This action might not be possible to undo. Are you sure you want to continue?
c _ K. Houston 2003
1 Complex Functions
In this section we will deﬁne what we mean by a complex func
tion. We will then generalise the deﬁnitions of the exponential,
sine and cosine functions using complex power series. To deal
with complex power series we deﬁne the notions of conver
gent and absolutely convergent, and see how to use the ratio
test from real analysis to determine convergence and radius
of convergence for these complex series.
We start by deﬁning domains in the complex plane. This
requires the prelimary deﬁnition.
Deﬁnition 1.1
The εneighbourhood of a complex number z is the set of com
plex numbers ¦w ∈ C : [z −w[ < ε¦ where ε is positive number.
Thus the εneigbourhood of a point z is just the set of points
lying within the circle of radius ε centred at z. Note that it
doesn’t contain the circle.
Deﬁnition 1.2
A domain is a nonempty subset D of C such that for every
point in D there exists a εneighbourhood contained in D.
Examples 1.3
The following are domains.
(i) D = C. (Take c ∈ C. Then, any ε > 0 will do for an ε
neighbourhood of c.)
(ii) D = C¸¦0¦. (Take c ∈ D and let ε =
1
2
([c[). This gives a
εneighbourhood of c in D.)
(iii) D = ¦z : [z − a[ < R¦ for some R > 0. (Take c ∈ C and let
ε =
1
2
(R −[c −a[). This gives a εneighbourhood of c in D.)
Example 1.4
The set of real numbers R is not a domain. Consider any
real number, then any εneighbourhood must contain some
complex numbers, i.e. the εneighbourhood does not lie in the
real numbers.
We can now deﬁne the basic object of study.
1
Deﬁnition 1.5
Let D be a domain in C. A complex function, denoted f : D →
C, is a map which assigns to each z in D an element of C, this
value is denoted f(z).
Common Error 1.6
Note that f is the function and f(z) is the value of the function
at z. It is wrong to say f(z) is a function, but sometimes people
do.
Examples 1.7
(i) Let f(z) = z
2
for all z ∈ C.
(ii) Let f(z) = [z[ for all z ∈ C. Note that here we have a
complex function for which every value is real.
(iii) Let f(z) = 3z
4
− (5 − 2i)z
2
+ z − 7 for all z ∈ C. All complex
polynomials give complex functions.
(iv) Let f(z) = 1/z for all z ∈ C¸¦0¦. This function cannot be
extended to all of C.
Remark 1.8
Functions such as sin x for x real are not complex functions
since the real line in C is not a domain. Later we see how to
extend the concept of the sine so that it is complex function
on the whole of the complex plane.
Obviously, if f and g are complex functions, then f + g,
f − g, and fg are functions given by (f + g)(z) = f(z) + g(z),
(f − g)(z) = f(z) − g(z), and (fg)(z) = f(z)g(z), respectively.
We can also deﬁne (f/g)(z) = f(z)/g(z) provided that g(z) ,= 0
on D. Thus we can build up lots of new functions by these
elementary operations.
The aim of complex analysis
We wish to study complex functions. Can we deﬁne differenti
ation? Can we integrate? Which theorems from Real Analysis
can be extended to complex analysis? For example, is there a
version of the mean value theorem? Complex analysis is es
sentially the attempt to answer these questions. The theory
will be built upon real analysis but in many ways it is easier
than real analysis. For example if a complex function is dif
ferentiable (deﬁned later), then its derivative is also differen
tiable. This is not true for real functions. (Do you know an ex
ample of a differentiable real function with nondifferentiable
derivative?)
2
Real and imaginary parts of functions
We will often use z to denote a complex number and we will
have z = x + iy where x and y are both real. The value f(z)
is a complex number and so has a real and imaginary part.
We often use u to denote the real part and v to denote the
imaginary part. Note that u and v are functions of z.
We often write f(x + iy) = u(x, y) + iv(x, y). Note that u is
a function of two real variables, x and y. I.e. u : R
2
→ R.
Similarly for v.
Examples 1.9
(i) Let f(z) = z
2
. Then, f(x +iy) = (x +iy)
2
= x
2
−y
2
+2ixy. So,
u(x, y) = x
2
−y
2
and v(x, y) = xy.
(ii) Let f(z) = [z[. Then, f(x + iy) =
_
x
2
+ y
2
. So, u(x, y) =
_
x
2
+ y
2
and v(x, y) = 0.
Exercises 1.10
Find u and v for the following:
(i) f(z) = 1/z for z ∈ C¸¦0¦.
(ii) f(z) = z
3
.
Visualising complex functions
In Real Analysis we could draw the graph of a function. We
have an axis for the variable and an axis for the value, and so
we can draw the graph of the function on a piece of paper.
For complex functions we have a complex variable (that’s
two real variables) and the value (another two real variables),
so if we want to draw a graph we will need 2 + 2 = 4 real
variables, i.e. we will have to work in 4dimensional space.
Now obviously this is a bit tricky because we are used to 3
space dimensions and ﬁnd visualising 4 dimensional space
very hard.
Thus, it is very difﬁcult to visualise complex functions. How
ever, there are some methods available:
(i) We can draw two complex planes, one for the domain and
one for the range.
3
(ii) The twovariable functions u and v can be visualised sep
arately. The graph of a function of two variables is a sur
face in three space.
u(x, y) = cos x + sin y and v(x, y) = x
2
−y
2
(iii) Make one of the variables time and view the graph as
something that evolves over time. This is not very helpful.
Deﬁning e
z
, cos z and sin z
First we will try and deﬁne some elementary complex func
tions to play with. How shall we deﬁne functions such as e
z
,
cos z and sin z? We require that their deﬁnition should coincide
with the real version when z is a real number, and we would
like them to have properties similar to the real versions of the
functions, e.g. sin
2
z + cos
2
z = 1 would be nice. However, sine
and cosine are deﬁned using trigonometry and so are hard to
generalise: for example, what does it mean for a triangle to
have an hypotenuse of length 2 + 3i? The exponential is de
ﬁned using differential calculus and we have not yet deﬁned
differentiation of complex functions.
However, we know from Real Analysis that the functions
can be described using a power series, e.g.,
sin x = x −
x
3
3!
+
x
5
5!
− =
∞
n=0
(−1)
n
x
2n+1
(2n + 1)!
.
Thus, for z ∈ C, we shall deﬁne the exponential, sine and
cosine of z as follows:
e
z
:=
∞
n=0
z
n
n!
,
sin z :=
∞
n=0
(−1)
n
z
2n+1
(2n + 1)!
,
cos z :=
∞
n=0
(−1)
n
z
2n
(2n)!
.
4
Thus,
e
3+2i
=
∞
n=0
(3 + 2i)
n
n!
= 1 + (3 + 2i) +
(3 + 2i)
2
2!
+
(3 + 2i)
3
3!
+ . . .
These deﬁnitions obviously satisfy the requirement that they
coincide with the deﬁnitions we know and love for real z, but
how can we be sure that the series converges? I.e. when we
put in a z, such as 3 + 2i, into the deﬁnition, does a complex
number comes out?
To answer this we will have to study complex series and
as the theory of real series was built on the theory of real
sequences we had better start with complex sequences.
Complex Sequences
The deﬁnition of convergence of a complex sequence is the
same as that for convergence of a real sequence.
Deﬁnition 1.11
A complex sequence ¸c
n
) converges to c ∈ C, if given any ε > 0,
then there exists N such that [c
n
−c[ < ε for all n ≥ N.
We write c
n
→ c or lim
n→∞
c
n
= c.
Example 1.12
The sequence c
n
=
_
4 −3i
7
_
n
converges to zero.
Consider
[c
n
−0[ = [c
n
[ =
¸
¸
¸
¸
_
4 −3i
7
_
n
¸
¸
¸
¸
=
¸
¸
¸
¸
4 −3i
7
¸
¸
¸
¸
n
=
_
_
25
49
_
n
=
_
5
7
_
n
.
So
[c
n
−0[ < ε ⇐⇒ (5/7)
n
< ε
⇐⇒ nlog(5/7) < log ε
⇐⇒ n >
log ε
log(5/7)
.
So, given any ε we can choose N to be any natural number
greater than log ε/ log(5/7). Thus the sequence converges to
zero.
Remark 1.13
Notice that a
n
= [c
n
− c[ is a real sequence, and that c
n
→ c if
and only if the real sequence [c
n
−c[ → 0. Hence, we are saying
something about a complex sequence using real analysis.
5
Paradigm 1.14
The remark above gives a good example of the paradigm
1
we
will be using. We can apply results from real analysis to pro
duce results in complex analysis. In this case we take the
modulus, but we can also take real and imaginary parts.
This is a key observation. Note it well!
Let’s apply the paradigm. The next proposition shows that
a sequence converges if and only its real and imaginary parts
do.
Proposition 1.15
Let c
n
= a
n
+ ib
n
where a
n
and b
n
are real sequences, and c =
a + ib. Then
c
n
→ c ⇐⇒ a
n
→ a and b
n
→ b.
Proof. [⇒] If c
n
→ c, then [c
n
−c[ → 0. But
0 ≤ [a
n
−a[ = [Re(c
n
) −Re(c)[ = [Re(c
n
−c)[ ≤ [c
n
−c[.
So by the squeeze rule [a
n
− a[ → 0, i.e. a
n
→ a. Similarly,
b
n
→ b.
[⇐] Suppose a
n
→ a and b
n
→ b, then [a
n
− a[ → 0, and
[b
n
−b[ → 0. We have
0 ≤ [c
n
−c[ = [(a
n
−a) + i(b
n
−b)[ ≤ [a
n
−a[ +[b
n
−b[.
The last inequality follows from the triangle inequality applied
to z = a
n
−a and w = i(b
n
−b). Because [a
n
−a[ → 0 and [b
n
−b[ → 0
we deduce [c
n
−c[ → 0, i.e. c
n
→ c.
HTTLAM 1.16
Try not to use the deﬁnition of convergence to prove that a
sequence converges.
Example 1.17
n
2
+ in
3
n
3
+ 1
=
n
2
n
3
+ 1
+ i
n
3
n
3
+ 1
→ 0 + i.1 = i.
Exercises 1.18
(i) Which of the following sequences converge(s)?
(n + 1)
5
n
5
i
and
_
5 −12i
6
_
n
.
(ii) Show that the limit of a complex sequence is unique.
1
Paradigm: a conceptual model underlying the theories and practice of a scientiﬁc subject.
(Oxford English Dictionary).
6
Complex Series
Now that we have deﬁned convergence of complex sequences
we can deﬁne convergence of complex series.
Deﬁnition 1.19
A complex series
∞
k=0
w
k
converges if and only if the sequence
¸s
n
) formed by its partial sums s
n
=
n
k=0
w
k
converges.
That is, the following sequences converges
s
0
= w
0
s
1
= w
0
+ w
1
s
2
= w
0
+ w
1
+ w
2
s
3
= w
0
+ w
1
+ w
2
+ w
3
.
.
.
Let’s apply the paradigm and give a result on complex series
using real series.
Proposition 1.20
Let w
k
= x
k
+ iy
k
where x
k
and y
k
are real for all k. Then,
∞
k=0
w
k
converges ⇐⇒
∞
k=0
x
k
and
∞
k=0
y
k
converge.
In this case
∞
k=0
w
k
=
∞
k=0
x
k
+ i
∞
k=0
y
k
.
Proof. Let a
n
=
n
k=0
x
k
, b
n
=
n
k=0
y
k
, and s
n
=
n
k=0
w
k
, and
apply Proposition 1.15. The second part of the statement
comes from equating real and imaginary parts.
Example 1.21
The series
∞
n=0
(−1)
n
i
n!
converges. Let x
k
= 0 and y
k
=
(−1)
k
k!
.
Then
x
k
= 0, obviously, and
(−1)
k
k!
= e
−1
.
Thus
∞
n=0
(−1)
n
i
n!
converges to i/e.
In real analysis we have some great ways to tell if a series
is convergent, for example, the ratio test and the integral test.
Can we use the real analysis tests in complex analysis? The
next theorem says we can, but ﬁrst let us make a deﬁnition.
Deﬁnition 1.22
We say
∞
k=0
w
k
is absolutely convergent if the real series
∞
k=0
[w
k
[ converges.
7
This deﬁnition is really the same as in Real Analysis, it has
merely been extended to complex numbers in a natural way.
Now for a very important theorem which says that if a series
is absolutely convergent, then it is convergent.
Theorem 1.23
If
∞
k=0
[w
k
[ converges, then
∞
k=0
w
k
converges.
This is a fantastic tool. Remember it. The assumption says
something about a real series (we know lots about these!) and
gives a conclusion about a complex series. Thus, we can apply
the ratio test or comparison test to the real series and say
something about the complex series. Great!
Proof. Let w
k
= x
k
+ iy
k
, with x
k
and y
k
real. Then
∞
k=0
[w
k
[
convergent implies that
∞
k=0
[x
k
[ is convergent (because 0 ≤
[x
k
[ = [Re(w
k
)[ ≤ [w
k
[ and we can apply the comparison test).
So the real series
∞
k=0
x
k
converges absolutely and we know
from Real Analysis I that this implies that
∞
k=0
x
k
converges.
Similarly, the series
∞
k=0
y
k
converges.
Then,
∞
k=0
w
k
=
∞
k=0
x
k
+ i
∞
k=0
y
k
, by Proposition 1.20.
HTTLAM 1.24
When asked to show a series converges, show it absolutely
converges.
Remark 1.25
Note that the converse to Theorem 1.23 is not true. We al
ready know this from Real Analysis. For example,
∞
k=0
(−1)
k
k
converges but
∞
k=0
¸
¸
¸
(−1)
k
k
¸
¸
¸ =
∞
k=0
1
k
diverges.
We now prove an inﬁnite version of the triangle inequality.
Lemma 1.26
Suppose that
∞
k=0
w
k
converges absolutely. Then
¸
¸
¸
¸
¸
∞
k=0
w
k
¸
¸
¸
¸
¸
≤
∞
k=0
[w
k
[.
Proof. For n ≥ 1,
¸
¸
¸
¸
¸
∞
k=0
w
k
¸
¸
¸
¸
¸
=
¸
¸
¸
¸
¸
∞
k=0
w
k
−
n
k=0
w
k
+
n
k=0
w
k
¸
¸
¸
¸
¸
≤
¸
¸
¸
¸
¸
∞
k=0
w
k
−
n
k=0
w
k
¸
¸
¸
¸
¸
+
¸
¸
¸
¸
¸
n
k=0
w
k
¸
¸
¸
¸
¸
≤
¸
¸
¸
¸
¸
∞
k=0
w
k
−
n
k=0
w
k
¸
¸
¸
¸
¸
+
n
k=0
[w
k
[ .
8
As n → ∞ then obviously, [
∞
k=0
w
k
−
n
k=0
w
k
[ → 0, hence the
result.
Deﬁnition 1.27
A complex power series is a sum of the form
m
k=0
c
k
z
k
, where
c
k
∈ C and m is possibly inﬁnite.
Such a power series is a function of z. Much of the theory of
differentiable complex functions is concerned with power se
ries, because as we shall see later, any differentiable complex
function can be represented as a power series.
Radius of Convergence
Just as with real power series we can have complex power
series that do not converge on the whole of the complex plane.
Example 1.28
Consider the series
∞
0
z
n
, where z ∈ C. We know for z = 1
this series does not converge because then we have
∞
0
1
n
=
∞
0
1 = 1 + 1 + 1 + . . . .
We also know it converges for z = 0, because
∞
0
0
n
=
∞
0
0 = 0 + 0 + 0 + = 0. Hopefully, you remember from
Real Analysis I that for real z the power series converges only
for −1 < z < 1.
So, for which complex values of z does it converge? Let us
use the ratio test. Let a
n
= [z
n
[. Then
¸
¸
¸
¸
a
n+1
a
n
¸
¸
¸
¸
=
[z
n+1
[
[z
n
[
= [z[.
As n → ∞ we have [z[ → [z[, because there is no dependence
on n. So by the ratio test the series
a
n
converges if [z[ < 1,
diverges if [z[ > 1 and for [z[ = 1 we don’t know what will
happen. So
z
n
converges absolutely, and hence converges,
for [z[ < 1.
That the set of complex numbers for which the series con
verges is given by something of the form [z[ < R for some R is
a general phenomenon, as the next theorem shows.
Theorem 1.29
Let
∞
0
a
n
z
n
be some complex power series. Then, there exists
R, with 0 ≤ R ≤ ∞, such that
∞
0
a
n
z
n
_
converges absolutely for [z[ < R,
diverges for [z[ > R.
9
Proof. The proof is similar to that for real power series used
in Real Analysis I. Stewart and Tall also have a good proof, see
p5657.
HTTLAM 1.30
Given a power series, immediately ask ‘What is its radius of
convergence?’
Exercise 1.31
Show that
∞
0
z
n
/n has radius of convergence 1.
In the last exercise note that for z = −1 the series converges,
but for z = 1 the series diverges, (both these fact should be
well known from Real Analysis). This tells us that for [z[ = 1
we can get some values of z for which the series converges and
some for which the series diverges.
Sine, cosine, and exponential are deﬁned for all complex
numbers
Let us now return to showing that the sine, cosine and expo
nential functions are deﬁned on the whole of C.
Example 1.32
(I’ll do this example in great detail. The next example will be
more like the solution I would expect from you.)
The series e
z
=
∞
n=0
z
n
n!
converges for all z ∈ C.
For any z ∈ C let a
n
=
¸
¸
z
n
n!
¸
¸
. We want
∞
n=0
a
n
to converge, so
we use the ratio test on this real series. We have
¸
¸
¸
¸
a
n+1
a
n
¸
¸
¸
¸
=
¸
¸
¸
¸
_¸
¸
¸
¸
z
n+1
(n + 1)!
¸
¸
¸
¸
_¸
¸
¸
¸
z
n
n!
¸
¸
¸
¸
_¸
¸
¸
¸
=
¸
¸
¸
¸
z
n+1
z
n
n!
(n + 1)!
¸
¸
¸
¸
=
[z[
n + 1
→ 0 as n → ∞.
The last part is true because for ﬁxed z the real number [z[ is
of course a ﬁnite constant.
So by the ratio test
∞
n=0
a
n
=
∞
n=0
¸
¸
z
n
n!
¸
¸
converges. Thus by
Theorem 1.23 the series
∞
n=0
z
n
n!
converges for all z ∈ C.
The following is an example with some of the small detail
missing. This is how I would expect the solution to be given if
I had set this as an exercise.
10
Example 1.33
The series sin z =
∞
n=0
(−1)
n
z
2n+1
(2n + 1)!
converges for all z ∈ C.
Let a
n
=
¸
¸
¸
¸
(−1)
n
z
2n+1
(2n + 1)!
¸
¸
¸
¸
. Then
¸
¸
¸
¸
a
n+1
a
n
¸
¸
¸
¸
=
¸
¸
¸
¸
z
2(n+1)+1
(2(n + 1) + 1)!
¸
¸
¸
¸
_¸
¸
¸
¸
z
2n+1
(2n + 1)!
¸
¸
¸
¸
=
¸
¸
¸
¸
z
2n+3
(2n + 3)!
_
z
2n+1
(2n + 1)!
¸
¸
¸
¸
=
¸
¸
¸
¸
z
2
(2n + 3)(2n + 2)
¸
¸
¸
¸
=
[z[
2
(2n + 3)(2n + 2)
→ 0 as n → ∞.
So by the ratio test the complex series converges absolutely,
and hence converges.
Exercise 1.34
Prove that cos z converges for all z.
Properties of the exponential
We have deﬁned the exponential function and shown that is
deﬁned on all of C, let’s now look at its properties. Most of
these you may already from Numbers and Proofs, but the
proofs may not have been rigorous.
Theorem 1.35
(i) e
¯ z
= e
z
, for all z ∈ C.
(ii) e
iz
= cos z + i sin z, for all z ∈ C.
(iii) e
z+w
= e
z
e
w
, for all z, w ∈ C.
(iv) e
z
,= 0, for all z ∈ C.
(v) e
−z
= 1/e
z
, for all z ∈ C.
(vi) e
nz
= (e
z
)
n
, for all z ∈ C and n ∈ Z.
(vii) [e
z
[ = e
Re(z)
, for all z ∈ C.
(viii) [e
iy
[ = 1, for all y ∈ R.
Proof. (i) We have
e
¯ z
=
∞
n=0
(¯ z)
n
n!
=
∞
n=0
(z
n
)
n!
=
∞
n=0
z
n
n!
= e
z
.
11
(ii) Exercise. (Just put iz into the power series and separate
the real and imaginary parts.)
(iii) This will be delayed until we deal with differentiability.
(iv) Note that e
z
and e
−z
both exist. We have
e
z
e
−z
= e
z−z
by (iii),
= e
0
= 1, by calculation.
Thus, e
z
cannot be zero.
(v) This is obvious from the proof of (iv).
(vi) Follows from repeated application of (iii).
(vii) We have
[e
z
[
2
= e
z
e
z
by deﬁnition,
= e
z
e
z
by (i),
= e
z+z
by (iii),
= e
2Re(z)
=
_
e
Re(z)
_
2
by (vi).
As both [e
z
[ and e
Re(z)
are real and positive we deduce that (vii)
is true.
(viii) From (vii) we get [e
iy
[ = e
Re(iy)
= e
0
= 1.
Corollary 1.36
(i) e
2πi
= 1.
(ii) (De Moivre’s Theorem) (cos θ + i sin θ)
n
= cos nθ + i sin nθ for
all θ ∈ R.
The proofs are left as simple exercises. Part (i) is one of the
best theorems in mathematics. It relates so many different
important numbers, e,
√
−1, π, and of course 1 and 2, in a
simple expression.
Warning! 1.37
We have not shown that e
zw
= (e
z
)
w
for all z, w ∈ C. This is
because we have not yet deﬁned a
b
for all complex a and b.
Consider z = 2πi and w = i. Then (e
z
)
w
= (e
2πi
)
i
= 1
i
. What
could 1
i
be?
2
Exercise 1.38
Prove that
sin z =
e
iz
−e
−iz
2i
and cos z =
e
iz
+ e
−iz
2
.
2
The astute reader may say ‘deﬁne it to be e
−2π
.’
12
Another property of the complex exponential is that it is
periodic.
Theorem 1.39
For any complex numbers z and w we have
e
z
= e
w
⇐⇒ z −w = 2πin for some n ∈ Z.
Proof. [⇒] Let z −w = x + iy where x and y are real. Then,
e
z
= e
w
⇐⇒ e
z
/e
w
= 1
⇐⇒ e
z−w
= 1
⇐⇒ e
x+iy
= 1 (∗)
=⇒ [e
x+iy
[ = 1, (the implication does not reverse!)
⇐⇒ [e
x
e
iy
[ = 1
⇐⇒ [e
x
[[e
iy
[ = 1
⇐⇒ [e
x
[ = 1
⇐⇒ e
x
= 1, since the exponential of a real number is positive,
⇐⇒ x = 0.
By (∗) we know that e
x+iy
= 1, so e
iy
= 1 as x = 0. Then,
e
iy
= 1 ⇐⇒ cos y + i sin y = 1
⇐⇒ cos y = 1 and sin y = 0
⇐⇒ y = 2πn for some n ∈ Z.
So z −w = x + iy = 0 + i.2πn = 2πin.
[⇐] Suppose that z −w = 2πin for some n ∈ Z. Then, e
z−w
=
e
2πin
= (e
2πi
)
n
= 1
n
= 1. But from the working in the earlier part
of the proof we know this is equivalent to e
z
= e
w
.
This theorem has serious repercussions for deﬁning the in
verse of e
z
, i.e. deﬁning the log function.
Deﬁnition of the complex logarithm
We all know that the real exponential function has an inverse
function called log
e
or just ln. Is there an inverse for the com
plex exponential?
Well, to deﬁne the real log of a number x we want some
unique number y such that e
y
= x, that is the crux of the deﬁ
nition of inverse. So let’s suppose we have a complex number
w, then we want some z such that e
z
= w. Let’s investigate
this.
Proposition 1.40
For complex numbers z and w ,= 0 we have
e
z
= w ⇐⇒ z = ln [w[ + i(arg(w) + 2kπ), for some k ∈ Z.
13
Proof. [⇒] Write z = x+iy, so we get e
x+iy
= e
x
(cos y+i sin y) = w.
Now let us take the modulus of both sides:
[e
x+iy
[ = [w[
[e
x
[[e
iy
[ = [w[
[e
x
[ = [w[
e
x
= [w[
log
e
e
x
= log
e
[w[ (using the real log function)
x = ln [w[.
Now suppose that w = r(cos θ +i sin θ) for some real r and θ, i.e.
r = [w[ and θ = arg(w). Then e
z
= w implies that r = ln [w[ and
y = θ +2πk for some k ∈ Z. So z has the form in the statement.
[⇐] If z = ln [w[ + i(arg(w) + 2πk) for some k ∈ Z then
e
z
= e
ln w+i(arg(w)+2πk)
= e
ln w
e
i arg(w)
e
2πik
= [w[e
i arg(w)
_
e
2πi
_
k
= [w[e
i arg(w)
= w.
Example 1.41
Solve e
z
= 1 + i
√
3.
Solution: Let w = 1 + i
√
3. The modulus of w is [w[ =
_
1
2
+
√
3
2
=
√
4 = 2. By drawing a picture (or through careful
use of calculator) we can see that arg(w) =
π
3
+ 2nπ, n ∈ Z. So
z = ln 2 + i
_
π
3
+ 2nπ
_
, n ∈ Z.
HTTLAM 1.42
Notice how well working out the modulus and argument serves
us. Conclusion: calculate modulus and argument.
Common Error 1.43
Don’t forget the 2kπ with the argument.
Exercise 1.44
Solve e
2iz
= i. (It’s not i(π/2 + 2kπ).)
So does the theorem allow us to deﬁne the log of a complex
number? Yes, if we deﬁne the log to be the complex number
with −π < arg(w) ≤ π. (The point is that if we have a w then
the proposition gives us lots of zs to choose from. If z is such
that e
z
= w, then z + 2πi will work just as well (e
z+2πi
= e
z
e
2πi
=
e
z
.1 = e
z
= w). Thus, there is some ambiguity and we make
a choice.) However, if we are trying to solve an equation and
take the log of both sides using this deﬁnition, then we may be
14
losing solutions. So in fact the best deﬁnition is to make the
function multivalued. This is something we will not go into in
great depth just now.
Another worked example
Example 1.45
Solve the equation sin z = 2.
Solution: We can rewrite this as
e
iz
−e
−iz
2i
. Let w = e
iz
. Then
the equation becomes
1
2i
_
w −
1
w
_
= 2. So,
w
2
−1 = 4iw
w
2
−4iw −1 = 0
w =
4i ±
_
(4i)
2
+ 4
2
=
4i ±
√
−12
2
= 2i ±
√
−3
= (2 ±
√
3)i.
Now, e
iz
= w = (2 ±
√
3)i, so
iz = ln [w[ + i arg(w)
= ln [(2 ±
√
3)i[ + i
_
π
2
+ 2nπ
_
= ln [2 ±
√
3[ + i
_
π
2
+ 2nπ
_
z =
1
i
_
ln [2 ±
√
3[ + i
_
π
2
+ 2nπ
__
= −i
_
ln [2 ±
√
3[ + i
_
π
2
+ 2nπ
__
=
_
π
2
+ 2nπ
_
−i ln [2 ±
√
3[
=
_
π
2
+ 2nπ
_
−i ln(2 ±
√
3).
(The last equality is true because ln(2+
√
3) > 0 and ln(2−
√
3) >
0.)
HTTLAM 1.46
Note that in the above example we replaced e
iz
with another
complex number w, because we could then get a polynomial
equation.
15
Exercise 1.47
Show that
sin z = 0 ⇐⇒ z = kπ, k ∈ Z,
cos z = 0 ⇐⇒ z =
1
2
(2k + 1)π, k ∈ Z.
These results will be used later.
Summary
• Paradigm: Complex analysis is developed by reducing to
real analysis, often through taking the modulus.
• We deﬁne exponential, sine and cosine by power series.
• If
∞
n=0
[w
k
[ converges, then
∞
n=0
w
k
converges.
• Apply the ratio test, comparison test, etc, to the modulus
of terms of a complex series to determine convergence.
• For power series use the ratio test to ﬁnd radius of con
vergence.
16
2 Complex Riemann Integration
In a later section we deﬁne contour integration, that is inte
gration over a complex variable. This notion is fundamental in
complex analysis. But let us ﬁrst generalise integration and
differentiation to complexvalued functions of a real variable.
A complexvalued function of a real variable is a map f :
S → C, where S ⊆ R. E.g. If f(t) = (2 + 3i)t
3
, t ∈ R, then
f(1) = 2 + 3i ∈ C.
Such a function is different to a complex function. A complex
valued function of a real variable takes a real number and pro
duces a complex number. A complex function takes a complex
number from a domain and produces a complex number.
Differentiation of complex valued real functions
Suppose that f : R →C is given by f(t) = (1 +3i)t
2
. If we deﬁne
f
(t) using the standard deﬁnition:
f
(t) = lim
δ→0
f(t + δ) −f(t)
δ
, (δ ∈ R),
then we get f
(t) = 2(1 + 3i)t. That is, in this case, the rule
f
(t) = nct
n−1
for f(t) = ct
n
, holds even though c is complex.
Basically, all similar rules work in this way, any constants
can be real or complex. So, for instance,
d
dt
e
ct
= ce
ct
,
d
dt
sin(ct) = c cos(ct),
d
dt
cos(ct) = −c sin(ct).
Exercise 2.1
Let φ(x) = 3x
3
+ 2ix −i + tan((4 + 2i)x). Then,
φ
(x) =
Remark 2.2
This is not the same as differentiation with respect to a com
plex variable.
3
That will come later.
Complex Riemann Integrals
Now we shall integrate complexvalued functions with respect
to one real variable. We shall do this with a bit more care than
differentiation.
3
They do behave in much the same way though.
17
Deﬁnition 2.3
Let g : [a, b] → C be given g(t) = u(t) + iv(t), We say that g
is complex Riemann integrable (abbreviated CRI) if both u
and v are RI as real functions, and we deﬁne
_
b
a
g(t) dt by
_
b
a
g =
_
b
a
g(t) dt =
_
b
a
u(t) dt + i
_
b
a
v(t) dt.
Example 2.4
_
b
a
e
3it
dt =
_
b
a
cos 3t dt + i
_
b
a
sin 3t dt
=
1
3
[sin 3b −sin 3a] +
i
3
[−cos 3b + cos 3a]
=
1
3i
_
−e
3ib
+ e
3ia
_
=
i
3
_
e
3ib
−e
3ia
_
.
Many properties of CRI can be derived from the corre
sponding properties for RRI by considering the real and imag
inary parts. For example, if u and v are continuously differ
entiable, then since g
= u
+ iv
we get a version of the Funda
mental Theorem of Calculus:
_
b
a
g
=
_
b
a
u
+ i
_
b
a
v
= [u(b) −u(a)] + i[v(b) −v(a)] = g(b) −g(a).
Other standard methods, such as substitution also work.
Obviously, separating functions into real and imaginary parts
can get a bit tedious. Fortunately, just as for differentiation
above, we can use standard integrals, replacing real constants
by complex ones. It is not difﬁcult to prove the following,
where a and C are complex constants.
Example 2.5
_
at
n
dt = a
t
n+1
n + 1
+ C,
_
e
at
dt =
e
at
a
+ C,
_
sin at dt = −
1
a
cos(at) + C,
_
cos at dt =
1
a
sin(at) + C.
18
So a lot of the time we can use standard integrals to calculate
complexvalued integrals with respect to a real variable t.
Exercise 2.6
Calculate
_
2
0
t
2
−it
3
−cos(2t) dt.
Triangle inequality for CRI
We need the following result later. Its format should be fa
miliar from real analysis, the only difference here is that the
functions can be complexvalued.
Lemma 2.7
If g : [a, b] →C is CRI and [g[ is RRI, then
¸
¸
¸
¸
_
b
a
g(t) dt
¸
¸
¸
¸
≤
_
b
a
[g(t)[ dt.
Proof. If LHS = 0, then the statement is trivial. Hence, as
sume LHS ,= 0. Let α = [
_
b
a
g[/
_
b
a
g. (Hence [α[ = 1).
So,
¸
¸
¸
¸
_
b
a
g(t) dt
¸
¸
¸
¸
= α
_
b
a
g(t) dt
=
_
b
a
Re (αg(t)) dt + i
_
b
a
Im(αg(t)) dt
=
_
b
a
Re (αg(t)) dt, because LHS is real,
≤
_
b
a
[αg(t)[ dt, as Re(z) ≤ [z[,
=
_
b
a
[α[ [g(t)[ dt
=
_
b
a
[g(t)[ dt, as [α[ = 1.
Summary
• We can integrate and differentiate complexvalued func
tions of real variables in the same way as realvalued
functions of real variables.
19
3 Contours
In the next section deﬁne integration along a contour in the
complex plane.
4
This is a fairly abstract process, the mean
ing of which usually takes a little time to understand. Fortu
nately, it is easy to do as it has similar properties to Riemann
integration of one real variable, and you have years of experi
ence of that.
In case you think that it is too abstract and not relevant to
real problems, then consider the integral
_
2π
0
e
cos θ
cos(nθ −sin nθ) dθ.
This is a seriously nasty integral! Imagine trying to solve it
via the methods we know. Using contour integration we shall
show that it is very simple to calculate.
First though we will deﬁne contours.
Contours
Deﬁnition 3.1
A contour (also called a path) is a continuous map γ : [a, b] →
C which is piecewise smooth, i.e. there exist a = a
0
< a
1
< a
2
<
< a
n
= b such that
(i) γ[[a
j−1
, a
j
] is differentiable, for all j,
(ii) γ
is continuous on [a
j−1
, a
j
], for all j.
(The left and right derivatives of γ at a
j
may differ.)
We say γ is closed if γ(a) = γ(b).
Warning! 3.2
A contour is not a complex function. It is a complexvalued
function of a real variable. Its image is usually some curve in
the plane.
Examples 3.3
(i) Straight line from α to β: This is γ : [0, 1] → C given by
γ(t) = α + t(β −α).
(ii) Circle of radius r based at the origin: γ : [0, 2π] → C given
by γ(t) = re
it
.
(iii) Circle of radius r based at z
0
: γ : [0, 2π] → C given by
γ(t) = z
0
+ re
it
.
4
Those of you who have done MATH2360 or MATH2420 will see that this is just a line
integral.
20
(iv) Circular arc of radius r based at w: γ(t) = w +re
it
, θ
1
≤ t ≤
θ
2
. (So 0 ≤ t ≤ 2π gives the circle above).
(v) Let α : [−1, π/2] →C be given by
α(t) =
_
t + 1, for −1 ≤ t ≤ 0,
e
it
for 0 ≤ t ≤ π/2.
Draw the image:
We draw an arrow to show the direction we go in.
(vi) Let γ : [0, 4π] →C be given by γ(t) = e
it
. Then the image of γ
is the unit circle centred at zero. The contour goes round
the circle twice. This subtlety will be important later.
HTTLAM 3.4
Given a contour, try to draw its image.
Common Error 3.5
There is often confusion between a contour and its image. A
contour is not a set of points in the complex plane, it is a map.
Consider the contours (ii) and (vi) above, taking r = 1 in
(ii). They have the same image, the unit circle. However, the
contours are different, one maps from [0, 2π], the other [0, 4π].
We do use the notation z ∈ γ later, by which we mean z ∈
γ([a, b]). Strictly speaking, writing z ∈ γ is incorrect because γ
is not a set.
Since contours are complexvalued functions of a real vari
ables, we can differentiate them, etc, with ease.
Summary
• A contour is a continuous map γ : [a, b] → C which is
piecewise smooth. It is a complexvalued function of a
real variable.
• Straight line from α to β: γ : [0, 1] → C given by γ(t) =
α + t(β −α).
• Circle of radius r based at z
0
: γ : [0, 2π] → C given by
γ(t) = z
0
+ re
it
.
21
4 Contour Integration
We now come to probably the most important deﬁnition in
complex analysis: contour integral. It is central to the mod
ule. If you don’t understand this section, then the rest of the
course will be a complete mystery to you.
Deﬁnition 4.1
Let f : D →C be a continuous complex function and γ : [a, b] →
C be a contour. Then, the integral of f along γ is
_
γ
f =
_
γ
f(z) dz :=
_
b
a
f(γ(t))γ
(t) dt.
Note that f(γ(t)) and γ
(t) are complexvalued functions of a
real variable, and hence so is their product. Thus we can
integrate this product.
Example 4.2
Let γ(t) = t + it
2
for 0 ≤ t ≤ 2 and f(z) = z. Then, γ
(t) = 1 + 2it,
and
_
γ
f =
_
2
0
(t + it
2
)(1 + 2it) dt
=
_
2
0
t + 2it
2
+ it
2
+ 2i
2
t
3
dt
=
_
2
0
t + 3it
2
−2t
3
dt
=
_
1
2
t
2
+
3it
3
3
−
2t
4
4
_
2
0
=
_
1
2
t
2
+ it
3
−
t
4
2
_
2
0
=
1
2
2
2
+ i2
3
−
2
4
2
= −6 + 8i.
22
Example 4.3
Let γ(t) = 2 + it
2
for 0 ≤ t ≤ 1 and f(z) = z
2
. Then,
_
γ
z
2
dz =
_
1
0
(2 + it
2
)
2
(2it) dt
=
_
1
0
d
dt
_
(2 + it
2
)
3
3
_
dt
=
_
(2 + it
2
)
3
3
_
1
0
=
(2 + i.1
2
)
3
3
−
(2 + i.0
2
)
3
3
=
1
3
_
(2 + i)
3
−8
_
= −2 +
11
3
i.
This is just the sort of example you need to be able to do with
ease.
Exercise 4.4
Draw the contours and calculate the integrals of the functions
along the contours.
(i) f
1
(z) = Re(z) and γ
1
(t) = t, 0 ≤ t ≤ 1.
(ii) f
2
(z) = Re(z) and γ
2
(t) = t + it, 0 ≤ t ≤ 1.
(iii) f
3
(z) = Re(z) and γ
3
(t) = 1 −t + i(1 −t), 0 ≤ t ≤ 1.
(iv) f
4
(z) = 1/z and γ
4
(t) = 2e
−it
, 0 ≤ t ≤ π.
(v) f
5
(z) = z
2
and γ
5
(t) = e
it
, 0 ≤ t ≤ π/2.
Can you justify the results in (ii) and (iii)? Can you make any
conjectures, say, involving f(z) = z
n
in (v)?
Remark 4.5
Note that in the deﬁnition of contour integral we only require
f to be continuous. The resulting integrand f(γ(t))γ
(t) is C
RI because it is continuous except possibly at ﬁnitely many
points where γ
(t) is discontinuous. In practice we subdivide
[a, b] into pieces [a
j−1
, a
j
] and calculate
_
b
a
f(γ(t))γ
(t) dt =
n
j=1
_
a
j
a
j−1
f(γ(t))γ
(t) dt.
23
Example 4.6
Let γ be as in Example 3.3(v). Find
_
γ
z
2
dz.
_
γ
z
2
dz =
_
π/2
−1
γ(t)
2
γ
(t) dt
=
_
0
−1
(t + 1)
2
.1 dt +
_
π/2
0
_
e
it
_
2
ie
it
dt
=
_
0
−1
(t + 1)
2
dt +
_
π/2
0
ie
3it
dt
=
_
1
3
(t + 1)
3
_
0
−1
+
_
i
3i
e
3it
_
π/2
0
=
_
1
3
−0
_
+
i
3i
[−i −1]
= −
i
3
.
Remarks 4.7
(i) Suppose f : D → C is a complex function such that f(x)
is real for x real, for example, sin x. If we take γ : [a, b] → C
given by γ(t) = t for a ≤ t ≤ b, then
_
γ
f(z) dz =
_
b
a
f(t)γ
(t) dt =
_
b
a
f(t) dt.
Thus, by taking a contour along the real line, we can see
that contour integration includes the theory of real inte
gration as a special case.
(ii) From a purely formal viewpoint, we can justify the deﬁni
tion of contour integral by saying that we are replacing z
by γ(t) so we need to replace dz by γ
(t)dt, (which can be
thought of as (dz/dt)dt).
24
Fundamental Example
Take the function deﬁned by f(z) = (z − w)
n
where n ∈ Z, (so
for n < 0 the map is not deﬁned at w). Let γ be a circle with
centre w and radius r > 0, i.e. γ(t) = w + re
it
, 0 ≤ t ≤ 2π. Then,
_
γ
(z −w)
n
dz =
_
2π
0
(γ(t) −w)
n
γ
(t) dt
=
_
2π
0
_
w + re
it
−w
_
n
. ire
it
dt
=
_
2π
0
r
n
e
int
. ire
it
dt
= ir
n+1
_
2π
0
e
i(n+1)t
dt
=
_
_
_
r
n+1
n + 1
_
e
i(n+1)t
¸
2π
0
= 0, if n ,= −1,
i
_
2π
0
1.dt = 2πi, if n = −1.
Thus
_
γ
1
z −w
dz = 2πi and
_
γ
(z −w)
n
dz = 0 for n ,= −1.
Note this well, this innocuous looking calculation will be
used to devastating effect later!
Summary
• The integral of f along γ is
_
γ
f =
_
γ
f(z) dz =
_
b
a
f(γ(t))γ
(t) dt.
•
_
γ
1
z −w
dz = 2πi, γ(t) = w + re
it
, 0 ≤ t ≤ 2π.
•
_
γ
(z −w)
n
dz = 0 for n ,= −1, γ(t) = w + re
it
, 0 ≤ t ≤ 2π.
25
5 Properties of Contour Integration
There are a number of well known properties of ordinary inte
gration:
_
b
a
λf(x) + µg(x) dx = λ
_
b
a
f(x) dx + µ
_
b
a
g(x) dx, λ, µ ∈ R,
_
b
a
f(x) dx =
_
c
a
f(x) dx +
_
b
c
f(x) dx, a < c < b,
_
b
a
f(x) dx = −
_
a
b
f(x) dx,
_
b
a
f(x) dx =
_
φ
−1
(b)
φ
−1
(a)
f(φ(y))φ
(y) dy (change of variables).
All of these have analogues in contour integration. We shall
now describe them.
In the following the functions will be continuous on some
domain D and the contours will be maps into D.
Linearity.
If f, g are continuous on D, γ a contour in D, λ, µ ∈ C, then
_
γ
λf + µg = λ
_
γ
f + µ
_
γ
g.
So this has the same format as in integration over a real vari
able.
The proof follows directly from the linearity property of Rie
mann integration. The details are as follows:
_
γ
λf + µg =
_
b
a
(λf + µg)(γ(t))γ
(t) dt
=
_
b
a
(λf(γ(t)) + µg(γ(t))) γ
(t) dt
= λ
_
b
a
f(γ(t))γ
(t) dt + µ
_
b
a
g(γ(t))γ
(t) dt
= λ
_
γ
f + µ
_
γ
g.
Integration over joins
What if produce a contour by doing one after another?
26
Deﬁnition 5.1
If γ
1
: [a, b] → C and γ
2
: [b, c] → C are two contours such that
γ
1
(b) = γ
2
(b), then their join (or sum) is the contour γ
1
+ γ
2
:
[a, c] →C given by
(γ
1
+ γ
2
) (t) =
_
γ
1
(t) a ≤ t ≤ b,
γ
2
(t) b ≤ t ≤ c.
The join is continuous by the glue rule.
Example 5.2
Consider Example 3.3(v). The contour α : [−1, π/2] →C is given
by
α(t) =
_
t + 1, for −1 ≤ t ≤ 0,
e
it
for 0 ≤ t ≤ π/2.
This can be produced from γ
1
(t) = t + 1 for −1 ≤ t ≤ 0 and
γ
2
(t) = e
it
for 0 ≤ t ≤ π/2.
Proposition 5.3
For any continuous f
_
γ
1
+γ
2
f =
_
γ
1
f +
_
γ
2
f.
Again, this follows from applying the deﬁnition and using a
CRI property:
_
b
a
=
_
c
a
+
_
b
c
.
Example 5.4
Find
_
γ
1
(z −w)
dz, where γ is the boundary of the square with
corners w±l ±il, starting at w+l −li and going anticlockwise.
So γ = γ
1
+ γ
2
+ γ
3
+ γ
4
. By Proposition 5.3 we have
_
γ
dz
z −w
=
4
i=1
_
γ
i
dz
z −w
.
27
To compute
_
γ
1
let γ
1
(t) = w + l + it, −l ≤ t ≤ l. We have,
_
γ
1
dz
z −w
=
_
l
−l
γ
t
(t)
γ
1
(t) −w
dt
=
_
l
−l
i
l + it
dt
=
_
l
−l
l −it
l
2
+ t
2
i dt
=
_
l
−l
t
l
2
+ t
2
dt + i
_
l
−l
l
l
2
+ t
2
dt
= 0 + i
_
tan
−1
t
l
_
l
−l
= i
π
2
.
Similarly
_
γ
2
=
_
γ
3
=
_
γ
4
=
π
2
i (check!). So,
_
γ
dz
z −w
= 2πi.
Notice that this coincides with the value of the integral when
γ is a circle, see the Fundamental Example.
Reverse contour
Deﬁnition 5.5
If γ : [a, b] → C is a contour, then its reverse is the contour
(−γ)(t) = γ(a + b −t), for a ≤ t ≤ b.
The point is that we do γ backwards. Instead of starting at
γ(a) we end there, etc.:
(−γ)(a) = γ(a + b −a) = γ(b),
(−γ)(b) = γ(a + b −b) = γ(a).
Proposition 5.6
For any continuous f, we have
_
−γ
f = −
_
γ
f.
Proof. Again we apply deﬁnitions and use RI properties, but
in this case we also need a simple change variables: s = a+b−t,
28
so dt = −ds. Note that (−γ)
(t) = (−1)γ(a + b −t). We have,
_
−γ
f =
_
b
a
f((−γ)(t))(−γ)
(t) dt
=
_
b
a
f(γ(a + b −t))(−1)γ
(a + b −t) dt
= −
_
a
b
f(γ(s))γ
(s) (−ds)
=
_
a
b
f(γ(s))γ
(s) ds
= −
_
b
a
f(γ(s))γ
(s) ds
= −
_
γ
f.
Example 5.7
See Exercise 4.4(ii) and (iii). Then γ
3
= −γ
2
as follows. We have
a = 0 and b = 1, so
(−γ
2
)(t) = γ
2
(0 + 1 −t) = γ
2
(1 −t) = (1 −t) + i(1 −t) = γ
3
(t).
This explains why
_
γ
3
f
3
(z) dz = −
_
γ
2
f
2
(z) dz.
Reparametrisation
Now we shall do the analogue of a change of variables.
Deﬁnition 5.8
Let γ : [a, b] → C be a contour and let φ : [c, d] → [a, b] be a
function such that
(i) φ is continuously differentiable ,
(ii) φ
(s) > 0 for all s ∈ [c, d],
(iii) φ(c) = a, and φ(d) = b.
Then the contour ¯ γ : [c, d] →C deﬁned by ¯ γ(s) = γ(φ(s)) is called
a reparametrisation of γ.
Example 5.9
Let γ(t) = e
it
for 0 ≤ t ≤ 2π. Let φ(s) = 2πs for 0 ≤ s ≤ 1. Then
¯ γ(s) = e
2πis
for 0 ≤ s ≤ 1.
We think of reparametrised contours as equivalent since we
have the following.
29
Proposition 5.10
If ¯ γ is a reparametrisation of γ, then
_
e γ
f =
_
γ
f,
for all continuous f.
Proof. We have,
_
e γ
f =
_
d
c
f(¯ γ(s)) ¯ γ
(s) ds
=
_
d
c
f(γ(φ(s)) γ
(φ(s)) φ
(s) ds
=
_
b
a
f(γ(t)) γ
(t) dt, using t = φ(s),
=
_
γ
f.
Thus, it does not matter whether we evaluate the integral
_
γ
1/(z − w) dz using γ(t) = e
it
for 0 ≤ t ≤ 2π or γ(t) = e
2πit
for 0 ≤ t ≤ 1.
Remarks 5.11
(i) If γ
1
: [a, b] → C and γ
2
: [c, d] → C are two contours such
that γ
1
(b) = γ
2
(c), then we can reparametrise γ
2
as
¯ γ
2
: [b, d + b −c] →C given by ¯ γ(s) = γ
2
(c −b + s).
Then γ
1
(b) = ¯ γ
2
(b) so we can form the join γ
1
+¯ γ
2
. Often we
abuse notation and write this simply as γ
1
+ γ
2
.
(ii) If γ is a simple (i.e. it doesn’t cross itself) closed con
tour with no orientation speciﬁed, then it is traversed an
ticlockwise.
(iii) If ¯ γ is a reparametrisation of γ, then it has the same image
as γ. The converse is not true.
30
Summary
• If f, g are continuous on D, γ a contour in D, λ, µ ∈ C, then
_
γ
λf + µg = λ
_
γ
f + µ
_
γ
g.
• If f is continuous on D, γ a contour in D, then
_
γ
1
+γ
2
f =
_
γ
1
f +
_
γ
2
f.
• If γ : [a, b] →C is a contour, then its reverse is the contour
(−γ)(t) = γ(a + b −t), for a ≤ t ≤ b.
• For f continuous, we have
_
−γ
f = −
_
γ
f.
• If ¯ γ is a reparametrisation of γ, then
_
e γ
f =
_
γ
f,
31
6 The Estimation Lemma
Recall for a real continuous function that
¸
¸
¸
¸
_
b
a
f(x) dx
¸
¸
¸
¸
≤ sup
x∈[a,b]
¦[f(x)[¦ (b −a).
We would like a complex version of this, that is, is there some
bound on [
_
γ
f(z) dz[? It is obvious that the ﬁrst part in the
product above can be generalised, but what does b − a cor
respond to? Those of you familiar with measure theory will
know it is the length of the interval from a to b, and so the
generalisation to complex analysis is the length of the contour
γ.
Deﬁnition 6.1
Let γ : [a, b] → C be a contour. The length of γ, denoted L(γ),
is deﬁned to be
L(γ) =
_
b
a
[γ
(t)[ dt.
This really does measure the length of the curve. Intuitively
speaking, γ
(t) is the velocity of a contour, so [γ
(t)[ is the
speed, and the integral of speed over time gives the length
of a path.
Example 6.2
Let γ(t) = α+t(β −α) with 0 ≤ t ≤ 1 be the straight line contour
from α to β.
We have
L(γ) =
_
1
0
[γ
(t)[ dt =
_
1
0
[β −α[ dt = [β −α[
_
1
0
dt = [β −α[[t]
1
0
= [β −α[.
So, the length of the contour from α to β is [β − α[, which is
reassuring.
Exercise 6.3
Show that the length of the contour given by traversing once
round the the circle of radius r based at the origin is 2π.
The next theorem is similar to one from real analysis:
Lemma 6.4 (Estimation Lemma)
Let f : D →C be a continuous complex function and γ : [a, b] →
D be a contour. Suppose that [f(z)[ ≤ M for all z ∈ γ. Then
¸
¸
¸
¸
_
γ
f(z) dz
¸
¸
¸
¸
≤ M L(γ).
32
Proof. We have,
¸
¸
¸
¸
_
γ
f(z) dz
¸
¸
¸
¸
=
¸
¸
¸
¸
_
b
a
f(γ(t))γ
(t) dt
¸
¸
¸
¸
≤
_
b
a
[f(γ(t))γ
(t)[ dt by Lemma 2.7
≤
_
b
a
M[γ
(t)[dt
= ML(γ).
Example 6.5
Show that
¸
¸
¸
¸
_
γ
e
z
z + 1
dz
¸
¸
¸
¸
≤
πR
R −1
where γ describes the semicircle from iR to −iR in the left
halfplane ¦z : Re(z) ≤ 0¦ and R > 1.
Solution: If z = x + iy lies on the image of γ, then x ≤ 0 and
so [e
z
[ = e
Re(z)
= e
x
≤ e
0
= 1. Also, when z lies on γ, [z[ = R so
R = [z + 1 −1[ ≤ [z + 1[ + 1. Thus [z + 1[ ≥ R −1. Hence,
¸
¸
¸
¸
e
z
z + 1
¸
¸
¸
¸
≤
1
R −1
, z ∈ γ.
We know that L(γ) = πR as γ is semicircle of radius R, so by
the Estimation Lemma (6.4)
¸
¸
¸
¸
_
γ
e
z
z + 1
dz
¸
¸
¸
¸
≤
1
R −1
L(γ) =
πR
R −1
.
Remarks 6.6
(i) The constant M always exists: On the image of γ the
function [f(z)[ will always be bounded because the map
t → [f(γ(t))[ is a continuous real function on [a, b], and
hence is bounded. So M = sup
z∈γ
¦[f(z)[¦ will do as a
bound. Anything bigger than this is also useful.
(ii) We can show some integral is zero by ﬁnding an M that
tends to zero or some contour, which can be made smaller,
so that L(γ) → 0.
Termwise integration of series
The lemma will be useful in a number of contexts. To begin
with, we use it to prove that we can integrate certain series
termbyterm. (We know that an inﬁnite series can be differ
entiated termbyterm.)
33
Corollary 6.7 (Termbyterm integration of series)
Let γ be a contour in a domain D. Let f : D →C and f
k
: D →C
be continuous complex functions, k ∈ N. Suppose that
(i)
∞
k=0
f
k
(z) converges to f(z) for all z ∈ γ;
(ii) there exist real constants M
k
such that [f
k
(z)[ ≤ M
k
for all
z ∈ γ;
(iii)
∞
k=0
M
k
converges.
Then,
∞
k=0
_
γ
f
k
(z) dz =
_
γ
∞
k=0
f
k
(z) dz =
_
γ
f(z) dz.
Proof. The second equality is obvious. We show the ﬁrst. Let
M =
∞
k=0
M
k
. We have
¸
¸
¸
¸
¸
_
γ
f(z) dz −
n
k=0
_
γ
f
k
(z) dz
¸
¸
¸
¸
¸
=
¸
¸
¸
¸
¸
_
γ
_
f(z) −
n
k=0
f
k
(z)
_
dz
¸
¸
¸
¸
¸
≤ sup
z∈γ
_¸
¸
¸
¸
¸
f(z) −
n
k=0
f
k
(z)
¸
¸
¸
¸
¸
_
L(γ)
= sup
z∈γ
_¸
¸
¸
¸
¸
∞
k=n+1
f
k
(z)
¸
¸
¸
¸
¸
_
L(γ)
≤ sup
z∈γ
_
∞
k=n+1
[f
k
(z)[
_
L(γ)
≤
_
∞
k=n+1
M
k
_
L(γ)
=
_
M −
n
k=0
M
k
_
L(γ)
→ 0 L(γ), as n → ∞,
= 0.
Thus, the ﬁrst equality is true.
Remark 6.8
I could have deﬁned uniform convergence and so on for com
plex series in order to state the theorem. Rather than waste
time doing so I just used a version of the Weierstrass Mtest in
the assumptions above. Hopefully, you remember from Real
Analysis II that this implies uniform convergence.
34
Example 6.9
For any contour γ the integral
_
γ
e
z
dz can be calculated by
termbyterm integration of the series for e
z
.
Let f
k
(z) =
z
k
k!
, then e
z
=
∞
k=0
f
k
(z), so condition (i) is fulﬁlled.
We have
[f
k
(z)[ =
¸
¸
¸
¸
z
k
k!
¸
¸
¸
¸
=
[z[
k
k!
.
But [z[ = [γ(t)[ and γ(t) is a continuous map from [a, b] to C so
its image must lie within an origincentred circle of radius R,
for some large enough R. Thus [z[ ≤ R for all z ∈ γ.
Hence, let M
k
=
R
k
k!
, then [f
k
(z)[ ≤ M
k
for all z ∈ γ. Therefore,
condition (ii) holds.
Also,
∞
0
M
k
=
∞
0
R
k
k!
= e
R
,
so
∞
0
M
k
converges. Condition (iii) holds.
Thus,
_
γ
e
z
dz =
_
γ
∞
k=0
z
k
k!
=
∞
k=0
_
γ
z
k
k!
dz.
This of course begs the question, ‘how do we ﬁnd
_
γ
z
k
k!
dz’? Ob
viously, we can apply the deﬁnition of integration, but there
are better ways, such as the Fundamental Theorem of Calcu
lus, as we shall see in the next two sections.
Summary
• The length of γ, denoted L(γ), is deﬁned to be
L(γ) =
_
b
a
[γ
(t)[ dt.
• Suppose that [f(z)[ ≤ M for all z ∈ γ. Then
¸
¸
¸
¸
_
γ
f(z) dz
¸
¸
¸
¸
≤ M L(γ).
• We can integrate termbyterm complex series that satisfy
a Weierstrass Mtest type condition.
35
7 Complex Differentiation
If we were inventing the theory of differentiation of complex
functions for the the ﬁrst time, then we might be tempted to
deﬁne complex differentiable to mean that the real and imag
inary parts of the function are differentiable with respect to
x and y. This would give us a theory, but not a great one; it
would be the same as theory of differentiable maps from R
2
to
R
2
.
Instead we develop a theory that really uses the complex
numbers. This gives a much richer theory. It looks a lot
better, but more importantly, it allows us to solve some hard
problems concerning real functions in a simple manner.
Continuity
Let’s ﬁrst give a deﬁnition of continuity.
Deﬁnition 7.1
Suppose that f : D →C is a complex function on a domain D.
We say f is continuous at c if lim
z→c
f(z) = f(c)
This is not remarkably different to real analysis.
Remark 7.2
This is equivalent to either of the following.
(i) For all ε > 0, there exists δ > 0 such that [f(z) − f(c)[ < ε,
whenever [z −c[ < δ.
(ii) For any sequence ¸z
n
), we have z
n
→ c implies that f(z
n
) →
f(c).
The proofs of these facts are the same as in real analysis.
Deﬁnition 7.3
We say f is continuous on D if f is continuous at c for all
c ∈ D.
We need some examples and the next result helps supply
some.
Proposition 7.4
Suppose f(x + iy) = u(x, y) + iv(x, y). Then, f is continuous if
and only if u and v are continuous.
Proof. Use Proposition 1.15.
Thus, f(z) = z
n
is continuous. Sums and products of contin
uous functions are continuous, and so on.
36
Differentiation
Now we come to the crucial deﬁnition, that of differentiation.
It doesn’t look different to the real situation, but the conse
quences are far more profound.
Deﬁnition 7.5
Let f : D → C be a complex function. Then, f is complex
differentiable at c if
lim
h→0
f(z + h) −f(z)
h
, h ∈ C
exists. We write f
(c) for this limit.
We say f is complex differentiable on D if it is complex
differentiable at c for all c ∈ D.
The deﬁnition looks the same as in the real case, but the fact
that h can go to zero from any direction in the complex plane
makes a huge difference, it imposes considerable restrictions.
Remark 7.6
Complex differentiable functions are also called holomoprhic
or analytic.
Example 7.7
The function f : C → C given by f(z) = z
2
is differentiable on
C.
Let c ∈ C. Then
lim
h→0
(c + h)
2
−c
2
h
= lim
h→0
c
2
+ 2ch + h
2
−c
2
h
= lim
h→0
2c + h
= 2c.
That is, f
(c) = 2c, as expected.
More generally, we have that, if f(z) = bz
n
for some complex
constant b, then f
(z) = nbz
n−1
. The proof of this is the same
as the real situation, at least symbolically. Of course, math
ematicians do not want to do anything as clumsy or ugly as
using ﬁrst principles. We would use a theorem such as the
following.
Proposition 7.8
Let f and g be complex functions on the domain D.
(i) If f and g are differentiable at c ∈ D, then so are
(a) f + g, and (f + g)
(c) = f
(c) + g
(c);
(b) fg, and (fg)
(c) = f(c)g
(c) + f
(c)g(c);
37
(c) f/g, and (f/g)
(c) =
g(c)f
(c) −f(c)g
(c)
g(c)
2
provided g(c) ,=
0.
(ii) (Chain Rule) Suppose f : D →C and g : E →C are complex
functions with f(D) ⊆ E. If f is differentiable at c and g is
differentiable at f(c), then g ◦ f is differentiable at c with
(g ◦ f)
(c) = g
(f(c))f
(c).
Proof. These are proved in the same way as the real case.
Warning! 7.9
Unlike continuity, complex differentiablity isn’t the same as
being differentiable with respect to two real variables. There
is a connection as we see shortly. In fact it is this difference
that makes complex analysis so different to real analaysis.
Exercise 7.10
Show that if f : D →C is differentiable at c, then f is continu
ous at c.
We don’t yet know that we can differentiate series term
byterm and so can’t immediately prove that e
z
, cos z and sin z
are differentiable, or that their derivatives are what we ex
pect them to be. It is possible to work ﬁnd their derivatives
from ﬁrst principles, (you can try this as an exercise if you
are keen), but will delay a proof till later and just state the
following.
Theorem 7.11
The elementary functions have the expected derivatives:
(i)
d
dz
e
z
= e
z
, for all z ∈ C.
(ii)
d
dz
sin z = cos z for all z ∈ C.
(iii)
d
dz
cos z = −sin z for all z ∈ C.
Exercise 7.12
Where are the following funtions not differentiable?
z
2
,
1
z
,
1
z
2
,
sin z
z(z
2
+ 1)
, tan z.
38
Summary
• The complex function f is continuous at c if lim
z→c
f(z) =
f(c).
• The complex function f is complex differentiable at c if
lim
h→0
f(z + h) −f(z)
h
, h ∈ C.
• Differentiablity =⇒ continuity.
• The elementary functions have the expected derivatives.
39
8 CauchyRiemann Equations
If f is a differentiable complex function, then the real and
imaginary parts are differentiable as real functions. (But not
vice versa, see earlier warning.)
Theorem 8.1
Suppose f(x + iy) = u(x, y) + iv(x, y), where z = x + iy, and that
f is differentiable at c = a + ib. Then the partial derivatives u
x
,
u
y
, v
x
and v
y
all exist at (a, b) and
f
(c) = u
x
(a, b) + iv
x
(a, b) = v
y
(a, b) −iu
y
(a, b).
Proof. We know
f(c + h) −f(c)
h
→ f
(c) as h → 0. Let h = k + il,
we get
u(a + k, b + l) −u(a, b) + iv(a + k, b + l) −iv(a, b)
k + il
→ f
(c) as k+il → 0.
So if we let h → 0 through real values, i.e. l = 0, then
u(a + k, b) −u(a, b) + iv(a + k, b) −iv(a, b)
k
→ f
(c) as k → 0.
Therefore,
u(a + k, b) −u(a, b)
k
→ Re(f
(c)) as k → 0. So u
x
(a, b)
exists and equals Re(f
(c)). Likewise, v
x
(a, b) = Im(f
(c)).
The second equation follows from letting h → 0 through
imaginary values, i.e. k = 0.
And now for a result that is fundamental in all complex
analysis courses.
Corollary 8.2 (CauchyRiemann equations)
If f(z) = u(x, y) + iv(x, y) is differentiable, then
u
x
= v
y
and v
x
= −u
y
.
Proof. Equate real and imaginary parts in the theorem above.
The two equations are called the CauchyRiemann equations
after two of the founders of complex analysis.
Note that the corollary says that if f is differentiable, then
the equations hold, but says nothing of the converse, which is
not true anyway!
Examples 8.3
(i) Let f(z) = z
2
= (x +iy)
2
. Then, u(x, y) = x
2
−y
2
and v(x, y) =
2xy. We see that
u
x
= v
y
= 2x and u
y
= −v
x
= −2y,
so the CR equations hold.
40
(ii) Let f(z) = [z[
2
= x
2
+ y
2
. Then u = x
2
+ y
2
and v = 0. Thus,
u
x
= 2x, v
y
= 0,
u
y
= 2y, v
x
= 0.
The only place where the CR equations are satisﬁed is
x = y = 0.
So f is differentiable nowhere, except possibly at 0. Is it
differentiable at 0? Well,
f(0 + h) −f(0)
h
=
[h[
2
−0
h
=
h
¯
h
h
=
¯
h → 0 as h → 0,
so f
(0) = 0.
Exercise 8.4
Show that the function f(z) = [z[ is not differentiable anywhere
in C.
Remark 8.5
The preceding exercise shows that complex differentiability is
imposing a stronger condition real differentiability, because
for real points of C the function f(x) = [x[ is differentiable,
provided x ,= 0. (You know this last fact well, I hope!).
The condition is stronger because we require f
(c) to exist
as h → 0 from all directions, not just real ones.
Remark 8.6
The converse to Corollary 8.2 is false. For example, let
f(z) =
_
0, if x = 0 or y = 0,
1, otherwise.
Then, the equations are satisﬁed, (check!), but f is not con
tinuous, so can’t be differentiable.
However, if u
x
, u
y
, v
x
, v
y
exist near c and are continuous at
c, and satisfy the CR equations, then f is differentiable at c.
This fact won’t be used later, but is useful to know.
Harmonic functions
Deﬁnition 8.7
Let w(x, y) be a C
2
function of two real variables
5
. Then w is a
harmonic function if it satisﬁes Laplace’s equation:
w
xx
+ w
yy
= 0.
Example 8.8
Let w(x, y) = x
2
−y
2
. Then w
xx
= 2 and w
yy
= −2, so w
xx
+w
yy
= 0.
5
A C
k
function is a function that is differentiable k times
41
Laplace’s equation is important in potential theory and many
other areas, and so, since they are solutions of it, harmonic
functions are of particular interest. Complex analysis provides
many examples of harmonic functions, as the next theorem
shows.
Theorem 8.9
If f : D → C is complex differentiable, and u and v are C
2

functions, then u and v are both harmonic:
u
xx
+ u
yy
= 0 and v
xx
+ v
yy
= 0.
Proof. We use the CR equations:
u
xx
+ u
yy
= (u
x
)
x
+ (u
y
)
y
= (v
y
)
x
+ (−v
x
)
y
= v
xy
−v
yx
= 0.
Similarly for v.
Conversely, given an harmonic function u there is locally
(i.e. in some εneighbourhood) a harmonic function v such
that f(x + iy) = u(x, y) + i(v, y) is complex differentiable. Such
a v is called a harmonic conjugate of u.
Example 8.10
Let u(x, y) = 2xy, then u is harmonic. We can construct v using
the CR equations (they’re very useful, aren’t they!) Because
2y = u
x
= v
y
=⇒ v = y
2
+ g(x) where g is a function of x,
and
2x = u
y
= −v
x
=⇒ v = −x
2
+ h(y) where h is a function of y,
we can deduce that v = y
2
−x
2
+ C where C is a constant.
Summary
• (CauchyRiemann equations) If f(z) = u(x, y) + iv(x, y) is
differentiable, then
u
x
= v
y
and v
x
= −u
y
.
• The function w : R
2
→ R is a harmonic function if it satis
ﬁes Laplace’s equation:
w
xx
+ w
yy
= 0.
• For every harmonic u there is a harmonic function v (called
a harmonic conjugate of u) such that f(x + iy) = u(x, y) +
i(v, y) is complex differentiable.
42
9 Fundamental Theorem of Calculus for Com
plex Functions
One of the best theorems in Real Calculus is the Fundamental
Theorem of Calculus. We now see this in a contour integration
setting.
Theorem 9.1 (Fundamental Theorem of Calculus)
Let f : D →C be a continuous complex function and γ : [a, b] →
D be a contour. Suppose there exists a complex differentiable
F : D →C such that F
= f. Then,
_
γ
f(z) dz = F(γ(b)) −F(γ(a)).
Proof. Let a = a
0
< a
1
< < a
n
= b be a dissection of [a, b]
such that γ
[[a
j−1
, a
j
] is continuous for all j. Then,
_
γ
f(z) dz =
_
γ
F
(z) dz
=
_
b
a
F
(γ(t))γ
(t) dt
=
n
j=1
_
a
j
a
j−1
F
(γ(t))γ
(t) dt
=
n
j=1
_
a
j
a
j−1
(F ◦ γ)
(t) dt
=
n
j=1
[(F ◦ γ)(t)]
a
j
a
j−1
by the usual FTC for RI functions,
= F(γ(b)) −F(γ(a)).
Example 9.2
Consider Example 4.3. Then, f(z) = z
2
, γ(0) = 2 and γ(1) = 2+i.
Obviously, F(z) =
1
3
z
3
is an antiderivative for f, i.e. F
(z) =
f(z). Then, by the FTC,
_
γ
f(z) dz =
1
3
_
(γ(1))
3
−(γ(0))
3
_
=
1
3
_
2
3
−(2 + i)
3
_
= −2 +
11
3
i.
43
How common are antiderivatives for continuous complex
functions? Do they always exist? For real continuous func
tions we know that the Riemann integral can be found and
this will be an antiderivative. Unfortunately, the analogue is
not true for continuous complex functions, not even differen
tiable ones. Consider this corollary of the FTC and the follow
ing example.
Corollary 9.3
With the assumptions of the above theorem suppose that γ is
any closed contour. Then
_
γ
f(z) dz = 0.
Proof. The deﬁnition of a closed contour is that γ(a) = γ(b).
So,
_
γ
f(z) dz = F(γ(b)) −F(γ(b)) = 0.
Example 9.4
Let f(z) = 1/z. Then, f is differentiable on D = C¸¦0¦. Let γ be
the unit circle round the origin, traversed once anticlockwise.
Then, by the fundamental example we know
_
γ
f(z) dz =
_
γ
1
z
dz = 2πi.
Therefore, if there existed an F : D →C such that F
= f, then
the corollary would be contradicted. Hence, antiderivatives do
not always exist.
Property of exponential
We can now prove a result we would expect to be true by anal
ogy with real analysis: if a function has zero derivative, then
it is constant.
First we need a deﬁnition.
Deﬁnition 9.5
A domain D is called connected if for each pair α, β ∈ D, there
exists a contour γ : [a, b] → D such that γ(a) = α and γ(b) = β.
So a domain is connected if we can draw a curve from any
point to any other.
Examples 9.6
(i) The domains D = C and D = C¸¦0¦ are both connected.
44
(ii) The domain D = C¸¦z : z is real¦ is not connected. There
is no way to construct a contour starting below and ﬁn
ishing above the real line, without crossing that line. But
the real line is not in D.
Theorem 9.7
Suppose that D is is connected domain, and f : D → C is
analytic, such that f
(z) = 0 for all z ∈ D. Then, f is constant.
Proof. Take any α and β in D. Then, as D is connected, there
exists a path γ : [a, b] → D, such that γ(a) = α and γ(b) = β. By
the FTC
f(β) −f(α) = f(γ(b)) −f(γ(a)) =
_
γ
f
=
_
γ
0 = 0.
Thus f(β) = f(α). Since these were general points of D, we
deduce that f is constant.
This theorem allows us to prove the property of e
z
in Theo
rem 1.35 we did not prove earlier.
Corollary 9.8
For all complex numbers z and w we have e
z+w
= e
z
e
w
.
Proof. Let f(z) = e
z
e
α−z
, where α is any complex number. By
the product rule we get f
(z) = −e
z
e
α−z
+ e
z
e
α−z
= 0. Thus, by
the theorem, the function is constant. So, if we let α = w + z,
then we get
f(w) = f(0)
e
w
e
(w+z)−w
= e
0
e
w+z+0
e
w
e
z
= e
w+z
.
* Existence of antiderivatives
The next proposition gives some conditions equivalent to the
existence of antiderivatives.
Proposition 9.9
Let f : D → C be a continuous complex function on a con
nected domain D. The following statements are equivalent.
(i) There exists an F : D →C such that F
= f.
(ii)
_
γ
f = 0 for every closed contour γ in D.
(iii)
_
γ
f only depends on the end points of γ for any contour γ
in D.
45
Proof. We shall prove (i)⇒(ii)⇒(iii)⇒(i).
(i)⇒(ii): This is just Corollary 9.3.
(ii)⇒(iii): Suppose that α and β are two points in D, then we
want to prove that the integral of f over any two contours γ
1
and γ
2
both starting at α and ﬁnishing at β are equal.
Consider the contour γ
1
− γ
2
. This is a closed contour be
cause it starts and ﬁnishes at α. (Draw a picture!) Hence, (ii)
implies that
_
γ
1
−γ
2
f = 0. We have
0 =
_
γ
1
−γ
2
f =
_
γ
1
f −
_
γ
2
f.
Hence (iii) is true.
(iii)⇒(i): We shall deﬁne a function F and show that it is
differentiable. Fix z
0
∈ D. Since D is connected, for any point
z ∈ D, there exists a contour γ from z
0
to z. We deﬁne
F(z) =
_
γ
f(w) dw.
[Note that this really is a function of z because γ has z as its
endpoint.]
Now we show that F
(z) = f(z). Let z
1
∈ D be any point.
Then by deﬁnition F is differentiable at z
1
with derivative f(z
1
)
if
lim
h→0
F(z
1
+ h) −F(z
1
)
h
= f(z
1
).
So, as D is a domain, there exists an ε
1
neighbourhood at
z
1
for some ε > 0. If [h[ < ε
1
, then there exists a contour
Λ : [0, 1] → D which gives the straight line from z
1
to z
1
+ h, i.e.
Λ(t) = z
1
+ ht.
Using our deﬁnition of F, we get
F(z
1
+ h) =
_
γ+Λ
f =
_
γ
f +
_
Λ
f.
Therefore,
F(z
1
+ h) −F(z
1
)
h
=
1
h
__
γ
f +
_
Λ
f −
_
γ
f
_
=
1
h
_
Λ
f(w) dw.
We have,
_
Λ
f(z
1
) dw = f(z
1
)
_
Λ
dw = f(z
1
) [w]
z
0
+h
z
0
= f(z
1
)h.
Hence,
F(z
1
+ h) −F(z
1
)
h
−f(z
1
) =
_
Λ
f(w) −f(z
1
)
h
dw.
46
We want the LHS to tend to zero as h → 0, so we use the
Estimation Lemma on the RHS. Since f is continuous, given
any ε > 0, there exists a δ > 0 such that [w − z
1
[ < δ implies
that [f(w) − f(z
1
)[ < ε. We can assume that δ < ε
1
. So, when
[z[ < δ and w ∈ Λ, we have
¸
¸
¸
¸
f(w) −f(z
1
)
h
¸
¸
¸
¸
<
ε
[h[
.
The length of Λ is [h[, so we deduce fromthe Estimation Lemma
that
¸
¸
¸
¸
_
Λ
f(w) −f(z
1
)
h
dw
¸
¸
¸
¸
≤
ε
[h[
[h[.
Thus,
¸
¸
¸
¸
F(z
1
+ h) −F(z
1
)
h
−f(z
1
)
¸
¸
¸
¸
≤ ε for all [h[ < δ.
Since ε was arbitrary we deduce that the LHS is zero. This
implies that F
(z
1
) = f(z
1
). So, F is what we were seeking.
Summary
• Fundamental Theorem of Calculus: Suppose f : D → C
is a continuous complex function and there exists F such
that F
= f. Then,
_
γ
f(z) dz = F(γ(b)) −F(γ(a)).
• Not all functions have an antiderivative.
• A domain D is called connected if for each pair α, β ∈ D,
there exists a contour γ : [a, b] → D such that γ(a) = α and
γ(b) = β.
47
10 Differentiability of Power Series
We have successfully deﬁned functions such as exp, sin and
cos by power series. Now we would like to show that they are
differentiable. We do this by investigating the differentiability
of power series with positive radius of convergence. We will
show later that if a function is differentiable at a point z
0
,
then near that point it can be given as a power series, so this
investigation wil be useful for more than just the elementary
functions.
We prove:
• A power series is differentiable in the obvious way: term
byterm.
• The derivative of a power series has the same radius of
convergence.
• Power series are inﬁnitely differentiable.
• The coefﬁcients of the power series can be given in terms
of the derivatives.
• The coefﬁcients of the power series are unique.
Suppose that
∞
n=0
a
n
z
n
is a series. Then, the obvious can
didate for the derivative is the one produced by termbyterm
differentiation
6
, i.e.
∞
n=0
na
n
z
n−1
. But, initially, we do do not
even know that this sequence converges.
So, we begin by proving that a power series and its ‘obvious
derivative’ have same radius of convergence, and then show
that this obvious derivative really is the derivative of the series.
Lemma 10.1
The series
∞
n=0
a
n
z
n
and
∞
n=0
na
n
z
n−1
have the same radius of
convergence.
Proof. The following proof is taken from Priestley p20.
If
∞
n=0
na
n
z
n−1
had radius of convergence R ,= 0, then it is
not difﬁcult to show that
∞
n=0
a
n
z
n
has the same radius of
convergence.
Therefore, we prove the converse: If
∞
n=0
a
n
z
n
had radius of
convergence R ,=, then so does
∞
n=0
na
n
z
n−1
.
Fix z with 0 < [z[ < R, and choose ρ such that [z[ < ρ < R.
Then
[na
n
z
n−1
[ =
n
[z[
_
[z[
ρ
_
n
[a
n
ρ
n
[.
6
Hopefully, studying maths has taught you that just because something is obvious doesn’t
mean it’s true!
48
The series
n([z[/ρ)
n
is easily shown to converge, by the ratio
test. (Exercise!).
Just as in the real case, if the series
b
n
converges, then
b
n
→ 0 as n → ∞ and so there exists a constant M such that
[b
n
[ ≤ M for all n. Thus, there exists a constant M such that
for all n
n
_
[z[
ρ
_
≤ M,
thus,
[na
n
z
n−1
[ ≤
M
[z[
[a
n
ρ
n
[,
and so by the comparison test,
na
n
z
n−1
is absolutely conver
gent. Hence, the result.
Now, let’s show that this really is the derivative of the series.
Theorem 10.2
Suppose that f(z) =
∞
n=0
a
n
z
n
has radius of convergence R >
0. Then, f
(z) =
∞
n=0
na
n
z
n−1
, for all [z[ < R.
Proof. Again we follow Priestley’s proof. We know that g(z) =
na
n
z
n−1
is welldeﬁned for [z[ < R. We shall show that f
(z)
exists and is equal to g(z).
[If we are thinking like mathematicians, then we know a
good way of doing this is to show that [f
(z) − g(z)[ = 0. In
other words we want
¸
¸
¸
¸
lim
h→0
f(z + h) −f(z)
h
−g(z)
¸
¸
¸
¸
= 0.
Certainly, this is true if
¸
¸
¸
¸
f(z + h) −f(z)
h
−g(z)
¸
¸
¸
¸
→ 0 as h → 0.
This is our method of attack.]
49
For any z and h such that [z[ < R and [z + h[ < R, we have
¸
¸
¸
¸
f(z + h) −f(z)
h
−g(z)
¸
¸
¸
¸
=
¸
¸
¸
¸
¸
∞
n=0
a
n
(z + h)
n
−z
n
h
−
∞
n=0
na
n
z
n−1
¸
¸
¸
¸
¸
=
¸
¸
¸
¸
¸
∞
n=1
a
n
_
(z + h)
n
−z
n
h
−nz
n−1
_
¸
¸
¸
¸
¸
= [h[
¸
¸
¸
¸
¸
∞
n=1
a
n
n
k=2
_
n
k
_
z
n−k
h
k−2
¸
¸
¸
¸
¸
use binom thm
≤ [h[
∞
n=1
1
2
n(n −1)[a
n
[
n−2
m=2
_
n −2
m
_
[z[
n−2−m
[h[
m
≤ [z[
∞
n=1
1
2
n(n −1)[a
n
[([z[ +[h[)
n−2
.
Fix z and choose ρ with [z[ < ρ < R. By Lemma 10.1
∞
n=1
n(n−
1)[a
n
[ρ
n−2
converges to K say. For [h[ < ρ −[z[,
¸
¸
¸
¸
f(z + h) −f(z)
h
−g(z)
¸
¸
¸
¸
≤
1
2
K[h[.
So as h → 0 we get
¸
¸
¸
f(z+h)−f(z)
h
−g(z)
¸
¸
¸ → 0 as required.
The proofs are quite technical, probably the most technical
in the course, and I wouldn’t expect you to reproduce them
in exams, but I would like you to understand them. These
two results will certainly be very important to the course, so
understand what they mean:
The derivative of a series can be found by termbyterm
differentation and the resulting series has the same radius of
convergence.
Examples 10.3
(i) We can show that
∞
0
z
n
=
1
1−z
for [z[ < 1. So, differentiat
ing both sides gives
∞
0
nz
n−1
=
1
(1−z)
2
for [z[ < 1.
(ii) The series S =
∞
n=1
sin nx
n
2
is absolutely convergent for x ∈
R but termbyterm differentiation gives
∞
n=1
cos nx
n
. This
clearly diverges at x = 0. Thus, S is not a power series!
Exercise 10.4
Prove Theorem 7.11.
50
We have seen that, if f is deﬁned on the disc of convergence,
then so is f
, and f
is a series. We can differentiate f
term
byterm to get f
and so on. This suggests the following:
Corollary 10.5
If f(z) =
∞
n=0
a
n
z
n
has radius of convergence R > 0, then f is
inﬁnitely differentiable.
Proof. By induction using the theorem.
Note that this is to be expected as a similar statement is true
for real power series. This result is one of the reasons that
power series are so great. Later, we will see that any com
plex function differentiable at a point can be given by a power
series.
Corollary 10.6
If f(z) =
∞
n=0
a
n
z
n
has radius of convergence R > 0, then a
k
=
f
(k)
(0)
k!
for all k.
Proof. By induction we can prove
f
(k)
(z) =
∞
n=0
n(n −1) . . . (n −k + 1)a
n
z
n−k
and this of course holds for [z[ < R.
Then,
f
(k)
(z) = (k(k −1) . . . (k −k + 1)a
k
)
+((k + 1)((k + 1) −1) . . . ((k + 1) −k + 1)a
k+1
z) + . . .
If we put z = 0 into this, then we get f
(k)
(0) = k!a
k
.
Again, you should already know this is true for real series.
Lemma 10.7 (Uniqueness Lemma)
Suppose for some R > 0,
∞
n=0
a
n
z
n
=
∞
n=0
b
n
z
n
, for all [z[ < R.
Then, a
n
= b
n
for all n.
Proof. Let f(z) =
∞
n=0
a
n
z
n
=
∞
n=0
b
n
z
n
. Then, by Corollary
10.6,
a
n
=
f
(n)
(0)
n!
and b
n
=
f
(n)
(0)
n!
.
Hence, a
n
= b
n
.
Note that the Uniqueness Lemma is not trivial. A priori we
don’t know that we can equate coefﬁcients for inﬁnite series.
51
Remark 10.8
Just like in Real Analysis it is possible to prove that the prod
uct of two power series S
1
and S
2
is again a power series with
radius of convergence at least the minimum of those for S
1
and
S
2
. Its coefﬁcients can be given in terms of those of S
1
and S
2
.
However, we shall delay proving this as later work will give
a particularly simple proof of this fact. (See Theorem 15.6.)
Power series about points other than zero
So far all our power series have been centred at the origin.
This is rather limiting. Just as in real analysis giving an ex
pansion about a different point is very useful. Let’s do that
now.
Suppose we are interested in the point z
0
∈ C. If we let h be
a complex variable, then in a neighbourhood of z
0
we can get
z
0
+ h to give any point. For example, if h is zero we get z
0
. If
[h[ < R for some R, then we have a disc around z
0
.
Suppose we have f(z
0
+ h) =
∞
0
a
n
h
n
, with the series con
vergent for [h[ < R. I.e. a power series in h about 0, but which
deﬁnes the function f around the point z
0
.
Let z = z
0
+h, well z
0
is a constant and h is a single variable
so z is a variable. By substitution, f(z) =
∞
0
a
n
(z −z
0
)
n
, which
converges for [h[ = [z −z
0
[ < R.
Thus we can deal with power series centred at a particular
point z
0
and we will have a disc of convergence centred at z
0
.
The main point is that we can prove results about power
series centred at zero and just translate to another point. So
for example, if we have the power series
a
n
(z −z
0
)
n
, then we
can let h = z−z
0
to get the series
a
n
h
n
and we can apply ratio
test, etc., to that. Then we translate back to say something
about the series
a
n
(z −z
0
)
n
.
For example, we can differentiate termbyterm:
Example 10.9
If f(z) =
a
n
(z − z
0
)
n
converges for [z − z
0
[ < R, then f
(z) =
na
n
(z −z
0
)
n−1
converges for [z −z
0
[ < R.
52
Proof: Let h = z − z
0
. Then g(h) =
a
n
h
n
converges for all
[h[ = [z −z
0
[ < R as f(z) = g(z −z
0
). We have
f
(z) =
d
dz
g(z−z
0
) = g
(z−z
0
)
d
dz
(z−z
0
) = g
(z−z
0
) =
na
n
(z−z
0
)
n
.
Summary
Let f(z) =
∞
n=0
a
n
z
n
have radius of convergence R > 0.
• f can be differentiated termbyterm to get the derivative.
• f
has the same radius of convergence as f.
• f is inﬁnitely differentiable.
• The coefﬁcients a
n
are unique and a
n
=
f
(n)
(0)
n!
for all n.
• We can translate these results to series of the following
form: f(z) =
∞
n=0
a
n
(z −z
0
)
n
, for [z −z
0
[ < R.
53
11 Win a Million Dollars!
The Riemann Zeta function
We deﬁne the Riemann Zeta function as follows:
ζ(s) =
∞
n=1
1
n
s
.
So, for example if s = 1, then we get ζ(1) =
∞
n=1
1
n
which we
know does not converge, so ζ(1) is not deﬁned. If s = 2 we
get ζ(2) =
∞
n=1
1
n
2
which does converge, and so on for all real
numbers s.
If we let s be complex, then what happens? We ﬁrst we
need to deﬁne n
s
when n is a real number and s is a complex
number, the only number we have done this for is e. We know
that n = e
log n
and so we could deﬁne n
s
by
n
s
= e
s log n
.
(Since obviously we want n
s
= (e
log n
)
s
= e
s log n
.)
Riemann was not the ﬁrst to study this type of function but
did leave us an interesting hypothesis.
The Riemann Hypothesis
We now come to the million dollar question. Where do the
roots of the function lie? I.e. the s ∈ C such that ζ(s) = 0.
What Riemann found is that all roots (apart from some triv
ial real ones) seem to have Re(s) =
1
2
, and so he conjectured
that all nontrivial roots have the property of being on this
line. This is called the Riemann Hypothesis and it is regarded
as being the most important unsolved problem in pure math
ematics.
Anyone who proves that the hypothesis is true can claim a
million dollars from the Clay Institute. You can ﬁnd the details
on the Web at http://www.claymath.org/ under the heading
Millenium Prize Problems.
There are six other problems from various areas of mathe
matics for which a million dollar prize is offered. One of them
(the Poincar´ e Conjecture) is explained in Homotopy and Sur
faces in Year 3.
For futher reading see Prime Time by Erica Klarreich in
New Scientist, Vol 168 issue 2264, 11/11/2000, p32, (Edward
Boyle Library Floor 11), for the connection with number theory
and the distribution of primes.
See also the classic text by E.C. Titchmarsh, The theory of
the RiemannZeta function.
54
12 Winding numbers
Cauchy’s theorem is one of the most remarkable theorems in
mathematics. To state it we need the notions of winding num
ber and interior point. Both these notions are intuitively sim
ple. We start with the winding number.
Deﬁnition 12.1
Let γ be a closed contour and w a point not on γ. Then the
winding number of γ about w, written n(γ, w), is the net num
ber of times that γ winds about w, with anticlockwise counted
positively.
Example 12.2
The winding numbers for points in the regions enclosed by the
contour are shown below.
Exercise 12.3
Find the winding numbers for points in the various regions.
55
We can calculate winding numbers mathematically.
Lemma 12.4 (Winding Number Lemma)
Let γ be a closed contour, w ∈ C¸γ. Then
n(γ, w) =
1
2πi
_
γ
dz
z −w
.
Proof. If we write γ(t) = w + r(t)e
iθ(t)
, then r(t) and θ(t) are
continuous, piecewise continuous differentiable functions on
[a, b]. We have,
_
γ
dz
z −w
=
_
b
a
γ
(t)
γ(t) −w
dt
=
_
b
a
r
(t)e
iθ(t)
+ iθ
(t)r(t)e
iθ(t)
r(t)e
iθ(t)
dt
=
_
b
a
r
(t)
r(t)
dt + i
_
b
a
θ
(t) dt
= [log r(t)]
b
a
+ i [θ(t)]
b
a
= log r(a) −log r(b) + i (θ(a) −θ(b)) .
Now, γ is a closed contour and so γ(a) = γ(b), i.e. r(a)e
iθ(a)
=
r(b)e
iθ(b)
. Equating the moduli and arguments we get r(a) = r(b)
and θ(b) = θ(a) + 2πk for some k ∈ Z.
Putting these results into the integral above,
_
γ
dz
z −w
= 2πik.
Moreover, θ(b)−θ(a) is the net increase in arg(γ(t)−w) as t runs
from a to b, so the number of times that γ winds round w is
θ(b) −θ(a)
2π
, i.e. k = n(γ, w).
Exercises 12.5
(i) Let γ be a closed contour and w ∈ C. Prove that n(−γ, w) =
−n(γ, w).
(ii) Let γ
1
and γ
2
be closed contours, so that the join can be
taken. Prove that n(γ
1
+ γ
2
, w) = n(γ
1
, w) + n(γ
2
, w)
An easy method of calculation
Calculating winding numbers can appear to be complicated to
calculate, you have to trace your ﬁnger round the curve bear
ing in mind how many revolutions have been made. And you
have to do it for each region. Fortunately, there is an easier
56
method. In the examples above note that when passing from
one region to another via an edge (rather than via a crossing
of two line) the winding number for points in the regions only
changes by 1.
Now, consider the complicated contour image drawn below
with a line passing through it.
Start at the left side of the line. Obviously the winding num
ber there is 0. As we go from left to right on the line we will
cross the contour. We apply the following rules:
• If we cross the contour so that it is travelling up, then the
winding number decreases by 1.
• If we cross the contour so that it is travelling down, then
the winding number increases by 1.
In the above diagram, the contour is travelling up when we
ﬁrst meet it, so the point in the region we pass into have
winding number −1. At the next crossing the contour is going
down, and so the winding number of points in the next region
is 0. We can carry this out for all regions on the line. To get
other regions we can use different lines.
Exercise 12.6
Find the winding numbers for the points in the diagram.
Justiﬁcation of the method
We can justify the method by considering two points w and z
that lie on opposite sides of a contour line that goes up. We
can assume that the contour starts and ﬁnishes at a point
near w and z. (If it didn’t we can do some reparametrisation
and join work.)
Now, take a loop C round w. Then, n(γ + C, w) = n(γ, z), as
the diagram shows. Thus,
n(γ + C, w) = n(γ, z)
n(γ, w) + n(C, w) =
n(γ, w) + 1 =
n(γ, w) = n(γ, z) −1.
This shows that if we pass from z to w, then the winding num
ber decreases by 1. Similarly, one can show an increase by 1
for a contour locally heading down.
57
Warning! 12.7
We have been using the image of the contour to calculate the
winding number of a point. Recall that the contour and its
image are different. Effectively, we have assumed that the
contour is traversed only once. If we go round the contour
twice, then the numbers calculated by eye have to be doubled.
More generally, if we go round k times, then we multiply the
‘by eye’ calculations by k.
Interior points
The method of counting the number of times a contour wraps
round a point helps deﬁne the notion of an interior point.
Deﬁnition 12.8
Let γ be a closed contour. The interior of γ is the set of points
w ∈ C¸γ for which n(γ, w) ,= 0.
We denote the set of interior points of γ by Int(γ).
Example 12.9
The interior points of the contour in the next diagram are
shaded.
Example 12.10
Let γ(t) = w + Re
2πit
, 0 ≤ t ≤ 1. Then, Int(γ) = ¦z : [z −w[ < R¦.
Summary
• The winding number is the number of times a contour
wraps round a point in an anticlockwise direction.
• Let γ be a closed contour, w ∈ C¸γ. Then
n(γ, w) =
1
2πi
_
γ
dz
z −w
.
• It is easy to calculate a winding number by eye.
• An interior point is any point with nonzero winding num
ber.
58
13 Cauchy’s (Fantastic) Theorem
We now come to the fundamental theorem in complex analy
sis. There is no analogue in real analysis. It has far reaching,
deep consequences, and most of what we will prove from now
on relies on this theorem.
Theorem 13.1 (Cauchy’s Theorem)
Let D ⊆ C be a domain, and f : D → C be a differentiable
complex function. Let γ be a closed contour such that γ and
its interior points lie in D.
Then,
_
γ
f = 0.
Remarks 13.2
(i) This is truly a great theorem. It refers to any domain in
C, any analytic function on D, and any contour with all
interior points in D. And it says that any integral aris
ing from this is zero. Thus, weak assumptions lead to a
strong conclusion.
(ii) It is important to note that the interior of γ lies within D.
If w is in the interior of γ but not in D, consider f(z) =
1/(z −w) which is analytic on D = C¸¦w¦, but
_
γ
f =
_
γ
dz
z −w
= 2πi n(γ, w) ,= 0,
so Cauchy’s theorem does not apply.
(iii) The proof of theorem is too complicated for the moment
and we will do it later. Note that we can’t just use the FTC
since we don’t know that f has an antiderivative on D.
(iv) At ﬁrst sight it may appear that the theorem will only
tell us about the behaviour of differentiable functions.
However, it has strong implications for nondifferentiable
functions as well.
Exercise 13.3
Using the contours in Exercises 2 Question 1, to which of the
following integrals does Cauchy’s theorem apply? (There is no
need to evaluate them.)
_
γ
1
[z[
2
dz,
_
γ
1
z
2
z −2
dz,
_
γ
2
z dz,
_
γ
2
+γ
3
sin
_
1
z −1
_
dz.
59
The calculation trick
Recall, the trick I performed where I calculated an integral
over a path, before the path had even been deﬁned. How was
this done?
The trick follows from Cauchy’s theorem and the funda
mental example, (recall that the example says that
_
γ
1
z
dz =
2πi where γ is a small circle round the origin).
I was trying to integrate
_
γ
a
z
dz, where a was a number cho
sen at random, and γ was a path chosen at random. However,
you were forced into choosing γ so that its winding number
was 1.
Now, we can take a small circle C going clockwise round
the origin, and we can make it so small that we can assume
it lies totally within the interior of γ. Now take another path
from the end of γ to the start of the circle C. Call this contour
β.
Consider the contour Γ = γ + β + C − β. This is a closed
contour. Its interior does not include the disc encircled by C.
So it doesn’t contain the origin. But a/z is differentiable on
D = C¸¦0¦, and the interior of Γ is a subset of D. Thus, we can
apply Cauchy’s theorem:
_
Γ
a
z
dz = 0
_
γ+β+C−β
a
z
dz = 0
_
γ
a
z
dz +
_
β
a
z
dz
_
C
a
z
dz +
_
−β
a
z
dz = 0
_
γ
a
z
dz +
_
β
a
z
dz
_
C
a
z
dz −
_
β
a
z
dz = 0
_
γ
a
z
dz +
_
C
a
z
dz = 0
_
γ
a
z
dz = −
_
C
a
z
dz
= a
_
−C
1
z
dz
= a 2πi
_
γ
a
z
dz = 2aπi
60
Thus we get the answer expected. This method – of cutting
out a disc where the function is not deﬁned – will be used
again later in more generality so make sure you understand
this simpler example.
Summary
• Let D ⊆ C be a domain, and f : D → C be a differentiable
complex function. Let γ be a closed contour such that γ
and its interior points lie in D.
Then,
_
γ
f = 0.
61
14 Strange Consequences of Cauchy’s Theorem
We will prove a number of surprising theorems that can be
deduced from Cauchy’s theorem.
(i) Cauchy’s Integral Formula: The value of a function at z
0
is determined by the values on a contour round z
0
.
(ii) Liouville’s Theorem: Any differential function bounded on
the whole of C is contant.
(iii) The Maximum Modulus Principle: The modulus of a func
tion on a domain achieves its maximum on the boundary
of the domain.
(iv) Fudamental Theorem of Algebra: Every complex polyno
mial has a complex root.
Cauchy’s Integral Formula
Theorem 14.1
Let D ⊆ C be a domain, and f : D → C be differentiable. Let
γ be a closed contour such that γ and its interior points lie in
D.
If w ∈ D¸γ, then
_
γ
f(z)
z −w
dz = 2πi n(γ, w)f(w).
Remarks 14.2
(i) If w is in the interior of γ, then
f(w) =
1
2πi n(γ, w)
_
γ
f(z)
z −w
dz.
This says that the values of f inside γ are completely de
termined by those on γ! Remarkable! This contrasts with
real analysis.
(ii) This behaviour is sometimes called ‘action at a distance’.
(iii) The special case f(z) = 1 for all z ∈ D gives the winding
number lemma.
Proof (of Theorem 14.1). Let γ
r
be the circular contour γ
r
=
w + re
it
, (0 ≤ t ≤ 2π), where r > 0 is sufﬁciently small so that γ
r
is contained in the interior of γ.
[Step 1] Let β be a contour in D¸¦w¦ from the start point of
γ to the start point of γ
r
.
62
The contour
¯ γ = γ + β + (−γ
r
) + + (−γ
r
)
. ¸¸ .
n(γ,w) times
+(−β),
where there are n(γ, w) copies of (−γ
r
), has winding number
zero about w, so by Cauchy’s theorem applied to
f(z)
z −w
on
D¸¦w¦,
_
e γ
f(z)
z −w
dz = 0
__
γ
+
_
β
−
_
γr
− −
_
γr
+
_
−β
_
f(z)
z −w
dz = 0
_
γ
f(z)
z −w
dz −n(γ, w)
_
γr
f(z)
z −w
dz = 0
_
γ
f(z)
z −w
dz = n(γ, w)
_
γr
f(z)
z −w
dz.
[Step 2] We shall now show that
lim
r→0
_
γr
f(z)
z −w
dz = 2πif(w).
We have,
_
γr
f(z)
z −w
dz = f(w)
_
γr
dz
z −w
+
_
γr
f(z) −f(w)
z −w
dz
= 2πif(w) +
_
γr
f(z) −f(w)
z −w
dz, by winding lemma.
Therefore
¸
¸
¸
¸
_
γr
f(z)
z −w
dz −2πif(w)
¸
¸
¸
¸
=
¸
¸
¸
¸
_
γr
f(z) −f(w)
z −w
dz
¸
¸
¸
¸
≤ sup
z∈γr
¸
¸
¸
¸
f(z) −f(w)
z −w
¸
¸
¸
¸
2πr,
by the Estimation Lemma. Now
f(z) −f(w)
z −w
→ f
(w) as z → w.
(Why does f
(w) exist?) Hence RHS → [f
(w)[.0 = 0 as r → 0.
Therefore,
lim
r→0
_
γr
f(z)
z −w
dz = 2πif(w).
Thus we can combine these steps to get
_
γ
f(z)
z −w
dz −2πin(γ, w)f(w) =
_
n(γ, w)
_
γr
f(z)
z −w
dz
_
−2πin(γ, w)f(w)
= n(γ, w)
__
γr
f(z)
z −w
dz −2πif(w)
_
.
63
The RHS tends to 0 as r → 0, but the LHS is independent of r
and so must be 0.
Examples 14.3
(i) Calculate
_
γ
sin z
z −π
dz, where γ(t) = 3e
−it
, 0 ≤ t ≤ 4π.
We apply CIF with f(z) = sin z and w = π. The point w lies
within the circle formed by γ, and n(γ, w) = −2, because γ
winds round w clockwise twice. (Draw a picture!)
Hence,
_
γ
sin z
z −π
dz = 2πi (−2) sin(π) = −4πi.
(ii) Let γ be a circle of radius 2 about 2, i.e. z such that
[z −2[ = 2.
Draw the contour and indicate where the integrand is not
differentiable.
Then,
_
z−2=2
e
z
z
2
−9
dz =
_
z−2=2
1
6
e
z
z −3
−
1
6
e
z
z + 3
dz
=
1
6
_
z−2=2
e
z
z −3
dz −
1
6
_
z−2=2
e
z
z + 3
dz
=
1
6
2πie
3
−0, by 14.1,
=
iπe
3
3
.
Liouville’s Theorem
The next theorem is also rather surprising.
Theorem 14.4
Suppose f is differentiable on the whole of C, and is bounded,
i.e. there exists M such that [f(z)[ ≤ M for all z ∈ C. Then, f
is constant.
64
Proof. [HTTLAM: We want to show the function is constant.
This is true if f(α) = f(0) for an arbitrary α. So we need to
show [f(α) −f(0)[ = 0 for every α ∈ C.]
Let α ∈ C. We let R ≥ 2[α[ and γ(t) = e
2πit
for 0 ≤ t ≤ 1 be a
circle of radius R round the origin. Then,
[f(α) −f(0)[ =
¸
¸
¸
¸
1
2πi
_
γ
f(z)
z −α
dz −
1
2πi
_
γ
f(z)
z
dz
¸
¸
¸
¸
, by CIF
=
¸
¸
¸
¸
1
2πi
¸
¸
¸
¸
¸
¸
¸
¸
_
γ
f(z)
_
1
z −α
−
1
z
_
dz
¸
¸
¸
¸
=
1
2π
¸
¸
¸
¸
_
γ
f(z)α
z(z −α)
dz
¸
¸
¸
¸
≤
1
2π
M sup
z∈γ
_¸
¸
¸
¸
α
z(z −α)
¸
¸
¸
¸
_
L(γ) by
≤
1
2π
M
[α[
R(R −[α[)
.2πR
=
M[α[
R −[α[
.
But R was effectively arbitrary, so as R → ∞, the LHS→ 0. As
neither f(α) nor f(0) depend on R we must have f(α)−f(0) = 0.
That is, f is constant.
Remark 14.5
Contrast this with real analysis. Being bounded does not im
ply constant, for example, consider sin. This is differentiable
on all R and [ sin x[ ≤ 1 for all x ∈ R but it sin is not constant.
Exercise 14.6
Show cos and sin are not bounded on C.
Maximum Modulus Principle
Theorem 14.7
Let f : D → C be a differentiable function and γ be a closed
contour such that Int(γ) ⊂ D.
If [f(z)[ ≤ M for all z ∈ γ, then [f(w)[ ≤ M for all w ∈ Int(γ).
Remark 14.8
The theorem says that the modulus of a function within the
interior of a contour is never bigger than the modulus of the
function on the contour. In other words the maximum modu
lus occurs on the boundary of a region.
65
Proof. This proof contains a nice trick. We shall apply CIF to
w ∈ Int(γ) and f(z)
k
, where k is a natural number:
f(w)
k
=
1
2πin(γ, w)
_
γ
f(z)
k
z −w
dz.
Deﬁne the distance from w to γ by
dist(γ, w) = inf¦[z −w[ : z ∈ γ¦.
Then, obviously [z −w[ ≥ dist(γ, w) for all z ∈ γ.
So,
[f(w)[
k
≤
1
2π [n(γ, w)[
M
k
dist(γ, w)
L(γ), by the Estimation Lemma,
since
[f(w)[
k
[z −w[
≤
M
k
dist(γ, w)
on γ.
Therefore,
[f(w)[ ≤
_
L(γ)
2π [n(γ, w)[ dist(γ, w)
_
1/k
M.
Now, we use the fact that lim
k→∞
x
1/k
= 1 for all x > 0, (this
is true because x
1/k
= exp(
1
k
ln x) → exp(0) = 1). Thus, letting
k → ∞ gives [f(w)[ ≤ M.
Fundamental Theorem of Algebra
We are now in a position to prove the Fundamental Theorem
of Algebra, (which is in fact not really a theorem of algebra but
of analysis!)
Theorem 14.9
Every polynomial p(z) = z
n
+ a
n−1
z
n−1
+ . . . a
1
z + a
0
, n ≥ 1, has a
root in C.
Proof. Suppose not and derive a contradiction. Since p(z) ,= 0
for all z ∈ C the function f deﬁned by f(z) = 1/p(z) is differen
tiable on all of C. Now, for z ,= 0,
¸
¸
¸
¸
p(z)
z
n
¸
¸
¸
¸
=
¸
¸
¸1 +
a
n−1
z
+ +
a
0
z
n
¸
¸
¸ → 1 as [z[ → ∞.
So there exists an R such that [z[ ≥ R implies that
¸
¸
¸
¸
p(z)
z
n
¸
¸
¸
¸
≥
1
2
.
This in turn means
[f(z)[ =
¸
¸
¸
¸
1
p(z)
¸
¸
¸
¸
≤
2
[z
n
[
≤
2
R
n
for [z[ > R.
66
By the Maximum Modulus Principle applied to f with γ being
the standard circle of radius R round the origin we have [z[ <
R =⇒ [f(z)[ ≤
2
R
n
.
So [f(z)[ ≤
2
R
n
on all of C. By Liouville’s theorem, f is con
stant, and hence so is p. This is a contradiction, so p has a
root.
Summary
• Cauchy’s theorem can be used to prove surprising theo
rems that have no analogues in real analysis.
• Cauchy’s Integral Formula: The value of a function at z
0
is determined by the values on contours round z
0
.
• Liouville’s Theorem: Any differential function bounded on
the whole of C is contant.
• The Maximum Modulus Principle: The modulus of a func
tion on a domain achieves its maximum on the boundary
of the domain.
• Fudamental Theorem of Algebra: Every complex polyno
mial has a complex root.
67
15 Taylor’s Theorem
Taylor’s theorem is a consequence of Cauchy’s theorem but is
so important we give it a separate section.
Theorem 15.1 (Taylor’s theorem for complex functions)
Suppose f : D → C is a differentiable function. Let z
0
∈ D and
R = dist(z
0
, C¸D) (which equals inﬁnity if D = C.) Then, there
exists a Taylor expansion of f about z
0
, i.e. there exist a
n
∈ C
such that
f(z) =
∞
0
a
n
(z −z
0
)
n
, for all [z −z
0
[ < R.
Furthermore,
a
n
=
f
(n)
(z
0
)
n!
=
1
2πi
_
Cr
f(z)
(z −z
0
)
n+1
dz,
where C
r
is a small circle about z
0
and r is any number with
0 < r < R.
Proof. Fix r with 0 < r < R. (We will replace z with z
0
+ h so
that we have a power series in h centred at 0). For each h with
[h[ < r, Cauchy’s Integral Formula gives
f(z
0
+ h) =
1
2πi
_
Cr
f(z)
z −(z
0
+ h)
dz.
Now, with [h[ < r and [z −z
0
[ = r,
f(z)
z −(z
0
+ h)
=
f(z)
(z −z
0
)
_
1 −
h
z−z
0
_ =
f(z)
z −z
0
∞
n=0
_
h
z −z
0
_
n
=
∞
n=0
f(z)h
n
(z −z
0
)
n+1
.
Now we shall integrate termbyterm, (see Corollary 6.7), us
ing g(z) =
f(z)
z −z
0
, and g
k
(z) =
f(z)h
n
(z −z
0
)
n+1
. Then, deﬁne M
k
by
M
k
= sup
z−z
0
=r
[g
k
(z)[ = sup
z−z
0
=r
¸
¸
¸
¸
f(z)h
n
(z −z
0
)
n+1
¸
¸
¸
¸
= sup
z−z
0
=r
[f(z)[
[h[
n
r
n+1
,
so,
∞
k=0
M
k
< ∞ as [h[ < r. (Use the ratio test). Thus, by
68
Corollary 6.7,
_
Cr
f(z)
z −(z
0
+ h)
dz =
∞
k=0
_
Cr
f(z)h
n
(z −z
0
)
n+1
dz
2πif(z
0
+ h) =
∞
k=0
__
Cr
f(z)
(z −z
0
)
n+1
dz
_
h
n
f(z
0
+ h) =
∞
k=0
_
1
2πi
_
Cr
f(z)
(z −z
0
)
n+1
dz
_
h
n
Therefore, if we substitute z = z
0
+h and let a
n
=
1
2πi
_
Cr
f(z)
(z −z
0
)
n+1
dz,
we get the series
f(z) =
∞
n=0
a
n
(z −z
0
)
n
.
That we also have a
n
=
f
(n)
(z
0
)
n!
follows from Corollary 10.6.
The theorem says that there exists an Rneighbourhood of z
0
upon which f is a power series. That is, f is analytic. This
explains why some authors use the word analytic to mean
complex differentiable:
differentiable and analytic are equivalent in complex analysis.
Analytic and differentiable are not equivalent in real analy
sis as we see in a later example.
The expression for a
n
is sometimes known as Cauchy’s for
mula for derivatives.
HTTLAM 15.2
When solving problems concerning differentiable complex func
tions, whether theoretical or practical ones, use the fact that
the function can be written locally as a power series.
Example 15.3
(i) Find the Taylor series of f(z) = (z + i)
3
at z = 0. We have,
f(z) = (z + i)
3
=⇒ f(0) = (0 + i)
3
= −i,
f
(z) = 3(z + i)
2
=⇒ f
(0) = 3(0 + i)
2
= −3,
f
(z) = 6(z + i) =⇒ f
(0) = 6(0 + i) = 6i,
f
(3)
(z) = 6 =⇒ f
(3)
(0) = 6,
f
(4)
(z) = 0 =⇒ f
(4)
(0) = 0.
So,
f(z) = f(0) + f
(0)(z −0) +
f
(0)
2!
(z −0)
2
+
f
(3)
(0)
3!
(z −0)
3
= −i −3z + 3iz
2
+ z
3
.
69
Note that this expansion is valid for all z ∈ C as f is dif
ferentiable on all of C, i.e. R = dist(0, C¸C) = ∞.
(ii) Expand f(z) =
1
1 −z
about 0. It is easy to show that
f
(n)
(z) =
n!
(1−z)
n+1
, so f
(n)
(0) = n!.
Thus,
1
1 −z
= f(z) =
∞
n=0
f
(n)
(0)
n!
(z −0)
n
=
∞
n=0
n!
n!
z
n
=
∞
n=0
z
n
.
This is valid for all [z[ < 1 as f is differentiable on C¸¦1¦
so R = dist(0, C¸ (C¸¦1¦)) = dist(0, ¦1¦) = 1.
Note that this conﬁrms a result we already know.
Now we come to a very surprising and useful theorem, which
contrasts sharply with real differentiability.
Corollary 15.4 (Inﬁnite Differentiability)
Suppose f : D → C is differentiable on the domain D. Then, f
is inﬁnitely differentiable, i.e. it has derivatives of all orders.
Proof. Suppose z
0
is any point in D. From the theorem we
know that, in some neighbourhood of z
0
, f can be given as
a power series, and so is inﬁnitely differentiable by Corollary
10.5.
This is really astounding. It doesn’t happen for real func
tions as the next example illustrates.
Example 15.5
Suppose f : R →R is deﬁned by
f(x) =
_
x
2
x > 0
0 x ≤ 0.
Then, f is differentiable with derivative
f(x) =
_
2x x > 0
0 x ≤ 0.
This derivative is not differentiable at 0, so f is not inﬁnitely
differentiable.
Some facts about power series
We now use Taylor’s Theorem to prove some facts about series
which might otherwise be difﬁcult to demonstrate.
70
Theorem 15.6
Suppose f(z) =
∞
n=0
a
n
z
n
for all [z[ < R
1
and g(z) =
∞
n=0
b
n
z
n
for all [z[ < R
2
. Then, f(z)g(z) = (fg)(z) =
∞
n=0
c
n
z
n
for all
[z[ < min¦R
1
, R
2
¦, where c
n
=
n
k=0
a
k
b
n−k
.
Proof. Let R = min¦R
1
, R
2
¦. Then, f and g are complex dif
ferentiable for all [z[ < R. The product of two differentiable
functions is differentiable by the product rule. So, fg is dif
ferentiable. By Theorem 15.1 it has a series expansion for
[z[ < R, and by Corollary 10.6, we get
(fg)(z) =
∞
n=0
(fg)
(n)
(0)
n!
z
n
.
Now we can apply Leibnitz’s rule
(fg)
(n)
(0) =
n
k=0
n!
k!(n −k)!
f
(k)
(0)g
(n−k)
(0)
=
n
k=0
n!
f
(k)
(0)
k!
g
(n−k)
(0)
(n −k)!
= n!
n
k=0
a
k
b
n−k
.
Therefore,
(fg)(z) =
∞
n=0
c
n
z
n
, ([z[ < R), where c
n
=
n
k=0
a
k
b
n−k
.
Corollary 15.7
Suppose f and g are power series with positive radii of con
vergence. If g(0) ,= 0, then f/g has a power series expansion at
0 with positive radius of convergence.
Proof. Deﬁne h(z) = 1/g(z). By continuity of g and the fact
that g(0) ,= 0, there is a εneighbourhood of 0 upon which h is
differentiable. Thus, h has a power series expansion, and the
result follows from the theorem applied to fh.
Remark 15.8
The precise value of the radius of convergence will depend on
where g is zero.
* Morera’s Theorem
There is a partial converse to Cauchy’s theorem. It relies on
the earlier optional material on the Fundamental Theorem of
Calculus.
71
Theorem 15.9
Let f : D → C be a continuous map on a connected domain D
such that
_
γ
f = 0 for all contours f. Then, f is differentiable.
Proof. We know from Theorem 9.9 that there exists F : D →C
such that F
= f. But by Corollary 15.4 the derivative of a
differentiable function is itself differentiable.
Summary
• If a complex function is differentiable on a domain, then
at every point there is power series expansion valid on
some εneighbourhood.
• Differentiable complex functions are inﬁnitely differentiable.
• If S
1
and S
2
are power series with radii of convergence R
1
and R
2
, then S
1
S
2
is a power series with radius of conver
gence min¦S
1
, S
2
¦.
• If S is a power series at z
0
with positive radius of conver
gence, then 1/S has a power series expansion at z
0
pro
vided S(z
0
) ,= 0.
• There is a partial converse to Cauchy’s Theorem.
72
16 Zeros of functions
We have met a number of functions that are not analytic (i.e.
differentiable) at certain points. For example, 1/z is not ana
lytic at 0, because it is undeﬁned when z is zero. Therefore, it
seems natural to investigate zeros of functions ﬁrst.
Zeros
Deﬁnition 16.1
An analytic function f : D →C has a zero at z
0
if f(z
0
) = 0.
Examples 16.2
(i) The function sin has zeros at kπ, for all k ∈ Z.
(ii) By Theorem 1.35(iv) f(z) = e
z
has no zeros.
(iii) f(z) = z
3
(z
2
+ 1) has a zero at z = i, one at z = −i, and one
at z = 0.
Exercise 16.3
Find the zeros of:
(i) z
2
+ 9, (ii) e
z
2
+9
(z + 4i), (iii) e
z−2
−1.
We now know that if f : D → C is differentiable at z
0
, it has
a Taylor series expansion about z
0
: f(z) =
∞
n=0
a
n
(z − z
0
)
n
for
all z with [z −z
0
[ < R for some R > 0.
Deﬁnition 16.4
We say f has a zero of order m at z
0
if
a
0
= a
1
= = a
m−1
= 0 but a
m
,= 0.
Remark 16.5
Obviously, by Theorem 15.1, f has a zero of order m at z
0
if
and only if f
(j)
(z
0
) = 0 for all j < m and f
(m)
(z
0
) ,= 0.
In particular, if f(z
0
) = 0 and f
(z
0
) ,= 0, then f has a zero of
order 1.
Examples 16.6
(i) The function f(z) = z
2
has a zero of order 2 at 0.
(ii) The function f(z) = z(z +2i)
3
has a zero of order 1 at 0 and
one of order 3 at −2i.
(iii) More generally, suppose that f is a polynomial with a root
of multiplicity m at z
0
. Then, f has a zero of order m at z
0
.
73
Exercise 16.7
Find the zeroes and their orders of the following:
(i) (z−1)
3
(z+1) (ii) (z
2
+1)
2
, (iii) ze
z
2
, (iv) (2z−3i)
4
, (v) z
3
+1.
Exercise 16.8
Suppose that f : D → C is an analytic function with a zero
of order m at z
0
. Then, there exists a differentiable g and an
R > 0, such that f(z) = (z − z
0
)
m
g(z), for all [z − z
0
[ < R, and
g(z
0
) ,= 0.
Summary
• An analytic function f : D →C has a zero at z
0
if f(z
0
) = 0.
• Suppose f(z) =
∞
n=0
a
n
(z − z
0
)
n
for all z with [z − z
0
[ < R
for some R. Then f has a zero of order m at z
0
if
a
0
= a
1
= = a
m−1
= 0 but a
m
,= 0.
74
17 Poles
The notions of pole and residue are crucial to the application
of complex analysis to real problems, such as real integrals,
or Fourier and Laplace Transforms
7
. We deﬁne poles in this
section and show how to locate them in the next. After that
we deﬁne their residue.
A motivating example
Let us consider an example: Integrate
_
z=1
e
z
z
2
dz, where [z[ = 1
denotes the standard contour of a unit circle round the origin.
We can’t use Cauchy’s Integral Formula as the integrand is
not analytic at 0 and the Fundamental Theorem of Calculus
doesn’t help as we can’t see any obvious antiderivative.
Let’s just integrate termbyterm
8
and see what happens.
_
z=1
e
z
z
2
dz =
_
z=1
1
z
2
_
1 + z +
z
2
2!
+
z
3
3!
+ . . .
_
dz
=
_
z=1
_
1
z
2
+
1
z
+
1
2
+
z
6
+ . . .
_
dz
=
_
z=1
1
z
2
dz +
_
z=1
1
z
dz +
_
z=1
1
2
dz +
_
z=1
z
6
dz + . . .
= 0 + 2πi + 0 + 0 + . . . .
Notice that only the 1/z term mattered. By the Fundamen
tal Example, all the other terms were irrelevant. This is key
to understanding the terms pole and residue we are going to
deﬁne.
Laurent expansions
First we shall look at functions, like the integrand e
z
/z
2
exam
ple above, that can be represented by a series with negative
powers.
Recall for an analytic function we can represent it as
f(z) =
∞
n=0
a
n
(z −z
0
)
n
for [z −z
0
[ < R.
Compare this with the following series.
7
These transforms are used in Applied Mathematics. If you know what they are, then ﬁne.
If not, then don’t worry about it.
8
Obviously, we should check that we can do this!
75
Deﬁnition 17.1
Suppose that f : D → C is complex function. Then we say f
has a Laurent expansion at z
0
if there exist a
n
∈ C and R > 0,
such that
f(z) =
∞
n=−∞
a
n
(z −z
0
)
n
for all z with 0 < [z −z
0
[ < R.
Note that the limits go from −∞ to ∞, and that the expansion,
does not necessarily equal f(z
0
) when z = z
0
, (which may be
undeﬁnable anyway).
Example 17.2
(i) f(z) = 1/z has an expansion at 0 with a
n
= 0 for n ,= −1,
a
−1
= 1, and R = ∞.
(ii) The expansion of f(z) = e
z
/z
2
at 0,
e
z
z
2
=
1
z
2
+
1
z
+
1
2
+
z
6
+ . . .
has a
n
=
1
(n + 2)!
for n ≥ 0, a
−1
= a
−2
= 1, and a
j
= 0 for
j ≤ −3. The radius is R = ∞.
(iii) Any power series with a positive radius of convergence is
a Laurent expansion. Hence, any function differentiable
at z
0
has a Laurent expansion.
Exercise 17.3
Find a Laurent expansion for f(z) =
sin z
z
4
deﬁned on C¸¦0¦.
The deﬁnition of a pole
Deﬁnition 17.4
Suppose that f : D → C is a complex function. We say that
f has a pole of order N at z
0
∈ D, if there exists a Laurent
expansion at z
0
of the form
f(z) =
∞
n=−N
a
n
(z −z
0
)
n
(0 < [z −z
0
[ < R), with a
−N
,= 0.
Poles are also called singularities.
If there is no N ≥ 0 such that a
n
= 0 for all n < −N, then we
say f has an essential singularity at z
0
.
Remark 17.5
The point is that for a pole of ﬁnite order there are only a ﬁnite
number of negative exponents of z in the series.
76
Examples 17.6
(i) order(1/z
5
) = 5 at 0.
(ii) order(e
z
/z
2
) = 2 at 0 as f(z) =
1
z
2
+
1
z
+
1
2
+
z
6
+ . . . .
(iii) order
_
i
(z −4i)
3
+ (z −4i)
2
−2i(z −4i)
5
_
= 3 at 4i.
(iv) f(z) = exp(1/z) has an essential singularity at 0 because
f(z) =
∞
n=0
(1/z)
n
n!
=
∞
n=0
z
−n
n!
=
0
n=−∞
z
n
(−n)!
,
and this obviously has an inﬁnite number of terms with
negative exponent.
Remarks 17.7
(i) If such a Laurent series exists, then the coefﬁcients are
unique. To see this look at (z − z
0
)
N
f(z), this is a power
series and so has unique coefﬁcients.
(ii) Suppose that f(z) has a Laurent expansion with N = 0,
i.e.
f(z) = a
0
+ a
1
(z −z
0
) + a
2
(z −z
0
)
2
+ . . . for 0 < [z −z
0
[ < R.
Then, we can deﬁne f(z
0
) = a
0
and thus make f analytic
at z
0
.
If we can do this, we say f has a removable singularity.
Example 17.8
The function f(z) =
sin z
z
is deﬁned on C¸¦0¦, and we have
f(z) =
sin z
z
=
1
z
_
z −
z
3
3!
+
z
5
5!
−. . .
_
= 1 −
z
2
3!
+
z
4
5!
−. . . .
So deﬁne f(0) = 1 to make f analytic on all of C. That is,
f(z) =
_
sin z
z
, for z ,= 0,
1, for z = 0.
Exercise 17.9
Show that (e
z
−1)/z has a removable singularity.
77
Summary
• An expansion of f of the form
f(z) =
∞
n=−∞
a
n
(z −z
0
)
n
for all z with 0 < [z −z
0
[ < R, is called a Laurent expansion
at z
0
.
• We say that f has a pole of order N at z
0
∈ D, if there
exists a Laurent expansion at z
0
of the form
f(z) =
∞
n=−N
a
n
(z −z
0
)
n
(0 < [z −z
0
[ < R), with a
−N
,= 0.
78
18 How to ﬁnd poles
It is hard to give nontrivial examples of poles without using
some theory. You may have noticed that the order is somehow
connected with the order of multiplicity of the polynomial in a
denominator. The following puts that intuition into a precise
mathematical statement, and does it for general functions not
just polynomials.
Lemma 18.1
Suppose that f(z) =
p(z)
q(z)
, where
(i) p is analytic at z
0
, and p(z
0
) ,= 0,
(ii) q is analytic at z
0
, and has a zero of order N at z
0
.
Then, f has a pole of order N at z
0
.
Proof. By assumption, we have q(z) = (z − z
0
)
N
r(z) where r(z)
is analytic with r(z
0
) ,= 0 (see Exercise 16.8). Then, g(z) =
p(z)
r(z)
is analytic at z
0
, so by Theorem 15.1 it has a power series
expansion:
g(z) =
∞
n=0
a
n
(z −z
0
)
n
, valid for [z −z
0
[ < R, some R > 0.
Thus,
f(z) =
p(z)
q(z)
=
p(z)
(z −z
0
)
N
r(z)
=
g(z)
(z −z
0
)
N
=
1
(z −z
0
)
N
∞
n=0
a
n
(z −z
0
)
n
=
a
0
(z −z
0
)
N
+
a
1
(z −z
0
)
N−1
+
a
2
(z −z
0
)
N−2
+ . . .
But,
a
0
= g(z
0
) =
p(z
0
)
r(z
0
)
,= 0,
because p(z
0
) ,= 0 by assumption and r(z
0
) ,= 0 by deﬁnition.
So as a
0
,= 0, we deduce that f has a pole of order N at z
0
.
79
Common Error 18.2
The above lemma gives a method for ﬁnding a pole and its
order. The deﬁnition of a pole is that given in Deﬁnition 17.4.
It is not the multiplicity of the denominator in an expression.
Do not confuse the deﬁnition of an object with a process by
which we ﬁnd the object.
Remark 18.3
Recall that a zero of order N of a polynomial is just a root of
multiplicity N.
Example 18.4
(i) The function
sin z
(z −3)
2
has a pole of order 2 at z = 3, since
sin is analytic and sin 3 ,= 0, and (z − 3)
2
is analytic with a
root of multiplicity 2 at 3.
(ii) The quotient
z(z −1)
4
(z
2
−2z + 5)
2
has poles of order 2 at z = 1±2i.
This is because we can factorize the polynomial z
2
−2z +5
to get (z − (1 + 2i))(z − (1 − 2i)). The resulting roots are
repeated as
(z
2
−2z + 5)
2
= [(z −(1 + 2i))]
2
[(z −(1 −2i))]
2
.
Exercises 18.5
Find the poles and their orders for
(a)
1 + e
z
z(z
2
−1)
, (b)
1 + z
z
3
−2z
2
, (c)
e
z
2
−1
z
5
(It ain’t 5!)
Simple poles
Poles of order 1 are given a special name.
Deﬁnition 18.6
A pole of order 1 is called a simple pole.
Example 18.7
(i) 1/z has a simple pole at 0.
(ii)
sin z
z
2
has a simple pole at 0:
sin z
4z
2
=
z −
z
3
3!
+
z
5
5!
−. . .
4z
2
=
1
4z
−
z
3!4
+
z
3
5!4
−. . . .
80
There is a useful way of locating simple poles.
Theorem 18.8
Suppose f(z) =
p(z)
q(z)
where,
(i) p is analytic at z
0
, and p(z
0
) ,= 0,
(ii) q is analytic at z
0
, q(z
0
) = 0 and q
(z
0
) ,= 0.
Then, f has a simple pole at z
0
.
Proof. By Remark 16.5 q has a zero of order 1 at z
0
. Thus,
by Lemma 18.1, f has a pole of order 1 at z
0
, i.e. the pole is
simple.
Examples 18.9
(i) The function f(z) =
cos z
z
has a simple pole at 0. We have
p(z) = cos z and q(z) = z, so p(0) = 1, q(0) = 0, and q
(0) = 1.
(ii) The function f(z) =
z + 3
sin z
has a simple pole at 0. We have
p(z) = z + 3 and q(z) = sin z, so p(0) = 3, q(0) = 0, and
q
(0) = cos 0 = 1.
Summary
• Suppose that f(z) =
p(z)
q(z)
, where
(i) p is analytic at z
0
, and p(z
0
) ,= 0,
(ii) q is analytic at z
0
, and has a zero of order N at z
0
.
Then, f has a pole of order N at z
0
.
• A simple pole is a pole of order 1.
81
19 Residues
We saw earlier that to integrate
e
z
z
2
=
1
z
2
+
1
z
+
1
2
+
z
6
. . .
we could use the Laurent expansion about zero to integrate
termbyterm. The only information that mattered was the
coefﬁcient of the z
−1
term. Let us now give that term an ofﬁcial
name.
Deﬁnition 19.1
Suppose that f has a pole at z
0
. The residue of f at z
0
is
the coefﬁcient a
−1
in the expansion
∞
n=−N
a
n
(z − z
0
)
n
. This is
denoted res(f, z
0
).
Examples 19.2
(i) res(e
z
/z
2
, 0) = 1.
(ii) res(z
2
, 0) = 0. (The residue is always 0 at differentiable
points.)
(iii) From earlier,
sin z
4z
2
=
1
4z
−
z
3!4
+
z
3
5!4
−. . .
so res
_
sin z
4z
2
, 0
_
=
1
4
.
Remark 19.3
The residue at a point is unique because the coefﬁcients of a
Laurent expansion with a ﬁnite number of negative terms are
unique.
The next theorem gives a hint of where we are going with
residues: They allow us to quickly calculate integrals. This
theorem is really just generalising the example of e
z
/z
2
used
at the start of the section on poles.
Theorem 19.4 (Cauchy’s Residue Theorem for a Circle)
Suppose that the analytic function f : D → C has a pole at p.
Let γ(p, r) denote the circle of radius r round p. Then, there
exists R > 0 such that
_
γ(p,r)
f(z) dz = 2πi res(f, p),
for all 0 < r < R.
82
Proof. At p the function f has a Laurent expansion f(z) =
∞
n=−N
a
n
(z −p)
n
, for 0 < [z −p[ < R, some R > 0. Therefore, let
r be such that 0 < r < R, and then
_
γ(p,r)
f(z) dz =
_
γ(p,r)
∞
n=−N
a
n
(z −p)
n
dz
=
∞
n=−N
_
γ(p,r)
a
n
(z −p)
n
dz.
The latter equality follows from Theorem 6.7 (you can check
this!). But, by the Fundamental Example,
_
γ(p,r)
a
n
(z −p)
n
dz = a
n
_
γ(p,r)
(z −p)
n
dz
=
_
a
−1
2πi, n = −1
0, n ,= −1.
Thus,
_
γ(p,r)
f(z) dz = 2πia
−1
= 2πi res(f, p).
Remark 19.5
The theorem says that to calculate certain integrals all we
have to do is calculate the residue of a function at a point.
Example 19.6
Let γ be the unit circle round the origin. We take f(z) = e
z
/z
2
from earlier, then res(f, 0) = 1 has already been calculated. So
_
γ
e
z
/z
2
= 2πi res(f, 0) = 2πi, as we have already seen.
We shall generalise the theorem further, but in the next
section we will see investigate some methods for calculating
residues.
Summary
• The residue of f at z
0
is the coefﬁcient a
−1
in the expan
sion
∞
n=−N
a
n
(z −z
0
)
n
. This is denoted res(f, z
0
).
• Suppose that the analytic function f : D →C has a pole at
p. Let γ(p, r) denote the circle of radius r round p. Then,
there exists R > 0 such that
_
γ(p,r)
f(z) dz = 2πi res(f, p),
for all 0 < r < R.
83
20 Bluffer’s Guide to Calculating Residues
Residues are an important part of any course on Complex
Analysis. They allow us to easily calculate otherwise com
plicated integrals – without doing any integration!
This section contains the main methods for calculating the
residues of poles. You should know them and their proofs.
Method 1: Simple Poles
Suppose f has a simple pole (i.e. order 1) at w then
res(f, w) = lim
z→w
(z −w)f(z).
Proof. As f has a simple pole at w we have
f(z) =
∞
n=−1
a
n
(z −w)
n
.
So,
lim
z→w
(z −w)f(z)
= lim
z→w
(z −w)
∞
n=−1
a
n
(z −w)
n
= lim
z→w
(z −w)
_
a
−1
z −w
+ a
0
+ a
1
(z −w) + a
2
(z −w)
2
+ . . .
_
= lim
z→w
a
−1
+ a
0
(z −w) + a
1
(z −w)
2
+ a
2
(z −w)
3
+ . . .
= a
−1
= res(f, w).
Examples 20.1
(i) The function f(z) =
sin(z)
z −3
has a pole of order 1 at z = 3, so
res(f, 3) = lim
z→3
(z−3)f(z) = lim
z→3
(z−3)
sin(z)
z −3
= lim
z→3
sin(z) = sin(3).
(ii) The function f(z) =
z
1 −cos z
has a pole of order 1 at z = 0,
so
res(f, 0) = lim
z→0
z
z
1 −cos z
= lim
z→0
z
2
2 sin
2
(z/2)
= lim
z→0
4 (z/2)
2
2 sin
2
(z/2)
= 2 lim
z→0
_
z/2
sin(z/2)
_
2
= 2
_
lim
z→0
z/2
sin(z/2)
_
2
= 2.
84
Method 2: Some quotients (Really good method!)
Suppose f(z) =
p(z)
q(z)
with p and q analytic, p(w) ,= 0, q(w) = 0
and q
(w) ,= 0. Then,
res(f, w) =
p(w)
q
(w)
.
Proof. By Theorem 18.8 f has a simple pole at w, so
res(f, w) = lim
z→w
(z −w)f(z), by Method 1,
= lim
z→w
(z −w)
p(z)
q(z)
= lim
z→w
p(z)
__
q(z)
z −w
_
= lim
z→w
p(z)
__
q(z) −q(w)
z −w
_
, as q(w) = 0,
=
lim
z→w
p(z)
lim
z→w
q(z) −q(w)
z −w
=
p(w)
q
(w)
, by deﬁnition of differentiation.
Remark 20.2
The conditions imply that f has a simple pole, so the method
will only work in this case.
Examples 20.3
(i) The function f(z) =
1
1 −z
3
has a pole at z = 1. We have
p(z) = 1 and q(z) = 1 − z
3
so q
(1) = −3 1
2
= −3. Thus
res(f, 1) =
p(1)
q
(1)
=
1
−3
= −
1
3
.
(ii) The function f(z) =
2z
2
1 + z
4
has a pole at z = e
iπ/4
(and
more besides, but we’ll ignore them). Let p(z) = 2z
2
and
q(z) = 1 + z
4
. Then p(e
iπ/4
) = 2e
iπ/2
,= 0; q
(z) = 4z
3
so
q
(e
iπ/4
) = 4e
3iπ/4
,= 0. So
res
_
f, e
iπ/4
_
=
p(e
iπ/4
)
q
(e
iπ/4
)
=
2e
iπ/2
4e
3iπ/4
=
e
−iπ/4
2
=
1
4
(1 −i)
√
2.
85
Method 3: Poles of order N
We now generalise Method 1 (which is the case N = 1 in the
following).
Suppose that f has a pole of order N at w. Then
res(f, w) =
1
(N −1)!
lim
z→w
d
N−1
dz
N−1
_
(z −w)
N
f(z)
_
.
Proof. By assumption f(z) =
∞
n=−N
a
n
(z −w)
n
for some a
i
∈ C
and 0 < [z −w[ < R. Thus,
f(z) =
∞
n=−N
a
n
(z −w)
n
(z −w)
N
f(z) = (z −w)
N
_
a
−N
(z −w)
N
+ +
a
−1
z −w
+ a
0
+ a
1
(z −w) + . . .
_
= a
−N
+ a
−N+1
(z −w) + + a
−1
(z −w)
N−1
+a
0
(z −w)
N
+ . . . .
Therefore,
1
(N −1)!
lim
z→w
d
N−1
dz
N−1
_
(z −w)
N
f(z)
_
=
1
(N −1)!
lim
z→w
d
N−1
dz
N−1
_
a
−N
+ a
−N+1
(z −w) + . . .
+a
−1
(z −w)
N−1
+ a
0
(z −w)
N
+ . . .
_
=
1
(N −1)!
lim
z→w
¦(N −1)! a
−1
+ N! a
0
(z −w) + . . . ¦
=
1
(N −1)!
(N −1)! a
−1
= a
−1
= res(f, w).
Examples 20.4
(i) Let f(z) =
_
z + i
z −i
_
2
. By Lemma 18.1 this has a pole of
order 2 at z = i. Hence,
res(f, i) =
1
(2 −1)!
lim
z→i
d
2−1
dz
2−1
_
(z −i)
2
_
z + i
z −i
_
2
_
= 1. lim
z→i
d
dz
(z + i)
2
= lim
z→i
2(z + i)
= 2(i + i)
= 4i.
86
(ii) The function f(z) =
e
e
z
z
2
has a pole of order 2 at z = 0,
(again by Lemma 18.1). So,
res(f, 0) =
1
1!
lim
z→0
d
dz
z
2
e
e
z
z
2
= lim
z→0
d
dz
e
e
z
= lim
z→0
e
z
.e
e
z
= e.
Method 4: Direct expansion
We can use this as a last resort. We expand functions as power
series, etc., and then calculate coefﬁcients.
Examples 20.5
(i) Let f(z) =
ze
z
(z −1)
3
. This has a pole of order 3 at z = 1. So,
expand about w = 1: (i.e. let z = w + h and take h as our
variable),
f(1 + h) =
(1 + h)e
1+h
(1 + h −1)
3
= (1 + h)
e.e
h
h
3
=
e
h
3
(1 + h)
_
1 + h +
h
2
2!
+
h
3
3!
+ . . .
_
=
e
h
3
_
1 + 2h +
3
2
h
2
+
2
3
h
3
+ . . .
_
=
e
h
3
+
2e
h
2
+
3e
2h
+
2e
3
+ . . .
So res(f, 1) =
3e
2
.
(ii) ∗ Find res
_
π cot πz
z
2
, 0
_
.
We expand about w = 0:
π cot π(0 + h)
(0 + h)
2
=
π cot πh
h
2
=
π
h
2
cos πh
sin πh
=
π
h
2
(1 −(πh)
2
/2! + . . . )
(πh −(πh)
3
/3! + . . . )
=
1
h
3
(1 −π
2
h
2
/2 + . . . )
(1 −π
2
h
2
/6 + . . . )
.
Consider the parts in brackets, this quotient will be a
power series and so we need to work out its coefﬁcients.
87
We need only work out the coefﬁent for h
2
since division
by the h
3
will give the residue coefﬁcient.
We work with the bottom bracket ﬁrst. Suppose that
(1 −π
2
h
2
/6 + . . . )
−1
= b
0
+ b
1
h + b
2
h
2
+ . . . . Then,
_
1 −π
2
h
2
/6 + . . .
_ _
1 −π
2
h
2
/6 + . . .
_
−1
= 1
_
1 −π
2
h
2
/6 + . . .
_ _
b
0
+ b
1
h + b
2
h
2
+ . . .
_
= 1
b
0
+ b
1
h + b
2
h
2
−
π
2
h
2
6
_
b
0
+ b
1
h + b
2
h
2
_
+ . . . = 1
b
0
+ b
1
h +
_
b
2
−
π
2
6
b
0
_
h
2
+ . . . = 1.
Equating coefﬁcients gives b
0
= 1, b
1
= 0 and b
2
−
π
2
6
b
0
= 0,
i.e. b
2
=
π
2
6
.
Therefore,
π cot πh
h
2
=
1
h
3
_
1 −
π
2
h
2
2
+ . . .
__
1 −
π
2
h
2
6
+ . . .
_
−1
=
1
h
3
_
1 −
π
2
h
2
2
+ . . .
__
1 +
π
2
h
2
6
+ . . .
_
=
1
h
3
_
1 −
π
2
3
h
2
+ . . .
_
=
1
h
3
−
π
2
3h
+ . . .
Thus res(f, 0) = −
π
2
3
.
Obviously, this method can get a bit calculation heavy.
Summary
• Simple pole: res(f, w) = lim
z→w
(z −w)f(z).
• Some quotients: Suppose p(w) ,= 0, q(z) = 0 and q
(w) ,= 0.
Then,
res
_
p
q
, w
_
=
p(w)
q
(w)
.
• Pole of order N:
res(f, w) =
1
(N −1)!
lim
z→w
d
N−1
dz
N−1
_
(z −w)
N
f(z)
_
.
• Calculate expansions directly, equate coefﬁcients, etc.
88
21 Cauchy’s Residue Formula
The next theorem shows that we can calculate certain com
plex integrals by merely calculating the residues and winding
numbers at poles. Thus we have a very simple ﬁveﬁnger
exercise method for calculating integrals. Hooray!
Theorem 21.1 (Cauchy’s Residue Formula)
Let D be domain and γ a closed contour such that Int(γ) ⊆ D.
Let ¦p
1
, . . . , p
m
¦ ∈ D¸γ and f : D → C be analytic with poles at
p
1
, . . . , p
m
. Then
_
γ
f(z) dz = 2πi
m
j=1
n(γ, p
i
) res(f, p
j
).
Proof. We can assume that all the poles are in the interior of
γ. Now, consider the following sketch.
Let C
j
be a circle of radius ε > 0 at p
j
, taken clockwise.
Then, there exists ε > 0 such that C
j
⊆ Int(γ) and C
j
∩ C
l
= ∅
for all j and l with j ,= l and so that C
j
lies within the disc
upon which the Laurent expansion for f at p
j
holds.
There exist contours β
j
from some point on the image of γ
to the start of C
j
. We can deﬁne γ
j
as the part of γ from the
start of β
j
to β
j+1
. So γ = γ
1
+ + γ
m
.
Next, we deﬁne a contour ¯ γ by
¯ γ = β
1
+ n(γ, p
1
)C
1
−β
1
+ γ
1
+β
2
+ n(γ, p
2
)C
2
−β
2
+ γ
2
+ . . .
+β
m
+ n(γ, p
m
)C
m
−β
m
+ γ
m
.
This is a path such that f is analytic on Int(¯ γ), (just like in
the calculation trick I did.) So by Cauchy’s Theorem we have
_
e γ
f = 0. But
_
e γ
f = n(γ, p
1
)
_
C
1
f + n(γ, p
2
)
_
C
2
f + + n(γ, p
m
)
_
Cm
f +
_
γ
1
+...γm
f
0 =
_
m
j=1
n(γ, p
j
)
_
C
j
f
_
+
_
γ
1
+...γm
f.
Note that the integrals over all the β
j
cancel. So, as γ = γ
1
+
89
+ γ
m
we get,
0 =
_
m
j=1
n(γ, p
j
)
_
C
j
f
_
+
_
γ
f
−
m
j=1
n(γ, p
j
)
_
C
j
f =
_
γ
f
m
j=1
n(γ, p
j
)
_
−C
j
f =
_
γ
f.
But, by Theorem 19.4,
_
−C
j
f = 2πi res(f, p
j
). Thus the proof is
complete.
There we have it, one of the best theorems in mathematics!
Examples 21.2
Evaluate the integral
_
γ
1
z
2
+ (i −1)z −i
dz, where γ is a square
with sides of length 4 centred at the origin, oriented anti
clockwise.
Solution: Let f(z) =
1
z
2
+ (i −1)z −i
=
1
(z + i)(z −1)
. There
fore, f has simple poles at z = 1 and z = −i. Let q(z) =
z
2
+(i−1)z−i, then q
(z) = 2z+i−1. We have q
(1) = 2+i−1 = 1+i,
and q
(−i) = −2i + i −1 = −1 −i. Thus,
res(f, 1) =
1
1 + i
=
1 −i
2
and res(f, −i) =
1
−1 −i
=
−1 + i
2
.
We have n(γ, 1) = n(γ, −i) = 1, so by Cauchy’s Residue Theorem
_
γ
1
z
2
+ (i −1)z −i
dz = 2πi ( res(f, 1) + res(f, −i))
= 2πi
_
1 −i
2
+
−1 + i
2
_
= 0.
Summary
• Let D be domain and γ a closed contour such that Int(γ) ⊆
D. Let ¦p
1
, . . . , p
m
¦ ∈ D¸γ and f : D → C be analytic with
poles at p
1
, . . . , p
m
. Then
_
γ
f(z) dz = 2πi
m
j=1
n(γ, p
i
) res(f, p
j
).
90
22 Evaluation of Deﬁnite Real Integrals
We shall now see that complex analysis allows us to solve, in
a simple manner, problems involving real functions that are
not easy to solve with real methods.
The basic idea
We want to calculate
_
∞
−∞
f(x) dx where f is a real function. The
basic method is actually quite simple, but it has a number of
parts, so may initially appear complicated. First, I’ll outline
the ideas behind it.
Let us note that we can calculate the real integral as a con
tour integral. Let Γ
1
(t) = t, then Γ
(t) = 1 for −R ≤ t ≤ R.
So,
_
Γ
1
f(z)dz =
_
R
−R
f(t).1 dt =
_
R
−R
f(t) dt.
If we let R → ∞, then we usually obtain the real integral
_
∞
−∞
f(x) dx (but see Remark 22.3 later).
This doesn’t really get us much further forward. However,
if we take a contour Γ
R
from R to −R via a path in the complex
plane and join it to Γ
1
, then we get a closed contour. Thus, we
can use Cauchy’s Residue Theorem to calculate the integral.
Consider the following diagram in the complex plane, where
Γ
R
is the semicircle in the upper halfplane.
If we integrate round the contour Γ = Γ
1
+ Γ
R
, then we get
_
Γ
f(z) dz =
_
Γ
1
f(z) dz +
_
Γ
R
f(z) dz.
Now, letting R → ∞, we get
lim
R→∞
_
Γ
f(z) dz = lim
R→∞
_
Γ
1
f(z) dz + lim
R→∞
_
Γ
R
f(z) dz
= lim
R→∞
_
R
−R
f(t) dt + lim
R→∞
_
Γ
R
f(z) dz
=
_
∞
−∞
f(x) dx + lim
R→∞
_
Γ
R
f(z) dz.
Now, the LHS is a contour integral for a closed contour and
so by Cauchy’s Residue Formula we can probably work it out
using some method. The RHS contains the integral we want.
Great! It also contains another term. Not so great! However,
if that term is zero, then we have a way of calculating a real
integral from residues!
91
Can we show that the integral
_
Γ
R
f(z) dz → 0 as R → ∞?
The following lemma gives us a supply of examples with this
property.
Lemma 22.1 (Jordan’s Lemma)
Suppose p and q are polynomials with deg(p) ≤ deg(q) − 2,
Γ
R
(t) = Re
it
, 0 ≤ t ≤ π is a semicircular contour of radius
R, and a ≥ 0.
Then,
_
Γ
R
p(z)
q(z)
e
iaz
dz → 0 as R → ∞.
Proof. As usual, we apply the Estimation Lemma. We have,
¸
¸
e
iaz
¸
¸
= e
Re(iaz)
= e
Re(ia(x+iy))
= e
Re(−ay+iax)
= e
−ay
≤ 1,
the latter inequality holds because y ≥ 0 for z ∈ γ, and because
a ≥ 0.
By the Fundamental Theorem of Algebra, the polynomial p
can be written as p(z) = c
deg p
j=1
(z − α
j
), where c is some con
stant and α
j
is a root of p. Similarly, q(z) = d
deg q
j=1
(z − β
j
), for
some d and β
j
.
We then have,
[p(z)[ = [c[
deg p
j=1
[z −α
j
[ ≤ [c[
deg p
j=1
([z[ +[α
j
[) ≤ [c[
deg p
j=1
(R +[α
j
[),
when z ∈ Γ
R
, (i.e. [z[ = R). Note that the RHS is a polynomial
in R of degree deg p.
Similarly,
[q(z)[ = [d[
deg q
j=1
([z −β
j
[) ≥ [d[
deg q
j=1
(R −[β
j
[),
for z ∈ Γ and R ≥ max¦[β
j
[¦.
Thus, for large R, by the Estimation Lemma,
¸
¸
¸
¸
_
Γ
R
p(z)
q(z)
e
iaz
dz
¸
¸
¸
¸
≤
[c[
deg p
j=1
(R +[α
j
[)
[d[
deg q
j=1
(R −[β
j
[)
πR.
So, if deg p ≤ deg q −2, then the RHS→ 0 as R → ∞. Hence, the
RHS is equal to 0.
Remark 22.2
We include an e
iaz
term as this will be useful in determining
integrals involving sines and cosines.
92
Remark 22.3
The integral
_
∞
−∞
f(x) dx is actually deﬁned to be
lim
b→∞
a→−∞
_
b
a
f(x) dx
and in fact, may be different to
lim
R→∞
_
R
−R
f(x) dx.
Deﬁnition 22.4
The integral lim
R→∞
_
R
−R
f(x) dx is called the principal part of
the integral, and is denoted pv
_
∞
−∞
f(x) dx.
The principal part of an integral may exist even if
_
∞
−∞
f(x) dx
does not. E.g. f(x) = x. However, if both integrals exist, then
they are equal. In all our examples we may as well assume
that both exist, but note that in practice we should check that
this is the case.
Real integrals of the form
_
∞
−∞
p(x)
q(x)
dx
For real integrals of the form
_
∞
−∞
p(x)
q(x)
dx, where p and q are
polynomials with q(x) ,= 0 for all real x, the method is the
following.
(i) Deﬁne Γ = Γ
1
+ Γ
R
, where
Γ
1
(t) = t for −R ≤ t ≤ R,
and
Γ
R
(t) = Re
it
for 0 ≤ t ≤ π,
taking R large enough so that Γ contains all the poles of
p(z)
q(z)
that lie in the upper halfplane.
(ii) Calculate
_
Γ
p(z)
q(z)
dz using Cauchy’s Residue Theorem.
(iii) Show that
_
Γ
R
p(z)
q(z)
dz → 0 as R → ∞, using Jordan’s Lemma.
(iv) Conclude that
_
∞
−∞
f(x) dx and lim
R→∞
_
Γ
f(z) dz are equal.
It is probably best if we do an example.
93
Example 22.5
Find the integral
_
∞
−∞
dx
(x
2
+ 1)
2
.
(i) Take Γ as the join of two contours above.
(ii) Let f(z) =
1
(z
2
+ 1)
2
. Then f(z) =
1
(z
2
+ 1)
2
=
1
(z −i)
2
(z + i)
2
.
So f has poles of order 2 at z = i and z = −i. Only the z = i
pole lies in the interior of Γ (when R > 1).
By Method 3 we get
res(f, i) =
1
(2 −1)!
d
dz
_
(z −i)
2
1
(z −i)
2
(z + i)
2
_¸
¸
¸
¸
z=i
=
d
dz
_
1
(z + i)
2
_¸
¸
¸
¸
z=i
=
_
−2
1
(z + i)
3
_¸
¸
¸
¸
z=i
= −2
1
(i + i)
3
=
1
4i
= −
i
4
.
(iii) By Jordan’s Lemma (as deg q = deg(z
2
+ 1)
2
= 4, deg p =
deg 1 = 0, and a = 0)
_
Γ
R
1
(z
2
+ 1)
2
dz → 0 as R → ∞.
(iv) Thus, as
_
Γ
f(z) dz =
_
Γ
1
f(z) dz +
_
Γ
R
f(z) dz
2πi res(f, i) =
_
R
−R
1
(x
2
+ 1)
2
dx +
_
Γ
R
f(z) dz
2πi
_
−
i
4
_
=
_
∞
−∞
1
(x
2
+ 1)
2
dx + 0 as R → ∞
π
2
=
_
∞
−∞
1
(x
2
+ 1)
2
dx
Can you calculate this integral using the standard methods of
real analysis?
94
Real integrals of the form
_
∞
−∞
_
sin x
cos x
_
p(x)
q(x)
dx
We can apply a similar method to integrals of the form
_
∞
−∞
_
sin x
cos x
_
p(x)
q(x)
dx
by making a good choice of complex function to integrate.
Let us try an example that really shows the power of com
plex analysis.
Example 22.6
Calculate the integral
_
∞
−∞
cos x
x
2
−2x + 2
dx.
The key here is to chose f(z) =
e
iz
z
2
−2z + 2
. (Remember that
e
iz
= cos z + i sin z so we will get the cos x, but we will also get a
sin x term along the real axis. Surprisingly, this turns out to
be a bonus, not a problem  just watch!)
(i) Let Γ be as in the previous example, so Γ = Γ
1
+ Γ
R
.
(ii) We want to evaluate
_
Γ
f(z) dz
ﬁrst, so we use Cauchy’s Residue Theorem.
We know that e
iz
has no zeros and no poles, so all the poles
arise from zeros of the denominator: z
2
−2z + 2.
z
2
−2z + 2 = 0 ⇐⇒ z =
2 ±
√
4 −8
2
= 1 ±i.
Let R >
√
2, so 1+i is inside the contour, 1−i is outside it. The
diagram shows the location of poles and the contour Γ
R
.
The pole is simple so
_
Γ
e
iz
z
2
−2z + 2
dz = 2πi res
_
e
iz
z
2
−2z + 2
, 1 + i
_
= 2πi
e
i(1+i)
2(1 + i) −2
= 2πi
e
−1+i
2i
= πe
−1
e
i
=
π
e
(cos 1 + i sin 1).
95
(iii) By Jordan’s Lemma
_
Γ
R
e
iz
z
2
−2z + 2
dz → 0 as R → ∞.
(iv) Now we shall calculate the integrals along the two parts
of the contour. Along Γ
1
we get
_
Γ
1
e
iz
z
2
−2z + 2
dz =
_
R
−R
e
ix
x
2
−2x + 2
dx
=
_
R
−R
cos x
x
2
−2x + 2
dx + i
_
R
−R
sin x
x
2
−2x + 2
dx.
Therefore,
lim
R→∞
_
Γ
f(z) dz = lim
R→∞
_
Γ
1
f(z) dz + lim
R→∞
_
Γ
R
f(z) dz
= lim
R→∞
__
R
−R
cos x
x
2
−2x + 2
dx + i
_
R
−R
sin x
x
2
−2x + 2
dx
_
+ 0
π
e
(cos 1 + i sin 1) =
_
∞
−∞
cos x
x
2
−2x + 2
dx + i
_
∞
−∞
sin x
x
2
−2x + 2
dx.
Equating real parts we get
_
∞
−∞
cos x
x
2
−2x + 2
dx =
π
e
cos 1.
Remark 22.7
Note that using the imaginary parts we get
_
∞
−∞
sin x
x
2
−2x + 2
dx =
π
e
sin 1.
Thus, we have gained more information than we were looking
for. What a method!
Summary
(i) Deﬁne Γ = Γ
1
+ Γ
R
by
Γ
1
(t) = t for −R ≤ t ≤ R, and Γ
R
(t) = Re
it
for 0 ≤ t ≤ π.
Take R large enough to include the poles of f that lie in
the upper halfplane.
(ii) Calculate
_
Γ
f(z) dz using Cauchy’s Residue Theorem.
(iii) Show that
_
Γ
R
f(z) dz → 0 as R → ∞, using Jordan’s Lemma.
(iv) Relate
_
∞
−∞
f(x) dx to
_
Γ
f(z) dz.
96
23 Integrals with sines and cosines
Let us consider the nasty integral
_
2π
0
e
cos θ
cos(nθ −sin nθ) dθ.
from Section 3.
If we consider contour integration, then is there some func
tion and contour such that
_
γ
f gives the above? Let γ(t) = e
iθ
,
for 0 ≤ θ ≤ 2π and f(z) = e
z
. Then, with our knowledge of poles
and residues we see that
_
γ
e
z
z
n+1
dz = 2πi res(e
z
/z
n+1
, 0) =
2πi
n!
So
_
γ
e
z
z
n+1
dz =
_
2π
0
e
e
iθ
e
iθ(n+1)
ie
iθ
dθ
2πi
n!
= i
_
2π
0
e
e
iθ
e
−iθn
dθ
2π
n!
=
_
2π
0
e
e
iθ
−inθ
dθ
=
_
2π
0
e
cos θ+i sin θ−inθ
dθ
=
_
2π
0
e
cos θ
e
i(sin θ−nθ)
dθ
=
_
2π
0
e
cos θ
(cos(sin θ −nθ) + i sin((sin θ −nθ)) dθ.
By equating real and imaginary parts we see that
_
2π
0
e
cos θ
cos(sin θ −nθ) dθ =
2π
n!
and because cos is an even function we deduce that
_
2π
0
e
cos θ
cos(nθ −sin θ) dθ =
2π
n!
Rather spectacular, wouldn’t you agree?
Integrals of the form
_
2π
0
f(cos θ, sin θ) dθ
Theorem 23.1
Let γ(t) = e
it
, for 0 ≤ t ≤ 2π. Then
_
2π
0
f(cos θ, sin θ) dθ = −i
_
γ
f
_
z + z
−1
2
,
z −z
−1
2i
_
z
−1
dz.
97
Proof. Let z = e
iθ
= γ(θ). Then cos θ =
z + z
−1
2
and sin θ =
z −z
−1
2i
. We also have γ
(θ) = ie
iθ
.
Thus,
−i
_
γ
f
_
z + z
−1
2
,
z −z
−1
2i
_
z
−1
dz = −i
_
2π
0
f(cos θ, sin θ)e
−iθ
ie
iθ
dθ
=
_
2π
0
f(cos θ, sin θ) dθ.
Remark 23.2
We can change the limits of integration by letting γ(t) = e
it
where a ≤ t ≤ a + 2π and a ∈ R. So we get integrals of the form
_
a+2π
a
f(cos θ, sin θ) dθ.
A common example is a = −π, so we get
_
π
−π
.
Example 23.3
Evaluate the real integral
_
2π
0
1
13 + 12 cos t
dt.
Solution: Applying the theorem we get, for γ(t) = e
it
, 0 ≤ t ≤
2π,
_
2π
0
1
13 + 12 cos t
dt = −i
_
γ
1
z (13 + 6 (z + 1/z))
dz
= −i
_
γ
1
6z
2
+ 13z + 6
dz.
Let g(z) =
1
6z
2
+ 13z + 6
. This has poles at −
3
2
and −
2
3
. Of
these, only −
2
3
lies within the unit circle given by γ. Hence, we
calculate the residue for this pole:
res
_
g, −
2
3
_
=
1
12z + 13
¸
¸
¸
¸
z=−2/3
=
1
12
_
−
2
3
+ 13
_
+ 13
=
1
5
.
Therefore, by Cauchy’s Residue Theorem,
_
2π
0
1
13 + 12 cos t
dt = −i 2πi
1
5
=
2π
5
.
Summary
• Let γ(t) = e
it
, for 0 ≤ t ≤ 2π. Then
_
2π
0
f(cos θ, sin θ) dθ = −i
_
γ
f
_
z + z
−1
2
,
z −z
−1
2i
_
z
−1
dz.
98
Deﬁnition 1.5 Let D be a domain in C. A complex function, denoted f : D → C, is a map which assigns to each z in D an element of C, this value is denoted f (z). Common Error 1.6 Note that f is the function and f (z) is the value of the function at z. It is wrong to say f (z) is a function, but sometimes people do. Examples 1.7 (i) Let f (z) = z 2 for all z ∈ C. (ii) Let f (z) = z for all z ∈ C. Note that here we have a complex function for which every value is real. (iii) Let f (z) = 3z 4 − (5 − 2i)z 2 + z − 7 for all z ∈ C. All complex polynomials give complex functions. (iv) Let f (z) = 1/z for all z ∈ C\{0}. This function cannot be extended to all of C. Remark 1.8 Functions such as sin x for x real are not complex functions since the real line in C is not a domain. Later we see how to extend the concept of the sine so that it is complex function on the whole of the complex plane. Obviously, if f and g are complex functions, then f + g, f − g, and f g are functions given by (f + g)(z) = f (z) + g(z), (f − g)(z) = f (z) − g(z), and (f g)(z) = f (z)g(z), respectively. We can also deﬁne (f /g)(z) = f (z)/g(z) provided that g(z) = 0 on D. Thus we can build up lots of new functions by these elementary operations. The aim of complex analysis We wish to study complex functions. Can we deﬁne differentiation? Can we integrate? Which theorems from Real Analysis can be extended to complex analysis? For example, is there a version of the mean value theorem? Complex analysis is essentially the attempt to answer these questions. The theory will be built upon real analysis but in many ways it is easier than real analysis. For example if a complex function is differentiable (deﬁned later), then its derivative is also differentiable. This is not true for real functions. (Do you know an example of a differentiable real function with nondifferentiable derivative?)
2
Real and imaginary parts of functions We will often use z to denote a complex number and we will have z = x + iy where x and y are both real. The value f (z) is a complex number and so has a real and imaginary part. We often use u to denote the real part and v to denote the imaginary part. Note that u and v are functions of z. We often write f (x + iy) = u(x, y) + iv(x, y). Note that u is a function of two real variables, x and y. I.e. u : R2 → R. Similarly for v. Examples 1.9 (i) Let f (z) = z 2 . Then, f (x + iy) = (x + iy)2 = x2 − y 2 + 2ixy. So, u(x, y) = x2 − y 2 and v(x, y) = xy. (ii) Let f (z) = z. Then, f (x + iy) = x2 + y 2 and v(x, y) = 0. Exercises 1.10 Find u and v for the following: (i) f (z) = 1/z for z ∈ C\{0}. (ii) f (z) = z 3 . Visualising complex functions In Real Analysis we could draw the graph of a function. We have an axis for the variable and an axis for the value, and so we can draw the graph of the function on a piece of paper. For complex functions we have a complex variable (that’s two real variables) and the value (another two real variables), so if we want to draw a graph we will need 2 + 2 = 4 real variables, i.e. we will have to work in 4dimensional space. Now obviously this is a bit tricky because we are used to 3 space dimensions and ﬁnd visualising 4 dimensional space very hard. Thus, it is very difﬁcult to visualise complex functions. However, there are some methods available: (i) We can draw two complex planes, one for the domain and one for the range. x2 + y 2 . So, u(x, y) =
3
How shall we deﬁne functions such as ez . cos z and sin z First we will try and deﬁne some elementary complex functions to play with. e. we shall deﬁne the exponential. y) = cos x + sin y and v(x. u(x. we know from Real Analysis that the functions can be described using a power series. for z ∈ C. Deﬁning ez . cos z and sin z? We require that their deﬁnition should coincide with the real version when z is a real number. (2n + 1)! Thus. what does it mean for a triangle to have an hypotenuse of length 2 + 3i? The exponential is deﬁned using differential calculus and we have not yet deﬁned differentiation of complex functions. This is not very helpful. sine and cosine of z as follows: ∞ e z := n=0 ∞ zn . The graph of a function of two variables is a surface in three space.. e. sin2 z + cos2 z = 1 would be nice. However. and we would like them to have properties similar to the real versions of the functions. y) = x2 − y 2 (iii) Make one of the variables time and view the graph as something that evolves over time. (2n)! sin z := n=0 ∞ cos z := n=0 (−1)n 4 . However.g.g.(ii) The twovariable functions u and v can be visualised separately. x3 x5 + − ··· = sin x = x − 3! 5! ∞ (−1)n n=0 x2n+1 . (2n + 1)! z 2n . n! (−1)n z 2n+1 . sine and cosine are deﬁned using trigonometry and so are hard to generalise: for example.
log(5/7) 4 − 3i 7 n 4 − 3i 7 n converges to zero. into the deﬁnition. n! 2! 3! These deﬁnitions obviously satisfy the requirement that they coincide with the deﬁnitions we know and love for real z. does a complex number comes out? To answer this we will have to study complex series and as the theory of real series was built on the theory of real sequences we had better start with complex sequences..13 Notice that an = cn − c is a real sequence. we are saying something about a complex sequence using real analysis.11 A complex sequence cn converges to c ∈ C. if given any ε > 0. Thus the sequence converges to zero. Deﬁnition 1.e. So. 5 .. Remark 1. such as 3 + 2i. Hence. when we put in a z. n 4 − 3i = 7 n = 25 49 = 5 7 n . but how can we be sure that the series converges? I. Complex Sequences The deﬁnition of convergence of a complex sequence is the same as that for convergence of a real sequence. given any ε we can choose N to be any natural number greater than log ε/ log(5/7).Thus. then there exists N such that cn − c < ε for all n ≥ N . We write cn → c or limn→∞ cn = c. and that cn → c if and only if the real sequence cn − c → 0.12 The sequence cn = Consider cn − 0 = cn  = So cn − 0 < ε ⇐⇒ ⇐⇒ ⇐⇒ (5/7)n < ε n log(5/7) < log ε log ε n> . Example 1. ∞ e 3+2i = n=0 (3 + 2i)n (3 + 2i)2 (3 + 2i)3 = 1 + (3 + 2i) + + + .
(Oxford English Dictionary). (ii) Show that the limit of a complex sequence is unique.e. 6 . The next proposition shows that a sequence converges if and only its real and imaginary parts do. i. bn → b. This is a key observation. then an − a → 0.e. HTTLAM 1. Example 1. The last inequality follows from the triangle inequality applied to z = an −a and w = i(bn −b). [⇐] Suppose an → a and bn → b. but we can also take real and imaginary parts.17 n2 + in3 n2 n3 = 3 +i 3 → 0 + i.18 (i) Which of the following sequences converge(s)? (n + 1)5 n5 i and 5 − 12i 6 n . an → a. cn → c. Because an −a → 0 and bn −b → 0 we deduce cn − c → 0. n3 + 1 n +1 n +1 Exercises 1. In this case we take the modulus. We can apply results from real analysis to produce results in complex analysis. Proof. and c = a + ib. Similarly.16 Try not to use the deﬁnition of convergence to prove that a sequence converges. i.15 Let cn = an + ibn where an and bn are real sequences. Then cn → c ⇐⇒ an → a and bn → b. Proposition 1. But 0 ≤ an − a = Re(cn ) − Re(c) = Re(cn − c) ≤ cn − c.Paradigm 1.14 The remark above gives a good example of the paradigm1 we will be using. and bn − b → 0. 1 Paradigm: a conceptual model underlying the theories and practice of a scientiﬁc subject.1 = i. So by the squeeze rule an − a → 0. [⇒] If cn → c. then cn − c → 0. Note it well! Let’s apply the paradigm. We have 0 ≤ cn − c = (an − a) + i(bn − b) ≤ an − a + bn − b.
In this case wk = k=0 k=0 xk + i k=0 yk . Let an = n xk .15. Proof. but ﬁrst let us make a deﬁnition. . Proposition 1. the following sequences converges s0 s1 s2 s3 = = = = . and k=0 k=0 k=0 apply Proposition 1.21 ∞ (−1)n i The series converges. w0 w0 + w1 w0 + w1 + w2 w0 + w1 + w2 + w3 Let’s apply the paradigm and give a result on complex series using real series. the ratio test and the integral test. . The second part of the statement comes from equating real and imaginary parts. and sn = n wk . n=0 n! (−1)k .19 A complex series ∞ wk converges if and only if the sequence k=0 sn formed by its partial sums sn = n wk converges. k! n Thus ∞ (−1) i converges to i/e. Deﬁnition 1. Let xk = 0 and yk = n=0 n! (−1)k Then xk = 0. Example 1. obviously.22 ∞ We say k=0 wk is absolutely convergent if the real series ∞ k=0 wk  converges. bn = n yk . k! In real analysis we have some great ways to tell if a series is convergent. for example. k=0 That is. Can we use the real analysis tests in complex analysis? The next theorem says we can. Then.20 Let wk = xk + iyk where xk and yk are real for all k.Complex Series Now that we have deﬁned convergence of complex sequences we can deﬁne convergence of complex series. Deﬁnition 1. ∞ ∞ ∞ wk converges ⇐⇒ k=0 k=0 ∞ ∞ xk and k=0 ∞ yk converge. 7 . and = e−1 .
We now prove an inﬁnite version of the triangle inequality.25 Note that the converse to Theorem 1. The assumption says something about a real series (we know lots about these!) and gives a conclusion about a complex series.23 If ∞ wk  converges. For n ≥ 1.23 is not true. show it absolutely converges. k=0 Then. Thus. We alk ready know this from Real Analysis. Proof. by Proposition 1. we can apply the ratio test or comparison test to the real series and say something about the complex series. k=0 k=0 k=0 HTTLAM 1. Remark 1. So the real series ∞ xk converges absolutely and we know k=0 from Real Analysis I that this implies that ∞ xk converges. the series ∞ yk converges. ∞ ∞ n n wk k=0 = k=0 ∞ wk − k=0 n wk + k=0 n wk wk k=0 n ≤ k=0 ∞ wk − k=0 n wk + wk + k=0 k=0 ≤ k=0 wk − 8 wk  .20. For example. . Let wk = xk + iyk . it has merely been extended to complex numbers in a natural way.This deﬁnition is really the same as in Real Analysis. then it is convergent. Lemma 1. Then ∞ ∞ wk ≤ k=0 k=0 wk . with xk and yk real. ∞ (−1) k=0 k converges but ∞ (−1)k k=0 k = ∞ 1 k=0 k diverges. Then ∞ wk  k=0 convergent implies that ∞ xk  is convergent (because 0 ≤ k=0 xk  = Re(wk ) ≤ wk  and we can apply the comparison test). This is a fantastic tool. ∞ wk = ∞ xk + i ∞ yk .26 Suppose that ∞ k=0 wk converges absolutely. Theorem 1.24 When asked to show a series converges. k=0 Similarly. Now for a very important theorem which says that if a series is absolutely convergent. then k=0 ∞ k=0 wk converges. Remember it. Great! Proof.
 result. where Such a power series is a function of z. with 0 ≤ R ≤ ∞. So by the ratio test the series an converges if z < 1. So.As n → ∞ then obviously. for z < 1. Then. for which complex values of z does it converge? Let us use the ratio test. 9 . where z ∈ C.28 Consider the series ∞ z n . Let an = z n . Example 1. as the next theorem shows. because 0 0 ∞ 0 0 = 0 + 0 + 0 + · · · = 0. an z n  As n → ∞ we have z → z... because as we shall see later.27 A complex power series is a sum of the form ck ∈ C and m is possibly inﬁnite. any differentiable complex function can be represented as a power series. So z n converges absolutely. Then an+1 z n+1  = = z. That the set of complex numbers for which the series converges is given by something of the form z < R for some R is a general phenomenon. there exists 0 R. diverges if z > 1 and for z = 1 we don’t know what will happen. you remember from Real Analysis I that for real z the power series converges only for −1 < z < 1. hence the Deﬁnition 1. because there is no dependence on n. such that ∞ an z n 0 converges absolutely for z < R.. Theorem 1. m k k=0 ck z . 0 ∞ n = We also know it converges for z = 0. We know for z = 1 0 this series does not converge because then we have ∞ 1n = 0 ∞ 1 = 1 + 1 + 1 + . ∞ k=0 wk − n k=0 wk  → 0. Much of the theory of differentiable complex functions is concerned with power series. diverges for z > R. and hence converges. Hopefully. Radius of Convergence Just as with real power series we can have complex power series that do not converge on the whole of the complex plane.29 Let ∞ an z n be some complex power series.
32 (I’ll do this example in great detail. Thus by n=0 n=0 n! n Theorem 1. = zn n! The last part is true because for ﬁxed z the real number z is of course a ﬁnite constant. We have an+1 an z n+1 (n + 1)! z n+1 n! = z n (n + 1)! z = n+1 → 0 as n → ∞. Sine. see p5657. but for z = 1 the series diverges.30 Given a power series.) n The series ez = ∞ z converges for all z ∈ C. HTTLAM 1. 0 In the last exercise note that for z = −1 the series converges. cosine and exponential functions are deﬁned on the whole of C. The proof is similar to that for real power series used in Real Analysis I. We want ∞ an to converge. Example 1. and exponential are deﬁned for all complex numbers Let us now return to showing that the sine.31 Show that ∞ z n /n has radius of convergence 1. The next example will be more like the solution I would expect from you. 10 . n=0 n! The following is an example with some of the small detail missing. This tells us that for z = 1 we can get some values of z for which the series converges and some for which the series diverges. immediately ask ‘What is its radius of convergence?’ Exercise 1. cosine. n So by the ratio test ∞ an = ∞ z converges.23 the series ∞ z converges for all z ∈ C.Proof. n=0 n! n For any z ∈ C let an = z . so n=0 n! we use the ratio test on this real series. (both these fact should be well known from Real Analysis). This is how I would expect the solution to be given if I had set this as an exercise. Stewart and Tall also have a good proof.
35 ¯ (i) ez = ez . Theorem 1. for all y ∈ R. Most of these you may already from Numbers and Proofs. (ii) eiz = cos z + i sin z. and hence converges. n! 11 . So by the ratio test the complex series converges absolutely. for all z ∈ C. (iv) ez = 0. w ∈ C. but the proofs may not have been rigorous. Then (2n + 1)! = = = = → z 2(n+1)+1 z 2n+1 (2(n + 1) + 1)! (2n + 1)! 2n+3 2n+1 z z (2n + 3)! (2n + 1)! z2 (2n + 3)(2n + 2) z2 (2n + 3)(2n + 2) 0 as n → ∞.34 Prove that cos z converges for all z. Exercise 1.33 The series sin z = Let an = (−1)n an+1 an ∞ (−1)n n=0 2n+1 z 2n+1 converges for all z ∈ C. (2n + 1)! z . for all z ∈ C. (viii) eiy  = 1. (vii) ez  = eRe(z) . (v) e−z = 1/ez . Properties of the exponential We have deﬁned the exponential function and shown that is deﬁned on all of C.Example 1. for all z ∈ C and n ∈ Z. for all z ∈ C. (vi) enz = (ez )n . for all z. (i) We have ∞ e = n=0 z ¯ (¯)n z = n! ∞ n=0 (z n ) = n! ∞ n=0 zn = ez . Proof. (iii) ez+w = ez ew . let’s now look at its properties. for all z ∈ C. for all z ∈ C.
ez cannot be zero. e. π. (vi) Follows from repeated application of (iii). It relates so many different √ important numbers. This is because we have not yet deﬁned ab for all complex a and b. ez ez by (i). The proofs are left as simple exercises.36 (i) e2πi = 1. e2Re(z) 2 eRe(z) by (vi). (ii) (De Moivre’s Theorem) (cos θ + i sin θ)n = cos nθ + i sin nθ for all θ ∈ R. (Just put iz into the power series and separate the real and imaginary parts. (viii) From (vii) we get eiy  = eRe(iy) = e0 = 1.(ii) Exercise. Thus. Warning! 1. 2 astute reader may say ‘deﬁne it to be e−2π .’ 12 .37 We have not shown that ezw = (ez )w for all z. As both ez  and eRe(z) are real and positive we deduce that (vii) is true. w ∈ C. Then (ez )w = (e2πi )i = 1i . (iv) Note that ez and e−z both exist. (v) This is obvious from the proof of (iv). ez+z by (iii). Part (i) is one of the best theorems in mathematics. Corollary 1. Consider z = 2πi and w = i.) (iii) This will be delayed until we deal with differentiability. by calculation.38 Prove that sin z = 2 The eiz − e−iz 2i and cos z = eiz + e−iz . in a simple expression. What could 1i be? 2 Exercise 1. −1. We have ez e−z = ez−z by (iii). = e0 = 1. (vii) We have ez 2 = = = = = ez ez by deﬁnition. and of course 1 and 2.
Then. So let’s suppose we have a complex number w. eiy = 1 ⇐⇒ ⇐⇒ ⇐⇒ cos y + i sin y = 1 cos y = 1 and sin y = 0 y = 2πn for some n ∈ Z. So z − w = x + iy = 0 + i.e. since the exponential of a real number is positive. for some k ∈ Z. But from the working in the earlier part of the proof we know this is equivalent to ez = ew .40 For complex numbers z and w = 0 we have ez = w ⇐⇒ z = ln w + i(arg(w) + 2kπ). Then. Deﬁnition of the complex logarithm We all know that the real exponential function has an inverse function called loge or just ln. (the implication does not reverse!) x iy e e  = 1 ex eiy  = 1 ex  = 1 ex = 1. x = 0. [⇐] Suppose that z − w = 2πin for some n ∈ Z. [⇒] Let z − w = x + iy where x and y are real. to deﬁne the real log of a number x we want some unique number y such that ey = x.Another property of the complex exponential is that it is periodic.39 For any complex numbers z and w we have ez = ew ⇐⇒ z − w = 2πin for some n ∈ Z. that is the crux of the deﬁnition of inverse. i. ez = ew ⇐⇒ ⇐⇒ ⇐⇒ =⇒ ⇐⇒ ⇐⇒ ⇐⇒ ⇐⇒ ⇐⇒ ez /ew = 1 ez−w = 1 ex+iy = 1 (∗) x+iy e  = 1. This theorem has serious repercussions for deﬁning the inverse of ez .2πn = 2πin. Is there an inverse for the complex exponential? Well. 13 . then we want some z such that ez = w. Theorem 1. deﬁning the log function. Proof. Then. Proposition 1. so eiy = 1 as x = 0. Let’s investigate this. ez−w = n e2πin = (e2πi ) = 1n = 1. By (∗) we know that ex+iy = 1.
so we get ex+iy = ex (cos y+i sin y) = w. If z is such that ez = w.1 = ez = w). n ∈ Z. Then ez = w implies that r = ln w and y = θ + 2πk for some k ∈ Z. if we deﬁne the log to be the complex number with −π < arg(w) ≤ π. if we are trying to solve an equation and take the log of both sides using this deﬁnition.41√ Solve ez = 1 + i 3. [⇐] If z = ln w + i(arg(w) + 2πk) for some k ∈ Z then ez = eln w+i(arg(w)+2πk) = eln w ei arg(w) e2πik k = wei arg(w) e2πi = wei arg(w) = w. Now let us take the modulus of both sides: ex+iy  ex eiy  ex  ex loge ex x = = = = = = w w w w loge w (using the real log function) ln w. (The point is that if we have a w then the proposition gives us lots of zs to choose from. √ Solution: Let w = 1 + i 3. n ∈ Z. By drawing a picture (or through careful π use of calculator) we can see that arg(w) = + 2nπ. Thus. [⇒] Write z = x+iy. i.42 Notice how well working out the modulus and argument serves us. Common Error 1. r = w and θ = arg(w). 3 HTTLAM 1.Proof. The modulus of w is w = √ 2 √ 12 + 3 = 4 = 2. So 3 π z = ln 2 + i + 2nπ . then we may be 14 . there is some ambiguity and we make a choice.e. Now suppose that w = r(cos θ + i sin θ) for some real r and θ. Example 1.44 Solve e2iz = i. Exercise 1. So z has the form in the statement. (It’s not i(π/2 + 2kπ).) However. then z + 2πi will work just as well (ez+2πi = ez e2πi = ez .) So does the theorem allow us to deﬁne the log of a complex number? Yes. Conclusion: calculate modulus and argument.43 Don’t forget the 2kπ with the argument.
So.46 Note that in the above example we replaced eiz with another complex number w. because we could then get a polynomial equation.) HTTLAM 1.losing solutions.45 Solve the equation sin z = 2. Another worked example Example 1. 15 . Then 2i = 2. w2 − 1 = 4iw w2 − 4iw − 1 = 0 4i ± w = (4i)2 + 4 √2 4i ± −12 = 2 √ = 2i ± −3 √ = (2 ± 3)i. So in fact the best deﬁnition is to make the function multivalued. Solution: We can rewrite this as the equation becomes 1 2i w− 1 w eiz − e−iz . 2 √ √ (The last equality is true because ln(2+ 3) > 0 and ln(2− 3) > 0. Let w = eiz . eiz = w = (2 ± √ 3)i. Now. This is something we will not go into in great depth just now. so iz = ln w + i arg(w) √ π = ln (2 ± 3)i + i + 2nπ 2 √ π = ln 2 ± 3 + i + 2nπ 2 √ 1 π z = ln 2 ± 3 + i + 2nπ i 2 √ π + 2nπ = −i ln 2 ± 3 + i 2 √ π + 2nπ − i ln 2 ± 3 = 2 √ π = + 2nπ − i ln(2 ± 3).
47 Show that sin z = 0 cos z = 0 ⇐⇒ ⇐⇒ z = kπ. 2 These results will be used later. comparison test. • For power series use the ratio test to ﬁnd radius of convergence. • Apply the ratio test. • We deﬁne exponential. k ∈ Z. etc. 1 z = (2k + 1)π. 16 . then wk converges. to the modulus of terms of a complex series to determine convergence. Summary • Paradigm: Complex analysis is developed by reducing to real analysis. often through taking the modulus. • If ∞ n=0 ∞ n=0 wk  converges. k ∈ Z.Exercise 1. sine and cosine by power series.
Differentiation of complex valued real functions Suppose that f : R → C is given by f (t) = (1 + 3i)t2 . in this case. dt Exercise 2. that is integration over a complex variable. 3 They do behave in much the same way though. 17 . If we deﬁne f (t) using the standard deﬁnition: f (t + δ) − f (t) . all similar rules work in this way. δ→0 δ then we get f (t) = 2(1 + 3i)t. where S ⊆ R. (δ ∈ R). If f (t) = (2 + 3i)t3 . any constants can be real or complex. E. A complex function takes a complex number from a domain and produces a complex number. φ (x) = Remark 2. But let us ﬁrst generalise integration and differentiation to complexvalued functions of a real variable. holds even though c is complex. A complexvalued function of a real variable takes a real number and produces a complex number. the rule f (t) = nctn−1 for f (t) = ctn . dt d sin(ct) = c cos(ct). then f (1) = 2 + 3i ∈ C. That is. So. for instance. Complex Riemann Integrals Now we shall integrate complexvalued functions with respect to one real variable.2 Complex Riemann Integration In a later section we deﬁne contour integration.2 This is not the same as differentiation with respect to a complex variable.1 Let φ(x) = 3x3 + 2ix − i + tan((4 + 2i)x). This notion is fundamental in complex analysis. t ∈ R. f (t) = lim d ct e = cect . dt d cos(ct) = −c sin(ct). Then.g. A complexvalued function of a real variable is a map f : S → C. Such a function is different to a complex function. We shall do this with a bit more care than differentiation.3 That will come later. Basically.
For example. a 18 . Obviously.3 Let g : [a. where a and C are complex constants. separating functions into real and imaginary parts can get a bit tedious. if u and v are continuously differentiable.5 tn+1 + C. such as substitution also work. Fortunately. Other standard methods. and we deﬁne a g(t) dt by b b b b g= a a g(t) dt = a u(t) dt + i a v(t) dt. b] → C be given g(t) = u(t) + iv(t). It is not difﬁcult to prove the following. We say that g is complex Riemann integrable (abbreviated CRI) if both u b and v are RI as real functions.4 b b b e3it dt = a a cos 3t dt + i a sin 3t dt i 1 [sin 3b − sin 3a] + [− cos 3b + cos 3a] 3 3 1 = −e3ib + e3ia 3i i 3ib e − e3ia . just as for differentiation above. Example 2. Example 2. replacing real constants by complex ones. then since g = u + iv we get a version of the Fundamental Theorem of Calculus: b b b g = a a u +i a v = [u(b) − u(a)] + i[v(b) − v(a)] = g(b) − g(a). = 3 = Many properties of CRI can be derived from the corresponding properties for RRI by considering the real and imaginary parts. n+1 atn dt = a eat dt = eat + C. we can use standard integrals. a 1 sin at dt = − cos(at) + C.Deﬁnition 2. a 1 cos at dt = sin(at) + C.
Its format should be familiar from real analysis. Proof. as α = 1. Lemma 2. Hence. b b g(t) dt a = α a b g(t) dt b = a b Re (αg(t)) dt + i a Im(αg(t)) dt = a b Re (αg(t)) dt. If LHS = 0. Triangle inequality for CRI We need the following result later. asb b sume LHS = 0. Let α =  a g/ a g. the only difference here is that the functions can be complexvalued. (Hence α = 1). then the statement is trivial. So. b] → C is CRI and g is RRI.6 2 Calculate 0 t2 − it3 − cos(2t) dt. a b ≤ = a b α g(t) dt g(t) dt. then b b g(t) dt ≤ a a g(t) dt. αg(t) dt.7 If g : [a. as Re(z) ≤ z. 19 . because LHS is real. Exercise 2.So a lot of the time we can use standard integrals to calculate complexvalued integrals with respect to a real variable t. a = Summary • We can integrate and differentiate complexvalued functions of real variables in the same way as realvalued functions of real variables.
(ii) Circle of radius r based at the origin: γ : [0. Its image is usually some curve in the plane. then consider the integral 2π ecos θ cos(nθ − sin nθ) dθ. (ii) γ is continuous on [aj−1 .1 A contour (also called a path) is a continuous map γ : [a. 4 Those of you who have done MATH2360 or MATH2420 will see that this is just a line integral.e.) We say γ is closed if γ(a) = γ(b). it is easy to do as it has similar properties to Riemann integration of one real variable. b] → C which is piecewise smooth. (iii) Circle of radius r based at z0 : γ : [0. Fortunately. 2π] → C given by γ(t) = reit . there exist a = a0 < a1 < a2 < · · · < an = b such that (i) γ[aj−1 . Warning! 3.4 This is a fairly abstract process. for all j. 2π] → C given by γ(t) = z0 + reit .2 A contour is not a complex function.3 Contours In the next section deﬁne integration along a contour in the complex plane. Examples 3. i. 20 . 0 This is a seriously nasty integral! Imagine trying to solve it via the methods we know. 1] → C given by γ(t) = α + t(β − α). Using contour integration we shall show that it is very simple to calculate. In case you think that it is too abstract and not relevant to real problems. aj ]. and you have years of experience of that. aj ] is differentiable. It is a complexvalued function of a real variable. for all j.3 (i) Straight line from α to β: This is γ : [0. First though we will deﬁne contours. Contours Deﬁnition 3. the meaning of which usually takes a little time to understand. (The left and right derivatives of γ at aj may differ.
b] → C which is piecewise smooth. They have the same image.5 There is often confusion between a contour and its image. 1] → C given by γ(t) = α + t(β − α). eit for 0 ≤ t ≤ π/2. 4π] → C be given by γ(t) = eit . etc. The contour goes round the circle twice. 2π]. (v) Let α : [−1. • Straight line from α to β: γ : [0.4 Given a contour. writing z ∈ γ is incorrect because γ is not a set. Common Error 3. We do use the notation z ∈ γ later. by which we mean z ∈ γ([a. for − 1 ≤ t ≤ 0. Since contours are complexvalued functions of a real variables. 4π]. the other [0. Strictly speaking.(iv) Circular arc of radius r based at w: γ(t) = w + reit . Consider the contours (ii) and (vi) above. HTTLAM 3. This subtlety will be important later. Then the image of γ is the unit circle centred at zero. (So 0 ≤ t ≤ 2π gives the circle above). It is a complexvalued function of a real variable. we can differentiate them. try to draw its image. taking r = 1 in (ii). 21 . b]). with ease. A contour is not a set of points in the complex plane. 2π] → C given by γ(t) = z0 + reit . θ1 ≤ t ≤ θ2 . the unit circle. We draw an arrow to show the direction we go in. (vi) Let γ : [0. π/2] → C be given by α(t) = Draw the image: t + 1. one maps from [0. Summary • A contour is a continuous map γ : [a. the contours are different. it is a map. • Circle of radius r based at z0 : γ : [0. However.
b] → C be a contour. Then. the integral of f along γ is b f= γ γ f (z) dz := a f (γ(t))γ (t) dt.1 Let f : D → C be a continuous complex function and γ : [a. 22 . It is central to the module.4 Contour Integration We now come to probably the most important deﬁnition in complex analysis: contour integral. Then. Example 4. If you don’t understand this section. Deﬁnition 4. then the rest of the course will be a complete mystery to you. γ (t) = 1 + 2it.2 Let γ(t) = t + it2 for 0 ≤ t ≤ 2 and f (z) = z. Thus we can integrate this product. Note that f (γ(t)) and γ (t) are complexvalued functions of a real variable. and hence so is their product. and 2 f = γ 0 2 (t + it2 )(1 + 2it) dt t + 2it2 + it2 + 2i2 t3 dt 0 2 = = 0 t + 3it2 − 2t3 dt 1 2 3it3 2t4 t + − 2 3 4 4 2 0 2 0 = = 1 2 t t + it3 − 2 2 1 2 24 = 2 + i23 − 2 2 = −6 + 8i.
Then. say. 0 ≤ t ≤ 1. The resulting integrand f (γ(t))γ (t) is CRI because it is continuous except possibly at ﬁnitely many points where γ (t) is discontinuous.02 )3 − 3 3 1 (2 + i)3 − 8 3 11 −2 + i. (v) f5 (z) = z 2 and γ5 (t) = eit . 1 z dz = γ 0 1 2 (2 + it2 )2 (2it) dt d dt (2 + it2 )3 3 1 = 0 dt = = = = (2 + it2 )3 3 0 (2 + i.3 Let γ(t) = 2 + it2 for 0 ≤ t ≤ 1 and f (z) = z 2 . (ii) f2 (z) = Re(z) and γ2 (t) = t + it. (iv) f4 (z) = 1/z and γ4 (t) = 2e−it .4 Draw the contours and calculate the integrals of the functions along the contours. 3 This is just the sort of example you need to be able to do with ease. aj ] and calculate b n aj f (γ(t))γ (t) dt = a j=1 aj−1 f (γ(t))γ (t) dt.Example 4. 0 ≤ t ≤ π/2. 0 ≤ t ≤ 1. involving f (z) = z n in (v)? Remark 4. (iii) f3 (z) = Re(z) and γ3 (t) = 1 − t + i(1 − t). (i) f1 (z) = Re(z) and γ1 (t) = t. Can you justify the results in (ii) and (iii)? Can you make any conjectures. 0 ≤ t ≤ 1.12 )3 (2 + i. 0 ≤ t ≤ π. Exercise 4.5 Note that in the deﬁnition of contour integral we only require f to be continuous. In practice we subdivide [a. 23 . b] into pieces [aj−1 .
6 Let γ be as in Example 3. by taking a contour along the real line. (ii) From a purely formal viewpoint. 3 Remarks 4. for example. b] → C given by γ(t) = t for a ≤ t ≤ b.1 dt + 0 π/2 eit 2 ieit dt = −1 (t + 1)2 dt + 0 0 ie3it dt π/2 0 i 3it 1 = (t + 1)3 + e 3 3i −1 1 i = − 0 + [−i − 1] 3 3i i = − . 24 . then b b f (z) dz = γ a f (t)γ (t) dt = a f (t) dt. z 2 dz = γ −1 0 γ(t)2 γ (t) dt π/2 = −1 0 (t + 1)2 .7 (i) Suppose f : D → C is a complex function such that f (x) is real for x real. Find π/2 γ z 2 dz. we can see that contour integration includes the theory of real integration as a special case. sin x. If we take γ : [a. (which can be thought of as (dz/dt)dt).3(v). we can justify the deﬁnition of contour integral by saying that we are replacing z by γ(t) so we need to replace dz by γ (t)dt. Thus.Example 4.
0 ≤ t ≤ 2π. γ(t) = w + reit . if n = −1. z−w γ(t) = w + reit . Let γ be a circle with centre w and radius r > 0.Fundamental Example Take the function deﬁned by f (z) = (z − w)n where n ∈ Z. 0 ≤ t ≤ 2π. = n+1 2π i 0 1. 25 . ireit dt rn eint .e.dt = 2πi. γ(t) = w + reit . 2π (z − w)n dz = γ 0 2π (γ(t) − w)n γ (t) dt w + reit − w 0 2π n = = 0 . if n = −1. z−w γ γ Note this well. Then. i. this innocuous looking calculation will be used to devastating effect later! Thus Summary • The integral of f along γ is b f= γ γ f (z) dz = a f (γ(t))γ (t) dt. • γ 1 dz = 2πi. ireit dt 2π = irn+1 ei(n+1)t dt 0 n+1 r 2π ei(n+1)t 0 = 0. 1 dz = 2πi and (z − w)n dz = 0 for n = −1. 0 ≤ t ≤ 2π. • γ (z − w)n dz = 0 for n = −1. (so for n < 0 the map is not deﬁned at w).
µ ∈ C. We shall now describe them. γ a contour in D. λ. f (x) dx. All of these have analogues in contour integration. Linearity.5 Properties of Contour Integration There are a number of well known properties of ordinary integration: b b b λf (x) + µg(x) dx = λ a b a c f (x) dx + µ a b g(x) dx. φ−1 (b) f (x) dx = a φ−1 (a) f (φ(y))φ (y) dy (change of variables). The details are as follows: b λf + µg = γ a b (λf + µg)(γ(t))γ (t) dt (λf (γ(t)) + µg(γ(t))) γ (t) dt a b b = = λ f (γ(t))γ (t) dt + µ a a g(γ(t))γ (t) dt = λ γ f +µ γ g. So this has the same format as in integration over a real variable. µ ∈ R. a < c < b. Integration over joins What if produce a contour by doing one after another? 26 . λ. If f. then λf + µg = λ γ γ f +µ γ g. The proof follows directly from the linearity property of Riemann integration. g are continuous on D. In the following the functions will be continuous on some domain D and the contours will be maps into D. f (x) dx = a b a f (x) dx + c a f (x) dx = − a b b f (x) dx.
α(t) = eit for 0 ≤ t ≤ π/2. The contour α : [−1.3 we have dz = z−w 4 γi γ i=1 dz . Again. c] → C given by (γ1 + γ2 ) (t) = γ1 (t) a ≤ t ≤ b.3 For any continuous f f= γ1 +γ2 γ1 f+ γ2 f.1 If γ1 : [a. then their join (or sum) is the contour γ1 + γ2 : [a. b] → C and γ2 : [b. z−w 27 .2 Consider Example 3. where γ is the boundary of the square with γ (z − w) corners w ± l ± il.4 1 Find dz.3(v). for − 1 ≤ t ≤ 0. this follows from applying the deﬁnition and using a b c b CRI property: a = a + c . Example 5. starting at w + l − li and going anticlockwise. γ2 (t) b ≤ t ≤ c. Example 5. Proposition 5. c] → C are two contours such that γ1 (b) = γ2 (b). By Proposition 5. This can be produced from γ1 (t) = t + 1 for −1 ≤ t ≤ 0 and γ2 (t) = eit for 0 ≤ t ≤ π/2.Deﬁnition 5. The join is continuous by the glue rule. So γ = γ1 + γ2 + γ3 + γ4 . π/2] → C is given by t + 1.
Proposition 5. z−w γ Notice that this coincides with the value of the integral when γ is a circle. Again we apply deﬁnitions and use RI properties. b] → C is a contour.To compute γ1 let γ1 (t) = w + l + it. So. but in this case we also need a simple change variables: s = a+b−t. The point is that we do γ backwards. l −l l −l l γ1 dz = z−w = = γt (t) dt γ1 (t) − w i dt l + it l − it i dt l2 + t2 t dt + i 2 + t2 l t l l −l l −l −l l = −l l2 l dt + t2 = 0 + i tan−1 Similarly γ2 = γ3 π = i . then its reverse is the contour (−γ)(t) = γ(a + b − t). −l ≤ t ≤ l. Proof. Reverse contour Deﬁnition 5. 2 dz = 2πi.5 If γ : [a.: (−γ)(a) = γ(a + b − a) = γ(b).6 For any continuous f . we have f =− −γ γ f. for a ≤ t ≤ b. etc. We have. (−γ)(b) = γ(a + b − b) = γ(a). 28 . Instead of starting at γ(a) we end there. 2 π = γ4 = i (check!). see the Fundamental Example.
29 . Then γ3 = −γ2 as follows.8 Let γ : [a.so dt = −ds. We have. γ = − Example 5. We have a = 0 and b = 1.7 See Exercise 4. Deﬁnition 5. (iii) φ(c) = a. d] → [a. b] → C be a contour and let φ : [c. b] be a function such that (i) φ is continuously differentiable . Note that (−γ) (t) = (−1)γ(a + b − t). We think of reparametrised contours as equivalent since we have the following. Reparametrisation Now we shall do the analogue of a change of variables. so (−γ2 )(t) = γ2 (0 + 1 − t) = γ2 (1 − t) = (1 − t) + i(1 − t) = γ3 (t). d].9 Let γ(t) = eit for 0 ≤ t ≤ 2π. d] → C deﬁned by γ(s) = γ(φ(s)) is called a reparametrisation of γ. (ii) φ (s) > 0 for all s ∈ [c. b f = −γ a b f ((−γ)(t))(−γ) (t) dt f (γ(a + b − t))(−1)γ (a + b − t) dt a a = = − f (γ(s))γ (s) (−ds) b a = b f (γ(s))γ (s) ds b = − a f (γ(s))γ (s) ds f. Then the contour γ : [c. Let φ(s) = 2πs for 0 ≤ s ≤ 1. Then γ(s) = e2πis for 0 ≤ s ≤ 1. This explains why γ3 f3 (z) dz = − γ2 f2 (z) dz.4(ii) and (iii). and φ(d) = b. Example 5.
d f = γ e c d f (γ(s)) γ (s) ds f (γ(φ(s)) γ (φ(s)) φ (s) ds c b = = a f (γ(t)) γ (t) dt. Proof.11 (i) If γ1 : [a.10 If γ is a reparametrisation of γ. then it has the same image as γ. d] → C are two contours such that γ1 (b) = γ2 (c). f. Remarks 5. it does not matter whether we evaluate the integral 1/(z − w) dz using γ(t) = eit for 0 ≤ t ≤ 2π or γ(t) = e2πit γ for 0 ≤ t ≤ 1. Often we abuse notation and write this simply as γ1 + γ2 . for all continuous f . b] → C and γ2 : [c. (ii) If γ is a simple (i. 30 . (iii) If γ is a reparametrisation of γ.Proposition 5. using t = φ(s).e. The converse is not true. d + b − c] → C given by γ(s) = γ2 (c − b + s). then f= γ e γ f. γ = Thus. it doesn’t cross itself) closed contour with no orientation speciﬁed. then it is traversed anticlockwise. then we can reparametrise γ2 as γ2 : [b. We have. Then γ1 (b) = γ2 (b) so we can form the join γ1 + γ2 .
γ a contour in D. for a ≤ t ≤ b. • If γ is a reparametrisation of γ. then λf + µg = λ γ γ f +µ γ g. we have f =− −γ γ f. b] → C is a contour. • If f is continuous on D. then f= γ1 +γ2 γ1 f+ γ2 f. λ. 31 . µ ∈ C.Summary • If f. • For f continuous. then its reverse is the contour (−γ)(t) = γ(a + b − t). γ a contour in D. • If γ : [a. then f= γ e γ f. g are continuous on D.
denoted L(γ). and the integral of speed over time gives the length of a path. The next theorem is similar to one from real analysis: Lemma 6. Exercise 6. and so the generalisation to complex analysis is the length of the contour γ. Suppose that f (z) ≤ M for all z ∈ γ.6 The Estimation Lemma Recall for a real continuous function that b f (x) dx ≤ sup {f (x)} × (b − a).b] We would like a complex version of this. b] → C be a contour. γ 32 . is there some bound on  γ f (z) dz? It is obvious that the ﬁrst part in the product above can be generalised. Deﬁnition 6. b] → D be a contour. that is. Intuitively speaking. This really does measure the length of the curve. γ (t) is the velocity of a contour. a x∈[a. So. but what does b − a correspond to? Those of you familiar with measure theory will know it is the length of the interval from a to b.4 (Estimation Lemma) Let f : D → C be a continuous complex function and γ : [a.1 Let γ : [a.2 Let γ(t) = α + t(β − α) with 0 ≤ t ≤ 1 be the straight line contour from α to β. is deﬁned to be b L(γ) = a γ (t) dt. We have 1 1 1 L(γ) = 0 γ (t) dt = 0 β − α dt = β − α 0 dt = β − α[t]1 0 = β − α. Example 6. so γ (t) is the speed. which is reassuring.3 Show that the length of the contour given by traversing once round the the circle of radius r based at the origin is 2π. The length of γ. the length of the contour from α to β is β − α. Then f (z) dz ≤ M L(γ).
so by the Estimation Lemma (6. Anything bigger than this is also useful. z ∈ γ. which can be made smaller. so that L(γ) → 0. we use it to prove that we can integrate certain series termbyterm. Termwise integration of series The lemma will be useful in a number of contexts. then x ≤ 0 and so ez  = eRe(z) = ex ≤ e0 = 1. and hence is bounded.4) ez 1 πR dz ≤ L(γ) = .) 33 . Example 6. So M = supz∈γ {f (z)} will do as a bound. z = R so R = z + 1 − 1 ≤ z + 1 + 1. Hence. Thus z + 1 ≥ R − 1. Also. z+1 R−1 We know that L(γ) = πR as γ is semicircle of radius R.5 Show that γ πR ez dz ≤ z+1 R−1 where γ describes the semicircle from iR to −iR in the left halfplane {z : Re(z) ≤ 0} and R > 1. b]. Solution: If z = x + iy lies on the image of γ. when z lies on γ. b f (z) dz γ = a b f (γ(t))γ (t) dt f (γ(t))γ (t) dt by Lemma 2. We have. (ii) We can show some integral is zero by ﬁnding an M that tends to zero or some contour.6 (i) The constant M always exists: On the image of γ the function f (z) will always be bounded because the map t → f (γ(t)) is a continuous real function on [a.Proof. To begin with. (We know that an inﬁnite series can be differentiated termbyterm. ez 1 ≤ .7 a b ≤ ≤ a M γ (t)dt = M L(γ). z+1 R−1 R−1 γ Remarks 6.
= 0. Remark 6. We have k=0 n n f (z) dz − γ k=0 γ fk (z) dz = γ f (z) − k=0 n fk (z) fk (z) k=0 dz × L(γ) ≤ sup z∈γ f (z) − ∞ = sup z∈γ k=n+1 ∞ fk (z) fk (z) k=n+1 ∞ × L(γ) × L(γ) ≤ sup z∈γ ≤ k=n+1 Mk n × L(γ) × L(γ) = M− k=0 Mk → 0 × L(γ). as n → ∞. Hopefully. The second equality is obvious. (iii) Then. you remember from Real Analysis II that this implies uniform convergence. (ii) there exist real constants Mk such that fk (z) ≤ Mk for all z ∈ γ. Let f : D → C and fk : D → C be continuous complex functions.7 (Termbyterm integration of series) Let γ be a contour in a domain D.8 I could have deﬁned uniform convergence and so on for complex series in order to state the theorem. the ﬁrst equality is true. 34 .Corollary 6. ∞ ∞ fk (z) dz = k=0 γ γ k=0 fk (z) dz = γ f (z) dz. We show the ﬁrst. Suppose that (i) ∞ k=0 fk (z) converges to f (z) for all z ∈ γ. k ∈ N. Rather than waste time doing so I just used a version of the Weierstrass M test in the assumptions above. ∞ k=0 Mk converges. Thus. Let M = ∞ Mk . Proof.
Therefore. denoted L(γ). Rk . such as the Fundamental Theorem of Calculus. let Mk = k! condition (ii) holds.Example 6. ‘how do we ﬁnd Summary • The length of γ. but there are better ways. so condition (i) is fulﬁlled. ∞ ∞ zk zk z e dz = = dz. b] to C so its image must lie within an origincentred circle of radius R. k! k=0 We have zk zk fk (z) = = . Hence. Also. Condition (iii) holds. we can apply the deﬁnition of integration. γ • We can integrate termbyterm complex series that satisfy a Weierstrass M test type condition. ∞ ∞ Rk Mk = = eR .9 For any contour γ the integral γ ez dz can be calculated by termbyterm integration of the series for ez . k! γ γ k=0 k! k=0 γ ∞ 0 zk dz’? Obγ k! viously. k! 0 0 so Mk converges. Then f (z) dz ≤ M L(γ). for some large enough R. is deﬁned to be b L(γ) = a γ (t) dt. Thus z ≤ R for all z ∈ γ. • Suppose that f (z) ≤ M for all z ∈ γ. Thus. then e = fk (z). k! k! But z = γ(t) and γ(t) is a continuous map from [a. ∞ zk z Let fk (z) = . as we shall see in the next two sections. then fk (z) ≤ Mk for all z ∈ γ. This of course begs the question. 35 .
(ii) For any sequence zn . The proofs of these facts are the same as in real analysis. and so on.3 We say f is continuous on D if f is continuous at c for all c ∈ D.15. we have zn → c implies that f (zn ) → f (c). Remark 7.4 Suppose f (x + iy) = u(x. y). it allows us to solve some hard problems concerning real functions in a simple manner. Continuity Let’s ﬁrst give a deﬁnition of continuity. there exists δ > 0 such that f (z) − f (c) < ε. Use Proposition 1. Instead we develop a theory that really uses the complex numbers. but more importantly. We say f is continuous at c if limz→c f (z) = f (c) This is not remarkably different to real analysis. f (z) = z n is continuous.1 Suppose that f : D → C is a complex function on a domain D. then we might be tempted to deﬁne complex differentiable to mean that the real and imaginary parts of the function are differentiable with respect to x and y. It looks a lot better. Then. (i) For all ε > 0. y) + iv(x. Deﬁnition 7. 36 . it would be the same as theory of differentiable maps from R2 to R2 . Proposition 7.2 This is equivalent to either of the following. We need some examples and the next result helps supply some. but not a great one. whenever z − c < δ. Thus. Proof. Deﬁnition 7. f is continuous if and only if u and v are continuous.7 Complex Differentiation If we were inventing the theory of differentiation of complex functions for the the ﬁrst time. Sums and products of continuous functions are continuous. This would give us a theory. This gives a much richer theory.
that of differentiation.5 Let f : D → C be a complex function. Remark 7. f is complex differentiable at c if h→0 lim f (z + h) − f (z) . as expected. and (f g) (c) = f (c)g (c) + f (c)g(c). we have that. The proof of this is the same as the real situation. More generally. if f (z) = bz n for some complex constant b. It doesn’t look different to the real situation. at least symbolically. We say f is complex differentiable on D if it is complex differentiable at c for all c ∈ D. and (f + g) (c) = f (c) + g (c). but the consequences are far more profound. We write f (c) for this limit. That is.Differentiation Now we come to the crucial deﬁnition. Of course. it imposes considerable restrictions. (i) If f and g are differentiable at c ∈ D. Proposition 7. f (c) = 2c.8 Let f and g be complex functions on the domain D. then f (z) = nbz n−1 . Then (c + h)2 − c2 c2 + 2ch + h2 − c2 = lim h→0 h→0 h h = lim 2c + h lim h→0 = 2c. Example 7.7 The function f : C → C given by f (z) = z 2 is differentiable on C. but the fact that h can go to zero from any direction in the complex plane makes a huge difference. (b) f g.6 Complex differentiable functions are also called holomoprhic or analytic. Let c ∈ C. h∈C h exists. The deﬁnition looks the same as in the real case. Then. mathematicians do not want to do anything as clumsy or ugly as using ﬁrst principles. then so are (a) f + g. We would use a theorem such as the following. Deﬁnition 7. 37 .
It is possible to work ﬁnd their derivatives from ﬁrst principles. and (f /g) (c) = 0. Warning! 7. 2. then g ◦ f is differentiable at c with (g ◦ f ) (c) = g (f (c))f (c). We don’t yet know that we can differentiate series termbyterm and so can’t immediately prove that ez .9 Unlike continuity. but will delay a proof till later and just state the following. z z z(z 2 + 1) 38 . dz (i) Exercise 7. for all z ∈ C. dz d (iii) cos z = − sin z for all z ∈ C. In fact it is this difference that makes complex analysis so different to real analaysis. (you can try this as an exercise if you are keen). 1 1 sin z . complex differentiablity isn’t the same as being differentiable with respect to two real variables. dz d (ii) sin z = cos z for all z ∈ C.(c) f /g. Proof. If f is differentiable at c and g is differentiable at f (c). Theorem 7. Exercise 7. These are proved in the same way as the real case. cos z and sin z are differentiable.12 Where are the following funtions not differentiable? z2. tan z. or that their derivatives are what we expect them to be. .11 The elementary functions have the expected derivatives: d z e = ez . There is a connection as we see shortly. g(c)f (c) − f (c)g (c) provided g(c) = g(c)2 (ii) (Chain Rule) Suppose f : D → C and g : E → C are complex functions with f (D) ⊆ E. then f is continuous at c.10 Show that if f : D → C is differentiable at c.
h h→0 • Differentiablity =⇒ continuity.Summary • The complex function f is continuous at c if limz→c f (z) = f (c). • The elementary functions have the expected derivatives. 39 . • The complex function f is complex differentiable at c if lim f (z + h) − f (z) . h ∈ C.
vx (a. y) + iv(x.2 (CauchyRiemann equations) If f (z) = u(x. 40 .) Theorem 8. b) − iuy (a.1 Suppose f (x + iy) = u(x. Corollary 8.e. which is not true anyway! Examples 8. Likewise.8 CauchyRiemann Equations If f is a differentiable complex function. vx and vy all exist at (a. We know we get f (c + h) − f (c) → f (c) as h → 0. Equate real and imaginary parts in the theorem above. h u(a + k. b) k exists and equals Re(f (c)). y) = x2 − y 2 and v(x. b) Therefore.e. y). b) − u(a. k u(a + k. Let h = k + il. k = 0. i. (But not vice versa. And now for a result that is fundamental in all complex analysis courses. then u(a + k. → Re(f (c)) as k → 0. The two equations are called the CauchyRiemann equations after two of the founders of complex analysis. b + l) − u(a. see earlier warning. where z = x + iy. So ux (a. y) = 2xy. b) → f (c) as k → 0.3 (i) Let f (z) = z 2 = (x + iy)2 . u(x. We see that ux = vy = 2x and uy = −vx = −2y. then ux = vy and vx = −uy . so the CR equations hold. b) − u(a. Then. b) = Im(f (c)). and that f is differentiable at c = a + ib. y) + iv(x. uy . Note that the corollary says that if f is differentiable. i. b) + iv(a + k. b + l) − iv(a. The second equation follows from letting h → 0 through imaginary values. k + il So if we let h → 0 through real values. y) is differentiable. l = 0. but says nothing of the converse. b) + iv(a + k. b) + ivx (a. then the real and imaginary parts are differentiable as real functions. Proof. Proof. b) and f (c) = ux (a. b) − iv(a. b) → f (c) as k+il → 0. b) = vy (a. b). Then the partial derivatives ux . then the equations hold.
For example. Then wxx = 2 and wyy = −2.4 Show that the function f (z) = z is not differentiable anywhere in C.6 The converse to Corollary 8. 1. The condition is stronger because we require f (c) to exist as h → 0 from all directions. This fact won’t be used later. but f is not continuous. Then. if ux .(ii) Let f (z) = z2 = x2 + y 2 . uy = 2y. h h h so f (0) = 0. vx = 0. ux = 2x.7 Let w(x. Exercise 8. otherwise. Then u = x2 + y 2 and v = 0. because for real points of C the function f (x) = x is differentiable. vx . if x = 0 or y = 0.5 The preceding exercise shows that complex differentiability is imposing a stronger condition real differentiability. (You know this last fact well. vy exist near c and are continuous at c. vy = 0. provided x = 0. Then w is a harmonic function if it satisﬁes Laplace’s equation: wxx + wyy = 0. Remark 8. so can’t be differentiable. so wxx +wyy = 0. then f is differentiable at c. However. except possibly at 0. y) = x2 −y 2 . ¯ f (0 + h) − f (0) h2 − 0 hh ¯ = = = h → 0 as h → 0. (check!). The only place where the CR equations are satisﬁed is x = y = 0. Remark 8.8 Let w(x. let f (z) = 0.2 is false. Example 8. So f is differentiable nowhere. but is useful to know. the equations are satisﬁed. 5A C k function is a function that is differentiable k times 41 . Thus. and satisfy the CR equations. uy . y) be a C 2 function of two real variables5 . I hope!). Is it differentiable at 0? Well. Harmonic functions Deﬁnition 8. not just real ones.
and 2x = uy = −vx =⇒ v = −x2 + h(y) where h is a function of y. Example 8. y) is differentiable. y) is complex differentiable. and u and v are C 2 functions. y) + i(v. Complex analysis provides many examples of harmonic functions. • The function w : R2 → R is a harmonic function if it satisﬁes Laplace’s equation: wxx + wyy = 0. and so.Laplace’s equation is important in potential theory and many other areas. harmonic functions are of particular interest. 42 . since they are solutions of it. Conversely. y) + iv(x. then ux = vy and vx = −uy .9 If f : D → C is complex differentiable. Such a v is called a harmonic conjugate of u. • For every harmonic u there is a harmonic function v (called a harmonic conjugate of u) such that f (x + iy) = u(x. Theorem 8. We use the CR equations: uxx + uyy = (ux )x + (uy )y = (vy )x + (−vx )y = vxy − vyx = 0. then u and v are both harmonic: uxx + uyy = 0 and vxx + vyy = 0. given an harmonic function u there is locally (i. then u is harmonic. as the next theorem shows. Summary • (CauchyRiemann equations) If f (z) = u(x.e. y) + i(v. we can deduce that v = y 2 − x2 + C where C is a constant.10 Let u(x. aren’t they!) Because 2y = ux = vy =⇒ v = y 2 + g(x) where g is a function of x. y) = 2xy. Similarly for v. We can construct v using the CR equations (they’re very useful. y) is complex differentiable. in some εneighbourhood) a harmonic function v such that f (x + iy) = u(x. Proof.
Then. i. γ(0) = 2 and γ(1) = 2+i. Example 9. f (z) = z 2 . f (z) dz = γ γ b F (z) dz F (γ(t))γ (t) dt a n aj = = j=1 n F (γ(t))γ (t) dt aj−1 aj = j=1 n aj−1 (F ◦ γ) (t) dt [(F ◦ γ)(t)]aj by the usual FTC for RI functions. aj ] is continuous for all j. by the FTC. Let a = a0 < a1 < · · · < an = b be a dissection of [a. b] such that γ [aj−1 . 1 Obviously.3. Then. Then. f (z) dz = F (γ(b)) − F (γ(a)).1 (Fundamental Theorem of Calculus) Let f : D → C be a continuous complex function and γ : [a. Suppose there exists a complex differentiable F : D → C such that F = f .e. We now see this in a contour integration setting. F (z) = f (z). aj−1 j=1 = = F (γ(b)) − F (γ(a)).9 Fundamental Theorem of Calculus for Complex Functions One of the best theorems in Real Calculus is the Fundamental Theorem of Calculus.2 Consider Example 4. Theorem 9. F (z) = 3 z 3 is an antiderivative for f . Then. γ Proof. 3 43 . f (z) dz = γ 1 (γ(1))3 − (γ(0))3 3 1 3 = 2 − (2 + i)3 3 11 = −2 + i. b] → D be a contour.
4 Let f (z) = 1/z. Then f (z) dz = 0. z Therefore. not even differentiable ones. there exists a contour γ : [a. the analogue is not true for continuous complex functions. if there existed an F : D → C such that F = f . Hence. γ Example 9. then the corollary would be contradicted. Deﬁnition 9. Property of exponential We can now prove a result we would expect to be true by analogy with real analysis: if a function has zero derivative. f (z) dz = F (γ(b)) − F (γ(b)) = 0. So. γ Proof. Consider this corollary of the FTC and the following example. Then. Examples 9. f is differentiable on D = C\{0}. b] → D such that γ(a) = α and γ(b) = β. The deﬁnition of a closed contour is that γ(a) = γ(b).3 With the assumptions of the above theorem suppose that γ is any closed contour. then it is constant. Corollary 9. 44 . by the fundamental example we know f (z) dz = γ γ 1 dz = 2πi.6 (i) The domains D = C and D = C\{0} are both connected.5 A domain D is called connected if for each pair α. So a domain is connected if we can draw a curve from any point to any other. First we need a deﬁnition. antiderivatives do not always exist. Then. Unfortunately.How common are antiderivatives for continuous complex functions? Do they always exist? For real continuous functions we know that the Riemann integral can be found and this will be an antiderivative. β ∈ D. traversed once anticlockwise. Let γ be the unit circle round the origin.
So.8 For all complex numbers z and w we have ez+w = ez ew . This theorem allows us to prove the property of ez in Theorem 1. Proof. f is constant. There is no way to construct a contour starting below and ﬁnishing above the real line.35 we did not prove earlier. Corollary 9. Then. Take any α and β in D. as D is connected. 45 . there exists a path γ : [a. b] → D. By the product rule we get f (z) = −ez eα−z + ez eα−z = 0.7 Suppose that D is is connected domain. Then. By the FTC f (β) − f (α) = f (γ(b)) − f (γ(a)) = γ f = γ 0 = 0. then we get e e w (w+z)−w f (w) = f (0) = e0 ew+z+0 ew ez = ew+z . Thus. by the theorem. if we let α = w + z. such that γ(a) = α and γ(b) = β. where α is any complex number. f only depends on the end points of γ for any contour γ in D. Proof. such that f (z) = 0 for all z ∈ D. without crossing that line.(ii) The domain D = C\{z : z is real} is not connected. we deduce that f is constant. (i) There exists an F : D → C such that F = f . But the real line is not in D. The following statements are equivalent.9 Let f : D → C be a continuous complex function on a connected domain D. and f : D → C is analytic. * Existence of antiderivatives The next proposition gives some conditions equivalent to the existence of antiderivatives. Since these were general points of D. the function is constant. (ii) (iii) γ γ f = 0 for every closed contour γ in D. Proposition 9. Let f (z) = ez eα−z . Thus f (β) = f (α). Theorem 9.
Proof. We shall prove (i)⇒(ii)⇒(iii)⇒(i). (i)⇒(ii): This is just Corollary 9.3. (ii)⇒(iii): Suppose that α and β are two points in D, then we want to prove that the integral of f over any two contours γ1 and γ2 both starting at α and ﬁnishing at β are equal. Consider the contour γ1 − γ2 . This is a closed contour because it starts and ﬁnishes at α. (Draw a picture!) Hence, (ii) implies that γ1 −γ2 f = 0. We have 0=
γ1 −γ2
f=
γ1
f−
γ2
f.
Hence (iii) is true. (iii)⇒(i): We shall deﬁne a function F and show that it is differentiable. Fix z0 ∈ D. Since D is connected, for any point z ∈ D, there exists a contour γ from z0 to z. We deﬁne F (z) =
γ
f (w) dw.
[Note that this really is a function of z because γ has z as its endpoint.] Now we show that F (z) = f (z). Let z1 ∈ D be any point. Then by deﬁnition F is differentiable at z1 with derivative f (z1 ) if F (z1 + h) − F (z1 ) = f (z1 ). lim h→0 h So, as D is a domain, there exists an ε1 neighbourhood at z1 for some ε > 0. If h < ε1 , then there exists a contour Λ : [0, 1] → D which gives the straight line from z1 to z1 + h, i.e. Λ(t) = z1 + ht. Using our deﬁnition of F , we get F (z1 + h) =
γ+Λ
f=
γ
f+
Λ
f.
Therefore, F (z1 + h) − F (z1 ) 1 = h h We have, f (z1 ) dw = f (z1 )
Λ Λ
f+
γ Λ
f−
γ
f
=
1 h
f (w) dw.
Λ
dw = f (z1 ) [w]z0 +h = f (z1 )h. z0
Hence, F (z1 + h) − F (z1 ) − f (z1 ) = h
46
Λ
f (w) − f (z1 ) dw. h
We want the LHS to tend to zero as h → 0, so we use the Estimation Lemma on the RHS. Since f is continuous, given any ε > 0, there exists a δ > 0 such that w − z1  < δ implies that f (w) − f (z1 ) < ε. We can assume that δ < ε1 . So, when z < δ and w ∈ Λ, we have f (w) − f (z1 ) ε < . h h The length of Λ is h, so we deduce from the Estimation Lemma that f (w) − f (z1 ) ε dw ≤ h. h h Λ Thus, F (z1 + h) − F (z1 ) − f (z1 ) ≤ ε for all h < δ. h
Since ε was arbitrary we deduce that the LHS is zero. This implies that F (z1 ) = f (z1 ). So, F is what we were seeking. Summary • Fundamental Theorem of Calculus: Suppose f : D → C is a continuous complex function and there exists F such that F = f . Then, f (z) dz = F (γ(b)) − F (γ(a)).
γ
• Not all functions have an antiderivative. • A domain D is called connected if for each pair α, β ∈ D, there exists a contour γ : [a, b] → D such that γ(a) = α and γ(b) = β.
47
10
Differentiability of Power Series
We have successfully deﬁned functions such as exp, sin and cos by power series. Now we would like to show that they are differentiable. We do this by investigating the differentiability of power series with positive radius of convergence. We will show later that if a function is differentiable at a point z0 , then near that point it can be given as a power series, so this investigation wil be useful for more than just the elementary functions. We prove: • A power series is differentiable in the obvious way: termbyterm. • The derivative of a power series has the same radius of convergence. • Power series are inﬁnitely differentiable. • The coefﬁcients of the power series can be given in terms of the derivatives. • The coefﬁcients of the power series are unique. Suppose that ∞ an z n is a series. Then, the obvious cann=0 didate for the derivative is the one produced by termbyterm ∞ n−1 . But, initially, we do do not differentiation6 , i.e. n=0 nan z even know that this sequence converges. So, we begin by proving that a power series and its ‘obvious derivative’ have same radius of convergence, and then show that this obvious derivative really is the derivative of the series. Lemma 10.1 The series ∞ an z n and n=0 convergence.
∞ n=0
nan z n−1 have the same radius of
Proof. The following proof is taken from Priestley p20. If ∞ nan z n−1 had radius of convergence R = 0, then it is n=0 ∞ n not difﬁcult to show that n=0 an z has the same radius of convergence. Therefore, we prove the converse: If ∞ an z n had radius of n=0 convergence R =, then so does ∞ nan z n−1 . n=0 Fix z with 0 < z < R, and choose ρ such that z < ρ < R. Then n n z nan z n−1  = an ρn . z ρ
6 Hopefully, studying maths has taught you that just because something is obvious doesn’t mean it’s true!
48
[If we are thinking like mathematicians. We know that g(z) = nan z n−1 is welldeﬁned for z < R. then we know a good way of doing this is to show that f (z) − g(z) = 0. h Certainly. M an ρn . for all z < R. there exists a constant M such that for all n z n ≤ M. nan z n−1  ≤ and so by the comparison test. Just as in the real case. Thus. Theorem 10. then bn → 0 as n → ∞ and so there exists a constant M such that bn  ≤ M for all n.] 49 . We shall show that f (z) exists and is equal to g(z). if the series bn converges. ρ thus.The series n(z/ρ)n is easily shown to converge. f (z) = ∞ nan z n−1 . this is true if f (z + h) − f (z) − g(z) → 0 as h → 0. by the ratio test. Again we follow Priestley’s proof.2 Suppose that f (z) = ∞ an z n has radius of convergence R > n=0 0. the result. gent. (Exercise!). Hence. let’s show that this really is the derivative of the series. In other words we want h→0 lim f (z + h) − f (z) − g(z) = 0. h This is our method of attack. Then. n=0 Proof. z nan z n−1 is absolutely conver Now.
0 nz (ii) The series S = n=1 sin nx is absolutely convergent for x ∈ n2 ∞ cos nx . we have f (z + h) − f (z) − g(z) h = = n=1 ∞ (z + h)n − z n an − nan z n−1 h n=0 n=0 ∞ ∞ ∞ an (z + h)n − z n − nz n−1 h n = h n=1 ∞ an k=2 n n−k k−2 z h use binom thm k n−2 ≤ h n=1 ∞ n−2 1 n(n − 1)an  zn−2−m hm 2 m m=2 ≤ z n=1 1 n(n − 1)an (z + h)n−2 . probably the most technical in the course. By Lemma 10. This n n=1 clearly diverges at x = 0. 50 . differentiat∞ 1 n−1 = (1−z)2 for z < 1. Examples 10. h 2 So as h → 0 we get f (z+h)−f (z) h n(n − − g(z) → 0 as required. Thus. f (z + h) − f (z) 1 − g(z) ≤ Kh. and I wouldn’t expect you to reproduce them in exams. The proofs are quite technical.3 (i) We can show that ing both sides gives ∞ ∞ 0 1 z n = 1−z for z < 1. so understand what they mean: The derivative of a series can be found by termbyterm differentation and the resulting series has the same radius of convergence.4 Prove Theorem 7. So. For h < ρ − z. These two results will certainly be very important to the course.11.1 1)an ρn−2 converges to K say. S is not a power series! R but termbyterm differentiation gives Exercise 10.For any z and h such that z < R and z + h < R. 2 ∞ n=1 Fix z and choose ρ with z < ρ < R. but I would like you to understand them.
Again. A priori we don’t know that we can equate coefﬁcients for inﬁnite series. n! n! Hence. Then. an = bn . (k − k + 1)ak ) + ((k + 1)((k + 1) − 1) . Lemma 10. . . Note that this is to be expected as a similar statement is true for real power series. Proof. if f is deﬁned on the disc of convergence.6. an z n = ∞ n n=0 bn z . by Corollary f (n) (0) f (n) (0) an = and bn = . f (k) (z) = (k(k − 1) . . then we get f (k) (0) = k!ak . we will see that any complex function differentiable at a point can be given by a power series. k! Proof. Proof. . 51 . Note that the Uniqueness Lemma is not trivial.5 If f (z) = ∞ an z n has radius of convergence R > 0. ∞ n=0 ∞ n n=0 bn z . Corollary 10. . .6 If f (z) = ∞ an z n has radius of convergence R > 0. an = bn for all n. then f is n=0 inﬁnitely differentiable. (n − k + 1)an z n−k and this of course holds for z < R. ((k + 1) − k + 1)ak+1 z) + . ∞ an z n = n=0 Then. This result is one of the reasons that power series are so great.7 (Uniqueness Lemma) Suppose for some R > 0. Then. Let f (z) = 10. Later. you should already know this is true for real series. . and f is a series. We can differentiate f termbyterm to get f and so on. This suggests the following: Corollary 10. By induction using the theorem. for all z < R. By induction we can prove ∞ f (k) (z) = n=0 n(n − 1) .We have seen that. If we put z = 0 into this. then ak = n=0 f (k) (0) for all k. then so is f . .
Then we translate back to say something about the series an (z − z0 )n . Suppose we have f (z0 + h) = ∞ an hn . If h < R for some R. 52 . then f (z) = nan (z − z0 )n−1 converges for z − z0  < R.9 If f (z) = an (z − z0 )n converges for z − z0  < R. if we have the power series an (z − z0 )n ..Remark 10. Thus we can deal with power series centred at a particular point z0 and we will have a disc of convergence centred at z0 . Its coefﬁcients can be given in terms of those of S1 and S2 . we can differentiate termbyterm: Example 10. etc. Let’s do that now. we shall delay proving this as later work will give a particularly simple proof of this fact.) Power series about points other than zero So far all our power series have been centred at the origin.6. with the series con0 vergent for h < R. Let z = z0 + h. which 0 converges for h = z − z0  < R. but which deﬁnes the function f around the point z0 . However. So for example.8 Just like in Real Analysis it is possible to prove that the product of two power series S1 and S2 is again a power series with radius of convergence at least the minimum of those for S1 and S2 . then in a neighbourhood of z0 we can get z0 + h to give any point. By substitution. well z0 is a constant and h is a single variable so z is a variable. if h is zero we get z0 . then we n can let h = z −z0 to get the series an h and we can apply ratio test. The main point is that we can prove results about power series centred at zero and just translate to another point. Just as in real analysis giving an expansion about a different point is very useful. then we have a disc around z0 . (See Theorem 15. f (z) = ∞ an (z − z0 )n . I. This is rather limiting. to that. Suppose we are interested in the point z0 ∈ C. For example. If we let h be a complex variable. For example. a power series in h about 0.e.
• f is inﬁnitely differentiable. • f can be differentiated termbyterm to get the derivative. n=0 • The coefﬁcients an are unique and an = 53 . f (n) (0) for all n. Summary Let f (z) = ∞ n=0 an z n have radius of convergence R > 0. • f has the same radius of convergence as f . We have f (z) = d d g(z−z0 ) = g (z−z0 ) (z−z0 ) = g (z−z0 ) = dz dz nan (z−z0 )n . n! • We can translate these results to series of the following form: f (z) = ∞ an (z − z0 )n .Proof: Let h = z − z0 . Then g(h) = an hn converges for all h = z − z0  < R as f (z) = g(z − z0 ). for z − z0  < R.
54 . Anyone who proves that the hypothesis is true can claim a million dollars from the Clay Institute. for the connection with number theory and the distribution of primes. for example if s = 1. Vol 168 issue 2264.11 Win a Million Dollars! The Riemann Zeta function We deﬁne the Riemann Zeta function as follows: ∞ ζ(s) = n=1 1 . One of them (the Poincar´ Conjecture) is explained in Homotopy and Sure faces in Year 3. The Riemann Hypothesis We now come to the million dollar question. If s = 2 we 1 get ζ(2) = ∞ n2 which does converge. This is called the Riemann Hypothesis and it is regarded as being the most important unsolved problem in pure mathematics. You can ﬁnd the details on the Web at http://www. For futher reading see Prime Time by Erica Klarreich in New Scientist. ns 1 So. then what happens? We ﬁrst we need to deﬁne ns when n is a real number and s is a complex number. There are six other problems from various areas of mathematics for which a million dollar prize is offered. We know that n = elog n and so we could deﬁne ns by ns = es log n . Where do the roots of the function lie? I. p32. and so on for all real n=1 numbers s. so ζ(1) is not deﬁned.e. Titchmarsh. 11/11/2000.claymath. (Edward Boyle Library Floor 11). then we get ζ(1) = ∞ n which we n=1 know does not converge. and so he conjectured 2 that all nontrivial roots have the property of being on this line. the only number we have done this for is e.org/ under the heading Millenium Prize Problems.C. If we let s be complex.) Riemann was not the ﬁrst to study this type of function but did leave us an interesting hypothesis. The theory of the RiemannZeta function. What Riemann found is that all roots (apart from some trivial real ones) seem to have Re(s) = 1 . the s ∈ C such that ζ(s) = 0. (Since obviously we want ns = (elog n )s = es log n . See also the classic text by E.
Deﬁnition 12. written n(γ. Then the winding number of γ about w. w). Both these notions are intuitively simple.3 Find the winding numbers for points in the various regions. 55 .1 Let γ be a closed contour and w a point not on γ. is the net number of times that γ winds about w. Example 12. with anticlockwise counted positively. Exercise 12.2 The winding numbers for points in the regions enclosed by the contour are shown below.12 Winding numbers Cauchy’s theorem is one of the most remarkable theorems in mathematics. To state it we need the notions of winding number and interior point. We start with the winding number.
piecewise continuous differentiable functions on [a. w) = 1 2πi dz . Then n(γ. z−w γ Proof. i. Prove that n(−γ. w ∈ C\γ. Putting these results into the integral above. θ(b)−θ(a) is the net increase in arg(γ(t)−w) as t runs from a to b.e.4 (Winding Number Lemma) Let γ be a closed contour. so the number of times that γ winds round w is θ(b) − θ(a) . Now. z−w γ Moreover. Lemma 12. dz = z−w = a b b a b γ γ (t) dt γ(t) − w r (t)eiθ(t) + iθ (t)r(t)eiθ(t) dt r(t)eiθ(t) b r (t) dt + i θ (t) dt r(t) a = a = [log r(t)]b + i [θ(t)]b a a = log r(a) − log r(b) + i (θ(a) − θ(b)) . (ii) Let γ1 and γ2 be closed contours. k = n(γ. r(a)eiθ(a) = r(b)eiθ(b) . so that the join can be taken. Prove that n(γ1 + γ2 .5 (i) Let γ be a closed contour and w ∈ C. then r(t) and θ(t) are continuous. i. dz = 2πik. Equating the moduli and arguments we get r(a) = r(b) and θ(b) = θ(a) + 2πk for some k ∈ Z.We can calculate winding numbers mathematically. γ is a closed contour and so γ(a) = γ(b). you have to trace your ﬁnger round the curve bearing in mind how many revolutions have been made. If we write γ(t) = w + r(t)eiθ(t) .e. And you have to do it for each region. w) + n(γ2 . We have. b]. 2π Exercises 12. there is an easier 56 . w) An easy method of calculation Calculating winding numbers can appear to be complicated to calculate. w). w) = −n(γ. w). w) = n(γ1 . Fortunately.
consider the complicated contour image drawn below with a line passing through it. w) n(γ. n(γ + C. so the point in the region we pass into have winding number −1. Now. z) = = = n(γ. n(γ + C. Justiﬁcation of the method We can justify the method by considering two points w and z that lie on opposite sides of a contour line that goes up. w) = n(γ. Then.6 Find the winding numbers for the points in the diagram. 57 . w) = n(γ. Obviously the winding number there is 0. w) + n(C. We can carry this out for all regions on the line. and so the winding number of points in the next region is 0.method. then the winding number increases by 1. We can assume that the contour starts and ﬁnishes at a point near w and z. To get other regions we can use different lines. the contour is travelling up when we ﬁrst meet it. z) − 1. (If it didn’t we can do some reparametrisation and join work. Similarly. • If we cross the contour so that it is travelling down. then the winding number decreases by 1. then the winding number decreases by 1. z). w) + 1 n(γ. as the diagram shows. In the examples above note that when passing from one region to another via an edge (rather than via a crossing of two line) the winding number for points in the regions only changes by 1. As we go from left to right on the line we will cross the contour. At the next crossing the contour is going down. This shows that if we pass from z to w. Thus. take a loop C round w. We apply the following rules: • If we cross the contour so that it is travelling up. w) n(γ.) Now. Start at the left side of the line. Exercise 12. one can show an increase by 1 for a contour locally heading down. In the above diagram.
If we go round the contour twice.Warning! 12. z−w γ • It is easy to calculate a winding number by eye. w ∈ C\γ. Example 12. 58 .9 The interior points of the contour in the next diagram are shaded.8 Let γ be a closed contour. We denote the set of interior points of γ by Int(γ). Example 12. Then n(γ.10 Let γ(t) = w + Re2πit . Interior points The method of counting the number of times a contour wraps round a point helps deﬁne the notion of an interior point. Recall that the contour and its image are different. w) = 1 2πi dz .7 We have been using the image of the contour to calculate the winding number of a point. The interior of γ is the set of points w ∈ C\γ for which n(γ. Int(γ) = {z : z − w < R}. 0 ≤ t ≤ 1. we have assumed that the contour is traversed only once. • An interior point is any point with nonzero winding number. Summary • The winding number is the number of times a contour wraps round a point in an anticlockwise direction. then the numbers calculated by eye have to be doubled. Effectively. Deﬁnition 12. if we go round k times. w) = 0. then we multiply the ‘by eye’ calculations by k. • Let γ be a closed contour. More generally. Then.
z dz.13 Cauchy’s (Fantastic) Theorem We now come to the fundamental theorem in complex analysis. It refers to any domain in C. w) = 0.1 (Cauchy’s Theorem) Let D ⊆ C be a domain. Theorem 13. and any contour with all interior points in D.) z2 1 z2 dz. sin dz. z−w so Cauchy’s theorem does not apply. but f= γ γ dz = 2πi n(γ. And it says that any integral arising from this is zero. There is no analogue in real analysis. to which of the following integrals does Cauchy’s theorem apply? (There is no need to evaluate them. Exercise 13. Then. deep consequences. Remarks 13. z−1 γ1 γ1 z − 2 γ2 γ2 +γ3 59 . (ii) It is important to note that the interior of γ lies within D. γ f = 0. However. dz. Note that we can’t just use the FTC since we don’t know that f has an antiderivative on D. If w is in the interior of γ but not in D.2 (i) This is truly a great theorem. it has strong implications for nondifferentiable functions as well. Let γ be a closed contour such that γ and its interior points lie in D. It has far reaching. weak assumptions lead to a strong conclusion. (iv) At ﬁrst sight it may appear that the theorem will only tell us about the behaviour of differentiable functions. Thus. and most of what we will prove from now on relies on this theorem. any analytic function on D. (iii) The proof of theorem is too complicated for the moment and we will do it later.3 Using the contours in Exercises 2 Question 1. consider f (z) = 1/(z − w) which is analytic on D = C\{w}. and f : D → C be a differentiable complex function.
How was this done? The trick follows from Cauchy’s theorem and the funda1 mental example. a I was trying to integrate dz. Now take another path from the end of γ to the start of the circle C. where a was a number choγ z sen at random. So it doesn’t contain the origin. Its interior does not include the disc encircled by C. Thus. However. and the interior of Γ is a subset of D. Consider the contour Γ = γ + β + C − β.The calculation trick Recall. Call this contour β. But a/z is differentiable on D = C\{0}. This is a closed contour. (recall that the example says that dz = γ z 2πi where γ is a small circle round the origin). Now. the trick I performed where I calculated an integral over a path. we can apply Cauchy’s theorem: a dz z a dz z a dz z a dz z a dz z a dz z = 0 = 0 = 0 = 0 = 0 a dz C z 1 = a dz −C z = a × 2πi = − Γ γ+β+C−β a a a dz + dz dz + γ z β z C z −β a a a dz + dz dz − γ z β z C z β a dz + γ z C γ γ a dz = 2aπi z 60 . and we can make it so small that we can assume it lies totally within the interior of γ. and γ was a path chosen at random. we can take a small circle C going clockwise round the origin. before the path had even been deﬁned. you were forced into choosing γ so that its winding number was 1.
Then. and f : D → C be a differentiable complex function. γ f = 0. Let γ be a closed contour such that γ and its interior points lie in D. Summary • Let D ⊆ C be a domain. 61 .Thus we get the answer expected. This method – of cutting out a disc where the function is not deﬁned – will be used again later in more generality so make sure you understand this simpler example.
and f : D → C be differentiable. (iii) The special case f (z) = 1 for all z ∈ D gives the winding number lemma. w) f (z) dz. (iv) Fudamental Theorem of Algebra: Every complex polynomial has a complex root. 62 . [Step 1] Let β be a contour in D\{w} from the start point of γ to the start point of γr .1 Let D ⊆ C be a domain. (0 ≤ t ≤ 2π). z−w γ Remarks 14.2 (i) If w is in the interior of γ. (ii) Liouville’s Theorem: Any differential function bounded on the whole of C is contant. Cauchy’s Integral Formula Theorem 14. where r > 0 is sufﬁciently small so that γr is contained in the interior of γ. If w ∈ D\γ. (ii) This behaviour is sometimes called ‘action at a distance’. w)f (w). (iii) The Maximum Modulus Principle: The modulus of a function on a domain achieves its maximum on the boundary of the domain. Proof (of Theorem 14. then f (z) dz = 2πi n(γ. (i) Cauchy’s Integral Formula: The value of a function at z0 is determined by the values on a contour round z0 .1). then f (w) = 1 2πi n(γ. Let γr be the circular contour γr = w + reit . Let γ be a closed contour such that γ and its interior points lie in D. z−w γ This says that the values of f inside γ are completely determined by those on γ! Remarkable! This contrasts with real analysis.14 Strange Consequences of Cauchy’s Theorem We will prove a number of surprising theorems that can be deduced from Cauchy’s theorem.
z−w z∈γr = Therefore γr f (z) dz − 2πif (w) z−w f (z) − f (w) → f (w) as z → w. f (z) dz = f (w) γr z − w dz f (z) − f (w) + dz z−w γr z − w γr f (z) − f (w) = 2πif (w) + dz. z−w (Why does f (w) exist?) Hence RHS → f (w). Now γ f (z) dz − 2πin(γ. has winding number f (z) zero about w. z−w We have. z−w 63 . by winding lemma. γ e + γ β − γr −··· − γr + −β γ f (z) dz − n(γ. w) γr f (z) dz − 2πin(γ. w)f (w) z−w f (z) dz − 2πif (w) . w) copies of (−γr ). w) z−w γr γ f (z) dz z−w f (z) dz z−w f (z) dz z−w f (z) dz z−w = 0 = 0 = 0 = n(γ. r→0 γ z − w r Thus we can combine these steps to get by the Estimation Lemma. so by Cauchy’s theorem applied to on z−w D\{w}.w) times where there are n(γ. n(γ. z−w γr f (z) − f (w) dz z−w γr f (z) − f (w) ≤ sup × 2πr. w) γr f (z) dz.0 = 0 as r → 0.The contour γ = γ + β + (−γr ) + · · · + (−γr ) +(−β). z−w [Step 2] We shall now show that r→0 lim γr f (z) dz = 2πif (w). Therefore. w)f (w) = z−w n(γ. w) γr = n(γ. f (z) lim dz = 2πif (w).
and is bounded. z−π γ We apply CIF with f (z) = sin z and w = π. 6 iπe3 = . Then. 0 ≤ t ≤ 4π.e. z such that z − 2 = 2. f is constant. Examples 14.1.e. Then. but the LHS is independent of r and so must be 0. Theorem 14.4 Suppose f is differentiable on the whole of C. 64 . and n(γ. i.3 sin z (i) Calculate dz. z−π (ii) Let γ be a circle of radius 2 about 2. w) = −2. 3 Liouville’s Theorem The next theorem is also rather surprising. where γ(t) = 3e−it . i. z−2=2 ez dz = z2 − 9 1 ez 1 ez − dz 6z +3 z−2=2 6 z − 3 1 ez 1 ez = dz − dz 6 z−2=2 z − 3 6 z−2=2 z + 3 1 = 2πie3 − 0. (Draw a picture!) Hence.The RHS tends to 0 as r → 0. γ sin z dz = 2πi × (−2) × sin(π) = −4πi. by 14. because γ winds round w clockwise twice. The point w lies within the circle formed by γ. Draw the contour and indicate where the integrand is not differentiable. there exists M such that f (z) ≤ M for all z ∈ C.
That is.6 Show cos and sin are not bounded on C. Maximum Modulus Principle Theorem 14. Then.] Let α ∈ C. then f (w) ≤ M for all w ∈ Int(γ). for example. This is differentiable on all R and  sin x ≤ 1 for all x ∈ R but it sin is not constant. So we need to show f (α) − f (0) = 0 for every α ∈ C. f is constant. If f (z) ≤ M for all z ∈ γ. As neither f (α) nor f (0) depend on R we must have f (α)−f (0) = 0.2πR 2π R(R − α) M α . [HTTLAM: We want to show the function is constant. consider sin. This is true if f (α) = f (0) for an arbitrary α. 65 .7 Let f : D → C be a differentiable function and γ be a closed contour such that Int(γ) ⊂ D. Remark 14. so as R → ∞. by CIF 2πi γ z − α 2πi γ z 1 1 1 f (z) − dz 2πi γ z−α z 1 f (z)α dz 2π γ z(z − α) 1 α M sup L(γ) by 2π z∈γ z(z − α) 1 α M .Proof. R − α But R was effectively arbitrary. In other words the maximum modulus occurs on the boundary of a region.5 Contrast this with real analysis. f (α) − f (0) = = = ≤ ≤ = 1 f (z) f (z) 1 dz − dz . Remark 14. Being bounded does not imply constant.8 The theorem says that the modulus of a function within the interior of a contour is never bigger than the modulus of the function on the contour. We let R ≥ 2α and γ(t) = e2πit for 0 ≤ t ≤ 1 be a circle of radius R round the origin. the LHS→ 0. Exercise 14.
Thus.Proof. zn z z So there exists an R such that z ≥ R implies that This in turn means f (z) = 1 2 2 ≤ n ≤ n for z > R. Then. p(z) z  R 66 p(z) 1 ≥ . w) for all z ∈ γ. letting k → ∞ gives f (w) ≤ M . n ≥ 1. Since p(z) = 0 for all z ∈ C the function f deﬁned by f (z) = 1/p(z) is differentiable on all of C. where k is a natural number: f (w)k = 1 2πin(γ. (which is in fact not really a theorem of algebra but of analysis!) Theorem 14. . obviously z − w ≥ dist(γ. w) M. f (w)k ≤ since Mk 1 L(γ). So. We shall apply CIF to w ∈ Int(γ) and f (z)k . n z 2 . Fundamental Theorem of Algebra We are now in a position to prove the Fundamental Theorem of Algebra. w) f (z)k dz. w) f (w)k Mk ≤ on γ.9 Every polynomial p(z) = z n + an−1 z n−1 + . p(z) an−1 a0 = 1+ + · · · + n → 1 as z → ∞. . (this 1 is true because x1/k = exp( k ln x) → exp(0) = 1). Now. by the Estimation Lemma. Suppose not and derive a contradiction. w) 1/k Therefore. 2π n(γ. Now. f (w) ≤ L(γ) 2π n(γ. z−w γ Deﬁne the distance from w to γ by dist(γ. This proof contains a nice trick. z − w dist(γ. we use the fact that limk→∞ x1/k = 1 for all x > 0. w) dist(γ. w) = inf{z − w : z ∈ γ}. Proof. w) dist(γ. a1 z + a0 . for z = 0. has a root in C.
• Liouville’s Theorem: Any differential function bounded on the whole of C is contant. and hence so is p. f is conR stant. 67 . • The Maximum Modulus Principle: The modulus of a function on a domain achieves its maximum on the boundary of the domain. so p has a root. • Cauchy’s Integral Formula: The value of a function at z0 is determined by the values on contours round z0 . R 2 So f (z) ≤ n on all of C. Summary • Cauchy’s theorem can be used to prove surprising theorems that have no analogues in real analysis. • Fudamental Theorem of Algebra: Every complex polynomial has a complex root. This is a contradiction.By the Maximum Modulus Principle applied to f with γ being the standard circle of radius R round the origin we have z < 2 R =⇒ f (z) ≤ n . By Liouville’s theorem.
Then. ∞ k=0 Mk < ∞ as h < r. C\D) (which equals inﬁnity if D = C. Fix r with 0 < r < R. z − (z0 + h) Cr Now.) Then. there exists a Taylor expansion of f about z0 . usf (z) f (z)hn ing g(z) = . (z − z0 )n+1 Now we shall integrate termbyterm. Furthermore. for all z − z0  < R. Proof. Thus.e. there exist an ∈ C such that ∞ f (z) = 0 an (z − z0 )n . (z − z0 )n+1 r z−z0 =r so. (see Corollary 6. f (z) f (z) = z − (z0 + h) (z − z0 ) 1 − ∞ h z−z0 f (z) = z − z0 ∞ n=0 h z − z0 n = n=0 f (z)hn . i. and gk (z) = . by 68 . (Use the ratio test).7). deﬁne Mk by z − z0 (z − z0 )n+1 Mk = sup gk (z) = sup z−z0 =r z−z0 =r f (z)hn hn = sup f (z) n+1 . (We will replace z with z0 + h so that we have a power series in h centred at 0). For each h with h < r. Let z0 ∈ D and R = dist(z0 . Cauchy’s Integral Formula gives f (z0 + h) = 1 2πi f (z) dz.15 Taylor’s Theorem Taylor’s theorem is a consequence of Cauchy’s theorem but is so important we give it a separate section.1 (Taylor’s theorem for complex functions) Suppose f : D → C is a differentiable function. (z − z0 )n+1 Cr where Cr is a small circle about z0 and r is any number with 0 < r < R. Theorem 15. an = 1 f (n) (z0 ) = n! 2πi f (z) dz. with h < r and z − z0  = r.
(z − z0 )n+1 f (z0 + h) = k=0 1 2πi Cr Therefore. f (z) = f (0) + f (0)(z − 0) + = −i − 3z + 3iz 2 + z 3 . Analytic and differentiable are not equivalent in real analysis as we see in a later example.3 (i) Find the Taylor series of f (z) = (z + i)3 at z = 0.7. f is analytic. Example 15.2 When solving problems concerning differentiable complex functions. 3(0 + i)2 = −3.Corollary 6. f (0) f (3) (0) (z − 0)2 + (z − 0)3 2! 3! . 0. whether theoretical or practical ones. That is. f (z) dz = z − (z0 + h) 2πif (z0 + h) = k=0 ∞ Cr ∞ Cr Cr k=0 ∞ f (z)hn dz (z − z0 )n+1 f (z) dz hn n+1 (z − z0 ) f (z) dz hn (z − z0 )n+1 1 2πi f (z) dz. That we also have an = (z0 ) follows from Corollary 10. This explains why some authors use the word analytic to mean complex differentiable: differentiable and analytic are equivalent in complex analysis. if we substitute z = z0 +h and let an = we get the series f (z) = n=0 ∞ Cr an (z − z0 )n . 69 f (n) = = = = = (z + i)3 3(z + i)2 6(z + i) 6 0 =⇒ =⇒ =⇒ =⇒ =⇒ f (0) f (0) f (0) f (3) (0) f (4) (0) = = = = = (0 + i)3 = −i. HTTLAM 15. f (z) f (z) f (z) f (3) (z) f (4) (z) So. The expression for an is sometimes known as Cauchy’s formula for derivatives. n! The theorem says that there exists an Rneighbourhood of z0 upon which f is a power series. 6(0 + i) = 6i. 6.6. We have. use the fact that the function can be written locally as a power series.
Suppose z0 is any point in D. which contrasts sharply with real differentiability. f can be given as a power series.e.4 (Inﬁnite Differentiability) Suppose f : D → C is differentiable on the domain D.Note that this expansion is valid for all z ∈ C as f is differentiable on all of C.e. in some neighbourhood of z0 . This derivative is not differentiable at 0.5 Suppose f : R → R is deﬁned by f (x) = x2 x > 0 0 x ≤ 0. Example 15. {1}) = 1. i. f is differentiable with derivative f (x) = 2x x > 0 0 x ≤ 0. and so is inﬁnitely differentiable by Corollary 10. f is inﬁnitely differentiable. Corollary 15. It is easy to show that n=0 f (n) (0) (z − 0)n = n! ∞ n=0 n! n z = n! ∞ zn. n=0 This is valid for all z < 1 as f is differentiable on C\{1} so R = dist(0. Then. 1 = f (z) = 1−z ∞ 1 about 0.5. i. it has derivatives of all orders. Some facts about power series We now use Taylor’s Theorem to prove some facts about series which might otherwise be difﬁcult to demonstrate. R = dist(0. (ii) Expand f (z) = Thus. Then. so f (n) (0) = n!. so f is not inﬁnitely differentiable. This is really astounding. 70 . C\C) = ∞. From the theorem we know that. Now we come to a very surprising and useful theorem. C\ (C\{1})) = dist(0. Note that this conﬁrms a result we already know. It doesn’t happen for real functions as the next example illustrates. Proof. 1−z n! f (n) (z) = (1−z)n+1 .
Deﬁne h(z) = 1/g(z). By continuity of g and the fact that g(0) = 0.6 Suppose f (z) = ∞ an z n for all z < R1 and g(z) = ∞ bn z n n=0 n=0 ∞ for all z < R2 . n! Now we can apply Leibnitz’s rule n (f g)(n) (0) = k=0 n n! f (k) (0)g (n−k) (0) k!(n − k)! n! f (k) (0) g (n−k) (0) k! (n − k)! ak bn−k . where cn = k=0 n ak bn−k . we get ∞ (f g)(z) = n=0 (f g)(n) (0) n z . and the result follows from the theorem applied to f h. * Morera’s Theorem There is a partial converse to Cauchy’s theorem. (z < R). By Theorem 15. and by Corollary 10. f and g are complex differentiable for all z < R.Theorem 15.8 The precise value of the radius of convergence will depend on where g is zero.7 Suppose f and g are power series with positive radii of convergence. If g(0) = 0. Then. where cn = n ak bn−k .6. Corollary 15. then f /g has a power series expansion at 0 with positive radius of convergence. Remark 15. f (z)g(z) = (f g)(z) = cn z n for all n=0 z < min{R1 . Proof. there is a εneighbourhood of 0 upon which h is differentiable. R2 }. It relies on the earlier optional material on the Fundamental Theorem of Calculus. k=0 Proof. The product of two differentiable functions is differentiable by the product rule. ∞ n (f g)(z) = n=0 cn z .1 it has a series expansion for z < R. h has a power series expansion. Thus. = k=0 n = n! k=0 Therefore. Let R = min{R1 . 71 . R2 }. So. f g is differentiable. Then.
S2 }. Then. f is differentiable. • Differentiable complex functions are inﬁnitely differentiable.9 Let f : D → C be a continuous map on a connected domain D such that γ f = 0 for all contours f . • If S is a power series at z0 with positive radius of convergence. then S1 S2 is a power series with radius of convergence min{S1 . Summary • If a complex function is differentiable on a domain. then 1/S has a power series expansion at z0 provided S(z0 ) = 0. • There is a partial converse to Cauchy’s Theorem.9 that there exists F : D → C such that F = f .4 the derivative of a differentiable function is itself differentiable. • If S1 and S2 are power series with radii of convergence R1 and R2 . But by Corollary 15.Theorem 15. 72 . Proof. then at every point there is power series expansion valid on some εneighbourhood. We know from Theorem 9.
We now know that if f : D → C is differentiable at z0 .35(iv) f (z) = ez has no zeros. Remark 16.16 Zeros of functions We have met a number of functions that are not analytic (i. if f (z0 ) = 0 and f (z0 ) = 0. suppose that f is a polynomial with a root of multiplicity m at z0 .1. 73 . (ii) ez 2 +9 (z + 4i).4 We say f has a zero of order m at z0 if a0 = a1 = · · · = am−1 = 0 but am = 0. and one at z = 0. (iii) f (z) = z 3 (z 2 + 1) has a zero at z = i. (ii) By Theorem 1. In particular. For example. f has a zero of order m at z0 . Zeros Deﬁnition 16. Examples 16. (iii) ez−2 − 1.e. Therefore. Then. 1/z is not analytic at 0. because it is undeﬁned when z is zero. Examples 16.6 (i) The function f (z) = z 2 has a zero of order 2 at 0. Exercise 16. for all k ∈ Z. differentiable) at certain points.2 (i) The function sin has zeros at kπ.3 Find the zeros of: (i) z 2 + 9. (ii) The function f (z) = z(z + 2i)3 has a zero of order 1 at 0 and one of order 3 at −2i. it seems natural to investigate zeros of functions ﬁrst. one at z = −i.1 An analytic function f : D → C has a zero at z0 if f (z0 ) = 0. then f has a zero of order 1. Deﬁnition 16. f has a zero of order m at z0 if and only if f (j) (z0 ) = 0 for all j < m and f (m) (z0 ) = 0. by Theorem 15. (iii) More generally. it has a Taylor series expansion about z0 : f (z) = ∞ an (z − z0 )n for n=0 all z with z − z0  < R for some R > 0.5 Obviously.
Then f has a zero of order m at z0 if a0 = a1 = · · · = am−1 = 0 but am = 0. • Suppose f (z) = ∞ an (z − z0 )n for all z with z − z0  < R n=0 for some R. there exists a differentiable g and an R > 0.8 Suppose that f : D → C is an analytic function with a zero of order m at z0 .7 Find the zeroes and their orders of the following: (i) (z −1)3 (z +1) (ii) (z 2 +1)2 . (v) z 3 +1.Exercise 16. Summary • An analytic function f : D → C has a zero at z0 if f (z0 ) = 0. 2 (iv) (2z −3i)4 . for all z − z0  < R. 74 . (iii) zez . and g(z0 ) = 0. Then. such that f (z) = (z − z0 )m g(z). Exercise 16.
.17 Poles The notions of pole and residue are crucial to the application of complex analysis to real problems. . where z = 1 2 z=1 z denotes the standard contour of a unit circle round the origin. 7 These transforms are used in Applied Mathematics. By the Fundamental Example. then don’t worry about it. that can be represented by a series with negative powers. . then ﬁne. If you know what they are. After that we deﬁne their residue. Let us consider an example: Integrate ez dz = z2 z2 z3 + + . Recall for an analytic function we can represent it as ∞ f (z) = n=0 an (z − z0 )n for z − z0  < R. . dz 2! 3! z=1 1 1 1 z = + + + + . . . or Fourier and Laplace Transforms7 . such as real integrals. . This is key to understanding the terms pole and residue we are going to deﬁne. 6 Notice that only the 1/z term mattered. dz z2 z 2 6 z=1 1 1 1 = dz + dz + dz + 2 z=1 z z=1 z z=1 2 = 0 + 2πi + 0 + 0 + . we should check that we can do this! 75 . Laurent expansions First we shall look at functions. A motivating example ez dz. 1 z2 1+z+ z=1 z=1 z dz + . like the integrand ez /z 2 example above. . Let’s just integrate termbyterm8 and see what happens. . We deﬁne poles in this section and show how to locate them in the next. If not. We can’t use Cauchy’s Integral Formula as the integrand is not analytic at 0 and the Fundamental Theorem of Calculus doesn’t help as we can’t see any obvious antiderivative. Compare this with the following series. 8 Obviously. all the other terms were irrelevant.
Example 17. does not necessarily equal f (z0 ) when z = z0 .2 (i) f (z) = 1/z has an expansion at 0 with an = 0 for n = −1. a−1 = 1. Then we say f has a Laurent expansion at z0 if there exist an ∈ C and R > 0. and aj = 0 for (n + 2)! j ≤ −3. We say that f has a pole of order N at z0 ∈ D. (which may be undeﬁnable anyway). z2 z z 2 6 1 for n ≥ 0. (ii) The expansion of f (z) = ez /z 2 at 0. and R = ∞. The radius is R = ∞.5 The point is that for a pole of ﬁnite order there are only a ﬁnite number of negative exponents of z in the series.. then we say f has an essential singularity at z0 . any function differentiable at z0 has a Laurent expansion. and that the expansion.Deﬁnition 17. has an = (iii) Any power series with a positive radius of convergence is a Laurent expansion. Hence. If there is no N ≥ 0 such that an = 0 for all n < −N . Note that the limits go from −∞ to ∞.3 sin z Find a Laurent expansion for f (z) = 4 deﬁned on C\{0}. with a−N = 0. such that ∞ f (z) = n=−∞ an (z − z0 )n for all z with 0 < z − z0  < R.. z The deﬁnition of a pole Deﬁnition 17.1 Suppose that f : D → C is complex function. Poles are also called singularities.4 Suppose that f : D → C is a complex function. Exercise 17. a−1 = a−2 = 1. Remark 17. if there exists a Laurent expansion at z0 of the form ∞ f (z) = n=−N an (z − z0 )n (0 < z − z0  < R). 76 . 1 1 1 z ez = 2 + + + + .
9 Show that (ez − 1)/z has a removable singularity. for 0 < z − z0  < R. That is.8 sin z The function f (z) = is deﬁned on C\{0}. f (z) = sin z . then the coefﬁcients are unique. To see this look at (z − z0 )N f (z)..7 (i) If such a Laurent series exists. 2 z z 2 6 = 3 at 4i. we say f has a removable singularity. for z = 0. n! (−n)! n=−∞ 0 and this obviously has an inﬁnite number of terms with negative exponent... .6 (i) order(1/z 5 ) = 5 at 0. . this is a power series and so has unique coefﬁcients.. Exercise 17. we can deﬁne f (z0 ) = a0 and thus make f analytic at z0 .. Example 17.e. 3! 5! So deﬁne f (0) = 1 to make f analytic on all of C. i. If we can do this.. (ii) Suppose that f (z) has a Laurent expansion with N = 0. i + (z − 4i)2 − 2i(z − 4i)5 (z − 4i)3 (iv) f (z) = exp(1/z) has an essential singularity at 0 because ∞ f (z) = n=0 (1/z)n = n! ∞ n=0 z −n zn = . and we have z f (z) = sin z z 1 z3 z5 = z− + − . f (z) = a0 + a1 (z − z0 ) + a2 (z − z0 )2 + . Remarks 17. for z = 0. z 1.Examples 17. 77 ... Then. (ii) order(ez /z 2 ) = 2 at 0 as f (z) = (iii) order 1 1 1 z + + + + . z 3! 5! z2 z4 = 1− + − .
is called a Laurent expansion at z0 .Summary • An expansion of f of the form ∞ f (z) = n=−∞ an (z − z0 )n for all z with 0 < z − z0  < R. if there exists a Laurent expansion at z0 of the form ∞ f (z) = n=−N an (z − z0 )n (0 < z − z0  < R). • We say that f has a pole of order N at z0 ∈ D. with a−N = 0. 78 .
f (z) = = p(z) q(z) p(z) (z − z0 )N r(z) g(z) = (z − z0 )N = 1 an (z − z0 )n (z − z0 )N n=0 a0 a1 a2 = + + + . and does it for general functions not just polynomials. some R > 0. and has a zero of order N at z0 . So as a0 = 0. Lemma 18. The following puts that intuition into a precise mathematical statement. Thus. because p(z0 ) = 0 by assumption and r(z0 ) = 0 by deﬁnition. and p(z0 ) = 0. You may have noticed that the order is somehow connected with the order of multiplicity of the polynomial in a denominator. valid for z − z0  < R. Proof.18 How to ﬁnd poles It is hard to give nontrivial examples of poles without using some theory.1 p(z) Suppose that f (z) = . f has a pole of order N at z0 .. we deduce that f has a pole of order N at z0 . 79 . we have q(z) = (z − z0 )N r(z) where r(z) p(z) is analytic with r(z0 ) = 0 (see Exercise 16.1 it has a power series expansion: ∞ g(z) = n=0 an (z − z0 )n . Then. g(z) = r(z) is analytic at z0 . By assumption. Then. (z − z0 )N (z − z0 )N −1 (z − z0 )N −2 a0 = g(z0 ) = p(z0 ) = 0.. r(z0 ) ∞ But. where q(z) (i) p is analytic at z0 .8). (ii) q is analytic at z0 . so by Theorem 15.
(ii) sin z has a simple pole at 0: z2 z3 z5 + − . It is not the multiplicity of the denominator in an expression. 4z 3!4 5!4 z− 80 (ii) The quotient 1+z (b) 3 .6 A pole of order 1 is called a simple pole. and (z − 3)2 is analytic with a root of multiplicity 2 at 3. Do not confuse the deﬁnition of an object with a process by which we ﬁnd the object.3 Recall that a zero of order N of a polynomial is just a root of multiplicity N . Remark 18. since (z − 3)2 sin is analytic and sin 3 = 0. Example 18.Common Error 18.. The resulting roots are repeated as (z 2 − 2z + 5)2 = [(z − (1 + 2i))]2 [(z − (1 − 2i))]2 . Deﬁnition 18.7 (i) 1/z has a simple pole at 0.4... (z 2 − 2z + 5)2 This is because we can factorize the polynomial z 2 − 2z + 5 to get (z − (1 + 2i))(z − (1 − 2i)). z − 2z 2 ez − 1 (c) (It ain’t 5!) z5 2 .. (a) z(z 2 − 1) Simple poles Poles of order 1 are given a special name.4 (i) The function sin z has a pole of order 2 at z = 3.5 Find the poles and their orders for 1 + ez . Exercises 18..2 The above lemma gives a method for ﬁnding a pole and its order. sin z 3! 5! = 4z 2 4z 2 1 z z3 = − + − . Example 18. The deﬁnition of a pole is that given in Deﬁnition 17. z(z − 1)4 has poles of order 2 at z = 1±2i.
q(0) = 0. q(0) = 0. We have z p(z) = cos z and q(z) = z. q(z) (i) p is analytic at z0 . f has a pole of order N at z0 .5 q has a zero of order 1 at z0 . We have sin z p(z) = z + 3 and q(z) = sin z. Proof. Theorem 18. (ii) q is analytic at z0 . so p(0) = 3.1. i. f has a simple pole at z0 . (ii) The function f (z) = z+3 has a simple pole at 0.9 cos z (i) The function f (z) = has a simple pole at 0. f has a pole of order 1 at z0 .There is a useful way of locating simple poles. 81 . and q (0) = cos 0 = 1. and has a zero of order N at z0 . (ii) q is analytic at z0 .8 p(z) Suppose f (z) = where. q(z0 ) = 0 and q (z0 ) = 0. and p(z0 ) = 0. the pole is simple. where q(z) (i) p is analytic at z0 . Thus.e. and q (0) = 1. and p(z0 ) = 0. Summary • Suppose that f (z) = p(z) . Then. Examples 18. so p(0) = 1. Then. by Lemma 18. • A simple pole is a pole of order 1. By Remark 16.
The residue of f at z0 is n the coefﬁcient a−1 in the expansion ∞ n=−N an (z − z0 ) . 4z 2 4z 3!4 5!4 so res sin z . (The residue is always 0 at differentiable points. 4 Remark 19. This theorem is really just generalising the example of ez /z 2 used at the start of the section on poles.2 (i) res(ez /z 2 . r) denote the circle of radius r round p. Theorem 19.3 The residue at a point is unique because the coefﬁcients of a Laurent expansion with a ﬁnite number of negative terms are unique.. 1 z z3 sin z = − + − . there exists R > 0 such that f (z) dz = 2πi res(f. Then. Let γ(p. γ(p. z0 ). 2 z z z 2 6 we could use the Laurent expansion about zero to integrate termbyterm.) (iii) From earlier.0 4z 2 1 = .4 (Cauchy’s Residue Theorem for a Circle) Suppose that the analytic function f : D → C has a pole at p. The next theorem gives a hint of where we are going with residues: They allow us to quickly calculate integrals.. This is denoted res(f. Examples 19. 82 . (ii) res(z 2 .. 0) = 0..19 Residues We saw earlier that to integrate ez 1 1 1 z = 2 + + + . Deﬁnition 19.r) for all 0 < r < R. The only information that mattered was the coefﬁcient of the z −1 term. Let us now give that term an ofﬁcial name. 0) = 1. p).1 Suppose that f has a pole at z0 .
r) for all 0 < r < R. let r be such that 0 < r < R. Therefore. Summary • The residue of f at z0 is the coefﬁcient a−1 in the expann sion ∞ n=−N an (z − z0 ) . • Suppose that the analytic function f : D → C has a pole at p. there exists R > 0 such that f (z) dz = 2πi res(f. 0) = 1 has already been calculated. So ez /z 2 = 2πi res(f. and then ∞ f (z) dz = γ(p.5 The theorem says that to calculate certain integrals all we have to do is calculate the residue of a function at a point. for 0 < z − p < R.7 (you can check this!). Example 19. γ(p. n = −1 0. At p the function f has a Laurent expansion f (z) = ∞ n n=−N an (z − p) . by the Fundamental Example. some R > 0.r) f (z) dz = 2πia−1 = 2πi res(f. p). then res(f.r) n=−N ∞ an (z − p)n dz an (z − p)n dz. p). Then. n = −1. z0 ). 0) = 2πi. as we have already seen.r) γ(p.r) = The latter equality follows from Theorem 6.r) γ(p. Let γ(p. This is denoted res(f. = Thus. Remark 19. n=−N γ(p.6 Let γ be the unit circle round the origin. We take f (z) = ez /z 2 from earlier. but in the next section we will see investigate some methods for calculating residues. an (z − p)n dz = an γ(p. r) denote the circle of radius r round p.Proof. γ We shall generalise the theorem further. γ(p. 83 . But.r) (z − p)n dz a−1 × 2πi.
z→3 z−3 z (ii) The function f (z) = has a pole of order 1 at z = 0.20 Bluffer’s Guide to Calculating Residues Residues are an important part of any course on Complex Analysis. z→w = a−1 = res(f. . . = lim (z − w) Examples 20. 3) = lim(z−3)f (z) = lim (z−3) z→3 z→3 sin(z) = lim sin(z) = sin(3). So. Method 1: Simple Poles Suppose f has a simple pole (i. w). w) = lim (z − w)f (z). As f has a simple pole at w we have ∞ f (z) = n=−1 an (z − w)n . z→w lim (z − w)f (z) ∞ = lim (z − w) z→w n=−1 an (z − w)n a−1 + a0 + a1 (z − w) + a2 (z − w)2 + . z→w Proof. so z−3 res(f.1 sin(z) (i) The function f (z) = has a pole of order 1 at z = 3. . .e. You should know them and their proofs. They allow us to easily calculate otherwise complicated integrals – without doing any integration! This section contains the main methods for calculating the residues of poles. order 1) at w then res(f. 1 − cos z so res(f. 0) = lim z z→0 z z2 4 (z/2)2 = lim = lim 1 − cos z z→0 2 sin2 (z/2) z→0 2 sin2 (z/2) z/2 sin(z/2) 84 2 = 2 lim z→0 z/2 = 2 lim z→0 sin(z/2) 2 = 2. z→w z−w = lim a−1 + a0 (z − w) + a1 (z − w)2 + a2 (z − w)3 + . .
q (1) −3 3 (ii) The function f (z) = 2z 2 has a pole at z = eiπ/4 (and 1 + z4 more besides. eiπ/4 = √ p(eiπ/4 ) 2eiπ/2 e−iπ/4 1 = 3iπ/4 = = (1 − i) 2. We have 1 − z3 p(z) = 1 and q(z) = 1 − z 3 so q (1) = −3 × 12 = −3. q(w) = 0 q(z) and q (w) = 0. Examples 20. w) = p(w) . q (w) Remark 20. q (w) Proof.2 The conditions imply that f has a simple pole. by deﬁnition of differentiation. Thus p(1) 1 1 res(f. Suppose f (z) = res(f. Then p(eiπ/4 ) = 2eiπ/2 = 0.Method 2: Some quotients (Really good method!) p(z) with p and q analytic. Let p(z) = 2z 2 and q(z) = 1 + z 4 .8 f has a simple pole at w. Then. p(w) = 0. as q(w) = 0. z→w = lim (z − w) = = = = p(z) z→w q(z) q(z) lim p(z) z→w z−w q(z) − q(w) . 1) = = =− . by Method 1. so res(f. So res f. w) = lim (z − w)f (z). q (eiπ/4 ) 4e 2 4 85 . q (z) = 4z 3 so q (eiπ/4 ) = 4e3iπ/4 = 0.3 (i) The function f (z) = 1 has a pole at z = 1. lim p(z) z→w z−w limz→w p(z) q(z) − q(w) limz→w z−w p(w) . but we’ll ignore them). so the method will only work in this case. By Theorem 18.
1 this has a pole of z−i order 2 at z = i. . w). . . . lim z→i z→i (z − i) 2 z+i z−i 2 d (z + i)2 dz = lim 2(z + i) = 2(i + i) = 4i. . } − 1)! z→w 1 (N − 1)! a−1 − 1)! Examples 20. N (z − w) z−w + a−N +1 (z − w) + · · · + a−1 (z − w)N −1 +a0 (z − w)N + . i) = lim 2−1 (2 − 1)! z→i dz = 1.Method 3: Poles of order N We now generalise Method 1 (which is the case N = 1 in the following). (N = (N = = (N (N = a−1 = res(f. Hence. Then res(f. . 86 . (z − w)N f (z) = (z − w)N = a−N Therefore. Thus. . Suppose that f has a pole of order N at w. − 1)! z→w dz +a−1 (z − w)N −1 + a0 (z − w)N + . 1 lim {(N − 1)! a−1 + N ! a0 (z − w) + . w) = 1 dN −1 lim N −1 (z − w)N f (z) .4 z + i 2 (i) Let f (z) = . . By assumption f (z) = and 0 < z − w < R. By Lemma 18. . 1 dN −1 lim (z − w)N f (z) z→w dz N −1 − 1)! 1 dN −1 lim N −1 a−N + a−N +1 (z − w) + . . (N − 1)! z→w dz ∞ n=−N Proof. . ∞ an (z − w)n for some ai ∈ C f (z) = n=−N an (z − w)n a−N a−1 + ··· + + a0 + a1 (z − w) + . 1 d2−1 res(f.
. We expand functions as power series.5 zez (i) Let f (z) = . d ez 1 d 2 ee z res(f. . h3 2 3 e 2e 3e 2e + 2+ + + . 3 h h 2h 3 z z π cot πz . h (1 − π 2 h2 /6 + . ) 1 (1 − π 2 h2 /2 + . z2 We expand about w = 0: π cot π(0 + h) π cot πh = (0 + h)2 h2 π cos πh = 2 h sin πh π (1 − (πh)2 /2! + . Examples 20.eh (1 + h) 3 h e h2 h3 (1 + h) 1 + h + + + . ) Consider the parts in brackets. .0 . h3 2! 3! 3 2 e 1 + 2h + h2 + h3 + . 1) = (ii) ∗ Find res 3e .1).ee (ii) The function f (z) = 2 has a pole of order 2 at z = 0.. this quotient will be a power series and so we need to work out its coefﬁcients. z→0 dz z→0 dz z→0 1! z Method 4: Direct expansion We can use this as a last resort. z (again by Lemma 18. and then calculate coefﬁcients. etc. 0) = lim z 2 = lim e = lim ez . . So. . 87 . 2 (1 + h)e1+h (1 + h − 1)3 e.. f (1 + h) = = = = = So res(f. let z = w + h and take h as our variable). (z − 1)3 expand about w = 1: (i.e. . ) = 2 h (πh − (πh)3 /3! + . .ee = e.. . . ) = 3 . . So. .. This has a pole of order 3 at z = 1.
6 π 2 h2 1+ + .. 3 + . b2 = π cot πh 1 = 3 1− 2 h h 1 = 3 1− h 1 = 3 1− h 1 π2 = 3− h 3h Thus res(f. 6 −1 π2 .e. Then. .We need only work out the coefﬁent for h2 since division by the h3 will give the residue coefﬁcient.. = 1 b0 + b1 h + b2 h2 − 6 π2 b0 + b1 h + b2 − b0 h2 + . 6 Therefore.. . 3 Obviously.. w) = limz→w (z − w)f (z). .w = . . . = 1 π 2 h2 b0 + b1 h + b2 h2 + . We work with the bottom bracket ﬁrst... b0 + b1 h + b2 h + . .. . . Summary • Simple pole: res(f. Then.. π2 b0 = 0. 0) = − π 2 h2 + . w) = lim (z − w)N f (z) . p p(w) res . • Some quotients: Suppose p(w) = 0.. . 1 − π 2 h2 /6 + . . equate coefﬁcients. Suppose that −1 (1 − π 2 h2 /6 + . . this method can get a bit calculation heavy. . 2 −1 = 1 1 − π h /6 + . . ... z→w dz N −1 (N − 1)! • Calculate expansions directly. i. 88 . q(z) = 0 and q (w) = 0. . 2 π2 2 h + . etc. 6 π 2 h2 1− + . 6 Equating coefﬁcients gives b0 = 1. = 1. .. ) = b0 + b1 h + b2 h2 + . b1 = 0 and b2 − π2 . 2 π 2 h2 + . q q (w) • Pole of order N : 1 dN −1 res(f. 2 2 1 − π 2 h2 /6 + . .
. Now. So. So γ = γ1 + · · · + γm . But γ e f = n(γ. taken clockwise. pm . p2 ) C2 f + · · · + n(γ. . . +βm + n(γ. Then. consider the following sketch. Let {p1 . pm ) Cm f+ γ1 +. Note that the integrals over all the βj cancel. Next.) So by Cauchy’s Theorem we have f = 0. pm } ∈ D\γ and f : D → C be analytic with poles at p1 . Let Cj be a circle of radius ε > 0 at pj . . . p1 )C1 − β1 + γ1 +β2 + n(γ.γm f. Then m f (z) dz = 2πi γ j=1 n(γ. we deﬁne a contour γ by γ = β1 + n(γ. Thus we have a very simple ﬁveﬁngerexercise method for calculating integrals. pj ). Proof.21 Cauchy’s Residue Formula The next theorem shows that we can calculate certain complex integrals by merely calculating the residues and winding numbers at poles..1 (Cauchy’s Residue Formula) Let D be domain and γ a closed contour such that Int(γ) ⊆ D. pj ) Cj f + γ1 +. as γ = γ1 + 89 . Hooray! Theorem 21. there exists ε > 0 such that Cj ⊆ Int(γ) and Cj ∩ Cl = ∅ for all j and l with j = l and so that Cj lies within the disc upon which the Laurent expansion for f at pj holds. p1 ) γ e m C1 f + n(γ. . (just like in the calculation trick I did. . There exist contours βj from some point on the image of γ to the start of Cj . . pi ) res(f.. We can deﬁne γj as the part of γ from the start of βj to βj+1 . . pm )Cm − βm + γm . p2 )C2 − β2 + γ2 + ..γm f 0 = j=1 n(γ.. This is a path such that f is analytic on Int(γ). We can assume that all the poles are in the interior of γ. .
. f has simple poles at z = 1 and z = −i. and q (−i) = −2i + i − 1 = −1 − i. res(f. pj ) Cj f + γ f − j=1 m n(γ. γ Summary • Let D be domain and γ a closed contour such that Int(γ) ⊆ D. one of the best theorems in mathematics! Examples 21. . . . 1) + res(f. . −i) = = . then q (z) = 2z+i−1. Let q(z) = z 2 +(i−1)z−i. pj ) j=1 −Cj f = But. where γ is a square 2 γ z + (i − 1)z − i with sides of length 4 centred at the origin.2 Evaluate the integral 1 dz. 90 . pj ) Cj f = γ f f. Let {p1 . pj ). complete. −Cj f = 2πi res(f. pm } ∈ D\γ and f : D → C be analytic with poles at p1 . 1) = n(γ. Thus the proof is There we have it. . −i)) z 2 + (i − 1)z − i 1 − i −1 + i = 2πi + 2 2 = 0. oriented anticlockwise. Then m f (z) dz = 2πi γ j=1 n(γ. 1) = 1 1−i 1 −1 + i = and res(f. Thus. by Theorem 19. m 0 = j=1 m n(γ. pj ). −i) = 1. γ n(γ.· · · + γm we get. pm . so by Cauchy’s Residue Theorem 1 dz = 2πi ( res(f.4. . . 1 1 Solution: Let f (z) = 2 = . Therez + (i − 1)z − i (z + i)(z − 1) fore. pi ) res(f. 1+i 2 −1 − i 2 We have n(γ. We have q (1) = 2+i−1 = 1+i.
First. However. The RHS contains the integral we want. Now. If we integrate round the contour Γ = Γ1 + ΓR . Thus. then Γ (t) = 1 for −R ≤ t ≤ R. so may initially appear complicated. Let Γ1 (t) = t.22 Evaluation of Deﬁnite Real Integrals We shall now see that complex analysis allows us to solve. ΓR Now. problems involving real functions that are not easy to solve with real methods. we get R→∞ lim f (z) dz = Γ R→∞ lim f (z) dz + lim Γ1 R −R R→∞ f (z) dz ΓR = = R→∞ ∞ lim f (t) dt + lim f (x) dx + lim R→∞ f (z) dz ΓR −∞ R→∞ f (z) dz. where ΓR is the semicircle in the upper halfplane. the LHS is a contour integral for a closed contour and so by Cauchy’s Residue Formula we can probably work it out using some method. then we get f (z) dz = Γ Γ1 f (z) dz + ΓR f (z) dz. Great! It also contains another term.1 dt = −R f (t) dt. −∞ This doesn’t really get us much further forward. letting R → ∞. we can use Cauchy’s Residue Theorem to calculate the integral. R R ∞ f (z)dz = Γ1 −R f (t). The basic method is actually quite simple. but it has a number of parts. then we get a closed contour.3 later). Let us note that we can calculate the real integral as a contour integral. if that term is zero. Not so great! However. if we take a contour ΓR from R to −R via a path in the complex plane and join it to Γ1 . I’ll outline the ideas behind it. in a simple manner. then we have a way of calculating a real integral from residues! 91 . then we usually obtain the real integral ∞ f (x) dx (but see Remark 22. Consider the following diagram in the complex plane. So. If we let R → ∞. The basic idea We want to calculate −∞ f (x) dx where f is a real function.
the polynomial p can be written as p(z) = c deg p (z − αj ). for z ∈ Γ and R ≥ max{βj }. Similarly. where c is some conj=1 stant and αj is a root of p. As usual. eiaz = eRe(iaz) = eRe(ia(x+iy)) = eRe(−ay+iax) = e−ay ≤ 1. p(z) iaz e dz → 0 as R → ∞. for large R. deg p deg p deg p p(z) = c j=1 z − αj  ≤ c j=1 (z + αj ) ≤ c j=1 (R + αj ). deg q deg q q(z) = d j=1 (z − βj ) ≥ d j=1 (R − βj ). Lemma 22. Then. ΓR (t) = Reit . By the Fundamental Theorem of Algebra. if deg p ≤ deg q − 2.Can we show that the integral ΓR f (z) dz → 0 as R → ∞? The following lemma gives us a supply of examples with this property. the RHS is equal to 0. Similarly. the latter inequality holds because y ≥ 0 for z ∈ γ. z = R). and because a ≥ 0. for j=1 some d and βj . and a ≥ 0. Remark 22. Note that the RHS is a polynomial in R of degree deg p. when z ∈ ΓR . 0 ≤ t ≤ π is a semicircular contour of radius R. c p(z) iaz e dz ≤ q(z) d deg p j=1 (R + αj ) deg q j=1 (R − βj ) × πR. Thus. q(z) = d deg q (z − βj ). We have.1 (Jordan’s Lemma) Suppose p and q are polynomials with deg(p) ≤ deg(q) − 2. we apply the Estimation Lemma. 92 . We then have. then the RHS→ 0 as R → ∞. Hence. ΓR q(z) Proof.e. by the Estimation Lemma. (i.2 We include an eiaz term as this will be useful in determining integrals involving sines and cosines. ΓR So.
93 . f (x) = x. if both integrals exist. the method is the following.Remark 22. q(z) (ii) Calculate Γ p(z) dz using Cauchy’s Residue Theorem. q(z) ∞ −∞ (iii) Show that ΓR (iv) Conclude that f (x) dx and limR→∞ Γ f (z) dz are equal.g. −R Deﬁnition 22. E. q(z) p(z) dz → 0 as R → ∞. then they are equal. taking R large enough so that Γ contains all the poles of p(z) that lie in the upper halfplane. may be different to R R→∞ lim f (x) dx.4 R The integral limR→∞ −R f (x) dx is called the principal part of ∞ the integral. where p and q are −∞ q(x) polynomials with q(x) = 0 for all real x. using Jordan’s Lemma. but note that in practice we should check that this is the case. and ΓR (t) = Reit for 0 ≤ t ≤ π. For real integrals of the form (i) Deﬁne Γ = Γ1 + ΓR . In all our examples we may as well assume that both exist.3 ∞ The integral −∞ f (x) dx is actually deﬁned to be b b→∞ a→−∞ lim f (x) dx a and in fact. It is probably best if we do an example. and is denoted pv −∞ f (x) dx. Real integrals of the form ∞ p(x) −∞ q(x) ∞ ∞ dx p(x) dx. where Γ1 (t) = t for − R ≤ t ≤ R. However. The principal part of an integral may exist even if −∞ f (x) dx does not.
dx . 2 2 −∞ (x + 1) (i) Take Γ as the join of two contours above. 1 1 1 (ii) Let f (z) = 2 . Then f (z) = 2 = . 2 2 2 (z + i)2 (z + 1) (z + 1) (z − i) So f has poles of order 2 at z = i and z = −i. Only the z = i pole lies in the interior of Γ (when R > 1). By Method 3 we get res(f, i) = = = = = = 1 d 1 (z − i)2 2 (z + i)2 (2 − 1)! dz (z − i) d 1 dz (z + i)2 z=i 1 −2 (z + i)3 z=i 1 −2 (i + i)3 1 4i i − . 4
Example 22.5 Find the integral
∞
z=i
(iii) By Jordan’s Lemma (as deg q = deg(z 2 + 1)2 = 4, deg p = deg 1 = 0, and a = 0) (z 2 1 dz → 0 as R → ∞. + 1)2
ΓR
(iv) Thus, as f (z) dz =
Γ Γ1 R −R ∞
f (z) dz +
ΓR
f (z) dz
2πi res(f, i) = 2πi − i 4 = π = 2
−∞ ∞ −∞
(x2
1 dx + f (z) dz + 1)2 ΓR 1 dx + 0 as R → ∞ 2 + 1)2 (x 1 dx 2 + 1)2 (x
Can you calculate this integral using the standard methods of real analysis?
94
Real integrals of the form
∞ −∞
sin x cos x
p(x) q(x)
dx
We can apply a similar method to integrals of the form
∞ −∞
sin x cos x
p(x) dx q(x)
by making a good choice of complex function to integrate. Let us try an example that really shows the power of complex analysis. Example 22.6 Calculate the integral cos x dx. − 2x + 2 −∞ eiz . (Remember that The key here is to chose f (z) = 2 z − 2z + 2 eiz = cos z + i sin z so we will get the cos x, but we will also get a sin x term along the real axis. Surprisingly, this turns out to be a bonus, not a problem  just watch!) (i) Let Γ be as in the previous example, so Γ = Γ1 + ΓR . (ii) We want to evaluate x2 f (z) dz
Γ ∞
ﬁrst, so we use Cauchy’s Residue Theorem. We know that eiz has no zeros and no poles, so all the poles arise from zeros of the denominator: z 2 − 2z + 2. √ 2± 4−8 2 z − 2z + 2 = 0 ⇐⇒ z = = 1 ± i. 2 √ Let R > 2, so 1 + i is inside the contour, 1 − i is outside it. The diagram shows the location of poles and the contour ΓR .
The pole is simple so eiz dz = 2πi res z 2 − 2z + 2 = 2πi eiz ,1 + i z 2 − 2z + 2
Γ
ei(1+i) 2(1 + i) − 2 e−1+i = 2πi 2i −1 i = πe e π = (cos 1 + i sin 1). e
95
(iii) By Jordan’s Lemma eiz dz → 0 as R → ∞. z 2 − 2z + 2
ΓR
(iv) Now we shall calculate the integrals along the two parts of the contour. Along Γ1 we get eiz dz = z 2 − 2z + 2 =
−R R −R R
Γ1
eix dx x2 − 2x + 2 cos x dx + i 2 − 2x + 2 x
R −R
x2
sin x dx. − 2x + 2
Therefore,
R→∞
lim
f (z) dz =
Γ
R→∞
lim lim
f (z) dz + lim
Γ1 R
R→∞
f (z) dz
ΓR
R cos x sin x dx + i dx + 0 2 − 2x + 2 2 − 2x + 2 R→∞ −R x −R x ∞ ∞ π cos x sin x (cos 1 + i sin 1) = dx + i dx. 2 − 2x + 2 2 − 2x + 2 e −∞ x −∞ x
=
Equating real parts we get
∞ −∞
cos x π dx = cos 1. x2 − 2x + 2 e
Remark 22.7 Note that using the imaginary parts we get
∞ −∞
sin x π dx = sin 1. x2 − 2x + 2 e
Thus, we have gained more information than we were looking for. What a method! Summary (i) Deﬁne Γ = Γ1 + ΓR by Γ1 (t) = t for − R ≤ t ≤ R, and ΓR (t) = Reit for 0 ≤ t ≤ π. Take R large enough to include the poles of f that lie in the upper halfplane. (ii) Calculate (iii) Show that (iv) Relate
∞ −∞ Γ
f (z) dz using Cauchy’s Residue Theorem. f (z) dz → 0 as R → ∞, using Jordan’s Lemma.
Γ
ΓR
f (x) dx to
f (z) dz.
96
wouldn’t you agree? Integrals of the form 2π 0 f (cos θ. If we consider contour integration. 97 . By equating real and imaginary parts we see that 2π ecos θ cos(sin θ − nθ) dθ = 0 2π n! and because cos is an even function we deduce that 2π 2π ecos θ cos(nθ − sin θ) dθ = n! 0 Rather spectacular. Then 2π f (cos θ. then is there some function and contour such that γ f gives the above? Let γ(t) = eiθ . 0 from Section 3. for 0 ≤ t ≤ 2π. for 0 ≤ θ ≤ 2π and f (z) = ez . 2 2i z −1 dz.1 Let γ(t) = eit . with our knowledge of poles and residues we see that 2πi ez dz = 2πi res(ez /z n+1 . sin θ) dθ Theorem 23.23 Integrals with sines and cosines 2π Let us consider the nasty integral ecos θ cos(nθ − sin nθ) dθ. Then. sin θ) dθ = −i 0 γ f z + z −1 z − z −1 . 0) = n+1 n! γ z So ez γ 2π z n+1 dz = 0 ee iθ eiθ(n+1) ieiθ dθ 2π 2πi iθ = i ee e−iθn dθ n! 0 2π 2π iθ = ee −inθ dθ n! 0 2π = 0 2π ecos θ+i sin θ−inθ dθ ecos θ ei(sin θ−nθ) dθ 0 2π = = 0 ecos θ (cos(sin θ − nθ) + i sin((sin θ − nθ)) dθ.
2 2i 2π z −1 dz = −i 0 2π f (cos θ. 2 2i z −1 dz. We also have γ (θ) = ieiθ . Hence. 0 ≤ t ≤ 2π. a π A common example is a = −π. So we get integrals of the form a+2π f (cos θ. Then 2π f (cos θ. so we get −π . we calculate the residue for this pole: res g. for 0 ≤ t ≤ 2π. 2i Thus. 13 + 12 cos t 0 Solution: Applying the theorem we get. Example 23. only the unit circle given by γ.3 Evaluate the real integral 1 dt. sin θ) dθ = −i 0 γ f z + z −1 z − z −1 . 98 . 2π 0 2π 1 dt = −i 13 + 12 cos t = −i γ γ 1 dz z (13 + 6 (z + 1/z)) 1 dz.2 We can change the limits of integration by letting γ(t) = eit where a ≤ t ≤ a + 2π and a ∈ R. Of 2 3 12 −2 3 1 1 = . Then cos θ = z − z −1 . 5 + 13 + 13 Therefore. by Cauchy’s Residue Theorem. sin θ) dθ. for γ(t) = eit . 13 + 12 cos t 5 5 Summary • Let γ(t) = eit . = 0 Remark 23. −i γ z + z −1 and sin θ = 2 f z + z −1 z − z −1 . sin θ) dθ. − 2 3 = 1 12z + 13 = z=−2/3 6z 2 + 13z + 6 2 − 3 lies within . 2π 0 1 1 2π dt = −i × 2πi × = . 2 + 13z + 6 6z Let g(z) = 1 these. This has poles at − 3 and − 2 . Let z = eiθ = γ(θ). sin θ)e−iθ ieiθ dθ f (cos θ.Proof.