therefore, the limit (7.15) and the estimate (7.16) combine to imply
Bα (z) Bα (z) − Bα (α)

f ′ (α) ≪ lim = lim = Bα′ (α).
z→α z − α z→α z−α
Since f is arbitrary, we conclude that
max |f ′ (α)| = Bα′ (α).

f
Problem 7.12
Proof. We imitate the proof of Proposition 7.3. In fact, we introduce the auxiliary function
g(z) = (z 2 + R2 )2 f (z).
For any z with |z| = R and Im z ≥ 0, we have

|z − iR| 2
|(z − iR)2 f (z)| ≤ = sec2 θ
|Re z|
for some θ ∈ [0, π4 ] so that

|(z − iR)2 f (z)| ≤ 2. (7.17)
Similarly, if |z| = R and Im z ≤ 0, then we obtain
|(z + iR)2 f (z)| ≤ 2. (7.18)
Combining the estimates (7.17) and (7.18), for all z with |z| = R, we get
|g(z)| = |(z + iR)2 | · |(z − iR)2 | · |f (z)| ≤ 2 × (|z| + |R|)2 = 8R2 .
Since g is nonconstant analytic in D(0; R), Theorem 6.13 (The Maximum Modulus Theorem)
implies that
|g(z)| ≤ 8R2
in D(0; R) and so
8R2
|f (z)| ≤ (7.19)
|z 2 + R2 |2
for all z ∈ D(0; R). If we take R → ∞, then we deduce from the inequality (7.19) that f (z) = 0
for all z ∈ C, completing the proof of the problem.
Problem 7.13
Proof.
(a) Let Γ be the boundary of a rectangle R in C and z ∈ C. We want to apply Theorem 7.4
(Morera’s Theorem), so we check its hypotheses into four steps.
– Step 1: The continuity of an auxiliary function. Define ϕ : C × [0, 1] → C by


 sin zt
 , if t 6= 0;
ϕ(z, t) = t (7.20)


z, if t = 0.
We claim that ϕ is continuous on C×[0, 1]. To this end, we need the following lemma:
Lemma 7.2
Suppose that ζ is complex. Then we have
sin ζ
lim = 1.
ζ→0 ζ
91
Proof of Lemma 7.2. Notice that

sin ζ sin ζ − sin 0
lim = lim = cos 0 = 1
ζ→0 ζ ζ→0 ζ−0
as required.
It is clear from the definition (7.20) that ϕ(z, t) is continuous for every z ∈ C and
t ∈ (0, 1], so it suffices to check its continuity at t = 0. In fact, for every fixed z ∈ C,
we get from Lemma 7.2 that
sin zt
lim ϕ(z, t) = z lim = z = ϕ(z, 0).
t→0 zt→0 zt
t6=0 t6=0
This shows that ϕ is continuous on C × [0, 1] as desired.

– Step 2: f is continuous on C. With the aid of the auxiliary function (7.20), we
can write Z 1
f (z) = ϕ(z, t) dt
0
so that Z 1
|f (z) − f (ω)| ≤ |ϕ(z, t) − ϕ(ω, t)| dt. (7.21)
0
Given ǫ > 0. By Step 1, there exists a δ > 0 such that |(z, t) − (ω, s)| < δ implies
|ϕ(z, t) − ϕ(ω, s)| < ǫ.
Particularly, if we take t = s, then |z − ω| < δ implies
|ϕ(z, t) − ϕ(ω, t)| < ǫ
and then the inequality (7.21) can be reduced to
|f (z) − f (ω)| < ǫ.
In other words, f is continuous on C.

– Step 3: Absolute Convergence of the Double Integral. To evaluate the integral
Z Z Z 1
sin zt
f (z) dz = dt dz, (7.22)
Γ Γ 0 t
we need to interchange the order of integration of the double integral (7.22). As
suggested by the example given in [4, pp. 98, 99], we have to establish the absolute
convergence of the double integral (7.22). To this end, we fix z ∈ C. Now Lemma 7.2
ensures that there exists a M > 0 such that
sin ζ
≤M (7.23)
ζ
for all 0 < |ζ| ≤ |z|. If ζ = tz, where t ∈ (0, 1], then the inequality (7.23) implies that
sin zt
≤ M |z|
t
which gives Z 1
sin zt
dt ≤ M |z|. (7.24)
0 t
Since Γ is a rectangle, it must be of finite length. Since the function F (z) = |z| is
continuous on the compact set Γ, F is bounded there. Hence we follow from the
inequality (7.24) and Theorem 4.10 (The M -L Formula) that
Z Z 1 Z
sin zt
dt dz ≤ M |z| dz < ∞.
Γ 0 t Γ
– Step 4: The Application of Theorem 7.4 (Morera’s Theorem). For each

fixed t ∈ [0, 1], since sin zt is entire, we follow from Theorem 4.14 (The Rectangle
Theorem) that Z
sin zt dz = 0.
Γ
Thus we establish from this and Step 3 that
Z Z Z 1 Z 1Z Z 1 Z
sin zt sin zt 1
f (z) dz = dt dz = dz dt = sin zt dz dt = 0.
Γ Γ 0 t 0 Γ t 0 t Γ
Finally, we conclude that f is entire.
(b) Fix z ∈ C and let R = |z|. By Problem 3.22, we have
∞
sin zt X (−1)n z 2n+1 2n
= t , (7.25)
t (2n + 1)!
n=0
where t ∈ (0, 1]. It is clear that tz ∈ D(0; R), so we see that

(−1)n z 2n+1 2n R2n+1
t ≤ .
(2n + 1)! (2n + 1)!
Since
R2n+3 (2n + 1)! R2
lim sup × = lim sup = 0,
n→∞ (2n + 3)! R2n+1 n→∞ (2n + 3)(2n + 2)
the Ratio Test [27, Theorem 6.8, p. 77] implies that the series
X∞
R2n+1
n=0
(2n + 1)!
converges. Thus the Weierstrass M -test [28, Theorem 10.5, p. 3] asserts that the series
(7.25) converges uniformly for all t ∈ (0, 1]. Consequently, this fact allows us to interchange
the integration and summation in the following calculation
Z 1X ∞
(−1)n z 2n+1 2n
f (z) = t dt
0 n=0 (2n + 1)!
X∞ Z
(−1)n z 2n+1 1 2n
= t dt
(2n + 1)! 0
n=0
X∞
(−1)n z 2n+1
= . (7.26)
n=0
(2n + 1)(2n + 1)!
Since z is arbitrary, the power series representation (7.25) of f holds throughout its domain
of convergence. To determine the radius of convergence, since
(2n + 1)(2n + 1)! 2n + 1
lim = lim = 0,
n→∞ (2n + 3)(2n + 3)! n→∞ (2n + 3)2 (2n + 2)
we apply Problem 2.13 and Theorem 2.8 to conclude that the power series (7.26) converges
for all z ∈ C. In other words, f is entire.
93
Remark 7.1
(a) In Step 3 above, the reason why absolute convergence of the double integral implies
the interchange of order of integration depends on the application of an advanced tool
called Fubini Theorem, see [23, Theorem 8.8, pp. 164, 165].
(b) We also remark that Problem 7.13(a) is actually a special case of the complex version
of the so-called Leibniz’s Rule: Suppose that ϕ(z, t) is defined on Ω × [0, 1], where
Ω is open in C. If ϕ(z, t) is analytic in z for each t and ϕ is continuous on Ω × [0, 1],
then the function f : Ω → C given by
Z 1
f (z) = ϕ(z, t) dt
0
is analytic. This statement can be found in Conway [7, Exercise 2, pp. 73, 74] or Stein
and Shakarchi [24, Theorem 5.4, p. 56]. The tools Stein and Shakarchi applied there
are Riemann sums and Theorem 7.6. The corresponding real version can be found in
[22, Theorem 9.42, pp. 236, 237].
Problem 7.14
Proof.
(a) Clearly, we have

Z 1 Z 1Z z Z z Z 1
sin zt
f (z) = dt = cos ζt dζ dt = cos ζt dt dζ.
0 t 0 0 0 0
We let Z 1
sin ζ
g(ζ) = cos ζt dt =
0 ζ
which is entire. Now by the proof of Theorem 4.15 (The Integral Theorem), we conclude
immediately that Z z Z 1
′ d
f (z) = g(ζ) dζ = g(z) = cos zt dt.
dz 0 0
(b) Applying Theorem 2.9 to the power series (7.26) and then using Problem 3.22, we conclude
that Z 1
X∞
′ (−1)n z 2n sin z
f (z) = = = cos zt dt.
(2n + 1)! z 0
n=0
Problem 7.15
Proof. It is reasonable to assume that z0 6= z1 . For every real x, we have
g(x) = z0 + xeiθ
x|z1 − z0 | · eiArg (z1 −z0 )

= z0 +
|z1 − z0 |
x(z1 − z0 )
= z0 +
|z1 − z0 |
x x
= 1− z0 + z1 . (7.27)
|z1 − z0 | |z1 − z0 |
Therefore, the formula (7.27) implies exactly that g(R) = L. This completes the proof of the
problem.
Problem 7.16
Proof. Let Ω+ = {z ∈ C | Im z > 0} and Ω− = {z ∈ C | Im z < 0}. Since f is analytic in Ω+ ∪ R

and f is real for real z, Theorem 7.8 (The Schwarz Reflection Principle)c implies that f can be
extended to an entire function F such that F (z) = f (z) on Ω+ . Since f is bounded in Ω+ ∪ R,
F is bounded in C. By Theorem 5.10 (Liouville’s Theorem), F and then f is constant which
Problem 7.17
Proof. Since f is entire, it deduces from Theorem 5.5 (The Taylor Expansion of an Entire
Function) that
∞
X f (k) (0) k
f (z) = z (7.28)
k!
k=0
for all z ∈ C. Since C is certainly a region symmetric with respect to R and f is real on R,
Corollary 7.9 shows that
f (z) = f (z) (7.29)
for all z ∈ C. Combining the two expressions (7.28) and (7.29), we obtain
∞
X ∞
X ∞
X
f (k) (0) k f (k)(0) f (k) (0)
z = zk = zk
k! k! k!
k=0 k=0 k=0
for all z ∈ C. By Theorem 6.9 (The Uniqueness Theorem), we have
f (k) (0) f (k) (0)

= (7.30)
k! k!
for all k = 0, 1, 2, . . ..
Next, we consider
∞
X f (k) (0) k
g(z) = if (iz) = ik+1 · z .
k!
k=0
c
Here we use the stronger form of the Schwarz Reflection Principle given in [7, §1, pp. 210 - 213] or [24, p. 60]
because the form of the principle in our book is only applicable to “C-analytic function f in a region D” whose
definition assumes that D is a bounded region.
95
Since f is imaginary on the imaginary axis, g is real on the real axis. Thus the preceding
paragraph shows that
f (k) (0) ik+1 f (k) (0) f (k) (0)
ik+1 = = ik+1 · . (7.31)
k! k! k!
If k is even, then k = 2m so that ik+1 = (−1)m i = −(−1)m i = −i2m+1 = −ik+1 and the
expression (7.31) reduces to
f (k) (0) f (k) (0)
=− . (7.32)
k! k!
Combining the results (7.30) and (7.32), we conclude that
f (k) (0)
=0
k!
for even k. Hence the representation (7.28) can be written as
∞
X f (2m+1) (0) 2m+1
f (z) = z
(2m + 1)!
m=0
which is indeed an odd function. This completes the proof of the problem.
Problem 7.18
Proof. Let z = x + iy, ω = γ(z) = u + iv and ω ∗ = γ(z). Since the line u = v can be
parameterized by γ(t) = t + it, where t ∈ R, we have z + iz = u + iv so that
(x + iy) + i(x + iy) = u + iv

(x − y) + i(x + y) = u + iv.
u+v v−u
Thus we have x = 2 and y = 2 so that
u + v v − u u + v v−u u + v v − u
ω ∗ = γ(z) = γ +i = −i +i −i = v + iu
2 2 2 2 2 2
as required, completing the proof of the problem.
Problem 7.19
Proof. Let D = {z ∈ C | |z| < 1 and Im z > 0} and D ∗ = {z ∈ C | |z| > 1 and Im z > 0}. By
Problem 1.17, we know that the mapping
z−1
ω=i (7.33)
z+1
maps the upper unit circle C + = {z ∈ C | |z| = 1 and Im z > 0} into the negative real axis R− .
In fact, it is easy to show that this mapping is bijective. Therefore, its inverse
i+ω
z= (7.34)
i−ω
maps R− one-one and onto C + .

We claim that the mapping (7.34) maps the region H = {ω ∈ C | Re ω < 0 and Im ω < 0}
onto D.d Let ω = a + ib, where a < 0 and b < 0. Then direct computation gives
a + (b + 1)i 1 − a2 − b2 2a
z= = 2 2
−i 2
−a − (b − 1)i a + (b − 1) a + (b − 1)2
and so
(1 − a2 − b2 )2 + 4a2 1 + 2a2 + a4 − 2b2 + 2a2 b2 + b4
|z|2 = = . (7.35)
[a2 + (b − 1)2 ]2 [a2 + (b − 1)2 ]2
Further simplification of the expression (7.35) implies that |z| < 1 if and only if
b[a2 + (b − 1)2 ] < 0
or equivalently, b < 0. This means that the mapping (7.34) maps H into D. Next, given
z = x + iy ∈ D, where y > 0 and x2 + y 2 < 1. Now the expression (7.33) implies that
x + iy − 1 x − 1 + iy 2y 1 − x2 − y 2
ω=i =i =− − i . (7.36)
x + iy + 1 x + 1 + iy (x + 1)2 + y 2 (x + 1)2 + y 2
Since y > 0 and x2 + y 2 < 1, the expression (7.36) ensures that Re ω < 0 and Im ω < 0.
Consequently, the mapping is onto and our claim follows.
Define F : H → D by
i + ω
F (ω) = f
i−ω
which is obviously analytic in H and is real for ω < 0. Let H ∗ = {ω ∈ C | Re ω < 0 and Im ω > 0},
the left upper half-plane. By Theorem 7.8 (The Schwarz Reflection Principle), F has an analytic
extension G : H ∪ R− ∪ H ∗ → C which is defined by

 F (ω), if ω ∈ H ∪ R− ;
G(ω) = (7.37)
 ∗
F (ω), if ω ∈ H .
By the expression (7.34), we know thate

1 i+ω
= .
z i−ω
Now the analysis of the above claim also shows that if ω ∈ H ∗ , then Im z > 0 and |z| > 1. Hence
the analytic function (7.37) can be expressed as

 f (z), if z ∈ D ∪ C + ;
z − 1  
g(z) = G i =
z+1  1

 f , if z ∈ D ∗ .
z
This concludes the proof of the problem.
Problem 7.20
d
Actually, it is easily seen that H is the left lower half-plane. For example, if ω = −1 − i, then z = − 15 + 52 i
which belongs to D.
e
By the example on [4, pp. 102, 103], z1 is the reflection of the point z across the curve C + .
97
Proof. Assume that f : D(0; 1) → C was a nonconstant analytic function in D(0; 1) and f (z)
is real for all z ∈ C(0; 1). By Definition 3.3, f is continuous on the closed unit disc D(0; 1).
Similar to the proof of Problem 7.19, we can show that the mapping T : C → C defined by
z−i
T (z) =
z+i
maps the sets U = {z ∈ C | Im z > 0} and R onto D(0; 1) and C(0; 1) respectively. Therefore,
the composition F = f ◦ T : U → C is also nonconstant and continuous. Furthermore, F is
analytic in U and F (z) takes real values for real z. Thus Theorem 7.8 (The Schwarz Reflection
Principle)f implies that F can be extended to a nonconstant entire function G defined by

 F (z), if z ∈ U ∪ R;
G(z) = (7.38)

F (z), if z ∈ U ∗,
where U ∗ = {z ∈ C | z ∈ U }. Since D(0; 1) is compact, we know that there exists a positive

constant M such that |F (z)| = |f (T (z))| ≤ M for all z ∈ U ∪ R. Therefore, this fact and the
expression (7.38) imply that
|G(z)| = |f (T (z))| ≤ M
for all z ∈ C. By Theorem 5.10 (Liouville’s Theorem), G is constant and hence f is also constant,
a contradiction. We complete the proof of the problem.
Problem 7.21
Proof. Assume that f (x) = |x| for all x ∈ R. Now, since f is analytic in the bounded region
Ω+ = {z ∈ C | |z| < 1 and Im z > 0} and f is continuous on the upper semi-disc, we observe
from Theorem 7.8 (The Schwarz Reflection Principle) that f can be extended to an analytic
function g in D(0, 1).
Since g is analytic in D(0; 1), we see that
g(h) − g(0) f (h) − f (0) h−0

lim = lim = lim =1
h→0 h h→0 h h→0 h
h>0 h>0 h>0
but
g(h) − g(0) f (h) − f (0) −h − 0
lim = lim = lim = −1.
h→0 h h→0 h h→0 h
h<0 h<0 h<0
Hence these mean that g′ (0) does not exist and this contradiction shows that f (x) = |x| for all
x ∈ R is impossible, completing the analysis of the problem.
Problem 7.22
f
Of course, we also use the stronger version of the principle here.
Proof. Suppose that the two original horizontal lines are Im z = a and Im z = b and they are
mapped by f to the horizontal lines Im z = c and Im z = d respectively, where a, b, c, d ∈ R. In
other words, we have
f (x + ia) = u + ic and f (x + ib) = v + id, (7.39)
where x, u, v ∈ R.
Of course, the two functions
g1 (z) = f (z + ia) − ic and g2 (z) = f (z + ib) − id
are entire and the expressions (7.39) give
g1 (x) = f (x + ia) − ic = u and g2 (x) = f (x + ib) − id = v.
In other words, g1 and g2 are real on the real axis. Now Corollary 7.9 leads to us that
f (z + ia) − ic = g1 (z) = g1 (z) = f (z + ia) + ic (7.40)
and
f (z + ib) − id = g2 (z) = g2 (z) = f (z + ib) + id. (7.41)
Let z = x + iy. If we apply the formula (7.40) to f (z), then we have
f (z) = f (x + iy) = f (x + i(y − a) + ia) = f (x − i(y − a) + ia) + 2ic (7.42)
Next, we apply the formula (7.41) to the expression (7.42) and get
f (z) = f (x − i(y − 2a + b) + ib) + 2ic

= f (x + i(y − 2a + b) + ib) + 2ic − 2id
= f (x + iy + 2i(b − a)) + 2i(c − d)
= f (z + 2i(b − a)) + 2i(c − d). (7.43)
By taking derivative with respect to z to both sides of the expression, we conclude that
f ′ (z + 2i(b − a)) = f ′ (z)
for all z ∈ C, i.e., f ′ is periodic which completes the proof of the problem.
Problem 7.23
Proof. Suppose that P and Q are two parallelograms and f maps P onto Q. Let L1 , L2 , L3
and L4 be the sides of P , where L1 is parallel to L2 and L3 is parallel to L4 . Similarly, we let
ℓ1 , ℓ2 , ℓ3 and ℓ4 be the sides of Q, where ℓ1 is parallel to ℓ2 and ℓ3 is parallel to ℓ4 .
Suppose that f (L1 ) = ℓ1 . Generally speaking, if the lines L and ℓ make the angles θ and φ
with the positive real axis respectively, then L′ = Le−iθ and ℓ′ = ℓe−iφ are two horizontal lines.
Next, it is easy to see that the function
F (z) = e−iφ f (ze−iθ ) (7.44)
is entire and satisfies

F (L′ ) = ℓ′ .
99
In other words, without loss of generality, we may assume that
L1 : Im z = a and ℓ1 : Im ζ = c.
Since L1 and ℓ1 are parallel to L2 and ℓ2 respectively, we also have
L2 : Im z = b and ℓ2 : Im ζ = d.
In other words, F maps L1 : Im z = a and L2 : Im z = b onto ℓ1 : Im ζ = c and ℓ2 : Im ζ = d

respectively. Thus it follows from Problem 7.22 that F ′ is periodic with period ω1 = 2i(b − a).
By the definition (7.44) (with θ = φ = 0), f ′ is periodic with period 2i(b − a).
Since P and Q are parallelograms, L1 is not parallel to L3 and ℓ1 is not parallel to ℓ3 . Let
α(6= 0, π) and β(6= 0, π) be the angles which transform L3 and ℓ3 onto horizontal lines L′3 and
ℓ′3 respectively. In other words, we may assume that
L′3 = L3 e−iα : Im z = A and L4 = L4 e−iα : Im z = B,
where A, B ∈ R. Similarly, we have
ℓ′3 = ℓ3 e−β : Im ζ = C and ℓ′4 = ℓ4 e−β : Im ζ = D,
where C, D ∈ R. Now the function G(z) = e−iβ f (ze−α ) is entire such that G(L′3 ) = ℓ′3 and
G(L′4 ) = ℓ′4 . By Problem 7.22 again, G′ is period with period ω2 = 2i(B − A) and so f ′ is
periodic with period 2i(B − A)e−α . If 2i(b − a) = 2i(B − A)e−α , then we have α = 0 or π which
is impossible. Hence the two periods are linearly independent.
Since f ′ has two linearly independent periods ω1 and ω2 , every z ∈ C can be expressed as
z = xω1 + yω2 for some x, y ∈ R. This implies that
f ′ (z) = f ′ (xω1 + yω2 ) = f ′ ({x}ω1 + {y}ω2 + [x]ω1 + [y]ω2 ) = f ′ ({x}ω1 + {y}ω2 ), (7.45)
where 0 ≤ {x}, {y} < 1. Since every {x}ω1 + {y}ω2 is a point of the (compact) parallelogram
with vertices 0, ω1 , ω2 , ω1 + ω2 , f ′ is bounded on this parallelogram. Combining this fact with
the representation (7.45), we conclude that f ′ is a bounded entire function. Hence we apply
Theorem 5.10 (Liouville’s Theorem) to f ′ to conclude that f ′ is constant. Finally, Problem 5.15
guarantees that f is a linear polynomial. This completes the proof of the problem.
CHAPTER 8
Simply Connected Domains
Problem 8.1
Proof. Let S be a star-like region and α ∈ S. Pick z ∈ C \ S and consider γ(t) = tz + (1 − t)α,
where t ≥ 1. Clearly, γ is a continuous curve, γ(1) = z. Since z − α 6= 0, we must have
γ(t) = α + t(z − α) which gives
lim γ(t) = ∞.
t→∞
Thus z is connected to ∞. We claim that γ([1, ∞]) ⊆ C \ S. Otherwise, there exists a t1 > 1
such that z1 = γ(t1 ) ∈ S. Since S is a star-like region and γ is a line segment, it must be true
that
γ([1, t1 ]) ⊆ S.
However, this implies that z ∈ S, a contradiction. Hence the point z1 does not exist and
our claim follows. By Definition 8.1, S is simply connected and it completes the proof of the
problem.
Remark 8.1
(a) One may use the concept of homotopy of paths from topology to define simply
connectivity and prove Problem 8.1, see [24, p. 96; Exercise 21, p. 107].
(b) Furthermore, we note that Burckel [6, p. 344] provides 15 equivalent definitions of a
simply connected region. In fact, some are defined with respect to its analytic senses
and the others are defined in the topological aspects of the region.
Problem 8.2
Proof. Let S be a nonempty convex region. Pick any point α ∈ S. Since S is convex, the line
segment connecting α and z ∈ S is contained in S for every z ∈ S. By the definition, S is
star-like. Since S is assumed to be a region, Problem 8.1 ensures immediately that S is simply
connected, completing the proof of the problem.
101
Chapter 8. Simply Connected Domains 102
Problem 8.3
Proof. Since S is open in C, for each z ∈ C, there exists a δz > 0 such that D(z; 2δz ) ⊆ S.
Obviously, we have [
C⊆ D(z; δz ).
z∈C
Since C is compact, there are finitely many points z1 , z2 , . . . , zN in C such that

N
[
C⊆ D(zk ; δzk ). (8.1)
k=1
Let δ = min(δz1 , δz2 , . . . , δzN ) > 0. We claim that the annulus
B = {ω ∈ C | r − δ ≤ |ω − α| ≤ r + δ}
is contained in S. To see this, we let ω ′ be the projection of the point ω ∈ B on C. Now the set
relation (8.1) ensures that ω ′ ∈ D(zk ; δzk ) for some 1 ≤ k ≤ N , see Figure 8.1 below.
Figure 8.1: The geometry of the circle C and the annulus B.
Then the triangle inequality implies that
|ω − zk | ≤ |ω − ω ′ | + |ω ′ − zk | < δ + δzk < 2δzk .
In other words, we have established that ω ∈ D(z; 2δzk ) ⊆ S and thus B ⊆ S. We have
completed the proof of the problem.
103
Problem 8.4
Proof. Rewrite Se = T ∪ I, where

n 1 o
T = x + i sin 0<x≤1 and I = {iy | − 1 ≤ y < ∞}.
x
We note that the set T is part of the Topologist’s sine curve studied in Problem 1.23. Since
Se contains all its limit pointsa , it is closed in C and therefore, S is open in C.
1
Given x ∈ (0, 1] and ǫ > 0. Let n ∈ N be large enough such that nπ < x. Since
1
lim sin = sin nπ = 0,
1
t→ nπ t
the point x + i sin x1 ∈ T ⊆ Se is connected to the point 1

nπ ∈ T ⊆ Se by the continuous curve
1
γ1 : [ nπ , x] → T given by
1 1
γ1 (t) = x + − t + i sin 1 .
nπ x + nπ −t
Recall the definition from [22, Exercise 20, p. 20] that
e = inf d(γ1 (t), z).

d(γ1 (t), S)
z∈Se
1 1 1
If we choose n > πǫ , then since d(( nπ , 0), (0, 0)) = nπ < ǫ, we obtain
e < ǫ.
d(γ1 (x), S)
1
Next, if we define γ : [ nπ , ∞) → Se by
 h 1 i

 γ (t), if t ∈ ,x ;

 1
nπ




γ(t) = 1 2−t (8.2)
 · , if t ∈ (x, 2];

 nπ 2 − x





i(t − 2), if t > 2,
1 1
then γ is a continuous curve on [ nπ , ∞). Furthermore, if t ∈ (x, 2], then γ(t) < nπ so that
e < ǫ.
d(γ(t), S)
1 1
Since γ( nπ ) = γ1 ( nπ ) = x + i sin x1 and
lim γ(t) = lim i(t − 2) = ∞,

t→∞ t→∞
we see that our curve (8.2) satisfies the requirements of Definition 8.1 and hence D is simply
connected, completing the proof of the problem.
a 1
Recall that (0, 0) is a limit point of the set {x + i sin x
| 0 < x ≤ 1}.
Problem 8.5
Proof. Let R = {x + iy | a < x < b and c < y < d} be a rectangle containing z, Γ = ∂R and γ
be a continuous curve connecting z to ∞. We further set
E = {t ∈ [0, ∞) | γ(t) ∈ R} and t0 = sup E.
Since γ(0) = z ∈ R, E is not empty so that t0 exists in R. If γ(t0 ) ∈ Γ, then there is nothing
to prove. Therefore, without loss of generality, we may assume that γ(t0 ) ∈
/ Γ in the following
discussion.
Suppose that t0 ∈ E, i.e., γ(t0 ) ∈ R. Since R is open in C, there exists a ǫ > 0 such that
D(γ(t0 ); ǫ) ⊆ R. Since γ is continuous on [0, ∞), there exists a δ > 0 such that |t − t0 | < δ
implies
γ(t) ∈ D(γ(t0 ); ǫ).
Particularly, we have γ(t0 + 2δ ) ∈ R which means that t0 + δ
2 ∈ E but this contradicts the fact
that t0 = sup E.
Now the observation in the previous paragraph forces that t0 ∈ C \ E. The assumption
γ(t0 ) ∈
/ Γ means that {γ(t0 )} ∩ Γ = ∅. Since Γ is compact and {γ(t0 )} is closed in C, it
follows from the well-known fact [22, Exercise 21, p. 101] that one can find a ǫ1 > 0 such that
d(γ(t0 ), Γ) = 2ǫ1 > 0. Consequently, we have
D(γ(t0 ); ǫ1 ) ∩ R = ∅. (8.3)
Again, the continuity of γ implies that there is a δ1 > 0 such that |t − t0 | < 2δ1 must imply
γ(t) ∈ D(γ(t0 ); ǫ1 ).
In particular, γ(t0 − δ1 ) ∈ D(γ(t0 ); ǫ1 ) or equivalently, γ(t0 − δ1 ) ∈

/ R by the result (8.3). Since
t0 − δ1 < t0 , the definition of t0 shows that there is a p ∈ E such that t0 − δ1 < p < t0 . Obviously,
we have |p − t0 | < δ1 < 2δ1 , so γ(p) ∈ D(γ(t0 ); ǫ1 ) and then the set relation (8.3) ensures that
γ(p) ∈/ R. However, p ∈ E certainly implies that γ(p) ∈ R, a contradiction. Hence we conclude
that γ(t0 ) ∈ Γ which completes the proof of the problem.
Remark 8.2
We note that only the continuity of γ is required in the proof of Problem 8.5.
Problem 8.6
Proof. Suppose that Γ is a simple closed polygonal path contained in a simply connected domain
D. Recall from the proof of Lemma 8.3 that if Γ has only two levels, then it is the boundary
of a single rectangle R. Thus it is natural to define the “inside” of Γ to be the points of the
rectangle R. Assume that the “inside” of a simple closed polygonal path Γ ⊆ D can be defined
up to k(≥ 2) levels and the “inside” of Γ is contained in D.
Let Γ′ be a simple closed polygonal path of (k + 1) levels, see Figure 8.2 below. Then we can
write
Γ′ = Γ ∪ ∂R, (8.4)
105
where Γ and ∂R are simple closed polygonal paths and boundaries of rectangles R respectively.
Thus the “inside” of Γ′ is just the union of the “inside” of Γ and the points of R.
Figure 8.2: The “inside” of a simple closed polygonal path.
Therefore, if Γ′ is contained in the simply connected domain D, then so are Γ and ∂R. By
the assumption step, the “inside” of Γ is contained in D. Next, if z ∈ R, then we must have
z ∈ D. Otherwise, we have z ∈ D c . On the one hand, we know that D c is closed in C, ∂R is
compact and D c ∩ (∂R) = ∅, so there exists a δ > 0 such that
d(∂R, D c ) = δ > 0. (8.5)
On the other hand, we recall from Definition 8.1 that z is connected within ǫ to ∞, i.e., there
exists a continuous curve γ : [0, ∞) → ∞ such that d(γ(t), D c ) < δ for all t ∈ [0, ∞). However,
Problem 8.5 means that we can find a t0 ∈ [0, ∞) such that γ(t0 ) ∈ ∂R and
d(γ(t0 ), D c ) < δ,
contradicting the result (8.5). Therefore, R lies in D and we observe from the set relation
(8.4) that the “inside” of Γ′ must lie in D too. By induction, our desired result follows and we
complete the proof of the problem.
Problem 8.7
Proof. By Definition 4.13, let z : [0, 1] → C be a non-simple closed polygonal path. Let p11 , p12
be a pair of points such that 0 < p11 < p12 < 1 and z(s) = z(t) for all s, t ∈ [p11 , p12 ], see Figure
8.3 for an example. In this example, we may define z1 : [0, 1] → C and γ1 : [0, 1] → C by

 z(2p11 t), if t ∈ [0, 12 ];
z1 (t) =

z(p12 + 2(1 − p12 )(t − 12 )), if t ∈ [ 21 , 1]
and
γ1 (t) = z(p11 + (p12 − p11 )t).
Then it is easy to see that both z1 and γ1 are continuous on [0, 1] so that we can write
z = z1 ∪ γ1 ,
where z1 is a simple closed polygonal path and γ1 is a line segment traversed twice in opposite
directions.b
Figure 8.3: A decomposition of a closed polygonal path.
Generally speaking, if there are k pairs of points {p11 , p12 , p21 , p22 , . . . , pk1 , pk2 } such that
z(sj ) = z(tj ) for all sj , tj ∈ [pj1 , pj2 ] and 1 ≤ j ≤ k. Then the above argument can be repeated
to decompose z into
z = (z1 ∪ z2 ∪ · · · ∪ zk ) ∪ (γ1 ∪ γ2 ∪ · · · ∪ γk ),
where each pair of zj and γj composes of simple closed polygonal paths and line segments
traversed twice in opposite directions. Hence we have completed the proof of the problem.
Problem 8.8
Proof. Let D be the whole plane minus the non-negative real axis. Then D is simply connected
and 0 ∈/ D. Since −1 ∈ D and log(−1) = log | − 1| + iArg (−1) = πi, Theorem 8.8 says that the
function Z z
dζ
f (z) = πi + (8.6)
−1 ζ
is an analytic branch of log z in D.
/ R+ , then we have Arg z ∈ (0, 2π). If z lies on the negative axis, then z = −|z|
Next, since z ∈
and Arg z = π so thatc
Z −|z|
dζ −|z|
f (z) = πi + = πi + log |ζ| = log |z| + πi. (8.7)
−1 ζ −1
Thus we have Im (log z) = Im (f (z)) = Arg z = π. If z ∈ D \ R− , then the line integral (8.6)
goes from −1 to −|z| and then from −|z| to z to get
Z −|z| Z Z
dζ dζ dζ
f (z) = πi + + = log |z| + πi + , (8.8)
−1 ζ C ζ C ζ
b
We remark that there are examples consisting of several simple closed polygonal paths (e.g. two rectangles
intersecting at one of their vertices only) or line segments (see the horizontal and vertical line segments at
z(pk1 ) = z(pk2 ) in Figure 8.3).
c
The integral in the expression (8.7) is the usual Riemann integral.
107
where C is the circular arc from connecting −|z| and z clockwise, see Figure 8.4. If C is
Figure 8.4: The circular arc C connecting −|z| and z clockwise.
parameterized as γ(t) = |z|ei(π−t) , where 0 ≤ t ≤ π − Arg z, then we deduce from the expression
(8.8) that
Z π−Arg z
1
f (z) = log |z| + πi + i(π−t)
· [−i|z|ei(π−t) ] dt
0 |z|e
Z π−Arg z
= log |z| + πi − i dt
0
= log |z| + πi − i(π − Arg z)
= log |z| + iArg z
which implies that Im (log z) = Arg z ∈ (0, 2π). This completes the proof of the problem.
Problem 8.9
Proof. Let D be the plane minus the non-positive real axis. By the analysis on [4, p. 115], we
can define an analytic branch of log z in D by
log z = log |z| + iArg z, (8.9)
where Arg z ∈ (−π, π). Next, we define
f (z) = ez log z . (8.10)
Since log z is analytic in D, f is also analytic in D. Besides, if x > 0, then f (x) = ex log x = xx .
In other words, the function f defined by the expression (8.10) satisfies our requirements.
Using the expressions (8.9) and (8.10) directly, we see that
π
f (i) = ei log i = ei(log |i|+iArg (i)) = e− 2
and
π
f (−i) = e−i log −i = e−i(log |−i|+iArg (−i)) = e− 2 .
Finally, let z = x + iy ∈ D. If y = 0, then we have f (x) = f (x), so we assume that y 6= 0 so
that z = x − iy ∈ D. In addition, we have
log z = log |z| + iArg z = log |z| − iArg z = log |z| + iArg z = log z
which gives
f (z) = exp(z · log z) = exp(z log z) = exp(z log z) = f (z).
CHAPTER 9
Isolated Singularities of an Analytic Function
Problem 9.1
Proof. Assume that z0 was a removable singularity of f . By Definition 9.2, there exists an
analytic function g defined in a neighborhood of z0 such that f (z) = g(z) for all z 6= z0 . Now
the continuity of g yields
lim |f (z)| = lim |g(z)| = |g(z0 )| < ∞
z→z0 z→z0
which contradicts our hypothesis. Consequently, z0 is not a removable singularity of f .

Next, we assume that z0 was an essential singularity of f . Since f (z) → ∞ as z → z0 , for
every ǫ > 0, there exists a M > 0 such that |f (z)| > M for all z ∈ D(z0 ; ǫ). Therefore, it
implies that f (D(z0 ; ǫ)) ∩ D(0; M ) = ∅. In other words, f (D(z0 ; ǫ)) is not dense in C which
contradicts Theorem 9.6 (The Casorati-Weierstrass Theorem). Hence z0 cannot be an essential
singularity of f and Definition 9.2 forces that f has a pole at z0 . This completes the analysis of
the problem.
Problem 9.2
Proof. The answer is negative. By the hypothesis, we get

1
lim |f (z)| = lim e |z| = ∞,
z→0 z→0
so Problem 9.1 implies that f has a pole at 0 and then f has the form
∞
X
f (z) = an z n (9.1)
n=−k
for some k ∈ N. Therefore, the form (9.1) implies that

M
|f (z)| ∼
|z|k
as z → 0 for some positive constant M . This is certainly a contradiction and we complete the
109
Chapter 9. Isolated Singularities of an Analytic Function 110
Problem 9.3
Proof. Suppose that f is entire and injective. Then f is nonconstant. Consider the function
g : C \ {0} → C defined by
g(z) = f (z −1 )
which has a singularity at 0. Since f is entire, Theorem 5.5 (The Taylor Expansion of an Entire
Function) says that
∞
X
f (ω) = ak ω k
k=0
for some ak ∈ C and then

0
X
g(z) = ak z k .
k=−∞
If g has a removable singularity at 0, then a−1 = a−2 = · · · = 0 which implies that g(z) = a0 .
Consequently, f is also a constant which contradicts to our hypothesis. Therefore, g has either
a pole or an essential singularity at 0.
Assume that the singularity was essential. By Theorem 9.6 (The Casorati-Weierstrass The-
orem), the set g(D(0; 1) \ {0}) is dense in C. Since g is analytic in the region D(2; 12 ), Problem
7.2 implies that g(D(2, 12 )) is also a region. In particular, g(D(2, 21 )) is open in C. Thus we have
1
g(D(0; 1) \ {0}) ∩ g D 2; 6= ∅.
2
In other words, there exists z1 ∈ D(0; 1) \ {0} and z2 ∈ D(2; 12 ) such that z1 6= z2 but
f (z1−1 ) = g(z1 ) = g(z2 ) = f (z2−1 ).
This contradicts the fact that f is injective and hence the singularity is not essential.
The previous analysis ensures that the singularity of g at 0 is a pole, so f must be a polyno-
mial. Suppose that f (z) = a0 + a1 z + · · · + an z n . By Theorem 5.12 (The Fundamental Theorem
of Algebra), f has n roots counting multiplicity. Since f is injective, it cannot have any two
distinct roots. In other words, f must have the form
f (z) = a(z − z0 )n
for some a, z0 ∈ C and a 6= 0. If n ≥ 2, then the fact

2πi
f (z0 + 1) = a = f (z0 + e n )
shows that f is not injective, a contradiction. Now we have n = 1 and so f (z) = a(z − z0 ) which
is a linear polynomial as desired. We end the proof of the problem.
Problem 9.4
111
1
Proof. If f is analytic at the origin, then f is entire. For |z| > 1, we have |z| < |z| so that
p
|f (z)| < 2 |z| < 2|z|
there. Next, since |f (z)| is continuous on the compact set D(0; 1), |f (z)| is bounded by a positive
constant M there. These two facts imply that one can find a positive constant M ′ such that
|f (z)| ≤ M ′ |z|
holds for all z ∈ C. Hence Theorem 5.11 (The Extended Liouville Theorem) guarantees that
f (z) = A + Bz for some A, B ∈ C. Now the triangle inequality gives
p
2 |z| > |f (z)| = |Bz + A| ≥ |B| · |z| − |A|
for large z and this forces that B = 0. In other words, f (z) = A in this case.
Suppose that f has an isolated singularity at 0. Then we consider its Laurent expansion at
0:
X∞
f (z) = ak z k .
k=−∞
Let R > 0. Now if k 6= 0, then we consider the circle C = C(0; R) and we apply Theorem 4.10
(The M -L Formula) to obtain
Z
1 f (ω) 1 1 √ 1 1 √ 1
|ak | = dω ≤ × 2πR × R + √ = R + √ . (9.2)
2πi C(0;R) ω k+1 2π Rk+1 R Rk R
If k > 0, then the inequality (9.2) shows that
1 √ 1
lim |ak | ≤ lim R + √ =0
R→∞ R→∞ Rk R
which means a1 = a2 = · · · = 0. If k < 0, then the inequality (9.2) gives
1 √ 1
lim |ak | ≤ lim R + √ = 0.
R→0 R→0 Rk R
In other words, a−1 = a−2 = · · · = 0. Hence we have shown that f (z) = a0 in this case,
Problem 9.5
Proof. If g ≡ 0, then there is nothing to prove. Thus all the zeros of g are isolated. Furthermore,
we have | fg(z)
(z)
| ≤ 1 in a deleted neighborhood of a zero of g, so Corollary 9.4 asserts that all the
singularities of fg are removable. As a result, fg can be extended to an entire function. By the
continuity of the extended function, it is also true that
f (z)
≤1
g(z)
f
for all z ∈ C. Now Theorem 5.10 (Liouville’s Theorem) leads to us that g is a constant which
Problem 9.6
1
Proof. It is easy to see that e z 6= 0 for all z ∈ C. Then 0 is a missing value. Let ω ∈ C \ {0}.
Set z1 = log ω = log |ω| + i(Arg ω + 2πk), where Arg ω ∈ (−π, π] and k ∈ Z. Certainly, we have
1
e z = elog ω = ω.
In addition, it is evident that

1
z=
log |ω| + i(Arg ω + 2πk)
1 log |ω| − i(Arg ω + 2πk)
= ·
log |ω| + i(Arg ω + 2πk) log |ω| − i(Arg ω + 2πk)
log |ω| − i(Arg ω + 2πk)
=
(log |ω|)2 + (Arg ω + 2πk)2
which gives
1
|z| = p . (9.3)
(log |ω|)2 + (Arg ω + 2πk)2
Hence the modulus (9.3) will ensure that 0 < |z| < 1 for all large enough k. This completes the
Problem 9.7
Proof. For simplicity, we assume that z0 = 0 in the following discussion. Suppose that we have
∞
X ∞
X
k
f (z) = ak z and g(z) = bp z p , (9.4)
k=−m p=−n
where ak , bp ∈ C for all k = −m, −m + 1, . . . and p = −n, −n + 1, . . .. If m 6= n, then it is easy

to see from the forms (9.4) that f + g has a pole of order max{m, n} at 0. If m = n, then the
order of the pole at 0 depends on the coefficients ak and bp . For example, if an = −bn , then the
order of pole of f + g at 0 is less than n. Anyhow the order of pole of f + g at 0 is less than or
equal to n.
By Definition 9.2, let’s write
A(z) P (z)
f (z) = and g(z) = , (9.5)
z m B(z) z n Q(z)
where A, B, P and Q are analytic at 0, A(0), B(0), P (0), Q(0) 6= 0. It follows from the forms
(9.5) that
A(z)P (z)
f (z) · g(z) = m+n ,
z B(z)Q(z)
where A(0)P (0) 6= 0 and B(0)Q(0) 6= 0. Hence f · g has a pole of order m + n at 0. Furthermore,
the form (9.5) also implies that
f (z) A(z) z n Q(z) A(z)Q(z)
= m × = m−n (9.6)
g(z) z B(z) P (z) z B(z)P (z)
113
which shows that fg has a zero of order n − m at 0 if n > m and similarly, it has a pole of order
m − n at 0 if m > n. Finally, if m = n, then the representation (9.6) reduces to
f (z) A(z)Q(z)
=
g(z) B(z)P (z)
so that
f (z) A(z)Q(z)
lim z · = lim z · = 0.
z→0 g(z) z→0 B(z)P (z)
f
By Theorem 9.3 (The Riemann’s Principle of Removable Singularities), g has a removable
singularity at 0. This completes the proof of the problem.
Problem 9.8
Proof. Suppose that z0 is an essential singularity of f . By Theorem 9.6 (The Casorati-Weierstrass

Theorem), f (D(z0 ; n1 ) \ {z0 }) is dense in C for every n ∈ N. In other words, it is true that
1 1
f D z0 ; \ {z0 } ∩ D 0; 6= ∅ (9.7)
n n
for every n ∈ N. Equivalently, the set relation (9.7) asserts that there is an αn ∈ D(z0 ; n1 ) \ {z0 }
such that f (αn ) ∈ D(0; n1 ) for each n ∈ N. Hence we obtain the result that both αn → z0 and
f (αn ) → 0 hold. Similarly, we also have
1 1
f D z0 ; \ {z0 } ∩ D n; 6= ∅
n n
for every n ∈ N. Consequently, one can find a βn ∈ D(z0 ; n1 )\{z0 } such that f (βn ) ∈ D(n, n1 ) for
each n ∈ N. Since f (βn ) > n − n1 holds for all n ∈ N, we conclude that βn → z0 and f (βn ) → ∞
as n → ∞.
Conversely, we suppose that there exist sequences {αn } and {βn } tending to z0 such that
f (αn ) → 0 and f (βn ) → ∞. If z0 was a removable singularity of f , then lim f (z) exists by
z→z0
Definition 9.2 or by its Laurent expansion about z0 . However, this means that
lim f (αn ) = lim f (βn )

n→∞ n→∞
which is impossible. If it was a pole of f , then Definition 9.2 implies that
lim f (z) = ∞
z→z0
which contradicts to our hypothesis. Hence they assert that z0 is an essential singularity of f ,
Problem 9.9
Proof.
(a) We know that

1 1 1
= 2 2 = 2 . (9.8)
z4 +z 2 z (z + 1) z (z + i)(z − i)
By Definition 9.2, it has a pole of order 2 at 0 and poles of order 1 at ±i.
(b) We note that cot z = cos z
sin z and sin z is zero if and only if z = nπ, where n ∈ Z. Thus cot z
is analytic in a deleted neighborhood of nπ. By Lemma 7.2, we establish
cos z cos(ζ + nπ) cos ζ
lim (z − nπ) · = lim ζ · = lim ζ · = 1. (9.9)
z→nπ sin z ζ→0 sin(ζ + nπ) ζ→0 sin ζ
Furthermore, we have
cos z cos ζ
lim (z − nπ)2 · = lim ζ × ζ · = 0.
z→nπ sin z ζ→0 sin ζ
Therefore, it follows from Theorem 9.5 that cot z has a pole of order 1 at nπ, where n ∈ Z.
(c) By similar analysis as part (b), csc z has a pole of order 1 at nπ, where n ∈ Z.
(d) The function
exp( z12 )
f (z) =
z−1
has singularities at 0 and 1. It is easily seen that f is analytic in a deleted neighborhood
of 1. Since
1
lim (z − 1)f (z) = lim exp 2 = e and lim (z − 1)2 f (z) = 0, (9.10)
z→1 z→1 z z→1
Theorem 9.5 ensures that f has a simple pole at 1.

For the singularities at 0, we consider the Laurent expansion of f at 0. In fact, we know
that 1 1 1 1
exp 2 = 1 + 2 + 4 + · · · and = −1 − z − z 2 − · · · .
z z z z−1
Thus we have
1 1
f (z) = −(1 + z + z 2 + · · · ) × 1 + 2 + 4 + · · · . (9.11)
z z
Since every term in the brackets of the expression (9.11) has positive coefficient, the re-
sulting expansion of f will have infinitely many nonzero terms in its principal part. By
the notes on [4, p. 125], we conclude that f has an essential singularity at 0.
Hence we end the proof of the problem.
Problem 9.10
Proof. We note that z 2 + 1 = (z − i)(z + i) so that

1 1 1
f (z) = = · . (9.12)
(z 2 + 1)2 (z − i) (z + i)2
2
1
Let F (z) = (z+i)2 . Then F is clearly analytic in a neighbourhood of z = i. By Theorem 6.5
(The Power Series Representation for Functions Analytic in a Disc), we know that
∞
X F (k) (i)
F (z) = (z − i)k . (9.13)
k!
k=0
115
Combining the two expressions (9.12) and (9.13), we get

∞
X ∞
F (k) (i) k−2 F (i) F ′ (i) X F (k+2) (i)
f (z) = (z − i) = + + (z − i)k .
k! (z − i)2 z−i (k + 2)!
k=0 k=0
1
Since F (i) = (2i)2
= − 14 and F ′ (i) = − 4i , the principal part of the Laurent expansion of f is
given by
1 i
2
− − .
4(z − i) 4(z − i)
Problem 9.11
Proof.
(a) We have
∞ ∞
1 1 1 2 4 1 X k 2k
X
= 2 = 2 · (1 − z + z − · · · ) = 2 (−1) z = (−1)k+1 z 2k .
z4 + z2 z (1 + z 2 ) z z
k=0 k=−1
(b) Suppose that

∞
exp( z12 ) X
= ak z k .
z−1
k=−∞
Since
1 1 1 1
exp 2 = 1 + + + ··· and = −(1 + z + z 2 + · · · ),
z 1!z 2 2!z 4 z−1
we have
1 1 ∞
X
2
− 1+ + + · · · (1 + z + z + · · · ) = ak z k .
1!z 2 2!z 4
k=−∞
For k ≥ 0, we obtain
1 1
ak = − 1 + + + · · · = −e.
1! 2!
Let n ∈ N. If k = −2n, then we get
h1 1 i 1 1
a−2n = − + + · · · = −e + 1 + + · · · + .
n! (n + 1)! 2! (n − 1)!
Similarly, if k = −2n + 1, then we conclude
h1 1 i 1 1
a−2n+1 = − + + · · · = −e + 1 + + · · · + .
n! (n + 1)! 2! (n − 1)!
(c) We note that

1 1 1 1
= = ·
z2 −4 (z − 2)(z + 2) z − 2 4(1 + z−2
4 )
which implies that
1 1
∞
X k ∞
X (−1)k+1
k z−2
2
= (−1) = (z − 2)k .
z −4 4(z − 2) 4 4k+2
k=0 k=−1
We have completed the analysis of the problem.

Problem 9.12
Proof. We know that

1 1 1
f (z) = − + .
z 2−z 1−z
If |z| < 1, then we have
1
= 1 + z + z2 + · · · . (9.14)
1−z
1
If |z| > 1, then |z| < 1 and thus
1 1 1 1 1 1 1
=− 1 = − 1 + + 2
+ · · · = − − 2 − ··· . (9.15)
1−z z(1 − z ) z z z z z
Combining the formulas (9.14) and (9.15), we see that

 ∞
 X

 zk , if |z| < 1;



 k=0
1
= (9.16)
1−z  −1
X


 z k , if |z| > 1.
 −

k=−∞
Similarly, we have

1 X z k
∞



 , if |z| < 2;

1 1 1  2 k=0 2

= × z = (9.17)
2−z 2 1− 
 1 X z k
2 −1



 −2

2
, if |z| > 2.
k=−∞
(a) Now the formulas (9.16) and (9.17) imply that
1 h 1 X z k X k i 1 X 1
∞
1
∞ ∞ ∞ X
f (z) = − + z = 1 − k+1 z k = 1 − k+2 z k .
z 2 2 z 2 2
k=0 k=0 k=0 k=−1
(b) In this case, the formulas (9.16) and (9.17) give
1 h 1 X z k i
∞ −1
X −2
X ∞
X 1 k
f (z) = − − zk = − zk − z .
z 2 2 2k+2
k=0 k=−∞ k=−∞ k=−1
(c) Finally, we have
1 h 1 X z k i 1 X
−1 −1
X −1 −2
X
k 1 k 1
f (z) = − z = − 1 z = − 1 zk .
z 2 2 z 2k+1 2k+2
k=−∞ k=−∞ k=−∞ k=−∞

117
Problem 9.13
Proof. For n = 1, 2, . . . , k, we notice that
z an − 1 = (z − 1)Pn (z),
where Pn (z) = z an −1 + z an −2 + · · · + z + 1. Thus we have
1
R(z) = .
(z − 1)k P1 (z)P2 (z) · · · Pk (z)
Now the product P (z) = P1 (z)P2 (z) · · · Pk (z) is a polynomial in z, so Theorem 5.12 (The
Fundamental Theorem of Algebra) ensures that there exists a δ > 0 such that C(1; δ) does not
contain any zero of P (z). Thus we observe easily from Corollary 9.10 that if C = C(1; δ), then
Z Z 1
1 R(z) 1 P (z)
c−k = −k+1
dz = dz. (9.18)
2πi C (z − 1) 2πi C z−1
Since P 1(z) is analytic inside and on C, we may apply Theorem 6.4 (The Cauchy Integral Formula)
to the last integral (9.18) to obtain
1 1
c−k = = .
P (1) ak ak−1 · · · a1
Problem 9.14
Proof. Suppose that

∞
X
f (z) = ak z k . (9.19)
k=−∞
Then we have
∞
X
f (−z) = (−1)k ak z k . (9.20)
k=−∞
f (z)−f (−z)
Since f is odd, we have f (z) = 2 . Thus it follows from the two representations (9.19)
and (9.20) that
∞
X ∞
X ∞
X
ak + (−1)k+1 ak k
ak z k = z = a2k+1 z 2k+1 .
2
k=−∞ k=−∞ k=−∞
Consequently, it means that a2k = 0 for all k ∈ Z. This completes the proof of the problem.
Problem 9.15
Proof.
(a) We observe from the expression (9.8) that
1 A1 A2 B C A1 z(z 2 + 1) + A2 (z 2 + 1) + Bz 2 (z + i) + Cz 2 (z − i)
= + + + =
z4 + z2 z z2 z − i z + i z 2 (z − i)(z + i)
which means that
A1 z(z 2 + 1) + A2 (z 2 + 1) + Bz 2 (z + i) + Cz 2 (z − i) = 1
1 1
for all z ∈ C. If z = 0, then A2 = 1. If z = i, then B = − 2i . If z = −i, then C = 2i .
Finally, if z = 1, then A1 = 0. Consequently, we conclude that
1 1 1 1
= 2− + .
z4 +z 2 z 2i(z − i) 2i(z + i)
(b) Similar as part (a), we have
1 1 1
= − .
z2 +1 2i(z − i) 2i(z + i)
Problem 9.16
Proof. We follow the given hint. Assume that there were ω ∈ C and δ > 0 such that |f (z)−ω| > δ
for all z ∈ D. Then we have
1 1
<
f (z) − ω δ
throughout D. By Corollary 9.4, the function
1
g(z) =
f (z) − ω
has a removable singularity at z0 . Therefore, g is analytic at z0 . Now g is analytic in the region

D ∪ {z0 } and g(zn ) = 0 for all n ∈ N, where zn → z0 . It follows from Theorem 6.9 (The
Uniqueness Theorem) that g ≡ 0 in D ∪ {z0 }, i.e., f (z) = ∞ throughout D which is impossible.
Problem 9.17
Proof. Suppose that D0 (0; 1) = D(0; 1) \ {0} and g(z) = sin 1z .
(a) We know that

1 1 1 1
sin = − 3
+ − ··· , (9.21)
z z 3!z 5!z 5
119
so it has an essential singularity at 0 and Theorem 9.6 (The Casorati-Weierstrass Theorem)

ensures that g(D0 (0; 1)) is dense in C. In other words, given ω ∈ C, there exists a sequence
{zn } ⊆ D0 (0; 1) such that g(zn ) → ω1 as n → ∞. Since f = g1 , we conclude that
1
lim f (zn ) = lim =ω
n→∞ n→∞ g(zn )
which means that f (D0 (0; 1)) is also dense in C.

1
(b) Let zn = nπ , where n ∈ Z. Clearly, we follow from Lemma 7.2 that
1 z − zn
lim (z − zn ) csc = lim
z→zn z z→zn sin 1z
1 ζ − nπ
=− · lim
nπ ζ→nπ sin ζ
1 ω
=− · lim n
nπ ω→0 (−1) sin ω
(−1)n+1
=−
nπ
6= 0
and
1 z − zn
lim (z − zn )2 csc = lim (z − zn ) · = 0.
z→zn z z→z n sin 1z
It follows from Theorem 9.5 that csc z1 has a simple pole at every zn . Since zn → 0 as
n → ±∞. By Problem 9.16, we see that f (D0 (0; 1)) is dense in C.
Problem 9.18
Proof. If f is a polynomial, then Theorem 5.12 (The Fundamental Theorem of Algebra) ensures
that f takes every value in the plane C. Suppose that f is not a polynomial and we consider
g(z) = f ( 1z ). Similar to the proof of Problem 9.3, it can be shown that g has either a pole or an
essential singularity at 0.
If 0 is a pole of g, then we have
∞
X
g(z) = an z n
n=−k
for some k ∈ N so that

k
X
f (z) = an z n . (9.22)
n=−∞
Since f is analytic at 0, the principal part of the Laurent expansion (9.22) disappears, i.e.,
a−1 = a−2 = · · · = 0. In other words, f reduces to a polynomial, a contradiction. Therefore,
0 is an essential singularity of g and Theorem 9.6 (The Casorati-Weierstrass Theorem) implies
that g(C \ {0}) is dense in C. Since g(C \ {0}) ⊆ f (C), the set f (C) is also dense in C and we
complete the analysis of the problem.
CHAPTER 10
The Residue Theorem
Problem 10.1
Proof.
1
(a) By Problem 9.9(a), z 4 +z 2 has a pole of order 2 at 0 and poles of order 1 at ±i. Furthermore,
we apply [4, Eqn. (1), p. 129] to the expression (9.8) to get
1 1 1 1 i
Res 4 ; i = lim(z − i) × 4 = lim =− =
z + z2 z→i z + z 2 z→i z 2 (z + i) 2i 2
and
1 1 1 1 i
Res 4 2
; −i = lim (z + i) × 4 2
= lim 2 = =− .
z +z z→−i z +z z→−i z (z − i) 2i 2
Since we know
1 1 1 1 1
= 2× = 2 (1 − z 2 + z 4 − · · · ) = 2 − 1 + z 2 − · · · ,
z4 +z 2 z 1+z 2 z z
Definition 10.1 implies that
1
Res ; 0 = 0.
z4 + z2
(b) Recall from Problem 9.9(b) that cot z has a simple pole at nπ for every n ∈ Z. Applying
[4, Eqn. (1), p. 129] to the limit (9.9), we conclude that
Res (cot z; nπ) = 1
for all n ∈ Z.
(c) By Problem 9.9(c), csc z has a simple pole at nπ for every n ∈ Z. Then we have
z − nπ 1 1
lim (z − nπ) csc z = lim = lim sin z−sin nπ
= = (−1)n .
z→nπ z→nπ sin z − sin nπ z→nπ z−nπ
cos nπ
1
exp( )
(d) By Problem 9.9(d), the function z−1z2 has a simple pole at 1 and an essential singularity
at 0. By the limit (9.10), we conclude at once that
exp( 1 )
z2
Res ; 1 = e.
z−1
121
Chapter 10. The Residue Theorem 122
Next, we get from Problem 9.11(b) (with n = 1) that

exp( 1 )
z2
Res ; 0 = −e + 1.
z−1
(e) We note that

1 1
= ,
z2 + 3z + 2 (z + 1)(z + 2)
so it has simple poles at −1 and −2. According to [4, Eqn. (1), p. 129], we see that
1 1 1
Res 2 ; −1 = lim (z + 1) × 2 = lim =1
z + 3z + 2 z→−1 z + 3z + 2 z→−1 z + 2
and
1 1 1
Res ; −2 = lim (z + 2) × 2 = lim = −1.
z 2 + 3z + 2 z→−2 z + 3z + 2 z→−2 z + 1
(f) We know from the condition of Problem 9.17(a) that sin 1z has an essential singularity at
0. Thus we follow from the expression (9.21) that
1
Res sin ; 0 = 1.
z
3
(g) It is clear that the function ze z has an essential singularity at 0. By [4, Example (iv), p.
124], we have
3 1 9
Res (ze z ; 0) = 1 2 = 2.
2 · (3)
(h) By the quadratic formula, the function

1
az 2 + bz + c
√
2
b −4ac
has simple poles at z± = −b± 2a b
if b2 − 4ac 6= 0 and a pole of order 2 at z0 = − 2a if
2
b − 4ac = 0. Therefore, we have
1 1
Res 2
; z+ = lim (z − z+ )
az + bz + c z→z+ a(z − z+ )(z − z− )
1
= lim
z→z+ a(z − z− )
1
=
a(z+ − z− )
1
=√
2
b − 4ac
and
1 1 1
Res 2
; z− = lim (z − z− ) = −√ .
az + bz + c z→z− a(z − z+ )(z − z− ) b2 − 4ac
Finally, we get
1 dh 1 i d 1
2
Res ; z0 = (z − z0 ) × = = 0.
az 2 + bz + c dz a(z − z0 )2 z=z0 dz a z=z0

123
Problem 10.2
Proof.
(a) Obviously, cot z is analytic in the simply connected domain D(0; 1 + δ) except 0 for some
small δ > 0 and C(0; 1) ⊆ D(0; 1 + δ) is a closed curve such that 0 lies inside C(0; 1), it
follows from Theorem 10.5 (The Cauchy’s Residue Theorem) and Problem 10.1(b) that
Z
cot z dz = 2πiRes (cot z; 0) = 2πi.
|z|=1
2πki
1
(b) The function f (z) = (z−4)(z 3 −1) has simple poles at 4 and zk = e 3 , where k = 1, 2, 3.
Now f is analytic in the simply connected domain D(0; 2 + δ) except z1 , z2 , z3 for some
small δ > 0 and C(0; 2) ⊆ D(0; 2 + δ) is a closed curve such that zk lie inside C(0; 2) and
4 lies outside C(0; 2). If B(z) = (z − 4)(z 3 − 1), then B ′ (z) = (z 3 − 1) + 3z 2 (z − 4). Thus
we obtain from [4, Eqn. (1), p. 129] that
1 1 1
Res (f (z); 4) = = and Res (f (z); zk ) = ,
43 −1 63 3zk2 (zk − 4)
1
where k = 1, 2, 3. Since zk3 = 1, we have zk2 = zk so that
zk
Res (f (z); zk ) = ,
3(zk − 4)
where k = 1, 2, 3. Next, recall that z1 + z2 + z3 = z1 z2 + z1 z3 + z2 z3 = 0 and z1 z2 z3 = 1,

so direct computation gives
z1 z2 z3
+ +
z1 − 4 z2 − 4 z3 − 4
z1 (z2 − 4)(z3 − 4) + z2 (z1 − 4)(z3 − 4) + z3 (z1 − 4)(z2 − 4)
=
(z1 − 4)(z2 − 4)(z3 − 4)
3z1 z2 z3 − 8(z1 z2 + z1 z3 + z2 z3 ) + 16(z1 + z2 + z3 )
=
z1 z2 z3 − 4(z1 z2 + z1 z3 + z2 z3 ) + 16(z1 + z2 + z3 ) − 64
1
=− .
21
Hence Theorem 10.5 (The Cauchy’s Residue Theorem) leads to us that
Z
dz 1 z1 z2 z3
3
= 2πi 0 × + 1 × + 1 × + 1 ×
|z|=2 (z − 4)(z − 1) 63 3(z1 − 4) 3(z2 − 4) 3(z3 − 4)
2πi
=− .
63
(c) Now sin z1 has an essential singularity at 0 and it is analytic in the simply connected domain
D(0; 1 + δ) except 0 for some small δ > 0 and C(0; 1) ⊆ D(0; 1 + δ) is a closed curve such
that 0 lies inside C(0; 1). Combining Problem 10.1(f) and Theorem 10.5 (The Cauchy’s
Residue Theorem), we establish
Z
1 1
sin dz = 2πiRes sin ; 0 = 2πi.
|z|=1 z z
(d) By using Problem 10.1(g) and Theorem 10.5 (The Cauchy’s Residue Theorem) directly,
we see that Z
3 3
ze z dz = 2πiRes (ze z ; 0) = 9πi.
|z|=2
Problem 10.3
1
Proof. It is clear that the function f (z) = [1−exp(−z)] n is analytic everywhere except zk = 2πki,
where k ∈ Z. If C is a regular closed curve surrounding only the singularity z0 = 0, then it
deduces from Theorem 10.5 (The Cauchy’s Residue Theorem) that
Z
1 1 dz
Res −z n
; 0 = . (10.1)
(1 − e ) 2πi C (1 − e−z )n
By the substitution ω = ω(z) = 1 − e−z , the integral in the formula (10.1) becomes
Z Z
dz dω
−z )n
= n (1 − ω)
, (10.2)
C (1 − e C ′ ω
where C ′ = ω(C).
1
The function ωn (1−ω) has (simple) poles at 0 and 1. Now we need to construct a suitable C
′
so that the curve C surrounds only the pole 0 but not 1. To this end, we consider the rectangle
R = {z = x + iy | −1 ≤ x ≤ 1 and − π2 ≤ y ≤ π2 }.
Now its boundary C = ∂R is a regular closed curve. Obviously, we have
ω = ω(z) = 1 − e−x−iy = (1 − e−x cos y) + ie−x sin y
which implies that

|ω(z) − 1| = e−x ,
where −1 ≤ x ≤ 1, so we have actually
1
≤ |ω(z) − 1| ≤ e
e
for all z ∈ R. Furthermore, Re ω(z) = 1 − e−x cos y implies that Re ω(z) > 1 if and only if
cos y < 0 if and only if π2 < y < 3π
2 . In other words, we have Re ω(z) ≤ 1 for all z ∈ R which
assures us that ω(R) is a subset of the left half of the annulus A centred at 1 with inner and
outer radii e−1 and e respectively. Finally, it is easy to see from the definition that the curves
n π πo
−1 −1
γ1 = ω(1 + iy) = (1 − e cos y) + ie sin y − ≤ y ≤
2 2
and n π πo
γ2 = ω(−1 + iy) = (1 − e cos y) + ie sin y − ≤ y ≤
2 2
are exactly the inner arc and the outer arc of A respectively. Similarly, the curves
n π o
γ3 = ω x + i = 1 + ie−x − 1 ≤ x ≤ 1
2
125
and n π o
γ4 = ω x − i = 1 − ie−x − 1 ≤ x ≤ 1
2
are the remaining boundary of the annulus A. Consequently, we obtain the desired result that
the regular closed curve C ′ = ω(C) surrounds only the origin.
Hence it deduces from Theorem 10.5 (The Cauchy’s Residue Theorem) that
Z
dω 1
n
= Res ; 0 = 1. (10.3)
C ′ ω (1 − ω) ω n (1 − ω)
Now our desired result follows immediately from the comparison of the expressions (10.1), (10.2)
and (10.3), completing the proof of the problem.
Problem 10.4
Proof. We know that f (z) = (z + z1 )2m+1 has a pole of order 2m+1 at 0 for every m = 0, 1, 2, . . ..
By Theorem 10.5 (The Cauchy’s Residue Theorem), we establish
Z 1 2m+1 1 2m+1
z+ dz = 2πiRes z + ;0 .
|z|=1 z z
By the binomial theorem, we have
1 2m+1
2m+1
X 2m+1
X
2m+1 2m+1−k −k
z+ = Ck z z = Ck2m+1 z 2m+1−2k .
z
k=0 k=0
2m+1
If k = m + 1, then Cm+1 is the coefficient of z −1 so that
Z 1 2m+1 2m+1 2m+1
z+ dz = 2πiCm+1 = 2πiCm ,
|z|=1 z
Problem 10.5
Proof. Suppose that f (ω) is analytic in the region D, D contains the regular closed curve C and
f (ω) p(ω) − p(z)
F (ω) = · ,
p(ω) ω−z
where ω ∈ D\{ω1 , ω2 , . . . , ωn } and z ∈ C. Intuitively, one may apply Corollary 10.6 and Theorem
10.11 (The Generalized Cauchy Integral Formula) directly but we can’t do that because they
require the domain of the analyticity of f should be simply connected. Thus we have to search
another versions of the results that fit our case. In fact, such variations can be found in [3,
Theorem 3.8.6, p. 213; Theorem 5.1.2, p. 294].
Since C encloses the distinct points ω1 , ω2 , . . . , ωn , it follows from [3, Theorem 5.1.2, p. 294]
that
Z X n Xn f (ω) p(ω) − p(z)
F (ω) dω = 2πi Res (F (ω); ωk ) = 2πi Res · ; ωk . (10.4)
C p(ω) ω−z
k=1 k=1
p(ω)−p(z)
Since p is a polynomial of degree n, direct computation shows that the quotient ω−z is a
polynomial of degree n − 1. Thus if we may let
n−1
X
p(ω) − p(z)
= Qm (ω)z m ,
ω−z
m=0
where Qk (ω) is a polynomial in ω, then we get from [4, Eqn. (1), p. 129] that
f (ω) p(ω) − p(z) f (ω) · p(ω)−p(z)
ω−z

Res · ; ωk = Res ; ωk
p(ω) ω−z (ω − ω1 )(ω − ω2 ) · · · (ω − ωn )
p(ωk ) − p(z)
f (ωk ) ·
ωk − z
= Y
(ω − ωj )
j6=k
n−1
X f (ωk ) · Qm (ωk ) m
= Y z . (10.5)
m=0 (ω − ωj )
j6=k
Using the expressions (10.4) and (10.5), we conclude that

Z Xn n−1
X f (ωk ) · Qm (ωk )
1
P (z) = F (ω) dω = Y zm
2πi C (ω − ωj )
k=1 m=0
j6=k
which is a polynomial in z of degree n − 1.

Besides, we follow from the definition and [3, Theorem 3.8.6, p. 213] that
Z Z
1 f (ω) p(ω) 1 f (ω)
P (ωk ) = · dω = dω = f (ωk )
2πi C p(ω) ω − ωk 2πi C ω − ωk
for every k = 1, 2, . . . , n. We complete the analysis of the problem.
Problem 10.6
Proof. For small h, we observe that

Z Z
f (z + h) − f (z) φ(ω) 1 1 φ(ω)
= − dω = dω.
h γ h ω − z − h ω − z γ (ω − z − h)(ω − z)
Then we have
Z
′ f (z + h) − f (z) φ(ω)
f (z) = lim = lim dω. (10.6)
h→0 h h→0 γ (ω − z − h)(ω − z)
If γ : ω = ω(t) with a ≤ t ≤ b, then it follows from Definition 4.3 that

Z Z b
φ(ω)
G(h) = dω = F (t, h) dt, (10.7)
γ (ω − z − h)(ω − z) a
where
φ(ω(t))
F (t, h) = ω̇(t).
[ω(t) − z − h][ω(t) − z]
127
We claim that G is a continuous function of h. To this end, since z ∈ / γ, we obtain from

[22, Exercise 20, p. 101] that d(z, γ) = inf |ω − z| > 0. Furthermore, we know that d(z, γ) is
ω∈γ
a continuous function in z, so we also have d(z + h, γ) > 0 for sufficiently small h > 0. These
facts imply that
1
[ω(t) − z − h][ω(t) − z]
is a continuous function of t and h. Finally, the continuity of φ on γ and the piecewise continuity
of ω̇(t) on [a, b] ensure that F (t, h) is a piecewise continuous function of t and h. By the definition
(10.7), we have
Z b
|G(h) − G(h′ )| ≤ |F (t, h) − F (t, h′ )| dt,
a
so G is a continuous function of h, as desired.
Consequently, it is legitimate by the above claim that we can interchange the limit and the
integral in the expression (10.6) and get
Z Z
φ(ω) φ(ω)
f ′ (z) = lim dω = dω
γ h→0 (ω − z − h)(ω − z) γ (ω − z)2
which proves the first assertion.

For the second assertion, suppose that f is analytic in a simply connected domain D and γ
is a regular closed curve contained in D. By Definition 4.3, γ is compact. Since f is obviously
continuous on γ, the first assertion can be applied here to conclude
Z
′ 1 f (ω)
f (z) = dω,
2πi γ (ω − z)2
where z lies inside γ. Suppose that we have

Z
(k) k! f (ω)
f (z) = dω.
2πi γ (ω − z)k+1
Using similar attack as proving the first assertion,

Z
f (k) (z + h) − f (k) (z) k! 1h 1 1 i
= f (ω) · − dω.
h 2πi γ h (ω − z − h)k+1 (ω − z)k+1
Recall that
ak+1 − bk+1 = (a − b)(ak + ak−1 b + · · · + abk−1 + bk ),
so we see that
1 1 h
k+1
− k+1
= ×
(ω − z − h) (ω − z) (ω − z − h)(ω − z)
h 1 1 1 i
+ + · · · +
(ω − z − h)k (ω − z − h)k−1 (ω − z) (ω − z)k
and thus
f (k) (z + h) − f (k) (z)
f (k+1) (z) = lim
h→0
Z h
k! 1h 1 1 i
= lim f (ω) · − dω
h→0 2πi γ h (ω − z − h)k+1 (ω − z)k+1
Z
k! 1h 1 1 i
= f (ω) · lim − dω
2πi γ h→0 h (ω − z − h)k+1 (ω − z)k+1
Z
k! 1
= f (ω) · lim ×
2πi γ h→0 (ω − z − h)(ω − z)
h 1 1 1 i
+ + · · · + dω
(ω − z − h)k (ω − z − h)k−1 (ω − z) (ω − z)k
Z
k! k+1
= f (ω) · dω
2πi γ (ω − z)k+2
Z
(k + 1)! f (ω)
= k+2
dω.
2πi γ (ω − z)
Hence Theorem 10.11 (The Generalized Cauchy Integral Formula) follows immediately from the
induction. It completes the proof of the problem.
Problem 10.7
Proof. It is impossible that f (z) ≡ 0 on C because it contradicts our hypothesis. Assume that
f had more than one zero, say z1 and z2 . Let r be so large that z1 and z2 lie inside C(0; r). By
the hypothesis again, we know that both z1 and z2 must be real. Suppose that z1 > z2 which
gives
r > z1 > z2 > −r.
Furthermore, we know from the hypothesis that f (±r) are the only real values on C(0; r). Thus
if f (r) = 0, then we can select another r > r1 so that D(0; r1 ) contains z1 , z2 , i.e.,
r > r1 > z1 > z2 > −r1 > −r.
This process will stop in a finite number of steps, otherwise, one can find a decreasing sequence
{rn } such that
r > r1 > r2 > · · · > z1 > z2 > · · · > −r2 > −r1 > −r.
Since the sequence {rn } is obviously bounded, it follows from [22, Theorem 3.14, p. 55] that
it converges. Since f (rn ) = 0 for all n ∈ N, Theorem 6.9 (The Uniqueness Theorem) ensures
that f ≡ 0 in C, a contradiction. Hence we may suppose that our chosen r satisfies f (r) 6= 0.
Similarly, we also assume that f (−r) 6= 0.
Next, we define g : (0, π) → R by
g(θ) = f (reiθ ).
Since f is entire, g is a continuous function of θ and the previous paragraph shows that its
imaginary part is always positive or negative, i.e., {f (reiθ ) | 0 < θ < π} lies entirely in the upper
half-plane or the lower half-plane. Similarly, the set {f (reiθ ) | π < θ < 2π} lies entirely in the
upper half-plane or the lower half-plane. Since f is analytic in D(0; r) and f (z) 6= 0 on C(0; r),
we obtain from Corollary 10.9 (The Argument Principle) that
1
Z(f ) = ∆Arg f (z) (10.8)
2π
as z travels around the circle C(0; r) from re0·i to re2πi . Since ∆Arg f (z) ≤ π in the upper
half-plane or the lower half-plane, we establish from the formula (10.8) that
Z(f ) ≤ 1,

129
Problem 10.8
Proof.
(a) If z is a zero of f , then |f (z)| ≥ |g(z)| implies that g(z) = 0 and so f (z) + g(z) = 0, a
contradiction. In other words, f has no zero on γ so that 1 + fg is well-defined on γ. Now
we can follow the proof of Theorem 10.10 (Rouché’s Theorem) to get
Z d(1 + g )
1 f g
Z(f + g) = Z(f ) + g = Z(f ) + n 1+ (γ); 0 .
2πi γ 1+ f f
By the hypothesis, we know that |(1 + fg ) − 1| ≤ 1 on γ, so (1 + fg )(γ) remains within or on

the disc of radius 1 centred at 1. In addition, we observe that 1 + fg(z)
(z) = 0 if and only if
g
f (z) + g(z) = 0, so the analytic function 1 + f will not touch the origin. Finally, we note
that γ is a closed curve. Hence these observations ensure that (1 + fg )(γ) is also a closed
curve lying entirely in the right half-plane without surrounding 0. As a consequence, we
have g
n 1+ (γ); 0 = 0,
f
so Rouché’s Theorem holds in this case.
(b) Let f (z) = 2z 4 and g(z) = z 5 + 1. If e5iθ + 2e4iθ + 1 = 0 for some θ ∈ [0, 2π], then
|eiθ + 2| = 1 or
(2 + cos θ)2 + sin2 θ = 1. (10.9)
After simplification, the equation (10.9) reduces to cos θ = −2 which is impossible. Hence
f (z) + g(z) = z 5 + 2z 4 + 1 6= 0 on C(0; 1). Since |g(z)| ≤ 2 = |f (z)|, part (a) implies that
Z(z 5 + 2z 4 + 1) = Z(2z 4 ) = 4 in the unit disc.
Problem 10.9
Proof.
(a) Now the functions f (z) = 3ez and g(z) = −z which are analytic inside and on the regular
closed curve C(0; 1). Since |f (z)| = 3|ez | ≥ 3e−1 > |z| for all z ∈ C(0; 1), Theorem 10.10
(Rouché’s Theorem) implies that
Z(f1 ) = Z(f ) = Z(3ez ) = 0
inside C(0; 1).
(b) The functions f (z) = −z and g(z) = 31 ez are analytic inside and on the regular closed
curve C(0; 1). Since |f (z)| = |z| = 1 > 3e ≥ |g(z)| on C(0; 1), it follows from Theorem
10.10 (Rouché’s Theorem) that
Z(f2 ) = Z(f ) = Z(−z) = 1
inside C(0; 1).

(c) The functions f (z) = z 4 and g(z) = −5z + 1 are analytic inside and on the regular closed
curve C(0; 2). Since |f (z)| = |z 4 | = 32 > |5z − 1| = |g(z)| on C(0; 2), it follows from
Theorem 10.10 (Rouché’s Theorem) that
Z(f3 ) = Z(f ) = Z(z 4 ) = 4 (10.10)
inside C(0; 2). Similarly, the functions F (z) = −5z and G(z) = z 4 + 1 are analytic inside
and on the regular closed curve C(0; 1). Since |F (z)| = |5z| = 5 > |z 4 + 1| = |G(z)|, it
follows from Theorem 10.10 (Rouché’s Theorem) that
Z(f3 ) = Z(F ) = Z(5z) = 1 (10.11)
inside C(0; 1). Combining the results (10.10) and (10.11), Z(f3 ) = 4 − 1 = 3 inside
1 ≤ |z| ≤ 2.
(d) Suppose that f (z) = −5z 4 + 3z 2 and g(z) = z 6 − 1. Both f and g are analytic inside and
on the regular closed curve C(0; 1). The triangle inequality implies that
|f (z)| = | − z 4 + 3z 2 | ≥ |3z 2 | − |z 4 | = 2 and |g(z)| ≤ 2
on |z| = 1. Put z = eiθ into |f (z)| = 2 to get | − 5e2iθ + 3| = 2. Then we derive from this
that cos 2θ = 1 so that θ = 0, π. Consequently, |f (z)| = 2 if and only if z = ±1. Since
|g(±1)| = 0 < 2, it means that we always have |f (z)| > |g(z)| on C(0; 1). By Theorem
10.10 (Rouché’s Theorem), we see that
Z(f4 ) = Z(f ) (10.12)

q
inside C(0; 1). Notice that f (z) = 0 if and only if z = 0 (repeated roots) and z = ± 35 ,
so all the zeros of f lie inside C(0; 1) and hence the formula (10.12) gives Z(f3 ) = 4 inside
C(0; 1).
Problem 10.10
Proof. We take f (z) = λ − z and g(z) = −e−z . It is clear that z = λ is the only zero of f (z) and
it lies in the right half-plane. Let R > 2λ and we consider the rectangle in the right half-plane
SR = {z = x + iy | 0 ≤ x ≤ R and −R ≤ y ≤ R}.
Now the boundary γR = ∂SR is easily seen to be a regular closed curve and SR becomes the
right half-plane as R → ∞. Furthermore, our functions f and g are entire.
Let z ∈ γR . If x = 0, then z = iy and so we observe that
p
|λ − iy| = λ2 + y 2 ≥ λ > 1 = |e−iy |. (10.13)
√
Next, we suppose that x > 0. Since z = x + iR for 0 < x ≤ R, we have |z| = x2 + R2 > R so
that the triangle inequality gives
|λ − z| ≥ |z| − λ > λ > 1 ≥ |e−z |. (10.14)

131
p
Finally, if z = R + iy, then we also have |z| = R2 + y 2 ≥ R and thus the inequality (10.14)
remains valid in this case. Consequently, the inequalities (10.13) and (10.14) together imply
that
|f (z)| > |g(z)|
for all z ∈ γR . Hence it yields from Theorem 10.10 (Rouché’s Theorem) that
Z(λ − z − e−z ) = Z(f + g) = Z(f ) = 1
inside γR and the desired result follows if we let R → ∞. This completes the proof of the
problem.
Problem 10.11
Proof. If f is a nonzero constant, then there is nothing to prove. Without loss of generality, we
′ (z)
may assume that f is nonconstant. Let F (z) = z m ff (z) . Since f 6= 0 on γ, F is analytic inside
and on γ except for isolated singularities inside γ. According to [3, Theorem 5.1.2, p. 294], we
see immediately that
Z ′ Z X
1 m f (z) 1
z dz = F (z) dz = Res (F ; zk ),
2πi γ f (z) 2πi γ
k
where the zk denote the isolated singularities of F inside γ which are exactly the zeros of f
inside γ. By Lemma 7.1, there exists a function g analytic inside and on the regular closed curve
γ such that f (z) = (z − zk )pk g(z) and g(zk ) 6= 0. Therefore, we have
f ′ (z) pk z m z m g′ (z)
F (z) = z m = + . (10.15)
f (z) z − zk g(z)
If zk = 0, then F is actually analytic at 0 so that Res (F ; zk ) = 0. Otherwise, F has a simple

pole at each zero of f . Thus it yields from [4, Eqn. (1), p. 129] and the expression (10.15) that
Res (F ; zk ) = lim (z − zk )F (z) = pk zkm .

z→zk
Z ′ (z) Z X
1 mf 1
z dz = F (z) dz = pk zkm ,
2πi γ f (z) 2πi γ k
where the sum takes over all zeros of f inside γ. This completes the proof of the problem.
Problem 10.12
Proof. On the one hand, we have ez 6= 0 for all z ∈ D(0; R) with R > 0, so there exists a ǫR > 0
such that
|ez | > ǫR (10.16)
k Rk
for all z ∈ D(0; R). On the other hand, since | zk! | ≤ k! and the series
∞
X zk
k!
k=0
is obviously convergent, we deduce from Theorem 1.9 (The Weierstrass M -Test) that Pn (z) → ez
uniformly in D(0; R). In other words, there exists an N ∈ N such that n ≥ N implies
|Pn (z) − ez | < ǫR (10.17)
for all z ∈ D(0; R). By combining the inequalities (10.16) and (10.17), it is easy to see that
|Pn (z) − ez | < |ez |
holds for all n ≥ N and z ∈ D(0; R). As both Pn (z)−ez and ez are analytic in D(0; R), Theorem
10.10 (Rouché’s Theorem) asserts that
Z(Pn (z)) = Z(Pn (z) − ez + ez ) = Z(ez ) = 0
inside C(0; R), completing the proof of the problem.
Problem 10.13
Proof.
(a) Direct computation gives
(1 − z)P (z) = a0 + a1 z + a2 z 2 + · · · + an z n − a0 z − a1 z 2 − a2 z 3 − · · · − an z n+1

Xn
= a0 + (ak − ak−1 )z k − an z n+1 . (10.18)
k=1
Let R > 1. If |z| = R, then we have |z|n+1 > |z|k for all k = 0, 1, 2, . . . , n. Therefore, it
follows from the expression (10.18) and the fact an ≥ an−1 ≥ · · · ≥ a0 ≥ 0 that
n
X
|(1 − z)P (z) − (−an z n+1 )| = a0 + (ak − ak−1 )z k
k=1
h Xn i
< a0 + (ak − ak−1 ) · |z n+1 |
k=1
= | − an z n+1 |.
By Theorem 10.10 (Rouché’s Theorem), we conclude that
Z((1 − z)P (z)) = Z(−an z n+1 ) = n + 1
inside C(0; R). By letting R → 1, all zeros of (1 − z)P (z) (except z = 1) and then the
zeros of P (z) lie in C(0; 1).
133
(b) We consider
1
P (z) = (10.19)
(1 − z)2
which is analytic in D(0; 1). Since P (k) (z) = (k + 1)!(1 − z)−k−2 for every k = 0, 1, 2, . . .,
we observe from Theorem 6.5 (The Power Series Representation for Functions Analytic in
a Disc) that
P (k) (0)
Ck = =k+1
k!
and so
X∞
P (z) = (k + 1)z k (10.20)
k=1
in D(0; 1).
Now if |z| < 1, then |1 − z| < 2 and thus it follows from the definition (10.19) that
1
|P (z)| > . (10.21)
4
Next, since the power series (10.20) has radius of convergence 1, we recall from [4, p. 27]
that the series Pn (z) → P (z) uniformly in the smallest disc D(0; ρ). Consequently, there
exists an N ∈ N such that n ≥ N implies
1
|Pn (z) − P (z)| ≤ (10.22)
4
for all z ∈ D(0; ρ). Combining the inequalities (10.21) and (10.22), we know that
|Pn (z) − P (z)| < |P (z)|
on C(0; ρ). Hence we apply Theorem 10.10 (Rouché’s Theorem) that if n ≥ N , then we
have
Z(Pn ) = Z(P )
inside C(0; ρ). Since P has no zeros in D(0; 1), we establish that Z(Pn ) = 0 inside C(0; ρ)
for all n ≥ N .
Problem 10.14
Proof. Let P (z) = z n + an−1 z n−1 + · · · + a1 z + a0 , where a0 , a1 , . . . , an−1 ∈ C. Let f (z) = z n

and g(z) = an−1 z n−1 + · · · + a1 z + a0 . On the regular closed curve C(0; R) for large enough R,
we know that
|g(z)| ≤ |an−1 | · |z|n−1 + |an−2 | · |z|n−2 + · · · + |a0 | ≤ M Rn−1 < Rn = |f (z)|,
where M is a positive constant. Then Theorem 10.10 (Rouché’s Theorem) applies in this case
to conclude that
Z(P ) = Z(f + g) = Z(f ) = n
inside C(0; R). It completes the proof of the problem.
Problem 10.15
Proof. Since |f | > |g| on γ, we get from the triangle inequality that
|f (z) + λg(z)| ≥ |f (z)| − λ|g(z)| ≥ |f (z)| − |g(z)| > 0 (10.23)
for all z ∈ γ and all λ ∈ [0, 1]. Thus J(λ) is well-defined on [0, 1]. If s, t ∈ [0, 1], then we have
Z ′ Z
1 f + sg ′ f ′ + tg ′ |s − t| f g′ − f ′ g
|J(s) − J(t)| = · − dz = · dz . (10.24)
2π γ f + sg f + tg 2π γ (f + tg)(f + sg)
Using the inequality (10.23), we know that
|(f + tg)(f + sg)| = |f + tg| · |f − sg| ≥ (|f | − |g|)2
on γ and then
f g′ − f ′ g |f g′ − f ′ g|
≤ (10.25)
(f + tg)(f + sg) (|f | − |g|)2
on γ.
By Definition 4.2, γ : [a, b] → C is continuous, so γ([a, b]) is compact and hence it is of finite
length L. Since f, g, f ′ and g ′ are analytic on γ, they are continuous functions on γ([a, b]) so
that all the sets f (γ([a, b])), g(γ([a, b])), f ′ (γ([a, b])) and g′ (γ([a, b])) are also compact. Thus
there exists a M > 0 such that
f, g, f ′ , g ′ ≪ M
throughout γ which shows that
|f g′ − f ′ g| ≤ 2M 2 (10.26)
throughout γ. By the inequality (10.23) again, there exists a ǫ > 0 such that
√
|f (z)| − |g(z)| ≥ ǫ>0 (10.27)
on γ. Substutiting the inequalities (10.26) and (10.27) into the inequality (10.25), we conclude
that
f g′ − f ′ g 2M 2
≤
(f + tg)(f + sg) ǫ
on γ. Now we apply Theorem 4.10 (The M -L Formula) to the inequality (10.24) to obtain
|s − t| 2M 2
|J(s) − J(t)| ≤ × × L = A|s − t|
2π ǫ
for some constant A. Hence J(λ) is a continuous function of λ.
Next, direct computation shows
f (λg)′ −f ′ (λg) λg ′
(f + λg)′ f ′ + λg ′ f′ f2 f ′ (1 + f )
= = + = +
f + λg f + λg f 1 + λgf
f 1 + λg
f
and then
Z Z Z (1 + λg ′
1 (f + λg)′ 1 f′ 1 f )
J(λ) = dz = dz + dz. (10.28)
2πi γ f + λg 2πi γ f 2πi γ 1 + λg
f
135
Our hypothesis definitely implies that |f (z)| > 0 on γ, so we are able to employ Theorem 10.8 to
conclude that the second integral in the expression (10.28) is an integer N . Next, the inequality
(10.23) implies that |1 + λg
f | > 0 on γ, so the third integral in the expression (10.28) is also
an integer Nλ by Theorem 10.8. In other words, J(λ) ∈ Z for every λ ∈ [0, 1]. Recall from
the previous paragraph that J(λ) is a continuous function of λ, so it must be a constant. In
particular, J(0) = J(1) and Theorem 10.8 establishes that
Z ′ Z
1 f 1 (f + g)′
Z(f ) = dz = dz = Z(f + g)
2πi γ f 2πi γ f + g
as required. We end the proof of the problem.
Problem 10.16
Proof. By the discussion on [4, p. 115], we know that the function

p 1 1 Z z 2ζ
2 2
F (z) = z − 1 = exp log(z − 1) = exp dζ (10.29)
2 2 √2 ζ 2 − 1
√
is analytic in C \ (−∞, 1] because f (z) = z 2 − 1 is analytic there and f ( 2) 6= 0. Suppose that
D = C \ [−1, 1]. Then D is open in C.
• Step 1: F is continuous on D. If Im z > 0, then the path of integration in the

integral (10.29) is taken to be in the upper half-plane. Similarly, if Im z < 0, then the
path of integration in the integral (10.29) is taken to be in the lower half-plane. Since F is
analytic in C \ (−∞, 1], F is clearly continuous there. Therefore, it remains to show that
F is continuous on (−∞, −1), i.e., for every x0 ∈ (−∞, −1), we have
lim F (z).
F (x0 ) = z→x (10.30)
0
z∈D
√
paths from z to x0 and 2 in the
To this end, suppose that Im z > 0. Let γ1 and γ2 be the√
upper half-plane respectively. Let γ3 be the path from 2 to x0 in the lower half-plane.
See Figure 10.1 for an illustration.
Figure 10.1: The regular closed curve C surrounding ±1.

√
Let C = γ1 − γ2 − γ3 be the regular closed curve connecting 2 and x0 in D. Then C must
surround ±1. Since the function g(z) = z 2 − 1 is analytic and nonzero on C, it observes
from Corollary 10.9 (The Argument Principle) that
Z
2ζ
2
dζ = 4πi
C ζ −1
and then Z Z
1 2ζ 1 2ζ
2
dζ − dζ = 2πi
2 γ1 ζ −1 2 γ2 +γ3 ζ2 −1
which implies that
1 Z 2ζ 1
Z
2ζ
exp dζ = exp 2πi + dζ
2 γ1 ζ2 − 1 2 γ2 +γ3 ζ 2 − 1
1 Z 2ζ
= exp dζ .
2 γ2 +γ3 ζ 2 − 1
Hence we conclude that

lim F (z) =
z→x0
lim
z→x0
F (z).
z∈γ1 z∈γ2 +γ3
In other words, this proves the validity of (10.30) in the case Im z > 0. Now the case
Im z < 0 can be verified similarly and so we have established the desired result (10.30).
• Step 2: F is analytic in D. Let N be a large positive integer. By Step 1, we see that

F is continuous in the open set DN = D(0; N ) \ [−1, 1] ⊆ D and analytic there except
on the line segment (−N, −1), so Theorem 7.7 ensures that F is analytic throughout DN .
Since N is arbitrary and DN → D as N → ∞, we conclude that F is analytic in D.
Problem 10.17
Proof. Similar to Problem 10.16, we define

p
F (z) = 3 (z − 1)(z − 2)(z − 3)
1
= exp log[(z − 1)(z − 2)(z − 3)]
3
1 Z z [(ζ − 1)(ζ − 2)(ζ − 3)]′ 1
= exp dζ + log 6
3 4 (ζ − 1)(ζ − 2)(ζ − 3) 3
√ 1 Z z [(ζ − 1)(ζ − 2)(ζ − 3)]′
3
= 6 exp dζ (10.31)
3 4 (ζ − 1)(ζ − 2)(ζ − 3)
is analytic in C \ (−∞, 3] because f (z) = (z − 1)(z − 2)(z − 3) is analytic there and f (4) 6= 0. If
we define D = C \ [1, 3], then D is open in C.
• Step 1: F is continuous on D. If Im z > 0, then the path of integration in the integral

(10.31) is supposed to be in the upper half-plane. Similarly, if Im z < 0, then the path
of integration in the integral (10.31) is assumed to be in the lower half-plane. Since F is
137
analytic in C \ (−∞, 3], F is clearly continuous there. Therefore, it remains to show that
F is continuous on (−∞, 1), i.e., for every x0 ∈ (−∞, 1), we have
F (x0 ) = z→x
lim F (z). (10.32)
0
z∈D
To this end, suppose that Im z > 0. Similar to Figure 10.1, we let γ1 and γ2 be the paths
from z to x0 and 4 in the upper half-plane respectively. Furthermore, let γ3 be the path
from 4 to x0 in the lower half-plane.
Let C = γ1 − γ2 − γ3 be the regular closed curve connecting 4 and x0 in D. Then C must
surround 1, 2 and 3. Since the function g(z) = (z − 1)(z − 2)(z − 3) is analytic and nonzero
on C, it observes from Corollary 10.9 (The Argument Principle) that
Z
[(ζ − 1)(ζ − 2)(ζ − 3)]′
dζ = 6πi
C (ζ − 1)(ζ − 2)(ζ − 3)
and then
Z Z
1 [(ζ − 1)(ζ − 2)(ζ − 3)]′ 1 [(ζ − 1)(ζ − 2)(ζ − 3)]′
dζ − dζ = 2πi
3 γ1 (ζ − 1)(ζ − 2)(ζ − 3) 3 γ2 +γ3 (ζ − 1)(ζ − 2)(ζ − 3)
which implies that

1 Z [(ζ − 1)(ζ − 2)(ζ − 3)]′ 1 Z [(ζ − 1)(ζ − 2)(ζ − 3)]′
exp dζ = exp dζ .
3 γ1 (ζ − 1)(ζ − 2)(ζ − 3) 3 γ2 +γ3 (ζ − 1)(ζ − 2)(ζ − 3)

lim F (z) =
z→x0
lim
z→x0
F (z),
z∈γ1 z∈γ2 +γ3
i.e., the limit (10.32) exists in the case Im z > 0. Now the case Im z < 0 can be verified
similarly and so we have proved that the limit (10.32) exists.
• Step 2: F is analytic in D. Let N be a large positive integer. Now our F is continuous

in the open set DN = D(0; N ) \ [1, 3] ⊆ D and analytic there except on the line segment
(−N, 1). Hence Theorem 7.7 guarantees that F is analytic throughout DN and then
throughout D as desired if we take N → ∞.

CHAPTER 11
Applications of the Residue Theorem to the
Evaluation of Integrals and Sums
Problem 11.1
Proof.
(a) It is a Type I integral, so we have

Z ∞
x2 dx z2
2 2
= 2πiRes ; i . (11.1)
−∞ (1 + x ) (1 + z 2 )2
Since i is a pole of order 2 of the function
z2
f (z) = ,
(z + i)2 (z − i)2
we see that
z2 dh i d h z2 i i
2
Res ; i = (z − i) f (z) = =− . (11.2)
(1 + z 2 )2 dz z=i dz (z + i)2 z=i 4
Combining the expressions (11.1) and (11.2), we conclude immediately that

Z ∞
x2 dx π
2 )2
= .
−∞ (1 + x 2
x2
(b) Since (x2 +4)2 (x2 +9)
is an even function, we have
Z ∞ Z ∞
x2 dx 1 x2 dx
2 2 2
=
0 (x + 4) (x + 9) 2 −∞ (x2 + 4)2 (x2 + 9)
which is a Type I integral, so we have

Z ∞ h
x2 dx 1 z2
= · 2πi Res ; 3i
0 (x2 + 4)2 (x2 + 9) 2 (z 2 + 4)2 (z 2 + 9)
z2 i
+ Res ; 2i . (11.3)
(z 2 + 4)2 (z 2 + 9)
139
Chapter 11. Applications of the Residue Theorem to the Evaluation ... 140
z2
Here 3i and 2i are a simple pole and a pole of order 2 of the function f (z) = (z 2 +4)2 (z 2 +9)
respectively.
z2
Since (z − 2i)2 f (z) = (z+2i)2 (z 2 +9)
, we have
z2 dh z2 i 13i
Res ; 2i = =− . (11.4)
(z 2 + 4)2 (z 2 + 9) dz (z + 2i)2 (z 2 + 9) z=2i 200
Using [4, Eqn. (1), p. 129], we see that

z2 z2
Res ; 3i =
(z 2 + 4)2 (z 2 + 9) 2(z 2 + 4) · (2z)(z 2 + 9) + (z 2 + 4)2 · (2z) z=3i
3
=− . (11.5)
50i
Therefore, if we put the residues (11.4) and (11.5) back into the expression (11.3), then
we obtain Z ∞
x2 dx 13π 3π π
2 2 2
− = .
0 (x + 4) (x + 9) 200 50 200
1
(c) Since x4 +x2 +1 is an even function, we have
Z ∞ Z ∞
dx 1 dx
=
0 x4 + x2 + 1 2 −∞ x4 + x2 + 1
which is a Type I integral, so we have

Z ∞ h i
dx 1 1 1
= · 2πi Res ; z1 + Res ; z2 , (11.6)
0 x4 + x2 + 1 2 z4 + z2 + 1 z4 + z2 + 1
where z1 = cis π3 and z2 = cis 2π

3 . Since both z1 and z2 are simple poles of f (z) =
1
z 4 +z 2 +1 ,
[4, Eqn. (1), p. 129] gives
1 1 1
Res ; z1 = = √ (11.7)
z4 + z2 + 1 4z13 + 2z1 −3 + i 3
and
1 1 1
Res 4 2
; z2 = 3 = √ . (11.8)
z +z +1 4z2 + 2z2 3+i 3
Hence, by substituting the residues (11.7) and (11.8) into the expression (11.6), we obtain
Z ∞
dx 1 1 √3π
= πi √ + √ = .
0 x4 + x2 + 1 −3 + i 3 3 + i 3 6
sin x
(d) It is clear that it is a Type II integral. Since the function x(1+x2 )
is even, we have
Z ∞ Z ∞ Z ∞
sin x dx 1 sin x dx 1 eix
2
= 2
= Im dx.
0 x(1 + x ) 2 −∞ x(1 + x ) 2 −∞ x
eiz
Since z has a simple pole at 0, we follow the idea of the example on [4, p. 146], it is true
that
Z Z
∞
sin x dx 1 ∞
eix − 1 1 h eiz − 1 i
= Im dx = Im 2πiRes ;i .
0 x(1 + x2 ) 2 −∞ x(1 + x2 ) 2 z(1 + z 2 )
141
Using [4, Eqn. (1), p. 129], we know that

eiz − 1 eiz − 1 e−1 − 1
Res ; i = =−
z(1 + z 2 ) (1 + z 2 ) + 2z 2 z=i 2
and hence Z ∞
sin x dx 1 −1
(1 − e−1 )π
= Im (1 − e )πi = .
0 x(1 + x2 ) 2 2
(e) Again, it is a Type II integral. Since the integrand is even, we can write
Z ∞ Z
cos x dx 1 ∞ cos x dx 1 h eiz i
= = Re 2πiRes ;i .
0 1 + x2 2 −∞ 1 + x2 2 1 + z2
By [4, Eqn. (1), p. 129], we have
eiz eiz e−1
Res ; i = =
1 + z2 2z z=i 2i
which implies Z ∞
cos x dx π
2
= .
0 1+x 2e
(f) It is a Type III (A) integral, so
Z ∞ 2
X log z
dx
= − Res ; zk ,
0 x3 + 8 z3 + 8
k=0
where zk = 2 exp[ (2k+1)πi

3 ] for k = 0, 1, 2. Since they are simple poles of z 31+8 , it follows
from [4, Eqn. (1), p. 129] that
log z log z
Res 3 ; zk =
z +8 3z 2 z=zk
zk log zk
=−
24
zk (log |zk | + iArg zk )
=−
24
(2 + i 3 ) exp (2k+1)πi
(2k+1)π
3
=− .
12
Z ∞
dx (2 + i π3 ) exp πi3 (2 + iπ) exp(πi) (2 + i 5π 5πi
3 ) exp 3
= + +
0 x3 + 8 12 12 12
h √ √ i
1 πi 1 3i 5πi 1 3i
= 2+ + − (2 + iπ) + 2 + −
12 √ 3 2 2 3 2 2
1 4 3π
= ×
12
√ 6
3π
= .
18
(g) It is a Type III (C) integral. Thus
Z ∞ α−1 z α−1
2πi(α−1) x
[1 − e ] dx = 2πiRes ; −1 = 2πi(−1)α−1 = 2πie(α−1)πi
0 1 + x 1 + z
which implies that
Z ∞ α−1
x 2πie(α−1)πi −2πieαπi −2πi π
dx = 2πi(α−1)
= 2απi
= −απi απi
= .
0 1 + x 1 − e 1 − e e − e sin(απ)
(h) The integral is of Type IV, so [4, Eqn. (5), p. 150] gives
Z 2π Z
dx 1 dz
= z+z −1 2
×
(2 + cos x)2 (2 + iz
0 |z|=1 2 )
Z
4 z dz
=
i
|z|=1 (z 2
+ 4z + 1)2
z √
= 8πRes ; 3 − 2 . (11.9)
(z 2 + 4z + 1)2
z √ z
If f (z) = (z 2 +4z+1)2
= √
(z+2− 3)2 (z+2+ 3)2
, then we know that
z √ dh √ 2 i
Res ; 3 − 2 = (z + 2 − 3) f (z) √
(z 2 + 4z + 1)2 dz z=−2+ 3
dh z i
= √ √
dz (z + 2 + 3)2 z=−2+ 3
√
3
= . (11.10)
18
Now it yields from the expressions (11.9) and (11.10) that
Z 2π
√ √
dx 3 4 3π
= 8π × = .
0 (2 + cos x)2 18 9
(i) It is a Type IV integral, so [4, Eqn. (5), p. 150] implies

Z Z
2π
sin2 x dx − 14 (z − z1 )2 dz
= 3 1 × iz
0 5 + 3 cos x |z|=1 5 + 2 (z + z )
Z
i (z 2 − 1)2 dz
= 2 2
2|z|=1 z (3z + 10z + 3)
i X (z 2 − 1)2
= × 2πi Res 2 2 ; zk , (11.11)
2 z (3z + 10z + 3)
k
where zk are the zeros of z 2 (3z 2 + 10z + 3) inside the unit circle |z| = 1. Therefore, the
formula (11.11) reduces to
Z h
2π
sin2x dx (z 2 − 1)2 (z 2 − 1)2 1 i
= −π Res 2 2 ; 0 + Res 2 2 ;− .
0 5 + 3 cos x z (3z + 10z + 3) z (3z + 10z + 3) 3
Notice that
(z 2 − 1)2 d h (z 2 − 1)2 i 10
Res ; 0 = =−
z 2 (3z 2 + 10z + 3) dz 3z 2 + 10z + 3 z=0 9
and
(z 2 − 1)2 1 (z 2 − 1)2 8
Res ; − = = .
z 2 (3z 2 + 10z + 3) 3 2z(3z 2 + 10z + 3) + z 2 (6z + 10) z=− 31 9
Thus we get
Z 2π 10 8 2π
sin2 x dx
= −π − + = .
0 5 + 3 cos x 9 9 9
143
(j) The integral is of Type IV and then

Z 2π Z
dx 1 dz
= 1 1 × iz
0 a + cos x |z|=1 a + 2 (z + z )
Z
2 dz
=
i |z|=1 z 2 + 2az + 1
X 1
= 4π Res 2 ; zk , (11.12)
z + 2az + 1
k
where zk are zeros of z 2 + 2az + 1 lying inside the unit circle. We note that z 2 + 2az + 1 = 0
if and only if p
z = −a ± a2 − 1.
√ √
If a > 1,√then | − a + a2 − 1| < 1√but | − a − a2 − 1| > 1. Similarly, if a < −1, then
| − a − a2 − 1| < 1 but | − a + a2 − 1| > 1. Therefore, we get from the expression
(11.12) that
 p
 1
Z 2π 
 4πRes ; −a + a2 − 1 , if a > 1;
dx  z 2 + 2az + 1
=
0 a + cos x   1 p
 4πRes
 ; −a − a2 − 1 , if a < −1
 z 2 + 2az + 1
 1

 4π × 2z + 2a z=−a+√a2 −1 , if a > 1;

=

 1

 4π × , if a < −1
√
2z + 2a z=−a− a2 −1
 2π

 √ , if a > 1;

 a2 − 1
=

 2π

 −√ , if a < −1.
a2 − 1
Problem 11.2
Proof. By the power series expansion of e2iz around 0, we have
e2iz − 1 − 2iz 4i
2
= −2 − z + · · · .
z 3
2iz
Since e −1−2iz
z2
has a removable singularity at 0, it is entire. Let CR be the closed contour used
on [4, p. 143], we follow from Theorem 6.3 (The Closed Curve Theorem) that
Z
e2iz − 1 − 2iz
dz = 0.
CR z2
Recall that Z Z Z R
= + .
CR ΓR −R
Therefore, we get
Z R Z
e2ix − 1 − 2ix e2iz − 1 − 2iz
dx + dz = 0. (11.13)
−R x2 ΓR z2
On the one hand, if z ∈ ΓR , then we have
e2iz − 1 e−2Im z + 1 2
2
≤ 2
≤ 2
z R R
and Theorem 4.10 (The M -L Formula) implies that
Z
e2iz − 1 2 2π
dz ≤ 2 × πR = .
ΓR z2 R R
Z
e2iz − 1
lim dz = 0. (11.14)
R→∞ ΓR z2
By Definition 4.3, we know that
Z Z π Z π
dz iReiθ
= dθ = i dθ = iπ (11.15)
ΓR z 0 Reiθ 0
for every R > 0. Now it follows from the results (11.14) and (11.15) that
Z
e2iz − 1 − 2iz
lim dz = −2i · iπ = 2π. (11.16)
R→∞ ΓR z2
On the other hand, since

Z 0 Z R
e2ix − 1 − 2ix e−2ix − 1 + 2ix
dx = dx,
−R x2 0 x2
the first integral in the expression (11.13) can be written as

Z R Z 0 Z R
e2ix − 1 − 2ix e2ix − 1 − 2ix e2ix − 1 − 2ix
dx = dx + dx
−R x2 −R x2 0 x2
Z R −2ix Z R
e − 1 + 2ix e2ix − 1 − 2ix
= dx + dx
0 x2 0 x2
Z R
e2ix + e−2ix − 2
= dx
0 x2
Z R
−4 sin2 x
= dx. (11.17)
0 x2
Hence, by putting the results (11.16) and (11.17) into the expression (11.13), we conclude that
Z ∞
sin2 x
−4 dx + 2π = 0
0 x2
so that Z ∞
sin2 x π
2
dx = .
0 x 2
145
Problem 11.3
Proof. Let CR be the given contour of large enough radius R. Furthermore, let ΓR , ηR and ηR ′ be
2π
the circular arc, the ray on the real axis and the ray with angle n with the direction indicated
in the exercise respectively. Note that if R is large enough, then the point exp( πi
n ) lies inside
CR .
On the one hand, since exp( πi n
n ) is a simple zero of the equation z + 1 = 0, Theorem 10.5
(The Cauchy’s Residue Theorem)a implies that
Z 1 πi
dz
n
= 2πiRes ; exp
CR 1 + z 1 + zn n
1
= 2πi × n−1
nz z=exp( πin
)
z
= −2πi ×
n z=exp( πin
)
−2πi πi
= exp . (11.18)
n n
On the other hand, if z ∈ ΓR , then z = Reiθ with 0 ≤ θ ≤ 2π n n n
n . Since |z +1| ≥ |z| −1 = R −1
on ΓR for large R, Theorem 4.10 (The M -L Formula) gives
Z
dz 2πR 1
n
≤ × n . (11.19)
ΓR 1 + z n R −1
Since n ≥ 2, the inequality (11.19) shows that
Z
dz
lim = 0. (11.20)
R→∞ ΓR 1 + zn
Next, it is obvious that
Z ∞ Z Z
dx dx dz
= lim = lim . (11.21)
0 1 + xn R→∞ ηR 1 + xn R→∞ ηR 1 + zn
′ , we have z = (R − x) exp( 2πi ) with 0 ≤ x ≤ R so that
Finally, on ηR n
Z Z R 2πi Z 0 2πi Z R dx
dz exp( n ) dx exp( 2πi
n ) dy
= − = = − exp . (11.22)
′ 1 + zn
ηR 0 1 + (R − x)
n
R 1 + yn n 0 1+x
n
Recall that CR = ΓR + ηR + ηR ′ , we substitute the results (11.20), (11.21) and (11.22) into the
expression (11.18) to get

Z Z Z πi
dz dz dz 2πi
lim + lim + lim =− exp
R→∞ ηR 1 + z n R→∞ η′ 1 + z n R→∞ ΓR 1 + z n n n
R
Z ∞ Z ∞
dx 2πi dx 2πi πi
− exp = − exp
0 1 + xn n 0 1 + xn n n
Z ∞
dx 2πi exp( πi
n)
= − ·
0 1+x n n 1 − exp( 2πi n )
π
= .
n sin πn
a
Of course, we apply the version [3, Theorem 5.1.2, p. 294] of Asmar and Grafakos here.
Problem 11.4
Proof.
(a) By the substitution y = ax, we have
Z ∞ Z ∞ Z
cos ax 3 cos y dy a3 ∞ cos y dy
dx = a =
0 (1 + x2 )2 0 (a2 + y 2 )2 2 −∞ (a2 + y 2 )2
which is a Type II integral. Since
Z ∞ h i
cos y dy eiz
2 2 2
= Re 2πiRes ; ai
−∞ (a + y ) (a2 + z 2 )2
n d h eiz i o
= Re 2πi
dz (z + ai)2 z=ai
(a + 1)e−a π
= ,
2a3
we have Z ∞
cos ax (a + 1)e−a π
dx = .
0 (1 + x2 )2 4
(b) Using the same strategy of Problem 11.3, we see that
Z z2 πi 3πi
z 2 dz πi
10
= 2πiRes ; exp = − exp . (11.23)
CR 1 + z 1 + z 10 10 5 10
On ΓR , similar to the inequality (11.19), we have
Z
z 2 dz πR R2
10
≤ ×
ΓR 1 + z 5 R10 − 1
so that Z
z 2 dz
lim = 0. (11.24)
R→∞ ΓR 1 + z 10
On ηR , we have
Z ∞ Z Z
x2 dx x2 dx z 2 dz
= lim = lim . (11.25)
0 1 + x10 R→∞ ηR 1 + x10 R→∞ ηR 1 + z 10
′ , we have
On ηR
Z Z R 3πi Z R x2 dx
z 2 dz (R − x)2 exp( 3πi
5 )
= − dx = − exp . (11.26)
′ 1 + z 10
ηR 0 1 + (R − x)2 5 0 1+x
10
After putting the results (11.24), (11.25) and (11.26) into the expression (11.23), we es-
tablish that
h 3πi i Z ∞ x2 dx πi 3πi
1 − exp = − exp
5 0 1 + x10 5 10
Z ∞ 2
x dx πi 1
10
=− × 3πi 3πi
0 1 + x 5 exp(− 10 − exp( 10 )
)
πi 1
=− ×
5 −2i sin 3π
10
π
= .
10 sin 3π
10
147
(c) We note that

Z 2π Z Z ez
iθ dz z ez
exp(e ) dθ = e = −i dz = 2πRes ; 0 = 2π.
0 |z|=1 iz |z|=1 z z
Problem 11.5
Proof. It is clear that

Z Z
2π
2m 1 1 2m dz
(cos x) dx = × z +
0 |z|=1 4m z iz
Z
1 (z 2 + 1)2m
= m dz
i4 |z|=1 z 2m+1
2π (z 2 + 1)2m
= m Res ;0 .
4 z 2m+1
2m
X
2 2m
Since (z + 1) = Ck2m z 4m−2k , we have
k=0
2m
X 2m
X
(z 2 + 1)2m 2m 2m−2k−1
2m
Cm
= C k z = + Ck2m z 2m−2k−1 . (11.27)
z 2m+1 z
k=0 k=0
k6=m
Hence we immediately conclude from the expansion (11.27) that

Z 2π
2π 2m
(cos x)2m dx = m Cm ,
0 4
Problem 11.6
Proof. This is a Type I integral so that

Z ∞
dx 1
2 n+1
= 2πiRes ; i . (11.28)
−∞ (1 + x ) (1 + z 2 )n+1
Since f (z) = (z + i)−n−1 (z − i)−n−1 has a pole of order n + 1 at i, we know from the formula
on [4, p. 130] that
1 1 dn
Res ; i = · (z + i)−n−1
(1 + z 2 )n+1 n! dz n z=i
1
= · (−n − 1)(−n − 2) · · · (−2n)(z + i)−2n−1
n! z=i
1 (−1)n (n + 1)(n + 2) · · · (2n)
= ×
n! (2i)2n+1
(2n)!
= −i ·
(n!)2 · 22n+1
1 · 3 · 5 · · · (2n − 1)
= −i · . (11.29)
2 · 4 · 6 · · · (2n) · 2
Hence our desired result follows immediately if we combine the expressions (11.28) and (11.29).
Problem 11.7
′ as described in Problem 11.3 with the central

Proof. Consider the contour CR = ΓR + ηR + ηR
π
angle 4 , see Figure 11.1 below.
π
Figure 11.1: The contour CR with central angle 4.
2
Since f (z) = eiz is entire and CR is a closed curve, Theorem 4.16 (The Closed Curve
Theorem) ensures that
Z Z Z Z
iz 2 iz 2 iz 2 2
e dz + e dz + e dz = eiz dz = 0. (11.30)
ΓR ηR ′
ηR CR
(a) On ΓR , if z = Rcis θ, where 0 ≤ θ ≤ π4 , then

2 2
|eiz | = e−R sin 2θ
. (11.31)
Recall from [22, Exercise 7, p. 197] that 2x π

π ≤ sin x ≤ x ≤ π for all x ∈ [0, 2 ], so if we put
x = 2θ, then we have
4θ
≤ sin 2θ ≤ π (11.32)
π
for every θ ∈ [0, π4 ]. By applying the result (11.32) to the expression (11.31) we conclude
that
iz 2 2 4θ
|e | ≤ exp − R ·
π
on ΓR and thus Lemma 4.9 gives
Z Z π
4
iz 2 2
e dz = ei[z(θ)] · R(− sin θ + i cos θ) dθ
ΓR 0 | {z }
This is G(θ).
149
Z 4θ
π
4
≤ R exp − R2 · dθ
0 | {z π }
This is |G(θ)|.
π π
2 4θ 4
= −R · exp − R ·
4R2 π 0
π −R2
= (1 − e )
4R
which obviously tends to 0 as R → ∞.
(b) On ηR , we know that

Z Z R Z R Z R
iz 2 ix2 2
e dz = e dx = cos x dx + i sin x2 dx. (11.33)
ηR 0 0 0
π
′ , we have z = (R − x)ei 4 with 0 ≤ x ≤ R so that
Next, on ηR
Z Z R Z R
iz 2 i π4 2 i π4 2
e dz = −e exp[−(R − x) ] dx = −e e−y dy. (11.34)
′
−ηR 0 0
Hence we put the result of part (a), the expressions (11.33) and (11.34) into the equation
(11.30) to obtain
hZ 2
Z
2
i
lim eiz dz + eiz dz = 0
R→∞ ηR ′
ηR
Z ∞ Z ∞ Z ∞
π 2
cos x2 dx + i sin x2 dx = ei 4 e−x dx. (11.35)
0 0 0
Recall the fact [11, §3.321, Eqn. (3), p. 336] that

Z ∞ √
−x2 π
e dx = ,
0 2
so the equation (11.35) becomes
Z ∞ Z ∞ √ √
2 2 π π 2π
cos x dx = sin x dx = · cos = .
0 0 2 4 4
Remark 11.1
The integrals considered in Problem 11.7(b) are called Fresnel integrals.
Problem 11.8
Proof. Let z1 , z2 , . . . , zm be the poles of f (i.e., the zeros of Q). Let R be large enough so that all
poles of f lie inside the circle CR = C(0; R). Then we follow from Theorem 10.5 (The Cauchy’s
Residue Theorem) that
Z Xm P (z)
P (z)
dz = 2πi Res ; zk . (11.36)
CR Q(z) Q(z)
k=1
Let P (z) = an z n + an−1 z n−1 + · · · + a0 and Q(z) = bm z m + bm−1 z m−1 + · · · + b0 , where m − n ≥ 2

and an bm 6= 0. If M = max{|a0 |, |a1 |, . . . , |an |}, then it is clear that M > 0 and
|P (z)| ≤ M |z|n (11.37)
for large |z|. Furthermore, we recall from Problem 1.26 that

|bm | m
|Q(z)| ≥ |z| (11.38)
2
for large enough |z|. Hence if |z| = R is large enough, then we yield from the inequalities (11.37)
and (11.38) that
P (z) 2M
≤
Q(z) |bm |Rm−n
on C(0; R). Thus we deduce from Theorem 4.10 (The M -L Formula) that
Z
P (z) 2M 4M π
dz ≤ m−n
× 2πR = →0
CR Q(z) |bm |R |bm |Rm−n−1
as R → ∞ because m − n − 1 ≥ 1, and hence the equation (11.36) becomes

m
X P (z) Z
1 P (z)
Res ; zk = lim dz = 0.
Q(z) 2πi R→∞ CR Q(z)
k=1
Problem 11.9
Proof.
1
(a) This is a Type I sum. Let f (z) = 1+z 2 . Then we have
∞
X h π cot πz π cot πz π cot πz i
f (n) = − Res ; 0 + Res ; i + Res ; −i . (11.39)
n=−∞
1 + z2 1 + z2 1 + z2
n6=0
π cot πz
Since 0 and ±i are simple poles of 1+z 2
, we establish from [4, Eqn. (1), p. 129] that
π cot πz π
Res 2
; 0 = = 1,
1+z 2z tan πz + π(1 + z 2 ) sec2 πz z=0
π cot πz π cot πz
Res ; i = Res ; −i
1 + z2 1 + z2
π cot πi
= .
2i
By the identities of sin z and cos z on [4, p. 41], we have
i(e−π + eπ )
cot πi = ,
e−π − eπ
so we get from the formula (11.39) that
∞
X 1 π cot πi π(e−π + eπ ) π(e2π + 1)
2 = − 1 + = −1 + = −1 + .
n2 + 1 i eπ − e−π e2π − 1
n=1
151
As a consequence, we have
1h π(e2π + 1) i
∞
X 1
= − 1 + .
n=1
n2 + 1 2 e2π − 1
1
(b) Again it is a Type I sum, so if f (z) = z4
which has a pole of order 4 at 0, then we have
∞
X π cot πz
f (n) = −Res ;0 .
n=−∞
z4
n6=0
Since the Laurent series of cot πz around 0 is given byb
1 πz π 3 z 3 2 5 5
cot πz = − − − π z + ··· ,
π 3 45 945
we obtain at once that
∞
X 1 π4
= .
n4 90
n=1
1
(c) This is a Type II sum. Suppose that f (z) = 1+z 2
which has simple poles at ±i. By [4,
Eqn. (1), p. 129], we have
X∞
(−1)n π csc πz π csc πz
= −Res ; i − Res ; −i
n=−∞
1 + n2 1 + z2 1 + z2
n6=0
π csc πz π csc πz
=− −
2z z=i 2z z=−i
π
=−
i sin πi
2π
= π
e − e−π
which implies that
∞
X (−1)n e2π + 2πeπ − 1
= .
1 + n2 2(e2π − 1)
n=1
Problem 11.10
Proof.
(a) Figure 11.2 shows the square CN = γ1 + γ2 + γ3 + γ4 considered in the problem. Let
A = N + 12 . On γ1 , we have z = A + it with t ∈ [−A, A]. Therefore, we obtain from
Definition 4.3 that Z Z A
dz N i dt
= (−1) .
γ1 z sin πz −A (A + it) cos itπ
b
See [16, Exercise A(5)(a), p. 730].
Figure 11.2: The square CN with vertices ±(N + 12 ) ± i(N + 12 ).
Similarly, on −γ3 , we have z = −A + it with t ∈ [−A, A] so that

Z Z Z A
dz dz N i dt
=− = −(−1) . (11.40)
γ3 z sin πz −γ3 z sin πz −A (A − it) cos itπ
Using the substitution t = −u to the last integral (11.40), we see that

Z Z A
dz N i du
= −(−1) .
γ3 z sin πz −A (A + iu) cos iuπ
Consequently, we have Z Z
dz dz
+ =0
γ1 z sin πz γ3 z sin πz
and then Z Z Z
dz dz dz
= + . (11.41)
CN z sin πz γ2 z sin πz γ4 z sin πz
On γ2 or γ4 , we have z = t ± iA, where t ∈ [−A, A], so
1 2 2
≤√ ≤ Aπ − e−Aπ )
z sin πz 2 2
A + t · |e−Aπ Aπ
−e | A(e
for large enough A and so for large N . By Theorem 4.10 (The M -L Formula), we conclude
that Z Z
dz 4 dz 4
≤ Aπ −Aπ
and ≤ Aπ
γ2 z sin πz e − e γ4 z sin πz e − e−Aπ
which tend to 0 as A → ∞ or equivalently as N → ∞. By the relation (11.41), we see at
once that Z
dz
lim = 0.
N →∞ CN z sin πz
1
(b) Similar to Type II series, if f (z) = 2z−1 which has a simple pole at 12 , then we have
Z h 1 i
N
X
π dz (−1)n π
= 2πi + Res ;
CN (2z − 1) sin πz 2n − 1 (2z − 1) sin πz 2
n=−(N −1)
h X
N
(−1)n π 1 (−1)N i
= 2πi + Res ; −
2n − 1 (2z − 1) sin πz 2 2N − 1
n=−N
153
h XN
(−1)n π 1 (−1)N i
= 2πi 2 + Res ; − , (11.42)
2n − 1 (2z − 1) sin πz 2 2N − 1
n=1
where CN is the rectangle with vertices (N + 12 ) ± i(N + 12 ) and −(N − 21 ) ± i(N + 12 ) for
large positive integer N , see Figure 11.3.
Figure 11.3: The rectangle CN with vertices (N + 12 ) ± i(N + 12 ) and −(N − 21 ) ± i(N + 12 ).
Let A = N + 21 , if z = A + it with |t| ≤ A, then we get

Z Z A
π dz (−1)N i dt
=
γ1 (2z − 1) sin πz 2 −A (N + it) cos iπt
and if z = −(N − 21 ) + it with |t| ≤ A, then we have

Z Z A
π dz (−1)N i dt
=− .
γ3 (2z − 1) sin πz 2 −A (N + it) cos iπt
Hence, by similar strategy applied in part (a), we can show that

Z
π dz
lim = 0,
N →∞ CN (2z − 1) sin πz
so the equation (11.42) reduces to

∞
X (−1)n+1 π 1
2 = Res ;
2n − 1 (2z − 1) sin πz 2
n=1
π
=
2 sin πz + π(2z − 1) cos πz z= 21
π
=
2
which definitely implies that
X∞
1 1 1 (−1)n+1 π
1− + − + ··· = = .
3 5 7 n=1
2n − 1 4

Problem 11.11
e kz
Proof. Suppose that f (z) = 1+e z and consider the rectangle ΓR with vertices at ±R and ±R+2πi
as shown in Figure 11.4. It is clear that 1 + ez vanishes inside the rectangle if and only if z = πi.
Figure 11.4: The rectangle ΓR with vertices at ±R and ±R + 2πi.
Thus we represent
z − πi z − πi 1
(z − πi)f (z) = ekz · z
= ekz · z = ekz ·
1+e e − eπi ez −eπi
z−πi
which implies
1
lim (z − πi)f (z) = lim ekz · ez −eπi
= −ekπi .
z→πi z→πi
z−πi
By Theorem 10.5 (The Cauchy’s Residue Theorem), we see that

Z ekz
ekz
dz = 2πiRes ; πi = −2πiekπi . (11.43)
ΓR 1 + ez 1 + ez
Now it is obvious that, on γ4 , we have

Z Z R
ekz ekx
dz = dx. (11.44)
γ4 1 + ez −R 1 + ex
Similarly, we know that points on −γ2 have the form z = x + 2πi with −R ≤ x ≤ R, so it follows
from Proposition 4.7 that
Z Z Z R Z R
ekz ekz ek(x+2πi) ekx
dz = − dz = − dx = −e2kπi dx. (11.45)
γ2 1 + ez −γ2 1 + ez −R 1 + ex+2πi −R 1 + ex
Next, suppose that z = R + it ∈ γ1 with t ∈ [0, 2π]. For large enough R, the triangle inequality
implies that
eR
|1 + eR+it | ≥ |eR+it | − 1 ≥
2
so that
ekz ek(R+it)
= ≤ 2e(k−1)R .
1 + ez 1 + eR+it
155
Hence Theorem 4.10 (The M -L Formula) gives

Z
ekz
z
dz ≤ 4πe(k−1)R
γ1 1 + e
which ensures that the integral along γ1 tends to 0 as R → ∞ because k < 1. By similar
analysis, the integral along γ3 tends to 0 as R → ∞ because k > 0. Combining these facts and
the expressions (11.44) and (11.45), this assures us that
Z ∞
2kπi ekx
(1 − e ) x
dx = −2πiekπi
−∞ 1 + e
Z ∞
ekx ekπi
x
dx = −2πi ·
−∞ 1 + e 1 − e2kπi
1
= 2πi · kπi
e − e−kπi
π
= .
sin kπ
Problem 11.12
Proof. We consider the keyhole contour Kǫ,M used on [4, p. 147], but this time we let M → 1.
By Theorem 10.5 (The Cauchy’s Residue Theorem), we have
Z X
f (z) log z dz = 2πi Res (f (z) log z; zk ),
Kǫ,M k
where the sum is taken over all the poles of f inside Kǫ,M . Since f is analytic for |z| ≤ 1, f has
no poles so that Z
f (z) log z dz = 0 (11.46)
Kǫ,M
for any 0 < ǫ < M < 1.

By Theorem 4.10 (The M -L Formula), we have
Z
f (z) log z dz ≤ πǫ max |f (z) log z|. (11.47)
Cǫ Cǫ
Since f is continuous at 0 and | log z| < | log |z||+2π = | log ǫ|+2π, we obtain from the inequality
(11.47) that Z
f (z) log z dz ≤ Aπǫ(| log ǫ| + 2π) → 0 (11.48)
Cǫ
as ǫ → 0, where A is a positive constant.
√ Next, we√recall that CM is the circular arc of radius
M traced counterclockwise from M 2 − ǫ2 + iǫ to M 2 − ǫ2 − iǫ, so it is true that
Z Z
lim f (z) log z dz = f (z) log z dz. (11.49)
ǫ→0
M →1 CM |z|=1
Finally, since Z Z 1
lim f (z) log z dz = f (x) log x dx (11.50)
ǫ→0
M →1 I1 0
and similarly, Z Z 1
lim f (z) log z dz = − f (x)(log x + 2πi) dx. (11.51)
ǫ→0
M →1 I2 0
Since Kǫ,M = CM + I2 + Cǫ + I1 , we put all the results (11.48), (11.49), (11.50) and (11.51) into
the equation (11.46) to conclude that
Z
lim f (z) log z dz = 0
ǫ→0
M →1 Kǫ,M
Z Z Z
lim f (z) log z dz + lim f (z) log z dz + lim f (z) log z dz = 0
ǫ→0 ǫ→0 ǫ→0
M →1 CM M →1 I1 M →1 I1
Z Z 1 Z 1
f (z) log z dz + f (x) log x dx − f (x)(log x + 2πi) dx = 0
|z|=1 0 0
Z Z 1
1
f (z) log z dz = f (x) dx.
2πi |z|=1 0
Problem 11.13
Proof. We note that Cn3n is the coefficient of z n in (1 + z)3n . Therefore, we have

Z
3n 1 (1 + z)3n
Cn = dz
2πi C z n+1
where C is a simple closed contour surrounding the origin. Similar to the analysis of [4, Example
1, p. 155], we choose C to be the circle C(0; 21 ) because
(1 + z)3 (1 + |z|)3 27
≤ = <1
8z 8|z| 32
throughout C(0; 12 ) and the convergence of the series

∞
X (1 + z)3n
n=0
(8z)n
is uniform there. Hence we have

∞
X ∞ Z
1 1 X (1 + z)3n dz
Cn3n · = ·
8n 2πi C(0; 1
) (8z)n z
n=0 n=0 2
Z hX (1 + z)3n i dz
∞
1
= ·
2πi C(0; 1 ) n=0 (8z)n z
2
Z
1 1 dz
= ·
2πi C(0; 1 ) 1 − (1+z)3 z
2 8z
Z
1 8 dz
=
2πi C(0; 1 ) 8z − (1 + z)3
2
Z
4i dz
= 3 2
π C(0; ) z + 3z − 5z + 1
1
2
157
Z
4i dz
= . (11.52)
π C(0; 21 ) (z − 1)(z 2 + 4z − 1)
√
Since z 2 + 4z − 1 has a simple zero at −2 ± 5, the application of Theorem 10.5 (The Cauchy’s
Residue Theorem) shows that the expression (11.52) further reduces to
∞
X 1 4i 1 √
Cn3n · = · 2πiRes ; −2 + 5
8n π (z − 1)(z 2 + 4z − 1)
n=0
1
= −8 · √
z2
+ 4z − 1 + (z − 1)(2z + 4) z=−2+ 5
−8
= √ √
2 5(−3 + 5)
4
=√ √
5(3 − 5)
√
5+3 5
= .
5
Problem 11.14
Proof. The result is trivial if x = 0, so we assume that x 6= 0. Similar to [4, Example 1, p. 155],
we have
∞ Z h
(1 + z)2 x in dz
X∞
2n n 1 X
Cn x = · , (11.53)
2πi C z z
n=0 n=0
where C is a simple contour surrounding the origin. Particularly, we choose C = C(0; 1). It
follows from the hypothesis |x| < 41 that
(1 + z)2 x |1 + z|2
≤ |x| ≤ 4|x| < 1
z |z|
throughout C(0; 1). Therefore, the convergence is uniform and we are able to interchange the
order of summation and integration in the expression (11.53) and get
Z ∞ h Z
(1 + z)2 x in
∞
X X
1 dz 1 dz
Cn2n xn = · =− ,
n=0
2πi C(0;1) n=0 z z 2πi C(0;1) (1 + z)2 x − z
Since (1 + z)2 x − z = 0 is the same as
xz 2 + (2x − 1)z + x = 0
whose solutions are given by √

1 − 2x ± 1 − 4x
z± = .
2x
1
Now the hypothesis |x| < 4 again implies that 0 < 1 − 4x < 2 so that
√ √
1 − 2x 1 + 2 − 2x 1− 2 − 2x 1 − 2x
< z+ < and < z− <
2x 2x 2x 2x
if x > 0 as well as
√ √
1+ 2 − 2x 1 − 2x 1 − 2x 1 − 2 − 2x
< z+ < and < z− <
2x 2x 2x 2x
if x < 0. Direct computation shows that z+ > 1 if x > 0 and z+ < −1 if x < 0. Hence it follows
from Theorem 10.5 (The Cauchy’s Residue Theorem) that
∞
X 1
Cn2n xn = −Res ; z−
n=0
xz 2 + (2x − 1)z + x
1
=−
2xz + (2x − 1) z=z−
1
=√
1 − 4x
which completes the proof of the problem.
Problem 11.15
Proof. To find the maximum of a2 b, where a2 + b2 = 4 and 0 ≤ a, b ≤ 2. We define the function

f (b) = (4 − b2 )b = 4b − b3 . Then we have
max a2 b = max f (b).

0≤a,b≤2 0≤b≤2
√
2 3
√ √ √
Note that f ′ (b) = 4 − 3b2 = 0 if and only if b = 3 and f ′′ ( 2 3 3 ) = −6 · 2 3
3 = −4 3 < 0, so f
√
2 3
attains its maximum at b = 3 and its maximum value is given by
2√3 √
16 3
max f (b) = f = . (11.54)
0≤b≤2 3 9
If |z| = 1, then we see from the figure provided in the problem that |z − 1|2 + |z + 1|2 = 4.
Thus we may let a = |z − 1| and b = |z + 1| so that our a and b satisfy the hypotheses in the
previous paragraph. Consequently, we conclude from the value (11.54) that
√
2 16 3
|(z − 1) (z + 1)| ≤ ,
9
completing the analysis of the problem.
Problem 11.16
Proof. Since part (b) considers

√ points on the circle of radius R > 0, we do this general case in
part (a) and then take R = 2 to obtain the desired result. Let z ∈ C(0; R), a = |z + 1| and
b = |z − 1|. Now the points z, −z, 1 and −1 form a parallelogram in the plane C.
159
(a) It is well-known that the sum of the squares of the diagonals of parallelogram is equal to
the sum of the squares of its sides.c In other words, we have (2R)2 + 22 = 2(a2 + b2 ) which
reduces to
a2 + b2 = 2(1 + R2 ), (11.55)
p √ √
where 0 ≤ a, b ≤ 2(1 + R2 ). If R = 2, then we have a2 = 6 − b2 and 0 ≤ a, b ≤ 6. To
2
find the maximum of a2 b , we define
6b − b3
f (b) = ,
2
√
where b ∈ [0, 6]. Then their maximums are the same, i.e.,
a2 b
max√ = max √ f (b).
a,b∈[0, 6] 2 b∈[0, 6]
2 √ √
Now f ′ (b) = 6−3b = 0 if and only if b = 2 and f ′′ (b) = −3b so that f ′′ ( 2) < 0.
2 √
Consequently, f attains its maximum at b = 2 and the maximum value is
√
√ 2(6 − 2) √
max√ f (b) = f ( 2) = =2 2
0≤b≤ 6 2
which gives the desired result.
(b) In the general case, we note that R > 0 and
[2(1 + R2 ) − b2 ]b
f (b) = ,
R2
p
where 0 ≤ b ≤ 2(1 + R2 ). As usual, we have
2(1 + R2 ) − 3b2
f ′ (b) = =0
R2
if and only if r
2(1 + R2 )
b= .
3
Since f ′′ (b) = − R6b2 , we must have
r r
2(1 + R2 )
6 2(1 + R2 )
′′
f =− 2 · < 0.
3 R 3
q
2)
Hence f attains its maximum at b = 2(1+R 3 and the maximum value is given by
r √
2(1 + R2 ) 3
4 6 (1 + R2 ) 2
max
√ f (b) = f = · . (11.56)
0≤b≤ 2(1+R2 ) 3 9 R2
To find the minimum of the expression (11.56), we define F : (0, ∞) → (0, ∞) by

3
(1 + R2 ) 2
F (R) = .
R2
c
Read [22, Exercise 17, p. 23].
Direct computation gives √

′ (R2 − 2) 1 + R2
F (R) =
R3
√
and then F ′ (R) = 0 if and only if R = 2.
√ It can be easily seen from the First Derivative
Test that F attains its minimum at R = 2 and actually,
3 √
(1 + 2) 2 3 3
min F (R) = = . (11.57)
0<R<∞ 2 2
Finally, by putting the value (11.57) back into the expression (11.56), we see that
√ √
(z − 1)2 (z + 1) 4 6 3 3 √
max = max
√ f (b) ≥ · = 2 2.
|z|=R z2 0≤b≤ 2(1+R2 ) 9 2
Problem 11.17
Proof.
(a) Note that Ckn and (−1)k Ck3n are the coefficients of z k and z −k in (1 + z)n and (1 − z −1 )3n
respectively. Therefore, the sum
n
X
(−1)k Ck3n Ckn (11.58)
k=0
is the constant term in the product (1 + z)n (1 − z −1 )3n . Using [4, Eqn. (4), p. 123], it is
given by
Z Z
1 (1 + z)n (1 − z −1 )3n 1 [(z − 1)3 (1 + z)]n
dz = dz, (11.59)
2πi C z 2πi C z 3n+1
where C = C(0; R) for R > 0. Combining the expression (11.58) and the integral (11.59),
we get
X n Z
k 3n n 1 [(z − 1)3 (1 + z)]n
(−1) Ck Ck = dz. (11.60)
2πi C z 3n+1
k=0
(b) Similar to Problem 11.16, we have to find the value
(z − 1)3 (z + 1)
max .
|z|=R z3
a3 b
Let a = |z − 1| and b = |z + 1|. To find the maximum of R3
, where a and b satisfy the
condition (11.55), we define
a3 b 2
[2(1 + R2 ) − b2 ]3 b2
f (b) = = , (11.61)
R3 R6
p √
where 0 ≤ b ≤ 2(1 + R2 ). Particularly, we put R = 3 into the formula (11.61) to get
b2 (b2 − 8)3
f (b) = − ,
27
161
√
where 0 ≤ b ≤ 8. Consider the function F (u) = −u(u − 8)3 with domain [0, 8]. Elemen-
tary calculus shows that
F ′ (u) = −4(u − 8)2 (u − 2),
so F ′ (u) = 0 if and only if u = 2, 8. Now the First Derivative Test ensures that F takes
the maximum at u = 2 and its maximum value is given by
max F (u) = F (2) = 432

0≤u≤8
so that
r
(z − 1)3 (z + 1) a3 b r 432
max = max √ 3 = max√ f (b) = = 4. (11.62)
|z|=R z3 0≤a,b≤ 8 3 2 0≤b≤ 8 27
Combining this fact (11.62) and Theorem 4.10 (The M -L Formula), we conclude that
Z
1 [(z − 1)3 (1 + z)]n 1 (z − 1)3 (z + 1) n 1 √
dz ≤ · max · √ · 2π 3 = 4n .
2πi C(0;√3) z 3n+1 2π |z|=R z3 3
Finally, the formula (11.60) guarantees that

n
X
(−1)k Ck3n Ckn ≤ 4n .
k=0
Now we complete the proof of the problem.

CHAPTER 12
Further Contour Integral Techniques
Problem 12.1
Proof.
(a) Let z = ω − 1. Then I becomes the line I ′ given by ω(t) = 1 + it, where −∞ < t < ∞.
Furthermore, we have Z Z ω
ez −1 e
3
dz = e 4
dω. (12.1)
I (z + 1) I′ ω
Using similar argument as [4, Example 1, pp. 161, 162], we see that
Z ω eω 2πi
e
4
dω = 2πiRes ;0 = . (12.2)
I′ ω ω4 5!
Combining the expressions (12.1) and (12.2), we assert that
Z
ez πi
3
dz = .
I (z + 1) 60e
(b) We use the same contour considered in [4, Example 1, pp. 161, 162] to write
Z 1+iR z Z az
a az
2
dz + 2
dz = 2πiRes ;0 , (12.3)
1−iR z CR z z2
where R > 3. Since az = ez log a , if z = x + iy for x ≤ 1, then we have
|az | = |ez log a | = |ex log a | = ex log a ≤ max(1, elog a ) = max(1, a)
in the left half-plane as R → ∞. As a result, Theorem 4.10 (The M -L Formula) implies
that Z
az
lim dz = 0
R→∞ CR z 2
and hence the equation (12.3) reduces to

Z z az ez log a
a
2
dz = 2πiRes ; 0 = 2πiRes ; 0 = 2πi log a
I z z2 z2
because
z log a (z log a)2
ez log a = 1 + + + ··· .
1! 2!
163
Chapter 12. Further Contour Integral Techniques 164
Remark 12.1
There is a general formula for the integral
Z x0 +i∞
az F (z) dz
x0 −i∞
in [16, Eqn. (4.13.4), p. 952] for the case a ≥ 1 and x0 ∈ R.
Problem 12.2
√
Proof. The function 4z 2 − 8z + 3 has zeros at z = 12 and z = 32 , so 4z 2 − 8z + 3 is analytic in
√
the plane minus [ 12 , 32 ].a Since 4z 2 − 8z + 3 ∼ 2z for large z, it follows that
Z Z
dz 1 dz
lim √ = lim = πi. (12.4)
R→∞ |z|=R 2
4z − 8z + 3 2 R→∞ |z|=R z
Now Figure 12.1 shows the formations of the contours C(0; R) and C(0; 2).
Figure 12.1: The formations of the contours C(0; R) and C(0; 2).
Note that
C(0; R) − C(0; 2) = Γ1 + Γ2 + (γ1 + γ2 ) = (Γ1 + I1 + γ1 + I2 ) + (Γ2 − I2 + γ2 − I1 ).

1
Since (4z 2 − 8z + 3)− 2 is analytic in a simply connected domain containing the smooth closed
curve Γ1 + I1 + γ1 + I2 , Theorem 8.6 (The General Closed Curve Theorem) ensures that
Z
dz
√ = 0. (12.5)
2
4z − 8z + 3
Γ1 +I1 +γ1 +I2
a
Refer to Problem 10.16.
165
Similarly, we also have Z

dz
√ = 0. (12.6)
Γ2 −I2 +γ2 −I1 4z 2 − 8z + 3
Therefore, we follow from the results (12.5) and (12.6) that
Z Z
dz dz
√ = √
C(0;2) 4z 2 − 8z + 3 C(0;R) 4z 2 − 8z + 3
and hence the limit (12.4) implies

Z
dz
√ = πi.
C(0;2) 4z 2 − 8z + 3
Problem 12.3
Proof. Let R > 0 be large and ǫ > 0 be small. We consider the closed contour ΓR formed by
the following curves and lines:
• γ1 : [−R, R] → C is the parabola defined by
γ1 (t) = 1 − t2 + it;
• the straight line γ2 connecting γ1 (R) = 1 − R2 + iR and the point ǫ exp(iθR ), where
θR = Arg γ1 (R);
• γ3 : [θR , θ−R ] → C defined by

γ3 (t) = ǫeit ,
where θ−R = Arg γ1 (−R);
• the straight line γ4 connecting ǫ exp(iθ−R ) to γ1 (−R) = 1 − R2 − iR.
This contour is shown in Figure 12.2 below:
Figure 12.2: The closed contour formed by γ1 , γ2 , γ3 and γ4 .

It is clear that f (z) = ez log z is analytic inside and on the closed contour ΓR and f has no
singularities inside ΓR , Theorem 10.5 (The Cauchy’s Residue Theorem), we find
Z
ez log z dz = 0
ΓR
and thus Z Z Z Z
z z z
e log z dz = − e log z dz − e log z dz − ez log z dz. (12.7)
γ1 γ2 γ3 γ4
√
Notice that both −γ2 and γ4 have the same form γ : [ǫ, R4 − R2 + 1] → C given by
γ(t) = teiθ ,
R R
where θ satisfies either tan θ = 1−R 2 or tan θ = − 1−R2 . When R → ∞ and ǫ → 0, in the case
of −γ2 , we have θ → −π and for γ4 , we find θ → π. By Proposition 4.7 and then Definition 4.3,
we have
Z Z
f (z) dz = − f (z) dz
γ2 −γ2
Z √
R4 −R2 +1
=− f (−γ2 (t))(−γ2 )′ (t) dt
ǫ
Z √
R4 −R2 +1
=− eiθ f (teiθ ) dt
ǫ
Z √
R4 −R2 +1
= −eiθ exp(teiθ ) log(teiθ ) dt
ǫ
Z √
R4 −R2 +1
iθ
= −e exp(teiθ )(log t + iθ) dt
ǫ
so that Z Z ∞
lim f (z) dz = e−t (log t − iπ) dt. (12.8)
R→∞ γ2 0
ǫ→0
Similarly, we also have

Z Z √ Z √
R4 −R2 +1 R4 −R2 +1
f (z) dz = f (γ4 (t))γ4′ (t) dt =e iθ
exp(teiθ )(log t + iθ) dt
γ4 ǫ ǫ
which gives Z Z ∞
lim f (z) dz = − e−t (log t + iπ) dt. (12.9)
R→∞ γ4 0
ǫ→0
By [11, §4.331, Eqn. (1), p. 571], we know that

Z ∞
e−t log t dt = −C,
0
where C is the Euler’s constant, so the integrals (12.8) and (12.9) can be split into two convergent
integrals and then
Z Z Z ∞
lim f (z) dz + lim f (z) dz = 2πi e−t dt = 2πi.
R→∞ γ2 R→∞ γ4 0
ǫ→0 ǫ→0
167
Substitute this result into the equation (12.7), we see that

Z Z
z
lim e log z dz = 2πi − lim ez log z dz. (12.10)
R→∞ γ1 R→∞ γ3
ǫ→0 ǫ→0
Finally, if z = γ3 (t) = ǫeit , where t ∈ [θR , θ−R ] ⊆ [−π, π], then we know that |γ3 (t)| ≤ 2πǫ
and
|f (z)| = | exp(ǫeit ) log(ǫeit )| = eǫ cos t | log ǫ + it| ≤ e1 (| log ǫ| + 2π).
Since f is obviously continuous on γ3 , Theorem 4.10 (The M -L Formula) implies that
Z
f (z) dz ≤ 2πe1 (| log ǫ| + 2π)ǫ → 0
γ3
as ǫ → 0. Hence we conclude from the expression (12.10) that

Z
ez log z dz = 2πi
γ
which ends the proof of the problem.
Problem 12.4
Proof. Using exactly the same argument as the proof of [4, Example 3, pp. 163, 164], we have
n
X Z
n n 13 1 1 π
(−1) (Ck ) = [f (z)] 3 · dz, (12.11)
2πi C sin πz
k=0
where C is any contour in −1 < Re z < n + 1 and

sin πz
f (z) = .
πz(1 − z)(1 − 2z ) · · · (1 − nz )
Furthermore, we can split the integral (12.11) as

Z 1 Z n+ 1 +i∞
1 h − 2 −i∞ i
n
X
n n 13
2 1 π
(−1) (Ck ) = + [f (z)] 3 · dz , (12.12)
2πi − 1 +i∞ n+ 1 −i∞ sin πz
k=0 2 2
If z = n − ω, then we have
n! sin π(n − ω)
f (n − ω) =
π(n − ω)[(ω + 1 − n)(ω + 2 − n) · · · (ω − 1)(ω − 0)]
(−1)n+1 n! sin πω
= n
π(n − ω)(−1) [−ω(1 − ω) · · · (n − 2 − ω)(n − 1 − ω)]
n! sin πω
=
πω[(1 − ω)(2 − ω) · · · (n − 2 − ω)(n − 1 − ω)(n − ω)]
sin πω
= (1−ω) 2−ω (n−2−ω) (n−1−ω) (n−ω)
πω[ 1 · 2 · · · n−2 · n−1 · n ]
sin πω
=
πω(1 − ω)(1 − ω2 ) · · · (1 − ωn )
= f (ω).
Therefore, we see that

Z n+ 12 +i∞ Z − 21 −i∞
1 π 1 π
[f (z)] 3 · dz = [f (ω)] 3 · dω
n+ 21 −i∞ sin πz − 21 +i∞ sin πω
and when Re z = − 21 , we deduce from the expression (12.12) that
X∞ Z − 1 −i∞
1 1 2 1 π
(−1)n (Ckn ) 3 ≤ [f (z)] 3 · dz
π − 1 +i∞ sin πz
k=0 2
1 2
2 1 Z dz
3
≤ × π3 × √ 2
π n+1 Re z=− 21 (sin πz) 3
√
3 Z
2 dz
= √3
√6 2 . (12.13)
π · n + 1 Re z=− 2 (sin πz) 3
1
Note that if z = − 12 + iy, then the definition of sin z gives
1 − πi −πy πi 1
| sin πz| = (e 2 − e 2 +πy ) = (e−πy + eπy ).
2i 2
Consequently, if y ∈ [0, ∞), then | sin πz| ≥ 12 eπy . If y ∈ (−∞, 0), then we have | sin πz| ≥ 21 e−πy .
Now these imply that
Z √ h Z 0 2π Z ∞ i
dz 3 y − 2π y
2 = 4 · e 3 dy + e 3 dy
Re z=− 12 (sin πz) 3 −∞ 0
√ h 3 2π 0 3 2π ∞i
e3y e− 3 y
3
= 4· −
2π −∞ 2π 0
√3 3
= 4· .
π
Hence it follows from the inequality (12.13) that
∞ √
X 3
2 √ 3 6
n n 31 3
(−1) (Ck ) ≤ √ √6
· 4· = √ √
k=0
3
π· n+1 π π π· 6 n+1
3
which guarantees
∞
X 1
(−1)n (Ckn ) 3 → 0
k=0
as n → ∞, completing the proof of the problem.
Problem 12.5
Proof.
(a) Using the same argument as in the proof of [4, Example 3, pp. 163, 164], we know that
Z 3 Z n+ 3 +i∞ p
1 h − 4 −i∞ i
∞
X p 4 π
(−1)k Ckn = + f (z) · dz . (12.14)
2πi − 3 +i∞ n+ 3 −i∞ sin πz
k=0 4 4
169
p
As we have shown in Problem 12.4, the integrand f (z) · sinππz is invariant under the
substitution z 7→ n − z, it suffices to consider the first integral in the expression (12.14).
On the line Re z = − 43 , we have z = − 34 + iy and
z z z z
z(1 − z) 1 − ··· 1 − = |z| · |1 − z| · 1 − · · · 1 −
2 n 2 n
3 3 3 3
≥ · 1+ · 1+ ··· 1 +
4 4 4·2 4n
Yn
3 3
= · 1+ . (12.15)
4 4k
k=1
Recall the binomial theorem ([22, Exercise 22, p. 201]) that

∞
X ∞
X
α α(α − 1) · · · (α − n + 1) n
(1 + x) = Cnα xn = Cnα = x .
n=0 n=0
n!
1
If we put x = k and α = 34 , then we obtain
1 34 3 3 5 3
1+ =1+ − 2
+ 3
− ··· ≤ 1 + (12.16)
k 4k 32k 128k 4k
3
because |Cn4 | is a decreasing function of n. Combining the inequalities (12.15) and (12.16),
we see immediately that
z z 3 Y
n
1 43
z(1 − z) 1 − ··· 1 − ≥ · 1+
2 n 4 k
k=1
3 h Y k + 1 i 34
n
= ·
4 k
k=1
3 3
= · (n + 1) 4 . (12.17)
4
Hence it concludes that
Z − 34 −i∞ p Z − 34 −i∞
p
π 1 | sin πz| π dz
f (z) · dz ≤ · p z z
·
− 34 +i∞ sin πz π − 34 +i∞ |πz(1 − z)(1 − 2 ) · · · (1 − n )| sin πz
Z − 3 −i∞ √
1 4 2 π dz
≤ · √ 3 ·
√
π − 3 +i∞ 3(n + 1) 8 sin πz
4
Z − 3 −i∞
2 4 dz
=√ 3
√
3π(n + 1) 8 − 34 +i∞ sin πz
3
≤ An− 8
for some constant A > 0.
(b) Let δ > 0 be small. Suppose that −t = −1 + δ so that t ∈ (0, 1). Now our integration
becomes along Re z = −t and Re z = n + t. On Re z = −t, since t ∈ (0, 1), instead of the
inequality (12.17), we establish
z z
n
Y t
z(1 − z) 1 − ··· 1 − ≥t 1+
2 n k
k=1
hY
n
1 it
≥t 1+
k
k=1
= (1 − δ)(n + 1)1−δ
≥ (1 − δ)n1−δ . (12.18)
1 πy
Recall the facts from the proof of Problem 12.4 that | sin πz| ≥ 2e if y ∈ [0, ∞) and
| sin πz| ≥ 12 e−πy if y ∈ (−∞, 0), where y = Im z.
Now instead of (12.14), we have
Z Z n+1−δ+i∞ p
1 h −1+δ−i∞ i
∞
X
k
p n π
(−1) Ck = + f (z) · dz .
2πi −1+δ+i∞ n+1−δ−i∞ sin πz
k=0
Applying similar argument we have employed in part (a), we follow from the inequality
(12.18) that
X∞ Z
k
p n 1 −1+δ−i∞ p π dz
(−1) Ck ≤ f (z) ·
π −1+δ+i∞ sin πz
k=0
Z −1+δ−i∞
1 dz
=√ √ p
π −1+δ+i∞ sin πz × z(1 − z)(1 − 2z ) · · · (1 − nz )
Z
1 dz
≤p √
π(1 − δ)n1−δ Re z=−1+δ sin πz
A
≤p
(1 − δ)n1−δ
for some positive constant A.
Problem 12.6
1
Proof. Suppose that x > 0 and h(t) = tt−1 > 0 on [0, ∞). Since ext ≥ 1 + xt, we observe from
the definition that
Z ∞ xt Z ∞ hZ 1 Z ∞ i
e dt
g(x + 2πi) = f (x) = dt ≥ x =x h(t) dt + h(t) dt . (12.19)
0 tt 0 tt−1 0 1
Since t1−t is bounded on [0, 1], the integral

Z 1 Z 1
h(t) dt = t1−t dt
0 0
is convergent and then the convergence of the integral

Z ∞
dt
t t−1
0
depends on the convergence of the second integral inside the brackets (12.19). To this end, we
study the convergence of the series
∞
X ∞
X 1
h(n) = . (12.20)
nn−1
n=1 n=1
171
1
Let an = nn−1
. Then it is easy to see that
√ 1
α = lim sup n
an = lim 1 = 0,
n→∞ n→∞ n1− n
so [27, Theorem 6.7, p. 76] ensures that the series (12.20) converges. Since h(t) ≥ 0 and h
decreases monotonically on [1, ∞), it follows from [28, Theorem 1.10, p. 37] that the second
integral in the brackets (12.19) is convergent. Therefore, there is a positive constant M such
that
g(x + 2πi) ≥ M x
which shows certainly that g(x + 2πi) → ∞ as x → ∞. This completes the analysis of the
problem.
CHAPTER 13
Introduction to Conformal Mapping
Problem 13.1
Proof. Suppose that z0 6= 0 and z1 , z2 ∈ C are distinct numbers such that f (z1 ) = f (z2 ). If
z2 = 0, then it forces that z1 = 0. Therefore, without loss of generality, we may assume that
z2 6= 0. Then we can write
h z k i
1
f (z1 ) − f (z2 ) = z2k −1
z2
which means that f (z1 ) = f (z2 ) if and only if z1 = z1 (n) = z2 exp( 2nπi
k ) if and only if
2nπ
Arg z1 − Arg z2 = (13.1)
k
for some n ∈ {1, 2, . . . , k − 1}. Hence the equation (13.1) insures that if δ > 0 is chosen such
that D(z0 ; δ) lies in the sector S = {z ∈ C | α < Arg z < β}, where 0 < β − α < 2π k (see Figure
13.1 below), then f is locally 1-1 at z0 by Definition 13.3.
Figure 13.1: The sector S and the disc D(z0 ; δ).
More precisely, if θ = Arg z0 , then we may take α = θ − πk and β = θ + πk so that the ray
connecting the origin and the point z0 bisects the sector S. This completes the proof of the
problem.
173
Chapter 13. Introduction to Conformal Mapping 174
Problem 13.2
Proof. If x = c, then we have

|ω| = | exp(c + iy)| = ec
which is a circle centred at 0 with radius ec . Next, if y = c, then ω = ex · eic which is a ray
with angle c and ex > 0. (Note that when x runs through R, ex runs through (0, ∞).) We have
Problem 13.3
Proof. We remark that throughout this chapter, we denote H to be the upper half-plane.
(i) Note that f1 (z) = z+2 maps the strip S onto the strip S1 = {z | 0 < Re z < 3} conformally.
Next, the mapping f2 (z) = π3 f1 (z) maps S1 onto the strip S2 = {z | 0 < Re z < π}
conformally. Furthermore, the mapping f3 (z) = if2 (z) sends S2 onto the vertical strip
S4 = {z | 0 < Im z < π} and the mapping f4 (z) = ef3 (z) maps S4 onto H conformally.
By Theorem 13.16, the mapping f5 (z) = ff44 (z)+i
(z)−i
conformally maps H onto T . Hence a
required conformal mapping f is given by
exp( πi
3 (z + 2)) − i
f (z) = .
exp( πi
3 (z + 2)) + i
(ii) Theorem 13.23 implies that the (unique) bilinear transformation f mapping −2, 0, 2 into
−1, 0, 2 is given by
(f (z) − 0)(2 + 1) (z − 0)(2 + 2)
=
(f (z) + 1)(2 − 0) (z + 2)(2 − 0)
which reduces to
4z
f (z) = .
6−z
(iii) The mapping f1 (z) = z 4 maps S onto H conformally. Next, the mapping f2 (z) = log f1 (z)
maps H onto the strip {z | 0 < Im z < π} conformally. Finally, f3 (z) = π1 f2 (z) conformally
maps the strip {z | 0 < Im z < π} onto the strip {z | 0 < Im z < 1}. Hence a required
conformal mapping f is given by
4
f (z) = log z.
π
1
(iv) The mapping f1 (z) = iz 2 sends S onto the upper semi-disc S1 = {z | |z| = 1 and Im z > 0}
(z−1)2
conformally. By [4, Example 1, p. 180], the mapping f2 (z) = − 4(z+1) 2 sends S1 confor-
z−i
mally onto H. According to Theorem 13.6, f3 (z) = z+i is a conformal mapping of H onto
the T . Therefore, a required conformal mapping f is given by
h √ i h √ i−1
(i z − 1)2 (i z − 1)2
f (z) = − √ −i × − √ +i .
4(i z + 1)2 4(i z + 1)2

175
Problem 13.4
Proof. Let the region between the two circles be G. The two circles touch internally at 2,
1
so [4, Example 1, p. 181] indicates that the mapping f1 (z) = z−2 sends G conformally onto
two parallel lines. Since f1 (0) = − 2 and f1 (−2) = − 4 , the two parallel lines are Re z = − 14
1 1
and Re z = − 21 . Then the mapping f2 (z) = z−2 1

+ 12 = 2(z−2)
z
maps G onto the infinite strip
S = {z ∈ C | 0 < Re z < 14 } conformally. Next, the mapping f3 (z) = 4πif2 (z) = 2πiz
z−2 sends G
conformally onto the strip T = {z ∈ C | 0 < Im z < π} and so the mapping
2πiz
f4 (z) = exp f3 (z) = exp
z−2
sends G onto H conformally. Finally, Theorem 13.16 shows that the mapping f given by
2πi
exp( z−2 )−i
f (z) = 2πi
exp( z−2 )+i
maps G onto the unit disc D(0; 1) conformally, completing the proof of the problem.
Problem 13.5
Proof. Let S = {z = x + iy ∈ C | x > 0 and 0 < y < 1}. Then f1 (z) = πiz sends S conformally
onto S1 = {z = x + iy, | −π < x < 0 and y > 0}. Next, f2 (z) = πiz + π2 will map S onto the
semi-infinite strip n o
S2 = z ∈ C − π2 < Re z < π2 and Im z > 0
conformally. We know that the Schwarz-Christoffel transformation f (z) = sin z maps S2 onto
H, so the composition f3 = f ◦ f2 maps S onto H conformally. Since
π
f3 (z) = f (f2 (z)) = sin + πiz = cos(πiz),
2
we observe from Theorem 13.16 that the mapping
cos(πiz) − i
g(z) =
cos(πiz) + i
sends S conformally onto the unit disc D(0; 1). This ends the proof of the problem.
Problem 13.6
Proof. By [4, Example 1, p. 180], the mapping f defined by
(z − 1)2
f (z) = −
4(z + 1)2
sends S onto H conformally. Now Theorem 13.16 ensures that the mapping g given by
f (z) − i h (z − 1)2 i h (z − 1)2 i−1

g(z) = = − − i × − + i
f (z) + i 4(z + 1)2 4(z + 1)2
maps S conformally onto the unit disc D(0; 1) and we complete the proof of the problem.
Problem 13.7
Proof. Suppose that U, V, W are regions and there exist conformal mappings f : U → V ,
g : V → W such that f (U ) = V and g(V ) = W . In other words, we have U ∼ V and
V ∼ W.
• Reflexive Property: The identity map id : U → U is certainly bijective. Furthermore,

it satisfies id ′ (z) = 1 for all z ∈ U , so Theorem 13.4 implies that id is conformal, i.e.,
U ∼ U.
• Symmetric Property: Since f : U → V is a conformal mapping, Theorem 13.8 shows

that F = f −1 : V → U is also a conformal mapping. Since f is bijective, F is also bijective.
Thus we conclude that V ∼ U .
• Transitive Property: We know that h = g ◦ f : U → W is bijective because f and g

are bijective. Since f and g are analytic, Problem 3.3 ensures that h is also analytic. By
Definition 13.9, h is a conformal mapping which means that U ∼ W .
According to the definition, “conformal equivalence” is an equivalence relation which completes

Problem 13.8
Proof.
(a) Suppose that R = {z = x+iy | a ≤ x ≤ b and c ≤ y ≤ d} and ∂R = γ1 +γ2 +γ3 +γ4 , where
γ1 and γ3 are the horizontal lines and γ2 and γ4 are the vertical sides.a Let f (z) = az + b,
where a 6= 0. It is clear that f : ∂R → f (∂R) is bijective. By Theorem 13.11, f (γ1 ),
f (γ2 ), f (γ3 ) and f (γ4 ) are lines. Since f is conformal at the four vertices, f (γ1 ) and f (γ3 )
are parallel lines such that they are perpendicular to both f (γ2 ) and f (γ4 ). Consequently,
f (∂R) is also a polygon.
(b) Let R and R′ = f (R) be the said rectangles. The hypotheses imply that f is nonconstant.
By a rotation and a translation if necessary, we may assume that points z of R have the
form 0 ≤ Re z ≤ a and 0 ≤ Im z ≤ b for some positive constants a and b. Denote 0, v1 , v2
and v3 to be its vertices. Similarly, we may also assume that if z ∈ R′ , then 0 ≤ Re z ≤ c
and 0 ≤ Im z ≤ d for some positive constants c and d with vertices 0, v1′ , v2′ and v3′ . Figure
13.2 shows the setting for this problem.
a
If not, then we can consider the rectangle R′ = eiθ R for some θ whose sides are either horizontal or vertical.
177
Figure 13.2: The rectangles R and R′ .
Suppose that f (z0 ) ∈ {0, v1′ , v2′ , v3′ }. By Problem 7.3, we have z0 ∈ ∂R. We claim that
if z0 ∈ / {0, v1′ , v2′ , v3′ }. Otherwise, we assume, without loss of
/ {0, v1 , v2 , v3 }, then f (z0 ) ∈
generality, that f (z0 ) = v2′ . Now we want to compare ∠γ1 , γ2 and ∠f (γ1 ), f (γ2 ), see Figure
13.3 below for an illustration:
Figure 13.3: The angles of ∠γ1 , γ2 and ∠f (γ1 ), f (γ2 ).
By Theorems 13.4 and 13.7, we have ∠γ1 , γ2 = kπ, where k is the least positive integer
such that f (k) (z0 ) 6= 0. However, as the interior angle at v2′ is π2 , so we will have
∠γ1 , γ2 6= ∠f (γ1 ), f (γ2 ),
a contradiction. Thus this proves our claim.

By Theorem 7.1 (The Open Mapping Theorem), we have the observation that interior
points of R are mapped by f to interior points of R′ . Assume that there was a ω ∈ R◦
such that f (ω) ∈ ∂R. For each n ∈ N, the continuity of f ensures that there exists a
δn > 0 such that zn ∈ D(ω; δn ) ⊆ R◦ implies
1
|ω − f (zn )| < .
n
Thus we can find a bounded sequence {zn } in R◦ such that f (zn ) → ω as n → ∞. By
the Bolzano-Weierstrass Theorem, {zn } has a convergent subsequence with limit z. Now
z ∈ R◦ is impossible because it contradicts to our mentioned observation. Hence f maps
∂R to ∂R′ .
Finally, since a rectangle is also a parallelogram, we conclude from Problem 7.23 that f is
a linear polynomial.
Problem 13.9
Proof. We check the definition [8, pp. 16, 17]. Suppose that f1 (z) = ca11z+d
z+b1
1
and f2 (z) = a2 z+b2
c2 z+d2 ,
where a1 d1 − b1 c1 6= 0 and a2 d2 − b2 c2 6= 0. Then it is easy to check that
a1 f2 (z) + b1
f1 (f2 (z)) =
c1 f2 (z) + d1
a2 z+b2
a1 · c2 z+d2 + b1
= a2 z+b2
c1 · c2 z+d2 + d1
a1 (a2 z + b2 ) + b1 (c2 z + d2 )
=
c1 (a2 z + b2 ) + d1 (c2 z + d2 )
(a1 a2 + b1 c2 )z + (a1 b2 + b1 d2 )
=
(c1 a2 + d1 c2 )z + (c1 b2 + d1 d2 )
and
(a1 a2 + b1 c2 )(c1 b2 + d1 d2 ) − (a1 b2 + b1 d2 )(c1 a2 + d1 c2 )

= a1 d1 (a2 d2 − b2 c2 ) + b1 c1 (c2 b2 − d2 a2 )
= (a1 d1 − b1 c1 )(a2 d2 − b2 c2 ) > 0.
Thus f1 ◦ f2 is also a bilinear transformation.

If f1 , f2 and f3 are bilinear transformations, then it is true that
(f1 ◦ f2 ) ◦ f3 = f1 ◦ (f2 ◦ f3 )
as compositions of functions. Obviously, the identity id (z) = z is a bilinear transformation

satisfying
f ◦ id = id ◦ f = f.
az+b dz−b
Finally, we notice from [4, p. 177] that if f (z) = cz+d , then f −1 (z) = −cz+a and therefore,
f ◦ f −1 = f −1 ◦ f = id .
Hence bilinear mappings form a group under composition which completes the analysis0 of the
problem.
Problem 13.10
Proof. The mappings are bilinear, so Theorem 13.11 ensures that they map the unit circle onto
a circle or a line.
(a) Suppose that z = eiθ . Then we have ω = cis (−θ) which means that |ω| = 1.
(b) Since ω(−1) = − 12 and ω has a pole at 1, its image is the vertical line Im z = − 21 .
179
(c) Notice that ω = ω2 ◦ ω1 , where ω1 (z) = z − 2 and ω2 = z1 . It is easily seen that
ω1 (C(0; 1)) = C(−2; 1).
Since 1 6= | − 2|, it follows from the proof of Lemma 13.10 that ω2 (C(−2; 1)) = C(β; R),
where
−2 2 1 1
β= 2 2
=− and R = 2
= .
| − 2| − 1 3 | − 2| − 1 3
Hence we have completed the proof of the problem.
Problem 13.11
Proof. We deduce from Theorem 13.15 that f has the form

z−α
f (z) = eiθ ,
1 − αz
where |α| < 1. Since 0 = f (0) = eiθ α, we have α = 0 and thus f (z) = eiθ z. Since f ′ (0) = eiθ > 0,
θ = 0 which implies that f (z) = z as desired. This completes the proof of the problem.
Problem 13.12
Proof. Note that f1 , f2 : D → D(0; 1) are bijective conformal mappings. Then the mapping
f2−1 : D(0; 1) → D is also bijective and conformal. By Problem 13.7, we see that the mapping
f = f1 ◦ f2−1 : D(0; 1) → D(0; 1) is an automorphism. Clearly, f (0) = f1 (f2−1 (0)) = f1 (z0 ) = 0.
By Problem 3.3 and Proposition 3.5, we obtain
1 f1′ (z0 )
f ′ (0) = f1′ (f2−1 (0)) · (f2−1 )′ (0) = f1′ (f2−1 (0)) · = > 0.
f2′ (f2−1 (0)) f2′ (z0 )
Hence f satisfies the hypotheses of Problem 13.11 so that f (z) ≡ z, i.e., f1 ≡ f2 which ends the
Problem 13.13
Proof. Suppose that D(z1 ; r1 ), D(z2 ; r2 ) are discs and U1 , U2 are half-planes, where r1 , r2 > 0.
Let U be H. As what we have used in Problem 13.7, D1 ∼ D2 means that D1 and D2 are
conformally equivalent.
• Case (i): D(z1 ; r1 ) ∼ D(z2 ; r2 ). Let f be this conformal mapping. Now the mapping
gk (z) = r1k (z − zk ) maps D(zk ; rk ) conformally onto the unit disc D(0; 1), where k = 1, 2.
It is obvious that each gk is a bilinear transformation. Since f = g2−1 ◦ g1 , it follows from
Problem 13.9 that f is also in the form of a bilinear transformation.
• Case (ii): D(z1 ; r1 ) ∼ U1 . Let f be this conformal mapping. It is easily seen that there
exist constants θ and b such that the mapping h(z) = eiθ (z + b) sends U1 conformally onto
U . It deduces from Theorem 13.16 that the mapping φ(z) = z−i z+i is a conformal mapping
of U onto D(0; 1). We know that f = h ◦ φ ◦ g, where g(z) = r11 (z − z1 ). Since the
−1 −1
inverses h−1 and φ−1 are bilinear transformations, Problem 13.9 guarantees that our f is
in the form of a bilinear transformation.
• Case (iii): U1 ∼ U2 . Let f be this conformal mapping. Suppose that h1 (z) = eiθ1 (z + b1 )
and h2 (z) = eiθ2 (z +b2 ) map U1 and U2 conformally onto U respectively, where θ1 , θ2 , b1 , b2
are some constants. Then we have f = h−1 2 ◦ h1 which is a bilinear transformation by
Problem 13.9.
Consequently, any conformal mapping of a half-plane or disc onto another half-plane or disc must
be in the form of a bilinear transformation. This completes the analysis of the problem.
Problem 13.14
Proof. Let g = −f . Then we have
(−a)z + (−b)
g(z) =
cz + d
and (−a)d − (−b)c = −(ad − bc) > 0. It follows from Theorem 13.17 that g is an automorphism
of H. Therefore, f maps H conformally onto the lower half-plane, completing the proof of the
problem.
Problem 13.15
Proof. Let Π1 be the first quadrant. Now Theorem 13.17 shows that an automorphism of H is
of the form
az + b
h(z) = ,
cz + d
√
where ad − bc > 0. We note that f (z) = z is a conformal mapping of H onto Π1 . Similarly,
g(z) = z 2 is a conformal mapping of Π1 onto H. Since the composition of two conformal
mappings is also a conformal mapping, the mapping F = f ◦h◦g : Π1 → Π1 is an automorphism.
Hence we conclude that r
az 2 + b
F (z) = f (h(g(z))) = .
cz 2 + d
Problem 13.16
181
z−i
Proof. We follow the given hint. Let h1 = f −1 ◦eiθ ◦f and h2 = f −1 ◦e−iθ g ◦f , where f (z) = z+i
z−α
and g(z) = eiθ ( 1−αz ) with |α| < 1. By the properties of composition of functions, we obtain
h = f −1 ◦ g ◦ f = f −1 ◦ eiθ ◦ e−iθ g ◦ f = (f −1 ◦ eiθ ◦ f ) ◦ (f −1 ◦ e−iθ g ◦ f ) = h1 ◦ h2 . (13.2)

z−i i(z+1)
Since f (z) = z+i , we have f −1 (z) = −z+1 . Therefore, we see that
z−i
−iθ z+i −α (1 − α)z − (1 + α)i
e g(f (z)) = z−i
=
1− α( z+i ) (1 − α)z + (1 + α)i
and direct computation shows

(1−α)z−(1+α)i
(1−α)z+(1+α)i + 1
h2 (z) = i · (1−α)z−(1+α)i
− (1−α)z+(1+α)i +1
(1 − α)z − (1 + α)i + (1 − α)z + (1 + α)i
=i·
−(1 − α)z + (1 + α)i + (1 − α)z + (1 + α)i
[2 − (α + α)]z + (α − α)i
=i·
(α − α)z + [2 + (α + α)]i
(1 − Re α)z + Im α
= .
(Im α)z + (1 + Re α)
Recall that |α| < 1, so (1 − Re α)(1 + Re α) − (Im α)2 = 1 − [(Re α)2 + (Im α)2 ] = 1 − |α|2 > 0.
In other words, h2 is a bilinear transformation.
If θ = π, then h1 (z) = −z and we obtain from the formula (13.2) that h = −h2 and thus this
insures that h is in the form of a bilinear transformation. Suppose that θ 6= π. Then we have
cos θ 6= −1 and cos 2θ 6= 0. Now simple algebra shows that
eiθ ( z−i
z+i ) + 1
h1 (z) =
−eiθ ( z−i
z+i ) + 1
eiθ (z − i) + z + i
=i·
−eiθ (z − i) + z + i
(1 + eiθ )z + (1 − eiθ )i
=i·
(1 − eiθ )z + (1 + eiθ )i
(2 cos 2θ )z − (2i sin θ2 )i
=i·
(−2i sin 2θ )z + (2 cos θ2 )i
(1 + cos θ)z + sin θ
= .
(− sin θ)z + (1 + cos θ)
As (2 + cos θ)2 + sin2 θ = 5 + 4 cos θ > 0, h1 is in the form of a bilinear transformation. By

Problem 13.9, h is also a bilinear transformation in this case and we have completed the proof
of the problem.
Problem 13.17
Proof.
z−1
(a) Now z+1 = z if and only if z 2 = −1 if and only if z = ±i.
z
(b) Similarly, z+1 = z if and only if z 2 = 0 if and only if z = 0.
Problem 13.18
Proof. By Lemma 13.20, we know that

(z − z2 )(z3 − z1 )
T (z) =
(z − z1 )(z3 − z2 )
is the (unique) bilinear mapping sending z1 , z2 and z3 to ∞, 0 and 1 respectively. Next, Theorem
13.11 implies that the image of the circle or line containing z1 , z2 and z3 under T is either a
circle or a line. Since ∞, 0 and 1 lie on the real axis, the image must be the real axis. Therefore,
T (z4 ) is real if and only if z4 lies on the circle or the line containing z1 , z2 and z3 , completing
Problem 13.19
Proof.
(a) By Theorem 13.23, we conclude that
2(ω − i) −2(z − i)
= .
(ω + 1)(1 − i) (z − 1)(−1 − i)
After simplification, we conclude that ω = − 1z .
(b) By Theorem 13.23 again, we get
2i(ω − i) 2iz
=
iω i(z + i)
which gives ω = z + i.
(c) We know that
(ω − 0)( 13 − ω1 ) ω( 3ω1 1 − 1)
lim = lim 1 ω = 3ω,
ω1 →∞ (ω − ω1 )( 1 − 0)
3 ( ω1 − 1)
ω1 →∞
3
so Theorem 13.23 implies that
(z − i)(2i + i) z−i
3ω = =3·
(z + i)(2i − i) z+i
z−i
which reduces to ω = z+i .
Problem 13.20
Proof. Figure 13.4 shows the conformal mapping f of the region between the two circles (the
light blue part) and the annulus (the light orange part).
183
Figure 13.4: The conformal mapping of the region between the two circles and the annulus.
We know from Theorem 13.15 that

z−α
f (z) = eiθ · (13.3)
1 − αz
is an automorphism of D(0; 1), where |α| < 1 and θ ∈ R. Now it remains to determine α so
that f maps D( 14 ; 14 ) conformally onto D(0; a). Obviously, α 6= 0; otherwise, f = eiθ · id which
is impossible. Now the hypothesis and the fact that f maps C(0; 1) onto itself certainly force
that our f must map C( 14 ; 14 ) onto the circle C(0; a). By choosing θ appropriately in the formula
(13.3), we may assume that f ( 21 ) = a. Since f is conformal, we have f (0) = −a so that
1
−α
2
= e−iθ a = α
1 − α2
4α = 1 + |α|2
√
2 − 4α + 1 = 0 which implies that α = 2 ± 3.
and thus α must be real. Consequently,
√ we have α
Since |α| < 1, we obtain α = 2 − 3 which gives the desired conformal mapping. This completes
Problem 13.21
Proof. If ζ = cosh z, then we can show easily from Proposition 3.5 or from [16, Eqn. (8), p. 201]
directly that
d 1 1 1 1
(cosh−1 z) = d
= =p =√ .
dz dζ (cosh ζ)
sinh ζ cosh2 ζ − 1 z2 − 1
Therefore, we conclude that

Z z
−1 dζ
f (z) = cosh z= p .
0 ζ2 − 1
By [16, Eqn. (8), p. 201] again, we know that cosh−1 z = i cos−1 z. Using [3, Example 1.8.8, p.
91; Exercise 48, p. 93] we can derive the identity sin−1 z + cos−1 z = π2 , so we establish
πi
f (z) = cosh−1 z = − i sin−1 z. (13.4)
2
We note from [4, pp. 187, 188] that sin−1 z maps H conformally onto the semi-infinite strip
{z ∈ C | − π2 < Re z < π2 and Im z > 0}, thus the equation (13.4) ensures that our mapping f
sends H conformally onto the semi-infinite strip {z ∈ C | Re z > 0 and 0 < Im z < π}.b We have
Problem 13.22
Proof. Applying the Schwarz-Christoffel transformation and idea as well as the notations of [4,
Sect. III, pp. 191, 192], we choose a1 = −1, a2 = 1 and a3 > 1. Since the exterior angles of an
isosceles right triangle are 3π 3π π
4 , 4 and 2 , we obtain
3π 3π π
α1 π = , α2 π = and α3 π =
4 4 2
3
which give α1 = α2 = 4 and α3 = 12 . Thus the mapping
Z z
dζ
f (z) = 3 3 1 (13.5)
0 (ζ + 1) 4 (ζ − 1) 4 (ζ − a3 ) 2
is a conformal mapping sends H onto an isosceles right triangle. In fact, we can omit the factor
1
(ζ − a3 ) 2 in the formula (13.5), so we can reduce the formula (13.5) to
Z z
dζ
f (z) = 3
0 (ζ 2 − 1) 4
which is our required mapping. This ends the proof of the problem.
Problem 13.23
Proof. Take a = −1, b = 0, c = 1 and replace d by ∞ in [4, Eqn. (3), p. 189] to get
Z z
dζ
f (z) = p ,
0 ζ(1 − ζ 2 )
p
which sends H conformally onto a rectangle, where the branch of ζ(1 − ζ 2 ) is taken to be the
one that is positive when 0 < ζ < 1, see Figure 13.5 below:
Figure 13.5: The conformal mapping of H onto a rectangle.

b
The reader is suggested to refer to [13, Maps 5.53, 5.54, pp. 173, 174].
185
Recall from [4, p. 191] that the point at infinity is mapped by f onto one of the vertices
of the rectangle. We claim that this rectangle is actually a square. To this end, we know that
f (0) = 0 and
1
f ′ (z) = p >0
z(1 − z 2 )
on the interval 0 < z < 1, so f maps the interval [0, 1] onto the interval [0, A], i.e., f (1) = A or
Z 1
dx
A= p .
0 x(1 − x2 )
As z crosses over the point 1, the argument of f ′ (z) will increase by π2 . Next, we consider the
path of integration [0, ∞]. As z crosses “over” the point ∞ (from ∞ to −∞), the argument of
f ′ (z) will also increase by π2 so that f (∞) = A + iB. Since
Z ∞
dx
A + iB = f (∞) = f (1) + p ,
1 x(1 − x2 )
we get Z ∞
dx
iB = p . (13.6)
1 x(x2 − 1)
Applying the substitution x = y1 to the integral (13.6), it can be shown easily that B = A which
means that the rectangle is in fact a square. This completes the proof of the problem.
CHAPTER 14
The Riemann Mapping Theorem
Problem 14.1
Proof. By the discussion on [4, p. 195], we know that

Z
g(ζ) dζ = 0 (14.1)
C
for any closed curve C in D, so by using similar argument as in the proof of Theorem 8.5, we
conclude that our F is well-defined and analytic in D.
Let z1 , z2 ∈ C and γ be a (smooth) curve from z1 to z2 . Furthermore, suppose that γ1 and
γ2 are curves from z0 to z1 and z2 respectively. Then γ1 + γ − γ2 forms a closed curve and so
we observe from the equation (14.1) that
Z Z Z
g(ζ) dζ − g(ζ) dζ = − g(ζ) dζ
γ1 γ2 γ
or equivalently, Z
F (z2 ) − F (z1 ) = g(ζ) dζ. (14.2)
γ
If we write g = u + iv and z(t) = x(t) + iy(t), then we apply Definition 4.3 to the expression
(14.2) to get
Z t2
F (z2 ) − F (z1 ) = (u − iv)ż(t) dt
t1
Z t2
= (u − iv)( dx + i dy)
t1
Z t2 Z t2
= (u dx + v dy) + i (u dy − v dx), (14.3)
t1 t1
where t1 and t2 are the values such that z1 = z(t1 ) and z2 = z(t2 ).
If Re F (z) = C1 for some constant C1 on γ, then the representation (14.3) implies that
Z t2
(u dx + v dy) = 0 (14.4)
t1
187
Chapter 14. The Riemann Mapping Theorem 188
and if we represent g = u + iv in the vector form (u, v), then the result (14.4) means that the
velocity vector (u, v) is orthogonal to the tangle vector (x′ (t), y ′ (t)), completing the proof of the
problem.
Problem 14.2
Proof. It follows from the representation (14.3) that if Im F (z) = C2 for a constant C2 , then we
have Z t2
(u dy − v dx) = 0
t1
which means that the velocity vector (u, v) is orthogonal to the vector (−y ′ (t), x′ (t)). Since
(x′ (t), y ′ (t)) · (−y ′ (t), x′ (t)) = 0, the flow g is parallel to the tangent vector (x′ (t), y ′ (t)), i.e., g
is tangent to Im F (z) = C2 . This completes the proof of the problem,
Problem 14.3
Proof.
(a) By Problem 14.1, we take z0 = 0 to get

Z z Z z
ζ2 z z2
F (z) = g(ζ) dζ = ζ dζ = = . (14.5)
0 0 2 0 2
Let c ∈ R. If z = x + iy, then z 2 = x2 − y 2 + 2ixy. By Problem 14.2, the streamlines of g

is Im F (z) = c. Therefore, we follow from the expression (14.5) that the streamlines of g
are
xy = c.
(b) For simplicity, we take z0 = 1 so that

Z z Z z
dζ
F (z) = g(ζ) dζ = = log z = log |z| + iArg z.
1 1 ζ
In this case, Im F (z) = Arg z. Hence the streamlines of g are rays from the origin.
Problem 14.4
Proof. We follow the given hint. Let C be a real constant. Let D be the exterior of D(0; 1) and
for simplicity, we pick I = [0, 1] to be the interval for consideration. Suppose that F : D → C \ I
is a conformal mapping with F (z) ∼ z as z → ∞. Thus we may suppose that
A1 A2
F (z) = z + A0 + + 2 + ··· ,
z z
189
where Ak ∈ C for all k = 0, 1, 2, . . .. If z = eiθ with θ ∈ [−π, π], then we notice that
Ak z −k = (Re Ak + iIm Ak )(cos kθ − i sin kθ)

= (Re Ak cos kθ + Im Ak sin kθ) + i(−Re Ak sin kθ + Im Ak cos kθ). (14.6)
Now Im F (eiθ ) = C, so the expression (14.6) implies that

∞
X
sin θ + Im A0 + (−Re Ak sin kθ + Im Ak cos kθ) = C. (14.7)
k=1
1 iz
By the identities sin z = 2i (e − e−iz ) and cos z = 12 (eiz + e−iz ), we can rewrite (14.7) as
∞
X
ck eikθ = C, (14.8)
−∞
where 
 Im Ak + iRe Ak

 , if k ≥ 2;

 2





 Im A1 − (1 − Re A1 )i

 if k = 1;

 2




ck = Im A0 if k = 0; (14.9)





 Im A1 + (1 − Re A1 )i



 if k = −1;

 2





 Im A|k| − iRe A|k| ,

if k ≤ −2.
2
We apply the factsa 
Z π  1, if n = 0;
1
einx dx =
2π −π 
0, if n = ±1, ±2, . . .
to the representation (14.8) to get

 C, if k = 0;
ck =

0, if k ∈ Z \ {0}.
Therefore, the definition (14.9) asserts that Re A1 = 1, Re Ak = 0 for all k ≥ 2 and Im Ak = 0

for all k ≥ 1. In other words, we have established that
1
F (z) = z + A0 + .
z
Remark 14.1
In fact, the mapping in Problem 14.5 is classically called the Joukowsky transform, see
[3, Example 7.1.8, pp. 408-410] or [9, Exercise 15, pp. 66, 67].
a
See [22, Eqn. (61) & (62), pp. 185, 186]
Problem 14.5
Proof.
(a) Let f (z) = 2z + z1 and let D and D ′ be the exteriors of D(0; 1) and the said ellipse
respectively. Since f ′ (z) = 2 − z12 , f ′ (z) = 0 if and only if z = ± √12 ∈
/ D. Thus f is
conformal throughout D by Theorem 13.4. Take z = eiθ , so
f (eiθ ) = 2(cos θ + i sin θ) + (cos θ − i sin θ) = 3 cos θ + i sin θ
which means that n o

x2
f (C(0; 1)) = (x, y) + y2 = 1 .
9
As a conformal map, f is continuous on the connected set D so that f (D) is connected.
In addition, since f (z) → ∞ as z → ∞, f must map D into D ′ . It remains to show that
the mapping f : D → D ′ is surjective, but it is easy to see because if (x, y) ∈ D ′ satisfies
x2 2 x
9 + y > 1, then the point z = 3 + iy lies in D. Hence the f is actually a surjective
conformal mapping.
(b) Without loss of generality, we let the real line segment be [−2, 2]. By [4, Example 2, p.
197], g(z) = z + 1z maps D conformally onto C \ [−2, 2]. Thus the composition
g ◦ f −1 : D ′ → C \ [−2, 2]
is a desired conformal mapping.

Consequently, we have completed the proof of the problem.
Problem 14.6
Proof. Since f (z0 ) ∈ D(0; 1), Theorem 13.15 implies that the mapping of the form
z − f (z )
0
h(z) = eiθ (14.10)
1 − f (z0 )z
is an automorphism of D(0; 1), where θ ∈ R. By Problem 13.7, the composition g = h ◦ f is a
conformal mapping of R onto D(0; 1). Obviously, the expression (14.10) gives
g(z0 ) = h(f (z0 )) = 0.
Furthermore, direct computation gives
[1 − f (z0 ) · f (z)]f ′ (z) − [f (z) − f (z0 )] · [−f (z0 ) · f ′ (z)]

g′ (z) = eiθ ·
[1 − f (z0 ) · f (z)]2
so that
f ′ (z0 )
g′ (z0 ) = eiθ · .
1 − |f (z0 )|2
Now if we take θ = −Arg f ′ (z0 ), then we obtain the desired result that g′ (z0 ) > 0. We end the
191
Problem 14.7
Proof. We define g : R → U by g(z) = f (z). We claim that g is analytic in R. To this end, let
z = x + iy ∈ R and f (z) = u(x, y) + iv(x, y), where u and v are real-valued functions. Then we
have
g(z) = f (z) = u(x, −y) + iv(x, −y) = u(x, −y) − iv(x, −y) = U (x, y) + iV (x, y),
where U (x, y) = u(x, −y) and V (x, y) = −v(x, −y). Since f is analytic in R, Theorem 6.6
and Corollary 2.10 together imply that u and v have C 1 partial derivatives. In particular,
gx = Ux + iVx and gy = Uy + iVy are continuous. By Proposition 3.1, we have
∂u(x, y) ∂v(x, y) ∂u(x, y) ∂v(x, y)
= and =−
∂x ∂y ∂y ∂x
which give
∂U (x, y) ∂u(x, −y) ∂v(x, −y) ∂V (x, y)
= = =
∂x ∂x ∂(−y) ∂y
and
∂V (x, y) ∂(−v(x, −y)) ∂u(x, −y) ∂U (x, y)
= = =−
∂x ∂x ∂(−y) ∂y
Hence it follows from Proposition 3.2 that g is analytic at z, so the claim is true.
Since R is symmetric respect to R, it is easy to see that the map z 7→ z is bijective in R.
Since f : R → U is the Riemann mapping, it is bijective. Finally, since g is the composition of
bijective maps, it is also bijective. By Definition 13.9, g is a conformal mapping. Recall that z0
is real, thus we obtain
g(z0 ) = f (z0 ) = f (z0 ) = 0.
Furthermore, for every ω ∈ R, the definition of analyticity of g guarantees that
g(z) − g(ω)
g′ (ω) = lim
z→ω z−ω
f (z) − f (ω)
= lim
z→ω z−ω
f (z) − f (ω)
= lim
z→ω z−ω
h f (z) − f (ω) i
= lim
z→ω z−ω
h f (ζ) − f (ω) i
= lim
ζ→ω ζ −ω
= f ′ (ω). (14.11)
Particularly, if we take ω = z0 , then we deduce from the expression (14.11) and the hypotheses
that g ′ (z0 ) = f ′ (z0 ) = f ′ (z0 ) > 0. By the Riemann Mapping Theorem, the mapping f is unique,
so we conclude that f (z) = g(z) which means f (z) = f (z) for all z ∈ R, as desired. This
Problem 14.8
Proof. By Theorem 13.16, every conformal mapping f of H onto D(0; 1) is of the form
z − α
f (z) = eiθ , (14.12)
z−α
where Im α > 0.
(a) Since f (−1) = 1, f (0) = i and f (1) = −1, we deduce from the formula (14.12) that
1 + α α 1 − α
eiθ = 1, eiθ · = i and eiθ = −1. (14.13)
1+α α 1−α
Suppose that α = a + ib. Then the first and the third equations (14.13) yield that
a2 + b2 = 1. Besides, the first two equations (14.13) imply that
a − b + 1 = 0. (14.14)
Finally, the last two equations (14.13) establish that
a + b − 1 = 0. (14.15)
Solving the equations (14.14) and (14.15), we get a = 0 and b = 1, i.e., α = i. Using the
second expression (14.13), we find that eiθ = −i. Hence the desired conformal mapping is
given by
z−i iz + 1
f (z) = −i · =− .
z+i z+i
(b) The condition f (i) = 0 implies that α = i. Next, the condition f (1) = 1 implies that
eiθ = i. Hence the required conformal mapping is
z−i iz + 1
f (z) = i · = .
z+i z+i
Problem 14.9
Proof. If R = C, then the mapping f : C → C defined by f (z) = z − z1 + z2 is conformal at

every z ∈ C because f ′ (z) = 1 by Theorem 13.4. Since f (z1 ) = z2 , it satisfies our requirements.
Next, we suppose that R 6= C. Since R is simply connected, the Riemann Mapping Theorem
ensures that there exist conformal mapping ϕ1 : R → D(0; 1) and ϕ2 : R → D(0; 1) such that
ϕ1 (z1 ) = ϕ2 (z2 ) = 0. By Problem 13.7, the mapping ϕ = ϕ−1 2 ◦ ϕ1 : R → R is a conformal
mapping and we have
ϕ = ϕ−1 −1
2 (ϕ1 (z1 )) = ϕ2 (0) = z2 ,
completing the analysis of the problem.
Problem 14.10
193
Proof. Assume that f : C → R was a conformal mapping. By the Riemann Mapping Theorem,
there exists a conformal mapping g : R → D(0; 1). By Problem 13.7, h = g ◦ f : C → D(0; 1)
is conformal, i.e., h is a bounded entire function which certainly contradicts Theorem 5.10
(Liouville’s Theorem). This completes the proof of the problem.
Problem 14.11
Proof.
(a) Note that G = {g : R → U | g is analytic and g ′ (z0 ) > 0}. By the proof of Part (B) of the
Riemann Mapping Theorem [4, p. 202], for every g ∈ G, we see easily that
1
g′ (z0 ) = |g ′ (z0 )| < ,
δ
where δ is a positive number independent of g such that D(z0 ; δ) ⊂ R. In conclusion, we
find that
1
sup g′ (z0 ) = M ∗ ≤ .
g∈G δ
(b) Following the same argument as the proof of Part (B), we can verify that there exists a
function Φ ∈ G such that Φ′ (z0 ) = M ∗ . Next, if Φ(z0 ) = α with 0 < |α| < 1, then the
map gb : R → U defined by
Φ(z) − α
gb(z) =
1 − αΦ(z)
is clearly analytic, i.e., gb ∈ G. Besides, we have
Φ′ (z0 )
gb′ (z0 ) = > Φ′ (z0 ),
1 − |α|2
a contradiction. Therefore, we conclude that Φ(z0 ) = 0. Let ϕ : R → U be the Riemann
mapping function with ϕ(z0 ) = 0 and ϕ′ (z0 ) = M > 0. Notice that F ⊆ G, so it happens
that Φ′ (z0 ) = M ∗ ≥ M > 0. Now we consider the map
f = Φ ◦ ϕ−1 : U → U.
As a composition of two analytic functions, f is also analytic in U and |f (z)| < 1 on U .

Simple algebra shows that
f (0) = Φ(ϕ−1 (0)) = Φ(z0 ) = 0.
Consequently, our f satisfies the hypotheses of Theorem 7.2 (Schwarz’s Lemma), so we see
that |f (z)| ≤ |z| and |f ′ (0)| ≤ 1.
On the other hand, by combining Proposition 3.5 and Problem 3.3, we know that
Φ′ (z0 ) M∗
f ′ (0) = = ≥ 1.
ϕ′ (z0 ) M
Hence according to Theorem 7.2 (Schwarz’s Lemma), it is true that f (z) = z, i.e., f = id .
By the definition, we finally get
Φ=ϕ
so that Φ is injective.
Problem 14.12
Proof.
(a) By similar idea as the proof of Problem 13.6, the mapping f : S → U defined by
−(1 + i)z 2 + 2(1 − i)z − (1 + i)

f (z) =
−(1 − i)z 2 + 2(1 + i)z − (1 − i)
1−i
z 2 − 2( 1+i )z + 1
= 1−i 2 1−i
1+i z − 2z + 1+i
z 2 + 2iz + 1
=
−iz 2 − 2z − i
z 2 + 2iz + 1
=i· 2 (14.16)
z − 2iz + 1
is conformal. Since S and U are Jordan regions by Definition 14.2, we follow from Theorem
14.3 (The Carathéodory-Osgood Theorem) that f can be extended to a homeomorphism
between S and U .
√
(b) Since the roots of z 2 − 2i + 1 = 0 are z = (1 ± 2)i ∈ / S, the representation (14.16)
ensures that f is analytic on S. However, it follows from the representation (14.16) that
the inverse f −1 : U → S has the form
√
−1 (−1 + iz) + 2 − 2z 2
f (z) =
i−z
which has a simple pole at −i ∈ U , so f −1 is not analytic on U .

Problem 14.13
Proof. Figure 14.1 shows the shape of a Norman window N . Without loss of generality, we may
suppose that 0 lies on ∂N .
Figure 14.1: The shape of a Norman window N .

195
Assume that f : U → N was analytic and surjective. By Theorem 7.1 (The Open Mapping
Theorem)b , f will map a z0 ∈ C(0; 1) onto the origin. Now we may further suppose that z0 = 1.
If f ′ (1) 6= 0, then Theorem 13.4 implies that f maps the rays I1 = {1 + it | 0 ≤ t ≤ ∞} and
I2 = {1 − it | 0 ≤ t ≤ ∞} onto two curves whose tangent lines form a straight angle,
∠f (I1 ), f (I2 ) = π, (14.17)
but it is impossible because the angle at 0 is just π2 . Next, if f ′ (1) = 0, then Theorem 13.7 will
ensure that, instead of the value (14.17), we have
∠f (I1 ), f (I2 ) = kπ,
where k is the least positive integer such that f (k) (1) 6= 0. However, this is also impossible. Hence
no such analytic and surjective map f exists. This completes the proof of the problem.
b
See also the proof of Problem 13.8.
CHAPTER 15
Maximum-Modulus Theorems for Unbounded
Domains
Problem 15.1
Proof. Without loss of generality, we may assume that f is non-constant. We basically follow
the proof of Theorem 15.1. Suppose first that D = {z ∈ C | Re z > 0}, f (z) ≪ 1 on ∂D,
f (z) ≪ log z for all z ∈ D and z0 ∈ D. Consider the auxiliary function
f N (z)
h(z) = , (15.1)
z+1
where N ∈ N. By the hypothesis, we have |h(z)| ≤ 1 along the imaginary axis. In addition, if
z = x + iy ∈ D, then x > 0 and so
p p
|z + 1| = x2 + y 2 + 2x + 1 ≥ x2 + y 2 = |z|.
Therefore, for all z ∈ D with |z| = R > 0, we have

N
| log z|N [(log R)2 + (Arg z)2 ] 2
|h(z)| ≤ ≤ . (15.2)
R R
Since −π ≤ Arg z ≤ π, we can take R large enough so that the bound (15.2) reduces to
√
2N (log R)N
|h(z)| ≤ ≤1 (15.3)
R
for all z in the right semi-disc DR = {z ∈ D | |z| ≤ R}. Since h is certainly analytic in DR and
DR is compact, Theorem 6.13 (The Maximum Modulus Theorem) implies that |h(z0 )| ≤ 1. By
the definition (15.1), we achieve
f N (z0 )
≤1
z0 + 1
or equivalently,
1
|f (z0 )| ≤ |1 + z0 | N .
Since N is arbitrary, we take N → ∞ to conclude that |f (z0 )| ≤ 1 as desired.
197
Chapter 15. Maximum-Modulus Theorems for Unbounded Domains 198
Next, we suppose that D is an arbitrary region. Then we consider

f (z) − f (a)
g(z) = ,
z−a
where a is a fixed point in D. By Proposition 6.7, g is C-analytic in D. Furthermore, the
hypothesis implies that, for large |z|,
|f (z) − f (a)| | log z| + | log a|
|g(z)| = ≤
|z − a| |z| − |a|
so that g(z) → 0 as z → ∞. Consequently, there exists a constant M > 0 such that |g(z)| ≤ M
on D. Similar to the previous paragraph, we set DR = {z ∈ D | |z| ≤ R} and h(z) = f N (z)g(z),
where R > 0 and N ∈ N. Thus it must be true that |h(z)| ≤ M on ∂DR for sufficiently large
R. Hence it follows from Theorem 6.13 (The Maximum Modulus Theorem) that |h(z0 )| ≤ M
for every z0 ∈ D. If g(z0 ) 6= 0, then we get
1 1
h(z) N MN
|f (z0 )| ≤ ≤ 1 .
g(z) |g(z0 )| N
By taking N → ∞, we get |f (z0 )| ≤ 1. According to Theorem 6.9 (The Uniqueness Theorem),
since f is nonconstant, the set S = {z ∈ D | g(z) = 0} is discrete, so we must have
|f (z)| ≤ 1 (15.4)
on D \ S. Finally, the analyticity of f ensures that the bound (15.4) holds in D. Hence our
conclusion of Theorem 15.1 is still valid in this case.
We see that our argument works because of the inequality (15.3). Thus if f (z) ≪ P (log z)
in D, where P (ω) is a polynomial in ω of degree n ≥ 1, then there exists a positive constant M
such that
M N (log R)nN
|h(z)| ≤ ≤1
R
holds in DR = {z ∈ D | |z| ≤ R} for sufficiently large enough R. This completes the proof of the
problem.
Problem 15.2
Proof. We consider
π π
D1 = {z ∈ C | − < Arg z < } and F (z) = exp(z 2 ).
4 4
On z = r exp(± π4 i), we see that z 2 = ±ir 2 so that |F (z)| = | exp(±ir 2 )| = 1, i.e., F is bounded
on ∂D1 . However, it is easy to check that
2
F (x) = ex → ∞
as x → ∞. Consequently, F is unbounded in D1 .
Suppose that F is the class of all functions f analytic in D1 such that f is bounded on ∂D1
but f (z) is unbounded in D1 . Let f (z) be an analytic function “smaller” than F (z) in D1 in
the sense that for every ǫ > 0, there exists a constant Aǫ such that
|f (z)| ≤ Aǫ exp(ǫ|z|2 )
199
for all z ∈ D1 . Now Corollary 15.5 with α = π2 implies immediately that this function f must be
bounded in the region D1 . Combining this fact and the observation in the previous paragraph,
2
we conclude that F ∈ F but f ∈ / F . Hence F (z) = ez is the “smallest” analytic function in
this sense.
3π
By replacing z by e−i 4 z, we see at once that the function
3π 3π 2
G(z) = F (e−i 4 z) = exp(e−i 2 z 2 ) = eiz
is the desired “smallest” non-constant analytic function in D, completing the proof of the prob-
lem.
Problem 15.3
Proof. Let F (z) = exp(ez ) and D = {x + iy | x ∈ R and − π2 < y < π2 }. Clearly, F is C-analytic
in D. If z = x ± i π2 , then we have
ez = ±iex .
In other words, ez maps ∂D onto the imaginary axis which implies that
|F (z)| = | exp(ez )| = | exp(±iex )| = 1
on ∂D, proving the first assertion. However, we note that
F (log x) = exp(elog x ) = ex → ∞
as x → ∞. In other words, F is unbounded in the region D.

Suppose that f (z) is an analytic function “smaller” than F (z) in D in the sense that for
every ǫ > 0, there exists a constant Aǫ such that
|f (z)| ≤ Aǫ exp(ǫ|ez |)
for all z ∈ D. In this case, the analytic function g(z) = f (log z) satisfies
|g(z)| ≤ Aǫ eǫ|z|
in the region D. By Theorem 15.4 (The Phragmén-Lindelöf Theorem), g(z) and then f (z) are
also bounded throughout D.
Suppose that F is the class of all functions f analytic in D such that f is bounded on ∂D
but f (log z) is unbounded in D. Then the first assertion says immediately that F ∈ F but the
previous paragraph means that any analytic function f “smaller” than F implies that f ∈ / F.
Problem 15.4
Proof. Assume that there were constants A and B such that
|g(z)| ≤ A exp(|z|B ) (15.5)

Chapter 15. Maximum-Modulus Theorems for Unbounded Domains 200
in C.a As given by the hint, we choose a positive integer N such that N > 2B and then we
π
divide the plane into N wedges of equal angles α = N . Let W1 , W2 , . . . , WN be these wedges.
Since g is entire, it is C-analytic in each Wk . Next, the hypothesis ensures that
|g(z)| ≤ Mk (15.6)
on ∂Wk for some positive constant Mk . Furthermore, it is clear from the assumption (15.5) that
N π
|g(z)| ≤ A exp(|z|B ) ≤ A exp(|z| 2 ) = A exp(|z| 2α ). (15.7)
π π 2α
Given ǫ > 0. Then the inequality |z| 2α ≤ ǫ|z| α holds if and only if |z| ≥ ǫ− π . Thus the
inequality (15.7) gives
π
|g(z)| ≤ A exp(ǫ|z| α ) (15.8)
2α 2α
if z ∈ Wk and |z| ≥ ǫ− π . If z ∈ Wk and |z| < ǫ− π , then since ex is strictly increasing, it is
easy to see from the inequality (15.7) that
|g(z)| ≤ A exp(ǫ−1 ). (15.9)
Let Aǫ = max(A, A exp(ǫ−1 )). We conclude from the inequalities (15.8) and (15.9) that
π
|g(z)| ≤ Aǫ exp(ǫ|z| α )
for all z ∈ Wk . In other words, our function g satisfies the hypotheses of Corollary 15.5. Hence
the inequality (15.6) holds throughout Wk and then g is bounded in C. Therefore, Theorem
5.10 (Liouville’s Theorem) implies that g is constant which contradicts our hypothesis. This
completes the analysis of the problem.
a
By Definition 16.12, g is of finite order.
CHAPTER 16
Harmonic Functions
Problem 16.1
Proof. Since f is analytic, Theorem 16.2 says that u and v are harmonic, i.e, uxx + uyy = 0 and
vxx + vyy = 0. Since
(u + v)xx + (u + v)yy = (uxx + uyy ) + (vxx + vyy ) = 0,
u+v is harmonic by Definition 16.1. By direct computation, we have (uv)xx = uxx v+2ux vx +uvxx
and (uv)yy = uyy v + 2uy vy + uvyy whose sum is
(uv)xx + (uv)yy = 2(ux vx + uy vy ). (16.1)
By Proposition 3.1, the equation (16.1) reduces to 0, so uv is also harmonic by Definition 16.1,
Problem 16.2
Proof. Suppose that u is a harmonic function. By Theorem 16.3, ux is the real part of an
analytic function f . As an analytic function, f is infinitely differentiable. Hence u is also
infinitely differentiable. Let g = ux . Then we deduce from [28, Theorem 15.12, p. 146] that
∂ ∂ ∂
gxx + gyy = uxx + uyyx = uxx + uxyy = (uxx + uyy ) = 0.
∂x ∂x ∂x
Consequently, ux is harmonic. Similarly, uy is also harmonic and we have completed the proof
of the problem.
Problem 16.3
201
Chapter 16. Harmonic Functions 202
Proof. Let f = u2 . Obviously, we have fx = 2uux and fxx = 2(uuxx + u2x ). Similarly, we also
have fyy = 2(uuyy + u2y ). Thus we obtain
fxx + fyy = 2u(uxx + uyy ) + 2(u2x + u2y ) = 2(u2x + u2y ) ≥ 0
so that fxx + fyy = 0 if and only if ux = uy = 0 if and only if u is a constant by Theorem 1.10,
a contradiction. This ends the proof of the problem.
Problem 16.4
2x
Proof. Let u = log(x2 + y 2 ). Then we get ux = x2 +y 2
and
2(y 2 − x2 )
uxx = . (16.2)
(x2 + y 2 )2
Similarly, we get
2(x2 − y 2 )
uyy = . (16.3)
(x2 + y 2 )2
Thus the sum of the formulas (16.2) and (16.3) certainly imply that uxx + uyy = 0, i.e., u is
harmonic in D = C \ {0}.
Assume that u was the real part of an analytic function f = u + iv in D. We note that
u = log(x2 + y 2 ) = 2 log |z|, so we write f = 2g, where g = log |z| + i v2 is analytic in D.
Therefore, g is an analytic branch of log z up to an imaginary constant in D. Next, recall from
Problem 8.8 that an analytic branch of log z can be defined with 0 < Arg z < 2π in the plane
C \ [0, ∞). This fact forces that g is not continuous on [0, ∞) which contradicts the assumption
that g is analytic in D. Hence no such g and then f exists and we complete the analysis of the
problem.
Problem 16.5
Proof.
(a) Note that x = r cos θ and y = r sin θ, so we obtain
∂x ∂x ∂y ∂y
= cos θ, = −r sin θ, = sin θ and = r cos θ.
∂r ∂θ ∂r ∂θ
These imply that
∂u ∂u ∂x ∂u ∂y ∂u ∂u
= · + · = cos θ + sin θ (16.4)
∂r ∂x ∂r ∂y ∂r ∂x ∂y
and thus
∂2u 2
2 ∂ u ∂2u 2
2 ∂ u
= cos θ + 2 cos θ sin θ + sin θ . (16.5)
∂r 2 ∂x2 ∂x∂y ∂y 2
Similarly, we have
∂u ∂u ∂u
= −r sin θ + r cos θ
∂θ ∂x ∂y
203
so that
∂2u ∂u ∂u 2 ∂2u 2
2 2 ∂ u 2 ∂ u
= −r cos θ + sin θ + r sin θ − 2 cos θ sin θ + cos θ . (16.6)
∂θ 2 ∂x ∂y ∂x2 ∂x∂y ∂y 2
Combining the expressions (16.4), (16.5) and (16.6), we conclude that
1 1
urr + ur + 2 uθθ = uxx + uyy . (16.7)
r r
Now if u(r, θ) is a harmonic function depending on r alone, then uθθ = 0 so that Laplace’s
equation (16.7) becomes
1
urr + ur = 0 (16.8)
r
as desired.
(b) The differential equation (16.8) can be written in the form (rur )r = 0 which means that
rur = a for some constant a, or equivalently ur = ar . By integration, we establish
u(r, θ) = a log r + b
for some constant b.
Problem 16.6
Proof. We start with the following form of Poisson Formula [1, Theorem 22, p. 168]:
Z
1 1 − |a|2 dζ
U (a) = 2
U (ζ) (16.9)
2π |ζ|=1 |ζ − a| iζ
where U is C-harmonic in D(0; 1) and a ∈ D(0; 1). By Theorem 13.16, the map ϕ : H → D(0; 1)
given by
z−i
ϕ(z) = (16.10)
z+i
is conformal and surjective. Besides, it is easy to check that ϕ maps ∂H = R onto C(0; 1) \ {1}.
Now we consider the map U = u ◦ ϕ−1 : D(0; 1) → R. Since H is simply connected, Theorem
16.3 ensures that u = Re f for some analytic function f on H. Write f = u + iv so that
f ◦ ϕ−1 = u ◦ ϕ−1 + iv ◦ ϕ−1 .
Since both f and ϕ−1 are analytic, its composition f ◦ ϕ−1 : D(0; 1) → C is also analytic
in D(0; 1) and we follow from Theorem 16.2 that u ◦ ϕ−1 is harmonic in D(0; 1). Since u is
continuous on R and bounded, u ◦ ϕ−1 can be made to be continuous at 1 so that it is actually
continuous on C(0; 1). In other words, u◦ϕ−1 is C-harmonic in D(0; 1). Thus we put U = u◦ϕ−1
into the formula (16.9) to get
Z
−1 1 1 − |a|2 dζ
u(ϕ (a)) = 2
u(ϕ−1 (ζ)) . (16.11)
2π |ζ|=1 |ζ − a| iζ
Put ζ = ϕ(t) and a = ϕ(z) into the formula (16.11) and then using the expression (16.10) and
2i
the fact ϕ′ (t) = (t+i)2 to establish
Z
−1 1 1 − |ϕ(z)|2 2
u(x + iy) = u(z) = u(ϕ (a)) = · u(t) · 2 dt. (16.12)
2π R |ϕ(t) − ϕ(z)|2 t +1
x2 +(y−1)2
Since | z−i 2
z+i | = x2 +(y+1)2
and
t−i z−i 2 |2y + 2(t − x)i|2 4[(t − x)2 + y 2 ]

|ϕ(t) − ϕ(z)|2 = − = = ,
t+i z+i (1 + t2 )[x2 + (y + 1)2 ] (1 + t2 )[x2 + (y + 1)2 ]
the expression (16.12) becomes

Z ∞h
1 x2 + (y − 1)2 i (1 + t2 )[x2 + (y + 1)2 ] 2
u(x + iy) = 1− 2 · · u(t) · 2 dt
2π −∞ x + (y + 1)2 4[(t − x)2 + y 2 ] t +1
Z ∞
1 (y + 1)2 − (y − 1)2
= u(t) dt
4π −∞ (t − x)2 + y 2
Z
1 ∞ y · u(t)
= dt
π −∞ (t − x)2 + y 2
which is our desired result. We have completed the proof of the problem.
Problem 16.7
Proof. Following the idea of [4, Example i, p. 232], since z 3 is analytic in D(0; 1), Re (z 3 ) is
harmonic in D(0; 1) by Theorem 16.2. Note that
Re (z 3 ) = x3 − 3xy 2 . (16.13)
Since x2 + y 2 = 1, the equation (16.13) reduces to Re (z 3 ) = 4x3 − 3x on the boundary C(0; 1).
Hence if we choose
1 1
u(x, y) = [Re (z 3 ) + 3x] = (x3 − 3xy 2 + 3x),
4 4
then it is harmonic in D(0; 1) by Definition 16.1 and satisfies u(x, y) = x3 on C(0; 1). This ends
Problem 16.8
Proof. By [4, Example ii, pp. 232, 233], we know that
3 1 z − 1
u(z) = − Arg ,
2 π z+1
where |z| ≤ 1. If u(x, y) = k, where k ∈ [0, 1], then we get

z − 1 3
θk = Arg =π −k . (16.14)
z+1 2
3π π
It is clear that as k runs through [0, 1], the formula (16.14) indicates that θk runs from 2 to 2.
Suppose that z = x + iy. Simple algebra gives
z−1 x2 + y 2 − 1 2y
= +i .
z+1 (x + 1)2 + y 2 (x + 1)2 + y 2
205
By the definition, we know that

2y
tan θk =
x2 + y2 − 1
and then the value (16.14) implies that
2y 1
=
x2 2
+y −1 tan kπ
or equivalently
x2 + y 2 − (2 tan kπ)y − 1 = 0. (16.15)
It is trivial that (−1, 0) and (1, 0) always lie on the locus represented by the equation (16.15).
Furthermore, if k = 12 , then we conclude from the equation (16.15) that y = 0 and x = ±1. If
k = 0 or 1, then the equation (16.15) reduces to the unit circle C(0; 1).
Figure 16.1: The level curves of u(x, y) = k for k ∈ (0, 21 ) ∪ ( 12 , 1).
Next, suppose that k ∈ (0, 12 ) ∪ ( 21 , 1). Then the centre and the radius of the circle (16.15)
are given by
p = (0, tan kπ) and | sec kπ|
respectively. In fact, for 0 < k < 21 , we know that tan kπ > 0 and thus the centre lies on
the positive imaginary axis which implies that the level curve u(x, y) = k is the lower circular
segment of the circle C(p, sec kπ). This is exactly the dark red arc in Figure 16.1 lying inside
the unit disc D(0; 1). Similarly, for 21 < k < π, then tan kπ < 0 so that its centre lies on the
negative imaginary axis. In this case, the level curve u(x, y) = k is the upper circular segment
of the circle C(p, | sec kπ|), see the purple arc lying inside the unit disc D(0; 1) in Figure 16.1.
Problem 16.9
Proof. By Problem 16.6 with the function


 1, if t > 0;
u(t) =

0, if t < 0,
we get Z ∞
1 y
u(x + iy) = dt. (16.16)
π 0 (t − x)2 + y 2
Using the substitution y tan u = t − x and the identity tan−1 (−x) = − tan−1 x for all x ∈ R, we
can reduce the expression (16.16) to the following form
1 1 x
u(x + iy) = + tan−1
2 π y
which is our desired harmonic function. This ends the proof of the problem.

Problem 16.10
Proof. Let S be the semi-infinite strip {z = x + iy | − π2 < x < π2 and y > 0}. The graph of the
the temperature problem with the prescribed boundary values is shown in Figure 16.2. We note
that this is a Dirichlet problem for the region S with its boundary in the z plane. Our method
of solution is to obtain a new Dirichlet problem for the upper half-plane H with its boundary in
the ω plane.
Figure 16.2: The temperature problem with the prescribed boundary values.
Recall from §13.3 that the function ω = f (z) = sin z maps S conformally onto H and the
interval [− π2 , π2 ] is mapped onto [−1, 1]. On z = − π2 + iy with y > 0, we have
1
sin z = − cos(iy) = − (ey + e−y ).
2
Therefore, f maps the vertical line z = − π2 + iy onto (−∞, −1). Similarly, it can be shown easily
that sin z maps the vertical line z = π2 + iy with y > 0 onto (1, ∞). Consequently, the mapping
f (z) = sin z transforms this boundary value problem into another boundary value problem in
H.
207
Stimulating by [4, Example ii, pp. 232, 233], we consider the mapping
ζ = log(ω 2 − 1) = log |ω 2 − 1| + iArg (ω 2 − 1)
which is analytic in H. By the definition, we know that
Arg (ω 2 − 1) = Arg (ω + 1) + Arg (ω − 1) = θ2 + θ1 ,
where θ1 and θ2 are shown in Figure 16.3 below.
Figure 16.3: The angle Arg (ω 2 − 1).
Clearly, the function Im ζ = Arg (ω 2 − 1) is harmonic in H by Theorem 16.2. In addition,

since Arg (ω 2 − 1) = Arg (ω + 1) + Arg (ω − 1), we actually obtain


 2π, if ω ∈ (−∞, −1);



Arg (ω 2 − 1) = π, if ω ∈ (−1, 1); (16.17)





0, if ω ∈ (1, ∞).
These facts show that, in H, Im ζ = Arg (ω 2 − 1) is harmonic with the prescribed boundary
values (16.17). Hence the composition function
1
u(x, y) = Arg (sin2 z − 1)
π
is harmonic in S with the prescribed boundary values
 π
 2, if z = − 2 + iy with y > 0;




u(x, y) = 1, if z ∈ (−1, 1);





0, if z = π2 + iy with y > 0.
Problem 16.11
Proof. If the number of zeros of f (z) = ez − P (z) was finite, then we deduce from Theorem
16.13 that f has the form
ez − P (z) = P1 (z)eP2 (z) (16.18)
for some polynomials P1 and P2 . If deg P2 ≥ 2, then the equation (16.18) implies that
ez − P (z)
lim P1 (z) = lim = 0.
z→∞ z→∞ eP2 (z)
By Problem 1.26, P1 (z) ≡ 0. In this case, the expression (16.18) becomes ez = P (z) which
is impossible. If deg P2 = 0, then P2 is a constant and so the expression (16.18) yields that
ez = Q(z) for some polynomial Q, a contradiction. Hence we must have deg P2 = 1 and
furthermore, P2 (z) = z. Now we rewrite the expression (16.18) as
[1 − P1 (z)]ez = P (z). (16.19)
If P1 (z) 6= 1, then we follow from the expression (16.19) that ez is a rational function, a contra-
diction. Thus P1 (z) ≡ 1 and so P (z) = 0 by the expression (16.19), but this contradicts to our
hypothesis.
For the entire function g(z) = sin z − P (z), if sin z − P (z) = P1 (z)eP2 (z) for some polynomials
P1 and P2 , then we have
sin z − P (z)
lim P (z) = z→∞
z→∞ 1
lim =0
z∈R z∈R
eP2 (z)
which contradicts Problem 1.26. Thus we must have P1 (z) ≡ 0 and then sin z − P (z) = 0 which
is another contradiction. Now we complete the analysis of the problem.
Problem 16.12
Proof. Suppose that f does not have infinitely many zeros. By Theorem 16.13, f is in the form
f (z) = Q(z)eP (z) , (16.20)
where P and Q are polynomials. Since f is non-vanishing, the representation (16.20) and
Theorem 5.12 (The Fundamental Theorem of Algebra) force that Q is a non-zero constant and
thus we may assume further that Q ≡ 1, i.e., f (z) = eP (z) .

Let S = k ∈ R lim |f (z)| exp(−|z|k ) = 0 and P (z) = an z n + an−1 z n−1 + · · · + a0 , where
z→∞
an 6= 0. Then it is easy to see that
|P (z)|
lim =1
z→∞ |an | · |z|n
so that
|f (z)|
lim = 1. (16.21)
z→∞ exp(|an | · |z|n )
Assume that n > j. Since j = inf S, the definition of infimum means that n is not a lower
bound of S, i.e., there exists a k ∈ S such that n > k. Therefore, we have exp(|z|−n ) < exp(|z|−k )
for large z which gives
|f (z)| |f (z)|
lim = lim =0
z→∞ exp(|z|n ) z→∞ exp(|z|k )
209
and this means that n ∈ S, but it implies the contrary result n > j > n. In other words, we
must have n ≤ j. If n < j, then we can find a k0 ∈ (n, j). Combining this fact and the limit
(16.21), it is trivial to see that
|f (z)| |f (z)| exp(|an | · |z|n )
lim k
= lim n
× = 0.
z→∞ exp(|z| 0 ) z→∞ exp(|an | · |z| ) exp(|z|k0 )
Consequently, it yields that k0 ∈ S which implies that k0 ≥ j, a contradiction. Hence it must
been that n = j and this completes the analysis of the problem.
Problem 16.13
Proof. Assume that sin z − z 6= c for some c ∈ C and for all z ∈ C. On the one hand, we have
sin z − z 6= c + 2π (16.22)
on C. Otherwise, if sin z0 − z0 − 2π = c for some z0 ∈ C, then we have
sin(z0 + 2π) − (z0 + 2π) = sin z0 − z0 − 2π = c,
a contradiction. On the other hand, since the function f (z) = sin z − z is clearly an entire
function of finite order, the Little Picard Theorem ensures that f assumes other values infinitely
many times, but this contradicts the observation (16.22). Hence we obtain the desired result
that sin z − z = c has a solution. This completes the proof of the problem.
Problem 16.14
Proof. Since g0 (0) = 0, we only consider the function

∞
X n
X
fk (z) fk (z)
F (z) = and Fn (z) = ,
gk (k) gk (k)
k=1 k=1
where n ∈ N. Since each fk is entire, every Fn is also entire. Let K be a compact set of C and
z ∈ K. Then it can be shown by induction easily that
k
..
k.
|fk (z)| ≤ exp(exp(· · · exp(|Re z|)) and | {z } .
gk (k) = k (16.23)
| {z }
k exponential factors k + 1 terms
Since K is compact, the set {|Re z| | z ∈ K} must be bounded by a positive constant M . Suppose
that N is a positive integer such that N ≥ 2 max(e, M ). In this case, N must satisfy
M 1
≤
N 2
so that
N N
eM ≤ e 2 ≤ N 2 .
Consequently, we achieve
f1 (z) eM 1 1
≤ N ≤ N ≤
g1 (z) N N2 2
and because N ≥ 2, so
M 1 N
f2 (z) ee exp( 21 N N ) N 2N 1 1
≤ NN ≤ N ≤ N =
√ ≤ 2.
g2 (z) N N N N N N N 2
N
By induction, it can be shown that k ≥ N implies
fk (z) 1
≤ k.
gk (k) 2
Now Theorem 1.9 (The Weierstrass M -Test) can be applied to conclude that Fn → F uniformly
in K. Finally, since K is taken arbitrary, it follows from Theorem 7.6 that F is entire. This
proves the first assertion.
For the second assertion, we note that if z = x > 0, then F (x) is a series of positive terms
which gives
fn (x)
|F (z)| = F (x) ≥
gn (n)
for every n ∈ N. Given that k ∈ N, since {gn (n)} is a strictly increasing sequence of positive
integers, the binomial theorem implies that for all large positive integers N , we have
egN+k (N +k) = (1 + α)gN+k (N +k)

[gN +k (N + k) − 1]α2
≥ · gN +k (N + k)
2
3
> kgN +k (N + k), (16.24)
2
where α = e − 1 > 1. Now, for each fixed j ∈ N, we know from the representation (16.23) that
gj (j) ≥ 1, so we may take xj = egj (j) > 1. Combining the inequality (16.24) and the fact
fj+2 (xN +k ) = exp(exp(exp · · · exp(xN +k ))) = exp(exp(exp · · · exp[gN +k (N + k)])),

| {z } | {z }
(j + 2) terms (j + 3) terms
we obtain
fj+2 (xN +k )
F (xN +k ) = ≥ exp(exp(exp · · · exp[gN +k (N + k)]))
gj+2 (j + 2) | {z }
(j + 2) terms
for all large positive integers N . Therefore, the result
log(log(log · · · log[F (xN +k )])) ≥ exp(exp[gN +k (N + k)])

| {z }
j terms
h3 i
> exp kgN +k (N + k)
2
hk i
= exp[kgN +k (N + k)] · exp gN +k (N + k)
2
> exp[kgN +k (N + k)]
= xkN +k
holds for all large positive integers N . By the definition, F is not of j-fold exponential order for
any positive integer j. This proves the second assertion and then we complete the proof of the
problem.
CHAPTER 17
Different Forms of Analytic Functions
Problem 17.1
Proof. By Definition 17.1, we get
N
Y 1
PN = 1−
k2
k=2
N h
Y (k − 1)(k + 1) i
=
k2
k=2
1·3 2·4 3·5 (N − 3)(N − 1) (N − 2)N (N − 1)(N + 1)
= × × × ··· × × ×
2·2 3·3 4·4 (N − 2)(N − 2) (N − 1)(N − 1) N ·N
1 N +1
= ×
2 N
1
so that PN → 2 as N → ∞, completing the proof of the problem.
Problem 17.2
Proof. By Definition 17.1, we see that
N h
Y (−1)k i
PN = 1+
k
k=2
N h
Y k + (−1)k i
=
k
k=2
3 2 5 4 N − 1 + (−1)N −1 N + (−1)N
= × × × × ··· × × . (17.1)
|2 3 4 5 {z N − 1 N }
(N − 1) terms
211
Chapter 17. Different Forms of Analytic Functions 212
Obviously, we know that



 N −1
N + (−1)N  N , if N is odd;

=
N 

 N + 1 , otherwise.

N
Hence the expression (17.1) reduces to

 1,
 if N is odd;
PN =
 N + 1 , otherwise.

N
Now we conclude that PN → 1 as N → ∞. This completes the proof of the problem.
Problem 17.3
Proof. For every N ∈ N, we consider

N
X i Xh
N
i i i
log 1 + = log 1 + + iArg 1 +
n n n
n=1 n=1
XN X N
i 1
= log 1 + +i arctan . (17.2)
n n
n=1 n=1
Since ex ≥ 1 + x for x ≥ 0, we obtain log(1 + x) ≤ x and then

r
i 1 1 1 1
log 1 + = log 1 + 2 = log 1 + 2 ≤ 2 . (17.3)
n n 2 n 2n
Applying the Comparison Test [27, Theorem 6.6, p. 76] to the inequality (17.3), we conclude
that
X∞
i
log 1 + (17.4)
n=1
n
∞
Y i
converges. By Proposition 17.2, the product 1+ converges.
n
n=1
∞
Y i
Assume that the product 1+ was convergent. Then Proposition 17.2 tells us that
n
n=1
∞
X i
the series log 1 + converges. Applying this fact with the convergence of the series (17.4)
n
n=1
to the sum (17.2), we see that the series
∞
X 1
arctan (17.5)
n=1
n
is also convergent. However, since arctan n1 ≥ π

for all n ≥ 1, the Comparison Test implies
4n
∞
Y i
that the series (17.5) is divergent, a contradiction. Hence the product 1+ is divergent.
n
n=1
213
Problem 17.4
Proof. Using similar argument as in the proof of Proposition 17.3, we take N ∈ N such that
k > N implies |zk | < 21 . Then, for every k > N , we have
zk2 zk3 1 z
k
log(1 + zk ) − zk = − + − · · · = zk2 − + − ··· (17.6)
2 3 2 3
so that
1 zk
| log(1 + zk ) − zk | = |zk |2 · − + − ···
2 3
1 |z | |z |2
k k
≤ |zk |2 · + + + ···
2 3 4
1 1 1
≤ |zk |2 · + + + ···
2 4 8
= |zk |2 .
∞
X ∞
X
Hence the series [log(1 + zk ) − zk ] and then the series [log(1 + zk ) − zk ] are convergent.
k=N +1 k=1
∞
X
Since zk converges, we conclude that the series
k=1
∞
X
log(1 + zk )
k=1
∞
Y
is convergent. By Proposition 17.2, the product (1 + zk ) converges which completes the proof
k=1
of the problem.
Problem 17.5
∞
Y
(−1) k
Proof. Let zk = √ ,
k
where k = 2, 3, . . .. Assume that the product (1 + zk ) was convergent.
k=2
By Proposition 17.2, the series
∞
X
log(1 + zk ) (17.7)
k=2
is convergent. Using the formula (17.6), if k ≥ 4, then we havea
1 h 1 (−1)k 1 i
log(1 + zk ) − zk = · − + √ − + ···
k 2 3 k 4k
1 1 1
≤ · − + √
k 2 3 k
a
Notice that all zk and 1 + zk are real numbers.
1 1 1
≤ · − +
k 2 6
1
≤−
3k
or equivalently,
1
zk − log(1 + zk ) ≥ . (17.8)
3k
∞
X ∞
X
Thus the convergence of zk and the series (17.7) imply that the series [zk − log(1 + zk )]
k=2 k=4
is convergent, so it follows from the inequality (17.8) and the Comparison Test that the series
∞
X 1
k
k=4
∞
Y
is convergent, a contradiction. Consequently, the product (1 + zk ) must be divergent, com-
k=2
pleting the proof of the problem.
Problem 17.6
Proof. Now we claim that

N−1 N −1
PN = (1 + z)(1 + z 2 ) · · · (1 + z 2 ) = 1 + z + z2 + · · · + z2
for every N ∈ N. The case for N = 1 is trivial. Suppose that

k−1 k −1
Pk = (1 + z)(1 + z 2 ) · · · (1 + z 2 ) = 1 + z + z2 + · · · + z2 (17.9)
for some k ∈ N. If N = k + 1, then the assumption step (17.9) implies that

k−1 k
Pk+1 = [(1 + z)(1 + z 2 ) · · · (1 + z 2 )](1 + z 2 )
k −1 k
= (1 + z + z 2 + · · · + z 2 )(1 + z 2 )
k −1 k k +1 k+1 −1
= 1 + z + z2 + · · · + z2 + z2 + z2 + · · · + z2 .
Hence our claim follows from induction.

If K is a compact subset of D(0; 1), then there exists a 0 < δ < 0 such that K ⊆ D(0; δ). In
∞
X
D(0; δ), we see that the series z k converges uniformly to 1−z
1 b
, so it yields that
k=0
∞
X 1
PN → zk =
1−z
k=0
uniformly on D(0; δ) and in particularly, on K. Since K is arbitrary, we know from Definition

∞
Y k
7.5 that the product (1 + z 2 ) converges uniformly on compacta to 1−z 1
in |z| < 1. This
k=0
b
See also Problem 2.19 and [4, p. 27].
215
Problem 17.7
Proof. Since λk → ∞, Theorem 17.7 (The Weierstrass Product Theorem) ensures the existence
X∞
1
of such an entire function g. Since converges, we deduce from the discussion on [4, p.
k2
k=1
245] that the function
∞
Y z
g(z) = 1− 2 (17.10)
k
k=1
satisfies our requirements, completing the analysis of the problem.
Problem 17.8
Proof. Recall from Problem 3.22 that

∞
X (−1)k
z3 z5
sin z = z − + − ··· = z 2k+1 ,
3! 5! (2k + 1)!
k=0
so we obtain √ ∞
sin(π z) X (−1)k π 2k k
√ = z . (17.11)
π z (2k + 1)!
k=0
(−1)k π 2k
If ak = (2k+1)! , then
ak+1 −π 2
lim = lim = 0.
k→∞ ak k→∞ (2k + 3)(2k + 2)
By Problem √ 2.13 and Theorem 2.8, the power series (17.11) converges everywhere, i.e., the
function π z z) is in fact entire. Since the zeros of sin z are kπ for all k ∈ Z, the zeros of
sin(π
√
√
sin(π z) are k2 for all k ∈ N. In other words, one solution to Problem 17.7 is given by the
entire function (17.11). This ends the proof of the problem.

Problem 17.9
Proof. By the inspiration of [4, Example 3, p. 247], an entire function with a single zero at
every k + 12 with k ∈ Z is given by
nY
∞ h
z z io n Y ∞ h
z z io
f (z) = 1− exp × 1 − exp
k=0
k + 12 k + 12 k=0
−k − 12 −k − 12
∞ h
Y z z z z i
= 1− exp 1 + exp −
k=0
k + 12 k + 12 k + 21 k + 21
∞ h
Y 4z 2 i
= 1− .
(2k + 1)2
k=0
Note that cos πz is an entire function having simple zeros at every k + 12 , where k ∈ Z. Therefore,
we have
∞ h
Y 4z 2 i
cos πz = C 1− (17.12)
(2k + 1)2
k=0
for some nonzero constant C. Put z = 0 into the expression (17.12), we get immediately that
C = 1. Hence we have completed the proof of the problem.
Problem 17.10
Proof.
(a) Applying similar idea of [4, Example 1, p. 246], an entire function g with zeros at every
positive integer is given by
∞
Y z z
g(z) = 1− ek .
k
k=1
Next, we define
1 Y∞
1 1
f (z) = g = 1− e k(1−z)
1−z k(1 − z)
k=1
which is clearly analytic in |z| < 1 and f (z) = 0 if and only if z = 1 − k1 , where k ∈ N.
(b) Suppose that {zk } is a sequence of distinct numbers such that zk → z0 as k → ∞. Then the
1
sequence {λk = zk −z 0
} satisfies λk → ∞ as k → ∞. By Theorem 17.7 (The Weierstrass
Product Theorem), one can find an entire function g such that g(z) = 0 if and only if
z = λk for every k ∈ N. Now if we define
1
f (z) = g ,
z − z0
then f is analytic in |z| < |z0 | and f (z) = 0 if and only if z = z1 , z2 , . . ..
Problem 17.11
Proof. Since f (z) = ϕ(z, t) satisties the hypotheses of Theorem 17.9, the function F (z) is ana-
lytic in a region D. Thus we get
Z Z Z b
′ 1 F (ζ) 1 f (ζ)
F (z) = dζ = dt dζ, (17.13)
2πi C (ζ − z)2 2πi C a (ζ − z)
2
where C is a disc contained in D with centre z. As suggested by the question, we switch the
order of integration in the formula (17.13) to obtain
Z b Z
′ 1 f (ζ)
F (z) = dζ dt. (17.14)
a 2πi C (ζ − z)2
217
Using the formula at the bottom of [4, p. 80], we see that

Z
′ 1 f (ζ)
ϕz (z, t) = f (z) = dζ.
2πi C (ζ − z)2
After the substitution of this formula into the expression (17.14), we reduce it to
Z b
′
F (z) = ϕz (z, t) dt
a
which is our desired formula and we end the proof of the problem.
Problem 17.12
Proof. Given that ǫ > 0 which will be determined later. We write

Z β Z x−ǫ Z x+ǫ
h(u)y h(u)y h(u)y
2 2
du = 2 2
du + 2 2
du
α (u − x) + y α (u − x) + y x−ǫ (u − x) + y
Z β
h(u)y
+ 2 2
du, (17.15)
x+ǫ (u − x) + y
where h is a continuous function on [α, β] and x ∈ (α, β). Since [α, β] is compact, h is bounded by
a positive constant M . Furthermore, on [α, x−ǫ], it is easy to see that (u−x)2 +y 2 ≥ ǫ2 +y 2 ≥ ǫ2 .
Therefore, we get
Z x−ǫ Z x−ǫ Z x−ǫ
h(u)y h(u)y M |y| M |y|(β − α)
2 2
du ≤ 2 2
du ≤ 2
du ≤ . (17.16)
α (u − x) + y α (u − x) + y α ǫ ǫ2
Similarly, it is true that
Z β
h(u)y M |y|(β − α)
du ≤ . (17.17)
x+ǫ (u − x)2 + y 2 ǫ2
Finally, since the function

y
g(u) =
(u − x)2 + y 2
is integrable and does not change sign on [x − ǫ, x + ǫ], we observe from the Weighted Mean
Value Theorem for Integrals [5, Exercise 17, p. 215] that
Z x+ǫ Z x+ǫ
h(u)y y du
2 2
du = h(ξ) 2 2
(17.18)
x−ǫ (u − x) + y x−ǫ (u − x) + y
for some ξ ∈ [x − ǫ, x + ǫ]. We apply the substitution u = x + y tan θ to the integral on the
right-hand side of the expression (17.18), we have
Z x+ǫ Z tan−1 (ǫ/y)
h(u)y ǫ
du = h(ξ) dθ = 2h(ξ) tan−1 . (17.19)
x−ǫ (u − x)2 + y 2 tan−1 (−ǫ/y) y
√
Finally, we take ǫ = 4 y and we put the results (17.16), (17.17) and (17.19) into the expression
(17.15) to get
Z β
h(u)y 1
lim 2 2
du = lim 2h(ξ) tan−1 3 .
y→0 α (u − x) + y y→0 y4
3
Since tan−1 (y − 4 ) → π
2 and ξ → x as y → 0, we conclude that
Z β
h(u)y
lim du = πh(x),
y→0 α (u − x)2 + y 2
Problem 17.13
Proof. According to direct integration, it is clear that

Z 1
dt log(1 − z)
f (z) = =− . (17.20)
0 1 − zt z
Recall from [4, p. 115] that log z is analytic in C \ (−∞, 0], so log(1 − z) and then f is analytic
in D = C \ [1, ∞). Suppose that γ is a simple closed curve encircling the point z = 1. Since
the function F (z) = 1 − z is analytic inside and on γ and F (z) 6= 0 on γ, Corollary 10.9 (The
Argument Principle) implies that
Z
1 1 F ′ (z)
∆Arg (1 − z) = dz = 1
2π 2πi γ F (z)
which means that ∆Arg (1 − z) = 2πi as z traverses along γ. Hence it follows from the repre-
sentation (17.20) that f has a “jump” of 2πi
x as z crosses from the upper half-plane to the lower
half-plane through any point x > 1. This completes the analysis of the problem.
Problem 17.14
Proof.
(a) By the definition, |φ(n)| = φ(n) ≤ n. Since |nz | = nRe z , we obtain

X∞ X∞ ∞ ∞
φ(n) φ(n) X n X 1
z
= z
≤ Re z
= .
n=1
n n=1
|n | n=1
n n=1
n z−1
Re
By [27, Theorems 6.6, 6.10, pp. 76, 77], the series

∞
X φ(n)
n=1
nz
converges absolutely for Re z − 1 > 1 or equivalently, Re z > 2.
(b) By [4, Example, pp. 254, 255], we have

∞
X ∞
X ∞ ∞
φ(n) 1 X φ(n) X cn
ζ(z) = · = ,
nz nz nz nz
n=1 n=1 n=1 n=1
where X n X
cn = φ = φ(n) = n.
d
d|n d|n
219
Consequently, we note that

∞
X ∞
X
φ(n) 1
ζ(z) = = ζ(z − 1)
nz nz−1
n=1 n=1
for Re z > 2.
CHAPTER 18
Analytic Continuation; The Gamma and Zeta
Functions
Problem 18.1
Proof. Let D1 = C \ {(x, 0) | x ≤ 0} and D2 = C \ {(0, y) | y ≤ 0}. Define g1 : D1 → C by
g1 (z) = log |z| + iArg z, (18.1)
where Arg z ∈ (−π, π). Similarly, we define g2 : D2 → C by
g2 (z) = log |z| + iArg z, (18.2)
where Arg z ∈ (− π2 , 3π
2 ). Obviously, we have
eg1 (z) = eg2 (z) = z,
so g1 and g2 are analytic branches of log z by Definition 8.7. If we let D be the first quadrant
and define f : D → C by
f (z) = log |z| + iArg z,
where Arg z ∈ (0, π2 ), then each of g1 and g2 is an analytic continuation of f . Suppose that ζ is
a point lying in the third quadrant. On the one hand, we know from the definition (18.1) that
π
Im g1 (ζ) = Arg ζ ∈ − π, − .
2
On the other hand, the definition (18.2) gives
3π
Im g2 (ζ) = Arg ζ ∈ π, .
2
Therefore, we establish that g1 (ζ) 6= g2 (ζ), completing the proof of the problem.
Problem 18.2
221
Chapter 18. Analytic Continuation; The Gamma and Zeta Functions 222
Proof.
∞
X
(a) Assume that there was a point |z0 | = 1 such that the series an z0n converges. Without
n=0
loss of generality, we may assume that z0 = 1. Otherwise, if z0 = eiθ0 for some θ0 ∈ [0, 2π],
then we have
X∞ ∞
X
an z0n = (an einθ0 ) · 1n .
n=0 n=0
Thus it must be true that an → 0 as n → ∞. Given ǫ > 0. Then there exists an N ∈ N
such that |an | < ǫ for all n ≥ N . Therefore, for every z ∈ D(0; 1), we see that
∞
X N
X −1 ∞
X N
X −1
ǫ
|f (z)| ≤ |an | · |z|n = |an | · |z|n + |an | · |z|n ≤ |an | +
1 − |z|
n=0 n=0 n=N n=0
which gives, for 0 < x < 1,

lim (1 − x)f (x) ≤ ǫ.
x→1−
x∈R
Since ǫ is arbitrary, we actually have (1 − x)f (x) → 0 as x → 1− and x ∈ R. However, this

contradicts the hypothesis that f has a pole at z = 1. Hence the power series diverges at
every point on C(0; 1).
(b) Suppose that the radius of convergence of

∞
X
f (z) = an z n (18.3)
n=0
is R > 0 and f has a pole at z0 = Reiθ0 . Then the function

∞
X
g(z) = f (Rz) = (an Rn )z n (18.4)
n=0
has radius of convergence 1 and g has a pole at z = eiθ0 . By the argument in part (a), we
may assume that the pole of g is at z = 1. Hence part (a) implies that the power series
(18.4) diverges at every point on C(0; 1) which is equivalent to saying that the power series
(18.3) diverges at all points on C(0; R).
Problem 18.3
∞
X
Proof. Define f (z) = (−1)n an z n and its radius of convergence to be R < ∞. Then we have
n=0
∞
X
f (−z) = an z n
n=0
with R as its radius of convergence. Now the function f (−z) satisfies the hypotheses of Theorem
18.3, so it has a singularity at z = R. Hence we conclude that f (z) has a singularity at z = −R
which ends the proof of the problem.
223
Problem 18.4
Proof.
(a) By the application of Problem 18.5 in advance, we have
Z ∞
1 1 2
√
3
= 1 e−nt t− 3 dt.
n Γ( 3 ) 0
If |z| < 1, then we derive
X∞ X∞ h Z ∞ i
zn 1 n −nt − 23
√ = z e t dt
n=1
3
n
n=1
Γ( 31 ) 0
Z ∞hX ∞ i
1 −t n − 32
= (ze ) t dt
Γ( 13 ) 0 n=1
Z ∞ −2
1 t 3 · ze−t
= dt
Γ( 13 ) 0 1 − ze−t
Z ∞ −2
z t 3
= dt (18.5)
Γ( 13 ) 0 et − z
Z N −2
z t 3
= 1 lim t
dt. (18.6)
Γ( 3 ) ǫ→0 ǫ e − z
N →∞
− 32
Let f (t) = et and g(t) = t . Clearly, f and g are continuous real-valued functions on
[ǫ, N ]. Furthermore, f ′ (t) = et > 0 is also continuous on [ǫ, N ]. By Proposition 17.10, the
integral (18.6) is analytic in C \ [eǫ , eN ]. Since ǫ → 0 and N → ∞, the integral (18.5) is
actually analytic in C \ [1, ∞) as required.
(b) Now we apply the second well-known formula on [4, p. 262], we have
Z ∞
1
= e−nt sin t dt
n2 + 1 0
which show that if |z| < 1, then

∞
X X∞ hZ ∞ i
zn −t n
= (ze ) sin t dt
n=0
n2 + 1 n=0 0
Z ∞hX∞ i
= (ze−t )n sin t dt
0 n=0
Z ∞ t
e sin t
= dt (18.7)
0 et − z
Z N t
e sin t
= lim dt. (18.8)
ǫ→0 ǫ et − z
N →∞
et
Again, if we let f (t) = and g(t) = et sin t,
then both are continuous real-valued functions
′
on [ǫ, N ] and f > 0 is also continuous on [ǫ, N ]. Hence we derive from Proposition 17.10
that the integral (18.8) is analytic outside [eǫ , eN ]. Finally, by letting ǫ → 0 and N → ∞,
the integral (18.7) is analytic outside [1, ∞).
Problem 18.5
Proof. Suppose that x = nt. Thus it follows from the definition [4, Eqn. (1), p. 265] that
Z ∞ Z ∞ p−1 Z ∞
−nt p−1 −x x dx 1 Γ(p)
e t dt = e · p−1 · = p e−x xp−1 dx = p
0 0 n n n 0 n
for p > 0. This completes the proof of the problem.
Problem 18.6
Proof. By the substitution x = t2 , we deduce from the definition [4, Eqn. (1), p. 265] and then
√
the fact Γ( 21 ) = π (see [4, p. 268]) that
Z ∞ Z √
−t2 1 ∞ −x − 1 1 1 π
e dt = e x dx = Γ
2 =
0 2 0 2 2 2
as required, completing the proof of the problem.
Problem 18.7
Proof. Define Z n t n
Γn (z) = tz−1 1 − dt,
0 n
where Re z > 0 and n ∈ N. Then we have
Z nh Z ∞
−t t n i z−1
Γ(z) − Γn (z) = e − 1− t dt + e−t tz−1 dt. (18.9)
0 n n
It is easy to see that the second integral in the expression (18.9) tends to 0 as n → ∞.
Let t ≤ n, i.e., nt ≤ 1. By the power series expansion of e−x , we see that
t t t2 t3
e− n = 1 − + − + ···
n 2!n2 3!n3
so that on the one hand,
t
t t2 t3 t2 1 t t2
e− n − 1 − = − + · · · = − + · · · ≤ .
n 2!n2 3!n3 n2 2! 3!n 2n2
We derive from the given identity with a = exp(− nt ) and b = 1 − t
n that
t n h (n − 1)t i t
t t t2 et2
e−t − 1 − ≤ n exp − · e− n − 1 − ≤ ne−t · e n · 2 ≤ e−t · .
n n n 2n 2n
Consequently, it implies that
Z nh Z n Z n
−t t n i z−1 z−1 −t
t n e
e − 1− t dt = |t | · e − 1 − dt ≤ e−t tRe z+1 dt.
0 n 0 n 2n 0
225
By the definition, we know that

Z n Z ∞
lim e−t tRe z+1 dt = e−t tRe z+1 dt = Γ(Re z + 2) < ∞
n→∞ 0 0
so that Z nh t n i z−1
lim e−t − 1 − t dt = 0
n→∞ 0 n
and then it yields from the expression (18.9) immediately that
Γ(z) = lim Γn (z),

n→∞
Problem 18.8
Proof. Suppose that z = x > 0. If we take the logarithm on both sides of the product formula
∞
−1 γx
Y x −x
Γ (x) = xe 1+ e k,
k
k=1
then we arrive at
∞ h
X x xi
− log Γ(x) = log x + γx + log 1 + −
k k
k=1
and thusa
d Xh x xi
∞
Γ′ (x) 1
− = +γ+ log 1 + − . (18.10)
Γ(x) x dx k k
k=1
Now termwise differentiation is permitted for the series (18.10) provided that the differentiated
series is uniformly convergent. To see this, we consider the series
dh x xi X 1 1
∞
X ∞ ∞
X x
log 1 + − = − =− .
dx k k x+k k k(x + k)
k=1 k=1 k=1
Let x ∈ [ 12 , 1]. Then we have k(x + k) ≥ k2 which gives

∞
X X 1 ∞
x
≤ .
k(x + k) k2
k=1 k=1
By the Comparison Test, the differentiated series is uniformly convergent on [ 12 , 1]. Hence it
follows from [28, Theorem 10.8, p. 4] that we can take term by term differentiation to get
X ∞
Γ′ (x) 1 x
− = +γ− . (18.11)
Γ(x) x k(x + k)
k=1
Put x = 1 into the expression (18.11) and use the fact Γ(1) = 1, we conclude that
−Γ′ (1) = 1 + γ − 1 = γ
which means that Γ′ (1) = −γ, completing the proof of the problem.
a
The logarithm derivative ψ(z) = (log Γ(z))′ is called the digamma function.
Problem 18.9
Proof. Recall from the factb

1 1 1
1 − z ζ(z) = 1 + z + z + · · ·
2 3 5
that
2
f (z) = 1 − ζ(z)
2z
1 1
= 1− z
ζ(z) − z ζ(z)
2 2
1 1 1 1 1
= 1+ z
+ z
+ · · · − z − z − z − ···
3 5 2 4 6
1 1 1
= 1 − z + z − z + ··· .
2 3 4
By Theorem 18.9, f is analytic in C \ {1} and furthermore
2 2
lim (z − 1)f (z) = lim 1 − z (z − 1)ζ(z) = lim 1 − z lim (z − 1)ζ(z) = 0 · 1 = 0,
z→1 z→1 2 z→1 2 z→1
so f has a removable singularity at z = 1. Therefore, f is entire which completes the proof of
the problem.
Problem 18.10
Proof. Let {pk } be the set of all primes. Since ζ has a singularity at z = 1, we have ζ(1+ǫ) → ∞
as ǫ → 0+ so that the identity [4, Eqn. (5), p. 269] ensures that
∞
Y 1
1− (18.12)
pk
k=1
diverges to 0. Clearly, we observe that

∞
X ∞
X
1 1
≤ < ∞.
p2k k2
k=1 k=1
∞
X1
Assume that −was convergent. Then Problem 17.4 guarantees that the product (18.12)
pk
k=1
is convergent which is a contradiction. Hence the series
∞
X 1
−
pk
k=1
must diverge, completing the proof of the problem.
Remark 18.1
For another proof of Problem 18.10, please refer to [26, Problem 8.10, p. 180].
b
Read [4, p. 268].
CHAPTER 19
Applications to Other Areas of Mathematics
Problem 19.1
Proof. Let f (z) = z − tan z and N ∈ N. Consider the square SN with vertices N π(±1 ± i).
Then f is meromorphic in SN , so Theorem 10.8 implies that
Z Z
1 tan2 z 1 sec2 z − 1
dz = dz = Zf − Pf . (19.1)
2πi ∂SN tan z − z 2πi ∂SN tan z − z
On the one hand, if we plot the graphs of y = x and y = tan x on [−N π, N π], then we see that
the straight line y = x has one and only one intersection with the curve y = tan x in each of the
following intervals
[−(k + 1/2)π, −kπ] and [kπ, (k + 1/2)π],
where k = 1, 2, . . . , N − 1, see Figure 19.1 for an example.
Figure 19.1: The intersections of y = x and y = tan x.
227
Chapter 19. Applications to Other Areas of Mathematics 228
Furthermore, the power series of tan z (see [1, p. 184]) indicates that f (0) = f ′ (0) = f ′′ (0) = 0
but f (3) (0) 6= 0, so f has a zero of order 3 at z = 0. In conclusion, if we denote Z′f to be the
number of nonreal zeros of f , then we establish that
Zf = Z′f + 2(N − 1) + 3 = Z′f + 2N + 1 (19.2)

sin z
inside SN . On the other hand, since tan z = cos z , we have Pf = Zcos z inside SN . It is clear that
1
cos(n + 2 )π = 0 for all −N ≤ n ≤ N − 1, we know that
Pf = 2N (19.3)
inside SN .
Putting the values (19.2) and (19.3) back into the expression (19.1), we get
Z
1 tan2 z
dz = 1 + Z′f . (19.4)
2πi ∂SN tan z − z
Now it remains to estimate the integral in the formula (19.4). On the vertical side Re z = N π
and y = Im z ∈ [−N π, N π], we have
|eiz − e−iz | |e−y − ey |

| tan z| = = < 1. (19.5)
|eiz + e−iz | |e−y + ey |
Similarly, the bound (19.5) holds on the other vertical side. Next, if we take N large enough,
then along the horizontal side x = Re z ∈ [−N π, N π] and Im z = N π, we obtain
|e−N π eix − eN π e−ix | |e2N π − e2ix | 1 + e−2N π 10

| tan z| = −N π ix N π −ix
= 2N π 2ix
≤ −2N π
≤ .
|e e +e e | |e +e | 1−e 9
Consequently, we always have | tan z| ≤ 10 9 on ∂SN for large enough N . By the triangle inequal-
10
ity, | tan z − z| ≥ |z| − | tan z| ≥ N − 9 on ∂SN so that
tan2 z 100
≤
tan z − z 81N − 90
on ∂SN . Since the length of ∂SN is 8N , it follows from Theorem 4.10 (The M -L Formula) that
Z
1 tan2 z 400N
dz ≤ <2
2πi ∂SN tan z − z π(81N − 90)
for large enough N . Combining this estimate with the formula (19.4), we conclude at once that
Z′f = 0, completing the proof of the problem.
Problem 19.2
Proof. By the hypothesis, we know that

Z Z
dz
f2 (z) dz → −
CN CN z
as N → ∞. Applying Corollary 10.9 (The Argument Principle) to the function f (z) = z to get
Z
dz
= 2πi.
CN z
229
Therefore, we conclude that Z

f2 (z) dz → −2πi
CN
as N → ∞.
By Problem 19.1, all the zeros of tan z−z are real. Denote the set of its zeros by {0, x1 , x2 , . . .}.
2
Thus it suffices to sum the values of sinx2 x at {0, x1 , x2 , . . .}. Figure 19.2 illustrates that the curves
2
y = sinx2 x and y = tan x − x intersect on the set {x1 , x2 , . . .}.
sin2 x
Figure 19.2: The intersections of y = x2
and y = tan x − x.
According to Theorem 10.5 (The Cauchy’s Residue Theorem), we have

Z hX
∞ i
f2 (z) dz = 2πi Res (f2 ; xk ) + Res (f2 ; i) + Res (f2 ; −i) + Res (f2 ; 0)
CN k=1
hX
∞
sin2 xk i
= 2πi + Res (f 2 ; i) + Res (f 2 ; −i) + Res (f 2 ; 0) (19.6)
x2k
k=1
Since tan(±i) − (±i) 6= 0, we deduce from [4, Eqn. (1), p. 129] that
−1 1 + e2
Res (f2 ; i) = =− . (19.7)
2i(tan i − i) 4
Similarly, we have
1 + e2
Res (f2 ; −i) = −. (19.8)
4
Next, recall from the proof of Problem 19.1 that tan z − z has a zero of order 3 at z = 0,
so the function f2 (z) has a simple pole at z = 0. Therefore, it follows from [1, p. 184] that
3 5
tan z − z = z3 + 2z
15 + · · · , so we have
1
f2 (z) = 2z 3
(1 + z 2 )( 3z + 15 + ···)
and then [4, Eqn. (1), p. 129] implies that
Res (f2 ; 0) = 3. (19.9)
Substituting all the residues (19.7), (19.8) and (19.9) into the formula (19.6), we establish
∞
X sin2 xk 1 + e2
− + 3 = −1
x2k 2
k=1
∞
X sin2 xk e2 − 7
=
x2k 2
k=1
so that
sin2 x X
∞
sin2 xk sin2 x
Var =2 + lim = e2 − 5.
x2 x2k x→0 x2
k=1
Problem 19.3
Proof. To apply the idea of §19.1, we note that all the zeros of ez − z are simple. Next, we recall
that f ′ /f has residue at every simple zero of f , so we consider the function
1 ez − 1
f (z) = · .
z 2 ez − z
Now it is clear from Theorem 10.5 (The Cauchy’s Residue Theorem) that
X 1 Z
1
= f (z) dz − Res (f ; 0), (19.10)
zk2 2πi CN
where CN is the square described in §19.1 and the summation counts all zeros of ez − z inside
CN . Since f has a double pole at z = 0, we have
d 2
Res (f ; 0) = [z f (z)] =1
dz z=0
and then the expression (19.10) reduces to

X 1 Z
1
= f (z) dz − 1. (19.11)
zk2 2πi CN
It remains to find the estimate of the integral. In fact, it is obvious that |z|2 ≥ N 2 π 2 on CN .
Let z = x + iN π for x ∈ [−N π, N π]. The triangle inequality implies that |ez − 1| ≤ ex + 1 on
[−N π, N π]. Furthermore, if we take N to be even, then
q
|e − z| = [(−1)N ex − x]2 + N 2 π 2 ≥ ex − x > 0.
z
so that
ez − 1 ex + 1
≤ (19.12)
ez − z ex − x
ex +1
on [−N π, N π]. Let F (x) = ex −x . Then we have
1 − xex
F ′ (x) = .
(ex − x)2
231
Thus F ′ (x0 ) = 0 if and only if x0 ex0 = 1. Clearly, x0 ∈ (0, 1), so we can establish from the First
Derivative Test that F attains its absolute maximum at x = x0 and the maximum value is given
by
ex0 + 1 x0 (ex0 + 1) 1
F (x0 ) = x0 = x
= > 0. (19.13)
e − x0 x0 (e 0 − x0 ) 1 − x0
Combining the inequality (19.12) and the result (19.13), we obtain the bound
ez − 1 1
z
≤ <∞ (19.14)
e −z 1 − x0
on the horizontal line segment z = x + iN π for x ∈ [−N π, N π]. Actually, the same bound
(19.14) also holds on the horizontal line segment z = x − iN π for x ∈ [−N π, N π].
On the vertical line segment z = N π + iy for y ∈ [−N π, N π], we see that
p
|ez − 1| = e2N π − 2eN π cos y + 1 ≤ eN π + 1 (19.15)
and
q
|ez − z| = (eN π cos y − N π)2 + (eN π sin y − y)2
q
= e2N π − 2eN π (N π cos y + y sin y) + N 2 π 2 + y 2
p
≥ e2N π − 4N πeN π
eN π
≥ (19.16)
2
for large even positive integer N . Therefore, it yields from the estimates (19.15) and (19.16)
that
ez − 1 1
≤ 2 1 + ≤4
ez − z eN π
on z = N π+iy for y ∈ [−N π, N π], where N is a sufficiently large even positive integer. Similarly,
if z = −N π + iy for y ∈ [−N π, N π], then
p
|ez − 1| = e−2N π − 2e−N π cos y + 1 ≤ e−N π + 1 ≤ 2
and
q
|ez − z| = (e−N π cos y + N π)2 + (e−N π sin y − y)2
q
= e−2N π + 2e−N π (N π cos y − y sin y) + N 2 π 2 + y 2
Nπ
≥
2
for large even positive integer N . Consequently, we get
ez − 1 4
z
≤
e −z Nπ
on z = −N π + iy for y ∈ [−N π, N π], where N is a sufficiently large even positive integer.

Hence the analysis in the previous two paragraphs guarantee that there exists a positive
constant M such that
ez − 1
≤M
ez − z
on CN for large enough even positive integer N . Now Theorem 4.10 (The M -L Formula) asserts
that Z
M 8M
f (z) dz ≤ 2 · 8N π ≤
CN N Nπ
for large enough even positive integer N so that
Z
lim f (z) dz = 0
N →∞ CN
and then the expression (19.11) implies that

∞
X 1
= −1.
zk2
k=1
Problem 19.4
Proof. The argument in §19.3 can be applied exactly the same here except the last two equations
on [4, p. 278] are given by
α2 z 2 a2 z 2
1 + αz + + · · · = eαz = 1 + a1 z + + ··· ,
2! 2!
γ 2z2 b2 z 2
1 + γz + + · · · = eγz = 1 + b1 z + + ··· .
2! 2!
Thus, we have immediately that a1 = α and ak = αk = ak1 for all k = 2, 3, . . .. Similarly, we
obtain b1 = γ and bk = γ k = bk1 for all k = 2, 3, . . .. Consequently, for every pair a1 , b1 ≥ 0 with
a1 + b1 = 2, it gives a solution ak = ak1 and bk = bk1 to the system of equations in §19.3. This
Problem 19.5
Proof. Similar to the discussion in §19.4, we have the equation

z z a1 z a2 z ak
= + + · · · + . (19.17)
1−z 1 − z d1 1 − z d2 1 − z dk
Assume that d1 was relatively prime to d2 , d3 , . . . , dk . Thus if z → exp( 2πi

d1 ) 6= 1, then we must
2πid
have z dj → exp( d1 j ) 6= 1 for all j = 2, 3, . . . , k. In other words, as z → exp( 2πi
d1 ), the first
term on the right-hand side of the equation (19.17) will approach infinity while all the remaining
terms tend to a finite limit, a contradiction. Hence it is impossible to do so which ends the proof
of the problem.
Problem 19.6
233
Proof.
(a) The cases for n = 0 and n = 1 are trivial. Assume that
ck ≤ 3k and ck+1 ≤ 3k+1 (19.18)
for some k ∈ N. If n = k + 1, then the assumption (19.18) certainly implies
ck+2 = ck+1 + 2ck ≤ 3k+1 + 2 · 3k = 3k (3 + 2) = 5 · 3k ≤ 3k+2 .

1
By induction, cn ≤ 3n holds for all n ≥ 0. Since lim sup |cn | n ≤ lim sup 3 = 3, it derives
n→∞ n→∞
from Theorem 2.8 that the radius of convergence of F (z) ≥ 31 .
(b) Direct computation gives

∞
X
(1 − z − 2z 2 )F (z) = (1 − z − 2z 2 ) cn z n
n=0
∞
X ∞
X ∞
X
= cn z n − cn z n+1 − 2cn z n+2
n=0 n=0 n=0
∞
X ∞
X ∞
X
= c0 + c1 z + cn z n − c0 z − cn z n+1 − 2cn z n+2
n=2 n=1 n=0
∞
X ∞
X ∞
X
=1+ cn+2 z n+2 − cn+1 z n+2 − 2cn z n+2
n=0 n=0 n=0
X∞
=1+ (cn+2 − cn+1 − 2cn )z n+2
n=0
=1
which implies that
1 2 1
F (z) = 2
=− + . (19.19)
1 − z − 2z 3(2z − 1) 3(z + 1)
(c) By expressing the two rational functions in the expression (19.19) as power series, we
follow that
1 1 1 1
F (z) = − · + ·
3 z − 21 3 z − (−1)
∞ ∞
2X n n 1X
= 2 z + (−1)n z n
3 n=0 3 n=0
∞
X 2n+1 + (−1)n
= zn.
3
n=0
Hence we find that

2n+1 + (−1)n
cn =
3
for all n = 0, 1, 2, . . ..
Problem 19.7
Proof. We note that

∞
X
(1 − z)F (z) = (1 − z) cn z n
n=1
∞
X ∞
X
= cn z n − cn z n+1
n=1 n=1
∞
X ∞
X
= c1 z + cn z n − cn z n+1
n=2 n=1
∞
X X∞
=z+ cn+1 z n+1 − cn z n+1
n=1 n=1
X∞
=z+ (cn+1 − cn )z n+1
n=2
∞
X
= n2 z n . (19.20)
n=1
∞
X 1
By differentiating zn = , we have
n=0
1−z
X ∞
z
= nz n .
(1 − z)2 n=0
Furthermore differentiation gives

X ∞
z2 + z
= n2 z n . (19.21)
(1 − z)3
n=0
Combining the expressions (19.20) and (19.21), we get
z2 + z 1 3 2
F (z) = = − + .
(1 − z)4 (1 − z)2 (1 − z)3 (1 − z)4
Using the binomial series, we have
X ∞ X ∞ X∞
1 1 1
= Cnn+1 z n , = Cnn+2 z n and = Cnn+3 z n .
(1 − z)2 (1 − z)3 (1 − z)4
n=0 n=0 n=0
Hence we obtain
cn = Cnn+1 − 3Cnn+2 + 2Cnn+3

3(n + 1)(n + 2) (n + 1)(n + 2)(n + 3)
= n+1− +
2 3
n(n + 1)(2n + 1)
= ,
6
235
Problem 19.8
Proof. Following the idea in §19.4 II, we write
Y3 4
Y
1 1 1 1
R(z) = 3
· 4
= · ,
1−z 1−z n=1
z − α m=1 z − β m
n
where α = exp( 2πi 2πi 3

3 ) and β = exp( 4 ) = i. Since R has a double pole at z = 1(= α = β ) and
4
simple poles at its remaining singularities, R has a partial fraction decomposition of the form
X bj 5
a1 a2
R(z) = + + ,
z − 1 (z − 1)2 z − zj
j=1
where z1 = α, z2 = α2 , z3 = i, z4 = −1 and z5 = −i.

Next, we know that
d 5 1
a1 = [(z − 1)2 R(z)] = , a2 = lim (z − 1)2 R(z) =
dz z=1 24 z→1 12
and 
 1

 − 2 4 , if j = 1, 2;

 3zj (1 − zj )
bj =

 1


 − , if j = 3, 4, 5.
4zj3 (1 − zj3 )
Simple calculation gives
√ √
1 3i 1 3i 1 i 1 1 i
b1 = − , b2 = + , b3 = − − , b4 = and b5 = − + .
6 18 6 18 8 8 8 8 8
Consequently, we have
√ √
1 3i 1 3i
5 1 6 − 18 6 + 18 − 81 − 8i
R(z) = + + + +
24(z − 1) 12(z − 1)2 z−α z−α z−i
1 i
1 − +
+ + 8 8. (19.22)
8(z + 1) z+i
Finally, by comparing the coefficients of z n on both sides of the expression (19.22), we achieve
√ 1 √3i
5 n + 1 1 3i n+1
C(n) = − + − − ·α − + · αn+1
24 12 6 18 6 18
1 i (−1)n 1 i n+1
+ + · (−i)n+1 + + − ·i .
8 8 8 8 8
Problem 19.9
Proof. Let z = x + iy. Then we notice that

X 1 X 1 1X 1 1 1 px
= ≤ = = · . (19.23)
npnz npnx 2 pnx 2px (px − 1) p2x 2(px − 1)
n≥2 n≥2 n≥2
√ px
Since p ≥ 2 and x > 21 , we have px > 2 which implies that 2(px −1) < 2. Therefore, the
inequality (19.23) reduces to
X 1 2 2
nz
< 2x ≤ √ < ∞. (19.24)
np p p
n≥2
As a result, we have
X 1 X 2
nz
< . (19.25)
np p2x
p prime p prime
n≥2
Suppose that K is a compact subset of Re z > 12 and F = {z ∈ C | Re z ≤ 12 }. Now F is closed

in C and F ∩ K = ∅. By [22, Exercise 21, p. 101], there exists a δ > 0 such that d(K, F ) > 2δ.
Furthermore, we have K ⊆ {z ∈ C | Re z ≥ 12 + δ}. Given a prime p ≥ 2, for every n ∈ N,
consider the function
1
fn,p (z) = nz
np
which is clearly analytic in Re z > 12 . Now it follows from Theorem 1.9 (The Weierstrass M -
X∞
Test) and the inequality (19.24) that the sequence fn,p (z) converges uniformly to a function
n=2
fp (z) in Re z > 12 . Hence we conclude from Theorem 7.6 that fp (z) is analytic in Re z > 21 .
1
Next, if Re z ≥ 2 + δ, then we get from the inequality (19.25) that
X X X 2 X X∞
1 2 1
|fp (z)| ≤ < ≤ < 2 <∞
npnz p2x p1+2δ n=1
n 1+2δ
p prime p prime p prime p prime
n≥2
which means that the series X

fp (z)
p prime
converges to a function f (z) uniformly in Re z ≥ 12 + δ, and then in K, by Theorem 1.9 (The

Weierstrass M -Test). Finally, we apply Theorem 7.6 to establish the fact that f (z) is analytic
in Re z > 12 , as desired. This completes the proof of the problem.
Index
B H
Bolzano-Weierstrass Theorem, 71, 85, 177 homotopy of path, 101
C
I
Chain Rule, 38
inversion through the unit circle, 22
D
digamma function, 225 J
Dirichlet’s Test, 28 Joukowsky transform, 189
E
L
Eneström and Kakeya Theorem, 16
Leibniz’s Rule, 93
Euler’s constant, 166
lemniscate of Bernoulli, 10
F
Fresnel integral, 149 T
Fubini Theorem, 93 Topologist’s sine curve, 18, 103
237
Index 238
Bibliography
[1] L. Ahlfors, Complex Analysis, 3rd ed., Mc-Graw Hill Inc., 1979.
[2] T. M. Apostol, Calculus Vol. 1: One-Variable Calculus, with an Introduction to Linear

Algebra, 2nd ed., John Wiley & Sons, Inc., 1967.
[3] N. H. Asmar and L. Grafakos, Complex Analysis with Applications, Cham, Springer Inter-
national Publishing, 2018.
[4] J. Bak and D. J. Newman, Complex Analysis, 3rd ed., New York, N. Y.: Springer, 2010.
[5] R. G. Bartle and D. R. Sherbert, Introduction to Real Analysis, 4th ed., John Wiley &
Sons, Inc., 2011.
[6] R. B. Burckel, An Introduction to Classical Complex Analysis, Vol. 1, Basel: Birkhäuser

Basel, 1979.
[7] J. B. Conway, Fucntions of One Complex Variable, 2nd ed., New York: Springer-Verlag,
1978.
[8] D. S. Dummit and R. M. Foote, Abstract Algebra, 3rd ed., Hoboken, N. J.: Wiley, 2004.
[9] E. Freitag and B. Rolf, Complex Analysis, 2nd ed., London: Springer, 2009.
[10] T. W. Gamelin, Complex Analysis, New York, N. Y.: Springer, 2001.
[11] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series and Products, 7th ed., San
Diego: Academic Press, 2007.
[12] G. H. Hardy, A Course of Pure Mathematics, 10th ed., Cambridge University Press, 2002.
[13] P. K. Kythe, Handbook of Conformal Mappings and Applications, Boca Raton, Florida:
CRC Press, 2019.
[14] S. Lang, Complex Analysis, 4th ed., New York, N. Y.: Springer, 1999.
[15] J. D. Lawrence, A Catalog of Special Plane Curves, New York: Dover, 1972.
[16] I. H. Lin, Classical Complex Analysis: A Geometric Approach - Vol. 1, Singapore: World
Scientific Publishing Company Co. Pte. Ltd., 2010.
[17] M. Marden, Geometry of Polynomials, 2nd ed., Mathematical Surveys, No. 3, Providence,
R. I.: Amer. Math. Soc., 1966.
[18] Q. G. Mohammad, On the zeros of polynomials, Amer. Math. Monthly, Vol. 72, No. 6, pp.
631 - 633, 1965.
[19] J. R. Munkres, Topology, 2nd ed., Upper Saddle River, N.J.: Prentice-Hall, 2000.
239
Bibliography 240
[20] T. Needham, Visual Complex Analysis, Oxford: Clarendon Press, 1997.
[21] G. Pólya and G. Szegö, Problems and Theorems in Analysis II: Theory of Functions. Zeros.
Polynomials. Determinants. Number Theory. Geometry, Berlin, Heidelberg: Springer, 1998.
[22] W. Rudin, Principles of Mathematical Analysis, 3rd ed., Mc-Graw Hill Inc., 1976.
[23] W. Rudin, Real and Complex Analysis, 3rd ed., Mc-Graw Hill Inc., 1987.
[24] E. M. Stein and R. Shakarchi, Complex Analysis, Princeton, N. J.: Princeton University
Press, 2003.
[25] T. Tao, Analysis II, 3rd ed., Singapore: Springer, 2016.
[26] K. W. Yu, A Complete Solution Guide to Principles of Mathematical Analysis, Ama-

zon.com, 2018.
[27] K. W. Yu, Problems and Solutions for Undergraduate Real Analysis I, Amazon.com, 2018.
[28] K. W. Yu, Problems and Solutions for Undergraduate Real Analysis II, Amazon.com, 2019.

A Complete Solution Guide To Complex Analysis 3rd Bak

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Complete Solution Guide To Complex Analysis 3rd Bak

Uploaded by

Copyright:

Available Formats

A Complete Solution Guide to

by Kit-Wing Yu, PhD

ISBN: 978-988-74155-0-3 (eBook)

• A Complete Solution Guide to Real and Complex Analysis I.

• Problems and Solutions for Undergraduate Real Analysis I.

• Problems and Solutions for Undergraduate Real Analysis II.

• Mock Tests for the ACT Mathematics.

• A Complete Solution Guide to Principles of Mathematical Analysis.

The features of this book are as follows:

• Necessary lemmas with proofs are provided.

so that I will post the updated errata on my website

from time to time.

1.1 The graph of z1 z2 and z1 z2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

7.1 The graph of the set S. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

8.1 The geometry of the circle C and the annulus B. . . . . . . . . . . . . . . . . . . 102

10.1 The regular closed curve C surrounding ±1. . . . . . . . . . . . . . . . . . . . . . 135

13.1 The sector S and the disc D(z0 ; δ). . . . . . . . . . . . . . . . . . . . . . . . . . . 173

14.1 The shape of a Norman window N . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

16.1 The level curves of u(x, y) = k for k ∈ (0, 21 ) ∪ ( 12 , 1). . . . . . . . . . . . . . . . . 205

19.1 The intersections of y = x and y = tan x. . . . . . . . . . . . . . . . . . . . . . . 227

List of Figures viii

1 The Complex Numbers 1

2 Functions of the Complex Variable z 23

4 Line Integrals and Entire Functions 47

5 Properties of Entire Functions 55

6 Properties of Analytic Functions 67

7 Further Properties of Analytic Functions 83

8 Simply Connected Domains 101

9 Isolated Singularities of an Analytic Function 109

10 The Residue Theorem 121

11 Applications of the Residue Theorem to the Evaluation ... 139

12 Further Contour Integral Techniques 163

13 Introduction to Conformal Mapping 173

14 The Riemann Mapping Theorem 187

15 Maximum-Modulus Theorems for Unbounded Domains 197

16 Harmonic Functions 201

17 Different Forms of Analytic Functions 211

18 Analytic Continuation; The Gamma and Zeta Functions 221

19 Applications to Other Areas of Mathematics 227

(a) It is clear that

(a) Then z1 + z2 = (a + c) + (b + d)i = (a + c) − (b + d)i = a − bi + c − di = z1 + z2 .

(d) If z = A + Bi, then we have z = A − Bi = A + Bi = z.

We complete the proof of the problem.

Proof. Suppose that z = rcis θ.

(a) For every n ∈ Z, we have z n = r n cis nθ so that |z n | = r n = |z|n .

(b) Since z = rcis (−θ), we have z · z = r 2 cis (θ − θ) = r 2 = |z|2 .

|Re z| + |Im z| ≥ |z|.

In conclusion, an equality occurs if and only if Re z = 0 or Im z = 0.

This completes the proof of the problem.

(a) By Problem 1.7(b) and then Problem 1.4(a), we have

|z1 + z2 |2 = (z1 + z2 )(z1 + z2 ) = (z1 + z2 )(z1 + z2 ). (1.1)

|z1 + z2 |2 = |z1 |2 + |z2 |2 + 2Re (z1 z2 ) (1.2)

which gives the triangle inequality.

This completes the proof of the problem.

(a) Let z1 = a + bi and z2 = c + di, where a, b, c, d ∈ Z. Thus z1 z2 = ac − bd + (ad + bc)i,

(a2 + b2 )(c2 + d2 ) = (ac − bd)2 + (ad + bc)2 . (1.4)

If we define u = ac − bd and v = ad + bc, then we get the desired result.

(a2 + b2 )(c2 + d2 ) = (ac + bd)2 + (ad − bc)2 . (1.5)

ad = bc and ac = bd. (1.6)

u2 − s2 = (u − s)(u + s) = (−2bd)(2ca) = −4abcd.

Since a, b, c and d are nonzero, we have u2 6= s2 . Similarly, we can show that

u2 − t2 = (u − t)(u + t) = (a2 − b2 )(c2 − d2 ).

By the hypotheses, we know that u2 6= t2 . Furthermore, we also have v 2 6= s2 and v 2 6= t2 .