1 COMPLEX ANALYSIS UG JAN-APR 2021.

NOTES

2 MANOJ KUMMINI

3 Contents
4 Outline 2
5 Lecture 1. Preliminaries 2
6 Lecture 2. Differentiability 4
7 Lecture 3. Power series 6
8 Lecture 4. Analytic functions 9
9 Lecture 5. Exponential and logarithmic functions 12
10 Lecture 6. Path integrals, I 14
11 Lecture 7. Path integrals, II 15
12 Exercises 16
13 Lecture 8. Absolute value of a path integral 16
14 Exercises 18
15 Lecture 9. Primitives 18
16 Exercises 20
17 Lecture 10. Cauchy integral theorem, I 20
18 Exercises 21
19 Lecture 11. Cauchy integral theorem, II 21
20 Exercises 22
21 Lecture 12. Cauchy integral theorem, III 23
22 Lecture 13. Index of a point 25
23 Exercises 26
24 Lecture 14. Cauchy integral formula 26
25 Exercises 27
26 Lecture 15. Holomorphic functions are analytic. 28
27 Exercises 30
28 Lecture 16. Morera’s theorem, Liouville’s theorem 30
29 Exercises 31
30 Lecture 17. Isolated singularities 31
31 Exercises 33
32 Lecture 18. Local mapping 34
33 Exercises 35
34 Lecture 19. Maximum principle, definite integrals etc. 35
35 Exercises 36
36 Lecture 20. Conformality 36
37 Exercises 38
38 Lecture 21. Riemann sphere 38
39 Exercises 39
40 Lecture 22. Moebius transformations 39
41 Exercises 40
1

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 2 MANOJ KUMMINI

42 Lecture 23. Moebius transformations, continued. 41

43 Exercises 42
44 Lecture 24. Singularity at infinity 42
45 Lecture 25. Automorphisms of the complex plane 43
46 Exercises 45
47 Lecture 26. Automorphisms of the Riemann sphere 46
48 Exercises 47
49 Lecture 27. Review of path homotopy 47
50 Exercises 49
51 Lecture 28. General version of Cauchy integral theorem. 49
52 Exercises 50
53 Exercises 50
54 References 50

55 Outline
56 These are notes from an undergraduate course on complex analysis during Jan–Apr 2020 at
57 CMI.
58 (1) Ahlfors, Complex Analysis.
59 (2) Conway, Functions of one complex variable.
60 (3) Kodaira, Complex Analysis.
61 (4) Lang, Complex Analysis.
62 (5) Rodrı́guez, Kra and Gilman, Complex Analysis, in the spirit of Lipman Bers (2nd ed.).
63 Lecture 1. Preliminaries
64 C as the ring R[𝑋 ]/(𝑋 2 + 1). Write 𝚤 for the image of 𝑋 in C.
65 Let 𝑐 ∈ C; then there exist unique 𝑎, 𝑏 ∈ R such that 𝑐 = 𝑎 + 𝑏𝚤. We call 𝑎 the real part of 𝑐
66 and 𝑏 the imaginary part of 𝑐, and write 𝑎 = <(𝑐) and 𝑏 = =(𝑐). If 𝑓 : 𝐴 −→ C is a function
67 (𝐴 being some set), then we write <(𝑓 ) and =(𝑓 ), respectively, for the functions 𝐴 −→ R,
68 𝑎 ↦→ <(𝑓 (𝑎)) and 𝑎 ↦→ =(𝑓 (𝑎)).
The function | · | : C −→ R≥0 , 𝑧 ↦→ (<(𝑧)) 2 + (=(𝑧)) 2 is called the modulus or the absolute
√︁
69
70 value function. This gives a metric on C: take the distance between 𝑐, 𝑐 0 ∈ C to be |𝑐 − 𝑐 0 |.
71 The function C −→ R2 , 𝑐 ↦→ (<(𝑐), =(𝑐)) gives an isomorphism of real vector spaces and a
72 homeomorphism1 of metric spaces (with R2 given the usual metric). Therefore C is a complete
73 metric space.
74 A subset 𝐴 ⊆ C is connected if there are no open subsets 𝑈 and 𝑉 of C such that 𝐴 = (𝐴 ∩
75 𝑈 ) ∪ (𝐴 ∩ 𝑉 ) with (𝐴 ∩ 𝑈 ) ≠ ∅ ≠ (𝐴 ∩ 𝑉 ) and (𝐴 ∩ 𝑈 ∩ 𝑉 ) = ∅.
76 Let 𝐴 ⊆ C, and 𝑧 0, 𝑧 1 ∈ 𝐴. A path in 𝐴 from 𝑧 0 to 𝑧 1 is a continuous function 𝛾 : [0, 1] −→ 𝐴
77 such that 𝛾 (𝑖) = 𝑧𝑖 , 𝑖 = 0, 1. Say that 𝐴 is path-connected if for every 𝑧 0, 𝑧 1 ∈ 𝐴, there is a path
78 from 𝑧 0 to 𝑧 1 .
79 1.1. Proposition. An open subset of C is connected if and only if it is path-connected.
80 Proof is left as an exercise.
81 1.2. Definition. By a domain, we mean a connected open subset of C.
1Let 𝑋 and 𝑌 be topological spaces, and 𝑓 : 𝑋 −→ 𝑌 a function. We say that 𝑓 is a homeomorphism if it is
bijective and continuous, and its inverse function (which exists since 𝑓 is bijective) is continuous.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 3

82 When we talk of limits and convergence in C, these are with respect to the metric topology.
83 In particular, a sequence of Í complex numbers is convergent if and only if it is a Cauchy se-
2
84 quence. Consider a series 𝑖∈N Í𝑛 𝑖 𝑎 of complex numbers. The sequence of partial sums for this
85 series is the sequence 𝑠𝑛 = 𝑖=0 𝑎𝑖 , 𝑛 ∈ N. We sayÍthat the series converges if the sequence
86 𝑠 0, 𝑠 1, 𝑠 2, .Í. . converges. Now suppose that the series 𝑖∈N |𝑎𝑖 | of real numbers converges. (We
87 say that 𝑖∈N 𝑎𝑖 is absolutely Í𝑛 convergent if this happens.) Let 𝜖 > 0; then there exists 𝑁 such that
88 for every 𝑛 ≥ 𝑚 > 𝑁 , 𝑖=𝑚 |𝑎𝑖 | < 𝜖. Therefore |𝑠𝑛 − 𝑠𝑚 | < 𝜖, i.e., the sequence (𝑠𝑛 ) is Cauchy.
Hence 𝑖∈N 𝑎𝑖 is convergent. We have now shown that every absolutely convergent series is
Í
89
90 convergent.
91 1.3. Notation. Hereafter, when we write a complex number 𝑐 = 𝑎 + 𝑏𝚤, it should be understood
92 that 𝑎 = <(𝑐) and 𝑏 = =(𝑐). Similarly, when we write 𝑓 = 𝑢 + 𝑣𝚤 for a C-valued function 𝑓 ,
93 𝑢 = <(𝑓 ) and 𝑦 = =(𝑓 ). 
94 1.4. Notation. For 𝑅 ∈ R+ ∪{+∞} and𝑐 ∈ C, we denote by 𝐵𝑐,𝑅 the open disc {𝑧 ∈ C : |𝑧−𝑐 | < 𝑅}
95 and by 𝐵𝑐,𝑅 , its closure in C.
96 Exercises.
97 1.1 Show that every connected open subset of R𝑛 is path-connected. The “topologist’s sine
98 curve”, i.e., the closure of
1
  
𝑥, sin | 𝑥 ∈ (0, 1)
𝑥
99 inside R2 is connected but not path-connected. (It is not open in R2 ).
100 1.2 Show that for every positive integer 𝑛, R𝑛 with the usual metric is a complete metric
101 space.
102 1.3 (Polar coordinates). For a nonzero 𝑐 ∈ C, there exist unique 𝑟 ∈ R+ and non-unique
103 𝜃 ∈ R so that 𝑐 = 𝑟 (cos 𝜃 + 𝚤 sin 𝜃 ). (We still do not know what 𝜋 is, or that 𝑒𝚤𝜃 =
104 (cos 𝜃 + 𝚤 sin 𝜃 ).) We refer to 𝜃 as an argument of 𝑐.
105 1.4 We think of 𝑧 as the ‘coordinate’ for C; This is related to the cartesian coordinates (𝑥, 𝑦)
106 of R2 by 𝑥 = <(𝑧) and 𝑦 = =(𝑧). We can also define another coordinate 𝑧, ¯ with the
107 property that 𝑧 = 𝑎 + 𝑏𝚤 (𝑎, 𝑏 ∈ R) is the same as the point given by 𝑧¯ = 𝑎 − 𝑏𝚤. Let 𝑛 be a
108 positive integer; expressÍthe equation 𝑧𝑛 = 𝑧¯𝑛 in polar coordinates and solve.
109 1.5 Prove the ratio test: Let 𝑖∈N 𝑎𝑖 be a series of non-zero real numbers. If
𝑎𝑖+1
𝐿 := lim
𝑖→∞ 𝑎𝑖
110 exists, then the series converges if 𝐿 < 1 and diverges if 𝐿 > 1.
1.6 Prove the root test: Let 𝑖∈N 𝑎𝑖 be a series of real numbers. Let
Í
111
1
𝐿 := lim sup |𝑎𝑛 | 𝑛 .
𝑛→∞
112 Then the series converges if 𝐿 < 1 and diverges if 𝐿 > 1.
1.7 Consider the series 𝑎𝑛 with
Í
113

 1,𝑎 𝑛 = 0,




𝑎𝑛 = 2 , 𝑛 odd,
𝑛−1

 𝑎𝑛−1 , 𝑛 ≥ 2 even.


 8
2By N, we mean {0, 1, 2, . . .}.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 4 MANOJ KUMMINI

114 Show that the ratio test is inconclusive, while the root test concludes that the series con-
115 verges.

116 Lecture 2. Differentiability

117 2.1. Definition. Let 𝑐 ∈ C and 𝑓 a (complex-valued) function defined in an open disc around 𝑐.
118 Say that 𝑓 is (complex-)differentiable at 𝑐 if
𝑓 (𝑐 + ℎ) − 𝑓 (𝑐)
lim
ℎ→0 ℎ
119 exists. If this is the case, we call this limit the derivative of 𝑓 at 𝑐, and denote it by 𝑓 0 (𝑐).
120 2.2. Remark. We will not explicitly say “complex-differentiable”, hereafter, for C-valued func-
121 tions from subsets of C. When we refer to such a function as being “differentiable”, it should
122 be understood as “complex-differentiable”.
123 By a constant function we mean a function of the form C −→ C, 𝑧 ↦→ 𝑐 for some 𝑐 ∈ C. It
124 is immediate that constant functions are differentiable. The identity function on C (i.e., the
125 map 𝑧 ↦→ 𝑧) is differentiable. We could also consider the restrictions of these functions to
126 some open 𝑈 ⊆ C. Before we construct more examples, we need some to see some rules of
127 differentiation.
128 2.3. Remark (Rules of differentiation). Let 𝑐 ∈ C, 𝑓 and 𝑔 functions defined on a neighbour-
129 hood 3 of 𝑐 and differentiable at 𝑐, ℎ a function defined on a neighbourhood of 𝑓 (𝑐) and differ-
130 entiable at 𝑓 (𝑐), and 𝛼 ∈ C. Then
131 (1) (𝑓 + 𝛼𝑔) 0 (𝑐) = 𝑓 0 (𝑐) + 𝛼𝑔0 (𝑐).
132 (2) (𝑓 𝑔) 0 (𝑐) = 𝑓 0 (𝑐)𝑔(𝑐) + 𝑓 (𝑐)𝑔0 (𝑐).
133 (3) (ℎ 0 0 0
 ◦ 𝑓 ) (𝑐) = ℎ (𝑓 (𝑐))𝑓 (𝑐).
0 𝑓 0 (𝑐)
1
134 (4) 𝑓 (𝑐) = − 𝑓 (𝑐) 2 if 𝑓 (𝑐 1 ) ≠ 0 for every 𝑐 1 in a neighbourhood of 𝑐.

135 2.4. Example. We can now construct two more examples of differentiable functions. Let 𝑝 (𝑋 ), 𝑞(𝑋 ) ∈
136 C[𝑋 ] with 𝑞(𝑋 ) ≠ 0. The function
C −→ C, 𝑧 ↦→ 𝑝 (𝑧)
137 (i.e., the polynomial 𝑝 evaluated at 𝑧) is differentiable at all points in C. Such functions are
138 called polynomial functions. Let 𝑈 = {𝑧 ∈ C | 𝑞(𝑧) ≠ 0}. Since the set of zeros of 𝑞(𝑋 ) is finite,
139 𝑈 is open. The function
𝑝 (𝑧)
𝑈 −→ C, 𝑧 ↦→
𝑞(𝑧)
140 is differentiable at every point in 𝑈 . These are called rational functions.
141 2.5. Remark. Let 𝑐 ∈ C and 𝑓 a (complex-valued) function defined in an open disc around 𝑐. If
142 𝑓 is differentiable at 𝑐, then it is continuous at 𝑐. To see this, note that
𝑓 (𝑐 + ℎ) − 𝑓 (𝑐)
𝑓 0 (𝑐) · 0 = lim · lim ℎ = lim 𝑓 (𝑐 + ℎ) − 𝑓 (𝑐).
ℎ→0 ℎ ℎ→0 ℎ→0

3Let 𝑋 be a topological space and 𝑥 ∈ 𝑋 . A neighbourhood of 𝑥 in 𝑋 is a subset 𝑉 of 𝑋 such that there exists an
open subset 𝑈 of 𝑋 such that 𝑥 ∈ 𝑈 ⊆ 𝑉 .

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 5

143 2.6. Example. Let 𝑓 be a real-valued function defined in an open disc around 𝑐 ∈ C. Suppose
144 that 𝑓 is differentiable at 𝑐. Then, taking ℎ to be real, we see that
𝑓 (𝑐 + ℎ) − 𝑓 (𝑐)
𝑓 0 (𝑐) = lim
ℎ→0 ℎ
145 is real. On the other hand, taking ℎ = 𝚤𝑡 to be purely imaginary, we get
𝑓 (𝑐 + 𝚤𝑡) − 𝑓 (𝑐)
𝑓 0 (𝑐) = lim
𝑡→0 𝚤𝑡
146 is purely imaginary. Hence 𝑓 0 (𝑐) = 0. We will see this in a general context later. 
147 2.7. Definition. Let 𝑈 ⊆ C be a domain, and 𝑓 : 𝑈 −→ C. Say that 𝑓 is holomorphic on 𝑈 if it is
148 (complex-)differentiable at every point in 𝑈 . A function 𝑓 : C −→ C that is holomorphic on C
149 is called entire.
150 2.8. Theorem. Let 𝑈 be a domain, 𝑓 : 𝑈 −→ C and 𝑐 = 𝑎 + 𝑏𝚤 ∈ 𝑈 . Write 𝑓 as 𝑢 (𝑥, 𝑦) + 𝑣 (𝑥, 𝑦)𝚤.
151 Then 𝑓 is complex-differentiable at 𝑐, if and only if 𝑢 and 𝑣 are differentiable at (𝑎, 𝑏) (as functions from
152 R2 −→ R) and their partial derivatives satisfy the Cauchy-Riemann equations
(2.9) 𝑢𝑥 (𝑎, 𝑏) = 𝑣𝑦 (𝑎, 𝑏) and 𝑢𝑦 (𝑎, 𝑏) = −𝑣 𝑥 (𝑎, 𝑏).
153 Further, when this happens, 𝑓 0 (𝑐) = 𝑢𝑥 (𝑎, 𝑏) + 𝚤𝑣 𝑥 (𝑎, 𝑏) = 𝑣𝑦 (𝑎, 𝑏) − 𝚤𝑢𝑦 (𝑎, 𝑏).
154 (Here 𝑢𝑥 (𝑎, 𝑏) is the partial derivative 𝜕𝑥 (𝑎, 𝑏), etc.)
𝜕𝑢

155 Proof. Write ℎ = Δ𝑥 + 𝚤Δ𝑦 and 𝑓 (𝑐 + ℎ) − 𝑓 (𝑐) = Δ𝑢 + 𝚤Δ𝑣. Assume that 𝑓 is differentiable at 𝑐.
156 Write 𝑓 0 (𝑐) = 𝑝 + 𝚤𝑞; then
Δ𝑢 + 𝚤Δ𝑣 = (𝑝 + 𝚤𝑞)(Δ𝑥 + 𝚤Δ𝑦) + 𝑟 (Δ𝑥 + 𝚤Δ𝑦),
where 𝑟 (ℎ) is a complex-valued function defined in a neighbourhood of 0 ∈ C, but possibly
not at 0, such that lim 𝑟 (ℎ)
ℎ = 0. Write 𝑟 (𝑧) = 𝑟 1 (𝑧) + 𝚤𝑟 2 (𝑧). Thus
ℎ→0

Δ𝑢 = 𝑝Δ𝑥 − 𝑞Δ𝑦 + 𝑟 1 (ℎ);

Δ𝑣 = 𝑞Δ𝑥 + 𝑝Δ𝑦 + 𝑟 2 (ℎ).

157 Let 𝜖 > 0. Then there exists 𝛿 > 0 such that for each ℎ ∈ 𝐵 0,𝛿 r{0}, | 𝑟 (ℎ)
ℎ | < 𝜖; since 𝑟 1 = <(𝑟 )
𝑟 1 (ℎ) 𝑟 2 (ℎ) 𝑟 1 (ℎ) (ℎ)
158 and 𝑟 2 = =(𝑟 ), | ℎ | < 𝜖 and | ℎ | < 𝜖. Therefore lim |ℎ| = 0 and lim 𝑟 2|ℎ| = 0. Hence 𝑢
ℎ→0 ℎ→0
159 and 𝑣 are differentiable at (𝑎, 𝑏) and (2.9) are satisfied.
Conversely, assume that 𝑢 and 𝑣 are differentiable at (𝑎, 𝑏) and that (2.9) are satisfied. Write
𝑝 = 𝑢𝑥 (𝑎, 𝑏) and 𝑞 = 𝑣 𝑥 (𝑎, 𝑏). Then
Δ𝑢 = 𝑝Δ𝑥 − 𝑞Δ𝑦 + 𝑟 1 (ℎ);
Δ𝑣 = 𝑞Δ𝑥 + 𝑝Δ𝑦 + 𝑟 2 (ℎ),
𝑟 1 (ℎ) 𝑟 2 (ℎ)
160 where lim = 0 and lim = 0. Write 𝑟 (𝑧) = 𝑟 1 (𝑧) + 𝚤𝑟 2 (𝑧). Then 𝑓 (𝑐 + ℎ) − 𝑓 (𝑐) =
ℎ→0 |ℎ| ℎ→0 |ℎ|
𝑟 (ℎ)
161 (𝑝 + 𝚤𝑞) (Δ𝑥 + 𝚤Δ𝑦) + 𝑟 (𝑧). Note that lim = 0, by the triangle inequality. Hence 𝑓 0 (𝑐) exists
ℎ→0 ℎ
162 and equals 𝑝 + 𝚤𝑞. 
163 Satisfying the Cauchy-Riemann equations alone is not a sufficient condition, in general, for
164 𝑓 to be differentiable at a point; see the exercises.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 6 MANOJ KUMMINI

165 2.10. Remark. Write 𝑓𝑥 = 𝑢𝑥 + 𝑣 𝑥 𝚤 and 𝑓𝑦 = 𝑢𝑦 + 𝑣𝑦𝚤 (wherever the partial derivatives on the
166 right are defined). The Cauchy-Riemann equations can be rephrased in a more concise way, as
167 𝑓𝑥 = −𝚤 𝑓𝑦 . Another description is given in the exercises.
168 Exercises.
169 2.1 Prove the rules of differentiation mentioned in class.
170 2.2 Let (
𝑧 5 |𝑧| −4, if 𝑧 ≠ 0;
𝑓 (𝑧) =
0, otherwise.
171 Write 𝑧 = 𝑥 + 𝑦𝚤 and determine <(𝑓 ) and =(𝑓 ) as functions of the real variables 𝑥 and
172 𝑦. Show that these satisfy the Cauchy-Riemann equations at 𝑧 = 0 ∈ C. Show that the
173 limit
𝑓 (ℎ)
lim
ℎ→0 ℎ
174 does not exist by considering first ℎ = 𝑟 and then ℎ = (1 + 𝚤)𝑟 , with 𝑟 ∈ R. Hence 𝑓 is
175 not differentiable.
2.3 Define
1
𝑓𝑧 = (𝑓𝑥 − 𝚤𝑓𝑦 ), and
2
1
𝑓𝑧¯ = (𝑓𝑥 + 𝚤 𝑓𝑦 )
2
176 wherever the RHS is defined.
177 (1) Treating 𝑧 and 𝑧¯ as independent coordinates, show that this definition agrees with
178 the formula one would get from applying the chain rule for the substitution 𝑥 = 𝑧+¯
2 ,
𝑧

2𝚤 .
𝑦 = 𝑧−¯ 𝑧
179
180 (2) If 𝑓 is differentiable at 𝑐, then 𝑓 0 (𝑐) = 𝑓𝑧 (𝑐); the Cauchy-Riemann equations sim-
181 plify to give 𝑓𝑧¯ (𝑐) = 0.
182 2.4 If 𝑓 = 𝑧𝑚 𝑧¯𝑛 , with 𝑚, 𝑛 ≥ 0, then 𝑓𝑧 = 𝑚𝑧𝑚−1𝑧¯𝑛 and then 𝑓𝑧¯ = 𝑛𝑧𝑚 𝑧¯𝑛−1 . Extend this to
183 ‘polynomials’ in 𝑧 and 𝑧. ¯
184 2.5 Show that the function
( 2
𝑥𝑦 (𝑥+𝑦𝚤)
𝑓 (𝑥 + 𝑦𝚤) = 𝑥 2 +𝑦 4 , if(𝑥, 𝑦) ≠ (0, 0)
0, otherwise
185 is not differentiable at 0.
186 2.6 Let 𝑓 (𝑧) be a function defined in a neighbourhood of 𝑐 ∈ 𝐶. Show that 𝑓 (𝑧) is differen-
187 tiable at 𝑐 if and only if 𝑓 (𝑧)
¯ is differentiable at 𝑐.
¯
188 2.7 (Cauchy-Riemann equations in polar coordinates) Write 𝑓 = 𝑢 + 𝚤𝑣, and express 𝑢 and
189 𝑣 as (real-valued) functions of 𝑟 and 𝜃 . Since 𝑥 = 𝑟 cos 𝜃 and 𝑦 = 𝑟 sin 𝜃 , we have 𝑢𝑟 =
190 𝑢𝑥 cos 𝜃 + 𝑢𝑦 sin 𝜃 , 𝑣𝑟 = 𝑣 𝑥 cos 𝜃 + 𝑣𝑦 sin 𝜃 , 𝑢𝜃 = −𝑢𝑥 𝑟 sin 𝜃 + 𝑢𝑦 𝑟 cos 𝜃 , 𝑣𝜃 = −𝑣 𝑥 𝑟 sin 𝜃 +
191 𝑣𝑦 𝑟 cos 𝜃 . Therefore the Cauchy-Riemann equations are
𝑟𝑢𝑟 = 𝑣𝜃 ; 𝑟𝑣𝑟 = −𝑢𝜃 .

192 Lecture 3. Power series

193 3.1. Definition. A (formal) power series in the variable 𝑧 is an expression of the form
∑︁
𝑎𝑛 𝑧 𝑛
𝑛∈N

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 7

where the 𝑎𝑛 are complex numbers. AÍformal power series 𝑛∈N 𝑎𝑛 𝑧𝑛 is said to converge (re-
Í
194
195 spectively, diverge) at 𝑐 ∈ C if the series 𝑛∈N 𝑎𝑛𝑐 𝑛 of complex numbers converges (respectively,
196 diverges). For a𝑈 ⊆ C, a power series is said to converge on𝑈 if it converges at 𝑐 for every 𝑐 ∈ 𝑈 .
197 We will often drop the word ‘formal’ while talking about power series.
3.2. Definition. The radius of convergence of the series 𝑛∈N 𝑎𝑛 𝑧𝑛 is
Í
198
  −1
1
lim sup |𝑎𝑛 | 𝑛 .
𝑛→∞
199 (Here, we mean that the radius of convergence is 0 (respectively +∞) if the lim sup is +∞ (re-
200 spectively, 0).) A series is said to be convergent if its radius of convergence is positive.
3.3. Theorem. Let 𝑅 be the radius of convergence of the series 𝑛∈N 𝑎𝑛 𝑧𝑛 .
Í
201
202 (1) It converges absolutely in 𝐵 0,𝑅 ; in particular, it converges in 𝐵 0,𝑅 .
203 (2) For every 0 ≤ 𝜌 < 𝑅, the sequence of functions
𝑚
∑︁
𝑐 ↦→ 𝑎𝑛𝑐 𝑛 , 𝑚 ∈ N,
𝑛=0

204 converges uniformly in 𝐵 0,𝜌 .

205 (3) For every 𝑐 ∈ C r 𝐵 0,𝑅 , the series is unbounded at 𝑐.
We will often abuse terminology
Í and call a power series a function on 𝐵 0,𝑅 , by
Í 𝑛
206 𝑛∈N 𝑎𝑛 𝑧
207 which we mean the function 𝑐 ↦→ 𝑛∈N 𝑎𝑛𝑐 on 𝐵 0,𝑅 .
𝑛

208 Proof of Theorem. (1): It suffices to prove the assertion about absolute convergence, i.e., that
∑︁
|𝑎𝑛 ||𝑧|𝑛
𝑛∈N
209 converges whenever |𝑧| < 𝑅. WithoutÍloss of generality, the 𝑎𝑛 are non-negative real numbers;
210 we want to show that for 0 ≤ 𝑥 < 𝑅, 𝑛∈N 𝑎𝑛 𝑥 𝑛 converges. Let 𝑥 < 𝑦 < 𝑅. There exists 𝑁 ∈ N
1
such that 𝑎𝑛𝑛 < 𝑦1 for every 𝑛 ≥ 𝑁 ; hence 𝑎𝑛 𝑥 𝑛 < (𝑥/𝑦)𝑛 for every 𝑛 ≥ 𝑁 . Hence 𝑛∈N 𝑎𝑛 𝑥 𝑛
Í
211
212 converges.
Í𝑚(2): Let𝑛𝜌 < 𝜎 < 𝑅. Then, as earlier, |𝑎𝑛 𝑧 | ≤ (𝜌/𝜎) for all sufficiently large 𝑛. Write 𝑠𝑚 (𝑧) =
213
𝑛 𝑛

214
Í𝑛=0 𝑎𝑛 𝑧 . Let 𝜖 Í
> 0. Then there Í exists 𝑁 such that for every 𝑚 > 𝑘 Í ≥ 𝑁 , |𝑠𝑚 (𝑧) − 𝑠𝑘 (𝑧)| =
215 | 𝑛=𝑘+1 𝑎𝑛 𝑧 | ≤ Í 𝑛=𝑘+1 |𝑎𝑛 𝑧 | ≤ 𝑛=𝑘+1 (𝜌/𝜎) < 𝜖, since the series 𝑛∈N (𝜌/𝜎)𝑛 converges.
𝑚 𝑛 𝑚 𝑛 𝑚 𝑛

216 Note that by (1), 𝑛∈N 𝑎𝑛 𝑧𝑛 converges in 𝐵 0,𝑅 to give a function 𝑓 (𝑧) on 𝐵 0,𝑅 . By taking 𝑚 → ∞
(keeping 𝑘 fixed), we see that |𝑓 (𝑧) − 𝑠𝑘 (𝑧)| < 𝜖, i.e., 𝑛∈N 𝑎𝑛 𝑧𝑛 converges uniformly on 𝐵 0,𝜌 .
Í
217
1
218 (3): Let |𝑐 | > 𝑦 > 𝑅. Then there are arbitrarily large 𝑛 such that 𝑎𝑛𝑛 > 𝑦1 . Hence
lim |𝑎𝑛𝑐 𝑛 | ≠ 0,
𝑛−→∞
219 so the series does not converge. 
3.4. Remark. The radius
Í of convergence of the complex power series 𝑛∈N 𝑎𝑛 𝑧𝑛 and that of
Í
220
221 the real power series 𝑛∈N |𝑎𝑛 |𝑥 𝑛 are the same. Hence the tests for determining the radius of
222 convergence of real power series can be used to determine the radius of convergence of complex
223 power series also.
3.5. Example. The radius of convergence of 𝑖=1 𝑛 is 1. The series does not coverge at 𝑧 = 1,
Í∞ 𝑧 𝑛
224
225 but converges at 𝑧 = −1.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 8 MANOJ KUMMINI

3.6. Proposition. Let 𝑛∈N 𝑎𝑛 𝑧𝑛 beÍa convergent power series with radius of convergence 𝑅. Then it is
Í
226
227 holomorphic on 𝐵 0,𝑅 , with derivative 𝑛∈N 𝑛𝑎𝑛 𝑧𝑛−1 . Further, the radius of convergence of the derivative
228 is 𝑅.
Proof. We will first prove that the radius of convergence of the series 𝑛∈N 𝑛𝑎𝑛 𝑧𝑛−1 is 𝑅. Indeed,
Í
229
1 1 1 1
lim sup (𝑛|𝑎𝑛 |) 𝑛 = lim 𝑛 𝑛 lim sup |𝑎𝑛 | 𝑛 = lim sup |𝑎𝑛 | 𝑛 = 1/𝑅.
𝑛 𝑛 𝑛 𝑛
Write 𝑓 (𝑧) and 𝑓1 (𝑧) respectively for the functions 𝑛∈N 𝑎𝑛 and 𝑛∈N 𝑛𝑎𝑛 𝑧𝑛−1 on 𝐵 0,𝑅 . We
Í Í
230 𝑧𝑛
231 want to show that 𝑓 0 (𝑐) = 𝑓1 (𝑐) for every 𝑐 ∈ 𝐵 0,𝑅 . Let 𝑐 ∈ 𝐵 0,𝑅 . We will show that
𝑓 (𝑐 + ℎ) − 𝑓 (𝑐)
lim − 𝑓1 (𝑐) = 0.
ℎ→0 ℎ
Write 𝑠𝑛 (𝑧) = 𝑛𝑖=0 𝑎𝑖 𝑧𝑖 . Then 𝑠𝑛0 (𝑧) = 𝑛𝑖=1 𝑖𝑎𝑖 𝑧𝑖−1 . Write 𝑅𝑛 (𝑧) = 𝑓 (𝑧) − 𝑠𝑛 (𝑧) on 𝐵 0,𝑅 . Then for
Í Í
232
233 every sufficiently small 𝑟 and every 𝑧 ∈ 𝐵𝑐,𝑟 ,
   
𝑓 (𝑧) − 𝑓 (𝑐) 𝑠𝑛 (𝑧) − 𝑠𝑛 (𝑐) 0 0
𝑅𝑛 (𝑧) − 𝑅𝑛 (𝑐)
− 𝑓1 (𝑐) = − 𝑠𝑛 (𝑐) + 𝑠𝑛 (𝑐) − 𝑓1 (𝑐) + .
𝑧 −𝑐 𝑧 −𝑐 𝑧 −𝑐
234 Choose 𝑟 above such that |𝑐 | + 𝑟 < 𝜌 < 𝑅. Let 𝜖 > 0.
235 Since
Í∞ Í∞ Í∞ ∞ 𝑖−1
𝑖=𝑛+1 𝑎𝑖 𝑧 − 𝑖=𝑛+1 𝑎𝑖 𝑐 𝑖=𝑛+1 𝑎𝑖 (𝑧 − 𝑐 )
𝑖 𝑖 𝑖 𝑖
𝑅𝑛 (𝑧) − 𝑅𝑛 (𝑐) ∑︁ ∑︁
= = = 𝑎𝑖 𝑧 𝑗 𝑐 𝑖−1−𝑗 ,
𝑧 −𝑐 𝑧 −𝑐 𝑧 −𝑐 𝑖=𝑛+1 𝑗=0

236 we see that

∞
𝑅𝑛 (𝑧) − 𝑅𝑛 (𝑐) ∑︁
≤ 𝑖 |𝑎𝑖 |𝜌 𝑖−1 .
𝑧 −𝑐 𝑖=𝑛+1
We already observed that 𝑚=1 𝑚𝑎𝑚 𝑧𝑚−1 converges in 𝐵 0,𝑅 . The same argument shows that
Í∞
237
238 there exists 𝑛 0 such that for each 𝑛 > 𝑛 0 ,
𝑅𝑛 (𝑧) − 𝑅𝑛 (𝑐) 𝜖
< .
𝑧 −𝑐 3
239 Similarly, there exists 𝑛 1 such that for each 𝑛 > 𝑛 1 ,
𝜖
𝑠𝑛0 (𝑐) − 𝑓1 (𝑐) < .
3
240 Fix 𝑛 ≥ max{𝑛 0, 𝑛 1 }. There exists 𝛿 > 0 such that for all 𝑧 ∈ 𝐵𝑐,𝛿
𝑠𝑛 (𝑧) − 𝑠𝑛 (𝑐) 𝜖
− 𝑠𝑛0 (𝑐) < .
𝑧 −𝑐 3
241 Hence for all 𝑧 ∈ 𝐵𝑐,𝛿
𝑓 (𝑧) − 𝑓 (𝑐)
− 𝑓1 (𝑐) < 𝜖.
𝑧 −𝑐
242 Therefore 𝑓 0 = 𝑓1 on 𝐵 0,𝑅 . 
3.7. Corollary. With notation as in the proposition, write 𝑓 (𝑧) for the function 𝑛∈N 𝑎𝑛 𝑧𝑛 on 𝐵 0,𝑅 . Then
Í
243
244 for every 𝑘 ≥ 1, the derivative 𝑓 (𝑘) (𝑧) of 𝑓 (𝑧) exists on 𝐵 0,𝑅 . Moreover, for every 𝑘 ∈ N, 𝑘!𝑎𝑘 = 𝑓 (𝑘) (0).
245 Proof. Immediate from the proposition. 
3.8. Proposition. Let 𝑛∈N 𝑎𝑛 𝑧𝑛 be a convergent power Í series𝑛 such that 𝑎𝑚 ≠ 0 for some 𝑚. Then there
Í
246
247 exists 𝑅 > 0 such that for every 𝑐 ∈ 𝐵 0,𝑅 with 𝑐 ≠ 0, 𝑛≥1 𝑎𝑛𝑐 ≠ 0.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 9

248 Proof. Let𝑚 be the smallest integer such that 𝑎𝑚 ≠ 0. Write the given series as 𝑧 𝑚Í
𝑛∈N 𝑎𝑛+𝑚 𝑧 .
𝑛

There exists 𝑅 > 0 such that 𝑛∈N 𝑎𝑛+𝑚𝑐 𝑛 ≠ 0 for every 𝑐 ∈ 𝐵 0,𝑅 , by continuity. Now note that
Í
249
250 for every 𝑐 ∈ 𝐵 0,𝑅 , 𝑐𝑚 = 0 only if 𝑐 = 0. 

251 Exercises.
252 3.1 Read the statement of the Weierstrass 𝑀-test in Ahlfors, Chapter 2, Section 2.3 and un-
253 derstand its proof.
254 3.2 All the exercises in Ahlfors, Chapter 2, Section 2.4 (‘Power series’)
3.3 Show that the radius of convergence of 𝑖∈N 𝑎𝑖 𝑧𝑖 is
Í
255
∑︁
sup{𝑟 ∈ R | 𝑟 ≥ 0, |𝑎𝑖 |𝑟 𝑖 converges}.
𝑖∈N

3.4 Let 𝑖∈N 𝑎𝑖 andÍ 𝑖∈N 𝑏𝑖 be convergent series of complex numbers, Í and 𝛼, 𝛽Í∈ C. Show
Í Í
256
257 that the series 𝑖∈N (𝛼𝑎𝑖 + 𝛽𝑏𝑖 ) is convergent and its value is 𝛼 𝑖∈N 𝑎𝑖 + 𝛽 𝑖∈N 𝑏𝑖 .
258 3.5 Prove the properties of limits superior and inferior listed in Rodrı́guez, Kra and Gilman,
259 Section 3.1.1. 1
1
260 3.6 Show that lim𝑛→∞ 𝑛 𝑛 = 1 and that for every Í 𝑘, lim𝑛𝑛→∞ 𝑘 = 1
𝑛
𝑛

3.7 The set C[[𝑧]] of all formal power series 𝑛∈N 𝑎𝑛 𝑧 form a commutative ring with
∑︁ ∑︁ ∑︁
addition : 𝑎𝑛 𝑧 𝑛 + 𝑏𝑛 𝑧 𝑛 = (𝑎𝑛 + 𝑏𝑛 )𝑧𝑛 ;
𝑛∈N 𝑛∈N 𝑛∈N
∑︁ ∑︁ 𝑛
∑︁ ∑︁
multiplication : 𝑎𝑛 𝑧 𝑛 · 𝑏𝑛 𝑧 𝑛 = (𝑎𝑘 𝑏𝑛−𝑘 )𝑧𝑛 .
𝑛∈N 𝑛∈N 𝑛∈N 𝑘=0

261 It contains C as a subring identified with the ‘constant’ power series: 𝑐 ↔ 𝑐 + 0𝑧 + 0𝑧 2 +

· · · . If 𝑎 0 ≠ 0, then 𝑛∈N 𝑎𝑛 𝑧𝑛 has an inverse in C[[𝑧]].
Í
262
263 3.8 Í
The subset C{𝑧} of C[[𝑧]] consisting of all the convergent power series is a subring. If
Í𝑛∈N 𝑎𝑛 𝑧𝑛 ∈ C{𝑧} and 𝑎 0 ≠ 0, then its inverse in C[[𝑧]] in fact belongs to C{𝑧}. (Hint:
264
𝑛

265 𝑛∈N 𝑎𝑛 𝑧 converges to something non-zero in a neighbourhood of 0.)

266 3.9 (Some ring-theoreticÍproperties of C[[𝑧]] and of C{𝑧}, not relevant for this course.) The
267 map C[[𝑧]] −→ C, 𝑛∈N 𝑎𝑛 𝑧𝑛 ↦→ 𝑎 0 is a surjective ring homomorphism; its kernel is
268 generated by 𝑧; hence the ideal 𝔪 generated by 𝑧 is a maximal ideal. Every element of
269 C[[𝑧]] r 𝔪 is invertible in C[[𝑧]], so 𝔪 is the unique maximal ideal of C[[𝑧]]. If 𝐼 is
270 a proper ideal of C[[𝑧]], then 𝐼 = 𝔪𝑡 (i.e., the ideal generated by 𝑧𝑡 ) for some 𝑡 ≥ 1.
271 Similar statements for C{𝑧} also.

272 Lecture 4. Analytic functions

273 4.1. Definition. Let 𝑈 be a domain. We say that 𝑓 : Í𝑈 −→ C is (complex-)analytic Í if for every 𝑐 ∈
274 𝑈 , there exist 𝛿 > 0 and a convergent power series 𝑎
𝑖∈N 𝑖 𝑧 𝑖 such that 𝐵
𝑐,𝛿 ⊆ 𝑈 , 𝑖∈N 𝑎𝑖 (𝑧 −𝑐)
𝑖

converges on 𝐵𝑐,𝛿 and 𝑓 (𝜁 ) = 𝑖∈N 𝑎𝑖 (𝜁 − 𝑐) for every 𝜁 ∈ 𝐵𝑐,𝛿 .

Í 𝑖
275

276 4.2. Remark. The coefficients 𝑎𝑖 in the expansion of 𝑓 as a power series centred at 𝑐 ∈ 𝑈 might
277 depend on 𝑐. It might not be possible to choose 𝑎𝑖 that will work at every 𝑐 ∈ 𝑈 . This is not
278 surprising. We have seen that for any power series centred at 𝑐 ∈ 𝐶, the set of points at which
279 it converges contains an open disc 𝐵𝑐,𝑅 and is contained inside the closed disc 𝐵𝑐,𝑅 . However 𝑈
280 might not be of this shape. We will see one such𝑈 (occurring in a natural way) when we discuss
281 branches of the logarithm, later.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 10 MANOJ KUMMINI

282 4.3. Remark. Let 𝑈 be a domain and 𝑓 : 𝑈 −→ C an analytic function. Then for every 𝑘 ≥ 1,
283 𝑓 (𝑘) (𝑧) is an analytic function on 𝑈 . Moreover, for every 𝑐 ∈ 𝑈 , there exists a neighbourhood
284 on which ∑︁
𝑓 (𝑧) = 𝑓 (𝑛) (𝑐)(𝑧 − 𝑐)𝑛 .
𝑛∈N

285 4.4. Remark. Every analytic function is holomorphic. After proving a version of the Cauchy
286 integral formula for a disc (Theorem 14.1), we will show that every holomorphic function is
287 analytic (Corollary 15.5). This is not the same situation for functions from R to R. For every
288 positive integer 𝑘, there exist 𝑓 : R −→ R such that the 𝑘th order derivative 𝑓 (𝑘) exists, but
289 is not continuous, so in particular 𝑓 (𝑘+1) does not exist. There are functions 𝑓 : R −→ R such
290 that 𝑓 (𝑘) exists for every positive integer 𝑘 (such functions are called smooth functions) but 𝑓 is
291 smooth but not real-analytic, i.e., 𝑓 does not have a power-series expansion on its domain.
4.5. Proposition (Lang, Chapter II, §4, Theorem 4.1). Let 𝑛∈N 𝑎𝑛 𝑧𝑛 be a convergent power series
Í
292
293 with radius of convergence 𝑅. Then it is analytic on 𝐵 0,𝑅 .
Proof. Write 𝑓 (𝑧) = 𝑛∈N 𝑎𝑛 𝑧𝑛 on 𝐵 0,𝑅 . Let 𝑐 ∈ 𝐵 0,𝑅 . We want to show that 𝑓 can be represented
Í
by a convergent power series centred at 𝑐 in a neighbourhood of 𝑐. To see this, choose 𝜖 > 0
such that 𝐵𝑐,𝜖 ⊆ 𝐵 0,𝑅 . On 𝐵𝑐,𝜖 , we can write
∞
∑︁
𝑓 (𝑧) = 𝑎𝑛 (𝑧 − 𝑐 + 𝑐)𝑛
𝑛=0
∞ ∑︁
𝑛  
∑︁ 𝑛
= 𝑎𝑛𝑐 𝑛−𝑘 (𝑧 − 𝑐)𝑘 .
𝑛=0 𝑘=0
𝑘
294 Claim: !
∞ ∑︁
∞  
∑︁ 𝑛
𝑔(𝑧) := 𝑎𝑛𝑐 𝑛−𝑘 (𝑧 − 𝑐)𝑘 .
𝑘
𝑘=0 𝑛=𝑘
295 converges and equals 𝑓 (𝑧) in 𝐵𝑐,𝜖 .
296 To prove the claim, let 𝑧 ∈ 𝐵𝑐,𝜖 . Note that |𝑐 | + |𝑧 − 𝑐 | < 𝑅. Hence the series
∞ ∑︁𝑛  
∑︁
𝑛
∑︁ 𝑛
|𝑎𝑛 |(|𝑐 | + |𝑧 − 𝑐 |) = |𝑎𝑛 ||𝑐 |𝑛−𝑘 |𝑧 − 𝑐 |𝑘 .
𝑛 𝑛=0
𝑘
𝑘=0
297 converges. (Recall that inside the open disc of convergence, we have absolute convergence.)
298 Hence we can change the order of summation:
∞ ∑︁𝑛   ∞ ∑︁
∞  
∑︁ 𝑛 𝑛−𝑘 𝑘
∑︁ 𝑛
|𝑎𝑛 ||𝑐 | |𝑧 − 𝑐 | = |𝑎𝑛 ||𝑐 |𝑛−𝑘 |𝑧 − 𝑐 |𝑘 .
𝑛=0
𝑘 𝑘
𝑘=0 𝑘=0 𝑛=𝑘
299 Therefore 𝑔(𝑧) converges absolutely in 𝐵𝑐,𝜖 . The same argument also shows that 𝑔(𝑧) = 𝑓 (𝑧) on
300 𝐵𝑐,𝜖 . 
301 4.6. Proposition. Let 𝑈 be a domain and 𝑓 an analytic function on 𝑈 that is not identically zero. Then
302 the zeros of 𝑓 are isolated, i.e. for every 𝑐 ∈ 𝑈 with 𝑓 (𝑐) = 0, there exists 𝜖 > 0 such that 𝐵𝑐,𝜖 ⊆ 𝑈 and
303 𝑓 (𝜁 ) ≠ 0 for every 𝜁 ∈ 𝐵𝑐,𝜖 r {𝑐}.
304 Proof. Let 𝐴 = {𝑐 ∈ 𝑈 | 𝑓 (𝑐) ≠ 0}. Since 𝑓 is continuous and not identically zero, 𝐴 is open
305 and non-empty. We may assume that 𝐴 ≠ 𝑈 . Write 𝐴 for the closure of 𝐴 in 𝑈 . We want to
306 show that the points in 𝑈 r 𝐴 are isolated. We will show the following:

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 11

307 (1) For each 𝑐 ∈ 𝐴 r 𝐴, there exists 𝜖 > 0 such that 𝐵𝑐,𝜖 r {𝑐} ⊆ 𝐴.
308 (2) 𝐴 = 𝑈 .
Í Let 𝑐 ∈ 𝐴 r𝑛 𝐴. Let 𝑛∈N 𝑎𝑛 𝑧 be a convergent power series and 𝜖 > 0 such that 𝑓 (𝑧) =
Í 𝑛
309
310 𝑛∈N 𝑎𝑛 (𝑧 − 𝑐) on 𝐵𝑐,𝜖 ⊆ 𝑈 . Since 𝐵𝑐,𝜖 ∩ 𝐴 ≠ ∅, it follows that 𝑓 is not identically zero on 𝐵𝑐,𝜖 .
311 Therefore there exists 𝑚 such that 𝑎𝑚 ≠ 0. By Proposition 3.8, we may assume that 𝑓 (𝜁 ) ≠ 0
312 for every 𝜁 ∈ 𝐵𝑐,𝜖 r {𝑐}.
313 We now show that 𝐴 = 𝑈 . By way of contradiction, assume that 𝐴 ≠ 𝑈 . We will show that
314 𝑈 r𝐴 is closed. Let 𝑐 ∈ 𝑈 be a limit point of 𝑈 r𝐴. Now, if 𝑐 ∈ 𝐴 r𝐴, then by above, there exists
315 𝜖 > 0 such that 𝐵𝑐,𝜖 r {𝑐} ⊆ 𝐴. If 𝑐 ∈ 𝐴, then there exists 𝜖 > 0 such that 𝐵𝑐,𝜖 ⊆ 𝐴. In both
316 cases, we cannot have a sequence in 𝑈 r 𝐴 converging to 𝑐. Hence 𝑐 ∈ 𝑈 r 𝐴, so it is closed.
317 This now leads to a contradiction, since 𝑈 is connected and both 𝐴 and 𝑈 r 𝐴 are non-empty
318 and closed. Therefore 𝐴 = 𝑈 . 
319 Exercises.
320 4.1 Let 𝑈 be a domain, 𝑐 0 ∈ C and 𝜏 : 𝑈 −→ C be the map 𝑐 ↦→ 𝑐 + 𝑐 0 . Then 𝜏 is continuous
321 and injective; the inverse of 𝜏 on Im(𝜏) (which exists since 𝜏 is injective) is continuous.
322 Im(𝜏) is a domain. If 𝑓 is holomorphic (respectively, analytic) on 𝑈 , then 𝑓 𝜏 −1 is holo-
323 morphic (respectively, analytic) on Im(𝜏). (Using this, we can ‘translate’ many questions
324 about the local behaviour of holomorphic or analytic functions at 𝑐 ∈ 𝑈 to that of holo-
325 morphic or analytic functions at 0, in an appropriate neighbourhood of 0.)
326 4.2 Prove analogous statements when 𝑓 is replaced by the composite 𝑓 ◦ [𝜁 ↦→ 𝑐𝜁 ] where 𝑐
327 is a (fixed) non-zero complex number.
328 4.3 Let 𝑓 : 𝑈 −→ C be analytic on a domain 𝑈 . Show that if 𝑓 (𝑘) (𝑧) = 0 for every 𝑧 ∈ 𝑈 ,
329 then 𝑓 is given by a polynomial of degree at most 𝑘, hence, 𝑓 can be extended to an entire
330 function as follows:
331 (1) There is a nonempty open subset of 𝑈 on which 𝑓 is given by a polynomial 𝑝 of de-
332 gree at most 𝑘.
333 (2) 𝑓 − 𝑝 is zero on a nonempty open subset of 𝑈 , so it is zero on 𝑈 .
334 4.4 Let 𝑓 : 𝑈 −→ C be analytic on a domain 𝑈 , not identically zero. Let 𝐴 = {𝑐 ∈ 𝑈 |
335 𝑓 (𝑛) (𝑐) = 0 for every 𝑛 ∈ N}. 𝐴 is closed, since {𝑐 ∈ 𝑈 | 𝑓 (𝑛) (𝑐) = 0} is closed, for
336 every 𝑛 ∈ N. 𝐴 is open, since, for every 𝑐 ∈ 𝐴, there is a neighbourhood in 𝑈 on which
337 𝑓 is identically zero, and,
Í hence, this𝑛 neighbourhood is a subset of 𝐴. Thus 𝐴 = ∅. Now
338 let 𝑐 ∈ 𝑈 and 𝑓 (𝑧) = 𝑛∈N 𝑎𝑛 (𝑧 − 𝑐) in a neighbourhood of 𝑐. Then there exists 𝑘 such
339 that 𝑎𝑘 ≠ 0. Thus there exists a neighbourhood 𝑉 of 𝑐 in 𝑈 such that 𝑓 (𝑧) ≠ 0 for every
340 𝑧 ∈ 𝑉 r {𝑐}.
−2
341 4.5 Consider the function 𝑓 (𝑥) = 𝑒 −𝑥 in a neighbourhood of 0 in R. Show that 𝑓 (𝑘) exists
342 in a neighbourhood of 0 and that 𝑓 (𝑘) (0) = 0 for every 𝑘 ≥ 0. Hence 𝑓 is not real-
343 analytic in a neighbourhood of 0. This example was discovered by Cauchy and Hamilton.
4.6 (Not relevant for this course.) Let 𝑈 be a domain and A (𝑈 ) the set of analytic functions
on 𝑈 . It is a commutative ring with
addition : (𝑓 + 𝑔)(𝑧) = 𝑓 (𝑧) + 𝑔(𝑧);
multiplication : (𝑓 𝑔)(𝑧) = 𝑓 (𝑧)𝑔(𝑧).
344 It contains C as the subring of the constant functions on 𝑈 . It is an integral domain.
345 For 𝑐 ∈ 𝑈 , the set 𝔪𝑐 := {𝑓 ∈ A (𝑈 ) | 𝑓 (𝑐) = 0} is a maximal ideal of A (𝑈 ). There is
346 a ring homomorphism A (𝑈 ) −→ C{𝑧 − 𝑐} (the ring of convergent power series in the
347 variable 𝑧 − 𝑐) which factors through the localisation A (𝑈 )𝔪𝑐 .

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 12 MANOJ KUMMINI

348 Lecture 5. Exponential and logarithmic functions

349 By 𝑒 𝑧 or exp(𝑧) we mean an analytic function 𝑓 (𝑧) that 𝑓 0 (𝑧) = 𝑓 (𝑧) and 𝑓 (0) = 1. Suppose
350 that this
Í has a solution. Then in aÍ
neighbourhood of 0, it can be written as a convergent power
351 series 𝑛∈N 𝑎𝑛 𝑧 . Since 𝑓 (𝑧) = 𝑛∈N 𝑛𝑎𝑛 𝑧𝑛−1 and 𝑓 (0) = 𝑎 0 = 1, we conclude by induction
𝑛 0

352 that 𝑎𝑛 = 𝑛!1 . Since

𝑎𝑛+1
lim =0
𝑛→∞ 𝑎𝑛

we see that the series 𝑛∈N 𝑧𝑛! converges everywhere on C.

Í 𝑛
353

354 5.1. Proposition. (1) 𝑒 𝑧 𝑒 −𝑧 = 1 for every 𝑧 ∈ C.

355 (2) In particular, 𝑒 ≠ 0 for every 𝑧 ∈ C.
𝑧

356 (3) 𝑒 (𝑧+𝑐) = 𝑒 𝑧 𝑒 𝑐 for every 𝑧, 𝑐 ∈ C.

357 Proof. (1) NOTE: A priori 𝑒 −𝑧 is not 𝑒1𝑧 , but just the composite function [𝑧 ↦→ exp(𝑧)] ◦ [𝑧 ↦→
358 −𝑧]. Hence 𝑒 −𝑧 is analytic on C,4 and, hence, so is 𝑒 𝑧 𝑒 −𝑧 . Its derivative is 0, so it is a constant
359 function.5 Now note that its value at 0 is 1.
360 (2) Follows immediately from (1).
361 (3) Fix 𝑐 and consider
𝑒 (𝑧+𝑐)
ℎ(𝑧) = 𝑐
𝑒
6
362 as a function of 𝑧. It is analytic on C and ℎ (𝑧) = ℎ(𝑧) and ℎ(0) = 1. Hence ℎ(𝑧) = 𝑒 𝑧 .
0 
From the exponential function, we can define sin(𝑧) and cos(𝑧):
𝑒𝚤𝑧 + 𝑒 −𝚤𝑧
cos 𝑧 =
2
𝑒𝚤𝑧 − 𝑒 −𝚤𝑧
sin 𝑧 =
2𝚤

2 4
363 5.2. Proposition. (1) cos 𝑧 = 1 − 𝑧2! + 𝑧4! − · · · .
3 5
364 (2) sin 𝑧 = 𝑧 − 𝑧3! + 𝑧5! − · · · .
365 (3) 𝑒𝚤𝑧 = cos 𝑧 + 𝚤 sin 𝑧.
366 (4) cos2 𝑧 + sin2 𝑧 = 1.
367 (5) cos(−𝑧) = cos 𝑧.
368 (6) sin(−𝑧) = − sin 𝑧.
369 (7) cos0 (𝑧) = − sin 𝑧.
370 (8) sin0 (𝑧) = cos 𝑧.
371 (9) If 𝑥 is real, then the new definitions of 𝑒 𝑥 , cos 𝑥, sin 𝑥 agree with the definitions in the case of real
372 numbers.
373 (10) 𝑒 𝑥+𝚤𝑦 = 𝑒 𝑥 (cos 𝑦 + 𝚤 sin 𝑦). In particular, 𝑒𝚤𝜋 = 1.
374 Proof of the above proposition is left as an exercise.
375 5.3. Definition. Let 𝑧 ∈ C r {0}. By an argument arg 𝑧 of 𝑧, we mean an real number 𝜃 such
376 that 𝑧 = |𝑧|𝑒𝚤𝜃 . Define the principal argument Arg 𝑧 of 𝑧 to be the argument in (−𝜋, 𝜋].
4Exercise 4.2
5Exercise 4.3
6Exercise 4.1

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 13

377 5.4. Definition. For a fixed choice of arg 𝑧, we often write log 𝑧 for log |𝑧| + 𝚤 arg 𝑧. Define
Log 𝑧 = log |𝑧| + 𝚤 Arg 𝑧
378 on C r (−∞, 0].
379 5.5. Remark. Let 𝑧 ∈ C r {0}. If 𝜃 1 and 𝜃 2 are arguments of 𝑧, then 𝜃 1 − 𝜃 2 is a multiple of 2𝜋.
380 Note that 𝑒 log 𝑧 = 𝑧.
381 5.6. Proposition. Log 𝑧 is holomorphic on C r (−∞, 0] with derivative 1/𝑧.
382 Proof. On the given domain, the real and imaginary parts of Log 𝑧, viz., log |𝑧| and Arg 𝑧 are
383 differentiable functions of the reals coordinates 𝑥 and 𝑦. Hence it suffices to check that they
384 satisfy the Cauchy-Riemann equations. For this, use the version in polar coordinates: 𝑢 = log 𝑟 ,
385 𝑣 = 𝜃 . Hence 𝑟𝑢𝑟 = 1 = 𝑣𝜃 and 𝑟𝑣𝑟 = 0 = −𝑢𝜃 .
386 Since Log 𝑧 is holomorphic, we can use differentiate 𝑒 Log 𝑧 = 𝑧 to get 𝑒 Log 𝑧 (Log0 𝑧) = 1, i.e.,
387 (Log0 𝑧) = 1/𝑧. 
388 Let 𝑈 be a domain and 𝑓 a continuous function on 𝑈 . We say that 𝑓 is a branch of the logarithm
389 on 𝑈 if 𝑒 𝑓 (𝑧) = 𝑧 for every 𝑧 ∈ 𝑈 . A branch of the logarithm 𝑓 on 𝑈 is principal if 𝑓 (𝑧) = Log(𝑧)
390 for every 𝑧 ∈ 𝑈 ∩ C r (−∞, 0].
391 5.7. Proposition. The power series
∞
∑︁ (𝑧 − 1)𝑛
(−1)𝑛−1
𝑛=1
𝑛
392 is the principal branch of the logarithm in 𝐵 1,1 .
393 Proof. The given power series has Í radius of convergence 1, so it defines an analytic function
394 𝑓 (𝑧) on 𝐵 1,1 . Note that 𝑓 (𝑧) = 𝑛∈N (−1) (𝑧 − 1)𝑛 = 1/𝑧. (Exercise: check last equality.) Let
0 𝑛

395 𝑔(𝑧) = 𝑒 𝑓 (𝑧) . Then 𝑔0 (𝑧) = 𝑒 𝑓 (𝑧) /𝑧 and 𝑔00 (𝑧) = 0. Hence 𝑔0 (𝑧) = 𝛼 a constant.7 Since 𝑔0 (1) = 1,
396 it follows that 𝑓 (𝑧) is a branch of the logarithm. Since Log 1 = 𝑓 (1), it follows that Log 𝑧 = 𝑓 (𝑧),
397 because two branches differ by an integer multiple of 2𝜋𝚤. 
398 Exercises.
399 5.1 Verify the properties of sin 𝑧 and cos 𝑧 listed in class.
3 2 2 4
400 5.2 Show that for 𝑥 > 0, 𝑥 − 𝑥6 < sin 𝑥 < 𝑥 and that 1 − 𝑥2 < cos 𝑥 < 1 − 𝑥2 + 𝑥24 . (Hint: Use
401 the fact that sin 𝑥 < 1 and
√ cos 𝑥 < 1 and integrate sin and cos alternately.)
402 5.3 Since cos 0 = 1 and cos( 3) < 0, there is a smallest real number 𝜃 such that cos 𝜃 = 0.
403 Then sin 𝜃 = ±1. One can then define 𝜋 := 2𝜃 .
404 5.4 Expand 𝑧1 as a power series around 𝑧 = 1. Find its radius of convergence.
405 5.5 Show that Log(𝑧 1𝑧 2 ) = Log(𝑧 1 ) + Log(𝑧 2 ) + 𝛿 for an appropriate 𝛿.
406 5.6 Let 𝑈 be a domain not containing 0 and 𝑓 and 𝑔 branches of the logarithm on 𝑈 . Show
407 that the function ℎ(𝑧) := (𝑓 (𝑧) −𝑔(𝑧))/(2𝜋𝚤) on 𝑈 takes only integer values, by showing
408 that 𝑒 2𝜋𝚤ℎ(𝑧) = 1. Hence there exists 𝑛 ∈ Z such that ℎ(𝑧) = 𝑛 for every 𝑧. Hence 𝑓 (𝑧) −
409 𝑔(𝑧) = 2𝑛𝜋𝚤. Conversely, if 𝑓 (𝑧) − 𝑔(𝑧) = 2𝑛𝜋𝚤 for some 𝑛, and 𝑓 (𝑧) is a branch of the
410 logarithm if and only if 𝑔(𝑧) is.
7We cannot prove this with the results we proved so far. It is true that 𝑔(𝑧) is analytic, being the composite of
the two analytic functions 𝑓 (𝑧) and 𝑒 𝑧 ; we can then use Exercise 4.3 from the last section. However the proof that
the composite of two analytic functions is analytic is long, and we will not discuss it in class. Instead we use the
fact (easily provable, using the chain rule) that the composite of two holomorphic functions is holomorphic and
Proposition 7.5. The proof of Proposition 7.5 does not refer to anything in this section, so our argument is not
circular.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 14 MANOJ KUMMINI

411 5.7 Let 𝑈 be a domain not containing 0 and 𝑓 a branch of the logarithm on 𝑈 . Show that 𝑓
412 is holomorphic on 𝑈 as follows. If 𝑐 ∈ 𝑈 r (∞, 0], then there is a neighbourhood 𝐵𝑐,𝑅
413 which does not intersect (∞, 0]; on that neighbourhood, 𝑓 (𝑧) differs from Log(𝑧) by a
414 holomorphic function, so 𝑓 (𝑧) is holomorphic. If𝑐 ∈ 𝑈 ∩(∞, 0], then ‘rotate the domain
415 on which Log is holomorphic’ by an appropriate 𝜃 by using the function Log(𝑒𝚤𝜃 𝑧) − 𝚤𝜃 .
416 Conclude that 𝑓 0 (𝑧) = 1/𝑧.
417 Lecture 6. Path integrals, I
418 6.1. Definition. Let 𝑈 be a domain. A path (also called an arc) in 𝑈 is a continuous map 𝛾 :
419 [𝑎, 𝑏] −→ 𝑈 . Let 𝛾 be a path. Say that 𝛾 is closed if 𝛾 (𝑏) = 𝛾 (𝑎). By −𝛾, we mean the function
420 [𝑎, 𝑏] −→ 𝑈 , 𝑡 ↦→ 𝛾 (𝑎 + 𝑏 − 𝑡), and call it the opposite path of 𝛾. Say that 𝛾 is differentiable if the
421 functions [𝑎, 𝑏] −→ R, 𝑡 ↦→ <(𝛾 (𝑡)) and 𝑡 ↦→ =(𝛾 (𝑡)) are in 𝐶 1 ([𝑎, 𝑏]). For a differentiable
422 path𝛾, write𝛾 0 (𝑡) for (<(𝛾 (𝑡))) 0 +𝚤 (=(𝛾 (𝑡))) 0. Say that𝛾 is piecewise differentiable if there exists
423 a partition 𝑎 = 𝑡 0 < 𝑡 1 < · · · < 𝑡𝑛 = 𝑏 such that 𝛾 | [𝑡𝑖 ,𝑡𝑖+1 ] is in 𝐶 1 ([𝑡𝑖 , 𝑡𝑖+1 ]) for every 0 ≤ 𝑖 < 𝑛;
424 we also say that the partition 𝑎 = 𝑡 0 < 𝑡 1 < · · · < 𝑡𝑛 = 𝑏 is good for 𝛾 to denote this fact.
425 6.2. Proposition. Between any pair of points in a domain, there exists a piecewise-differentiable path
426 connecting them.
427 Proof. Let 𝑈 be a domain and 𝑐 ∈ 𝑈 . We show that the set
𝐴 := {𝜁 ∈ 𝑈 | there exists a piecewise-differentiable path from 𝑐 to 𝜁 }
428 is both open and closed. Let 𝜁 ∈ 𝐴. Then there exists 𝑅 > 0 such that 𝐵𝜁 ,𝑅 ⊆ 𝑈 . For every
429 𝜁 0 ∈ 𝐵𝜁 ,𝑅 , the radial straight line joining 𝜁 and 𝜁 0 extends a piecewise-differentiable path from
430 𝑐 to 𝜁 ; hence 𝐵𝜁 ,𝑅 ⊆ 𝐴. Hence 𝐴 is open. Now let 𝑝 ∈ 𝐴. Let 𝑟 > 0. Let 𝜁 ∈ 𝐵𝑝,𝑟 ∩ 𝐴. Then the
431 radial straight line joining 𝜁 and 𝑝 extends a piecewise-differentiable path from 𝑐 to 𝜁 ; hence
432 𝑝 ∈ 𝐴, so 𝐴 is closed.
433 Note that 𝑐 ∈ 𝐴, so 𝐴 ≠ ∅. Now, since 𝑈 is connected, we see that 𝐴 = 𝑈 . 
434 6.3. Definition. Let 𝑎 < 𝑏 ∈ R and 𝑓 : [𝑎, 𝑏] −→ C a continuous function. Define
∫ 𝑏 ∫ 𝑏 ∫ 𝑏
𝑓 (𝑡)d𝑡 = <(𝑓 (𝑡))d𝑡 + 𝚤 =(𝑓 (𝑡))d𝑡 .
𝑎 𝑎 𝑎
435 6.4. Lemma. Let 𝑎 = 𝑠 0 < 𝑠 1 < · · · < 𝑠𝑚 = 𝑏 and 𝑎 = 𝑡 0 < 𝑡 1 < · · · < 𝑡𝑛 = 𝑏 be good
436 partitions for a piecewise-differentiable path 𝛾 : [𝑎, 𝑏] −→ 𝑈 . Let 𝑢 0, . . . , 𝑢𝑘 be distinct elements of
437 {𝑠 0, . . . , 𝑠𝑚 , 𝑡 1, . . . , 𝑡𝑛−1 } arranged in the ascending order. Then the partition 𝑎 = 𝑢 0 < · · · < 𝑢𝑘 = 𝑏 is
438 good for 𝛾.
439 Proof. We need to show that 𝛾 | [𝑢𝑖 ,𝑢𝑖+1 ] is in 𝐶 1 ([𝑢𝑖 , 𝑢𝑖+1 ]). Note that there exists 𝑗 such that
440 [𝑢𝑖 , 𝑢𝑖+1 ] ⊆ [𝑠 𝑗 , 𝑠 𝑗+1 ] or [𝑢𝑖 , 𝑢𝑖+1 ] ⊆ [𝑡 𝑗 , 𝑡 𝑗+1 ]; this proves the assertion. 
441 6.5. Definition. Let 𝑈 be a domain and 𝑓 : 𝑈 −→ C. Let 𝛾 : [𝑎, 𝑏] −→ 𝑈 be a piecewise
442 differentiable path, with a good partition 𝑎 = 𝑡 0 < 𝑡 1 < · · · < 𝑡𝑛 = 𝑏. Define
∫ 𝑛−1 ∫ 𝑡𝑖+1
∑︁
𝑓 (𝑧)d𝑧 = 𝑓 (𝛾 (𝑡))𝛾 0 (𝑡)d𝑡 .
𝛾 𝑖=0 𝑡𝑖

443 This is independent of the choice of the good partition.

444 Exercises.
6.1 Show that the definition of 𝛾 𝑓 (𝑧)d𝑧 (Definition 6.5) does not depend on the choice of
∫
445

446 the partition.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 15

447 6.2 Let 𝛾 1 : [𝑎 1, 𝑏 1 ] −→ C and 𝛾 2 : [𝑎 2, 𝑏 2 ] −→ C be paths in C such that 𝛾 1 (𝑏 1 ) = 𝛾 2 (𝑎 2 ).

448 Define a new path 𝛾˜2 : [𝑏 1, 𝑏 2 − 𝑎 2 + 𝑏 1 ] −→ C by setting 𝛾˜2 (𝑡) = 𝛾 2 (𝑡 + 𝑎 2 − 𝑏 1 ). Note
449 that the images of 𝛾 2 and 𝛾˜2 are the same; this is an example of reparametrization of a
450 path, discussed in the next lecture. Define the concatenation of 𝛾 1 and 𝛾˜2 to be the path
451 𝛾 : [𝑎 1, 𝑏 2 − 𝑎 2 + 𝑏 1 ] −→ C
(
𝛾 1 (𝑡), 𝑡 ∈ [𝑎 1, 𝑏 1 ],
𝑡 ↦→
𝛾˜2 (𝑡), 𝑡 ∈ [𝑏 1, 𝑏 2 − 𝑎 2 + 𝑏 1 ].

452 Show that if 𝛾 1 and 𝛾 2 are piecewise-differentiable paths, then so is 𝛾. This is used in the
453 proof of Proposition 6.2 to a piecewise-differentiable path from 𝑐 to 𝜁 0 (while showing
454 that 𝐴 is open) and from 𝑐 to 𝑝 (while showing that 𝐴 is closed).
455 6.3 Let 𝑈 be a domain. A piecewise-linear path in 𝑈 is a continuous function 𝛾 : [𝑎, 𝑏] −→ 𝑈
456 such that there exists a partition 0 = 𝑡 0 < 𝑡 1 < . . . < 𝑡𝑛 = 𝑏 such that 𝛾 | [𝑡𝑖 ,𝑡𝑖+1 ] : 𝑡 ↦→
(𝑡−𝑡𝑖 )𝛾 (𝑡𝑖+1 )+(𝑡𝑖+1 −𝑡)𝛾 (𝑡𝑖 )
457
(𝑡𝑖+1 −𝑡𝑖 ) . Show that for every pair points in 𝑈 , there is a piecewise-linear
458 path joining them.

459 Lecture 7. Path integrals, II

460 7.1. Definition. Let 𝑈 be a domain and 𝛾 : [𝑎, 𝑏] −→ 𝑈 a path. A reparametrization of 𝛾 is a
461 path of the form 𝛾 ◦ 𝜏 : [𝑎0, 𝑏 0] −→ 𝑈 where 𝜏 : [𝑎0, 𝑏 0] −→ [𝑎, 𝑏] is a continuous piecewise
462 differentiable non-decreasing surjective function.
463 Note that Im(𝛾) = Im(𝛾 ◦ 𝜏). The next example shows that this is not sufficient.
464 7.2. Example. Let 𝛾 : [0, 1] −→ C, 𝑡 ↦→ 𝑒 2𝜋𝚤𝑡 . Then 𝛾 1 : [0, 2] −→ C, 𝑡 ↦→ 𝑒 𝜋𝚤𝑡 is a
465 reparametrization of 𝛾. To see this, let 𝜏1 : [0, 2] −→ [0, 1] be the map 𝑡 ↦→ 2𝑡 ; then 𝛾 1 = 𝛾 ◦ 𝜏1 .
466 On the other hand, 𝛾 2 : [0, 2] −→ C, 𝑡 ↦→ 𝑒 2𝜋𝚤𝑡 , is not a reparametrization of 𝛾. Intuitively, 𝛾 2
467 involves going round the circle twice, while 𝛾 involves going round only once.
7.3. Discussion (invariance under reparametrization). Let𝑈 be a domain and𝛾 : [𝑎, 𝑏] −→ 𝑈 a
path and 𝜏 : [𝑎0, 𝑏 0] −→ [𝑎, 𝑏] a continuous piecewise differentiable non-decreasing surjective
function. Write 𝛾˜ = 𝛾 ◦ 𝜏. Let 𝑎0 = 𝑠 0 < 𝑠 1 < · · · < 𝑠𝑚 = 𝑏 0 be a good partition for 𝜏 and
𝑎 = 𝑡 0 < 𝑡 1 < · · · < 𝑡𝑛 = 𝑏 be a good partition for 𝛾. Let 𝑢 0 < . . . < 𝑢𝑘 be the distinct
elements of {𝑠 0, . . . , 𝑠𝑚 } ∪ {𝜏 −1 (𝑡 0 ), . . . , 𝜏 −1 (𝑡𝑛 )}. Then 𝑎0 = 𝑢 0 < · · · < 𝑢𝑘 = 𝑏 0 is good for 𝛾.
˜ Let
𝑎 = 𝑣 0 < · · · < 𝑣𝑙 = 𝑏 be the distinct elements of {𝜏 (𝑢 0 ), . . . , 𝜏 (𝑢𝑘 )}; this is good for 𝛾. Thus,
∫ 𝑘−1 ∫ 𝑢𝑖+1
∑︁
𝑓 (𝑧)d𝑧 = 𝑓 (𝛾˜ (𝑠))𝛾˜0 (𝑠)d𝑠
𝛾˜ 𝑖=0 𝑢𝑖
𝑘−1 ∫ 𝑢𝑖+1
∑︁
= 𝑓 (𝛾 (𝜏 (𝑠)))𝛾 0 (𝜏 (𝑠))𝜏 0 (𝑠)d𝑠
𝑖=0 𝑢𝑖
𝑙−1 ∫ 𝑣𝑖+1
∑︁
= 𝑓 (𝛾 (𝑡))𝛾 0 (𝑡)d𝑡
𝑖=0 𝑣𝑖
∫
= 𝑓 (𝑧)d𝑧.
𝛾

468 Question: where did we use the hypothesis that 𝜏 is a non-decreasing function? 

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 16 MANOJ KUMMINI

469 7.4. Discussion (integration along the opposite path). Let 𝑈 be a domain and 𝛾 : [𝑎, 𝑏] −→ 𝑈
470 a piecewise differentiable path, with a good partition 𝑎 = 𝑠 0 < . . . < 𝑠𝑛 = 𝑏. For 0 ≤ 𝑖 ≤ 𝑛,
471 write 𝑡𝑖 = (𝑎 + 𝑏) − 𝑠𝑛−𝑖 . Then
−𝛾 | [𝑡𝑖 ,𝑡𝑖+1 ] = 𝛾 | [𝑠𝑛−𝑖−1,𝑠𝑛−𝑖 ] ◦ (𝑡 ↦→ (𝑎 + 𝑏) − 𝑡) .
Therefore 𝑎 = 𝑡 0 < . . . < 𝑡𝑛 = 𝑏 is a good partition for −𝛾.
∫ 𝑛−1 ∫ 𝑡𝑖+1
∑︁
𝑓 (𝑧)d𝑧 = 𝑓 ((−𝛾)(𝑡))(−𝛾) 0 (𝑡)d𝑡
−𝛾 𝑖=0 𝑡𝑖
𝑛−1 ∫ 𝑡𝑖+1
∑︁
= 𝑓 (𝛾 (𝑎 + 𝑏 − 𝑡)) (𝛾 0 (𝑎 + 𝑏 − 𝑡)) (−1)d𝑡
𝑖=0 𝑡𝑖
𝑛−1
∑︁ ∫ 𝑠𝑛−𝑖−1
= 𝑓 (𝛾 (𝑠))𝛾 0 (𝑠)d𝑠
𝑖=0 𝑠𝑛−𝑖
∫
=− 𝑓 (𝑧)d𝑧. 
𝛾

472 7.5. Proposition. Let 𝑈 be a domain and 𝑓 holomorphic on 𝑈 . If 𝑓 0 is identically zero on 𝑈 , then 𝑓 is a
473 constant function.
474 Proof. Let 𝑐 1, 𝑐 2 ∈ 𝑈 . We want to show that 𝑓 (𝑐 1 ) = 𝑓 (𝑐 2 ). Let 𝛾 : [𝑎, 𝑏] −→ 𝑈 a piecewise-
475 differentiable path with 𝛾 (𝑎) = 𝑐 1 and 𝛾 (𝑏) = 𝑐 2 . Let 𝑎 = 𝑡 0 < . . . < 𝑡𝑛 = 𝑏 be a good partition
476 for 𝛾. It suffices to show that 𝑓 (𝛾 (𝑡𝑖 )) = 𝑓 (𝛾 (𝑡𝑖+1 )). Replacing 𝑎 by 𝑡𝑖 and 𝑏 by 𝑡𝑖+1 , we may
477 assume that 𝛾 is differentiable on [𝑎, 𝑏].
478 The function
𝑔 : [𝑎, 𝑏] −→ C, 𝑡 ↦→ 𝑓 (𝛾 (𝑡))
479 is differentiable, with derivative 𝑔0 (𝑡) = 𝑓 0 (𝛾 (𝑡))𝛾 0 (𝑡) = 0. Hence 𝑓 (𝑐 2 ) = 𝑔(𝑏) = 𝑔(𝑎) =
480 𝑓 (𝑐 1 ). 
481 Exercises.
482 (1) Check that in Example 7.2, 𝛾 2 is not a reparametrization of 𝛾.
483 (2) Read Discussion 7.3 about reparametrization and understand where we used the hy-
484 pothesis that 𝜏 is a non-decreasing function.

485 Lecture 8. Absolute value of a path integral

486 8.1. Lemma. Let 𝑓 : [𝑎, 𝑏] −→ C be a continuous function and 𝑐 ∈ C. Then
∫ 𝑏 ∫ 𝑏
𝑐 𝑓 (𝑡)d𝑡 = 𝑐 𝑓 (𝑡)d𝑡 .
𝑎 𝑎
487 Proof. Write 𝑓 (𝑡) = 𝑢 (𝑡) + 𝚤𝑣 (𝑡) and 𝑐 = 𝛼 + 𝚤𝛽. Then both sides of the asserted equality are
488 equal to
∫ 𝑏 ∫ 𝑏
(𝛼𝑢 (𝑡) − 𝛽𝑣 (𝑡)) d𝑡 + 𝚤 (𝛼𝑣 (𝑡) + 𝛽𝑢 (𝑡)) d𝑡 . 
𝑎 𝑎
489 8.2. Corollary. Let 𝑓 : [𝑎, 𝑏] −→ C be a continuous function. Then
∫ 𝑏 ∫ 𝑏
𝑓 (𝑡)d𝑡 ≤ |𝑓 (𝑡)| d𝑡 .
𝑎 𝑎

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 17

Proof. Without loss of generality, we may assume that 𝑓 (𝑡)d𝑡 ≠ 0. Let 𝜃 be an argument of
∫𝑏
490
𝑎
𝑓 (𝑡)d𝑡. Then
∫𝑏
491
𝑎
∫ 𝑏  ∫ 𝑏 
−𝚤𝜃
𝑓 (𝑡)d𝑡 = < 𝑒 𝑓 (𝑡)d𝑡
𝑎 𝑎
∫ 𝑏  
= < 𝑒 𝑓 (𝑡) d𝑡
−𝚤𝜃
(by Lemma 8.1)
𝑎
∫ 𝑏
≤ |𝑓 (𝑡)| d𝑡 . 
𝑎
492 8.3. Definition. Let 𝛾 : [𝑎, 𝑏] −→ C be a piecewise-differentiable path and 𝑓 a C-valued func-
8
493 tion
∫ defined and continuous on Im(𝛾). The integral of 𝑓 with respect to arc length denoted by
494
𝛾
𝑓 |d𝑧| is
∫ 𝑏
𝑓 (𝛾 (𝑡))|𝛾 0 (𝑡)|d𝑡 .
𝑎
495 8.4. Proposition. Let 𝛾 : [𝑎, 𝑏] −→ C be a piecewise-differentiable path and 𝑓 a C-valued function
496 defined and continuous on Im(𝛾). Then
∫ ∫
𝑓 |d𝑧| = 𝑓 |d𝑧|.
−𝛾 𝛾

497 Proof. We repeat the argument from Discussion 7.4, with suitable changes.
∫ 𝑛−1 ∫ 𝑡𝑖+1
∑︁
𝑓 (𝑧)|d𝑧| = 𝑓 ((−𝛾)(𝑡))|(−𝛾) 0 (𝑡)|d𝑡
−𝛾 𝑖=0 𝑡𝑖
𝑛−1 ∫ 𝑡𝑖+1
∑︁
= 𝑓 (𝛾 (𝑎 + 𝑏 − 𝑡))|𝛾 0 (𝑎 + 𝑏 − 𝑡)|d𝑡
𝑖=0 𝑡𝑖
𝑛−1
∑︁ ∫ 𝑠𝑛−𝑖−1
= − 𝑓 (𝛾 (𝑠))𝛾 0 (𝑠)d𝑠
𝑖=0 𝑠𝑛−𝑖
∫
= 𝑓 (𝑧)|d𝑧|. 
𝛾

498 8.5. Proposition. Let 𝛾 : [𝑎, 𝑏] −→ C be a piecewise-differentiable path and 𝑓 a C-valued function
499 defined and continuous on Im(𝛾). Then
∫ ∫
𝑓 d𝑧 ≤ |𝑓 ||d𝑧|.
𝛾 𝛾

500 Proof. Use Corollary 8.2 to see that

∫ ∫ 𝑏 ∫ 𝑏 ∫
𝑓 d𝑧 = 0
𝑓 (𝛾 (𝑡))𝛾 (𝑡)d𝑡 ≤ 0
|𝑓 (𝛾 (𝑡))||𝛾 (𝑡)|d𝑡 = |𝑓 ||d𝑧|. 
𝛾 𝑎 𝑎 𝛾

8.6. Definition. Let 𝛾 be a piecewise-differentiable path. The arc length of 𝛾 is

∫
501
𝛾
|d𝑧|.
8Many textbooks, including Ahlfors, also use 𝑓 d𝑠 denote this, but we will avoid this usage, since sometimes
∫
𝛾
we use 𝑠 to denote a real or complex variable.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 18 MANOJ KUMMINI

502 8.7. Corollary. Let𝛾 : [𝑎, 𝑏] −→ C be a piecewise-differentiable path and 𝑓 a C-valued function defined
503 and continuous on Im(𝛾). Let 𝐶 ≥ max{|𝑓 (𝑧)| : 𝑧 ∈ Im(𝛾)}. Write 𝐿 for the arc length of 𝛾. Then
∫
𝑓 d𝑧 ≤ 𝐶𝐿
𝛾

504 Proof. Observe

∫ that if 𝑔 is a real-valued continuous function on Im(𝛾) taking non-negative real
505 values, then 𝛾 𝑔|d𝑧| is a non-negative real number. Now apply this observation with 𝑔 = 𝐶 − |𝑓 |
506 to see that ∫ ∫
𝑓 d𝑧 ≤ |𝑓 ||d𝑧| ≤ 𝐶𝐿. 
𝛾 𝛾

507 Exercises.
508 (1) Show that the arc length of a piecewise-linear path is the sum of the lengths of the line
509 segments in it. (See Exercise 6.3 in Lecture 6.)
510 (2) Let 𝛾 : [𝑎, 𝑏] −→ C be a piecewise-differentiable path. Then the arc length of 𝛾 is the
511 supremum of the set
( 𝑛−1 )
∑︁
|𝛾 (𝑡𝑖+1 ) − 𝛾 (𝑡𝑖 )| : 𝑛 ≥ 1, 𝑎 = 𝑡 0 < 𝑡 1 < . . . < 𝑡𝑛 = 𝑏 .
𝑖=0

512 (3) Let 𝛾 be a piecewise-differentiable path in C and 𝛾˜ a reparametrization. Show that the
513 arc lengths of 𝛾 and 𝛾˜ equal each other.

514 Lecture 9. Primitives

515 9.1. Definition. Let 𝑈 be a domain and 𝑓 : 𝑈 −→ C. A primitive 𝐹 of 𝑓 on 𝑈 is a (holomorphic)
516 function 𝐹 : 𝑈 −→ C such that 𝐹 0 = 𝑓 on 𝑈 .
517 Note that 𝑓 need not have a primitive on 𝑈 ; see Proposition 9.4.
518 9.2. Proposition. If 𝐹 1 and 𝐹 2 are primitives of a function 𝑓 on a domain 𝑈 , then 𝐹 1 − 𝐹 2 is a constant
519 function.
520 Proof. Note that (𝐹 1 − 𝐹 2 ) 0 = 𝐹 10 − 𝐹 20 = 0; now apply Proposition 7.5 to 𝐹 1 − 𝐹 2 . 
521 9.3. Example. Let𝑚 ∈ Z and 𝑓 (𝑧) = 𝑧𝑚 (wherever it can be defined). If𝑚 ≥ 0, then 𝑧𝑚+1 /(𝑚+1)
522 is a primitive of 𝑧𝑚 on C. If 𝑚 < −1, then 𝑧𝑚+1 /(𝑚 + 1) is a primitive of 𝑧𝑚 on C r {0}. Now
523 suppose 𝑚 = −1. If there is a branch of the logarithm on 𝑈 , then it is a primitive of 𝑓 (𝑧).
524 (Branches of the logarithm are holomorphic, with derivative 𝑧1 ; see exercise in Section 5.) Hence
1
525
𝑧 has a primitive on C r (−∞, 0], while, using the next proposition, one of the exercises will
526 show that it does not have a primitive on C r {0}. 
527 9.4. Proposition. Let𝑈 be a domain and 𝑓 : 𝑈 −→ C be a continuous function. Then the following are
528 equivalent:
529 (1) 𝑓 has a primitive on 𝑈 .
530 ∫ There exists a function 𝐹 : 𝑈 −→ C such that for every piecewise-differentiable path𝛾 : [𝑎, 𝑏] −→
(2)
531 𝑈 , 𝛾 𝑓 (𝑧)d𝑧 = 𝐹 (𝛾 (𝑏)) − 𝐹 (𝛾 (𝑎)).
(3) For every piecewise-differentiable closed path 𝛾 in 𝑈 , 𝛾 𝑓 (𝑧)d𝑧 = 0.
∫
532

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 19

Proof. (1) =⇒ (2): Let 𝐹 be a primitive of 𝑓 on 𝑈 . Then 𝛾 𝑓 (𝑧)d𝑧 = 𝛾 𝐹 0 (𝑧)d𝑧; we want to show
∫ ∫
533

534 that its value is 𝐹 (𝛾 (𝑏)) − 𝐹 (𝛾 (𝑎)). Let 𝑎 = 𝑡 0 < 𝑡 1 < · · · < 𝑡𝑛 = 𝑏 be a good partition for 𝛾. It
535 suffices to show that for every 𝑖
∫ 𝑡𝑖+1
𝐹 0 (𝛾 (𝑡))𝛾 0 (𝑡)d𝑡 = 𝐹 (𝛾 (𝑡𝑖+1 )) − 𝐹 (𝛾 (𝑡𝑖 )).
𝑡𝑖
536 Without loss of generality, we may assume that𝛾 is a differentiable path. Write 𝐺 = 𝐹 ◦𝛾. Then
537 𝐺 0 (𝑡) = 𝐹 0 (𝛾 (𝑡))𝛾 0 (𝑡) = 𝑓 (𝛾 (𝑡))𝛾 0 (𝑡) is a continuous function, so we see that
∫ 𝑏
𝐺 0 (𝑡)d𝑡 = 𝐺 (𝑏) − 𝐺 (𝑎)
𝑎
538 by evaluating its real and imaginary parts (which are continuous, and, hence the fundamental
539 theorem of calculus applies).
540 (2) =⇒ (1): We prove that 𝐹 is a primitive of 𝑓 on 𝑈 . Let 𝑐 ∈ 𝑈 . Let 𝜖 > 0. We want to show
541 that there exists 𝛿 > 0 such that for all ℎ ∈ C with |ℎ| < 𝛿,
𝐹 (𝑐 + ℎ) − 𝐹 (𝑐)
(9.5) − 𝑓 (𝑐) < 𝜖
ℎ
542 First, choose∫𝛿 such that 𝐵𝑐,𝛿 ⊆ 𝑈 . Then, for every ℎ with 𝑐 + ℎ ∈ 𝐵𝑐,𝛿 , we can evaluate 𝐹 (𝑐 +
543 ℎ) − 𝐹 (𝑐) as 𝜏 𝑓 (𝑧)d𝑧, where 𝜏 is the function
[0, 1] −→ C, 𝑡 ↦→ 𝑡 (𝑐 + ℎ) + (1 − 𝑡)𝑐.
(That is, we are going from 𝑐 to 𝑐 + ℎ along the line segment joining 𝑐 to 𝑐 + ℎ at a constant
speed.) Write 𝑓 (𝑧) = 𝑓 (𝑐) + 𝜙 (𝑧) on 𝐵𝑐,𝛿 . Using one of the exercises (or equation (3) of Ahlfors,
Chapter 4, Section 1.1 (‘Line integrals’)) we see that
∫ ∫ 1
𝜙 (𝑧)d𝑧 = 𝜙 (𝜏 (𝑡))𝜏 0 (𝑡)d𝑡
𝜏 0
∫ 1
= |ℎ| |𝜙 (𝑡 (𝑐 + ℎ) + (1 − 𝑡)𝑐)|d𝑡
0
544 Since 𝑓 is continuous, we may assume, possibly replacing 𝛿 by a smaller real number, that
545 |𝜙 (𝑧)| < 𝜖 for every 𝑧 ∈ 𝐵𝑐,𝛿 . Now
∫ ∫
𝐹 (𝑐 + ℎ) − 𝐹 (𝑐) 𝑓 (𝑧)d𝑧 𝑓 (𝑐) · ℎ + 𝜙 (𝑧)d𝑧
− 𝑓 (𝑐) = 𝜏 − 𝑓 (𝑐) = 𝜏
− 𝑓 (𝑐) < 𝜖,
ℎ ℎ ℎ
546 thus proving (9.5).
547 ★ mk: [Rewriting the above argument using integration w.r.t. arc length:] Since 𝑓 is
548 continuous, we may assume, possibly replacing 𝛿 by a smaller real number, that |𝜙 (𝑧)| < 𝜖 for
549 every 𝑧 ∈ 𝐵𝑐,𝛿 . By Corollary 8.7 ∫
𝜙 (𝑧)d𝑧 < 𝜖 |ℎ|.
𝜏
550 Now
∫ ∫
𝐹 (𝑐 + ℎ) − 𝐹 (𝑐) 𝜏
𝑓 (𝑧)d𝑧 𝑓 (𝑐) · ℎ + 𝜏
𝜙 (𝑧)d𝑧
− 𝑓 (𝑐) = − 𝑓 (𝑐) = − 𝑓 (𝑐) < 𝜖,
ℎ ℎ ℎ
551 thus proving (9.5).
552 (2) ⇐⇒ (3): Exercise. 

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 20 MANOJ KUMMINI

553 Exercises.
554 (1) Let 𝑓 : [𝑎, 𝑏] −→ C be a function. Show that
∫ 𝑏 ∫ 𝑏
𝑓 d𝑡 ≤ |𝑓 |d𝑡
𝑎 𝑎

555 (This is proved in equation (3) of Ahlfors, Chapter 4, Section 1.1 (‘Line integrals’).)
556 (2) Prove the assertion (2) ⇐⇒ (3) in Proposition 9.4.
(3) Let 𝑟 be a positive real number. Let 𝛾 : [0, 2𝜋] −→ C, 𝑡 ↦→ 𝑟𝑒𝚤𝑡 . Show that 𝛾 (1/𝑧)d𝑧 =
∫
557

558 2𝜋. On the other hand, if 𝛾 is a piecewise-differentiable∫closed path that avoids some
559 ray in C (i.e., {𝑟𝑒𝚤𝛼 | 𝑟 ∈ R, 𝑟 ≥ 0} for some fixed 𝛼) then 𝛾 (1/𝑧)d𝑧 = 0.

560 Lecture 10. Cauchy integral theorem, I

561 10.1. Theorem (Cauchy integral theorem for a rectangle). (Ahlfors, Chapter 4, Section 1.4, Theorem
562 2, p. 109) Let 𝑈 be a domain and 𝑓 a holomorphic function on 𝑈 . Let 𝑅 ⊆ 𝑈 be a rectangle. Then
∫
𝑓 (𝑧)d𝑧 = 0.
𝜕𝑅
563 Note that 𝜕𝑅 is the union of four line segments, parallel to the real and imaginary axes. It
564 is thought of as a closed curve in 𝑈 , starting from one corner, and going once along the line
565 segments.
566 Proof. Proof given in Ahlfors (due to Goursat). 
567 The following lemma should help clarify the estimation of |𝜂 (𝑅𝑛 )| in equation (16) and the
568 following paragraph on page. 111 of Ahlfors’ book. In the proof of (9.5), we estimated
∫
𝜙 (𝑧)d𝑧
𝜏
569 where 𝜏 is a linear path, i.e., a line segment parametrized by a linear function. We want to do
570 a similar for 𝜕𝑅𝑛 , which is a piecewise-linear path.
571 10.2. Lemma. Let𝑈 be a domain,𝑔 : 𝑈 −→ C a continuous function and𝛾 : [𝑎, 𝑏] −→ 𝑈 a piecewise-
572 linear path, i.e., a continuous function such that there exists a partition 𝑎 = 𝑡 0 < 𝑡 1 < . . . < 𝑡𝑛 = 𝑏 such
(𝑡−𝑡𝑖 )𝛾 (𝑡𝑖+1 )+(𝑡𝑖+1 −𝑡)𝛾 (𝑡𝑖 )
573 that 𝛾 | [𝑡𝑖 ,𝑡𝑖+1 ] : 𝑡 ↦→ (𝑡𝑖+1 −𝑡𝑖 ) . Then
∫ 𝑛−1 ∫
∑︁ |𝛾 (𝑡𝑖+1 ) − 𝛾 (𝑡𝑖 )| 𝑡𝑖+1
𝑔(𝑧)d𝑧 ≤ |𝑔(𝛾 (𝑡))|d𝑡 .
𝛾 𝑖=0
𝑡 𝑖+1 − 𝑡 𝑖 𝑡 𝑖

574 In particular, if 𝐶 ≥ |𝑔(𝑧)| for every 𝑧 ∈ Im(𝛾), then

∫ 𝑛−1
∑︁
𝑔(𝑧)d𝑧 ≤ 𝐶 |𝛾 (𝑡𝑖+1 ) − 𝛾 (𝑡𝑖 )|.
𝛾 𝑖=0

575 Proof. Since

∫ 𝑛−1 ∫
∑︁ 𝑡𝑖+1
𝑔(𝑧)d𝑧 = 𝑔(𝛾 (𝑡))𝛾 0 (𝑡)d𝑡 .
𝛾 𝑖=0 𝑡𝑖
576 it follows that ∫ 𝑛−1 ∫
∑︁ 𝑡𝑖+1
𝑔(𝑧)d𝑧 ≤ |𝑔(𝛾 (𝑡))||𝛾 0 (𝑡)|d𝑡 .
𝛾 𝑖=0 𝑡𝑖

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 21

577 Now note that

𝛾 (𝑡𝑖+1 ) − 𝛾 (𝑡)
𝛾 0 (𝑡) =
𝑡𝑖+1 − 𝑡𝑖
578 on [𝑡𝑖 , 𝑡𝑖+1 ], proving the first assertion. The second assertion follows immediately from the
579 first. 
580 10.3. Corollary. With notation as in Ahlfors’ book, |𝜂 (𝑅𝑛 )| ≤ 𝜖𝐿𝑛𝑑𝑛 .
Proof. Note that
∫ ∫ ∫ ∫
[𝑓 (𝑧) − 𝑓 (𝑧 ) − (𝑧 − 𝑧 ) 𝑓 (𝑧 )] d𝑧 =
∗ ∗ 0 ∗ ∗
𝑓 (𝑧)d𝑧 − 𝑓 (𝑧 ) d𝑧 − 𝑓 (𝑧 )
0 ∗
(𝑧 − 𝑧 ∗ )d𝑧.
𝜕𝑅𝑛 𝜕𝑅𝑛 𝜕𝑅𝑛 𝜕𝑅𝑛
= 𝜂 (𝑅𝑛 )
581 since 1 and (𝑧 − 𝑧 ∗ ) have primitives on C. Hence we want to estimate
∫
|𝜂 (𝑅𝑛 )| = [𝑓 (𝑧) − 𝑓 (𝑧 ∗ ) − (𝑧 − 𝑧 ∗ ) 𝑓 0 (𝑧 ∗ )] d𝑧 .
𝜕𝑅𝑛
582 Note that 𝑛 is large enough so that
|𝑓 (𝑧) − 𝑓 (𝑧 ∗ ) − (𝑧 − 𝑧 ∗ )𝑓 0 (𝑧 ∗ )| < 𝜖 |𝑧 − 𝑧 ∗ | < 𝜖𝑑𝑛 .
583 for all 𝑧 ∈ 𝜕𝑅𝑛 . Now apply Lemma 10.2. ★ mk: [Or, directly ] 𝜕𝑅𝑛 is a piecewise-linear path,
584 and its arc length is the length 𝐿𝑛 of the perimeter of 𝑅𝑛 (Exercise 1 of Lecture 8). Now apply
585 Corollary 8.7. 
586 Exercises.
587 (1) Show that in the proof of the theorem (with notation as in Ahlfors’ book),
Ù
𝑅𝑛 = 1.
𝑛

588 Lecture 11. Cauchy integral theorem, II

589 11.1. Theorem (Cauchy integral theorem for a disc). (Ahlfors, Chapter 4, Section 1.5, Theorem 4,
590 p.113) Let 𝑈 be an open disc, 𝑓 a holomorphic function on 𝑈 . Then 𝑓 has a primitive on 𝑈 . In particu-
591 lar, ∫
𝑓 (𝑧)d𝑧 = 0,
𝛾
592 for every piecewise-differentiable closed path 𝛾 in 𝑈 .
593 Proof. The second assertion follows from the first and Proposition 9.4; therefore we will prove
594 the first. Without loss∫ of generality, we may assume that 𝑈 is centred at 0 (Exercise). Define
595 𝐹 : 𝑈 −→ C by 𝜁 ↦→ 𝜎 𝑓 (𝑧)d𝑧, where 𝜎 is the piecewise-differentiable path from 0 to 𝜁 that
596 goes from 0 to (<(𝜁 ), 0) (the line segment parallel to the real axis) and from there to 𝜁 (the line
597 segment parallel to the imaginary axis).
598 We will show that 𝐹 is holomorphic on 𝑈 with 𝐹 0 = 𝑓 . Let 𝑐 ∈ 𝑈 . Let 𝜖 > 0. We want to show
599 that there exists 𝛿 > 0 such that for all ℎ ∈ C with |ℎ| < 𝛿,
𝐹 (𝑐 + ℎ) − 𝐹 (𝑐)
(11.2) − 𝑓 (𝑐) < 𝜖.
ℎ
There exists 𝛿 > 0 such that 𝐵𝑐,𝛿 ⊆ 𝑈 , that |𝑓 (𝑧) − 𝑓 (𝑐)| < 𝜖2 for every 𝑧 ∈ 𝐵𝑐,𝛿 , since 𝑓 is
continuous. Let ℎ ∈ 𝐵 0,𝛿 . Let 𝜎 (respectively, 𝜎1 ) be the piecewise-differentiable path from 0 to

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 22 MANOJ KUMMINI

𝑐 (respectively,𝑐 +ℎ) that goes from 0 to (<(𝑐), 0) (respectively, to (<(𝑐 +ℎ), 0)) and from there
to 𝑐 (respectively 𝑐 +ℎ). Let 𝜏 be the piecewise-differentiable path from 𝑐 to (<(𝑐 +ℎ), =(𝑐)) and
from there to𝑐+ℎ. Applying Theorem 10.1 to the rectangle with vertices (<(𝑐), 0), (<(𝑐+ℎ), 0),
(<(𝑐 + ℎ), =(𝑐)) and 𝑐 we see that
∫ ∫ ∫
𝐹 (𝑐 + ℎ) = 𝑓 (𝑧)d𝑧 = 𝑓 (𝑧)d𝑧 + 𝑓 (𝑧)d𝑧
𝜎1 𝜎 𝜏
∫
= 𝐹 (𝑐) + 𝑓 (𝑧)d𝑧.
𝜏
Write 𝜙 (𝑧) = 𝑓 (𝑧) − 𝑓 (𝑐) on 𝐵𝑐,𝛿 . Now,
∫ ∫ ∫
𝑓 (𝑧)d𝑧 = 𝑓 (𝑐)d𝑧 + 𝜙 (𝑧)d𝑧
𝜏 𝜏 𝜏
∫
= 𝑓 (𝑐) [(𝑐 + ℎ) − 𝑐] + 𝜙 (𝑧)d𝑧
𝜏
∫
= ℎ𝑓 (𝑐) + 𝜙 (𝑧)d𝑧.
𝜏
600 (We have used the fact constant functions have primitives on C.) Hence we can rewrite (11.2)
601 as
∫
𝜙 (𝑧)d𝑧
(11.3) 𝜏
< 𝜖.
ℎ
602 Let 𝜏1 (respectively 𝜏2 ) be the piecewise-differentiable path from 𝑐 to (<(𝑐 + ℎ), =(𝑐)) (re-
603 spectively, from (<(𝑐 + ℎ), =(𝑐)) to 𝑐 + ℎ). Then 𝜏 as the concatenation of 𝜏1 and 𝜏2 . Therefore
604 the arc length of 𝜏 is at most |<(ℎ)| + |=(ℎ)| < 2|ℎ|. Now apply Corollary 8.7 after noting that
605 |𝜙 (𝑧)| < 𝜖/2 on Im(𝛾) to obtain (11.3). 

606 Exercises.
607 (1) Show that in the proof of Theorem 11.1, we can assume that the centre of𝑈 is 0 as follows:
608 Let 𝑐 be the centre of 𝑈 . Let 𝜏 : 𝑈 −→ C be the function 𝑧 ↦→ 𝑧 − 𝑐. Let 𝑈∫1 = Im(𝜏).
609 Then 𝜏 maps 𝑈 homeomorphically to 𝑈 1 . Let 𝑓1 = 𝑓 ◦𝜏 −1 and 𝛾 1 = 𝜏 ◦𝛾. Then 𝛾 𝑓 (𝑧)d𝑧 =
∫
610 𝑓 (𝑧)d𝑧.
𝛾1 1
611 (2) Depending on the generality of Green’s theorem that you are familiar with, one can es-
612 tablish a version of Cauchy integral theorem, as follows. Let 𝛾 be a Jordan curve in C
613 (i.e., a closed piece-wise differentiable path that is injective, except at the end-points).
614 Let 𝑈 be a domain that contains 𝛾 and the open subset of C bounded by 𝛾. Let 𝑓 be a
615 holomorphic function on 𝑈 . Write 𝑧 = 𝑥 + 𝚤𝑦, 𝑓 = 𝑢 (𝑥, 𝑦) + 𝚤𝑣 (𝑥, 𝑦). We showed that if
616 𝑓 is holomorphic, then 𝑢 and 𝑣 are differentiable on 𝑈 .
617 (a) 𝑓 d𝑧 = 𝑢d𝑥 − 𝑣d𝑦 + 𝚤 (𝑣d𝑥 + 𝑢d𝑦). ∫
618 (b) Suppose that 𝑓 0 is continuous. Then 𝛾 𝑓 d𝑧 = 0.
619 (3) Let 𝑈 be a domain containing 𝐵 0,1 and 𝛾 : [0, 1] −→ C, 𝑡 ↦→ 𝑒 2𝜋𝚤𝑡 . Compute
1
∫
1
d𝑧
𝛾 𝑧− 2

620 as follows. (Note that neither is the integrand holomorphic on 𝑈 nor is 𝛾 centred at 21 ,
621 so earlier arguments do not apply immediately.)

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 23

622 (a) Let 0 < 𝑟  1 and 𝜎 : [0, 1] −→ C, 𝑡 ↦→ 21 + 𝑟𝑒 2𝜋𝚤𝑡 . Compute

1
∫
1
d𝑧
𝜎 𝑧− 2

623 (b) For 0 < 𝜖  1, define the following four points and paths in 𝑈 : 𝑝 ∈ Im 𝛾 with
624 <(𝑝) > 0 and =(𝑝) = 𝜖; 𝑞 ∈ Im 𝛾 with <(𝑞) > 0 and =(𝑞) = −𝜖; 𝑎 ∈ Im 𝜎 with
625 <(𝑝) > 21 and =(𝑝) = 𝜖; 𝑏 ∈ Im 𝜎 with <(𝑞) > 21 and =(𝑞) = −𝜖; 𝛾 1 from 𝑝 to 𝑞
626 counter-clockwise, following the same path as 𝛾; 𝛾 1 from 𝑎 to 𝑏 counter-clockwise,
627 following the same path as 𝜎; 𝜏1 from 𝑎 to 𝑝, parallel to the real axis; 𝜏2 from 𝑏 to
628 𝑞, parallel to the real axis. Let Γ be the closed piecewise differentiable path at 𝑝
629 obtained by concatenating 𝛾 1 , −𝜏2 , −𝜎1 and 𝜏1 . Show that
1
∫
1
d𝑧 = 0.
Γ 𝑧− 2

630 (Hint: Γ ⊆ 𝑈 r { 21 + 𝑟 | 𝑟 ∈ R, 𝑟 ≥ 0}, on which 𝑧−1 1 has a primitive.)

2
(c) Show that
1 1
∫ ∫
lim 1
d𝑧 = 1
d𝑧.
𝜖→0 𝛾 𝑧 − 𝛾 𝑧− 2
1 2
1 1
∫ ∫
lim d𝑧 = d𝑧.
𝜖→0 𝜎 𝑧 − 1 𝑧 − 1
2
1
∫ 2
𝜎
1 1
∫
lim 1
d𝑧 = lim d𝑧
𝜖→0 𝜏 𝑧 − 𝜖→0 𝜏 𝑧 − 1
1 2 2 2

631 (d) Conclude that

1 1
∫ ∫
d𝑧 = d𝑧.
𝛾 𝑧 − 21 𝜎 𝑧 − 21
1
632 (e) Generalize the result, after replacing by an arbitrary 𝑐 ∈ 𝐵 0,1 .
2

633 Lecture 12. Cauchy integral theorem, III

634 General background: We need to show that if 𝑔(𝑧) is holomorphic on 𝐵𝑐,𝑅 (𝑅 > 0) and 𝛾 is
635 the closed path [0, 1] −→ C, 𝑡 ↦→ 𝑐 + 𝑟𝑒 2𝜋𝚤𝑡 (with 0 < 𝑟 < 𝑅) then
1
∫
𝑔(𝑧)
d𝑧 = 𝑔(𝜁 ).
2𝜋𝚤 𝛾 𝑧 − 𝜁
636 for every 𝜁 ∈ 𝐵𝑐,𝑟 . One way to evaluate the integral (a la Lang or Rodrı́guez-Kra-Gilman) is to
637 observe that 𝛾 can be ‘continuously deformed’ to a closed path 𝛾 1 : [0, 1] −→ C, 𝑡 ↦→ 𝜁 + 𝜌𝑒 2𝜋𝚤𝑡
638 with a small 𝜌 (so that 𝛾 1 is inside 𝐵𝑐,𝑟 ), and therefore we may try to evaluate the integral on 𝛾 1 .
639 Another option (a la Ahlfors) is to look at the function
𝑔(𝑧) − 𝑔(𝜁 )
𝑓 (𝑧) =
𝑧 −𝜁
640 and observe that it is holomorphic on 𝐵𝑐,𝑅 except at 𝜁 , where it has the property that lim𝑧→𝜁 (𝑧 −
641 𝜁 )𝑓 (𝑧) = 0. We now strengthen CIT 11.1 to include such functions with this property. In fact,
642 we will see later that we can extend 𝑓 to a holomorphic function which is defined also at 𝜁 .

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 24 MANOJ KUMMINI

643 12.1. Theorem. Ahlfors, p. 113, Theorem 5 (the version with ‘mild singularities’.) Let 𝑈 be an open disc,
644 𝑈 0 an open subset of 𝑈 obtained by omitting finitely many points of 𝑈 , 𝑓 a holomorphic function on 𝑈 0
645 and 𝛾 a closed path in 𝑈 0. Assume that lim𝑧→𝜁 (𝑧 − 𝜁 ) 𝑓 (𝑧) = 0 for every 𝜁 ∈ 𝑈 r 𝑈 0. Then
∫
𝑓 (𝑧)d𝑧 = 0.
𝛾

Proof. Without loss of generality,𝑈 is centred at 0. Define 𝐹 : 𝑈 0 −→ C, 𝜁 ↦→ 𝜎 𝑓 (𝑧)d𝑧, where

∫
646
647 𝜎 be an rectilinear path in 𝑈 0 (a path consisting of finitely many segments, parallel to the real
648 and the imaginary axes) from 0 to 𝜁 . (This path needs to avoid the points in 𝑈 r 𝑈 0.) We need
649 to show that
650 (1) the value of 𝐹 (𝜁 ) does not depend on the choice of 𝜎, for every 𝜁 ∈ 𝑈 0
651 (2) 𝐹 is holomorphic with 𝐹 0 = 𝑓 .
652 To prove the first assertion, we will prove an analogous version of Theorem 10.1, in which some
653 points in the interior of the rectangle are omitted; see Theorem 12.2 below. The holomorphicity
654 of 𝐹 can be proved exactly as in the proof of Theorem 11.1, since for every 𝜁 ∈ 𝑈 0, there exists
655 𝛿 > 0 such that 𝐵𝜁 ,𝛿 ⊂ 𝑈 0. 
656 12.2. Theorem (Cauchy integral theorem for a rectangle, with ‘mild singularities’). (Ahlfors,
657 Chapter 4, Section 1.4, Theorem 3, p. 111) Let 𝑈 be a domain, 𝑈 0 an open subset of 𝑈 obtained by omit-
658 ting finitely many points of𝑈 . Let 𝑓 a holomorphic function on𝑈 0 such that lim𝑧→𝜁 (𝑧 − 𝜁 ) 𝑓 (𝑧) = 0 for
659 every 𝜁 ∈ 𝑈 r 𝑈 0. Let 𝑅 ⊆ 𝑈 be a rectangle, such that 𝜕𝑅 ⊆ 𝑈 0. Then
∫
𝑓 (𝑧)d𝑧 = 0.
𝜕𝑅
660 Proof. Proof given in Ahlfors, p. 112. After subdividing 𝑅, we may assume that 𝑅 contains exactly
661 one element of 𝑈 r 𝑈 0; call this element 𝜁 . Let 𝑅0 ⊆ 𝑅 be a square of size 2𝑎 (with sides parallel
662 to the axes) with centre 𝜁 . Then
∫ ∫
𝑓 (𝑧)d𝑧 = 𝑓 (𝑧)d𝑧.
𝜕𝑅 𝜕𝑅0
663 Let 𝜖 > 0. We may choose 𝑎 such that
|(𝑧 − 𝜁 )𝑓 (𝑧)| < 𝜖
664 for every 𝑧 ∈ 𝑅0 . Therefore for each 𝑧 ∈ 𝜕𝑅0 , |𝑧 − 𝜁 | > 𝑎 and so
𝜖
|𝑓 (𝑧)| < .
𝑎
665 The length of the perimeter of 𝑅0 is 8𝑎. Hence by Corollary 8.7
∫
𝜖
𝑓 (𝑧)d𝑧 < · 8𝑎 = 8𝜖.
𝜕𝑅0 𝑎
666 Therefore ∫
𝑓 (𝑧)d𝑧 = 0. 
𝜕𝑅
667 12.3. Remark. We will later see that in this situation, lim𝑧→𝜁 𝑓 (𝑧) exists for every 𝜁 ∈ 𝑈 0. Hence
668 we can extend the function to a holomorphic function on 𝑈 , by setting 𝑓 (𝜁 ) = lim𝑧→𝜁 𝑓 (𝑧) for
669 every 𝜁 ∈ 𝑈 0. The proof of this result will require some knowledge about the local behaviour
670 of holomorphic functions, for which we need to know this result. Otherwise, the argument
671 would be circular.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 25

672 Lecture 13. Index of a point

673 The following proposition generalizes Exercise 3 of Lecture 11.
674 13.1. Proposition. Let 𝛾 : [𝑎, 𝑏] −→ C a closed piecewise differentiable path. Let 𝜁 ∈ C r Im 𝛾. Then
675 there exists 𝑛(𝜁 , 𝛾) ∈ Z such that
d𝑧
∫
= 𝑛(𝜁 , 𝛾) · 2𝜋𝚤.
𝛾 𝑧 −𝜁

676 Proof. For 𝑠 ∈ [𝑎, 𝑏], write

𝛾 0 (𝑡)d𝑡
∫ 𝑠
ℎ(𝑠) = .
𝑎 𝛾 (𝑡) − 𝜁
677 This is a continuous function on [𝑎, 𝑏]. Since 𝛾 0 (𝑡) is continuous except on a finite subset of
678 [𝑎, 𝑏],
𝛾 0 (𝑠)
ℎ0 (𝑠) =
𝛾 (𝑠) − 𝜁
679 on the complement of that finite set. Therefore
𝛾 (𝑡) − 𝜁
ℎ 1 (𝑡) :=
𝑒 ℎ(𝑡)
680 is differentiable except on a finite subset of [𝑎, 𝑏]. Note that
𝛾 0 (𝑡) (𝛾 (𝑡) − 𝜁 )ℎ0 (𝑡)
ℎ01 (𝑡) := ℎ(𝑡) − = 0.
𝑒 𝑒 ℎ(𝑡)
681 Since ℎ 1 (𝑠) is continuous, it is constant, so
𝛾 (𝑡) − 𝜁
𝑒 ℎ(𝑡) =
𝛾 (𝑎) − 𝜁
682 for every 𝑡 ∈ [𝑎, 𝑏]. Since𝛾 (𝑎) = 𝛾 (𝑏), we conclude that 𝑒 ℎ(𝑎) = 𝑒 ℎ(𝑏) = 1. Therefore there exists
683 𝑛(𝜁 , 𝛾) ∈ Z such that
d𝑧
∫
= ℎ(𝑏) = 𝑛(𝜁 , 𝛾) · 2𝜋𝚤. 
𝛾 𝑧 −𝜁

684 13.2. Lemma. Let 𝛾 : [𝑎, 𝑏] −→ C be a closed piecewise differentiable path. Then the function C r
685 Im 𝛾 −→ Z, 𝜁 ↦→ 𝑛(𝜁 , 𝛾) is locally constant.
686 Proof. Let 𝜁 ∈ C r Im 𝛾. We want to show that there exists 𝛿 > 0 such that for every 𝜁 0 ∈ 𝐵𝜁 ,𝛿 ,
d𝑧 d𝑧
∫ ∫
(13.3) = 0
.
𝛾 𝑧 −𝜁 𝛾 𝑧 −𝜁
687 Let 𝛿 > 0 be such that 𝐵𝜁 ,𝛿 ∩ Im 𝛾 = ∅. Let 𝜁 0 ∈ 𝐵𝜁 ,𝛿 . Let 𝑓 (𝑧) : C r {𝜁 0 } −→ C be the function
𝑧 −𝜁
𝑓 (𝑧) = .
𝑧 − 𝜁0
688 Let 𝐿 be the line segment joining 𝜁 and 𝜁 0 and 𝑈 = C r 𝐿. Then 𝑓 (𝑈 ) ∩ (−∞, 0] = ∅, i.e., for
689 every 𝑧 ∈ 𝑈 , =(𝑓 (𝑧)) ≠ 0 or <(𝑓 (𝑧)) > 0 (Exercise). Hence we can define
𝑔 : C r 𝐿 −→ C, 𝑧 ↦→ Log(𝑓 (𝑧)).
690 Note that 𝑔 is holomorphic on 𝑈 and that
𝑓 0 (𝑧) 1 1
𝑔0 (𝑧) = = − .
𝑓 (𝑧) 𝑧 − 𝜁 𝑧 − 𝜁 0

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 26 MANOJ KUMMINI

691 Since 𝑈 is a domain and 𝛾 is a closed path in 𝑈 , it follows that

1 1
∫  
− d𝑧 = 0,
𝛾 𝑧 −𝜁 𝑧 − 𝜁0
692 establishing (13.3). 
693 The following corollary recovers Exercise 3 of Lecture 11.
694 13.4. Corollary. Let𝑐 ∈ C, 𝑟 > 0 and𝛾 : [0, 1] −→ C the path 𝑡 ↦→ 𝑐 +𝑟𝑒 2𝜋𝚤𝑡 . Then for every 𝜁 ∈ 𝐵𝑐,𝑟 ,
695 𝑛(𝜁 , 𝛾) = 1.
696 Proof. Note that 𝑛(𝑐, 𝛾) = 1. Let 𝑈 = {𝜁 ∈ 𝐵𝑐,𝑟 | 𝑛(𝜁 , 𝛾) = 1}. By the lemma, 𝑈 and 𝐵𝑐,𝑟 r 𝑈 are
697 open. Since 𝐵𝑐,𝑟 is connected and 𝑈 non-empty, 𝐵𝑐,𝑟 = 𝑈 . 

698 Exercises.
699 (1) Show that 𝑓 (𝑈 ) ∩ (−∞, 0] = ∅ in the proof of Lemma 13.2.
700 (2) Let 𝛾 : [𝑎, 𝑏] −→ C be a piecewise-differentiable path. Show that 𝑛(𝜁 , 𝛾) = 0 for all
701 𝜁 ∈ C with |𝜁 |  0.
702 (3) Ahlfors, Chapter 4, Section 2.1 (‘Index of a point ...’), Exercise 3 (p. 118). (proof of the
703 Jordan curve theorem).
704 (4) Here is another proof of Lemma 13.2. Let 𝜁 , 𝜖 be such that 𝐵𝜁 ,𝜖 ⊆ C r Im(𝛾). Let 0 <
705 𝛿  𝜖 and 𝜁 0 ∈ 𝐵𝜁 ,𝛿 . For every 𝑧 ∈ Im(𝛾),
1 1 𝜁 − 𝜁0 𝛿
− 0
= 0
< .
𝑧 −𝜁 𝑧 −𝜁 (𝑧 − 𝜁 )(𝑧 − 𝜁 ) 𝜖 (𝜖 − 𝛿)
706 Let 𝐿 be the arc length of 𝛾. Then
1 1
∫  
𝛿𝐿
− d𝑧 < < 2𝜋 .
𝛾 𝑧 −𝜁 𝑧 − 𝜁0 𝜖 (𝜖 − 𝛿)

707 Lecture 14. Cauchy integral formula

708 14.1. Theorem (Cauchy integral formula for a circular path). (Ahlfors, Chapter 4, (22), p.119, for
709 circles.) Let 𝑈 be a domain, 𝑐 ∈ 𝑈 , 𝑟 > 0 such that 𝐵𝑐,𝑟 ⊆ 𝑈 . Let 𝜁 1, . . . , 𝜁𝑚 ∈ 𝑈 and 𝑈 0 = 𝑈 r
710 {𝜁 1, . . . 𝑧𝑚 }. Let 𝑓 be a holomorphic function on𝑈 0 such that lim𝑧→𝜁𝑖 (𝑧 − 𝜁𝑖 ) 𝑓 (𝑧) = 0 for every 1 ≤ 𝑖 ≤
711 𝑚. Let 𝛾 be the circular path on the boundary of 𝐵𝑐,𝑟 . Then for all 𝜁 ∈ 𝐵𝑐,𝑟 ∩ 𝑈 0,
1
∫
𝑓 (𝑧)d𝑧
𝑓 (𝜁 ) = .
2𝜋𝚤 𝛾 𝑧 − 𝜁
712 Proof. Let
𝑓 (𝑧) − 𝑓 (𝜁 )
𝑔(𝑧) = .
𝑧 −𝜁
713 Then𝑔 is holomorphic on𝑈 r{𝜁 , 𝜁 1, . . . , 𝜁𝑚 }, and lim𝑧→𝑎 (𝑧−𝑎)𝑔(𝑧) = 0 for every𝑎 ∈ {𝜁 , 𝜁 1, . . . , 𝜁𝑚 }.
714 Therefore ∫ ∫
𝑓 (𝑧)d𝑧 𝑓 (𝜁 )d𝑧
= .
𝛾 𝑧 −𝜁 𝛾 𝑧 −𝜁

715 Now apply Corollary 13.4. 

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 27

716 14.2. Lemma. Let𝛾 be a piecewise-differentiable closed path in C. Let 𝑔 : Im(𝛾) −→ C be a continuous
717 function. For positive integers 𝑛, define 𝐹𝑛 : C r Im(𝛾) −→ C by
∫
𝑔(𝜁 )
𝑧 ↦→ d𝜁 .
𝛾 (𝜁 − 𝑧)
𝑛

718 Then for each 𝑛 ≥ 1, 𝐹𝑛 is holomorphic on C r Im(𝛾) with 𝐹𝑛0 = 𝑛𝐹𝑛+1 .

719 Proof. We will prove that 𝐹 1 is holomorphic with 𝐹 10 = 𝐹 2 . For the rest, read the proof of Ahlfors,
720 Chapter 4, Section 2.3 (‘Higher derivatives’), Lemma 3.
721 Let 𝑧 0, 𝜖 be such that 𝐵𝑧0,𝜖 ⊆ C r Im(𝛾). Let 0 < 𝛿  𝜖 and 𝑧 ∈ 𝐵𝑧0,𝛿 .
722 Step 1: lim𝑧→𝑧0 𝐹 1 (𝑧) = 𝐹 1 (𝑧 0 ). Proof: Note that
∫
𝑔(𝜁 )
(14.3) 𝐹 1 (𝑧) − 𝐹 1 (𝑧 0 ) = (𝑧 − 𝑧 0 ) d𝜁 .
𝛾 (𝜁 − 𝑧)(𝜁 − 𝑧 0 )

723 For every 𝜁 ∈ Im(𝛾), |(𝜁 − 𝑧 0 )| > 𝜖 and |(𝜁 − 𝑧)| > 𝜖 − 𝛿. By Proposition 8.5
∫
𝛿
|𝐹 1 (𝑧) − 𝐹 1 (𝑧 0 )| < |𝑔(𝜁 )||d𝜁 |.
𝜖 (𝜖 − 𝛿) 𝛾
724 Therefore lim𝑧→𝑧0 |𝐹 1 (𝑧) − 𝐹 1 (𝑧 0 )| = 0.
725 Step 2: 𝐹 10 (𝑧 0 ) = 𝐹 2 (𝑧 0 ). Proof: Consider the function
∫
𝑔(𝜁 )
𝐺 (𝑧) = d𝜁
𝛾 (𝜁 − 𝑧)(𝜁 − 𝑧 0 )

726 on C r Im(𝛾). Applying the previous step with 𝑔(𝜁 )/(𝜁 − 𝑧 0 ) replacing 𝑔(𝑧), we see that
727 lim𝑧→𝑧0 𝐺 (𝑧) = 𝐺 (𝑧 0 ) = 𝐹 2 (𝑧 0 ). Now by (14.3)
𝐹 1 (𝑧) − 𝐹 1 (𝑧 0 )
lim = lim 𝐺 (𝑧) = 𝐹 2 (𝑧 0 ). 
𝑧→𝑧 0 𝑧 − 𝑧0 𝑧→𝑧 0

728 14.4. Corollary. With notation as in Theorem 14.1,

∫
(𝑛) 𝑛! 𝑓 (𝑧)d𝑧
𝑓 (𝜁 ) = .
2𝜋𝚤 𝛾 (𝑧 − 𝜁 )𝑛+1
729 In particular, 𝑓 is infinitely complex-differentiable on 𝑈 0.
730 Proof. Write
∫
𝑓 (𝑧)d𝑧
𝐹𝑛+1 = .
𝛾 (𝑧 − 𝜁 )𝑛+1
1 (𝑛)
731 By Theorem 14.1, 𝐹 1 = 𝑓 . By Lemma 14.2, 𝐹𝑛+1 = 𝑛! 𝑓 . 

732 Exercises.
733 (1) Complete the proof of Lemma 14.2. (Ahlfors, Chapter 4, Section 2.3 (‘Higher deriva-
734 tives’), Lemma 3 (p. 121))
735 (2) Let 𝑈 be a domain and 𝛾 a piecewise-differentiable path in 𝑈 . If 𝑓𝑛 is a sequence of
736 continuous functions on 𝑈 converging uniformly to 𝑓 , then
∫ ∫
lim 𝑓𝑛 (𝑧)d𝑧 = 𝑓 (𝑧)d𝑧.
𝑛 𝛾 𝛾

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 28 MANOJ KUMMINI

If converges uniformly to 𝑓 , then

Í
737 𝑛 𝑓𝑛
∑︁ ∫ ∫
𝑓𝑛 (𝑧)d𝑧 = 𝑓 (𝑧)d𝑧.
𝑛 𝛾 𝛾

738 (3) Let 𝑟 ∈ R and 𝑓𝑟 : R2 −→ R, (𝑥, 𝑦) ↦→ (𝑥 2 + 𝑦 2 − 1)𝑟 . Show that for each 𝑟 , 𝑓𝑟 is a real-
739 analytic function. 𝑓𝑟 (𝑝) does not depend on 𝑟 if 𝑝 ∈ 𝜕𝐵 0,1 , but depends on 𝑟 if 𝑝 ∈ 𝐵 0,1 .
740 This is in contrast with the behaviour of holomorphic functions (on a domain containing
741 𝐵 0,1 ).

742 Lecture 15. Holomorphic functions are analytic.

743 In this lecture, we will prove that holomorphic functions are analytic.
744 15.1. Lemma. Let 𝑈 be a domain, 𝑐 ∈ 𝑈 and 𝑓 holomorphic on 𝑈 0 := 𝑈 r {𝑐}. Then the following are
745 equivalent:
746 (1) lim 𝑓 (𝑧) exists (in C).
𝑧→𝑐
747 (2) lim (𝑧 − 𝑐)𝑓 (𝑧) = 0.
𝑧→𝑐
748 (3) there exists a holomorphic function 𝑓˜ on 𝑈 such that 𝑓˜|𝑈 0 = 𝑓 ;
749 Morever, in this situation, 𝑓˜ is uniquely determined by 𝑓 .
750 Proof. (1) =⇒ (2): lim (𝑧 − 𝑐) 𝑓 (𝑧) = lim (𝑧 − 𝑐) lim 𝑓 (𝑧) = 0.
𝑧→𝑐 𝑧→𝑐 𝑧→𝑐
751 (2) =⇒ (3): Let 𝑟 > 0 be such that 𝐵𝑐,𝑟 ⊆ 𝑈 . Let 𝛾 : [0, 1] −→ 𝑈 be the path 𝑡 ↦→ 𝑐 + 𝑒 2𝜋𝚤𝑡 .
752 Define 𝑓˜ : 𝑈 −→ C by (
𝑓 (𝜁 ), if 𝜁 ∈ 𝑈 0,
˜
𝑓 (𝜁 ) = 𝑓 (𝑧)d𝑧
, if 𝜁 = 𝑐.
∫
𝛾 𝑧−𝑐
753 We need to show that 𝑓˜ is holomorphic on 𝑈 ; for which it suffices to check that it is differ-
754 entiable at 𝑐. We may therefore restrict our attention to 𝐵𝑐,𝑟 . Using Cauchy integral formula
755 (Theorem 14.1) for 𝑈 0 ∩ 𝐵𝑐,𝑟 , we can rewrite 𝑓˜ on 𝐵𝑐,𝑟 as
∫
˜ 𝑓 (𝑧)d𝑧
𝑓 (𝜁 ) =
𝛾 𝑧 −𝜁

756 Now apply Lemma 14.2, 𝑓˜ is holomorphic on 𝐵𝑐,𝑟 .

757 (3) =⇒ (1): lim = 𝑓˜(𝑐).
𝑧→𝑐
758 Proving uniqueness is left as an exercise. 
759 15.2. Definition. Let 𝑈 be a domain, 𝑐 ∈ 𝑈 and 𝑓 holomorphic on 𝑈 0 := 𝑈 r {𝑐}. We say that 𝑐
760 is a removable singularity of 𝑓 if the equivalent conditions of the previous lemma are satisfied.
761 15.3. Theorem. Let 𝑈 be a domain, 𝑐 ∈ 𝑈 and 𝑓 a holomorphic on 𝑈 . Let 𝑛 ∈ N. Then there exists a
762 holomorphic function 𝑓𝑛 (𝑧) on 𝑈 such that
𝑛−1 (𝑘)
∑︁ 𝑓 (𝑐)
𝑓 (𝜁 ) = (𝜁 − 𝑐)𝑘 + (𝜁 − 𝑐)𝑛 𝑓𝑛 (𝜁 )
𝑘!
𝑘=0

763 on 𝑈 . Let 𝛾 be the circular path around the boundary of 𝐵𝑐,𝑅 where 𝑅 > 0 is such that 𝐵𝑐,𝑅 ⊆ 𝑈 . Then
1
∫
𝑓 (𝑧)d𝑧
𝑓𝑛 (𝜁 ) =
2𝜋𝚤 𝛾 (𝑧 − 𝑐)𝑛 (𝑧 − 𝜁 )

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 29

764 on 𝐵𝑐,𝑅 .
𝑓 (𝑧)−𝑓 (𝑐)
765 Proof. (Ahlfors, Chapter 4, Section 3.1, pp. 124ff.) The function 𝑧−𝑐 (on 𝑈 r {𝑐}) has a
766 removable singularity at 𝑧 = 𝑐, so there exists a holomorphic function 𝑓1 (𝑧) on 𝑈 such that
𝑓 (𝑧)−𝑓 (𝑐)
767 𝑓1 (𝑧) = 𝑧−𝑐 on 𝑈 r {𝑐}. Repeating this argument for 𝑓1 , and by induction, we see that
768 for each positive integer 𝑘, there exists a holomorphic function 𝑓𝑘+1 on 𝑈 such that 𝑓𝑘+1 (𝑧) =
𝑓𝑘 (𝑧)−𝑓𝑘 (𝑐)
769
𝑧−𝑐 on 𝑈 r {𝑐}. Putting this together, we get the following:
𝑓 (𝑧) = 𝑓 (𝑐) + (𝑧 − 𝑐) 𝑓1 (𝑧)
= 𝑓 (𝑐) + (𝑧 − 𝑐) 𝑓1 (𝑐) + (𝑧 − 𝑐) 2 𝑓2 (𝑧)
= 𝑓 (𝑐) + (𝑧 − 𝑐) 𝑓1 (𝑐) + (𝑧 − 𝑐) 2 𝑓2 (𝑐) + · · · + (𝑧 − 𝑐)𝑛−1 𝑓𝑛−1 (𝑐) + (𝑧 − 𝑐)𝑛 𝑓𝑛 (𝑧)

770 Note that 𝑓 (𝑘) (𝑧) is the 𝑘-th order derivative of (𝑧 − 𝑐)𝑘 𝑓𝑘 (𝑧), so 𝑓 (𝑘) (𝑐) = 𝑘!𝑓𝑘 (𝑐). This proves
771 the first assertion.
772 Now note that on 𝐵𝑐,𝑅
1
∫
𝑓𝑛 (𝑧)
𝑓𝑛 (𝜁 ) = d𝑧
2𝜋𝚤 𝛾 𝑧 − 𝜁
(15.4)
1 1
∫ 𝑛−1 (𝑘) ∫
𝑓 (𝑧) ∑︁ 𝑓 (𝑐)
= d𝑧 − d𝑧.
2𝜋𝚤 𝛾 (𝑧 − 𝑐) (𝑧 − 𝜁 )
𝑛 2𝜋𝚤𝑘! 𝛾 (𝑧 − 𝑐)𝑛−𝑘 (𝑧 − 𝜁 )
𝑘=0

773 Let 𝜁 1, 𝜁 2 ∈ 𝐵𝑐,𝑅 and

1
∫
𝐺𝑚 (𝜁 1, 𝜁 2 ) := d𝑧.
𝛾 (𝑧 − 𝜁 1 )𝑚 (𝑧 − 𝜁2)
774 as a function of 𝜁 1 , with 𝜁 2 fixed. We first show that 𝐺 1 (𝜁 1, 𝜁 2 ) = 0. First assume that 𝜁 1 ≠ 𝜁 2 .
775 Then
1 1
∫
d𝑧 = (𝑛(𝜁 1, 𝛾) − 𝑛(𝜁 2, 𝛾)) = 0.
𝛾 (𝑧 − 𝜁 1 )(𝑧 − 𝜁 2 ) 𝜁1 − 𝜁2
1
776 Now assume that 𝜁 1 = 𝜁 2 . Since (𝑧−𝜁 1 ) 2has a primitive on C r {0}, we see that

1
∫
2
d𝑧 = 0.
𝛾 (𝑧 − 𝜁 1 )
1
777 By Lemma 14.2 applied to the function (𝑧−𝜁 2)
, we see that 𝐺𝑚 for 𝑚 ≥ 2 are successive deriva-
778 tives of 𝐺 1 (thought of as a function of 𝜁 1 ), so 𝐺𝑚 = 0 for each 𝑚 ≥ 1. Therefore in (15.4), we
779 obtain
1
∫
d𝑧 = 0
𝛾 (𝑧 − 𝑐)
𝑛−𝑘 (𝑧 − 𝜁 )

780 for each 𝑘 = 0, . . . , 𝑛 − 1, thus completing the proof of the theorem. 

781 15.5. Corollary. Let 𝑈 be a domain and 𝑓 holomorphic on 𝑈 . Let 𝑐, 𝑅 be such that 𝐵𝑐,𝑅 ⊆ 𝑈 . Then
∑︁ 𝑓 (𝑘) (𝑐)
𝑓 (𝜁 ) = (𝜁 − 𝑐)𝑘
𝑘!
𝑘∈N

782 on 𝐵𝑐,𝑅 . In particular, every holomorphic function on 𝑈 is analytic on 𝑈 .

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 30 MANOJ KUMMINI

783 Proof. Let 𝛾 be the circular path on the boundary of 𝐵𝑐,𝑅 . To prove the assertion, it suffices to
784 show that for every 𝜖 > 0, there exists 𝑁 such that for every 𝑛 > 𝑁 ,
𝑛 1
∫
𝑓 (𝑧)d𝑧
(𝜁 − 𝑐) < 𝜖.
2𝜋𝚤 𝛾 (𝑧 − 𝑐)𝑛 (𝑧 − 𝜁 )
785 Let 𝑀 = sup{𝑓 (𝑧) | 𝑧 ∈ Im(𝛾)}. Then
𝑛 1
∫
𝑓 (𝑧)d𝑧 𝑀 |𝜁 − 𝑐 |𝑛
(𝜁 − 𝑐) ≤ ≤ .
2𝜋𝚤 𝛾 (𝑧 − 𝑐)𝑛 (𝑧 − 𝜁 ) 𝑅𝑛−1 (𝑅 − |𝜁 − 𝑐 |)
786 The assertion now follows, since |𝜁 − 𝑐 | < 𝑅. 
787 Exercises.
788 (1) Prove the uniqueness of 𝑓˜ in Lemma 15.1.
789 (2) Show that 𝑘th order derivative of (𝑧 − 𝑐)𝑘 𝑔(𝑧) at 𝑧 = 𝑐 is 𝑘!𝑔(𝑐).
790 (3) With fixed 𝜁 2 ∈ 𝐵𝑐,𝑅 , use an appropriate result to conclude that 𝐺 1 (𝜁 1, 𝜁 2 ) is a holomor-
791 phic function of 𝜁 1 ∈ 𝐵𝑐,𝑅 , the proof of Theorem 15.3. Then show that 𝐺 1 (𝜁 2, 𝜁 0 ) = 0 by
792 taking a limit.
793 (4) Let 𝑐 be a removable singularity of 𝑓 (which is defined on 𝑈 r {𝑐} for some open neigh-
794 bourhood 𝑈 of 𝑐). Show that there exists 𝑚 ∈ N and a holomorphic function 𝑓1 on 𝑈
795 such that 𝑓 (𝑧) = (𝑧 − 𝑐)𝑚 𝑓1 (𝑧) such that 𝑓1 (𝑐) ≠ 0.
796 (5) Let 𝑈 be a domain and 𝑓 a holomorphic function on 𝑈 . Then the zeros of 𝑓 are isolated.
797 Show that 𝑓 has only finitely many zeroes in any compact subset of 𝑈 . If 𝑐 ∈ 𝑈 is a zero,
798 then there exists a unique positive integer 𝑚 such that 𝑓 (𝑧) = (𝑧 −𝑐)𝑚 𝑓1 (𝑧) on 𝑈 , where
799 𝑓1 is holomorphic on 𝑈 and 𝑓1 (𝑐) ≠ 0. It is called the order (or multiplicity) of the zero at
800 𝑐.

801 Lecture 16. Morera’s theorem, Liouville’s theorem

802 ∫16.1. Corollary (Morera’s theorem). Let 𝑈 be a domain and 𝑓 : 𝑈 −→ C a continuous function. If
803
𝛾
𝑓 (𝑧)d𝑧 = 0 for every closed piecewise-differentiable path 𝛾 in 𝑈 , then 𝑓 is analytic.

804 Proof. By Corollary 15.5, it suffices to show that 𝑓 is holomorphic. By Proposition 9.4, 𝑓 has
805 a primitive 𝐹 on 𝑈 . Since 𝐹 is holomorphic, it is infinitely complex-differentiable by Corol-
806 lary 14.4; in particular, 𝑓 = 𝐹 0 is holomorphic. 
807 16.2. Proposition (Liouville’s theorem). Every bounded entire function is constant.
808 Proof. Let 𝑓 : C −→ C be a bounded holomorphic function. Let 𝑀 ∈ R be such that |𝑓 (𝑧)| < 𝑀
809 for every 𝑧 ∈ C. Let 𝜁 ∈ C and 𝑟 > 0. By Corollary 14.4,
1
∫
𝑓 (𝑧)
0
𝑓 (𝜁 ) = d𝑧
2𝜋𝚤 𝜕𝐵𝜁 ,𝑟 (𝑧 − 𝜁 ) 2
810 Hence |𝑓 0 (𝜁 )| ≤ 𝑀𝑟 −1 , so 𝑓 0 = 0 on C. Now apply 7.5. 
811 16.3. Theorem (Fundamental theorem of algebra). C is an algebraically closed field.
812 We need to show that complex polynomials of positive degree have a zero. First we prove a
813 lemma about such polynomials.
814 16.4. Lemma. Let 𝑓 ∈ C[𝑋 ] be a polynomial of positive degree. Then for every positive real number 𝑀,
815 there exists 𝑅 > 0 such that |𝑓 (𝑧)| > 𝑀 for every 𝑧 with |𝑧| > 𝑅.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 31

Proof. Write 𝑓 (𝑋 ) = 𝑑𝑖=0 𝑎𝑖 𝑋 𝑖 , with 𝑑 > 0 and 𝑎𝑑 ≠ 0. For every 𝑧 ∈ C, |𝑓 (𝑧)| ≥ |𝑎𝑑 ||𝑧|𝑑 −
Í
816

𝑖=0 |𝑎𝑖 ||𝑧| . (Use 𝑎 𝑧 = 𝑓 (𝑧)− 𝑖=0 𝑎𝑖 𝑧 .) The assertion now follows from Exercise 1 below. 
Í𝑑−1 𝑖 𝑑 𝑑 Í𝑑−1 𝑖
817

818 Proof of Theorem 16.3. Let 𝑓 ∈ C[𝑋 ] be a polynomial of positive degree. We want to show that
819 there exists 𝑐 ∈ C such that 𝑓 (𝑐) = 0. By way of contradiction, assume that this is false. Hence
1
820 𝑔(𝑧) = 𝑓 (𝑧) is an entire function. We now claim that 𝑔 is bounded. Assume the claim. Then 𝑔
821 and, therefore, 𝑓 are constant functions by Proposition 16.2, contradicting the hypothesis that
822 𝑓 has positive degree.
823 To prove the claim, assume, on the contrary, that for each positive integer 𝑛, there exists
824 𝑐𝑛 ∈ C such that |𝑔(𝑐𝑛 )| > 𝑛. Then |𝑓 (𝑐𝑛 )| < 𝑐1𝑛 . If the sequence 𝑐𝑛 is bounded (i.e., con-
825 tained in a compact subset of C), then it would have a convergent subsequence 𝑐𝑛𝑚 , 𝑚 ≥ 1. (We
826 implicitly assume that the function 𝑚 ↦→ 𝑛𝑚 is an increasing function.) Then 𝑓 (lim𝑚 𝑐𝑛𝑚 ) =
827 lim𝑚 𝑓 (𝑐𝑛𝑚 ) = 0, a contradiction. Hence for every positive real number 𝑅, there exists 𝑛 such
828 that |𝑐𝑛 | > 𝑅. Now use Lemma 16.4 to obtain a contradiction. 

829 Exercises.
(1) Let 𝑑𝑖=0 𝑏𝑖 𝑋 𝑖 ∈ R[𝑋 ] with 𝑏𝑑 > 0. Then for every positive real number 𝑀, there exists
Í
830
831 𝑅 > 0 such that 𝑔(𝑥) > 𝑀 for every 𝑥 ∈ R with 𝑥 > 𝑅.
832 (2) Show that Lemma 16.4 does not hold for entire functions in general, by looking at the
833 exponential function.

834 Lecture 17. Isolated singularities

835 Let 𝑈 be a domain, 𝑐 ∈ 𝑈 and 𝑓 a holomorphic function on 𝑈 r {𝑐}. We say that 𝑐 is an
836 isolated singularity of 𝑓 . Recall that an isolated singularity 𝑐 is said to be a removable singularity
837 if lim (𝑧 − 𝑐)𝑓 (𝑧) = 0 (Definition 15.2).
𝑧→𝑐
838 (We do not rule out the situation that 𝑓 is defined or is differentiable at 𝑐.)
839 17.1. Definition. An isolated singularity 𝑐 is said to be a pole of 𝑓 if it is not a removable singu-
840 larity of 𝑓 and it is a removable singularity of 1/𝑓 .
841 17.2. Example. 𝑧𝑚 with 𝑚 < 0 has a pole at 0.
842 17.3. Remark. With notation as above, let 𝑈 be a neighbourhood of 𝑐 such that 𝑓 is defined and
843 holomorphic on 𝑈 r {𝑐}. Since the zeros of a holomorphic function are isolated (use Corol-
844 lary 15.5 and Proposition 4.6), we may assume, by replacing 𝑈 by a smaller neighbourhood if
845 necessary, that 𝑓 (𝜁 ) ≠ 0 for each 𝜁 ∈ 𝑈 , 𝜁 ≠ 𝑐. Hence we can talk of 𝑓1 in 𝑈 r {𝑐} and consider
846 whether 𝑐 is a removable singularity or not.
847 17.4. Proposition. Let 𝑐 be a pole of 𝑓 . Then
1
848 (1) lim = 0.
𝑧→𝑐 𝑓 (𝑧)
849 (2) There exists a positive integer 𝑁 and a neighbourhood 𝑉 of 𝑐 in 𝑈 such that
∞
∑︁
𝑓 (𝑧) = 𝑎𝑘 (𝑧 − 𝑐)𝑘
𝑘=−𝑁

850 on 𝑉 r {𝑐}.
851 (3) For every positive real number 𝑀, there exists 𝛿 > 0 such that |𝑓 (𝑧)| > 𝑀 for every 𝜁 ∈ 𝐵𝑐,𝛿 .

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 32 MANOJ KUMMINI

852 Proof. (1): By Lemma 15.1, there exists a neighbourhood 𝑉 of 𝑐 and a holomorphic function 𝑔 on
1 1
853 𝑉 such that 𝑔(𝑧) = 𝑓 (𝑧) on 𝑉 r {𝑐}. If 𝑔(𝑐) ≠ 0, then lim 𝑓 (𝑧) = 𝑔(𝑐) , which would imply that 𝑓
𝑧→𝑐
854 has a removable singularity at 𝑐. Hence lim 𝑔(𝑧) = 𝑔(𝑐) = 0.
𝑧→𝑐
855
1
(2): Since 𝑐 is a removable singularity of 𝑓1 , we can write 𝑓 (𝑧) = (𝑧 − 𝑐) 𝑁 𝑓1 (𝑧) for some 𝑁 ∈ N
856 and a holomorphic function 𝑓1 (𝑧) in a neighbourhood of 𝑐 with 𝑓1 (𝑐) ≠ 0. (See Exercise 4 in
1 1
857 Lecture 15.) Since lim 𝑓 (𝑧) = 0, 𝑁 > 0. Note that 𝑓1 (𝑧) is holomorphic in a neighbourhood of 𝑐,
𝑧→𝑐
858 so it admits a convergent power series expansion around 𝑐.
859 (3): Exercise. 
860 The next two propositions characterise zeros and poles of holomorphic functions by looking
861 at the limit of |𝑧 − 𝑐 |𝑛 |𝑓 (𝑧)| for various 𝑛. Their proofs are left as exercises.
862 17.5. Proposition. Let 𝑈 be a domain, 𝑐 ∈ 𝑈 and 𝑓 a non-zero holomorphic function on 𝑈 r {𝑐}. Then
863 the following are equivalent:
864 (1) 𝑓 can be extended to a holomorphic function 𝑓˜ on 𝑈 with 𝑓˜(𝑐) = 0;
865 (2) lim 𝑓 (𝑧) = 0;
𝑧→𝑐
866 (3) there exist 𝑚, 𝑛 ∈ Z with 𝑚 < 0 and 𝑛 < 0 such that lim |𝑧 − 𝑐 |𝑚 |𝑓 (𝑧)| = 0 and lim |𝑧 −
𝑧→𝑐 𝑧→𝑐
867 𝑐 |𝑛 |𝑓 (𝑧)| = ∞
868 (4) there exists 𝑁 ∈ Z with 𝑁 < 0 such that lim |𝑧 − 𝑐 |𝑚 |𝑓 (𝑧)| = 0 for every 𝑚 > 𝑁 and lim |𝑧 −
𝑧→𝑐 𝑧→𝑐
869 𝑐 |𝑛 |𝑓 (𝑧)| = ∞ for every 𝑛 < 𝑁 .
870 17.6. Proposition. Let 𝑈 be a domain, 𝑐 ∈ 𝑈 and 𝑓 a holomorphic function on 𝑈 r {𝑐}. Then the
871 following are equivalent:
872 (1) 𝑐 is pole of 𝑓 ;
873 (2) there exist 𝑚, 𝑛 ∈ N such that lim |𝑧 − 𝑐 |𝑚 |𝑓 (𝑧)| = 0 and lim |𝑧 − 𝑐 |𝑛 |𝑓 (𝑧)| = ∞
𝑧→𝑐 𝑧→𝑐
874 (3) there exists 𝑁 ∈ Z with 𝑁 > 0 such that lim |𝑧 − 𝑐 |𝑚 |𝑓 (𝑧)| = 0 for every 𝑚 > 𝑁 and lim |𝑧 −
𝑧→𝑐 𝑧→𝑐
875 𝑐 |𝑛 |𝑓 (𝑧)| = ∞ for every 𝑛 < 𝑁 .
876 17.7. Definition. Let 𝑈 be a domain, 𝑐 ∈ 𝑈 and 𝑓 a holomorphic function on 𝑈 r {𝑐}. Say that
877 𝑐 is an essential singularity of 𝑓 if it is not a removable singularity or a pole of 𝑓 .
878 17.8. Proposition. 𝑐 is an essential singularity of 𝑓 if and only if lim |𝑧 − 𝑐 |𝑛 |𝑓 (𝑧)| does not exist for
𝑧→𝑐
879 any 𝑛 ∈ Z.
880 Proof. ‘If’: by definition. ‘Only if’: By way of contradiction, assume that lim |𝑧−𝑐 | 𝑁 |𝑓 (𝑧)| exists.
𝑧→𝑐
881 Then 𝑐 is a removable singularity of (𝑧 − 𝑐) 𝑁 𝑓 (𝑧). If 𝑁 ≤ 0, then 𝑐 is a removable singularity
882 of 𝑓 . If 𝑁 > 0, then 𝑐 is a pole of 𝑓 . 
883 17.9. Proposition. Let𝑈 be a domain, 𝑐 ∈ 𝑈 and 𝑓 a holomorphic function on𝑈 r {𝑐}. Suppose that 𝑐
884 is an essential singularity of 𝑓 . Then for every 𝐴 ∈ C, every 𝜖 > 0 and every 𝛿 > 0, there exists 𝜁 ∈ 𝐵𝑐,𝜖
885 such that |𝑓 (𝜁 ) − 𝐴| < 𝛿.
886 Proof. By way of contradiction, let 𝐴 ∈ C, 𝜖 > 0, 𝛿 > 0 be such that for every 𝜁 ∈ 𝐵𝑐,𝜖 r {𝜁 },
887 |𝑓 (𝜁 ) − 𝐴| ≥ 𝛿. Then for every 𝑛 < 0, lim |𝑧 − 𝑐 |𝑛 |𝑓 (𝑧) − 𝐴| = ∞, so 𝑐 is not an essential
𝑧→𝑐
888 singularity of 𝑓 (𝑧) − 𝐴. Then there exists 𝑚 > 0 such that lim |𝑧 − 𝑐 |𝑚 |𝑓 (𝑧) − 𝐴| = 0. Note that
𝑧→𝑐
lim |𝑧 − 𝑐 | |𝑓 (𝑧)| ≤ lim |𝑧 − 𝑐 | |𝑓 (𝑧) − 𝐴| + lim |𝑧 − 𝑐 |𝑚 |𝐴| = 0
𝑚 𝑚
𝑧→𝑐 𝑧→𝑐 𝑧→𝑐
889 so 𝑐 is not an essential singularity of 𝑓 . 

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 33

890 17.10. Definition. Let𝑈 be a domain. By a meromorphic function on𝑈 , we mean a a holomorphic
891 function 𝑓 : 𝑈 0 −→ C where 𝑈 0 ⊆ 𝑈 , and points in 𝑈 r 𝑈 0 are isolated in 𝑈 and are poles of 𝑓 .
892 17.11. Example. If 𝑓 is a holomorphic function on a domain𝑈 then its zeros are isolated, by Corol-
893 lary 15.5 and Proposition 4.6; hence 𝑓1 is meromorphic on 𝑈 . E.g., 𝑧1 is a meromorphic function
894 on C. Every rational function is meromorphic on every domain in C.
895 Exercises.
896 (1) Let 𝑐 be a pole of 𝑓 . For every positive real number 𝑀, there exists 𝛿 > 0 such that
897 |𝑓 (𝑧)| > 𝑀 for every 𝜁 ∈ 𝐵𝑐,𝛿 .
898 (2) Prove Proposition 17.5.
899 (3) Prove Proposition 17.6.
900 (4) Let 𝑈 be a domain, 𝑐 ∈ 𝑈 and 𝑓 holomorphic on 𝑈 r {𝑐}. Suppose that 𝑐 is a pole of 𝑓 .
901 Write
∞
∑︁
𝑓 (𝑧) = 𝑎𝑘 (𝑧 − 𝑐)𝑘
𝑘=−𝑁

902 in a punctured neighbourhood 𝑉 r {𝑐} of 𝑐, with 𝑁 > 0 and 𝑎 −𝑁 ≠ 0. Let 𝑟 > 0 be such
903 that 𝐵𝑐,𝑟 ⊆ 𝑉 . Let 𝛾 : [0, 1] −→ 𝑉 be the path 𝑡 ↦→ 𝑐 + 𝑟𝑒 2𝜋𝚤𝑡 . Then
∫
𝑓 d𝑧 = 2𝜋𝚤𝑎 −1 .
𝛾

904 We say that 𝑎 −1 is the residue of 𝑓 at 𝑐, and denote it by Res 𝑓 (𝑐).

905 (5) Let 𝑈 be a disc, 𝑓 a meromorphic function on 𝑈 and 𝛾 a piecewise-differentiable closed
906 path in 𝑈 . Let {𝜁 𝑗 } be the poles of 𝑓 . Assume that 𝛾 does not pass through 𝜁 𝑗 for any 𝑗.
907 Show that 𝑛(𝜁 𝑗 , 𝛾) = 0 except for finitely many 𝑗 and that
1
∫ ∑︁
𝑓 d𝑧 = 𝑛(𝜁 𝑗 , 𝛾) Res 𝑓 (𝜁 𝑗 ).
2𝜋𝚤 𝛾 𝑗

(6) Let 𝑈 be a domain and M (𝑈 ) be the set of meromorphic functions on 𝑈 . For 𝑓 , 𝑔 ∈

M (𝑈 ) and 𝑐 ∈ 𝑈 , let

(𝑓 + 𝑔)(𝑐) = lim (𝑓 (𝑧) + 𝑔(𝑧))

𝑧→𝑐
(𝑓 𝑔)(𝑐) = lim (𝑓 (𝑧)𝑔(𝑧))
𝑧→𝑐

908 Show that 𝑐 is a pole of 𝑓 +𝑔 if and only if 𝑐 is a pole of 𝑓 or of 𝑔. State and prove a similar
909 characterization of poles of 𝑓 𝑔.
910 (7) (Not relevant for this course.) Let 𝑈 be a domain. Then M (𝑈 ) is the field of fractions of
911 A (𝑈 ).
912 (8) Let 𝑁 be an integer and suppose that 𝑐 is a removable singularity of (𝑧 − 𝑐) 𝑁 𝑓 (𝑧). If
913 𝑁 ≤ 0, then 𝑐 is a removable singularity of 𝑓 (𝑧). If 𝑁 > 0, then 𝑐 is a pole of 𝑓 (𝑧) of
914 order ≤ 𝑁 .
1
915 (9) 𝑒 𝑧 has an essential singularity at 0.
916 (10) Let 𝑓 (𝑧), 𝑔(𝑧) be a holomorphic functions defined in a neighbourhood of 𝑐 ∈ C. Sup-
917 pose that 𝑓 (𝑧) has an essential singularity at 𝑐. Show that 𝑓 (𝑧)𝑔(𝑧) has an essential
918 singularity at 𝑐.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 34 MANOJ KUMMINI

919 Lecture 18. Local mapping

920 18.1. Example. Consider the holomorphic function 𝑓 (𝑧) = 𝑧𝑚 , 𝑚 > 0, on C. It has a zero at
921 𝑧 = 0, of order 𝑚. Note that for every 𝑏 ∈ C, there exist 𝑚 solutions (counted with multiplicity)
922 for the equation 𝑓 (𝑧) = 𝑏. 
923 We now show that every holomorphic function exhibits the same behaviour locally. Here is
924 the result:
925 18.2. Proposition. Let 𝑈 be a domain and 𝑓 a non-constant holomorphic function on 𝑈 . Let 𝜁 ∈ 𝑈 .
926 Write 𝑎 = 𝑓 (𝜁 ). Let𝑚 be the order of the zero of 𝑓 (𝑧) − 𝑎 at 𝑧 = 𝜁 . Then for every 0 < 𝜖  1, there exists
927 𝛿 > 0 such that for every 𝑏 ∈ 𝐵𝑎,𝛿 , there exists 𝑚 solutions in 𝐵𝜁 ,𝜖 to the equation 𝑓 (𝑧) = 𝑏.
928 As an immediate corollary, we get the following:
929 18.3. Corollary. Let 𝑈 be a domain and 𝑓 a non-constant holomorphic function on 𝑈 . Then 𝑓 (𝑈 ) is an
930 open subset of C. In other words, every non-constant holomorphic function is an open map.
931 Proof. With notation as in Proposition 18.2, 𝐵 𝑓 (𝜁 ),𝛿 ⊆ 𝑓 (𝐵𝜁 ,𝜖 ) for every 𝜁 ∈ 𝑈 and every 0 <
932 𝜖  1. Since the open discs 𝐵𝜁 ,𝜖 form a basis for the topology of 𝑈 , 𝑓 is open. 
933 Before proving Proposition 18.2, we need to develop a method to count zeros. In the above
934 example with 𝑓 = 𝑧𝑚 , we note that if 𝛾 is a closed piecewise-differentiable curve in C not pass-
935 ing through 0, then
d𝑧
∫ 0 ∫
𝑓 (𝑧)
d𝑧 = 𝑚 = 2𝜋𝚤 · 𝑛(0, 𝛾) · 𝑚.
𝛾 𝑓 (𝑧) 𝛾 𝑧
1
∫ 𝑓 0 (𝑧)
936 In particular, if𝛾 is a circular path such that 0 is in the bounded region, then𝑚 = 2𝜋𝚤 𝛾 𝑓 (𝑧)
d𝑧.
937 18.4. Proposition. Let 𝑈 be a disc and 𝑓 holomorphic on 𝑈 . Let {𝜁 𝑗 } be the distinct zeros of 𝑓 ; denote
938 the order of 𝜁 𝑗 by 𝑚 𝑗 . Let 𝛾 be a closed piecewise differentiable path in 𝑈 , not passing through any of the
939 𝜁 𝑗 . Then
940 (1) 𝑛(𝜁 𝑗 , 𝛾) = 0 for all but finitely many 𝑗.
(2)
1
∫ 0
𝑓 (𝑧) ∑︁
d𝑧 = 𝑛(𝜁 𝑗 , 𝛾) · 𝑚 𝑗 .
2𝜋𝚤 𝛾 𝑓 (𝑧) 𝑗

941 Proof. We will use Exercise 5 from Lecture 15 in this proof. (1): Since Im 𝛾 is compact, there
942 exists an open subset 𝑉 of 𝑈 such that its closure 𝑉 in C contains Im 𝛾 and is a subset of 𝑈 .
943 Note that 𝑓 has only finitely many zeros in 𝑉 . For any 𝜁 ∈ C r𝑉 , 𝑛(𝜁 , 𝛾) = 0, since the function
1
944
𝑧−𝜁 is holomorphic on 𝑉 .
945 (2): By (1), we
Î may assume that the (distinct) zeros of 𝑓 are 𝜁 1, . . . , 𝜁𝑟 , with orders 𝑚 1, . . . 𝑚𝑟 .
946 Write 𝑓 (𝑧) = 𝑟𝑗=1 (𝑧 − 𝜁 𝑗 )𝑚 𝑗 𝑔(𝑧) where 𝑔 is holomorphic on an open set 𝑉 containing Im 𝛾 and
947 does not have any zeros. Then
𝑓 0 (𝑧) ∑︁ 𝑚 𝑗 𝑔0 (𝑧)
𝑟
= + .
𝑓 (𝑧) 𝑗=1
𝑧 − 𝜁 𝑗 𝑔(𝑧)
𝑔 0 (𝑧)
948 The assertion now follows from noting that 𝑔(𝑧) is holomorphic on 𝑉 . 
949 Proof of Proposition 18.2. Let 0 < 𝜖  1. Then 𝐵𝜁 ,𝜖 ⊆ 𝑈 . Let 𝛾 : [0, 1] −→ 𝑈 , 𝑡 ↦→ 𝜁 + 𝜖𝑒 2𝜋𝚤𝑡 . Let
950 Γ = 𝑓 ◦ 𝛾. Moreover, since 𝜖  1, we may assume that 𝜁 is the only solution to 𝑓 (𝑧) = 𝑎 in 𝐵𝜁 ,𝜖 ;
951 in particular, 𝑎 ∉ Im Γ. Let 𝛿 > 0 be such that 𝐵𝑎,𝛿 ∩ Im Γ = ∅.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 35

Let 𝑏 ∈ 𝐵𝑎,𝛿 . Let {𝜁 𝑗 } be the distinct zeros of 𝑓 (𝑧) − 𝑏 in 𝐵𝜁 ,𝜖 , with 𝑚 𝑗 the order of 𝜁 𝑗 . Then
1 𝑓 0 (𝑧) 1 d𝑤
∑︁ ∫ ∫
𝑛(𝜁 𝑗 , 𝛾) · 𝑚 𝑗 = d𝑧 = = 𝑛(𝑏, Γ);
𝑗
2𝜋𝚤 𝛾 𝑓 (𝑧) − 𝑏 2𝜋𝚤 Γ 𝑤 − 𝑏
1 𝑓 0 (𝑧) 1 d𝑤
∫ ∫
𝑚= d𝑧 = = 𝑛(𝑎, Γ).
2𝜋𝚤 𝛾 𝑓 (𝑧) − 𝑎 2𝜋𝚤 Γ 𝑤 − 𝑎
952 In both rows, the first equality is by Proposition 18.4 and the second by the substitution 𝑤 =
953 𝑓 (𝑧). Observe that 𝑛(𝑎, Γ) = 𝑛(𝑏, Γ) since 𝑎 and 𝑏 belong to
Í the same connected component of
954 C r Im(Γ). Now note that 𝑛(𝜁 𝑗 , 𝛾) = 1 for each 𝜁 𝑗 . Hence 𝑗 𝑚 𝑗 = 𝑚. 

955 Exercises.
956 (1)

957 Lecture 19. Maximum principle, definite integrals etc.

958 In this short lecture, we tie various loose ends.
959 19.1. Proposition. Let 𝑈 be a domain and 𝑓 a non-constant holomorphic function on 𝑈 .
960 (1) There does not exist 𝜁 ∈ 𝑈 such that |𝑓 (𝜁 )| = sup{|𝑓 (𝑧)| : 𝑧 ∈ 𝑈 }
961 (2) Assume that𝑈 is bounded and that 𝑓 can be extended to a continuous function 𝑓˜ on𝑈 . There exists
962 𝜁 ∈ 𝜕𝑈 such that | 𝑓˜(𝜁 )| = sup{|𝑓 (𝑧)| : 𝑧 ∈ 𝑈 }.
963 Proof. (1): Let 𝜁 ∈ 𝑈 . For every 0 < 𝜖  1, there exists 𝛿 such that 𝐵 𝑓 (𝜁 ),𝛿 ⊆ 𝑓 (𝐵𝜁 ,𝜖 ), by
964 Proposition 18.2. Now note that there exists 𝑏 ∈ 𝐵 𝑓 (𝜁 ),𝛿 such that |𝑏 | > |𝑓 (𝜁 )|.
965 (2): Since 𝑈 is compact, there exists 𝜁 ∈ 𝑈 such that | 𝑓˜(𝜁 )| = sup{| 𝑓˜(𝑧)| : 𝑧 ∈ 𝑈 } =
966 sup{|𝑓 (𝑧)| : 𝑧 ∈ 𝑈 }. 
967 Here is a generalization of Proposition 18.4 to meromorphic functions.
968 19.2. Proposition. Let𝑈 be a disc and 𝑓 meromorphic on𝑈 . Let 𝑎𝑖 be the distinct zeros of 𝑓 , with orders
969 𝑙𝑖 , respectively; let 𝑏 𝑗 be the distinct poles of 𝑓 , with orders 𝑚 𝑗 , respectively. Let 𝛾 be a closed piecewise
970 differentiable path in 𝑈 , not passing through any of the 𝑎𝑖 and any of the 𝑏 𝑗 . Then
1
∫ 0
𝑓 (𝑧) ∑︁ ∑︁
d𝑧 = 𝑙𝑖 · 𝑛(𝑎𝑖 , 𝛾) − 𝑚 𝑗 · 𝑛(𝑏 𝑗 , 𝛾)
2𝜋𝚤 𝛾 𝑓 (𝑧) 𝑖 𝑗

971 Proof. As in the proof of Proposition 18.4, we may assume that the number of zeroes and the
972 number of poles are finite. Hence we may write
Ö Ö
𝑓 (𝑧) = (𝑧 − 𝑎𝑖 )𝑙𝑖 (𝑧 − 𝑏𝑏 ) −𝑚 𝑗 𝑔(𝑧)
𝑖 𝑗
finite finite

973 in some open subset 𝑉 containing Im 𝛾, where 𝑔(𝑧) is holomorphic on 𝑉 and does not have any
974 zeros. Hence
𝑓 0 (𝑧) ∑︁ 𝑙𝑖 ∑︁ 𝑚𝑖 𝑔0 (𝑧)
= − + ,
𝑓 (𝑧) 𝑖
𝑧 − 𝑎𝑖 𝑗
𝑧 − 𝑏 𝑗 𝑔(𝑧)
finite finite
975 from which the assertion follows. 
976 We now look at an example of evaluating definite real integrals using complex integration.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 36 MANOJ KUMMINI

977 19.3. Example. Integrate

d𝜃
∫ 𝜋

0 𝑎 + cos 𝜃
978 where 𝑎 > 1 is a real number. Write 𝑏 for its value. Note that
∫ 2𝜋
d𝜃
2𝑏 =
0 𝑎 + cos 𝜃
Write 𝑧 = 𝑒𝚤𝜃 . Then d𝜃 = −𝚤 d𝑧 1 1
𝑧 and cos 𝜃 = 2 𝑧 + 𝑧 . Let 𝛾 : [0, 2𝜋] −→ C, 𝜃 ↦→ 𝑒
2𝜋𝚤𝜃 . Then


979

d𝑧
∫
2𝑏 = −𝚤 2
𝛾 𝑧 + 2𝑎𝑧 + 1
1
√ √
980 The meromorphic function 𝑧 2 +2𝑎𝑧+1 has two poles 𝛼 = −𝑎 + 𝑎 2 − 1 and 𝛽 = −𝑎 − 𝑎 2 − 1. Since
981 |𝛼 | < 1, 𝑛(𝛼, 𝛾) = 1; Since |𝛽 | > 1, 𝑛(𝛼, 𝛾) = 0; Note that
1 1 1 1
 
= − .
𝑧 2 + 2𝑎𝑧 + 1 𝛼 − 𝛽 𝑧 − 𝛼 𝑧 − 𝛽
982 Hence
(−𝚤)(2𝜋𝚤) 𝜋
2𝑏 = , i.e., 𝑏 = √ . 
𝛼 −𝛽 𝑎2 − 1
983 Exercises.
984 (1)

985 Lecture 20. Conformality

986 20.1. Definition. Let 𝑛 ≥ 2 be an integer, 𝑈 ⊆ R𝑛 an open subset and 𝑝 ∈ 𝑈 . A function
987 𝑓 : 𝑈 −→ R𝑛 is said to be conformal at 𝑝 if it is differentiable at 𝑝 and it preserves angles and
988 orientation at 𝑝. We say that 𝑓 is conformal on 𝑈 if it is conformal at 𝑝 for every 𝑝 ∈ 𝑈 .
989 What does this mean? Let 𝑒 1, . . . , 𝑒𝑛 denote the standard basis for R𝑛 , and let 𝑥 1, . . . , 𝑥𝑛 be the
990 coordinates of R𝑛 with respect to this basis. Write 𝑓 = (𝑓1, . . . , 𝑓𝑛 ), with respect to this basis.
991 Let 1 ≤ 𝑖 ≤ 𝑛. Consider the curve 𝛾 : (−𝜖, 𝜖) −→ 𝑈 , 𝑡 ↦→ 𝑝 + 𝑡𝑒𝑖 . Since 𝑓 is differentiable at 𝑝,
992 the composite curve 𝑓 𝛾 is differentiable at 0, with derivative
 𝜕𝑓1 (𝑝) 
 𝜕𝑥𝑖 
 𝜕𝑓2 
 𝜕𝑥 (𝑝) 
 𝑖 
 .. 
 . 
 𝜕𝑓𝑛 
 (𝑝) 
 𝜕𝑥𝑖 
993 This is the 𝑖th column of the jacobian matrix 𝐽 of 𝑓 at 𝑝.
994 The jacobian of 𝑓 at 𝑝 gives the map d𝑓 : Ω(𝑈 )𝑝 −→ Ω(R𝑛 ) 𝑓 (𝑝) , when these are identified
995 with R𝑛 . Here, Ω(−) is the cotangent bundle, and Ω(−)𝑞 the cotangent space at 𝑞.
996 Saying that 𝑓 preserves angles at 𝑝 is same as saying that 𝐽 preserves angles, i.e.,
𝑣 ·𝑤 𝐽 𝑣 · 𝐽𝑤
=
|𝑣 ||𝑤 | |𝐽 𝑣 ||𝐽𝑤 |
997 for every non-zero 𝑣, 𝑤 ∈ R . We now have the following:
𝑛

998 20.2. Proposition. Let 𝐽 be an 𝑛 ×𝑛 real matrix. Then 𝐽 preserves angles if and only if there exist 𝜆 > 0
999 and an orthogonal matrix 𝐴 such that 𝐽 = 𝜆𝐴.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 37

1000 Proof. ‘Only if’: Note that 𝐽𝑒𝑖 and 𝐽𝑒 𝑗 are orthogonal to each other if 𝑖 ≠ 𝑗, so 𝐽𝑒𝑖 · 𝐽𝑒 𝑗 = 0 for
1001 𝑖 ≠ 𝑗. Write 𝜆𝑖 = |𝐽𝑒𝑖 |, 1 ≤ 𝑖 ≤ 𝑛. Note that 𝜆𝑖 > 0 for each 𝑖. Let 𝐴 be the matrix whose
1002 𝑖th column is 𝐽𝑒𝜆𝑖𝑖 . Since the columns of 𝐽 are orthogonal to each other, it follows that 𝐴 is an
1003 orthogonal matrix. Therefore |𝐴𝑣 | = |𝐴𝑡 𝑣 | = |𝑣 | for every 𝑣 ∈ R𝑛 .
1004 Now consider the linear transformation 𝐴𝑡 𝐽 . For every 𝑣, 𝑤 ∈ R𝑛 ,
𝐴𝑡 𝐽 𝑣 · 𝐴𝑡 𝐽𝑤 𝐽 𝑣 · 𝐴𝐴𝑡 𝐽𝑤 𝐽 𝑣 · 𝐽𝑤 𝑣 ·𝑤
= = = ,
|𝐴 𝐽 𝑣 ||𝐴 𝐽𝑤 |
𝑡 𝑡 |𝐽 𝑣 ||𝐽𝑤 | |𝐽 𝑣 ||𝐽𝑤 | |𝑣 ||𝑤 |
1005 i.e., 𝐴𝑡 𝐽 preserves angles. Note that 𝐴𝑡 𝐽𝑒𝑖 = 𝜆𝑖 𝐴𝑡 𝐴𝑒𝑖 = 𝜆𝑖 𝑒𝑖 for each 𝑖, i.e., 𝐴𝑡 𝐽 is a diagonal
1006 matrix (with respect to the basis 𝑒𝑖 ). Hence it must be a multiple of 𝐼𝑛 (Exercise), i.e, 𝜆𝑖 = 𝜆 𝑗 for
1007 all 𝑖, 𝑗. Set 𝜆 = 𝜆𝑖 .
1008 ‘If’:
𝐽 𝑣 · 𝐽𝑤 𝜆𝐴𝑣 · 𝜆𝐴𝑤 𝐴𝑣 · 𝐴𝑤 𝑣 · 𝐴𝑡 𝐴𝑤 𝑣 ·𝑤
= = = = . 
|𝐽 𝑣 ||𝐽𝑤 | |𝜆𝐴𝑣 ||𝜆𝐴𝑤 | |𝐴𝑣 ||𝐴𝑤 | |𝑣 ||𝑤 | |𝑣 ||𝑤 |
1009 An orientation on 𝑈 is a choice of a basis (i.e. a non-zero vector) in ∧𝑛 Ω(𝑈 ). Since we have
1010 already looked at the jacobian matrix with respect to 𝑥 1, . . . , 𝑥𝑛 , let us take d𝑥 1 ∧ · · · ∧ d𝑥𝑛 . Then
1011 the induced map ∧𝑛 Ω(𝑈 )𝑝 −→ ∧𝑛 Ω(R𝑛 ) 𝑓 (𝑝) is multiplication by det 𝐽 . To preserve orientation
1012 is to say that det 𝐽 > 0.
1013 We summarise this discussion as follows.
1014 20.3. Proposition. Let 𝑛 ≥ 2 be an integer, 𝑈 ⊆ R𝑛 an open subset and 𝑓 : 𝑈 −→ R𝑛 a differentiable
1015 function. Then 𝑓 is conformal on𝑈 if and only if the jacobian matrix of 𝑓 at 𝑝 is a multiple of an orthogonal
1016 matrix and its determinant is positive, for every 𝑝 ∈ 𝑈 .
1017 We now restrict out attention to dimension 2. Let 𝑥, 𝑦 be coordinates of R2 . Write 𝑓 = (𝑢, 𝑣).
1018 Then  
𝑢𝑥 (𝑝) 𝑢𝑦 (𝑝)
𝐽 =
𝑣 𝑥 (𝑝) 𝑣𝑦 (𝑝)
1019 Since the columns are orthogonal to each other, there exists 𝜆 ≠ 0 such that 𝑢𝑦 (𝑝) = −𝜆𝑣 𝑥 (𝑝)
1020 and 𝑣𝑦 (𝑝) = 𝜆𝑢𝑥 (𝑝). Then det 𝐽 = 𝜆(𝑢𝑥 (𝑝) 2 + 𝑣 𝑥 (𝑝) 2 ), so 𝜆 > 0. Thus
 
𝑢𝑥 (𝑝) −𝜆𝑣 𝑥 (𝑝)
𝐽 =
𝑣 𝑥 (𝑝) 𝜆𝑢𝑥 (𝑝)
1021 For the rows of 𝐽 to be orthogonal, 𝜆 2 = 1, so 𝜆 = 1. Hence
 
𝑢𝑥 (𝑝) −𝑣 𝑥 (𝑝)
𝐽 =
𝑣 𝑥 (𝑝) 𝑢𝑥 (𝑝)
1022 Summarising this, we get the following relation between conformality and holomorphicity.
1023 20.4. Proposition. Let 𝑈 ⊆ C be open and 𝑓 : 𝑈 −→ C. Then the following are equivalent:
1024 (1) 𝑓 is holomorphic and 𝑓 0 (𝑧) has no zeroes on 𝑈 ;
1025 (2) 𝑓 is conformal on 𝑈 .
1026 20.5. Remark. Some books might require conformal maps to be injective, by definition.
1027 20.6. Remark. A conformal map preserves angles and orientation only, but not length. To see
1028 this, let 𝑈 ⊆ C be a domain and 𝑓 holomorphic on 𝑈 . Suppose that 𝑓 0 (𝑝) ≠ 0. Let 𝛾 : (−𝜖, 𝜖) :
1029 𝑈 be a 𝐶 1 -path with 𝛾 (0) = 𝑝. Write Γ = 𝑓 𝛾. Then |Γ0 (0)| = |𝑓 0 (𝑝)||𝛾 0 (0)|. Hence the length of
1030 an infinitesimal arc through 𝑝 gets multiplied by |𝑓 0 (𝑝)|.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 38 MANOJ KUMMINI

1031 Exercises.
1032 (1) Show that the map 𝑧 ↦→ 𝑧 preserves angles, but not orientation.
1033 (2) Orientation in the case of R2 . Let 𝑣 1, 𝑣 2 be a basis of R2 . Plot them as vectors based at 0.
1034 We can think of orientation as the direction (clockwise, or counter-clockwise) in which
1035 we have to go from 𝑣 1 to 𝑣 2 traversing the smaller of the angles between them. (One of
1036 these angles must be in (0, 𝜋); this is the smaller angle.) Let 𝑓 : R2 −→ R2 be a linear
1037 transformation. Show that 𝑓 preserves orientation if and only if the direction in which
1038 one has to traverse smaller angle from 𝑓 (𝑣 1 ) to 𝑓 (𝑣 2 ) is the same as the direction in which
1039 one has to traverse the smaller angle from 𝑣 1 to 𝑣 2 .
1040 (3) Show that if 𝑓 : 𝑈 −→ C is conformal, then for every 𝑝 ∈ 𝑈 , it maps a neighbourhood
1041 of 𝑝 homeomorphically onto its image.
1042 (4) Show that if 𝑈 ⊆ C is a domain and 𝑓 is an injective holomorphic function on 𝑈 , then
1043 𝑓 is conformal.

1044 Lecture 21. Riemann sphere

1045 We want to discuss Moebius transformations next.
1046 Consider the unit sphere 𝑆 2 in R3 , with the map
𝑥 1 + 𝚤𝑥 2
𝜎 : 𝑆 2 r {(0, 0, 1)} −→ C, (𝑥 1, 𝑥 2, 𝑥 3 ) ↦→ .
1 − 𝑥3
1047 What is this map? Identify the hyperplane 𝑥 3 = 0 with C, with 𝑥 1 as the real part and 𝑥 2 the
1048 imaginary part. Then 𝜎 ((𝑥 1, 𝑥 2, 𝑥 3 )) is the point where the line through (0, 0, 1) and (𝑥 1, 𝑥 2, 𝑥 3 )
1049 meets C. I.e., we need to solve for 𝜆 in
(1 − 𝜆)(0, 0, 1) + 𝜆(𝑥 1, 𝑥 2, 𝑥 3 ) = (𝑦1, 𝑦2, 0).
1
1050 Hence 𝜆 = 1−𝑥 3
. This gives the above description of 𝜎. 𝜎 is a homeomorphism, with 𝑆 2 given
1051 the subspace topology of R3 . Points of 𝑆 2 r (0, 0, 1) with 𝑥 3 > 0 are mapped to C r 𝐵 0,1 , the
1052 points with 𝑥 3 < 0 are mapped to 𝐵 0,1 . With this, 𝑆 2 is a one-point compactification of R2 = C
1053 (Exercise). The map 𝜎 is called stereographic projection.
1054 By the Riemann sphere, we mean 𝑆 2 , with a complex manifold structure given on it. Cover 𝑆 2
1055 with two open subsets, 𝑈 := 𝑆 2 r {(0, 0, 1)} and 𝑉 := 𝑆 2 r {(0, 0, −1)}. We identify 𝑈 with C,
1056 using 𝜎. Note that 𝜎 ((0, 0, −1)) = 0. We can now define 𝜏 : 𝑉 −→ C by
(
0, if 𝑝 = (0, 0, 1)
𝜏 (𝑝) = 1
𝜎 (𝑝) , otherwise.

1057 This identifies 𝑉 with C, and on C r {0}, the map 𝜏𝜎 −1 is 𝑧 ↦→ 𝑧1 , which is a biholomorphic
1058 map, i.e., a bijective holomorphic map whose inverse is holomorphic. We will write C b for the
1059 Riemann sphere.
1060 Using the Riemann sphere, we can reinterpret the notion of poles. Identify C with 𝑈 using
1061 𝜎, and write ∞ for the point (0, 0, 1). This is sometimes called the point at infinity (w.r.t this
1062 identification). Let 𝑝 ∈ C and 𝑓 a holomorphic function defined in a neighbourhood 𝑊 of 𝑝,
1063 with a pole at 𝑝. The function 𝑓 : 𝑊 r {𝑝} −→ C = 𝑈 extends to a function 𝑓˜ : 𝑊 −→ 𝑆 2 , with
1064 𝑓˜(𝑝) = ∞. Shrink 𝑊 so that 𝑓 does not have a zero in 𝑊 . Hence Im 𝑓˜ ⊆ 𝑉 . The composite 𝜏 𝑓˜
1
1065 is the holomorphic map 𝑧 ↦→ 𝑓 (𝑧) on 𝑊 .

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 39

1066 Exercises.
1067 (1) Show that the map 𝜎 in the definition of the Riemann sphere is a homeomorphism and
1068 that the Riemann sphere is a one-point compactification of R2 = C.
1069 (2) Let 𝑐 ∈ (−1, 1). Show that 𝜎 maps 𝑆 2 ∩ {𝑥 3 = 𝑐} to a circle in C and 𝑆 2 ∩ {𝑥 3 < 𝑐} to the
1070 open set bounded by the circle.
1071 Lecture 22. Moebius transformations
1072 References for this lecture are Ahlfors Chapter 2, Section 1.4, and Chapter 3, Section 3. See
1073 also Rodrı́guez, Kra and Gilman, Chapter 8, especially the early parts.
1074 NOTE: We identify C with C b r {∞} through the stereographic projection 𝜎. When you read
1075 this lecture and the next, you should keep this in mind. Sometimes, we will switch between C
1076 and its image under 𝜎 without explicitly mentioning it.
1077 By a Moebius transformation, we mean a meromorphic function on C given by a rational func-
1078 tion of the form
𝑎𝑧 + 𝑏
𝑓 (𝑧) =
𝑐𝑧 + 𝑑
1079 where 𝑎, 𝑏, 𝑐, 𝑑 are complex numbers with 𝑎𝑑 ≠ 𝑏𝑐.
1080 22.1. Remark. We make the following observations (notation as above):
1081 (1) 𝑓 (𝑧) is holomorphic on C if and only if 𝑐 = 0; otherwise 𝑓 has exactly one pole, at − 𝑑𝑐 .
1082 (2) 𝑓 (𝑧) is injective on the complement of the pole. (Exercise)
1083 (3) We can think of 𝑓 as being given by
  
𝑎 𝑏 𝑧
𝑐 𝑑 1
1084 We can extend 𝑓 to bijective function from Cb to C,
b still denoted by 𝑓 , by setting
(
𝑓 (∞) = ∞, if 𝑐 = 0;
𝑓 (− 𝑐 ) = ∞ and 𝑓 (∞) = 𝑐 , otherwise.
𝑑 𝑎

1085 22.2. Proposition. The extended function 𝑓 : C b is a homeomorphism.

b −→ C
1086 Proof. The extended function is bijective (check). I will show that it is continuous and open,
1087 treating the two cases 𝑐 = 0 and 𝑐 ≠ 0 separately for your convenience.
𝑐 = 0: Then 𝑑 ≠ 0. Replacing 𝑎 by 𝑑𝑎 and 𝑏 by 𝑑𝑏 , we may assume that 𝑓 is given by the
polynomial 𝑎𝑧 + 𝑏. The inverse function is 𝑧 ↦→ 𝑧−𝑏𝑎 . It is continuous and open. Hence 𝑓 is a
homeomorphism of C to itself. Now consider the extension of 𝑓 to C. b Let 𝑈 be an open subset
of C.
b We want to show that 𝑓 (𝑈 ) and 𝑓 −1 (𝑈 ) are open. If 𝑈 ⊂ C, then 𝑓 (𝑈 ) and 𝑓 −1 (𝑈 ) are
open. Otherwise, i.e. if ∞ ∈ 𝑈 , then C
b r 𝑈 is compact and, hence, closed, so
 
𝑓 (𝑈 ) = C
br𝑓 C b r𝑈
 
−1 −1 b
𝑓 (𝑈 ) = C r 𝑓
b C r𝑈
1088 are open. (Note: 𝑓 is bijective.)
1089 𝑐 ≠ 0: Let us compute 𝑓 −1 : Write 𝑤 = 𝑎𝑧+𝑏
𝑐𝑧+𝑑 . Rewrite as (𝑐𝑧 + 𝑑)𝑤 = (𝑎𝑧 + 𝑏), so we get
1090 (𝑐𝑤 − 𝑎)𝑧 = −𝑑𝑤 + 𝑏. Define a meromorphic function 𝑔 : C −→ C by
−𝑑𝑧 + 𝑏
𝑔(𝑧) = .
𝑐𝑧 − 𝑎

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 40 MANOJ KUMMINI

1091 Note that      

−𝑑 𝑏 𝑎 𝑏 𝑎 𝑏 −𝑑 𝑏
= −(𝑎𝑑 − 𝑏𝑐)𝐼 2 =
𝑐 −𝑎 𝑐 𝑑 𝑐 𝑑 𝑐 −𝑎
1092 Hence on C r {− 𝑑𝑐 , 𝑎𝑐 }, (𝑓 𝑔)(𝑧) = (𝑔𝑓 )(𝑧), i.e, 𝑔 is the inverse of 𝑓 . It is easy to check that
1093 its extension to C b is the inverse of 𝑓 (on C).
b Since C b is a metric space, we can check that 𝑓 is
1094 continuous by taking limits. Let 𝜁 ∈ C. b Let 𝜁𝑖 be a sequence converging to 𝜁 . If 𝜁 = ∞, then
1095 |𝜎 (𝜁𝑖 )| → ∞ (see the description of 𝜎 in the previous lecture). Hence 𝑓 (𝜁𝑖 ) → 𝑎𝑐 . Similarly, if
1096 𝜁 = − 𝑑𝑐 , then |𝑓 (𝜁𝑖 )| → ∞. Hence 𝑓 is continuous. This applies to every Moebius transforma-
1097 tion, including to 𝑔 = 𝑓 −1 . Hence 𝑓 is a homeomorphism. 
1098 We will come back to Moebius transformations in the next lecture. For now, we look at ra-
1099 tional functions, in general.
1100 Let 𝑝, 𝑞 ∈ C[𝑧] be relatively prime non-zero polynomials. Let𝑚 = deg 𝑝 and 𝑛 = deg 𝑞. Then
1101 we get a meromorphic function
𝑝 (𝑧)
𝑓 (𝑧) =
𝑞(𝑧)
1102 on C. We can extend 𝑓 to C b If 𝜁 ∈ C and 𝑞(𝜁 ) ≠ 0, 𝑓 (𝜁 ) = 𝑝 (𝜁 ) (nothing
b as follows. Let 𝜁 ∈ C.
𝑞(𝜁 )
𝑝 (𝜁 )
1103 new here). For 𝜁 ∈ C is a zero of 𝑞, then for every sequence 𝜁𝑖 −→ 𝜁 , −→ ∞, so we can
| 𝑞(𝜁𝑖𝑖 ) |
1104 define 𝑓 (𝜁 ) = ∞.
1105 Now assume 𝜁 = ∞. Write 𝑝 (𝑧) = 𝑎𝑚 𝑧𝑚 + · · · + 𝑎 0 and 𝑞(𝑧) = 𝑏𝑛 𝑧𝑛 + · · · + 𝑏 0 . If 𝑚 > 𝑛,
𝑝 (𝜁 )
1106 then | 𝑞(𝜁𝑖𝑖 ) | −→ ∞, so we can define 𝑓 (∞) = ∞. If 𝑚 = 𝑛, then we can define 𝑓 (∞) = 𝑏𝑎𝑚𝑚 , since
𝑝 (𝜁𝑖 ) 𝑝 (𝜁 )
1107 −→ 𝑏𝑎𝑚𝑚 . If 𝑚 < 𝑛, then 𝑞(𝜁𝑖𝑖 ) −→ 0, so define 𝑓 (∞) = 0. Consider the neighbourhood
𝑞(𝜁𝑖 )
1108 𝑉 ⊆C b of ∞, from the last lecture. We identify 𝑉 with C using 𝜏. Since 𝜁𝑖 → ∞, 𝜏 (𝜁𝑖 ) −→ 0.
1109 On a neighbourhood 𝑊 ⊆ 𝑉 of ∞, 𝑓 is holomorphic, and has the form
𝑎𝑚 ( 𝑧1 )𝑚 + 𝑎𝑚−1 ( 𝑧1 )𝑚−1 + · · · + 𝑎 0 𝑎𝑚 + 𝑎𝑚−1𝑧 1 + · · · + 𝑎 0𝑧𝑚
= 𝑧𝑛−𝑚 .
𝑏𝑛 ( 𝑧1 )𝑛 + 𝑏𝑛−1 ( 𝑧1 )𝑛−1 + · · · + 𝑏 0 𝑏𝑛 + 𝑏𝑛−1𝑧 1 + · · · + 𝑏 0𝑧𝑛
1110 Hence the order of the zero of 𝑓 at ∞ is 𝑛 − 𝑚.
1111 Exercises.
1112 (1) Let 𝑎, 𝑏, 𝑐, 𝑑 be complex numbers such that 𝑎𝑑 − 𝑏𝑐 ≠ 0. Let 𝑈 = C if 𝑐 = 0; let 𝑈 =
1113 C r {− 𝑑𝑐 }, otherwise. Show that the function
𝑎𝑧 + 𝑏
𝑓 : 𝑈 −→ C, 𝑧 ↦→
𝑐𝑧 + 𝑑
1114 is injective. Hint: Look at the linear map C −→ C2 given by the matrix
2
 
𝑎 𝑏
.
𝑐 𝑑
1115 (2) Complete the proof (by filling in the missing steps) of Proposition 22.2.
1116 (3) Read Ahlfors, Chapter 2, Section 1.4 about partial fraction expansions. Do Exercise 1 of
1117 that section.
1118 (4) Let 𝑝, 𝑞 ∈ C[𝑧] be relatively prime non-zero polynomials. Assume without loss of gen-
1119 erality that deg 𝑞 > 0. Let 𝑓 be the meromorphic function
𝑝 (𝑧)
𝑞(𝑧)

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 41

1120 on C. Show that 𝑓 0 is meromorphic and the poles of 𝑓 0 are exactly the poles of 𝑓 . If 𝜁 is
1121 a pole of 𝑓 of order 𝑚, then the order of the pole of 𝑓 0 at 𝜁 is 𝑚 + 1.

1122 Lecture 23. Moebius transformations, continued.

1123 References for this lecture are Ahlfors, Chapter 3, Section 3.2 and Rodrı́guez, Kra and Gilman,
1124 Section 8.1.
1125 Let 𝑎, 𝑏, 𝑐, 𝑑 ∈ C with 𝑎𝑑 − 𝑏𝑐 ≠ 0. For every 𝜁 ∈ C, 𝜁 ≠ 0,
𝑎𝑧 + 𝑏 𝜁 𝑎𝑧 + 𝜁𝑏
=
𝑐𝑧 + 𝑑 𝜁𝑐𝑧 + 𝜁𝑑
1126 as meromorphic functions on C. Hence we may assume that 𝑎𝑑 −𝑏𝑐 = 1. In other words, every
1127 Moebius transformation can be represented by an element of SL2 (C). Hereafter, we will make
1128 this assumption.
1129 Recall that a Moebius transformation is typically only a meromorphic function on C, but a
1130 well-defined function on C. b

1131 23.1. Lemma. Assume that the Moebius transformation

𝑎𝑧 + 𝑏
𝑓 (𝑧) = (with 𝑎𝑑 − 𝑏𝑐 = 1)
𝑐𝑧 + 𝑑
1132 b fixes 0, 1 and ∞, then it is the identity map: 𝑎 = 𝑑 = 1, 𝑏 = 𝑐 = 0.
(considered as a function on C)
1133 Proof. Since 𝑓 (∞) = ∞, and 𝑓 is bijective, 𝑓 is holomorphic on C, and the only zero of 𝑓 is 0.
1134 Hence 𝑐 = 0. Without loss of generality, we may assume that 𝑑 = 1, i.e, 𝑓 is given by a linear
1135 polynomial 𝑎𝑧 + 𝑏. Since we have assumed that 𝑎𝑑 − 𝑏𝑐 = 1, 𝑎 = 1. Since 0 is the only zero of 𝑓 ,
1136 𝑓 is the identity map. 
1137 23.2. Proposition. Let 𝜁 0, 𝜁 1, 𝜁 ∞ be distinct points on C.
b Then there is a unique Moebius transformation
𝑎𝑧 + 𝑏
𝑧 ↦→
𝑐𝑧 + 𝑑
1138 such that 𝜁 0 ↦→ 0, 𝜁 1 ↦→ 1 and 𝜁 ∞ ↦→ ∞.
1139 Proof. Let
𝑧 − 𝜁0 𝜁1 − 𝜁∞
𝑓 (𝑧) = .
𝑧 − 𝜁∞ 𝜁1 − 𝜁0
1140 It is a Moebius transformation, with 𝑓 (𝜁 0 ) = 0, 𝑓 (𝜁 1 ) = 1 and 𝑓 (𝜁 ∞ ) = ∞. Let 𝑔(𝑧) be any
1141 Moebius transformation such that 𝑔(𝜁 0 ) = 0, 𝑔(𝜁 1 ) = 1 and 𝑔(𝜁 ∞ ) = ∞. Then 𝑔𝑓 −1 and 𝑓 𝑔−1 are
1142 Moebius transformations, fixing 0, 1 and ∞. Hence 𝑔𝑓 −1 = 𝑓 𝑔−1 = id. Therefore 𝑔 = (𝑓 −1 ) −1 =
1143 𝑓. 
1144 23.3. Definition. Let 𝜁 , 𝜁 0, 𝜁 1, 𝜁 ∞ be distinct points on C.
b Their cross-ratio is the image of 𝜁 under
1145 the unique Moebius transformation that sends 𝜁 0 to 0, 𝜁 1 to 1 and 𝜁 ∞ to ∞. We will denote the
1146 cross-ratio by (𝜁 , 𝜁 0, 𝜁 1, 𝜁 ∞ ).
1147 In other words
𝜁 − 𝜁0 𝜁1 − 𝜁∞
(𝜁 , 𝜁 0, 𝜁 1, 𝜁 ∞ ) = .
𝜁 − 𝜁∞ 𝜁1 − 𝜁0
1148 23.4. Proposition. Let 𝜁 , 𝜁 0, 𝜁 1, 𝜁 ∞ be distinct points on C
b and 𝑓 a Moebius transformation. Then
(𝑓 (𝜁 ), 𝑓 (𝜁 0 ), 𝑓 (𝜁 1 ), 𝑓 (𝜁 ∞ )) = (𝜁 , 𝜁 0, 𝜁 1, 𝜁 ∞ ).

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 42 MANOJ KUMMINI

1149 Proof. Let 𝑔 be the unique Moebius transformation that sends 𝜁 0 to 0, 𝜁 1 to 1 and 𝜁 ∞ to ∞. Then
(𝜁 , 𝜁 0, 𝜁 1, 𝜁 ∞ ) = 𝑔(𝜁 ).
1150 Then the Moebius transformation 𝑔𝑓 −1 sends 𝑓 (𝜁 0 ) to 0, 𝑓 (𝜁 1 ) to 1 and 𝑓 (𝜁 ∞ ) to ∞.
(𝑓 (𝜁 ), 𝑓 (𝜁 0 ), 𝑓 (𝜁 1 ), 𝑓 (𝜁 ∞ )) = 𝑔𝑓 −1 (𝑓 (𝜁 )) = 𝑔(𝜁 ) = (𝜁 , 𝜁 0, 𝜁 1, 𝜁 ∞ ). 
1151 23.5. Proposition. Let 𝜁 , 𝜁 0, 𝜁 1, 𝜁 ∞ be distinct points on C. Then the cross ratio (𝜁 , 𝜁 0, 𝜁 1, 𝜁 ∞ ) is real if
1152 and only if the points lie on a circle or a straight line.
1153 Proof. We start with the following observation. Let𝑎, 𝑏, 𝑐 ∈ C be distinct. Then they are collinear
1154 if and only if the (non-zero) elements 𝑎 − 𝑏, 𝑎 − 𝑐 ∈ C are linearly dependent over R if and only
1155 if 𝑎−𝑏
𝑎−𝑐 is real.
1156 We now prove the proposition. ‘If’: Suppose that 𝜁 , 𝜁 0, 𝜁 1, 𝜁 ∞ lie on a straight line. Then
𝜁 − 𝜁0 𝜁1 − 𝜁∞
and
𝜁 − 𝜁∞ 𝜁1 − 𝜁0
1157 are real numbers, so the cross ratio is a real number.
1158 If the four points lie on a circle, the proof involves a calculation with the angle between the
1159 line segments 𝑧 − 𝜁 0 and 𝑧 − 𝜁 ∞ , and similarly between 𝜁 − 𝜁 0 and 𝜁 − 𝜁 ∞ . Please see the file
1160 rkg p204.pdf uploaded in moodle.
𝜁 −𝜁
1161 ‘Only if’: First assume that 𝜁 0, 𝜁 1, 𝜁 ∞ are collinear, then 𝜁11 −𝜁∞0 is a real number, and, if, further,
𝜁 −𝜁
1162 the cross-ratio is real, then 𝜁 −𝜁∞0 is real, i.e., 𝜁 , 𝜁 0, 𝜁 ∞ are collinear.
1163 If 𝜁 0, 𝜁 1, 𝜁 ∞ are not collinear, then we need to consider the circle containing these points and
1164 show, using the a calculation of angles, that 𝜁 also lies on the same circle. Please see the file
1165 rkg p204.pdf uploaded in moodle. 
1166 23.6. Corollary. A Moebius transformation maps circles and straight lines to circles and straight lines.
1167 Exercises.
1168 (1) Show that the composite of two Moebius transformations is a Moebius transformation.
1169 Lecture 24. Singularity at infinity
1170 In Lecture 22, we looked at extending rational functions to ∞ and the resulting singularity
1171 at ∞.
1172 24.1. Definition. Let 𝑓 be an entire function. We say that 𝑓 has a removable singularity (respec-
1173 tively, a pole, an essential singularity) at ∞ if the function 𝑓 ( 𝑧1 ) has a removable singularity (re-
1174 spectively, a pole, an essential singularity) at 0.
1175 Note that an entire function has a power series expansion that is convergent everywhere;
1176 take 𝑐 = 0 and 𝑅 = ∞ in Corollary 15.5.
1177 24.2. Proposition. An entire function has a removable singularity at ∞ if and only if it is constant.
1178 Proof. Let 𝑓 be an entire function. If it is constant, it has a removable
Í singularity at Í
∞. Con-
1179 versely assume that it has a removable singularity at ∞. Write 𝑓 (𝑧) = 𝑛∈N 𝑎𝑛 𝑧𝑛 . Then 𝑛∈N 𝑎𝑛 𝑧 −𝑛
1180 has a removable singularity at 0. Therefore
∑︁
lim 𝑎𝑛 𝑧 −𝑛+1 = 0.
𝑧→0
𝑛∈N
1181 Hence 𝑎𝑛 = 0 if −𝑛 + 1 < 0, i.e., 𝑓 (𝑧) = 𝑎 0 + 𝑎 1𝑧. Then 𝑎 1 = lim𝑧→0 𝑧𝑓 ( 𝑧1 ) = 0. Hence 𝑎𝑛 = 0 for
1182 each 𝑛 > 0, i.e., 𝑓 is constant. 

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 43

1183 24.3. Proposition. Let 𝑓 be an entire function. Then the following are equivalent:
1184 (1) 𝑓 has a pole at ∞;
1185 (2) For every 𝑀 > 0 there exists 𝑅 > 0 such that |𝑓 (𝑧)| > 𝑀 for all |𝑧| > 𝑅;
1186 (3) 𝑓 is a non-constant polynomial.
1187 Proof. (1) =⇒ (2): Apply Proposition 17.4 (3) to the function 𝑓 ( 𝑧1 ) at its pole 0, to see that for
1188 every 𝑀 > 0 there exists 𝑟 > 0 such that |𝑓 ( 𝑧1 )| > 𝑀 for all |𝑧| < 𝑟 . Take 𝑅 = 𝑟1 .
1189 (2) =⇒ (1): By Proposition 17.9, 𝑓 does not have an essential singularity at ∞. By Proposi-
1190 tion 24.2, 𝑓 does not have a removable singularity at ∞.
1191 (1) =⇒ (3): Note that 𝑓 is a non-constant function. We have already established that for
1192 every 𝑀 > 0 there exists 𝑅 > 0 such that |𝑓 (𝑧)| > 𝑀 for all |𝑧| > 𝑅; hence the zeros of 𝑓 are
1193 in a compact subset of C, so 𝑓 has only finitely Î𝑛 many zeros, say, 𝑐 1, . . . , 𝑐𝑛 of orders 𝑚 1, . . . , 𝑚𝑛
1194 respectively. Therefore we can write 𝑓 (𝑧) = 𝑖=1 (𝑧 − 𝑐𝑖 ) 𝑔(𝑧) where 𝑔(𝑧) is an entire function
𝑚 𝑖

1195 without any zeros.

1196 It suffices to show that 𝑔 is constant. Assume the contrary. By hypothesis, 𝑓 does not have
1197 an essential singularity at ∞. Therefore by Exercise 10 of Lecture 17 and Proposition 24.2, we
1198 see that 𝑔 has a pole at ∞. Therefore for every 𝑀 > 0 there exists 𝑅 > 0 such that |𝑔(𝑧)| > 𝑀
1
1199 for all |𝑧| > 𝑅. Hence 𝑔(𝑧) which is an entire function is bounded, so it is constant by Liouville’s
1200 theorem (Proposition 16.2), contradicting the hypothesis that 𝑔 is not constant.
1201 (3) =⇒ (1): Write 𝑓 (𝑧) = 𝑎 0 + 𝑎 1𝑧 + · · · + 𝑎𝑛 𝑧𝑛 with 𝑛 > 0 and 𝑎𝑛 ≠ 0. Then 𝑓 ( 𝑧1 ) =
1202 𝑎𝑛 𝑧 −𝑛 + · · · + 𝑎 0 has a pole of order 𝑛 at 0. 
24.4. Corollary. Let 𝑓 (𝑧) = 𝑛∈N 𝑎𝑛 𝑧𝑛 be an entire function with 𝑎𝑛 ≠ 0 for infinitely many 𝑛. Then
Í
1203
1204 for every 𝐴 ∈ C, every 𝑅 > 0, every 𝛿 > 0, there exists 𝜁 ∉ 𝐵 0,𝑅 such that |𝑓 (𝜁 ) − 𝐴| < 𝛿.
1205 Proof. Using the above propositions, we see that 𝑓 ( 𝑧1 ) has an essential singularity at 0. Now
1206 apply Proposition 17.9 to 𝑓 ( 𝑧1 ) at 0. 
1207 Lecture 25. Automorphisms of the complex plane
1208 This lecture is based on Rodrı́guez, Kra and Gilman, Sections 8.1, 8.2.
1209 25.1. Definition. Let 𝑈 ⊆ C be a domain. By an automorphism of 𝑈 , we mean a holomorphic
1210 function 𝑓 : 𝑈 −→ 𝑈 such that there exists 𝑔 : 𝑈 −→ 𝑈 such that 𝑓 𝑔 = 𝑔𝑓 = id𝑈 .
1211 25.2. Proposition. Let𝑈 ⊆ C be a domain. Let 𝑓 : 𝑈 −→ 𝑈 be a bijective function. Then the following
1212 are equivalent:
1213 (1) 𝑓 is holomorphic;
1214 (2) 𝑓 is biholomorphic, i.e, 𝑓 and 𝑓 −1 are holomorphic;
1215 (3) 𝑓 and 𝑓 −1 are conformal.
1216 Proof. If 𝑓 and 𝑓 −1 are conformal, then they are holomorphic. Conversely, if 𝑓 is bijective and
1217 holomorphic, then it is conformal, since if 𝑓 0 (𝑐) = 0 for some 𝑐, then 𝑓 would not be injective in
1218 an neighbourhood of 𝑐. Therefore it remains to show that if 𝑓 is holomorphic, then 𝑓 −1 is holo-
1219 morphic. We will think of 𝑈 as a subset of R2 and show that 𝑓 −1 is differentiable as a function
1220 of two real variables and that the Cauchy-Riemann equations are satisfied (Theorem 2.8)
1221 We first show that 𝑓 −1 is differentiable as a function of two real variables. Let 𝑝 ∈ 𝑈 and
1222 write 𝑞 = 𝑓 (𝑝). Note that 𝑓 0 (𝑝) ≠ 0, for, otherwise, 𝑓 would not be injective on 𝐵𝑝,𝛿 r {𝑝} for
1223 some 0 < 𝛿  1. Write 𝑓 = 𝑢 + 𝚤𝑣. Hence the derivative of (𝑢, 𝑣) at 𝑝 is the jacobian matrix
 " 𝜕𝑢 𝜕𝑢
#
(𝑝) − (𝑝)
 𝜕𝑢 𝜕𝑣
(𝑝) (𝑝)
𝐽 (𝑓 , 𝑝) := 𝜕𝑥
𝜕𝑢
𝜕𝑥
𝜕𝑣 = 𝜕𝑥 𝜕𝑦
.
𝜕𝑦 (𝑝) 𝜕𝑦 (𝑝) 𝜕𝑢
𝜕𝑦 (𝑝) 𝜕𝑢
𝜕𝑥 (𝑝)

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 44 MANOJ KUMMINI

1224 We have used the Cauchy-Riemann equations for 𝑓 . From Theorem 2.8, we see that
 2  2
2 𝜕𝑢 𝜕𝑣
0
|𝑓 (𝑝)| = (𝑝) + (𝑝) = det 𝐽 (𝑓 , 𝑝).
𝜕𝑥 𝜕𝑥
1225 Therefore 𝐽 (𝑓 , 𝑝) is invertible since 𝑓 0 (𝑝) ≠ 0. Note that (𝑢, 𝑣) is continuously differentiable,
1226 since 𝑓 is complex-analytic. Now using the inverse function theorem (e.g., Rudin, Principles
1227 of Mathematical Analysis, Chapter 9), we see that 𝑓 −1 is differentiable in a neighbourhood of
1228 𝑞, as a function of two real variables.
1229 Now to show that 𝑓 −1 satisfies the Cauchy-Riemann equations, observe that
𝐽 (𝑓 −1, 𝑞) 𝐽 (𝑓 , 𝑝) = 𝐼 i.e. 𝐽 (𝑓 −1, 𝑞) = (𝐽 (𝑓 , 𝑝)) −1 .
1230 Since 𝐽 (𝑓 , 𝑝) is a non-zero real matrix of the form
 
𝑎 −𝑏
𝑏 𝑎
1231 (such matrices are invertible) its inverse 𝐽 (𝑓 −1, 𝑞) too is of the same form (Exercise). Hence 𝑓 −1
1232 satisfies the Cauchy-Riemann equations. 
1233 25.3. Proposition. The map C −→ C, 𝑧 ↦→ 𝑎𝑧 + 𝑏, where 𝑎, 𝑏 ∈ C, 𝑎 ≠ 0 is an automorphism, with
1234 inverse 𝑧 ↦→ 𝑎1 (𝑧 − 𝑏). Conversely, if 𝑓 : C −→ C is an automorphism, then there exist 𝑎, 𝑏 ∈ C, 𝑎 ≠ 0
1235 such that 𝑓 (𝑧) = 𝑎𝑧 + 𝑏 for every 𝑧 ∈ C.
1236 Proof. The map 𝑧 ↦→ 𝑎𝑧 +𝑏 (with 𝑎 ≠ 0) is bijective and holomorphic, i.e., an automorphism. Its
1237 inverse is the map 𝑧 ↦→ 𝑎1 (𝑧 − 𝑏). Now assume that 𝑓 : C −→ C is an automorphism. Since 𝑓
1238 is entire, it has a convergent power series expansion,
Í valid everywhere on C. (In Corollary 15.5,
1239 we can take 𝑐 = 0 and 𝑅 = ∞.) Write 𝑓 (𝑧) = 𝑛∈N 𝑎𝑛 𝑧𝑛 . Suppose that 𝑎𝑛 ≠ 0 for infinitely
1240 many 𝑛. Then by Corollary 24.4 we see that for each 𝑅 > 0, the set 𝑓 (C r 𝐵 0,𝑅 ) is dense in C.
1241 However, since 𝑓 is injective, 𝑓 (𝐵 0,1 ) is non-empty but
𝑓 (C r 𝐵 0,1 ) ∩ 𝑓 (𝐵 0,1 ) = ∅.
1242 Hence 𝑎𝑛 = 0 for all 𝑛  0. Write 𝑓 (𝑧) = 𝑎 0 + 𝑎 1𝑧 + · · · + 𝑎𝑚 𝑧𝑚 with 𝑚 ≥ 0 and 𝑎𝑚 ≠ 0. Since
1243 𝑓 is injective, it is not constant, so 𝑚 ≥ 1.
1244 We need to show that 𝑚 = 1. By way of contradiction, assume that 𝑚 > 1. Then deg 𝑓 0 > 0,
1245 so there exists 𝜁 ∈ C such that 𝑓 0 (𝜁 ) ≠ 0. Hence there exists a neighbourhood of 𝜁 on which 𝑓
1246 is not injective, a contradiction. 
1247 25.4. Example. The map
𝑧 −𝚤
𝑧 ↦→
𝑧 +𝚤
1248 the upper half plane {𝑧 ∈ C | =(𝑧) > 0} to the open unit disc. If 𝑧 = 𝑥 + 𝑦𝚤, with 𝑦 > 0, then
𝑥 + (𝑦 − 1)𝚤
<1
𝑥 + (𝑦 + 1)𝚤
1249 since |𝑦 − 1| < |𝑦 + 1|. It is a Moebius transformation, and holomorphic on the upper half plane,
1250 so the map is conformal.
1251 (The Riemann mapping theorem says that every simply connected domain 𝑈 ( C is biholo-
1252 morphic to the open unit disc. Above, we have given a specific map that works for the upper
1253 half plane.) 
1254 We now prove the Schwarz lemma, which describes maps from 𝐵 0,1 to itself.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 45

1255 25.5. Proposition. Let 𝑓 : 𝐵 0,1 −→ 𝐵 0,1 is holomorphic with 𝑓 (0) = 0, then |𝑓 (𝑧)| ≤ |𝑧| for every
1256 𝑧 ∈ 𝐵 0,1 and |𝑓 0 (0)| ≤ 1. If |𝑓 (𝑧)| = |𝑧| for some 𝑧 ∈ 𝐵 0,1 , 𝑧 ≠ 0 or if |𝑓 0 (0)| = 1, then 𝑓 (𝑧) = 𝑐𝑧 for
1257 some constant 𝑐 with |𝑐 | = 1.
Proof. Write 𝑓 (𝑧) = 𝑛=1 𝑎𝑛 𝑧𝑛 . Let
Í∞
1258
(
𝑓 (𝑧)
, if 𝑧 ≠ 0,
𝑔(𝑧) = 𝑧
𝑎 1, if 𝑧 = 0.
1259 (Note that 𝑓 0 (0) = 𝑔(0).) Then, for every 0 < 𝑟 < 1, and every 𝑧 with |𝑧| = 𝑟 ,
𝑓 (𝑧) |𝑓 (𝑧)| 1
|𝑔(𝑧)| = = ≤ .
𝑧 |𝑧| 𝑟
1260 Hence by the maximum principle (Proposition 19.1) |𝑔(𝑧)| ≤ 𝑟1 for every 𝑧 ∈ 𝐵 0,𝑟 for every
1261 0 < 𝑟 < 1. Hence |𝑔(𝑧)| ≤ 1 for every 𝑧 ∈ 𝐵 0,1 . Hence |𝑓 (𝑧)| ≤ |𝑧| for each 𝑧 ∈ 𝐵 0,1 and
1262 𝑓 0 (0) = 𝑔(0) ≤ 1.
1263 Now suppose that |𝑓 (𝑧)| = |𝑧| for some 𝑧 ∈ 𝐵 0,1 , 𝑧 ≠ 0 or that |𝑓 0 (0)| = 1. Equivalently,
1264 |𝑔(𝑧)| = 1 for some 𝑧 ∈ 𝐵 0,1 , then 𝑔 is a constant function, again by the maximum principle. 
1265 As an immediate corollary, we get a property of holomorphic automorphisms of the unit
1266 disc. There is a more precise statement, characterising holomorphic automorphisms of the
1267 unit disc. The proof of the general statement is not difficult, but we will skip it. If you are
1268 interested, you can look at Rodrı́guez, Kra and Gilman, Section 8.2, Theorem 8.18, or Lang,
1269 Complex Analysis, Chapter VII, Section 2.
1270 25.6. Corollary. Let 𝑓 : 𝐵 0,1 −→ 𝐵 0,1 be a bijective holomorphic map. Then it is a Moebius transforma-
1271 tion.
1272 Proof. Let 𝑐 = 𝑓 (0). Then check that the Moebius transformation
𝑧 −𝑐
𝑔 : 𝑧 ↦→
1 − 𝑐𝑧
1273 is an holomorphic automorphism of 𝐵 0,1 . Hence 𝑔𝑓 a holomorphic automorphism of 𝐵 0,1 , with
1274 𝑔𝑓 (0) = 0. Therefore it suffices to show that 𝑔𝑓 is a Moebius transformation. Replacing 𝑓 by
1275 𝑔𝑓 , we may assume that 𝑓 (0) = 0. Write 𝑤 = 𝑓 (𝑧). Then
|𝑧| = |𝑓 −1 (𝑤)| ≤ |𝑤 | = |𝑓 (𝑧)| ≤ |𝑧|
1276 by the use of the the Schwarz Lemma (Proposition 25.5) once for 𝑓 −1 and then for 𝑓 . Hence
1277 |𝑓 (𝑧)| = |𝑧| for every 𝑧 ∈ 𝐵 0,1 . Again by Proposition 25.5, there exists 𝑎 with |𝑎| = 1 such that
1278 𝑓 (𝑧) = 𝑎𝑧. Hence 𝑓 is a Moebius transformation. 
1279 Exercises.
1280 (1) Show that the inverse of a non-zero real matrix of the form
 
𝑎 −𝑏
𝑏 𝑎
1281 is a matrix of the above form. (Hint: after appropriate scaling, this matrix representa-
1282 tion rotation in R2 .)
1283 (2) Determine the images of horizontal and vertical lines in the upper half plane under the
1284 map in Example 25.4.
1285 (3) Find the inverse of the map in Example 25.4.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 46 MANOJ KUMMINI

1286 (4) Show that the map

𝑧−1
𝑧 ↦→ −𝚤
𝑧+1
1287 conformally maps the upper half disc ({|𝑧| < 1, =𝑧 > 0}) to the first quadrant.
1288 (5) Find a conformal map from the upper half unit disc to the unit disc.
1289 (6) Show that C and 𝐵 0,1 are not biholomorphic to each other.

1290 Lecture 26. Automorphisms of the Riemann sphere

1291 We want to understand holomorphic functions from C b to itself. In Lecture 21, we described
1292 the following open covering of C:
b (Here we change our notation a little bit.) 𝑈 0 := C b r {∞},
1293 𝑈 ∞ := C
b r {0}, 𝜎 : 𝑈 0 −→ C the stereographic projection map, and 𝜏 : 𝑈 ∞ −→ C.

1294 26.1. Definition. Let 𝑓 : C b −→ C b and 𝑝 ∈ C b and 𝑞 = 𝑓 (𝑝). Say that 𝑓 is differentiable at 𝑝 if 𝑓˜ is
1295 differentiable at 𝜁 , where


 (𝜎 𝑓 𝜎 −1, 𝜎 (𝑝)) if 𝑝 ∈ 𝑈 0 and 𝑞 ∈ 𝑈 0 ;
 (𝜏 𝑓 𝜎 −1, 𝜎 (𝑝)) if 𝑝 ∈ 𝑈 0 and 𝑞 = ∞;


( 𝑓˜, 𝜁 ) =


 (𝜎 𝑓 𝜏 −1, 𝜏 (𝑝)) if 𝑝 = ∞ and 𝑞 ∈ 𝑈 0 ;

 (𝜏 𝑓 𝜏 −1, 𝜏 (𝑝)) if 𝑝 = ∞ = 𝑞.


1296 Say that 𝑓 is holomorphic if it is differentiable at 𝑝 for each 𝑝 ∈ C. b
1297 It might appear that we have given preference to 𝑈 0 over 𝑈 ∞ while making the above defini-
1298 tion. This is not the case. For example, suppose that {𝑝, 𝑞} ⊆ 𝑈 0 ∩ 𝑈 ∞ . Then the following are
1299 equivalent:
1300 (1) 𝜎 𝑓 𝜎 −1 is differentiable at 𝜎 (𝑝);
1301 (2) 𝜎 𝑓 𝜏 −1 is differentiable at 𝜏 (𝑝);
1302 (3) 𝜏 𝑓 𝜎 −1 is differentiable at 𝜎 (𝑝);
1303 (4) 𝜏 𝑓 𝜏 −1 is differentiable at 𝜏 (𝑝).
1304 Assume that 𝜎 𝑓 𝜎 −1 is differentiable at 𝜎 (𝑝); let us show that 𝜎 𝑓 𝜏 −1 is differentiable at 𝜏 (𝑝).
1305 Note that, in a neighbourhood of 𝜏 (𝑝),
1
 
−1 −1
𝜎 𝑓 𝜏 = (𝜎 𝑓 𝜎 ) ◦ 𝑧 ↦→
𝑧
1306 so 𝜎 𝑓 𝜏 −1 is differentiable at 𝜏 (𝑝).
1307 26.2. Proposition. Let 𝑓 : C b be a bijective holomorphic map. Then 𝑓 −1 is holomorphic.
b −→ C

1308 Proof. Let 𝑝 ∈ C b and 𝑞 = 𝑓 (𝑝). We will assume that 𝑝 ∈ 𝑈 0 and 𝑞 ∈ 𝑈 0 ; the other cases can be
1309 handled similarly. Hence we need to show that 𝜎 𝑓 −1𝜎 −1 is differentiable at 𝜎 (𝑞). But note that
1310 𝜎 𝑓 −1𝜎 −1 is the inverse of the bijective holomorphic map 𝜎 𝑓 𝜎 −1 , so 𝜎 𝑓 −1𝜎 −1 is differentiable at
1311 𝜎 (𝑞). 
1312 26.3. Definition. By an automorphism of C,
b we mean a bijective holomorphic map from C
b to
1313 itself.
1314 The special linear group SL2 (C) is the group of 2 × 2 complex matrices with determinant 1. It
1315 is a group, under usual matrix multiplication. Let 𝑓 and 𝑔 be Moebius transformations:
𝑎𝑧 + 𝑏 𝑎0𝑧 + 𝑏 0
𝑓 (𝑧) = and 𝑔(𝑧) = 0 .
𝑐𝑧 + 𝑑 𝑐 𝑧 + 𝑑0

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 47

1316 Then 0 0
𝑎 𝑎𝑐 0𝑧+𝑑 (𝑎𝑎0 + 𝑏𝑐 0)𝑧 + (𝑎𝑏 0 + 𝑏𝑑 0)
𝑧+𝑏
0 +𝑏
𝑓 (𝑔(𝑧)) = 0 0 = ,
𝑐 𝑎𝑐 0𝑧+𝑑
𝑧+𝑏
0 +𝑑
(𝑐𝑎0 + 𝑑𝑐 0)𝑧 + (𝑐𝑏 0 + 𝑑𝑑 0)
1317 which is a Moebius transformations. Hence the set of Moebius transformations form a group
1318 M under composition.
1319 Note that the group operation in SL2 (C) is given by:
𝑎 𝑏 𝑎0 𝑏 0
  0
𝑎𝑎 + 𝑏𝑐 0 𝑎𝑏 0 + 𝑏𝑑 0
  
=
𝑐 𝑑 𝑐0 𝑑0 𝑐𝑎0 + 𝑑𝑐 0 𝑐𝑏 0 + 𝑑𝑑 0
1320 Hence there is a group homomorphism
 
𝑎 𝑏 𝑎𝑧 + 𝑏
SL2 (C) −→ M, ↦ → .
𝑐 𝑑 𝑐𝑧 + 𝑑
1321 We have seen in Lecture 23 that every Moebius transformation can be represented by an ele-
1322 ment of SL2 (C). Hence the above group homomorphism is surjective, with kernel {±𝐼 }. The
1323 group
SL2 (C)/{±𝐼 }
1324 is usually written PSL2 (C).
1325 26.4. Proposition. Every element of M = PSL2 (C) is a meromorphic conformal automorphism of C.
b
1326 Conversely, every a meromorphic conformal automorphism of C
b is given by an element of M.

1327 Proof. It is easy to check that elements of M are automorphisms. Conversely, let 𝑓 be an auto-
1328 morphism of C. b If 𝑓 (∞) = ∞, then 𝑓 (𝑧) = 𝑎𝑧 + 𝑏 by Proposition 25.3. If 𝑓 (∞) = 𝑐 ≠ ∞, then
1
1329 let 𝑔(𝑧) = 𝑧−𝑐 . Now 𝑔𝑓 is an automorphism, and fixes ∞, so it is in M. Hence 𝑓 ∈ M. 
1330 Exercises.
1331 (1) Let 𝜁 , 𝜁 0, 𝜁 1, 𝜁 ∞ be distinct points of 𝐶b and 𝜎 a permutation of four symbols. Determine
1332 the relation between the cross ratio (𝜁 , 𝜁 0, 𝜁 1, 𝜁 ∞ ) and (𝜎 (𝜁 , 𝜁 0, 𝜁 1, 𝜁 ∞ )).

1333 Lecture 27. Review of path homotopy

1334 Reference for this section is Munkres, Topology, the chapter on covering spaces and funda-
1335 mental groups.
Let 𝑈 be an open subset of C. Let 𝛾, 𝜂 : [𝑎, 𝑏] −→ 𝑈 be two paths with 𝛾 (𝑎) = 𝜂 (𝑎) and
𝛾 (𝑏) = 𝜂 (𝑏). We say that 𝛾 and 𝜂 are path-homotopic to each other if there exists a continuous
map 𝐻 : [0, 1] × [𝑎, 𝑏] −→ 𝑈 such that
𝐻 (0, 𝑡) = 𝛾 (𝑡) for all 𝑡 ∈ [𝑎, 𝑏];
𝐻 (1, 𝑡) = 𝜂 (𝑡) for all 𝑡 ∈ [𝑎, 𝑏];
𝐻 (𝑠, 0) = 𝛾 (0) = 𝜂 (0) for all 𝑠 ∈ [0, 1];
𝐻 (𝑠, 1) = 𝛾 (1) = 𝜂 (1) for all 𝑠 ∈ [0, 1].
1336 We can think of this as a continuously varying family of paths [𝑎, 𝑏] −→ 𝑈 parameterised
1337 by [0, 1] such that for every member of the family is a path from 𝛾 (0) to 𝛾 (1). We say that 𝐻
1338 is a path-homotopy between 𝛾 and 𝜂. We will say that a closed path 𝛾 is null-homotopic if 𝛾 and
1339 the constant path 𝑒𝛾 (0) at 𝛾 (0) (i.e, the map [𝑎, 𝑏] −→ 𝑈 , 𝑡 ↦→ 𝛾 (0)) are path-homotopic to
1340 each other. Note that being path-homotopic is an equivalence relation; we will refer to the
1341 equivalence classes under this relation as path-homotopy classes.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 48 MANOJ KUMMINI

1342 We say that𝑈 is simply-connected if for every closed path in𝑈 is null-homotopic. For example,
1343 C is simply connected, but C r {0} is not.
The result we want to prove is that if 𝑓 is holomorphic on 𝑈 , then 𝛾 𝑓 d𝑧 depends only on
∫
1344

1345 the path-homotopy class of 𝛾. However, there is (at least) one issue that needs to be sorted out:
1346 even if 𝛾 and 𝜂 are path-homotopic piecewise-differentiable paths, with path homotopy 𝐻 , the
1347 paths∫𝐻 (𝑠, −) need not be piecewise-differentiable for 𝑠 ∈ (0, 1). Hence we need to understand
1348 what 𝜏 𝑓 d𝑧 is, when 𝜏 is merely a continuous path.
1349 Reference for the remainder of this section is Lang, Complex Analysis, Chapter III, Section
1350 4.
1351 27.1. Lemma. Let𝑈 ⊆ C be an open set and𝛾 : [𝑎, 𝑏] −→ 𝑈 a continuous path in𝑈 . Then there exists
1352 𝑟 > 0 such that for every 𝑥 ∈ Im(𝛾) and for every 𝑦 ∈ C r 𝑈 , |𝑥 − 𝑦| > 𝑟 .
1353 Proof. The function 𝛿 : 𝑈 −→ R, 𝑡 ↦→ inf {|𝛾 (𝑡) − 𝑦| : 𝑦 ∈ C r 𝑈 is attained by some 𝑦 ∉ 𝑈 ,
1354 since it suffices to consider 𝑦 lying inside a closed and bounded subset of C. It is continuous:
1355 Let 𝑡𝑛 → 𝑡. let 𝑦, 𝑦𝑛 ∉ 𝑈 be such that 𝛿 (𝑡) = |𝛾 (𝑡) − 𝑦| and 𝛿 (𝑡𝑛 ) = |𝛾 (𝑡𝑛 ) − 𝑦𝑛 |. Let 𝜖 > 0.
1356 Then there exists 𝑁 such that for every 𝑛 > 𝑁 , |𝛾 (𝑡𝑛 ) − 𝛾 (𝑡)| < 𝜖. Hence 𝛿 (𝑡) < 𝛿 (𝑡𝑛 ) + 𝜖 and
1357 𝛿 (𝑡𝑛 ) < 𝛿 (𝑡) + 𝜖, for every 𝑛 > 𝑁 . Hence 𝛿 (𝑡𝑛 ) → 𝛿 (𝑡). Therefore there exists 𝑡 0 ∈ [𝑎, 𝑏] such
1358 that 𝛿 (𝑡 0 ) = inf {𝛿 (𝑡) | 𝑡 ∈ [𝑎, 𝑏]}. Since 𝑈 is open, 𝛿 (𝑡 0 ) > 0. 
1359 27.2. Discussion. For now, assume that 𝛾 is piecewise-differentiable. Let 𝜖 > 0 be small
1360 enough such that 𝐵𝛾 (𝑡),𝜖 ⊆ 𝑈 for every 𝑡 ∈ [𝑎, 𝑏]; such an 𝜖 exists by Lemma 27.1. Since 𝛾
1361 is uniformly continuous, there exists 𝛿 > 0 such that 𝛾 (𝐵𝑡,𝛿 ) ⊆ 𝐵𝛾 (𝑡),𝜖 . Then there exist
1362 (1) a partition 𝑎 = 𝑡 0 < 𝑡 1 < . . . < 𝑡𝑛 = 𝑏 such that
1363 (a) 𝑡𝑖+1 − 𝑡𝑖 < 𝛿;
1364 (b) 𝛾 is differentiable on (𝑡𝑖 , 𝑡𝑖+1 );
1365 (2) a covering of Im(𝛾) by open discs 𝐵𝑖 , 0 ≤ 𝑖 ≤ 𝑛 − 1 such that 𝛾 ([𝑡𝑖 , 𝑡𝑖+1 ]) ⊆ 𝐵𝑖 .
For 0 ≤ 𝑖 ≤ 𝑛 − 1, let 𝑔𝑖 be a primitive of 𝑓 in 𝐵𝑖 . Then
∫ 𝑡𝑖+1
𝑓 (𝛾 (𝑡))𝛾 0 (𝑡)d𝑡 = 𝑔𝑖 (𝛾 (𝑡𝑖+1 )) − 𝑔𝑖 (𝑡𝑖 ), and hence,
𝑡𝑖
∫ 𝑛−1
∑︁
𝑓 d𝑧 = 𝑔𝑖 (𝑡𝑖+1 ) − 𝑔𝑖 (𝑡𝑖 ). 
𝛾 𝑖=0

In view of the discussion above, we can extend the definition of 𝛾 𝑓 d𝑧 to continuous paths
∫
1366

1367 𝛾 as follows. Note that we did not use all the information about the partition, in the above
1368 discussion.
1369 27.3. Definition. Let 𝑈 be a domain and 𝛾 : [𝑎, 𝑏] −→ 𝑈 a continuous path. Let 𝑎 = 𝑡 0 < 𝑡 1 <
1370 . . . < 𝑡𝑛 = 𝑏 be a partition and 𝐵 0, . . . , 𝐵𝑛−1 be open discs in 𝑈 such that 𝛾 ([𝑡𝑖 , 𝑡𝑖+1 ]) ⊆ 𝐵𝑖 for
1371 every 0 ≤ 𝑖 ≤ 𝑛 − 1. Let 𝑔𝑖 be a primitive of 𝑓 on 𝐵𝑖 . Define
∫ 𝑛−1
∑︁
𝑓 d𝑧 = 𝑔𝑖 (𝛾 (𝑡𝑖+1 )) − 𝑔𝑖 (𝛾 (𝑡𝑖 )).
𝛾 𝑖=0

27.4. Proposition. The definition of 𝛾 𝑓 d𝑧 is independent of the choice of the partition 𝑎 = 𝑡 0 < 𝑡 1 <
∫
1372

1373 . . . < 𝑡𝑛 = 𝑏, the open discs 𝐵𝑖 , 0 ≤ 𝑖 ≤ 𝑛 − 1 and the primitives 𝑔𝑖 , 0 ≤ 𝑖 ≤ 𝑛 − 1 of 𝑓 .

Proof. For the sake of clarity, we will write 𝐼 ({𝑡𝑖 }, {𝐵𝑖 }, {𝑔𝑖 }) to denote 𝑛−1
Í
1374 𝑖=0 𝑔𝑖 (𝛾 (𝑡𝑖+1 ))−𝑔𝑖 (𝛾 (𝑡𝑖 ))
1375 ˜
in Definition 27.3. Let {𝑡˜𝑖 }, {𝐵𝑖 } and {𝑔˜𝑖 } another set of choices.

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 COMPLEX ANALYSIS UG JAN-APR 2021. NOTES 49

1376 Step 1: Assume that {𝑡˜𝑖 } = {𝑡𝑖 }. Then 𝛾 ([𝑡𝑖 , 𝑡𝑖+1 ]) ⊆ 𝐵𝑖 ∩ 𝐵˜𝑖 , on which 𝑔𝑖 and 𝑔˜𝑖 are primitives
1377 of 𝑓 . Then there exists 𝑐𝑖 ∈ C such that 𝑔𝑖 (𝑧) − 𝑔˜𝑖 (𝑧) = 𝑐𝑖 for every 𝑧 ∈ 𝐵𝑖 ∩ 𝐵˜𝑖 . Hence

𝑔𝑖 (𝛾 (𝑡𝑖+1 )) − 𝑔𝑖 (𝛾 (𝑡𝑖 )) = 𝑔˜𝑖 (𝛾 (𝑡𝑖+1 )) − 𝑔˜𝑖 (𝛾 (𝑡𝑖 ))

1378 so 𝐼 ({𝑡𝑖 }, {𝐵𝑖 }, {𝑔𝑖 }) = 𝐼 ({𝑡𝑖 }, {𝐵˜𝑖 }, {𝑔˜𝑖 }).

1379 Step 2: Assume that {𝑡˜𝑖 } is a refinement of {𝑡𝑖 }. Then the covering {𝐵𝑖 } and the primitives
1380 {𝑔𝑖 } (which were defined for {𝑡𝑖 }) induced a covering and primitives with respect to {𝑡˜𝑖 }. More
1381 precisely, if 𝑡˜𝑗𝑖 = 𝑡𝑖 and 𝑡˜𝑗𝑖+1 = 𝑡𝑖+1 , then use 𝐵𝑖 and 𝑔𝑖 for the intervals [𝑡˜𝑗 , 𝑡˜𝑗+1 ], 𝑗𝑖 ≤ 𝑗 < 𝑡 𝑗𝑖+1 . We
1382 abuse notation and continue to use {𝐵𝑖 } and {𝑔𝑖 } for the induced covering and primitives. It is
1383 easy to see that 𝐼 ({𝑡𝑖 }, {𝐵𝑖 }, {𝑔𝑖 }) = 𝐼 ({𝑡˜𝑖 }, {𝐵𝑖 }, {𝑔𝑖 }). By the earlier case, 𝐼 ({𝑡˜𝑖 }, {𝐵𝑖 }, {𝑔𝑖 }) =
1384 𝐼 ({𝑡˜𝑖 }, {𝐵˜𝑖 }, {𝑔˜𝑖 }).
Step 3: Consider the general case. Let {𝑡ˆ𝑖 } be a common refinement of the {𝑡𝑖 } and the {𝑡˜𝑖 }.
As in Step 2, the covering {𝐵𝑖 } and the primitives {𝑔𝑖 } induce a covering and primitives on {𝑡ˆ𝑖 },
which we abuse notation and denote by {𝐵𝑖 } and {𝑔𝑖 }. Similarly, we get {𝐵˜𝑖 } and {𝑔˜𝑖 } from the
partition {𝑡˜𝑖 }. Then

𝐼 ({𝑡𝑖 }, {𝐵𝑖 }, {𝑔𝑖 }) = 𝐼 ({𝑡ˆ𝑖 }, {𝐵𝑖 }, {𝑔𝑖 })

= 𝐼 ({𝑡ˆ𝑖 }, {𝐵˜𝑖 }, {𝑔˜𝑖 })
= 𝐼 ({𝑡˜𝑖 }, {𝐵˜𝑖 }, {𝑔˜𝑖 })

1385 where the first and the third equalities follow from Step 2 and the second one from Step 1. 

1386 We emphasise that order to make sense of Definition 27.3, we need to know that holomor-
1387 phic functions on discs have primitives (Theorem 11.1).

1388 Exercises.

1389 Lecture 28. General version of Cauchy integral theorem.

1390 Let 𝑈 be a domain.
1391 The following lemma is Lang, Complex Analysis, Chapter III, Section 4, Lemma 4.3.

1392 28.1. Lemma. Let 𝛾 and 𝜂 be two continuous paths [𝑎, 𝑏] −→ 𝑈 . Assume that there exist a partition
1393 𝑎 = 𝑡 0 < 𝑡 1 < . . . < 𝑡𝑛 = 𝑏 and open discs 𝐵 0, . . . , 𝐵𝑛−1 in 𝑈 such that 𝛾 ([𝑡𝑖 , 𝑡𝑖+1 ]) ⊆ 𝐵𝑖 and
1394 𝜂 ( [𝑡𝑖 , 𝑡𝑖+1 ]) ⊆ 𝐵𝑖 for every 0 ≤ 𝑖 ≤ 𝑛 − 1. Then
∫ ∫
𝑓 d𝑧 = 𝑓 d𝑧.
𝛾 𝜂

1395 Proof. Write 𝑧𝑖 = 𝛾 (𝑡𝑖 ) and 𝑤𝑖 = 𝜂 (𝑡𝑖 ). Let 𝑔𝑖 be a primitive of 𝑓 on 𝐵𝑖 , 0 ≤ 𝑖 ≤ 𝑛 − 1. Since 𝑔𝑖+1
1396 and 𝑔𝑖 are primitives of 𝑓 on 𝐵𝑖+1 ∩ 𝐵𝑖 , the function 𝑔𝑖+1 − 𝑔𝑖 is constant on 𝐵𝑖+1 ∩ 𝐵𝑖 . It follows
1397 that
𝑔𝑖+1 (𝑧𝑖+1 ) − 𝑔𝑖 (𝑧𝑖+1 ) = 𝑔𝑖+1 (𝑤𝑖+1 ) − 𝑔𝑖 (𝑤𝑖+1 )

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)

 50 MANOJ KUMMINI

for all 0 ≤ 𝑖 ≤ 𝑛 − 1. Hence

∫ ∫ 𝑛−1
∑︁
𝑓 d𝑧 − 𝑓 d𝑧 = [𝑔𝑖 (𝑧𝑖+1 ) − 𝑔𝑖 (𝑧𝑖 ))] − [𝑔𝑖 (𝑤𝑖+1 ) − 𝑔𝑖 (𝑤𝑖 ))]
𝛾 𝜂 𝑖=0
𝑛−1
∑︁
= [𝑔𝑖+1 (𝑧𝑖+1 ) − 𝑔𝑖 (𝑧𝑖 ))] − [𝑔𝑖+1 (𝑤𝑖+1 ) − 𝑔𝑖 (𝑤𝑖 ))]
𝑖=0
= [𝑔𝑛 (𝑧𝑛 ) − 𝑔0 (𝑧 0 )] − [𝑔𝑛 (𝑤𝑛 ) − 𝑔0 (𝑤 0 )]
=0
1398 since 𝑧 0 = 𝑤 0 and 𝑧𝑛 = 𝑤𝑛 . 
1399 28.2. Theorem. Let 𝛾 and 𝜂 be path-homotopic continuous paths [𝑎, 𝑏] −→ 𝑈 . Let 𝑓 be holomorphic
1400 on 𝑈 . Then ∫ ∫
𝑓 d𝑧 = 𝑓 d𝑧.
𝛾 𝜂

1401 Proof. Let 𝐻 = [0, 1] × [𝑎, 𝑏] −→ 𝑈 be a path homotopy. Since Im(𝐻 ) is compact, there exists
1402 𝑟 > 0 such that for every 𝑥 ∈ Im(𝐻 ) and for every 𝑦 ∈ C r 𝑈 , |𝑥 − 𝑦| > 𝑟 , as in Lemma 27.1.
1403 Hence there exists 𝜖 such that 𝐵𝑥,𝜖 ⊆ 𝑈 for every 𝑥 ∈ Im(𝐻 ). Since 𝐻 is uniformly continuous,
1404 there exists 𝛿 > 0 such that for every 𝑝 ∈ [0, 1] × [𝑎, 𝑏], 𝐻 (𝐵𝑝,𝛿 ) ⊆ 𝐵𝐻 (𝑝),𝜖 . Hence there exist
1405 partitions 0 = 𝑠 0 < 𝑠 1 < . . . < 𝑠𝑚 = 1 and 𝑎 = 𝑡 0 < 𝑡 1 < . . . < 𝑡𝑛 = 𝑏 such that for each 𝑖, 𝑗,
1406 there exists an open disk 𝐵𝑖,𝑗 such that 𝐻 ([𝑠𝑖 , 𝑠𝑖+1 ] × [𝑡 𝑗 , 𝑡 𝑗+1 ]) ⊆ 𝐵𝑖,𝑗 ⊆ 𝑈 . (For example, choose
1407 the 𝑠𝑖 and the 𝑡 𝑗 such that the diagonal of the rectangle [𝑠𝑖 , 𝑠𝑖+1 ] × [𝑡 𝑗 , 𝑡 𝑗+1 ] has length at most
1408 2𝛿.) For 𝑖 = 0, . . . , 𝑚, define paths 𝛾𝑖 : [𝑎, 𝑏] −→ 𝑈 by 𝛾𝑖 (𝑡) = 𝐻 (𝑠𝑖∫, 𝑡). Note that ∫ 𝛾 0 = 𝛾 and
1409 𝛾𝑚 = 𝜂. We now induct on 𝑖 and apply the lemma to conclude that 𝛾 𝑓 d𝑧 = 𝛾 𝑓 d𝑧 for every
𝑖
1410 1 ≤ 𝑖 ≤ 𝑚. 
28.3. Corollary. If𝑈 is simply connected, then 𝛾 𝑓 d𝑧 = 0 for every holomorphic 𝑓 on𝑈 and every closed
∫
1411

1412 path 𝛾.
1413 Exercises.
1414 (1) Let 𝑈 ⊆ C be a bounded domain. Show that it is simply connected if and only if C r 𝑈
1415 is connected.
1416 (2) Let 𝛾 be a closed path in C, not passing through 0. Assume further that there exists a
1417 ray through the origin {𝑟𝜁 | 𝑟 ∈ R, 𝑟 > 0, 𝜁 ∈ C, 𝜁 ≠ 0} that does not intersect Im(𝛾).
1418 Find a simply connected
∫ 1 domain 𝑈 containing 𝛾 that admits a branch of the logarithm.
1419 Conclude that 𝛾 𝑧 d𝑧 = 0.
1420 (3) Let 𝑈 be a domain. Show that 𝑈 is simply connected if and only if for every closed path
1421 𝛾 in 𝑈 and every 𝜁 ∉ 𝑈 , 𝑛(𝜁 , 𝛾) = 0.
1422 Exercises
1423 References

1424 Chennai Mathematical Institute, Siruseri, Tamilnadu 603103. India

1425 Email address: mkummini@cmi.ac.in

[git] ˆ Branch: master @ a933e41 ˆ Release: (2021-04-20)