Lecture Notes on Complex Analysis

by Stephen New

0
 Chapter 1. Complex Numbers

1.1 Definition: A complex number is a vector in R2 . The complex plane, denoted

by C, is the set of complex numbers:
  
2 x
C=R = x ∈ R, y ∈ R .
y
         
0 1 0 x 0
In C we usually write 0 = ,1= ,i= ,x= , iy = yi = and
0 0 1 0 y
 
x
x + iy = x + yi = .
y

If z = x+iy with x, y ∈ R then x is called the real part of z and y is called the imaginary
part of z, and we write
Re z = x , and Im z = y .
1.2 Definition: We define the sum of two complex numbers to be the usual vector sum:
     
a c a+c
(a + ib) + (c + id) = + = = (a + c) + i(b + d) ,
b d b+d

where a, b ∈ R. We define the product of two complex numbers by setting i2 = −1 and

by requiring the product to be commutative and associative and distributive over the sum:

(a + ib)(c + id) = ac + iad + ibc + i2 bd = (ac − bd) + i(ad + bc) .

1.3 Example: Let z = 2 + i and w = 1 + 3i. Find z + w and zw.

Solution: z + w = (2 + i) + (1 + 3i) = (2 + 1) + i(1 + 3) = 3 + 4i, and zw = (2 + i)(1 + 3i) =
2 + 6i + i − 3 = −1 + 7i.
1.4 Example: Show that every non-zero complex number has a unique inverse z −1 and
find a formula for the inverse.
Solution: We let z = a+ib where a, b ∈ R with a2 +b2 6= 0, and we solve (a+ib)(x+iy) = 1
to find z −1 = x + iy:

(a + ib)(x + iy) = 1 ⇐⇒ (ax − by) + i(ay + bx) = 1

        
ax − by 1 a −b x 1
⇐⇒ = ⇐⇒ =
bx + ay 0 b a y 0
   −1       
x a −b 1 1 a b 1 1 a
⇐⇒ = = 2 = 2
y b a 0 a + b2 −b a 0 a + b2 −b
1
⇐⇒ x + iy = 2 (a − ib) .
a + b2
a b
Thus (a + ib)−1 = − i .
a2 + b2 a2 + b2

1
1.5 Notation: For z, w ∈ C we use the following notation:
1 w
−z = −1z , w − z = w + (−z) , = z −1 and = wz −1 .
z z
(4 − i) − (1 − 2i)
1.6 Example: Find .
1 + 2i
(4 − i) − (1 − 2i) 3+i
Solution: = = (3 + i)(1 + 2i)−1 = (3 + i)( 51 − 25 i) = 1 − i.
1 + 2i 1 + 2i
1.7 Note: The set of complex numbers is a field under the operations of addition and
multiplication. This means that for all u, v and w in C we have
u+v =v+u
(u + v) + w = u + (v + w)
0+u=u
u + (−u) = 0
uv = vu
(uv)w = u(vw)
1u = u
u u−1 = 1 if u 6= 0
u(v + w) = uv + uw
1.8 Definition: If z = x + iy with x, y ∈ R then we define the conjugate of z to be
z = x − iy .
and we define the length (or magnitude) of z to be
p
|z| = x2 + y 2 .
1.9 Note: For z and w in C the following identities are all easy to verify.
z=z
z + z = 2Re z , z − z = 2iIm z
2
zz = |z| , |z| = |z|
z+w =z+w , zw = z w , |zw| = |z||w|
1.10 Note: We do not have inequalities between complex numbers. We can only write
a < b or a ≤ b in the case that a and b are both real numbers. But there are several
inequalities between real numbers which concern complex numbers. For z ∈ C and w ∈ C,
|Re (z)| ≤ |z| , |Im (z)| ≤ |z|
|z + w| ≤ |z| + |w| , this is called the triangle inequality
|z + w| ≥ |z| − |w|
The first two inequalities follow from the fact that |z|2 = |Re (z)|2 + |Im (z)|2 . We can then
prove the triangle inequality as follows: |z +w|2 = (z +w)(z +w) = |z|2 +|w|2 +(wz +zw) =
|z|2 + |w|2 + 2Re (zw) ≤ |z|2 + |w|2 + 2|zw| = |z|2 + |w|2 + 2|z||w| = (|z| + |w|)2 . The last
inequality follows from the triangle inequality since |z| = |z + w − w| ≤ |z + w| + |w| and
|w| = |z + w − z| ≤ |z + w| + |z|. (Alternatively, the last two inequalities can be proven
using the Law of Cosines).

2
 
1.11 Example: Given complex numbers a and b, describe the set z ∈ C |z−a| < |z−b| .
Solution: Geometrically, this is the set of all z such that z is closer to a than to b, so it is
the half-plane which contains a and lies on one side of the perpendicular bisector of the
line segment ab.

1.12 Example: Given a complex number a, describe the set z ∈ C 1 < |z − a| < 2 .
 
Solution: z| z − a| = 1 is the circle  centred at a of radius 1 and z| z − a| = 2 is the
circle centred at a of radius 2, and z ∈ C 1 < |z − a| < 2 is the region between these
two circles. Such a region is called an annulus.
1.13 Example: Show that every non-zero complex number has exactly two complex
square roots, and find a formula for the two square roots of z = x + iy.
Solution: Let z = x + iy where x, y ∈ R with x and y not both zero. We need to solve
w2 = z for w ∈ C. Write w = u + iv with u, v ∈ R. We have
w2 = z ⇐⇒ (u + iv)2 = x + iy ⇐⇒ (u2 − v 2 ) + i(2uv) = x + iy
⇐⇒ u2 − v 2 = x and 2uv = y .

To solve this pair of equations for u, square both sides of the second equation to get
4u2 v 2 = y 2 , then multiply the first equation by 4u2 to get 4u4 − 4u2 v 2 = 4x u2 , that is
4u4 − 4x u2 − y 2 = 0. By the quadratic formula,
p p
4x ± 16x 2 + 16y 2 x ± x2 + y 2
u2 = = .
8 2
In the case that y 6= 0, we must use the + sign so that the right side is non-negative, so
we obtain s p
x + x2 + y 2
u=± .
2
A similar calculation gives s p
−x + x2 + y 2
v=± .
2
All four choices of sign will satisfy the equation u2 − v 2 = x, but to satisfy 2uv = y notice
that when y > 0, u and v have the same sign, and when y < 0, u and v have the opposite
sign. It remains only to consider the case that y = 0, and we leave this case as an exercise.
The final result is that
 s p s p 

 x+ x +y 2 2 2
−x + x + y  2

 ± +i , if y > 0,
2 2







 s
 p s p 
x+ x +y 2 2 2
−x + x + y  2
w= ± −i , if y < 0,



 2 2
√






 ± x , if y = 0 and x > 0,
 p
± i |x| , if y = 0 and x < 0 .


1.14
√ Note: When working with real numbers, for 0 < x ∈ R it is customary to write
x or x1/2 to denote the unique positive square √ root1/2 of x. When working with complex
numbers, for 0 6= z ∈ C we sometimes √ write z or z to denote one of the two square
1/2
roots of z, and we sometimes write z or z to denote both square roots of z.

3
 √
1.15 Example: Find 3 − 4i.
Solution: Using the formula derived in the previous example, we have

√ √ √
q q  q q 
3+ 32 +42 −3+ 32 +42 3+5 −3+5
3 − 4i = ± 2 −i 2 =± 2 −i 2 = ±(2 − i) .

1.16 Note: The Quadratic Formula can be used for complex numbers. Indeed for
a, b, c, z ∈ C with a 6= 0 we have
 2  2
2 b c 2 2 b b b c
az + bz + c = 0 ⇐⇒ z + z + = 0 ⇐⇒ z + z+ − + =0
a a 2a 2a 2a a
2  2 √
b2 − 4ac b2 − 4ac

b b c b
⇐⇒ z + = − = ⇐⇒ z + =
2a 2a a 4a2 2a 2a
√
−b + b2 − 4ac
⇐⇒ z = ,
2a
√
where b2 − 4ac is being used to denote both square roots in the case that b2 − 4ac 6= 0.
1.17 Example: Solve i z 2 − (2 + 3i)z + 5(1 + i) = 0.
Solution: By the Quadratic Formula, we have
p √
(2 + 3i) +(2 + 3i)2 − 20i(1 + i) (2 + 3i) + −5 + 12i + 20 − 20i
z= =
√ 2i 2i
(2 + 3i) + 15 − 8i
=
2i

and by the formula for square roots we have

√ √ √
q q  q q 
15+ 152 +82 −15+ 152 +82 15+17 −15+17
15 − 8i = ± 2 −i 2 = ± 2 − i 2 = ± (4 − i)

and so
(2 + 3i) ± (4 − i) 6 + 2i −2 + 4i
z= = or = 1 − 3i or 2 + i .
2i 2i 2i
1.18 Definition: If z 6= 0, we define the angle (or argument) of z to be the angle θ(z)
from the positive x-axis counterclockwise to z. In other words, θ(z) is the angle such that


z = |z| cos θ(z) + i sin θ(z) .

1.19 Note: We can think of the angle θ(z) in several different ways. We can require, for
example, that 0 ≤ θ(z) < 2π so that the angle is uniquely determined. Or we can allow
θ(z) to be any real number, in which case the angle will be unique up to a multiple of 2π.
Then again, we can think of θ(z) as the infinite set of real numbers θ(z) = {θ0 +2πk|k ∈ z},
that is we can regard θ(z) as an element of R/2π, the set of real numbers modulo 2π.

4
1.20 Notation: For θ ∈ R (or for θ ∈ R/2π) we shall write

eiθ = cos θ + i sin θ .

1.21 Note: If z 6= 0 and we have x = Re (z), y = Im (z), r = |z| and θ = θ(z) then

x = r cos θ , y = r sin θ
p y
r = x2 + y 2 , tan θ = , if x 6= 0
x
1
z = reiθ , z = r e−i θ , z −1 = e−i θ
r
We say that x + i y is the cartesian form of z and rei θ is the polar form.
1.22 Example: Let z = −3 − 4i. Express z in polar form.
Solution: We have |z| = 5 and tan θ(z) = 43 . Since θ(z) is in the third quadrant, we have
−1
θ(z) = π + tan−1 34 . So z = 5ei(π+tan (4/3)) .
−1
1.23 Example: Let z = 10ei tan 3
. Express z in cartesian form.
  √ √
Solution: z = 10 cos(tan−1 3) + i sin(tan−1 3) = 10 √110 + i √310 = 10 + 3 10 i.

1.24 Example: Find a formula for multiplication in polar coordinates.

Solution: For z = reiα and w = eiβ we have zw =
rs(cos α + i sin α)(cos β + i sin
β) =
(cos α cos β − sin α sin β) + i(sin α cos β + cos α sin β) = rs cos(α + β) + i sin(α + β) and
so we obtain the formula
reiα seiβ = rs ei(α+β) .
1.25 Note: An immediate consequence of the above example is that

(r ei θ )n = rn ei nθ

for r, θ ∈ R and for n ∈ Z. This result is known as De Moivre’s Law.

1.26 Example: Find (1 + i)10 .
Solution: This can be done in cartesian coordinates using the binomial theorem√(which
holds for complex
√ numbers), but √ it is easier in polar coordinates. We have 1 + i = 2ei π/4
so (1 + i)10 = ( 2ei π/4 )10 = ( 2)10 ei 10π/4 = 32ei π/2 = 32i.
1.27 Example: Find a formula for the nth roots of a complex number. In other words,
given z = reiθ , solve wn = z.
Solution: Let w = seiα . We have wn = z ⇐⇒ (seiα )n = reiθ ⇐⇒ sn ei nα = reiθ ⇐⇒
√ θ + 2πk
sn = r and nα = θ + 2πk for some k ∈ Z ⇐⇒ s = n r and α = for some k ∈ Z.
n
Notice that when z 6= 0 there are exactly n solutions obtained by taking 0 ≤ k < n. So we
obtain the formula
√
(r ei θ )1/n = n r ei (θ+2πk)/n , k ∈ {0, 1, . . . , n − 1} .
√ √
In particular, (r ei θ )1/2 = ± r ei θ/2 . For p0 < a ∈ R we have z 2 = a ⇐⇒ z = ± a, and
for 0 > a ∈ R we have z 2 = a ⇐⇒ z = ± |a| i.

5
1.28 Note: When √ working with complex numbers, for 0 6= z ∈ C and for 0 < n ∈ Z, we
sometimes
√ write n
z or w1/n to denote one of the n solutions to wn = z, and we sometimes
write n z or z 1/n to denote the set of all nth roots.
1.29 Note: For z, w ∈ C, the rule

(zw)1/n = z 1/n w1/n

does hold provided that z 1/n is used to denote the set of all nth roots, but it does not
always hold when z 1/n is used to denote one of the nth roots. Consider the following
amusing “proof” that 1 = −1:
√ p √ √
1 = 1 = (−1)(−1) = −1 −1 = i2 = −1 .
√
3
1.30 Example: Find −2 + 2i.
√
Solution: Note that −2 + 2i = 2 2 ei 3π/4 , and so the formula for nth roots gives

√ √
q
3 3
−2 + 2i = 2 2 e3π/4
√
2 ei(π/4+ 3 k) , k ∈ {0, 1, 2}
2π
=
√ √ √
= 2 ei π/3 , 2 ei 11π/12 , 2 ei 19π/12 .

1.31 Note: The remaining examples in this chapter illustrate situations in which we
can use complex numbers as a tool to help solve certain problems which only involve real
numbers.
1.32 Example: Let x0 = 1 and x1 = 1, and for n ≥ 2 let xn = 2xn−1 − 5xn−2 . Find a
closed-form formula for xn .
Solution: If a sequence xn satisfies the recursion formula a xn + b xn−1 + c xn−2 = 0 and if
the associated quadric az 2 + bz + c = 0 has distinct roots α and β, then it can be shown
that xn = Aαn + Bβ n for some constants A and B (if you have not seen this fact before,
then try to prove it by induction). For √
the given sequence, the associated quadratic is
2 2± 4−20
z − 2z + 5 = 0 which has roots z = 2 = 1 ± 2i, and so we have

xn = A(1 + 2i)n + B(1 − 2i)n

for some constants A and B. To get x0 = 1 and x1 = 1, we need A + B = 1 and

A(1 + 2i) + B(1 − 2i) = 1. Solving these two equations gives A = B = 21 , so we have
 √
n √
n  √
( 5)n
1 1
5 e−i θ ei nθ + e−i nθ
n


xn = 2 ((1 + 2i) + (1 − 2i)n ) = 2 5 ei θ + = 2
√ n
( 5) √
n
= 2 (2 cos nθ) = 5 cos nθ

where θ = θ(1 + 2i) = tan−1 2. Thus we obtain

√
xn = ( 5)n cos n tan−1 2 .

6
 n  
X 3n
1.33 Example: Find .
i=0
3i
 √   √ 
Solution: Let α = ei 2π/3 . Note that 1 + α + α2 = 1 + − 12 + 23 i + − 21 − 23 i = 0.
By the Binomial Theorem we have
           
3n 3n 3n 3n 3n 3n 3n
(1 + 1) = + + + + + ··· +
0 1 2 3 4 3n
           
3n 3n 3n 3n 3n 3n
(1 + α)3n = + α + α2 + + α + ··· +
0 1 2 3 4 3n
           
3n 3n 3n 3n 3n 3n
(1 + α2 )3n = + α2 + α + + α2 + · · · +
0 1 2 3 4 3n
n  
3n 3n 2 3n
X 3n
Adding these three equations gives (1 + 1) + (1 + α) + (1 + α ) = 3 . Note
i=0
3i
√ √
that 1 + α = 1 − 21 + 23 i = 12 + 23 i = ei π/3 and similarly 1 + α2 = e−i π/3 , and so
n  
X 3n 
3n
3n 
= 13 (1 + 1)3n + (1 + α)3n + (1 + α2 )3n = 31 23n + ei π/3 + e−i π/3


i=0
3i

23n + 2(−1)n
= 1
3 23n + ei nπ + e−i nπ = .
3
1.34 Note: The Fundamental Theorem of Algebra (which we shall prove later in this
course) states that every non-constant polynomial over C has a root in C. It follows that
every such polynomial factors into linear factors over C. If a polynomial f (x) has real
coefficients, and α is a complex root of f so that f (α) = 0, then we have f (α) = f (α) = 0
so that α is also a root of f . Notice that in this case
(x − α)(x − α) = x2 − (α + α)x + αα = x2 − 2Re (α) + |α|2 ,
which has real coefficients. It follows that every non-constant polynomial over R factors
into linear and quadratic factors over R.
1.35 Example: Let f (x) = x4 + 2x2 + 4. Solve f (z) = 0 for z ∈ C, factor f (z) over the
complex number, and then factor f (x) over the real numbers.
√
Solution: By the quadratic formula, f (z) = 0√when z 2 = −1 ± 3 i or in polar coordinates
z = 2e±i 2π/3 . Thus the roots of f are z = ± 2e±i π/3 , and so f factors over C as
√ √ √ √
z 4 + 2z 2 + 4 = z − 2ei π/3 z − 2e−i π/3 z + 2ei π/3 z + 2e−i π/3 .





√ √ √ √ √
Since z − 2ei π/3 z − 2e−i π/3 = z 2 − 2 z + 2 and z + 2ei π/3 z + 2e−i π/3 =





√
z 2 + 2 z + 2, we see that over R, f factors as
√ √
f (x) = (x2 − 2 x + 2)(x2 + 2 x + 2) .
1.36 Note: Historically, complex numbers first arose in the study of cubic equations. An
equation of the form ax3 + bx2 + cx + d = 0, where a, b, c, d ∈ C with a 6= 0 can be solved as
follows. First, divide by a to obtain an equation of the form x3 + Bx2 + Cx + D = 0. Next,
make the substitution y = x+ B3 and rewrite the equation in the form y 3 +py +q = 0. Then
p p3 −3
make the substitution y = z − 3z to convert the equation to the form z 3 + q − 27 z = 0.
3
3
Finally, multiply by z to obtain z +qz 6 3
− p27 3
and solve for z using the Quadratic Formula.

7
1.37 Example: Let f (x) = x3 + 3x2 + 4x + 1. Note that f 0 (x) = 3x2 + 6x + 4 =
3(x + 1)2 + 1 > 0, so f is increasing and hence has exactly one real root. Find the real
root of f .
Solution: Let y = x+1. Then x3 +3x2 +4x+1 = (y−1)3 +3(y−1)2 +4(y−1)+1 = y 3 +y−1.
Try y = z + rz −1 with r = − 31 , so we have y 3 + y − 1 = (z − 31 z −1 )3 + (z − 13 z −1 ) − 1 =
1 −3 1
z 3 − 1 − 27 z . We solve z 6 − z 3 − 27 = 0 using the quadratic formula, and obtain
√ 31 q √ 31 r √ 31 q √
31
1± 27 3 1+
−1
q
1 3 2 1 3 2(1− ) 3 1−
3
z = 2 . If z = 2
27
then r z = − 3 √ 31 = − 3 31
27
= 2
27
.
1+ 27 1− 27
q √ q √
3 1− 31 3 1+ 31
−1
Similarly, if z = 2
27
then r z = 27
. In either case we have y = z + rz −1 =
q √ q √ q√ 2 q√
31 31 31 31
27 −1
3 1+ 3 1− 3 +1 3

2
27
+ 2
27
, and x = y − 1 = 27
2 − 2 − 1. (We did not use complex
numbers in this example).
1.38 Example: Find the three real roots of f (x) = x3 − 3x + 1.
Solution: Let x = z+rz −1 with r = 1 so that f (x) = (z+z −1 )√3 −3(z+z −1 )+1 = z 3 +1+z −3 .
Multiply by z 3 and solve z 6 + z 3 + 1 = 0 to get z 3 = −1±2 3 i = e±i 2π/3 . If z 3 = ei 2π/3
then z = ei 2π/9 , ei 8π/9 or ei 14π/9 and so x = z + z −1 = z + z = 2Re (z) = 2 cos( 2π 9 ),
8π 14π 3 −i 2π/3
2 cos( 9 ) or 2 cos( 9 ). If z = e then we obtain the same values for x. Thus the
three real roots are 2 cos(40 ), −2 cos(20 ) and 2 cos(80◦ ).
◦ ◦

8
 Chapter 2. Complex Functions

2.1 Definition: Let X and Y be sets. We say that f is a function (or a map) from X
to Y , and we write f : X → Y , when to each element x ∈ X there is assigned a unique
element y = f (x) ∈ Y . The set X is called the domain of f , and the image (or range)
of f is the set 
Image (f ) = f (X) = f (x) x ∈ X .

More generally, for U ⊆ X, the image of U under f is the set f (U ) = f (x) x ∈ U . For
V ⊆ Y , the inverse image of V under f is the set

f −1 (V ) = x ∈ X f (x) ∈ U .


The graph of f is the set


Graph (f ) = (x, y) ∈ X × Y x ∈ X, y = f (x) .

We say that f is a multi-function from X to Y , and we use the same notation f : X → Y ,

when f is a function from X to the set of all subsets of Y .
2.2 Note: A map f : U ⊆ R → R can be visualized by drawing a picture of its graph,
which is a curve in R2 .
2.3 Note: A map f : U ⊆ R → C can be visualized by drawing its image, which is
typically a curve in C.
2.4 Example: The line segment from a ∈ C to b ∈ C is the image of the map

z(t) = a + t(b − a) , 0 ≤ t ≤ 1 .

2.5 Example: The circle centred at a ∈ C with radius r > 0 is the image of the map

z(t) = a + r ei t , 0 ≤ t ≤ 2π .

2.6 Example: Describe and sketch the image of the map z(t) = (1 + i t)2 .
Solution: We can sketch the image of any map z(t) simply by plotting points. Try plotting
the points z(t) for t = −2, −1, 0, 1, 2. For this particular map, we can eliminate the
parameter t to describe the image: z(t) = (1 + i t)2 = (1 − t2 ) + i (2t) so we have x = 1 − t2
and y = 2t, and so x = 1 − 41 y 2 . This shows that the image is the parabola x = 1 − 41 y 2 .

t z(t)
−2 −3 − 4i
−1 −2i
0 1
1 2i
2 3 + 4i

9
2.7 Example: Describe and sketch the image of the map z(t) = sin(2t)ei t .
Solution: We have z(t) = r(t)ei θ(t) where r(t) = sin(2t) and θ(t) = t. Plot the points
π
r(t)ei θ(t) for t = 12 k, k = 0, 1, 2, · · · , 24 on a polar grid (the cartesian grid consists of
vertical lines x = const. and horizontal lines y = const., while the polar grid consists of
cirles r = const. and rays θ = const.). You will see that the curve is a four-leafed rose: it
consists of one loop in each of the four quadrants.

θ=t r = sin(2t)
0 0
π/12 √1/2
π/6 3/2
π/4 √1
π/3 3/2
5π/6 1/2
π/2 0

2.8 Note: To visualize a map f : U ⊆ C → R we can draw the level curves (also called
contour lines). These are the inverse images f −1 (u) = {z ∈ C|f (z) = u} of constant
values u ∈ R, and they are typically curves in U ⊆ C. We can also use the level curves of
f to help draw its graph, which is a surface in R3 .
2.9 Example: Describe the level curves and the graph of the map u = f (z) = Re (z).
Solution: Foru ∈ R we have f −1 (u) = {u + i y | y ∈ R}, which is the line x = u. Also,
Graph (f ) = (x, y, z) ∈ R3 u = x , which is the plane through the origin perpendicular
to the vector (1, 0, −1).
2.10 Example: Sketch some level curves and sketch the graph of u = f (z) = |z|2 .
Solution: For u ∈ R we have f −1 (u) = {x + i y | x2 + y 2 = u}. When u < 0, this is empty,
√
when u = 0 it is the origin, and when u > 0 it is the circle about the origin of radius u .
Also, we have Graph (f ) = (x, y, z) u = x2 + y 2 , which is a paraboloid.

10
2.11 Example: Sketch some level curves of u = f (z) = Re (1/z).
x
Solution: We have u(x + i y) = 2 . When u = 0 we have x = 0, and when u 6= 0 we
x + y2
x
have 2 = u ⇐⇒ x = u x2 + u y 2 ⇐⇒ x2 − ux + y 2 = 0 ⇐⇒ (x − 2u 1 2
) + y 2 = 4u1 2
x + y2
1 1
so the level curve u =constant is the circle centred at ( 2u , 0) with radius 2u . These circles
all go through the origin. If you sketch several of them you will see that they form the
pattern which is made by the electric field of a dipole (a small bar magnet).

2.12 Note: To visualize a map f : U ⊆ C → C we can sketch the images of various curves
in the domain (if z = x + i y then we usually draw the images of the lines x = const. and
y = const. while if z = r ei θ then we draw the images of the circles r = const. and the
rays θ = const.). Alternatively, we can draw the inverse images of various curves in the
range (if w = f (z) with w = u + i v then we might draw the inverse images of the lines
u = const. and v = const.)
2.13 Example: Give a geometric description of the map w(z) = a z + b where a ∈ C and
b ∈ C. Sketch the images of the lines x = −1, 0, 1 and y = −1, 0, 1 when z = x + i y and
a = 1 + 2i and b = 4 + 3i.
Solution: If a = rei α and z = s ei β then a z = (r s)ei(α+β) , so multiplying z by a has
the effect of scaling z by a factor of r = |a| and rotating the result about the origin by
the angle α = θ(a). Adding b is the same as translating by b. This geometric description
shows that the three vertical lines x = −1, 0, 1 will be sent to the three lines which are
parallel to a i = −2 + i and which pass through the points w(−1) = 3 + i, w(0) = 4 + 3i
and w(1) = 5 + 5i, respectively, and the three horizontal lines y = −1, 0, 1 are sent
to the three lines parallel to a = 1 + 2i through w(−i) = 6 + 2i, w(0) = 4 + 3i and
w(i) = 2 + 4i, respectively. This can also be shown algebraically. For example, the
vertical line x = c is given parametrically by z(t) = c + i t, t ∈ R, and it is sent to
w(z(t)) = a(c + i t) + b = ac + b + i at = w(c) + at, which is the line through w(c) parallel
to i a.

11
2.14 Example: Let w(z) = z 4 . Describe the images of the circles r = const. and the rays
θ = const. where z = r ei θ . Also, sketch the image of the line x = 1, where z = x + i y.

Solution: We have w = (r ei θ )4 = r4 ei 4θ , so if w = s ei φ then we have s = r4 and φ = 4θ.

Thus the circle r = c is mapped to the circle s = c4 and the ray θ = α is mapped to the
ray φ = 4α. The line x = 1 is given parametrically by z = 1 + i t and it is mapped to
the curve w(t) = (1 + i t)4 = 1 + 4t i − 6t2 − 4t3 i + t4 = (1 − 6t2 + t4 ) + i (4t − 4t3 ), so
its image is the curve given parametrically by u(t) = 1 − 6t2 + t4 and v(t) = 4t − 4t3 .
The u-intercepts occur when v = 0, √ that is when t = 0, ±1 and the v-intercepts occur
when u = 0, that is when t = 3 ± 2 2. Also, We have u0 (t) = −12t + 4t3 = 4t(t2 − 3)
2

and v 0 (t) = 4 − √
12t2 = 4(1 − 3t2 ), and so the curve is vertical when u0 (t) = √0, that
0
is when t = 0, ± 3 and it is horizontal when v√ that is when t = ±1/ 3. To
(t) = 0, √
sketchpthe curve, plot the points when t = 0, ±1/ 3, ±1, ± 3, ±2, and perhaps also when
√
t = ± 3 ± 2 2.

t u v
−2
√ 7 25
√
− 3 −8 8 3
−1√ −4 √0
−1/ 3 −8/9 −8 3/9
0√ 1 √0
1/ 3 −8/9 8 3/9
√1 −4 0√
3 −8 −8 3
2 7 −25

1
2.15 Example: Let w(z) = . Describe the images of the circles r = const. and the rays
z
θ = const., and then describe the images of the lines x = const. and y = const.

1
Solution: If z = r ei θ and w = s ei φ then we have w = −i θ
= 1r ei θ so that s = 1r
re
and φ = θ. This map is known as the inversion in the unit circle: the circle r = c is
mapped to the circle s = 1/c while the ray θ = α is mapped to itself. If z = x + i y
and w = u + i v then the vertical line x = c is given parametrically by z(t) = c + i t
c + it c
and it is sent to w(z(t)) = 2 2
, so its image is the curve given by u(t) = 2 and
c +t c + t2
t
v(t) = 2 . When c = 0 we have u = 0 and v = t/t2 = 1/t, so the line x = 0 (excluding
c + t2
the origin) is mapped to the line u = 0 (excludind the origin). When c 6= 0, we can use the
expression for u(t) to solve for t to get t2 = (c − u c2 )/u and then we can substitute this
into the expression v 2 (t) = t2 /(c2 + t2 )2 and simplify to get v 2 = 1c u − u2 or equivalently
1 2 1 2 1
(u − 2c ) + v 2 = ( 2c ) . Thus the image of the line x = c, c 6= 0 is the circle centred at 2c
1
with radius 2|c| , excluding the origin. Similarly, the image of the horizontal line y = c is
1 1
the circle centred at 2c i with radius 2|c| , excluding the origin.

12
2.16 Definition: We define the exponential function by

ex+i y = ex ei y = ex cos y + i ex sin y .

We also write exp(z) = ez .

2.17 Note: It is not hard to check that the exponential function has the following prop-
erties for all complex numbers z and w:

e0 = 1
e−z = 1/ez , enz = (ez )n , n ∈ Z
ez+w = ez ew , ez−w = ez /ew
ez = ew ⇐⇒ w = z + i 2πk for some k ∈ Z

2.18 Example: Let w(z) = ez . Describe the images of the lines x = const. and y = const.
where z = x + i y.

Solution: We have w = ex ei y, so if w = r ei θ then we have r = ex and θ = y. So the

vertical line x = c is mapped to the circle r = ec , and the horizontal line y = c is mapped
to the ray θ = c. Notice that the domain of ez is all of C while the range is C \ {0}.

2.19 Definition: We define the trigonometric functions by

ei z − e−i z ei z + e−i z sin z

sin z = , cos z = , tan z =
2i 2 cos z

and sec z = 1/ cos z, csc z = 1/ sin z and cot z = cos z/ sin z. We define the hyperbolic
functions by

ez − e−z ez + e−z sinh z

sinh z = , cosh z = , tanh z =
2 2 cosh z

and coth z = cosh z/ sinh z.

13
2.20 Note: It is not hard to verify the following properties, where z, w ∈ C:

sin(z + 2π) = sin z , cos(z + 2π) = cos z

sin(−z) = − sin z , cos(−z) = cos z
2 2
sin z + cos z = 1
sin(z + w) = sin z cos w + cos z sin w , sin(2z) = 2 sin z cos z
cos(z + w) = cos z cos w − sin z sin w , cos(2z) = cos2 z − sin2 z
sinh(−z) = − sinh z , cosh(−z) = cosh z
cosh2 z − sinh2 z = 1
sinh(z + w) = sinh z cosh w + cosh z sinh w , sinh(2z) = 2 sinh z cosh z
cosh(z + w) = cosh z cosh w + sinh z sinh w , cosh(2z) = cosh2 z + sinh2 z

In fact all of the trigonometric identities and hyperbolic identies which hold for real num-
bers also hold for complex numbers. Here are some more properties:

sinh(z + i 2π) = sinh z , cosh(z + i 2π) = cosh z

sinh(i z) = i sin z , cosh(i z) = cos z
sin(i z) = i sinh z , cos(i z) = cosh z
sin(x + i y) = sin x cosh y + i cos x sinh y , | sin(x + i y)|2 = sin2 x + sinh2 y
cos(x + i y) = cos x cosh y − i sin x sinh y , | cos(x + i y)|2 = cos2 x + sinh2 y
sinh(x + i y) = sinh x cos y + i cosh x sin y , | sinh(x + i y)|2 = sinh2 x + sin2 y
cosh(x + i y) = cosh x cos y + i sinh x sin y , | cosh(x + i y)|2 = sinh2 x + cos2 y

2.21 Example: Find sin( π6 + i ln 2).

Solution: We have
√ √
sin( π6 + i ln 2) = sin( π6 ) cosh(ln 2) + i cos( π6 ) sinh(ln 2) = 1
2 · 5
4 +i 2
3
· 3
4 = 5+3 3 i
8 .

2.22 Example: Solve cosh z = −2.

Solution: If z = x + i y then we have cosh z = cosh x cos y + i sinh x sin y, so we have
cosh z = −2 when cosh x cos y = −2 and sinh x sin y = 0. We cannot have sinh x = 0, since
if sinh x = 0 then x = 0 so cosh x cos y = cos y 6= −2. So we must have sin y = 0 and so
y = kπ for some k ∈ Z and we have cos y = ±1. To have cosh x cos y = −2, we must have
cos y = −1 and cosh x = 2 (since cosh x is always positive). We can solve cosh√ x = 2 as
x −x x 2 x x
follows: cosh x =√2 ⇐⇒ e + e = 4 ⇐⇒ (e )√− 4e + 1 = 0 ⇐⇒ e √= 2 ± 3 so we
have x = ln(2 ± 3) or equivalently x = ± ln(2 + 3). Thus z = ± ln(2 + 3) + i (π + 2πk)
for some k ∈ Z.
2.23 Example: Let w(z) = sin z. Describe the images of the lines x = const. and
y = const. where z = x + i y.
Solution: The vertical line x = c is given parametrically by z(t) = c + i t and it is mapped
to the curve w(t) = sin(c + i t) = sin c cosh t + i cos c sinh t. If w = u + i v then we have
u(t) = sin c cosh t and v(t) = cos c sinh t. Using the identity cosh2 t − sinh2 t = 1 we obtain

14
 u2 v2
− = 1, provided that t 6= π2 k, k ∈ Z. This is the equation of a hyperbola.
sin2 c cos2 c
The image of the line x = c will be one of the two branches of this hyperbola; when sin c
is positive u(t) is also positive and the image is the branch on the right; when sin c is
negative, the image is the branch on the left. When sin c = 0 (so that c = πk), the image
is the line u = 0, that is, the v-axis. When cos c = 0, the image lies on the line v = 0 (the
u-axis) and it is either the interval [1, ∞) (when sin c = 1) or else the interval (−∞, −1]
(when sin c = −1).
The horizontal line y = c is given parametrically by z(t) = t + i c and it is mapped to
w(t) = sin t cosh c + i cos t sinh c so we have u(t) = sin t cosh c and v(t) = cos t sinh t. Since
u2 v2
sin2 t + cos2 t = 1 we have + = 1. The line y = c is mapped to this ellipse,
cosh2 c sinh2 c
unless c = 2πk i in which case the image can be seen to be the line segment [−1, 1] on the
u-axis.
If you sketch a few of these hyperbolas and ellipses, you will get a nice picture showing
two orthogonal families of curves. You will see that the domain and the range of sin z are
both equal to C.

2.24 Definition: Let X and Y be sets and let f : X → Y . We say that f is one-to-one,
written as f is 1:1, (or that f is injective) when for every y ∈ Y there exists at most
one x ∈ X such that f (x) = y. We say that f is onto (or surjective) when for every
y ∈ Y there exists at least one x ∈ X such that f (x) = y. We say that f is invertible (or
bijective) when f is both one-to-one and onto, that is when for every y ∈ Y there exists
exactly one x ∈ X such that f (x) = y.
When a map f from X to Y is invertible, it has a unique inverse function from Y
to X, denoted by f −1 , which is defined by

f (x) = y ⇐⇒ f −1 (y) = x

or equivalently by
f (f −1 (y)) = y , f −1 (f (x)) = x
for all x ∈ X and y ∈ Y . When f from X to Y is not invertible, we define its inverse
multi-function from Y to X given by

f −1 (y) = x ∈ X f (x) = y .


2.25 Note: When a map f : X → Y is 1:1 (but perhaps not onto), the map f : X → f (X)
is both 1:1 and onto, and hence invertible. When a map f : X → Y is not 1:1, then
sometimes we can find a subset U ⊂ X such that the restriction f : U ⊂ X → Y is 1:1,

15
and then the map f : U → f (U ) is invertible. In this case, the inverse function of the
map f : U → f (U ) is called a branch of the inverse multi-function. When working with
complex-valued functions of a complex variable, we shall sometimes use the notation f −1
to denote the inverse multi-function of f , and we shall sometimes use the notation f −1 to
denote some branch of the inverse multi-function.
2.26 Example: In real variable calculus, to define sin−1 x it is customary to restrict the
domain of sin x to − π2 ≤ x ≤ π2 so that it becomes 1:1. If we thought instead of sin−1 x as
a multi-function then for example we would have sin−1 ( 12 ) = π6 + 2πk, 5π

6 + 2πk k ∈ Z .

2.27 Example: The polar coordinates map f : (r, θ) r > 0, θ ∈ R → R2 given by
f (r, θ) = (r cos θ, r sin θ) is not 1:1. We can make it 1:1 by restricting the domain in various
ways. For example for any θ0 ∈ R, if we restrict the domain to

{(r, θ)|r > 0, θ0 < θ < θ0 + 2π}

then f becomes 1:1 and its inverse is given by f −1 (x, y) = (r, θ) where r = |x + i y| and
θ = θ(x + i y) with θ0 < θ(x + i y) < θ0 + 2π. Alternatively, if we think of f −1 as a multi-
function, then we can still write f −1 (x, y) = (r, θ) where r = |x + i y| and
 θ = θ(x + i y),
but this time θ(x + i y) denotes an infinite set of the form θ(x + i y) = θ0 + 2π k k ∈ Z
with say 0 ≤ θ0 < 2π.
2.28 Definition: The inverse of the exponential function ez is the logarithmic function
(or the logarithmic multi-function), denoted by log z.
2.29 Example: Find a formula for log z.
Solution: Let z = r ei θ and w = u + i v. Then w = log z ⇐⇒ ew = z ⇐⇒ eu ei v = r ei θ ,
which happens when eu = r and v = θ + 2πk for some k ∈ Z. Thus

log r ei θ = ln r + i(θ + 2πk), k ∈ Z .

This is the formula for the multi-valued logarithm. We can obtain a branch of the logarithm
by restricting the domain of the exponential function
 in various ways. For example, for
any θ0 ∈ R, if we restrict its domain to the set rei θ r > 0, θ0 < θ < θ0 + 2π , then it
becomes invertible with inverse function

log rei θ = ln r + i θ , where θ0 < θ < θ0 + 2π .

2.30 Example: Find log(1 − i)

√ √
Solution: log(1 − i) = log( 2e−i π/4 ) = ln 2 + i (− π4 + 2πk), k ∈ Z.
2.31 Note: For the multi-valued logarithm, you should convince yourself that the follow-
ing formulas make sense and they all hold:

elog z = z
log(z w) = log z + log w
log(z/w) = log z − log w
2.32 Definition: We can use the logarithm to define complex exponents: given a ∈ C
we define
z a = exp(a log z) .

16
2.33 Example: Find i−2i .
Solution: i−2i = exp(−2i log i) = exp(−2i (i ( π2 + 2πk)) = exp(π + 4πk), k ∈ Z.
2.34 Example: Find a formula for z 2/3 .
Solution: Write z = r ei θ . Then we have

z 2/3 = exp( 23 log z) = exp 2

3 ln r + i(θ + 2π k) = r2/3 ei(2θ+4πk)/3 , k ∈ Z .

Notice that there are three distinct values of z 2/3 obtained by taking k ∈ {0, 1, 2}.
2.35 Note: Check that

z n = exp(n log z) is single valued for 0 < n ∈ Z

z 1/n = exp n1 log z takes n values for 0 < n ∈ Z

z −a = (z a )−1

2.36 Definition: The inverse trigonometric functions are denoted by sin−1 z, cos−1 z,
tan−1 z and so on. The inverse hyperbolic functions are denoted by sinh−1 z, cosh−1 z,
tanh−1 z and so on.
2.37 Note: Since the trigonometric and the hyperbolic functions are defined using the
exponential function, their inverses can be expressed in terms of the logarithmic function:
 p 
sin−1 z = −i log i z + 1 − z 2
 p 
−1 2
cos z = −i log z + z − 1
i i+z
tan−1 z = log
2 i−z
 p 
−1 2
sinh z = log z + z + 1
 p 
cosh−1 z = log z + z 2 − 1
1 1+z
tanh−1 z = log
2 1−z

where the square roots are double valued. Let us derive the formula for sin−1 z. We have
w = sin−1 z ⇐⇒ √ z = sin w ⇐⇒ z = (e − e
iw −iw iw 2 iw
√ ⇐⇒
(e ) − 2iz(e ) − 1 = 0
)/2i
iw 2 2
⇐⇒ e = iz + 1 − z so we obtain iw = log iz + 1 − z , as required.
2.38 Example: Find cosh−1 (−2).
Solution: We already did this in example 2.22, but let us do it again using the above
logarithmic formula. We have
√ √ √
cosh−1 (−2) = log(−2 ± 3) = log((2 ± 3)ei π ) = ln(2 ± 3) + i (π + 2πk), k ∈ Z .

17
 Chapter 3. Sets, Limits and Continuity

3.1 Notation: Given a, b ∈ C and r ∈ R, let L(a, b) denote the line segment from a to b

L(a, b) = {a + t(b − a)|0 ≤ t ≤ 1} ,

let S(a, r) denote the circle about a of radius r


S(a, r) = z ∈ C |z − a| = r ,

and let D(a, r), D(a, r) and D∗ (a, r) denote the open disc, the closed disc, and the punc-
tured disc, centred at a of radius r

D(a, r) = z ∈ C |z − a| < r

D(a, r) = z ∈ C |z − a| ≤ r
D∗ (a, r) = z ∈ C 0 < |z − a| < r


We use analogous notation for subsets of Rn . For example in R3 , S(a, r) is the sphere of
radius r centered at a and D(a, r) is the open ball of radius r centered at a.
3.2 Definition: Let E ⊆ C (or more generally E ⊆ Rn ).
(a) A point a ∈ E is an interior point of E when ∃r > 0 D(a, r) ⊆ E.
(b) The interior of E, denoted by E 0 , is the set of all interior points of E.
(c) E is open when every point of E is an interior point, in other words when E = E 0 .
(d) A point a ∈ C is a limit point of E when ∀r > 0 D∗ (a, r) ∩ E 6= ∅.
(e) The closure of E, denoted by E, is the union of E with the set of all its limit points.
(f) E is closed when every limit point of E is an element of E, that is when E = E.
(g) The boundary of E is the set ∂E = E \ E 0 .
(h) The complement of E is the set E c = {z ∈ C|z ∈ / E}.
3.3 Example: For a, b ∈ R, we have L(a, b) = [a, b], L(a, b)0 = (a, b), L(a, b) = [a, b]
and ∂L(a, b) = {a, b}. For a, b ∈ C, L(a, b) is closed, L(a, b)0 = ∅, L(a, b) = L(a, b) and
∂L(a, b) = L(a, b).
3.4 Example: In C, the open disc D(a, r) and the punctured disc D∗ (a, r) are both open,
while the closed disc D(a, r) is closed. Their interiors are D(a, r)0 = D(a, r)0 = D(a, r) and
D∗ (a, r)0 = D∗ (a, r). Their closures are all equal: D(a, r) = D(a, r) = D∗ (a, r) = D(a, r).
Their boundaries are ∂D(a, r) = ∂D(a, r) = S(a, r) and ∂D∗ (a, r) = S(a, r) ∪ {0}.

3.5 Example: The annulus A = z ∈ C r < |z − a| ≤ R is neither open nor closed.
Its interior is the open annulus A0 = z ∈ C r < |z − a| < R , its closure is the closed

annulus A = z ∈ C r ≤ |z − a| ≤ R , and its boundary is the union of the two circles
∂A = S(a, r) ∪ S(a, R).
3.6 Example: Let E = 11 , 12 , 13 , · · · . Then E is neither open nor closed: its interior is


empty E 0 = ∅, its only limit point is 0, its closure is E = E ∪ {0}, and its boundary is
equal to its closure ∂E = E.

18
3.7 Theorem: Let E ⊆ C (or more generally, let E ⊆ Rn ). Then
(a) E is open iff E c is closed and E is closed iff E c is open.
(b) E 0 is open and if U ⊆ E is open then U ⊆ E 0 .
(c) E is closed and if E ⊆ K with K closed then E ⊆ K.
Proof: We prove the first half of part (a) and we prove part (b). Suppose first that E
is open. We claim that E c is closed. Let a be a limit point of E c . We must show that
a ∈ E c . Suppose, for a contradiction, that a ∈ E. Since E is open, we can choose r > 0
so that D(a, r) ⊆ E. Then we have D∗ (a, r) ⊂ D(a, r) ⊆ E, so D∗ (a, r) ∩ E c = ∅, which
contradicts the fact that a is a limit point of E c .
Conversely, suppose that E c is closed, and suppose, for a contradiction, that E is
not open. Since E is not open, we can choose a ∈ E such that there is no r > 0 with
D(a, r) ⊆ E. Then for all r > 0 we have D(a, r) ∩ E c 6= ∅. Since a ∈ / E c , this implies that
for all r > 0 we have D∗ (a, r) ∩ E c 6= ∅ and so a is a limit point of E c . Since E c is closed,
we have a ∈ E c , which contradicts the fact that a ∈ E. This completes the proof of the
first half of part (a).
Now let us prove part (b). First we shall show that if U ⊆ E is open then U ⊆ E 0 .
Suppose that U is open with U ⊆ E. Let a ∈ U . We must show that a ∈ E 0 . Since U is
open we can choose r > 0 so that D(a, r) ⊆ U . Since U ⊆ E we have D(a, r) ⊆ E, and so
a ∈ E 0 , as required.
Next we shall show that E 0 is open. Let a ∈ E 0 . Choose r > 0 so that D(a, r) ⊆ E.
Since D(a, r) is open and D(a, r) ⊆ E, we have D(a, r) ⊆ E 0 by the result in the previous
paragraph. Thus E 0 is open, and this completes the proof of part (b).
3.8 Theorem: Let {Eα |α ∈ A} be a (possibly infinite) family of sets in C (or more
n
generally,
T in
R S let c{F1 , F2 , · S
) and finite family of sets in C (or in Rn ). Then
· · , Fn }
be a T
c c
(a) Eα = (Eα ) and Eα = (Eα c ).
α∈A α∈A α∈A α∈A
S n
T
(b) If the sets Eα and Fi are open then Eα and Fi are open.
α∈A i=1
Sn T
(c) If the sets Eα and Fi are closed then Fi and Eα are closed.
i=1 α∈A

Proof: We prove the first half of part (a) and we prove part (b). To prove the first half of
part (a), note that
 \ c \
x∈ Eα ⇐⇒ x ∈ / Eα ⇐⇒ ∃ α ∈ A , x ∈ / Eα
α∈A α∈A
[
⇐⇒ ∃ α ∈ A , x ∈ Eα c ⇐⇒ x ∈ Eα c .
α∈A
S
To prove part (b), suppose first that the sets Eα are all open. Let a ∈ Eα . Choose
α∈A
α ∈ A so that a ∈ Eα .SSince Eα is open,
S we can choose r > 0 so that D(a, r) ⊆ Eα , and
then D(a, r) ⊆ Eα ⊆ Eα . Thus Eα is open, as required.
α∈A α∈A
n
T
Next suppose that the sets F1 , F2 , · · · , Fn are all open. Let a ∈ Fi . For each i,
i=1
choose ri > 0 so that D(a, ri ) ⊆ Fi . Let r = min{r1 , r2 , · · · , rn }. Then for all i we have
n
T n
T
D(a, r) ⊆ D(a, ri ) ⊆ Fi , and so D(a, r) ⊆ Fi . Thus Fi is open, as required.
i=1 i=1

19
3.9 Definition: Let E ⊆ C (or more generally, let E ⊆ Rn ).
(a) E is bounded when ∃ r > 0 E ⊆ D(0, r).
(b) E is convex when ∀ a, b ∈ E L(a, b) ⊆ E.
(c) We say that two open sets U and V separate E when E ⊆ U ∪V , U ∩V = ∅, U ∩E 6= ∅
and V ∩ E 6= ∅. We say that E is connected when there do not exist two such open sets.
(d) E is compact when every open cover of E admits a finite subcover, in other words,
S n
S
when for every collection of open sets Uα , α ∈ A with E ⊆ Uα , we have E ⊆ Uαi
α∈A i=1
for some α1 , α2 , · · · , αn ∈ A.
3.10
 Example: The segment L(a, b) and the disc D(a, r) are both bounded. The line
z ∈ C |z − a| = |z − b| and the half-plane z ∈ C |z − a| < |z − b| are both unbounded.
3.11 Example: The sets L(a, b), D(a, r) and D(a, r) are all convex. The punctured disc
D∗ (a, r) is not convex; for example if u = a − 2r and v = a + 2r then u, v ∈ D∗ (a, r) but
L(u, v) 6⊆ D∗ (a, r). Also, the half-annulus E = z ∈ C 1 < |z| < 2, Re (z) > 0 is not
convex; for example, take u and v to be 21 ± i.
3.12 Example: Each of the sets L(a, b), D(a, r), D(a, r) and D∗ (a, r) is a connected set.
The union L(0, 1) ∪ L(i, i + 1) is not connected, because it is separated by the disjoint open
sets U = {z|Im (z) < 12 } and V = {z|Im (z) > 21 }. The union D(−1, 1) ∪ D(1, 1) is not
connected since it is separated by the two discs D(−1, 1) and D(1, 1). On the other hand,
the union D(−1, 1) ∪ D(1, 1) is connected.
3.13 Theorem: (The Heine-Borel Theorem) Let E ⊆ C (or more generally, let E ⊆ Rn ).
Then the following are equivalent:
(a) E is compact.
(b) Every infinite subset of E has a limit point in E.
(c) E is closed and bounded.
Proof: This theorem is quite difficult to prove and we omit the proof.
3.14 Example: The above theorem makes it easy to recognize compact sets. The segment
L(a, b) and the closed disc D(a, r) are both compact (since they are closed and bounded).
The open disc D(a, r) is not compact, since it is not closed. The line z ∈ C |z−a| = |z−b|
is not compact, since it is not bounded.
3.15 Example: The set E = 11 , 12 , 13 , · · · is not closed (since E 6= E) so it is not


compact. This means that we must be able to find an open cover with no finite subcover.
Indeed, notice that the distance between two neighbouring points in E is n1 − n+11 1
= n(n+1) ,
1 1


so if we let Un = D n , n(n+1) then each point of E lies in exactly one of the discs Un .
∞
S
Thus the discs U1 , U2 , U3 , · · · cover E, that is E ⊆ Un , but if we remove even one of
n=1
these discs then the remaining discs will not cover E.
However, the set K = E = E ∪ {0} is closed and bounded and hence compact, so any
open cover of K must
have a finite subcover. Consider the1 cover {U0 , U1 , U2 , · · ·} where
1 1


Un = D n , n(n+1) for n > 0 (as above) and U0 = D 0, N where N is some positive
integer. Since n1 ∈ U0 whenever n > N , we see that U0 , U1 , U2 , · · · , UN is a finite

N
S
subcover, that is, E ⊆ Un
n=0

20
3.16 Definition: Let f : U ⊆ C → C (or more generally, let f : U ⊆ Rm → Rn ), and let
a be a limit point of U . We write

lim f (z) = b or f (z) → b as z → a

z→a

when for all  > 0 there exists δ > 0 such that for all z ∈ U we have

0 < |z − a| < δ =⇒ |f (z) − b| <  ,

or equivalently z ∈ D∗ (a, δ) =⇒ f (z) ∈ D(b, ). We write

lim f (z) = ∞ or f (z) → ∞ as z → a

z→a

when for all R > 0 there exists δ > 0 such that for all z ∈ U

0 < |z − a| < δ =⇒ |f (z)| > R .

1
3.17 Example: Use the definition to show that lim c = c, lim z = a and lim = ∞.
z→a z→a z→a z−a
Solution: Let f (z) = c. Given  > 0, choose δ to be any positive real number. Then for
any z (and in particular, for any z with 0 < |z − a| < δ) we have |f (z) − c| = |c − c| = 0 < .
Now let f (z) = z. Given  > 0, choose δ = . Then

0 < |z − a| < δ =⇒ |f (z) − a| = |z − a| < δ =  .

1 1
Finally, let f (z) = z−a . Given R > 0, choose δ = R. Then

1 1
0 < |z − a| < δ =⇒ |f (z)| = > = R.
|z − a| δ


3.18 Theorem: Let
f : U ⊆ C → C, and let a be a limit point of U . Let u(z) = Re f (z)
and v(z) = Im f (z) so that f (z) = u(z) + i v(z) with u, v : U → R. Then lim f (z) exists
z→a
if and only if lim u(z) and lim v(z) both exist, and in this case,
z→a z→a

lim f (z) = lim u(z) + i lim v(z) .

z→a z→a z→a

An analogous result holds for functions f : U ⊆ Rm → Rn .

Proof: Suppose first that lim f (z) exists, say lim f (z) = b = s + i t with s, t ∈ R. Note
z→a z→a
that u(z) − s = Re (f (z) − b) so we have |u(z) − s| ≤ |f (z) − b|. So, given  > 0 choose δ > 0
such that 0 < |z − a| < δ=⇒|f (z) − b| <  and then we have |u(z) − s| ≤ |f (z) − b| < .
This shows that lim u(z) = s. Similarly lim v(z) = t.
z→a z→a
Next, we suppose that lim u(z) and lim v(z) exist, say lim u(z) = s and lim v(z) = t
z→a z→a z→a z→a
and let b = s + i t. By the triangle inequality we have |f (z) − b| ≤ |u(z) − s| + |v(z) + t|. So,
given  > 0 we choose δ > 0 such that 0 < |z −a| < δ=⇒ |u(z)−s| < 2 and |v(z)−t| < 2 .

Then we have |f (z) − b| ≤ |u(z) − s| + |v(z) − t| < . This shows that lim f (z) = b.
z→a

21
3.19 Theorem: Let f, g : U ⊆ C → C, let a be a limit point of U . Suppose lim f (z) = p
z→a
and lim g(z) = q, and let c ∈ C. Then
z→a
(a) lim c f (z) = c p
z→a
(b) lim f (z) ± g(z) = p ± q
z→a
(c) lim f (z)g(z) = p q
z→a
(d) lim f (z)/g(z) = p/q provided that q 6= 0
z→a
(e) lim |f (z)| = |p|.
z→a
Analogous results hold for functions f : U ⊆ Rm → Rn , except for parts (c) and (d).
Proof: These can be proven in the same way as the corresponding results for real-valued
functions. For example, to prove part (c), we can use the equality
f (z)g(z) − pq = (f (z) − p)(g(z) − q) + (f (z) − p) q + (g(z) − q) p .
Given  > 0, choose δ > 0 so that whenever 0 < |z − a| < δ we have |f (z) − p| < 3 ,
p

|g(z) − q| < 3 , |f (z) − p| |q| < 3 and |g(z) − q| |p| < 3 . Then for 0 < |z − a| < δ,
p

|f (z)g(z) − pq| ≤ |f (z) − p| |g(z) − q| + |f (z) − p| |q| + |g(z) − q| |p| < 3 3 + 3 + 3 =  .

p p

z 2 − 2z + 5
3.20 Example: Let f (z) = . Find lim f (z).
z − 1 − 2i z→1+2i

Solution: We have
(z − (1 + 2i))(z − (1 − 2i))
lim f (z) = lim
z→1+2i z→1+2i (z − (1 + 2i))
= lim (z − (1 − 2i)) = (1 + 2i) − (1 − 2i) = 4i .
z→1+2i

3.21 Example: For each z ∈ C∗ = C \ {0}, let θ(z) be the angle of z chosen so that
0 ≤ θ(z) < 2π. Then θ : C∗ → [0, 2π). Show that if a ≥ 0 the lim θ(z) does not exist.
z→a

Solution: Suppose (for a contradiction) that lim θ(z) does exist, say lim θ(z) = b. Let
z→a z→a
 = π2 and choose δ > 0 so that z ∈ D∗ (a, δ)=⇒|θ(z) − b| <  = π
2. Since a ± i 2δ ∈ D∗ (a, δ),
we have |θ(a ± i 2δ ) − b| <  = π2 , and so
θ(a + i 2δ ) − θ(a − i 2δ ) ≤ θ(a + i 2δ ) − b + θ(a − i 2δ ) − b < π
2 + π
2 = π.
On the other hand, since a ≥ 0, we have 0 < θ(a + i 2δ ) ≤ π
2 and 3π
2 ≤ θ(a − i 2δ ) < 2π, so
θ(a + i 2δ ) − θ(a − i 2δ ) ≥ 3π
2 − π
2 =π
which gives the desired contradiction.
3.22 Definition: Let f : U ⊆ C → C (or more generally, let f : U ⊆ Rm → Rn ), and let
a ∈ U . We say that f is continuous at a when for all  > 0 there exists δ > 0 such that
for all z ∈ U we have
|z − a| < δ =⇒ |f (z) − f (a)| < 
or equivalently z ∈ D(a, δ) =⇒ f (z) ∈ D(f (a), ). In the case that a is a limit point of U ,
this is equivalent to
lim f (z) = f (a) .
z→a
We say that f is continuous in U when f is continuous at every point a ∈ U .

22



3.23 Theorem: Let f : U ⊆ C → C, let u(z) = Re f (z) and v(z) = Im f (z) . Then
f is continuous at a ∈ U if and only if both u and v are continuous at a. An analogous
result holds for f : U ⊆ Rm → Rn .

Proof: This follows from Theorem 3.18.

3.24 Theorem: Let f, g : U ⊆ C → C. Suppose f and g are both continuous at a ∈ U

and let c ∈ C. Then
(a) cf is continuous at a
(b) f ± g is continuous at a
(c) f g is continuous at a
(d) f /g is continuous at a, provided that g(a) 6= 0.
(e) |f | is continuous at a.
Analogous results hold for f, g : U ⊆ Rm → Rn , except for parts (c) and (d).

Proof: This follows from Theorem 3.19.

3.25 Theorem: Let f : U ⊆ C → V ⊆ C and let g : V ⊆ C → C. If f is continuous at

a ∈ U and g is continuous at f (a) ∈ V then g ◦ f is continuous at a.
An analogous result holds for f : U ⊆ Rl → V ⊆ Rm and g : V ⊆ Rm → Rn .

Proof: Suppose that f is continuous at a ∈ U and that g is continuous at b where b = f (a).

Given  > 0, first choose δ0 so that for w ∈ V we have w ∈ D(b, δ0 ) =⇒ g(w) ∈ D(g(b), ),
and then choose δ > 0 so that for z ∈ U we have z ∈ D(a, δ) =⇒ f (z) ∈ D(b, δ0 ). Then
for z ∈ U we have

z ∈ D(a, δ) =⇒ f (z) ∈ D(b, δ0 ) =⇒ g(f (z)) ∈ D(g(b), ) = D(g(f (a)), ) .

3.26 Example: Let U = {r ei θ |r > 0, 0 < θ < 2π}. Let θ : U → (0, 2π) be the angle
function. Show that θ is continuous in U .

Solution: Write z = x + i y with x, yp∈ R. For
Im (z) > 0, the angle function is given
−1 2 2
by the formula θ(x + i y) = cos x/ x + y . This formula expresses θ(x + i y) using
sums, products, quotients and composites of known continuous functions, and so it must
be continuous, by parts b) and c) of the above theorem. Thus θ(z) is continuous at all
points z with Im (z) > 0.
Similarly, for Re (z) < 0, θ(z) is given by the formula θ(x+i y) = π +tan−1 y/x , and


p
for Im (z) < 0 we have θ(x + i y) = 2π − cos−1 x/ x2 + y 2 . These are both continuous

and so θ(z) is continuous for all z ∈ U .

3.27 Note: If we choose θ(z) ∈ [0, 2π) for all z ∈ C∗ , then we have seen (in example
3.21) that for a > 0, lim θ(z) does not exist, so θ : C∗ → [0, 2π) is not continuous in C∗ .
z→a
In fact it is impossible to choose θ(z) ∈ R so that θ : C∗ → R is continuous in C∗ . As in
the above example, we must restrict the domain to make the angle function continuous.
Indeed, for any α ∈ R, if we restrict the domain to Uα = {r ei θ |r > 0, α < θ < α + 2π}
and choose θ(z) with α < θ(z) < α + 2π then θ : Uα → (α, α + 2π) will be continuous.

23
3.28 Note: We have found formulas for the real and imaginary parts of the identity
f (z) = z, the exponential f (z) = ez , the trigonometric functions, and the hyperbolic
functions. These formulas reveal that they are all continuous in their domains. Also,
any branch of the logarithm log z = ln |z| + i θ(z) is continuous provided that θ(z) is
chosen to be continuous. The inverse trigonometric and inverse hyperbolic functions can
all be expressed in terms of the logarithm, and so they are also continuous provided that
θ(z) is chosen to be continuous. Any complex function which can be expressed using sums,
products, quotients and composites of the above functions will be continuous in its domain.
3.29 Theorem: Let f : U ⊆ C → C (or let f : U ⊆ Rm → Rn ) where U is open. Then
(a) f is continuous if and only if f −1 (V ) is open for every open set V ⊆ C.
(b) If f is continuous and E ⊆ U is connected then f (E) is connected.
(c) If f is continuous and K ⊆ U is compact then f (K) is compact.
(d) If f is continuous and K ⊆ U is compact then |f (x)| attains its extreme values on K.
Proof: We prove part (a). Suppose first that f is continuous and let V ⊆ C be open.
We must show that f −1 (V ) is open, so given a ∈ f −1 (V ) we need to show there exists
δ > 0 such that D(a, δ) ⊆ f −1 (V ). Let a ∈ f −1 (V ), which means that f (a) ∈ V .
Choose  > 0 so that D(f (a), ) ⊆ V . Then choose δ > 0 so that D(a, δ) ⊆ U and
f (D(a, δ)) ⊆ D(f (a), ) ⊆ V . Then we have D(a, δ) ∈ f −1 (V ).
Next we suppose that f −1 (V ) is open for every open set V ⊆ C. We want to show
that f is continuous at every point a ∈ U . Let a ∈ U . Given  > 0, let V = D(f (a), ).
Then a ∈ f −1 (V ) and since V is open, f −1 (V ) is open, so we can choose δ > 0 such
that D(a, δ) ⊆ f −1 (V ). Then we will have f (D(a, δ)) ⊆ D(f (a), ). This shows that f is
continuous.
To prove part (b) we suppose that f is continuous and that E is connected, and
(looking for a contradiction) we shall suppose that f (E) is not connected. Say f (E) ⊆
V ∪ W where V and W are disjoint open sets which separate E. You may check that the
open sets f −1 (V ) and f −1 (W ) separate E giving a contradiction.
To prove part (c), suppose that f is continuous and that K ⊆ U is compact. We wish
to show that f (K) is compact. Let {Vα } be an open cover of f (K). Since f is continuous,
each of the sets f −1 (Vα ) will be open, and you can check that they cover K. Since K is
compact, we can find a finite subcover, say K ⊆ f −1 (Vα1 ) ∪ · · · ∪ f −1 (Vαn ). You may check
that this implies that we have f (K) ⊆ Vα1 ∪ · · · ∪ Vαn , so {Vα } has a finite subcover.
Part (d) follows from part c), because if f is continuous and K ⊆ U is compact then
|f | will also be continuous and so |f |(K) = |f (z)| z ∈ K is a compact set in R. Any
closed and bounded set in R includes its extreme values.

24
 Chapter 4. Derivatives

4.1 Note: From now on, we shall always use the letter U to denote an open set.
4.2 Definition: Recall that for a function f : U ⊆ R → R we define

f (x) − f (a)
f 0 (a) = lim
x→a x−a

provided the limit exists, and then we say that f is differentiable at x = a and f 0 (a)
is called the (real) derivative of f at a. Equivalently, we see that f is differentiable at
f (x) − f (a)
x = a if there exists a real number f 0 (a) such that lim − f 0 (a) = 0. This last
x→a x−a
|R(x)|
= 0, where R(x) = f (x) − f (a) + f 0 (a)(x − a) .


condition can be rewritten as lim
x→a |x − a|
In this way we obtain a definition which applies to functions f : U ⊆ Rn → Rm .
A function f : U ⊆ Rn → Rm is differentiable at x = a if there exists an m × n
|R(x)|

matrix Df (a) such that lim = 0, where R(x) = f (x) − f (a) + Df (a)(x − a) .
x→a |x − a|
The matrix Df (a) is called the (real) derivative matrix of f at x = a. We say that
f : U ⊆ Rn → Rm is differentiable in U if it is differentiable at every point a ∈ U .
For a map f : U ⊆ Rn → R, the j th partial derivative of f is given by

∂f
fxj (a) = (a) = g 0 (0) ,
∂xj

if it exists, where g(t) = f (a + t ej ) with ej denoting the j th standard basis vector in Rn .

We now recall (without proof) some theorems from vector calculus.
4.3 Theorem: Let f
: U ⊆ Rn → Rm , and let fi be the components of f so that
f (x) = f1 (x), · · · fm (x) . Then if f is differentiable at x = a then the partial derivatives
∂fi
∂xj all exist and
 ∂f1 ∂f1 ∂f1 
∂x1 (a) ∂x2 (a) . . . ∂x n
(a)
Df (a) = 
 .. .. .. 
. . . 
∂fm ∂fm ∂fm
∂x1 (a) ∂x2 (a) · · · ∂xn (a)

4.4 Theorem: If f : U ⊆ Rn → Rm is C 1 in U , which means that the partial derivatives

∂fi
∂xj all exist and are continuous in U , then f is differentiable in U .

4.5 Theorem: If f : U ⊆ Rn → Rm is differentiable at a then f is continuous at a.

4.6 Theorem: If f, g : U ⊆ Rn → Rm are both differentiable at x = a, then
(a) D(c f )(a) = c Df (a) where c ∈ R.
(b) D(f ± g)(a) = Df (a) ± Dg(a).
(c) (The Product Rule) If m = 1 then D(f g)(a) = Df (a)g(a) + f (a)Dg(a).

(d) (The Quotient Rule) If m = 1 then D(f /g)(a) = Df (a)g(a) − f (a)D(a) /g 2 (a).

25
4.7 Theorem: (The Chain Rule) If f : U ⊆ Rn → Rm is differentiable at a, and
g : V ⊆ Rm → Rl is differentiable at f (a) then h(x) = g(f (x)) is differentiable at a and
h0 (a) = g 0 (f (a))f 0 (a).
4.8 Theorem: (The Inverse Function Theorem) If f : U ⊆ Rn → Rm is C 1 in U and
Df (a) is invertible, then we can restrict the domain of f
to some open set V ⊆ U with
a ∈ V so that f is invertible, g = f −1 is C 1 , and Dg f (x) = Df (x)−1 .
 
x1 (t)
4.9 Example: For a differentiable map f : U ⊆ R → Rn given by f (t) =  ... , we
 
have xn (t)
x1 0 (t)
 

Df (t) = f 0 (t) =  ...  .

 

xn 0 (t)
The vector Df (a) = f 0 (a) is the tangent vector to the curve x = f (t) at the point f (a).
In particular, for a differentiable map f : U ⊆ C → C given by f (t) = z(t) = x(t) + i y(t),
we have
Df (t) = z 0 (t) = x0 (t) + i y 0 (t) .
4.10 Example: For a differentiable map f : U ⊆ Rn → R we have
 
∂f ∂f ∂f
Df (x) = ∂x1 (x) ∂x2 (x) . . . ∂xn (x) .

We define the gradient of f at a to be ∇f = Df (x)T . Given a point a ∈ U and a vector

v ∈ Rn , we define the directional derivative Dv f (a) of f at a with respect to v as follows.
Choose any curve α : R → U with α(0) = a and α0 (0) = v, and set β(t) = f (α(t)). By the
.
chain rule, we have β 0 (t) = Df (α(t))α0 (t) and so β 0 (0) = Df (a)v = ∇f (a) v. We define
Dv f (a) = β 0 (0) = Df (a)v = ∇f (a) v . .
Notice that the gradient ∇f (a) is perpendicular to the level set f (x) = f (a). To see this,
choose any curve x(t) with x(0) = a and with f (x(t)) = f (a) (so that x(t) lies in the
level set). Then by the chain rule we have Df (x(t))x0 (t) = 0, and setting t = 0 gives
.
Df (a)x0 (0) = 0 or equivalently ∇f (a) x0 (a) = 0. Thus ∇f (a) is perpendicular to x0 (0).
4.11 Example: Given a differentiable map f : U ⊆ Rn → Rm , notice that the ith row of
the matrix Df (a) is equal to Dfi (a) = ∇fi (a)T , where fi is the ith component of f . So
the ith row is perpendicular to the level set fi (x) = fi (a).
4.12 Example: For a differentiable map
f : U ⊆ Rn → Rm , we denote the j th column
∂f
of the matrix Df (a) by fxj (a) or by ∂xj
, so we have
 ∂f1 
∂xj
∂f
fxj (a) = (a) =  ...  .
 
∂xj ∂f m
∂xj

Notice that this is equal to the tangent vector to the curve β(t) = f (a + t ej ), where ej is
0
the j th standard

basis vector; indeed if α(t) = a 0+ t ej so α(0) =

a0 and α0 (0) = ej , and if
β(t) = f α(t) , then by the Chain Rule we have β (t) = Df α(t) α (t) so β (0) = Df (a) ej ,
which is the j th column of Df (a).

26
4.13 Example: For a differentiable map f : U ⊆ C → C given by w(z) = u(z) + i v(z)
with z = x + i y, we have  
ux (a) uy (a)
Df (a) = .
vx (a) vy (a)
   
ux uy
The columns fx = and fy = are the tangent vectors to the curves f (a+t) and
vx vy
f (a+i t) respectively, and the rows Du = ( ux uy ) and Dv = ( vx vy ) are perpendicular
to the level curves u = u(a) and v = v(b) respectively.
4.14 Example: Let f be the polar coordinates map (x, y) = f (r, θ) = (r cos θ, r sin θ).
Then    
xr xθ cos θ −r sin θ
Df (r, θ) = = .
yr yθ sin θ r cos θ
√
√

π π π 3/2 √ −1
At (r, θ) = (2, 6 ), we have (x, y) = f (2, 6 ) = ( 3, 1) and Df (2, 6 ) = .
1/2 3
π π π
Below on the left, is a picture showing the images of the lines r  6, 3,
=√0, 1, 2and θ = 0,  2
3/2 −1
(the images are circles and rays), and the tangent vectors fr = and fθ = √
−1/2 3
√
are shown at the point √ (x, y) = ( 3,√ 1). On the right, there is a picture showing the
level curves x = 0, 1, 3 and y = 0, √ 1, 3 (they are multiples√of r = sec θ and r = csc θ),
and the gradient vectors Dx = 2 −1 and Dy = ( 12


3 3 ) are shown at the point
(r, θ) = (2, π6 ) draw This map f is not 1 : 1 so it does not have an inverse, but since the
matrix Df (2, π6 ) is invertible, we know that we can make f invertible by restricting its
domain. If g = f −1 near the point (r, θ) = (2, π3 ), then we have
√ −1  √
√

3/2 −1
√ 1 3 √1
Dg( 3, 1) = = .
1/2 3 2 −1/2 3/2

This can also be verified by finding an explicit formula for g,pfor example if we restrict the
π π −1


domain of f to r > 0, − 2 < θ < 2 then (r, θ) = g(x, y) = 2 2
x + y , tan (y/x) .

4.15 Note: We now wish to interpret the real derivative matrix Df of a map f : U ⊆
C → C in terms of complex numbers.

27
 2 × 2 matrix
4.16 Note: A real   A corresponds to two complex numbers in the following
a b
two ways. Let A = , and write z = x + i y with x, y ∈ R. Then we have
c d
     
x a b
Az = A = x+ y
y c d
= (a + i c)x + (b + i d)y
.
z+z z−z
= (a + ic) + (b + i d)
2 2i
= 2 (a + d) + i (c − b) z = 21 (a − d) + i (c + b) z
1

 
x
Thus we have A = px + qy = uz + vz where p and q are the columns of A, that is
y
u and v are given by u = 12 (a + d) + i (c − b) = 21 (p − i q)


p = a + i c and q = b + i d, and
and v = 12 (a − d) + i (c + b) = 21 (p + i q). Note also that p = u + v and q = i (u − v).

Conversely, given u = α + i β and v = γ + i δ, with α, β, γ, δ ∈ R, we have

  
α −β x
u z = (α + i β)(x + i y) = (αx − βy) + i (βx + αy) =
β α y
  
γ δ x
vz = (γ + i δ)(x − i y) = (γx + δy) + i (δx − γy) =
δ −γ y
   
α −β γ δ
and so u z + v z = Az where A = + .
β α δ −γ
4.17 Definition: For a map f : U ⊆ C → C given  by f (z) = u(z)+i
 v(z) with z = x+i y,
ux (a) uy (a)
wich is differentiable at a ∈ U so that Df (a) = , we define
vx (a) vy (a)

fx (a) = ux (a) + i vx (a) = fz (a) + fz (a)

fy (a) = uy (a) + i vy (a) = i (fz (a) − fz (a))
fz (a) = 21 (fx (a) − i fy (a)) = 1


2 (ux (a) + vy (a)) + i (vx (a) − uy (a))
fz (a) = 12 (fx (a) + i fy (a)) = 1


2 (ux (a) − vy (a)) + i (uy (a) − vx (a)) .

Note that if fz (a) = α + i β and fz (a) = γ + i δ with α, β, γ, δ ∈ R then

   
α −β γ δ
Df (a) = + .
β α δ −γ

When w = f (z), other notations for these include

∂f ∂w
fx = ∂x = wx = ∂x
∂f ∂w
fy = ∂y = wy = ∂y
∂f ∂w
fz = ∂z = ∂f = wz = ∂z = ∂w
∂f ∂w
fz = ∂z = ∂f = wz = ∂z = ∂w .

28
4.18 Note: For f : U ⊆ C → C with a ∈ U , then from the above note and definition, we
see that
f is differentiable at a
⇐⇒ there exists a real 2 × 2 matrix Df (a) such that
|R(z)|

lim = 0, where R(z) = f (z) − f (a) + Df (a)(z − a)
z→a |z − a|
⇐⇒ there exist two complex numbers fx (a) and fy (a) such that
|R(z)|

lim = 0, where R(z) = f (z) − f (a) + fx (a)Re (z − a) + fy (a)Im (z − a)
z→a |z − a|
⇐⇒ there exist two complex numbers fz (a) and fz (a) such that
|R(z)|

lim = 0, where R(z) = f (z) − f (a) + fz (a)(z − a) + fz (z − a) .
z→a |z − a|

∂z ∂z ∂z ∂z ∂a ∂a
4.19 Example: Show that ∂z = ∂z = 1, ∂z = ∂z = 0, and ∂z = ∂z = 0, where a ∈ C.
Solution: If f (z) = z, then
 we have f (x  + i y)= u(x, y) + i v(x, y), where u(x, y) = x and
ux uy 1 0
v(x, y) = y. So Df = = , fx = ux + i vx = 1, fy = uy + i vy = i,
vx vy 0 1
fz = 21 (f + x − i fy ) = 1 and fz = 12 (fx − i fy ) = 0.  
1 0
If f (z) = z, then we have u(x, y) = x and v(x, y) = −y. So Df = , fx = 1,
0 −1
fy = −i, fz = 0 and fz = 1.
If f (z) = a ∈ C then u(x, y) = Re (a) and v(x, y) = Im (a). So Df = 0 and hence
fx = fy = fz = fz = 0.
4.20 Theorem: Let f : U ⊆ C → C be differentiable in U . For α = x, y, z or z we have
(a) (cf )α = c fα
(b) (f ± g)α = fα ± fα
(c) (The Product Rule) (f g)α = fα g + f gα
(d) (The Quotient Rule) (f /g)α = (fα g − f gα )/g 2 , when g 6= 0
Proof: We prove the product rule, and leave the other parts as an exercise. We write
f = u+i v and g = s+i t where u, v, s and t are real-valued. Then f g = (us−vt)+i (ut+vs).
The Product Rule in Theorem 4.6 applies to the functions u, v, s and t, so we have

(f g)x = (us − vt)x + i (ut + vs)x

= (ux s + usx − vx t − vtx ) + i (ux t + utx + vx s + vsx )
= (ux + i vx )(s + i t) + (u + i v)(sx + i tx )
= fx g + f gx

Similarly, (f g)y = fy g + f gy . Then, using this result, we have

(f g)z = 12 (f g)x − i (f g)y

1


= 2 (fx g + f gx ) − i (fy g + f gy )
1 1
= 2 (fx − i fy ) g + f 2 (gx − i gy )
= fz g + f gz ,

and similarly, (f g)z = fz g + f gz .

29
4.21 Theorem: For f : U ⊆ C → C, define f : U → C by f (z) = f (z). Then we have
f z = fz and f z = fz .
 
ux uy
Proof: Write f = u+i v with u and v real-valued. Then f = u−i v so Df =
−vx −vy
and hence f x = ux − i vx = fx and f y = uy − i vy = fy . So we have

f z = 12 (f x − i f y ) = 12 (fx − i fy ) = 12 (fx + i fy ) = fz ,

and similarly, f z = fz .
4.22 Theorem: (The Chain Rule) Suppose f : U → V ⊆ C and g : V → C are both
differentiable, and let h(z) = g(f (z)). Then h is differentiable, and if we write w = f (z)
and q = g(w), then     
qz qz qw qw wz wz
= .
qz qz qw qw wz wz
∂q ∂q ∂w ∂q ∂w ∂q ∂q ∂w ∂q ∂w
Equivalently, we have = + and = + .
∂z ∂w ∂z ∂w ∂z ∂z ∂w ∂z ∂w ∂z
Proof: Write z = x + i y, f (z) = w = u + i v and g(w) = q = s + i t. Then

1


qz = 2 (sx + ty ) + i (tx − sy )
1


= 2 (su ux + sv vx + tu uy + tv vy ) + i (tu ux + tv vx − su uy − sv vy ) .

On the other hand

qw wz + qw wz = qw wz + qw wz
1
(su + tv ) + i (tu − sv ) 21 (ux + vy ) + i (vx − uy )



= 2
1
(su − tv ) + i (tu + sv ) 12 (ux − vy ) − i (vx + uy ) .



+ 2

Expanding and simplifying this last expression shows that qz = qw wz + qw wz . Similarly,

we can show that qz = qw wz + qw wz .
4.23 Example: Let f (z) = z 2 + 3zz. Find fz and fz .
Solution: We solve this using two methods. First, by Example 4.19 and Theorem 4.20, we
can calculate ∂f ∂f
∂z and ∂z using all the same rules that we use to find partial derivatives of
real functions of two real variables. We have fz = 2z + 3z and fz = 3z.
Our second solution is to express f in terms of real variables, and then use Definition
4.17. We have f (z) = f(x + i y) = (x + i y)2 + 3(x + i y)(x − i y) = (4x2 + 2y 2 ) + i (2xy),
8x 4y
and so Df = . Thus we have fx = 8x + i 2y and fy = 4y + i 2x, and so
2y 2x

fz = 21 (fx − i fy ) = 21 (8x + i 2y − i 4y + 2x) = 5x − i y = 5 z+z z−z

2 − i 2i = 2z + 3z
fz = 21 (fx + i fy ) = 21 (8x + i 2y + i 4y − 2x) = 3x + i 3y = 3z .

30
 (z + z)z
4.24 Example: Let f (z) = . Find fz (1 + i) and fz (1 + i).
2 + zz
Solution: By the Product and Quotient rules, we have
∂f (z + (z + z))(2 + zz) − (z + z)zz ∂f (3 + i)(4) − (2)(2) 2+i
= 2
, so (1 + i) = 2
=
∂z (2 + zz) ∂z (4) 4
and
∂f (z)(2 + zz) − (z + z)z 2 ∂f (1 + i)(4) − (2)(2i) 1
= 2
, so (1 + i) = 2
= .
∂z (2 + zz) ∂z (4) 4
4.25 Example: Let w = f (z) = iz + z, let q = g(w) = w2 − w, and let h(z) = g(f (z)).
Find hz (1 + 2i) and hz (1 + 2i).
Solution: We provide three solutions to this problem. Our first solution uses the Chain
Rule in theorem 4.22. We have
   
fz fz i 1
( hz hz ) = ( gw gw ) = ( 2w −1 ) = ( 2wi − 1 2w + i ) .
fz fz 1 −i
When z = 1 + 2i we have w = f (z) = i(1 + 2i) + (1 − 2i) = −1 − i and so we obtain
hz = 2wi − 1 = 2(−1 − i) i − 1 = 1 − 2i and hz = 2w + i = 2(−1 − i) + i = −2 − i.
Our second solution is to expand the composite g(f (z)) so that we can avoid using
the Chain Rule. We have h(z) = g(f (z)) = (iz + z)2 − (−iz + z) = −z 2 + 2izz + z 2 + i z − z.
Thus we have hz = −2z + 2iz − 1 so hz (1 + 2i) = −2(1 + 2i) + 2i(1 − 2i) − 1 = 1 − 2i and
we have hz = 2iz + 2z + i so hz = 2i(1 + 2i) + 2(1 − 2i) + i = −2 − i.
The third solution is to express f , g and h in terms of real variables. Write z = x + i y,
w = f (z) = u + i v and q = h(z) = s + i t. Then f (x + i y) = i(x + i y) + (x − i y) =
(x − y) + i (x − y) so u = x − y and v = x − y, and g(u + i v) = (u + i v)2 − (u − i v) =
(u2 − v 2 − u) + i (2uv − v) so s = u2 − v 2 − u and t = 2uv + v. By the Chain Rule for real
variables,
      
sx sy 2u − 1 −2v 1 −1 2u − 2v − 1 −2u + 2v + 1
= =
tx ty 2v 2u + 1 1 −1 2u + 2v + 1 −2u − 2v − 1

so hz = 21 (2u−2v−1)+(−2u−2v−1) +
2i (2u+2v+1)−(−2u+2v+1) =
(−2v−1)+i (2u)

and hz = 12 (2u−2v−1)−(−2u−2v−1) + 2i (2u+2v+1)+(−2u+2v+1) = 2u+i (2v+1).

When z = 1 + 2i, we have w = f (z) = i(1 + 2i) + (1 − 2i) = −1 − i, so u = v = −1 and
hence hz (1 + 2i) = (−2v − 1) + i (2u) = 1 − 2i and hz (1 + 2i) = 2u + i (2v + 1) = −2 − i.
4.26 Definition: Let f : U ⊆ C → C. We define
f (z) − f (a)
f 0 (a) = lim
z→a z−a
provided that the limit exists, and in this case we say that f is holomorphic at z = a
and thet f 0 (a) is the derivative of f at a. Equivalently, we say that f is holomorphic at
z = a if there exists a complex number f 0 (a) such that
|S(z)|
lim = 0,
z→a|z − a|
where S(z) = f (z) − f (a) + f 0 (a)(z − a) . We say that f is holomorphic in U if it is


df
holomorphic at every point in U . When w = f (z) we also write f 0 = dz = w0 = dw
dz .

31
4.27 Definition: For f : U ⊆ C → C we define

f (z) − f (a)
f × (a) = lim
z→a z−a
provided the limit exists, and if so we say that f is conjugate-holomorphic at z = a.
Equivalently, f is conjugate-holomorphic at a if there exists a complex number f × (a) such
|T (z)|
= 0 where T (z) = f (z) − f (a) + f × (a)(z − a) .


that lim
z→a |z − a|

4.28 Theorem: Let f : U ⊆ C → C and let a ∈ U .

(a) f is holomorphic at a
⇐⇒ f is differentiable at a and fz (a) = 0.
⇐⇒ f is differentiable at a and ux (a) = vy (a) and uy (a) = −vx(a) 
α −β
⇐⇒ f is differentiable at a and Df (a) is of the form Df (a) = .
β α
In this case we have f 0 (a) = fz (a) = ux + i vx = α + i β.
(b) f is conjugate-holomorphic at a
⇐⇒ f is differentiable at a and fz (a) = 0.
⇐⇒ f is differentiable at a and ux (a) = −vy (a) and uy (a) = vx(a) 
γ δ
⇐⇒ f is differentiable at a and Df (a) is of the form Df (a) = .
δ −γ
In this case we have f × (a) = fz (a) = ux + i vx = γ + i δ.
Proof: This follows immediately from Definition 4.17, Note 4.18 and Definitions 4.26, 4.27.
4.29 Definition: The two differential equations ux = vy and uy = −vx are called the
Cauchy-Riemann equations. Note that if f : U ⊆ C → C is C 1 in U , then it is
differentiable in U , and if f also satisfies the Cauchy-Riemann equations in U , then it is
holomorphic in U .
4.30 Example: Let f (z) = z 2 + 2|z|2 . Determine where f is holomorphic and where it is
conjugate-holomorphic.
Solution: We have f (z) = z 2 + 2zz, so fz = 2z + 2z = 4Re (z), and fz = 2z. Thus
f is conjugate-holomorphic when fz = 4Re (z) = 0, that is along the y-axis, and f is
holomorphic when fz = z = 0, that is at the origin.
4.31 Theorem: If f : U ⊆ C → C is holomorphic (or conjugate-holomorphic) at a then
f is continuous at a.
 
f (z) − f (a)
(z − a) = f 0 (a) · 0 = 0.


Proof: We have lim f (z) − f (a) = lim
z→a z→a z−a
4.32 Theorem: If f, g : U ⊆ C → C are both be holomorphic at a, then
(a) (cf )0 (a) = c f 0 (a)
(b) (f ± g)0 (a) = f 0 (a) ± g 0 (a)
(c) (f g)0 (a) = f 0 (a)g(a) + f (a)g 0 (a)
 0
f f 0 (a)g(a) − f (a)g 0 (a)
(d) (a) = , provided g(a) 6= 0.
g g 2 (a)
Similar results hold when f and g are both conjugate-holomorphic.
Proof: This follows immediately from Theorem 4.20.

32
4.33 Theorem: (The Chain Rule) Let f, h : U ⊆ C → V ⊆ C and let g, k : V → C with
f and g holomorphic and h and k conjugate-holomorphic. Then
(a) g ◦ f is holomorphic with (g ◦ f )0 (z) = g 0 (f (z))f 0 (z).
(b) k ◦ f is conjugate-holomorphic with (h ◦ f )× (z) = h× (f (z))f 0 (z).
(c) g ◦ h is conjugate-homorphic with (f ◦ h)× (z) = f 0 (h(z))h× (z).
(d) k ◦ h is holomorphic with (k ◦ h)0 (z) = k × (h(z))h× (z).

Proof: These all follows from the Chain Rule in Theorem 4.22. We prove part (a). Write
∂q
w = f (z) and q = g(w). Since f and g are holomorphic, we have ∂w∂z = 0 and ∂w = 0. So
∂q ∂q ∂w ∂q ∂w
by the Chain Rule in Theorem 4.22 , we have ∂z = ∂w ∂z + ∂w ∂z = 0, which shows that
∂q ∂q ∂w ∂q ∂w ∂q ∂w
g ◦ f is holomorphic, and ∂z = ∂w ∂z + ∂w ∂z = ∂w ∂z .

4.34 Theorem: (The Inverse Function Theorem) If f : U ⊆ C → C is holomorphic in

U and f 0 (a) 6= 0 then we can make f invertible by restricting its domain, and then the
inverse function g = f −1 will be holomorphic near f (a) with g 0 (f (z)) = 1/f 0 (z). A similar
result holds when f is conjugate-holomorphic.

Proof: We give a proof which uses the Inverse Function for real functions, under the
additional assumption that f 0 (z) is continuous in U (we shall prove later that when f is
holomorphic in U , its derivative is also holomorphic, and hence continuous). Suppose that
f is holomorphic in U with f 0 (z)= α(z) +i β(z), where α and β are continuous, and that
α −β
f 0 (a) 6= 0. Then we have Df = , and ux = vy = α and uy = −vx = β. Since
β α
α and β are continuous in U , f is C 1 in U . Also, since f 0 (a) = α(a) + i β(a) 6= 0 we have
Df (a) = α(a)2 + β(a)2 6= 0, so Df (a) is invertible. By the Inverse Function Theorem 4.8,
we can restrict the domain of f so that it becomes invertible and has a C 1 inverse g with
Dg(f (z)) = Df (z)−1 . Note that

α(z) β(z)
 −1 !

α(z) −β(z) α2 (z)+β 2 (z) α2 (z)+β 2 (z)
Dg f (z) = = −β(z) α(z) .
β(z) α(z)
α2 (z)+β 2 (z) α2 (z)+β 2 (z)

Since g is C 1 in U and satisfies the Cauchy-Riemann Equations in U , it is holomorphic in

U , and we have

α(z) −β(z) 1
g 0 f (z) =


+ i = .
α2 (z) + β 2 (z) α2 (z) + β 2 (z) f 0 (z)

4.35 Theorem: The maps z n , n ∈ Z, the exponential map ez , the trigonometric fuctions
and the hyperbolic functions are all holomorphic in their domains. Also, any continuous
branch of the logarithm log z, (with an open domain) is holomorphic. We have
(a) (z n )0 = n z n−1 , where n ∈ Z.
(b) (ez )0 = ez .
(c) (sin z)0 = cos z, (cos z)0 = − sin z, (tan z)0 = sec2 z.
(d) (sinh z)0 = cosh z, (cosh z)0 = sinh z, (tanh z)0 = sech2 z.
1
(e) (log z)0 = for any branch of log z.
z

33
Proof: For Part (a), let f (z) = z n , 0 < n ∈ Z. Then we have

wn − z n
f 0 (z) = lim = lim wn−1 + wn−2 z + · · · + w z n−2 + z n−1 = n z n−1 .


w→z w−z w→z

1
For f (z) = z n with n < 0, say n = −m, we have f (z) = zm , so by the Quotient Rule

−m z m−1
f 0 (z) = = −m z −m−1 = n z n−1 .
z 2m

For Part (b), let f (z) = ez and write z = x + i y and f (z) = u(z) + i v(z). Then
   
x+i y x x ux uy ex cos y −ex sin y
f (z) = e = e cos y + i e sin y , Df = = ,
vx vy ex sin y ex cos y

and we see that f is holomorphic in C with f 0 (z) = ex cos y + i ex sin y = ez .

For part (c), we only derive the formula for the derivative of sin z, but we do this in
two ways. One way is to let f (z) = sin z and write z = x + i y and f (z) = u(z) + i v(z).
Then we have
   
ux uy cos x cosh y sin x sinh y
f (z) = sin x cosh y + i cos x sinh y , Df = =
vx vy − sin x sinh y cos x cosh y

and so f is holomorphic in C and f 0 (z) = cos x cosh y − i sin x sinh y = cos(z).

Another way is to apply Part (b) and the differentiation rules in Theorem 4.32 b) to
the definition of sin z. Indeed
1 iz 1 1
(sin z)0 = (e − e−iz )0 = (ieiz + ie−iz ) = (eiz + e−iz ) = cos z .
2i 2i 2

We leave the other formulas in Parts (c) and (d) as an exercise.

For Part (e), let f (z) = log z and write z = x + pi y and f (z) = u(z) + i v(z). Then
1
since log(z) = ln |z| + i θ(z), we have u(x + i y) = ln x + y = 2 ln(x2 + y 2 ) and
2 2

 y
 tan−1 + 2π k , if x > 0


 x
 x
cos−1


 p + 2π k , if y > 0
x + y2
2


v(x + i y) = θ(x + i y) = y

 tan−1 + π + 2π k , if x < 0


 x

 x
 − cos−1 + 2π k , if y < 0 .

 p
x2 + y 2

Verify that using any one of the four formulas for v(x + i y) = θ(x + i y) gives
   x y 
ux uy x2 +y 2 x2 +y 2
Df = = −y x
vx vy x2 +y 2 x2 +y 2

y
and so f is holomorphic with f 0 (z) = x
x2 +y 2 − i x2 +y 1
2 = z.

34
4.36 Example: The two above theorems show that elementary complex functions can be
differentiated much like the real elementary functions. For example, let f (z) = (z 3 esin z )5 ,
then f 0 (z) = 5(z 3 esin z )4 (3z 2 esin z + z 3 esin z cos z).
4.37 Example: Let f (z) = z 2 − 2z + 3. Then f 0 (z) = 2z − 2 and we have f (2) = 3 and
f 0 (2) = 2. Since f 0 (2) 6= 0, we can restrict the domain of f so that it is invertible. Let g
be the inverse function. Find g 0 (3).
1 1
Solution: By the Inverse Function Theorem, we have g 0 (3) = = .
f 0 (2) 2
4.38 Example: Find a formula for the derivative of one branch of z w , where w ∈ C.
Solution: Let z w = exp(w log z), where log z is a branch of the logarithm. Then

w w zw
(z w )0 = exp(w log z) = .
z z

Notice that this is similar to the familiar formula (z w ) = w z w−1 ; the familiar formula has
the disadvantage that it does not specify which branch of z w−1 we should use.


4.39 Example: Let f (z) = sin z 2 + (1 + i)z . Find fz and fz .
Solution: We have f (z) = w(v(u(z))), where u(z) = z 2 + (1 + i)z, v(u) = sin u and
w(v) = v. Note that uz = 2z, uz = (1 + i), vu = cos u, vu = 0, wv = 0 and wv = 1. By
the Chain Rule, vz = vu uz + vu uz = 2z cos u and also vz = vu uz + vu uz = (1 + i) cos u.
Using the Chain Rule again, we have wz = wv vz + wv v z = v z = vz = (1 − i)cos u and also


wz = wv vz + wv v z = v z = vz = 2z cos u. Thus fz = (1 − i)cos u = (1 − i)cos z 2 + (1 + i)z


and fz = 2z cos u = 2z cos z 2 + (1 + i)z .
An alternate solution is to note that for z = x + i y we have

ez = ex−i y = ex (cos y − i sin y) = ez ,

z 2 +(1−i)z


and so from the definition of sin z we also
have sin(z) = sin z. Thus f (z) = sin
and so fz (z) = (1 − i) cos z 2 + (1 − i)z and fz (z) = 2z cos z 2 + (1 − i)z .

35
 Chapter 5. Conformal Maps

5.1 Note: Later on we shall see that every holomorphic function is C ∞ , which means that
all partial derivatives of all orders exist (and are continuous). For this chapter we shall
assume that all functions are C 2 , which means that all the second order partial derivatives
of f (namely uxx , uxy , uyx , uyy , vxx , vxy , vyx and vyy ) exist and are continuous. We shall
also use the fact that for C 2 functions, we always have uxy = uyx and vxy = vyx .
5.2 Definition: A map f : U ⊆ Rn → Rn is said to preserve orientation at x = a
when |Df (a)| > 0, and it is said to reverse orientation at a if |Df (a)| < 0.
5.3 Note: Let f : U ⊆ C → C. If f is holomorphic  at z = a and f 0 (a) 6= 0 then f

α −β
preserves orientation at a, since |Df (a)| = det = α2 + β 2 > 0. On the other
β α
hand, if f is conjugate-holomorphic
  at a with f × (a) 6= 0 then f reverses orientation at a
γ δ
since |Df (a)| = det = −(γ 2 + δ 2 ) < 0.
δ −γ
5.4 Definition: A map f : U ⊆ Rn → Rn is called an isometry when it preserves
distance, that is if |f (x) − f (y)| = |x − y| for all x, y ∈ Rn . Using some linear algebra, one
can show that f is an isometry if and only if f is of the form f (x) = Ax + b for some vector
b ∈ Rn and some orthogonal n × n matrix A (A is orthogonal means that AT A = I).
5.5 Note: Since
  the 2 × 2 orthogonal
 matrices are the matrices either of the form
cos θ − sin θ cos θ sin θ
or of the form , we see that the isometries in R2 are
sin θ cos θ sin θ − cos θ
the maps f which are either of the form f (z) = az + b or of the form f (z) = az + b for
some a, b ∈ C with |a| = 1.
5.6 Definition: A map f : U ⊆ Rn → Rn is called a similarity of scaling factor k > 0
when it scales distances by a factor of k, that is if |f (x) − f (y)| = k|x − y| for all x, y ∈ Rn .
It is not hard to see that f is a similarity of scaling factor k if and only if k1 f is an isometry.
5.7 Note: A map f : U ⊆ C → C is a similarity of scaling factor k > 0 if and only if f
is either of the form f (z) = az + b or of the form f (z) = az + b for some a, b ∈ C with
|a| = k.
5.8 Note: Let f : U ⊆ Rn → Rm be differentiable at a. Given a vector v ∈ Rn ,
choose a curve
α(t) with α(0) = a and α0 (0) = v. The image of α under f is the curve
β(t) = f α(t) . By the Chain Rule, we have β 0 (0) = Df (α(0))α0 (0) = Df (a)v, so we say
that f sends the vector v at a to the vector w = Df (a)v at f (a).
5.9 Definition: A map f : U ⊆ Rn → Rm is called conformal at a when it preserves
angles between curves at a, or to be precise, f is conformal at a when

(Df v) · (Df w) v·w

=
|Df v| |Df v| |v| |w|

for all vectors v, w ∈ Rn . We say f is conformal in U when it is conformal at every a ∈ U .

36
5.10 Note: Using linear algebra, one can show that f is conformal at a if and only if
Df (a)T Df (a) = k I for some k > 0. We shall show only that the latter implies the former;
suppose that Df T Df = k I with k > 0. Then

(Df v) · (Df w) = (Df v)T (Df w) = v T Df T Df w = v T k I w = k v T w = k v · w

p √ √
and in particular |Df v| = (Df v) · (Df v) = k |v|, and similarly |Df w| = √ k |w|. It
follows that f is conformal; f behaves locally like a similarity of scaling factor k.
5.11 Example: The steriographic projection from the unit sphere, with the north
pole removed, to the complex plane is the map φ which sends the point (x, y, z) on the
sphere to the point of intersection (u, v) of the line through (x, y, z) and the plane z = 0.
Find a formula for φ and φ−1 , and show that stereographic projection is conformal.

Solution: The line through (0, 0, 1) and (x, y, z) is given by (0, 0, 1) + t(x, y, z − 1), t ∈ R.
The point of intersection of this line with the plane z = 0 occurs when 1+t(z −1) = 0, that
1 x y
is when t = 1/(1−z). The point of intersection is (0, 0, 1)+ 1−z (x, y, z −1) = ( 1−z , 1−z , 0),
so we have  
x y
(u, v) = φ(x, y, z) = , .
1−z 1−z
Given (u, v) on the other hand, the line through (0, 0, 1) and (u, v) is given by α(t) =
(0, 0, 1) + t(u, v, −1) = (tu, tv, 1 − t). The point of intersection with the unit sphere occurs
when |α(t)| = 1, so we need (tu)2 + (tv)2 + (1 − t)2 = 1, that is t2 u2 + t2 v 2 − 2t + t2 = 0, or
t (tu2 + tv 2 + t − 2) = 0. The point of intersection occurs when t = u2 +v2 2 +1 , so we obtain
the formula

u2 + v 2 − 1
 
−1 2u 2v
(x, y, z) = φ (u, v) = , , .
u2 + v 2 + 1 u2 + v 2 + 1 u2 + v 2 + 1

Now let us show that φ−1 is conformal. We have

   
xu xv −u2 + v 2 + 1 −2uv
2
Dφ−1 =  yu yv  = 2  −2uv u2 − v 2 + 1 
(u + v 2 + 1)2
zu zv 2u 2v

37
and a quick calculation yields
 
−1 T −1 4 1 0
(Dφ ) (Dφ )= 2
(u + v 2 + 1)2 0 1

Note that near the point (u, v), φ−1 behaves like a similarity of scaling factor 2/(u2 +v 2 +1).
5.12 Theorem: Let f : U ⊆ C → C.
(a) f is conformal at a if and only if either f is holomorphic at a with f 0 (a) 6= 0, in which
case f preserves orientation, or f is conjugate-holomorhic at a with f × (a) 6= 0, in which
case f reverses orientation.
(b) If U is connected, then f is conformal in U if and only if either f is homorphic in U with
f 0 (z) 6= 0 for all z ∈ U , in which case f preserves orientation, or f is conjugate-holomorphic
in U with f × (z) 6= 0 for all z ∈ U , in which case f reverses orientation.
Proof: To prove part (a), note that f is conformal at a if and ony if Df (a) is a positive scalar
multiple of  Since the 2 ×
 an orthogonalmatrix.  2 orthogonal matrices are the matrices of
cos θ − sin θ cos θ sin θ
the form or , we see that f is conformal if and only if
 sin θ cos θ  sin θ
 − cos θ
α −β γ δ
Df = or Df = for some α, β or γ, δ ∈ R not both equal to zero.
β α δ −γ
Part (b) involves a subtle point: if f is conformal in U then how do we know that f
cannot be holomorphic at some points a ∈ U and conjugate-holomorphic at other points?
It is for this reason that we must assume that U is connected. Since we have assumed that
all functions in this chapter are C 2 we know that ux , uy , vx and vy are all continuous and
so |Df | = ux vy − uy vx is also continuous. At each point a ∈ U we have |Df (a)| 6= 0, so
|Df | is a continuous map from U to R∗ . Since U is connected, we know that |Df |(U ) is
also connected and lies in R∗ . This implies that either |Df (a)| > 0 for all a or |Df (a)| < 0
for all a.
5.13 Note: If f : U ⊆ C → C is holomorphic at a with f (a) = b and f 0 (a) = r ei θ ,
where r > 0, then by the definition of the (complex) derivative, for z near a we have
f (z) ∼
= f (a) + f 0 (a)(z − a) = b + reiθ (z − a). This shows that locally, f behaves like the
following similarity: translate by −a, rotate by θ, scale by r, then translate by b.
5.14 Example: Let f (z) = z 2 . Then f is holomorphic in C and f 0 (z) = 2z so f 0 (z) 6= 0
in C∗ . Hence f (z) = z 2 is conformal in C∗ and preserves orientation. Verify directly that
f preserves the oriented angle from α(t) = i + t to β(t) = i + (1 + i) t.
   
1 1
Solution: We have α(0) = β(0) = i, α0 (0) = 1 = and β 0 (0) = 1 + i = , so the
0 1
angle from α0 (0) to β 0 (0) is π4 . The images are γ(t) = f (α(t)) = (i+t)2 = (t2 −1)+i 2t (this
is the parabola u = 41 v 2 − 1) and δ(t) = f (β(t)) = (i + (1 + i)t)2 = −(1 + 2t) + i (2t + 2t2 )
(check that this is the parabola v = 21 u2 − 12 ). Note that γ(0) = δ(0)=  −1, so the two
0
parabolas intersect at −1. We have γ 0 (t) = 2t + 2i so γ 0 (0) = 2i = and we have
  2
−2
δ 0 (t) = −2 + i (4t + 2) so δ 0 (0) = −2 + 2i = . So the angle from γ 0 (0) to δ 0 (0) is π4 .
2
Notice also that α and β meet at i, and we have f (i) = −1 and f 0 (i) = 2i = 2ei π/2 .
So near z = i, f can be approximated as follows: translate by −i, rotate by π2 , scale by 2,
then translate by −1. Indeed, this is precisely what happens to the tangent vectors.

38
5.15 Definition: Let u : U ⊆ C → C. The (2 dimensional) Laplacian is the differential
operator ∇2 given by
∇2 u = uxx + uyy .
The map u is called harmonic in U when it is C 2 and satisfies Laplace’s equation

∇2 u = 0 .

5.16 Note: There are several functions from physics which satisfy Laplace’s equation.
Steady state temperature (in a homogeneous medium), electrostatic potential (in a vacc-
ume) and the velocity potential for a steady flow of fluid (irrotational and indecompressible)
al satisfy Laplace’s equation.
−1
5.17 Example: As an exercise, you should check that the map u(x, y, z) = p
x2 + y 2 + z 2
satisfies the 3 dimensional Laplace equation uxx + uyy + uzz = 0, but that the map
1
u(x, y) = − p does not satisfy the 2 dimensional Laplace equation. The first map
x2 + y 2
u represents the electic potential surrounding a point charge in R3 , but the second map u
does not represent the potential which surrounds
p a long straight wire. On the other hand,
you can check that the map u(x, y) = ln x2 + y 2 does satisfy the 2 dimensional Laplace
equation, and this map u does represent the potential surrounding a wire.
5.18 Theorem: If f (z) = u(z) + i v(z) is holomorphic (or conjugate-holomorphic) in U
then u and v are both harmonic functions. When f = u + i v is holomorphic, we say that
v is the harmonic conjugate of u.
Proof: The Cauchy-Riemann equations ux = vy and uy = −vx imply that

uxx = (ux )x = (vy )x = vyx = vxy = (vx )y = (−uy )y = −uyy

and likewise vxx = −uyx = −uxy = −vyy .

5.19 Example: Let f (z) = ez . Verify that u is harmonic, where u = Re (f ).
Solution: Since ex+i y = ex (cos y + i siny), we have u(x + i y) = ex cos y. So ux = ex cos y
and uxx = ex cos y, while uy = −ex sin y and uyy = −ex cos y = −uxx .

39
5.20 Example: Let f (z) = z 3 . Verify that u is harmonic, where u = Re (f ).
Solution: We have f (x + i y) = (x + i y)3 = (x3 − 3xy 2 ) + i (3x2 y − y 3 ) so u = x3 − 3xy 2 .
We have ux = 3x2 − 3y 2 so uxx = 6x and uy = 3x2 − 6xy and so vyy = −6x = −uxx .
5.21 Note: There is a partial converse to the above note which says that for certain
sets U , for example when U is convex, if u is harmonic in U then there exists a harmonic
function v such that the map f = u + i v is holomorphic in U . The following example
shows how to find v.
5.22 Example: Let u = 2x2 + 3xy − 2y 2 . Check that u is harmonic in C, and find a
harmonic conjugate v.
Solution: We have ux = 4x + 3y, uxx = 4, uy = 3x − 4y and uyy = −4 = −uxx , so u
is harmonic. To find v such that u + i v is holomorphic, we need to find v such that the
Cauchy-Riemann equations ux = vy and uy = −vx are satisfied. To get vy = ux = 4x + 3y
we must take v = 4x + 3y dy = 4xy + 23 y 2 + c(x). Then we have vx = 4y + c0 (x). To
R

get vx = −uy = 4y − 3x we need to have c0 (x) = −3x, so we choose c(x) = − 23 x2 . In this

way we obtain v = 4xy + 32 (y 2 − x2 ). The function f = u + i v should be holomorphic, and
indeed you can check that f (z) = 2 − 32 i z 2 .

5.23 Example: A long strip of heat conducting material is modelled by the set
S = {x + i y|0 < y < 1} .
Find the steady state temperature u(x+i y) at each point in the strip given that the bottom
edge is held at a constant temperature of a◦ and the top edge is held at b◦ . Describe the
isotherms, that is the curves of constant temperature.
Solution: We must find a map u : S → R which is continuous on S and harmonic in S
such that u(x, 0) = a and u(x, 1) = b for all x. We can take
u(x + i y) = a + (b − a)y .
It is easy to see that u is harmonic, indeed uxx = uyy = 0. Also notice that u is the
imaginary part of the holomorphic map f (z) = a i + (b − a)z. The isotherm u = c is the
horizontal line c = a + (b − a)y, or y = c−a
b−a .
5.24 Theorem: If u : U ⊆ C → R is harmonic and if f : V ⊆ C → U ⊆ C is holomorphic
then u ◦ f is harmonic.
Proof: Write x + i y = f (s + i t), u = u(x + i y), and v = u ◦ f . The chain rule gives
vs = ux xs + uy ys vt = ux xt + uy yt .
Using the chain rule and the product rule, we obtain
vss = (uxx xs + uxy ys )xs + ux xss + (uyx xs + uyy ys )ys + uy yss
vtt = (uxx xt + uxy yt )xt + ux xtt + (uyx xt + uyy yt )yt + uy ytt
Adding these, using the fact that uxy = uyx we obtain
vss +vtt = uxx (xs 2 +xy 2 )+uyy (ys 2 +yt 2 )+uxy (2xs ys +2xt yt )+ux (xss +ytt )+uy (yss +ytt ) .
Since f is holomorphic, the Cauchy-Riemann equations xs = yt and xt = −ys imply that
(ys 2 + yt 2 ) = (xs 2 + xt 2 ) and that (2xs ys + 2xt yt ) = 0 and that x and y are each harmonic
so that (xss + xtt ) = 0 and (yss + ytt ) = 0. So we are left with
vss + vtt = (uxx + uyy )(xs 2 + xy 2 ) .
Finally, since u is harmonic, we have (uxx + utt ) = 0 and hence vss + vtt = 0.

40
5.25 Note: We shall now consider problems of the following kind: given an open set
U ⊆ C, find a harmonic function u : U → R which satisfies some given condition on the
boundary ∂U ; this kind of problem is called a boundary value problem. We solved
an easy boundary value problem in example 5.23, in which the open set was the strip
S = {x + i y|0 < y < 1}. The above theorem allows us to use a solution to one boundary
value problem on a set U to obtain a solution to another problem on a set V by mapping
the set U to the set V using a holomorphic map.
5.26 Example: The upper half-plane H = {x + i y|y > 0} is a model for a large heat
conducting plate. Find the steady state temperature v(z) at each point in the plate if the
temperature along the bottom edge is held at a◦ for x > 0 and b◦ for x < 0. Also, describe
the isotherms.
Solution: Notice that we can map the strip S = {x + i y|0 < y < 1} (which appeared in
the example 5.23) to the half-plane H = {x + i y|y > 0} using the map f (z) = eπ z . The
bottom edge of S is mapped to the positive x-axis, and the top edge of S is mapped to the
negative x-axis. To map H back to S we use the inverse map g(z) = π1 log z, where log z
is the branch of the logarithm given by log z = ln |z| + iθ(z) where 0 ≤ θ(z) ≤ π.
From example 5.23, the map u(z) = Im (a i + (b − a)z) is harmonic in the strip S with
u = a when y = 0 and u = b when y = 1. To solve our problem in H, we take v = u ◦ g.
To be explicit, we take
 
b−a b−a
v(z) = Im a i + log z = a + θ(z) ,
π π
b−a c−a
where 0 ≤ θ(z) ≤ π. The isotherm u = c is the ray c = a + π θ(z) or θ(z) = b−a π.
5.27 Example: Find the steady state temperature u(z) inside a circular plate modelled
by the disc U = D(0, 1), given that the top half of the boundary is held at a◦ = 1◦ and the
bottom half is held at b◦ = 5◦ . In particular, find the temperature at the point 21 i. Also
describe the isotherm u = 2.
z+1
Solution: The map f1 (z) = maps the disc D(0, 1) to the disc D( 12 , 12 ), and it sends the
2
top half of the boundary of the first disc to the top half of the boundary of the second. The
1
map f2 (z) = maps the disc D( 21 , 12 ) to the half-plane H1 = {x + i y|x > 1}, and it maps
z
the top half of the boundary of the disc to the bottom half {1 + i y|y < 0} of the boundary
of H1 . The map f3 (z) = z −1 translates the half-plane H1 to H0 = {x +i y|x > 0}. Finally
the map f4 (z) = i z rotates H0 to the half-plane H = {x + i y|y > 0} sending the bottom
half of the boundary of H0 to the right half {x > 0} of the boundary of H. So we can use
our solution v(z) from the previous example to obtain the solution  u = v ◦ f 4 ◦ f3◦ f2 ◦ f
1.
2 2 1−z
To be explicit, we have f2 (f1 (z)) = and f3 (f2 (f1 (z))) = −1 =
  1+z 1+z 1+z
1−z
and f4 (f3 (f2 (f1 (z)))) = i , so our solution is
1+z
   
b−a 1−z 4 1−z
u(z) = a + θ i =1+ θ i ,
π 1+z π 1+z
where 0 ≤ θ i 1−z 1−z 1−z π


1+z ≤ π. Since θ(i 1+z ) = θ( 1+z ) + 2 , we have
 
4 1−z
u(z) = 3 + θ ,
π 1+z

41
where − π2 ≤ θ 1−z π 1


1+z ≤ 2. In particular, the temperature at 2 i is

1− 12 i
∼
u( 12 i) = 3 + 4 4 3−4i 4
tan−1 − 4
= 1.82◦ .



π θ 1+ 12 i
=3+ π θ 5 =3+ π 3

To find the isotherm u = 2, we recall that the corresponding isotherm v = c = 2 in

−1
example 5.26 was the ray θ(z) = c−a 2−1 π
b−a π = 5−1 π = 4 . This ray is rotated by f4 (z) = −iz
π −1 π
to the ray θ(z) = − 4 , then translated by f3 = z + 1 to the ray θ(z − 1) = − 4 , this ray is
the portion below the x-axis of the line whose nearest point to the origin is 12 (1 + i) and so
it is mapped by f2 −1 (z) = 1/z to the portion√
above the x-axis of the circle with diameter
2 1−i 2
0, 1+i = 1 − i, that is the circle |z − 2 | = 2 , and finally this arc is translated and scaled
√
by the map f1 −1 (z) = 2z − 1 to the portion√above the x-axis of the circle |z + i| = 2.
Thus the isotherm u = 2 is the arc |z +i| = 2, z ∈ D(0, 1).
−1
We also remark that θ( 1−z 1−z
1+z ) = Im log( 1+z ) = −2 Im (tanh z).

z+1
f1 (z) = 2 f2 (z) = 1/z f4 (f3 (z)) = i(z − 1)

5.28 Example: Find the steady state temperature u(z) in the plate shaped like the semi-
infinite strip U = {x + i y| − 1 < x < 1, y > 0} given that the temperature along the
bottom edge and the right edge is held at a◦ = 10◦ and the temperature along the left
edge of the boundary is held at b◦ = 40◦ . Also, find the temperature at z = i.
Solution: The map f1 (z) = π2 z widens the strip U by a factor of π2 , and then the map
f2 (z) = sin z sends the strip to the half plane H = {x + i y|y > 0}. The left edge of the
boundary of U is mapped to the portion of the real axis with x < −1. Lastly, the map
f3 (z) = z + 1 sends H to itself and it sends the portion of the real axis with x < −1
to the portion with x < 0. We can again use our solution v(z) from example 5.26 to
obtain the solution to this problem. We take u = u ◦ f3 ◦ f2 ◦ f1 . To be explicit, we have
f3 (f2 (f1 (z))) = 1 + sin π2 z and so

b−a 30
θ 1 + sin( π2 z) = 10 + θ 1 + sin( π2 z) ,



u(z) = a +
π π

where 0 ≤ θ 1 + sin( π2 z) ≤ π. In particular, we have

30
θ 1 + sin(i π2 ) = 10 + 30
θ 1 + i sinh π2 = 10 + 30
tan−1 sinh π2 ∼= 21.1◦ .




u(i) = 10 + π π π

5.29 Example: Find the steady state temperature v(z) at each point on a plate modelled
by the half-plane H = {x + i y|y > 0} given that the temperature along the boundary is
held constant at a◦ for x > 1, b◦ for −1 < x < 1 and at c◦ for x < −1.

42
Solution: We can use the fact that the sum of two harmonic maps will also be harmonic.
We use the solution from example 5.26 to get one harmonic map v1 in H with v1 = a along
the portion of the x-axis with x > 1 and v1 = b along the portion with x < 1, and we get
another harmonic map v2 in H with v2 = 0 along the portion of the x-axis with x > 1 and
v2 = c − b along the portion with x < −1. Then we add them to get v = v1 + v2 . To be
b−a c−b
explicit, v1 (z) = a + θ(z − 1) and v2 (z) = θ(z + 1) and so
π π
b−a c−b
v(z) = a + θ(z − 1) + θ(z + 1) ,
π π
where 0 ≤ θ(z − 1), θ(z + 1) ≤ π.
5.30 Example: Find the steady-state temperature u(z) in the semi-circular plate mod-
elled by U = {x + i y|x2 + y 2 < 1, y < 0} given that the temperature along the boundary
√
is held constant at a◦ = 5◦ when y = 0 and x > 0, and at b◦ = 10◦ when y = − 1 − x2
and at c◦ = 20◦ when y = 0 and x < 0. In particular, find the temperature at z = − 21 i.
Solution: The map f1 = 1/z sends U to the region V above the x-axis and outside the unit
circle V = {x + i y|x2 + y 2 > 1, y > 0}. Then f2 (z) = log z, the branch of the logarithm
with 0 ≤ θ ≤ π, maps V to the semi-infinite strip W = {x + i y|x > 0, 0 < y < π}.
We rotate the strip by 90◦ using f3 (z) = i z then shift it to the right by π2 using the
map f4 (z) = z + π2 (so that its base is centred at the origin), and then we use the map
f5 (z) = sin z to map the strip to the half-plane H = {x + i y|y > 0}. The portions of the
boundary which are to be held constant at a◦ , b◦ and c◦ are mapped to the portions of the
x-axis with x > 1, −1 < x < 1 and x < −1 respectively, so we can use our solution v(z)
from the previous example. Our solution is u = v ◦ f5 ◦ . . . ◦ f1 . To be explicit, we have
f5 (f4 (f3 (z))) = sin(i z + π2 ) = cos(i z) = cosh z, and f5 (f4 (f3 (f2 (z)))) = cosh(log z) =
elog z + e− log z z + z1 1
z +z 1 + z2
= , and so (f5 ◦ . . . ◦ f1 )(z) = = . Our solution is
2 2 2 2z
1 + z2 1 + z2
   
b−a c−b
u(z) = a + θ −1 + θ +1
π 2z π 2z
1 + z2 1 + z2
   
5 10
=5+ θ −1 + θ +1 .
π 2z π 2z
3/4
10 3/4
In particular, u(−i/2) = 5+ π5 θ 5
θ −1+i 43 + 10 3




−i −1 + π θ −i +1 = 5+ π π θ(1+i 4 =
−1 3 ∼
5 + π5 π − tan−1 43 + 10 −1 3 5 ◦


π tan 4 = 10 + π tan 4 = 11.0 .

π
f1 (z) = 1/z f2 (z) = log z f4 (f3 (z)) = iz + 2

43
5.31 Note: All of the above examples can be re-worded so that they are asking us to find
the electrostatic potential in a certain region given that the voltage along the boundary is
held constant. If u is the electrostatic potential in a region, then the electric field E is
defined by
E = −∇u .
If f is holomorphic and u = Re (f ) and v = Im (f ), then we have ∇u = ux + i uy = fz and
∇v = vx + i vy = −uy + i ux = i(ux + i uy ) = i ∇u = i fz .
5.32 Example: Find the electostatic potential and the electric field at each point inside
a long hollow metal cylinder, with unit radius, made up of two semi-cylindrical pieces
separated by thin strips of insulating material, with one piece held at a potential of 1 Volt,
and the other at 5 Volts. In particular, find the electrostatic potential and the electric field
at points along the centre of the cylinder.

Solution: The cross-section of the cylinder is modelled by the

 unit disc U = D(0, 1). As in
4 1−z
example 5.27, the electric potential is u(z) = 3 + π θ 1+z . Note that u = Re (f ), where
4 1+z −2
f (z) = 3 − π4 i log 1−z 8i


1+z . The electric field is E = −∇u = −fz = π i 1−z (1+z)2 = π(1−z 2 ) .
8
In particular, we have u(0) = 3 and E(0) = π i.
5.33 Example: Find all solutions v(z) to Laplace’s equation in C∗ such that v(r ei θ ) =
f (r) for some function f (the solution will model the electrostatic potential at each point
around a long charged rod).
Solution: The exponential function maps C onto C∗ . If v(z) is harmonic in C∗ then
u(z) = v(ez ) will be harmonic in C. If v is of the form v(r ei θ ) = f (r) then we have
u(x + i y) = v(ex eiy ) = f (ex ). Since u is independent of y, Laplace’s equation becomes
uxx = 0, and the only solutions are of the form u(x + i y) = ax + b = Re (az + b) for some
a, b ∈ R. Thus we have v(z) = u(log z) = Re (a log z + b) = a ln |z| + b.
5.34 Example: Find the electrostatic potential v(z) and the electric field E(z) at each
point inside a long grounded cylinder, of unit radius, which encloses a charged wire centred
inside the cylinder.
Solution: We look for a harmonic map v(z) defined on the punctured disc U = D∗ (0, 1)
with v(z) = 0 when |z| = 1. From the previous example, we can take v(z) = a ln |z|. The
constant a depends on the charge per unit length and on the choice of units. In fact

v(z) = −2kq ln |z| ,

N m2
where q is the charge on the rod in coulombs per meter and k ∼ = 9 × 109 C2 . Since
v = Re (f ), where f (z) = −2kq log(z), we have E(z) = −fz = 2kq/z.

44
5.35 Example: A charged wire at x = 0, y = 1 lies inside the region in space given by
y > 0, and the boundary of the region is grounded. Find the potential u(z) at each point
in the region and around the wire.
Solution: Let U be the punctured half-plane U = {z|Im z > 0, z 6= i}. The map f1 (z) =
z + i maps U to the set V = {z|Im z > 1, z 6= 2i}, the map f2 (z) = 1/z maps V to the
punctured disc W = D∗ (− 21 i, 21 ), and the map f3 (z) = 2z + i maps W to the punctured
disc D∗ (0, 1). So our solution is u = v ◦ f3 ◦ f2 ◦ f1 where v(z) = −2kq ln |z| is the solution
from the previous example. Check that
z−i
u(z) = −2kq ln .
z+i

f1 (z) = z + i f2 (z) = 1/z f3 (z) = 2z + i

5.36 Note: The velocity field F of a flow (of perfect fluid) and the velocity potential v
are related (like the electric field and electric potential) by

F = −∇v .

5.37 Example: Find the velocity potential v(z) of the constant flow with velocity field
F (x + i y) = c in the upper half plane H = {x + i y|y > 0}.
Solution: We must have F = −∇v so we need c = −(vx + i vy ), that is vx = −c and vy = 0.
Since vy = 0, v is independent of y, and since vx = −c we have

v = −c x = Re (−c z) .

We could add a constant to this solution.

5.38 Example: Use the previous example to find the velocity potential for the region
U = {x + i y|x2 + y 2 > 1, y > 0} given that as z → ∞ the flow tends to the constant flow
F = k. Also, determine the speed of the flow near z = i, that is, at the top of the bump.
Solution: As in example 5.32, the map f (z) = cosh(log z) = 21 (z + 1/z) sends the region
U to the upper half-plane H = {x + i y|y > 0}. We use the potential v from the previous
example, and we take u(z) = v(f (z)) = Re g(z), where g(z) = − 2c (z + 1/z). The velocity
field is F = −gz = 2c (1 − 1/z 2 ). As z → ∞ we have F (z) → c/2 so we must take c = 2k.
Thus our solution is

v(z) = Re − k(z + z −1 ) , F (z) = k(1 − 1/z 2 ) .

We have F (i) = 2k, so the velocity at the top of the bump is twice the velocity at ∞.

45
 Chapter 6. Integration

6.1 Definition: Let I be an interval in R (I could be open, closed or half-open). Say

I = ha, b where a, b ∈ R ∪ {±∞} with a < b and where h and i denote either open or
closed brackets depending on whether a and b are open or closed endpoints of I. A function
f : I → C is called piecewise continuous when there exist points si ∈ R ∪ {±∞} with
a = s0 < s1 < · · · < sn = b and there exist continuous functions

g1 : hs0 , s1 ] → C , gi : [si−1 si ] → C for 1 < i < n , and gn : [sn−1 , sn i → C

such that f (t) = g(t) for all t ∈ (si−1 , si ). A function f : I → C is called piecewise C 1
when f is differentiable and f 0 is piecewise continuous on I.
6.2 Definition: Let f : I ⊆ R → C be piecewise continuous, where I is an interval, let
u(t) = Re f (t) and v(t) = Im f (t), and let t1 , t2 ∈ I. We define the integral of f from t1
to t2 to be Z t2 Z t2 Z t2 Z t2
f= f (t) dt := u(t) dt + i v(t) dt .
t1 t1 t1 t1

6.3 Remark: It is also possible to define the definite integral as a limit of Riemann sums,
but we shall not do this here.
Z θ
it
6.4 Example: Let f (t) = e for t ∈ R. For θ ∈ R, find α(t) dt.
0
Solution: We have
Z θ Z θ Z θ Z θ


f (t) dt = cos t + i sin t dt = cos t dt + i sin t dt
0 0 0 0
h iθ h iθ
= sin t + i − cos t = sin θ + i (1 − cos θ) = i − i ei θ .
0 0

Note that as θ varies, this traces out the circle centered at i of radius 1.
6.5 Theorem: (Linearity) If f, g : I ⊆ R → C are piecewise continuous, where I is an
interval with t1 , t2 ∈ I, and c ∈ C then
Z t2 Z t2 Z t2 Z t2 Z t2
cf = c f and (f + g) = f+ g.
t1 t1 t1 t1 t1

Proof: This follows from Linearity for real-valued functions.

6.6 Theorem: (Decomposition) If f : I ⊆ R → C is piecewise continuous, where I is an
interval with t1 , t2 , t3 , · · · , tn ∈ I then
Z tn Z t2 Z t3 Z tn
f= f+ f + ··· + f.
t1 t1 t2 tn−1

Proof: This follows from the Decomposition Theorem for real-valued functions.

46
6.7 Theorem: (Change of Parameter) Let s : I ⊆ R → R be piecewise C 1 (this means
that s = s(t) is continuous and s0 (t) is piecewise continuous) where I is an interval, let
J = s(I), which is also an interval, let f : J ⊆ R → R be piecewise continuous, and let
t1 , t2 ∈ I. Then
Z t2 Z s(t2 )
0
f (s(t))s (t) dt = f (s) ds .
t1 s(t1 )

Proof: This follows from the Change of Variables Theorem for real-valued functions.
6.8 Theorem: (Estimation) Let f : [t1 , t2 ] ⊂ R → C be piecewise continuous. Then
Z t2 Z t2
f (t) dt ≤ |f (t)| dt .
t1 t1
Z t2 Z t2 Z t2
Proof: Write f (t) dt in polar coordinates as f (t) dt = f (t) dt ei θ . Then
t1 t1 t1
Z t2 Z t2 Z t2 Z t2
−i θ −i θ
e−iθf (t) dt ,


f (t) dt = e f (t) dt = e f (t) dt = Re
t1 t1 t1 t1

where the last equality holds since for r ≥ 0 we have r = |Re (r)|. We can then use the
Estimation Theorem for real-valued functions to obtain
Z t2 Z t2 Z t2 Z t2
−iθ −iθ iθ




Re e f (t) dt = Re e f (t) dt ≤ Re e f (t) dt ≤ f (t) dt ,
t1 t1 t1 t1

since |Re (ei θf (t))| ≤ |eiθf (t)| = |f (t)|.

6.9 Theorem: (The Fundamental Theorem of Calculus) Let f, g : I ⊆ R → C where I is
an interval with t1 , t2 ∈ I, f is piecewise continuous, g is differentiable, and g 0 = f . Then
Z t2
f (t) dt = g(t2 ) − g(t1 ) .
t1

Proof: Let u(t) = Re g(t) and v(t) = Im g(t) so that g(t) = u(t) + i v(t) and f (t) = g 0 (t) =
u0 (t)+i v 0 (t). Then, using the Fundamental Theorem of Calculus for real-valued functions,
we have
Z t2 Z t2 Z t2 Z t2
0 0 0
f (t) dt = u (t) + i v (t) dt = u (t) dt + i v 0 (t) dt
t1 t1 t1 t1



= u(t2 ) − u(t1 ) + i v(t2 ) − v(t1 ) = g(t2 ) − g(t1 ) .

6.10 Example: With the help of the Fundamental Theorem of Calculus, we can find the
integral of example 6.4 as follows
Z θ h iθ
ei t dt = − i ei t = −i ei θ + i .
0 t=0

6.11 Definition: Let a, b ∈ U ⊆ C. A path (or curve) from a to b in U is a piecewise

C 1 function α : [t1 , t2 ] ⊆ R → U with α(t1 ) = a and α(t2 ) = b. In the case that a = b, we
say that α is a loop at a in U .

47
6.12 Example: For a, b ∈ C, the path
α(t) = a + (b − a) t , for 0 ≤ t ≤ 1
traces the line segment from a to b.
6.13 Example: For a ∈ C and 0 < r ∈ R, the loop
α(t) = a + r ei t , for 0 ≤ t ≤ 2π
traces the circle of radius r centered at a.
6.14 Definition: The arclength of a path α : [t1 , t2 ] → U ⊆ C is given by
Z t2
L(α) = |α0 (t)| dt .
t1

We remark that the arclength L(α) exists and is finite, since α0 is piecewise continuous.
6.15 Example: Find the arclength of the path α(t) = t2 + i t3 , 0 ≤ t ≤ 1.
Solution: We have
Z 1 Z 1 Z 1 p
0 2
L(α) = |α (t)| dt = |2t + i 3t | dt = 4t2 + 9t4 dt
0 0 0
1 13
1 √
Z p Z
= t 4 + 9t2 dt = 18 u du
0 4
13
√ √

1 1


= 27 u u = 27 13 13 − 8 .
4

6.16 Definition: Given a path α : [t1 , t2 ] → U ⊆ C and a continuous function f : U → C

we define the path integral of f along α to be
Z Z Z t2
f= f (z) dz := f (α(t))α0 (t) dt .
α α t1

6.17 Remark: It is possible to define the path integral as a limit of Riemann sums, but
we shall not do this here.
6.18 Remark: The complex path integral is related to real path integrals in the following
way. Write z = α(t) = x(t) + i y(t) and f (z) = u(z) + i v(z) with x, y, u, v ∈ R. Then
Z Z t2 Z t2
0
u(α(t)) + i v(α(t)) x0 (t) + i y 0 (t) dt



f (z) dz = f (α(t))α (t) dt =
α t1 t1
Zt2 Z t2
0 0
= u(α(t))x (t) − v(α(t))y (t) dt + i v(α(t))x0 (t) + u(α(t))y 0 (t) dt
t1 t1
Z Z
= (u dx − v dy) + i (v dx + u dy) .
α α

This can easily be remembered by defining dz = dx + i dy and then writing

Z Z Z Z
f (z) dz = (u + i v)(dx + i dy) = (u dx − v dy) + i (v dx + u dy) .
α α α α
Z
In a similar way we could define the path integral f (z) dz, where dz = dx − i dy
α

48
 Z
6.19 Example: Find c where a, b, c ∈ C and α(t) = a + (b − a) t for 0 ≤ t ≤ 1.
α

Solution: We have
Z Z 1 Z 1 h i1
0
c dz = c α (t) dt = c (b − a) dt = c (b − a) t = c (b − a) .
α 0 0 t=0

Z
6.20 Example: Find z 2 dz where α(t) = 2 + (−1 + i) t for 0 ≤ t ≤ 1.
α

Solution: We have
Z Z 1
2
Z 1
2
2
z dz = α(t) α0 (t) dt = 2 + (−1 + i) t (−1 + i) dt
α 0 0
h
1

3 i 1 1
(1 + i)3 − 23 = 13 (−2 + 2i − 8) = − 10 2


= 3 2 + (−1 + i) t = 3 3 + 3 i.
0

6.21 Theorem: (Linearity) Let α be a path in U ⊆ C let f, g : U → C be continuous,

and let c ∈ C. Then
Z Z Z Z Z
cf = c f and (f + g) = f+ g.
α α α α α

Proof: This follows from the Linearity Theorem 6.5.

6.22 Theorem: (Decomposition) Let α : [t1 , t2 ] → U ⊆ C be a path, let n be a positive
integer, let t1 = s0 < s1 < s2 < · · · < sn = t1 , for each i = 1, 2, · · · , n, let αi be the
restriction of α to the interval [si−1 , si ], and let f : U → C be continuous. Then
Z Z Z Z
f= f+ f + ··· + f.
α α1 α2 αn

Proof: This follows from the Decomposition Theorem 6.6.

6.23 Definition: For a path α : [t1 , t2 ] → U ⊂ C, we define the inverse path α−1 by

α−1 (t) = α(t1 + t2 − t) , for t1 ≤ t ≤ t2 .

Note that α−1 has the same image as α, but it traces this image in the opposite direction
with α−1 (t1 ) = α(t2 ) and α−1 (t2 ) = α(t1 ).
6.24 Theorem: (Change of Direction) Let α : [t1 , t2 ] → U ⊆ C be a path, and let
f : U → C be continuous. Then
Z Z
f =− f.
α−1 α

Proof: This theorem is a special case of the following more general theorem.

49
6.25 Theorem: (Change of Parameter) Let s : [t1 , t2 ] ⊂ R → [s1 , s2 ] ⊂ R be invertible
and piecewise C 1 . Note that s must be monotonic, and if s is increasing then we have
s1 = s(t1 ) and s2 = s(t2 ), while if s is decreasing then we have s1 = s(t2 ) and s2 = s(t1 ).
Let α : [s1 , s2 ] → U ⊆ C be a path, let β(t) = α(s(t)) for t1 ≤ t ≤ t2 , and let f : U → C
be continuous. Then Z Z
f (z) dz = ± f (z) dz ,
β α

where we use + when s is increasing and we use − when s is decreasing.

Proof: Since β(t) = α(s(t)), the Chain Rule gives β 0 (t) = α0 (s(t))s0 (t), and so
Z Z t2 Z t2 Z s(t2 )
0 0 0
f= f (β(t))β (t) dt = f (α(s(t)))α (s(t))s (t) dt = f (α(s))α0 (s) ds
β t1 t1 s(t1 )

by the Change
Z s2 of Parameter TheoremZ 6.7. When s is increasing, the integral on the right
is equal to f (α(s))α0 (s) ds = f , but when s is decreasing, the integral on the right
s1
Z s1 αZ s2 Z
0 0
is equal to f (α(s))α (s) ds = − f (α(s))α (s) ds = − f .
s2 s1 α

6.26 Remark: We use the Decomposition Theorem, the Change of Direction Theorem
and the Change of Parameter Theorem implicitly when we join two or more paths together
to form a single path or loop. For example, let a, b ∈ U ⊆ C, let f : U → C be continuous,
let α and β be two paths from a to b in U , and let γ be a loop which follows α then β −1 .
Then no matter how we choose to parametrize γ, we have
Z Z Z
f= f− f.
γ α β

This fact makes it unnecessary to find an explicit formula for γ(t), such as the following:
if α : [t1 , t2 ] → U and β : [t3 , t4 ] → U , then one specific parametrization for a loop γ which
follows α then β −1 is given by
(
α(t) , t 1 ≤ t ≤ t2
γ(t) =
β(t2 + t4 − t) , t2 ≤ t ≤ t2 + t4 − t3 .

6.27 Theorem: (Estimation) Let α : [t1 , t2 ] → U ⊆ C be a path, let L = L(α), and let
M = max f (z) . Then
z=α(t)

Z Z t2
f (z) dz ≤ f (α(t))α0 (t) dt ≤ M L .
α t1

Proof: This follows from the Estimation Theorem 6.8.

6.28 Definition: Let f, g : U ⊆ C → C. If g 0 (z) = f (z) for all z ∈ U then we write
Z
g= f

and we say that g is an antiderivative of f in U .

50
6.29 Note: Note that since complex functions have the same derivative formulas as
real functions, they also have the same antiderivative formulas. For example, we can use
Integration by Parts or the Substitution Rule to find an antiderivative.
2
6.30 Example: Find an antiderivative for f (z) = z 3 ez .
Solution: Make the substitution w = z 2 , dw = 2z dz, then use Integration by Parts with
u = 21 w, du = 12 dw, v = ew and dv = ew dw to get
Z Z Z
3 z2 1 1 1 w
z e dz = 2w e
w
dw = 2w e
w
− 2e dw = 12 w ew − 12 ew
1 1

2
= 2 (w − 1) ew = 2 z 2 − 1 ez .

6.31 Note: Let U ⊆ C be a non-empty connected open set. If V ⊂ U is non-empty

and open, and U \ V is also open, then we must have V = U since otherwise U would be
separated by the open sets V and U \ V .
6.32 Theorem: Let U ⊆ C be a non-empty connected open set. Let f, g : U → C be
holomorphic with f 0 = g 0 in U . Then there is a constant c ∈ C such that f = g + c in U .
Proof: Let a ∈ U , let c = f (a) − g(a), let h = f − g, and let
 
V = z ∈ U f (z) = g(z) + c = z ∈ U h(z) = c .

We must show that V = U . By the above note, we can do this by showing that V is
non-empty and that both V and U \ V are open. The set V is clearly non-empty since
a ∈ V . To see that V \ U is open, note that V = h−1 (c) so we have −1


U \ V = h C \ {c} .
−1


Since h is continuous and C \ {c} is open, the set h C \ {c} is open by Theorem 3.29.
It remains to show that V is open.
Let w ∈ V , so we have h(w) = c, and choose r > 0 so that D(w, r) ⊆ U . We claim
that D(w, r) ⊆ V . Let z ∈ D(w, r). We must show that f (z) = g(z) + c, or equivalently
that h(z) = c. Note that since f 0 = g 0 and h = f − g we have h0 = 0. Let p be the
point with Re (p) = Re (z) and Im (p) = Im (w). Let u = Re (h) and v = Im (h) so that
h = u + i v. On the horizontal line through w, given by α(t) = w + t for t ∈ R, we have

d d d
u α(t) + i v α(t) = h(α(t) = h0 α(t) α0 (t) = 0.





dt dt dt
d d




Since dt u α(t)

= 0 and dt v α(t) = 0, it follows

that the real-valued functions u α(t)
and v α(t) are both constant and so h α(t) is constant. In particular, h(p) = h(w) = c.
A similar argument involving the vertical line through p, which is given by β(t) = p + it
for t ∈ R, shows that h(z) = h(p) = c.

6.33 Example: Let Uα = r ei θ |r > 0, α < θ < α + 2π , and let f (z) = 1/z. Then the
antiderivatives of f in Uα are the maps of the form g(z) = log z +c where log z = |z|+i θ(z)
with α < θ(z) < α + 2π. However, f (z) does not have an antiderivative in C∗ because
none of the maps g(z) can be extended continuously to C∗ .

51
6.34 Theorem: (The Fundamental Theorem of Calculus) Let α : [t1 , t2 ] → U ⊆ C be a
path in U , and let f, g : U → C with f continuous and g holomorphic with g 0 = f in U .
Then Z h iα(t2 )
f = g(z) = g(α(t2 )) − g(α(t1 )) .
α α(t1 )

In particular, if α is a loop then Z

f = 0.
α

Proof: Let h(t) = g(α(t)). By the Chain Rule, h0 (t) = Dg(α(t))α0 (t) = g 0 (α(t))α0 (t), and
so by the Fundamental Theorem of Calculus 6.9, we have
Z Z Z t2 Z t2
0 0 0
f= g = g (α(t))α (t) dt = h0 (t) dt = h(t2 ) − h(t1 ) = g(α(t2 )) − g(α(t1 )) .
α α t1 t1
Z
When α is a loop we have α(t1 ) = α(t2 ), so g(α(t1 )) = g(α(t2 )), and hence f = 0.
α
6.35 Example: Using the Fundamental Theorem of Calculus, we can solve example 6.20
as follows
Z h iα(1) h i1+i
z 2 dz = 13 z 3 = 13 z 3 = 13 (−2 + 2i) − (8) = − 10 2


3 + 3 i.
α α(0) 2

Z
2
6.36 Example: Find z 3 ez dz where α(t) = 1 + 2 ei t , 2π
3 ≤ t ≤ π.
α

√
Solution: Since α 2π3 = 3 i and α(π) = −1, by the Fundamental Theorem of Calculus,
using the antiderivative calculated in example 6.30, we have

2 −1
Z h i
2
z 3 ez dz = 12 (z 2 − 1) ez √ = 0 − 12 (−4)e−3 = e23 .
α 3i

Z
6.37 Example: Find sin3 z sec2 z dz where α(t) = i + ei t for − π2 ≤ t ≤ π
2.
α
Solution: We make the substitution u = cos z, du = − sin z dz to get

sin3 z dz (1 − cos2 z) sin z dz 1 − u2

Z Z Z Z
3 2
sin z sec z dz = = = − du
cos2 z cos2 z u2
Z
1 1
= 1− u2 du = u + u = cos z + sec z ,

and so
Z h iα(π/2) h i2i
3 2
sin z sec z dz = cos z + sec z = cos z + sec z
α α(−π/2) 0
 2 −2  4
2e2
= e +e
2 + e2 +e2
−2 − (1 + 1) = e2e+1
2 + e4 +1 − 2 .

52
6.38 Definition: For a path α : [t1 , t2 ] → U ⊆ C and a point a ∈ C which does not
lie on the path α, we define the winding number η(α, a) of α about a as follows. We
write α(t) = a + r(t)ei θ(t) where r(t) = |α(t) − a| and θ(t) is chosen continuously with
0 ≤ θ(t1 ) < 2π (it can be shown that the map θ(t) exists and is uniquely determined), and
then we set
θ(t2 ) − θ(t1 )
η(α, a) = .
2π
If α is a loop then we have α(t1 ) = α(t2 ) and so ei θ(t1 ) = ei θ(t2 ) and hence θ(t2 ) − θ(t1 )
will be a multiple of 2π. Thus for a loop α, we have η(α, a) ∈ Z.
6.39 Example: It is not hard to find the winding number η(α, a) from a picture of the
path α. For example, for α and a as shown below, we can choose values t = si (as shown).
Then θ(s0 ) ∼ = π4 , and then θ(t) increases (since we move counterclockwise around a) with
θ(s1 ) = π4 , θ(s2 ) = π, θ(s3 ) = 3π 5π
2 , θ(s4 ) = 2π and θ(s5 ) = 2 , and then θ(t) reaches its
maximum at θ(s6 ) ∼ 11π
= 4 and begins to decrease (since we now begin moving clockwise
around a) with θ(s7 ) = 5π ∼ 7π
2 , θ(s8 ) = 2π and finally θ(s9 ) = 4 . Thus we have
7π π
θ(s9 ) − θ(s0 ) ∼ 4 − 4 3
η(α, a) = = = 4 .
2π 2π

6.40 Example: If α is the pretzel curve α(t) = r(t)ei θ(t) , where r(t) = (2 + cos 3t) and
θ(t) = 2t with 0 ≤ t ≤ 2π (as shown below), then the winding number of α about 0
is η(α, 0) = θ(2π)−θ(0)
2π = 4π−0
2π = 2. The winding number about other points is hard to
compute from the given formula for α, but is easy to find using a sketch of the curve. For
example we have η(α, 2) = η(α, 2ei 2π/3 ) = η(α, 2ei 4π/3 ) = 1 and η(α, 4) = 0.

53
6.41 Theorem: (Winding Number) For the path α(t) = a + r(t)ei θ(t) with r(t) > 0 and
t1 ≤ t ≤ tt , we have
Z iα(t2 )
dz r(t2 ) h
= ln + 2π i η(α, a) = ln |z − a| + 2π i η(α, a) .
α z−a r(t1 ) α(t1 )

In particular, when α is a loop we have r(t1 ) = r(t2 ) so

Z
dz
2π i η(α, a) =
α z−a
Proof: We have
Z t2 Z t2 0 i θ Z t2 0 Z t2
α0 (t) r e + i r θ 0 ei θ
Z
dz r
= dt = i θ
dt = dt + i θ0 dt
α z−a t1 α(t) − a t1 re t1 r t1
h it2 h i t2

= ln r(t) + i θ(t) = ln r(t2 ) − ln r(t1 ) + i θ(t2 ) − θ(t1 )
t1 t1
r(t2 )
= ln + 2π i η(α, a) .
r(t1 )
Z
5z + 2 3
for 0 ≤ t ≤ 83 .

i πt
6.42 Example: Find 2 2
dz where α(t) = 1 + 8 t e
α z (z + 1)
Solution: First we sketch the path α by making a table of values and plotting points on a
polar grid.
3
t θ = πt r =1+ 8 t
0 0 1
1/3 π/3 9/8
2/3 2π/3 5/4
1 π 11/8
4/3 4π/3 3/2
5/3 5π/3 13/8
2 2π 7/4
7/3 7π/3 15/8
8/3 8π/3 2

8

√ 5z + 2
Notice that α(0) = 1 and α 3 = −1 + 3 i. Now we decompose the function
z 2 (z + 1)2
A B C D 5z + 2
into partial fractions. In order to get + 2 + + = we need
z z z+1 (z + 1)2 z 2 (z + 1)2
Az(z +1)2 +B(z +1)2 +Cz 2 (z +1)+Dz 2 = 5z +2 . Equate coefficients to get the equations
A + C = 0, 2A + B + C + D = 0, A + 2B = 5 and B = 2. Solve these to get A = 1, B = 2,
C = −1 and D = −3. Using the Winding Number Theorem, we have
Z Z
5z + 2 1 2 1 3
2 2
dz = + 2− − dz
α z (z + 1) α z z z + 1 (z + 1)2
h i−1+√3 i √ h i−1+√3 i
= ln 21 + 2π i η(α, 0) − z2 − ln 23 − 2π i η(α, −1) + z+13
1 1
√
4 2√
= ln 2 + 2π i −
3 −1+ 3 i
+ 2 − ln 23 − 2π i 45 + −1+3√3 i − 23
√ √ √
= ln 2 + 8π3 i + 1
2 (1 + 3 i) + 2 − ln 3 5π 3
2 − 2 i − 4 (1 + 3 i) − 23
   √ 
= ln √4 + 1 − π − 3 i.
3 4 6 4

54
 Z
6.43 Example: Find sin3 z sec2 z dz where α(t) = i + ei t for − π2 ≤ t ≤ π
2.
α
Solution: We make the substitution u = cos z, du = − sin z dz to get

sin3 z dz (1 − cos2 z) sin z dz 1 − u2

Z Z Z Z
3 2
sin z sec z dz = = = − du
cos2 z cos2 z u2
Z
1 1
= 1− u2 du = u + u + c = cos z + sec z + c ,

and so
Z h iα(π/2) h i2i
sin3 z sec2 z dz = cos z + sec z = cos z + sec z
α α(−π/2) 0
 2 −2  4
2e2
= e +e
2 + e2 +e2
−2 − (1 + 1) = e2e+1
2 + e4 +1 − 2 .

55
 Chapter 7. Cauchy’s Integral Formulas

7.1 Remark: Let U ⊆ C be open, and let α be a path which runs counterclockwise
around the boundary of a closed convex set E ⊂ U . Recall that Green’s theorem (for real
path integrals) states that if u, v : U → R are C 1 maps, then
Z Z Z
u dx + v dy = (vx − uy ) dx dy .
α E

Let f : U → C be holomorphic, and let u = Re (f ) and v = Im (f ). If we suppose that u

and v are C 1 , then Green’s Theorem and the Cauchy-Riemann equations imply that
Z Z Z Z
f (z) dz = (u + i v)(dx + i dy) = u dx − v dy + i v dx + u dy
α α α α
Z Z Z Z
= (−vx − uy )dx dy + i (ux − vy )dx dy = 0 .
E E

We shall now prove a series of theorems which generalize this result (which is known as
Cauchy’s Theorem) and which do not require the assumption that u and v are C 1 .
Indeed, we shall be able to show that every holomorphic map is C ∞ .
∞
T
7.2 Lemma: Let K0 ⊇ K1 ⊇ K2 ⊇ · · · be non-empty compact sets. Then Kn 6= ∅.
n=0
∞
Kn = ∅. For n ∈ Z+ , let Un = Kn c . Note
T
Proof: Suppose, for a contradiction, that
n=0
∞
T ∞
 S c
that each Un is open and U1 ⊆ U2 ⊆ U3 ⊆ · · ·. We have ∅ = K0 ∩ Kn = K0 ∩ Un
n=1 n=1
∞
S 
and so K0 ⊆ Un . Thus U1 , U2 , U3 , · · · is an open cover of K0 . Since K0 is compact,
n=1
we can choose a finite subcover, say K0 ⊆ Un1 ∪ Un2 ∪ · · · ∪ Unl with n1 < n2 < · · · < nl .
Since Un1 ⊆ Un2 ⊆ · · · ⊆ Unl we have K0 ⊆ Unl . But then K0 ∩ Unl c = ∅, that is
K0 ∩ Knl = ∅, and this is not possible since ∅ =
6 Knl ⊆ K0 .
7.3 Theorem: (Cauchy’s Theorem in a Triangle) Suppose that f : U ⊆ C → C is
holomorphic in U . Let ∆ be a closed solid triangle
Z in U and let α be a loop which goes
once around the boundary of the triangle. Then f = 0.
α
R
Proof: Let I = α f and set I0 = I, ∆0 = ∆, α0 = α and L0 = L(α). Divide ∆ into four
similar triangles ∆01 , ∆02 , ∆03 and ∆04 , let α01 , . . . , α04 be loops around these triangles,
R
and let I0j = α0j f for j = 1, 2, 3, 4. Choose k so that I0k is the largest of these, and
then set I1 = I0k , ∆1 = ∆0k , α1 = α0k and L1 = L(α1 ). Since the triangles ∆0j are half
as big as ∆0 we have L0 = 2L1 . Also, since I1 ≥ I0j for all j, we have
Z 4 Z
X 4 Z
X n
X
I0 = f = f ≤ f = I0j ≤ 4I1 .
α0 j=1 α0j j=1 α0j j=1

55
Next we subdivide ∆1 into four similar triangles ∆11 , ∆12 , ∆13 and ∆14 , and repeat the
procedure. In this way we obtain a sequence of similar triangles ∆0 ⊃ ∆1 ⊃ . . . with a loop
αk around each triangle, and we have I0 ≤ 4I1 ≤ 42 I2 ≤ . . . and L0 = 2L1 = 22 L2 = . . .,
R
where Ik = αk f and Lk = L(αk ). By the above lemma, we can choose a point a which
lies in all of the compact sets ∆k .
Now let  > 0. Since f is holomorphic at a, we can choose δ so that for |z − a| < δ
we have f (z) − f (a) + f 0 (a)(z − a) ≤ |z − a|. Choose N so that for n ≥ N we have

∆n ⊂ D(a, δ), and note that for all z ∈ ∆n we have |z − a|
≤ Ln . So for z ∈ ∆n we

0
have |z − a| < δ which implies f (z) − f (a) + f (a)(z − a) ≤ |z − a|
< Ln . Since
fR (a) + f 0 (a)(z − a) has an antiderivative, namely f (a) zR+ f 0 (a) 21 z 2 − az , we know
that
f (a) + f 0 (a)(z − a) dz = 0 so we have α f (z) dz = α f (z) − f (a) + f 0 (a)(z − a) dz.
R
αn
Writing Mn = max f (z) − (f (a) + f 0 (a)(z − a) < Ln , the Estimation Theorem gives


z=α(t)

L0 2
Z Z
f (z) − f (a) + f 0 (a)(z − a) dz ≤ Mn Ln ≤  Ln 2 =  n .


In = f (z) dz =
αn αn 4

Thus I0 ≤ 4n In ≤ L0 2 . Since  was arbitrary, we must have I0 = 0.

7.4 Theorem: (Cauchy’s Theorem in a Convex Region) Suppose that f : U ⊆ C → C
is holomorphicZin U , where U is open and convex. Then f has an antiderivative in U .
Consequently, f = 0 for all loops α in U .
α
Z
Proof: Choose any point a ∈ U . For each z ∈ U set g(z) = f where α is the line
α
segment from a to z (that is α(t) = a + (z − a) t , 0 ≤ t ≤ 1). We claim that g 0 (z) = f (z)
for all z ∈ U . Indeed, given h ∈ C (small enough so that z + h ∈ U ) we let β be the
line segment from z to z + h and Rwe let γ beRthe line segment from z + h to a, so by the
definition of g we have g(z + h) = γ −1 f = − γ f , and by Cauchy’s Theorem in a Triangle
R R R
we have α f + β f + γ f = 0, and so

g(z + h) − g(z)
Z Z
1 
f (z) − = f (z) + f (w) dw + f (w) dw
h h α γ
Z
1
= f (z) − f (w) dw
h β
Z Z
1 1
= f (z) dw − f (w) dw
h β h β
Z
1
= f (z) − f (w) dw
h β
≤ max f (z) − f (w) .
w=β(t)

As h → 0 we have w = β(t) → z and so |f (z) − f (w)| → 0, since f is continuous.

56
7.5 Definition: Let α, β : [t1 , t2 ] → U ⊆ C be paths with α(t1 ) = β(t1 ) = a and
α(t2 ) = β(t2 ) = b. A path-homotopy (or deformation of paths) from α to β in U is a
continuous map F : [t1 , t2 ] × [0, 1] → U such that F (t, 0) = α(t) and F (t, 1) = β(t) for all
t, and also F (t1 , s) = a and F (t2 , s) = b for all s. If such a homotopy exists, then we say
that α is (path)-homotopic to β in U and we write α ∼ = β in U . Note that for each fixed
s ∈ [0, 1], the map Fs (t) := F (t, s) is a continuous curve from a to b.

F β b

α
7.6 Example: In a convex set U ⊆ C we can find a path-homotopy between any two
paths α, β : [t1 , t2 ] → U with α(t1 ) = β(t1 ) and α(t2 ) = β(t2 ). Indeed, we can take


F (t, s) = α(t) + s β(t) − α(t) .

7.7 Theorem: (Cauchy’s Theorem for Z If f : U ⊆ C → C is holomorphic and α

Z Paths)
and β are homotopic paths in U then f= f.
α β

Proof: Say α, β : [p, q] → U . Choose a path-homotopy F : [p, q] × [0, 1] → U from α to

β in U . Choose partitions p = t0 < t1 < . . . tk = q and 0 = s
1 < s2 < . . . sl = 1 with
the property that for each i, j the image F [ti−1 , ti ] × [sj−1 , sj ] is contained in a convex
set which lies in U . (To prove that such partitions can be found, you must use the fact
that [p, q] × [0, 1] is compact). For each i and j, let αi be the restriction of α to [ti−1 , ti ],
let βi be the restriction of β to [ti−1 , ti ], let ai be the line segment from α(ti−1 ) to α(ti ),
let bi be the line segment from β(ti−1 ) to β(ti ), and let γij be the polygonal loop around
the polygon with vertices at F (ti−1 , sj−1 ), F (ti , sj−1 ), F (ti , sj ) and F (ti−1 , sj ). Then by
Cauchy’s Theorem for convex sets, we have
Z Z Z Z Z
f= f , f= f and f =0.
αi ai βi bi γij

When we consider all of the paths ai −1 , bi and γij , every line segment occurs twice, once
in each direction, and so the path integrals all cancel with each other to give
XZ XZ XZ
0= f− f+ f
i bi i ai i,j γij
XZ XZ
= f− f
i βi i αi
Z Z
= f− f
β α

57
7.8 Example: Let α, β : [0, π] → C∗ be given by α(t) = ei t and β(t) = e−i t . Show that
α and β are not homotopic in C∗ .

Let f (z)R= 1/z.R Then f is holomorphic in C∗ and we have α f = i π and

R
Solution:
R
β
f = −i π. Since α f 6= β f we know that α is not homotopic to β.

7.9 Definition: Let α, β : [t1 , t2 ] → U ⊆ C be loops in U . A loop-homotopy (or

deformation of loops) from α to β in U is a continuous map F : [t1 , t2 ] × [0, 1] → U such
that F (t, 0) = α(t) and F (t, 1) = β(t) for all t and F (t1 , s) = F (t2 , s) for all s. If such a
homotopy exists, we say that α is (loop)-homotopic to β in U and we write α ∼ β in U .

F
α

7.10 Example: In a convex set U ⊆ C, any two loops are homotopic.
Indeed, given loops
α, β : [t1 , t2 ] → U we can take F (t, s) = α(t) + s β(t) − α(t) .

7.11 Theorem: (Cauchy’s Theorem for Z If f : U ⊆ C → C is holomorphic and α

Z Loops)
and β are homotopic loops in U , then f= f.
α β

Proof: The proof is the same as the proof of Cauchy’s Theorem for Paths.

7.12 Example: Let α, β : [t1 , t2 ] → C∗ be loops. Show that if η(α, 0) 6= η(β, 0) then α
and β are not homotopic in C∗ .
R
Solution: Let f (z) = 1/z. Then f is holomorphic and we have α f = 2π i η(α, 0) and
f = 2π i η(β, 0), so if η(α, 0) 6= η(β, 0) then α and β cannot be homotopic in C∗ .
R
β

7.13 Definition: A set U ⊂ C is called simply connected when U is connected and

any two loops α, β : [t1 , t2 ] → U are homotopic in U . Roughly speaking, a connected set
will be simply connected if it doesn’t have any holes in it.

7.14 Example: Any convex set is simply connected, but C∗ is not.

Z Region) If U ⊆ C is a simply
7.15 Theorem: (Cauchy’s Theorem in a Simply Connected
connected open set and if f : U → C is holomorphic, then f = 0 for every loop α in U .
α

Proof: Since U is simply connected, any loop α : [t1 , t2 ] → U will be homotopic to the
Z Z Z t2
constant loop κ given by κ(t) = α(t1 ) for all t, so f = f = f (α(a))κ0 (t) dt = 0
α κ t1
since κ0 (t) = 0.

58
7.16 Theorem: (Cauchy’s Integral Formulas) Let U ⊆ C be a convex open set, let a ∈ U ,
let f be holomorphic inZ U and let α be a loop in U \ {a}. Then
f (z)
(a) 2π i η(α, a) f (a) = dz .
α z−a Z
(n) (n) f (z)
(b) All the derivatives f (a) exist, and 2π i η(α, a) f (a) = n! n+1
dz .
α (z − a)
Proof: First we prove part (a). For any  > 0, let α denote the path α (t) = a+(α(t)−a).
Note that α ∼ α in U \ {a}, indeed a homotopy is given by F (t, s) = α(t) + s(α (t) − α(t)).
f (z) − f (a)
Also note that the map is holomorphic in U \ {a}. So we have
z−a
Z Z Z
f (z) f (z) f (a)
dz − 2π i η(α, a) f (a) = dz − dz
α z−a α z−a α z−a
f (z) − f (a) f (z) − f (a)
Z Z
= dz = dz ≤ M L(α ) ,
α z−a α z−a
f (z) − f (a) f (z) − f (a)
where M = max . As  → 0 we have → f 0 (a) so M → |f 0 (a)|,
z=α (t) z−a z−a
and also L(α ) =  L(α) → 0 Z
(n) f (z)
We prove part (b) inductively. Suppose 2π i η(α, a) f (a) = n! n+1
dz.
α (z − a)
Then we have
 (n)
f (a + h) − f (n) (a)
 Z
n! f (z) f (z)
2π i η(α, a) = n+1
− dz
h h α (z − (a + h)) (z − a)n+1
Z  
n! 1 1
= f (z) − dz
h α (z − (a + h))n+1 (z − a)n+1
Z Z
(n + 1)! 1
= f (z) n+2
dw dz ,
h α λ (z − w)
where λ is the line segment from a to a + h. So we have
f (n) (a + h) − f (n) (a)
Z
f (z)
L := 2π i η(α, a) − (n + 1)! n+2
dz
h α (z − a)
Z Z Z 
(n + 1)! 1 f (z)
= f (z) n+2
dw dz − h n+2
dz
h α λ (z − w) α (z − a)
Z Z 
(n + 1)! 1 1
= f (z) n+2
dw − h dz
h α λ (z − w) (z − a)n+2
Z Z
(n + 1)! 1 1
= f (z) n+2
− dw dz
h α λ (z − w) (z − a)n+2
Z Z Z
(n + 2)! 1
= f (z) n+3
du dw dz ,
h α λ τ (z − u)
where τ is the line segment from a to w. Choose r > 0 so that D(a, 2r) ⊆ U \ Image (α),
and let |h| < r. For w ∈ Image (λ) and u ∈ Image (τ ) we have w between a and a + h, and
1 1
u between a and w, so u ∈ D(a, r), and so we have |z − u| ≥ r hence ≤ . By the
|z − u| r
(n + 2)! 1
Estimation Theorem, L ≤ L(α) max |f (z)| |h| |h| n+3 → 0 as |h| → 0.
|h| z=α(t) r

59
 Z
it z+1
7.17 Example: Let α(t) = 2 e for 0 ≤ t ≤ 2π and let f (z) = 2 . Find f (z) dz.
z +1 α
Solution: We shall find the integral in several ways. First, we shall use partial fractions. To
z+1 z+1 A B
write 2 = in the form + we need A(z − i) + B(z + i) = z + 1
z +1 (z + i)(z − i) z+i z−i
i+1 i−1
for all z. Setting z = i gives B(2i) = i + 1 so B = = . Setting z = −i gives
2i 2
−i + 1 1+i
A(−2i) = −i + 1 so A = = . And so we have
−2i 2
1−i 1−i
Z Z Z
z+1 1+i dz dz 1+i
2
dz = + = 2π i η(α, −i) + 2π i η(α, i)
α z +1 2 α z+i 2 α z−i 2 2
1+i 1−i
= 2π i + 2π i = π (i − 1) + π (i + 1) = 2π i .
2 2
Now we shall find the integral again by immitating the proof of Cauchy’s integral
formula. Notice that f is holomorphic except at z = ±i. Let α1 be the loop around the
top half of thecircle, and let α2 be the loop aroundthe bottom half, to be explicit, we
2 ei t for 0 ≤ t ≤ π 1 − π2 t for 0 ≤ t ≤ π
take α1 (t) = 2 and α2 (t) = and then we
R πR t − 3 for
R π ≤ t ≤ 2π 2 ei π for π ≤ t ≤ 2π
will have α f = α1 f + α2 f . Next we deform the paths α1 and α2 into the circular paths
σ1 and σ2 , where σ1 (t) = i + r ei t and σ2 (t) = −i + r ei t for 0 ≤ t ≤ 2π, where 0 < r < 1.
We have
Z 2π Z 2π
1 + i + r ei t
Z Z
0
f= f= f (σ1 (t)) σ1 (t) dt = it + r 2 ei 2t + 1
i reit dt
α1 σ1 0 0 −1 + 2i re
Z 2π
1+i
→ dt = π(1 − i) as r → 0 ,
0 2
and we have
2π 2π
1 − i + r eit
Z Z Z Z
0
f= f= f (σ2 (t)) σ2 (t) dt = it 2 i 2t
i reit dt
α2 σ2 0 0 −2i re + r e
2π
1−i
Z
−→ dt = π (i − 1) as r → 0 .
0 −2
Z
Thus f = 2π i.
α
Finally, we shall compute the integral a third time using Cauchy’s formula. Taking
α1 and α2 as above, we have
Z Z Z
(z + 1)/(z + i) F (z) i+1
f= dz = dz = 2π i F (i) = 2π i = π(i + 1) ,
α1 α1 z−i α1 z − i 2i
where F (z) = (z + 1)/(z + i), and
(z + 1)/(z − i) 1−i
Z Z Z
G(z)
f= dz = dz = 2π i G(−i) = 2π i = π (i − 1) ,
α2 α2 z+i α2 z + i −2i
Z
where G(z) = (z + 1)/(z − i). Again we obtain f = 2π i.
α

60
 it ez R
7.18 Example: Let α(t) = 2 e for 0 ≤ t ≤ 2π, and let f (z) = 2 . Find α f .
z −1
Solution: Of the three methods we used above, only the third works easily here. Notice
that f is holomorphic except at z = ±1. Let α1 be the loop aroundR the
R rightR half of the
circle, and let α2 be the loop around the left half, so we have α f = α1 f + α2 f . Then
we have
ez /(z + 1)
Z Z Z
F (z) e
f= dz = dz = 2π i F (1) = 2π i = i π e ,
α1 α1 z−1 α1 z − 1 2
and
ez /(z − 1) e−1
Z Z Z
G(z)
f= dz = dz = 2π i G(−1) = 2π i = −iπe−1 .
α2 α2 z+1 α2 z + 1 −2

So the integral of f over α is equal to i π e − 1e .

Z
it z+1
7.19 Example: Let α(t) = 2 e for 0 ≤ t ≤ 2π and let f (z) = 3 . Find f.
z (z − 1)2 α
Solution: We shall solve this integral using two methods. First we use partial fractions.
z+1 A B C D E
To write f in the form 3 2
= + 2 + 3 + + we need to have
z (z − 1) z z z z−1 (z − 1)2
Az 2 (z − 1)2 + Bz(z − 1)2 + C(z − 1)2 + Dz 3 (z − 1) + Ez 3 = z + 1 for all z. Equating
coefficients gives fives 5 equations: A + D = 0, −2A + B − D + E = 0, A − 2B + C = 0,
B − 2C = 1 and C = 1. Solving these gives A = 5, B = 3, C = 1, D = −5 and E = 2. So
Z Z
5 3 1 5 2
f= + 2+ 3− + dz
α α z z z z − 1 (z − 1)2


= 2π i 5η(α, 0) − 5η(α, 1) = 2π i(5 − 5) = 0 .

Now we compute the integral again using Cauchy’s formulas. Notice that f is holo-
morphic except at z = 0, 1. Let α1 be the loop around the portion of the circle which lies
1 1
to the right
R of the
R line yR = 2 and let α0 be the loop around the portion to the left of y = 2 ,
so that α f = α1 f + α0 f . We have

(z + 1)/z 3
Z Z Z
F (z) 2π i 00
f= dz = dz = F (0) .
α0 α0 (z − 1)2 α0 z 3 2!
z+1 −z − 3 2z + 10
From F (z) = 3
, we calculate F 0 (z) = 3
and F 00 (z) = to get
(z − 1) Z (z − 1) (z − 1)4
F 00 (0) = 10, so we have f = 10π i. Also,
α1

(z + 1)/z 3
Z Z Z
G(z) 2π i 0
f= dz = dz = G (1) .
α1 α1 (z − 1)2 α1 (z − 1) 2 1!
−2z − 3
Z
z+1
From G(z) = 3
we find G0 (z) = to get G0 (1) = −5, so we have f = −10π i.
zZ z4 α2

Again we obtain f = 0.
α

61
  
1
ei t − e−i t = 2i1 1


7.20 Note: If α(t) = ei t for 0 ≤ t ≤ 2π, then sin t = 2i α(t) − α(t)
−i t

1 
1 it
and cos t = 2 e + e = 2 α(t) + α(t) , and α0 (t) = i ei t = iα(t). It follows that
1

Z 2π Z


1
1 1
, 12 z + 1


f (sin t, cos, t) dt = f 2i z− z z dz .
0 α iz
Sometimes we can use this equality to solve a real integral involving trigonometric (or
hyperbolic) functions, by converting it to a path integral.
Z π
dt
7.21 Example: Find .
0 2 + cos t
Solution: Let α(t) = ei t for 0 ≤ t ≤ 2π. Then
Z π 1
1 2π i z dz −i dz
Z Z Z
dt dt 1
= = 1 1

= 2
0 2 + cos t 2 0 2 + cos t 2 α 2+ 2 z+ z α z + 4z + 1
−i dz
Z Z
F (z) dz
= √
√
= √
α z − (−2 + 3) z − (−2 − 3) α z − (−2 + 3)
√
= 2π i F (−2 + 3) = 2π i 2−i√ = √
3
π
3
,
−i
where F (z) = √ .
z − (−2 − 3)
Z 2π
dt
7.22 Example: Find .
0 3 + sin2 t
Solution: Let α(t) = ei t for 0 ≤ t ≤ 2π. Then
Z 2π 1
iz dz
Z Z Z
dt 4i z dz 4i z dz
2 = 2 = 2 4 2
= 4 2
α −12z + (z − 2z + 1) α z − 14z + 1
3 + sin t 1 1


α 3−
4 z− z
0
Z
4i z dz
= √
√
√
√

α z − (2 + 3) z + (2 + 3) z − (2 − 3) z + (2 − 3)
Let α1 be the loop around√ the right side of the unit√circle and let α2 be the loop around
the left side so that 2 − 3 lies inside α1 and −2 + 3 lies inside α2 . Then for
4i z
F (z) = √
√
√
and
z − (2 + 3) z + (2 + 3) z + (2 − 3)
4i z
G(z) = √
√
√
,
z − (2 + 3) z + (2 + 3) z − (2 − 3)
we have
Z 2π Z Z
dt F (z) dz G(z)
= √ + √
0 3 + sin2 t α1 z − (2 − 3) α2 z + (2 − 3)
 √ √ 
= 2π i F (2 − 3) + H(−2 + 3)
√ √ !
(4i)(2 − 3) (4i)(−2 + 3)
= 2π i √ √ + √ √
((−2 3)(4)(4 − 2 3) (−4)(2 3)(−4 + 2 3)
 
= −π −2 3 + −2 3 = √π3 .
1√ 1√

62
 Z ∞
7.23 Note: Sometimes we can solve an improper real integral of the form u(x) dx by
0
finding a complex-valued function f (z) of a complex variable, whose real (or imaginary)
part extends u(x), and then integrating f (z) around a large loop which follows the positive
x-axis, for example the loop which follows the line λ(t) = t for −R ≤ t ≤ R, then the
semicircle σ(t) = Rei t for 0 ≤ t ≤ π, where R is some large positive real number.

−R λ R
Z ∞
dx
7.24 Example: Find I = .
0 x4 +1
Solution: Let γ be the loop which follows the line λ(t) = t for −R ≤ t ≤ R then the
semicircle σ(t) = ei t for 0 ≤ t ≤ π. Then
Z Z Z
dz dz dz
4
= 4
+ 4
.
γ z +1 λ z +1 σ z +1

We have
Z Z
dz dz
=




γ z4 + 1 γ z−e
i π/4 z + ei π/4 z − ei 3π/4 z + ei 3π/4
Z Z
F (z) dz G(z) dz 
i π/4

i 3π/4


= i π/4
+ i 3π/4
= 2π i F e + G e
γ1 z − e γ2 z − e
 
1 1
= 2π i √ √ + √ √
(2eiπ/4 )( 2)( 2 i) (− 2)( 2 i)(2ei 3π/4 )
   
π 1 1 1
= 2 √
2
− 2i +
√ √
2
+ 2 i = √π2 ,
√1

where γ1 is a loop around the right half of γ, and γ2 is a loop around the left half of γ, so
that ei π/4 lies inside γ1 and ei 3π/4 lies inside γ2 , and F (z) and G(z) are as expected. Also
Z Z R
dz dt
4
= 4
−→ 2I as R → ∞
λ z +1 t=−R t + 1

and π Z π
iR ei t dt iR ei t
Z Z
dz
4
= it 4
≤ 4 i 4t + 1
dt
σ z +1 t=0 (Re ) + 1 t=0 R e
Z π
R dt πR
≤ 4
= 4 −→ 0 as R → ∞ .
t=0 R − 1 R −1
Since Z Z Z
π dz dz dz
√ = 4
= 4
+ −→ 2I + 0 as R → ∞ ,
2
γ z +1 λ z +1 σ z4+1
π
it follows that I = √
2 2
.

63
 Z ∞
dx
7.25 Example: Find I = .
0 (x2 + 1)3
Solution: Let λ, σ and γ be as in the previous example. Then
F 00 (i)
Z Z Z
dz dz F (z) dz πi 3π
2 3
= 3 3
= 3
= 2π i = = 8
γ (z + 1) γ (z − i) (z + i) γ (z − i) 2! (2i)5
where F (z) = 1/(z + i)3 , so F 0 (z) = −3/(z + i)4 and F 00 (z) = 12/(z + i)5 . Also,
Z Z R
dz dt
2 3
= 2 3
−→ as R → ∞
λ (z + 1) t=−R (t + 1)

and
π π
iR ei t dt
Z Z Z
dz R dt πR
=
3 ≤ = −→ 0 as R → ∞
σ (z + 1)2
2
t=0 (Rei t )2 + 1 0 (R2− 1) 3 (R − 1)3
2

Since
Z Z Z
3π dz dz dz
8 = = + −→ 2I + 0 as R → ∞ ,
γ (z + 1)3
2
λ (z + 1)3
2
σ (z 2 + 1)3
3π
it follows that I = 16 .
Z ∞
dx
7.26 Example: Find I = .
0 x3+1
Solution: Note that z 3 + 1 = (z + 1) z − ei π/3 z − e−i π/3 . Let γ be the loop which

follows the line λ then the arc σ then the line µ−1 , where λ(t) = t for 0 ≤ t ≤ R,
σ(t) = R ei t for 0 ≤ t ≤ 2π
3 , and µ(t) = t e
i 2π/3
for 0 ≤ t ≤ R, so that the point eiπ/3 lies
inside the loop γ. Then for F (z) = 1/ (z + 1)(z − e−i π/3 ) we have


Z Z
dz F (z)
= 2π i F ei π/3


3
= i π/3
γ z +1 γ z−e
2π i 2π

= 3 e−i pi/6 .
2π
= √
√ = √
3 3 3 1
2 + 2 i ( 3 i) 3 2 +2i
Also, we have
Z Z R
dz dt
3
= 3
−→ I as R → ∞ ,
λ z +1 t=0 t + 1
Z Z R i 2π/3
dz e dt
3
= 3
−→ ei 2π/3 I as R → ∞ ,
µ z + 1 t=0 t + 1
and
π π
i R ei t dt
Z Z Z
dz R dt πR
= ≤ = −→ 0 as R → ∞ .
σ z3 + 1 0 (Rei t )3 + 1 0 R3 − 1 R3 − 1
Since
Z Z Z Z
dz dz dz dz
2π
3 e −i π/6
= 3
= 3
+ 3
− 3
−→ I − ei 2π/3 I as R → ∞ ,
γ z +1 λ z +1 σ z +1 µ z +1
 √  √
2π
e−iπ/6 = 1 − ei 2π/3 I = 32 − 23 i I = 3 e−iπ/6 I. Thus I = 32π


it follows that 3
√ .
3

64
 Z ∞
cos x
7.27 Example: Find I = dx.
0 x2 + 1
∞
ei z ei z ei z
Z
cos x
Solution: Write I = dx. Let f (z) = = , let F (z) = ,
0 x2 + 1 z2 + 1 (z − i)(z + i) z+i
and let α be the loop that follows the line λ given by λ(t)
it
R = tR for −R R ≤ t ≤ R, then the
semicircle σ given by σ(t) = R e for 0 ≤ t ≤ π. Then α f = λ f + σ f and we have

e−1
Z Z
F (z) dz
f= = 2π i F (i) = 2π i · = πe ,
α α z − i 2i
Z Z R it Z R Z R
e dt cos t sin t
f= 2
= 2
dt + i 2
dt −→ 2I as R → ∞ , and
λ −R t + 1 −R t + 1 −R t + 1
Z π i Reit Z π
i Rei t
Z
e R dt πR
f = it 2
dt ≤ 2
= 2 −→ 0 as R → ∞ ,
σ 0 (Re ) + 1 0 R −1 R −1
it
since ei Re = ei(R cos t+i R sin t) = e−R sin t+i R cos t = e−R sin t ≤ 1 for 0 ≤ t ≤ π. It
folows that 2I = πe , so I = 2e
π
.
Z ∞
sin x
7.28 Example: Find I = dx.
0 x
Z ∞
sin x ei z
Solution: Write I = dx. Let f (z) = , and let α be the loop which follows
0 x z
first the line λ given by λ(t) = t for r < t < R, then the large semicircle σ given by
σ(t) = R ei t for 0 ≤ t ≤ R, then the line κ−1 where κ(t) = −t R for rR≤ t ≤RR, and
R thenRthe
small semicircle ρ−1 where ρ(t) = r ei t for 0 ≤ t ≤ π. Then α f = λ f + σ f − κ f − ρ f
and we have
Z
f = 0,
α
R R Z R
ei t
Z Z Z
cos t sin t
f= dt = dt + i dt ,
λ r t r t r t
R Z R Z R
e−i t
Z Z
cos t sin t
f =− dt = − dt + i dt ,
κ r t r t r t
Z Z Z R
sin t
f− f = 2i dt −→ 2i I as r → 0 and R → ∞ ,
λ κ r t
Z π iReit Z π Z π/2
i Reit
Z
e −R sin t
f = it
dt ≤ e dt = 2 e−R sin t dt
σ 0 Re 0 0
Z π/2 h iπ/2
−(2R/π)t π −(2R/π)t π
1 − e−R ≤ π


≤2 e dt = − R e =R R
0 0

−→ 0 as R → ∞ , and
Z π i reit Z π
i reit
Z
e
f= it
dt = i e−r sin t+i r cos t dt −→ iπ as r → 0 ,
ρ 0 r e 0
it it
where we used the fact that ei Re = e−R sin t+i R cos t and ei re = e−r sin t+i r cos t . It follows
that 2i I − iπ = 0, so I = π2 .

65
 Z ∞
ln x
7.29 Example: Find I = dx.
0 x2 +1
Z ∞
ln x log z log z
Solution: Write I = dx. Let f (z) = 2 and let F (z) = , where log z
0 +1 x2 z +1 z+i
is the branch of the logarithm given by log z = ln |z| + i θ(z) with − π2 < θ(z) < 3π2 , and
let α be the loop which follows first the line λ given by λ(t) = t for r < t < R, then the
large semicircle σ given by σ(t) = R ei t for 0 ≤ t ≤ R, then the line κ−1 where κ(t) = −t
−1 it
R r ≤Rt ≤ R,R and then
for R theR small semicircle ρ where ρ(t) = r e for 0 ≤ t ≤ π. Then
α
f = λ f + σ f − κ f − ρ f and we have

iπ
Z Z
F (z) dz 2
f= = 2π i F (i) = 2π i 2 = i π2 ,
α α z−i 2i
Z ZR
ln t
f= dt −→ I as r → 0 and R → ∞ ,
λ r +1 t2
Z Z R Z R Z R
ln t + i π ln t π dt
f =− 2
dt = − 2
dt − i 2
κ r t +1 r t +1 r t +1
Z ∞
π dx
−→ −(I + i J) as r → 0 and R → ∞ , where J = ,
0 x2 + 1
Z π Z π
(ln R + i t) i Reit dt
Z
(ln R + π)R π(ln R + π)R
f = it 2
dt ≤ 2
dt =
σ 0 (Re ) + 1 0 R −1 R2 − 1
−→ 0 as R → ∞ , and
Z π Z π
(ln r + i t) i reit
Z
(ln r + π)r π(ln r + π)r
f = dt ≤ dt =
ρ 0 (reit )2 + 1 0 1 − r2 1 − r2
−→ 0 as r → 0 .

2
π2
It follows that I + i J = i π2 so I = 0. Incidentally, we also find that J = 2 .
Z ∞
ln x
7.30 Example: Find I = dx.
0 (x2 + 1)3
Z ∞ Z ∞
ln x dx π dx log z
Solution: Write I = and J = . Let f (z) = 2 and
0 (x2 + 1)3 0 (x2+ 1) 3 (z + 1)3
log z
F (z) = , where log z is the branch of the logarithm given by log z = ln |z| + i θ(z)
(z + i)3
with − π2 < θ(z) < 3π 2 . Note that

1
(z + i)3 − 3(log z)(z + i)2 (z + i) − 3z log z
F 0 (z) = z
6
= , and
(z + i) z(z + i)4



00 (1 − 3 log z − 3)(z)(z + i)4 − (z + i) − 3z log z (z + i)4 + 4z(z + i)3
F (z) =
z 2 (z + i)8
(2 + 3 log z)(z)(z + i) + (z + i − 3z log z)(5z + i)
=− .
z 2 (z + i)5

66
Let α, λ, σ, κ and ρ be as in the previous two examples. Then
2 + i 3π 3π



F 00 (i) (i)(2i) + 2i + 2 (6i)
Z
2
2
f = 2π i = −π i · 5
= · · · = − π2 + i 3π
16 ,
α 2! −(2i)
Z Z R
ln t dt
f= 2 3
−→ I as r → 0 and R → ∞ ,
λ r (t + 1)
Z Z R
ln t + i π
f =− 2 3
dt −→ −I − i J as r → 0 and R → ∞ ,
κ r (t + 1)
Z π
(ln R + i t) i Reit
Z
π(ln R + π)R
f =
3 dt ≤ dt −→ 0 as R → ∞ , and
σ 0 (Reit )2 + 1 (R2 − 1)3
Z π
(ln r + i t) i reit
Z
π(ln r + π)r
f =
3 dt ≤ −→ 0 as r → 0 .
ρ 0 (reit )2 + 1 (1 − r2 )3
2
3π 2
It follows that 2I + i J = − π2 + i 3π π
16 , and so I = − 4 . We also find that J = 16 .
7.31
R Theorem: (Morera’s Theorem) Let f : U ⊆ C → C be continuous. Suppose that
α
f = 0 for every loop α in U . Then f is holomorphic in U .
Proof: Let a ∈ U . Choose r > 0 so that D(a, r) ⊆ U . Since D(a, r) is convex, the proof
of Cauchy’s Theorem in a Convex Set R shows that f has an antiderivative g in D(a, r)
(indeed, g may be defined by g(z) = λ f where λ is the line segment from a to z). Since
g is holomorphic in D(a, r), f = g 0 is also holomorphic in D(a, r) by Cauchy’s Integral
Formula.
7.32 Theorem: (Liouville’s Theorem) If f : C → C is holomorphic and bounded, then f
is constant.
Proof: Suppose that f is holomorphic in C with |f (z)| ≤ M for all z. Let a and b be any
two distinct points in C. Let α(t) = a + r|b − a|eit for 0 ≤ t ≤ 2π, where r > 1. Then
Z
1 f (z) f (z)
f (a) − f (b) = − dz
2π i α z − a z − b
a−b
Z
1
= f (z) dz
2π α (z − a)(z − b)
1 1 1
≤ 2π r|b − a| M |b − a|
2π r|b − a| (r − 1)|b − a|
M
= → 0 as r → ∞ .
r−1
7.33 Theorem: (The Fundamental Theorem of Algebra) Every non-constant polynomial
has a root in C.
Proof: Suppose, for a contradiction, that p is a non-constant polynomial with no roots.
Since p is a non-constant polynomial, we have p(z) → ∞ as z → ∞, and so we can choose
R large enough that when |z| ≥ R we have |p(z)| ≥ 1 and so 1/|p(z)| ≤ 1. Note that since p
has no roots, 1/p is holomorphic in C. In particular, 1/p is continuous in D(0, R) and so it
attains its maximum value. Since 1/p is bounded in D(0, R) and |1/p| ≤ 1 outside D(0, R),
we know that 1/p is bounded in C. By Liouville’s Theorem, 1/p must be a constant. But
this implies that p is constant, giving the desired contradiction.

67
 Chapter 8. Power Series

8.1 Definition: A sequence of complex numbers is a function f : {k, k + 1, k + 2 . . .} → C

where k ∈ Z. We usually write f (n) as an and we denote the sequence f by {an }n≥k or
simply by {an }. For a ∈ C, we say that the sequence {an } converges to a, and we write


lim an = a or we write an → a
n→∞

when for all  > 0 there exists N ∈ Z such that n ≥ N ⇒ an ∈ D(a, ). If the sequence
converges to some a ∈ C, then we say it converges, otherwise we say it diverges. We
say that the sequence {an } diverges to ∞, and write


lim an = ∞ or an → ∞
n→∞

when for all R > 0 there exists N ∈ Z such that n ≥ N ⇒ an ∈

/ D(0, R).

n
8.2 Example: If an = 1/n then an → 0. If bn = 2 + 12 (1 + i) then bn → 2. If
cn = (1 + i)n then cn → ∞. If dn = in then {dn } diverges.
8.3 Theorem: Let an = xn + i yn and let a = x + i y ∈ C. Then an → a if and only if
xn → x and yn → y.
Proof: Suppose first that an → a. Note that (xn − x) = Re (an − a) so |xn − x| ≤ |an − a|.
So given  > 0 we choose N ∈ Z so that n ≥ N ⇒ |an − a| < , and then for n ≥ N
we have |xn − x| ≤ |an − a| < . This shows that xn → x. Similarly, we can show
that yn → y. Conversely, suppose that xn → x and that yn → y. By the triangle
inequality we have |an − a| ≤ |xn − x| + |yn − y|. So
given  > 0 we choose N ∈ Z
so that n ≥ N ⇒ |xn − x| < 21  and |yn − y| < 21  . Then for n ≥ N we will have
|an − a| ≤ |xn − x| + |yn − y| < . This shows that an → a.
8.4 Theorem: Let {an } and {bn } be sequences with an → a and bn → b and let c ∈ C.
Then
(a) (c an ) → c a
(b) (an ± bn ) → a ± b
(c) (an bn ) → ab
(d) (an /bn ) → a/b, provided that b 6= 0 (and hence bn 6= 0 for large n)
(e) |an | → |a|
All parts except (c) and (d) hold for sequences in Rn .
Proof: We shall only prove part (c) (the proofs of the other parts are similar). We write
an = xn + i yn , a = x + i y, bn = un + i vn and b = u + i v. We suppose that an → a
and bn → b so that by Theorem 8.3 we have xn → x, yn → y, un → u and vn → v.
We have an bn = (xn + i yn )(un + i vn ) = (xn un − yn vn ) + i (xn vn + yn un ). From our
knowlege of sequences of real numbers, we know that (xn un − yn vn ) → xu − yv and that
(xn vn + yn un ) → xv − yu. By Theorem 8.3 again, we see that

an bn = (xn un − yn vn ) + i (xn vn + yn un ) → (xu − yv) + i (xv + yu) = ab .

68
 ∞
P P

8.5 Definition: We write an or simply an to denote the sequence {sl } where
n=k
l
P
sl = an . This kind of sequence is called a series, and the finite sums sl are called the
n=k P
partial sums. We say the series an converges or diverges according to whether the
P∞
sequence {sl } converges or diverges. We also write an to denote the limit of {sl }, if
n=k
∞
P
it exists, and we call the limit the sum of the series. If sl → s then we write an = s.
P P n=k
The series an is said to converge absolutely when the series |an | converges.
P P
8.6 Theorem: (Linearity) Suppose that an and bn converge and let c ∈ C. Then
∞
X ∞
X ∞
X ∞
X ∞
X
c an = c an and (an + bn ) = an + bn .
n=k n=k n=k n=k n=k

Proof: This is immediate from Theorem 8.4.

P
8.7 Theorem: (Convergence Tests) Let an be a series. Then
P
(a) If an converges then |an | → 0.
X∞ X∞
P P
(b) If |an | converges then an converges and an ≤ |an |.
n=k n=k
(c) (The Ratio Test)
an+1 P
(i) If lim < 1 then |an | converges.
n→∞ an
an+1 P
(ii) If ∃N ∈ Z s.t. n ≥ N ⇒ ≥ 1 then |an | 6→ 0 and so an diverges.
an
(d) (The Root Test)p P
(i) If lim n |an | < 1 then |an | converges.
n→∞ p P
(ii) If ∃N ∈ Z s.t. n ≥ N ⇒ n |an | ≥ 1 then |an | 6→ 0 and so an diverges.
an+1
Proof: We shall only prove the ratio test here. Suppose first that lim = p < 1.
n→∞ an
an+1
Choose r with p < r < 1. Choose N such that for n ≥ N we have ≤ r. Then
an
we have |aN +1 | ≤ r|aN |, aN +2 ≤ r|aN +1 | ≤ r2 |aN |, |aN +3 | ≤ r|aN +2 | ≤ r3 |aN | and so
on, andP so |ann−N| ≤ rn−N |aN | for all nP≥ N . Since 0 < r < 1, the real-valius geometric
series |aN |r converges, and so |an | converges by the comparison test for series of
positive real numbers.
On the other hand, if we suppose that there exists N ∈ Z such that for n ≥ N we
an+1
have an 6= 0 and ≥ 1 then we have |aN | ≤ |aN +1 | ≤ |aN +2 | ≤ · · · and so |an | 6→ 0.
an
P∞ 1
8.8 Example: The sum 2
converges by part (b) since for n ≥ 2 we have
n=0 (n + i)
1 1 P 1
|n + i| ≥ n − 1 so 2
≤ 2
, and we know that converges.
(n + i) (n − 1) (n − 1)2

69
 ∞
cn (z − a)n ,
P
8.9 Definition: A power series centred at a ∈ C is a series of the form
n=0
with cn ∈ C where, by convention, we take (z − a)0 = 1. A power series is a series for each
value of z ∈ C. It will converge for some values of z and diverge for others.
∞
z n = 1 + z + z 2 + · · · is a power series centred at
P
8.10 Example: The geometric series
n=0
l 1 − z l+1
zn = . For |z| < 1 we have z l → 0 as
P
a = 0. Its partial sums are given by sl =
n=0 1−z
1 ∞ 1
zn =
P
l → ∞ and so sl → hence . On the other hand, for |z| ≥ 1 we have
1−z n=0 1−z
∞
|z n | ≥ 1 for all n so |z n | 6→ 0 and hence z n diverges.
P
n=0
∞
cn (z − a)n be a power series.
P
8.11 Theorem: Let
n=0
(a) There exists a number R with 0 ≤ R ≤ ∞, called the radius of convergence of the
power series, such that
cn (z − a)n converges absolutely.
P
(i) if |z − a| < R then
(ii) if |z − a| > R then |cn (z − a)n | 6→ 0 and so cn (z − a)n diverges.
P

n cn (z − a)n−1 has the same radius of convergence R.

P
(b) The power series
∞
cn (z − a)n for z ∈ D(a, R) is
P
(c) When R > 0 then the function f defined by f (z) =
n=0
∞
Z ∞
X 1
0 n−1
cn (z − a)n+1 .
P
holomorphic with f (z) = n cn (z − a) and f (z) dz =
n=1 n=0
n + 1
(d) When R > 0, the above function f (z) has derivatives of all orders and the coefficients
∞
f (n) (a) X f (n) (a)
cn are given by cn = , so we have f (z) = (z − a)n .
n! n=0
n!
bn (z − a)n = cn (z − a)n for all z ∈ D(a, R) then bn = cn for all n.
P P
(e) When R > 0, if
Proof: We shall give the proof in the case that a = 0.
∞
cn wn converges, where w ∈ C
P
To prove part (a), we shall show that if
∞ n=0
cn z n converges absolutely for all z with |z| < |w|. Fix w ∈ C, suppose that
P
then
P n=0
cn wn converges, and let z ∈ C with |z| < |w|. Since cn wn converges, we know that
P
|cn w | → 0 as n → ∞ and so we can choose M > 0 so that M ≥ |cn wn | for all n. Then
n

we have
zn zn z n
|cn z n | = cn wn n = |cn wn | n ≤ M .
w w w
Since wz < 1, the series M wz converges and hence the series |cn z n | converges too
P P
by the comparison
 test
P (forn series of positive real terms). The radius of convergence is
R = sup |w| w ∈ C, cn w converges . If R = ∞ then the series converges P for all z.
Next we prove part (b). Let R be the radiusP of convergence of the series cn z n and
let S be the radius of convergence of the series n cn z n−1 . First we show P that R ≥ S.
If S 6= 0 then let zPbe any point P with |z| < S. Then by part (a), the series |n cn z n−1 |
n−1 1 n−1
converges, andP so |cn z |= n |n cn z | also converges by comparison, and hence
n n−1
P
|cn z | = |z| |cn z | also converges. This implies that R ≥ |z|. Since z was arbitrary,
we have R ≥ S.

70
 It is a bit harder to show that R ≤ S. If R 6= 0 then let z be any
n point with
n
0 < |z| < R. Choose ρ > 0 with |z| < ρ < R. We have |n cn z n−1 | = |z| |z|/ρ |cn ρn |. The
P
n
series (of positive real terms) n |z|/ρ converges by the Ratio Test, so we know that

n
n
n |z|/ρ → 0 and hence we can choose M > 0 so that M ≥ n |z|/ρ for all n. Then
M
we have |n cn z n−1 | ≤ |z| |cn ρn |. Since ρ < R we know that the series |cn ρn | converges,
P
PM n M
|cn ρn | also converges, and hence the series |n cn z n−1 |
P P
so the series |z| |cn ρ | = |z|
also converges by comparison. Thus S ≥ |z|, and since z was arbitrary, S ≥ R.
∞ ∞
cn z n and let g(z) = n cn z n−1 for all
P P
Now we prove part (c). Let f (z) =
n=0 n=1
z ∈ D(0, R). We claim that f 0 (z) = g(z). Given z ∈ D(0, R) choose r > 0 with |z| < r < R.
Then for |w| < r we have
∞ ∞
cn w n − cn z n
P P
∞
f (w) − f (z) n=0 n=0
X
− g(z) = − n cn z n−1
w−z w−z n=0
∞ ∞ ∞
X cn (wn − z n ) X n−1
X  wn − z n 
= − n cn z = cn − n z n−1
n=1
w−z n=1 n=2
w−z
∞
X  
= cn wn−1 + wn−2 z + · · · + w z n−2 + z n−1 − n z n−1
n=2
X∞  
n−2 n−3 n−4 2 n−2
= cn (w − z) w + 2w z + 3w z + · · · + (n − 1)z
n=2
∞
X  
n−2 n−3 n−2
≤ |cn ||w − z| |w| + 2 |w| |z| + · · · + (n − 1)|z|
n=2
X∞
|cn ||w − z| 1 + 2 + · · · + (n − 1) rn−2


≤
n=2
∞
|w − z| X
= n(n − 1) |cn | rn−2 .
2 n=2

cn z n , n cn z n−1 and − 1) cn z n−2 all have

P P P
But notice that by part (b), the series n(nP
the same radius of convergence R and so since r < R we know that n(n − 1)|cn | rn−2
converges. Thus |w −z| 12 n(n−1)|cn |rn−2 → 0 as w → z. This proves thatP f 0 (z) = g(z).
P
1 n+1
To complete the proof of part (c), note that by part (b) the power series n+1 cn z
has the same radius of convergence R, and by our above proof that f 0 = g, the function h
∞
1 n+1
is holomorphic in D(a, R) with h0 = f .
P
defined by h(z) = n+1 cn z
n=0
Part (d) follows from part (c). Indeed, if f (z) = c0 +c1 z+c2 z 2 +c3 z 3 +· · · then we have
f 0 (z) = c1 + 2 c2 z + 3 c3 z 2 + 4 c4 z 3 + · · ·, f 00 (z) = 2 · 1 c2 + 3 · 2 c3 z + 4 · 3 c4 z 2 + 5 · 4 c5 z 3 + · · ·
and f 000 (z) = 3 · 2 · 1 c3 + 4 · 3 · 2 c4 z + 5 · 4 · 3 c5 z 2 + · · · and so on, and we have f (0) = c0 ,
f 0 (0) = 1 c1 , f 00 (0) = 2! c2 , f 000 (0) = 3! c3 and so on. Using induction you can show that
f (n) (0) = n! cn .
Finally, note that part (e) follows immediately from part (d).

71
8.12 Theorem: (Taylor’s Theorem) If f (z) is holomorphic in D(a, R) and 0 < r < R ≤ ∞
then
∞
f (n) (a)
Z
X
n 1 f (z)
f (z) = cn (z − a) where cn = = n+1
dz ,
n=0
n! 2πi σ (z − a)
and where σ is the circle σ(t) = a + r ei t with 0 ≤ t ≤ 2π.
Proof: We give the proof in the case that a = 0. Fix z ∈ D(0, R) and choose r with
|z| < r < R. Then by Cauchy’s integral formula,
Z
1 f (w)
f (z) = dw
2π i σ w − z
Z
1 1 1
= f (w) dw
2π i σ w 1 − (z/w)
(z/w)N
Z   z 2  z N −1 
1 1 z
= f (w) 1+ + + ··· + + dw
2π i σ w w w w 1 − (z/w)
N −1
f (w)(z/w)N
Z Z
X 1 f (w) n 1
= z dw + dw
n=0
2π i σ wn+1 2π i σ w−z
N −1
X f (n) (0) n
= z + RN ,
n=0
n!
f (w)(z/w)N
Z
1
where RN = dw. Setting M = max |f (w)|, the estimation theorem
2π i σ w−z w=σ(t)
N
1 M (|z|/r)
gives |RN | ≤ 2πr. Since |z| < r, we have RN → 0 as N → ∞
2π (r − |z|)
8.13 Example: The elementary complex functions have the same derivative formulas as
their real counterparts, and so they have the same Taylor series centred at the origin (or
centred at any real number). For all z ∈ C we have
∞
z
X zn
e =
n=0
n!
∞ ∞
X z 2n+1n
X z 2n
sin z = (−1) cos z = (−1)n
n=0
(2n + 1)! n=0
(2n)!
∞ 2n+1 ∞
Xz X z 2n
sinh z = cosh z =
n=0
(2n + 1)! n=0
(2n)!
For |z| < 1 we have
∞ ∞ ∞
1 X 1 X 1 X
= zn = (−1)n z n = (−1)n z 2n
1 − z n=0 1 + z n=0 1 + z2 n=0
When |z| < 1, the principal branch of logarithm and inverse tangent are given by
∞ ∞
X zn X zn
log(1 − z) = − log(1 + z) = (−1)n+1
n=1
n n=1
n
∞
−1
X z 2n+1
tan (z) = (−1)n
n=0
2n + 1

72
For |z| < 1 and for a ∈ R, the principal branch of (1 + z)a is given by
∞  
a
X a a(a − 1) 2 a(a − 1)(a − 2) 3
(1 + z) = zn = 1 + a z + z + z + ···
n=0
n 2! 3!

This last power series is called the Binomial series.

8.14 Note: We should point out two important differences between Taylor series of com-
plex functions and Taylor series of real functions. The first difference is that holomorphic
∞
 is not the case for real C functions.
functions are always equal to their Taylor series. This
2
e−1/x , x6=0
The standard example is the real function f (x) = 0 , x=0 . This function is C ∞ at
x = 0, but all its derivatives vanish so its Taylor series is equal to 0. The second difference
we would like to mention is that a real function might be C ∞ in a large interval while
its Taylor series might converge only in a small interval, but notice that if a function is
holomorphic in an open disc, then its Taylor series will converge in the entire disc. An
example which illustrates this difference is the real function f (x) = 1/(1 + x2 ). This func-
tion is C ∞ for all x, but its Taylor series only converges for |x| < 1. The reason for this is
that when we extend f to the complex numbers, so f (z) = 1/(1 + z 2 ), then we find that
1
f (z) = so that f is holomorphic in C \ {±i}. The radius of convergence is
(z − i)(z + i)
equal to 1 because the disc D(0, 1) is the largest disc (centred at 0) which can be contained
in the domain of f (z).
8.15 Note: If f and g are both holomorphic at a then The product f g will also be
holomorphic at a. The coefficients of the Taylor series of f g at a are given by (f g)(n) (a)/n!,
and so they can be computed, using the product rule, from the coefficients of the Taylor
series for f and for g. One can show that the Taylor series at a for f g is obtained from the
Taylor series at a of f and of g by multiplying the power series together as if they were
polynomials. We have
∞ ∞ ∞  n
! ! 
X X X
n n
zn
P
an (z − a) bn (z − a) = ai bn−i
n=0 n=0 n=0 i=0

Also, if f and g are holomorphic at a and g(a) 6= 0, then we can solve the equation
hg = f for h to obtain the Taylor series of h = f /g centred at a from the Taylor series of
f and of g. This is equivalent to calculating f /g using long division as if the power series
were polynomials.
Also, if f is holomorphic at a and g is holomorphic at b = f (a) then the composite
g ◦ f is holomorphic at a and hence has a Taylor series centred at a. Using the chain rule,
one can show that the Taylor series for g ◦f at a can be computed by composing the Taylor
series of g at b with that of f at a as if the power series were polynomials.

73
 1
8.16 Example: Find the Taylor series at 0 for f (z) = .
(1 − z)2
Solution: We give several solutions. But first we note that since f (z) is holomorphic in
C \ {1}, we know that the Taylor series at 0 converges in D(0, 1).
For our first solution, we calculate the derivatives: f (z) = (1−z)−2 , f 0 (z) = 2(1−z)−3 ,
f (z) = 3!(1 − z)−4 , and so on. So f (0) = 1, f 0 (0) = 2, f 00 (0) = 3 and so on. Thus
00

f 0 (0) f 00 (0) 2 f 000 (0) 3

f (z) = f (0) + z+ z + z + · · · = 1 + 2z + 3z 2 + 4z 3 + · · ·
1! 2! 3!
Our second solution uses the Binomial series:
−2 (−2)(−3) (−2)(−3)(−4)
f (z) = (1 − z)−2 = 1 + (−z)1 + (−z)2 + (−z)3 + · · ·
1! 2! 3!
= 1 + 2z + 3z 2 + · · ·

1
Our third solution is to differentiate both sides of = 1 + z + z 2 + z 3 + · · · to obtain
1−z

f (z) = 0 + 1 + 2z + 3z 2 + · · ·

1
Our fourth solution is to mutiply the Taylor series for by itself as if it was a poly-
1−z
nomial to obtain
f (z) = (1 + z + z 2 + z 3 + · · ·)(1 + z + z 2 + z 3 + · · ·)
= 1 + (1 + 1)z + (1 + 1 + 1)z 2 + (1 + 1 + 1 + 1)z 3 + · · ·
= 1 + 2z + 3z 2 + 4z 3 + · · ·

8.17 Example: Find the Taylor series for f (z) = ez /(1 − z).
∞
P  P∞  ∞ n
z 1 1 n 1
n


z n . We can write
P P
Solution: We have f (z) = e 1−z = n! z z = n!
n=0 n=0 n=0 i=0
out the first few terms: f (z) = 1 + 2z + 52 z 2 + 8 3
3 z + 65
24 z
4
+ · · ·.
8.18 Example: Find the first few terms of the Taylor series about 0 for f (z) = tan z.
sin z
Solution: We have tan z = . We can use long division:
cos z
z + 13 z 3 + 2 5
15 z + ···

1 − 21 z 2 + 1 4
24 z + ··· z − 16 z 3 + 1 5
120 z − ···
z − 21 z 3 + 1 5
24 z − ···
1 3 1 5
3z − 30 z + ···
1 3 1 5
3z − 6z + ···
2 5
15 z + ···

We find that f (z) = z + 31 z 3 + 15

2 5
z + · · ·. We can also easily find the radius of convergence.
π
Since cos z = 0 ⇐⇒ z = 2 + π k for some k ∈ Z, we know that f (z) is holomorphic for
z 6= π2 + π k, so the radius of convergence is R = π2 .

74
 1
8.19 Example: Find the Taylor series centred at 2i for f (z) = .
z
∞
1 1 1 1 i 1 i X  i(z − 2i) n
Solution: f (z) = = = =− =−
z z − 2i + 2i 2i 1 + z−2i
2i
2 1 − i(z−2i) 2 n=0 2
2
∞
X  i n+1

= − (z − 2i)n . The disc of convergence is D(2i, 2).
n=0
2

8.20 Theorem: (The Local Identity Theorem) Let f and g be holomorphic in the disc
D(a, r). Let {an } be a sequence in D∗ (a, r) with an → a. If f (an ) = g(an ) for all n then
f (z) = g(z) for all z ∈ D(a, r).
Proof: Suppose that f (an ) = g(an ) for all n. Let h = f − g. Then h(an ) = 0 for all n.
Since h is continuous at a, we have h(a) = 0. Since its holomorphic it is equal to its
∞
cn (z − a)n . We want to show that all the coefficients cn are zero.
P
Taylor series h(z) =
n=0
Suppose not, and say m is the smallest integer such that cm 6= 0. Let k(z) = h(z)(z −a)−m
Then we have k(z) = cm + cm+1 (z − a) + cm+2 (z − a)m+2 + · · ·, so k(z) is holomorphic
in D(a, r) and k(a) = cm 6= 0. Since k(z) is continuous with k(a) 6= 0, we can find
s > 0 such that k(z) 6= 0 for all z ∈ D(a, s). But since (z − a)m 6= 0 in D∗ (a, s) and
since h(z) = k(z)(z − a)m , this would imply that h(z) 6= 0 in D∗ (a, s). This gives us a
contradiction since we assumed that h(an ) = 0 for all n.
8.21 Theorem: (The Identity Theorem) Let  U ⊆ C be a connected open set. Let
f, g : U → C be holomorphic in U . Let A = z ∈ U f (z) = g(z) . Suppose that A has a
limit point in U . Then f (z) = g(z) for all z ∈ U .
Proof: Let h = f − g so that h is holomorphic in U and A = h−1 (0) = z ∈ U h(z) = 0 .


We must show that h(z) = 0 for all z ∈ U , or equivalently that A = U . Let V be the set
of all limit points of A which lie in U . Note that V 6= ∅ since A has a limit point in U .
Note that V ⊆ A since for a ∈ U , if a ∈ / A, that is if h(a) 6= 0, then since h is continuous,
we can choose r > 0 so that h(z) 6= 0 for all z ∈ D(a, r), and we see that a is not a limit
point of A, that is a ∈/ V . Note that U \ V is open since if a ∈ U \ V , that is if a is not a
limit point of A, then we can choose r > 0 so that D∗ (a, r) is disjoint from A, and we see
that each z ∈ D(a, r) is not a limit point of A so we have D(a, r) ⊆ U \ V . Finally note
that V is open because given a ∈ V we can choose r > 0 so that D(a, r) ⊆ U and then,
by the Local Identity Theorem, we have h(z) = 0 for all z ∈ D(a, r) so that D(a, r) ⊆ V .
It follows that V = U (otherwise the open sets V and U \ V would separate U ) and hence
A = U (since V ⊆ A).
8.22 Lemma: Let f be holomorphic with |f | constant in the disc D(a, r). Then f is
constant.
Proof: Say |f (z)| = c for all z ∈ D(a, r). Let u = Re (f ) and v = Im (f ) so that f = u + iv.
Then we have u2 + v 2 = c2 . Differentiate with respect to x and to y to get the two
equations u ux + v vx = 0 and u uy + v vy = 0. The Cauchy Riemann Equations then
give u ux − v uy = 0 and v ux + u uy = 0. Eliminating uy from these two equations gives
(u2 + v 2 ) ux = 0, that is c2 ux = 0. Eliminating ux gives c2 uy = 0. If c = 0 then we have
|f (z)| = c = 0 so f (z) = 0 for all z ∈ D(a, r). If c 6= 0 then we obtain ux = 0 and uy = 0
for all x + iy ∈ D(a, r), and hence f is constant.

75
8.23 Theorem: (The Local Maximum Modulus Theorem) Let f be holomorphic in the
disc D(a, r). Suppose that f (z) ≤ f (a) for all z ∈ D(a, r). Then f is constant.
Proof: Fix s with 0 < s < r. Let σ(t) = a + s ei t for 0 ≤ t ≤ 2π. By Cauchy’s Integral
Formula, we have
2π 2π
f (a + s ei t ) i s ei t
Z Z Z
1 f (z) 1 1
f (a) = dz = i t
dt = f (a + s ei t ) dt .
2π i σ z−a 2π i 0 se 2π 0

Since |f (z)| ≤ |f (a)| for all z ∈ D(a, r), the Estimation Theorem gives
Z 2π
1
f (a) ≤ f (a + s ei t ) dt ≤ f (a)
2π 0

Z 2π
1
and so we must have |f (a)| = f (a + s ei t ) dt. It follows that
2π 0

Z 2π  
f (a) − f (a + s ei t ) dt = 0 .
0

Since the integrand is continuous and non-negative, it must be identically zero so we have
f (a + s ei t ) = f (a) for all t, that is |f (z)| = |f (a)| for all z with |z − a| = s. Since s
was fixed but arbitrary, we have |f (z)| = |f (a)| for all z ∈ D(a, r). By the above lemma,
it follows that f is constant.
8.24 Theorem: (The Maximum Modulus Theorem) Let U ⊆ C be a bounded connected
open set. Let f : U → C be holomorphic in U and continuous on U . Then |f | attains its
maximum on ∂U .
Proof: Since |f | is continuous on U and U is compact, |f | attains its maximum on U . Since
U = U ∪ ∂U , |f | attains its maximum either on ∂U or on U . Consider the case that |f |
attains its maximum at a point a ∈ U . Choose r > 0 small enough so that D(a, r) ⊆ U . By
the Local Maximum Modulus Theorem, f is constant in D(a, r). By the Identity Theorem,
f is constant in U . Since f is constant in U and continuous on U , it is constant on U .
Since f is constant, it attains its maximum at all points, including points in ∂U .

76
 Chapter 9. Laurent Series and Residues

9.1 Note: We have studied power series. We are also interested in series of the form

∞
X −1
X ∞
X ∞
X ∞
X
n n n n
cn (z − a) = cn (z − a) + cn (z − a) c = c−n w + cn (z − a)n .
n=−∞ n=−∞ n=0 n=1 n=0

where we have written w = 1/(z − a). If the first series has radius of convergence 1/R and
the second has radius of convergence S, then the first converges when |w| < 1/R, that is
when |z − a| > R, and the second converges for |z − a| < S. They both converge in the
annulus A = z ∈ C R < |z − a| < S . The next theorem shows that every function which
is holomorphic in an annulus can be expressed as a series of this form.
9.2 Theorem: (Laurent’s Theorem) Let 0 ≤ R < ρ < S ≤ ∞ and let a ∈ C. Suppose
that f is holomorphic in the annulus A = z ∈ C R < |z − a| < S . Then for all z ∈ A,

∞ Z
X
n 1 f (z)
f (z) = cn (z − a) where cn = dz ,
n=−∞
2πi σ (z − a)n+1

where σ is the circle σ(t) = a + ρ ei t with 0 ≤ t ≤ 2π. In particular, we have

Z
1
c−1 = f (z) dz .
2πi σ

Proof: To simplify notation, we take a = 0, so A = {z|R < |z| < S}. For z ∈ A pick r and
s so that R < r < |z| < s < S. Again to simplify notation, suppose that Im (z) > 0. Let
α be the loop in A which follows the semicircle counterclockwise from s to −s, then the
line segment from −s to −r, then the semicircle clockwise from −r to r, and then the line
segment from r to s. Let β be the loop which follows the line segment from s to r, then
the semicircle clockwise from r to −r, then the line segment from −r to −s, and then the
semicircle counterclockwise from −s to s.

z
α
0

Z
f (w)
Since η(α, z) = 1 and η(β, z) = 0, Cauchy’s theorem tells us that dw = 2πi f (z)
Z α w − z
f (w)
and dw = 0. Also, since the integrals along the line segments cancel, we have
β w−z

77
Z Z Z Z
f (w) f (w) f (w) f (w)
dw+ dw = dw− dw, where σr and σs are the circles
α w−z β w−z σs w − z σr w − z
σr (t) = r ei t and σs (t) = s ei t for 0 ≤ t ≤ 2π. So we have
Z Z 
1 1 1
f (z) = f (w) dw − f (w) dw
2πi σs w−z σr w−z
Z 
−1 1
Z
1 1 1
= f (w) dw − f (w) dw
2πi σs w 1 − wz σr z 1 − wz
∞ ∞
!
zn wm
Z Z
1 X X
= f (w) dw + f (w) dw
2πi σs n=0
wn+1 σr m=0
z m+1
∞ Z ∞ Z
!
1 X f (w) z n X f (w)wm
= dw + dw
2πi n=0 σs wn+1 m=0 σr
z m+1
∞ Z −1  Z
!
1 X f (w)  X f (w) 
= dw z n + n+1
dw z n
2πi n=0 σ wn+1 n=−∞ σ w
∞  Z 
1 X f (w)
= n+1
dw z n
2πi n=−∞ σ w

In the second last equlity, we replaced m by −n − 1, and we used the fact that each of the
loops σs and σr is homotopic to σ in A. The interchange of summation and integration in
the third equality should be justified. We can justify it as follows. For any positive integer
N we have
∞  N −1
(z/w)N
Z Z 
1 X z
n 1 X z
n
f (w) dw = f (w) + dw
σs w n=0 w σs w n=0 w 1 − (z/w)
N −1 Z
1 zn 1 (z/w)N
X Z
= f (w) dw + f (w) dw
n=0 σs
w wn σs w 1 − (z/w)

∞ Z
X zn
As N → ∞ the first term tends to the infinite sum f (w) n+1 dw and the second
n=0 σs w
term may be estimated using the Estimation Theorem:

(z/w)N (|z|/s)N
Z
f (w) dw ≤ max f (w) 2πs → 0
σs w−z |w|=s (s − |z|)

as N → ∞ since (|z|/s) < 1.

78
 1
9.3 Example: Let f (z) = . Note that f is holomorphic except at z = 0 and
z(z 2 + 4)
z = ±2i. In particular, f is holomorphic in the annulus A = {z|0 < |z| < 2} and in the
annulus B = {z|2 < |z| < ∞} and also in the annulus C = {z|0 < |z − 2i| < 2}. Find the
Laurent series Rof f (z)
R in A andR in B and in C. Also, use the Laurent series to find the
path integrals α f , β f and γ f , where α, β and γ are the circles α(t) = e , β(t) = 3 ei t
it

and γ(t) = 2i + ei t for 0 ≤ t ≤ 2π.

∞ ∞
1 1 1 X  2n X
n z (−1)n 2n−1
Solution: We have f (z) = = (−1) = z . This
4z 1 + (z/2)2 4z n=0 2 n=0
4n+1
is theZLaurent series for f in A. Since the coefficient of z −1 in this series is c−1 = 41 , we
1
have f = 2πi c−1 = π i.
α 2
∞ ∞
1 1 1 X n 2 2n
X
(−1)n 4n z −2n−3 . This


Also, we have f (z) = 3 2
= 3 (−1) =
z 1 + (2/z) z n=0 z n=0 Z
−1
is the Laurent series for f in B. Since the coefficient of z is c−1 = 0, we have f = 0.
β
In the third annulus we write
1 1 1 1 1 1
f (z) = =
z − 2i z + 2i z z − 2i (z − 2i) + 4i (z − 2i) + 2i
1 1 1 1 1
=
z − 2i 4i 1 + 4i 2i 1 + z−2i
z−2i
2i
∞  n X∞  n
1 1 X
n z − 2i n z − 2i
=− (−1) (−1)
8 z − 2i n=0 4i n=0
2i
 
∞ n
1 1 X X (−1)j (−1)n−j 
=− 
j n−j
(z − 2i)n
8 z − 2i n=0 j=0 (4i) (2i)
 
∞ n n
1 X X (−1) 
=− (z − 2i)n−1
8 n=0 j=0 in 2n+j
 
∞ n n
1 X i X 1 
=− (z − 2i)n−1
8 n=0 2n j=0 2j
∞
1 X in (2n+1 − 1)
=− (z − 2i)n−1
8 n=0 22n
Z
−1
This is the Laurent series in C. The coefficient of (z − 2i) is c−1 = − 81 so f = − 41 πi.
γ

9.4 Note: It should be remarked that all three of the path integrals in the above example
are easy to compute using Cauchy’s integral formula. In the following example, however,
its easier to use the Laurent series to find the path integral.

79
 1
9.5 Example: Let f (z) = 4 . Since sinh z = 0 when z = kπi, k ∈ Z we see that f
z sinh z
is holomorphic except at z = kπi. Find the firstRfew terms of the Laurent series for f in
the annulus A = {z|0 < |z| < π}, and hence find σ f where σ is the circle σ(t) = ei t with
0 ≤ t ≤ 2π.
1 1 1 1
Solution: We have f (z) = 4 = 4 . We use long division:
z sinh z 1 4
z z(1 + 61 z 2 + 120 z + · · ·)

1 − 16 z 2 + 7 4
360 z + ···

1 + 16 z 2 + 1 4
120 z + ··· 1 + 0 z2 + 0 z4 + · · ·
1 + 16 z 2 + 1 4
120 z + ···
− 61 z 2 − 1 4
120 z + ···
− 16 z 2 − 1 4
36 z + ···
7 5
360 z + ···
Z
−5 1 −3 7 −1 7 7π
We find that f (z) = z − 6z + 360 z + · · ·. Since c−1 = 360 , we have f= 180 i.
σ
9.6 Definition: Suppose that f is holomorphic in an open set U which contains the
punctured disc D∗ (a, R). We define the multiplicity of f at a, denoted by m(f, a), as
follows. Say the Laurent series of f in D∗ (a, R) is given by
∞
X
f (z) = cn (z − a)n .
n=−∞

If cn = 0 for all n ∈ Z so that f is identically zero, then we define m(f, a) = ∞. If for

all N ∈ Z there exists n < N with cn 6= 0 then we define m(f, a) = −∞. Otherwise, we
define m(f, a) be the smallest integer m such that cm 6= 0.
Let m = m(f, a). When m = −∞ we say that f has an essential singularity at a.
When m < 0 we say that f has a pole at a of order |m|. A pole of order 1 is also called
a simple pole. When m ≥ 0 we say that f has a removable singularity at a, and in this
case we shall extend f so that it is holomorphic in the disc D(a, r) by setting f (a) = c0 .
If m > 0 then we say that f has a zero at a of order m. A zero of order 1 is also called a
simple zero. In any of these cases, we define the residue of f at a to be Res (f, a) = c−1 .
If σ is the circle σ(t) = a + r ei t for 0 ≤ t ≤ 2π where 0 < r < R then we have
Z
1
Res (f, a) = c−1 = f (z) dz
2πi σ

9.7 Note: If f has a removable singularity at a, then of course we have lim f (z) = c0 .
z→a
If f has a pole at a then its not hard to show that lim f (z) = ∞. If f has an essential
z→a
singularity at a, then the limit lim f (z) does not exist, and in fact there is a (dificult)
z→a
theorem called Picard’s Theorem which states that for all  > 0 the image f D∗ (a, ) is

either equal to C or to C \ {p} for some point p.

80
9.8 Definition: We say that f is meromorphic in the open set U when there exist
finitely many points a1 , a2 , · · · , an ∈ U such that f is holomorphic in U \ {a1 , a2 , · · · an }
and f has a pole at each of the points ak . We remark that in this case we can extend f to
a holomorphic map f : U → C b by setting f (ak ) = ∞ for each k.

9.9 Theorem: (The Residue Theorem) Let U be an open set and let a1 , a2 , · · · , an ∈ U .
Let f be holomorphic in U \ {a1 , a2 , · · · , an }. Let α be a loop in U \ {a1 , a2 , · · · , an } which
is homotopic in U to a constant loop. Then
Z n
X
f (z) dz = 2πi η(α, ak ) Res (f, ak ) .
α k=1

Proof: Choose R > 0 so that each punctured disc D∗ (ak , R) lies inside U \ {a1 , a2 , · · · , an }.
Inside each of these punctured discs, f will be equal to its Laurent series. For fixed k, we
write
X∞
f (z) = cn (z − ak )n = pk (z) + hk (z) , where
n=−∞
−1
X ∞
X
n
pk = cn (z − ak ) and hk = cn (z − ak )n
n=−∞ n=0

(pk is called the principal part of f , and hk is called the holomorphic part of f at
the point ak ). We have Res (f, ak ) = c−1 = Res (pk , ak ). Notice that hk is holomorphic in
the disc D(ak , R) (not Zjust the punctured disc) and notice that pk is holomorphic in all of
C \ {ak } and we have pk (z) dz = 2πi η(α, ak ) c−1 . Now we let
α

n
X
g(z) = f (z) − pk (z) .
k=1

Although f was only holomorphic in U \ {a1 . . . , an }, X

the map g is holomorphic
X in all of
∗
U , indeed in D (ak , R) we have g(z) = f (z) − pk (z) − pj (z) = hk (z) − pj (z). Since
j6=k j6=k
α is homotopic to a constant loop in U , we have
Z Z n
X Z n Z
X
0= g(z) dz = f (z) − pk (z) dz = f (z) dz − pk (z) dz
α α k=1 α k=1 α
Z n
X
= f (z) dz − 2πi η(α, ak ) Res (f, ak )) .
α k=1

81
9.10 Note: We describe two situations in which it is easy to calculate Res (f, a). Suppose
g(z)
first that f (z) = where g(z) is holomorphic in an open set U containing a. Let
(z − a)n+1
σ be a circle in U centred at a. Then Cauchy’s Integral Formula gives

g (n) (a)
Z
1 g(z)
Res (f, a) = dz = .
2πi σ (z − a)n+1 n!

g(z)
Suppose next that f (z) = where g and h are holomorphic in an open set U containing
h(z)
a with h(a) = 0 and h0 (a) 6= 0. Then we can write h(z) = (z−a)k(z) with k(a) = h0 (a) 6= 0.
Note that k(z) 6= 0 in a small disc D(a, r) and g(z)/k(z) is holomorphic in this disc. We
g(z)/k(z)
have f (z) = with g(z)/k(z) holomorphic in D(a, r) so if σ is a circle in D(a, r)
(z − a)
centred at a then
Z
1 g(z)/h(z) g(a) g(a)
Res (f, a) = dz = = 0 .
2πi σ z − a k(a) h (a)

9.11 Example: Let α be a loop in D(0, 3) with η(α, 0) = 3, η(α, π2 ) = −1 and η(α, − π2 ) = 1,
(z + 1) ez
Z
and let f (z) = . Find f (z) dz.
z cos z α
Solution: Notice that f is holomorphic in C except at z = 0 and z = π2 + kπ for k ∈ Z.
In particular, f is holomorphic in D(0, 3) except at z = 0 and z = ± π2 . So by the Residue
Theorem, Z  
f (z) dz = 2πi 3 Res (f, 0) − Res (f, π2 ) + Res (f, − π2 ) .
α

To find Res (f, 0), note that we can write f (z) = g(z) z where g(z) = (z + 1) ez / cos z
(which is holomorphic at 0) so by the first part of the above note

Res (f, 0) = g(0) = 1 .

To find Res f, π2 and Res f, − π2 , note that we can write f (z) = g(z)/h(z) where now

g(z) = (z + 1)ez /z (which is holomorphic at ± π2 ) and h(z) = cos z (which has a simple
zero at ± π2 ). By the second part of the above note, we have

π π
+ 1 eπ/2 π2



π

g 2
2 2

π/2
Res f, = 0 π = = − 1 + π e , and
2
h 2
− sin π2
g − π2 − π2 + 1 e−π/2 − π2




Res f, − π2 = = 1 − π2 e−π/2 .



=
h0 − π2 π



− sin − 2

Thus we obtain
Z
f (z) dz = 2πi 3 + (1 + π2 )eπ/2 + (1 − π2 )e−π/2 .


α

82
9.12 Theorem: (Zeros and Poles) Let f be meromorphic in U . Let A be the set of all
zeros and poles of f in U . Let α be a loop in U \ A which is homotopic in U to a constant
loop. Then Z 0
f (z) X
dz = 2πi η(α, a) m(f, a) .
α f (z) a∈A

Proof: I may include a proof later.

9.13 Theorem: (Rouché’s Theorem) Let f and g be holomorphic in U . Let α be a loop
in U . Suppose that f (z) − g(z) < f (z) for all z ∈ Image (α). Then
X X
η(α, a) m(f, a) = η(α, a) m(g, b)
a∈A b∈B

where A is the set of zeros of f in U and B is the set of zeros of g in U .

Proof: I may include a proof later.
9.14 Theorem: Let f : U ⊆ C → C be holomorphic and injective. Then f is conformal.
Proof: I may include a proof later.
9.15 Theorem: (The Open Mapping Theorem) Let f : U ⊆ C → C be holomorphic and
non-constant. Then f (U ) is open.
Proof: I may include a proof later.
9.16 Theorem: (The Inverse Function Theorem) Let f : U ⊂ C → C be holomorphic
with f 0 (a) 6= 0 where a ∈ U . Then there exists r > 0 such that the restriction of f to
D(a, r) is invertible, and its inverse g = f −1 is holomorphic with

1
g 0 (w) =

f g(w)


for all w ∈ f D(a, r)
Proof: I may include a proof later.

83