Escolar Documentos
Profissional Documentos
Cultura Documentos
Daniel Alpay∗
January 6, 2007
Abstract
This document is an ongoing draft which contains solutions of exercises
for the service course on Introduction to the theory of functions of a
complex variable, given by the department of mathematics at Ben-Gurion
University. References are given to the books from where some of the
exercises are taken, but there is no need to look into these books. All the
relevant information is in the notes.
Contents
1 Complex numbers 3
1.1 Complex numbers and polynomials . . . . . . . . . . . . . . . . . 3
1.2 Series of complex numbers . . . . . . . . . . . . . . . . . . . . . . 11
1.3 Infinite products . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.4 Geometric interpretation . . . . . . . . . . . . . . . . . . . . . . . 19
1.5 Moebius maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1
4 Further applications of Cauchy’s formula 65
4.1 The maximum principle and Schwarz’ lemma . . . . . . . . . . . 65
4.2 Liouville’s theorem and the fundamental theorem of algebra . . . 71
4.3 Harmonic functions . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.4 Zeroes of analytic functions . . . . . . . . . . . . . . . . . . . . . 81
4.5 Analytic continuation . . . . . . . . . . . . . . . . . . . . . . . . 85
6 Conformal mappings 96
7 Problems: 98
7.1 Hyperbolic distance . . . . . . . . . . . . . . . . . . . . . . . . . 98
2
1 Complex numbers
1.1 Complex numbers and polynomials
Exercise 1.1 Let a = eiα and b = eiβ with α and β real numbers. Show that
a+b α−β
= −i cot
a−b 2
α−β
cos
a+b 2
= i .
1 − ab sin 2 α+β
At this stage of the course we do not know the fundamental theorem of algebra,
which states that every polynomial of degree n has n roots (counting multiplic-
ity). Still we do know the following: if p(z) is a polynomial and if p(z0 ) = 0,
then we can factor out (z − z0 ) from p(z), that is, we can write
p(z) = (z − z0 )q(z), (1.1)
where q(z) is a polynomial of degree n − 1. In particular, if we know that
z0 , . . . , zn−1 are the roots of p(z) (say, all different, for the present applications
below), then
n−1
Y
p(z) = an (z − zk ),
k=0
n
where an is the coefficient of z in p(z).
3
Exercise 1.3 Check (1.1).
Pn
Solution: Let p(z) = k=0 ak z k . Since p(z0 ) = 0 we have
with
n `=k−1
! n−1 n
!
X X X X
q(z) = ak z ` z0k−1−` = z` ak z0k−1−` .
k=1 `=0 `=0 k=`+1
z 4 − 5z 3 + 18z 2 − 17z + 13
1 + z + · · · + z7 = 0
1 − z2 + z4 − z6 = 0
(1 − z)n = (1 + z)n
(1 − z)n = z n .
1
Solution: Let c = maxj |cj | j . By hypothesis c > 0. Let z be a root of the
polynomial equation
z n + c1 z n−1 + · · · + cn = 0, (1.2)
4
z
and let u = . Dividing both sides of (1.2) by cn we obtain:
c
c1 n−1 cn
un + u + · · · + n = 0.
c c
But |cj | ≤ cj and so this last equation leads to:
Assume that |u| ≥ 2. Then 1/|u| ≤ 1/2. Dividing both sides of (1.3) by |u|n
leads to
1 1
1≤ + ··· + n
|u| |u|
1 1
≤ + ··· + n
2 2
< 1,
1
which is a contradiction. Thus |u| < 2, that is |z| < 2 maxj |cj | j . 2
Pn−1
Using the third identity decompose the polynomial p(z) = k=0 z 2k into ele-
mentary factors and prove the identity
n−1 √
Y kπ n
sin( ) = n−1 . (1.4)
2n 2
k=1
Solution: We prove the third equality. The roots of order 2n of unity are
2kπ ikπ
zk = exp = exp , k = 0, . . . , 2n − 1.
2n n
5
We have z0 = 1 and zn = −1. The other roots are purely imaginary and appears
in pairs since p(z) = z 2n − 1 has real coefficients (and thus, p(w) = 0 → p(w) =
0). The roots from k = 1 to k = n − 1 are all different and so the roots of p(z)
are, besides 1 and −1,
zk and zk , k = 1, . . . , n − 1.
Thus
n−1
Y
p(z) = (z + 1)(z − 1) (z − zk )(z − zk )
k=1
n−1
Y
= (z + 1)(z − 1) (z 2 − 2Re zk + 1)
k=1
kπ
which concludes the proof since Re zk = cos( ).
n
Using the formula for the sum of a geometric series we obtain
n−1
X 1 − z 2n
p(z) = z 2k = ,
1 − z2
k=0
and hence, using the previous arguments to prove the third equality and the
third equality itself we have:
n−1 n−1
Y ikπ −ikπ Y
p(z) = (z − exp )(z − exp )= (z 2 − 2Re zk + 1).
n n
k=0 k=1
Recall that
kπ kπ
1 − cos( ) = 2 sin2 ( ).
n 2n
Hence
n−1
Y kπ
n= (2 − 2 cos( ))
n
k=1
n−1
Y kπ
= 4 sin2 ( )
2n
k=1
n−1
Y kπ
= 4n−1 sin2 ( )
2n
k=1
6
and hence the result by taking the square root of both sides since the numbers
kπ
sin( ) > 0 for k = 1, . . . , n − 1. 2
2n
n−1
! n−1
! n−1
X X X
cos(a + kb) +i sin(a + kb) = (cos(a + kb) + sin(a + kb))
k=0 k=0 k=0
n−1
X
= ei(a+kb)
k=0
1 − einb
= eia
1 − eib
nb nb nb
i −i i
e 2 e 2 −e 2
= eia
b b b
i −i i
e 2 e 2 −e 2
nb b sin nb
i(a+ − )
= e 2 2 2
b
sin
2
To prove the second formula, we first recall Newton’s binomial formula:
n
n
X n
(α + β) = αk β n−k , (1.7)
j
k=0
7
n ! n !
X n X n
cos(a + kb) + i sin(a + kb) =
j k
k=0 k=0
n
X n
= ei(a+kb)
k
k=0
n !
ia
X n ikb
= e e
k
k=0
= e (1 + eib )n
ia
bn b b n
i −i i
= eia e 2 e 2 + e 2
nb n
i(a+ ) b
= e 2 2 cos
2
n nb
n b i(a+ )
= 2 cos e 2 ,
2
More generally, (see for instance []) one has for a, b, r ∈ R and n a positive
integer
n−1
X cos a − r cos(a − b) − rn cos(a + nb) + rn+1 cos(a + (n − 1)b)
rk cos(a + kb) =
1 + r2 − 2r cos b
k=0
n−1
X sin a − r sin(a − b) − rn sin(a + nb) + rn+1 sin(a + (n − 1)b)
rk sin(a + kb) =
1 + r2 − 2r cos b
k=0
8
4
X
C + iS = 2n+1
0
4
X
= (2 )n
0
1 − (2 )5
=
1 − 2
1 − 10
=
1 − 2
− 11
=
1 − 2
+1
=
1 − 2
1
=
1−
1 − ∗
=
(1 − )(1 − ∗ )
1 − ∗
=
2(1 − Re )
π π
(1 − cos ) − i sin
= 11 11 .
π
2(1 − cos )
11
π
1 sin
Hence C = and S = − 11 2
2 π .
2(1 − cos )
11
In the same vein, but more difficult is the following exercise, taken from the
same book ([12, p. 79]).
9
Exercise 1.11 Let
an = {k ∈ N ; , 0 ≤ 3k ≤ n}
bn = {k ∈ N ; , 0 ≤ (3k + 1) ≤ n}
cn = {k ∈ N ; , 0 ≤ (3k + 2) ≤ n} .
Show that
X n 2n + 2(−1)n cos( 2nπ
3 )
= (1.11)
3k 3
k∈an
X 2n + 2(−1)n cos( 2(n+1)π
n 3 )
= (1.12)
3k + 1 3
k∈bn
X n 2n + 2(−1)n cos( 2(n+2)π
3 )
= (1.13)
3k 3
k∈cn
10
1.2 Series of complex numbers
Related to this section see also Exercises 3.28 and 3.29 in Section 3.4.
converge?
Solution: For θ = 0 (mod 2π) we have eikθ = 1 and the series diverges. Let us
now assume that θ 6= 0 (mod 2π). Then |1 − eiθ | > 0 and we have
m
X 1 − e(m+1)θ
| eikθ | = | |
1
1 − eiθ
2 def.
≤ iθ
= M.
|1 − e |
Thus condition (a) in Abel’s theorem is met. The sequence √1n decreases and
so the second condition is also met, and the series converges. 2
The books of Polya and Szegö [16]–[?] are a mine of exercises, most of them
quite challenging; here are some of them (see [16, Exercises 36, 37, 38 p.110]).
11
2
P
Since n≥1 zn converges we have that the series
X X
(x2n − yn2 ) = Re zn2
≥1 n≥1
(1.14) does not necessarily hold when the xn are not in the right half–plane. take
(−1)n P
for instance xn = √ . The series n≥1 xn converges (use Abel’s theorem),
n
but n≥1 x2n is the harmonic series and is divergent. To get a counterexample
P
to the claim when the xn are not in the right half–plane, take then for instance
einθ
zn = √ n = 1, 2, . . .
n
12
converge. But the series with term |zn |2 = diverge. 2
n
Exercise
P k 1.15 Find a sequence P of complex numbers such that all the series
k
z
n n converge but all the series n |zn | diverge, k = 1, 2, . . ..
e2πiθn
zn = ,
ln(n + 1)
where θ is irrational. 2
Exercise 1.16 Let 0 < α < π/2 and let zn = ρn eiθn be a sequence of complex
numbers all different from 0 and such that
−α ≤ θn ≤ α.
P P
Show that the series n zn and n |zn | converge or diverge at the same time.
12
Solution: (see [16, p. 305]). Let zn = xn + iyn . By hypothesis
xn
|zn | ≤ ,
cos α
from which follows the result. 2
The following is a classical exercise (as most of the exercises presented in these
notes);
Exercise 1.17 Compute, for r real number of absolute value strictly less than
1 and θ ∈ R, the sums
X∞
rn sin(nθ)
0
and
∞
X
rn cos(nθ).
0
There are (at least two ways to prove this formula). The easiest (but logically not
at this place in the notes) is to resort to the general theorem on differentiation
of power series in their disk of convergence and to differentiate the power series
of 1/(1 − z). The more direct one is as follows: start from
1 − z N +1
1 + z + · · · + zN = .
1−z
This is an identity between two polynomials and we can differentiate to obtain
(N + 1)z N (1 − z) + z N +2
2 N −1 1
1 + z + 2z + · · · + N z = − .
(1 − z)2 (1 − z)2
You may also decide to prove the above formula by induction if you do not allow
differentiation at this stage. Then multiply both sides by z and let N → ∞
13
to obtain (1.17). This equation will be instrumental to compute (1.15). To
compute (1.16) use
∞
X 2
n(n − 1)z n−2 = .
2
(1 − z)3
14
1.3 Infinite products
Exercise 1.18 Let |q| < 1. Show that the infinite product
∞
Y
(1 − q ` z)
`=1
Solution: We have
n
X
an = a1 + (a` − a`−1 )
`=2
and hence
n
X ∞
X
|an | ≤ |a1 | + |a` − a`−1 | ≤ |a1 | + |a` − a`−1 |
`=2 `=2
Since a1 = 0 we have
∞
X
g(z) = 1 + z n (−an−1 + an ),
n=2
15
and so, for |z| < 1, we have:
∞
X
|g(z) − 1| ≤ |z|n | − an−1 + an |
n=2
∞
X
≤ |z|2 | − an−1 + an |.
n=2
for n ≥ n0 . Thus, by theorem, the product converges and vanishes only at the
points z = 2, 3, . . .. 2
Remark 1.20 Sequences which satisfy (1.18) are called of bounded variation.
For instance, the sequence 1, 1/2, 1/3, . . . is of bounded variation.
The following exercise is taken from the book of Choquet on topology [3, p
315]. The purpose of the exercise is to build a bi–periodic function on C (thus
it has a lattice of periods). In a later chapter we will see the relationships of
this exercise to the theory of elliptic functions. See Exercise ??.
a1 · · · an = b1 · · · bn ,
S(a1 z) · · · S(an z)
and let M (z) = . Show that M (kz) = M (z).
S(b1 z) · · · S(bn z)
(e) Set G(z) = M (ez ). What can we say about G?
Solution: Since |k| > 1 the series with general term z/k n is absolutely con-
vergent for any z ∈ C. Thus, by the result proved in class the infinite product
16
converges in C∗ for every z not equal to −k n , n = 1, 2, . . .
We now turn to (c). From (b) we have P (k/z) = (1 + 1/z)P (1/z), and replacing
z by kz in the above expression,
1 kz
P (1/z) = (1 + )P (1/kz) and hence P (1/kz) = P (1/z).
kz 1 + kz
Thus:
S(kz) = P (kz)P (1/kz)(1 + kz)
kz
= (1 + z)P (z)P (1/z) (1 + kz)
1 + kz
= (1 + z)P (z)P (1/z)kz
= kzS(z).
since we assumed a1 · · · an = b1 · · · bn .
(e) Let ω ∈ C be such that k = exp ω. Since |k| > 1 the numbers ω and 2πi are
linearly independent over Z. We cannot find m and n such that mω + 2πin = 0.
Indeed, if there are such m and n, then emω = e−2πin = 1 and so k m = 1
contradicting the assumption |k| > 1. Moreover, we have
where we used (d) with ez in place of z. Thus, G(z) is bi–periodic since ω and
2π are linearly independent over Z. 2
17
The next exercise deals with a very important family of functions, called Blaschke
products.
Exercise 1.22 Let z0 , z1 , . . . be a series of numbers in the open unit disk dif-
ferent from 0 and such that
∞
X
(1 − |zn |) < ∞. (1.19)
n=0
We have:
|zn | zn − z
an (z) = −1
zn 1 − zzn
|zn |zn − |zn |z − zn + |zn |2 z
=
zn (1 − zzn )
(|zn | − 1)(|zn |z + zn )
=
zn (1 − zzn )
|zn |
(|zn | − 1)( z + 1)
zn
= .
(1 − zzn )
Thus, for |z| < 1,
2
|an (z)| ≤ (1 − |zn |)
1 − |z|
and (1.21) holds since (1.19) holds. 2
Remark 1.23 Products of the form (1.20) are called Blaschke products, and
play an important role in complex analysis; see for instance [9], [6], [18].
18
1.4 Geometric interpretation
Exercise 1.24 Characterize and draw the sets of points in the plane R2 such
that:
a) |z − 1 + i| = 1.
b) z 2 + z 2 = 2.
c) |z − i| = |z + i|.
d) |z|2 + 3z + 3z + 5 = 0.
e) |z|2 + 3z + z + 5 = 0.
f ) |z| > 1 − Re z.
g) z 2 + 3z + 3z + 5 = 0.
√
h) Re (z(1 − i) < 2.
Solution: Cases (a), (b) are circles. More precisely, (a) is the circle with center
(1, −1) and radius 1. Equation (b) can be rewritten as x2 + y 2 = 1, and so we
obtain the unit circle. Case (c) is the line orthogonal to the interval (0, 1) and
(0, −1) and passing by the middle of this interval, i.e. it is just the real line.
More misleading are (d), (e) and (g). The equations look like the equation of a
circle but this is not the case. For (d) we have the empty set. Indeed, we have
x2 + y 2 + 4x + y + 5 + iy = 0
x2 + y 2 + 4x + y + 5 = 0 and y = 0.
|z − z0 | = λ|z − z1 |
19
where λ > 0, λ 6= 1 and z0 6= z1 is the circle with center and radius
z0 − λ 2 z1 λ
and |z0 − z1 |
1 − λ2 1 − λ2
respectively.
20
1.5 Moebius maps
Exercise 1.26 Let w be of modulus strictly less than 1. Prove that
z−w
1 − zw < 1 ⇐⇒ |z| < 1.
z−ν
Exercise 1.27 For |ν| < 1, define bν (z) = . Show that
1 − zν
Exercise 1.28 Given two triples of complex numbers (z1 , z2 , z3 ) and (w1 , w2 , w3 ),
show that the map z 7→ w defined by
w − w1 w3 − w2 z − z1 z3 − z2
· = · (1.22)
w − w2 w3 − w1 z − z2 z3 − z1
is such that w(z` ) = w` for ` = 1, 2, 3.
z − z1 z3 − z2 w3 − w1
w1 − w2 · ·
z − z2 z3 − z1 w3 − w2
w(z) = z − z1 z3 − z2 w3 − w1
1− · ·
z − z2 z3 − z1 w3 − w2
z3 − z2 w3 − w1
(z − z2 )w1 − w2 (z − z1 ) · ·
z3 − z1 w3 − w2
= z3 − z2 w3 − w1 .
(z − z2 ) − (z − z1 ) · ·
z3 − z1 w3 − w2
Hence, z 7→ w(z) is indeed a Moebius map. One checks that w(z` ) = z` for
` = 1, 2, 3 by direct computation:
z3 − z2 w3 − w1
(z1 − z2 )w1 − w2 (z1 − z1 ) · ·
z3 − z1 w3 − w2
w(z1 ) = z3 − z2 w3 − w1
(z1 − z2 ) − (z1 − z1 ) · ·
z3 − z1 w3 − w2
(z1 − z2 )w1
=
(z1 − z2 )
= w1 .
21
z3 − z2 w3 − w1
(z2 − z2 )w1 − w2 (z2 − z1 ) · ·
z3 − z1 w3 − w2
w(z2 ) = z3 − z2 w3 − w1
(z2 − z2 ) − (z2 − z1 ) · ·
z3 − z1 w3 − w2
z3 − z2 w3 − w1
−w2 (z2 − z1 ) · ·
z3 − z1 w3 − w2
= z3 − z2 w3 − w1
−(z2 − z1 ) · ·
z3 − z1 w3 − w2
= w2 .
z3 − z2 w3 − w1
(z3 − z2 )w1 − w2 (z3 − z1 ) · ·
z3 − z1 w3 − w2
w(z3 ) = z3 − z2 w3 − w1
(z3 − z2 ) − (z3 − z1 ) · ·
z3 − z1 w3 − w2
w3 − w1
(z3 − z2 )w1 − w2 (z3 − z2 ) ·
w3 − w2
= w3 − w1
(z3 − z2 ) − (z3 − z2 ) ·
w3 − w2
w3 − w1
w1 − w2 ·
w3 − w2
= w3 − w1
1−
w3 − w2
w1 (w3 − w2 ) − w2 (w3 − w1 )
=
w3 − w2 − w3 + w1
w3 (w1 − w2 )
=
w1 − w2
= w3 .
2
Exercise 1.29 For which k ∈ R is the image of the circle |z − 1| = k under the
z−3
Moebius map f (z) = a line? Find the equation of the line.
1 − 2z
z−3 w+3
Solution: Set w = . Then, z = . The condition |z − 1| = k
1 − 2z
1 + 2w
w+3
becomes thus − 1 = k, i.e.,
1 + 2w
|w + 3 − (1 + 2w)| = k|1 + 2w|,
which can be rewritten as
1
|2 − w| = 2k|w + |.
2
22
1
Hence, we obtain a line if and only if k = . The equation of the line is
2
1 3
|2 − w| = |w + |, i.e. x = .
2 4
If one wants only the value of k but not the equation of the line, a shorter way
is as follows: we know that the image of the circle is either a line or a circle.
1
It will be a line if and only if it is not a bounded set, i.e. if and only if z =
2
1 1
belongs to the circle |z − 1| = k, i.e. | − 1| = k. Hence k = . 2
2 2
az + b
Exercise 1.30 Show that the non-trivial Moebius map w = ϕ(z) =
cz + d
maps the open unit disk into itself if and only if
At this stage we pause and remark the following: we necessarily have |c| < |d|.
Indeed, if |d| = |c|, the above can be rewritten as
(bd − ac) 6= 0.
If it is equal to 0, then on the one hand (1.25) leads to |b| < |a| and on the other
hand we have
bd − ac = 0 =⇒ |b||d| = |a||c|
|a|
=⇒ |d| = |c|
|b|
which together with |d| = |c| leads to |a| = |b|.
23
d
If |c| > |d|, the point −is in D, and ϕ has a pole at this point. Thus the image
c
of D by ϕ cannot be bounded. So |c| < |d|. We divide both sides of (1.24) by
|d|2 − |c|2 and obtain
|b|2 − |a|2
2 bd − ac
|w| − 2Re w + < 0.
|d|2 − |c|2 |d|2 − |c|2
Completing the square we obtain
2 2 2
2
w − bd − ac < − |b| − |a| + bd − ac .
|d|2 − |c|2 |d|2 − |c|2 |d|2 − |c|2
We have:
2
|b|2 − |a|2 bd − ac −(|b|2 − |a|2 )(|d|2 − |c|2 ) + |bd − ac|2
− 2 + =
|d| − |c|2 |d|2 − |c|2 (|d|2 − |c|2 )2
|ad − bc|2
= .
(|d|2 − |c|2 )2
Thus the image of the open unit disk is the circle of center
bd − ac
w0 =
|d|2 − |c|2
and radius
|ad − bc|
r0 = .
|d|2 − |c|2
This circle will be included in the open unit disk if and only if
|w0 | + r0 ≤ 1,
Multiplying both sides by the strictly positive number |d|2 − |c|2 we obtain
(1.23). 2
a = 1, , b = −w, c = −w, d = 1,
24
2 The Cauchy–Riemann equations and related
topics
2.1 Continuous functions
Any nonzero complex number z admits a logarithm, that is there always exists
a number w such that z = exp w. To see this write z = ρ exp(iθ). It suffices to
take
w = ln ρ + iθ.
On the other hand, it is false that there always exists a continuous logarithm
function on a given set Ω. The same problem holds for square root functions.
These are illustrated in the following exercise, taken from [8, p. 41]
Exercise 2.1 (a) Show that there is no function from C \ {0} into itself such
that
f (zw) = f (z)f (w)
(2.1)
f (z)2 = z
25
In the first case g(z) ≡ 1 implies that f (z)f (w) = f (zw) and we obtain a con-
tradiction with (a). In the second case it suffices to replace f by −f to obtain
a contradiction.
(c) Assume that there is a function h(z) continuous on C \ {0} such that
exp(h(z)) = z for all z ∈ C \ {0}. Set g(z) = exp(h(z)/2). Then h(z) is
continuous on C \ {0} and h(z)2 = z, a contradiction with (b).
where z = ρ exp(iθ) and where −π < θ < π is continuous (in fact differentiable,
with derivative 1/z) and satisfies exp(ln z) = z in C \ {0}. The difference be-
tween c) and (d) is that the set C \ {0} is not simply connected whereas C \ R−
is simply connected.
We will prove in the sequel that in any simply connected domain a non-vanishing
analytic function (that is a function which has everywhere a derivative) has an
analytic (and in particular continuous) logarithm. We will also give necessary
and sufficient conditions for an analytic logarithm to exist when the set is not
simply connected.
2
26
2.2 Cauchy–Riemann equations and applications
We know from the course that the Cauchy–Riemann equations at a point do
not imply differentiability at the given point. The next three exercises illustrate
this phenomenon. In all three exercises, the point under consideration is z = 0.
The first example is not very strong since the function is not even continuous at
the point where the Cauchy–Riemann equations hold. The second is a bit more
involved. The function is continuous, but all radial derivative at the origin (the
point where the Cauchy–Riemann equations hold)
f (eiθ ) − f (0)
lim , where θ is fixed,
→0 eiθ
exist, but depend on the angle. In the last exercise, the function is continuous
all radial derivative are the same at the origin, but the function still is not dif-
ferentiable.
Differentiability in an open set is not insured even if one assumes that the
Cauchy–Riemann equations hold on the given open set. Still, it is well to recall
the not so well known Looman–Menchoff theorem (see e.g. [15, Chapter I, §6].
If f is continuous in the open set Ω and if the Cauchy–Riemann equations hold
there, then f is analytic in Ω.
Exercise 2.2 (see [8, p. 66]). Let
exp − 1 , if z 6= 0
f (z) = z4
0 if z = 0
iπ
1
while for z = ρe 4 , f (z) = exp 4 and so
ρ
iπ
lim f (ρe 4 ) = ∞
ρ→0
27
Since f (z) is not continuous at the origin in particular it is not differentiable.
We show that the Cauchy–Riemann equations hold at z = 0. Write
1
f (z) = exp − = u(x, y) + iv(x, y).
(x + iy)4
We have
1
u(x, 0) = exp − 4 ,
x
v(x, 0) ≡ 0,
1
u(0, y) = exp − 4 ,
y
v(0, y) ≡ 0.
The next exercise is quite standard; it can be found e.g. in [4, Exercise 7 p. 48].
Exercise 2.3 Let 2
z
if z 6= 0
f (z) = z
0 if z = 0.
Show that:
(a) f is continuous at the origin.
(b) The Cauchy–Riemann equations hold at the origin.
(c) f is not differentiable at the origin.
Solution: Since
z2
|f (z) − f (0)| = | − 0|
z
|z 2 |
=
|z|
= |z|,
28
we get that
lim |f (z) − f (0)| = lim |z| = 0.
z→0 z→0
Hence f is continuous at the origin.
Write
(x − iy)2
f (z) = = u(x, y) + iv(x, y).
x + iy
Since
f (x) = x and f (iy) = iy
we obtain
u(x, 0) = x, v(x, 0) = 0, u(0, y) = 0, and v(0, y) = y,
for x 6= 0 and y 6= 0. Moreover, since f (0) = 0, we have u(0, 0) = v(0, 0) = 0.
Thus
u(x, 0) − u(0, 0) v(0, y) − u(0, 0)
lim = 1 and lim = 1.
x→0 x y→0 y
∂u ∂v
Thus (0, 0) and (0, 0) exist and are equal to 1. Hence the first Cauchy–
∂x ∂y
Riemann equation holds at z = 0. The second one is proved in the same way.
We now prove that f is not differentiable at the origin. Write z = eiθ . Then,
f (z) − f (0) z2 2 e−2iθ
= 2 = 2 2iθ = e−4iθ .
z z e
Hence, for a fixed θ
f (z) − f (0)
lim = lim e−4iθ = e−4iθ .
→0 z →0
f (z) − f (0) π
Thus limz→0 = 1 when θ = 0 and = −1 when θ = . So
z 4
f (z) − f (0)
limz→0 does not exists and f is not differentiable at the origin.
z
2
29
Solution: Recall that for any two real numbers a and b,
2|ab| ≤ a2 + b2 . (2.2)
We have
|x|3 |y| p 2 1p 2
|f (z) − f (0)| = 6 2
x + y2 ≤ x + y2 ,
x +y 2
where we have used (2.2) with a = x3 and b = y. Thus
1p 2 |z|
lim |f (z) − f (0)| ≤ lim x + y 2 = lim = 0.
z→0 z→0 2 z→0 2
Thus f is continuous at the origin. This proves (a). To prove (b) we compute
with θ fixed and z = eiθ (so that x = cos θ and y = sin θ):
We now show that f is not differentiable at the origin. Take z = t + it3 . Then
f (z) − f (0) t3 t3 (t3 − it) 1 t2 − i
= 6 3
= .
z 2t (t + it ) 2 1 + it2
This expression tends to −i/2 when t → 0 and so is different from the limit
along the rays and so f is not differentiable at the origin.
Finally, since the real and imaginary part of f (z) are given by
x3 y 2 −x4 y
u(x, y) = and v(x, y) = ,
x6+ y2 x6 + y 2
we have
Hence the various partial derivatives of first order all exist at the origin and are
equal to 0, that is, the Cauchy–Riemann equations hold at the origin. 2
after these exercises, the following is quite easy: (see [10, Execice 1 p. 12]).
30
Exercise 2.5 Can a function of a complex variable which is defined and con-
tinuous for |z| < 1 be such that it is only differentiable at the origin? Can a
function which is continuous in a region be differentiable only along certain lines
of that region?
The next example is on the positive side:
Exercise 2.6 Where is the function
f (z) = x2 + y 2 + 2ixy
differentiable, and compute the derivative at these points.
Solution: We have f (z) = u(x, y) + iv(x, y) with
u(x, y) = x2 + y 2 and v(x, y) = 2xy.
The functions u(x, y) and v(x, y) are polynomials and in particular differentiable
(in the sense of functions of two real variables) everywhere and so the Cauchy–
Riemann equations give a necessary and sufficient condition (as opposed to only
necessary in general) for f (z) to be differentiable (in the complex variable sense).
We have
∂u
(x, y) = 2x,
∂x
∂u
(x, y) = 2y,
∂y
∂v
(x, y) = 2y,
∂x
∂v
(x, y) = 2x.
∂y
Hence the Cauchy–Riemann equations are
2x = x and 0 = −y.
Thus y = 0, and x is arbitrary, and the Cauchy–Riemann equations are satisfied
at any real point. The function f (z) is differentiable on the real line. By the
∂u(x, y) ∂u(x, y)
formula F 0 (z) = −i we have
∂x ∂y
f 0 (x) = 2x − i2y = 2x since y = 0.
2
31
Solution: Here u(x, y) = x2 and v(x, y) = xy, and as in the previous exer-
cise, the existence of a derivative at a given point is equivalent to the Cauchy–
Riemann equations being satisfied at that point. We have now
We have
∂u
(x, y) = 2x,
∂x
∂u
(x, y) = 0,
∂y
∂v
(x, y) = y,
∂x
∂v
(x, y) = x.
∂y
Hence the Cauchy–Riemann equations are
2x = y and 0 = −x.
Thus x = y = 0 and there are satisfied only at the origin. The function f (z) is
∂u(x, y) ∂u(x, y)
differentiable only at z = 0. By the formula F 0 (z) = −i we
∂x ∂y
0
have f (0) = 0. 2
32
We have
1 0 0 −1
0 1 1 0 2 2
det
a 0 b 0 = a + b > 0
0 a 0 b
since a and b are not simultaneously equal to 0. It follows that
∂u ∂u ∂v ∂v
(x, y) = (x, y) = (x, y) = (x, y) = 0,
∂x ∂y ∂x ∂y
and u and v are constant in Ω since Ω is connected. So f is a constant function.
2
33
2.3 Derivatives
Exercise 2.10 Let f be a continuous function on the interval [0, 1]. From the
definition of the derivative show that the function
Z 1
F (z) = eizt f (t)dt
0
For the next exercise, see [2, Exercise 3.41 p. 79]. The aim is to show that Show
that continuous logarithm on an open set is analytic.
34
Exercise 2.11 Let Ω be an open set and f be a function continuous on Ω and
such that
exp(f (z)) = z, z ∈ Ω.
Show that f is analytic in Ω.
35
2.4 Analytic functions
The following exercise is taken from []. For yet another proof, see, which uses
much deeper machinery, but allows f to vanish in Ω, see Exercise 5.13. see
Exercise 5.13.
f (z)2 − f (z0 )2
By hypothesis the limit limz→z0 exists and, in view of the con-
z − z0
tinuity of f , the limit limz→z0 f (z) + f (z0 ) = 2f (z0 ). Since f (z0 ) 6= 0 we have
f (z) − f (z0 )
that limz→z0 exists, and is equal to
z − z0
f (z)2 − f (z0 )2
limz→z0
z − z0
.
2f (z0 )
2
Solution: The idea is to show that ϕ is equal to a power series centered at the
origin and with radius of convergence at least 1 (in fact, for x not identically
equal to 0, the radius of convergence is exactly 1, but we will not prove it here).
We then conclude by the theorem on the analyticity of power series in their disk
of convergence. The key is to use Weierstrass theorem to interchange integra-
tion and infinite sum at an appropriate place of the argument.
36
For |z| < 1 we can write
eit + z eit − z + 2z
it
=
e −z eit − z
2z
= 1 + it
e −z
1
= 1 + 2ze−it
1 − e−it z
∞
!
X
−it −it n
= 1 + 2ze (e z)
n=0
∞
X
=1+2 z n+1 e−i(n+1)t ,
n=0
and so
∞
eit + z X
m(t) = m(t) + fn (t),
eit − z n=0
where
fn (t) = 2z n+1 e−i(n+1)t m(t).
Then fn is a continuous function of t in the interval [0, 2π] and
We have used Weierstrass’s theorem to go from the last to one line to the last
line.
37
By definition of fn we have
Z 2π
fn (t)dt = 2z n+1 an+1 ,
0
1
with Z 2π
aj = e−ijt m(t)dt.
0
Then,
2π
X
ϕ(z) = a0 + 2an+1 z n+1 , (2.6)
n=0
The argument above was made for any |z| < 1. Thus (2.6) converges for all
|z| < 1 and by the theorem on the analyticity of convergent power series in
their disk of convergence, ϕ is analytic in the open unit disk. 2
(2) Assume that the function m is real valued and that moreover it takes positive
values: m(t) ≥ 0 on [0, 1]. Show that
38
2.5 Elementary functions
Exercise 2.16 Show that
sin(2x) sinh(2y)
tan z = +i .
cos(2x) + cosh(2y) cos(2x) + cosh(2y)
See also Exercise 4.13 in connection with the above exercise.
The following exercise is also a classic. To attack it you have to bear in mind
the following: all polynomial trigonometric identities (by this I mean identities
which do not involve fractional powers or infinite series), which are correct on
the real line still hold in the complex plane. One way to see this is using the
notion of analytical continuation, which we have not yet studied (see Section
4.4). The other, more direct and elementary, consists in checking directly the
presumed identity from the definitions
exp(iz) + exp(−iz) exp(iz) − exp(−iz)
cos z = , sin z = ,
2 2i
and the fact that the exp function is multiplicative. For instance to prove that
Show directly (that is, without resorting to the maximum principle) that | sin z|
has no local maximum.
39
Solution: If z = 0, then fn (0) = 1 for every n and so is the limit. We now
assume that z 6= 0. Set
fn (z) = cos(z/2) cos(z/22 ) · · · cos(z/2n ).
We use the identity (2.7) to show by induction that
sin z
fn (z) = .
2n sin(z/2n )
For n = 1 the claim is true since
sin(z)
f1 (z) = cos(z/2) = .
2 sin(z/2)
Assume the claim true at n. Then
sin(z/2n+1 )fn+1 (z) = fn (z) sin(z/2n+1 ) cos(z/2n+1 )
sin(z/2n )
= fn (z)
2
sin(z) sin(z/2n )
= n n
2 sin(z/2 ) 2
sin(z)
= n+1 ,
2
where we have used (2.7) to go from the first line to the second and the induction
hypothesis to go from the second line to the third. Thus
sin(z) 1
fn+1 (z) = ,
2n+1 sin(z/2n+1 )
and the induction hypothesis is true at rank n + 1. since it holds at rank n = 1
it holds for every positive integer.
Thus
−1
sin(z/2n+1 )
sin(z)
lim fn (z) = lim .
n→∞ n→∞ z z/2n+1
The function sin(z) is differentiable at every point z and its derivative is sin(z).
Thus, by definition of the derivative
sin(z/2n+1 )
sin0 (0) = lim ,
n→∞ z/2n+1
and so
sin(z)
lim fn (z) = , (z 6= 0).
n→∞ z
sin(z)
The formula also holds for z = 0 when one extends the function to be 1
z
at the origin. 2
40
3 Cauchy’s theorem and first applications
3.1 Cauchy’s theorem and integrals
Exercise 3.1 Compute C (x2 − iy 2 )dz where C is the upper semicircle: z(t) =
R
where the integral are along any path linking the points z1 to z2 .
The next exercise is taken from []
Exercise 3.3 Let γ(t) = x(t) + iy(t) be a path and let γ ∗ be the path given by
def.
γ ∗ (t) = γ(t)∗ = x(t) − iy(t).
Show that Z Z
f (z)dz = f (z)dz. (3.1)
γ γ∗
In the following three exercises, the Jordan lemma plays an important role. It
reads as follows:
41
Lemma 3.4 It holds that
Z π2 Z π
lim e−R sin t dt = 0 and lim e−R sin t dt = 0.
R→∞ 0 R→∞ 0
It is a direct consequence of the Jordan inequality (see for instance [17, §19.5
p.224], [?, p. 114]):
2 sin t π
≤ ≤ 1 where 0 < t ≤ ,
π t 2
which leads to Z π
π
e−R sin t dt < .
0 R
See [4, p. 187]
The Fresnel integrals
Z Z
cos(t2 )dt and sin(t2 )dt (3.3)
R R
sin 1 R R sin(t2 )
The first term tends to - as R → ∞ and the integral 1 dt in
2 4t2
absolutely convergent since
sin(t2 )
1
4t2 ≤ 4t2 .
R∞
Thus 0 cos(t2 )dt converges. The proof of the convergence of the other Fresnel
integral is of course similar.
Exercise 3.5 For R > 0 consider the closed contour ΓR = γ1,R + γ2,R + γ3,R
where:
(i) γ1,R is the interval [0, R].
42
(ii) γ2,R is the arc of the circle of radius R and centered at the origin, with
angle varying between 0 and π/4.
iπ
(iii) γ3,R is the interval linking the origin to the point R exp .
4
Compute the Fresnel integrals by computing the integral of the function exp(−z 2 )
over the contour ΓR and letting R → ∞. It is assumed known that
Z √
2 π
e−t dt = .
R 2
that is,
Z Z Z
exp −z 2 dz + exp −z 2 dz + exp −z 2 dz = 0, ∀R > 0. (3.4)
γ1,R γ2,R γ3,R
π
exp −z 2 dz = 0. Indeed, for θ ∈ [0,
R
We now show that limR→∞ γ2,R
] we have
4
4
cos(2θ) ≥ 1 − θ. (3.5)
π
43
Thus
Z Z π
exp −z 2 dz = 4 exp − R2 (cos(2θ) + i sin(2θ)) iR exp(iθ)dθ
γ2,R 0
Z π
≤ R 4 exp − R2 cos(2θ) dθ
0
Z π
4 2 4θ
≤R exp −R (1 − ) (where we use (3.5))
0 π
Z π
4θ
= R exp(−R ) 4 exp R2
2
dθ
0 π
π
2 θ=
π 4R θ 4
= R exp(−R2 ) 2 exp
4R π θ=0
2 π 2
= R exp(−R ) 2 exp(R ) − 1
4R
π
1 − exp(−R2 )
=
4R
→ 0 as R → ∞.
R∞ 2
Since limR→∞ γ1,R exp(−z 2 )dz = 0 e−t dt < ∞, the limit
R
Z
lim exp(−R2 )dz
R→∞ γ3,R
i.e Z √
2 π
exp(−z )dz = − .
γ3,R 4
But
Z Z R
2 iπ 2 iπ
exp(−z )dz = exp − exp( )(R − t) (−1) exp( )dt
γ3,R 0 2 4
Z R
1+i
=( √ ) exp(−i(R − t)2 )(−1)dt
2 0
Z ∞
1+i
→ −( √ ) (cos(t2 ) − i sin(t2 ))dt as R → ∞,
2 0
where to go from the last-to-one line to the last line we made the change of
variable t− 7→ R − t. Hence
Z ∞ Z ∞ √
1+i 2 2 π
−( √ ) cos(t )dt − i sin(t )dt = − .
2 0 0 4
44
Thus
∞ ∞
Z Z r
π
cos(t2 )dt = sin(t2 )dt = ,
0 0 32
r
π
and the Fresnel integrals are equal to twice this number, i.e. . 2
8
We note that
Z ∞ Z ∞ Z ∞
2 2 2
e−t cos t2 dt + i e−t sin t2 dt = e−t (1−i)
dt.
0 0 0
This suggests to take the following closed contour ΓR = γ1,R +γ2,R +γ3,R where:
(i) γ1,R is the interval [0, R].
(ii) γ2,R is the arc of the circle of radius R and centered at the origin, with
angle varying between 0 and π/8.
(iii) γ3,R is the interval linking the origin to the point R exp iπ
8 .
By Cauchy’s theorem,
Z
2
e−z (1−i)
dz = 0 for all R > 0,
ΓR
Thus Z Z
−z 2 (1−i) 2
lim e dz = − lim e−z (1−i)
dz.
R→∞ γ1,R R→∞ γ3,R
45
A parametrization of γ3,R is given by
π
γ(t) = ei 8 (R − u), u ∈ [0, R],
R ∞ sin t
Exercise 3.7 Compute 0
dt using Cauchy’s theorem as follows: inte-
t
eiz
grate the function on the closed contour defined below and let R → ∞ and
z
→ 0. The contour is built from four parts (both and R are strictly positive
numbers):
(i) γ1,R, is the real interval [, R].
(ii) γ2,R is the half-circle of radius R, centered at the origin, and which lies in
the upper half-plane.
46
(iii) γ1,R, is the real interval [−R, −].
(iv) γ4, is the half-circle of radius , centered at the origin, and which lies in
the upper half-plane.
R
Solution: An abridged solution is as follows: First consider γ2,R f (z)dz. We
have
Z Z
π it
iR cos t−R sin t iRe
f (z)dz = e dt
Reit
γ2,R 0
Z π
e−R sin t dt
≤
0Z π
2
−R sin t
= 2 e dt
0
−→ 0
by Jordan’s lemma.
Thus
−R R
eit eit
Z Z
lim dt + dt = π
→0 − it it
R→∞
But
−R R R
eit eit
Z Z Z
sin t
dt + dt = 2 dt,
− it it t
and hence the result: Z ∞
sin t π
dt = . (3.6)
0 t 2
2
47
3.2 Primitives, logarithms and squareroots
Exercise 3.8 Let Ω be the complex plane from which are removed the half lines
x = k, x ≥ 0 for k = 0, 1, 2, 3, 4, . . . , 2007. Show that there is a function analytic
in Ω such that
f (z)2007 = z(z − 1)(z − 2)(z − 3)(z − 4) · · · (z − 2007).
Q2007
Solution: The set Ω is simply connected. The function `=0 (z − `) does not
vanish there. Thus, it has an analytic logarithm and analytic roots of any order.
2
sin z
Exercise 3.9 The function has no primitive in C \ {0}.
z2
Solution: Indeed, by Cauchy’s formula
Z
sin z
2
dz = 2πi(cos 0) 6= 0.
|z|=1 z
sin z
Note that more generally, for p ∈ N the function will have a primitive in
zp
C \ {0} if and only if p is odd.
Exercise 3.10 Does the function
1
f (z) =
z2 + 1
has a primitive in Ω where:
(i) Ω = C \ {−i, i}.
(ii) Ω = C \ [−i, i],
where [−i, i] denotes the closed interval [−i, i], ,that is;
[−i, i] = {−i + 2ti , t ∈ [0, 1]} .
(iii) Ω = C \ {z = iy , y ∈ R and |y| ≥ 1} . In this last case, show that on the
real line the primitive is F (x) = arctan(x) when we fix F (0) = 0
z
Same questions for the function . In case (iii), show that on the real line
z2 +1
1
F (x) = ln(x2 + 1).
2
Exercise 3.11 Let a, b ∈ C and p, q ∈ N. Find a necessary and sufficient
condition for the function
sin z cos z
f (z) = a −b q
zp z
48
to have a primitive in C \ {0}.
Exercise 3.13 (a)Let Ω be a connected open set and let f be analytic in Ω and
such that
|1 − f (z)| < 1 ∀ z ∈ Ω.
Show that
f 0 (z)
Z
dz = 0
γ f (z)
for all closed contours γ in Ω.
(b) Show that f has an analytic logarithm in Ω.
Exercise 3.14 Let f analytic and not vanishing in r0 < |z − z0 | < r1 . Show
that for r0 < r < r1 ,
f 0 (z)
Z
1
∈ Z.
2πi |z−z0 |=r f (z)
Solution: Without loss of generality and to keep the notation simple we set
z0 = 0 and r = 1. The integral to compute is
Z 2π 0 it
1 f (e ) it
e dt.
2π 0 f (eit )
Recall that
f (eis )0 = ieis f 0 (eis ). (3.7)
Set for 0 ≤ s ≤ 2π
s
f 0 (eit ) it
Z
1
g(s) = f (eis ) exp −i e dt .
2π 0 f (eit )
49
and hence the result. 2
Exercise 3.15 Let f be analytic in the annulus 1 < |z| < 2 and not vanishing
there. Show that there is an integer n ∈ Z and a function g analytic in 1 <
|z| < 2 such that
f (z) = z n eg(z) .
Similarly:
Exercise 3.16 Let f be analytic in the domain Ω which consists of the plane,
from which are removed the open unit disk and the disk of center 5 and radius
3, and not vanishing there. Show that there exist numbers n1 and n2 in Z and
a function g analytic in Ω such that
50
3.3 Cauchy’s formula and applications
Exercise 3.17 Let Ω be a simply connected open set and let γ be a closed simple
smooth curve in Ω. Let z0 6∈ γ and let f be analytic in Ω. Prove that
f 0 (z)
Z Z
f (z)
dz = 2
dz. (3.8)
γ z − z0 γ (z − z0 )
Exercise 3.18 Let f be analytic in |z| < 1 + for some > 0. Show that
Z 2π it
1 e +z
f (z) = iIm f (0) + Re f (eit )dt. (3.9)
2π 0 eit − z
Solution: We have
Z 2π it Z 2π it
1 e +z it 1 2e − eit + z
f (e )dt = f (eit )dt
4π 0 eit − z 4π 0 eit − z
Z 2π Z 2π
1 eit it 1
= f (e )dt − f (eit )dt
2π 0 eit − z 4π 0
f (0)
= f (z) − .
2
On the other hand, for a given z ∈ D, Cauchy’s formula applied to the function
1 + ζz
g(ζ) = f (ζ)
1 − ζz
leads to
2π
1 + eit z
Z
1
f (eit ) dt = f (0),
2π 0 1 − eit z
51
and so
2π 2π
eit + z e−it + z
Z Z
1 1
f (eit ) dt = f (eit )dt
4π 0 eit − z 04π e−it − z
Z 2π
1 + eit z
1
= f (eit ) dt
4π 0 1 − eit z
f (0)
= ,
2
and hence the result since
Z 2π it
1 e +z
Re f (eit )dt =
2π 0 eit − z
Z 2π it Z 2π it
1 1 e +z it 1 e +z it )dt .
= f (e )dt + f (e
2 2π 0 eit − z 2π 0 eit − z
2
Exercise 3.19 Let f be analytic in |z| < 1 + for some > 0 and assume that
Re f (eit ) ≥ 0 for t ∈ [0, 2π]. Show that:
1 − |z| 1 + |z|
Re f (0) ≤ Re f (z) ≤ Re f (0), ∀|z| < 1. (3.10)
1 + |z| 1 − |z|
52
Integrating these inequalities we obtain (3.10) since
Z 2π
1
Ref (0) = Re f (eit )dt = Re f (0).
2π 0
eit + e−it
Solution: We have cos t = and so:
2
Z 2π Z 2π it 2p
e + e−it
cos2p tdt = dt
0 0 2
Z 2π 2it
(e + 1)2p
= dt
0 22p e2pit
(z 2 + 1)2p
Z
1
= pi dz
4 |z|=1 z 2p+1
But
2p
f (2p) (0) = (2p)! ,
p
since
2p
X
2` 2p
f (z) = z ,
`
`=0
and so
53
2π
(z 2 + 1)2p
Z Z
2p 2πi 1
cos tdt = p dz
0 4 i 2πi |z|=1 z 2p+1
(2p)
2πi f (0)
=
4p i (2p)!
2π 2p
= p .
4 p
2
R 2π
We note that 0 cos2p+1 tdt = 0. We also note that one can apply the residue
theorem to compute
(z 2 + 1)2p
Z
dz.
|z|=1 z 2p+1
The next exercise has a long history and be found in numerous places (see for
instance [, ], [], []).
Exercise 3.23 Let f = u + iv be analytic in |z| < 1 + with > 0, and assume
that f (0) = 0. Show that
Z 2π Z 2π
4
u(cos t, sin t) dt ≤ 36 v(cos t, sin t)4 dt. (3.11)
0 0
In particular, Z 2π
Re f 4 (eit )dt = 0.
0
Since
Re f 4 = Re (u + iv)4 = u4 + v 4 − 6u2 v 2 ,
we obtain
Z 2π Z 2π
(u(cos t, sin t)4 + v(cos t, sin t)4 )dt = 6 u(cos t, sin t)2 v(cos t, sin t)2 dt,
0 0
and in particular
Z 2π Z 2π
u(cos t, sin t)4 dt ≤ 6 u(cos t, sin t)2 v(cos t, sin t)2 dt.
0 0
54
Taking square of both sides and applying the Cauchy–Schwarz inequality to the
expression on the right, we obtain:
Z 2π 2 Z 2π 2
4 2 2
u(cos t, sin t) dt ≤ 36 u(cos t, sin t) v(cos t, sin t) dt
0 0
Z 2π Z 2π
≤ 36 u(cos t, sin t)4 dt v(cos t, sin t)4 dt .
0 0
R 2π
If 0 u(cos t, sin t)4 dt > 0 we divide both sides of the last inequality by this
R 2π
expression and obtain (3.11). If 0 u(cos t, sin t)4 dt = 0, then (3.11) is trivially
satisfied. 2
55
Solution: Take r > 0. Cauchy’s formula gives us
1 Z f (z)dz
|fn | =
2πi |z|=r z n+1
Z 2π
f (reit )rieit
1
= dt
2πi 0 rn+1 ei(n+1)t
Z 2π
f (reit )rieit
1
≤ dt
2π 0 rn+1 ei(n+1)t
Z 2π
M rer
1
≤ dt
2π 0 rn+1
er
= M n.
r
Thus
er
|fn | ≤ M inf .
r>0 r n
r
The minimum of the function r 7→ ren is for r = n. Plugging r = n in the
formula one obtains the required estimate. 2
The next exercise is taken from [?]. It is quite difficult if given without hints
towards the solution.
Exercise 3.27 Let f be an analytic function in the open unit disk and assume
that Z 2π
sup |f 0 (reit )|dt < M for some M > 0. (3.14)
r∈[0,1) 0
R1
Show that 0
|f (x)|dx < ∞.
P∞
Hints: Write f (z) = 0 an z n and give an upper bound on |an | using (3.14)
R1
and then give an upper bound to 0 |f (x)|dx using
∞
X
|f (x)| ≤ |an |xn , x > 0.
0
56
3.4 Power series
Exercise 3.28 Given a sequence
Pn a` , ` = 1, 2, . . . of complex numbers such that
1
lim sup |a` | ` < 1, let Sn = 1 a` . Show that
∞ ∞
X 1 X
Sn z n = a` z ` .
0
1−z 0
Particular cases: a` = 1, a` = `, a` = 1` , a` = `2 .
a0 = 1
a1 = 1
an+2 = an+1 + an .
Show that
∞
X 1
an z n = .
0
1 − z − z2
57
In particular the imaginary part of the above expression is 0 for all ρ ∈ (0, R),
that is:
∞
X √
an ρn sin(nπ 2) = 0 ∀ρ ∈ (0, R).
n=1
n
nz
P∞
Exercise 3.31 Define F (z) = − n=1 (−1) . Show that F is analytic in
n
the open unit disk and that
Set G(z) = (1 + z) exp(−F (z)). The function G is analytic in the open unit disk
and we have:
G0 (z) = −(1 + z)F 0 (z) exp(−F (z)) + exp(−F (z)) = 0, |z| < 1.
Thus G(z) = G(0) = 1 and exp(F (z)) = 1 + z in the open unit disk. 2
The function F (z) is denoted by ln(1 + z). It is the analytic extension to the
open unit disk of the function ln(1+x) of calculus defined on (−1, 1) by the same
power series. For another proof of Exercise 3.31 using analytic continuation, see
Exercise 4.32.
58
Show that fα is analytic in the open unit disk (or is a polynomial when α is a
natural number) and that the following hold:
αfα (z)
fα0 (z) = ,
1+z
fα (z)fβ (z) = fα+β (z), |z| < 1.
α(α − 1) · · · (α − n + 1)
an =
n!
are non zero. Since
|an+1 | |α − n|
= −→ 1 as n −→ ∞
|an | n+1
α(α − 1) · · · (α − n + 1) α(α − 1) · · · (α − n)
+ =
(n − 1)! n!
α(α − 1) · · · (α − n + 1)
= (n + α − n)
n!
= αan ,
and hence the result. To prove the second equality, it suffices to differentiate
the function H = fα fβ − fα+β . We have:
59
P∞
Exercise 3.33 Let F (z) = n=0 an z n be a convergent power series with con-
vergence radius bigger or equal to 1. Show that
Z 1 F (z) − F (0)
f or z 6= 0
F 0 (tz)dt = z
0 F 0 (0) f or z = 0.
Solution: For |z| < 1 ≤ R and t ≤ 1 we have |zt| < 1 ≤ R and so by the
theorem on differentiability of a power series we have
∞
X
F 0 (tz) = nan tn−1 z n−1
n=1
n−1 n−1
for |z| < 1. Set fn (t) = nan−1 t z and Mn = n|an−1 ||z|n−1 . Since the
radius of convergence of the series of the derivative is also R we have
∞
X
Mn < ∞ for |z| < 1 ≤ R.
n=1
P∞
Exercise 3.34 Let F (z) = n=1 an z n be a power series with radius of conver-
gence R (R may be infinite). Let a > 0, and let f be a continuous function from
[0, a] into C. Where is the function
Z a
g(z) = F (zt)f (t)dt
0
analytic. Compute its derivative, both as power series and in closed form.
60
Ra
Same question for H(z) = 1
F (z/t)f (t)dt, where now a > 1 and f is continu-
ous on [1, a].
Solution: The idea is the same as in Exercise 2.14. We just give the out-
line of the proof. For a fixed z we set fn (t) = an z n tn f (t). Then with M =
maxt∈[0,a] |f (t)| we have
where Z a
An = an tn f (t)dt.
0
Thus G has a development in power series in |z| < R/a and is analytic there.
Its derivative is (recall the theorem on differentiability of power series)
∞
X
G0 (z) = nAn z n−1 .
n=1
zn
The arguments for the function H are similar. Set now fn (z) = an f (t). With
tn
M = maxt∈[1,a] |f (f )| and since 1/a ≤ 1/t ≤ 1 for 1 ≤ t ≤ a we have for z ∈ C:
where Z a
f (t)
An = an dt.
1 tn
Thus H has a development in power series in |z| < R and is analytic there. Its
derivative is (recall the theorem on differentiability of power series)
∞
X
H 0 (z) = nAn z n−1 .
n=1
61
Another application of Weierstrass’ theorem leads to
Z a
f (t)
H 0 (z) = − F 0 (z/t) 2 dt.
0 t
62
3.5 An application of the fundamental theorem of calculus
for analytic functions
Exercise 3.35 Let f be analytic in a convex open set Ω and assume that
Re f 0 (z) > 0 in Ω. Show that f is one-to-one in Ω.
and so Z 1
f 0 (z1 + t(z2 − z1 ))dt > 0. (3.16)
0
(This last strict inequality holds also if z1 = z2 ). It follows from (3.16) and from
(3.15):
Z 1
0
f (z2 ) − f (z1 ) = (z2 − z1 ) Re f (z1 + t(z2 − z1 ))dt +
0
Z 1
0
+ Im f (z1 + t(z2 − z1 ))dt
0
63
Indeed, if f 0 (z0 ) 6= 0 then at least one of the numbers Re f 0 (z0 ) and Im f 0 (z0 )
is not zero. Without loss of generality we may assume that Re f 0 (z0 ) > 0
(otherwise replace f by −f or ±if depending on the case). By continuity,
Re f 0 (z) > 0 in an open disk around z0 . We can then apply the precedent result
since a disk is in particular convex.
Exercise 3.37 Assume that the analytic function f is one to one in Ω. Show
that the formula (see e.g. [2, p. 180]).
sf 0 (s)
Z
1
g(z) = ds (3.17)
2πi γ f (s) − z
sf 0 (s)
Z
1 1
g(f (z0 )) = ds
2π γ f (s) − f (z 0 s − z0
)
s − z0
sf 0 (s)
=
f (s) − f (z0 )
s − z0 s=z0
= z0 .
2
64
4 Further applications of Cauchy’s formula
4.1 The maximum principle and Schwarz’ lemma
Exercise 4.1 Let p(z) = z n + an−1 z n−1 + · · · + a0 . Prove that either p(z) ≡ z n
or that there exists a point ζ on the unit circle such that |p(ζ)| > 1.
where we used the maximum modulus principle and the hypothesis that |p(eiθ | ≤
1 for all θ ∈ [0, 2π]. But q(0) = 1 and so q(z) ≡ 1 by the maximum modulus
principle. This forces an−1 = · · · = a0 = 0 and so p(z) = z n . So we proved
that the negation of (P1 ) implies (P2 ). On the other hand, it is trivial that (P1 )
implies that (P2 ) does not hold. So (P1 ) and (P2 ) are equivalent. 2
Show that the function r 7→ M (r, p) is increasing and that the function r 7→
M (r, p)
is decreasing.
rn
Solution: Let r1 < r2 . We have
{z ; |z| ≤ r1 } ⊂ {z ; |z| ≤ r2 } ,
def.
M (r, p) = sup |p(z)| = sup |p(z)|,
|z|=r |z|≤r
65
and hence M (r, p) is increasing.
1
To prove the second claim, we consider the function g(z) = z n p( ):
z
1 1
M (r, g) = sup |g(z)| = sup nn |p( )| = rn M ( , p), (4.18)
|z|=r |z|=r z r
and so
1 M (r, p)
M ( , g) = . (4.19)
r rn
The function g(z) is a polynomial and so the function z 7→ M (r, g) is increasing
1
and so the map z 7→ M ( , g) is decreasing. This concludes the proof, thanks to
r
(4.19). 2
Exercise 4.3 Is there a function analytic in the open unit disk and such that
|f (z)| = e|z| there?
Solution: Assume by contradiction that such a function f exists. It does not
vanish and f −1 (z) = e−|z| ≤ 1 for z ∈ D. But |f (0)| = e0 = 1. The maximum
modulus principle implies that f is a constant. But no constant function can
satisfy |f (z)| = e|z| for all |z| < 1. 2
66
P∞
Hence, the series 0 f (z n ) converges absolutely and uniformly in |z| ≤ r for
r < 1. 2
Exercise 4.6 Let f be a function analytic from the open unit disk D into itself.
Assume that for some point z0 ∈ D it holds that
find f .
By Schwarz’ lemma,
F (z) = zG(z)
where G is analytic and contractive in D. Hence
(f ◦ ϕz0 )(z) = (ϕz0 )(zG(z))
zG(z) + z0
= .
1 + zG(z)z0
But
1 − |z0 |2 0
(f ◦ ϕz0 )0 (z) = f (ϕz0 (z)),
(1 − zz0 )2
67
and 0
1 − |z0 |2
zG(z) + z0
= (zG0 (z) + G(z)).
1 + zG(z)z0 (1 − zG(z)z0 )2
Hence,
1 − |z0 |2 0 1 − |z0 |2
f (ϕ(z)) = (zG0 (z) + G(z)).
(1 − zz0 )2 (1 − zG(z)z0 )2
Setting z = 0 and using the fact that f 0 (z0 ) = 1 we obtain
G(z0 ) = 1.
Exercise 4.7 Let f be analytic from D into D and assume that f (a) = f (b) = 0
for two different numbers a and b in D. Show that
z−a z−b
|f (z)| ≤
· .
1 − za 1 − zb
By Schwartz’ lemma,
F (z) = zG(z),
where G is analytic and contractive in the open unit disk. Replacing in this
equation z by ϕ−a (z) we obtain
and so
z−a
f (z) = g(z),
1 − za
where the function g(z) = G(ϕ−a (z)) is analytic and contractive in D. Further-
more, g(b) = 0 since f (b) = 0 and a 6 −b. We reiterate the same argument on g
and obtain
z−b
g(z) = h(z)
1 − zb
68
for some function h analytic and contractive in D. Thus
z−a z−a z−b
f (z) = g(z) = h(z),
1 − za 1 − za 1 − zb
and so
z − a z − b
|h(z)| ≤ z − a z − b .
|f (z) =
1 − za 1 − zb
1 − za 1 − zb
2
Exercise 4.8 Consider the function (2.5) and assume that m(t) ≥ 0 and that
R 2π
0
m(t)dt = 1. Show that
ϕ(z) − 1
ϕ(z) + 1 ≤ |z|, ∀z ∈ D.
Solution: We have
2π
1 − |z|2
Z
Re ϕ(z) = m(t)dt ≥ 0.
0 |eit − z|2
Using Exercise 3.19 one can prove the following surprising result; it appears
as an exercise (without any hint) in [9, Exercise 4, p. 40]. The proof is given
in [2, Exercise 5.36 p. 146]. We give the exercise and the solution with the
hypothesis that f is analytic in a neighborhood of the closed unit disk. This
can be dispensed with.
Exercise 4.9 Let h be analytic in |z| < 1 + for some > 0 and assume that h
does not vanish in |z| < 1 + and is bounded by 1 in modulus in D. Show that
69
Proof: Since h does not vanish in |z| < 1 + there is a function f analytic
there and such that h = e−f ; in particular
Since |h| ≤ 1 in the open unit disk we have that Re f (z) ≥ 0 there. Since f is
analytic in |z| < 1+ we can use Harnack’s inequalities (as proved in the present
set of exercises; recall that (3.10) is a special case of Harnack’s inequalities). Fix
R < 1 and ζ ∈ [0, 2π]. We have then
1 − |R| 1 + |R|
Re f (0) ≤ Re f (Reiζ ) ≤ Re f (0), ∀ζ ∈ [0, 2π].
1 + |R| 1 − |R|
≤ inf |h(z)|.
|z|=1/7
and thus, using the maximum modulus principle for h in |z| ≤ 1/5 and 1/h in
|z| ≤ 1/7 we obtain the result. 2
70
4.2 Liouville’s theorem and the fundamental theorem of
algebra
Since the real part of an analytic function is harmonic, the following exercise is
solved as Exercise 4.24.
Exercise 4.10 Show that an entire function f such that Re f (z) ≤ M for some
M ∈ M is constant.
Exercise 4.11 Show that there are no entire non constant function f such that
f (z + 1) = f (z) (4.20)
f (z + i) = f (z). (4.21)
Solution: By the periodicity conditions, it is enough to know the function f
on the closed square with corners (0, 0), (0, 1), (1, 0), (1, 1). On this square, the
function is bounded in modulus (by elementary property of continuous functions
on closed bounded sets). Thus, it is bounded on all the complex plane, and hence
constant by Liouville’s theorem. 2
Remark: there are biperiodic non constant meromorphic functions. These are
the elliptic functions. See Exercise 1.21.
Find P .
In the second case, we have that all the derivatives of P up to order N vanish
at z = 1/2. But, by the Taylor expansion at z = 1/2 and since P has degree N ,
N
X P (`) (1/2)
P (z) = (z − 1/2)` ,
`!
`=0
71
and so P ≡ 0 (and in particular no polynomial of degree N meets the require-
ment). 2
72
4.3 Harmonic functions
We recall that harmonic functions admits partial derivatives of all orders.
Exercise 4.13 Show that
sin x
u(x, y) =
cos x + cosh y
is harmonic in
Solution: A natural way is to first check that ∆u = 0 and then to solve the
Cauchy–Riemann equations. We will proceed in a different way. Recall that
exp(iz) + exp(−iz)
cosh(iz) = = cos z.
2
Thus
sin(z + z)/2
u(x, y) =
cos(z + z)/2 + cosh(z − z)/2i
sin(z + z)/2
=
cos(z + z)/2 + cos(z − z)/2
sin z/2 cos z/2 + sin z/2 cos z/2
=
cos z/2 cos z/2 − sin z/2 sin z/2 + cos z/2 cos z/2 + sin z/2 sin z/2
sin z/2 cos z/2 + sin z/2 cos z/2
=
2 cos z/2 cos z/2
sin z/2 sin z/2
= +
2 cos z/2 2 cos z/2
= Re tan(z/2).
Here we used that sin(z) = sin(z) and similarly for cos since the coefficients of
the McLaurin series of cos and sin are real, and thus
tan(z) = tan(z).
This is a well known formula; see []. A good exercise (see Exercise 4.15) is
to prove this formula for polynomials. The formula does not hold if u(x, y) is
not defined in a neighborhood of the origin, as one can see with the function
u(x, y) = ln(x2 + y 2 ).
73
sin x
Applying formula (4.22) to u(x, y) = we have u(x, y) = Re F (z)
cos x + cosh y
with
sin(z/2) sin(z/2)
F (z) = 2 =2 = tan(z/2).
cos(z/2) + cosh(z/2i) cos(z/2) + cos(z/2)
Exercise 4.15 Prove formula (4.22) for a polynomial
Solution: It suffices to check the formula for F (z) = z n . Since
z n = (x + iy)n
X
k n X n
= (−1) x2k y n−2k + i (−1) k
x2k+1 y n−2k−1 ,
2k 2k + 1
k,2k≤n k,2k+1≤n
that is
zn
n
X
k n
z = (−1) ,
2k 2n (−1)k
2k≤n
But from
0n = (1 − 1)n
X n X n
= 1n−2k (−1)2k + 1n−2k−1 (−1)2k−1
2k 2k + 1
2k≤n 2k+1≤n
we obtain
X n X n
=
2k 2k + 1
2k≤n 2k+1≤n
Hence,
X n X 2n
n
= = = 2n−1 .
2k 2k + 1 2
2k≤n 2k+1≤n
74
Exercise 4.16 Let u(x, y) and v(x, y) admit partial derivatives which are con-
tinuous in an open set Ω, and assume that u and v satisfy the Cauchy in Ω.
Assume moreover that u2 + v 2 6= 0 in Ω. Show that
∂u ∂v
u + v
∆ ∂x2 ∂x = 0 in Ω.
u + v2
One can build a number of exercises based on the same principle. Take F = u+iv
analytic in Ω and compute (for instance)
F 00 F F0
Re , Re , Im ...
F F0 F 00
In the computations, note that
∂2u ∂2u
F 00 (z) = − i
∂x2 ∂x∂y
75
you know ahead that the given function u is harmonic!).
y
∂2u
Z
∂v ∂u
(x, y) = 2
(x, t)dt − (x, 0)
∂x 0 ∂x ∂y
Z y 2
∂ u ∂u
= − 2
(x, t)dt − (x, 0)
0 ∂y ∂y
∂u ∂u ∂u
= (x, 0) − (x, y) − (x, 0)
∂y ∂y ∂y
∂u
= − (x, y).
∂y
76
To go from the first line to the second, we used that u is harmonic and hence
∂2u ∂2u
(x, y) = − (x, y).
∂x2 ∂y 2
2
More generally, one has the following formula for the harmonic conjugate:
Exercise 4.19 Let Ω be a simply connected set, and let u harmonic in Ω.
(a) Prove that Z
∂u ∂u
− dx + dy = 0
γ ∂y ∂x
for every simple closed contour γ.
(b) Fix a point (x0 , y0 ) ∈ Ω. Prove that the function
Z
∂u ∂u
v(x, y) = − dx + dy
γx,y ∂y ∂x
Exercise 4.20 Assume that u and v are conjugate harmonic in an open set Ω.
Let a and b be two real numbers. Show that the functions U (x, y) = au(x, y) −
bv(x, y) and V (x, y) = bu(x, y) + av(x, y) are also conjugate harmonic in Ω.
which ends the proof since by assumption the function f (z) = u(x, y) + iv(x, y)
is analytic in Ω. 2
Exercise 4.21 Assume that u and v are harmonic in an open set Ω and that
v is the harmonic conjugate of u in Ω. Define functions U and V by
2
−v(x,y)2
U (x, y) = eu(x,y) cos 2u(x, y)v(x, y)
2 2
−v(x,y)
V (x, y) = eu(x,y) sin 2u(x, y)v(x, y).
77
Solution: It is enough to check that the function F (z) = U (x, y) + iV (x, y) is
analytic in Ω. By hypothesis, the function f (z) = u(x, y) + v(x, y) is analytic
in Ω. One has
2
−v(x,y)2
U (x, y) + iV (x, y) = eu(x,y) cos 2u(x, y)v(x, y) +
u(x,y)2 −v(x,y)2
+ie sin 2u(x, y)v(x, y)
u(x,y)2 −v(x,y)2
= e (cos 2u(x, y)v(x, y) + i sin 2u(x, y)v(x, y))
u(x,y)2 −v(x,y)2 2iu(x,y)v(x,y)
= e e
u(x,y)2 −v(x,y)2 +2iu(x,y)v(x,y)
= e
2 2
= e(u(x,y)+iv(x,y)) = ef (z) ,
and so F is analytic in Ω.
u2x + u2y = 0.
Hence ux = uy = 0 and u is constant (on the connected components of its
domain of definition). 2
In the preceding exercise, one can replace u2 by ϕ(u) where ϕ(t) is any func-
tion of class C 2 (that is, with continuous second derivative), and whose second
derivative is nonzero, at the possible exception of a finite number of points (this
latter can also be relaxed). Indeed,
(ϕ(u))x = ux ϕ0 (u)
(ϕ(u))xx = (ux )2 ϕ00 (u) + uxx ϕ0 (u)
(ϕ(u))y = uy ϕ0 (u)
(ϕ(u))yy = (uy )2 ϕ00 (u) + uyy ϕ0 (u).
78
Hence,
and hence the result if ϕ00 vanishes at most at a finite number of points.
Exercise 4.23 Let u and v be harmonic in R2 and assume that v is the har-
monic conjugate of u. Assume that
u3 − 3uv 2 ≥ 0 in R2 .
Find u and v.
Solution: The function f (z) = u(x, y) + iv(x, y) is entire and so is the function
F (z) = f (z)3 . We have
Thus F 0 (z)F (z)2 ≡ 0. Since the zeros of F are isolated, we get that F 0 (z) = 0
first at those points where F (z) 6= 0 and then on all of C by continuity of F 0 .
Hence F (z) is a constant and so are u and v. 2
u(x, y) ≤ M ∀(x, y) ∈ R2 .
The function
H(z) = exp(u(x, y) + iv(x, y) − M )
79
is entire. It is bounded in the complex plane since
|H(z)| = eu(x,y)−M ≤ 1.
∂ ∂ ∂
Compute z, z and |z|.
∂z ∂z ∂z
80
4.4 Zeroes of analytic functions
Exercise 4.26 Let a0 , a1 , . . . be a sequence of complex numbers such that
∞
X
|an | < ∞. (4.25)
n=0
Exercise 4.27 Find all functions f analytic in the open unit disk and such that
f ( n1 ) = 0 for n = 2, 3, . . .
Solution: Since the zeros accumulate at the poit z = 0, the only function is
f (z) ≡ 0. 2
Exercise 4.28 Same question as in the previous exercise for each of the fol-
lowing conditions:
1
f = e−n (4.27)
n
1
|f | ≤ e−n . (4.28)
n
Is there a function analytic in the punctured disk such that (??) exist. Explain
the difference with the first question.
81
Solution: We consider the first case and show by contradiction that no such
function exists. If it exists, we not that it is not identically vanishing, since the
values at 1/n are different from 0. From (4.27) we obtain f (0) = limn→∞ f (1/n) =
0. Thus there is an integer p ≥ 0 and a function g analytic in the open unit disk
such that g(0) 6= 0 and
f (z) = z p g(z). (4.29)
Thus
f (1/n) = n−p g(1/n)
Since f (1/n) = e−n we obtain
g(1/n) = np e−n ,
and thus
g(0) = lim np e−n = 0.
n→∞
But g(0) 6= 0, and thus we obtain a contradiction and there is no function with
the required property.
We now turn to the second case. The function f (z) ≡ 0 answers the question.
We show that it is the only solution. The proof is quite similar as in the first
case. Assume that a solution which is not identically vanishing exists. Condition
(4.28) implies that
|f (0)| = lim |f (1/n)| = 0,
n→∞
|g(1/n)|
|f (1/n)| = ≤ e−n
np
and so
|g(1/n)| ≤ np e−n ,
once more leading to g(0) = 0, and hence to a contradiction.
In the last case we cannot assume that f is continuous at the origin. The
function f (z) = e−1/z meets conditions (4.27). 2
Exercise 4.29 Is there a function analytic in the open unit disk and such that
1
f (1/(2n)) = f (1/(2n − 1)) = , n = 2, 3, . . .
n
Solution: Assume such a function exists. The condition
1
f (1/(2n)) = , n = 2, 3, . . .
n
82
forces f (z) = 2z since both functions coincide at the points 1/2, 1/3, . . .. Simi-
larly, the second condition forces
2z
f (z) = ,
z+1
since both functions coincide at the points 1/2, 1/3, . . .. Hence we obtain a
contradiction, and no such function exist. 2
Exercise 4.30 Let f and g be analytic and not vanishing in the open unit disk
D. Assume that for all n ≥ 2,
f 0 ( n1 ) g 0 ( n1 )
1 = . (4.30)
f(n) g( n1 )
Show that there is a constant c such that f (z) = cg(z) for all z ∈ D.
Exercise 4.31 (a) Find all functions analytic in a neighborhood of the origin,
and for which there exists n0 ∈ N such that
f (1/n) + f 00 (1/n) = 0
for n ≥ n0 .
(b) Same question for
f (1/n) = f 00 (1/n).
f (z) + f 00 (z) = 0
83
in a neighborhood of the origin. Thus
for some complex numbers a and b (use analytic continuation from a real neigh-
borhood of the origin to a complex neighborhood of the origin). The formula
for f shows moreover that it extends to an entire function.
In case (b),
f (z) = a cosh z + b sinh z.
2
84
4.5 Analytic continuation
P∞ n zn
Exercise 4.32 Define F (z) = − n=1 (−1) . Show that F is analytic in
n
the open unit disk and that
Solution: The function F is analytic in the open unit disk (proved in Exercise
3.31 using a theorem on power series). The functions exp F (z) and 1 + z are
both analytic in the open unit disk. In the calculus class one proves (basically,
the same proof as the one in Exercise 3.31) that
85
5 Laurent expansions, singularities and applica-
tions
5.1 Laurent expansions
Exercise 5.1 Find the Laurent expansions for the following functions in cor-
responding domains:
1 1 1 2
a) ; 0 < |z−1| < 1; b) ; 0 < |z−i| < 2; c) sin2 ; 0 < |z|;
z(z − 1) (z 2 + 1)2 z z
1 − e−z sin z 1
d) ; 0 < |z|; e) ; |z − 2| 6= 0; f) ; 0 < |z| < 1.
z3 z−2 z 2 (z 2 + 1)
Exercise 5.2 Find four initial nontrivial terms of the Laurent expansions of
the functions
ez
2
z(z + 1)
in the domain 0 < |z| < 1.
86
The next exercise is taken from [5, Example 1 p. 77].
Exercise 5.7 Show that
∞
1 X 1
cosh z + = a0 + a` z ` + `
z z
`=1
where Z 2π
1
a` = cos(`t) cosh(2 cos t)dt.
2π 0
For a discussion of the above result, and much more information, see [11, §7]
87
5.2 Singularities
Exercise 5.9 Show that z = 0 is a removable singularity of the function
1 1
f (z) = − .
tan z sin z
Solution: We have
z cos z − 1
f (z) = .
sin z z
The point z = 0 is a removable singularity of z/ sin z since
sin z
lim = 1.
z→0 z
(To check these, note that the limit is equal to (sin0 (0)). It is also a removable
singularity of
cos z − 1 z z3
= − + · · · , z 6= 0.
z 2 4!
This last equation also shows that the origin is a first order zero of (cos z − 1)/z.
Thus z = 0 is a removable singularity of f , and moreover is a first order zero of
z. 2
z3 5
z− + z5! − · · ·
3!
f (z) =
z4
1 1 1 1
= − · + · z + ··· (5.1)
z3 3! z 5!
Hence, z = 0 is a pole of order 3. 2
sin z
Note: The function g(z) = z 3 f (z) = z has a removable singularity at 0 and
g(0) = 1 6= 0.
sin3 z
Exercise 5.11 Nature of z = 0 for the function f (z) = z
88
Solution: Since
z2 z 3
lim = lim = 1,
z→0 f (z) z→0 sin z
3 z3 z5 1 (3z)3 (3z)5
4 z− 3! + 5! − · · ·z 4 − 4 3z − 3! + 5! − ···
f (z) =
z
1
z 3 + z 5 14 5! (3 − 35 ) + · · ·
=
z
1 1
= z2 + z4 (3 − 35 ) + · · · (5.2)
4 5!
Hence z = 0 is a removable singularity for f , which has an analytic extension
to all of C. By abuse of notation we still denote by f this extension.
Furthermore (5.2) expresses that z = 0 is a zero of order 2. 2
Solution: From the arguments of Exercise 2.13 the function f has a deriva-
tive at every point where it does not vanish, and so f is analytic in Ω \
{w ∈ Ω ; f (w) = 0} . The points {w ∈ Ω ; f (w) = 0} are a priori isolated sin-
gularities of f . Since f is continuous at these points, these singularities are
removable, and f is analytic in all of Ω. 2
89
5.3 Rouché’s theorem
The following exercise is taken from []
Exercise 5.14 If f is analytic in |z| ≤ 1.2 and if |f (z)| < 1 on |z| = 1, the
equation f (z) = z n has exactly n solutions in |z| < 1.
The following exercise is also taken from Flanigan’s book (p. 303). It con-
sists in proving, for analytic functions, a very important theorem of topology,
Brouwer’s theorem, which is in fact true for continuous functions.
Exercise 5.16 Let g analytic in |z| < 1.2 and mapping the closed unit disk
into the open unit disk. Show that g has a fixed point, i.e there is z in the open
unit disk such that g(z) = z.
Hint: apply Rouché’s theorem with f (z) = z and F (z) = g(z) − z.
z 4 − 3z + 1 = 0
z 4 + z 3 − 4z + 1 = 0
Exercise 5.19 Prove that for real λ strictly greater than 1, there is a unique
solution to the equation
zeλ−z = 0
in the open unit disk.
90
5.4 Residue theorem
Exercise 5.20 Compute
2n
∞
X n
.
0
7n
we see that
(1 + z)n
Z
n 1
= dz. (5.3)
` 2πi |z|=1 z `+1
For z on the unit circle we have
(1 + z)2 4
7z < 7 < 1.
Thus:
2n
∞ ∞
n (1 + z)2n
Z
X 1 X 1
= dz
0
7n 2πi n=0 7 |z|=1 z n+1
n
∞ n
(1 + z)2
Z
1 X dz
= (by Weierstrass’ theorem)
2πi |z|=1 n=0 7z z
Z
1 1 dz
= 2
2πi |z|=1 (1 + z) z
1−
Z 7z
7 dz
=
2πi |z|=1 7z − (1 + z)2
Z
7 dz
= .
2πi |z|=1 5z − 1 − z 2
We apply the residue theorem to compute √ this last integral. The zeroes of the
equation 5z − 1 − z 2 = 0 are z± = (5 ± 21)/2. Thus
Z Z
7 dz 1 dz
=−
2πi |z|=1 5z − 1 − z 2 2πi |z|=1 (z − z− )(z − z+ )
1
= −Res , z−
(z − z− )(z − z+ )
1 1
− =√ ,
z− z+ 21
91
√
and so the sum is equal to 7/ 21:
2n
∞
X n 7
=√ .
0
7n 21
92
5.5 Computations of definite integrals using the residue
theorem
The first two exercises are taken from [],
Exercise 5.21 Let a, b ∈ R+ . Compute
Z
dx
2 2 2 2
R x + a )(x + b )
and we have:
0
1 2 2 −i
Res (f, ia) = =− =− = 3
(z + ia)2 z=ia (z + ia)3 z=ia (2ia)3 4a
Thus,
−i
Z
dx π
= 2πiRes (f, ia) = 2πi = 3.
R x2 + a2 )(x2 + b2 ) 4a3 2a
2
93
Solution: Let
z z
f (z) = = ,
(z 2 + z + 1)2 (z − z− ) (z − z+ )2
2
with √
−1 ± i 3
. z± =
2
√
The point z+ = (−1 + i 3)/2 is in the open upper half-plane, and we have that
Z
xdx
2 + x + 1)2
= 2πiRes (f, z+ )
R (x
z
= 2πi
(z − z− )2 z=z+
1 2z
= 2πi −
(z − z− )2 (z − z− )3 z=z+
1 2z+
= 2πi −
(z+ − z− )2 (z+ − z− )3
√ !
1 (−1 + i 3) 2π
= 2πi √ − √ = √ .
(i 3)2 (−i3 3) 3 3
2
4
But for z0 a zero of z + 1 we have
eiaz0
Res(f, z0 ) = ,
4z03
and so the integral is equal to
iπ 3iπ
!
e4 e 4 π √ √
2πi 3iπ + 9iπ = √ cos(a 2)/2 + sin(a 2)/2 .
4e 4 4e 4 2
2
94
Exercise 5.25 Show that
Z 2π
dt 2π
2
=
0 1 + 8 cos t 3
(b) Compute Z
dz
.
|z|=2 (z 1000 + 1)(z − 3)
1
R
Hint for (a): Compute |z|=R P (z)
and let R → ∞.
95
6 Conformal mappings
Exercise 6.1 Let D denotes the open unit disk and C+ denotes the open upper
half–plane. Show that the map
z − i(z 2 + 1)
ϕ(z) =
z + i(z 2 + 1)
96
But for z ∈ D+ we have
1
y>0 and > 1,
x2 + y2
and so
1
y −1 > 0,
x2 + y 2
that is Re ψ(z) > 0.
STEP 3: ψ is onto Cr .
Indeed, for w such that Re w > 0 consider solve the equation ψ(z) = w. We
have
z 2 + izw + 1 = 0,
and thus the two solutions are
√
w + ± w2 + 1
z=i .
2
They both are in the open upper half-plane and their product is 1. So one is of
modulus strictly less than 1, and thus in D+ . 2
97
7 Problems:
7.1 Hyperbolic distance
z−w
Let w ∈ ID and ϕw (z) = 1−z w̄ . It was shown in a previous exercise that the
functions cφw are biholomorphic mappings from the open unit disk onto itself
(with |c| = 1); it is a nontrivial fact that these are the only biholomorphic
mappings from D onto itself. The aim of the present problem is to prove:
still is in S.
(Hint: look first at the case a = 0.)
2. Use the preceding result to show the following: if f ∈ S and f (a) = y,
then,
y + ba (z)s(z)
f (z) = (7.4)
1 + y ∗ s(z)ba (z)
z−a
for some s ∈ S and where ba (z) = 1−za∗ .
z−w
3. Apply this result with a = x1 and y = y1 to a function ϕ(z) = α 1−zw ∗ for
98
6. Show that (7.2) holds for
99
References
[1] L. Ahlfors. Complex analysis. McGraw-Hill Book Co., third edition, 1978.
[2] R.B. Burckel. An introduction to classical complex analysis, Vol. 1.
Birkhäuser, 1979.
[3] G. Choquet. Topology, volume XIX of Pure and Applied Mathematics.
Academic Press, 1966.
[4] R.V. Churchill and J.W. Brown. Complex variables and applications. Fifth
edition. McGraw Hill, 1990.
[5] E.T. Copson. An introduction to the theory of functions of a complex vari-
able. Oxford Uinversity Press, 1972. First edition 1935.
[6] P.L. Duren. Theory of H p spaces. Academic press, New York, 1970.
[7] F.J Flanigan. Complex variables. Harmonic and analytic functions. Dover
books on advanced mathematics. Dover, 1983.
[8] E. Freitag and R. Busam. Complex analysis. Springer, 2005.
[9] K. Hoffman. Banach spaces of analytic functions. Dover Publications Inc.,
New York, 1988. Reprint of the 1962 original.
[10] K. Knopp. Problem book in the theory of functions. Volume II. Dover,
New–York, 1952.
[11] H. Koch. Einführung in die klassische Mathenatik I. Springer-BVerlag,
1986.
[12] M.A. Maingueneau. 30 semaines de khôlles en MATH, première partie.
Ellipses. Edition Marketing, 32 rue Bargue, 75015 Paris, 1994.
[13] J.H. Mathews and R.W. Howell. Complex analysis for mathematics and
engineering. Jones and Bartlett Publishers, Sudbury, Massachusetts, 1997.
[14] R. Narasimhan. Compact Riemann surfaces. Lectures in mathematics,
ETH Zürich. Birkäuser Verlag, Basel, 1992.
[15] R. Narasimhan and Y. Nievergelt. Complex analysis in one variable.
Birkhäuser Boston Inc., Boston, MA, second edition, 2001.
[16] G. Polya and G. Szegö. Problems and Theorems in Analysis I. Series.
Integral Calculus. Theory of functions. Springer, 1998. First published in
1924 as vol. 193 of the Grundlehren der mathematiken Wissenschaften.
[17] H.A. Priestley. Introduction to complex analysis. Oxford University Press,
second (revised) edition, 2003.
[18] W. Rudin. Real and complex analysis. Mc Graw Hill, 1982.
100
[19] L.I. Volkovyskii, G.L. Lunts, and I.G. Aramanovich. A collection of prob-
lems on complex analysis. Dover, 1991.
101