9 Complex integral

Here we finally get to the central point of our course — complex integral. Specifically, our task is
make sense of the following notation ∫
f (z)dz.
γ

There are three parts of this formula: the integral itself, and what we actually understand by
writing the integral sign; the curve or path γ; and finally the domain E, which should contain γ and
on which f must be holomorphic (remember that the goal is to study holomorphic functions, not
arbitrary ones). There are quite a few subtle geometric and topological points related to γ, E, and f ,
but I will try to get to the final result (Cauchy’s theorem, see the next section) skipping most of these
technical points. Full treatment should be looked for in a graduate textbook.

9.1 Path
By a path γ in complex analysis one usually calls a rectifiable curve. Loosely speaking, this is a
continuous curve that has a finite length (but what is length anyway?). We will be satisfied with a
somewhat much less strict definition of the path: Path γ is a continuous piecewise smooth function.
This means that 1) γ : [α, β] −→ C is continuous, i.e.,

γ(t) = x(t) + iy(t),

and x and y are continuous real functions; moreover the derivative γ ′ (t) = x′ (t) + iy ′ (t) exists at all
the points of [α, β] except, possible, a finite number of points t1 , t2 , . . . , tn , and if derivative exists then
it is a continuous function on its own.
Here are a few examples.
A path corresponding to the half circle through 1, i, −1 is given by γ(t) = eit , t ∈ [0, π].
A path corresponding to the line segment connecting points x0 + iy0 and x1 + iy1 is given by
γ(t) = (x0 + (x1 − x0 )t) + i(y0 + (y1 − y0 )t), t ∈ [0, 1], etc. More examples will be given below.
Quite often it is important to distinguish between the path itself (function t 7→ γ(t)) and its image
(the set of points {z ∈ C : z = γ(t), t ∈ [α, β]}), I will not do it in these lectures and will be using the
same letter γ for both the path and its image.
Note that the path has a direction, I will denote −γ the path that has the same image as γ but
the opposite direction.
The path is called closed if γ(α) = γ(β); simple if it has no self intersections (except for possibly
γ(α) = γ(β)); a simple closed path is often called a Jordan’s curve.

9.2 The integral

Now that we have some idea about paths I can define

Definition 9.1. ∫ ∫ β
f (z)dz = f (γ(t))γ ′ (t)dt,
γ α

Math 452: Complex Analysis by Artem Novozhilov⃝

c

e-mail: [email protected]. Spring 2019

1
 where on the right is a pair of usual familiar real integrals
∫ β ∫ β ∫ β
′ ′
f (γ(t))γ (t)dt = Re(f (γ(t))γ (t))dt + i Im(f (γ(t))γ ′ (t))dt.
α α α

Note that this definition of complex integral is perfectly fine even for the paths that are not differen-
tiable at every point, because the Riemann integral exists for piecewise continuous functions, and this
is what I required from γ ′ to satisfy. ∫
The definition may look a little arbitrary, and in this respect I would like to remark that γ f is
nothing else as two real line integrals, which were studied in Calc III, and which have quite natural
geometric and physical interpretations. Indeed, assuming f = u + iv and z = x + iy,
∫ ∫ β
f (z)dz = (u + iv)(x′ + iy ′ )dt =
γ α
∫ β ∫ β
′ ′
= (ux − vy )dt + i (vx′ + uy ′ )dt =
∫α
∫ α

= udx − vdy + i vdx + udy.

γ γ

Here is our first (and arguably most important) example.

Example 9.2. Let γ = ∂B(0, 1), i.e., the boundary of the unit disk, and n ∈ Z. Then
∫ {
n 0, n ̸= −1,
z =
γ 2πi, n = −1.

In this case I have γ(t) = eit , t ∈ [0, 2π], hence

∫ ∫ 2π
n
z dz = i eit(n+1) dt,
γ 0
∫
and the result follows from eat dt = a1 eat for any complex a.

Here are a few technical points which will be important in the following. The proofs rely on the
properties of real integrals, which I assume to be known.

1. ∫ ∫
f= f.
γ −γ
∫b ∫a
The proof follows from a =− b .

2. Let γ = γ1 ∪ γ2 and γ1 and γ2 are disjoint. Then

∫ ∫ ∫
f= f+ f.
γ γ1 γ2

Clearly one can generalize to any finite number of γ’s. Proof follows from the definition and
properties of real integrals.

2
 3. Reparametrization. This is one of the most important properties which informally says that
the complex integral depends only on the image of γ and its direction, but not on a concrete
parametrization of this image. To be specific: Let γ̃ : [α̃, β̃] −→ C be another path, and assume
that γ̃ = γ ◦ ψ, where function ψ : [α̃, β̃] −→ [α, β], onto, and has a positive derivative. Then
∫ ∫
f = f.
γ̃ γ

Proof. Due to Property 2 I can concentrate on an interval where both γ ′ and γ̃ ′ exist. Then
γ̃ ′ (t) = γ ′ (ψ(t))ψ ′ (t)
by the chain rule.
∫ ∫ β̃
f (z)dz = f (γ̃(t))γ̃ ′ (t)dt =
γ̃ α̃
∫ β̃
= f (γ(ψ(t)))γ ′ (ψ(t))ψ ′ (t)dt = (changing the variables) =
α̃
∫ β ∫
= f (γ(s))γ(s)ds = f (z)dz.
α γ


Property 3 is especially important because it allows to choose the simplest paramentrization of

our path for each smooth segment. Here is an example.
Example 9.3. Find ∫
z 2 dz,
γ
where γ = γ1 ∪ γ2 , γ1 is the line segment of the real axis from −R to R and γ2 is the half circle in the
upper half plane connecting points R and −R. Due to property 2 I can separate this integrals into
two: along γ1 and γ2 correspondingly. I also have
γ1 (t) = −R + 2Rt, t ∈ [0, 1],
and
γ2 (t) = Reit , t ∈ [0, π],
note that my t is not continuous along the whole γ because by Property 3 I can choose the parametriza-
tion that I like. Finally,
∫ ∫ 1 ∫ π
2
z dz = R (2t − 1) 2Rdt +
2 2
R2 e2it iReit dt =
γ 0 0
 π
1)3  
1
2R3 (2t − 3
 + R i e3it  =
=  
2 3 0 3 0
1 1 R3 3πi
= R3 + R3 + (e − 1) = 0.
3 3 3
The calculations in the last examples are nice to emphasize the importance of the properties of
complex integrals, but, strictly speaking, are not necessary. In the next subsection I will explain the
reasons for this.

3
9.3 Antiderivative
Judging by its title the Fundamental theorem of calculus is the central theorem in Calc I. Here I will
show how to formulate a complex analogue of this theorem, but would like to remark from the very
beginning that this “fundamental theorem” is a particular case of a much more general statement
(Cauchy’s theorem) and hence plays only an axillary role.
I start with the usual definition that complex function F is called an antiderivative of f in domain
E if F ′ (z) = f (z) for all z ∈ E. From our experience we have that, e.g., z 2 is an antiderivative for
2z in all C, z −1 is an antiderivative of −1/z 2 in C \ {0}, and Log z is an antiderivative of 1/z in E,
where E has no points on the negative real half-axis. Note that similar to the real case antiderivative
is not unique, but is determined (on a domain E) up to an additive complex constant. I have

Proposition 9.4. Let F be an antiderivative of f on domain E, and let γ ∈ E. Then

∫
f (z)dz = F (γ(β)) − F (γ(α)).
γ

Proof. I will assume that γ is smooth. The full version is obtained by invoking Property 2 from the
previous section.
∫ ∫ β
f (z)dz = F ′ (γ(t))γ ′ (t)dt = (s = γ(t)) =
γ α
∫ γ(β)
= F ′ (s)ds = F (γ(β)) − F (γ(α))
γ(α)

as required. 

And here is our first glimpse of Cauchy’s theorem as an immediate corollary to the previous.

Corollary 9.5. Let γ be a closed path, γ ∈ E, and f has an antiderivative F in E. Then

I
f (z)dz = 0.
γ
H
(Here I use the notation to emphasize that the integral is taken along a closed path.)
To use this corollary one needs to be sure that an antiderivative exists, which is not always clear.
The contrapositive, however, gives a direct recipe to prove that there is no antiderivative in some
domain E. For instance we know (see the fundamental example above) that
∫
1
dz = 2πi,
∂B(0,1) z

which implies that z1 has no antiderivative in any domain containing the unit circle.
Finally since z 2 is entire, and clearly has antiderivative z 3 /3 in the whole C, then the calculations
in Example 9.3 were not necessary since we integrated along a closed path.

4
9.4 One useful estimate
For real integrals we have (which is in a sense a generalized triangle inequality)
∫ b  ∫ b
 
 f (x)dx ≤ |f (x)|dx.

a a
The proof follows from the inequalities f (x) ≤ |f (x)| and −f (x) ≤ |f (x)| and direct integration. One
cannot assume that the same inequality holds for complex integrals. Indeed, assume that something
similar holds ∫  ∫ ∫ 2π
 
 1  1
 dz  ≤ dz = ieit dt = 0,
 ∂B(0,1) z  ∂B(0,1) |z| 0
which does not make any sense since we already know that this integral is equal 2πi. Here is the
correct way to estimate complex integrals.
Proposition 9.6. Let E be a domain, γ ∈ E, and M = maxz∈γ |f (z)|. Then
∫  ∫ β
 
 f (z)dz  ≤ M |γ ′ (t)|dt = M · length γ,
 
γ α
where I defined
∫ β
length γ = |γ ′ (t)|dt.
α
Proof. The proof is not very illuminating and a little technical, but it uses one trick, which is very
useful
∫  to remember while working with complex numbers. So I provide the details. I need to estimate
 
 γ f . I will use the fact that
∫ ∫ 
 
f =  f  eiθ ,
γ γ
∫
where θ is the argument of γ f . Therefore,
∫  ∫
 
 f  = e−iθ f.
 
γ γ
Now, the expression on the left is real, hence the expression on the right must be real and hence
coincides with its real part:
∫  ∫ ∫ β [ ]
 
 f  = Re e−iθ f (z)dz = Re f (γ(t))eiθ ′
γ (t) dt.
 
γ γ α
Now, the right hand side is real, and I can use the integral inequality for the real integrals
∫  ∫ β   ∫ β
   
 f ≤ iθ ′
f (γ(t))e γ (t) dt = |f (γ(t))||γ ′ (t)|dt,
 
γ α α
from where the required inequality follows. 
Example 9.7. Let γ be the semicircle in the upper half plane of radius R and f (z) = (z 4 + 1)−1 . I
have ∫  ∫ π 
   Rieit  R
 f (z)dz  ≤  
   R4 e4it + 1  dt ≤ |R4 − 1| π,
γ 0
where I also used the inverse triangle inequality |a − b| ≥ ||a| − |b||.