Ca Stdy
Ca Stdy
Ca Stdy
Preface
J.W. Brown and R.V. Churchill: Complex Variables and Applications (9th edition)
McGraw-Hill, New York, 2014,
and must be used in conjunction with it. In this study guide the chapter and section numbers correspond
with those in the textbook. However, if we are referring specifically to the study guide we make this clear
by saying “of this study guide”. Most of the guide consists of solutions to selected problems set in the
textbook, and since we make many references to various sections of the textbook, the guide will be of no
use to you unless you have the textbook in front of you.
Chapter 1: All the sections in this chapter. This chapter is mainly revision of the material on complex
numbers dealt with in the first year.
Chapter 2: Sections 13 to 28. However, the main thrust of sections 15, 16 and 18 is covered in more
detail in MAT2613 and MAT3711.
Chapter 3: Sections 30 to 39. Complex inverse trigonometric and hyperbolic functions will not be dealt
with in this course.
Chapter 4: All the sections in this chapter, in other words, sections 41 to 59. However, sections 41
and 42 are revision of material covered more extensively in MAT2615.
Chapter 5: Sections 60 to 73. Of this material, some sections are largely revision of material covered
in MAT2613.
Chapter 6: All the sections in this chapter, in other words, sections 74 to 84.
Chapter 7: Study only sections 85 to 90 and 92 to 94. Note that some important exercises at the end
of section 91 in fact refer to section 90 and are covered here in the study guide, although
the material of section 91 does not form part of the course.
As we have already explained, the guide consists mainly of solutions to some of the problems in the
textbook. Use these solutions as follows:
• Then obtain the numbers of the problems solved in the guide, BUT DO NOT READ THE SOLU-
TIONS. Try and solve these problems on your own.
iv
When reading the textbook and this guide you should have paper and pencil at hand. The author will have
left out certain steps, and you need to fill these in as you read – it is not possible to write a mathematics
book without leaving steps out and you should regard learning this “filling in” process as part of your
mathematical training.
PLEASE READ THE PREFACE TO THE TEXTBOOK as well as the “Road map to MAT3705” in this
study guide. Your textbook may come with an enclosed software package specifically for application to
complex analysis. The use of the software package is briefly explained in the preface to the textbook.
If you have access to a PC, using this package may help you to visualise certain important concepts.
However, bear in mind that software packages – no matter how slick – can never serve as a substitute for
hard work. As the well-known saying goes: “Genius is 10% inspiration and 90% perspiration”.
Dr L Lindeboom
v MAT3705/1
Chapter 1 of the textbook is mainly revisionary in nature and reviews the fundamentals of complex
numbers in addition to introducing the conceptual framework within which we will proceed with deeper
analysis in the subsequent chapters. The main function of chapter 1 is to set the scene for what follows,
and since you will already have covered much of this material in your first year, we will not pause too long
here.
In chapter 2 you are introduced to the idea of differentiability of complex functions before describing
complex analogues of some common elementary functions in chapter 3. Although much more can be said,
two overriding themes will emerge from our study of differentiability: first, the central role the Cauchy–
Riemann equations play in identifying the points at which a given function is differentiable, and second
the close link of differentiability to power series. As we shall see when investigating differentiability, the
Cauchy–Riemann equations are an extremely powerful and useful tool in eliminating a large number of
points where a given function cannot possibly be differentiable. However, they are not always as useful
in finding points where the function actually is differentiable.
Laplace’s equation plays a very important role in Applied Mathematics. One very valuable serendipity of
the complex theory of differentiability is that it provides us with a large number of solutions to Laplace’s
equation. Specifically if a complex function is differentiable on some domain, then both its real and
imaginary parts turn out to be solutions to Laplace’s equation on that same domain. With the theory of
differentiability in place, we proceed to investigate integration theory for complex functions. The theory
of complex integration is extremely elegant and once again closely linked to the concept of analyticity. In
chapter 4 we initiate our investigation of complex integration by first of all considering the problem of
existence of antiderivatives, and then (through the Cauchy-Goursat theorem) showing that analyticity of
a function is sufficient to guarantee the existence of antiderivatives. Building on this foundation we may
then derive the very powerful Cauchy integral formulas.
Besides yielding other unexpected bonuses such as Liouville’s theorem, the Fundamental theorem of Al-
gebra, and the Maximum modulus principle, the Cauchy integral formulas provide the tools we need to
proceed to the next phase of our programme in which we show that any function which is analytic on
vi
a region, except maybe at finitely many exceptional points, can effectively be broken up into powers of
(z − z0 ) on that region. We develop these ideas in a mathematically precise way in the form of the theory
of Taylor and Laurent series in chapter 5. For example, if we compare Taylor’s theorem to the corollary
on p. 214 of the textbook, it is clear that that a function is analytic at a point (i.e. differentiable in some
neighbourhood of a point) if and only if it can be written as a power series at that point. Ultimately we
will see that any function f which is differentiable in a neighbourhood of z0 , except perhaps at z0 itself,
can be written in the form
X∞
f (z) = an (z − z0 )n
n=−∞
around z0 . Therefore, terms of the form (z − z0 )n prove to be the basic building blocks of such functions.
Now let γ be a positively oriented circle centred at z0 . The easily established fact that
(
0 if n 6= −1
Z
(z − z0 )n dz =
γ 2πi if n = −1
(n ∈ Z) proves to be rather more significant than anticipated. Arguing formally we therefore have that
Z ∞
X Z
f (z) dz = an (z − z0 )n dz = 2πi a−1 .
γ n=−∞ γ
The integral of any such function may then be computed by effectively breaking the function up into its
component parts and integrating each component separately.
Having shown that any “nice” function may be broken up into powers of (z − z0 ) at each point z0 , we are
finally in a position to establish the promised residue theory in chapter 6. In short, the result basically
says that if a function f is of the form
∞
X
f (z) = an (z − z0 )n
n=−∞
inside and on a positively oriented closed curve γ (with z0 inside γ), then
Z
f (z) dz = 2πia−1 .
γ
For such functions the process of integration therefore consists of computing the coefficient a−1 (the so-
called residue of f at z0 ). As may be expected much of chapter 6 is then devoted to developing techniques
for computing the coefficient a−1 . This simple yet elegant principle of integrating by computing residues
enables us to effectively compute a wide range of diverse complex integrals.
Finally, in chapter 7 we turn to applications showing how the power of complex analysis may be used to
elegantly solve a number of highly non-trivial real integrals.
vii MAT3705/1
Contents
Page
Preface iii
Chapter 4. Integrals 63
Chapter 5. Series 89
CHAPTER 1
Complex Numbers
Introduction
The set of all complex numbers is essentially a two-dimensional extension of the field of real numbers. By
a complex number we mean a number comprising a real and an imaginary part. It can be written in the
form z = a + ib, where a and b are real numbers, and i is postulated to be the imaginary unit with the
property i2 = −1. The complex numbers clearly contain the real numbers – these may be identified with
the complex numbers for which the imaginary part is zero. Extending the field of numbers from the reals
to the complex numbers will enable us to solve equations such as z 2 + 1 = 0, which we weren’t able to
solve using only real numbers. Specifically with complex numbers a solution exists to every polynomial
equation of degree one or higher. However, although we gain a lot in passing to complex numbers, we
also lose something: in view of the fact that the field of complex numbers is two-dimensional, there is no
sensible way in which to order complex numbers. In other words, statements such as p ≥ q which seem so
natural when dealing with the reals, make no sense for complex numbers.
In this first chapter we revise the basic algebraic properties of complex numbers and look at two ways
of representing complex numbers (cartesian form and polar form). We also look at the process of com-
puting roots of complex numbers, and investigate the properties of the modulus and complex conjugation
functions.
Learning outcomes
After studying this chapter you should be able to:
• find and sketch the regions where complex inequalities are defined.
Sections to be studied
Study all the sections in chapter 1. The work in this chapter is mainly revisionary in that you will be
revising much of the material on complex numbers dealt with in the first year.
2
Below you will find solutions to some of the exercises in chapter 1. Be sure to attempt these yourself
before you work through the solutions.
Exercise 1, §2, p. 4
Verify that
√ √
(a) 2 − i − i 1 − 2i = −2i; (b) (2, −3) (−2, 1) = (−1, 8) ;
1 1
(c) (3, 1) (3, −1) , = (2, 1) ;
5 10
Solution:
(a)
√ √ √ h √ i
2 − i − i 1 − 2i = 2 − i + (−i) 1 − 2i
√ h √ √ i
= 2 − i + 0.1 − (−1) − 2 + i (−1) 1 + 0 − 2
√ √
= 2−i + − 2−i
√ √
= 2 − 2 + i (−1 − 1)
= −2i
(b)
(c)
1 1 1 1
(3, 1) (3, −1) , = [(3, 1) (3, −1)] ,
5 10 5 10
1 1
= (3.3 − 1 (−1) , 1.3 + 3 (−1)) ,
5 10
1 1
= (10, 0) ,
5 10
1 1 1 1
= 10. − 0. , 0. + 10.
5 10 5 10
= (2, 1)
3 MAT3705/1
Exercise 4, §2, p. 4
Verify that each of the two numbers z = 1 ± i satisfies the equation z 2 − 2z + 2 = 0.
Solution:
z = 1 + i yields
Similarly if z = 1 − i then
Exercise 5, §2, p. 5
Prove that multiplication is commutative, as stated in the second of equations (1), §2.
Solution:
Let z1 = x1 +iy and z2 = x2 +iy2 where x1 , x2 , y1 and y2 are real. Then using the fact that multiplication
of real numbers is commutative, we can show that
Exercise 7, §2, p. 5
Use the associative law for addition and the distributive law to show that
Solution:
z (z1 + z2 + z3 ) = z ((z1 + z2 ) + z3 ) by (2) of §2
= z (z1 + z2 ) + zz3 by (3) of §2
= (zz1 + zz2 ) + zz3 by (3) of §2
= zz1 + zz2 + zz3 by (2) of §2
Exercise 8, §2, p. 5
(a) Write (x, y) + (u, v) = (x, y) and point out how it follows that the complex number 0 = (0, 0) is
unique as an additive identity.
4
(b) Likewise, write (x, y) (u, v) = (x, y) and show that the number 1 = (1, 0) is a unique multiplicative
identity.
Solution:
(a) Let (u, v) be any complex number such that (x, y) + (u, v) = (x, y) for all (x, y) ∈ C. Then surely
(0, 0) = (0, 0) + (u, v) by assumption
= (u, v) by (4) of §2.
(b) Let (u, v) be given such that (x, y) (u, v) = (x, y) for all (x, y) ∈ C. Then
(1, 0) = (1; 0) (u, v) by assumption
= (u, v) (1, 0) by (1) of §2
= (u, v) by (4) of §2.
Exercise 1, §3, p. 7
Verify that
1 + 2i 2 − i 2 5 i
(a) + =− ; (b) = ;
3 − 4i 5i 5 (1 − i) (2 − i) (3 − i) 2
(c) (1 − i)4 = −4.
Solution:
(a)
1 + 2i 2 − i (1 + 2i) 5i + (3 − 4i) (2 − i)
+ =
3 − 4i 5i (3 − 4i) 5i
(−10 + 5i) + (2 − 11i)
=
20 + 15i
(−8 − 6i)
=
(20 + 15i)
(−2) (4 + 3i)
=
5 (4 + 3i)
2
= −
5
(b)
5 5
=
(1 − i) (2 − i) (3 − i) [(2 − 1) + i (−2 − 1)] (3 − i)
5
=
(1 − 3i) (3 − i)
5
=
((3 − 3) + i (−9 − 1))
5 10i
= − ×
10i 10i
50i
=
100
i
=
2
5 MAT3705/1
(c)
Exercise 4, §3, p. 7
Prove that if z1 z2 z3 = 0, then at least one of the three factors is zero.
Suggestion: Write (z1 z2 ) z3 = 0 and use a similar result (§3) involving two factors.
Solution:
Suppose z1 z2 z3 = 0. By the discussion at the start of §3, it is clear that if (z1 z2 ) z3 = 0 then one of z1 z2
and z3 is zero. If therefore z3 6= 0 we must then have z1 z2 = 0 in which case one of z1 and z2 must then
be zero. Therefore if z1 z2 z3 = 0, then at least one of z1 , z2 and z3 is zero.
Exercise 2, §5, p. 13
Verify inequalities (3), §4, involving Re z, Im z, and |z|.
Solution:
Let z = Re (z) + i Im (z). Now since Im (z)2 ≥ 0 we have that
Since by definition (
− Re (z) if Re (z) < 0
|Re (z)| =
Re (z) if Re (z) ≥ 0
we also have
Re (z) ≤ |Re (z)| .
(a) |z − 1 + i| = 1; (b) |z + i| ≤ 3;
6
Solution:
1 x
O
_
i
(b) 3 ≥ |z + i| = |z − (−i)|
This is the locus of all points within a distance of no more than 3 units from (−i) .
2i
x
_
i
3
_
4i
Exercise 6, §5, p. 14
Using the fact that |z1 − z2 | is the distance between two points z1 and z2 , give a geometric argument that
the equation |z − 1| = |z + i| represents the line through the origin whose slope is −1.
Solution:
|z − 1| = |z + i| is the set of all points equi–distant from 1 = (1, 0) and −i = (0, −1). The locus of
points is therefore a straight line. Note in particular that both (0, 0) and 21 , − 21 = 12 ((1, 0) + (0, −1)) are
(0−(− 1 ))
equidistant from (1, 0) and (0, −1) and hence on this line. The slope of the line is therefore 0− 12 = −1.
( 2)
Exercise 2, §6, p. 16
In each case sketch the set of points determined by the given condition:
Solution:
Re (z − i) = Re (x − i (y + 1))
= x
= Re z.
x
O 2
(b) Since 2z − i = 2z + i, we have that |2z − i| = 2z − i = |2z + i| . Thus we have that |2z + i| = 4 if
and only if |2z − i| = 4 if and only if |z − (i/2)| = 2. This is the circle of radius 2 centred at i/2.
_
5i
y 2
_i
2 x
O
_ 3_i
2
Exercise 3, §6, p. 16
Verify properties (3) and (4) of z in §6.
8
Solution:
Let z1 = x1 + iy1 and z2 = x2 + iy2 where x1 , x2 , y1 and y2 are real. Then z 1 = x1 − iy1 and z 2 = x2 − iy2 .
Therefore using the elementary properties of addition and multiplication we see that
and that
Exercise 5, §6, p. 16
Verify property (9) of moduli in §6.
Solution:
Note that 2
z1
z1 z1
z2 = z2 z2 by (7) of §6
z1 z1
= z2 z2 by (5) of §6
z1 z 1
=
z2 z 2 2
|z1 |
= |z2 | by (7) of§6
Exercise 9, §6, p. 16
By factoring z 4 − 4z 2 + 3 into two quadratic factors and then using inequality (2), §5 show that if z lies
Solution:
First note that
4
z − 4z 2 + 3 = z 2 − 3 z 2 − 1
= z 2 − 3 z 2 − 1
2 2
≥ |z| − 3 |z| − 1 .
Thus if |z| = 2 then z 4 − 4z 2 + 3 ≥ (4 − 3) (4 − 1) = 3, whence
1 1 1
z 4 − 4z 2 + 3 = |z 4 − 4z 2 + 3| ≤ 3 .
(a) z1 + z2 + · · · + zn = z 1 + z 2 + · · · + z n ;
(b) z1 z2 · · · zn = z 1 z 2 · · · z n .
Solution:
(a) By (2) in §6, z1 + z2 = z 1 + z 2 for all z1 , z2 ∈ C. Now suppose that for some fixed k ≥ 2 we have
that
z1 + z2 + z3 + · · · + zk = z 1 + z 2 + · · · + z k
Thus by induction
z1 + z2 + · · · + zn = z 1 + z 2 + · · · + z n for all n ≥ 2.
z 2 + z 2 = 2.
10
Solution:
Suppose z = x + iy where x, y ∈ R. Then by (6) of section 5
x2 − y 2 = 1 ⇔ Re (z)2 − Im (z)2 = 1
z+z 2 z−z 2
⇔ − =1
2 2i
(z + z)2 (z − z)2
⇔ + =1
4 4
z 2 + 2zz + z 2 + z 2 − 2zz + z 2 = 4
⇔
⇔ z2 + z2 = 2
Exercise 4, §9, p. 24
Solve the equation eiθ − 1 = 2 for θ (0 ≤ θ < 2π) and verify the solution geometrically.
Solution:
Rewrite eiθ − 1 = 2 as |(cos θ − 1) + i sin θ|2 = 4, which reduces to cos θ = −1. Since θ is to be in the
interval 0 ≤ θ < 2π, it follows that θ = π. This solution of the equation eiθ − 1 = 2 is geometrically
evident if we recall that eiθ lies on the circle |z| = 1 and that eiθ − 1 is the distance between the points
eiθ and 1. See the figure below.
y
iθ
e
O 1 x
Exercise 5, §9, p. 24
By writing the individual factors on the left in exponential form, performing the needed operations, and
finally changing back to rectangular coordinates, show that
√ √ √
(a) i 1 − 3i 3 + i = 2 1 + 3i ; (b) 5i/ (2 + i) = 1 + 2i;
√ √ −10 √
(c) ( 3 + i)6 = −64 ; (d) 1+ 3i = 2−11 −1 + 3i .
Solution:
√ √ √ √
3
For 3+i, r = 3 + 1 = 2. Select ϕ such that 3+i = 2 (cos ϕ + i sin ϕ), i.e. cos ϕ = 2 , sin ϕ = 12 .
√ π
We may let ϕ = π6 whence 3 + i = 2ei 6 . Thus
√ √ iπ π π
i 1 − 3i 3+i = e 2 .2e−i 3 .2ei 6
π
= 4ei( 3 )
π π
= 4 cos + i sin
3 3
√ !
1 3
= 4 + i
2 2
√
= 2 1 + 3i .
π √ √
(b) Clearly 5i = 5ei 2 . Now for 2 + i, r = 5. Thus if we select θ ∈ 0, π2 so that cos θ = √2
22 + 1 = 5
and sin θ = √15 , then we will have
√ √ √
2+i= 5 cos θ + i 5 sin θ = 5eiθ .
r sin θ
= 21 , θ = arctan 1
(Since here tan θ = r cos θ 2 will do the trick). Then
π
5i 5ei 2 1
= √ (where θ = arctan )
2+i 5eiθ 2
√ i(π/2−θ)
= 5e
√ π π
= 5 cos − θ + i sin −θ
√ 2 2
= 5 (sin θ + i cos θ) by trigonometric identities
√
1 2
= 5 √ +√ i by the way θ was chosen
5 5
= 1 + 2i.
√ √ √
(c) For 3 + i we have 3 + i = 2. Thus√ to write 3 + i in polar form we need to find θ with
√
3 + i = 2 (cos θ + i sin θ). Then cos θ = 23 , sin θ = 12 . Clearly θ = π6 i.e.
√ π
3 + i = 2ei 6 .
Then
√ π 6
( 3 + i)6 = (2)6 ei 6
= 2 6 ei π
= 64(cos π + i sin π)
= −64.
= 2 cos θ + i2 sin θ
12
√
3
√ π
(i.e. cos θ = 21 , sin θ = 2 ). We may therefore set θ = π
3 to get 1 + 3i = 2ei 3 . Then
√ −10 π −10
1+ 3i = 2−10 ei 3
π
= 2−10 ei(−10 3 )
π
= 2−10 ei(−3π− 3 )
π π
= 2−10 cos −3π − + i sin −3π −
3 3
√ !
1 3i
= 2−10 − +
2 2
√
= 2−11 −1 + 3i .
Exercise 7, §9, p. 24
Let z be a nonzero complex number and n a negative integer (n = −1, −2, . . .). Also, write z = reiθ and
m = −n = 1, 2, . . .. Using the expressions z m = rm eimθ and z −1 = (1/r) ei(−θ) , verify that (z m )−1 =
m m
z −1 and hence that the definition z n = z −1 in §8 could have been written alternatively as z n =
(z m )−1 .
Solution:
Here z = reiθ is any nonzero complex number and n a negative integer (n = −1, −2, . . .). Also, m = −n =
1, 2, . . .. By writing
−1 1
(z m )−1 = rm eimθ = m ei(−mθ)
r
and
1 i(−θ) m
m
m 1 1
z −1 = e = ei(−mθ) = m ei(−mθ) ,
r r r
m m
we see that (z m )−1 = z −1 . Thus the definition z n = z −1 can also be written as z n = (z m )−1 .
Exercise 8, §9, p. 24
Prove that two nonzero complex numbers z1 and z2 have the same moduli if and only if there are complex
numbers c1 and c2 such that z1 = c1 c2 and z2 = c1 c2 .
Solution:
First of all, given two nonzero complex numbers z1 and z2 , suppose that there are complex numbers c1
and c2 such that z1 = c1 c2 and z2 = c1 c2 . Since
Suppose, on the other hand, that we know only that |z1 | = |z2 |. We may write
we find that
θ1 + θ2 θ1 − θ2
c1 c2 = r1 exp i exp i = r1 exp (iθ1 ) = z1
2 2
and
θ1 + θ2 θ1 − θ2
c1 c2 = r1 exp i exp −i = r1 exp θ2 = z2 .
2 2
That is,
z1 = c1 c2 and z2 = c1 c2 .
(a) cos 3θ = cos3 θ − 3 cos θ sin2 θ; (b) sin 3θ = 3 cos2 θ sin θ − sin3 θ.
Solution:
We know from de Moivre’s formula that
or
cos3 θ + 3 cos2 θ (i sin θ) + 3 cos θ (i sin θ)2 + (i sin θ)3 = cos 3θ + i sin 3θ.
That is,
cos3 θ − 3 cos θ sin2 θ + i 3 cos2 θ sin θ − sin3 θ = cos 3θ + i sin 3θ.
By equating real parts and then imaginary parts here, we arrive at the desired trigonometric identities:
(a) cos 3θ = cos3 θ − 3 cos θ sin2 θ; (b) sin 3θ = 3 cos2 θ sin θ − sin3 θ.
(a) Use the binomial formula (§3) and de Moivre’s formula (§8) to write
n
!
X n
cos nθ + i sin nθ = cosn−k θ (i sin θ)k (n = 1, 2, . . .) .
k=0
k
14
and use the above sum to obtain the expression [compare Exercise 10(a)]
m
!
X n
cos nθ = (−1)k cosn−2k θ sin2k θ (n = 1, 2, . . .) .
k=0
2k
(b) Write x = cos θ and suppose that 0 ≤ θ ≤ π, in which case −1 ≤ x ≤ 1. Point out how it follows
from the final result in part (a) that each of the functions
Solution:
Since (i)4m = 1, (i)4m+1 = i, (i)4m+2 = −1 and (i)4m+3 = −i for each m ∈ Z, it follows that taking
the real part of the above sum is the same as summing over only even values of k. Hence if m
denotes the largest integer less than or equal to n/2, we get that
√ √
(b) Let 0 ≤ θ ≤ π and let x = cos θ. Then sin θ ≥ 0, and so sin θ = 1 − cos2 θ = 1 − x2 . Hence
is a polynomial of degree n. (To see that this really is of degree n and not something smaller, note
that for this polynomial the nth term is the term corresponding to p = r; that is (−1)p xn .)
Exercise 1, §11, p. 30
√
Find the square roots of (a) 2i; (b) 1 − 3i, and express them in rectangular coordinates.
Solution:
π
1 √ π
(a) Since 2i = 2 exp i 2 + 2kπ (k = 0, ±1, ±2, . . .), the desired roots are (2i) 2 = 2 exp i 4 + kπ
where k = 0, 1.
That is,
√ √ π √
i π4 π 1 i
c0 = 2e = 2 cos + i sin = 2 √ +√ =1+i
4 4 2 2
and
√ π
c1 = 2ei 4 eiπ = −c0 = − (1 + i) ,
√ √ 1 √
(b) Observe that 1− 3i = 2 exp i − π3 + 2kπ (k = 0, ±1, ±2, . . .). Hence 1 − 3i 2 = 2 exp i − π6 + kπ
where k = 0, 1.
The principal root is
√ ! √
√ −i π6
√ π π √ 3 i 3−i
c0 = 2e = 2 cos − i sin = 2 − = √ ,
6 6 2 2 2
√
3 3π 2kπ
ck = 8 exp(i ) exp(i ) where (k = 0, 1, 2)
12 3
Thus the principal cube root of z0 is
√
3 3π
c0 = 8 exp(i )
12
π
= 2 exp(i )
4
1 1
= 2( √ + i √ )
2 2
√
= 2(1 + i)
and since √
−1 + 3i
w3 =
2
√
√ −1 + 3i
c0 w3 = [ 2(1 + i)][ ]
2
1 √
= √ [(1 + i)(−1 + 3i)
2
√ √
−( 3 + 1) + ( 3 − 1)i
= √
2
17 MAT3705/1
Now
√ 2
(−1 + 3i)
w32 =
4√ √
−2 − 2 3i −1 − 3i
= =
4 2
so that
√
√ −1 − 3i
c0 w32 = [ 2(1 + i)][ ]
2
1 √
= √ [(1 + i)(−1 − 3i)
2
√ √
( 3 − 1) − ( 3 + 1)i
= √
2
Exercise 6, §9, p. 29
Find the four roots of the equation z 4 + 4 = 0 and use them to factor z 4 + 4 into quadratic factors with
real coefficients.
Solution
The four roots of the equation z 4 + 4 = 0 are the four fourth roots of the number −4. To find those roots,
we write −4 = 4 exp [i (π + 2kπ)] (k = 0 ± 1, ±2, . . .).
Then
√ √ π π
1 π kπ
(−4) = 4 2 exp i + = 2ei 4 eik 2 where k = 0, 1, 2, 3.
4 2
To be specific,
√ √ π √
i π4 π 1 1
c0 = 2e = 2 cos + i sin = 2 √ +√ = 1 + i,
4 4 2 2
π
c1 = c0 ei 2 = (1 + i) i = −1 + i,
c2 = c0 eiπ = (1 + i) (−1) = −1 − i,
3π
c3 = c0 ei 2 = (1 + i) (−i) = 1 − i.
z 4 + 4 = (z − c0 ) (z − c1 ) (z − c2 ) (z − c3 )
= [(z − c1 ) (z − c2 )] . [(z − c0 ) (z − c3 )]
= [(z + 1) − i] [(z + 1) + i] . [(z − 1) − i] [(z − 1) + i]
h i h i
= (z + 1)1 + 1 . (z − 1)2 + 1
= z 2 + 2z + 2 z 2 − 2z + 2 .
Exercise 7, §11, p. 31
Show that if c is any nth root of unity other than itself, then
1 + c + c2 + · · · + cn+1 = 0.
18
Solution:
Let c be any nth root of unity other than itself. With the aid of the identity
1−z n
(Exercise 9, §9), 1 + z + z 2 + · · · + z n−1 = 1−z , z 6= 0.
Then
1 − cn 1−1
1 + c + c2 + · · · + cn−1 = = = 0.
1−c 1−c
Exercise 1–3, §12, p. 34
Sketch the following sets and determine which are domains:
(a) |z − 2 + i| ≤ 1; (b) |2z + 3| > 4;
π
(e) 0 ≤ Argz ≤ 4 (z 6= 0) ; (f) |z − 4| ≥ |z| .
Solution:
(a) Write |z − 2 + i| ≤ 1 as |z − (2 − i)| ≤ 1 to see that this is the set of points inside and on the circle
centred at the point 2 − i with radius 1. It is closed and bounded. It is not a domain as it is not
open.
x
O y
_
2 i
(b) Write |2z + 3| > 4 as z − − 23 > 2 to see that the set in question consists of all points exterior to
the circle with center at − 32 and radius 2. It is open and connected and hence a domain. It is not
bounded.
y
_ _3 O x
2
19 MAT3705/1
(c) Write Im z > 1 as y > 1 to see that this is the half plane consisting of all points lying above the
horizontal line y = 1. It is open and connected and hence a domain. It is not bounded.
y =1
O x
(d) The set Im z = 1 is simply the horizontal line y = 1. It is closed, not open and hence not a domain.
It is not bounded.
y
y =1
O x
(e) The set 0 ≤ arg z ≤ π4 (z 6= 0) is indicated below. It is not open and hence not a domain. It is also
not closed since the boundary point 0 does not belong to the set (see (1) of section 6). It is not
bounded.
y
O x
(f) The set |z − 4| ≥ |z| can be written in the form (x − 4)2 + y 2 ≥ x2 + y 2 , which reduces to x ≤ 2.
The set is also geometrically evident since it consists of all points z such that the distance between
z and 4 is greater than or equal to the distance between z and the origin. This set is closed but not
open and so not a domain. It is not bounded.
_
|z 4|
| z|
O 2 4 x
20
Exercise 4, §12, p. 35
In each case, sketch the closure of the set:
1
≤ 12 ; (d) Re z 2 > 0.
(a) −π < arg z < π (z 6= 0) ; (b) |Re z| < |z| ; (c) Re z
Solution:
(a) The closure of the set −π < arg z < π (z 6= 0) is the entire plane.
O x
p
(b) We first write the set |Re z| < |z| as |x| < x2 + y 2 , or x2 < x2 + y 2 . But this last inequality is the
same as y 2 > 0, or |y| > 0. Hence the closure of the set |Re z| < |z| is the entire plane.
O x
1 z z x−iy 1 1 x
≤ 12 , or x2 − 2x + y 2 ≥ 0.
(c) Since z = zz = |z|2
= x2 +y 2
, the set Re z ≤ 2 can be written as x2 +y 2
2
Finally, by completing the square, we arrive at the inequality (x − 1) + y 2 ≥ 12 , which describes
the circle, together with its exterior, that is centered at z = 1 with radius 1. The closure of this set
is itself.
y
O z =1 x
(d) Since z 2 = (x + iy)2 = x2 − y 2 + i2xy, the set Re z 2 > 0 can be written as y 2 < x2 , or |y| < |x|.
21 MAT3705/1
The closure of this set consists of the lines y = ±x together with the shaded region shown below.
y
x
O
Exercise 5, §12, p. 35
Let S be the open set consisting of all points z such that |z| < 1 or |z − 2| < 1. State why S is not
connected.
Solution:
The set S consists of all points z such that |z| < 1 or |z − 2| < 1, as shown below.
z1 z2
O x
1 2
Since the point z = 1 is not in S, every polygonal line joining z1 and z2 must contain at least one point
that is not in S. Thus it is clear that S is not connected.
22
23 MAT3705/1
CHAPTER 2
Analytic Functions
Introduction
In this chapter we will assume that z = x + iy where x, y ∈ R. Any complex function f : C → C may
be decomposed into a real and an imaginary part f (z) = u (x, y) + iv (x, y) where u and v are real-valued
functions on C. In fact, since C is basically just a copy of R2 , we may think of u and v as real-valued
functions of two independent real variables. This simple little device of writing f as a combination of two
such 2-variable real functions allows one to apply the results and techniques of real analysis to complex
analysis. The essential idea behind this approach is to try and describe the nature and behaviour of the
complex function f in terms of the behaviour of the real functions u and v. A classic example of where this
was done with great success is provided by the Cauchy-Riemann equations (see sections 21 to 23). The
theory relating to these equations forms much of the heart of the theory of differentiability for complex
functions.
On a different note, this device of describing f in terms of u and v also allows one to think of a complex
function as a transformation of coordinates from the z = x + iy = (x, y) coordinate system to the
f (z) = u (x, y) + iv (x, y) = (u, v) coordinate system. So, yet another important way in which we can
gain some understanding of the behaviour of a specific complex function f is to study its behaviour as a
transformation of coordinates and to see how it transforms regions in the (x, y)-plane onto regions in the
(u, v)-plane.
Learning outcomes
After studying this chapter you should be able to:
• explain and apply the Cauchy-Riemann equations to determine where a function is differentiable;
Sections to be studied
Sections 13 to 28 have to be studied in chapter 2. However, the main thrust of sections 15, 16 and 18 is
covered in more detail in MAT2613 and MAT3711. For this reason we will use, but not greatly emphasise,
the material dealt with in these sections. However, since this material is essential background for what
follows, you should still take care to acquaint yourself with the ideas in these sections.
The proofs of the theorems need not be studied, although you should be familiar with both their formulation
and application.
25 MAT3705/1
Below you will find solutions to some of the exercises in §13 − 28 of chapter 2. Be sure to attempt these
yourself before you work through the solutions.
Exercise 1, §14, p. 43
For each of the functions below, describe the domain of definition that is understood:
1 1
(a) f (z) = ; (b) f (z) =Arg z ;
z2 +1
z 1
(c) f (z) = ; (d) f (z) = .
z+z 1 − |z|2
Solution:
1
(a) f (z) = is defined for all z except where z 2 = −1, i.e. except where z = ±i.
z2 +1
1
(b) z → is not defined where z = 0. In addition Arg(w) exists for every w 6= 0. Hence the composition
z
1
z → Arg is defined whenever z 6= 0.
z
z
(c) f (z) = is not defined where z + z = 0. But z + z = 2 Re (z). Hence f is here defined for all
z+z
z with Re (z) 6= 0.
1 2
(d) f (z) = 2 is defined whenever 1 − |z| 6= 0, i.e. whenever |z| =
6 1.
1 − |z|
Exercise 3, §14, p. 43
Suppose that f (z) = x2 − y 2 − 2y + i (2x − 2xy), where z = x + iy. Use the fact (§6) that
z+z z−z
x= and y =
2 2i
to express f (z) in terms of z, and simplify the result.
Solution:
z+z z−z
With z = x + iy we have x = and y = . Thus f (z) = x2 − y 2 − 2y + i (2x − 2xy) becomes
2 2i
2 2
1 1 1
f (z) = (z + z) − (z − z) − 2 (z − z)
2 2i 2i
1
+i [z + z] − [z + z] (z − z)
2i
1 2 1 2
z + 2zz + z 2 + z − 2zz + z 2
=
4 4
1 1 2
z − z2
− (z − z) + i (z + z) −
i 2
2
= z + i (z − z) + i (z + z)
= z 2 + 2iz.
26
Exercise 5, §14, p. 44
By referring to the discussion in §14 related to Fig.19, find a domain in the z plane whose image under the
transformation w = z 2 is the square domain in the w plane bounded by the lines u = 1, u = 2, v = 1 and
v = 2.
Solution:
Referring to Fig. 19, the mapping w = z 2 will map the hyperbolae 1 = 2xy and 2 = 2xy onto the lines
v = 1 and v = 2. In Example 4(d), §12, we saw that u = Re (w) = x2 − y 2 . Thus w = z 2 will also map
hyperbolae of the form c = x2 − y 2 onto lines of the form c = u. In particular it will map the region
bounded by 1 = x2 − y 2 , 2 = x2 − y 2 and 1 = 2xy, 1 = xy onto the square bounded by u = 1, u = 2,
v = 1, v = 2. However the region bounded by 1 = x2 − y 2 , 2 = x2 − y 2 and 1 = 2xy, 1 = xy consists
of two separate domains each of which maps onto the square bounded by u = 1, u = 2 and v = 1, v = 2
(see the sketch). To see that each maps onto the whole square observe that we can still cover all possible
values of v = 2xy and u = x2 − y 2 in the square if we insist that either x, y > 0 or x, y < 0.
1 = xy x_ y = 0
1 = 2 xy
2 _ 2
x y =2
2 _ 2
x y =1
x+y =0
27 MAT3705/1
Exercise 8, §14, p. 44
π
Sketch the region onto which the sector r ≤ 1, 0 ≤ θ ≤ 4 is mapped by the transformation (a) w = z 2 ; (b)
w = z 3 ; (c) w = z 4 .
Solution:
Let z = reiθ (0 ≤ r, 0 ≤ θ). Since z n = rn einθ and since 0 ≤ r ≤ 1 and 0 ≤ θ ≤ π4 if and only if
0 ≤ rn ≤ 1 and 0≤ nθ ≤ n π4 , it is clear that z → z n maps the sector 0 ≤ r ≤ 1, 0 ≤ θ ≤ π4 into the sector
0 ≤ ρ ≤ 1, 0 ≤ ϕ ≤ n π4 . To see that the map is onto note that given any w0 = ρ0 eiϕ0 with 0 ≤ ρ0 ≤ 1,
1
0 ≤ ϕ0 ≤ n π4 , we may set r0 = (ρ0 ) n and θ0 = ϕn0 . Then surely 0 ≤ r0 ≤ 1, 0 ≤ θ0 ≤ π4 and z0n = w0 ,
where z0 = r0 eiθ0 .
π
(a) z → z 2 maps the sector 0 ≤ r ≤ 1, 0 ≤ θ ≤ 4 onto the quarter–disc 0 ≤ ρ ≤ 1, 0 ≤ ϕ ≤ π2 ;
π 3π
(b) z → z 3 maps the sector 0 ≤ r ≤ 1, 0 ≤ θ ≤ 4 onto the sector 0 ≤ ρ ≤ 1, 0 ≤ ϕ ≤ 4 ;
(c) z → z 4 maps the sector 0 ≤ r ≤ 1, 0 ≤ θ ≤ π4 onto the region 0 ≤ ρ ≤ 1, 0 ≤ ϕ ≤ π, i.e. onto the
upper half of the disc |w| ≤ 1 where w = ρeiϕ .
Exercise 5, §18, p. 54
z 2
Show that the limit of the function f (z) = as z tends to 0 does not exist. Do this by letting nonzero
z
points z = (x, 0) and z = (x, x) approach the origin. [Note that it is not sufficient to simply consider
points z = (x, 0) and z = (0, y), as it was in Example 2, §15.]
Solution:
Let z = x + iy. Then
2
x2 − y 2 + i2xy
z 2
x + iy
= = .
z x − iy x2 − y 2 − i2xy
Now if z = (x, 0) = x + i0, then
x2
z 2
= 1. =
z x2
If on the other hand z = (x, x) = x + ix (y = x), then
z 2 i2x2
=− = −1.
z i2x2
z 2 z 2
Thus along the line z = (x, 0), tends to 1 as z → 0, whereas along the line z = (x, x), tends
z z
z 2
to −1 as z → 0. Clearly then lim can not exist.
z→0 z
Solution:
(a)
1 2
4z 2 4 z
lim = lim 2 (by (2) of the theorem)
z→∞ (z − 1)2 1
z→0 −1
z
4
= lim
z→0 (1 − z)2
= 4
(b) Now lim (z − 1)3 = 0 and hence (by (1) of the theorem) lim 1
3 = ∞.
z→1 z→1 (z−1)
1 2
z2 + 1 1 + z2
1 z +1
(c) Let f (z) = . Then f = 1 = . Since
z−1 z z −1
z − z2
1 1
lim 1 = lim
z 2 ) / (z − z2)
z→0 f z
x→0 (1 +
z − z2
= lim
z→0 1 + z 2
= 0
z2 + 1
lim = ∞.
z→∞ z − 1
Solution:
1
(a) Since here c = 0, we have T (z) = (az + b) . In this case
d
1 1
lim 1 = lim
z→0 1 a
z→0 T z d+b z
dz
= lim .
z→0 a + bdz
Moreover since
1 cz + d
lim = lim
z→− dc T (z) az + b
z→− dc
c − dc + d
d
= (a − + b 6= 0 since ad − bc 6= 0)
a − dc + b
c
= 0,
it follows that
lim T (z) = ∞.
z→− dc
Solution:
A set S is unbounded
⇔ for every R > 0 we can find w ∈ S so that w lies outside the circle |z| = R
Exercise 2, §20, p. 61
Use results in §20 to find f 0 (z) when
5
(a) f (z) = 3z 2 − 2z + 4; (b) f (z) = 2z 2 + i ;
4
1 + z2
z−1 1
(c) f (z) = z 6= − ; (d) f (z) = (z 6= 0) .
2z + 1 2 z2
Solutions:
5 4 4
(b) If f (z) = (z) = 2z 2 + i then f 0 (z) = 5 2z 2 + i (4z) = 20z 2z 2 + i .
z−1 1
(c) If f (z) = z 6= − then by the quotient rule
2z + 1 2
(2z + 1) .1 − 2. (z − 1)
f 0 (z) =
(2z + 1)2
3 1
= z 6= − .
(2z + 1)2 2
4
1 + z2
(d) If f (z) = then by the quotient and chain rules
z2
h 3 i 4
z 2 . 4 1 + z 2 .2z − [2z] . 1 + z 2
f 0 (z) =
z4
3
z2 4z 2 − 1 + z2
2z 1 +
=
z4
2 3 3z 2 − 1
2 1+z
= (z 6= 0) .
z3
Exercise 8, §20, p. 61
Use the method in Example 2, §19, to show that f 0 (z) does not exist at any point z when
(a) f (z) = z; (extra) (b) f (z) = Re z; (c) f (z) = Im z.
Solution:
∆w ∆z ∆x − i∆y
= = .
∆z ∆z ∆x + i∆y
∆w ∆x
Now as ∆z = ∆x + i∆y approaches 0 along the line (∆x, 0), = = 1 approaches 1. On the
∆z ∆x
∆w i∆y
other hand along the line (0, ∆y), =− approaches −1. Thus for any z
∆z i∆y
f (z + ∆z) − f (z) ∆w
f 0 (z) = lim = lim
∆z→0 ∆z ∆z→0 ∆z
fails to exist.
∆w ∆ (Re (z)) ∆x
= = .
∆z ∆z ∆x + i∆y
Along the line (∆x, 0) this tends to 1 as ∆z → 0 whereas along the line (∆x, ∆x) (i.e. ∆x = ∆y)
this tends to
∆x 1 1 1
= = −i as ∆z → 0.
∆x + i∆x 1+i 2 2
31 MAT3705/1
fails to exist.
1
Along the line (0, ∆y) this tends to = −i as ∆z → 0, whereas along the line ∆x = ∆y this tends
i
∆x 1 1 1
to = = − i as ∆z → 0. As before this ensures that for any z
∆x + i∆x 1+i 2 2
f (z + ∆z) − f (z) ∆w
f 0 (z) = lim = lim
∆z→0 ∆z ∆z→0 ∆z
fails to exist.
Exercise 9, §20, p. 62
Let f denote the function whose values are
2
(z)
f (z) =
when z 6= 0,
z
0 when z = 0.
Show that if z = 0, then ∆w/∆z = 1 at each nonzero point on the real and imaginary axes in the ∆z, or
∆w
(∆x, ∆y), plane. Then show that = −1 at each nonzero point (∆x, ∆x) on the line ∆y = ∆x in that
∆z
plane. Conclude from these observations that f 0 (0) does not exist. (Note that, to obtain this result, it is
not sufficient to consider only horizontal and vertical approaches to the origin in the ∆z plane.)
Solution:
For z = 0 we have
f (0 + ∆z) − f (0) f (∆z) − 0
=
∆z ∆z !
2
1 ∆z
=
∆z ∆z
2
∆z
=
∆z
∆x − i∆y 2
=
∆x + i∆y
∆x2 − ∆y 2 − i2∆x∆y
= .
(∆x2 − ∆y 2 ) + i2∆x∆y
On the real and imaginary axes (i.e. either ∆y = 0 or ∆x = 0) we get
f (0 + ∆z) − f (0)
= 1.
∆z
32
Exercise 1, §24, p. 70
Use the theorem in §21 to show that f 0 (z) does not exist at any point if
(a) f (z) = z; (b) f (z) = z − z; (c) f (z) = 2x + ixy 2 ; (d) ex e−iy .
Solution:
Let z = x + iy.
(a) For f (z) = z = x − iy we have that u (x, y) = x and v (x, y) = −y. Thus for all z = (x, y), ux = 1
and vy = −1. Clearly the Cauchy–Riemann equations then fail to hold for every z. (We always
have ux 6= vy .) Therefore f is nowhere differentiable.
(b) For f (z) = z − z = 2i Im (z) = 2iy we have u (x, y) = 0 and v (x, y) = 2y. Therefore ux = 0 and
vy = 2. As before the Cauchy–Riemann equations never hold since we always have ux 6= vy . Thus
f 0 (z) never exists.
ux = 2, uy = 0, vx = y 2 and vy = 2xy.
The equation uy = −vx will therefore hold precisely when 0 = −y 2 , i.e. when y = 0. However if
y = 0 then
2 = ux 6= vy = 2xy = 0.
Therefore the equations uy = −vx and ux = vy can never hold simultaneously. Hence f 0 (z) does
not exist at any z.
(d) For f (z) = ex e−iy = ex cos y−iex sin y we have u = ex cos y, v = −ex sin y. Hence ux = ex cos y, uy =
−ex sin y, vx = −ex sin y and vy = −ex cos y. Recall that ex 6= 0 for all x. Thus ex cos y = ux = vy =
−ex cos y will hold precisely when cos y = 0, i.e. when y = (2k + 1) π2 (k ∈ Z). On the other hand
−ex sin y = uy = −vx = ex sin y will hold precisely when sin y = 0, i.e. when y = kπ (k ∈ Z). Thus
as before the equations ux = vy and uy = −vx can never hold simultaneously. Therefore f 0 (z) fails
to exist for each z.
33 MAT3705/1
Exercise 3, §24, p. 71
From results obtained in §21 and §23, determine where f 0 (z) exists and find its value when
1
(a) f (z) = ; (b) f (z) = x2 + iy 2 ; (c) f (z) = z Im z.
z
Solution:
1
(a) Given f (z) = (z 6= 0) we write
z
1 z x − iy
f (z) = = = 2 .
z zz x + y2
x y
Thus here u = and v = − 2 . Now whenever (x, y) 6= (0, 0) we have
x2 +y 2 x + y2
y 2 − x2
ux = = vy
(x2 + y 2 )2
and
2xy
uy = − = −vx .
(x2 + y2)
Moreover ux , vy , uy , vx as given above are defined and continuous on all of C − {0}. Therefore by
the theorem in §23 of the textbook f 0 (z) exists whenever z 6= 0. At z = 0 f (z) is not defined and
hence f 0 (0) does not exist. Using differentiation rules it is easy to see that
1
f 0 (z) = − for all z 6= 0.
z2
(c) Given f (z) = z Im (z) we have f (z) = xy + iy 2 and hence u = xy and v = y 2 . Therefore ux = y,
uy = x, vx = 0, vy = 2y. Now y = ux = vy = 2y will hold precisely when y = 0 with x = uy =
−vx = 0 holding precisely when x = 0. Thus if z 6= 0 f 0 (z) does not exist since then at least one of
ux = vy and uy = −vx must fail. On the other hand if z = 0, we do have ux = vy and uy = −vx .
Since in addition we also have that ux , uy , vx , vy are continuous in a neighbourhood of z = 0
((x, y) = (0, 0)), f 0 (0) exists with
Exercise 5, §24, p. 71
Solve equations (2), §24, for ux and uy to show that
sin θ cos θ
ux = ur cos θ − uθ , uy = ur sin θ + uθ .
r r
34
Then use these equations and similar ones for vx and vy to show that, in §24, equations (4) are satisfied
at a point z0 if equations (6) are satisfied there. Thus complete the verification that equations (6), §24,
are the Cauchy–Riemann equations in polar form.
Solution
From (2) in §24 we know that
ur = ux cos θ + uy sin θ (a)
and
uθ = −ux r sin θ + uy r cos θ. (b)
Therefore
sin θ
ur cos θ − uθ = (ux cos θ + uy sin θ) cos θ
r
sin θ
− (−ux r sin θ + uy r cos θ)
r
2 2
= ux cos θ + sin θ
= ux
and
cos θ
ur sin θ + uθ = (ux cos θ + uy sin θ) sin θ
r
cos θ
+ (−ux r sin θ + yy r cos θ)
r
= uy sin2 θ + cos2 θ
= uy .
Similarly
sin θ cos θ
vr cos θ − vθ = vx and vr sin θ + vθ = vy .
r r
Therefore if
1 1
ur = vθ and uθ = −vr
r r
then
sin θ
ux = ur cos θ − uθ
r
1
= vθ cos θ − (−vr ) sin θ
r
cos θ
= vθ + vr sin θ
r
= vy
and
1
uy = ur sin θ + uθ cos θ
r
1
= vθ sin θ − vr cos θ
r
sin θ
= − vr cos θ − vθ
r
= −vx .
35 MAT3705/1
Exercise 6, §24, p. 71
Suppose that a function f (z) = u + iv is differentiable at a nonzero point z0 = r0 exp (iθ0 ). Use the
expressions for ux and vx found in Exercise 5, together with the polar form (6), §24, of the Cauchy–
Riemann equations, to show that f 0 (z0 ) can be written
Solution:
See the theorem in §21 for the requirements for f 0 (z) = ux + ivx to exists. Now since
sin θ sin θ
ux = ur cos θ − uθ and vx = vr cos θ − vθ
r r
(see exercise 5), it follows that
0 sin θ sin θ
f (z0 ) = ur cos θ − uθ + i vr cos θ − vθ .
r r
Now since f 0 (z0 ) exists, the Cauchy–Riemann equations hold. By exercise (5) above this is equivalent to
1 1
ur = vθ and uθ = −vr .
r r
Therefore
0 1 1
f (zθ ) = ur cos θ − uθ sin θ + i vr cos θ − vθ sin θ
r r
= ur cos θ + vr sin θ + i (vr cos θ − ur sin θ)
= (cos θ − i sin θ) (ur + ivr )
= e−iθ (ur + ivr ) .
Exercise 7, §24, p. 71
(a) With the aid of the polar form (6), §24, of the Cauchy–Riemann equations, derive the alternative
form
−i
f 0 (z0 ) = (uθ + ivθ )
z0
of the expression for f 0 (z0 ) found in Exercise 6.
(b) Use the expression for f 0 (z0 ) found in part (a) to show that the derivative of the function f (z) =
1
1/z (z 6= 0) in Exercise 3(a), §24, is f 0 (z) = − 2 .
z
36
Solution:
and
1 1
ur = vθ and uθ = −vr
r r
where–ever f 0 (z) exists. Hence f 0 (z0 ) may be written as
0 1 sin θ 1 sin θ
f (z0 ) = vθ cos θ − uθ +i − uθ cos θ − vθ
r r r r
1
= (−i) (cos θ − i sin θ) (uθ + ivθ )
r
1
= −i e−iθ (uθ + ivθ )
r
i
= − iθ (uθ + ivθ )
re
i
= − (uθ + ivθ )
z0
where zθ = reiθ .
i
f 0 (z) = − (uθ + ivθ )
z
i 1 1
= − − sin θ − i cos θ
z r r
11
= − (cos θ − i sin θ)
zr
11
= − e−iθ
zr
1 1
= −
z reiθ
1
= − 2.
z
Exercise 1, §26, p. 76
Apply the theorem in §23 to verify that each of these functions is entire:
(a) f (z) = 3x + y + i (3y − x) ; (b) f (z) = sin x cosh y + i cos x sinh y;
Solution:
(a) For f (z) = 3x + y + i (3y − x) we have u = 3x + y and v = 3y − x. For all (x, y) we now have that
ux = 3 = vy and uy = 1 = −vx . Since in addition ux , uy , vx , vy are continuous on all of C, f is
entire (analytic on all of C.)
(b) Given f (z) = sin x cosh y + i cos x sinh y we have u = sin x cosh y and v = cos x sinh y. For all (x, y)
we now see that
ux = cos x cosh y = vy
and
uy = sin x sinh y = −vx .
(c) Here f = u + iv where u = e−y sin x and v = −e−y cos x. But then
for each (x, y) with again each of ux , uy , vx and vy clearly continuous on all of C. Therefore f is
entire.
(d) Here
z 2 − 2 e−x eiy
f (z) =
x2 − y 2 − 2 + i2xy e−x (cos y − i sin y)
=
x − y 2 − 2 e−x cos y + 2xye−x sin y
2
=
+i 2xye−x cos y − x2 − y 2 − 2 e−x sin y
and hence
u = x2 − y 2 − 2 e−x cos y + 2xye−x sin y
and
v = 2xye−x cos y − x2 − y 2 − 2 e−x sin y.
Now
ux = 2xe−x cos y − x2 − y 2 − 2 e−x cos y + 2ye−x sin y − 2xye−x sin y = vy
and
uy = −2ye−x cos y − x2 − y 2 − 2 e−x sin y + 2xe−x sin y + 2xye−x cos y = −vx .
In addition all these first partials are continuous on all of C whence f is analytic on all of C.
Exercise 4, §26, p. 76
In each case, determine the singular points of the function and state why the function is analytic everywhere
except at those points:
38
2z + 1 z3 + i z2 + 1
(a) f (z) = ; (b) f (z) = ; (c) f (z) = .
z (z 2 + 1) z 2 − 3z + 2 (z + 2) (z 2 + 2z + 2)
Solution:
Each of the functions is a quotient of two polynomials. Since polynomials are analytic on all of C, each
of these functions will therefore be analytic where–ever the denominator is non–zero. At the zeros of the
denominator the function is not defined and hence these points are singular points of the function.
2z + 1
has singular points where z z 2 + 1 = 0, i.e. where either z = 0 or z = ±i.
(a) f (z) = 2
z (z + 1)
z3 + i
(b) f (z) = has singular points where 0 = z 2 − 3z + 2 = (z − 2) (z − 1), i.e. where z = 1, 2.
z 2 − 3z + 2
z2 + 1
(c) f (z) = has singular points where either z + 2 = 0 or z 2 + 2z + 2 = 0, i.e. where
(z + 2) (z 2 + 2z + 2)√
−2 ± 4 − 8
either z = −2 or z = = −1 ± i.
2
Exercise 6, §26, p. 77
Use results in §24 to verify that the function
Solution:
For g (z) = ln r + iθ (r > 0, 2π > θ > 0) where z = reiθ we have that u = ln r and v = θ. Clearly
ur = 1r , vθ = 1 and uθ = vr = 0 on the region (r > 0, 2π > θ > 0). Thus
1 1 1
ur = = vθ and uθ = 0 = vθ
r r r
with in addition each of ur , uθ , vθ , vr continuous on the entire region. Therefore by the theorem in §24
g 0 (z) exists in this region with
or rather that |z02 + 1| > 0 and 0 < Arg(z02 + 1) < π. From what we noted earlier g(z02 + 1) will therefore
be differentiable at each such point. Thus g(z 2 + 1) is differentiable, and hence analytic, on all of x > 0,
d
y > 0. It now easily follows from the chain rule that dz g(z 2 + 1) = [1/(z 2 + 1)].2z = 2z/(z 2 + 1) on this
region.
Exercise 7, §26, p. 77
Let a function f (z) be analytic everywhere in a domain D. Prove that f (z) must be constant in D if
f (z) is real–valued for all z in D , then f (z) must be constant throughout D.
Solution:
1. Let f be analytic and real–valued on all of D. Since f is real–valued, v = 0, and since in addition
f is analytic on D,
ux = vy = 0 and uy = −vx = 0.
But then
f 0 (z) = ux + ivx = 0
vx = −uy = Vx and vy = ux = Vy .
Since now
∂
(v − V ) = vx − Vx = 0
∂x
∂
(v − V ) = vy − Vy = 0,
∂y
we may argue as in the proof of the theorem in §25 to show that v − V is then constant.
ux = vy , uy = −vx
for u + iv and
vx = uy , vy = −ux
40
Exercise 2, §27, p. 79
Let the function f (z) = u (x, y) + iv (x, y) be analytic in a domain D, and consider the families of level
curves u (x, y) = c1 and v (x, y) = c2 , where c1 and c2 are arbitrary real constants. Prove that these
families are orthogonal. More precisely, show that if z0 = (x0 , y0 ) is a point in D which is common to two
particular curves u (x, y) = c1 and v (x, y) = c2 and if f 0 (z0 ) 6= 0, then the lines tangent to those curves
at (x0 , y0 ) are perpendicular.
Suggestion: Note how it follows from the equation u (x, y) = c1 and v (x, y) = c2 that
∂u ∂u dy ∂v ∂v dy
+ = 0 and + = 0.
∂x ∂y dx ∂x ∂y dx
Solution:
dy
Observe that the equation u (x, y) = c1 implicitly defines a function y of x. To compute the derivative
dx
of this function we differentiate the equation u (x, y) = c1 implicitly to get
∂u ∂u dy
+ = 0.
∂x ∂y dx
(We could also have used the chain rule for functions of two real variables to see that
∂u ∂u dy ∂u dx ∂u dy d
+ = + = c1 = 0.)
∂x ∂y dx ∂x dx ∂y dx dx
dy
In any case at any (x, y) the slope of the line tangent to u (x, y) = c1 is given by m1 = where
dx
∂u ∂u
+ m1 = 0. Similarly we can show that at any (x, y) on the curve v (x, y) = c2 the slope of the line
∂x ∂y
∂v ∂v
tangent to this curve at (x, y) is given by m2 where + m2 = 0.
∂x ∂y
Now let z0 be a point on the intersection of u (x, y) = c1 and v (x, y) = c2 at which f 0 (z0 ) 6= 0. Then
at z0 . Since 0 6= f 0 (z0 ) = ux (x0 , y0 ) + ivx (x0 , y0 ) , ux and vx can not both be zero at z0 = x0 + iy0 .
Suppose ux (x0 , y0 ) 6= 0. Then
1 uy (x0 , y0 )
= −
m1 ux (x0 , y0 )
vx (x0 , y0 )
=
vy (x0 , y0 )
= −m2 .
This proves that the tangent lines to u (x, y) = c1 and v (x, y) = c2 at z0 = x0 + iy0 are orthogonal.
(To see this recall that two straight lines y = m1 x + c1 and y = m2 x + c2 are orthogonal precisely when
m1 m2 = −1.)
41 MAT3705/1
.
.
CHAPTER 3
Elementary Functions
Introduction
Our primary objective in this chapter is to introduce complex analogs of the most important elementary
functions, to describe the basic properties of these functions, and to gain some experience in working
with these functions. The complex cousins of the well-known elementary real functions are both more
interesting and more tricky. For example, whereas the real-valued version of sin is bounded, the same is
not true of the complex version. Where for each positive real number x there is a unique real number
w = log(x) for which x = ew , for any given non-zero complex number z there are infinitely many possible
values we can ascribe to log(z). The complex version of the logarithm is therefore what we may call a
multi-valued function. To produce a complex version of the logarithm which ascribes one single value to
each element of its domain, we need to restrict the possible values of log(z) in a natural way, by taking a
so-called branch cut. These and other aspects are explained in this chapter.
Learning outcomes
After studying this chapter you should be able to:
• solve problems when the exponential, logarithmic, trigonometric and hyperbolic functions defined
on a complex variable are involved;
Sections to be studied
Study only sections 30 to 39 in chapter 3. Complex inverse trigonometric and hyperbolic functions will
not be dealt with in this course.
42
Below you will find solutions to some of the exercises in chapter 3. Be sure to attempt these yourself
before you work through the solutions.
Exercise 1, §30, p. 89
Show that
r
2 + πi e
(a) exp (2 ± 3πi) = −e2 ; (b) exp = (1 + i) ;
4 2
(c) exp (z + πi) = − exp z.
Solution
(a)
(b)
2 + πi 1 π
exp = exp +i
4 2 4
1
π π
= e 2 cos + i sin
4 4
√
1 1
= e √ +i √
2 2
r
e
= (1 + i)
2
(c)
Exercise 3, §30, p. 89
Prove that the function exp z̄ is not analytic anywhere.
43 MAT3705/1
Solution
Let z = x + iy. Then z̄ = x − iy and
Thus here
Therefore
ux = ex cos y
vx = −ex sin y
uy = −ex sin y
vy = −ex cos y.
can only hold when cos y = − cos y, i.e. when cos y = 0. Similarly
can only hold when sin y = 0. Since cos y and sin y can never simultaneously be zero, it follows that the
Cauchy–Riemann equations never hold and hence that exp (z̄) is nowhere differentiable.
Exercise 5, §30, p. 90
Write |exp (2z + i)| and exp iz 2 in terms of x and y. Then show that
Solution
On setting z = x + iy we see that
and
= exp −2xy + i x2 − y 2
= e−2xy .
44
Hence
= e2x + e−2xy .
Exercise 8, §30, p. 90
Find all values of z such that
(a) ez = −2;
(b) ez = 1 + i;
Solution
ex eiy = ez = −2 = 2eiπ
we must have
ex = 2 and y = π + 2nπ (n ∈ Z) .
1 1 1
cos θ + i sin θ = (1 + i) = √ + i √ ,
r 2 2
√
i.e. cos θ = √1 and sin θ = √1 . Clearly θ = π
will suffice. Hence 1 + i = 2eiπ/4 . Then
2 2 4
√
ex eiy = ez = 1 + i = 2eiπ/4
√ 1 π
if and only if ex = 2 (i.e. x = 2 ln 2) and y = 4 + 2nπ (n ∈ Z) . Thus ez = 1 + i precisely when
z = 12 ln 2 + i π4 + 2nπ , n ∈ Z.
(c) With z = x + iy, exp (2z − 1) = exp ((2x − 1) + i2y) . Thus as before
if and only if
e2x−1 = 1 and 2y = 2nπ, n ∈ Z.
1
Consequently exp (2z − 1) = 1 if and only if x = 2 (or equivalently 2x − 1 = ln 1 = 0) and
y = nπ (n ∈ Z) , that is z = 12 + inπ (n ∈ Z) .
45 MAT3705/1
Exercise 9, §30, p. 90
Show that exp (iz) = exp (iz̄) if and only if z = nπ (n = 0, ±1, ±2, . . .) . (Compare Exercise 4, §29)
Solution
Observe that
Thus
Then
Solution
(a) ez = ex eiy = ex (cos y + i sin y) is real if and only if sin y = 0 if and only if y = Im z = nπ (n ∈ Z) .
(b) Similarly ez = ex (cos y + i sin y) is pure imaginary if and only if cos y = 0 if and only if y = Im z =
(2n + 1) π2 (n ∈ Z) .
(b) y tends to ∞.
Solution
exp (x + iy) → 0.
(b) For any fixed x and w = exp (x + iy) it follows from (3) of §30 that
Thus as y increases, exp (x + iy) rotates anticlockwise around the circle centred at the origin with
radius ex .
are harmonic in D and why V (x, y) is, in fact, a harmonic conjugate of U (x, y) .
Solution
Since f (z) = u (x, y) + iv (x, y) is analytic in a domain D and exp (z) = ez analytic on all of C, the
composition
and
V (x, y) = Im (exp ◦f (z)) = eu(x,y) sin (v (x, y))
Exercise 1, §33, p. 95
Show that
47 MAT3705/1
(a) Log(−ei) = 1 − π2 i;
1
(b) Log(1 − i) = 2 ln 2 − π4 i.
Solution
(b) To find Arg(1 − i) we need to select θ in the fourth quadrant with −π < θ ≤ π so that tan θ = −1.
Clearly Arg(1 − i) = − π4 whence
Log (1 − i) = ln |1 − i| + iArg (1 − i)
1 π
= ln 2 2 − i
4
1 π
= ln 2 − i .
2 4
Exercise 4, §33, p. 95
Show that log i2 6= 2 log i when the branch
3π 11π
log z = ln r + iθ (r > 0, <θ< )
4 4
is used.
Solution
We have
log i2 = log(−1) = ln 1 + iπ = iπ
and
π 5
i = 1ei 2 +2π = ei 2 π
whence
5π
2 log i = 2(ln 1 + i( )) = i5π
2
so within the given branch
log i2 =
6 2 log i
Solution
So by (*)
1 π
log i 2 = ln 1 + i + kπ
π 4
= i + kπ (k = 0, 1 ) .
4
1 1
= i( π4 ) and log i 2 5π
and we have log i 2 =i 4
k=0 k=1
Thus the square roots of i are
π 5π
ei 4 and ei 4
Exercise 8, §33, p. 96
Find all roots of the equation log z = (π/2) i.
Solution
π
log (z) = (π/2) i ⇔ ln |z| + iθ = i (where z = |z| eiθ )
2
π
⇔ ln |z| = 0, ∈ arg (z)
2
π
⇔ |z| = 1, Arg (z) =
2
iπ/2
⇔ z=e =i
(a) the function Log(z − i) is analytic everywhere except on the half line y = 1 (x ≤ 0) ;
Solution
(a) By (5) of §33 Logz is analytic everywhere except where either z = 0 or Arg(z) = π. In terms of
Cartesian coordinates this means that Logz is analytic everywhere except on the non–positive real
axis. By contrast z → z − i is a 1 − 1 analytic map from C onto C. Hence the composition Log(z − i)
is analytic on all of C except where z − i = x + i (y − i) lies on the non–positive real–axis, i.e. except
where x ≤ 0 and y = 1.
49 MAT3705/1
(b) By a similar argument to that in (a) above we can show that Log(z + 4) is analytic on all of C except
where z + 4 = (x + 4) + iy is on the non–positive real axis, i.e. except where x ≤ −4 and y = 0.
Therefore
Log (z + 4)
z2 + i
will fail to be analytic on this portion and also where z 2 + i = 0. Now since z 2 = −i = e−iπ/2 if and
only if z 2 + i = 0, the roots of this equation will be
( 1
π √ − i √1 (k = 0)
z = ei(− 4 +kπ) = 2
1
2
1
− √2 + i √2 (k = 1) .
Thus Log(z + 4) / z 2 + i is analytic everywhere except where z = ± √12 (1 − i) and on the portion
Solution
Since
log (z − 1) = ln |z − 1| + i arg (z − 1)
we surely have
Re [log (z − 1)] = ln |z − 1|
1
= ln (x − 1)2 + y 2
2
1
= ln (x − 1)2 + y 2 .
2
Now for any z with z − 1 6= 0 (i.e. z 6= 1) we can find a branch of log (z − 1) such that the branch
cut is analytic at z (see (4) of §33 of the textbook). By Theorem 1 of §27, the real part, that is
1 2 2 , must be harmonic at each such z.
2 ln (x − 1) + y
Exercise 1, §34, p. 99
Show that if Re z1 > 0 and Re z2 > 0, then
Solution
Since Re (z1 ) > 0 and Re (z2 ) > 0, both z1 and z2 are in either the first or fourth quadrant. Consequently
π π
(∗) − < Arg (z1 ) < ,
2 2
π π
− < Arg (z2 ) < .
2 2
50
whence
Solution
For general nonzero complex numbers z1 , z2 we have −π < Arg(z1 ) ≤ π and −π < Arg(z2 ) ≤ π whence
Now as before
z1 z2 = |z1 | |z2 | exp (i (Arg (z1 ) + Arg (z2 )))
and hence
Arg (z1 z2 ) = Arg (z1 ) + Arg (z2 ) + 2N π
where N ∈ Z is chosen so that
Solution
√
(a) In polar form 1 + i = 2eiπ/4 . Hence
1
π
log (1 + i) = ln 2 2 + i + 2nπ
4
1 π
= ln 2 + i + 2nπ (n ∈ Z) .
2 4
Therefore
(a) (−i)i ;
√ 3πi
(b) 2e −1 − 3i
;
(c) (1 − i)4i .
Solution
√ √ √
(b) In polar form 2e −1 − 3i is of the form 2e −1 − 3i = 2e −1 − 3i eiθ = eeiθ where θ is an
√
angle in the third quadrant with tan θ = xy = 3. For our purposes we also need −π < θ ≤ π. Then
θ = − 2π
3 and hence
√ i
he
2π
Log −1 − 3i = ln e + i −
2 3
2π
= 1−i .
3
√ 3πi
The principal value of 2e −1 − 3i
is therefore given by
he √ i3πi he √ i
−1 − 3i = exp 3πiLog −1 − 3i
2 2
2π
= exp 3πi 1 − i
3
2
= exp 2π + i3π
= exp 2π 2 (cos 3π + i sin 3π)
= − exp 2π 2
√ π
(c) In polar form (1 − i) = 2e−i 4 . Thus
1
π
Log (1 − i) = ln 2 2 + i −
2
1 π
= ln 2 − i .
2 4
Solution:
For any z 6= 0 the principal value of |z|a is given by |z|a = exp (aLog |z|) = exp (a ln |z|)) . Moreover
in this case.
Solution:
π
In polar form i = 1.ei 2 and hence
π π
log (i) = ln 1 + i + 2nπ = i + 2nπ (n ∈ Z) .
2 2
Therefore
Now the only way these values can be the same for all n ∈ Z is if the 2nπb’s are the same for all n ∈ Z,
i.e. if Im (c) = b = 0. Thus |ic | has only one possible value whenever c is real (that is Im (c) = 0).
Solution:
We may mimic the proof of (5) at the end of §35 to get
Solution
Given that
it follows that
= sin2 x + sinh2 y
and
= cos2 x + sinh2 y.
Solution
It follows directly from what we proved in exercise (7) above that
q p
|sin z| = sin2 x + sinh2 y ≥ sin2 x = |sin x|
Solution
Again using what we showed in exercise (7) above it follows that
q
|sinh y| = sinh2 y
q
≤ sin2 x + sinh2 y = |sin z|
q
≤ cos2 x + sin2 x + sinh2 y
q
= 1 + sinh2 y
q
= cosh2 y
= cosh y.
(The last equality follows from the fact that cosh y ≥ 0 for all y.) By a similar argument to the above we
can also show that
|sinh y| ≤ |cos z| ≤ cosh y.
Exercise 11, §38, p. 108
Show that neither sin z̄ nor cos z̄ is an analytic function of z anywhere.
Solution
If z = x + iy, then z̄ = x − iy. From equation (13) of §37 it follows that
Thus for sin z̄, u = Re (sin z̄) = sin x cosh y and v = Im (sin z̄) = − cos x sinh y. Therefore
ux = cos x cosh y
vx = sin x sinh y
uy = sin x sinh y
vy = − cos x cosh y.
Clearly cos x cosh y = ux = vy = − cos x cosh y can only hold if cos x cosh y = 0. Since cosh y 6= 0 this
means that cos x = 0, i.e. that x = (2n + 1) π2 (n ∈ Z) . Similarly sin x sinh y = uy = −vx = − sin x sinh y
can only hold if sin x sinh y = 0. Since sin x 6= 0 when x = (2n + 1) π2 (n ∈ Z) , we must then have that
sinh y = 0, i.e. that y = 0. Thus by the Cauchy–Riemann equations the only points where sin z̄ may
be differentiable are the isolated points zn = (2n + 1) π2 (n ∈ Z) . For any z ∈ C the function sin z̄ can
therefore never be differentiable in an entire neighbourhood of z, that is sin z̄ is never analytic.
cos z̄ = u
e + ie
v
56
where
u
e = cos x cosh y,
ve = sin x sinh y.
The equation − sin x cosh y = u ex = vex = sin x cosh y only holds if sin x = 0, i.e. if x = nπ (n ∈ Z) . For
these values of x, cos x sinh y = uey = −evx = − cos x sinh y can only hold if y = 0. Thus the only points
where cos z̄ may be differentiable are the isolated points z = nπ (n ∈ Z) . As before it follows from this
that cos z̄ is nowhere analytic.
Solutions
From the equations we verified in exercise (12), it follows that
and
cos z̄ = cos x cosh y + i sin x sinh y = cos x cosh y − i sin x sinh y.
On applying (11) and (12) of §37, it now follows that sin z̄ = sin z and cos z̄ = cos z.
Solution
We use what we proved in exercise (13) above.
(b) Here
sin (iz) = sin iz = sin ((−i) z̄) = − sin (iz̄) .
Thus − sin (iz̄) = sin (iz) = sin (iz̄) can only hold if sin (iz̄) = 0, i.e. if iz̄ = nπ (n ∈ Z) . But iz̄ = nπ
if and only if z = imπ (m ∈ Z) (set m = −n), and hence we are done.
Solution
By (13) of §37 we will have sin z = cosh 4 whenever sin x cosh y = cosh 4 and cos x sinh y = 0. To find all
z for which sin z = cosh 4, we therefore need to solve the simultaneous equations
Thus y = 0 does not yield a solution and so we must have x = (2k + 1) π/2 (k ∈ Z) .
We now substitute these values into sin x cosh y to get the required solution. Now if x = (2k + 1) (π/2)
then (
1 if k is even
sin x =
−1 if k is odd.
However since cosh y = 12 (ey + e−y ) > 0, the case sin x = −1 must be excluded in the light of the fact
that we will then have that
sin x cosh y = − cosh y < 0 < cosh 4.
Hence we must have that x = (4n + 1) π/2 (n ∈ Z) . (Here we have set k = 2n.) For such an x we get
Now since cosh y = cosh (−y) , it follows that either y = 4 or y = −4 will be enough to ensure that
cosh y = cosh 4. Thus we finally conclude that sin z = cosh 4 ⇔ z = (4n + 1) π/2 ± i4 (n ∈ Z) .
Solution
By (14) of §37 cos z = 2 yields
cos x cosh y − i sin x sinh y = 2.
Now if sin x sinh y = 0 then either x = kπ (k ∈ Z) or y = 0. However if y = 0, then cos x cosh 0 = cos x ≤
1 < 2, and hence this does not yield a solution. Clearly we must have
x = kπ (k ∈ Z) .
Substituting this into cos x cosh y yields cos (kπ) cosh y = (−1)k cosh y. If k is odd we get no solution since
then cos (kπ) cosh y = − cosh y < 0 < 2. If k = 2n is even, then cos (kπ) cosh y = cosh y = 2 will hold
whenever y = ± arccosh (2) . Therefore cos z = 2 precisely when
Alternative
1 iz
e + e−iz = 2
cos z = z⇔
2
2
⇔ eiz − 4eiz + 1 = 0
√
iz 4± 16 − 4 √
⇔ e = =2± 3
2
√
⇔ e−y eix = eiz = 2 ± 3 ei0
√
⇔ e−y = 2 ± 3 and x = 2nπ (n ∈ Z)
(By the remark on top of p. 26 of the textbook.)
√
⇔ y = − ln 2 ± 3 and x = 2nπ (n ∈ Z)
√
⇔ z = 2nπ ± i ln 2 + 3
Solution
Solution
Using the expressions obtained in exercise (4) above we get
= sinh2 x + sin2 y.
Solution
Using the identity in (12), §39
p q q
|sinh x| = sinh x ≤ sinh x + cos y = |cosh z|2 = |cosh z|
2 2 2
Applying the results of exercise 9(b) of §38 (i.e. |cos z| = |cos(x + iy)| ≤ cosh y ) to cosh z = cos (iz) =
cos (−y + ix) we obtain
|cos iz| = |cos (−y + ix)| ≤ cosh x
We thus have
|sinh x| ≤ |cosh z| ≤ cosh x.
Solution
1 z
e − e−z = sin z = 0 ⇔ ez = e−z (×ez )
2
⇔ e2x ei2y = e2z = 1
⇔ 2x = 0, 2y = 2nπ (n ∈ Z)
(see the top of p. 26 of the textbook)
⇔ z = inπ (n ∈ Z)
1 z
e + e−z = cosh z = 0 ⇔ ez = −e−z (×ex )
2
⇔ e2z = −1
⇔ 2z = (2n + 1) πi (n ∈ Z)
π
⇔ z= + nπ i (n ∈ Z)
2
60
Solution
We recall that both cosh z and sinh z are analytic on all of C. In addition, from what we showed in exercise
(8) above, it is clear that cosh z and sinh z are never simultaneously zero. Therefore
sinh z
tanh z =
cosh z
is analytic at all points z where cosh z 6= 0, and zero where sinh z = 0. Therefore tanh z has singularities
where cosh z = 0 (i.e. where z = π2 + nπ i (n ∈ Z)) and zeros where z = nπi (n ∈ Z) .
Solution
Both the functions z → ez and z → sinh z are analytic on all of C and hence so is their composition
sinh (ez ) . By the Theorem of §27, Re (sinh (ez )) must therefore be harmonic on all of C. Now for any
w ∈ C it follows from (9), §39 that
and hence
Re (sinh (ez )) = sinh (ex cos y) cos (ex sin y) .
(a) sinh z = i;
(b) cosh z = 21 ;
Solution
(a)
1 z
e − e−z = i (×ez )
sinh z = i ⇔
2
⇔ (ez )2 − 2iez − 1 = 0
√
z 2i ± −4 + 4
⇔ e = =i
2
π
⇔ ez = ei 2
π
⇔ z=i + 2nπ (n ∈ Z) .
2
61 MAT3705/1
(b)
1 z 1
e + e−z = cosh z = ⇔ ez + e−z = 1 (×ez )
2 2
⇔ (ez )2 − ez + 1 = 0
√ √
z 1± 1−4 1 3
⇔ e = = ±i
2 2 2
In polar form √
1 3 π π π
+i = cos + i sin = ei 3
2 2 3 3
and √
1 3 π π π
−i = cos − + i sin − = e−i 3 .
2 2 3 3
Then √
x −iy z 1 3 π
e e =e = +i = ei 3
2 2
if and only if π
z =0+i + 2nπ (n ∈ Z) .
3
Similarly √
1 3 iπ
z
e = −i = e− 3
2 2
if and only if π
z = 0 + i − + 2nπ (n ∈ Z) .
3
Therefore
1 π
cosh z = ⇔ z = i 2nπ ± (n ∈ Z) .
2 3
Solution
Since cosh z = cos (iz) = − cos (iz + π) by (4) of §38 and (11) of §37, it follows that
Comparing this with what we showed in exercise (16) of §38, it follows that
√
cosh z = −2 ⇔ iz + π = 2kπ ± i ln 2 + 3
√
⇔ z = ± ln 2 + 3 + i (2n + 1) π (n = −k ∈ Z) .
62
63 MAT3705/1
CHAPTER 4
Integrals
Introduction
Our investigation starts with the theory of integrating complex functions along some given contour. Such
integrals are easy enough to define, but finding elegant ways to compute them is another matter. Once
again the concept of analyticity ends up playing a crucial role. The Cauchy-Goursat theorem tells us that
R
if a function is analytic on some domain, then for any closed contour C in that domain, C f (z)dz = 0. By
the Antiderivative theorem, this is equivalent to saying that such a function has an antiderivative on that
domain. Once we are assured of the existence of an antiderivative for a certain class of functions, then for
that class we can try to mimic the theory of integration as done in elementary calculus. However, the ram-
ifications of the Cauchy-Goursat theorem go far beyond the existence antiderivatives. The consequences of
this theorem include the very elegant Cauchy integration formulas (a very powerful integration tool), the
Maximum modulus principle (which tells us where to look for the maximum of |f (z)| on some bounded
closed region), Liouville’s theorem (which basically tells us how difficult it is for an analytic function to
be bounded), and even the Fundamental Theorem of Algebra (which guarantees that every polynomial
equation of degree one or higher has a solution.)
The ability to integrate a complex function on some given contour depends entirely on the extent to which
that function is analytic. If the function is very bad and not at all analytic on or near the contour, then
our only real option is to parametrise the contour and try to compute the integral from “first principles”.
If, however, the integrand, say f , does behave well in the sense of being analytic everywhere except for
some isolated points, we have many more options available to us. If the contour C is not closed and f is
R
analytic on some domain containing C, we may then use antiderivatives to compute C f (z)dz. If C is
R
closed and f is analytic on and inside C, then the Cauchy-Goursat theorem tells us that C f (z)dz = 0.
(Note that to be able to apply the Cauchy-Goursat theorem to some function f , it is not necessary for f to
have no singularities. All we need is for f to have no singularities on or inside C. For example, although
f (z) = z1 has a singularity at z = 0, this point is outside |z − 2| = 1 and so we still have
Z
1
dz = 0
|z−2|=1 z
by this theorem.)
(Of course, if we were integrating over the positively oriented circle |z| = 1 instead of |z| = 3, we would
not have needed partial fractions since in this case
sin z
only the singularity z = 0 lies inside |z| = 1. In particular this means that (z−2) , although not differentiable
at z = 2, is nevertheless analytic inside and on |z| = 1.
g(z)
But what if f is of the form f (z) = h(z) with both g and h analytic, but with h not a polynomial? (For
sin z
example, something like tan z = cos z is such a function.)
R
If in this case we want to compute the integral C f (z)dz on some closed contour C, there is no easy way
to reduce it to an application of the Cauchy integral formulas as we did above. So although such functions
are very nice, the technology we develop in this chapter cannot adequately deal with this class of functions.
Our theory of integration therefore clearly needs a bit more development. However, before we can further
refine the theory of integration, we need the background of the theory of power series expansions. This
we investigate in the next chapter, before returning to the theory of integration.
65 MAT3705/1
Learning outcomes
After studying this chapter you should be able to:
• explain and apply the Cauchy integral formula and its extensions;
Sections to be studied
You need to study all the sections in chapter 4, that is, sections 41 to 59. However, sections 41 and 42
are revision of material covered more extensively in MAT2615.
No proofs are required, but you should be able to be formulate and apply the theorems.
66
Below you will find solutions to some of the exercises in chapter 4. Be sure to attempt these yourself
before you work through the solutions.
Solution:
(a)
Z 1 Z 1
(1 + it)2 dt = [(1 − t2 ) + 2it]dt
0 0
1
= [t − t3 + it2 ] 10
3
1 2
= 1− +i= +i
3 3
(b)
Z 2 2 Z 2
1 1 2
− i dt = − 1 − i dt
1 t 1 t2 t
1
= [− − t − i2 ln t]|21
t
1
= − − i.2 ln 2
2
(c)
Z π π
6
i2t 1 i2t 6
e dt = e
0 i2 0
1 π π
= cos + i sin −1
i2 3 3
√ !
1 1 3
= − +i
i2 2 2
√
3 1
= +i
4 4
(d)
Z ∞ Z b
−zt
e dt = lim e−zt dt
0 b→∞ 0
1 −zt b
= lim − e
b→∞ z 0
1
1 − e−zt
= lim
b→∞ z
Z ∞
Note that e−zb = e− Re(z)b → 0 as b → ∞ since by hypothesis Re (z) > 0. e−zt dt =
Hence
0
1 1
lim 1 − e−zt = .
b→∞ z z
67 MAT3705/1
Evaluate the two integrals on the right here by evaluating the single integral on the left and then identifying
the real and imaginary parts of the value found.
Solution:
Z π Z π Z π
x x
e cos xdx + i e sin xdx = e(1+i)x dx
0 0 0
1 (1+i)x π
= e
1+i
0
1 π iπ
= e e −1
1+i
1 1
= −i (−eπ − 1)
2 2
Hence
Z π
x 1 1 π
e cos xdx = Re −i (−e − 1)
0 2 2
1
= − (eπ + 1)
Z π 2
x 1 1 π
e sin xdx = Im −i (−e − 1)
0 2 2
1 π
= (e + 1) .
2
.
Also, show that this function φ has a positive derivative, as required in the conditions following equation
(9), §43.
68
Solution:
With z and Z defined as before
!! !!
y y
z arctan p = 2 exp i arctan p
4 − y2 4 − y2
!! !!!
y y
= 2 cos arctan p + i sin arctan p
4 − y2 4 − y2
p !
4 − y2 y
= 2 +i
2 2
p
= 4 − y 2 + iy (−2 ≤ y ≤ 2) .
Note that if α = arctan √y i.e. tan α = √ y , then
4−y 2 4−y 2
2
y
α
_ 2
4 y
Finally note that with f (y) = √ y , it follows from the chain rule for real functions that
4−y 2
d 1
φ0 (y) = arctan (f (y)) = 0
2 f (y) .
dy 1 + f (y)
4
f 0 (y) = 3 .
(4 − y 2 ) 2
1 4 1
φ0 (y) = · =p 3 .
y2 4 − y2
1+ 4−y 2
(4 − y 2 ) 2
Clearly φ0 (y) > 0 on −2 < y < 2. (Strictly speaking φ0 (y) doesn’t actually exist at y = ±2.)
represents an arc C1 that intersects the real axis at the points z = 1/n (n = 1, 2, . . .) and z = 0.
69 MAT3705/1
(b) Verify that the arc C1 in part (a) is, in fact, a smooth arc.
Suggestion: To establish the continuity of y (x) at x = 0, observe that
π
0 ≤ x3 sin ≤ x3
x
when x > 0. A similar remark applies in finding y 0 (0) and showing that y 0 (x) is continuous at
x = 0.
Solution:
1
(a) For x = n we have
1 1 1
z = + iy
n n n
3 !
1 1 π
= +i sin 1
n n n
3
1 1
= +i sin (nπ)
n n
1
= .
n
For x 6= n1 (n ∈ Z), πx 6= nπ and hence sin πx 6= 0 (that is Im (z (x)) 6= 0). Thus for any x with
x 6= n1 (n ∈ Z) z (x), is not on the real axis. Clearly z intersects the real axis at precisely the points
0 and n1 (n ∈ Z).
(b) Since x3 , x1 , and sin x are all continuously differentiable for x 6= 0, it follows that x3 sin πx is
continuously differentiable when x 6= 0. Clearly the same is then true of z (x). It remains to show
that z 0 exists and is continuous at 0. Now since −h2 ≤ h2 sin πh ≤ h2 it is clear from the sandwich
π
theorem that h2 sin
h → 0 as h → 0 and hence that
z (h) − z (0)
z 0 (0) = lim
h→0 h π
= lim 1 + ih2 sin
h→0 h
= 1.
π π π
For x 6= 0 we have z 0 (x) = 1 + i 3x2 sin − πx cos . Since −3x2 ≤ 3x2 sin ≤ 3x2 ,
π x x x
−πx ≤ −πx cos ≤ πx it follows from the sandwich theorem that
x π π
z 0 (x) = 1 + i 3x2 sin − πx cos → 1 + i0 = z 0 (0). Thus z 0 is continuous at 0.
x x
where
1. f (z) = (z + 2) /z and C is
3. f (z) = π exp (πz) and C is the boundary of the square with vertices at the points 0, 1, 1 + i, and
i, the orientation of C being in the counterclockwise direction.
Solution of exercise 1:
(c) Here the contour is just the join of the two contours considered in (a) and (b). Thus the integral
turns out to be the sum of the integrals in (a) and (b). Hence
Z
z+2
dz = (−4 + i2π) + (4 + i2π) = i4π.
C z
71 MAT3705/1
Solution of Exercise 3:
C3
i 1+i
C4 C2
0 C1 1
Now
Z Z 1 1
πeπz dz = πeπt dt = eπt 0 = eπ − 1,
C1 0
Z Z 1
πeπz dz = πeπ(1−it) idt
C2 0
Z 1
= πeπ e−iπt idt
0
π −iπt 1
= −e e 0
= −e e−iπ − 1
= 2eπ ,
Z Z 1
πz
πe dz = πeπ((1−t)−i) (−1) dt
C3 0
Z 1
π −iπ
= −πe e e−πt dt
0
1
= −eπ e−πt 0
= eπ − 1,
and
Z Z 1
πz
πe dz = πe−πi(1−t) (−i) dt
C4 0
Z 1
−iπ
= −iπe eiπt dt
0
1
= eiπt 0
= −2.
72
Therefore
Z 4 Z
X
πeπz dz = πeπz dz
C n=1 Cn
= (e − 1) + 2eπ + (eπ − 1) + (−2)
π
= 4 (eπ − 1) .
Solution of Exercise 4:
Here
C : z = x + ix3 where − 1 ≤ x ≤ −1.
Hence Z Z 1
f x + ix3 1 + i3x2 dx.
f (z) dz =
C −1
Therefore
Z Z 0
f x + ix3 1 + i3x2 dx
f (z) dz =
C −1
1Z
f x + ix3 1 + i3x2 dx
+
0
Z 0 Z 1
1 + i3x2 dx + 4x3 1 + i3x2 dx
=
−1 0
0 1
= x + ix −1 + x4 + i2x6 0
3
= (1 + i) + 1 + i2
= 2 + i3.
where m and n are integers and C is the unit circle |z| = 1, taken counterclockwise.
Solution:
The circle |z| = 1 may be parametrised by z (θ) = eiθ , 0 ≤ θ ≤ 2π. By exercise 2(b) of §9 we have
z (θ) = eiθ = e−iθ . (To see this note that
Hence
Z Z 2π
n
m
z (z) dz = eimθ e−inθ ieiθ dθ
C 0
Z 2π
= i ei(m−n+1)θ dθ
0
2πi if m = n − 1
= 1 2π
ei(m−n+1)θ 0 if m 6= n − 1
(m − n + 1)
(
2πi if m = n − 1
=
0 if m 6= n − 1.
R dz
(a) Co = 2πi;
z − z0
(z − z0 )n−1 dz = 0 (n = ±1, ±2, . . .).
R
(b) C0
Solution:
(a)
Z Z π
dz 1
iθ
= iR e dθ
C0 z − z0 −π (z0 + R eiθ ) − z0
Z π
= i dθ = i2π
−π
(b)
Z Z π n−1
n−1
(z − z0 ) dz = z0 + R eiθ − z0 iR eiθ dθ
C0 −π
Z π
= i Rn einθ dθ
−π
1 n inθ π
= R e
n −π
1 n iπ n n
= R e − e−iπ
n
1 n
= R ((−1)n − (−1)n )
n
= 0
74
z − 1 ≥ |z|2 − 1 = 3, i.e.
2 1 1
≤ .
|z 2 − 1| 3
Therefore Z
1 1 π
z 2 − 1 dz ≤ 3 π = 3 .
C
Solution:
Here the length of C is the sum of the distances from 0 to 3i, 3i to −4 and −4 to 0, that is
y
3i
x
_
4 O
Now for any z on C we see from the sketch that Re (z) ≤ 0 and hence that |ez | = eRe(z) ≤ e0 = 1. Also
on C the point furthest away from the origin (i.e. the point where |z − 0| is a maximum) is −4. Therefore
on C |z| = |z| = |z − 0| ≤ |−4 − 0| = 4, whence |ez − z| ≤ |ez | + |z| ≤ 5. Consequently
Z
(ez − z) dz ≤ 5 × 12 = 60.
C
Solution:
The circumference of the circle CR :|z| = R is 2πR. Recall that −π < Arg(z) ≤ π. In addition since,
R > 1, ln R > 0. Therefore for any z with |z| = R we have
Consequently Z
Log (z) ln R + π 2π (ln R + π)
2
dz ≤ 2
2πR = .
CR z R R
obtained in Exercises 8(a) and 9(a) of §38 show that |sin z| ≥ 1 on the vertical sides of the square
and that |sin z| > sinh (π/2) on the horizontal sides. Thus show that there is a positive constant A,
independent of N , such that |sin z| ≥ A for all points z lying on the contour CN .
and hence that the value of this integral tends to zero as N tends to infinity.
Solution:
1
(a) By means of the inequality |sin z| ≥ |sin x| it follows that on the lines z = ± N + 2 π + iy we have
1
|sin z| ≥ sin ± N + π
2
1
= ± sin
N+ π
2
N
= (−1)
= 1.
76
Now note that |sin z| ≥ |sinh y|. Therefore on the lines z = x ± i N + 21 π we have
1
|sin z| ≥ sinh ± N +
π
2
1
= ± sinh N+ π
2
π
≥ sinh ( sinh is increasing on [0, ∞))
2
> 1.
and hence
1 1 4
z 2 sin z ≤ = .
1 2 2 (2N + 1)2 π 2
N+ 2 π
In addition the length of CN is
1
4×2 N + π = 4 (2N + 1) π.
2
Therefore
Z
dz 4 16
≤ 2 2 × 4 (2N + 1) π = (2N + 1) π .
CN z 2 sin z (2N + 1) π
→0 as N → ∞.
R dz
It follows that CN → 0 as N → ∞.
z 2 sin z
(a)
1+i
1 3 1+i
Z
2
z dz = z
0 3 0
1
= (1 + i)3
3
1
1 + 3i + 3i2 + i3
=
3
1 2
= (−2 + 2i) = (−1 + i )
3 3
77 MAT3705/1
(b)
Z π+2i z z π+2i
cos dz = 2 sin
2 2 0
0
π
= 2 sin +i −0
2 π π
= 2 sin cosh 1 + i cos sinh 1
2 2
= 2 cosh 1
1
= e+
e
(c)
Z 3
3
3 1 4 1 4
1 − (−1)4 = 0
(z − 2) dz = (z − 2) =
1 4 1 4
and where the path of integration is any contour from z = −1 to z = 1 that, except for its end points, lies
above the real axis.
Suggestion: Use an antiderivative of the branch
i π 3π
z = exp (i log z) |z| > 0, − < arg z <
2 2
of the same power function.
Solution:
Except for the endpoints the entire contour lies in the region 0 < θ < π, |z| > 0. This region lies
π 3π
in both (|z| > 0, −π < Arg (z) < π) and |z| > 0, − < arg (z) < . Hence if log z is the branch
2 2
π 3π
|z| > 0, − < arg (z) < , then for any z on the given contour (except the endpoints ±1) we will
2 2
have that Log(z) = log (z) and hence that
Therefore replacing the one branch of z i by the other in the integral will not change the value of the
integral. We may therefore use the branch
i π 3π
z = exp (i log z) |z| > 0, − < arg (z) < .
2 2
For the same branch of log z the branch
(1+i) π 3π
z = exp ((1 + i) log z) |z| > 0, − < arg (z) <
2 2
78
when the contour C is the circle |z| = 1, in either direction, and when
z2 1
(a) f (z) = ; (b) f (z) = ze−z ; (c) f (z) = 2 ;
z+3 z + 2z + 2
Solution:
Since Z Z
− f (z) dz = f (z) dz
C −C
it is clear that Z Z
f (z) dz = 0 ⇔ f (z) dz = 0.
C −C
We may therefore assume that C is positively oriented in each of (a) – (f).
z2
(a) The only point where f (z) = is not differentiable is z = −3. Therefore f is analytic inside
z+3
and on |z| = 1, and so by Cauchy’s theorem
z2
Z
dz = 0.
C z+3
79 MAT3705/1
by Cauchy’s theorem.
1 √
fails to be differentiable where z 2 +2z +2 = 0, i.e. where z = 12 −2 ± 4 − 8 =
(c) f (z) =
z2
+ 2z + 2 √
−1±i. However since |−1 ± i| = 2 > 1, both these points lie outside C and so by Cauchy’s theorem
Z
1
2
dz = 0.
C z + 2z + 2
1 1
(d) f (z) = sechz = fails to be differentiable where cosh z = 0, i.e. where z = i n + π (n ∈ Z).
cosh
z 2
1 π π
However i n + ≥ > 1 for each n ∈ Z. That is all these points lie outside the circle C and
2 2 2
so Z
sechzdz = 0
C
by Cauchy’s theorem.
sin z 1
(e) f (z) = tan z = fails to be differentiable where cos z = 0, that is where z = n + π (n ∈ Z).
cosz 2
1
π ≥ π2 > 1, all these points are outside the circle |z| = 1 and hence
Since n +
2
Z
tan zdz = 0.
C
(f) z → Logz fails to be differentiable on the line segment y = 0, x ≤ 0. (See (6) in §31.) Hence
z → Log(z + 2) fails to be differentiable precisely where y = 0 and x + 2 ≤ 0, i.e. where y = 0,
x ≤ −2. Since all these points lie outside the circle C, it follows that
Z
Log (z + 2) dz = 0.
C
1
(a) f (z) = 2 fails to be differentiable where 3z 2 + 1 = 0, i.e. where z = ±i √13 . Both these points
3z + 1
clearly lie inside C2 , that is in the region −1 < x < 1, −1 < y < 1.
z+2 z
(b) f (z) = fails to be differentiable where sin 2 = 0, that is where zn = 2nπ (n ∈ Z). Now for
sin z2
n = 0, z0 = 0 lies inside C2 . For n 6= 0 we have
√ iθ/2
π 3π
f (z) = re r > 0, − < θ <
2 2
1
of the multiple–valued function z 2 . Show that
Z
f (z) dz = 0
C
by evaluating separately the integrals of f (z) over the semicircle and the two radii which constitute C.
Why does the Cauchy–Goursat theorem not apply here?
Solution:
The
half–circle C1 from 1 to −1 (with |z| = 1) which lies in the upper half–plane, lies entirely in the region
π 3π
|z| > 0, − < θ < . By (5) of §34 in the textbook the branch
2 2
2 3 2 3 π 3π
z 2 = exp log z |z| > 0, − < arg (z) <
3 3 2 2 2
1 3
is then an antiderivative of z 2 on this region. Therefore using this branch of 23 z 2 , we conclude that
2 3 −1
Z
1
z dz =
2 z2
C1 3 1
2 3 3
= exp log (−1) − exp log (1)
3 2 2
2 3
= exp (iπ) − exp (0)
3 2
2 2
= − −i .
3 3
81 MAT3705/1
(The values of log (−1) and log (1) were computed in the examples of §32.) For any z on the line segment
(|z| > 0, arg (z) = π) we have
log (z) = ln |z| + iπ.
2 3 1 2
Z Z 1√
1
z 2 dz = tdt = t 2 = .
C3 0 3 0 3
Therefore
Z 3 Z
1 X 1
z dz =
2 z 2 dz = 0.
C n=1 Cn
1
Finally note that we may not use Cauchy’s theorem since z 2 fails to be differentiable at the point 0 on
the contour.
Solution:
Let R denote the interior of the contour C. Then with u = x and v = −y, it follows from (4) in §50 that
Z Z
1 1
zdz = (x − iy) dz
2i C 2i C
Z
1
= (u + iv) dz
2i C
ZZ ZZ
1
= (−vx − uy ) dA + i (ux − vy ) dA
2i R
R
Z Z ZZ
1
= (−0 − 0) dA + i (1 − (−1)) dA
2i R R
ZZ
= dA
R
as required.
Exercise 1, §57, p. 170
Let C denote the positively oriented boundary of the square whose sides lie along the lines x = ±2 and
y = ±2.Z Evaluate each of these integrals:
e−z dz
Z Z
cos z zdz
(a) ; (b) 2
dz ; (c) ;
C z − (πi/2) C z (z + 8) C 2z + 1
Z Z
cosh z tan (z/2)
(d) 4
dz ; (e) 2 dz (−2 < x0 < 2)
C z C (z − x0 )
Solution:
π
(a) Since e−z is an entire function and i is inside C, we have that
2
e−z
Z
−(iπ/2)
π dz = 2πie = 2πi (−i) = 2π
C z − i 2
cos z/ z 2 + 8
Z Z
cos z
2
dz = dz
C z (z + 8) C z
cos z
= 2πi
z 2 + 8 z=0
π
= i .
4
2πi d3
Z
cosh z
4
dz = cosh z
C z 3! dz 3
z=0
πi
= sinh 0
3
= 0.
sin z2
(e) tan (z/2) = fails to be differentiable where cos (z/2) = 0, that is where z = (2n + 1) π
cos z2
(n ∈ Z). However for any n, |(2n + 1) π| ≥ π > 2. Therefore all these points lie outside C and so
tan (z/2) is analytic (differentiable) inside and on C. By contrast x0 lies inside C. Thus
Z
tan (z/2) 2πi d
2 dz = tan (z/2)
C (z − x0 ) 1! dz
z=x0
= iπ sec2 (x0 /2) .
2z 2 − z − 2
Z
g (w) = 6 3) ,
dz (|w| =
C z−w
then g (2) = 8πi. What is the value of g (w) when |w| > 3?
Solution:
The point z = 2 clearly lies inside the circle C : |z| = 3. Therefore
2z 2 − z − 2
Z
2
g (z) = dz = 2πi 2z − z − 2 = 8πi.
C z−2 z=2
2z 2 − z − 2
Z
g (w) = dz = 0
C (z − w)
f 0 (z) dz
Z Z
f (z) dz
2 = 2.
C (z − z0 ) C (z − z0 )
84
Solution:
If f is analytic inside and on a simple closed contour then so is f 0 by theorem 1 of §57. Therefore on
applying (6) in §55 first to f with n = 1 and then to f 0 with n = 0, it follows that
f 0 (z)
Z Z
f (z) 0
2 dz = 2πif (z 0 ) = dz
C (z − z0 ) C (z − z0 )
for every z0 inside C. If now z0 is outside C, both f (z) /(z − z0 )2 and f 0 (z) / (z − z0 ) are analytic inside
and on C. For such z0 we then have by Cauchy’s theorem that
f 0 (z)
Z Z
f (z)
2 dz = 0 = dz.
C (z − z0 ) C (z − z0 )
Solution:
Let C be the unit circle z = eiθ (−π < θ ≤ π) centred at 0. Since eaz is an entire function, it follows that
eaz
Z
az
dz = 2πie = 2πi.
C z z=0
Since
in §56.
Solution:
We show how one may use the binomial formula and mathematical induction to prove (2) of §56. The
specific case n = 2 then corresponds to the solution of the above problem.
Let C be a simple closed contour and z0 a point inside C. Then by the theorem in §55
Z
1 f (z)
f (z0 ) = dz.
2πi C z − z0
where ∆z is small enough so that z0 + ∆z is inside C. Let M be the maximum of |f (z)| on C, L the
length of C and d the shortest distance from z0 to the points z of C. For any z ∈ C and 0 < |∆z| < d we
then have
|z − z0 | ≥ d |z − z0 − ∆z| ≥ |z − z0 | − |∆z| ≥ d − |∆z| .
Therefore in the above sum all the terms except the one with m = k tend to 0 as ∆z → 0. To conclude
the proof we show that for the integral in the m = k term we have
Z Z
f (z) dz f (z)
k+1
→ k+2
dz
C (z − z0 − ∆z) (z − z0 ) C (z − z0 )
as ∆z → 0. Therefore the term in (β) above tends to zero. Applying this to (α) it follows that
Z
(k + 1)! f (z)
as ∆z → 0. Therefore f (k+1) (z 0 ) exists and equals k+2
dz. Thus by induction
2πi C (z − z0 )
Z
(n) n! f (z)
f (z0 ) = dz
2πi C (z − z0 )n+1
for all n = 0, 1, 2, . . . .
Suggestion: Use Cauchy’s inequality (§57) to show that the second derivative f 00 (z) is zero everywhere in
the plane. Note that the constant MR in Cauchy’s inequality is less than or equal to A (|z0 | + R).
Solution:
Let f be given such that |f (z)| ≤ A |z| for all z. Then for any z on the positively oriented circle
CR : |z − z0 | = R we have |f (z)| ≤ A |z| = A |z + (z − z0 )| ≤ A (|z0 | + |z − z0 |) = A (|z0 | + R). If in
addition f is entire, Cauchy’s integration formulae ensure that for each z0
Z
2! f (z)
f (2) (z0 ) = dz.
2πi CR (z − z0 )3
As R → ∞, the right–hand side tends to 0 thereby proving that f (2) (z0 ) = 0. Since z0 was arbitrary this
shows that f (2) = 0, and hence that f (z) = a1 z + a0 . Finally note that since |f (z)| ≤ A |z| for all z, f (0)
must be 0, that is a0 = 0. Therefore f (z) = a1 z.
Suggestion: Apply Liouville’s theorem (§58) to the function g (z) = exp [f (z)].
Solution:
Let f (z) = u (x, y) + iv (x, y) be entire. Since ez is also entire, so is ef (z) . Now note that
f (z)
e = eu(x,y)
If therefore u (x, y) ≤ u0 for some u0 and all z, ef (z) is a bounded entire function since then
f (z)
e = eu(x,y) ≤ eu0
for all z. By Liouville’s theorem ef (z) , and therefore also ef (z) = eu(x,y) , must then be constant. Thus
u (x, y) = ln eu(x,y) is constant.
88
Solution:
If f (z0 ) = 0 for some z0 on the perimeter of R, then |f (z0 )| = 0 is surely the minimum of |f | in which
case we are done as |f (z)| > 0 for all z in the interior of R.
1
If f (z) 6= 0 on the perimeter of R as well, then
is continuous on R and analytic on its interior since the
f
same is true of f . By the maximum modulus principle f1 then assumes its maximum on the perimeter
Solution:
Since f (z) = u (x, y) + iv (x, y) is continuous on the closed bounded region R and analytic on its interior,
the same is true of ef (z) . By (4) of §30, ef (z) 6= 0 on R. Therefore by what we showed in exercise 2 above
f (z)
e = eu(x,y) assumes its minimum on the perimeter of R and not the interior. Since ln is increasing
on (0, ∞) and eu(x,y) > 0, the same must then be true of ln eu(x,y) = u (x, y).
Solution:
For f (z) = ez we clearly have Re [f (z)] = ex cos y. Now in the rectangular region 0 ≤ x ≤ 1, 0 ≤ y ≤ π
ex increases from 1 to e as x varies from 0 to 1, and cos y decreases from 1 to −1 as y varies from 0 to π.
Therefore ex cos y assumes a maximum of e at the point z = 1 + i0 = 1 and a minimum of −e at the point
z = 1 + iπ on the perimeter of this rectangle.
89 MAT3705/1
.
.
CHAPTER 5
Series
Introduction
The basic goal of this chapter is to investigate the extent to which complex functions can be written
as power series. The two theorems which form the cornerstones of this chapter are Taylor’s theorem
and Laurent’s theorem. Taylor’s theorem establishes yet another very important fact regarding analytic
functions. If a function is analytic in some region of the form |z − a| < R, then in that region the function
can be written as a power series centred at a. But by the corollary to Theorem 1 in section 71, the converse
is also true! Together these two results establish the fact that a function is analytic at a point if and only
if it can be written as a power series in some neighbourhood of that point. The analytic functions are
therefore precisely the class of functions which allow for power series expansions.
But what if we are interested in the behaviour of a function f at a given point a, and f is analytic all
around this point, but not at the actual point we are interested in? Is there anything that can be said
about such functions as far as power series expansions are concerned? Because f is not analytic at a itself,
Taylor’s theorem is no longer valid. However, all is not lost. In the absence of Taylor’s theorem, Laurent’s
theorem comes to our rescue. What this theorem tells us is that if f is analytic on a region of the form
r < |z − a| < R (which excludes the troublesome point a), then provided we allow negative powers of z − a,
we can in that region still write f as a series of powers of z − a. Such an expansion in powers of z − a
we will then call a Laurent series for f in the region r < |z − a| < R.
When we are studying the behaviour of functions around isolated singular points it is this theory of Laurent
series that proves to be especially helpful. So if we want to be able to study the behaviour of complex
functions at such points, we clearly need a bit of practice in dealing with Laurent series. The material in
section 73 will demonstrate a wide range of techniques which prove to be very useful in the computation
of Laurent series.
We close our overview of this chapter by presenting some hints on computing Laurent series of rational
functions. If approached, correctly, then at least for rational functions, this should be a reasonably
q(z)
straightforward process. In short, the idea runs as follows: If f (z) = p(z) , where p and q are polynomials,
needs to be expanded as a Laurent series centred at a (i.e. in powers of (z − a)) we can use Taylor’s
1
theorem to rewrite q(z) in powers of (z − a). Next we can use partial fractions to decompose p(z) into a
K
sum of terms of the form (z−b)n . Each of these terms is then separately written in Laurent form, and finally
90
K
all the various expansions for q(z) and the (z−b)n ’s combined algebraically to get the Laurent expansion
q(z) 1
for f (z) = p(z) . So all that remains is to explain how to write terms like (z−b)n in Laurent-form.
However, before doing so we note that a function f may have several Laurent expansions centred at
z = a, depending on the number of singularities it has. A careful look Laurent’s theorem reveals that if a
particular form of the Laurent expansion in powers of (z − a) holds at a point (say w0 ) then this form
will converge on the largest annulus of the form S < |z − a| < R (where 0 ≤ S < R) which contains w0
and on which f is still analytic! As soon as we leave the annulus and move beyond a point where f is
NOT analytic, the expansion changes. For example, if f has two singularities at say z0 and z1 with say
but not on the circles |z − a| = R0 and |z − a| = R1 as such, since these contain z0 and z1 . So by Laurent’s
theorem on each of the annuli |z − a| < R0 , R0 < |z − a| < R1 and R1 < |z − a| < ∞, f will have some
Laurent expansion. However, the expansion may be different on each of these sets!
1
We finally indicate how a term such as (z−b) (where b 6= a) may be written as a Laurent series centred at
1 1
z = a. The expansions for (z−b)2 , (z−b)3 , . . . may then be obtained from this one by differentiating the
1 1 3
expansion for (z−b) . (Note for example that (z−b)4
d
= − 16 dz 1
3 (z−b) .) Now we know that
∞
1 X
= 1 + w + w2 + ... = wn (|w| < 1).
1−w
n=0
If z−a
b−a < 1 (i.e. |z − a| < |b − a|) we may write
1 −1 1
=
(z − b) (b − a) 1 − ( z−a
b−a )
∞ ∞
1 −1 X (z − a)n X −1
= = (z − a)n .
(z − b) (b − a) (b − a)n (b − a)n+1
n=0 n=0
b−a b−a
If, on the other hand, z−a < 1 (i.e. |z − a| > |b − a|) we set w = ( z−a ) to get
!
1 1 1
= b−a
(z − b) (z − a) 1 − ( z−a )
∞
X (b − a)n
1
=
(z − a) (z − a)n
n=0
∞
X 1
= (b − a)m−1 .
(z − a)m
m=1
91 MAT3705/1
Suppose for example that we are asked to compute the Laurent series of
z 2 − 8z
(z − 2)2 (z + 1)
in each of the following regions: (i) |z − 1| < 1, and (ii) 1 < |z − 1| < 2. So for each of these regions we
must find an expansion in powers of (z − 1) that converges on
the given region.
First of all, note that by means of partial fractions we can see that
z 2 − 8z 1 4
2
= − .
(z − 2) (z + 1) z + 1 (z − 2)2
1 |z−1|
We first consider the term z+1 . If |z − 1| < 2 (equivalently −2 < 1), then
1 1
=
z+1 z−1+2
∞
1X z−1 n
=
2 −2
n=0
∞
X
n (z − 1) n
= (−1)
2n+1
n=0
1 (z − 1)1 (z − 1)2
= − + − ...
2 22 23
−2
If, however, |z − 1| > 2 (equivalently z−1 < 1), then
!
1 1 1
= −2
z+1 z−1 1 − ( z−1 )
∞
1 X −2 n
=
z−1 z−1
n=0
∞
X 2n
= (−1)n
(z − 1)n+1
n=0
1 2 4
= − 2
+ − ...
z−1 (z − 1) (z − 1)3
1
Now consider the term z−2 . If |z − 1| < 1, then
1 1
= (−1)
z−2 1 − (z − 1)
∞
X
= (−1) (z − 1)n
n=0
= −1 − (z − 1) − (z − 1)2 − ...
= −z − (z − 1)2 − ...
92
1
and if |z − 1| > 1 (equivalently |z−1| < 1), then
!
1 1 1
= 1
z−2 z−1 1 − z−1
∞
1 X 1
=
z−1 (z − 1)n
n=0
1 1 1
= + 2
+ + ...
z − 1 (z − 1) (z − 1)3
and
1 d 1
= −
(z − 2)2 dz z − 2
1 2 3
= 2
+ 3
+ + ...
(z − 1) (z − 1) (z − 1)4
∞
X (n + 1)
= whenever |z − 1| > 1.
(z − 1)n+2
n=0
In the region |z − 1| < 1 it will of course also hold that |z − 1| < 2. So in this region
z 2 − 8z 1 4
= −
(z − 2)2 (z + 1) z + 1 (z − 2)2
∞ ∞
! !
n
n (z − 1)
X X
n
= (−1) −4 (n + 1)(z − 1)
2n+1
n=0 n=0
1 1 1
= ( − 4) + (− − 8)(z − 1) + ( − 12)(z − 1)2 + ...
2 4 8
= −3.5 − 8.25(z − 1) − 11.875(z − 1)2 + ...
z 2 − 8z 1 4
= −
(z − 2)2 (z + 1) z + 1 (z − 2)2
∞ ∞
! !
X (z − 1)n X (n + 1)
= (−1)n n+1 −4
2 (z − 1)n+2
n=0 n=0
8 4 1 (z − 1) (z − 1)2
= ... − − + − + + ...
(z − 1)3 (z − 1)2 2 22 23
93 MAT3705/1
Learning outcomes
After studying this chapter you should be able to:
• represent a function f (z) analytic throughout an annular domain R1 < |z − z0 | < R2 centred at z0 as
a Laurent series (relative to C, a positively simple closed contour around z0 );
p(z) J K L
• decompose f (z) = (z−a)2 (z−b)
into a sum of the form f (z) = (z−a) ) + (z−a)2
+ (z−b) using partial
fractions;
• apply the rules of integration and differentiation to a power series to obtain a Laurent series of a
specific function f (z);
Sections to be studied
You need to study sections 60 to 73 in chapter 5. Some sections are largely revision of material dealt with
in MAT2613. You need not study the proofs of theorems in this chapter, although you should be familiar
with both their formulation and application.
94
Below you will find solutions to some of the exercises in §60 − 73 of chapter 5. Be sure to attempt these
yourself before you work through the solutions.
Solution:
For zn = n12 + i we have
1
|zn − i| =
.
n2
Given any ε > 0 we may select Nε ∈ N so that Nε > √1ε . Then
1 1 1
n ≥ Nε ⇒ ε > 2
≥ 2 ⇒ |zn − i| = 2 < ε.
Nε n n
Hence zn → i.
1
Alternatively note that Im (zn ) = yn = 1 is a constant sequence and that Re (zn ) = xn = 2 tends to zero.
n
Hence
lim zn = lim (xn + iyn ) = 0 + i1 = i.
n→∞ n→∞
Solution:
1
For n = 2k even, we have z2k = 1 + i 2 . This is a point in the first quadrant for which tan (θ2k ) =
4k
1
(where θ 2k = Arg(z 2k )). Hence
4k 2
1
θ2k = arctan → 0.
4k 2
1
Similarly for n = 2k − 1 odd, z2k−1 = 1 − i is a point in the fourth quadrant and hence here
(2k − 1)2
1
θ2k−1 = arctan − →0
(2k − 1)2
where θ2k−1 = Arg(z2k−1 ). Clearly (θn ) converges to 0.
Solution:
Let (zn ) and z be given. By the inequality in §5
Now if lim zn = z, then given ε > 0 we can find Nε so that |zn − z| < ε whenever n ≥ Nε . But by the
above inequality we then have that
when 0 < r < 1. (Note that these formulas are also valid when r = 0.)
Solution:
For z = reiθ (with 0 < r < 1) we have |z| = r < 1. Hence by the example in §61,
∞ ∞
X X 1 1
rn einθ = zn = = .
1−z 1 − reiθ
n=0 n=0
We proceed to compute the real and imaginary parts of the term 1/ 1 − reiθ .
1 1 (1 − r cos θ) + ir sin θ
= ×
1 − reiθ (1 − r cos θ) − ir sin θ (1 − r cos θ) + ir sin θ
(1 − r cos θ) + ir sin θ
=
(1 − r cos θ)2 + r2 sin2 θ
(1 − r cos θ) + ir sin θ
=
1 − 2r cos θ + r2 cos2 θ + sin2 θ
(1 − r cos θ) r sin θ
= 2
+i .
1 − 2r cos θ + r 1 − 2r cos θ + r2
∞
X ∞
X ∞
X ∞
X
Now since rn einθ = rn (cos nθ + i sin nθ) = rn cos nθ + i rn sin nθ, it easily follows that
n=0 n=0 n=0 n=0
∞ ∞
!
X X
rn cos nθ = Re rn einθ
n=0 n=0
1
= Re
1 − reiθ
1 − r cos θ
=
1 − 2r cos θ + r2
and similarly that
∞
X r sin θ
rn sin nθ = .
1 − 2r cos θ + r2
n=0
96
Solution:
From example 6 of §64 we know that
∞
X z 2n
cosh z = .
(2n)!
n=0
Therefore 2n !
∞ ∞
2
X z2 X z 4n+1
z cosh z =z = .
(2n)! (2n)!
n=0 n=0
Solution:
From the example in §61 we know that
∞
1 X
= wn if |w| < 1.
1−w
n=0
Therefore
∞ ∞
1 1 X X
= = (−w)n = (−1) wn if |w| < 1.
1+w 1 − (−w)
n=0 n=0
Consequently
!
z z 1
f (z) = 4
= z4
z +4 4 1+ 4
∞
zX z 4n
= (−1)n n
4 4
n=0
∞
X z 4n+1
= (−1)n
4n+1
n=0
∞
X z 4n+1
= (−1)n
22n+2
n=0
√
4
z
where the series converges whenever < 1, i.e. when |z| < 2.
4
Solution:
For f (z) = sinh z we have
f (2k) (z) = sinh z f (2k+1) (z) = cosh z
Solution:
By Taylor’s theorem the Taylor series centred at 0 (the Maclaurin series) for tanh z will exist on the largest
disc of the form |z| < R0 on which tanh z is still analytic. Therefore the largest circle within which the
Maclaurin series will converge is the circle |z| = R0 where R0 is the distance from 0 to the nearest point
sinh z
where tanh z fails to be analytic. Now tanh z = cosh z fails to be analytic where cosh z = 0, i.e. where
z = (2k + 1) πi (k ∈ Z). The singularities closest to 0 are ±πi and so the Maclaurin series converges
inside the circle |z| = |±πi| = π. The first three derivatives of f (z) = tanh z are
Therefore the coefficients of the first four terms of the Maclaurin series will be
f (1) (0) 1
f (0) = tanh 0 = 0, = sech2 0 = 1
1! 1!
f (2) (0) 1 f (3) (0) 1 1
= 0 = 0; = (4 × 0 + 2) = .
2! 2! 3! 3! 3
with the first two nonzero terms in the series being z + 13 z 3 + . . ..
Solution:
From representation (3) of §64 we know that
∞
X z 2n+1
sin z = (−1)n (|z| < ∞) .
(2n + 1)!
n=0
Now by Taylor’s theorem the coefficients of the z 2n and z 2n+1 terms in the Maclaurin series are exactly
for all n = 0, 1, 2, . . . .
X∞ z 2n+1
(b) From (3) of §64, sin z = (−1)n . Therefore
n=0 (2n + 1)!
∞ 2n+1 !
sin z 2 z2
1 1 X n
sin z 2 = 4
= (−1)
z4 z 4 z (2n + 1)!
n=0
∞
X z 4n−2
= (−1)n
(2n + 1)!
n=0
1 z 2 z 6 z 10
= − + − + · · · valid for 0 < |z < ∞|
z2 3! 5! 7!
Solution:
∞
X (−1)n 1
From (2) of §64 sin w = w2n+1 (|w| < ∞). Therefore on setting w = 2 we get
(2n + 1)! z
n=0
∞ 2n+1
(−1)n
1 X 1
z 2 sin = z2
z2 (2n + 1)! z 2
n=0
∞
X (−1)n 1
= . (0 < |z| < ∞)
(2n + 1)! z 4n
n=0
Extra example:
Derive the Laurent series representation
"∞ #
ez 1 X (z + 1)n 1 1
= + + (0 < |z + 1| < ∞) .
(z + 1)2 e
n=0
(n + 2)! z + 1 (z + 1)2
Solution:
∞
X X∞
1 n 1
For any w ∈ C we have ew = n! w . Therefore eez = ez+1 = (z + 1)n , whence
n=0 n!
n=0
" ∞ #
ez 1 1X 1
= (z + 1)n
(z + 1)2 (z + 1)2 e n=0 n!
" #
1 1 1 1 (z + 1) (z + 1)2
= + + + + + ···
e (z + 1)2 (z + 1) 2! 3! 4!
∞
" #
1 1 1 X 1 n
= + + (z + 1) .
e (z + 1)2 (z + 1) (n + 2)!
n=0
Solution:
We know from the example (3) in §65 that
1
= 1 + z + z2 + z3 + · · · (|z| < 1)
1−z
Therefore for 0 < |z| < 1 we surely have
1 1 2 3
= 1 + z + z + z + · · ·
z 2 (1 − z) z2
1 1
= 2
+ + 1 + z + z2 + · · ·
z z
∞
1 1 X n
= + + z .
z2 z
n=0
100
1
Now when < 1 (i.e. when 1 < |z| < ∞), it follows that
z
! 2 3 !
1 1 1 1 1 1 1
= − =− 1+ + + + ···
1−z z 1 − z1 z z z z
1 1 1
= − − 2 − 3 − ··· .
z z z
Therefore
∞
1 1 1 1 X 1
= − − − − · · · = − .
z 2 (1 − z) z3 z4 z5 zn
n=3
Solution:
z
To expand in powers of (z − 1) we may write
(z − 1) (z − 3)
!
1 1 1 1
= = − z−1
z−3 (z − 1) − 2 2 1− 2
∞ n
1 X (z − 1)
= − .
2 2n
n=0
z − 1
whenever < 1 (i.e. |z − 1| < 2). Therefore for 0 ≤ |z − 1| < 2 we get
2
∞
!
z z 1 X (z − 1)n
= −
(z − 1) (z − 3) z−1 2 2n
n=0
∞
!
(z − 1)n
1 X
= 1+ −
z−1 2n+1
n=0
∞
1 1 X 1 1
= − − + (z − 1)n
2 (z − 1) 2n+1 2n+2
n=0
∞
1 X 3
= − − (z − 1)n .
2 (z − 1) 2n+2
n=0
Extra example:
Write the two Laurent series in powers of z that represent the function
1
f (z) =
z (1 + z 2 )
Solution:
For −z 2 = |z|2 < 1 (i.e. |z| < 1) we have that
1 1 2 3
= 1 + −z 2 + −z 2 + −z 2 + · · ·
=
1 + z2 1 − (−z 2 )
= 1 − z2 + z4 − z6 + z8 − · · · .
Therefore
1 1
= − z + z3 − z5 + z7 − · · ·
z(1 + z 2 ) z
∞
1 X
= + (−1)n+1 z 2n+1 . (0 < |z| < 1)
z
n=0
If on the other hand − z12 < 1 (i.e. 1 < |z| < ∞ ), then
∞ ∞
1 n X (−1)n
1 1 1 X
= 2 = 2 − 2 = .
1 + z2 z 1 − − z12 z 2n+2
z z
n=0 n=0
Thus
∞ ∞ ∞
1 1 X (−1)n X (−1)n X (−1)n+1
= = =
z (1 + z 2 ) z z 2n+2 z 2n+3 z 2n+1
n=0 n=0 n=1
(a) Let z be any complex number, and let C denote the unit circle
w = eiφ ( − π ≤ φ ≤ π)
in the w plane. Then use that contour in expression (5), §66, for the coefficients in a Laurent series,
adapted to such series about the origin in the w plane, to show that
∞
z 1 X
exp w− = Jn (z) wn (0 < |w| < ∞),
2 w n=−∞
where Z π
1
Jn (z) = exp [−i (nφ − z sin φ)] dφ (n = 0, ±1, ±2, . . . ).
2π −π
(b) With the aid of Exercise 6,§42 regarding certain definite integrals of even and odd complex–valued
functions of a real variable, show that the coefficients in part (a) can be written
1 π
Z
Jn (z) = cos (nφ − z sin φ) dφ (n = 0, ±1, ±2, . . . ).
π 0
102
Solution:
w = eiθ − π < θ ≤ π.
For any fixed z ∈ C it then follows from the Theorem in §66 that we can write the function
z 1
fz (w) = exp w−
2 w
and
the function θ → cos (nθ − z sin θ) is even whereas θ → sin (nθ − z sin θ) is odd. Therefore by exercise
5 of §42 Z π
1 π
Z
1
cos (nθ − z sin θ) dθ = cos (nθ − z sin θ) dθ
2π −π π 0
and Z π
1
sin (nθ − z sin θ) dθ = 0.
2π −π
Consequently Z π
1
Jn (z) = cos (nθ − z sin θ) dθ.
π 0
∞
X (−1)n (n − 1)
1
= for 1 < |z − 1| < ∞
z2 (z − 1)n
n=2
Solution:
We know that
∞
1 X
= wn |w| < 1.
1−w
n=2
1
whenever |w| < 1. If next we set w = , it follows that
1−z
2 ∞ ∞
X (−1)n (n + 1)
z−1 1 X 1
= 2 = (n + 1) n =
z (1 − z) (z − 1)n
1
1− 1−z
n=0 n=0
1
whenever
< 1, i.e. when 1 < |z − 1|. On dividing through by (z − 1)2 we see that
1 − z
∞
X (−1)n (n + 1) X (−1)n (n − 1) ∞
1
= =
z2 (z − 1)n+2 (z − 1)n
n=0 n=2
104
(In the last equality we again used the fact that (−1)n = (−1)n+2 .)
Solution:
z − 2
Whenever < 1 (i.e. when |z − 2| < 2), it follows that
2
!
1 1 1 1 1
= =
2 1 + z−2
z 2
2 1 − −(z−2)
2
∞ n ∞
(−1)n
X 1 − (z − 2) X
= = (z − 2)n .
2 2 2n+1
n=0 n=0
From Theorem 2 in §71 we now conclude that for all |z − 2| < 2 we have
∞
!
d X (−1)n
1 d 1
= − =− (z − 2)n
z2 dz z dz 2n+1
n=0
∞ ∞
X (−1)n+1 X (−1)n
= n (z − 2)n−1 = (n + 1) (z − 2)n .
2n+1 2n+2
n=1 n=0
(In the last sum we made use of the fact that (−1)n = (−1)n+2 .)
Exercise 5, §72, p. 219
Prove that if cos z
when z 6= ±π/2
z − (π/2)2
2
f (z) = 1
− when z = ±π/2,
π
then f is an entire function.
Solution:
cos z 2 π 2
Since cos z is entire it is easy to see that is analytic (differentiable) whenever z −
π 2 2 6= 0.
z2 − 2
To see that cos z π
2 if z 6= ± 2
z2
f (z) = − (π/2)
1
− if z = ± π2
π
is in fact analytic on all of C, we therefore only need to check that it is differentiable at z = ± π2 . To do
this we expand cos z as a power series at z = π2 (i.e. in powers of z − π2 ). Recall that
∞
X w2n+1
sin w = (−1)n
(2n + 1)!
n=0
105 MAT3705/1
along a contour interior to the circle of convergence from w = 1 to w = z to obtain the representation
∞
X (−1)n+1
Logz = (z − 1)n (|z − 1| < 1) .
n
n=1
106
Solution:
From example 1 in §64 we know that
∞
1 X
= (−1)n (w − 1)n (|w − 1| < 1) .
w
n=0
1
Now by (6) in §31, Logz is an anti–derivative of on the region (|z| > 0, −π < Arg (z) < π). In particular
z
the disc |w − 1| < 1 is entirely contained in this region. If therefore z is in this disc and we integrate from
1 to z along a contour inside this disc, it follows from the Theorem 1 of §71 that
Z z
1
Logz = Logw|z1 = dw
1 w
Z zX ∞
= (−1)n (w − 1)n dw
1 n=0
∞ n+1 z
X (w − 1)
= (−1)n
n+1
n=0 1
∞ n+1
X (z − 1)
= (−1)n
n+1
n=0
∞
X (−1)n+1
= (z − 1)n (|z − 1| < 1) .
n
n=0
Solution:
Since Logz is analytic on all of (0 < |z| < ∞, −π < Arg(z) < π), it easily follows that Log(z) / (z − 1) is
analytic on this entire region except where z − 1 = 0, i.e. where z = 1. To see that
Logz when z 6= 1
f (z) = z−1
1 when z = 1,
is analytic on this entire region, we therefore only need to check that it is differentiable at z = 1. Now by
exercise 6 above we may write f (z) as a power series whenever 0 < |z − 1| < 1. We get
∞ ∞
!
n+1 n
1 n (z − 1) n (z − 1)
X X
f (z) = (−1) = (−1) .
z−1 n+1 n+1
n=0 n=0
1
Therefore f 0 (1) exists and equals − .
2
Exercise 8, §72, p. 220
Prove that if f is analytic at z0 and f (z0 ) = f 0 (z0 ) = · · · = f (m) (z0 ) = 0, then the function g defined by
the equations
f (z)
when z 6= z0 ,
(z − z0 )m+1
g (x) =
f (m+1) (z0 )
when z = z0
(m + 1)!
is analytic at z0 .
Solution:
Since f is analytic at z0 we can find some R > 0 so that f is analytic for all z with |z − z0 | < R. We show
that g is differentiable throughout |z − z0 | < R.
f (z)
For any z 6= z0 with |z − z0 | < R, g (z) is of the form g (z) = where f is analytic at each such
(z − z0 )m+1
z. Clearly g (z) is then also analytic (differentiable) at each such z.
To show that g is also differentiable at z0 we first look at the Taylor series of f at z0 . By Taylor’s theorem
∞
X
f (z) = cn (z − z0 )n
n=0
where
f (n) (z0 )
cn = .
n!
Since by assumption f (z0 ) = f 0 (z0 ) = · · · = f (m) (z0 ) = 0, we therefore get
and hence
f m+1 (z0 )
g (z0 + h) − g (z0 ) 1 f (z0 + h)
= −
h h (z0 + h − z0 )m+1 (m + 1)!
1
= [(cm+1 + cm+2 h + · · · ) − cm+1 ]
h
= cm+2 + cm+3 h + cm+4 h2 + · · ·
→ cm+2 as h → 0.
inside some circle |z − z0 | = R. Use Theorem 2 in §71, regarding term by term differentiation of such a
series, and mathematical induction to show that
∞
X (n + k)!
f (n)
(z) = an+k (z − z0 )k (n = 0, 1, 2, . . .)
k!
k=0
when |z − z0 | < R. Then, by setting z = z0 , show that the coefficients an (n = 0, 1, 2, . . .) are the coeffi-
cients in the Taylor series for f about z0 . Thus give and alternative proof of Theorem 1 in §72.
Solution:
Suppose that for some R > 0,
∞
X
f (z) = ak (z − z0 )k |z − z0 | < R.
k=0
(k + 1!)
Since k + 1 = , this can also be written as
k!
∞
X (k + 1)!
0
f (z) = ak+1 (z − z0 )k |z − z0 | < R.
k!
k=0
for all |z − z0 | < R. On applying Theorem 2 of §71 to f (m) above it then follows that
∞
d X (k + m)!
f (m+1)
(z) = ak+m (z − z0 )k
dz k!
k=0
∞
X (k + m)!
= kak+m (z − z0 )k−1
k!
k=1
∞
X (k + m)!
= ak+m (z − z0 )k−1
(k − 1)!
k=1
∞
X (k + (1 + m)!)
= ak+(1+m) (z − z0 )k
k!
k=0
109 MAT3705/1
for any n ∈ N. If now we evaluate these series at z0 , only the k = 0 term remains, whence
n!
f (n) (z0 ) = an = n!an n ∈ N.
0!
f (n) (z0 )
(By convention 0! = 1.) Therefore an = for each n ∈ N, that is the coefficients in the series
n!
∞
X
expansion f (z) = ak (z − z0 )k are precisely the Taylor–coefficients of f at z0 .
k=0
Solution:
1
Observe that csc z = with
sin z
∞
X (−1)n 2n+1 1 1
sin z = z = z − z3 + z5 − · · · .
(2n + 1)! 3! 5!
n=0
Now if 0 < |z| < π, then sin z is non-zero and so in this case we may write
1 1
csc z = = 1 3 1 5 1 7 .
sin z z− 3! z + 5! z − 7! z + ···
···
1 1 1 1
Therefore csc z = + z + 2 − 5! z 3 + · · · if 0 < |z| < π.
z 3! (3!)
110
Solution:
∞ ∞
X 1 n X
1 n
Recall that ez
−1 = z −1 = z z
n! z . The zeros of e − 1 are at all points z where e = 1, i.e.
n!
n=0 n=1
z = i2nπ (n ∈ Z). Thus if 0 < |z| < 2π, then ez − 1 is non-zero and so in this case we may write
1 1 1
= P∞ 1 n = 1 2 1 3 1 4 .
ez −1 n=1 n! z z+ 2! z + 3! z + 4! z + ···
By means of long division
1 1 1 1 3
z − 2 + 12 z − 720 z + ···
1 2 1 3 1 4
z+ 2! z + 3! z + 4! z + ··· 1
1 1 2 1 3 1 4
1+ 2! z + 3! z + 4! z + 5! z + · · ·
− 12 z − 61 z 2 − 24
1 3 1 4
z − 120 z − ···
1 1 2 1 3 1 4
− 2 z − 4 z − 12 z − 48 z − · · ·
1 2 1 3 1 4
12 z + 24 z + 80 z + · · ·
1 2 1 3 1 4
12 z + 24 z + 72 z + · · ·
1 4
− 720 z − ···
1 1 1 1 1 3
Thus =− + z− z + · · · if 0 < |z| < 2π.
ez−1 z 2 12 720
Exercise 5, §73, p. 225
Use the expansion
1 1 1 1 7
2
= 3− · + z + ··· (0 < |z| < π)
z sinh z z 6 z 360
which is an immediate consequeence of (8) in §73 and then use the method illustrated in example 4 §68,
to show that Z
dz πi
2
=−
C z sinh z 3
where C is the positively oriented unit circle |z| = 1.
Solution:
Observe that
1 1 11 7
− = + z + ···
z 2 sinh z
z 3 6 z 360
for all 0 < |z| < π. Since the positively oriented circle C : |z| = 1 falls within this region, it follows from
Theorem 1 of §71 that
Z Z Z Z
1 1 1 1 7
2
dz = 3
dz − dz + zdz + · · · .
C z sinh z C z 6 Cz 360 C
It then follows from example 4 of §68 that
Z
1 1
2
dz = 0 − 2πi + 0 + 0 + · · ·
C z sinh z 6
πi
= − .
3
111 MAT3705/1
Point out why this representation is valid in the indicated disk and why E2n+1 = 0 (n = 0, 1, 2, . . .). Then
show that
E0 = 1, E2 = −1, E4 = 5, and E6 = −61.
Solution:
1
Observe that the singularities of sechz = cosh z are at all points z where cosh = 0, i.e. where z =
π
2 + nπ i (n ∈ Z) (see the Theorem in §38). However none of these points are contained in the disk
|z| < π2 and hence cosh
1
z is therefore analytic on this disk. By Taylor’s theorem the Maclaurin series
expansion must then exist in this disk. Now from Example 6 of section 64 we know that
∞
X 1 2n 1 1
cosh z = z = 1 + z2 + z4 + · · · (|z| < ∞) .
(2n)! 2! 4!
n=0
1
This series has only even powers of z. Performing long division to extract the Maclaurin series of cosh z
from this one will therefore also yield a series with only even powers, that is the coefficients (and hence also
1
E2n+1 ) of all the odd powers in the Maclaurin expansion of cosh z must be zero. We proceed to compute
the first four non–zero terms in this expansion by means of long division.
1 − 21 z 2 + 24
5 4 61 6
z − 720 z + ···
1 2 1 4 1 6
1+ 2! z + 4! z + 6! z + ··· 1
1 + 2!1 z 2 + 24
1 4 1 6
z + 720 z + ···
− 12 z 2 − 24
1 4 1 6
z − 720 z − ···
1 2 1 4 1 6
− 2 z − 4 z − 48 z − · · ·
5 4 7 6
24 z +
360 z
5 4 5 6
24 z +
48 z
61 6
− 720 z − ···
61 6
− 720 z − · · ·
··
1 1 2 5 4 61 6
z + · · · |z| < π2 . The first four non–zero coefficients are therefore
Clearly cosh z = 1 − 2 z + 24 z − 720
E0 = 1, E2 1 E4 5 E6
2! = − 2 , 4! = 24 , 6! = −61
720 . This implies that E0 = 1, E2 = −1, E4 = 5, E6 = −61.
112
113 MAT3705/1
CHAPTER 6
in this region. To integrate f along C, we can now try to integrate term by term, and apply Cauchy’s
integral formulas to each of these termwise integrals to get
Z ∞ Z ∞ Z
X
n
X 1
f (z)dz = an (z − a) dz + bn dz
C C C (z − a)n
n=0 n=1
= 2πib1
1
− a)n dz = 0 when n 6= −1 with
R R
(Here we used the fact that C (z C (z−a) dz = 2πi.)
1
So computing the integral, boils down to computing the coefficient b1 of the (z−a) term in the Laurent
expansion of f . We call this coefficient b1 the residue of f at a. Taking this argument to its logical con-
clusion presents us with Cauchy’s residue theorem, which in its simplest form boils down to the statement
that
Z
f (z)dz = 2πi · {sum of the residues inside C}
C
It is very important to note that ONLY the residues of singularities inside C contribute to the value of
R
the integral C f (z)dz. Thus although f may have many other singularities besides those inside C, those
R
other singularities make no contribution whatsoever to the value of C f (z)dz. Now of course for us to
be able to make good use of this theorem, we need to be skilled at computing these residues. For this
reason a large part of chapter 6 is devoted to classifying singularities of complex functions, and developing
techniques for computing the residues at these singularities.
114
Learning outcomes
After studying this chapter you should be able to:
• understand the close relationship between poles and zeros as established in Theorems 1 and 2 of
§83.
Sections to be studied
You need to study all the sections in chapter 6, that, is sections 74 to 84. You need to know the statements
and applications of the important theorems in this chapter, but not the proofs.
115 MAT3705/1
Below you will find solutions to some of the exercises in §74 − 84 of chapter 6. Be sure to attempt these
yourself before you work through the solutions.
z − sin z
(c) ;
z
cot z
(d) ;
z4
sinh z
(e) .
z 4 (1− z2)
Solution:
1
(a) The point z = 0 is the only singularity of f (z) = inside C where C is the positively oriented
z + z2
1
circle |z| = . By (2) of §71 the residue is
2
Z
1 1
Res (f ) = dz
z=0 2πi C z + z 2
Z
1 1/ (1 + z)
= dz
2πi C z
1
= (Cauchy Integration Formulae)
1 + z z=0
= 1.
∞
X (−1)n 1
(b) Since cos w = w2n for all |w| < ∞, we may set w = to get
2n! z
n=0
∞
!
1 X (−1)n 1
z cos = z
z (2n)! z 2n
n=0
11 1 1 1 1
= z− + 3
− + ···
2! z 4! z 6! z 5
Clearly
1 1
Res z cos =− .
z=0 z 2
116
∞
X (−1)n 2n+1
(c) Since sin z = z (see §64) it follows that
(2n + 1)!
n=0
∞
!
1 1 X (−1)n 2n+1
(z − sin z) = z− z
z z (2n + 1)!
n=0
1 2 1 1
= z − z4 + z6 − . . .
3! 5! 7!
1
Therefore the residue (coefficient of the term) is
z
1
Res (z − sin z) = 0.
z=0 z
cos z
(d) We use long division to compute the first few terms of the Laurent series of cot z = . Since
sin z
∞
(−1)n
X
2n 1 2 1 4
cos z = z = 1 − z + z − ...
(2n)! 2 24
n=0
and
∞
(−1)n 2n+1
X 1 3 1 5
sin z = z = z− z + z − ... ,
(2n + 1)! 6 120
n=0
it follows that
1
z − 13 z − 451 3
z − ...
z − 61 z 3 + 1 5
120 z − ... 1 2 1 4
1 − 2 z + 24 z − . . .
1 − 61 z 2 + 120
1 4
z − ...
− 13 z 2 +1 4
30 z − ...
− 13 z 2 −1 4
18 z − ...
1 4
− 45 z + ...
1 4
− 45 z + ...
............
Therefore
cos z 1 1 1
cot z = = − z − z3 − . . .
sin z z 3 45
in which case
cot z 1 1 1 1 1
= 5− − ...
z4 z 3 z 3 45 z
But then
cot z 1
Res =− .
z=0 z4 45
∞
X n
z2 z2
In addition if |z| < 1 then 1/ 1 − = . Therefore if 0 < |z| < 1, then
n=0
sinh z 1 2 4
1 3 1 5
= 1 + z + z + ... z + z + z + ...
z 4 (1 − z 2 ) z4 3! 5!
1 1 1 1 1
= + 1+ + 1+ + z + ...
z3 3! z 3! 5!
Consequently
sinh z 1 7
Res =1+ = .
z=0 z (1 − z 2 )
4 3! 6
exp (−z)
(a) ;
z2
exp (−z)
(b) ;
(z − 1)2
1
(c) z 2 exp
z
z+1
(d) .
z 2 − 2z
Solution:
In each case let C be the positively oriented circle |z| = 3.
exp(−z)
(a) The only singularity of z2
is at z = 0.
Therefore Z
exp (−z) exp (−z)
dz = 2πi Res .
C z2 z=0 z2
∞
exp(−z) 1 P 1 n 1 1 1 1 exp(−z)
Since z2
= z2 n! (−z) = z2
− z + 2! − 3! z + · · · it follows that Res z2
= −1, and
n=0 z=0
hence that Z
exp (−z)
dz = −2πi.
C z2
exp(−z)
(b) The only singularity of (z−1)2
is at z = 1. To find the residue we compute the Laurent series at
z = 1. Since
∞
!
1 1 X 1
e −z
= e−(z−1) = (− (z − 1))n
e e n!
n=0
118
it follows that
exp (−z) 1 1 1 3
= · 1 − (z − 1) + (z − 1) − . . .
(z − 1)2 (z − 1)2 e 2!
1 1 1 1 1
= 2 − + − ...
e (z − 1) e (z − 1) 2e
Therefore Z
exp (−z) 1
2 dz = 2πi − e .
C (z − 1)
∞
!
1 1 n
2 1 X
z exp = z2
z n! z
n=0
1 11 1 1
= z2 + z + + + + ...
2! 3! z 4! z 2
Therefore
2 1 1 1
Res z exp = = .
z=0 z 3! 6
1
Since in addition 0 is the only singularity of z 2 exp , it follows that
z
Z
2 1 1 πi
z exp dz = 2πi = .
C z 6 3
z+1
(d) has singularities at both 0 and 2 both of which lie inside the circle C. Therefore
z 2 − 2z
Z
z+1 h i
2
dz = 2πi Res (f ) + Res (f ) .
C z − 2z z=0 z=2
1 3 3
= − − − z − ...
2z 4 8
At z = 2 we get
z+1 (z − 2) + 3 1
=
2 (z − 2) 1 + z−2
z (z − 2) 2
2 3 !
1 3 z−2 z−2 z−2
= + 1− + − + ...
2 2 (z − 2) 2 2 2
3 1 1 1
= − + (z − 2) − . . .
2 (z − 2) 4 8
Therefore Z
z+1 1 3
dz = 2πi − + = 2πi.
C z 2 − 2z 2 2
(a) By using the Maclaurin series for ez and referring to Theorem 1 in §71, which justifies the term by
term integration that is to be used, write the above integral as
∞ Z
X 1 n 1
z exp dz.
n! C z
n=0
(b) Apply the theorem in §76 to evaluate the integrals appearing in part (a) to arrive at the desired
result.
Solution:
Let C be the positively oriented circle |z| = 1. By example 1 of §65
∞
X 1 n
ez = z |z| < ∞.
n!
n=0
Z ∞
X 1 n
1
= e z z dz
C n!
n=0
∞ Z
X 1 1
= z n e z dz.
n! C
n=0
120
Therefore
1 1 n−2 1 1 1 1 1
z n e z = z n + z n−1 + z + ... + + + + ...
2! n! (n + 1)! z (n + 2)! z 2
Clearly 1
1
Res z n e z = .
z=0 (n + 1)!
1
Therefore since z = 0 is the only singular point of z n e z , it follows from the theorem in §76 that
Z
1
1
2πi
z n e z dz = 2πi Res z n e z = .
C z=0 (n + 1)!
Consequently
Z ∞ Z ∞
1 X 1 1 X 1
ez+ z dz = z n e z dz = 2πi .
C n! C n! (n + 1)!
n=0 n=0
z2
(b) ;
1+z
sin z
(c)
z
cos z
(d) ;
z
1
(e) .
(2 − z)3
Solution:
1
(a) The only singularity of z exp zis z = 0. For z = 0 we have
∞ ∞
1 1 n X 1 1−n
1 X
z ez = z = z |z| > 0.
n! z n!
n=0 n=0
1
Therefore the principal part of z e z is
∞ ∞
X 1 1−n X 1
z = z −k |z| > 0
n! (k + 1)!
n=2 k=1
z2
(b) has a singularity at z = −1. At this point the Laurent expansion is
1+z
z2 1 1
= ((z + 1) − 1)2 = − 2 + (z + 1)
1+z (z + 1) (z + 1)
1
and the principal part of the function is . Therefore z = −1 is a simple pole.
z+1
sin z
(c) has a singular point at z = 0. At this point
z
∞
!
sin z 1 X 1 1 2 1
= (−1)n z 2n+1 =1− z + z4 − . . .
z z (2n + 1)! 3! 5!
n=0
sin z
Therefore z = 0 is a removable singular point and the principal part of is 0.
z
cos z
(d) has a singular point at z = 0. At this point
z
∞
cos z 1 X (−1)2n 2n 1 1 1
= z = − z + z3 − . . .
z z (2n)! z 2! 4!
n=0
cos z 1
Therefore the principal part of is and hence z = 0 is a simple pole.
z z
1 1
(e) 3 = − has a singular point at z = 2. The principal part of the function at 2 is
(2 − z) (z − 2)3
1
− and z = 2 is therefore a pole of order 3.
(z − 2)3
Suggestion: As pointed out in §62, there is a Taylor series for f (z) about z0 since f is analytic there. Start
each part of this exercise by writing out a few terms of that series.
Solution:
Since f is analytic at z0 , Taylor’s Theorem assures us that for some R > 0
∞
X
f (z) = an (z − z0 )n |z − z0 | < R
n=0
where
f (n) (z0 )
an = n = 0, 1, 2, 3, . . .
n!
122
f (z) = a1 (z − z0 ) + a2 (z − z0 )2 + . . . |z − z0 | < R.
In this case
f (z)
= a1 + a2 (z − z0 ) + a3 (z − z0 )2 + . . . 0 < |z − z0 | < R.
(z − z0 )
Solution:
(a) z 2 + 9 can be written as z 2 + 9 = (z − 3i)(z + 3i) and has a zero of order 1 at 3i and also a zero of
order 1 at −3i. At these values z + 1 6= 0.
z+1
Therefore 2 has simple poles at z = ±3i. The residue is
z +9
z+1 z + 1 z + 1 3i + 1 1 1
Res ( 2 )= d = = = − i
z=3i z + 9 2 2z z=3i 6i 2 6
dz (z + 9)
z=3i
and
z+1 z + 1 z + 1 −3i + 1 1 1
Res ( 2 )= d
= = = + i
z=3i z + 9 (z 2 + 9) 2z z=−3i −6i 2 6
dz z=−3i
1 3
1 1
(c) z3 3
is nonzero at z = − whereas 2 z + has a zero of order 3 there. Therefore z = − is a
2 2 2
3
z
pole of order 3 of . Since
2z + 1
3 z 3
z 2
= 3
2z + 1 z + 12
z 3 1
with analytic at z = − , the residue is given by
2 2
3
1 d2 z 3
z
Res =
z=− 21 2z + 1 2! dz 2 2 z=− 1
2
1 6z
=
2! 8 z=− 1
2
3
= − .
16
exp (z)
(d) exp (z) is never zero whereas z 2 + π 2 = (z − iπ) (z + iπ) has simple zeros at ±iπ. Thus has
(z 2 + π 2 )
simple poles at ±iπ. The residues at these points are:
exp z exp z
Res =
z=iπ z 2 + π 2 d 2 2
dz (z + π )
z=iπ
exp (iπ)
=
2 (iπ)
i
=
2π
and
exp z exp (−iπ)
Res =
z=−iπ z2 + π2 2 (−iπ)
i
= − .
2π
Log z π + 2i
(b) Res 2 = .
z=i (z 2 + 1) 8
124
Solution:
1
(a) Here z 4 is defined by
1 1
z 4 = exp log z
4
where
log z = ln |z| + i arg (z)
1
for any z with |z| > 0, 0 < arg (z) < 2π. In particular this branch of z 4 exists and is analytic at
1
z = −1. With log z and z 4 as above we have
log (−1) = ln 1 + iπ = iπ
1 1
= √ + i√ .
2 2
1
By contrast z + 1 has a simple zero at z = −1 and hence z 4 / (z + 1) must have a simple pole there.
Thus
1 1
z4 z4 1 1 i
Res = d = (−1) 4 = √ + i √ .
z=−1 z + 1
dz (z + 1) 2 2
z=−1
for any z with |z| > 0 and −π < Arg(z) < π. In particular z = i lies in this region (note that
π
i = 1ei 2 in polar form) and hence Log z is analytic at z = i with
π π
Log (i) = ln 1 + i =i .
2 2
2 2
By contrast z 2 + 1 = (z − i)2 (z + i)2 has a double zero at z = i in which case Log z/ z 2 + 1 must
then have a double pole at z = i. Therefore since
(2i)2 1
− 2 (2i) i π2
i
=
(2i)4
π + 2i
= .
8
(a) |z| = 2;
(b) |z + 2| = 3.
Solution:
1
3
clearly has a simple pole at z = −4 and a pole of order 3 at z = 0. Therefore by theorem §80
z (z + 4)
1/z 3
1 1
Res 3
= d =−
z (z + 4) 64
dz (z + 4)
z=−4
z=−4
and
1 d2
1 1
Res 3
=
z=0 z (z + 4) 2! dz 2 (z + 4) z=0
1
= 3
(z + 4) z=0
1
= .
64
(a) Next observe that |0| < 2 and |−4| > 2. Of the two singular points only 0 is in the interior of
|z| = 2. Thus when integrating around this circle we get
Z
dz 1 πi
3 (z + 1)
= 2πi Res 3 (z + 1)
= .
C z z=0 z 32
(b) Since |0 + 2| < 3 and |−4 + 2| < 3, both singular points lie inside |z + 2| = 3. Thus for this circle
Z
dz 1 1
3
= 2πi Res + Res = 0.
C z (z + 1) z=0 z 3 (z + 1) z=−1 z 3 (z + 1)
126
Solution:
cosh πz
has singular points where z z 2 + 1 = 0, i.e. where z = 0, ±i. Since cosh πz is non–zero at
2
z (z + 1)
these points and each point is a simple zero of z z 2 + 1 = z (z − i) (z + i) , these points are simple poles
cosh πz
of . Thus
[z (z 2 + 1)]
cosh πz cosh πz
Res =
2
z=i z (z + 1) d 2
dz (z (z + 1)) z=i
cosh πz
=
3z 2 + 1 z=i
cosh πi
=
3 (i)2 + 1
cos π
=
−2
1
=
2
Similarly
cosh πz cosh (−iπ) 1
Res 2
= 2 =
z=−i z (z + 1) 3 (−i) + 1 2
and
cosh πz cosh 0
Res = = 1.
z=0 2
z (z + 1) 3 (0)2 + 1
Since all these points are inside |z| = 2, we have
Z
cosh πz 1 1
2
dz = 2πi + 1 +
C z (z + 1) 2 2
= 4πi.
Solution:
Since
d
sinh z = cosh (±iπ) = cos (±π) = −1 6= 0
dz z=±iπ
127 MAT3705/1
exp (zt)
sinh z clearly has simple zeros at ±iπ. Noting that exp (zt) is never zero, it follows that must
sinh z
have simple poles at z = ±iπ. Therefore
exp (zt) exp (zt)
Res + Res
z=πi sinh z z=−πi sinh z
= −2 cos πt.
Solution:
π
(a) The zeros of cos z are at zn = + nπ(n ∈ Z). Observe that since
2
d π
+ nπ = (−1)n 6= 0
(cos z) = sin
dz z= π +nπ 2
2
π z
for each n ∈ Z, cos z has simple zeros at zn = + nπ. Thus z sec z = has simple poles at each
2 cos z
of the zn ’s. Therefore
z
Res (z sec z) = d
z=zn
dz cos z
z=zn
π
2 + nπ
=
sin π2 + nπ
π
= (−1)n + nπ n ∈ Z.
2
sinh z π
(b) tanh z = has singular points where cosh z = 0, i.e. where zn = i (2n + 1) (n ∈ Z). At
cosh z 2
these points
d π
cosh z = sinh i (2n + 1) 6= 0.
dz z=i(2n+1) π 2
2
sinh z
Therefore cosh z has a simple zero and tanh z = a simple pole at each zn = i (2n + 1) π2
cosh z
(n ∈ Z) .
128
Solution:
sin z π
(a) tan z = has singular points where cos z = 0, i.e. where zn = (2n + 1) (n ∈ Z) . Now
cos z 2
π π
the inequality (2n + 1) < 2 only holds if n = 0, −1 and hence only ± lie inside the circle
2 2
|z| = 2. Since
d π
cos z = sin ± = ±1,
dz z=± π 2
2
π sin z
the points z = ± are simple zeros of cos z and hence simple poles of tan z = . Thus
2 cos z
sin z sin π2
Resπ tan z = d = =1
sin π2
dz cos z
z= 2 π
z= 2
and similarly
sin − π2
Res tan z = = 1.
z=− π2 sin − π2
Hence
Z " #
tan zdz = 2πi Resπ tan z+ Resπ tan z
C z= 2 z=− 2
= 4πi.
1 π
(b) , has singular points where sinh 2z = 0, i.e. where zn = i n (n ∈ Z) . Since
sinh 2z 2
d
sinh 2z = 2 cosh (i n π)
dz z=i n π
2
= 2 (−1)n n∈Z
129 MAT3705/1
1 π
it is clear that these points are simple zeros of sinh z and simple poles of . At each zn = i n
sinh z 2
1 1 (−1)n
Resπ = = .
z=i n 2
sinh 2z 2 cosh (i n π) 2
Of all these singular points only the ones corresponding to n = −1, 0, 1 lie inside the circle |z| = 2.
Consequently
1
Z " #
dz X 1
= 2πi Res
C sinh 2z z=zn sinh 2z
n=−1
1 1 1
= 2πi − + −
2 2 2
= −πi.
N
" #
(−1)n
Z
dz 1 X
= 2πi +2 .
CN z 2 sin z 6 n2 π 2
n=1
Then, using the fact that the value of this integral tends to zero as N tends to infinity (Exercise 8, §47),
point out how it follows that
∞
X (−1)n+1 π2
= .
n2 12
n=1
Solution:
1
has singular points where z = 0 and where sin z = 0, i.e. where zn = nπ (n ∈ Z) . Since
z 2 sin z
d
sin z = cos (nπ) 6= 0,
dz zn =nπ
sin z has simple zeros at each zn with z 2 having a zero of order 2 at z = 0. Thus z 2 sin z has a zero of
1
order 3 at z = 0 and simple zeros at the other zn ’s, whence 2 of course has a triple pole at 0 and
z sin z
∞
X (−1) n
simple poles at the other zn ’s. Recall that sin z = z 2n+1 . Hence
(2n + 1)!
n=0
1 1
= 1 2 1 4
z 2 sin z
z3 1− 3! z + 5! z + ...
130
1
= .
6
For each zn = nπ (n = ±1, ±2, . . .) we have
1 1/z 2
Res =
z=nπ z 2 sin z d
dz (sin z) z=nπ
1/z 2
=
cos z z=nπ
(−1)n
= .
(nπ)2
N
" #
1 X 1 1
= 2πi Res + Res + Res
z=0 z 2 sin z z=zn z 2 sin z z=z−n z 2 sin z
n=1
N
" #
1 X (−1)n
= 2πi +2 .
6 (nπ)2 n=1
This simplifies to
∞
X (−1)n+1 π2
= .
n2 12
n=1
131 MAT3705/1
Solution:
1 2 √
has singular points where z 2 − 1 = −3, i.e. where z 2 = 1 ± i 3. Now in polar form
(z 2 − 1)2 + 3
√ iπ √ iπ
1 + i 3 = 2e 3 and 1 − i 3 = 2e− 3 .
1
The square roots of these numbers are precisely the singular points of and correspond to
(z 2 − 1) + 3
√ !
√ √
r
iπ 3 1 3 1
2e 6 = 2 +i = + i√
2 2 2 2
! √
√ √
r
i 7π 3 1 3 1
2e 6 = 2 − −i =− − i√
2 2 2 2
√ ! r
√ iπ √ 3 1 3 1
2e− 6 = 2 −i = − i√
2 2 2 2
√ !
√ √
r
i5π 3 1 3 1
2e 6 = 2 − +i =− + i√ .
2 2 2 2
Of these points only r r
3 1 3 1
z0 = + i √ and − z 0 = − + i√
2 2 2 2
lie inside the given rectangle. All these points are simple poles and so
1 1
Res 2 = h i
z=z0 (z 2 − 1) + 3 d 2 2
dz (z − 1) + 3
z=z0
1
= 2
4 (z − 1) z z=z0
1
= q 2 q
3
4 2 + i √12 −1 3
2 + i √12
√
3 1
= − √ −i √
24 2 8 2
132
and
1 1
Res =
z=−z 0 (z 2 − 1)2 + 3 2
4 (z − 1) z z=−z 0
1
= q 2 q
4 − 2 + i 2 − 1 − 32 + i √12
3 √1
√
3 1
= √ −i √ .
24 2 8 2
Consequently
Z
dz 1 1
= 2πi Res + Res
C (z 2 − 1)2 + 3 (z 2 − 1)2 + 3
z=z0 z=−z 0 (z 2 − 1)2 + 3
i
= 2πi − √
4 2
π
= √ .
2 2
Exercise 8, §83, p. 254
Consider the function
1
f (z) = ,
[q (z)]2
where q is analytic at z0 , q (z0 ) = 0, and q 0 (z0 ) 6= 0. Show that z0 is a pole of order m = 2 of the function
f, with residue
q 00 (z0 )
B0 = .
[q 0 (z0 )]3
Suggestion: Note that z0 is a zero of order m = 1 of the function q, so that equation (3), §83, holds. Then
write
φ (z) 1
f (z) = 2 where φ (z) = .
(z − z0 ) [g (z)]2
The desired form of the residue B0 = φ0 (z0 ) can be obtained by showing that q 0 (z0 ) = g (z0 ) and
q 00 (z0 ) = 2g 0 (z0 ) .
Solution:
Since by hypothesis q has a simple zero at z0 , it follows from (3) of §83 that
2g 0 (z0 )
= − .
[g (z0 )]3
133 MAT3705/1
1
(b) f (z) = .
(z + z 2 )2
Solution:
d
(a) Since sin z = cos 0 = 1, sin z has a simple zero at z = 0. We may therefore directly apply
dz z=0
the results of Exercise 7 above to the function
1
f (z) = csc2 z =
sin2 z
to see that
d2
sin z
dz 2
Res csc2 z = − 3
z=0 d
dz sin z
z=0
sin z
=
cos3 z z=0
= 0.
d
(b) Observe that z + z2
has a zero at 0 whereas 2
z + z = 1. As in (a) we may therefore directly
dz z=0
apply the results of Exercise 7 above to
1
f (z) =
(z + z 2 )2
to conclude that
2
d 2
1 − dz 2 z + z
Res = 3
(z + z 2 )2
z=0
d
2
dz (z + z )
z=0
2
= −
[1 + 2z]3 z=0
= −2.
134
135 MAT3705/1
.
.
CHAPTER 7
Applications of Residues
Introduction
The importance of Cauchy’s residue theorem (discussed in chapter 6) extends far beyond complex analysis
itself. Residue theory has many important applications, including the computation of real integrals and
finding the roots of polynomials. Using a few clever tricks many classes of real integrals can be written
in a form where we can use residue theory to compute the integral. A number of these classes play a
very important role in physics, and are typically extremely difficult to compute by other means. The
largest part of chapter 7 is spent describing precisely how one goes about using residue theory to compute
these real integrals. A second important consequence of residue theory is Rouché’s theorem. By the
Fundamental theorem of Algebra we know that a polynomial of degree n, has n roots. Unfortunately, for
polynomials of degree 4 or higher, there is no simple formula for computing the roots. So knowing that the
roots exist is very nice, but where are they? Using Rouché’s theorem, we can approximate complicated
polynomials by simpler ones, and in this way at least get some idea of where to search for the roots of
such complicated polynomials.
Learning outcomes
After studying this chapter you should be able to:
R∞ 1
• apply residue theory to evaluate an improper integral using the complex roots, i.e. 0 x6 +1
dx;
• apply residue
Z theory to evaluate convergent improper integrals from Fourier analysis, i.e. integrals
∞
of the form f (x) cos ax dx;
−∞
• evaluate improper integrals of the form where Jordan’s lemma, i.e. where
Z
lim f (z)eiaz dz = 0, a > 0
R→∞ CR
is valid;
• use Rouché’s theorem to determine the number of roots, counting multiplicities, of a polynomial in
an annulus or inside a circle.
Sections to be studied
Study only sections 85 to 90 and 92 to 94 in chapter 7. Note that some important exercises at the end
of section 91 in fact refer to section 90 and are covered here in the study guide, although the material of
section 91 does not form part of the course.
137 MAT3705/1
Below you will find solutions to some of the exercises in chapter 7. Be sure to attempt these yourself
before you work through the solutions.
Solution:
1
The function f (z) = has singularities at ±i. We integrate f around the contour [−R, R] ∪
1 + z2
CR (R > 0) where
For R > 1 the singular point z = i is in the interior of [−R, R] ∪ CR . So by Cauchy’s residue theorem
Z R Z
f (t) dt + f (z) dz = 2πi Res f (z) .
−R CR z=i
and hence Z R Z
f (t) dt = π − f (z) dz.
−R CR
2 1 1
For z = R eit 2
we have z + 1 ≥ R − 1 whence 2
≤
2
. Then
1+z R −1
Z
1 1
2
dz ≤
πR
CR 1 + z R2 − 1
R
where πR is the length of CR . From this it is clear that CR f (z) dz → 0 as R → ∞. Consequently
Z ∞ Z R
1 1
dt = lim dt
−∞ 1 + t2 R→∞ −R 1 + t2
Z
1
= π − lim dz
R→∞ CR 1 + z 2
= π.
d 1
(Note: We didn’t actually need residues to compute this integral. Since arctan t = , it follows
dt 1 + t2
that
Z ∞ Z b
1 1
2
dt = lim dt
0 1 + t b→∞ 0 1 + t2
= lim arctan b − arctan 0
b→∞
π
= .)
2
Solution:
1
Consider f (z) = . Now f has poles of order 2 at z = ±i. We integrate f over the same contour
+ 1)2
(z 2
used in exercise (1) above. As before z = i is in the interior of [−R, R] ∪ CR when R > 1, in which case
Cauchy’s residue theorem tells us that
Z R Z
f (t) dt + f (z) dz = 2πi Res f (z) .
−R CR z=i
Now since
1 1 1
2 = ,
2
(z + 1) (z − i)2 (z + i)2
it follows from the theorem in §80 that
1 d 1
Res =
z=i (z 2 + 1)2 dz (z + i)2 z=i
−2
=
(z + i)3 z=i
i
= −
4
and hence that
Z R Z
1 i 1
2 2 dt = 2πi − 4 − 2 2 dz
−R (t + 1) CR (z + 1)
Z
π 1
= − 2 dz.
2 2
CR (z + 1)
as R → ∞. But then Z ∞ Z
1 π 1 π
2 2 dt = 2 − R→∞
lim 2 dz =
2
.
−∞ (t + 1) CR (z 2 + 1)
Since the integrand is an even function this yields
Z ∞
1 ∞
Z
1 1 π
2 2 dt = 2 2 2 dt = 4 .
0 (t + 1) −∞ (t + 1)
Solution:
1
Consider f (z) = . Such an f has singular points where z 4 = −1 = eiπ , i.e. where
1 + z4
iπ i3π i5π i7π
z0 = e 4 , z 1 = e 4 , z2 = e 4 , z3 = e 4 .
(See section 8.) Of these only z0 and z1 are in the upper half–plane. Integrate f over the same contour as
in exercise (1) above. Then for R > 1 it follows from Cauchy’s Residue Theorem that
Z R Z
1 1 1 1
dt + dz = 2πi Res + Res .
−R 1 + t4 CR 1 + z4 z=z0 1 + z 4 z=z1 1 + z 4
Therefore Z
1 1
dz ≤
πR → 0
CR 1 + z4 R4 − 1
as R → ∞. Consequently
Z ∞ Z
1 1 1 1 1 1
4
dt = 2πi − √ − i √ + √ −i √ − lim dz
−∞ 1 + t 4 2 4 2 4 2 4 2 R→∞ CR 1 + z 4
π
= √ ,
2
whence Z ∞ Z ∞
1 1 1 π
dt = dt = √
0 1 + t4 2 −∞ 1 + t4 2 2
1
since is an even function.
1 + t4
140
Solution: h 2 i
We integrate z 2 / z 2 + 9 z 2 + 4
around the positively oriented contour CR ∪ [−R, R] where
±3i, z = ±2i.
Of these only 3i and 2i will lie inside CR ∪ [−R, R] for R large enough. For R large it then follows from
the residue theorem that
Z Z R
(∗) f (z) dz + f (z) dz = 2πi Res (f (z)) + Res (f (z)) .
CR −R z=3i z=2i
2
Since z 2 + 9 z 2 + 4 = (z − 3i) (z + 3i) (z − 2i)2 (z + 2i)2 it follows that f (z) has a simple pole at
z2
Res (f ) = lim (z − 3i)
z=3i z→3i (z − 3i) (z + 3i) (z 2 + 4)2
z2
= lim
z→3i (z + 3i) (z 2 + 4)2
−9
=
6i (−9 + 4)2
3i
=
50
and by the Theorem in §80
z2
1 d
Res (f ) =
z=2i 1! dz (z 2 + 9) (z + 2i)2 z=2i
h i h i
z 2 + 9 (z + 2i)2 2z − 2z (z + 2i)2 + 2 z 2 + 9 (z + 2i) z 2
=
(z 2 + 9)2 (z + 2i)4
z=2i
13i
= −
200
h i
(Note that f (z) = (z − 2i)−2 g (z) where g (z) = z 2 / z 2 + 9 (z + 2i)2 .) Now for z = R eit we have
z + 9 z 2 + 42 = z 2 + 9 z 2 + 42
2
2
≥ |z|2 − 9 |z|2 − 4
2
= R2 − 9 R2 − 4 ,
141 MAT3705/1
π
R3 R3 π
Z
≤ dt = →0
0 (R2 − 9) (R2 − 4)2 (R2 − 9) (R2 − 4)2
as R → ∞. Applying all this to (∗) , it follows that
Z ∞ Z R
x2 x2
Z
2 2
dx = lim f (z) dz + 2 dx
−∞ (x + 9) (x + 4) R→∞ ΓR −R (x2 + 9) (x2 + 4)
3i 13i
= 2πi −
50 200
π
= .
100
Since the integrand is an even function it then follows that
Z ∞
x2 1 ∞ x2
Z
dx = dx
0 (x2 + 9) (x2 + 4)2 2 −∞ (x2 + 9) (x2 + 4)
π
= .
200
Now let CR be as in exercise (1) above. Then for R > 0 large enough it will follow from Cauchy’s Residue
theorem that
Z R Z
xdx 1
2 2
+ 2 2
dz = 2πi Res f (z) + Res f (z)
−R (x + 1) (x + 2x + 2) CR (z + 1) (z + 2z + 2) z=i
z=−1+i
(∗) i
= 2πi
10
π
=− .
5
Next note that for z on CR (i.e. |z| = R)
z + 1 ≥ |z|2 − 1 = R2 − 1
2
z + 2z + 2 ≥ z 2 + 2z − 2 ≥ |z|2 − 2 |z| − 2 = R2 − 2R − 2
2
and hence
z R
(z 2 + 1) (z 2 + 2z + 2) ≤ (R2 − 1) (R2 − 2R − 2)
as R → ∞. (Recall that the length of CR is πR.) If in (∗) above we let R → ∞, we then get
Z ∞
xdx π
2 2
=− .
−∞ (x + 1) (x + 2x + 2) 5
1 iπ
Now by Theorem 2 in §83 3
has a simple pole at e 3 with
1+z
1 1 z
Res = iπ = iπ
1 + z3 3z 2 z=e 3 3z 3 z=e 3
iπ
z=e 3
1 iπ
= − e3
3
iπ
(Here we used the fact that e 3 is a 3rd root of −1.)
Moreover
2π
Z Z3
1 1
dz ≤ Rdt
1+z
3 3 i3t
1+R e
CR 0
2π
Z3
R
≤ dt
R3 −1
0
2Rπ
= → 0 as R → ∞
3 (R3 − 1)
and
Z ZR
1 1 i2π
dz = i2π −e 3 dt
1 + z3 1 + [(R − t) e 3 ]3
LR 0
ZR
i2π 1
= −e 3 dt.
1 + (R − t)3
0
Setting x = R − t yields
Z ZR
1 i2π 1
3
dz = −e 3 dx.
1+z 1 + x3
LR 0
1−e 3 dx + dz = .
1 + x3 1+z 3 3
0 CR
1−e 3
3
dx =
1+x 3
0
144
or equivalently that
Z∞ iπ
! −iπ
1 π 2ie 3 e 3
dx = i2π × −iπ
1 + x3 3 e 3 −1 e 3
0
π 2i
= iπ −iπ
3 e3 −e 3
π 1
=
3 sin π3
π π
= cosec .
3 3
Exercise 10, §86, p.265
Let m and n be integers, where 0 ≤ m < n. Follow the steps below to derive the integration formula
Z ∞
x2m
π 2m + 1
dx = csc π .
0 x2n + 1 2n 2n
(a) Show that the zeros of the polynomial z 2n + 1 lying above the real axis are
(2k + 1) π
ck = exp i (k = 0, 1, 2, . . . , n − 1)
2n
z 2m 1
Res 2n
= − ei(2k+1)α (k = 0, 1, 2, . . . , n − 1) .
z=ck z +1 2n
(c) Use the final result in part (b) to complete the derivation of the integration formula.
145 MAT3705/1
Solution:
(a) The zeros of the polynomial z 2n + 1 are at all z for which z 2n = −1 = eiπ . Thus by the formula in
§10 the roots are
π 2kπ (2k + 1) π
zk = exp i + = exp i
2n 2n 2n
2k + 1
0≤ π ≤ π,
2n
We conclude that
(2k + 1) π
zk = exp i 0≤k ≤n−1
2n
(b) Let zk = exp i (2k+1)π
2n (0 ≤ k ≤ n − 1) be as in (a). By Theorem 2 in §83 each of these is a simple
z 2m
pole of (z 2n +1)
with
z 2m z 2m
Res 2n =
z=zk z +1 2nz 2n−1 z=zk
1
= (zk )2(m−n)+1
2n
1 h i(2k+1)π/2n i2(m−n)+1
= e
2n
1 −i(2n−1)π i(2k+1)α
= e e
2n
1
= − ei(2k+1)α
2n
where α = 2m+1
2n π. (Since 2n − 1 is odd, e−i(2n−1)π = −1). Given that
n−1
X 1 − zn
zk = z 6= 1
1−z
k=0
146
z 2m
(c) We now integrate f (z) = z 2n +1
over the curve [−R, R] ∪ CR where CR is as in exercise 1 above. For
R large enough (R > 1) the curve [−R, R] ∪ CR will contain all singular points of f (z) in the upper
half–plane. Therefore with zk as in (a), we have by (a) and (b) that
Z R
x2m z 2m
Z
2n + 1
dx + 2n + 1
dz
−R x CR z
n−1
X z 2m
= 2πi Res
z=zk z 2n + 1
k=0
π
=
n sin α
π (2m + 1) π
= cosec
n 2n
(where α = (2m + 1) π/2n). Now for z on CR (i.e. |z| = R) we have that z 2n + 1 ≥ |z|2n − 1 =
Therefore
z 2m R2m ÷ R2n
Z
2n
dz ≤ πR
CR z +1 R2n − 1 ÷ R2n
R−[2(n−m)−1] π
=
1 − R−2n
0
→ =0
1
as R → ∞. (To see that this is the case, recall that 2 (n − m) − 1 > 0 since n > m.) Letting R → ∞
it follows that
Z ∞
z 2m z 2m
Z
π (2m + 1) π
2n
dz = cosec − lim 2n
dz
−∞ z +1 n 2n R→∞ CR z +1
π (2m + 1) π
= cosec .
n 2n
147 MAT3705/1
Solution:
We integrate
eiz
f (z) =
(z 2 + a2 ) (z 2 + b2 )
around the contour CR ∪ [−R, R] where as before CR is the circle sector
CR : z (t) = R eit 0 ≤ t ≤ π.
The function f (z) has singularities where z 2 + a2 z 2 + b2 = 0, i.e. where z = ±ai, z = ±bi. Of these
only ai and bi lie inside CR ∪ [−R, R] for large R and so by the residue theorem
Z R Z
f (ẋ) dx + f (z) dz = 2πi Res (f ) + Res (f )
−R CR z=ai z=bi
for large R. The function f has simple poles at ai and bi and therefore by Theorem 2 in §83
eiz / z 2 + b2
Res (f ) = d
|z=ai
2 2
dz (z + a )
z=ai
eiz / z 2 + b2
= |z=ai
2z
e−a
=
2ai (−a2 + b2 )
e−a i
2a (a2 − b2 )
eiz / z 2 + a2
Res (f ) = d
|z=bi
2 2
dz (z + b )
z=bi
eiz / z 2 + a2
= |z=bi
2z
e−b
=
2bi (−b2 + a2 )
e−b i
= − .
2b (a2 − b2 )
148
e = e−R sin t
iz
≤ e0 = 1 for t ∈ [0, π] ,
|z|2 − a2 |z|2 − b2 = R2 − a2 R2 − b2 .
2
z + a2 z 2 + b2 ≥
Therefore
Z π iz(t)
eiz
Z e
iR eit dt
2 2 2 2
dz ≤ 2 2 2 2
(z + a ) (z + b ) 0 (R − a ) (R − b )
CR
Rπ
≤ →0
(R2 − a2 ) (R2 − b2 )
e−a i e−b i
= 2πi −
2a (a2 − b2 ) 2b (a2 − b2 )
π 1 1
= − .
(a2 − b2 ) beb aea
Finally recall that Re eix = cos x. If therefore we compare the real parts of the previous equality, we get
that Z ∞
cos x π 1 1
2 2 2 2
dx = 2 − .
−∞ (x + a ) (x + b ) (a − b2 ) beb aea
Solution:
eiaz
We integrate g (z) = (z 2 +b2 )2
around the contour CR ∪ [−R; R] where CR is parametrised by z (t) =
R eit (0 ≤ t ≤ π). The function g has singularities at ±bi. Of these only bi will lie inside CR ∪ [−R, R]
for large R > 0. Therefore by the residue theorem
Z R
eiax eiaz
Z
2 dx + 2 dz = 2πi z=bi
Res g (z) .
−R (x2 + b2 ) CR (z 2 + b2 )
Since
eiaz
1
g (z) = ,
(z − bi)2 (z + bi)2
z = bi is a double pole. By the theorem in §80
eiaz
d
Res g (z) = 2
z=bi dz (z + bi) z=bi
(z + bi)2 iaeiaz − 2 (z + bi) eiaz
=
(z + bi)4
z=bi
(ab + 1) e−ab
= .
i4b3
Therefore by equating real parts
R
π (ab + 1) e−ab eiaz
Z Z
cos ax
dx = − Re dz.
−R (x2 + b2 )2 2b3 CR (z 2 + b2 )2
2 2
Next observe that since eiaz = e−ay ≤ 1 whenever y ≥ 0, and z 2 + b2 ≥ |z|2 − b2 , it follows that
1
|g (z)| ≤
(R2 − b2 )2
whenever z ∈ CR . Consequently
Z Z
πR
Re g (z) dz ≤ g (z) dz ≤ →0
CR CR (R2 − b2 )2
Solution:
iaz
We integrate g (z) = (zze4 +4) over the contour [−R, R] ∪ CR where CR is the semicircle CR : z (t) =
R eit (0 ≤ t ≤ π) . The function g has singular points where z 4 = −4 = eiπ , i.e. at
√ π π
zk = 2ei( 4 +k 2 ) 0 ≤ k ≤ 3.
√
Of these z0 and z1 are in the upper half–plane. Therefore for R big enough R > 2 , Cauchy’s residue
theorem ensures that
Z R
xeiax zeiaz
Z
4
dx + 4
dz = 2πi Res g (z) + Res g (z) .
−R x + 4 CR z + 4 z=z0 z=z1
|z|4 − 4 = R4 − 4 whence
z R
z 4 + 4 ≤ R4 − 4 → 0 as R → ∞.
zeiaz
Z
dz → 0 as R → ∞,
CR z4 + 4
i.e.
zeiaz
Z
Im dz →0 as R → ∞.
CR z4 + 4
Now by Theorem 2 in §83
π
zeiaz z0 eiaz0 z02 eiaz0 2ei 2 eia(1+i) −i −a ia
Res 4 = = = = e e .
z=z0 z + 4 4z03 4z04 −16 8
Similarly
3π
zeiaz z 2 eiaz1 2ei 2 eia(−1+i) −2i −a −ia i
Res 4 = 1 4 = = e e = . e−a e−ia
z=z1 z + 1 4z1 −16 −16 8
Therefore
zeiaz zeiaz i −a ia −ia
1 −a
Res + Res = − e e − e = e sin a.
z=z0 z 4 + 4 z=z1 z 4 + 4 8 4
(Recall that 2i sin a = eia − e−ia .) Then
∞
zeiaz
Z Z
x sin ax 1 −a
dx = Im 2πi e sin a − lim dz
−∞ x4 + 1 4 R→∞ CR z 4 + 1
π −a
= e sin a.
2
151 MAT3705/1
Solution:
zeiz
We integrate g (z) = (z 2 +1)(z 2 +4) over the same contour as above. The function g has singular points
where z = ±i, z = ±2i. Of these z = i, z = 2i are in the upper half–plane. So for R > 2 Cauchy’s
residue theorem ensures that
Z R
xeix zeiz
Z
2 2
dx + 2 2
dz
−R (x + 1) (x + 4) CR (z + 1) (z + 4)
zeiz zeiz
= 2πi Res 2 + Res 2 .
z=i (z + 1) (z 2 + 4) z=2i (z + 1) (z 2 + 4)
zeiz
Z
lim dz = 0.
R→∞ CR (z 2 + 1) (z 2 + 4)
Solution:
iz
Consider the integral of the function g (z) = (z(z+1)e
2 +4z+5) on [−R, R] ∪ CR where as before CR is the upper
half of the circle |z| = R. This function has singular points where z 2 + 4z + 5 = 0, i.e. where z =
1
√
2 −4 ± 16 − 20 = −2 ± i. Since z = −2 + i is in the upper half–plane, it follows from Cauchy’s residue
theorem that
Z R
(x + 1) eix (z + 1) eiz (z + 1) eiz
Z
2
dx + 2
dz = 2πi Res ,
−R x + 4x + 5 CR z + 4z + 5 z=−2+i z 2 + 4z + 5
By Theorem 2 in §83
(z + 1) eiz (z + 1) eiz
Res =
z=−2+i z 2 + 4z + 5 2z + 4 z=−2+i
1 + i −1−2i
= e
2
e−1
= ((cos 2 + sin 2) + i (cos 2 − sin 2)) .
2
(z+1) √
Moreover since f (z) = (z 2 +4z+5)
is analytic outside the circle |z| = |−2 ± i| = 5, and since
(z + 1) R+1
z 2 + 4z + 5 ≤ R2 − 4R − 5
whenever |z| = R with the right hand side tending to 0 as R → ∞, it follows from the Theorem in §88
that
(z + 1) eiz
Z
lim dz = 0.
R→∞ CR z 2 + 4z + 5
(a) By integrating the function exp iz 2 around the positively oriented boundary of the sector 0 ≤ r ≤
y _i π
Re4
CR
x
O R
and
Z R Z R Z
1 2 2
sin x2 dx = √ e−r dr − Im eiz dz,
0 2 0 CR
π
where CR is the arc z = R eiθ
0≤θ≤ 4 .
(b) Show that the value of the integral along the arc CR in part (a) tends to zero as R tends to infinity
by obtaining the inequality
Z Z π
iz 2
R 2 −R2 sin θ
e dz ≤
e dθ
2
CR 0
(c) Use the results in parts (a) and (b), together with the known integration formula
Z ∞ √
2 π
e−x dx = ,
0 2
Solution:
2
Now since eiz is analytic on all of C, it follows from the Cauchy-Goursat theorem that
Z Z Z
2 2 2
0 = eiz dz + eiz dz + eiz dz
CR LR [0,R]
Z Z 0 Z R
2
iz 2 1 1 i 1
√ +i √1
2 t2
= e dz + e − √ −i √ dt + eit dt
2 2
CR −R 2 2 0
Z Z R Z R
2 1 1 2 2
= eiz dz − √ + i √ e−s ds + eit dt.
CR 2 2 0 0
and similarly
Z R Z R Z
1 2 2
sin t2 dt = √ e−s ds − Im eiz dz.
0 2 0 CR
iz 2 (t) = iR2 ei2t = i R2 cos 2t + iR2 sin 2t = −R2 sin 2t + iR2 cos 2t.
Therefore
Z Z π
4 2
iz 2 iz (t) it
e dz = e iR e dt
CR 0
Z π
4 2
≤ e−R sin 2t Rdt.
0
(c) Finally Z
iz 2
R π π
e dz ≤ 2
= →0
CR 2 2R 4R
155 MAT3705/1
2
eiz dz = 0, and hence we have form §60 that
R
as R → ∞. But then lim
R→∞ CR
Z Z
2 2
lim Re eiz dz = 0 = lim Im eiz dz .
R→∞ CR R→∞ CR
Moreover √
Z ∞
2 π
e−s ds = .
0 2
R∞ R∞ 2
√1 e−s ds = 1
pπ
If therefore we let R → ∞ in (a) we surely have 0 cos t2 dt = 2 0 2 2 and similarly
R∞ 1
pπ
0 sin t2 dt = 2 2.
Then, with the aid of the trigonometric identity 1 − cos (2x) = 2 sin2 x, point out how it follows that
∞
sin2 x
Z
π
2
dx = .
0 x 2
Solution:
(eiaz −eibz )
Consider the function g (z) = z2
. We integrate this function across the contour [−R, −ρ] ∪ Cρ ∪
[ρ, R] ∪ CR where
(see figure 108 in §89). The only singular point of the integrand is at z = 0 and so by the residue theorem
−ρ R
eiax − eibx eiaz − eibz eiax − eibx eiaz − eibz
Z Z Z Z
dx + dz + dx + dx = 0.
−R x2 Cρ z2 ρ x2 CR z2
It follows that
Z R iax Z −ρ iax
eiaz − eibz eiaz − eibz e − eibx e − eibx
Z Z
− dz − dz = dx + dx
Cρ z2 CR z2 ρ x2 −R x2
(set x = −s in the 2nd integral)
Z R iax Z R −ias
e − eibx e − eibs
= dx + ds
ρ x2 ρ s2
Z R
cos (ax) − cos (bx)
= 2 dx.
ρ x2
1 iax
e + e−iax , etc.)
(Set x = s and recall that cos ax =
2
156
Notice that
∞ n ∞ n
!
1 iaz ibz
1 X i n
X i n
e − e = (az) − (bz)
z2 z2 n! n!
n=0 n=0
∞
!
1 X(i)n n
= (a − bn ) z n
z2 n!
n=1
a2 − b2 i a3 − b3
i (a − b)
= − − z + ··· .
z 2! 3!
1
eiaz − eibz has a simple pole at 0 with residue i (a − b). Therefore by the theorem in §89
Therefore z2
eiaz − eibz
Z
lim dz = (−i (a − b)) πi = π (a − b) .
ρ→0 Cρ z2
Now for |z| = R , z12 = R12 → 0 as R → ∞. Therefore since z12 is analytic whenever |z| > 0, it follows
eiaz eibz
Z Z
2
dz and 2
dz
CR z CR z
Solution:
(− 12 ) log z
We integrate the function g (z) = e (z 2 +1) over the contour [−R, −ρ] ∪ Cρ ∪ [ρ, R] ∪ CR where Cρ and CR
are as in Fig.109 and where
π 3π
log (z) = ln |z| + i arg (z) − < arg (z) < .
2 2
We again integrate this function over [−R, −ρ]∪Cρ ∪[ρ, R]∪CR where Cρ and CR are as before. Remember
that log (z) = ln |z| + iπ on [−R, −ρ] and log (z) = ln |z| on [ρ, R]. In addition g has a singular point at
z = i which will lie inside the contour if 0 < ρ < 1 < R. It will then follow from the residue theorem that
By Theorem 2 in §83
exp − 21 log z
Res g (z) =
2z
z=i
z=i
exp − 12 i π2
ln 1 +
=
2i
1 − iπ
= e 4.
2i
1
= √ (1 − i)
2 2i
Thus
1 1
2πi Res g (z) = 2πi · √ (1 − i) = π √ (1 − i)
z=i 2 2i 2
Now for |z| = R
e(− 12 ) log z z − 12
= 2
2
z +1 z + 1
1
≤ √ .
R(R2 − 1)
1
1
since e(− 2 ) log z = exp(R(− 12 log z)) = exp(− 12 log |z|) = |z|− 2 .
Therefore Z
≤ √ 1
g (x) dz πR → 0 as R → ∞,
CR
R(R2 − 1)
(we can show this by L’Hospital’s theorem)
158
that is Z
lim g (z) dz = 0.
R→∞ CR
Similarly Z " 1
#
ρ− 2
g (x) dz ≤ 2 πρ.
ρ −1
Cρ
R
Since ρ ln ρ → 0 as ρ → 0 , this means that Cρ g (z) dz → 0 as ρ → 0 and hence that
Z
lim g (z) dz = 0.
ρ→0 Cρ
If therefore we let R → ∞ and ρ → 0 in the above integration formula it follows that (*)
1
∞
x− 2
Z
1
π √ (1 − i) = (1 − i) dx.
2 0 x2 + 1
i.e.
Z ∞
x− 2
1
π √12 (1 − i)
dx =
0 x2 + 1 1−i
π
= √
2
and
π 3π
− < arg (z) < .
2 2
The singular point z = i is the only singular point inside [−R, −ρ] ∪ Cρ ∪ [ρ, R] ∪ CR . In addition
log z = ln |z| + iπ on [−R, −ρ] and log z = ln |z| on [ρ, R]. Therefore by the residue theorem
−ρ
(ln |x| + iπ)2 R
(ln x)2
Z Z Z Z
2πi Res g (z) = dx + g (z) dz + dx + g (x) dz
z=i −R x2 + 1 Cρ ρ x2 + 1 CR
(Set s = −x in the 1st integral)
(ln x)2
Z R Z R Z R Z Z
2 1 ln x
= 2 2
dx − π 2
dx + 2πi 2
dx + g (z) dz + g (z) dz.
ρ x +1 ρ x +1 ρ (x + 1) Cρ CR
159 MAT3705/1
and hence Z 2
(log z)2 ln R + 3π
2
dz ≤ πR → 0
CR z 2 + 1 (R2 − 1)
ln R (ln R)2
as R → ∞. (To see this note that we can use L’Hospital’s theorem to show that both R and R tend
to 0 as R → ∞.) Similarly
Z
2
2
(log z) ln ρ + 3π
2
dz ≤ πρ → 0 as ρ → 0.
Cρ z 2 + 1 (ρ2 − 1)
(Again this can be seen by using L’Hospital’s theorem to show that both ρ ln ρ and ρ (ln ρ)2 tend to 0 as
ρ → 0.) Thus Z Z
lim g (z) dz = 0 = lim g (z) dz.
R→∞ CR ρ→0 Cρ
π2 ∞
(ln x)2 ∞
Z Z
π ln x
2πi( i) = 2 2
dx − π 2 ( ) + 2πi dx.
8 0 1+x 2 0 1 + x2
i.e.
π3 π3 ∞
(ln x)2 ∞
Z Z
ln x
− + =2 dx + 2πi dx
4 2 0 1 + x2 0 1 + x2
so that
π3 ∞
(ln x)2 ∞
Z Z
ln x
=2 dx + 2πi dx.
4 0 1 + x2 0 1 + x2
Comparing real and imaginary parts now reveals that
∞
(ln x)2 π3 ∞
Z Z
ln x
dx = and dx = 0.
0 1 + x2 8 0 1 + x2
160
Solution:
By (1), (2) and (3) of §92, the given integral becomes
Z 2π Z
dθ 1 1
= 1 1
dz
0 5 + 4 sin θ C 5 + 4 2i z − z iz
Z
dz
= 2 + 5iz − 2
C 2z
where C is the positively oriented circle |z| =
1. The integrand of
the complex integral has singular points
q
2
where 2z 2 + 5iz − 2 = 0, that is where z = 14 −5i ± (5i) + 16 = 14 (−5i ± 3i) = −2i, − 12 i. Of these
points only z = − 12 i lies inside C. Therefore by the residue theorem
!
Z 2π
dθ 1
= 2πi Res 2
.
0 5 + 4 sin θ z=− 12 i 2z + 5iz − 2
Solution:
By the remark at the end of §92 before example 2 we have that
1 n 1 1 n 1
cos nθ = z + n sin nθ = z − n
2 z 2i z
√ 1 1 1
integrand has singularities at 0, ± 2 and ± √ . Of all these singularities only 0, √ and − √ lie inside
2 2 2
1
the circle |z| = 1. The integrand has simple poles at ± √ and a pole of order 5 at 0. We may therefore
2
1
use Theorem 2 in §83 to find the residues at ± √ . By this result
2
" 2 #
z 6 + 1 /z 5
Res g (z) = d 4 2
z= √1
2 dz (2z − 5z + 2) √1
z=
2
" 2 #
z 6 + 1 /z 5
=
8z 3 − 10z
z= √1
2
27
= −
16
and similarly
" 2 #
z 6 + 1 /z 5
Res g (z) =
z=− √1 8z 3 − 10z
2 z=− √1
2
27
= − .
16
To find the residue at 0 we have by the theorem in §80 that
" 2 #
1 d4 z6 + 1
Res g (z) = − .
z=0 4! dz 4 (2z 4 − 5z 2 + 2)
z=0
However to avoid the pain of differentiating 4 times we rather look at the Laurent series of
2
z6 + 1
7 1 1
5 4 2
= z + 2z + 5 .
z (2z − 5z + 2) z (2z − 1) (z 2 − 2)
2
whence 2 ∞
!
z6 + 1
7 1 X 1 1
= z + 2z + 5 2k+1 − z 2k .
z 5 (2z 4 − 5z 2 + 2) z 3 2k+1
k=0
162
1
In the resulting expansion the term will be
z
1 1 1 63 1
8− z4 = .
z5 3 8 24 z
Therefore
63
Res g (z) = .
z=0 24
2
2π
cos2 3θ z6 + 1
Z Z
i
dθ = dz
0 5 − 4 cos 2θ 4 |z|=1 z 5 (2z 4 − 5z 2 + 2)
" #
i
= 2πi Res g (z) + Res g (z) + Res g (z)
4 z= √1 z=− √1 z=0
2 2
π 27 27 63
= − − − +
2 16 16 24
3π
= .
8
Alternative: In the above integral the difficulty posed by computing the residue at a pole of order 5 can
be avoided altogether if we use partial fractions. Observe that
1 1
= .
z 5 (2z 4 2
− 5z + 2) z 5 (z 2 − 2) 2 z − √1 z+ √1
2 2
1 A B C D E 1 F G Hz + I
= + 2 + 3 + 4 + 5 + + + 2
z 5 (2z 4 2
− 5z + 2) z z z z z 2 1
z − √2 1
z + √2 (z − 2)
21 5 1 8 1
B = D = I = 0, A = , C= , E= , F =G=− , H= .
8 4 2 3 24
2
On multiplying both sides by z 6 + 1 and integrating over |z| = 1 Cauchy’s integration formulae then
163 MAT3705/1
ensure that
2
z6 + 1
Z
dz
|z|=1 z 5 (2z 4 − 6z 2 + z)
2 2 2
z6 + 1 z6 + 1 z6 + 1
Z Z Z
21 5 1
= dz + dz + dz
8 |z|=1 z 4 |z|=1 z3 2 |z|=1 z5
2 2 2
z6 + 1 z6 + 1 z6 + 1
Z Z Z
4 4 1
− dz − dz + dz
3 |z|=1 z − √12 3 |z|=1 z + √1 24 |z|=1 (z 2 − 2)
2
5 2πi d2 1 2πi d4
21 6
2 6
2
= (2πi × 1) + z +1 + z + 1
8 4 z! dz 2 z=0 2 4! dz 4 z=0
!2 !2
1 6
6
4 4 1
− 2πi √ + 1 − 2πi −√ +1 +0
3 2 3 2
3πi
= − .
2
(The last integral is zero since there the integrand is analytic inside and on |z| = 1.) Therefore
2
2π
cos2 3θ z6 + 1
Z Z
i
dθ = 5 4 2
dz
0 5 − 4 cos 2θ 4 |z|=1 z (2z − 5z + z)
i 3πi 3π
= − = .
4 2 8
Solution:
By (1), (2) and (3) of §92, we see that
1 1
z2 +
Z π Z
cos 2θ 2 z 2 1
dθ = 1 dz
−π 1 − 2a cos θ + a2 C 1 − 2a 2 z + z1 + a2 iz
z4
i +1
Z
= dz
C 2z 2 (az 2 − (a2 + 1) z + a)
or a. None of these zeros are also zeros of z 4 + 1 and hence all are singular points of the integrand. Of
these points only 0 and a are inside C. Therefore the residue theorem tells us that
Z π
cos 2θ h i
2
dθ = 2πi Res g (z) + Res g (z)
−π 1 − 2a cos θ + a z=0 z=a
i(z 4 +1)
where g (z) = [2z 2 (az 2 −(a2 +1)z+a)]
. By the Theorem in §80
i z4 + 1
d
Res g (z) =
z=0 2 2
dz 2 (az − (a + 1) z + a)
z=0
3
1 i4z az − a + 1 z + a − i z 4 + 1 2az − a2 + 1
2 2
=
2 (az 2 − (a2 + 1) z + a)2
z=0
2
i a +1
= .
2 a2
Thus
" #
π
i a2 + 1 i a4 + 1
Z
cos 2θ
dθ = 2πi +
−π 1 − 2a cos θ + a2 2 a2 2 a2 (a2 − 1)
a4 − 1 + a4 + 1
= −π
a2 (a2 − 1)
2a2 π
= .
(1 − a2 )
π π
a2 π
Z Z
cos 2θ 1 cos 2θ
2
dθ = dθ = .
0 1 − 2a cos θ + a 2 −π 1 − 2a cos θ + a2 (1 − a2 )
165 MAT3705/1
Solution:
Again by (1), (2) and (3) of §92,
2n
π
1 2n 1 i (−1)n+1 z2 − 1
Z Z Z
2n 1
sin θdθ = z− dz = dz
−π C 2i z iz 22n c z 2n+1
2n
(z 2 −1)
where C is the positively oriented circle |z| = 1 and where n = 1, 2, . . .. The function z 2n+1
has a
singular point at z = 0, and hence by the residue theorem
2n 2n
z2 − 1 z2 − 1
Z
dz = 2πi Res .
C z 2n+1 z=0 z 2n+1
2n
2n
(−1)2n−k z 2k − (2n + 1)
P
=
k=0 k
1 2n
for all 0 < |z|. The coefficient of the term (corresponding to k = n) is (−1)n . That is
z n
2n
z2 − 1 2n (2n)!
Res (−1)n = (−1)n .
z=0 z 2n+1 n (n!)2
Therefore
π
i (−1)n+1
Z
2n (2n)! n
sin θdθ = 2πi (−1)
−π 22n (n!)2
(2n)!
= 2π .
22n (n!)2
Solution:
(a) f (z) = z 2 has a double zero at z = 0 and no poles inside |z| = 1. Hence here
∆C arg z 2 = 2π (2 − 0) = 4π.
(b) f (z) = f (z) = 1/z 2 has no zeros inside |z| = 1 and a double pole at z = 0 which lie inside |z| = 1.
Thus
∆C arg 1/z 2 = 2π (0 − 2) = −4π.
7
(2z−1)7
(c) f (z) = (2z−1)
z 3 has a triple pole at z = 0 and a zero of order 7 at z = 1
2 . Hence ∆ C arg z 3 =
2π (7 − 3) = 8π.
Solution:
Since f is analytic inside and on C, it clearly has no poles inside or on C. So by the theorem in §93
1
Z= ∆C arg f (z)
2π
where Z is the number of zeros (counting multiplicities) of f inside C. Now from the sketch it is clear
that as z traverses once around the contour C, f (z) will effectively circle the origin 3 times. That is
1
Therefore counting multiplicities f has 2π (6π) = 3 zeros inside C.
than 2π when a point z makes one cycle around C and recalling that ∆C arg f (z) is an integral multiple
of 2π, point out why the winding number of Γ with respect to the origin w = 0 must be zero.
Solution:
Let f and Γ be as stated and suppose that the ray w = reiα (0 ≤ r < ∞) does not intersect Γ. Since 0
is by assumption outside Γ and also on the ray, the entire ray must then be outside Γ (or else it would
have crossed Γ). Now let f (z) traverse around Γ starting from say f (z0 ) where arg f (z0 ) = ϕ0 and
α − 2π < ϕ0 < α. Although arg f (z) may increase or decrease as f (z) traverses Γ, the fact that f (z)
can never be on the ray w = reiα (0 ≤ r < ∞) ensures that arg f (z) can never cross either θ = α − 2π
or θ = α. (To see this note that a point w 6= 0 lies on the ray if and only if arg w = α + 2kπ (k ∈ Z).)
Clearly then |arg f (z) − arg f (z0 )| < 2π for each f (z) on Γ, whence
Since ∆C arg f (z) must be an integer multiple of 2π, this inequality can only hold if in fact
∆C arg f (z) = 0.
Solution:
(a) Let f (z) = z 6 − 5z 4 and g (z) = z 3 − 2z. Now f (z) = z 4 z 2 − 5 has 6 zeros counting multiplicities,
but only 4 are inside the circle |z| = 1. On the circle |z| = 1 we have |f (z)| ≥ 5 |z|4 − |z|6 = 4 and
|g (z)| ≤ |z|3 + 2 |z| = 3. So by Rouché’s theorem f (z) + g (z) = z 6 − 5z 4 + z 3 − 2z also has 4 zeros
inside |z| = 1.
(b) Let f (z) = 9 for all z and g (z) = 2z 4 − 2z 3 + 2z 2 − 2z. Clearly f has no zeros anywhere. In addition
if |z| = 1 then
|g (z)| ≤ 2 |z|4 + 2 |z|3 + 2 |z|2 + 2 |z| = 8 < 9 = |f (z)| .
Thus by Rouché’s theorem f (z) + g (z) = 2z 4 − 2z 3 + 2z 2 − 2z + 9 also has no zeros inside |z| = 1.
2z 5 − 6z 2 + z + 1 = 0
Solution:
First let f (z) = −6z 2 and g (z) = 2z 5 + z + 1. Here f has 2 zeros inside |z| = 1. Now for |z| = 1 we have
Hence by Rouché’s theorem f (z) + g (z) = 2z 5 − 6z 2 + z + 1 also has 2 zeros inside |z| = 1. Next let
f˜ (z) = 2z 5 and g̃ (z) = −6z 2 + z + 1. The polynomial f (z) = 2z 5 has 5 zeros inside |z| = 2 (all at z = 0.)
For |z| = 2 we have
|g̃ (z)| ≤ 6 |z| + |z| + 1 = 27 and f (z) = 2 |z|5 = 64.
2 ˜
So by Rouché’s theorem f˜ (z) + g̃ (z) = 2z 5 − 6z 2 + z + 1 also has 5 zeros inside |z| = 2. So in the region
inside |z| = 2 but not inside |z| = 1 (i.e. the region 1 ≤ |z| < 2), the polynomial 2z 5 − 6z 2 + z + 1 must
have exactly 5 − 2 = 3 roots.
Extra Exercise §94
Write f (z) = z n and g (z) = a0 + a1 z + · · · + an−1 z n−1 and use Rouché’s theorem to prove that any
polynomial
P (z) = a0 + a1 z + · · · + an−1 z n−1 + an z n (an 6= 0) ,
where n ≥ 1, has precisely n zeros, counting multiplicities. Thus give an alternative proof of the funda-
mental theorem of algebra.
Suggestion: Note that one can let an be unity. Then show that |g (z)| < |f (z)| on the circle |z| = R,
where R is sufficiently large and, in particular, larger than
Solution:
Let f (z) = an z n and g (z) = a0 + a1 z + · · · + an−1 z n−1 where an 6= 0. The polynomial f (z) clearly has n
zeros at 0. For |z| = R we have
|g (z)| |a0 | + |a1 | |z| + · · · + |an−1 | |z|n−1
≤
|f (z)| |an | |z|n
|a0 | 1 |a1 | 1 |an−1 | 1
= n
+ n−1
+ ··· +
|an | R |an | R |an | R
→ 0 as R → ∞.
Therefore we can find R0 > 0 so that for any R ≥ R0 we will have that |g (z)| < |f (z)| whenever |z| = R.
Thus by Rouché’s theorem
will have precisely n zeros inside any circle |z| = R with R ≥ R0 . Therefore on all of C a0 + a1 z + a2 z 2 +
· · · + an z n has precisely n zeros.
(a) Point out why the denominator in the integrand of the integral defining Φ (t) is never zero on C.
This ensures the existence of the integral.
(b) Let t and t0 be any two points in the interval 0 ≤ t ≤ 1, and show that
f g0 − f 0g
Z
|t − t0 |
|Φ (t) − Φ (t0 )| = dz .
2π
C (f + tg) (f + t0 g)
at points on C, show that there is a positive constant A, which is independent of t and t0 , such that
|Φ (t) − Φ (t0 )| ≤ A |t − t0 | .
(c) By referring to equation (8), §93, state why the value of the function Φ is, for each value of t in
the interval 0 ≤ t ≤ 1, an integer representing the number of zeros of f (z) + tg (z) inside C. Then
conclude from the fact that Φ is continuous, as shown in part (b), that f (z) and f (z) + g (z) have
the same number of zeros, counting multiplicities, inside C.
Solution:
(a) By assumption |g (x)| < |f (z)| for all z on the contour C. Thus for any 0 ≤ t ≤ 1 this implies that
whenever z is a point on C.
for all z ∈ C and all 0 ≤ s ≤ 1. In particular for t and t0 it then follows that
for all z on C. Since f, g are analytic on C, each of f, g, f 0 and g 0 are then continuous on C. (This
|f g 0 −f 0 g|
follows from for example Theorem 1 of §57 and the discussion at the end of §19.) Since (|f |−|g|)2
is
then defined and continuous on C, it follows from (6) of §18 that we can find a constant M > 0 so
that
|f g 0 − f 0 g|
≤ M for all z on C.
(|f | − |g|)2
Comparing this with (∗∗) above, it then follows from (6) of §47 that
where L is the length of C and M is clearly independent of t and t0 . By (∗) we then have
ML
|Φ (t) − Φ (t0 )| ≤ |t − t0 | .
2π
2πε
Now let ε < 0 be given and set δ = ML . Then
ML
|t − t0 | < δ ⇒ |Φ (t) − Φ (t0 )| < δ = ε.
2π
(c) For each fixed t, f (z) + tg (z) is analytic inside and on C (i.e. it has no poles inside or on C), and
hence it then follows from (8) of §93 that
f 0 (z) + tg 0 (z)
Z
1
Φ (t) = dz
2πi C f (z) + tg (z)
is an integer representing the number of zeros of f (z) + tg (z) inside C. But then Φ (t) must be
constant on [0, 1] since if Φ was not constant the fact that all its values are integers means that
it would then have to admit of a “jump discontinuity” somewhere on [0, 1]. But there can be no
such jump since Φ is continuous. Therefore t → Φ (t) is constant on [0, 1] as claimed. In particular
Φ (0) = Φ (1) , or in other words f and f + g have the same number of zeros inside C (counting
multiplicities of course).