MTH 407/607: COMPLEX ANALYSIS I

NIKITA AGARWAL

Abstract. These are lecture notes for the course taught in Semester 1, 2024-25.

1. Week 1
In this week, we will cover the first bullet: Complex numbers – powers and roots,
geometric representation, stereographic projection

1.1. The set of complex numbers. Define the set of complex numbers, denoted as C,
as the collection of all ordered tuples (a, b), where a, b ∈ R. The operations of addition
and multiplication are given by
(a, b) + (c, d) = (a + c, b + d)
(a, b)(c, d) = (ac − bd, ad + bc).
Verify that C is a field with additive identity (0, 0) and multiplicative identity (1, 0).
The map a 7→ (a, 0) is an injective map from R into C. Thus we can consider R as a
subset of C.
Further, putting i = (0, 1), we have (a, b) = a(1, 0) + b(0, 1) = a + bi. Observe that
i 2 = −1 so the equation x2 + 1 = 0 has a root in C but not in R. From now on, a
complex number will be denoted as a + bi instead of an ordered tuple. Generally we use
the variable z for a complex number. Observe that
(a + bi)(c + di) = ac + bdi 2 + (ad + bc)i = (ac − bd) + (ad + bc)i.
For a complex number z = x + yi, we define
• x to be the real part of z, denoted by Re(z) and y to be the imaginary part of z,
denoted by Re(z).
• |z| to be the absolute value of z, given by p the Euclidean distance between the
2
points (0, 0) and (x, y) in R . That is, |z| = x2 + y 2 .
• z to be the conjugate of z, given by x − yi.
Observe that
|z|2 = zz.
Q. Let R(z) be a rational function of z. Prove that R(z) = R(z) if all the coefficients of
R(z) are real.

1.2. Geometry of complex numbers. Each complex number z = x + yi can be iden-

tified with the tuple (Re(z), Im(z)) = (x, y) ∈ R2 . The geometry of R2 is thus used to
visualize complex numbers. We have the following (prove each one of them):
• The sum z + w of two complex numbers z, w is the fourth vertex of the parallel-
ogram with other three vertices 0, z and w.

Date: November 20, 2024.

1
2 NIKITA AGARWAL

• Parallelogram Law:
|z + w|2 + |z − w|2 = 2(|z|2 + |w|2 ).
That is, the sum of squares of the lengths of the sides of a parallelogram is equal
to the sum of the squares of the lengths of its diagonals.
• Triangle inequality:
|z + w| ≤ |z| + |w|.
Consequently,
||z| − |w|| ≤ |z − w|.
1.3. Polar representation and roots of complex numbers. Each complex number
z = x + yi can be expressed in terms of polar coordinates (r, θ), where x = r cos θ and
y = r sin θ. Observe that r equals the absolute value |z| of z and θ is the angle between
the positive real axis and the line segment joining 0 and z. Note that θ plus any multiple
of 2π will also work. The angle θ is called the argument of z, denoted as arg(z) (it is not
a function). We introduce the notation
eiθ = cos θ + i sin θ.
Then z = reiθ . Also note that
eiθ1 eiθ2 = (cos θ1 cos θ2 − sin θ1 sin θ2 ) + i(cos θ1 sin θ2 + sin θ1 cos θ2 )
= cos(θ1 + θ2 ) + i sin(θ1 + θ2 )
= ei(θ1 +θ2 ) .
and
e−iθ = cos(−θ) + i sin(−θ)
= cos θ − i sin θ
= eiθ .
Suppose zj = rj eiθj , for j = 1, 2. Then
z1 z2 = r1 r2 eiθ1 eiθ2 = r1 r2 ei(θ1 +θ2 ) .
Using induction, we have
z1 . . . zn = r1 . . . rn ei(θ1 +···+θn ) .
Setting zj = z = reiθ for all j, we get
z n = rn einθ ,
for all n ≥ 0.
Also, if z 6= 0, then r 6= 0 and
1 iθ
−1 e
z −1 = iθ
= r i(θ
= r−1 e−iθ .
re |e |
Thus, for all n ∈ Z, we have
z n = rn einθ .
This formula is known as the de Moivre’s formula.
We can now ask the following question: Given a complex number z 6= 0 and an integer
n ≥ 2, does there exist a complex number w such that wn = z? How many such w exist?
Let z = reiθ and w = Reiψ , then
Rn = r, nψ = θ + 2kπ, k ∈ Z.
 LECTURE NOTES ON COMPLEX ANALYSIS I 3

by the de Moivre’s formula. Hence w equals

r1/n ei(θ+2kπ)/n , 0 ≤ k ≤ n − 1.
We restricted the values of k to 0 to n − 1 since for other values of k, w equals to one of
the complex numbers from the above list. Each of these numbers satisfying wn = z are
known as the nth roots of z. We have just arrived at the following result.
Theorem 1.1. For each non-zero number z ∈ C, there are n distinct nth roots of z.
Observe that these roots are equally spaced on the circle centred at the origin with
radius r1/n .
Stereographic projection. In complex analysis, often we will deal with functions which
become infinite as the variable approaches a given point. We introduce the concept of
C∞ = C ∪ {∞} to deal with this situation. We also define a concept of distance on C∞
to discuss continuity of functions. We will now identify C∞ with the unit sphere S 2 in
R3 .
Let N = (0, 0, 1) be the north pole of S 2 . Identify C with the xy-plane, that is
x + iy ↔ (x, y, 0). Then C cuts S 2 along the equator. Join each w ∈ C with the north
pole N through a straight line. This line intersects S 2 at a unique point W 6= N . We
identify the point w ∈ C with the point W ∈ S 2 .
If |w| > 1, W lies in the northern hemisphere (positive z-coordinate). If |w| < 1, W
lies in the southern hemisphere (negative z-coordinate). If |w| = 1, W = w. We have
thus identified C with S 2 \ {N }.
The question now is what happens when |w| → ∞. Clearly W → N . We identify N
with ∞ ∈ C∞ . Thus C is represented as the sphere S 2 . We call C∞ as the extended
complex plane. It is also known as the Riemann sphere, after G.F.B. Riemann
(1826-1866). Here is the identification of C∞ with S 2 :
• Given w = x + iy ∈ C, W = (x1 , y1 , z1 ) ∈ S 2 , where
2x w+w 2y −i(w − w) |w|2 − 1
x1 = = , x2 = = , x3 = . (1)
|w|2 + 1 |w|2 + 1 |w|2 + 1 |w|2 + 1 |w|2 + 1
This is obtained by considering the line tN + (1 − t)z joining N and w, identifying
t for which tN + (1 − t)w ∈ S 2 .
• Given W = (x1 , y1 , z1 ) ∈ S 2 \ {N }, we get
x1 + iy1
w= .
1 − z1
This is obtained by considering the line tN +(1−t)W joining N and W , identifying
t for which tN + (1 − t)W for which the third coordinate is 0. This gives t =
−z1 /(1 − z1 ).
The map from C to S 2 \ {N } sending w to W and its inverse are continuous and thus
preserve all the topological properties such as convergence and compactness.
Let us now define distance between two points in C∞ . For w, w0 ∈ C, define the
distance between them as the distance between the corresponding W = (x1 , y1 , z1 ) and
W 0 = (x01 , y10 , z10 ) in R3 (usual distance). Let us look at the exact form. For w, w0 ∈ C,
p
dC∞ (w, w0 ) = (x1 − x01 )2 + (y1 − y10 )2 + (z1 − z10 )2
p
= 2 − 2(x1 x01 + y1 y10 + z1 z10 ) since W, W 0 ∈ S 2
2|w − w0 |
= p (using (1)).
(|w|2 + 1)(|w0 |2 + 1)
4 NIKITA AGARWAL

Moreover, for w ∈ C,
2
dC∞ (w, ∞) = p .
1 + |w|2
This distance metric is known as spherical metric.
Here is an interesting fact about stereographic projection.
Theorem 1.2. A circle on S 2 not passing through N corresponds to a circle on C.
Moreover, any circle on S 2 passing through N corresponds to a straight line in C.
Proof. Any circle on S 2 is a collection of points on S 2 when intersected by a plane. That
is, a circle on S 2 is the collection
{(x1 , y1 , z1 ) ∈ S 2 : Ax1 + By1 + Cz1 + D = 0},
where A2 + B 2 + C 2 6= 0. This circle is identified with the following set of points in C
(using (1)):
{x + iy : 2Ax + 2By + C(x2 + y 2 − 1) + D(x2 + y 2 + 1) = 0},
which is a circle in C provided C + D 6= 0. Note that C + D 6= 0 implies that the circle
on S 2 we started off with does not pass through N . If it does, then C + D = 0 in which
case, the set identified set on C is a straight line. 
 LECTURE NOTES ON COMPLEX ANALYSIS I 5

2. Week 2
The concepts of limits and continuity in C are imagining C as a metric space with
metric dC given by
dC (z1 , z2 ) = |z1 − z2 |,
which is same as the Euclidean distance between (x1 , y1 ) and (x2 , y2 ) in R2 where zj =
xj + iuyj , for j = 1, 2.
Definition 2.1. Let G be an open subset of C. Let f : G → C be a function. Let z0 ∈ G.
We say that f is differentiable at z0 if the limit
f (z) − f (z0 )
lim
z − z0
z→z0

exists. The limit, if it exists, is known as the derivative of f at z0 and is denoted by

f 0 (z0 ).

 − δ definition: For given  > 0, there exists δ > 0 such that

f (z) − f (z0 )
− f 0 (z0 ) < ,
z − z0
whenever 0 < |z − z0 | < δ.
Theorem 2.2. We have the following properties for differentiability.
• If f is differentiable at z0 , then f is continuous at z0 .
• (Sum and product rue) If f, g are differentiable at z0 , then f + g and f g also are,
and
(f + g)0 (z0 ) = f 0 (z0 ) + g 0 (z0 ), (f g)0 (z0 ) = f (z0 )g 0 (z0 ) + f 0 (z0 )g(z0 ).
(Quotient rule) In addition, if g(z0 ) 6= 0, then
 0
f f 0 (z0 )g(z0 ) − f (z0 )g 0 (z0 )
(z0 ) = .
g g(z0 )2
• (Chain rule) If f is differentiable at z0 and g is differentiable at f (z0 ), then g ◦ f
is differentiable at z0 and
(g ◦ f )0 (z0 ) = g 0 (f (z0 ))f 0 (z0 ).
Proof. We just prove the chain rule. We consider two cases: i) f 0 (z0 ) 6= 0 and f 0 (z0 ) = 0.
Let us look at the case when f 0 (z0 ) 6= 0. Then there exists  > 0 such that f (z) 6= f (z0 )
for all 0 < |z − z0 | < . Thus for all z satisfying 0 < |z − z0 | < , we can write
g ◦ f (z) − g ◦ f (z0 ) g ◦ f (z) − g ◦ f (z0 ) f (z) − f (z0 )
= .
z − z0 f (z) − f (z0 ) z − z0
Since f is differentiable at z0 , it is continuous at z0 . Therefore f (z) → f (z0 ). Hence,
since g is differentiable at f (z0 ), as z → z0 , we have
g ◦ f (z) − g ◦ f (z0 )
→ g 0 (f (z0 )).
f (z) − f (z0 )
Also using the fact that f is differentiable at z0 , we prove the chain rule.
Consider the other case when f 0 (z0 ) = 0. Since g is differentiable at f (z0 ), there exists
a constant C > 0 and  > 0 such that
|g(w) − g(f (z0 ))|
≤ C,
|z − f (z0 )|
6 NIKITA AGARWAL

for all w satisfying 0 < |w − f (z0 )| < . Also since f is continuous at z0 , there exists
δ > 0 such that
|f (z) − f (z0 )| < , whenever |z − z0 | < δ.
Combining the above, for all |z − z0 | < δ, we have
|g(f (z)) − g(f (z0 ))| ≤ C|f (z) − f (z0 )|.
Therefore
g ◦ f (z) − g ◦ f (z0 ) f (z) − f (z0 )
≤C .
z − z0 z − z0
We have proved the chain rule in this case as well. 
Example 2.3. It is immediate from the definition of derivative that the constant function
is differentiable everywhere with derivative 0 and the identity function is differentiable
everywhere with derivative 1.
By the application of product rule, for each n ∈ N, the function z 7→ z n is differentiable
everywhere with derivative nz n−1 .
Further, by the application of sum rule, any polynomial function z 7→ a0 + a1 z + a2 z 2 +
· · · + an z n is differentiable everywhere with derivative a1 + 2a2 z · · · + nan z n−1 .
p(z)
Moreover, any rational function z 7→ , where p, q are polynomial functions is
q(z)
defined provided q(z) 6= 0. Hence by the application of product rule, any rational function
is differentiable at all the points where it is defined.
Differentiability in complex and real sense. Suppose f = u+iv is a complex-valued
function defined an open set G ⊆ C, where u, v are its real and imaginary parts. For a
given point z0 = x0 + iy0 and a complex number c = a + ib, suppose f is differentiable at
z0 with f 0 (z0 ) = c.
Approaching z0 along the real axis, we get
f (z0 + h) − f (z0 )
c = lim
h→0, h∈R h
u(x0 + h, y0 ) − u(x0 , y0 ) v(x0 + h, y0 ) − v(x0 , y0 )
= lim + i lim
h→0, h∈R |h| h→0, h∈R |h|
= ux (x0 , y0 ) + ivx (x0 , y0 ).
Approaching z0 along the imaginary axis, we get
f (z0 + ih) − f (z0 )
c = lim
h→0, h∈R ih
u(x0 , y0 + h) − u(x0 , y0 ) v(x0 , y0 + h) − v(x0 , y0 )
= −i lim + lim
h→0, h∈R h h→0, h∈R h
= −iuy (x0 , y0 ) + vy (x0 , y0 ).
Thus we have the following result.
Theorem 2.4. A complex-valued function f = u + iv is differentiable at z0 if and only
if its real and imaginary parts u and v are differentiable at z0 with
ux (z0 ) = a, uy (z0 ) = −b, vx (z0 ) = b, vy (z0 ) = a.
Hence
f 0 (z0 ) = ux (z0 ) + ivx (z0 ) = vy (z0 ) − iuy (z0 ).
 LECTURE NOTES ON COMPLEX ANALYSIS I 7

Definition 2.5. (Cauchy-Riemann equations) The following two partial differential equa-
tions for the pair of real-valued functions are known as Cauchy-Riemann equations
ux = vy , uy = −vx .
Theorem 2.6. (Sufficient condition for differentiability) A complex-valued function f =
u + iv is differentiable at z0 if u and v exist and are continuous at z0 and satisfy the
Cauchy-Riemann equations at z0 .
The results follows from the following result from calculus of real functions: A real-
valued function is differentiable at a point x ∈ R2 if its partial derivatives exist and are
continuous at x ∈ R2 .
Definition 2.7. A complex valued function in an open subset G of C is said to be
holomorphic (or analytic) if it is differentiable at every point in G.
Theorem 2.8. A complex-valued function f = u + iv is differentiable at z0 if u and v
are differentiable at z0 and the Cauchy-Riemann equations are satisfied at z0 .
Example 2.9. Let the function f be holomorphic in the open set G. Prove that the
function g(z) = f (z) is holomorphic on the set G = {z | z ∈ G}.
Solution: Let z0 ∈ G. Then

g(z) − g(z0 ) f (z) − f (z0 )

=
z − z0 z − z0
f (z) − f (z0 )
=
z − z0
 
f (z) − f (z0 )
= .
z − z0

As z → z0 in G, we have z → z0 in G. Thus the RHS goes to f 0 (z0 ) since f is differentiable

at z0 . Hence
g 0 (z0 ) = f 0 (z0 ).
Theorem 2.10. (From multivariable calculus) Let G be an open and connected subset
of R2 . Let f = f (x, y) : G → R be a differentiable function with its partial derivatives
satisfying fx = fy = 0 at all points of G. Then f is a constant function.
Theorem 2.11. Let G be an open and connected subset of C. Let f : G → C be a
differentiable function with f 0 (z) = 0 for all z ∈ G. Then f is a constant function.
Proof. From the calculations above,
f 0 (z) = ux (x0 , y0 ) + ivx (x0 , y0 ) = vy (x0 , y0 ) − iuy (x0 , y0 ).
Hence ux = uy = vx = vy = 0 at (x0 , y0 ). Therefore u, v are constant functions. Thus f
is a constant function. 
Theorem 2.12. Let G be an open and connected subset of C. Let f : G → C be a
differentiable function with f (z) ∈ R for all z ∈ G. Then f is a constant function.
Proof. Here v = 0. Therefore by the CR equations, ux = uy = 0. Therefore u is a
constant function. Hence f is a constant function. 
8 NIKITA AGARWAL

Power series.
Definition 2.13. A power series is an expression of the form
X∞
an (z − a)n ,
n=0
where an , a, z ∈ C.
Goal: To find values of z ∈ C for which the power series converges.
Theorem 2.14. Let ∞ n
P
n=0 an (z − a) be a power series. Define R ∈ [0, ∞] (extended real
number) as
R−1 = lim sup |an |1/n .
Then
(1) For all z ∈ C with |z − a| < R, the series converges absolutely and uniformly to a
function f in B(a, r) for any 0 < r < R.
(2) If R ∈ (0, ∞], then f is continuous and differentiable in B(a, R) with
∞
X
0
f (z) = nan (z − a)n−1 .
n=1

(3) For all z ∈ C with |z − a| > R, the series diverges.

For proof, see Theorem 1.3 of Conway, 2nd Edition. Note that this result says nothing
about convergence on the circle |x − a| = R if 0 < R < ∞.
Definition 2.15. The P∞extended realn number R ∈ [0, ∞] is called the radius of convergence
of the power series n=0 an (z − a) .
Theorem 2.16. If a power series ∞ n
P
n=0 an (z − a) has a positive radius of convergence
0 < R ≤ ∞, then its sum defines a function f in the ball B(a, R). The function f is
infinitely differentiable in this ball. For each k ≥ 1,
X∞
(k)
f (z) = n(n − 1) . . . (n − k + 1)an (z − a)n−k .
n=k
(n)
f (a)
Moreover, an = .
n!
See Proposition 2.5 of Conway, 2nd Edition for a proof.
Theorem 2.17. Let ∞ n
P
n=0 an (z −a) be a power series. Assume that one of the following
hold:
i) limn |an+1 /an | = `, for some ` ∈ [0, ∞].
ii) limn |an |1/n = `, for some ` ∈ [0, ∞].
Then the radius of convergence of the given power series is given by R = 1/`.
Use Proposition 1.4 of Conway, 2nd Edition to prove this.
Example 2.18. Discuss convergence of the power series
X∞
nn z n .
n=1

This power series is about the point 0. Here an = nn . Since |an |1/n = n → ∞, the radius
of convergence of the power series is given by R = 1/∞ = 0. Thus the power series
diverges for all z ∈ C other than 0.
 LECTURE NOTES ON COMPLEX ANALYSIS I 9

Example 2.19. Discuss convergence of the power series

∞
X zn
.
n=0
n!
This power series is about the point 0. Here an = 1/n!. Since
an+1 1
= → 0,
an n+1
the radius of convergence of the power series is given by R = 1/0 = ∞. Thus the power
series is convergent on all of C.
Example 2.20. Discuss convergence of the power series
∞
X n3
n
(z − 1)n .
n=0
3
n3
This power series is about the point 1. Here an = 3n
. Since
an+1 (n + 1)3 3n 1
= → ,
an 3n+1 n3 3
the radius of convergence of the power series is given by R = 3. Thus the power series is
convergent for all z with |z − 1| < 3 and divergent for all z with |z − 1| > 3.
Need to look at the boundary |z − 1| = 3. That is, z = 1 + 3eiθ , θ ∈ [0, 2π). Thus the
series becomes ∞
X
n3 einθ .
n=0
th
The n term does not go to 0 for any θ. Hence the series diverges at each point on the
boundary.
Example 2.21. Discuss convergence of the power series
∞
X zn
.
n=1
n
Here an = 1/n. Since |an+1 /an | → 1, the radius of convergence is 1.
Need to look at the boundary |z| = 1. That is, z = eiθ , θ ∈ [0, 2π). Thus the series
becomes ∞
X einθ
.
n=0
n
The series diverges when θ = 0. For nonzero θ, we will use the Dirichlet’s test: If (an ) is a
real sequence and (bn ) is a sequence of complex numbers satisfying: i) (an ) is a decreasing
PN
sequence converging
P to 0, ii) There exists a constant M > 0 such that | n=1 bn | ≤ M ,
for all N . Then an bn converges.
Let us consider the above series on the boundary. Let θ ∈ (0, 2π) be a fixed angle. Let
an = 1/n and bn = einθ in the Dirichlet’s test. Then
N
X 1 − ei(N +1)θ 2
| bn | = ≤ = M.
n=0
1 − eiθ 1 − eiθ
Therefore, by the Dirichlet’s test, the series converges.
Since M diverges as θ approaches 0, the convergence of the series to a function on the
boundary with the angle θ = 0 removed is not uniform.
10 NIKITA AGARWAL

3. Week 3
Principal value and branches of arg(z).
• Recall that for z ∈ C \ {0}, if θ is the angle between the positive real axis and the
line joining 0 and z, then arg(z) takes values θ + 2nπ, for all n ∈ Z.
• Let G be an open and connected subset of C not containing the origin 0. By a
branch of arg(z), we mean a continuous function α on G such that for all z ∈ G,
α(z) is a value of arg(z).
• We call the particular value of arg(z) in (−π, π] the principal value of arg(z).
P∞ z n
Exponential and trigonometric functions. Consider the power series n=0 . It
n!
is better to write this series as
∞
X zn
1+ ,
n=1
n!
in which case, there is no issue at z = 0. Otherwise use the convention: 00 = 1.
The radius of convergence of this series is ∞. Hence the series converges at each complex
number z and converges absolutely and uniformly to a function in each compact subset
of C. Let us denote the series by ez = exp(z). This is known as the exponential series or
the exponential function. Here are some important properties of this function.
• We know that the function ez is infinitely differentiable with (using term by term
differentiation)
d z
e = ez .
dz
• For fixed a ∈ C, let g(z) = ez ea−z , for z ∈ C. Then g 0 (z) = 0 (by the product
rule), for all z ∈ C. Hence g is a constant function. Since g(0) = ea , we have
g(z) = ea , for all z ∈ C. Therefore, for all a, b ∈ C, by replacing a by a + b and z
by b, we have
ea+b = ea eb .
• Setting b = −a, we get 1 = ea e−a . Hence ez 6= 0, for all z ∈ C. Moreover
(ez )−1 = e−z .
• Since all the coefficients of the power series are real, we have
ez = ez .
Thus for purely imaginary z,
ez = e−z .
Set z = iθ, we have
|ez |2 = ez ez = 1.
More generally,
|ez |2 = ez ez = ez ez = e2Re(z) .
Thus
|ez | = eRe(z) .
• Let z = x + iy, then
ez = ex eiy = ex (cos y + i sin y).
 LECTURE NOTES ON COMPLEX ANALYSIS I 11

As for real functions, we define

∞ 2n ∞
X
n z
X z 2n+1
cos z = (−1) , sin z = (−1)n .
n=0
(2n)! n=0
(2n + 1)!
Each of these functions are infinitely differentiable since the series have infinite radius of
convergence. By term by term differentiation, we have
(cos z)0 = − sin z, (sin z)0 = cos z.
Here are some more properties.
• We have
eiz + e−iz eiz − e−iz
cos z = , sin z = .
2 2i
Thus cos2 z + sin2 z = 1, for all z ∈ C.
• We have
eiz = cos z + i sin z.
Thus, for θ ∈ R,
eiθ = cos θ + i sin θ (relationship b/w exponential and real trigonometric.)
We have recovered our earlier notation. Hence if z = x + iy, we have
|ez | = ex , arg(ez ) = Im(z).
A complex-valued function f is said to be periodic with period c ∈ C if f (z) = f (z +c),
for all z ∈ C. Observe that if we divide the complex plane into infinitely many horizontal
strips by the lines Im(z) = π(2n − 1), n ∈ Z, the exponential function behaves the same
in each of the strips. That is, ez = ez+2πi , for all z ∈ C.
Logarithm function. In real function theory, we define logarithm as the inverse of
exponential. This is possible since real exponential is an invertible map. In the complex
case, it is not. Let us see what happens if we define log z for z ∈ C so that it satisfies
ew = z, when w = log z. Since ew is never 0, we cannot define log z for z = 0. Now let
z ∈ C \ {0} and w = x + iy. Then
|z| = ex , arg(z) = y.
Hence x = log |z| and y = arg(z) + 2πn, n ∈ Z. Therefore the solution set of ew = z is
w∈
{log |z| + i(arg(z) + 2πn), n ∈ Z}.
Note here that log |z| is the usual real logarithm.
Definition 3.1. Let G be an open and connected subset of C. Let f : G → C be a
continuous function such that z = exp f (z), for all z ∈ G. Then f is known as a branch
of the logarithm.
We have the following result.
Theorem 3.2. Let G be an open and connected subset of C. Let f : G → C be a branch
of the logarithm log z on G. Then all the branches of the logarithm log z on G are given
by {f + 2πni, n ∈ Z}.
Proof. (The proof was not discussed in the class. Ask me if your are unable to follow.)
Firstly, f + 2πni is a branch of logarithm on G for each n ∈ Z.
Conversely, let g be another branch of the logarithm on G, then for all z ∈ G,
exp g(z) = z = exp f (z).
12 NIKITA AGARWAL

Hence for each z ∈ G, g(z) = f (z) + 2nπi. The n may depend on z ∈ G. We now prove
that the n is independent of z ∈ G. For this, consider the function h : G → Z defined as
f (z) − g(z)
h(z) = , z ∈ G.
2πi
Then h is a continuous function, since f, g are. Moreover h(z) ∈ Z for all z ∈ G. Since
G is connected, the image of h is connected in Z. The only connected subsets of Z are
singletons. Hence h is a constant function, let h(z) = m, for all z ∈ G. Thus
g(z) = f (z) + 2mπi, z ∈ G.

Let us see one branch of log z on some connected open subset of C. Let G = C \ {z ∈
R : z ≤ 0} (Slit the plane along the negative real axis of C). Let z = reiθ , where
θ ∈ (−π, π). Then define
f (z) = log r + iθ.
Prove that f is continuous on G. Hence f is a branch of the logarithm on G. This is
known as the principal branch of the logarithm.
Note that by the chain rule, we have the following result.
Theorem 3.3. A branch of the logarithm function is differentiable and its derivative is
1/z.
Proof. Differentiability follows by verifying CR equations in polar form (see HW 2).
Since (ew )0 = ew , for all w ∈ C and ef (z) = z for all z ∈ G, we have ef (z) f 0 (z) = 1, which
implies f 0 (z) = 1/z, for all z ∈ G. 
Complex powers. Let f be a branch of the logarithm on a connected open set G ⊂ C.
Let b ∈ C is a fixed complex number. Define g : G → C as
g(z) = exp(bf (z)).
If b ∈ Z, then
g(z) = (exp(f (z)))b = z b .
With this as the reference, we define a branch of z b as the function
z b = exp(b log z),
on an open connected set G on which there is a branch of log z. Hence
log(z b ) = b log z.
Square root function. For b = 1/2 in z b , we get the square root function. Observe
that, for the principal branch of the logarithm, for z = reiθ ,
1 1
z 1/2 = exp( log z) = exp( (log r + iθ)) = r1/2 eiθ/2 .
2 2
Hyperbolic functions. Using the exponential function, we can define other functions
known as hyperbolic functions as follows:
ez + e−z ez − e−z
cosh z = , sinh z = ,
2 2
sinh z cosh z
tanh z = , coth z = ,
cosh z sinh z
1 1
sech z = , csch z = .
cosh z sinh z
 LECTURE NOTES ON COMPLEX ANALYSIS I 13

The above defined functions are differentiable at the points wherever they are defined.
Let us understand the first two functions in more detail. Note that sinh z is 0 when
ez + e− z = 0. That is, e2z = 1, which implies that 2z is an integer multiple
 of 2πi.
 Thus,
1
the zeros of sinh z are nπi, for n ∈ Z. Similarly, the zeros of cosh z are n + πi, for
2
n ∈ Z.
Example 3.4. Let us compute all the values of (1 + i)i . We know that
(1 + i)i = exp(i log(1 + i)).
√ iπ/4
Since 1 + i = 2e , hence
√
log(1 + i) = log 2 + i(π/4 + 2πn), n ∈ Z.
Thus
√ √
exp(i log(1 + i)) = exp(−(π/4 + 2πn) + i log 2) = exp(−(π/4 + 2πn))exp(i log 2).
Therefore
 √ √ 
(1 + i)i = exp(−(π/4 + 2πn)) cos log 2 + i sin log 2 , n ∈ Z.

Möbius transformations.
Definition 3.5. A linear fractional transformation is a nonconstant function T of
the form
az + b
T (z) = ,
cz + d
where a, b, c, d ∈ C are constants. If ad − bc 6= 0, T is known as a Möbius transforma-
tion.
Theorem 3.6. The set of Möbius transformations form a group under composition. We
denote it as M.
Proof. We check the properties of a group.
• (Closure) Let T, S be two Möbius transformations given by
az + b a0 z + b 0
T (z) = , S(z) = 0 .
cz + d c z + d0
Then
a0 z + b 0
a0 0
+b (aa0 + bc0 )z + (ab0 + bd0 ) Az + B
T ◦ S = c0 z + d0 = 0 0 0 0
= ,
az+b (ca + dc )z + (cb + dd ) Cz + D
c 0 +d
c z + d0
with
AD − BC = (aa0 + bc0 )(cb0 + dd0 ) − (ab0 + bd0 )(ca0 + dc0 ) = (ad − bc)(a0 d0 − b0 c0 ) 6= 0.
Hence T ◦ S is a Möbius transformation.
• (Identity) Consider the Möbius transformation I(z) = z. It is easy to see that
T ◦ I = I ◦ T = T , for each Möbius transformation T .
• (Inverse) Give T as above, we need to check if there is a Möbius transformation
S such that T ◦ S = S ◦ T = I. Using the arguments above, we need to find
a0 , b0 , c0 , d0 such that a0 d0 − b0 c0 6= 0 and A = D = 1, B = C = 0. That is,
aa0 + bc0 = 1 = cb0 + dd0 , ab0 + bd0 = 0 = ca0 + dc0 .
14 NIKITA AGARWAL

This can be written as

   0    0  
a b 0 0 a 1 a d
 c d 0 0  c0  0  c0  1 −c
0 0 c d  b0  = 1 ⇒  b0  = ad − bc −b .
        
0 0 a b d0 0 d0 a
ad − bc
Clearly a0 d0 − bc = 6= 0. We have that
(ad − bc)2
dz − b
S(z) = ,
−cz + a
satisfies T ◦ S = I. Such a S is unique by the above arguments. Also check
S ◦ T = I. Hence S is an inverse of T .

Relation
 with matrices. Consider GL(2, C), the group of all invertible matrices.
  0 Each
a b0
  
a b a b
∈ GL(2, C) gives a Möbius tranformation T as above. Let , 0 0 ∈
c d c d c d
GL(2, C) with associated Möbius transformations T, S as above. Then the matrix product
  0 0  0
aa + bc0 ab0 + bd0
 
a b a b
=
c d c0 d0 ca0 + dc0 cb0 + dd0
gives the Möbius transformation T ◦ S. Hence the map
Φ : GL(2, C) → M
defined as  
a b az + b
Φ (z) = ,
c d cz + d
is a group homomorphism.
(An important observation) The preimage of T consists of all matrices of the form
   
a b
λ : λ ∈ C \ {0} .
c d
 LECTURE NOTES ON COMPLEX ANALYSIS I 15

4. Week 4
Recall the definition of Mobius transformations from the last week.

az + b αz + β
Theorem 4.1. (see HW 2) Let Sz = and T z = . Show that S = T if and
cz + d γz + δ
only if there is a non-zero complex number λ such that a = λα, b = λβ, c = λγ, a = λδ.
Consider a Möbius transformation T of the form
az + b
T (z) = .
cz + d
We can view T as a mapping from C∞ to C∞ with
T (∞) = a/c, T (−d/c) = ∞.
Note that a = c = 0 or d = c = 0 is not possible since ad − bc 6= 0.
Definition 4.2. There are some special types of Möbius transformations.
• Translation: T (z) = z + b.
• Dilation: T (z) = az with a 6= 0.
• Rotation: T (z) = eiθ z, for a real constant θ (special dilation).
• Inversion: T (z) = 1/z.
Theorem 4.3. A Möbius transformation is a composition of translations, dilations and
the inversion.
Proof. If c = 0, then d 6= 0. We are done since T (z) = (a/d)z + b/d.
Suppose c 6= 0. Let T1 (z) = z + d/c, T2 (z) = 1/z, T3 (z) = ((bc − ad)/c2 ) z, and T4 (z) =
z + a/c. Then T = T4 ◦ T3 ◦ T2 ◦ T1 . 
Fixed points of T . We need to solve for T (z) = z. That is solve for z in
cz 2 + (d − a)z − b = 0.
Hence T can have at most two fixed points unless T is the identity transformation (a =
1, b = 0, c = 0, d = 1).
Let T be a Möbius transformation and a, b, c be three distinct points in C∞ . Let
T (a) = α, T (b) = β and T (c) = γ. Let S be another map Möbius transformation with
S = T on the set {a, b, c}. Then S −1 ◦ T has three (distinct) fixed points a, b, c. Therefore
S −1 ◦ T = I. Hence S = T . Hence we have the following result.
Theorem 4.4. A Möbius transformation is uniquely determined by its image on three
distinct points in C∞ .
Let z2 , z3 , z4 be three distinct points in C∞ . The unique Möbius transformation T such
that
S(z2 ) = 1, S(z3 ) = 0, S(z4 ) = ∞
is given by
   
z − z3 z2 − z3
S(z) = / , if z2 , z3 , z4 ∈ C,
z − z4 z2 − z4
z − z3
S(z) = , if z2 = ∞,
z − z4
z2 − z4
S(z) = , if z3 = ∞,
z − z4
z − z3
S(z) = , if z4 = ∞.
z2 − z3
16 NIKITA AGARWAL

Definition 4.5. (Cross ratio) Let z1 ∈ C∞ . The image of z1 under the Möbius trans-
formation S with S(z2 ) = 1, S(z3 ) = 0, S(z4 ) = ∞, is known as the cross ratio of
z1 , z2 , z3 , z4 . It is denoted as (z1 , z2 , z3 , z4 ).
For instance, (z2 , z2 , z3 , z4 ) = 1 and (z3 , z2 , z3 , z4 ) = 0.
Theorem 4.6. Let z2 , z3 , z4 be three distinct points in C∞ . Let T be any Möbius trans-
formation. Then
(z1 , z2 , z3 , z4 ) = (T z1 , T z2 , T z3 , T z4 ),
for all z1 ∈ C∞ .
Proof. Let Sz = (z1 , z2 , z3 , z4 ) and S1 z = (z, T z2 , T z3 , T z4 ). Then S = S1 ◦ T by Theo-
rem 4.4. Hence
(z1 , z2 , z3 , z4 ) = Sz1 = S1 ◦ T z1 = S1 (T z1 ) = (T z1 , T z2 , T z3 , T z4 ).

Theorem 4.7. (Three-fold transitivity) Let z1 , z2 , z3 be three distinct points of C∞ and let
w1 , w2 , w3 be three distinct points of C∞ . Then there is a unique Möbius transformation
which takes zj to wj for j = 1, 2, 3.
Proof. Let T z = (z, z1 , z2 , z3 ) and Sz = (z, w1 , w2 , w3 ), for all z ∈ C. Then S −1 ◦ T works.
Uniqueness follows from Theorem 4.4. 
Definition 4.8. (Clircle) By a clircle, we mean a straight line in C together with the
point ∞, or a circle in C.
Note that a clircle in C∞ is the image of a circle on S 2 under the stereographic projec-
tion.
Theorem 4.9. A Mobius transformation maps clircles onto clircles.
Proof. The result is clearly true for a translation, dilation, and a rotation. We will just
prove it for the inversion.
The equation of a circle in C is given by: for some a ∈ C and r > 0,
|z − a| = r ⇒ |z|2 + βRe(z) + γIm(z) + δ = 0,
for suitable constants β, γ, δ ∈ R.
The equation of a straight line in C is of the form: for z = x + iy,
y = mx ⇒ βRe(z) + γIm(z) + δ = 0,
for suitable constants β, γ, δ ∈ R. Note that the lines passes through the origin if and
only if δ = 0.
Thus the equation of a clircle is of the form:
α|z|2 + βRe(z) + γIm(z) + δ = 0,
for suitable constants α, β, γ, δ ∈ R.
The image of this clircle under inversion is
α/|z|2 + βRe(z)/|z|2 − γIm(z)/|z|2 + δ = 0 ⇒ δ|z|2 + βRe(z) − γIm(z) + α = 0,
which is again a clircle. 
Theorem 4.10. Let z1 , z2 , z3 , z4 be four distinct points in C∞ . Then (z1 , z2 , z3 , z4 ) is a
real number if and only if the four points lie on a clircle.
Proof. Let S be the MT defined as Sz = (z, z2 , z3 , z4 ), for all z ∈ C. Since a MT maps
clircles onto clircles, the set of points {z1 , z2 , z2 , z4 } lie on a clircle if and only if the set of
points {Sz1 , Sz2 , Sz3 , Sz4 } = {Sz1 , 1, 0, ∞} lie on a clircle if and only if Sz1 ∈ R (since
the clircle containing the points 1, 0, ∞ is the extended real line R∞ ). 
 LECTURE NOTES ON COMPLEX ANALYSIS I 17

Conformality.
Definition 4.11. A curve in C is a continuous function γ : [a, b] → C. We will think of
curve as a subset of C, that is, {γ(t) t ∈ [a, b]} is a curve.
We say that a curve is differentiable at t0 ∈ (a, b) if both real and imaginary compo-
nents of γ are differentiable at t0 . That is,
γ(t) − γ(t0 )
lim
t→t0 t − t0
0
exists. Denote this limit as γ (t0 ).
We say that a curve is regular at t0 ∈ (a, b) if it is differentiable at t0 and γ 0 (t0 ) 6= 0.
If γ is regular at t0 , then it has a well-defined direction at t0 , by specifying the argument
of γ 0 (t0 ).
Suppose γ1 and γ2 be two curves in C. Suppose they intersect at a point γ1 (t1 ) = γ2 (t2 ).
Also suppose that γj is regular at tj , for j = 1, 2. The angle between γ1 and γ2 is defined
as
argγ20 (t2 ) − argγ10 (t1 ) = argγ20 (t2 )γ10 (t1 ).
The angle depends on the order in which we take the curves.
Let us look at the following observations. Let f be a holomorphic function on an open
set G in C. Let γ be a curve in G. Let γ be regular at t0 and f 0 (z0 ) 6= 0 for z0 = γ(t0 ).
Then
(f ◦ γ)0 (t0 ) = f 0 (z0 )γ 0 (t0 ) 6= 0.
Hence f ◦ γ is also regular at t0 . Moreover
arg(f ◦ γ)0 (t0 ) = argf 0 (z0 ) + argγ 0 (t0 ).
Thus we have the following result.
Theorem 4.12. Let f be a holomorphic function on an open set G in C. Let z0 ∈ G and
f 0 (z0 ) 6= 0. Let γ1 , γ2 be two curves such that γ1 (t1 ) = γ2 (t2 ) = z0 and that γj is regular
at tj , for j = 1, 2. Then the angle between f ◦ γ1 and f ◦ γ2 equals the angle between γ1
and γ2 .
Proof. For j = 1, 2,
arg(f ◦ γj )0 (tj ) = argf 0 (z0 ) + argγ 0 (tj ).
Hence
arg(f ◦ γ2 )0 (t2 ) − arg(f ◦ γ1 )0 (t1 ) = argγ 0 (t2 ) − argγ 0 (t1 ).

Example 4.13. Consider f (z) = z 2 . Note that f 0 (0) = 0. Consider two curves γ1 :
[0, 1] → C as t 7→ t and γ2 : [0, 1] → C as t 7→ teiα for α 6= 0. Clearly γ1 (0) = γ2 (0) = 0.
The angle between γ1 and γ2 at 0 is α, but the angle between f ◦ γ1 and f ◦ γ2 at f (0) = 0
is 2α.
18 NIKITA AGARWAL

5. Week 5
Definition 5.1. A complex-valued function f on an open set G in C is said to be
conformal at the point z0 ∈ G if it preserves the angles between pairs of regular curves
intersecting at z0 .
By the preceding theorem, if f is holomorphic and f 0 (z0 ) 6= 0, then f is conformal. Let
us look at the converse.
Theorem 5.2. Let f be a complex-valued function on an open set G in C, whose real
and imaginary parts have continuous first order partial derivatives. If f is conformal at
each point of G, then f is holomorphic on G and f 0 is never 0 on G.
We will come back to discuss the proof.
Theorem 5.3. A Mobius transformation T is conformal at all points z where T z ∈ C.
az + b ad − bc
Proof. Let T z = be a Mobius transformation. Then T 0 z = 6= 0, when
cz + d (cz + d)2
z 6= −d/c. 
Example 5.4. Let us compute a Mobius transformation mapping 0, 1, ∞ to 1, ∞, −i,
respectively.
Let T be such a map.
• Method 1: You need to know the form of cross-ratio (the map S) to apply this
method.
We know that
1+i z−1
(T z, 1, ∞, −i) = (T z, T −, T 1, T ∞) = (z, 0, 1, ∞) ⇒ = = 1 − z.
Tz + i 0−1
Hence
1+i −i(1 − z) + 1 + i iz + 1
T z = −i + = = .
1−z 1−z 1−z
az + b
• Method 2: Let T z = .
cz + d
Since T 1 = ∞, we can take the denominator of T to be z − 1. That is, c = 1,
d = −1.
az − 1
Since T 0 = 1, we have b = −1. Thus T z = .
z−1
Finally, T (∞) = −i implies a = −i. Hence
−iz − 1
Tz =
z−1
Example 5.5. Consider the MT
z+i
Tz = .
z−i
• Observe that T maps 0 7→ −1, i 7→ ∞ and −i 7→ 0. The clircle determined by
the points 0, i, −i is the extended imaginary axis denoted by iR∞ . The clircle
determined by their images −1, ∞, 0 is the extended real axis R∞ . Hence T takes
the extended imaginary axis to the extended real axis. That is,
T (iR∞ ) = R∞ .
• We will use these labels.
P+ : the right half plane
P− : the left half plane
H+ : the upper half plane
H− : the lower half plane.
 LECTURE NOTES ON COMPLEX ANALYSIS I 19

• Since T is a bijection, we have

T (P+ ) ∪ T (P− ) = H+ ∪ H− .
Moreover since T is continuous and P+ , P− are connected, T (P+ ), T (P− ) are con-
nected. Therefore, both T (P+ ) and T (P− ) are disjoint and are either H+ or H−
(since T (C∞ ) = C∞ ).
• Since T (1) = i, we have
T (P+ ) = H+ , T (P− ) = H− .
• The image of the unit circle S 1 (|z| = 1) under T contains T 1 = i, T i = ∞,
T (−i) = 0, hence
T (S 1 ) = iR∞ .
• Let D denotes the open disk in the complex plane bounded by S 1 and let E be
the extended exterior of S 1 . That is,
E = {z ∈ C : |z| > 1} ∪ {∞}.
By the connectedness argument and onto property of T , since T 0 = −1, we get
T (D) = P− , T (E) = P+ .
• The image of the extended real axis contains the points T 1 = i, T 0 = −1 and
T ∞ = 1. Hence
T (R∞ ) = S 1 .
By the connectedness argument and onto property of T , since T i = ∞, we get
T (H+ ) = E, T (H− ) = D.
• Let us now understand the image of the coordinate grid (horizontal lines and ver-
tical lines) under T .
Consider an extended vertical line L other than the imaginary axis. It is orthog-
onal to the extended real axis R∞ meeting it at ∞ (only). Hence, since T is
conformal, their images T (L) and T (R∞ ) = S 1 are orthogonal meeting at the
point T (∞) = 1. Moreover T (L) being a clircle and ∞ = T (i) ∈ / T (L) implies
T (L) is a circle. Thus T (L) is a circle orthogonal to S 1 at the point 1.
Further, the images T (L) are circles crossing the point 1 and tangent to the real
axis.
Similarly, the images of the extended horizontal lines, except for the extended real
axis, under T are the circles through the point 1 orthogonal to real axis.
Complex integration.
• A function φ : [a, b] → C is said to be piecewise continuous if it is continuous
at all but finitely many points of [a, b] and has one sided limit at each point of
discontinuity.
• Since then both Re φ and Im φ are Riemann-integrable, we define the integral of
φ over [a, b] as
Z b Z b Z b
φ(t) dt = Re φ(t) dt + i Im φ(t) dt.
a a a
Linearity of Riemann integral for real-valued functions implies linearity of integral
for complex-valued functions. That is,
Z b Z b Z b
(αφ + βψ)(t) dt = α φ(t) dt + β ψ(t) dt,
a a a
20 NIKITA AGARWAL

where φ, ψ : [a, b] → C are piecewise continuous functions and α, β ∈ C are

constants.
• We say that a function φ : [a, b] → C is said to be piecewise C 1 if it is continuous, it
is differentiable at all but finitely many points of [a, b], its derivative is continuous
at each point where it exists and its derivative has one-sided limits at each of the
finitely many points where it fails to exist.
• By the fundamental theorem of calculus for real-valued functions, we obtain the
following result:
Theorem 5.6. If a function φ : [a, b] → C is piecewise C 1 , then
Z b
φ0 (t) dt = φ(b) − φ(a).
a

• As for real-valued functions, we have the following result.

Theorem 5.7. If φ : [a, b] → C is piecewise continuous function, then
Z b Z b
φ(t) dt ≤ |φ(t)| dt.
a a
Rb
Proof. If a φ(t) dt = 0, we are done. Suppose it is not zero.
Rb Rb
Let λ = | a φ(t) dt|/ a φ(t) dt. Then
Z b Z b Z b Z b
φ(t) dt = λ φ(t) dt = λφ(t) dt = Re λφ(t) dt
a a a a
Z b Z b Z b
= Re (λφ(t)) dt ≤ |λφ(t)| dt = |φ(t)| dt.
a a a


• Let γ : [a, b] → C be a piecewise C 1 curve. We define the length of γ as
Z b
L(γ) = |γ 0 (t)| dt.
a

This formula is derived from taking supremum over all possible polygonal paths
inscribed in γ.
• Let γ : [a, b] → C be a piecewise C 1 curve. A complex-valued function f is said to
be defined on γ is it is defined on the image set of γ, which is {γ(t) : t ∈ [a, b]}.
In this case, we can define f ◦ γ : [a, b] → C. If f is continuous and defined
R on γ,
then f ◦ γ is continuous. We define the integral of f over γ, denoted by γ f (z) dz,
as
Z Z b
f (z) dz = f (γ(t))γ 0 (t) dt.
γ a

Clearly
Z Z Z
(αf + βg)(z) dz = α f (z) dz + β g(z) dz,
γ γ γ

where f, g are continuous functions defined on γ and α, β ∈ C are constants.

Moreover, if c ∈ (a, b) and γ1 = γ|[a,c] and γ2 = γ|[c,b] , then
Z Z Z
f (z) dz = f (z) dz + f (z) dz.
γ γ1 γ2
 LECTURE NOTES ON COMPLEX ANALYSIS I 21

• Let γ : [a, b] → C be a piecewise C 1 curve. Let f be a holomorphic function

defined on an open set containing γ. Assume that f 0 is continuous. Then
Z
f 0 (z) dz = f (γ(b)) − f (γ(a)).
γ

In particular, if γ is closed, that is, if γ(a) = γ(b), then γ f 0 (z) dz = 0.

R

The above follows since

Z Z b Z b
0 0 0
f (z) dz = f (γ(t))γ (t) dt = (f ◦ γ)0 (t) dt.
γ a a

Now use Theorem 5.6.

• (Reparametrization) Let γ : [a, b] → C be a piecewise C 1 curve. We say that
γ1 : [c, d] → C is a reparametrization of γ if there is an increasing piecewise C 1
function α : [c, d] → [a, b] such that γ1 = γ ◦ α. In this case,
Z Z
f (z) dz = f (z) dz,
γ γ1

since
Z Z d Z d
f (z) dz = f (γ1 (s))γ10 (s) ds = f (γ ◦ α(s))γ 0 (α(s))α0 (s) ds.
γ1 c c

Using the change of variable t = α(s), we get

Z d Z b Z
0 0 0
f (γ ◦ α(s))γ (α(s))α (s) ds = f (γ(t))γ (t) dt = f (z) dz.
c a γ
1
• Let γ : [a, b] → C be a piecewise C closed curve. Let c ∈ (a, b) and let d =
c + (b − a). Define γ1 : [c, d] → C as
(
γ(t), c ≤ t ≤ b,
γ1 (t) =
γ(t − b + a), b ≤ t ≤ d.
The curve remains the same but the initial point is shifted from γ(a) to γ(c).
• (Reverse of a curve) Let γ : [a, b] → C be a curve. The curve given by the function
−γ : [−b, −a] → C defined as −γ(t) = γ(−t) is known as the reverse of the curve
γ. Note that the direction of the curve is reversed. If γ is piecewise C 1 , then so
is −γ, and
Z Z −a Z a Z
0 0
f (z) dz = − f (γ(−t))γ (−t) dt = f (γ(s))γ (s) ds = − f (z) dz.
−γ −b b γ

• Let γ : [a, b] → C be a piecewise C 1 closed curve. Let f be a continuous complex-

valued function on γ. Then
Z Z b Z b
0
| f (z) dz| = | f (γ(t))γ (t) dt| ≤ |f (γ(t))||γ 0 (t)| dt ≤ M L(γ),
γ a a

where M = sup{f (γ(t)) : t ∈ [a, b]} < ∞.

• Let γ : [a, b] → C be a piecewise C 1 closed curve. Let (fn ) be a sequence of
continuous complex-valued functions on γ converging uniformly to f on γ. Then
Z Z Z
| f (z) dz − fn (z) dz| = | (f (z) − fn (z)) dz| ≤ Mn L(γ),
γ γ γ
22 NIKITA AGARWAL

where Mn is the maximum of |f − fn | on γ. Since fn → f uniformly on γ, we

have Mn → 0. Thus
Z Z
lim fn (z) dz = f (z) dz. (2)
n→∞ γ γ

Theorem 5.8. If ∞
P
n=1 fn = f (uniform limit), then
Z Z XN N Z
X ∞ Z
X
f (z) dz = lim fn (z)dz = lim fn (z)dz = fn (z)dz.
γ n→∞ γ n=1 n→∞ γ γ
n=1 n=1

Example 5.9. Let n ∈ Z be an integer. Integrate the function f (z) = z n around the
unit circle parametrized by the curve γ(t) = eit , for 0 ≤ t ≤ 2π.
For n 6= −1, z n is the derivative of the function z n+1 /n + 1 (defined in an open set
containing γ). Hence, for n 6= −1,
Z
z n dz = 0.
γ
Now consider n = −1, Z Z 2π
−1
z dz = e−it ieit = 2πi.
γ 0

Example 5.10. Let n ≥ 1 be an integer. Integrate the function f (z) = z −1 (z + z −1 )2n

around the unit circle parametrized by the curve γ(t) = eit , for 0 ≤ t ≤ 2π.
From the above discussion,
Z Z 2π
f (z) dz = f (γ(t))γ 0 (t) dt
γ 0
Z 2π
= e−it (eit + e−it )2n ieit dt
0
Z 2π
= i (2 cos t)2n dt
0
On the other hand,
2n   2n
−1
X 2n k −(2n−k) X 2(k−n)−1
f (z) = z z z = z .
k=0
k k=0
Using the fact that (
n 6= −1
Z
0,
zn =
γ 2πi, n = −1,
we get  
1.3 . . . (2n − 1)
Z
2n 1.2 . . . (2n)
f (z) dz = 2πi = 2πi = 2πi .
γ n (1.2 . . . n)2 1.2 . . . n
R
Comparing the two expressions for γ f (z) dz obtained, we get
Z 2π
1 1.3. . . . .(2n − 1)
(cos t)2n dt = .
2π 0 2.4. . . . .(2n)
Example 5.11. Prove that
Z ∞ √ p √
−t2 2 π 1+ 2
e cos t dt = .
0 4
 LECTURE NOTES ON COMPLEX ANALYSIS I 23
2
Let us integrate the function f (z) = e−z over the boundary of the region {z : 0 ≤
|z| ≤ R, 0 ≤ Arg(z) ≤ π/8}. The boundary γ (in counterclockwise direction) can be
written as the union of three curves (plot them on the complex plane):
γ1 : [0, R] → C given by γ1 (t) = t,
γ2 : [0, π/8] → C given by γ1 (t) = Reit ,
and the reverse of
γ3 : [0, R] → C given by γ3 (t) = teiπ/8 .
Thus, since γ is a closed curve (here we are using f = g 0 for some differentiable/holomorphic
function g), Z Z Z Z
0= f = f+ f− f.
γ γ1 γ2 γ3
Note the negative sign in the third term.
Now consider the three integrals one by one.
Z Z R Z ∞ √
−t2 2
f = e dt → e−t dt = π/2,
γ1 0 0

as R → ∞ (we are using the last identity above from your earlier courses). Now
Z
f ≤ MR L(γ2 ) → 0,
γ2
2/
√
as R → ∞, where MR = sup{|f (z)| : z ∈ γ2 } = e−R 2
and L(γ2 ) < ∞.
See HW 4.
24 NIKITA AGARWAL

6. Week 6
Cauchy’s theorem.
Theorem 6.1. (On a triangle) Let T be a triangle in C and let f be a holomorphic
function in an open set containing T and its interior (region enclosed by T ). Then
Z
f = 0.
T

The above result is sometimes known as Goursat’s lemma, after E. Goursat (late 1800s).

Proof. Let us integrate f over T in the counterclockwise direction, let I denote the inte-
gral.
• Divide T into 4 triangles by joining the midpoints of the sides of T , as shown in
the figure below. Then
X 4 Z
I= f.
j=1 T0j
R
Hence there is a j such that | T0j
f | ≥ |I|/4, otherwise
4 Z
X 4
X Z
|I| = | f| ≤ | f | < |I|,
j=1 T0j j=1 T0j

which is a contradiction.
Denote such a T0j by T1 . Again divideRthis process R for T1 to get four triangles
such that one of them, say T2 satisfies | T2 f | ≥ | T1 f |/4.
• Repeating this process, we get a sequence (Tn ) of decreasing triangles with
Z
| f | ≥ |I|/4n , n ≥ 1.
Tn

• Observe that L(Tn ) = L(T )/2n , for all n ≥ 1.

• Let Kn be the union of Tn and the region it encloses. Then Kn is compact,
Kn ⊃ Kn+1 , and the diameters of Kn tend to 0, as n → ∞. Hence ∩Kn = {z0 },
a singleton point (by the Cantor’s intersection theorem). Note that z0 ∈ T or z0
lies in the region enclosed by T . In any case, f is holomorphic at z0 .
• Thus
f (z) = f (z0 ) + f 0 (z0 )(z − z0 ) + R(z),
R(z)
where limz→z0 = 0.
Rz − z0
Observe that Tn (f (z0 ) + f 0 (z0 )(z − z0 )) = 0. Hence
Z Z
f= R(z).
Tn Tn

• Let n = supz∈Tn \{z0 } |R(z)/(z − z0 )|. Then n → 0 and for z ∈ Tn ,

|R(z)| ≤ n |z − z0 | ≤ n L(Tn ).
Hence Z
| R| ≤ n L(Tn )2 = n L(T )2 /4n .
Tn
 LECTURE NOTES ON COMPLEX ANALYSIS I 25

• This and above inequality imply

Z Z
n
|I| ≤ 4 | f| = | Rn | ≤ n L(T )2 → 0.
Tn Tn

Hence I = 0.

Theorem 6.2. (Existence of primitive) Let G be a convex open subset of C and let f be
a holomorphic function in G. Then there exists a homolorphic function g on G such that
g0 = f .
Proof. Fix a point z0 ∈ G.
• For z ∈ G, define Z
g(z) = f,
[z0 ,z]

where [z0 , z] denotes the line segment joining z0 to z. We prove that g 0 = f .

• Fix a point z1 ∈ G. Then
Z Z Z
f+ f+ f = 0,
[z0 ,z1 ] [z1 ,z] [z,z0 ]

by Cauchy’s theorem for a triangle.

• Hence Z
g(z) − g(z1 ) = f.
[z1 ,z]
Thus
g(z) − g(z1 )
Z
1
− f (z1 ) = (f (w) − f (z1 )) dw.
z − z1 z − z1 [z1 ,z]

• Since f is continuous at z1 , given  > 0, there exists δ > 0 such that |f (w) −
f (z1 )| < , whenever |w − z1 | < δ.
• Hence if |z − z1 | < δ, we have
g(z) − g(z1 )
Z
1 1
| − f (z1 )| = | (f (w) − f (z1 )) dw| ≤ L([z1 , z]) = .
z − z1 z − z1 [z1 ,z] |z − z1 |
Thus
g(z) − g(z1 )
lim = f (z1 ).
z→z1 z − z1
That is, g is differentiable at z1 and g 0 (z1 ) = f (z1 ).
• Note: the definition of g is independent of the point z0 up to addition of a constant
since if z00 is another point, then
Z Z Z
f+ f= f.
[z0 ,z00 ] [z00 ,z] [z0 ,z]

by Cauchy’s theorem for a triangle.


Corollary 6.2.1. (For a convex region)
R Let G be a convex open subset of C and let f be
a holomorphic function in G. Then γ f = 0, for each piecewise C 1 closed curve γ in G.

The result follows from the preceding result and γ f = γ g 0 = 0.

R R
26 NIKITA AGARWAL

Corollary 6.2.2. Let f be a continuous complex-valued function in a convex open subset

G of C such that the integral of f over every triangle contained in G is 0, then f has a
primitive in G.
In the preceding theorem, we used the Cauchy’s theorem for triangles to get that the
integral of f over every triangle contained in G is 0, and continuity of f . If we apriori
know that the integral of f over every triangle contained in G is 0, then we can bypass
the Cauchy’s theorem for triangles, and just use continuity of f to prove the preceding
theorem.
Theorem 6.3. (Cauchy’s formula for a circle) Let C be a counterclockwise oriented circle
and let f be a holomorphic function in an open set containing C and its interior. Then
Z
1 f (w)
f (z) = dw,
2πi C w − z
for all z in the interior of C (open disk enclosed by the circle C).
Proof. Let z0 be a point in the interior of C.
• Let 0 <  < d(z0 , C). Let C be the circle centered at z0 with radius  oriented
counterclockwise as shown in the figure.
• Let γj , j = 1, 2, 3, 4, be the closed curves, all oriented counterclockwise as shown
in the figure.
f (w)
• Then is holomorphic in a convex open set containing each γj . Thus
w − z0
Z
f (w)
dw = 0, j = 1, 2, 3, 4.
γj w − z0

• Since
4 Z Z Z
X f (w) f (w) f (w)
dw = dw − dw,
j=1 γj w − z0 C w − z0 C w − z0
we have Z Z
f (w) f (w)
dw = dw.
C w − z0 C w − z0
Also Z Z
1 1
dw = dw = 2πi.
C w − z0 C w − z0
• Consider
f (w) − f (z0 )
Z Z Z Z
1 f (w) 1 f (w) f (z0 ) 1 1
−f (z0 ) = − dw = dw.
2πi C w − z0 2πi C w − z0 2πi C w − z0 2πi C w − z0
f (w) − f (z0 )
• Let M denote the maximum of | | on C (finite since it is continuous
w − z0
on compact C ). Then
f (w) − f (z0 )
Z Z
1 f (w) 1 M L(C )
| dw − f (z0 )| = | dw| ≤ = M .
2πi C w − z0 2πi C w − z0 2π
• Finally, as  → 0, M → |f 0 (z0 )| since f is holomorphic at z0 . Hence M → 0.
Therefore Z
1 f (w)
f (z0 ) = dw.
2πi C w − z0

 LECTURE NOTES ON COMPLEX ANALYSIS I 27

Remark 6.4. (Mean value property) Let f be holomorphic in the disk |z − z0 | < R. For
0 < r < R, let Cr be the circle centered at z0 with radius r oriented in counterclockwise
direction. Parametrize Cr by γ(t) = z0 + reit , for t ∈ [0, 2π]. Then
Z 2π Z 2π
f (z0 + reit ) it
Z
1 f (w) 1 1
f (z0 ) = dw = it
rie dt = f (z0 + reit ) dt.
2πi Cr w − z0 2πi 0 re 2π 0
We have got a useful formula:
Z 2π
1
f (z0 ) = f (z0 + reit ) dt.
2π 0

In other words, the value of f at the center of Cr is the average of its values over Cr .
Definition 6.5. (Cauchy integral formula) Let γ : [a, b] → C be a piecewise C 1 curve,
and let φ be a continuous complex-valued function on γ. The Cauchy integral of φ over
γ is defined as the function f : C \ γ → C given by
Z
φ(w)
f (z) = dw, z ∈ C \ γ.
γ w−z

Remark 6.6. By the Cauchy’s formula for a circle, let γ be a counterclockwise oriented
circle C and let φ be a holomorphic function in an open set containing C and its interior.
Then Z
1 φ(w)
φ(z) = dw,
2πi C w − z
for all z in the interior of C (open disk enclosed by the circle C). That is, the Cauchy
integral formula for φ in the interior of C equals 2πiφ.
Let γ : [a, b] → C be a piecewise C 1 curve, and let φ be a continuous complex-valued
function on γ. Let z0 ∈ / γ and let R = d(z0 , γ) > 0. Let z be such that |z − z0 | < R.
Observe that
∞  n X ∞
1 1 1 X z − z0
= = = (w − z0 )−n−1 (z − z0 )n ,
w−z w − z0 − (z − z0 ) w − z0 n=0 w − z0 n=0

provided |z − z0 | < |w − z0 |. The series (in w) on the right converges uniformly on any
compact subset of this region as follows:
Let A be a compact subset such that for all w ∈ A, |z − z0 | < |w − z0 |. Hence m =
min{|w − z0 | : w ∈ A} > 0. Also R < m. Therefore for w ∈ A, we have
|(w − z0 )−n−1 (z − z0 )n | < m−n−1 Rn ,
for all n ≥ 1. Therefore by the Weierstrass M-test, the series (in w) converges uniformly
in A.
In particular, taking A = γ (which is a compact set) in the above discussion, the series
∞
X
(w − z0 )−n−1 (z − z0 )n
n=0

converges uniformly on γ to the function 1/(w − z).

Since the function φ is bounded (being continuous) on γ, we have
∞
X φ(w)(z − z0 )n
n=0
(w − z0 )n+1
28 NIKITA AGARWAL

φ(w)
converges uniformly to on γ. Hence we can integrate term by term (using Theo-
w−z
rem 5.8) to get
∞
φ(w)(z − z0 )n
Z Z
φ(w) X
f (z) = dw = n+1
dw, |z − z0 | < R. (3)
γw−z n=0 γ (w − z0 )

We have thus shown that f can be expressed as the power series ∞ n

P
n=0 an (z − z0 ) in the
disk |z − z0 | < R, where Z
φ(w)
an = n+1
.
γ (w − z0 )
By power series discussion, we have
f (n) (z0 )
Z
(n) φ(w)
an = ⇒ f (z0 ) = n! n+1
dw, z ∈ γ.
n! γ (w − z0 )
Observe the following:
Let f be a holomorphic function in an open set G. Let z0 ∈ G. For 0 < r < dist(z0 , C \
G), let Cr denote the circle centered at z0 with radius r, oriented counterclockwise. Then,
f
inside Cr , the function f coincides with the Cauchy integral of the function over Cr .
2πi
Hence we have the following result.
Theorem 6.7. Let f be a holomorphic function in an open set G. Let z0 ∈ G. For r > 0,
let Cr denote the circle centered at z0 with radius r, oriented counterclockwise. Then
• the function f is differentiable to all orders and for any positive integer n,
Z
(n) n! f (w)
f (z0 ) = dw, |z − z0 | < r < dist(z0 , C \ G).
2πi Cr (w − z0 )n+1
• in the disk |z − z0 | < dist(z0 , C \ G), the function f is represented by its Taylor’s
series centered at z0 .
Theorem 6.8. (Converse of Goursat’s Lemma, Theorem R 6.1) Let f be a continuous
complex-valued function in an open subset G of C. If T f = 0 for all triangles T which
are contained in G along with their interiors, then f is holomorphic on G.
Proof. It is enough to prove this result when G is an open disk since G is a union of open
disks and holomorphicity is a local property.
The given hypothesis imply that f has a primitive g, which is holomorphic.
Hence g is infinitely differentiable.
Thus, f = g 0 is holomorphic. 
Definition 6.9. A function that is holomorphic on all of C is known as an entire function.
Theorem 6.10. (Liouville’s Theorem) The only bounded entire functions are constant
functions.
Proof. Let f be a bounded entire function and |f (z)| ≤ M , for all z ∈ C. Fix a point
z0 ∈ C and let R > 0. Let CR denotes the circle |z − z0 | = R oriented counterclockwise.
Then Z
0 1 f (w)
f (z0 ) = dw.
2πi CR (w − z0 )2
Hence
1 M M
|f 0 (z0 )| ≤ 2
2πR = .
2π R R
Since the above is true for all R > 0, we get f 0 (z0 ) = 0. Hence f is a constant function. 
 LECTURE NOTES ON COMPLEX ANALYSIS I 29

Theorem 6.11. (Fundamental Theorem of Algebra) Every nonconstant polynomial with

complex coefficients has a complex root.
Proof. Let p(z) = an z n + an−1 z n−1 + · · · + a1 z + a0 be a nonconstant polynomial of degree
n with complex coefficients having no complex roots.
Let f = 1/p. Then f is an entire function.
Since an 6= 0, we have for z 6= 0,
 
n an−1 a1 a0
p(z) = an z 1 + + ··· + + .
an z an z n−1 an z n
Hence  
n |an−1 | |a1 | |a0 |
|p(z)| ≥ |an z | 1 − − ··· − − .
|an z| |an z n−1 | |an z n |
There exists R > 0 such that whenever |z| ≥ R, the term in the parentheses on the right
is at least 1/2. Therefore
1 2
|f (z)| = ≤ , |z| ≥ R.
|p(z)| |an |Rn
Also, f being a continuous function is bounded in the compact set |z| ≤ R. Thus, f
is a bounded entire function, and hence is a constant function. We have arrived at a
contraction to the assumption that p is a nonconstant polynomial. 
The above is an existential proof. The problem of finding accurate roots is a major
theme in numerical analysis.
Definition 6.12. Let f be a holomorphic function on an open subset G of C. A point
z0 ∈ C is known as a zero of f of order (or multiplicity) m ≥ 1, if f (n) (z0 ) = 0, for all
0 ≤ n ≤ m − 1, and f (m) (z0 ) 6= 0.
If z0 is a zero of f of order (or multiplicity) m ≥ 1, the Taylor’s series of f about z0
has the form ∞
X
an (z − z0 )n ,
n=m
f (n) (z0 )
where an = , for n ≥ m, and am 6= 0.
n!
We can express f (z) = (z − z0 )m g(z), where
(
(z − z0 )−m f (z), z ∈ G \ {z0 },
g(z) =
am , z = z0 ,
is holomorphic in G.
Since the function g is nonzero at z0 , it is nonzero in a disk around z0 . Hence we have
the following result.
Theorem 6.13. Zeros of finite order of a holomorphic function are isolated.
What about zeros of infinite order?
The set of all zeros of f of infinite order is an open set since the Taylor series expansion
of f around each zero of f of infinite order is 0. By the above argument, the set of all
points of G that are not zeros of f of infinite order is an open set. If G is connected, one
of these sets is empty. If G is not connected, and if f has a zero z0 of infinite order, then
f is 0 in the connected component of G containing z0 .
30 NIKITA AGARWAL

7. Week 7
There was a problem session and Quiz 1 was conducted in this week. Below are the
pictures for Theorems 6.1, 6.2 and 6.3.
 LECTURE NOTES ON COMPLEX ANALYSIS I 31

8. Week 8
Theorem 8.1. Let f be a holomorphic function, not the zero function, in a connected
open subset G of C. Then each zero of f is of finite order, and f −1 (0), the zero set of f
has no limit points in G.
Proof. If f has a zero of infinite order, then f is identically 0 in G since G is connected
and open. Therefore all the zeros of f are of finite order. Moreover, since all zeros of
finite order are isolated, the set of zeros, f −1 (0) has no limit point. 
Theorem 8.2. (Identity theorem) Let f and g be holomorphic functions in a connected
open subset G of C. If the set D = {z ∈ G : f (z) = g(z)} has a limit point in G, then
f = g.
Proof. Let h = f − g. Then h is holomorphic on G. The set D is the set of zeros of h.
By the preceding theorem, D does not a limit point in G unless h = 0. 
Theorem 8.3. (Maximum modulus principle) Let f be a nonconstant holomorphic func-
tion in an open connected subset G of C. Then |f | does not attain a local maximum in
G.
Proof. We will prove the result by contradiction. Suppose |f | attains a local maxima at
z0 ∈ G.
• Let r > 0 be such that the closed disk |z − z0 | < r is contained in G and |f (z0 )| ≥
|f (z)|, for all z in this disk.
• If f (z0 ) = 0, then f = 0 in the disk (which has a limit point in G). Hence by
the identity theorem, f is identically 0 in G, which is a contradiction since f is
nonconstant. Thus f (z0 ) 6= 0.
• Let λ = |f (z0 )|/f (z0 ). By the mean value property (Remark 6.4),
Z 2π
1
f (z0 ) = f (z0 + reit ) dt.
2π 0
Hence
Z 2π Z 2π
1 it 1
Re λf (z0 + reit ) dt.


|f (z0 )| = λf (z0 ) = λf (z0 + re ) dt =
2π 0 2π 0
1 R 2π 1 R 2π
The last equality is because |f (z0 )| = 0
λf (z0 +reit ) dt implies λf (z0 +
2π 2π 0
reit ) dt is real.
We now have
Z 2π
1
|f (z0 )| − Re λf (z0 + reit ) dt = 0.



(4)
2π 0
• Since
Re λf (z0 + reit ) ≤ |λf (z0 + reit )| = |f (z0 + reit )| ≤ |f (z0 )|, t ∈ [0, 2π],

the integrand in the integral in (4) is nonnegative and continuous, and we get
|f (z0 )| = Re λf (z0 + reit ) , t ∈ [0, 2π],

which can be rewritten as

|f (z0 )| = Re (λf (z)) , |z − z0 | = r.
• Since Re(λf (z)) ≤ |λf (z)| = |f (z)| ≤ |f (z0 )| = Re(λf (z)), for all |z − z0 | = r, we
get λf (z) is real and
λf (z) = |f (z0 )|, |z − z0 | = r.
32 NIKITA AGARWAL

• Substituting λ, we get f (z) = f (z0 ), for all |z − z0 | = r.

• Since the circle |z − z0 | = r has a limit point in G, by the identity theorem, f is a
constant function on G. This contradicts the hypothesis that f is a nonconstant
function.

Theorem 8.4. (Schwarz’s lemma) Let D be the open unit disk |z| < 1. Let f : D → D
be a holomorphic function such that f (0) = 0. Then |f (z)| ≤ |z|, for all z ∈ D. The
inequality is strict at all points other than the origin, except in the case when f has the
following form f (z) = λz, for all z ∈ D, for some fixed λ of modulus 1.
Proof. Define g on D as 
 f (z)
, 0 < |z| < 1,
g(z) = z
f 0 (0), z = 0.
Clearly g is a holomorphic function.
• For 0 < r < 1 and |z = r|, we get |g(z)| = |f (z)/z| ≤ 1/r. Since |g| is continuous
on the closed disk |z| ≤ r, it attains its maxima in this disk. The maxima cannot
be attained in the open disk |z| < r, otherwise it will be a local maxima, and will
violate the Maximum modulus principle. Hence |g(z)| ≤ 1/r, for all |z| ≤ r.
Note: I missed discussing the matter in red in the class. This is
crucial otherwise the conclusion in the next bullet does not follow.
• Since the above argument is true for all 0 < r < 1, we have |g(z)| ≤ 1, for all
|z| < 1.
• Therefore |f (z)| ≤ |z|, for all z ∈ D.
• Furthermore if |f (z0 )| = |z0 | for some z0 ∈ D \ {0}, then |g(z0 )| = 1. Since
|g(z)| ≤ 1 on D, z0 is a local maxima of |g| in D. Thus, by the Maximum
modulus principle, g is a constant function, say g(z) = λ, for all z ∈ D, which
implies f (z) = λz, for all z ∈ D. Clearly |z0 | = |f (z0 )| = |λz0 | = |λ||z0 |, implies
|λ| = 1 since z0 6= 0.

The following result has already been used in Schwarz’s lemma. It is worth writing it
down as a theorem.
Theorem 8.5. Let G be a bounded open subset of C. Let f be a continuous function on
the closure of G, and is holomorphic in G. Then
max{|f (z)| : z ∈ Cl(G)} = max{|f (z)| : z ∈ Bd(G)}.
Here Cl(G) denotes the closure of G and Bd(G) denotes the boundary of G.
Proof. If f is a constant function, the result follows trivially. Suppose f is nonconstant.
Since G is bounded, there exists a point of global maxima for |f | in Cl(G). Let z0 be a
point of global maxima.
By the Maximum modulus principle, |f | does not have a local maximum in G. Hence
z0 ∈ Bd(G). 
Theorem 8.6. (Morera’s Theorem) Let f be a continuous complex-valued function in an
open subset G of C such that the integral of f over every triangle contained in G is 0,
then f is analytic in G.
Proof. Recall Corollary 6.2.2 at this point. We proved that if G is convex open with the
above hypothesis then f has a primitive in G.
 LECTURE NOTES ON COMPLEX ANALYSIS I 33

We prove now that f is analytic in each open disk contained in G which implies that f
is analytic in G. Thus, without loss of generality, assume that G is an open disk.
Further, if f has a primitive in G, say g satisfying g 0 = f . Then g is holomorphic, and
by Theorem 6.7, g is infinitely differentiable, and therefore f = g 0 is analytic.
Thus the theorem follows from the arguments given above. 
In the following statement, for a subset G of C, G∗ = {z : z ∈ G}, G+ = {z ∈ G :
Im(z) > 0}, G− = {z ∈ G : Im(z) < 0}, and G0 = {z ∈ G : Im(z) = 0}.
Theorem 8.7. (Schwarz reflection principle) Let G be an open connected subset of C
such that G = G∗ . If f : G+ ∪ G0 → C is a continuous function which is holomorphic
in G+ and if f (x) ∈ R for all x ∈ G0 , then there is an holomorphic function g : G → C
such that g = f on G+ ∪ G0 .
Proof. Step 1: Define g.

Since G = G∗ , we have G∗− = G+ . Hence if z ∈ G− , then z ∈ G+ . Thus define

g : G → C as (
f (z), z ∈ G+ ∪ G0
g(z) =
f (z), z ∈ G− .
Step 2: Continuity of g. This step was not discussed in the class.

Let h : G− ∪ G0 → C be defined as h(z) = f (z), for all z ∈ G− ∪ G0 . Observe that

(G+ ∪ G0 )∗ = G− ∪ G0 . Since f is continuous in G+ ∪ G0 , h is continuous in G− ∪ G0 .
Moreover for z ∈ G0 ⊂ R, f (z) ∈ R, hence h(z) = f (z) = f (z). By pasting lemma (with
X = G, A = G+ ∪ G0 , B = G− ∪ G0 , Y = C), g is a continuous function.
Lemma 8.8. (Pasting lemma from topology without proof ) Let X = A ∪ B where A, B
are closed on X. Let f : A → Y and h : B → Y be continuous. If f (x) = h(x) for
all x ∈ A ∩ B, then the function g : X → Y given as g = f on A and g = h on B is
continuous on X.
Step 3: Differentiability of g in G+ ∪ G− .

By Example 2.9, g is holomorphic in G− . Since f is holomorphic in G+ , g is also holo-

morphic in G+ . Thus we just need to prove that g is holomorphic at all points in G0 .

Step 4: Differentiability of g in G0 .
• Let z0 ∈ G0 . Let R > 0 be such that B(z0 , R) ⊆ G. We prove that g is
holomorphic in B(z0 , R). For this, we use Morera’s theorem. Since we have
restricted the domain to B(z0 , R), we need to consider only triangles which are
contained in B(z0 , R), and hence the interiors of such triangles are also contained
in B(z0 , R).
• Any triangle in B(z0 , R) is of four types as shown:
Type 1: either contained in G+ or G− , or
Type 2: intersects both G+ and G0 but not G− , or
Type 3: intersects both G− and G0 but not G+ , or
Type 4: Rintersects
R each of G+ , G− and G0 .
• Clearly T g = T f = 0, for all triangles T Rof Type R 1 contained in G+ . If T is
a triangle of Type 1 contained in G− , then T g = T h = 0. So we focus on the
other types of triangles.
34 NIKITA AGARWAL

• By drawing all possible triangles of Type 4 (which intersect each ofR G+ , G− , and
G0 ), it is clear that if for any triangle TR of Types 2 and 3, we have T g = 0, then
for any triangle T of Type 4, we have T g = 0.
• Thus we only need to focus on triangles of Types 2 and 3.

Step 5: Focus on one particular triangle of Type 2.

• Let T = [a, b, c, a] be a triangle in G+ ∪ G0 such that [a, b] ⊂ G0 . Proof in other
cases is similar.
• Let ∆ be the triangle T along with its interior. Then g = f for all z ∈ ∆.
 LECTURE NOTES ON COMPLEX ANALYSIS I 35

• Since f is continuous in G+ ∪ G0 , it is uniformly continuous in ∆. Let  > 0 be

given. There exists δ > 0 such that for all z, w ∈ ∆ with |z − w| < δ, we have
|f (z) − f (w)| < .
• Let α, β be points on the edges [a, c], [b, c], respectively, such that |a − α| < δ
and |b − β| < δ. Consider the triangle T1 = [α, β, c, α] and quadrilateral Q =
[a, b, β, α, a]. Then Z Z Z
f= f+ f.
T T1 Q
Since f is
R holomorphic in G+ and T1 along with its interior is contained in G+ ,
we have T1 f = 0. Hence
Z Z
f= f.
T Q
• Let |f | ≤ M on ∆. Let L be the length of T .
• The quadrilateral Q has four edges [a, b], [b, β], [β, α] and [α, a]. Observe that
Z Z
| f | ≤ M |b − β| < M δ, | f | ≤ M |a − α| < M δ.
[b,β] [α,a]

• Let 0 ≤ t ≤ 1. Then
|tb + (1 − t)a − (tβ + (1 − t)α)| ≤ t|b − β| + (1 − t)|a − α| < δ.
• Further
Z Z
f+
[a,b] [β,α]f
Z 1 Z 1
= (b − a) f (tb + (1 − t)a) dt − (β − α) f (tβ + (1 − t)α) dt
0 0
Z 1
≤ |b − a| (f (tb + (1 − t)a) − f (tβ + (1 − t)α)) dt
0
Z 1
+|(b − a) − (β − α)| f (tβ + (1 − t)α) dt
0
≤ |b − a| + 2M δ.
• Hence Z Z
| f| = |
f | ≤ ` + 4M δ.
T Q
R
Choose δ <  and since  is arbitrary, we get T f = 0.

36 NIKITA AGARWAL

8.1. Solved problems – some discussed on Sept 25, 2024.

Example 8.9. Let C be the unit circle oriented counterclockwise. Evaluate the integral
Z  3
z−2
dz.
C 2z − 1
Solution: Let f (z) = (z − 2)3 which is an entire function. By Theorem 6.7, since 1/2 lies
in the interior of C,
Z 3
2 (z − 2)3
Z
(2) 2! f (z) 2!
f (1/2) = dz = dz.
2πi C (z − 1/2)3 2πi C (2z − 1)3
Hence 3
2πif (2) (1/2)
Z 
z−2 2πi6(−3/2) −9πi
dz = 4
= 4
= .
C 2z − 1 2 2 8
We have used f (2) (z) = 6(z − 2).
Example 8.10. Let a, b be complex numbers satisfying |a| < 1 < |b|. Let C be the unit
circle oriented counterclockwise. Evaluate
(z − b)m
Z
1
dz,
2πi C (z − a)n
for all pairs of integers m, n.

Solution: If n ≤ 0, the integrand is analytic in C, hence the integral is 0.

If n ≥ 1: Let f (z) = (z − b)m , which is analytic in the open disk {z ∈ C : |z| < |b|}.
Also a lies in the interior of C. Hence by the Theorem 6.7,
(n − 1)! (n − 1)! (z − b)m
Z Z
(n−1) f (z)
f (a) = n
dz = n
dz.
2πi C (z − a) 2πi C (z − a)
Hence 
0, n ≥ m + 2, m ≥ 0
(z − b)m f (n−1) (a) 
Z
1
n
dz = = m(m − 1) . . . (m − n + 2)(a − b)m−n+1 , n ≤ m + 1, m ≥ 0
2πi C (z − a) (n − 1)! m(m − 1) . . . (m − n + 2)(a − b)m−n+1 ,

m < 0.
Example 8.11. Prove that there is no holomorphic function f in the open unit disk such
that f (1/n) = 2−n , for n ≥ 2.

Solution: Suppose there is such a function. By continuity of f , f (0) = limn→∞ f (1/n) =

limn→∞ 2−n = 0. Hence 0 is a zero of f . This zero if of finite order since f is not identically
0. Therefore f (z) = z m g(z), for all z ∈ D, where m ≥ 1 and g(0) 6= 0. Now
g(0) = lim g(1/n) = lim nm f (1/n) = lim nm 2−n = 0,
n→∞ n→∞ n→∞
which is a contradiction.
Example 8.12. Let D be the open unit disk |z| < 1. Let u ∈ D. Prove that the Mobius
transformation φu given by
z−u
φu (z) = ,
1 − zu
maps D onto D.

Solution: For z = eiθ ,

eiθ − u 1 − ue−iθ
φu (eiθ ) = = = 1.
1 − eiθ u 1 − eiθ u
 LECTURE NOTES ON COMPLEX ANALYSIS I 37

Hence φu maps the unit circle onto itself (clircles maps to clircles). Taking G = D
and f = φu in Theorem 8.5, we get |φu (z)| ≤ 1, for all |z| ≤ 1. Hence the Mobius
transformation φw maps D onto D.
Alternately, since φu maps the unit circle onto itself, the interior D of the unit circle is
either mapped onto itself or the exterior of the unit circle. Since φw (0) = −u ∈ D, φu
maps D onto D.
Example 8.13. (Pick’s lemma) Let D be the open unit disk |z| < 1. Let f : D → D be
a holomorphic function. Prove that for any two points z, w ∈ D,
f (z) − f (w) z−w
≤ .
1 − f (z)f (w) 1 − zw
Solution: Let w ∈ D. The function
F = φf (w) ◦ f ◦ φ−1
w
is holomorphic from D to D and maps 0 to 0 since φu (u) = 0.
By Schwarz’s lemma, we have |F (z)| ≤ |z|, for all z ∈ D.
Let z ∈ D and z 0 ∈ D be such that φw (z) = z 0 . Hence
f (z) − f (w) z−w
|φf (w) ◦ f ◦ φ−1 0 0
w (z )| ≤ |z | ⇒ |φf (w) (f (z))| ≤ |φw (z)| ⇒ ≤ .
1 − f (z)f (w) 1 − zw
38 NIKITA AGARWAL

9. Week 9
Mid-Semester Exam week
10. Week 10
Mid-Semester break
 LECTURE NOTES ON COMPLEX ANALYSIS I 39

11. Week 11
Theorem 11.1. (Weierstrass convergence theorem) Let G be an open subset of C. Let
(fk )k≥1 be a sequence of holomorphic functions on G that converge locally uniformly in
(n)
G to a function f . Then f is holomorphic and for each n ≥ 1, the sequence (fk )k≥1
converges locally uniformly in G to f (n) .
Note: Locally uniformly means that fk converges uniformly to f in each compact
subset of G.
Recall, in R, even the differentiability of f fails. In fact, by the Weierstrass approxi-
mation theorem, any continuous function on R is a local uniform limit of a sequence of
polynomials.
Proof. Fix z0 ∈ G and let r > 0 be such that the disk |z − z0 | ≤ r is contained in G. Let
Cr be the circle |z − z0 | = r oriented counterclockwise.
• By Cauchy’s integral formula,
Z
1 fk (w)
fk (z) = dw,
2πi Cr w − z
for all |z − z0 | < r.
• Observe that for w ∈ Cr , |z − z0 | < r, we get |w − z| ≥ r − |z − z0 |. Hence, for
each z with |z − z0 | < r,
fk (w) f (w) fk (w) − f (w) |fk (w) − f (w)|
− = ≤ < ,
w−z w−z w−z r − |z − z0 |
for all k ≥ K, and for each w ∈ Cr (since fk converges to f uniformly on Cr
(compact subset of G)).
fk (w) f (w)
• Hence for each z with |z − z0 | < r, converges uniformly to on Cr .
w−z w−z
• Therefore by Equation (2),
Z Z
fk (w) f (w)
dw → dw,
Cr w − z Cr w − z
for each z with |z − z0 | < r, which implies
Z
1 f (w)
lim fk (z) = dw.
k→∞ 2πi Cr w − z
• Also since fk (z) → f (z), we get
Z
1 f (w)
f (z) = dw,
2πi Cr w−z
for all |z − z0 | < r.
• Since f is given by a Cauchy integral in |z − z0 | < r, it is holomorphic in this disk.
Since G can be covered by such disks, f is holomorphic on G.
• Fix n ≥ 1.
• Both f (n) (z) and f (n) (z) can be expressed as integrals in |z − z0 | < r. Using this
we get on |z − z0 | < r,
fk (w) − f (w)
Z
(n) (n) n!
fk (z) − f (z) = dw.
2πi Cr (w − z)n+1
• Let Mk denote the maximum of |fk (w)−f (w)| for w ∈ Cr . Then for |z −z0 | < r/2,
(n) 2n+1 n!Mk
fk (z) − f (n) (z) ≤ .
rn
40 NIKITA AGARWAL

(n)
• Since fk converges uniformly to f on Cr , we have Mk → 0. Thus fk (z) converges
uniformly to f (n) (z) on |z − z0 | < r/2.
(n)
We have thus proved that each z0 ∈ G has an open disk around it in which fk
(n)
converges uniformly to f (n) . We now prove that this implies that fk converges
locally uniformly to f (n) in G.
Let A be a compact subset of G. For each z ∈ A, let rz be such that B(z, rz ) ⊆
(n)
G and fk converges to f (n) uniformly in B(z, rz ). Since A is compact, the
open cover {B(z, rz )}z∈A of A has a finite subcover {B(zj , rzj )}j=1,...,s . Clearly
(n) (n)
fk converges to f (n) uniformly in ∪nj=1 B(zj , rzj ) which contains A. Hence fk
converges to f (n) uniformly on A.


11.1. Laurent series.

Definition 11.2. A series of the form ∞ n
P
n=−∞ an (z − z0 ) with an ∈ C is known as a
Laurent series centered at the point z0 ∈ C.
The series is said to converge at a point z if both the series ∞ n
P
n=0 an (z − z0 ) and
P1 n
P ∞ −n
n=−∞ an (z − z0 ) = n=1 a−n (z − z0 ) converge. In that case, the limit
N
X
lim an (z − z0 )n
N →∞
n=−N

exists and is defined to be the sum of the Laurent series.

Theorem 11.3. Let
1
R1 = lim sup |a−n |1/n , R2 = .
n lim supn |an |1/n
If R1 < R2 , the Laurent series converges absolutely and locally uniformly in the annulus
R1 < |z − z0 | < R2 . Moreover the limit function is holomorphic in this annulus.
Note: In the above, R1 = 0, R2 = ∞ cases are included.

Proof. The result follows since the series ∞ n

P
n=0 an (z−z
P∞ 0 ) converges absolutely and locally
−n
uniformly in the disk |z−z0 | < R2 , and the series n=1 a−n (z−z0 ) converges absolutely
and locally uniformly in the region |z − z0 | > R2 .
By the Weierstrass approximation Theorem 11.1, the limit function is holomorphic in the
annulus R1 < |z − z0 | < R2 . 

Let the limit be defined as the function f . For R1 < r < R2 , the series converges
uniformly on the circle Cr (circle centered at z0 with radius r). Thus we can integrate
the series term-by-term on Cr , to get
Z X∞ Z
f (z) dz = an (z − z0 )n dz = 2πia−1 .
Cr n=−∞ Cr
R
In the above, we have used the fact that Cr (z − z0 )n dz = 2πi when n = −1. For all
other n ∈ Z, the integral is 0.
Thus Z
1
a−1 = f (z) dz.
2πi Cr
 LECTURE NOTES ON COMPLEX ANALYSIS I 41

Remark : Observe that if f is holomorphic in B(z0 , R2 ), then a−1 = 0 (by Corollary 6.2.1).
In place of f (z), for k ∈ Z, repeating the
P∞same process to the function (z − z0 )−k−1 f (z)
which is the limit of the Laurent series n=−∞ an (z − z0 )n−k−1 , we get
Z ∞
X Z
−k−1
(z − z0 ) f (z) dz = an (z − z0 )n−k−1 dz = 2πiak .
Cr n=−∞ Cr

Hence Z
1
ak = (z − z0 )−k−1 f (z) dz.
2πi Cr
Thus the coefficients are uniquely determined by the limit function.
Remark : If f is holomorphic in B(z0 , R2 ), for k ≤ −1, ak = 0, and then the Laurent
series becomes the power series which we established in the earlier part of this course.
1
Example 11.4. Consider the function f (z) = . Note that
(z − 1)(z − 2)
1 1 1
f (z) = =− +
(z − 1)(z − 2) z−1 z−2
For |z| > 1,
∞ ∞
1 1 1 X −n X
− =−   =− z =− z −n .
z−1 1 z n=0
z 1− n=1
z
For |z| < 2,
∞
1 1 1X n n
=− =− z /2 .
z−2 2(1 − z/2) 2 n=0
Hence for 1 < |z| < 2, f has the following Laurent series expansion
∞ ∞
X
−n 1X n n
f (z) = − z − z /2 .
n=1
2 n=0

11.2. Cauchy integral near ∞. Via the stereographic projection of C ∪ {∞} onto S 2 ,
the point ∞ gets mapped to the north pole N . The region |z − z0 | > R in C ∪ {∞}
which contains ∞ gets mapped to an open set in S 2 around N . Hence the open sets of
the form |z − z0 | > R are known as neighbourhoods of the point ∞.
Recall the definition of Cauchy integral formula: Let γ : [a, b] → C be a piecewise C 1
curve, and let φ be a continuous complex-valued function on γ. The Cauchy integral of
φ over γ is defined as the function f : C \ γ → C given by
Z
φ(w)
f (z) = dw, z ∈ C \ γ.
γ w−z

Fix a point z0 ∈
/ γ. Let R > 0 be the maximum of |w − z0 | for w ∈ γ. Then γ ⊆ B(z0 , R).
Fix a point z with |z − z0 | > R. Observe that
 
∞
1 1 −1  1 X
= = =− (w − z0 )n−1 (z − z0 )−n ,

w−z w − z0 − (z − z0 ) z − z0 1 − w − z0

n=1
z − z0
provided |z − z0 | > |w − z0 |.
The series (in w) on the right converges uniformly on any compact subset of this region
as follows (see also the earlier discussion of Cauchy integral for holomorphic functions
after Remark 6.6 in red – more details added there):
42 NIKITA AGARWAL

Let A be a compact subset such that for all w ∈ A, |z − z0 | > |w − z0 |. Hence M =

max{|w − z0 | : w ∈ A} > 0. Also M < R. Therefore for w ∈ A, we have
|(w − z0 )n−1 (z − z0 )−n | < M n−1 R−n ,
for all n ≥ 1. Therefore by the Weierstrass M-test, the series (in w) converges uniformly
in A.
In particular, taking A = γ (which is a compact set) in the above discussion, the series
∞
X
− (w − z0 )n−1 (z − z0 )−n
n=1

converges uniformly on γ to the function 1/(w − z).

Now, since the function φ is bounded (being continuous) on γ, we have that the series
X∞
− φ(w)(w − z0 )n−1 (z − z0 )−n
n=1

φ(w)
converges uniformly to on γ. Hence we can integrate term by term (using Theo-
w−z
rem 5.8) to get
∞ Z
X 
f (z) = − φ(w)(w − z0 ) n−1
dw (z − z0 )−n , |z − z0 | > R. (5)
n=1 γ

We have
P thus shown that−nthe Cauchy integral f of φ over γ can be expressed as the Laurent
series ∞n=1 a−n (z − z0 ) , in |z − z0 | > R (with only negative powers of (z − z0 )), where
Z
a−n = φ(w)(w − z0 )n−1 dw, n ≥ 1.
γ

Theorem 11.5. (Used as a fact in the next result – differentiation under the integral
sign) Let F = F (r, t) : [r1 , r2 ] × [t1 , t2 ] → C be a continuous function having a continuous
first order partial derivative with respect to r. Let I : [r1 , r2 ] → C be defined as
Z t2
I(r) = F (r, t) dt.
t1
Then Z t2
d I(r) ∂
= F (r, t) dt.
dr t1 ∂r
Theorem 11.6. Let f be a holomorphic function in the annulus R1 < |z − z0 | < R2 .
R R1 < r < R2 , let Cr denote the circle |z − z0 | = r oriented counterclockwise. Then
For
Cr
f (z) dz is independent of r.
R
Note: If f is holomorphic in the disk |z − z0 | < R2 , then Cr f (z) dz = 0 for all r < R2
by Corollary 6.2.1.
R
Proof. Let I(r) = Cr f (z) dz.
Using the parametrization γ(t) = z0 + reit , 0 ≤ t ≤ 2π, of Cr , we get
Z 2π
Ir = f (z0 + reit )ireit dt.
0
Differentiating under the integral sign (see Theorem 11.5) gives
Z 2π
d I(r) ∂
f (z0 + reit )ireit dt.


=
dr 0 ∂r
 LECTURE NOTES ON COMPLEX ANALYSIS I 43

Now
∂ ∂
f (z0 + reit )ireit = f 0 (z0 + reit )ire2it + f (z0 + reit )ieit = f (z0 + reit )eit .



∂r ∂t
Thus Z 2π
d I(r) ∂
f (z0 + eit )eit dt = f (z0 + reit )eit |2π



= t=0 = 0.
dr 0 ∂t

Example 11.7. Consider the function f (z) = 1/z. It is holomorphic in the R annulus
|z| > 0 (z0 = 0, R1 = 0, R2 = ∞ in the above result). Then by Theorem 11.6, Cr f (z) dz
is independent of r. For r = 1, we know that
Z Z
f (z) dz = 1/z dz = 2πi.
C1 C1

Hence Z
f (z) dz = 2πi,
Cr
for all r > 0.
This example gives a contrast
R to Corollary 6.2.1 since the domain on which f is holomor-
phic is not convex. Hence Cr f (z) dz need not be 0.
Theorem 11.8. Let f be a holomorphic function in the annulus R1 < |z − z0 | < R2 .
For R1 < r < R2 , let Cr denote the circle |z − z0 | = r oriented counterclockwise. If
R1 < r1 < |z − z0 | < r2 < R2 , then
Z Z
1 f (w) 1 f (w)
f (z) = dw − dw.
2πi Cr2 w − z 2πi Cr1 w − z
Note: If f is holomorphic in the disk |z − z0 | < R2 , then we know that
Z
1 f (w)
f (z) = dw.
2πi Cr2 w − z
1 R f (w)
Moreover, Cr1
= 0 since the integrand is holomorphic in an open disk con-
2πi w−z
taining Cr1 and its interior. Hence we recover Theorem 6.3 which gives Cauchy’s formula
for a circle.
Proof. Fix a point z satisfying R1 < r1 < |z − z0 | < r2 < R2 .
Define a function g in the annulus R1 < |w − z0 | < R2 as follows:

 f (w) − f (z)
, w 6= z,
g(w) = w−z
f 0 (z), w = z.

Consider the function F (w) = f (w) − f (z) which is holomorphic in the annulus R1 <
|w − z0 | < R2 . Since F (z) = 0, we get F (w) = (w − z)G(w), where the function G(w) is
holomorphic in the annulus R1 < |w − z0 | < R2 . The function G is precisely g(w) defined
above. Therefore g is holomorphic in the annulus R1 < |w − z0 | < R2 .
Hence by Theorem 11.6,
Z Z
g(w) dw = g(w) dw.
Cr1 Cr2
44 NIKITA AGARWAL

f (w) − f (z)
Since z ∈
/ Cr1 ∪ Cr2 , we get g(w) = , for w ∈ Cr1 ∪ Cr2 . Substituting this in
w−z
the above equation and rearranging terms, we get
Z Z Z Z
f (w) f (z) f (w) f (z)
dw − dw = dw − dw.
Cr1 w − z Cr1 w − z Cr2 w − z Cr2 w − z

Upon rearranging the terms, we get

Z Z Z Z
f (w) f (w) 1 1
dw − dw = f (z) dw − f (z) dw.
Cr2 w − z Cr1 w − z Cr2 w − z Cr1 w − z

The first integral on the right equals 2πif (z) by the Cauchy integral formula for the
constant function 1, and the second integral on the right is 0 since the integrand is
holomorphic in an open disk containing Cr1 and its interior.
We have the required result. 
Theorem 11.9. Let f beP a holomorphic function in the annulus R1 < |z−z0 | < R2 . Then
there is a Laurent series ∞ n
n=−∞ an (z − z0 ) such that in the annulus R1 < |z − z0 | < R2
∞
X
f (z) = an (z − z0 )n .
n=−∞

The coefficients an are given by

Z
1
an = (z − z0 )−n−1 f (z) dz, n ∈ Z,
2πi Cr

for any r satisfying R1 < r < R2 .

Note: The above Laurent series is known as the Laurent series representation of f
at z0 .
Proof. By the preceding result, if R1 < r1 < |z − z0 | < r2 < R2 , then
Z Z
1 f (w) 1 f (w)
f (z) = dw − dw.
2πi Cr2 w − z 2πi Cr1 w − z
Hence f = f1 + f2 , where
Z
1 f (w)
f1 (z) = dw, |z − z0 | < r2 ,
2πi Cr2 w−z
Z
1 f (w)
f2 (z) = − dw, |z − z0 | > r1 .
2πi Cr1 w − z

The function f1 has a power series expansion ∞ n

P
n=0 an (z − z0 ) in the disk |z − z0 | < r2
(by (3)) with
Z
1
an = (z − z0 )−n−1 f (z) dz, n ≥ 0,
2πi Cr2
P−1 n
and the function f2 has a Laurent series expansion n=−∞ an (z − z0 ) in the region
|z − z0 | < r2 (by (5)) with
Z
1
an = (z − z0 )−n−1 f (z) dz, n ≤ −1.
2πi Cr1
 LECTURE NOTES ON COMPLEX ANALYSIS I 45

By Theorem 11.6, the integrals defining the coefficients an , n ∈ Z are independent of

r1 , r2 ∈ (R1 , R2 ). Hence for any r ∈ (R1 , R2 ),
Z
1
an = (z − z0 )−n−1 f (z) dz, n ∈ Z.
2πi Cr
Since r1 , r2 are arbitrary in (R1 , R2 ), the Laurent series representation holds in the annulus
R1 < |z − z0 | < R2 . 
11.3. Isolated singularities. Definitions follow.
• Punctured disk centered at z0 ∈ C: For z0 ∈ C and a real number R > 0, the
collection of points z ∈ C with 0 < |z − z0 | < R is known as a punctured disk
centered at z0 .

• Punctured plane: For z0 ∈ C, the set C \ {z0 } is known as the punctured plane
centered at z0 .

• Isolated singularity: Let f be a holomorphic function on an open subset G of

C. We say that z0 ∈ C is an isolated singularity of f if z0 ∈
/ G but there is a
punctured disk centered at z0 which is contained in G.
Remark 11.10. Let R be the distance between z0 and C\G. Then the function f
is holomorphic in the punctured disk 0 < |z −z0 | < R which is P
contained in G. By
Theorem 11.9, f has a Laurent series representation f (z) = ∞ n=−∞ an (z − z0 )
n

convergent in the punctured disk.

• Principal part: The series −1 n
P
n=−∞ an (z − z0 ) is known as the principal part of
the Laurent series of f at z0 .

• An isolated singularity is classified into three types as follows:

– Removable singularity: An isolated singularity z0 of f is known as a
removable singularity if the principal part of the LS of f at z0 is 0.
In this case, f can be extended holomorphically to the open set G ∪ {z0 } by
defining f (z0 ) = a0 .
46 NIKITA AGARWAL
 LECTURE NOTES ON COMPLEX ANALYSIS I 47

12. Week 12
12.1. Isolated singularities. Some concepts from the last lecture were revisited.
• Isolated singularity: Let f be a holomorphic function on an open subset G of
C. We say that z0 ∈ C is an isolated singularity of f if z0 ∈
/ G but there is a
punctured disk centered at z0 which is contained in G.
Example 12.1. The function f (z) = z −1 has an isolated singularity at 0 since it
is holomorphic in a punctured disk at 0.
The function f (z) = z −1 (z − 1)−1 has an isolated singularity at 0 since it is holo-
morphic in the punctured disk 0 < |z| < 1 and it also has an isolated singularity
at 1 since it is holomorphic in a punctured disk 0 < |z − 1| < 1.
Remark 12.2. Let R be the distance between z0 and C \ G. Then the function
f is holomorphic in the punctured disk 0 < |z − z0 | < R (annulus) which is
contained
P∞ in G. By Theorem 11.9, f has a Laurent series representation f (z) =
n
n=−∞ an (z − z0 ) convergent in the punctured disk.

• Principal part: The series −1 n

P
n=−∞ an (z − z0 ) is known as the principal part of
the Laurent series of f at z0 .

• An isolated singularity is classified into three types as follows:

– Removable singularity: An isolated singularity z0 of f is known as a
removable singularity if the principal part of the LS of f at z0 is 0.
In this case, f can be extended holomorphically to the open set G ∪ {z0 } by
defining f (z0 ) = a0 .
z
Example 12.3. Consider the function f (z) = z on C \ {2πiZ}. The
e − 1
function ez − 1 in the denominator is an entire function having a simple zero
at the origin. Hence ez − 1 = zg(z) for some entire function g with g(0) 6= 0.
Hence the function 1/g is holomorphic in a neighbourhood of 0, and coincides
with f in a punctured disk centered at 0. Thus 0 is a removable singularity
of f .
Further, for each integer n 6= 0, z0 = 2πin is an isolated singularity of f .
Observe that the denominator ez − 1 has a simple zero at z0 , hence ez − 1 =
(z − z0 )h(z) for some entire function h with h(z0 ) 6= 0. Hence the function
z/h(z)
z/h(z) is holomorphic in an open disk D about z0 . Since f (z) = in
z − z0
the punctured disk D \ {z0 }, f has a simple pole at z0 .
– Pole: An isolated singularity z0 of f is known as a pole of order m ≥ 1 if in
the Laurent series of f at z0 (as given above), a−m 6= 0 and an = 0, for all
n < −m.
In this case, the principal part of the Laurent series of f at z0 is a nonzero
rational function given by
a−m a−m+1 a−1 a−m + a−m+1 z + · · · + a−1 z m−1
+ + · · · + = .
(z − z0 )m (z − z0 )m−1 z − z0 (z − z0 )m
– Simple pole: A pole of order 1 is known as a simple pole.
Example 12.4. The function f (z) = z −m has a pole of order m at 0.
Remark 12.5. Suppose f defined on G has a pole of order m at z0 , then
the function g(z) = (z − z0 )m f (z) has a removable singularity at z0 and can
48 NIKITA AGARWAL

be extended holomorphically on G ∪ {z0 } by setting g(z0 ) = a−m 6= 0. Thus

f (z) = (z − z0 )−m g(z), where g is holomorphic in G ∪ {z0 } and nonzero at z0 .
Thus the function 1/f has a removable singularity at z0 and its holomorphic
extension to a neighbourhood of z0 has a zero of order m at z0 .
Thus if z0 is a pole of f , then
lim |f (z)| = ∞.
z→z0

– Essential singularity: An isolated singularity z0 of f is known as an essen-

tial singularity if it is neither a removable nor a pole.
In this case, the principal part of Laurent series of f at z0 has infinitely many
nonvanishing terms.
Example 12.6. The function f (z) = e1/z defined on the punctured plane centered
at 0 has an essential singularity at 0.
• Punctured disk centered at ∞: For a real number R ≥ 0, the collection of
points z ∈ C with |z| > R is known as a punctured disk centered at ∞.

• Isolated singularity at ∞: We say that a function f has an isolated singularity

at ∞ if there is a punctured disk centered at ∞ on which f is holomorphic. Then
the point ∞ is called a removable singularity, pole of order m, or an essential
singularity of f is the function f (1/z) has the origin 0 as a removable singularity,
pole of order m, or an essential singularity, respectively.
Example 12.7. The point ∞ is a removable singularity of the function 1/z, a
pole of order m of the function z m , and an essential singularity of the function ez .
Theorem 12.8. (Riemann’s theorem for removable singularity) Let f be a holomorphic
function in an open subset G of C having an isolated singularity at z0 . If f is bounded in
a punctured disk centered at z0 , then z0 is a removable singularity of f .
Proof. Let R > 0 be such that the punctured disk B(z0 , R) \ {z0 } is contained in G and
|f | be
Pbounded above by M > 0 in the punctured disk B(z0 , R) \ {z0 }.
∞
Let n=−∞ na (z − z0 )n be the Laurent series expansion of f in the punctured disk
B(z0 , R) \ {z0 }.
Let 0 < r < R. We know that
Z
1
a−n = (z − z0 )n−1 f (z) dz,
2πi Cr
where Cr is the circle |z − z0 | = r oriented counterclockwise. Thus
1 n−1
|a−n | ≤ r M L(Cr ) = rn M.
2π
Let n ≥ 1. Then taking r → 0+ , we get a−n = 0. 
Theorem 12.9. (Criterion for a pole) Let f be a holomorphic function in an open subset
G of C having an isolated singularity at z0 . If limz→z0 |f (z)| = ∞, then z0 is a pole of f .
Proof. Since limz→z0 |f (z)| = ∞, there exist R > 0 and M > 0 such that |f (z)| ≥ M for
z in the punctured disk 0 < |z − z0 | < R.
Hence the function 1/f is well-defined and bounded in the punctured disk 0 < |z−z0 | < R.
Thus by the preceding result, z0 is a removable singularity of 1/f and the holomorphic
extension g of 1/f in the disk |z − z0 | < R satisfied g(z0 ) = limz→z0 1/f (z) = 0.
Suppose z0 is a zero of g of order m ≥ 1.
Then g(z) = (z − z0 )m h(z), where h is a holomorphic function in the disk |z − z0 | < R
 LECTURE NOTES ON COMPLEX ANALYSIS I 49

with h(z0 ) 6= 0.
Thus h does not vanish in a disk |z − z0 | < r for some 0 < r < R.
1/h(z)
Substituting g = 1/f , we get f (z) = , in the punctured disk 0 < |z − z0 | < r.
(z − z0 )m
Since 1/h is holomorphic in the disk |z − z0 | < r, it has a power series expansion in it.
Thus z0 is a pole of f . The order of this pole is m since h(z0 ) 6= 0. 
Theorem 12.10. (Casorati-Weierstrass theorem) Let f be a holomorphic function in
an open subset G of C having an essential singularity at z0 . Then for any complex
number w, there exists a sequence (zn )n≥1 ⊆ G such that zn → z0 and f (zn ) → w.
Note: The above result implies that for every punctured disk D in G centered at z0 ,
Cl(f (D)) = C.
Proof. We prove the result by contradiction.
Suppose for some w ∈ C, there exist r > 0 and M > 0 such that |f (z) − w| > M for all
0 < |z − z0 | < r.
1
Hence the function g(z) = is holomorphic and bounded in the punctured disk
f (z) − w
0 < |z − z0 | < r.
Thus z0 is a removable singularity of g.
Note that f (z) = w + 1/g(z).
If g(z0 ) 6= 0, z0 is a removable singularity of f .
On the other hand, if g(z0 ) = 0, z0 is a pole of f . 

12.2. Residue theory.

Definition 12.11. (Index of a curve and winding number) Let γ be a closed
rectifiable curve in C. For a ∈
/ γ, the number
Z
1
n(γ; a) = (z − a)−1 dz,
2πi γ
is known as the index of γ with respect to the point a (or the winding number of γ around
a).
Example 12.12. 1) Let γ be a circle centered at a oriented counterclockwise, then
n(γ; a) = 1.
2) Let γ be the figure of eight given by attaching the circles γ1 given by |z − 1| = 1 in
counterclockwise direction and γ2 given by |z + 1| = 1 in clockwise direction (draw the
figure). Then for a ∈
/ γ,

−1, |a + 1| < 1

n(γ; a) = n(γ1 ; a) + n(γ2 ; a) = 1, |a − 1| < 1

0 otherwise.
Theorem 12.13. Let γ : [0, 1] → C be a closed rectifiable curve and a ∈
/ γ. Then n(γ; a)
is an integer.
Proof. We prove the result only for piecewise C 1 curve γ.
Define a function g : [0, 1] → C as
Z t
g(t) = (γ(s) − a)−1 γ 0 (s) ds.
0
50 NIKITA AGARWAL

Observe that g(0) = 0 and g(1) = 2πin(γ; a).

We also get
g 0 (t) = (γ(t) − a)−1 γ 0 (t), 0 < t < 1.
Hence
d −g(t)
(γ(t) − a) = e−g(t) γ 0 (t) − e−g(t) g 0 (t)(γ(t) − a) = 0.


e
dt
−g(t)
Hence e (γ(t)−a) is a constant function in [0, 1]. Thus e−g(1) (γ(1)−a) = e−g(1) (γ(1)−
a). Since γ(1) = γ(0) 6= a, we get e−g(1) = 1. Hence g(1) = 2πik, for some k ∈ Z. 
Theorem 12.14. Let γ : [0, 1] → C be a closed rectifiable curve and a ∈
/ γ. Then n(γ; a)
is constant for all a belonging to a component of G = C \ γ. Also n(γ; a) = 0, for all a
belonging to the unbounded component of G.
Proof. Define f : G → Z as f (a) = n(γ; a). We prove that f is a continuous function
which would give the required result.
Firstly observe that each component of G is open.
Let a ∈ G and r = dist(a, γ) > 0. Let b be such that |a − b| < δ < r (δ > 0 will be chosen
later). Then
Z
1
|f (a) − f (b)| = (z − a)−1 − (z − b)−1 dz
2π γ
a−b
Z
1
= dz
2π γ (z − a)(z − b)
|a − b|
Z
1
≤ dz .
2π γ (z − a)(z − b)

Since |z − a| ≥ r > r/2, |z − b| ≥ |z − a| − |a − b| > r/2. Hence

2δ
|f (a) − f (b)| < 2 L(γ).
πr
2
 
πr 
Given  > 0. Choose δ < min r, .
2δ
Since γ is a closed rectifiable curve, it is contained in B(0, r) for some r > 0. Hence G
contains all z with |z| > r. This also shows that G has a unique unbounded component.
Denote it as U .
(A different argument than what was discussed in the class) Let  > 0 be given and a ∈ U
(we will choose a shortly). Consider
Z
1 L(γ)
|n(γ; a)| = (z − a)−1 dz < .
2π γ 2 dist(a, γ)π
L(γ)
If a ∈ U is such that dist(a, γ) > , then |n(γ; a)| < . Such an a ∈ U can be chosen
2π
since U is unbounded. 
Definition 12.15.P (Residue) Letn f has an isolated singularity at z0 and has Laurent
series expansion ∞ n=−∞ n a (z − z0 ) near1
z0 . The residue of f at z0 is the coefficient a−1 .
It is denoted by Res(f ; z0 ).
Remark 12.16. 1) If z0 is a removable singularity of f , then the residue of f at z0 is 0.
Thus the residue is nonzero only when z0 is either a pole or an essential singularity.
1The Laurent series which is valid in the some punctured disk 0 < |z − z0 | < R, for some R > 0.
 LECTURE NOTES ON COMPLEX ANALYSIS I 51

2) Recall from our earlier discussion that if the Laurent series is valid in the punctured
disk 0 < |z − z0 | < R, then for all 0 < r < R,
Z
1
Res(f ; z0 ) = a−1 = f (z) dz,
2πi Cr

where Cr is the circle |z − z0 | = r oriented counterclockwise.

Remark 12.17. 1) Let z0 be a simple pole for f , then

(z − z0 )f (z) = a−1 + a0 (z − z0 ) + . . . .

Hence Res(f ; z0 ) = limz→z0 (z − z0 )f (z).

2) If z0 be a pole of order m ≥ 2 for f , then

(z − z0 )m f (z) = a−m + a−m+1 (z − z0 ) + · · · + a−1 (z − z0 )m−1 + . . . ,

in a punctured disk at z0 . Thus

dm−1
lim m−1
((z − z0 )m f (z)) = (m − 1)!a−1 .
z→z0 dz
Hence
1 dm−1
Res(f ; z0 ) = lim m−1 ((z − z0 )m f (z)) .
(m − 1)! z→z0 dz

z3
Example 12.18. 1) (not discussed in class) Compute the residue of f (z) = at
z−1
z0 = 1:
Since 1 is a simple pole for f , we get

Res(f ; 1) = lim (z − 1)f (z) = lim z 3 = 1.

z→1 z→1

1
2) Compute the residue of f (z) = at z0 = i:
(z 2
+ 1)2
Since i is a pole of order 2 for f , we get
 
d 2

d 1 −2 −i
Res(f ; i) = lim (z − i) f (z) = lim 2
= lim 3
= .
z→i dz z→i dz (z + i) z→i (z + i) 4

Theorem 12.19. (Cauchy’s theorem) Let G be an open subset of C. Let f be holomorphic

in G. Then γ f (z) dz = 0, for every closed rectifiable curve in G whose interior2 is
R

contained in G.

Proof. Let γ be a simple (does not intersect itself except at the end points) closed recti-
fiable curve in G.
Suppose γ(t) = x(t) + iy(t), a ≤ t ≤ b, be a parametrization of γ.
Let R be the interior of γ.
Also let f (x + iy) = u(x, y) + iv(x, y), where u and v are the real and imaginary parts of

2region enclosed by the curve

52 NIKITA AGARWAL

f . Then
Z Z b
f (z) dz = f (γ(t))γ 0 (t) dt
γ a
Z b
= ((u(x(t), y(t))x0 (t) − v(x(t), y(t))y 0 (t)) dt
a
Z b
+i (u(x(t), y(t))y 0 (t) + v(x(t), y(t))x0 (t)) dt
Z a Z
= (u dx − v dy) + i (u dy − v dx)
γ γ
Z Z
= (−vx − uy ) dx dy + i (ux − vy ) dx dy (Green’s theorem3)
R R
= 0 (Cauchy-Riemann equations).
Since any closed curve γ which is not simple can be expressed as a union of simple closed
curves, the result holds true for all closed rectifiable curves. 
Remark 12.20. (Notation) For two curves γ1 and γ2 , we use the notation γ1 + γ2 to
represent the curve by joining the curves γ1 and γ2 , if this is possible.
By −γ1 , we mean the reverse of the curve γ.
Theorem 12.21. (Extended Cauchy’s theorem) Let f be a holomorphic function in an
open subset G of C. Let γ1 , γ2 be two simple closed rectifiable curves in G both oriented
counterclockwise, enclosing a region which is contained in G. Then
Z Z
f (z) dz = f (z) dz.
γ1 γ2

Note: We are considering curves as shown in the figure below.

Proof. Look at the figure. Join γ1 and γ2 with two curves S2 (joining P to P 0 ) and S4
(joining Q0 to Q).
Let γ10 be the portion of the curve γ1 by removing S1 . Hence
Z Z Z
f= f− f.
γ10 γ1 S1

Let γ20 be the portion of the curve γ2 by removing −S3 . Hence

Z Z Z
f= f+ f.
γ20 γ2 S3
 LECTURE NOTES ON COMPLEX ANALYSIS I 53

Let S be the closed rectifiable curve given by S1 + S2 + S3 + S4 . Hence

Z Z Z Z Z
f= f+ f+ f+ f.
S S1 S2 S3 S4

Consider the closed rectifiable curve γ given by γ10 − S4 − γ20 − S2 .

Since the region enclosed by γ and also the region enclosed by S is contained in G, by
the Cauchy’s theorem, Z Z
f = 0, f = 0.
γ S
Now
Z Z Z Z Z Z Z Z Z Z Z
0= f= f− f− f− f= f− f− f− f− f− f.
γ γ10 S4 γ20 S2 γ1 S1 S4 γ2 S3 S2

Combining terms of integral over S1 , S2 , S3 , S4 to get integral over S. Using the above
equality for S, we get Z Z
f= f.
γ1 γ2

Corollary 12.21.1. For any simple closed rectifiable curve γ in C oriented counterclock-
wise enclosing a point a, n(γ; a) = 1.
Proof. Let R = dist(a, γ) > 0. Let 0 < r < R and let Cr be the circle |z − a| = r oriented
counterclockwise. By the extended Cauchy’s theorem, n(γ; a) = n(Cr ; a) = 1. 
Theorem 12.22. Let f be a holomorphic function on an open subset G of C with zeros
z1 , . . . , zk counting multiplicities. Let γ be a closed rectifiable curve oriented counterclock-
wise in G which does not pass through any of the zeros z1 , . . . , zk and whose interior is
contained in G. Then
Z 0 k
1 f (z) X
dz = n(γ; zj ).
2πi γ f (z) j=1

Proof. The function f can be written as f (z) = (z − z1 ) . . . (z − zk )g(z), where g is

holomorphic in G and g(z) 6= 0, for all z ∈ G. Thus
f 0 (z) 1 1 g 0 (z)
= + ··· + + , z 6= z1 , . . . , zk .
f (z) z − z1 z − zk g(z)
Since g 0 /g is holomorphic
R 0 in G and the interior of γ is contained in G, by the Cauchy’s
theorem, we get γ g /g = 0. Hence we obtain the desired result. 

Remark 12.23. 1) Let w ∈ C. Let f be a holomorphic function on an open subset G of

C. Let z1 , . . . , zk be the solutions of the equation f (z) = w counting multiplicities. Let γ
be a closed rectifiable curve in G which does not pass through any of the zeros z1 , . . . , zk
and whose interior is contained in G. Then
k
f 0 (z)
Z
1 X
dz = n(γ; zj ).
2πi γ f (z) − w j=1

Apply the preceding result to the function f (z) − w in place of f (z).

2) (Counting zeros) If γ is a simple closed rectifiable curve in G containing no zeros of
f , and whose interior is contained in G. Since n(γ; zj ) equals 1 if zj lies in the interior
54 NIKITA AGARWAL

of γ and equals 0 otherwise, by the preceding result, the number of zeros (counting with
multiplicities) of f in the interior of γ is given by
Z 0
1 f (z)
dz.
2πi γ f (z)
Example 12.24. Let γ be the circle |z| = 3 oriented counterclockwise. Let us compute
R 2z + 2
γ 2
dz.
z + 2z
Take f (z) = z 2 + 2z in the preceding remark on counting zeros. Since f has two zeros
0, −2 in the interior of γ, both with multiplicities 1, we get
Z
2z + 2
2
dz = 4πi.
γ z + 2z

The following theorem, known as the Residue theorem, extends the Cauchy’s theorem
to situations where the interior of the curve contains isolated singularities of f .
Theorem 12.25. (Residue theorem) Let f be a holomorphic function in a connected open
subset G of C. Let γ be a simple closed rectifiable curve oriented counterclockwise in G
such that the interior of γ is contained in G except for the isolated singularities z1 , . . . , zk
of f . Then
Z Xk
f = 2πi Res(f ; zj ).
γ j=1

Proof. Let us first consider the situation when the interior of γ contains only one singu-
larity z1 . Let r1 > 0 be such that the punctured disk 0 < |z − z1 | < r1 is contained in the
interior of γ and the Laurent series expansion of f at z1 is valid in it. Let C1 be the circle
 LECTURE NOTES ON COMPLEX ANALYSIS I 55

|z − z1 | = r1 /2 oriented counterclockwise. By the extended Cauchy’s theorem, since the

region enclosed by the curves γ and C1 is contained in G, we have
Z Z
f= f = 2πi Res(f ; z1 ),
γ C1

by Remark 12.16.
Now assume the result is true for a curve for which k − 1 singularities z1 , . . . , zk−1 lying
in its interior.
Let us consider the case when γ contains z1 , . . . , zk in its interior. Let rk > 0 be such that
the punctured disk 0 < |z − zk | < rk is contained in the interior of γ, does not contain
any of the singularities z1 , . . . , zk−1 , and the Laurent series expansion of f at zk is valid
in it. Let Ck be the circle |z − zk | = rk /2 oriented counterclockwise. Let Ck0 be the curve
obtained by removing −S4 from the circle Ck . Let γ 0 be the curve obtained by removing
S2 from the curve γ. Consider the simple closed rectifiable curve γ 00 starting at the point
P given by joining −S1 , −Ck0 , −S3 , and γ 0 . Since it encloses k − 1 singularities, we have
Z k−1
X
f = 2πi Res(f ; zj ).
γ 00 j=1

Observe that Z Z Z Z
f= f+ f+ f,
γ γ 00 Ck S
where S = S1 + S2 + S3 + S4 is a closed rectifiable curve whose interior is contained in
G. By the Cauchy’s theorem, Z
f = 0.
S
Further, by Remark 12.16, Z
f = 2πi Res(f ; zk ).
Ck
Hence we have the required result. 
1
Example 12.26. (Not discussed in the class) Consider the function f (z) = .
z(z 2+ 1)
1
Observe that f (z) = . The function f has three isolated singularities at
z(z + i)(z − i)
0, ±i, all of which are poles. Let us compute the residues of f at each of these points.
Since each of the poles are simple,
1
Res(f, 0) = lim zf (z) = 1, Res(f, i) = lim(z − i)f (z) = 2 = −1/2,
z→0 z→i 2i
Res(f, −i) = lim (z + i)f (z) = 1/2.
z→−i

Hence, by the Residue theorem, Z

f = 0,
C1
Z
f = 2πiRes(f ; 0) = 2πi,
C2
Z
f = 2πi (Res(f ; 0) + Res(f ; i)) = πi,
C3
Z
f = 2πi (Res(f ; 0) + Res(f ; −i)) = 3πi,
C4
56 NIKITA AGARWAL

Z
f = 2πi (Res(f ; 0) + Res(f ; i) + Res(f ; −i)) = 2πi,
C5
 LECTURE NOTES ON COMPLEX ANALYSIS I 57

13. Week 13
Theorem 13.1. (Open mapping theorem) Let G be an open connected subset of C.
Let f be a nonconstant holomorphic function on G. Then f is an open map (image of
an open set in G is open).
Proof. Let U be open in G, z0 ∈ U and let w0 = f (z0 ). We prove that there exists an
open ball around w0 which is contained in f (U ).
Note that z0 is a zero of f (z) − w0 , let its multiplicity be m ≥ 1.
Since the zeros of a nonconstant holomorphic function are isolated, there exists δ > 0
such that f (z) − w0 has no zero in 0 < |z − z0 | ≤ δ contained in U .
Let Cδ denote the circle |z − z0 | = δ oriented counterclockwise. Let d = dist(w0 , f (Cδ )) >
0.
On the disk B(w0 , d/2), define the function
f 0 (z)
Z
1
N (w) = dz.
2πi Cδ f (z) − w
This is well-defined and is continuous in w: Let w, w0 ∈ B(w0 , d/2). For each z ∈ Cδ ,
we have |f (z) − w| ≥ |f (z) − w0 | − |w − w0 | > d − d/2 = d/2 and |f (z) − w0 | ≥
|f (z 0 ) − w0 | − |w − w0 | > d − d/2 = d/2. Let M = max{|f 0 (z)| : z ∈ Cδ }. Then
4M L(Cδ ) |w − w0 |
|N (w) − N (w0 )| = .
2πd2
Further since N (w) is the number of solutions counting multiplicity of the equation
f (z) = w, for z in the disk B(w0 , d/2), the function N : B(w0 , d/2) → Z. Hence N is
constant in the (connected) disk B(w0 , d/2).
Since N (w0 ) = m, we get N (w) = m for all w ∈ B(w0 , d/2).
Hence for each w ∈ B(w0 , d/2), there exists z ∈ B(z0 , δ) such that f (z) = w. Therefore
B(w0 , d/2) ⊆ f (U ). Thus f (U ) is open. 
13.1. Calculation of improper integrals using the Residue theorem.
Example 13.2. Show that Z ∞
x2 √
dx = π/ 2.
−∞ 1 + x4
z2
Consider the function f (z) = . It has 4 simple poles given by the fourth roots of
1 + z4
−1: eiθ , θ ∈ {π/4, 3π/4, 5π/4, 7π/4}. Let us denote these by z1 , z2 , z3 , z4 . Then
z12 1−i
Res(f ; z1 ) = lim (z − z1 )f (z) = = √ ,
z→z1 (z − z2 )(z − z3 )(z − z4 ) 4 2
z22 −1 − i
Res(f ; z2 ) = lim (z − z2 )f (z) = = √ .
z→z2 (z − z1 )(z − z3 )(z − z4 ) 4 2
Let R > 1 and γ be the closed curve obtained by joining γ1 (t) = t, −R ≤ t ≤ R and
γ2 (t) = Re2πit , 0 ≤ t ≤ 1/2. Then
Z
π
f = 2πi(Res(f ; z1 ) + Res(f ; z2 )) = √ .
γ 2
The above is true for all R > 1.
The integral of f on γ1 is
Z R
x2
4
dx,
−R 1 + x
58 NIKITA AGARWAL

which by taking R → ∞ gives the given improper integral.

The integral of f on γ2 is
Z 1/2
3 e4πit
2πiR dt.
0 1 + R4 e8πit
Observe that |1 + R4 e8πit | ≥ R4 − 1, for all t ∈ [0, 1/2]. Hence
Z 1/2
3 e4πit πR3
2πiR dt ≤ → 0,
0 1 + R4 e8πit R4 − 1
as R → ∞.
Remark 13.3. The technique used in the preceding example works to compute any
improper integral of the form Z ∞
P (x)
dx,
−∞ Q(x)
where P, Q are polynomials in x with the degree of Q greater than or equal to the degree
of P . Also Q having no zeros on the real axis.
Choose γ to be the closed boundary of a semicircle in the upper half plane which contains
all the poles of the rational function P (z)/Q(z) in the upper half plane.
Example 13.4. Show that for a > 0,
Z ∞
cos ax
2
dx = πe−a .
−∞ 1 + x

eiaz
Consider the function f (z) = . It has 2 simple poles i, −i. Let R > 1. Take γ as
1 + z2
in the preceding example. Then
Z
f = 2πiRes(f ; i) = lim(z − i)f (z) = πe−a .
γ z→i

The above is true for all R > 1.

Since R2 − 1 = |z|2 − 1 ≤ |1 + z 2 | and |eiaz | = e−a Im(z) ≤ 1 (since a > 0 and Im(z) ≥ 0)
for all z ∈ γ2 , the integral of f on γ2 satisfies
eiaz
Z
πR
2
dz ≤ 2 → 0,
γ2 1 + z R −1
as R → ∞.
The integral of f on γ1 is
Z R iax Z R Z R
e cos ax sin ax
2
dx = 2
dx + i 2
dx
−R 1 + x −R 1 + x −R 1 + x
The second integral is 0 since the integrand is an odd function in x.
By taking R → ∞ in the first integral, we obtain the required improper integral.
 LECTURE NOTES ON COMPLEX ANALYSIS I 59

Remark 13.5. The technique used in the preceding example works to compute any
improper integral of the form Z ∞
P (x) cos ax
dx,
−∞ Q(x)
where a > 0, P, Q are polynomials in x with the degree of Q greater than or equal to the
degree of P . Also Q having no zeros on the real axis.
Choose γ to be the closed boundary of a semicircle circle in the upper half plane which
contains all the poles of the rational function P (z)eiaz /Q(z) in the upper half plane.
Example 13.6. Show that Z ∞
log x
dx = 0.
0 1 + x2
Let us consider the branch g of the logarithm function on the set G = {z ∈ C \ {0} :
arg(z) ∈ (−π/2, 3π/2)} given by g(reiθ ) = log r + iθ.
log z
Consider the function f (z) = . It has 2 simple poles: ±i. Let 0 < r < 1 < R.
1 + z2

Consider the closed curve γ as shown in the figure containing the pole i of f with residue
π/4. Then by the Residue theorem,
π2i
Z
f= .
γ 2
Also
R π Z −r Z 0
log |x| + πi
Z Z Z
log x log R + iθ iθ log r + iθ iθ
f= dx+iR e dθ+ dx+ir e dθ.
γ r 1 + x2 0 1 + R2 e2iθ −R 1 + x2 π 1+r e
2 2iθ

Observe that
Z R Z −r Z R Z −r
log x log |x| + πi log x 1
2
dx + 2
dx = 2 2
dx + πi 2
dx.
r 1+x −R 1+x r 1+x −R 1 + x
R ∞ dx
Letting r → 0 and R → ∞ and 0 = π/2, we get
1 + x2
Z ∞ Z π Z π
log x log R + iθ iθ log r + iθ iθ
2 2
dx = − lim iR 2 2iθ
e dθ + lim+ ir 2 2iθ
e dθ.
0 1+x R→∞ 0 1+R e r→0 0 1+r e

The two terms on the right are both 0 since for ρ > 0 and θ ∈ [0, π],
Z π Z π
log ρ + iθ iθ πρ| log ρ| ρ πρ| log ρ| π2ρ
ρ 2 2iθ
e dθ ≤ + θ dθ = + → 0,
0 1+ρ e |1 − ρ2 | |1 − ρ2 | 0 |1 − ρ2 | 2|1 − ρ2 |
when ρ → ∞ or ρ → 0+ .
Example 13.7. Show that Z ∞
sin x π
dx = .
0 x 2
60 NIKITA AGARWAL

eiz
Consider the function f (z) = , which has a simple pole at 0. Consider the closed curve
z
γ as shown. Since the interior of γ does not contain the pole 0, by the Residue theorem,

R Z −r ix Z R
eix
Z Z Z Z Z Z
e sin x
0= f = dx + f+ dx + f = 2i dx + f+ f,
γ r x γR −R x γr r x γR γr
cos x
using the fact that is an odd function.
x
Consider now
Z Z π Z π Z π/2
iReiθ −R sin θ
f = i e dθ ≤ e dθ = 2 e−R sin θ dθ.
γR 0 0 0

Now use sin θ ≥ 2θ/π, for 0 ≤ θ ≤ π/2, we get

Z π/2
π(1 − e−R )
Z
f ≤2 e−2Rθ/π dθ = → 0,
γR 0 R
as R → ∞.
eiz − 1
Now observe that the function h(z) = has a removable singularity at 0, hence the
z
function h is bounded near 0. Let Mr be an upper bound for |h| on γr (r small), then
Z
h ≤ L(γr )Mr → 0,
γr
as r → 0. Hence Z Z Z
f= 1/z dz + h → πi,
γr γr γr
as r → ∞. Summarizing the above, we get the required integral.
 LECTURE NOTES ON COMPLEX ANALYSIS I 61

14. Week 14
Example 14.1. Show that for −1 < a < 1,
Z ∞
xa πa
dx = .
0 (1 + x)2 sin πa
Consider the branch of the function z a on G = C \ [0, ∞) given by
g(z) = ra eiaθ , for z = reiθ , θ ∈ (0, 2π).
g(z)
Let f (z) = defined on G. The function f has one pole −1 of order 2. Let
(1 + z)2

0 < δ < r < 1 < R. Consider the simple closed curve γ as shown in the figure which is
composed
√ of four curves: γ1 line segment joining r + iδ to R + iδ, γ2 portion of the circle
|z| = R2 + δ 2 oriented counterclockwise from the point R +√ iδ to R − iδ, γ3 line segment
joining R − iδ to r − iδ, and γ4 portion of the circle |z| = r2 + δ 2 oriented clockwise
from the point r − iδ to r + iδ.
The interior of γ has one pole −1 of f of order 2, hence by the Residue theorem,
Z
0
f = 2πiRes(f ; −1) = lim (z + 1)2 f (z) = 2πiaei(a−1)π = −2πiaeiaπ .

γ z→−1

The above is true for all 0 < δ < r < 1 < R.

The integral of f over the portion of circles γ2 and γ4 can be bounded above as follows:
√
Z
( R2 + δ 2 )a √
f ≤ √ 2π R2 + δ 2 → 0,
γ2 ( R2 + δ 2 − 1)2
as δ → 0 and R → ∞ since −1 < a < 1, and
√
Z
( r2 + δ 2 )a √
f ≤ √ 2π r2 + δ 2 → 0,
γ4 ( r2 + δ 2 − 1)2
as δ → 0 and r → 0 since −1 < a < 1.
Now Z R Z R
(t + iδ)a ta
Z
f= 2
dt → 2
dt,
γ1 r (1 + t + iδ) r (1 + t)
as δ → 0 (Lemma 14.2). Hence
Z ∞
ta
Z
f→ dt,
γ1 0 (1 + t)2
62 NIKITA AGARWAL

as r → 0 and R → ∞. Also,
R R
(t − iδ)a ta ei2πa
Z Z Z
f =− dt → − dt,
γ3 r (1 + t − iδ)2 r (1 + t)2
as δ → 0 (Lemma 14.2). Hence
∞
ta ei2πa
Z Z
f →− dt,
γ3 0 (1 + t)2
as r → 0 and R → ∞.
Combining all of the above, we get
∞
ta
Z
iaπ i2πa
−2πiae = (1 − e ) dt.
0 (1 + t)2
Hence
∞
ta −2πiaeiaπ
Z
πa πa
2
dt = i2πa
= iπa −iπa
= .
0 (1 + t) 1−e e −e sin πa
2i
Lemma 14.2. 1) For t ∈ [r, R] and δ ∈ [0, r], define the function

g(t + iδ)
, δ 6= 0,


2

k1 (t, δ) = (1 + t + iδ)
ta

 , δ = 0.
(1 + t)2


Then k1 is uniformly continuous on [r, R] × [0, r].

2) For t ∈ [r, R] and δ ∈ [0, r], define the function

g(t − iδ)
, δ 6= 0,


(1 + t − iδ)2

k2 (t, δ) =
ta ei2πa
, δ = 0.


(1 + t)2


Then k2 is uniformly continuous on [r, R] × [0, r].

Proof. Observe that g(t + iδ) → ta as δ → 0 and g(t − iδ) → ta ei2πa as δ → 0. Hence
the functions k1 , k2 are continuous in [r, R] × [0, r]. Since the domain of the functions are
compact, the functions are uniformly continuous. 
Definition 14.3. (Meromorphic function) Let G be an open subset of C. A complex
valued function f is known to be meromorphic function in G if it is defined and
analytic in G except for poles.
Remark 14.4. A meromorphic function f in G can be extended to a continuous function
from G to C∞ = C ∪ {∞} by defining f (z0 ) = ∞ whenever z0 ∈ G is a pole of f since
limz→z0 |f (z)| = ∞.
Theorem 14.5. (Argument principle) Let f be a meromorphic function in G. Let γ be
a simple closed rectifiable curve in G oriented counterclockwise, not containing any pole
or zero of f and such that the interior of γ is contained in G. Then
Z 0
1 f (z)
dz = Zf − Pf ,
2πi γ f (z)
where Zf denote the number of zeros of f and Pf denote the number of poles of f in the
interior of γ (both counting multiplicities).
 LECTURE NOTES ON COMPLEX ANALYSIS I 63

Proof. Let z1 , . . . , zn be the zeros of f and p1 , . . . , pm be the poles of f in the interior of

γ counting multiplicities. Observe that
f (z) = (z − z1 ) . . . (z − zn )g(z),
where g has no zeros in the interior of γ and has poles p1 , . . . , pm in the interior of γ.
Hence
(z − z1 ) . . . (z − zn )
f (z) = h(z),
(p − p1 ) . . . (p − pm )
where h is an analytic function with no zeros in the interior of γ. Hence
n m
f 0 (z) X 1 X 1 h0 (z)
= − + .
f (z) j=1
z − zj k=1 z − pk h(z)
Integrating over γ, we have the desired result since
Z Z
1 1
dz = dz = 1,
γ z − zj γ z − pk

and
h0 (z)
Z
dz = 0,
γ h(z)
by the Cauchy’s theorem. 
Remark 14.6. 1) If f is holomorphic in G, then the Argument principle gives Re-
mark 12.23.
f 0 (z)
2) If log f (z) was defined, then it would be the primitive of . Also log f (z) =
f (z)
log |f (z)| + iArg(f (z)).
By the argument principle, the integral is 2πiK for some K ∈ Z. Hence as z goes around
γ, log f (z) would change by 2πiK since |f (z)| comes back to its initial point.
Hence the argument of f (z) changes by 2πK.
Example 14.7. (Zeros but no poles) Let n ≥ 1. Let C be the unit circle with counter-
clockwise orientation. Compute
z n−1
Z
n
dz.
C 3z − 1
Let f (z) = 3z n − 1. Then f has no zero on C. Hence by the argument principle,
z n−1 3nz n−1
Z Z
2πi 1 2πi
n
dz = n
dz = n = 2πi/3,
C 3z − 1 3n 2πi C 3z − 1 3n
since the number of zeros of f in the interior of C is n.
R f 0 (z)
Example 14.8. (Both zeros and poles) Evaluate the integral C
dz for
f (z)
(z − 1)10 (z − 3)(2z − i)
f (z) = ,
(z − i)5 (z + i)7
where C is the circle |z| = 2 oriented counterclockwise.
The function f has a simple zero at i/2 and a zero at 1 with multiplicity 10, in the interior
of C. Hence Zf = 11. Further, it has a pole of order 5 at i and a pole of order 7 at −i,
in the interior of C. Hence Pf = 12. Thus
Z 0
f (z)
dz = 2πi(Zf − Pf ) = −2πi.
C f (z)
64 NIKITA AGARWAL

Theorem 14.9. (Rouche’s theorem) Let γ be a simple closed rectifiable curve in G. Let
f, g be two meromorphic functions in an open set containing the curve γ and its interior.
Let γ contains neither a pole nor zero of f and g. Let Zf , Zg be the number of zeros
and Pf , Pg be the number of poles of f, g, respectively in the interior of γ (counting
multiplicities). Let
|f (z) + g(z)| < |f (z)| + |g(z)|,
for all z ∈ γ, then
Zf − Pf = Zg − Pg .
Proof. On γ,
f (z) f (z)
+1 < + 1.
g(z) g(z)
If f (z)/g(z) = λ ≥ 0 (a nonnegative real number), then the above strict inequality
becomes λ + 1 < λ + 1, which is a contradiction, hence f (z)/g(z) is never a nonnegative
real number for z ∈ γ.
Let Ω = C \ [0, ∞).
Since (f /g)(γ) is a compact subset of the open set Ω, there exists an open set V in Ω
containing (f /g)(γ). Now U = G ∩ (f /g)−1 (V ) is an open set containing γ whose image
is contained in Ω.
Thus the meromorphic function f /g maps the open set U containing γ into Ω.
If h is a branch of the logarithm on Ω, then h(f /g) is a primitive of (f /g)0 /(f /g) in an
open set containing γ. Thus
Z Z Z
0 = (h(f /g)) = (f /g) /(f /g) = (f 0 /f − g 0 /g) = 2πi(Zf − Pf ) − 2πi(Zg − Pg ),
0 0
γ γ γ

by the argument principle if γ has counterclockwise orientation. There will be a negative

sign if the orientation of γ is clockwise, but that will not affect the final answer. 
Corollary 14.9.1. Let γ be a simple closed rectifiable curve in G. Let f, g be two mero-
morphic functions in an open set containing the curve γ and its interior. Let γ contains
neither a pole nor zero of f and g. Let Zf , Zg be the number of zeros and Pf , Pg be the
number of poles of f, g, respectively in the interior of γ (counting multiplicities). Let
|f (z) − g(z)| < |f (z)|,
for all z ∈ γ, then
Zf − Pf = Zg − Pg .
Note: If any of the inequality in the statement of Rouche’s theorem and its corollary:
|f (z) + g(z)| < |f (z)| + |g(z)|,
and
|f (z) − g(z)| < |f (z)|,
is true for all z ∈ γ, then neither f (z) nor g(z) is 0 for any z ∈ γ.
Remark 14.10. (Proof of the Fundamental theorem of algebra using Rouche’s theorem)
Let p(z) = a0 + a1 z + · · · + an−1 z n−1 + z n be a polynomial. Then
lim p(z)/z n = a0 /z n + a1 /z n−1 + · · · + an−1 /z + 1 = 1.
|z|→∞

Hence there exists R large enough such that

|p(z)/z n − 1| < 1,
 LECTURE NOTES ON COMPLEX ANALYSIS I 65

for |z| = R. Hence for |z| = R,

|p(z) − z n | < |z|n ≤ |p(z)| + |z|n .
Thus p has n zeros in |z| < R.
Clearly there are no zeros of p in |z| ≥ R, otherwise it will violate the above strict
inequality |p(z) − z n | < |z|n .
Example 14.11. Count the number of zeros with multiplicities of the polynomial p(z) =
z 7 + 4z 4 + z 3 + 1 in the regions |z| < 1 and 1 < |z| < 2.
Observe that
|z 7 + z 3 + 1| ≤ 3 < 4 = |4z 4 |,
on |z| = 1. Hence taking g(z) = p(z) and f (z) = 4z 4 , by the Corollary to the Rouche’s
theorem, we get that the number of zeros of p equal the number of zeros of f in the disk
|z| < 1 (counting multiplicites). Hence the number of zeros of p in the disk |z| < 1 is 4.
Now observe that
|4z 4 + z 3 + 1| ≤ 26 + 23 + 1 = 27 (1/2 + 1/24 + 1/27 ) < 27 = |z 7 |,
on |z = 2|. Hence taking g(z) = p(z) and f (z) = z 7 , by the Corollary to the Rouche’s
theorem, we get that the number of zeros of p equal the number of zeros of f in the disk
|z| < 2 (counting multiplicites). Hence the number of zeros of p in the disk |z| < 2 is 7.
Thus the number of zeros of f in the annulus 1 < |z| < 2 (counting multiplicites) is 3.
Example 14.12. Let f be a holomorphic function in the disk |z| < 1 and let |f (z)| < 1
on the circle |z| = 1. Then f has a unique fixed point in the disk |z| < 1.
Observe that
|f (z) − z + z| = |f (z)| < 1 = |z|,
on the circle |z| = 1. Hence by the Rouche’s theorem, the functions f (z) − z and z have
the same number of zeros in the disk |z| < 1. Thus f (z) = z at a unique z in the disk
|z| < 1.
Example 14.13. Prove that all the zeros of the polynomial p(z) = z 7 − 5z 3 + 12 lie in
the annulus 1 < |z| < 2.
Let g(z) = 12. On |z| = 1,
|p(z) − g(z)| = |z 7 − 5z 3 | ≤ |z|7 + 5|z|3 = 6 < 12 = |g(z)|.
Hence p and g have the same number of zeros in the disk |z| < 1. Also p has no zeros on
the circle |z| = 1. Thus p has no zero in the disk |z| < 1 since g has no zero in the disk
|z| < 1.
Let g(z) = z 7 . On |z| = 2,
|p(z) − g(z)| = | − 5z 3 + 12| ≤ 5|z|3 + 12 = 52 < 27 = |g(z)|.
Hence p and g have the same number of zeros in the disk |z| < 2. Thus p has 7 zeros in
the disk |z| < 2 since g has 7 zeros in the disk |z| < 2.
Observe that no zero of p lies on the circles |z| = 1 and |z| = 2 otherwise the above two
strict inequalities will be violated.
Since p is a polynomial of degree 7, it has 7 zeros. Thus all its zeros lie in the annulus
1 < |z| < 2.
Example 14.14. Prove that z 5 + 2 + ez has exactly three zeros in the left half-plane.
Let R > 3. Consider the closed curve γ which is composed of the line segment γ1 from
−iR to iR and the semi-circle γ2 of radius R from iR to −iR, as shown in the figure.
Let f (z) = z 5 + 2 + ez and g(z) = z 5 + 2.
66 NIKITA AGARWAL

On γ1 , we have z = iy, we have

p
|f (z) − g(z)| = |ez | ≤ 1 < 4 ≤ |g(z)| = y 10 + 4.
Also on γ2 , |z| = R. Hence
|g(z)| = |z 5 + 2| ≥ R5 − 2.
Also
|f (z) − g(z)| = |ez | ≤ 1.
Thus on γ,
|f (z) − g(z)| < |f (z)|.
Therefore f and g have the same number of zeros in the interior of γ. Since f has three
zeros in the interior of γ (for each R > 2) given by 21/5 ei(π+2kπ)/5 for k = 1, 2, 3, g also
has three zeros in the interior of γ. Since R > 2 is arbitrary, we have the required result.
Theorem 14.15. Let G be an open subset of C. If f is analytic on G except for poles,
then the poles cannot have a limit point in G.
Proof. Suppose z0 ∈ G is a limit point of poles. Then z0 is not an isolated singularity
since each of its neighbourhood contains a pole. Hence f is analytic at z0 (by hypothesis).
Thus f is well-defined in a neighbourhood of a, which is a contradiction. 
1
Example 14.16. Give the Laurent series expansion of f (z) = in the
z(z − 1)(z − 2)
annuli 0 < |z| < 1, 1 < |z| < 2, and 2 < |z| < ∞.
Since P
the annuli are centered at 0, we are looking for the Laurent series expansion of the
form ∞ n=−∞ an z .
n

Note that  
1 1 1
f (z) = + .
z z−2 1−z
Note that 
−1/2 1 P∞ n
P∞ z n
=− (z/2) = − n=0 n+1 , |z| < 2,


1 
1 − z/2 2 n=0
2
= n
z−2  1/z 1 P ∞ n
P −1 z
= (2/z) = n=−∞ n+1 , |z| > 2.

1 − 2/z z n=0 2

 LECTURE NOTES ON COMPLEX ANALYSIS I 67
P∞
1  n=0 z n , |z| < 1,
= −1/z 1 P∞ n
P−1 n
1−z  =− n=0 1/z = − n=−∞ z , |z| > 1.
1 − 1/z z
Hence we have the following Laurent series expansion in the given annuli.
• Annulus 0 < |z| < 1: The Laurent series expansion of f is
∞ ∞
! ∞ 
zn

1 X X
n
X −1
− n+1
+ z = n+2
+ 1 zn.
z n=0
2 n=0 n=−1
2
• Annulus 1 < |z| < 2: The Laurent series expansion of f is
∞ −1
! ∞ −2
1 X zn X
n
X zn X
− n+1
− z = − n+2
− zn.
z n=0
2 n=−∞ n=−1
2 n=−∞

• Annulus 2 < |z| < ∞: The Laurent series expansion of f is

−1 −1
! −2 
n

1 X z X
n
X 1
n+1
− z = n+2
− 1 zn.
z n=−∞ 2 n=−∞ n=−∞
2
68 NIKITA AGARWAL

15. Week 15
Some solved problems.
Example 15.1. Let p, q be polynomials of degree m, n, respectively. Suppose n > m + 1.
Let C be a circle whose interior contains all the roots of q. Prove that
Z
p(z)
dz = 0.
C q(z)
Solution: Since n > m + 1, the limit
p(z)
lim z 2
|z|→∞ q(z)
exists and is finite. Thus, there exists r > 0 and M > 0 such that for |z| ≥ r,
p(z) p(z) M
z2
≤M ⇒ ≤ 2.
q(z) q(z) |z|
Let R > 0 be any number such that R > r and the circle CR of radius R centered at
0 with the same orientation as C contains C in its interior. Since the region enclosed
between C and Cr does not contain any root of q, by the extended Cauchy’s theorem,
Z Z
p(z) p(z)
dz = dz.
C q(z) CR q(z)
such that Consider
Z Z
p(z) p(z) M 2πM
dz ≤ dz ≤ L(CR ) 2 = → 0,
CR q(z) CR q(z) R R
as R → ∞. Hence we have the desired result.
Example 15.2. (Partial fraction decomposition) Let f be a nonconstant rational func-
tion and let z1 , . . . , zp be its poles in C ∪ {∞}. Prove that f can be expressed as
f = f1 + · · · + fp , where fj is a rational function whose only pole is zj , j = 1, . . . , p.

Solution: Suppose f = P/Q, where P, Q are polynomials of degrees m, n, respectively.

Suppose m < n. Then lim|z|→∞ |f (z)| = 0. Hence all the poles of f are in C. We prove
that
f = f1 + · · · + fp ,
where fj is the principal part of the Laurent series decomposition of f at zj .
Let F = f − f1 − · · · − fp . We prove that F is the zero function.
Firstly, in a neighbourhood of zj , f − fj is analytic4 and also k6=j fk is analytic, thus
P
X
F = (f − fj ) − fk ,
k6=j

is analytic. Hence F is an entire function.

Since fj is a finite sum (number of terms at most order of the pole zj ) in which each term
is has a negative power of z − zj , we have lim|z|→∞ |fj (z)| = 0. Also, lim|z|→∞ |f (z)| = 0.
Thus lim|z|→∞ |F (z)| = 0. Therefore there exists r > 0 such that
|F (z)| ≤ 1, |z| ≥ r.
Also, |F (z)| ≤ K < ∞, for |z| ≤ r, since F is analytic.
Hence F is a bounded entire function. By Liouville’s theorem, it is an entire function.
Since lim|z|→∞ |F (z)| = 0, F is the zero function.
4If zj is a removable singularity of f − fj , define appropriate value of f − fj at zj to make it analytic.
 LECTURE NOTES ON COMPLEX ANALYSIS I 69

Observe that f has a pole at ∞ (that is, lim|z|→∞ |f (z)| = ∞) if and only if n < m.
If n < m, let z1 = ∞ and other poles are in C. By the division algorithm, we can express
f as
P = QR + S,
where R, S are polynomials with the degree of S strictly less than the degree of Q and R
is not a constant polynomial. Hence
f = R + S/Q.
We can now apply the partial fraction decomposition to the rational function S/Q (degree
S strictly less than the degree Q). Here R has a pole at ∞. By the above arguments,
S/Q = f2 + · · · + fp and let f1 = R.
If m = n, let P (z) = am z m + . . . and Q(z) = bm z m + . . . , where am bm 6= 0. Express
P = am (Q/bm ) + am (P/am − Q/bm ). Then
f = am /bm + am S/Q,
where S = (P/am − Q/bm ) has degree less than equal to m − 1. Now apply the earlier
arguments to S/Q to write it as f1 + · · · + fp . Club the term am /bm with f1 to get
f = (am /bm + am f1 ) + am f2 + · · · + am fp .
Note that am /bm + am f1 has only z1 as a pole.
Example 15.3. (An application of partial fraction decomposition) Compute the partial
fraction decomposition of
z4 + 1
f (z) = 3 .
z −1
By the division algorithm, z 4 + 1 = (z 3 − 1)z + (z + 1). Hence
z+1
f (z) = z + 3 .
z −1
z+1
We find the partial fraction decomposition of g(z) = 3 now. The poles are the third
z −1
roots of unity 1, ω, ω 2 , which are all simple. Note ω = ei2π/3 . The principal part of the
LS expansion of g at 1 is
z+1
limz→1 2
Res(f ; 1) limz→1 (z − 1)g(z) (z − z + 1) 2
= = = .
z−1 z−1 z−1 z−1
The principal part of the LS expansion of g at ω is
z+1 ω+1
limz→ω
Res(f ; ω) limz→ω (z − ω)g(z) 2
(z − 1)(z − ω ) (ω − 1)(ω − ω 2 )
= = = .
z−ω z−ω z−ω z−ω
The principal part of the LS expansion of g at ω 2 is
z+1 ω2 + 1
limz→ω2
Res(f ; ω 2 ) limz→ω2 (z − ω 2 )g(z) (z − 1)(z − ω) (ω 2 − 1)(ω 2 − ω)
= = = .
z − ω2 z − ω2 z − ω2 z − ω2
Hence
ω+1 ω2 + 1
2 (ω − 1)(ω − ω 2 ) (ω 2 − 1)(ω 2 − ω)
f (z) = z + + + .
z−1 z−ω z − ω2
70 NIKITA AGARWAL

Example 15.4. Let a, b ∈ C be such that 0 < |a| < |b|. Compute the Laurent series
1
expansion of the function in the annulus |a| < |z| < |b|.
(z − a)(z − b)
Observe that  
1 1 1 1
= − .
(z − a)(z − b) (a − b) z − a z − b
Now
∞ −1
1 1 1X X
= = n
(a/z) = a−(n+1) z n ,
z−a z(1 − a/z) z n=0 n=−∞

and
∞ ∞
1 −1 −1 X X
= = n
(z/b) = − b−(n+1) z n .
z−b b(1 − zb) b n=0 n=0
Hence the Laurent series expansion of f in the annulus |a| < |z| < |b| is
−1 ∞
!
1 X X
a−(n+1) z n + b−(n+1) z n .
(a − b) n=−∞ n=0

Example 15.5. (Another improper integral using the residue theorem) For a > 1, we
compute Z ∞
1
dx.
0 1 + xa
Let G = {reiθ : r > 0, |θ − π/a| < π}. The positive real axis lies in G. Consider
a branch g(z) of the function z a on G which takes the value 1 at z = 1. Consider the
function
1
f (z) = , z ∈ G.
1 + g(z)
Then f has a simple pole at eiπ/a .
Let r > 0. Let Ar be the circular arc {reiθ : 0 < θ < 2π/a}, oriented counterclockwise.

For 0 <  < 1 < R, let γ be the closed curve consisting of the interval [, R], followed by
the arc AR , followed by the line segment [Rei2π/a , ei2π/a ], followed by the arc −A .
By the residue theorem,
Z
f = 2πiRes(f ; eiπ/a ) = 2πi −eiπ/a /a


(1)
γ
 LECTURE NOTES ON COMPLEX ANALYSIS I 71

RR 1 i2π/a i2π/a 2πi/a R

R 1
Along [, R], we get 
dx, and along [Re , e ], we get e 
dx.
1 + xa 1 + xa
Along AR , Z
2πR
| f| ≤ a → 0,
AR R −1
as R → ∞.
Along A , Z
2π
| f| ≤ → 0,
A a−1
as  → 0.
Taking R → ∞ and  → 0 in (1), we get
Z ∞

1 2πi −eiπ/a /a π/a
dx = = .
0 1 + xa 1 − e2πi/a sin π/a
72 NIKITA AGARWAL

Definition 15.6. (Biholomorphism and Automorphism) Let U, V be open subsets of C.

A function f : U → V is known as a biholomorphism if f is a bijection and both f
and its inverse f −1 are holomorphic.
An automorphism of U is a biholomorphic map from U to U .
The collection of all automorphisms of U is denoted as Aut(U ) which forms a group under
composition of maps. It is nonempty since the identity map is an automorphism.
We study automorphisms on the unit disk D and the upper half plane H.
15.1. Automorphisms of the disk. Let us recall Schwarz’s lemma proved earlier.
Theorem 15.7. (Schwarz’s lemma) Let D be the open unit disk |z| < 1. Let f : D → D
be a holomorphic function such that f (0) = 0. Then
• |f (z)| ≤ |z|, for all z ∈ D,
• If for some nonzero z0 ∈ D, |f (z0 )| = |z0 |, then f (z) = eiθ z, for all z ∈ D, for
some fixed θ ∈ [0, 2π).
We have also proved the following:
• For a holomorphic function f : D → D,
1
|f 0 (z)| ≤ ,
1 − |z|
for all z ∈ D.
• Let u ∈ D. The Mobius transformation φu given by
z−u
φu (z) = ,
1 − zu
is an automorphism of D. Below is the image of some curves under the map φa
with a = (1 + i)/2.

• (Pick’s lemma) Let f : D → D be a holomorphic function. Then for any two

points z, w ∈ D,
f (z) − f (w) z−w
≤ .
1 − f (z)f (w) 1 − zw
Theorem 15.8. Let f be an automorphism of D. Then f is a Mobius transformation of
the form
f (z) = eiθ φu (z),
for u = f −1 (0) ∈ D and θ ∈ [0, 2π] which is unique modulo 2π.
 LECTURE NOTES ON COMPLEX ANALYSIS I 73

Proof. The map f ◦ φ−1

u is an automorphism of the disk. Moreover

(f ◦ φ−1
u )(0) = f (a) = 0.
Moreover
(f ◦ φ−1 −1
u ) (0) = (φu ◦ f
−1
)(0) = φu (u) = 0.
−1
Applying Schwarz’s lemma to both f ◦ φu and its inverse, we get
|f ◦ φ−1
u (z)| ≤ |z|, |φu ◦ f
−1
(z)| ≤ |z|, z ∈ D.
Taking z = f ◦ φ−1
u (w) in the second inequality, we get

|w| ≤ |f ◦ φ−1
u (w)|.
The above is true for all w ∈ D due to bijectivity of f ◦ φ−1
u .
Hence we get
|f ◦ φ−1
u (z)| = |z|, z ∈ D.
−1 iθ
By Schwarz’s lemma, f ◦ φu (z) = e z, for some θ ∈ [0, 2π).
Hence
f (z) = eiθ φu (z), z ∈ D.

Remark 15.9. Note u = f −1 (0). Hence u is uniquely determined. Further since
1 − |u|2
f 0 (z) = eiθ ,
(1 − zu)2
we have
f 0 (0)
eiθ = .
1 − |u|2
Thus θ is uniquely determined modulo 2π.
Thus the collection of automorphisms of D is in one-to-one correspondence with the
parameter space (u, θ) ∈ D × [0, 2π).
Corollary 15.9.1. An automorphism of D can be expressed as a Mobius transformation
of the form
az + b
,
bz + a
where |a|2 − |b|2 = 1.
Proof. Observe that
iθ/2
p p
e z/ 1 − |u| 2 − eiθ/2 u/ 1 − |u|2
eiθ φu (z) = p p .
−e−iθ/2 uz/ 1 − |u|2 + e−iθ/2 / 1 − |u|2
p p p
Take ap= eiθ/2 / 1 − |u|2 , b = −eiθ/2 u/ 1 − |u|2 , c = −e−iθ/2 u/ 1 − |u|2 , and d =
e−iθ/2 / 1 − |u|2 .
Clearly c = b and d = a. Also
|a|2 − |b|2 = 1.

74 NIKITA AGARWAL

16. Week 16 – Last week of classes

16.1. Automorphisms of the upper half plane. Let H be the upper half plane con-
sisting of all z ∈ C with Im(z) > 0. Consider the following Mobius transformations:
i−z 1−w
F (z) = , G(w) = i , z, w ∈ C.
i+z 1+w
Theorem 16.1. The map F : H → D is a biholomorphism with inverse G : D → H.
Proof. For z ∈ H, |z − i| < |z + i|, hence F (z) ∈ D.
Let w ∈ D. Let us compute Im(G(w)). Let w = u + iv. Observe that
1−w (1 − w)(1 + w) 1 − |w|2 − 2iv
G(w) = i =i = i .
1+w |1 + w|2 |1 + w|2
Hence
1 − |w|2
Im(G(w)) = > 0.
|1 + w|2
Thus G(w) ∈ H.
It is also easy to check that F (G(w)) = w and G(F (z)) = z. Thus we have the required
result. 
Remark 16.2. An interesting observation about the map F follows.
It is easy to check that F maps the extended real line onto the unit circle. Setting
z = x ∈ R, we get
i−x 1 − x2 2x
F (x) = = 2
+i .
i+x 1+x 1 + x2
Taking x = tan θ, we get
F (x) = cos 2θ + i sin 2θ = ei2θ .
As x goes from −∞ to ∞, θ goes from −π/2 to π/2. Thus F (x) goes from the open arc
on the unit circle starting at e−iπ ending at e−iuπ . The point at ∞ goes to −1.
Theorem 16.3. Let f : H → H be an automorphism of H. Then f is a Mobius
transformation which can be expressed as
az + b
f (z) = ,
cz + d
where a, b, c, d ∈ R satisfy ad − bc = 1.
Proof. The map F ◦ f ◦ G : D → D is an automorphism of D. Hence
F ◦ f ◦ G = eiθ φu ,
for some u ∈ D and θ ∈ [0, 2π). Thus f is the Mobius transformation
f = G ◦ (eiθ φu ) ◦ F.
Let f be of the form:
az + b
f (z) = ,
cz + d
where ad − bc 6= 0.
Since f maps H onto H, we deduce that f maps the boundary of H, the clircle R∞ onto
itself. We prove that a, b, c, d can be chosen to be in R.
Let us look at the following cases:
 LECTURE NOTES ON COMPLEX ANALYSIS I 75

• d = 0: Then bc 6= 0.
Since limz→∞ f (z) = a/c, we get a/c ∈ R.
Also since f (1) = (a + b)/c ∈ R and a/c ∈ R, we get b/c ∈ R.
Let a = λc and b = µc, for some λ, µ ∈ R, then
λcz + µc λz + µ
f (z) = = .
cz z
Note that all the coefficients are in R.
• c = 0: Then ad 6= 0.
Firstly f (0) = b/d ∈ R.
Also since f (1) = (a + b)/d ∈ R and b/d ∈ R, we get a/d ∈ R.
Let a = λd and b = µd, for some λ, µ ∈ R, then
λdz + µd
f (z) = = λz + µ.
d
Note that all the coefficients are in R.
• c 6= 0, d 6= 0:
Firstly f (0) = b/d ∈ R.
Also f (∞) = a/c ∈ R.
Since f −1 maps R∞ onto R∞ , we get f −1 (∞) = −d/c ∈ R.
Let a = δc, b = λd and d = µc, for some δ, λ, µ ∈ R. Then b = λµc. Thus
δcz + λµc δz + λµ
f (z) = = .
cz + µc z+µ
Note that all the coefficients are in R.
Next we prove that since f maps H to H, a, b, c, d ∈ R can be chosen so that ad−bc = 1.
For this, observe that
(az + b)(cz + d)
Im(f (z)) = Im
|cz + d|2
ac|z|2 + bcz + adz + bd
= Im
|cz + d|2
(ad − bc)Im(z)
=
|cz + d|2
> 0.
Hence ad − bc > 0. Now
a b
√ z+√
ad − bc ad − bc
f (z) = .
c d
√ z+√
ad − bc ad − bc
The coefficients satisfy the conditions in the statement of the theorem. 
Remark 16.4. We have just obtained that the group of matrices SL(2, Z) given by
  
a b
SL(2, Z) = : a, b, c, d ∈ R, ad − bc = 1 ,
c d
acts on the upper half plane H via Mobius transformations.
This action is in fact transitive. That is, for z, w ∈ H, there exists an automorphism
76 NIKITA AGARWAL

f : H → H such that f (z) = w.

Consider the map f = G ◦ φ−1
F (w) ◦ φF (z) ◦ F . Then

f (z) = G ◦ φ−1 −1
F (w) ◦ φF (z) (F (z)) = G ◦ φF (w) (0) = G(F (w)) = w.

16.2. Conformality. Recall conformality from Lecture 4. We proved the following.

• If f is holomorphic and f 0 (z0 ) 6= 0, then f is conformal at z0 .
• A Mobius transformation T is conformal at all points z where T z ∈ C.
16.2.1. Linear maps. Consider a linear transformation on R2 .
16.2.2. Conformality implies holomorphic. We stated the following result but did not
prove it. Let us prove it now.
Theorem 16.5. Let f be a complex-valued function on an open set G in C, whose real
and imaginary parts have continuous first order partial derivatives. If f is conformal at
each point of G, then f is holomorphic on G and f 0 is never 0 on G.
For the proof, we need some notations. For a complex variable z = z + iy and a complex-
valued function f = ui v (u, v are real, imaginary parts of f , respectively), since
z+z z−z
x= , y= ,
2 2i
we get
1 1 1
fz = fx xz + fy yz = (fx − ify ) = (ux + ivx − i(uy + ivy )) = (ux + vy + i(vx − uy )) ,
2 2 2
and
1 1 1
fz = fx xz + fy yz = (fx + ify ) = (ux + ivx + i(uy + ivy )) = (ux − vy + i(vx + uy )) .
2 2 2
The Cauchy-Riemann equations are thus equivalent to fz = 0. In which case,
1
fz = (ux + vy + i(vx − uy )) = fx = f 0 .
2
Proof. Let f = u + iv, where u and v are the real and imaginary parts of f . Let γ be a
differentiable curve in G. Let z0 ∈ γ and z0 = γ(t0 ). Then by the Chain rule,
(f ◦ γ)0 (t0 ) = fz (z0 )γ 0 (t0 ) + fz (z0 )γ 0 (t0 ).
Set
a = fz (z0 ), b = fz (z0 ).
Let z0 ∈ G. Let θ ∈ [0, 2π].
Consider two curves γ1 , which is a line segment through the point z0 in the horizontal
direction, and γ2 , which is a line segment through the point z0 in the direction eiθ .
Parameterize γ1 as t 7→ z0 + t and γ2 as t 7→ z0 + teiθ with γ1 (0) = γ2 (0) = z0 , if f
preserves angles, that is,
arg(aγ20 (t0 ) + bγ20 (0)) − arg(aγ10 (t0 ) + bγ10 (0)) = arg(γ20 (t0 )) − arg(γ10 (t0 )),
implies
arg(aeiθ + be−iθ ) − arg(a + b) = θ,
which further implies
arg(a + be−2iθ ) + θ − arg(a + b) = θ,

which implies
arg(a + be−2iθ ) = arg(a + b).
 LECTURE NOTES ON COMPLEX ANALYSIS I 77

As θ goes from 0 to 2π, the point a + be−2iθ moves along the circle centered at a with
radius |b| in the clockwise direction, whereas the point a + b is a fixed point on this circle.
Thus, their arguments will match if and only if b = 0. This implies
fz (z0 ) = 0.
Since z0 ∈ G is arbitrary, we have fz = 0. Hence f is holomorphic on G.
Also f 0 (z) = fz (z0 ) 6= 0 for all z0 ∈ G since otherwise (f ◦ γ)0 (t0 ) = 0 for all curves γ,
hence angles will not be preserved. 
Theorem 16.6. Let U, V be open subsets of C. Let f : U → V be a bijection. If f is
holomorphic on U then f 0 (z) 6= 0, for all z ∈ U .
Moreover, the inverse g : V → U of f is holomorphic.
Proof. Suppose f 0 (z0 ) = 0 for some z0 ∈ U . Let R > 0 be such that f 0 (z) 6= 0 for all
z ∈ B(z0 , R) and
f (z) = f (z0 ) + a(z − z0 )k + G(z),
f (k) (z0 )
where a = , k ≥ 2, and G has a zero at z0 of order at least k + 1 which is greater
k!
than or equal to 3.
Let w be a complex number (to be chosen shortly). Then
f (z) − f (z0 ) − w = F (z) + G(z),
where F (z) = a(z − z0 )k − w.
Also since G has a zero at z0 of order at least k + 1 and F (z0 ) = −w, we get
lim |G(z)/F (z)| = 0.
z→z0

Thus, there exists 0 < r < R such that |G(z)/F (z)| < 1 for all z with |z − z0 | ≤ r.
Choose w such that 0 < |w/a|1/k < r. Then F has k zeros in the disk B(z0 , r).
On the circle |z−z0 | = r, |G(z)| < |F (z)|. Hence, by Rouche’s theorem, both F and F +G
have the same number of zeros in the disk B(z0 , r). Thus (F + G)(z) = f (z) − f (z0 ) − w
has k ≥ 2 zeros in the disk B(z0 , r).
Since f 0 (z) 6= 0 for all z ∈ B(z0 , r), all the zeros of F + G in this disk are distinct.
Let z1 , z2 be two distinct zeros such that
(F + G)(z1 ) = (F + G)(z2 ) ⇒ f (z1 ) = f (z2 ) = f (z0 ) + w.
This contradicts that f is one-one.

Now we prove that g is holomorphic on V = f (U ). Let w0 = f (z0 ) ∈ V and w = f (z) is

near w0 . Note that z, z0 are unique preimages under f of w, w0 since f is injective.
If w 6= w0 ,
g(w) − g(w0 ) z − z0 1
= = .
w − w0 f (z) − f (z0 ) f (z) − f (z0 )
z − z0
0
Since f (z0 ) 6= 0, we can let z → z0 to get
g 0 (w0 ) = 1/f 0 (z0 ) = 1/f 0 (g(w0 )).

Corollary 16.6.1. Let U, V be open subsets of C. Let f : U → V be a bijection. Then
f is conformal on U if and only if f is a biholomorphism.
78 NIKITA AGARWAL

Proof. (⇒) Let f be conformal on U . By Theorem 16.5, f is holomorphic on U . Hence by

Theorem 16.6, the inverse g : V → U of f is holomorphic. Hence f is a biholomorphism.

(⇐) Let f be a biholomorphism. By Theorem 16.6, f 0 (z) 6= 0, for all z ∈ U . Hence

by Theorem 4.12, f is conformal on U . 
 LECTURE NOTES ON COMPLEX ANALYSIS I 79

17. Last lecture

Example 17.1. Prove that for 1 < a < ∞, the function z + a − ez has exactly one zero
in the left half-plane.

Solution: Let R > a + 1. Consider the closed curve γ which is composed of the line
segment γ1 from −iR to iR and the semi-circle γ2 of radius R from iR to −iR.
Let g(z) = z + a − ez and f (z) = z + a.
On γ1 , z = iy, we have
p
|f (z) − g(z)| = |ez | ≤ 1 < a ≤ |f (z)| = y 2 + a.
Also on γ2 , |z| = R. Hence
|f (z)| = |z + a| ≥ R − a.
Also
|f (z) − g(z)| = |ez | ≤ 1.
Thus on γ,
|f (z) − g(z)| < |f (z)|.
Therefore f and g have the same number of zeros in the interior of γ. Since f has one
zero in the interior of γ (for each R > a + 1) given by z = −a, g also has one zero in the
interior of γ. Since R > 2 is arbitrary, we have the required result.
What happens as a approaches 1?
Let z0 be a zero of g, then z0 +a−ez0 = 0. Taking complex conjugate, we get z0 +a−ez0 =
0. Hence z0 is also a zero of g. Thus the zeros of g occur in conjugate pairs.
Thus, for 1 < a < ∞, since there is only one zero of g in the left-half plane, it must be
real.
Let us consider the function g on the real line
g(x) = x + a − ex .
Since g 0 (x) = 1 − ex , the point 0 is a global maxima for g. Further since g(0) = a − 1 > 0,
limx→−∞ g(x) = −∞, there is a unique zero xa of g in the negative real axis. Clearly
since the function g increases as a increases, xa0 < xa when a < a0 .
As a → 1+ , xa → 0+ .
Example 17.2. Let f be an entire function. Suppose f does not assume any value in
the open disk B(z0 , R). Prove that f is a constant function.
Consequently, for a nonconstant entire function f , f (C) is dense in C.
Also, by the Open mapping theorem (Theorem 13.1), f (C) is open in C.
Moreover, since C is connected and f is continuous, f (C) is connected.
In fact a major theorem of complex analysis (Picard’s theorem) states that f (C) is either
all of C or C except a single point.

Solution: This is same as problem 1 of the mid-semester exam.

Example 17.3. Let f be holomorphic and satisfies |f (z)| ≤ M for z ∈ B(z0 , R). Prove
that if f has a zero at z0 , then
M
|f (z)| ≤
|z − z0 |, |z − z0 | < R.
R
Show that the equality holds at some point z 6= z0 only when f (z) is a constant multiple
of z − z0 .
80 NIKITA AGARWAL

Solution: Define h on B(0, 1) as

h(z) = f (z0 + Rz)/M.
Then |h(z)| ≤ 1, for all |z| < 1, and h(0) = f (z0 )/M = 0. Hence by Schwarz lemma,
|h(z)| ≤ |z|, |z| < 1.
That is
|f (z0 + Rz)/M | ≤ |z|, |z| < 1.
Let w = z0 + Rz, then for all w ∈ B(z0 , R),
|f (w)| ≤ M |w − z0 |/R.
Equality holds for some z 6= z0 implies
h(z) = eiθ z,
that is
f (z0 + Rz)/M = eiθ z.
Take w = z0 + Rz, we get
eiθ M
f (w) = (w − z0 ).
R
Thus if z0 is a zero of f of multiplicity at least 2, then for all w ∈ B(z0 , R) \ {z0 },
|f (w)| < M |w − z0 |/R.
Example 17.4. (Revisit) Prove that
Z ∞ √ p √
−t2 2 π 1+ 2
e cos t dt = .
0 4
2
Solution: Let us integrate the function f (z) = e−z over the boundary of the region
{z : 0 ≤ |z| ≤ R, 0 ≤ Arg(z) ≤ π/8}. The boundary γ (in counterclockwise direction)
can be written as the union of three curves (plot them on the complex plane):
γ1 : [0, R] → C given by γ1 (t) = t,
γ2 : [0, π/8] → C given by γ1 (t) = Reit ,
and the reverse of
γ3 : [0, R] → C given by γ3 (t) = teiπ/8 .
Thus, since γ is a closed curve (here we are using f = g 0 for some differentiable/holomorphic
function g), Z Z Z Z
0= f = f+ f− f − (1)
γ γ1 γ2 γ3
Note the negative sign in the third term.
Now consider the three integrals one by one.
Z Z R Z ∞ √
−t2 2
f = e dt → e−t dt = π/2,
γ1 0 0

as R → ∞ (we are using the last identity above from your earlier courses). Now
Z
f ≤ MR L(γ2 ) → 0,
γ2
 LECTURE NOTES ON COMPLEX ANALYSIS I 81

2
√
as R → ∞, where MR = sup{|f (z)| : z ∈ γ2 } = e−R / 2 and L(γ2 ) < ∞.
Finally consider the last term:
Z Z R Z R √
Z R √ √
−(teiπ/8 )2 iπ/8 −t2 (1+i)/ 2 iπ/8 2 2
f= e e dt = e e dt = e−t / 2 ei(t / 2+π/8) dt
γ3 0 0 0
Z R
2/
√ √ Z R
2/
√ √
= e−t 2
cos(t2 / 2 + π/8) dt + i e−t 2
sin(t2 / 2 + π/8) dt
0 0
Z R/21/4 Z R/21/4
−s2 2
=2 1/4
e 2
cos(s + π/8) ds + i 2 1/4
e−s sin(s2 + π/8) ds,
0 0
1/4
by taking s = t/2 .
Now substituting everything in equation (1), taking R → ∞ and identifying real and
imaginary parts, we get
Z ∞
√ 2
π/2 = 2 1/4
e−s cos(s2 + π/8) ds,
0
Z ∞
2
0 = 21/4 e−s sin(s2 + π/8) ds.
0
Thus Z ∞
2 π
e−s cos s2 cos π/8 − sin s2 sin π/8 ds =


,
0Z 2 21/4
∞
2
e−s cos s2 sin π/8 + sin s2 cos π/8 ds = 0.


0
From the second equation, substitute
Z ∞ Z ∞
−s2 2
2
e sin s ds = − tan π/8 e−s cos s2 ds,
0 0
in the first equation, we get
Z ∞ Z ∞
−s2 2 π
cos π/8 2
e cos s ds + sin π/8 tan π/8 e−s cos s2 ds = ,
0 0 2 21/4
which implies
Z ∞
2 π
(cos π/8 + sin π/8 tan π/8) e−s cos s2 ds = .
0 2 21/4
Since
cos π/8 + sin π/8 tan π/8 = 1/ cos π/8,
we get
Z ∞
p √ √ p √
−s2 π
2 π 2+ 2 π 1+ 2
e cos s ds = cos π/8 = = .
0 2 21/4 2 21/4 2 4