MATH8811: COMPLEX ANALYSIS

DAWEI CHEN

Contents
1. Classical Topics 2
1.1. Complex numbers 2
1.2. Differentiability 2
1.3. Cauchy-Riemann Equations 3
1.4. The Riemann Sphere 4
1.5. Möbius transformations 5
1.6. Integration 6
1.7. Cauchy’s Theorems and Applications 7
1.8. Properties of Analytic Functions 9
1.9. The Winding Number 11
1.10. Singularities 12
1.11. Laurent Series 13
1.12. Residue Theory 15
1.13. Normal Families 18
1.14. The Riemann Mapping Theorem 19
1.15. Schwarz Reflection 21
1.16. Analytic Continuation 23
1.17. Analytic Covering Maps 24
1.18. The Picard Theorems 25
1.19. Infinite Products 27
2. Riemann Surfaces 34
2.1. Preliminaries 35
2.2. Sheaves 37
2.3. Maps between sheaves 38
2.4. Stalks and germs 39
2.5. Sheaf cohomology 40
2.6. Holomorphic vector bundles 45
2.7. Vector bundles and locally free sheaves 46
2.8. Divisors 46
2.9. Line bundles 48
2.10. Sections of a line bundle 50
2.11. The Riemann-Roch formula 52
2.12. The Riemann-Hurwitz formula 53
2.13. Genus formula of plane curves 55
2.14. Base point free and very ample line bundles 56

Date: Spring 2020.

Most of the material is from [GH, T, S].
1
2

2.15. Canonical maps 59

2.16. Dimension of linear systems 60
References 62

1. Classical Topics
1.1. Complex numbers.√ A complex number is expressed as z = x + iy, where
x, y ∈ R and i2 = −1. We use C to denote the set of complex numbers, which
geometrically corresponds to the (x, y)-coordinate plane R2 . Define the conjugate
of z to be z = x − iy, and the norm of z satisfies |z|2 = zz = x2 + y 2 .
If z 6= 0, z can also be represented by the polar coordinate z = r(cos θ + i sin θ),
where r = |z|, cos θ = x/r and sin θ = y/r. By comparing power series expansions,
we conclude the Euler formula
eiθ = cos θ + i sin θ
for all θ ∈ R. In this form z = reiθ , which is convenient for multiplication:
z1 z2 = |z1 ||z2 |ei(θ1 +θ2 ) .
Such θ is called the argument of z, which is well-defined modulo multiples of 2π.
Note that z = e(log r)+iθ (where log r is the real log function), hence one can
define log z = log r + iθ, which is a multi-valued function
√ since θ and θ + 2πn for
n ∈ Z are both arguments of z. Similarly for z 6= 0, n z has n different values,
given by
r1/n ei(θ+2kπ)/n
for 0 ≤ k ≤ n − 1.
1.2. Differentiability. Let U be an open subset of R2 . Let f (z) : U → C be a
complex-valued function. We may also write
f (z) = u(z) + iv(z) = u(x, y) + iv(x, y),
2
where u, v : R → R. Recall the difference-quotient definition of derivative
f (w) − f (z)
f 0 (z) = lim .
w→z w−z
Definition 1.1. If f 0 (z) exists and continuous for all z ∈ U , we say that f is
holomorphic on U , and denote it by f ∈ H(U ). A function f ∈ H(C) is called
entire.
Example 1.2. Polynomials and ez are entire functions. A rational function (quo-
tient of two polynomials) is holomorphic away from the zeros of its denominator.
Definition 1.3. We say that f is analytic if f can be represented by a convergent
power series expansion on a neighborhood of every point in U , and we denote it by
f ∈ A(U ).
Later on we will show that H(U ) = A(U ) and the condition that f 0 (z) is con-
tinuous can be dropped from the definition of being holomorphic. Hence one can
freely interchange the notion of being analytic and holomorphic. Below we verify
one direction of their equivalence.
Lemma 1.4. In the above setting, we have A(U ) ⊂ H(U ).
 3

Proof. Suppose z0 ∈ U . For 0 < r  1 (e.g. being the radius of convergence),

∞
X
f (z) = an (z − z0 )n
n=0

converges absolutely and uniformly on (compact subsets of) the disk |z − z0 | < r,
so does the formal derivative series
X∞
nan (z − z0 )n−1 .
n=0
As in calculus, under these conditions differentiation is interchangeable with sum-
mation, hence
∞
X
f 0 (z) = nan (z − z0 )n−1 ,
n=0
which is convergent and continuous for |z − z0 | < r.
By the same argument we have
∞
X
f (k) (z) = n(n − 1) · · · (n − k + 1)an (z − z0 )n−k
n=0
f (n) (z0 )
for |z − z0 | < r, which shows that an = n! for all n ≥ 0. 
1.3. Cauchy-Riemann Equations. Abuse notation slightly by writing f (z) =
f (x + iy) = f (x, y). If f 0 (z) exists for some z = x + iy ∈ U , then one can approach
z along the x or y directions, respectively, and obtain that
f (z + h) − f (z) f (x + h, y) − f (x, y) ∂f
f 0 (z) = lim = lim = ,
h∈R
h→0
h h∈R
h→0
h ∂x
f (z + ih) − f (z) f (x, y + h) − f (x, y) ∂f
f 0 (z) = lim = lim −i = −i .
h∈R
h→0
ih h∈R
h→0
h ∂y
Hence we conclude one version of the Cauchy-Riemann equation
∂f ∂f
= −i .
∂x ∂y
If we write f (x, y) = u(x, y) + iv(x, y), then similarly we obtain that
f 0 (z) = ux + ivx = −iuy + vy ,
∂u
where ux = ∂x , etc. The Cauchy-Riemann equation is equivalent to the following
equations
ux = vy , uy = −vx .
If we treat f as a function from R2 to R2 , then the Jacobian matrix of f is
 
u uy
df = x .
vx vy
In this setting the Cauchy-Riemann equation is equivalent to the property that
df = ρA
for some ρ ≥ 0 and A ∈ SO(2, R). If ρ 6= 0, i.e. if f 0 (z) 6= 0, then df is the
composition of rotation by A and dilation by ρ, hence as a map U → R2 , f preserves
angles between smooth arcs through z and orientation. In general, a map preserving
4

angle and orientation at each point is called conformal. We thus conclude that for
f ∈ H(U ), if f 0 (z) 6= 0 for all z ∈ U , then f is conformal on U .
Conversely if the Cauchy-Riemann equations hold for f = u + iv, we can write
u(x + s, y + t) = u(x, y) + ux (x, y)s + uy (x, y)t + o(|s|, |t|),
v(x + s, y + t) = v(x, y) + vx (x, y)s + vy (x, y)t + o(|s|, |t|).
Using ux = vy , uy = −vx and writing w = s + it, we obtain
f (z + w) − f (z) = (u(x + s, y + t) − u(x, y)) + i(v(x + s, y + t) − v(x, y))
= ux (x, y)s + uy (x, y)t + i(vx (x, y)s + vy (x, y)t) + o(|w|)
= ux (x, y)s − vx (x, y)t + i(vx (x, y)s + ux (x, y)t) + o(|w|)
= (ux (x, y) + ivx (x, y))(s + it) + o(|w|)
= (ux (z) + ivx (z))w + o(|w|).
0
It follows that f (z) exists and equals ux (z) + ivx (z) = vy (z) − iuy (z) as desired.
We have thus proved the following result.
Theorem 1.5. A complex-valued function f (with continuous partials ux , uy , vx , vy )
is holomorphic iff it satisfies the Cauchy-Riemann equation.
Another way to record the Cauchy-Riemann equation is as follows. Treat z, z as
independent variables and write x = (z + z)/2, y = (z − z)/(2i). Then we have
 
∂f ∂f ∂x ∂f ∂y 1 ∂f ∂f
= + = −i ,
∂z ∂x ∂z ∂y ∂z 2 ∂x ∂y
 
∂f ∂f ∂x ∂f ∂y 1 ∂f ∂f
= + = +i .
∂z ∂x ∂z ∂y ∂z 2 ∂x ∂y
Then the Cauchy-Riemann equation is equivalent to
∂f
= 0,
∂z
which is also equivalent to
∂f ∂f
= .
∂z ∂x
Example 1.6. f (z) = z does not satisfy the Cauchy-Riemann equation, hence it
is not holomorphic.
1.4. The Riemann Sphere. Let C∞ be the one-point compactification of C, by
adding a point at ∞. The neighborhoods of ∞ are the complements of compact
sets in C. It is clear that C∞ is homeomorphic to the 2-sphere S 2 .
A complex structure can be given to C∞ such that it becomes a complex one-
dimensional manifold, i.e., a Riemann surface. Cover C∞ by two charts (C, z) and
(C∞ − {0}, z −1 ), both of which are isomorphic to C. On the overlap C∗ = C − {0},
the transition function is given by the map z 7→ z −1 , which is biholomorphic.
This description is equivalent to saying that C∞ is the complex projective line
CP1 . By definition, CP1 = (C2 − {(0, 0)})/C∗ , with homogeneous coordinates
{[x, y]}, where x, y ∈ C, not both zero, and [x, y] represents the equivalence class
of (x, y) modulo simultaneous scaling. Note that CP1 can be covered by two charts
U = {[x, y]} with y 6= 0 and V = {[x, y]} with x 6= 0. Then the coordinate on U
is u = x/y and on V is v = y/x, with the transition function v = 1/u, which is
exactly the same as above.
 5

From now on we will use C∞ , S 2 , or CP1 interchangeably for the notation of the
Riemann sphere.
Using the above charts, one can extend the notion of holomorphic maps from
C → C to C∞ → C∞ . There are three cases. If f (z0 ) = ∞ for some z0 ∈ C, then we
require that 1/f (z) is holomorphic around z0 . To make sense of being holomorphic
at z = ∞, if f (∞) 6= ∞, we require that f (1/z) is holomorphic around z = 0.
Finally if f (∞) = ∞, we require that 1/f (1/z) is holomorphic around z = 0.
Example 1.7. Any rational function on C extends to a holomorphic map from
C∞ → C∞ .
1.5. Möbius transformations. For a matrix
 
a b
A= ∈ GL(2, C),
c d
define the Möbius transformation associated to A by
az + b
TA (z) =
cz + d
as a map from C∞ → C∞ .
It is easy to check that TA ◦ TB = TA·B and TA−1 = TA−1 . In particular, TA is
biholomorphic and is thus an automorphism of C∞ .
Note that if A e = λA for λ ∈ C∗ , then T e = TA . Conversely if TA = T e for
A A
A, Ae ∈ SL(2, C), then

ad − bc ade − ebe
c
TA0 (z) = = TA0e(z) =
e
2
,
(cz + d) cz + d)2
(e e

hence cz + d = ±(e cz + d)
e since the numerators are 1 by assumption. It follows
that A and Ae are the same matrices in SL(2, C) possibly up to a choice of sign.
Therefore, we conclude the following.
Proposition 1.8. The family of all Möbius transformations is the same as PSL(2, C) =
SL(2, C)/{± Id}.
Lemma 1.9. Every Möbius transformation is the composition of four elementary
maps:
(1) Translation z 7→ z + z0 .
(2) Dilation z 7→ λz, λ > 0.
(3) Rotation z 7→ eiθ z, θ ∈ R.
(4) Inversion z 7→ 1/z.
Proof. If c = 0, then
a b
z+ ,
TA (z) =
d d
which is the composition of translation, dilation and possibly rotation.
If c 6= 0, then
bc − ad 1 a
TA (z) = 2 d
+ .
c z+ c c

Example 1.10. The Möbius transformation z 7→ z−1 z+1 takes the right half-plane
onto the unit disk. In particular, the imaginary axis iR maps onto the unit circle.
6

If we include all lines in R2 as circles (of infinite radius), then it is not hard to
verify the following.
Lemma 1.11. Möbius transformations take circles onto circles.
Proof. This can be checked using the four elementary maps in the composition of
a Möbius transformation. 
Note that the equation T (z) = z is (at most) a quadratic equation for any
Möbius transformation T . Hence T can have at most two fixed pointed unless it is
the identity. For instance, z 7→ z + 1 has only one fixed point at ∞, while z 7→ 1/z
has two fixed points at ±1.
Lemma 1.12. A Möbius transformation is determined completely by its action on
three distinct points on C∞ . Moreover, given distinct z1 , z2 , z3 ∈ C∞ , there exists a
unique Möbius transformation T such that T (z1 ) = 0, T (z2 ) = 1, and T (z3 ) = ∞.
Proof. If S and T are two Möbius transformations that agree on three distinct
points, then S · T −1 has at least three fixed points, hence S · T −1 is the identity
transformation, and S = T .
For the other statement, one can first apply Möbius transformations to assume
that zi are not ∞. Then define
z − z1 z2 − z3
T (z) = · ,
z − z3 z2 − z1
which does the job. 
Definition 1.13. The cross ratio of four distinct points z1 , z2 , z3 , z4 ∈ C∞ is
defined as
z1 − z3 z1 − z4 z1 − z3 z2 − z4
(z1 , z2 , z3 , z4 ) = : = · .
z2 − z3 z2 − z4 z2 − z3 z1 − z4
Lemma 1.14. The cross ratio is preserved under Möbius transformations. More-
over, four distinct points lie on a circle iff their cross ratio is real.
Proof. For a Möbius transformation T , let wj = T (zj ) for distinct zj for j =
2, 3, 4. Consider the cross ratios S1 (z) = (z, z2 , z3 , z4 ) and S2 ◦ T (z) = S2 (w) =
(w = T (z), w2 , w3 , w4 ) as Möbius transformations. Since S1 (z2 ) = 1, S1 (z3 ) = 0,
S1 (z4 ) = ∞, S2 ◦ T (z2 ) = S2 (w2 ) = 1, S2 (w3 ) = 0, S2 (w4 ) = ∞, we conclude that
S1 = S2 ◦ T , hence
(z1 , z2 , z3 , z4 ) = S1 (z1 ) = S2 ◦ T (z1 ) = (T (z1 ), T (z2 ), T (z3 ), T (z4 )).
for four distinct z1 , z2 , z3 , z4 ∈ C∞ . The other statement follows from taking four
points in the real axis (as a generalized circle). 
1.6. Integration. Recall that a smooth path is a C 1 -curve γ : [0, 1] → R2 , i.e., γ
is differentiable with continuous derivatives. In many cases we will also consider
piecewise smooth paths. If γ(0) = γ(1), we say that γ is closed.
Definition 1.15. For any complex-valued continuous function f on U ⊂ C and a
path γ in U , define
Z Z 1
f (z)dz = f (γ(t))γ 0 (t)dt.
γ 0
H
If γ is closed, then we also write γ f (z)dz for this integral.
 7

By the chain rule, the above definition is independent of parameterizations of γ

(as long as the orientation is preserved).
Example 1.16. Suppose f ∈ H(U ). Then
Z Z 1 Z 1
d
f 0 (z)dz = f 0 (γ(t))γ 0 (t)dt = f (γ(t))dt = f (γ(1)) − f (γ(0)).
γ 0 0 dt
In particular if γ is closed, then
I
f 0 (z)dz = 0.
γ

Example 1.17. For n 6= −1, z has an antiderivative (z n+1 )/(n + 1), hence for
n

any closed path γ ∈ C we have

I
z n dz = 0.
γ

If n = −1, consider the circle parameterized by γ(t) = re2πit for t ∈ [0, 1]. Then
I Z 1 Z 1
1 1 2πit 0
dz = 2πit
(re ) dt = 2πi dt = 2πi.
γ z 0 re 0

1.7. Cauchy’s Theorems and Applications. The next few results form a corner
stone of complex integration.
Theorem 1.18 (Cauchy’s Integration Theorem). Let γ1 and γ2 be two C 1 -paths
with same endpoints such that they are C 1 -homotopic in U . Then for any f ∈ H(U ),
we have Z Z
f (z)dz = f (z)dz.
γ1 γ2

Proof. We can triangulate the homotopy region into a union of small subregions
with boundary loops ηj such that
Z Z XI
f (z)dz − f (z)dz = f (z)dz.
γ1 γ2 j ηj
H
Hence it suffices to prove that η f (z)dz = 0 for sufficiently small loops η.
Recall that d(f (z)dz) = 0, hence f (z)dz is a closed form. By Poincaré’s lemma,
0
any closed form is locally exact,
H hence on the region bounded by η, f (z)dz = F (z)
for some F , and consequently η f (z)dz = 0.
Alternatively, write f (z) = u(x, y) + iv(x, y). Let V be the region with (positive
oriented) boundary curve η. Then
I I
f (z)dz = f (z)dz
η ∂V
I
= (u + iv)dx + (iu − v)dy
∂V
ZZ
= (−uy − ivy + iux − vx )dxdy
V
= 0,
where we used Green’s theorem and the Cauchy-Riemann equations in the last two
steps. 
8

Denote by D(z, r) the open disk centered at z of radius r.

Theorem 1.19 (Cauchy’s Integration Formula). Let D(z0 , r) ⊂ U and f ∈ H(U ).
Then I
1 f (w)
f (z) = dw
2πi γ w − z
for all z ∈ D(z0 , r), where γ(t) = z0 + re2πit is the boundary of D(z0 , r).
Proof. Fix z ∈ D(z0 , r) and consider the region Uε = D(z0 , r) − D(z, ε) for small ε.
Then the boundary of Uε consists of two circles that are homotopic in U . Applying
Cauchy’s Integration Theorem, we have
I I
1 f (w) 1 f (w)
0 = dw − dw
2πi ∂D(z0 ,r) w − z 2πi ∂D(z,ε) w − z
f (w) − f (z)
I I I
1 f (w) 1 f (z) 1
= dw − dw − dw
2πi ∂D(z0 ,r) w − z 2πi ∂D(z,ε) w−z 2πi ∂D(z,ε) w − z
I
1 f (w)
= dw + O(ε) − f (z),
2πi ∂D(z0 ,r) w − z
where we used f (w)−f (z)
is close to f 0 (z) and ∂D(z,ε) w−z1
H
w−z dw = 2πi. Now letting
ε → 0, we thus conclude the desired formula. 
Corollary 1.20. Every f ∈ H(U ) can be represented by a convergent power series
on D(z0 , r) for z0 ∈ U , where r = dist(z0 , ∂U ). Hence A(U ) = H(U ). Moreover,
the nth derivatives of f exist and they are all analytic on U .
Proof. Previously we showed that A(U ) ⊂ H(U ). Conversely for f ∈ H(U ), let γ
be a circle in U centered at z0 that encloses a point z in the interior. By Cauchy’s
Integration Formula, we have
I
1 f (w)
f (z) = dw
2πi γ w − z
I
1 f (w)
= dw
2πi γ (w − z0 ) − (z − z0 )
∞  n
f (w) X z − z0
I
1
= dw
2πi γ w − z0 n=0 w − z0
∞  I 
X 1 f (w)
= dw (z − z0 )n
n=0
2πi γ (w − z0 )n+1
where the interchange of summation and integration is due to absolute and uniform
convergence of the geometric series. Setting
I
n! f (w)
f (n) (z0 ) = dw
2πi γ (w − z0 )n+1
for any n ≥ 0, we thus conclude that
∞
X f (n) (z0 )
f (z) = (z − z0 )n
n=0
n!
converges on D(z0 , r) and f is analytic, with its nth derivative f (n) (z0 ) given by
the above expression, hence f (n) ∈ A(U ) for all n. 
 9

Corollary 1.21 (Cauchy’s Estimates). Let f ∈ H(U ) with |f (z)| ≤ M on U . Then

M n!
|f (n) (z0 )| ≤
dist(z0 , ∂U )n
for all n ≥ 0 and z0 ∈ U .
Proof. Let r = dist(z0 , ∂U ) (or close to). By the previous proof, we have
Z 1
n! f (z0 + re2πit )
|f (n) (z0 )| = r(e2πit )0 dt
2πi 0 rn+1 e2(n+1)πit
Z 1
n! f (z0 + re2πit )
= n
dt
r 0 e2nπit
n! 1 f (z0 + re2πit )
Z
≤ dt
rn 0 e2nπit
Z 1
M n!
≤ dt
rn 0
M n!
= .
rn


1.8. Properties of Analytic Functions.

Theorem 1.22 (Liouville’s Theorem). If f is a bounded entire function, then f is
constant.
More generally, if |f (z)| ≤ C(1 + |z|N ) for all z ∈ C with a fixed constant C and
integer N ≥ 0, then f is a polynomial of degree at most N .
Proof. For the first statement, applying Cauchy’s estimates, |f 0 (z)| ≤ C/r for a
constant C and arbitrarily large r. Hence f 0 (z) = 0 for all z, and f is constant.
For the other statement, take U = D(z, R). We have
C(1 + RN )(N + 1)!
|f (N +1) (z)| ≤ .
RN +1
As R → ∞, we conclude that f (N +1) (z) = 0 for all z, hence f is a polynomial of
degree at most N . 

Theorem 1.23 (Fundamental Theorem of Algebra). Every non-constant complex

polynomial has at least one zero in C.
Proof. Suppose p(z) is a nowhere vanishing, non-constant polynomial. Define
f (z) = 1/p(z). Then f ∈ H(C). As z → ∞, |p(z)| → ∞, hence |f (z)| → 0,
and f is bounded on C. By Liouville’s Theorem, f is constant, leading to a con-
tradiction. 

Theorem 1.24 (Uniqueness Theorem). Let f ∈ H(U ). Then the following are
equivalent:
(1) f ≡ 0.
(2) For some z0 ∈ U , f (n) (z0 ) = 0 for all n ≥ 0.
(3) The set {z ∈ U | f (z) = 0} has an accumulation point in U .
In particular, a non-constant holomorphic function has isolated zeros.
10

Proof. The implication (1) =⇒ (2), (3) is clear. Using the power series expansion of
f at z0 , we see that (2) =⇒ (1). Finally suppose (3) holds. Let {zn } be a sequence
of zeros of f in U that converges to z0 as n → ∞. Prove by contradiction. Suppose
N is the minimal nonnegative integer such that f (N ) (z0 ) 6= 0. Then
∞
X
f (z) = an (z − z0 )n = aN (z − z0 )N (1 + O(z − z0 ))
n=0
as z → z0 , where aN 6= 0. It implies that f (z) 6= 0 in a small punctured neighbor-
hood of z0 , contradicting that z0 is a limit of zeros of f . Hence we conclude that
(3) =⇒ (2). 
Theorem 1.25 (Open Mapping Theorem). For a non-constant function f ∈ H(U )
and every z0 ∈ U , there exists a positive integer n and a holomorphic function g
locally at z0 such that g(z0 ) 6= 0 and
f (z) = f (z0 ) + (z − z0 )n g(z).
We say that z0 is a zero of order n. In particular, f maps open sets to open sets.
Proof. Take n ≥ 1 to be the first nonzero derivative for f (n) (z0 ) and then we get
the desired expression. Since g(z0 ) 6= 0, locally around z0 one can take a single-
valued holomorphic nth root (not unique) h of g, i.e., g(z) = (h(z))n . Moreover, the
derivative of (z−z0 )h(z) is nonzero at z0 , hence (z−z0 )h(z) is a local diffeomorphism
by the usual inverse function theorem. Then f (z) − f (z0 ) is the composition of this
diffeomorphism with w 7→ wn , which implies that f maps open sets to open sets.
In fact, this proof shows that around a zero z0 of order n, f behaves like a branched
cover of degree n totally ramified at z0 (to be discussed later). 
Theorem 1.26 (Maximum Modulus Principle). Let U ⊂ C be a connected open
set and f ∈ H(U ). If there exists z0 ∈ U such that |f (z)| ≤ |f (z0 )| for all z ∈ U ,
then f is constant.
Proof. If f is not constant, then f maps U to an open set, hence f (z0 ) lies in the
interior of f (U ), contradicting that f has maximum modulus at z0 . 
Theorem
H 1.27 (Morera’s Theorem). Let f be a continuous function on U . If
γ
f = 0 for all closed paths γ in U , then f ∈ A(U ).
Proof. Fix z0 ∈ U . Define Z z
g(z) = f (w)dw
z0
which is independent of the integration path by the assumption on f . Since g 0 (z) =
f (z), g ∈ H(U ) = A(U ), hence all derivatives of g are analytic on U , including f
as the first derivative of g. 
Theorem
R 1.28. If f is continuous on U , then f admits an antiderivative on U iff
γ
f = 0 for all closed paths γ in U .
Proof. If F 0 (z) = f (z) on U , then γ f = 0 by the Fundamental Theorem of Cal-
R
R Rz
culus. Conversely if γ f = 0 for all closed paths γ in U , define F (z) = z0 f (w)dw
for a fixed z0 ∈ U . Then F (z) is well-defined and F 0 (z) = f (z). 
Finally we show that the condition of continuous derivatives can be dropped
from the definition of holomorphic functions.
 11

Theorem 1.29 (Goursat’s Theorem). Let f be a complex differentiable function

on U . Then f ∈ H(U ).
H
Proof. By Morera’s theorem, it suffices to show that γ f (z)dz = 0 for all closed
paths γ in U . Such an integration can be approximated
H by summations of integrals
over rectangles, hence it suffices to prove γ f (z)dz = 0 for γ being the boundary
(i)
of a rectangle R. Divide R into four rectangles R1 of equal size for i = 1, 2, 3, 4.
Then I XI
f (z)dz = f (z)dz,
(i)
∂R i ∂R1
(i) (1)
and there exists R1 , say R1 , such that
I I
f (z)dz ≤ 4 · f (z)dz .
(1)
∂R ∂R1
(1)
Now divide R1 and repeat this process. We obtain a sequence of rectangles
R ⊃ R1 ⊃ R2 ⊃ · · ·
(where we omit the upper scripts), such that
I I
n
f (z)dz ≤ 4 · f (z)dz
∂R ∂Rn

for all n. Note that |∂Rn | = 2−n |∂R|.

Let z0 be the limit point of Rn as n → ∞. For z close to z0 , by assumption
|(f (z) − f (z0 )) − f 0 (z0 )(z − z0 )| ≤ ε|z − z0 |
H
for ε  1. Since ∂Rn
(z − z0 )dz = 0, it follows that
I I
f (z)dz ≤ ε |z − z0 |dz ≤ ε|∂Rn |2 = ε4−n |∂R|2
∂Rn ∂Rn
where we used |z − z0 | ≤ |∂Rn | for z on the boundary of Rn and z0 inside Rn . We
thus conclude that I
f (z)dz ≤ ε|∂R|2 ,
∂R
hence it must be zero as ε → 0. 
1.9. The Winding Number. Recall that
I
1 1
1= dz
2πi γ z − z0
for any circle γ centered at z0 (or more generally any closed path homotopic to
γ without passing through z0 ). Such closed paths look like going around z0 once.
This motivates the following definition-theorem.
Theorem 1.30 (Winding Number). Let γ : [0, 1] → C be a closed path. Then for
any z0 not contained in the image of γ, the integral
I
1 1
n(γ; z0 ) = dz
2πi γ z − z0
is an integer, called the winding number or index of γ relative to z0 . It is constant
on each connected component of C − γ and is zero on the unbounded component
(that includes ∞).
12

Proof. Define
t
γ 0 (s)
Z
g(t) = ds
0 γ(s) − z0
for t ∈ [0, 1]. One checks that
d −g(t)
e (γ(t) − z0 ) = 0,
dt
hence
e−g(t) (γ(t) − z0 ) = e−g(0) (γ(0) − z0 ) = γ(0) − z0 .
Taking t = 1, we have
e−2πin(γ;z0 ) (γ(1) − z0 ) = γ(0) − z0 .
Since γ(1) = γ(0), we conclude that n(γ; z0 ) is an integer. Apparently n(γ; z0 ) is
locally constant, hence constant on each connected component of C − γ. If z0 is
outside of the region R bounded by γ, then 1/(z − z0 ) is holomorphic on R. It
follows that n(γ; z0 ) = 0 on the unbounded component of C − γ. 
Corollary 1.31. If γ1 and γ2 are homotopic closed paths in C\{z0 }, then n(γ1 ; z0 ) =
n(γ2 ; z0 ).
Theorem 1.32 (Winding Number Version of Cauchy’s Integral Formula). Let γ be
a closed path homotopic to a point in U , and let f ∈ H(U ). Then for any z0 ∈ U \γ,
we have Z
1 f (z)
n(γ; z0 ) · f (z0 ) = dz.
2πi γ z − z0
Proof. For fixed z0 , define
 f (z) − f (z0 )

z 6= z0
g(z) = z − z0
 0
f (z0 ) z = z0 .

R expansion of f at z0 , one checks that g ∈ H(U ), hence by

Using the power series
Cauchy’s Theorem, γ g(z)dz = 0. We thus conclude the desired formula. 
1.10. Singularities. Suppose f is analytic on a small disk except possibly at the
center z0 . We say that z0 is an isolated singularity of f . If one can assign a value
at z0 such that f becomes analytic at z0 , then z0 is called removable. If f (z) → ∞
as z → z0 , we say that z0 is a pole of f . For all other cases, z0 is called an essential
singularity.
Example 1.33. For an integer n, z n has a removable singularity at 0 iff n ≥ 0,
and it has a pole at 0 iff n < 0.
Example 1.34. The function e1/z has an essential singularity at 0, which can be
seen by taking z → 0 along the real and imaginary axes, respectively.
Proposition 1.35. Suppose f ∈ H(U \ {z0 }). Then
(1) z0 is removable iff limz→z0 (z − z0 )f (z) = 0.
(2) z0 is a pole iff there exists a positive integer n and h ∈ H(U ) with h(z0 ) 6= 0
h(z)
such that f (z) = (z−z 0)
n . In this case n is called the pole order of f at z0 .

(3) z0 is essential iff for all 0 < ε  1, the set f (D(z0 , ε)∗ ) is (open) dense in C,
where D(z0 , ε)∗ denotes the punctured disk at z0 contained in U .
 13

Proof. For (1), if z0 is removable, then the conclusion is clear. Conversely if

limz→z0 (z − z0 )f (z) = 0, define g(z) = (z − z0 )2 f (z) for z 6= z0 and g(z0 ) = 0.
Then g(z) is differentiable (including at z0 ). Hence by Goursat’s theorem g ∈ H(U ).
Moreover, g 0 (z0 ) = 0 by assumption, hence g has at least a double zero at z0 by its
power series expansion. It follows that h(z) = g(z)/(z − z0 )2 ∈ H(U ). Now define
f (z0 ) = h(z0 ). Since f = h on U \ {z0 }, it implies that z0 is a removable singularity
of f .
For (2), the “if” part is clear by definition. Conversely if z0 is a pole, then
g(z) = 1/f (z) has a removable singularity at z0 by (1), since limz→z0 g(z) = 0.
Hence g(z) = (z−z0 )n ge(z) for some n ≥ 1 and ge locally holomorphic with ge(z0 ) 6= 0.
h(z)
It follows that f (z) = (z−z 0)
n with h(z) = 1/e
g (z), which is well-defined and nonzero
at z0 .
For (3), first suppose f (D(z0 , ε)∗ ) ∩ D(w0 , δ) = ∅ for some w0 ∈ C and ε, δ > 0.
1
Then f (z)−w 0
∈ H(D(z0 , ε)∗ ) has a removable singularity at z0 by (1), which implies
that f (z) has either a removable singularity or a pole at z0 . Conversely, the density
of f (D(z0 , ε)∗ ) for every ε > 0 clearly violates the definition of having a removable
singularity or a pole at z0 . 
Remark 1.36. One can strengthen (3) by the Great Picard’s Theorem, which says
that if f has an essential singularity at z0 , then on any D(z0 , ε)∗ , f (z) takes on
all possible complex values, with at most a single exception, infinitely often. For
example, e1/z has an essential singularity at z = 0, and the function can take on
any value infinitely many times except 0.
Definition 1.37. If there exists a discrete set P ⊂ U such that f ∈ H(U \ P ) and
such that each point in P is a pole of f , we say that f is a meromorphic function
on U . The set of meromorphic functions on U is denoted by M(U ).
1.11. Laurent Series. Let A be an annulus
A = {z ∈ C | r1 < |z − z0 | < r2 }, 0 ≤ r1 < r2 ≤ ∞
where we allow the inner radius to be zero and the outer radius to be infinity.
Proposition 1.38 (Laurent Expansion). Suppose f ∈ H(A). Then there exist
unique an ∈ C such that
X∞
f (z) = an (z − z0 )n
n=−∞

which converges absolutely and uniformly on any compact subset of A. Moreover,

I
1 f (w)
an = dw
2πi |w−z0 |=r (w − z0 )n+1
for all n and any r1 < r < r2 .
P−1
Such series is called the Laurent series of f around z0 , and n=−∞ an (z − zn )n
is called its principal part.
Proof. For fixed z ∈ A, denote by γr and ηr the circles centered at z0 and at z of
radius r, respectively. Take s1 and s2 such that r1 < s1 < |z − z0 | < s2 < r2 , and
take ε small. Let c be a cycle defined by
c = γs2 − γs1 − ηε .
14

Since γs2 − γs1 is homotopic to ηε away from z, we have

I
1 f (w)
dw = 0,
2πi c w − z
which combined with Cauchy’s integration formula implies that
I I
1 f (w) 1 f (w)
f (z) = dw − dw.
2πi γs2 w − z 2πi γs1 w − z

Next we show that the integral over γs2 contributes to an for n ≥ 0 and the
integral over γs1 contributes to an for n < 0. Consider the expansion
∞
(z − z0 )n
X
if |w − z0 | > |z − z0 |


(w − z0 )n+1

1 1 
n=0
= = ∞
w−z (w − z0 ) − (z − z0 )  X (w − z0 )n

 − if |w − z0 | < |z − z0 |.
(z − z0 )n+1

n=0

Since these power series converge absolutely and uniformly on the respective inte-
gration paths, inserting them into the above integral and interchanging the sum-
mation and integration gives the desired expansion as well as the expression of
an .
Alternatively, the value of an can be verified by dividing the Laurent series by
(z − z0 )n+1 and then integrating. 

Note that if r1 = 0, then z0 becomes an isolated singularity of f .

P∞
Corollary 1.39. Suppose f (z) = n=−∞ an (z − z0 )n is the Laurent expansion of
f at an isolated singularity z0 , which converges on 0 < |z − z0 | < δ for some δ > 0.
Then
(1) z0 is removable iff an = 0 for all n < 0.
(2) z0 is a pole of order m > 0 iff an = 0 for all n < −m but a−m 6= 0.
(3) Otherwise z0 is an essential singularity.

Proof. Apply the previous properties of isolated singularities. 

Next we classify holomorphic functions from C∞ → C∞ .

Proposition 1.40. Suppose f : C∞ → C∞ is a holomorphic function and f 6≡ ∞.

Then f is a rational function, i.e. f = p/q for two polynomials p, q and q 6≡ 0.

Proof. If f (z) ∈ C for all z ∈ C∞ , then f is entire and bounded, hence constant by
Liouville’s theorem. Suppose f (z0 ) = ∞ for some z0 ∈ C (we may apply Möbius
transformations to assume that f (∞) 6= ∞, which preserve the set of rational
functions). Since limz→z0 f (z) = ∞, z0 is a pole of f . Note that f has finitely many
poles, since the poles are isolated (or equivalently the zeros of 1/f are isolated).
Now subtracting the principal parts of the Laurent series of f at each pole, we
obtain an entire function which is also bounded as its image is a compact subset of
C, hence is a constant. It follows that f is the sum of a constant and those principal
parts, hence f is a rational function. 
 15

P∞
1.12. Residue Theory. Suppose f (z) = n=−∞ an (z − z0 )n is the Laurent ex-
pansion of f at an isolated singularity z0 . Then the coefficient a−1 is called the
residue of f at z0 , and denote it by Res(f ; z0 ).
g(z)
Example 1.41. Suppose f (z) = z−z0 where g is holomorphic around z0 . Let
g(z) = g(z0 ) + a1 (z − z0 ) + · · ·
by the power series expansion of g at z0 . Then we see that
Res(f ; z0 ) = g(z0 ).
g(z)
Similarly if f (z) = (z−z0 )2 where g is holomorphic around z0 , then
Res(f ; z0 ) = g 0 (z0 ).
Theorem 1.42 (Residue Theorem). Let z1 , . . . , zm ∈ U , f ∈ H(U \ {zi }m
i=1 ), and
γ a closed path homotopic to a point in U such that γ does not pass through any
zi . Then
Z m
1 X
f (z)dz = n(γ; zi ) Res(f ; zi ).
2πi γ i=1

Proof. Each zi is P
an isolated singularity of f . Let fi be the principal part of f at
m
zi . Then g = f − i=1 fi ∈ H(U ), hence by Cauchy’s theorem
I m I
1 X 1
f (z)dz = fi (z)dz.
2πi γ i=1
2πi γ
Note that I
1 1
dz = n(γ; zi )
2πi γ (z − zi )k
for k = 1 and the integral is 0 for k ≥ 2. The desired formula follows from the
definition of residue. 

Example 1.43. Let γ be a circle with 1 and 2 inside γ. Then by the Residue
Theorem, Z
z+1
= 2πi(Res(f ; 1) + Res(f ; 2)).
γ (z − 1)(z − 2)
By the previous example, we have
1+1 2+1
Res(f ; 1) = = −2, Res(f ; 2) = = 3.
1−2 2−1
Suppose f is meromorphic at z0 (i.e. z0 is not an essential singularity). Then
f (z) = (z − z0 )m g(z) for some integer m and g(z0 ) 6= 0, where m = ordz0 (f ) is the
zero or pole order of f at z0 (we use −3 for a triple pole, etc in this context). Note
that
f0 m g0
= + ,
f z − z0 g
0
and the set of poles of f /f is contained in the set of zeros and poles of f . In
particular,
Res(f 0 /f ; z0 ) = ordz0 (f ).
Hence the residue theorem implies the following.
16

Theorem 1.44 (Argument Principle). Suppose f ∈ M(U ) and γ is a closed path

homotopic to a point in U such that γ does not meet the zero set Z and the pole
set P of f . Then
Z 0
1 f (z) X
dz = n(γ; z) ordz (f ).
2πi γ f (z)
z∈Z∪P

Remark 1.45. Note that

f 0 (z) f 0 (γ(t)) 0 (f ◦ γ(t))0
dz = γ (t)dt = dt.
f (z) f (γ(t)) f ◦ γ(t) − 0
By the definition of winding number, the LHS of the above theorem is equal to
n(f ◦ γ; 0), which measures the change of argument of the composed path f ◦ γ
around 0. Too see it, view β = f ◦ γ : [0, 1] → C as a closed path. Then we have
Z 1
β 0 (t)
Z
1 1 1
n(β; 0) = dz = dt.
2πi β z − 0 2πi 0 β(t) − 0

Theorem 1.46 (Rouche’s Theorem). Let f, g ∈ H(U ), and γ a simple closed

path in U such that its interior is contained in U . If f has no zero on γ and if
|f (z) − g(z)| < |g(z)| for all z on γ, then f and g have the same number of zeros
(counting with orders) inside γ.
Proof. Set h = f /g. Then h has no zeros or poles on γ. Since |h ◦ γ(t) − 1| < 1
for all t ∈ [0, 1] and the inequality |w − 1| < 1 cuts out the unit disk centered at
1, we conclude that 0 belongs to the unbounded component of C \ h ◦ γ([0, 1]), and
hence the winding number n(h ◦ γ; 0) = 0. Since γ is homotopic to a point in U ,
by the Argument Principle and the above remark, the number of zeros minus the
number of poles of h in the interior of γ (counting with orders) is zero. Note that
this is the number of zeros of f minus the number of zeros of g inside γ (counting
with orders), hence they are equal. 

Let us apply residues to evaluate more integrals. First recall

eiθ + e−iθ eiθ − e−iθ
cos θ = , sin θ = .
2 2i
Set z = eiθ . We also have
Z 2π Z 2π
g(eiθ ) iθ
Z
g(z)
dz = ie dθ = g(eiθ )dθ.
|z|=1 iz 0 ieiθ 0
Z 2π
dθ z − z −1
Example 1.47. Let us evaluate . Replace sin θ by where
0 2 + sin θ 2i
z = eiθ . The integrand becomes
1 2iz
−1
=
2 + (z − z )/2i 4iz + z 2 − 1
which is the function g above. Hence we need to evaluate
Z Z
2 2
2 + 4iz − 1
dz = √ √ dz.
|z|=1 z |z|=1 (z − ( 3 − 2)i)(z − (−2 − 3)i)
 17

√
The only pole inside the unit circle is ( 3 − 2)i. Hence by the Residue Theorem
the value of the integral is
2 2π
2πi · √ √ =√ .
( 3 − 2)i − (−2 − 3)i 3
Next we consider improper integrals. Denote by H the closed upper-half plane.
Let us first make the following observation.
Theorem 1.48. Suppose f is holomorphic away from a finite set of singularities,
has no singularities on R, and there exist R, C > 0 and p > 1 such that |f (z)| ≤
C|z|−p for {z ∈ H : |z| > R}, then
Z ∞ Xm
f (x)dx = 2πi Res(f ; zj ),
−∞ j=1

where z1 , . . . , zm are the singularities of f in H.

Proof. Let γr be the semicircle {z ∈ H : |z| = r} union the real line segment
−r ≤ x ≤ r (draw a picture), positively oriented. Then
Z Z r Z π
f (z)dz = f (x)dx + f (reit )ireit dt.
γr −r 0
it it 1−p
If r > R, then |f (re )ire | ≤ Cr , and hence
Z π
f (reit )ireit dt ≤ πCr1−p
0

which tends to zero as r → ∞. It follows that

Z Z r Z ∞
lim f (z)dz = lim f (x)dx = f (x)dx.
r→∞ γr r→∞ −r −∞

Then the desired formula follows from the Residue Theorem. 

Remark 1.49. In the statement of the above theorem, if we replace the upper-half
plane by the lower-half plane, then the conclusion still holds with the sign changed
to − on the right-hand side, because the lower semicircle used in the proof will have
negative orientation (draw a picture).
Z ∞
1
Example 1.50. Let us evaluate dx. Since the integrand f (z) is
−∞ (1 + x2 ) 2
−4
asymptotically z , the assumption in the above theorem holds. There are two
poles at ±i and only i is contained in the upper-half plane. Write
g(z)
f (z) = ,
(z − i)2
where g(z) = (z + i)−2 . Then
i
Res(f ; i) = g 0 (i) = − .
4
It follows that Z ∞
1 π
dx = 2πi Res(f ; i) = .
−∞ (1 + x2 )2 2
18

1.13. Normal Families. Let U be an open set in C and F a family of holomorphic

functions defined on U .
Definition 1.51. If each sequence in F has a subsequence which converges uni-
formly to a holomorphic function on every compact subset of U , then F is called a
normal family.
Definition 1.52. If there is an M such that |f (z)| ≤ M for every z ∈ U and every
f ∈ F, then F is called uniformly bounded on U .
Theorem 1.53 (Montel’s Theorem). Every uniformly bounded family on U is a
normal family.
Proof. We apply the idea of Arzela-Ascoli Theorem in real analysis. Let {fn } be a
uniformly bounded sequence of functions on U with |fn (z)| ≤ M with a fixed M
for all z ∈ U and all n. Choose points z1 , z2 , · · · in U with rational coordinates
(hence they are countable). Then there is a convergent subsequence of complex
numbers {f1n (z1 )}, since they are bounded by M . Then {f1n (z2 )} is also a bounded
sequence, so it has a convergent subsequence denoted by {f2n (z2 )}. Continuing this
process, we inductively construct a sequence of subsequences of {fn }
f11 , f12 , · · · , f1n , · · ·
f21 , f22 , · · · , f2n , · · ·
.. ..
. ··· . ···
each being a subsequence of the preceding one, such that {fkn (zj )}n is a convergent
sequence for each k and each j ≤ k. Then the diagonal sequence {fnn } converges
at every zj .
Set gn = fnn . We will show that {gn } converges uniformly on every compact
subset of U . For w ∈ U , choose r > 0 such that D(w, 2r) ⊂ U . For each z ∈ D(w, r),
we have D(z, r) ⊂ D(w, 2r) ⊂ U . By Cauchy’s Estimates applied to D(z, r), every
f ∈ F satisfies
M
|f 0 (z)| ≤ .
r
Then for any z, z 0 ∈ D(z, r), we have
Z z
0 M
|f (z) − f (z )| = f 0 (u)du ≤ |z − z 0 |
z 0 r
which in particular holds for all gn .
rε
Given ε > 0, set δ = 3M . If z ∈ D(w, r), there are points with rational coordi-
nates arbitrarily close to z. Let zj be such point with |z − zj | < δ. Then
M ε
|gn (z) − gn (zj )| <δ= .
r 3
Next since {gn } = {fnn } converges at zj , we choose N such that
ε
|gn (zj ) − gm (zj )| < , ∀ m, n ≥ N.
3
Then we have
ε ε ε
|gn (z)−gm (z)| ≤ |gn (z)−gn (zj )|+|gn (zj )−gm (zj )|+|gm (zj )−gm (z)| < + + =ε
3 3 3
 19

for all m, n ≥ N . It proves that {gn } converges uniformly on these closed disks
D(w, r), hence it converges uniformly on every compact subset K of U , as K can
be covered by finitely many such disks. 
1.14. The Riemann Mapping Theorem. Let U be a path-connected open sub-
set of C. Recall that U is simply connected if every loop contained in U is con-
tractible (i.e. the fundamental group of U is trivial).
Lemma 1.54. If U is simply connected and f is a nowhere vanishing holomorphic
function on U , then f has a holomorphic logarithm and a holomorphic nth root.
Proof. Suppose f ∈ H(U ) is nowhere zero. Then f 0 /f ∈ H(U ). Since U is simply
connected, for a fixed z0 ∈ U , we have
Z z 0
f (w)
F (z) = dw ∈ H(U ).
z0 f (w)

In particular, F 0 = f 0 /f and F (z0 ) = 0. Since f (z0 ) 6= 0, fix one value of log f (z0 )
(not unique) and define
h(z) = e−F (z)−log f (z0 ) f (z).
Then one checks that h0 (z) = 0 and h(z0 ) = 1, hence h(z) = 1 for all z ∈ U . It
follows that
f (z) = eF (z)+log f (z0 ) .
Hence the exponent can be used as log f . Take g(z) = e(log f )/n . Then g n = f . 
A complex valued function is called conformal at z0 (i.e. preserving angles of
two intersecting arcs), if it has a nonzero derivative at z0 .
Theorem 1.55 (Riemann Mapping Theorem). Every proper, simply connected,
non-empty open subset U of C is conformally equivalent to the open unit disk D.
We need some preparations before proving the theorem. First we determine the
conformal automorphisms of D. For w ∈ D, define
z−w
hw (z) =
1 − wz
as a special Möbius transformation, where hw (w) = 0 and hw (0) = −w. If |u| = 1,
then u−1 = u, and
u−w u−w u−w
= −1 = = 1.
1 − wu u −w u−w
It implies that hw maps the unit circle to itself. Since w is inside D and hw (w) = 0,
hw maps D onto itself, hence hw is a conformal automorphism of D.
Theorem 1.56. Every conformal automorphism h of the unit disk is of the form
z−w
h(z) = uhw (z) = u
1 − wz
for some u and w satisfying that |u| = 1 and |w| < 1.
Proof. Suppose h(w) = 0 for some w ∈ D. Consider h ◦ h−1 w , which is a conformal
automorphism of D and takes 0 to 0. By the Schwarz Lemma, h ◦ h−1 w (z) = uz for
some |u| = 1. Replacing z by hw (z), we conclude that h(z) = uhw (z). 
20

Recall that U is a proper, simply connected, non-empty open subset of C. Fix

z0 ∈ U and let F be the set of one-to-one conformal maps f of U into D such that
f (z0 ) = 0.
Lemma 1.57. The set F is non-empty.
Proof. Since U is proper in C, there is some λ 6∈ U . Then f (z) = z − λ is non-
vanishing on U , hence has a holomorphic square root g. Since f is one-to-one
and non-vanishing, so is g, and hence g is a one-to-one conformal map such that
0 6∈ g(U ). By the Open Mapping Theorem, g(U ) is open, so we can find a closed
disk D(w0 , r) ⊂ g(U ) with 0 < r < ∞. Moreover, the set g(U ) cannot contain both
u and −u for any u 6= 0, as f = g 2 is one-to-one. Thus the reflection D(−w0 , r) ∩
g(U ) = ∅. It follows that
|g(z) + w0 | > r
for all z ∈ U . Hence the function
r
p(z) =
g(z) + w0
is a one-to-one conformal map of U into the unit disk D. For the fixed point
z0 ∈ U , suppose that p(z0 ) = w. Compose p with the conformal automorphism hw
of D. We obtain a one-to-one conformal map hw ◦ p of U into D which takes z0 to
hw (p(z0 )) = hw (w) = 0. Hence F contains hw ◦ p. 
Lemma 1.58. Let U , z0 , and F be as above. Given any f ∈ F, if f does not map
U onto D, then there exists some g ∈ F such that |g 0 (z0 )| > |f 0 (z0 )|.
Proof. Suppose w ∈ D and w 6∈ f (U ). Recall that hw is a conformal automorphism
of D with hw (w) = 0. Then hw ◦ f (z) 6= 0 for all z ∈ U . Hence hw ◦ f has
a holomorphic square root q on U , i.e. q 2 = hw ◦ f . If q(z0 ) = λ, then λ2 =
hw ◦ f (z0 ) = hw (0) = −w, and
h0w (0) 0 1 − |w|2 0 1 − |λ|4 0
q 0 (z0 ) = f (z0 ) = f (z0 ) = f (z0 ).
2q(z0 ) 2λ 2λ
As before q is a one-to-one conformal map of U into D and q(z0 ) = λ 6= 0. So
g = hλ ◦ q maps z0 to 0, hence g ∈ F and satisfies
1 − |λ|4 1 + |λ|2 0
g 0 (z0 ) = h0λ (λ)q 0 (z0 ) = 2
f 0 (z0 ) = f (z0 ).
2λ(1 − |λ| ) 2λ
Since 0 < (1 − |λ|)2 = 1 + |λ|2 − 2|λ| and f 0 (z0 ) 6= 0 for f being conformal, we
conclude that |g 0 (z0 )| > |f 0 (z0 )|. 
Proof of the Riemann Mapping Theorem. Let z0 ∈ U and F be as above. Since F
is non-empty, set
m = sup{|f 0 (z0 )| : f ∈ F}.
Then m is either ∞ or a positive number, since f 0 is non-vanishing. Choose a
sequence {fn } in F such that
lim |fn0 (z0 )| = m.
n→∞
Since F is a uniformly bounded family, it is a normal family, hence there is a sub-
sequence of {fn } which converges uniformly on compact subsets of U to a function
h. It follows from Cauchy’s Estimates (leaving as an exercise) that fn0 in this subse-
quence converges uniformly on compact subsets of U to h0 , and hence h0 (z0 ) = m.
 21

Since m 6= 0, h is not constant. Since each fn is one-to-one, h is also one-to-one

(Hurwitz’s Theorem, another exercise). Since fn (z0 ) = 0 for all n, we get h(z0 ) = 0.
Hence h ∈ F and h0 (z0 ) = m is attained as the maximum. By the previous lemma,
h must map U onto D, hence h is a conformal equivalence of U to D. 
Remark 1.59. In the proof the only property of simply connected sets we used
is that every non-vanishing holomorphic function has a holomorphic square root.
Since this property is preserved under conformal equivalence, being simply con-
nected is also preserved under conformal equivalence, and the unit disk is simply
connected, we conclude that a connected open subset U in C is simply connected
iff every non-vanishing holomorphic function on U has a holomorphic square root.
Remark 1.60. If U is a simply connected open set with a simple closed curve
as boundary, and if h : U → D is a conformal equivalence, then h extends to a
continuous function from U to D with a continuous inverse. For a proof, see e.g.
Rudin’s book.
1.15. Schwarz Reflection. Given a function analytic on a region, one can ask if
it can be analytically continued outside of the domain. One way to extend is to
use some kind of symmetry. First we introduce a notation. For a subset V of C,
let V − = {z : z ∈ V }.
Theorem 1.61 (Schwarz Reflection Principle). Let U be an open set symmetric
about the x-axis. Let A = U ∩ R and V be the part of U in the open upper-half
plane (draw a picture). If f is analytic on V , continuous on V ∪ A, and real-valued
on A, then f can extend to an analytic function on all of U by using f (z) on V − .
Proof. First, using Cauchy-Riemann Equations one checks that g(z) = f (z) is
analytic on V − (exercise). By assumption, f and g agree on A, hence they define
a single continuous function h on U = V ∪ A ∪ V − such that h is analytic on U ,
except possibly at points of A.
Take a point z0 ∈ A and let D Rbe a disk neighborhood of z0 in W . One can show
that for any closed path γ in D, γ h(z)dz = 0 (exercise). It follows from Morera’s
Theorem that h is analytic on D. We thus conclude that h as an extension of f is
analytic on U . 
Identify the set of all lines and circles in C as the set of all circles in the Riemann
sphere S 2 .
Definition 1.62. Suppose h : W → V is a conformal equivalence between domains
in S 2 . Let I ⊂ W be an arc on a circle in S 2 . Then the image h(I) is called an
analytic curve.
By using a Möbius transformation, we may assume that the circle in the above
definition is R ∪ {∞} and I is either an open interval on the real line or R ∪ {∞}
itself. In the latter the curve will be a simple closed curve.
Let us describe explicitly reflection through an analytic curve. First, a function
f on an open set U is called conjugate analytic if f is analytic on U .
Definition 1.63. Suppose C is an analytic curve in U such that U \ C has two
connected components V and W . If ρ : U → U is a conjugate analytic function
which fixes each point of C and interchanges V and W , and if ρ ◦ ρ(z) = z for all
z ∈ U , then ρ is called a reflection through C on U .
22

Theorem 1.64. If C is an analytic curve in C, then there is a reflection ρ through

C on some domain V containing C. This reflection is unique in the sense that
another reflection through C on some neighborhood V1 of C must be equal to ρ on
the connected component of V ∩ V1 which contains C.

Proof. The idea is to reduce to the case when C lies in R and use the standard
conjugate z̄. Let h : W → V be a conformal equivalence that maps L onto C,
where L is either a line segment in R or R ∪ {∞}. Replacing W by a connected
component of W ∩ W − which contains L, we may assume that W = W − . Let
κ(z) = z. Then we get a reflection ρ through C on V by setting
ρ = h ◦ κ ◦ h−1 .
Since h is analytic and κ is conjugate analytic, one checks that ρ is conjugate
analytic (exercise). Moreover,
ρ ◦ ρ = κ ◦ κ = Id,
and ρ(z) = z on C because κ(z) = z on L and h−1 maps C to L. Thus ρ is a
reflection through C on V .
If ρ1 is another reflection through C on V1 , then ρ1 ◦ ρ is analytic on V ∩ V1 , since
the composition of two conjugate analytic functions is analytic (exercise). Moreover,
ρ1 ◦ ρ(z) = z on C, hence it must be equal to z on all connected components of
V ∩ V1 containing C, by the Uniqueness Theorem of analytic functions. It implies
that ρ1 = ρ−1 = ρ on this component. 

Now we can strengthen the Schwarz Reflection Principle as follows.

Theorem 1.65. If C is an analytic curve in a domain U such that U \ C has two

components V and W , and ρ is a reflection through C on U , then any function f
which is continuous on V ∪ C, analytic on V , and real-valued on C can extend to
an analytic function on all of U by using f (ρ(z)) on W .

Proof. The proof is the same as the proof of the Schwarz Reflection Principle, where
ρ plays the role of complex conjugate on U . Let

g(z) = f (ρ(z))
on W ∪ C. Then g is continuous on W ∪ C and analytic on W because it is the
composition of two conjugate analytic functions f and ρ. Since ρ fixes points on C
and f is real-valued on C, g and f agree on C, hence they define a single function
h on U such that h is continuous on U and analytic on U \ C. Arguing as before
by Morera’s Theorem, h as an extension of f is analytic on all of U . 

Example 1.66. Let us consider the reflection through an arc on the unit circle.
1 eiθ
The map ρ(z) = or ρ(reiθ ) = is defined and conjugate analytic on C \ {0}.
z r
It satisfies ρ ◦ ρ(z) = z. It also fixes each point on the unit circle. Suppose a
domain U meets the unit circle in an arc C and is taken to itself by ρ. Then the
unique reflection through C on U is the map ρ. For such U and C, the previous
theorem says that any function analytic on the part V of U lying on one side of C,
continuous on V ∪ C, and real-valued on C can extend analytically to all of U .
 23

1.16. Analytic Continuation. Here we discuss analytic continuation along curves.

For an analytic function f defined on an open disk D, we say that (f, D) is an an-
alytic function element, and is an analytic function element at w if w ∈ D. Two
analytic function elements (f1 , D1 ) and (f2 , D2 ) at w are said to be equivalent at
w if f1 = f2 on D1 ∩ D2 .
Definition 1.67. Let γ : [0, 1] → C be a curve and (f0 , D0 ) an analytic function
element at z0 = γ(0). Suppose there exists a partition 0 = t0 < t1 < · · · <
tn+1 = 1 of [0, 1] and a sequence of function elements (f1 , D1 ), . . . , (fn , Dn ) such
that γ([tj , tj+1 ]) ⊂ Dj and fj = fj+1 on Dj ∩ Dj+1 for all j. Then we say that
(fn , Dn ) is an analytic continuation of (f0 , D0 ) along γ (draw a picture).
Remark 1.68. By the Uniqueness Theorem (after refining the sequence of disks if
necessary), it is clear that the analytic continuation of f along γ, up to equivalence,
only depends on f and γ, not on the choice of the chain of disks.
Remark 1.69. Suppose (f0 , D0 ) is a function element at z0 and γ is a curve joining
z0 to w. If there is an open set U containing D0 and γ, and an analytic function f
on U such that f = f0 on D0 , then it is obvious that (f0 , D0 ) can be analytically
continued along γ using the restriction of f .
Suppose f ∈ A(U ). Locally around a point z0 ∈ U , there exists an antiderivative
F of f . Now for any smooth path γ : [0, 1] → U beginning with z0 , one can
analytically continue F along γ as an antiderivative by
Z
F (zt ) = F (z0 ) + f (w)dw
γ

where γ(0) = z0 and γ(t) = zt . However, this procedure does not necessarily lead
to a globally defined antiderivative F ∈ A(U ). For example, take f (z) = 1/z and
U = C∗ . If γ is a closed path traversing around 0 one time, then the starting and
ending log branches will differ by 2πi. On the other hand, if U is simply connected,
then by Cauchy’s Integration Theorem such a construction yields a globally defined
F . This holds in general for analytic continuation as follows.
Theorem 1.70 (Monodromy Theorem). Suppose γ0 and γ1 are homotopic curves
in an open set U joining z0 to w, and (f0 , D0 ) an analytic function element at z0 ,
with D0 ⊂ U . If (f0 , D0 ) can be analytically continued along every curve in U , then
the continuations of f0 along γ0 and γ1 are equivalent function elements at w.
Proof. Let {γs } be a one-parameter family of curves from z0 to w as a homotopy
from γ0 to γ1 . Then given any ε > 0, there is a δ > 0 such that ||γs − γr || < ε
whenever |s − r| < δ. Denote by φs the terminal function element for the continu-
ation along γs . We will show that for each r ∈ [0, 1], there is a δ > 0 such that φs
is equivalent to φr whenever |s − r| < δ.
Let 0 = t0 < t1 < · · · < tn+1 = 1 be a partition and (f1 , D1 ), . . . , (fn , Dn ) a
sequence of function elements defining φr = (fn , Dn ) as an analytic continuation
of (f0 , D0 ) along γr . Then γr ([tj , tj+1 ]) ⊂ Dj for j = 0, . . . , n. For each j, let
εj be the distance from γr ([tj , tj+1 ]) to the boundary of Dj . If ||γs − γr || < j ,
then γs ([tj , tj+1 ]) ⊂ Dj . Thus let ε = min{ε0 , . . . , εn }, and choose δ > 0 such
that ||γs − γr || <  whenever |s − r| < δ. Then for each s with |s − r| < δ, the
same partition and sequence of function elements (f1 , D1 ), . . . , (fn , Dn ) also defines
24

(fn , D0 ) as an analytic continuation of (f0 , D0 ) along γs . By the uniqueness remark,

it implies that φs is equivalent to φr whenever |r − s| < δ.
Now we define a function h on [0, 1] by setting h(s) = 0 if φs is equivalent to φ0
and h(s) = 1 otherwise. By the previous paragraph h is a continuous function and
h(0) = 0. Hence h(s) = 0 for all s and in particular h(1) = 0, which means φ1 is
equivalent to φ0 . 
Corollary 1.71. Analytic continuations (if they exist) are unique in a simply con-
nected open set.
Corollary 1.72. Suppose U is a simply connected open set and (f0 , D0 ) is an
analytic function element at z0 ∈ U with D0 ⊂ U . If (f0 , D0 ) can be analytically
continued along every curve in U , then there is a function f which is analytic on
U and equal to f0 on D0 .
Proof. If z ∈ U , since any two curves from z0 to z in U are homotopic, the Mon-
odromy Theorem implies that the terminal function elements of analytic continu-
ations of (f0 , D0 ) along them are equivalent in some neighborhood of z. Hence all
such analytic continuations determine the same function value at z, and we define
f (z) to be this value.
Clearly f0 (z) = f (z) for all z ∈ D0 . It remains to show that f ∈ A(U ). For
w ∈ U , let γ be a curve in U joining z0 to w, and let (Dn , fn ) be the terminal
function element of an analytic continuation of f along γ. If z ∈ Dn , then (Dn , fn )
is also the terminal element of a continuation of (f0 , D0 ) along a curve γ1 from z0
to z in U , where we extend γ to γ1 ending at z by joining γ with the line segment
from w to z. It follows by definition of f that f (z) = fn (z) for all z ∈ Dn (not
only for z = w). Since fn is analytic on Dn and w is an arbitrary point in U , we
conclude that f is analytic on all of U . 
1.17. Analytic Covering Maps. Suppose V and W are open subsets of C.
Definition 1.73. An analytic map h : V → W is called an (unramified) analytic
covering map if for each w0 ∈ W , there is a neighborhood A of w0 , contained in
W , such that h−1 (A) is the disjoint union of a collection {Bj } of open subsets of
W satisfying that h restricted to Bj is a conformal equivalence onto A for every j.
Example 1.74. The map exp : z → ez is an analytic covering map from C to
C \ {0} (exercise).
Example 1.75. The map z 7→ z n is an analytic covering map from C \ {0} to
C \ {0} (exercise).
Theorem 1.76 (Lifting Through an Analytic Covering Map). Let h : V → W
be an analytic covering map and U a simply connected domain in C. Then any
analytic function f : U → W can be lifted through h, i.e. there is an analytic map
g : U → V (not necessarily unique) such that f = h ◦ g (draw a diagram).
Proof. Fix z0 ∈ U and let w0 = f (z0 ). Choose a neighborhood A0 of w0 such that
h−1 (A0 ) is a disjoint union of open sets, each mapping onto A0 by h as a conformal
equivalence. Denote by B0 one of these open sets, and let h−1 0 be the inverse of
h : B0 → A0 . Choose an open disk D0 , centered at z0 and contained in f −1 (A0 ).
On D0 we define a function
g0 = h−10 ◦ f.
 25

Then g0 lifts f through h, but only on D0 . The rest of the proof is to show that the
analytic function element (g0 , D0 ) can be analytically continued along any path in
U beginning at z0 . If we can do this, then the Monodromy Theorem implies that
(g0 , D0 ) can be continued to an analytic function g on all of U (since U is simply
connected). Moreover, since f and h ◦ g agree on D0 , then they are the same on all
of U by the Uniqueness Theorem.
Let γ : [0, 1] → U be a path beginning at z0 . Let S be the subset of I = [0, 1]
such that (g0 , D0 ) can be analytically continued along γ as far as to s for s ∈ S.
Clearly if s ∈ S, then t ∈ S for all 0 ≤ t < s. If S = [0, 1], then we are done.
If S = [0, r] for some r < 1, then the terminal element is defined on a disk that
intersects (r, 1], hence S should be larger, a contradiction. The only remaining case
is S = [0, r) with r < 1, which cannot happen as we will prove below.
If S = [0, r), choose a neighborhood A of f (γ(r)) in W such that h−1 (A) is a
disjoint union of open sets Bj , each mapping onto A conformal equivalently under
h. Since f is continuous, we can choose an open disk D ⊂ U such that f (D) ⊂ A
and a point s ∈ [0, r) such that w = f (γ(s)) ∈ f (D) (draw a picture). Since
s ∈ S, the function element (g0 , D0 ) can be analytically continued along γ to
some (gn , Dn ) with γ(s) ∈ Dn . Since f = h ◦ g0 on D0 , again by the Uniqueness
Theorem we have f = h ◦ gn on Dn . In particular, w = f (γ(s)) = h(gn (γ(s))),
hence gn (γ(s)) ∈ h−1 (A), say, gn (γs ) ∈ Bk . Now define a new function element
(gn+1 , Dn+1 ) by setting Dn+1 = D and gn+1 = f ◦ h−1 k , where hk : Bk → A.
Then f = h ◦ gn+1 on Dn+1 . Note that gn (Dn ∩ Dn+1 ) ⊂ Bk , for otherwise the
connected open set Dn ∩ Dn+1 would be separated by gn−1 (Bk ) and the union of the
sets gn−1 (Bj ) for j 6= k. It follows that the two inverse functions of h used in the
definitions of gn and gn+1 agree on f (Dn ∩ Dn+1 ), which implies that gn = gn+1 on
Dn ∩Dn+1 . Since γ(r) ∈ Dn+1 , it means that (g0 , D0 ) can be analytically continued
to γ(r), hence r ∈ S, contradicting that S = [0, r). 
1.18. The Picard Theorems. We will apply the previously developed techniques
to prove the following two amazing theorems.
Theorem 1.77 (Little Picard Theorem). A non-constant entire function takes on
every value in C except possibly one.
Theorem 1.78 (Great Picard Theorem). An analytic function with an essential
singularity at z0 takes on every complex value, except possibly one, infinitely often
in every neighborhood of z0 .
The upshot for proving the Picard theorems is to construct an analytic covering
map from the unit disk to C with two points removed. We begin with some facts
about reflection through a circle.
Suppose C1 and C2 are two circles in the plane that intersect at two points. If
the tangents to the two circles are perpendicular at the two points, then we say
that the circles meet at right angles (draw a picture). Indeed if the tangents are
perpendicular at one intersection point, then they are perpendicular at the other
point (by symmetry to the line connecting their centers).
Theorem 1.79. If two circles C1 and C2 meet at right angles, and A is the arc of
C1 lying inside C2 , then reflection through A maps C2 onto itself (draw a picture).
Proof. There is a Möbius transformation h that maps C1 onto R ∪ ∞ and maps
the inside of C1 onto the upper-half plane H. We can further choose which point
26

on C1 is sent to ∞, and assume that it is not on A. Since Möbius transformations

are conformal and take circles to circles (or lines), the image of C2 under h meets
the x-axis at two points with right angles, hence h(C2 ) must be a circle such that
a diameter of h(C2 ) is contained in the x-axis. Then the reflection ρ through the
x-axis (z 7→ z) maps the inside of h(C2 ) onto itself. Now a reflection through the
arc A can be described as h−1 ◦ ρ ◦ h, and by the uniqueness result, this is the
only reflection through A. Since ρ takes the inside of h(C2 ) onto itself, reflection
through A takes the inside of C2 onto itself. 

Theorem 1.80. With C1 , C2 as above, the reflection through C1 takes any other
circle C, which meets C2 at right angles, to another circle which meets C2 at right
angles or a line through the center of C2 .
Proof. As discussed before a reflection through a circle is a Möbius transformation
followed by conjugation, hence it takes a circle C to either a circle or a line. The
previous theorem says that the reflection through C1 takes the inside of C2 onto
itself, hence the image of C under this reflection still meets C2 in two points. Since a
conformal map preserves angles and conjugation reverses angles, their composition
is also angle reversing. It follows that if two curves meet perpendicularly, then
the same holds for their image under a reflection. Therefore, reflection through C1
takes C to a circle or line which also meets C2 at right angles, and the only way
the image of C can be a line is when it passed through the center of C2 . 

Now we construct an analytic covering map from the open unit disk D onto
C \ {0, 1}. First note that the three points −1, eiπ/3 , and e−iπ/3 are equidistant
points on the unit circle C, hence we can join each pair of these points with an arc
of a circle that meets C at right angles. These three arcs bound an open curvilinear
triangle V0 (draw a picture), which is simply connected. By the Riemann Mapping
Theorem, there is a conformal equivalence h : V0 → H where H is the upper-half
plane. As remarked before, h extends to a continuous map V 0 → H ⊂ S 2 which
takes the boundary of V0 to the boundary of H in S 2 . By composing with a suitable
Möbius transformation, we may assume that h takes −1, eiπ/3 , and e−iπ/3 to ∞,
1, and 0, respectively.
Since h is real-valued on the three edges of V0 , by the circle version of the
Schwarz Reflection Principle, h has an analytic continuation by reflection across
each of these edges. This makes h being defined and analytic in the set V1 (draw a
picture). By the previous two theorems, each of the three new cells is contained in
D and bounded by three circles that meet C at right angles. Moreover, as h maps
V0 onto the upper-half plane, it maps each of the new cells of V1 onto the lower
half-plane.
We can now analytically continue h to larger and larger domains {Vn } by repeat-
ing the above process (draw a picture). One checks that the union of these Vn is D,
so we obtain an analytic function h which maps D onto a subset of C containing
the upper and lower open half-planes and the intervals (−∞, 0), (0, 1), and (1, ∞)
on R. The points ∞, 1, and 0 are not in the image of h, because every triangular
cell has all of its vertices on the unit circle coming from reflections of the vertices
of V0 that map to ∞, 1, and 0.
Theorem 1.81. The map h : D → C\{0, 1} described above is an analytic covering
map.
 27

Proof. Use the famous “Angels and Demons” illustration (print out the picture and
distribute). 

Theorem 1.82 (Little Picard Theorem). If f is an entire function and if there

are two distinct points in C that are not in the image of f , then f is constant.
Proof. Suppose the image of f misses two values, say 0 and 1. Then f is an analytic
function from C to C \ {0, 1}. Since the map h : D → C \ {0, 1} constructed above is
an analytic covering map and C is simply connected, we know that f can be lifted
through h to an analytic function g : C → D such that f = h ◦ g. But g becomes a
bounded entire function, hence a constant, which implies that f is constant. 

The proof of the Great Picard Theorem relies on the following technical result.
Theorem 1.83. Let U be a simply connected subset of C. Then a sequence of
analytic functions on U with values in C \ {0, 1} is either a normal family or they
converge uniformly to ∞.
Proof. See [T, Theorem 7.4.5] for a proof. 

Theorem 1.84 (Great Picard Theorem). Let f be a function which is analytic in a

neighborhood U of z0 except at z0 itself, where it has an essential singularity. Then
f takes on every value, except possibly one, infinitely often in every neighborhood
of z0 .
Proof. Without loss of generality, we may assume that z0 = 0. Prove by contradic-
tion. If the conclusion is false, then there is a disk D(0, r) in which f fails to take on
at least two values, say 0 and 1. Then f is an analytic function from D(0, r) \ {0}
to C \ {0, 1} with an essential singularity at 0. Define a sequence of functions
fn (z) = f (z/n)
for all positive integers n. By the previous theorem, {fn } is a normal family, i.e.
it has a subsequence converging uniformly on compact subsets of D(0, r) or they
converge uniformly to ∞. The former implies that f is bounded, hence has a
removable singularity at 0. The latter implies that 1/f is bounded, and hence has
a removable singularity at 0, which further implies that f has at worst a pole at 0.
Both cases contradict that f has an essential singularity at 0. 

1.19. Infinite Products. First, recall that for a half-open half-closed interval I of
length 2π, the log function associated to I is
logI z = log |z| + i argI z,
where argI z is the unique argument of z lying in I for a nonzero complex number z.
If I = (−π, π], we call it the principal branch of the log function and simply denote it
by log. In particular, if u and v both have positive real part, then im log u, im log v ∈
(−π/2, π/2), and im log uv ∈ (−π, π), hence in this case log u + log v = log uv, since
they cannot differ by a nonzero multiple of 2πi.
Qn
Definition 1.85. If {uk } is a sequence
Q∞ of complex numbers and pn = k=1 uk ,
we say that the infinite product k=1 uk converges to a complex number p if
limn→∞ pn = p.
28

Theorem 1.86. The P∞ infinite product converges to a non-zero number p if and only
if the infinite sums k=1 log uk converges to a number λ, where p = eλ . Moreover,
if the infinite series converges absolutely, then the infinite product is unchanged by
any rearrangement of the factors.
Proof. Define pn as above. First suppose p = limn→∞ pn exists and is nonzero.
Since the principal branch of log is continuous at 1, we have
lim log(pn /p) = 0.
n→∞

Then there exists an N such that

−π/4 < im(log(pn /p)) < π/4
for n ≥ N . It follows that pn /pm = (pn /p)(pm /p)−1 is in the right half-plane, i.e.
has positive real part for n, m ≥ N . In particular, un+1 = pn+1 /pn has positive
real part for n ≥ N . Thus we have
log(pn+1 /pN ) = log((pn /pN )un+1 ) = log(pn /pN ) + log un+1 .
Using this equation and an inductive argument beginning with n = N , we obtain
that
n
X
log(pn /pN ) = log uk
k=N +1
for all n > N . Since the left side converges as n → ∞, so does the right side, which
implies the convergence of the series.
Conversely if the infinite series converges, let
n
X
λn = log un
k=1

be its nth partial sum. Then limn→∞ λn = λ. Since pn = eλn and the exponential
function is continuous, the sequence {pn } converges to eλ .
Finally if the series converges absolutely, then each of its rearrangement con-
verges to the same number. It follows that each rearrangement of the infinite
product also converges to the same number. 

Now consider a sequence of functions {uk (z)} defined on a set S. We say that
the infinite product of {uk (z)} converges uniformly on S if the sequence {pn (z)} of
the partial products converges uniformly on S.
Theorem 1.87. Let {uk (z)} be a sequence of complex-valued functions defined and
bounded uniformly on a set S. If the series
∞
X
log uk (z)
k=1

converges uniformly to λ(z) on S, then the infinite product

∞
Y
uk (z)
k=1

converges uniformly to eλ(z) on S.

 29

Proof. As before, let λn (z) be the nth partial sum of log uk (z), and pn (z) the nth
partial product of uk (z). By assumption, λn (z) − λ(z) converges uniformly to 0 on
S. Since the exponential function is continuous, it implies that
pn (z)
= eλn (z)−λ(z)
p(z)
converges uniformly to 1.
Since the pn (z) are bounded uniformly on S, then <(λn (z)) = log |pn (z)| are
bounded uniformly on S, and hence <(λ) is bounded on S. It follows that p(z) =
eλ(z) is bounded on S. Hence pn = (pn /p)p converges uniformly to p on S. 

Theorem 1.88. Let {ak (z)} be a sequence of complex-valued functions defined on

a set S. If the series
X∞
|ak (z)|
k=1
converges uniformly on S, then the infinite product
∞
Y
(1 + ak (z))
k=1

converges uniformly on S, and each rearrangement of the product converges to the

same function. If the infinite product converges to p(z), then each zero of p(z) is a
zero, with the same order, of some finite product of the factors (1 + ak (z)).
Proof. For small w, say |w| < 1/2, one checks that
1
|w| ≤ | log(1 + w)| ≤ 2|w|.
2
If the series converges uniformly on S, then there is a K such that |ak (z)| ≤ 1/2
for all k ≥ K and all z ∈ S. It follows that one of the two series
∞
X ∞
X
| log(1 + ak (z))| and |ak (z)|
k=K k=K

converges uniformly
P∞ on S if and only if the other one does.P∞
Therefore, if k=1 |ak (z)| converges uniformly, then k=K log(1 + ak (z)) con-
verges uniformlyQand absolutely. By the previous theorems, this implies the uniform
∞
convergence
Q∞ of k=K (1 + ak (z)) to a function on S with no zeros. It follows that
k=1 (1+ak (z)) converges uniformly on S, the limit is unaffected by rearrangement
of the factors, and each of its zeros is a zero, with the same order, of the product
of the factors 1 + ak (z) for k < K. 

Example 1.89. Consider the infinite product

∞
Y
(1 + z 2 /k 2 )
k=1
2 2 2 2
P∞
on D(0, R). Since |z /k | ≤ R /k for all z ∈ D(0, R) and k=1 R2 /k 2 converges,
it follows that the series
∞
X |z 2 |
k2
k=1
30

converges uniformly on D(0, R). Hence by the above theorem, the infinite product
also converges uniformly on D(0, R) for each R, and hence on each bounded subset
of C.
Definition 1.90. Let f ∈ A(U ), not identically zero. Then the meromorphic
function f 0 /f is called the log derivative of f on U .
This definition arises formally from (log f )0 = f 0 /f .
Theorem 1.91. Let {fn } be a sequence of analytic functions on a connected open
set U . If {fn } converges uniformly to f (not identically zero) on U , then the
sequence {fn0 /fn } converges uniformly to f 0 /f on compact subsets of U \ S, where
S is the set of zeros of f .
Proof. Similar to one of the homework problems. 
Corollary 1.92. Let {uk } be a sequence of analytic functions on a connected open
set U . If the infinite product
∞
Y
f (z) = uk (z)
k=1
converges uniformly on compact subsets of U to a function f which is not identically
zero, then the infinite sum
∞
X u0k (z)
uk (z)
k=1
converges uniformly to f 0 /f on compact subsets of U \ S, where S is the set of zeros
of f .
Example 1.93. Consider the function
∞ 
z2
Y 
f (z) = πz 1− 2
k
k=1
where the infinite product converges uniformly on each bounded subset of C. Its
log derivative can be written as
∞
f 0 (z) 1 X 2z
= + .
f (z) z z 2 − k2
k=1
Since
2z 1 1
= + ,
z 2 − k2 z−k z+k
we obtain that
n
f 0 (z) X 1
= lim .
f (z) n→∞ z−k
k=−n
Indeed with some more work one can show that
f (z) = sin(πz).
Given a sequence of points in U such that they do not have any accumulation
point in U , can one construct an analytic function on U such that it has exactly
the points of this sequence as its zeros, with each zero order equal to the number
of times it appears in the sequence? In order to answer this question, we first
introduce the following functions.
 31

Denote by E0 (z) = 1 − z and

2
/2+···+z p /p
Ep (z) = (1 − z)ez+z
for p > 0. Note that Ep (1) = 0 as a simple zero (by checking the derivative nonzero
at 1), the only zero of Ep (z) occurs at z = 1, and z + z 2 /2 + · · · + z p /p is the pth
partial sum of the power series expansion of − log(1 − z) at z = 0.
Theorem 1.94. Ep (z) is an entire function and if |z| ≤ 1, then |Ep (z)−1| ≤ |z|p+1 .
Proof. Note that Ep (0) = 1 and
2
/2+···+z p /p
(1 − Ep (z))0 = −Ep0 (z) = z p ez+z .
Since the derivative has a zero of order p at z = 0, the function 1 − Ep (z) has
a zero of order p + 1 at z = 0. Note that the expansion of ez at z = 0 and
z + z 2 /2 + · · · + z p /p have all coefficients as non-negative real numbers. It follows
that the expansion of
1 − Ep (z)
h(z) =
z p+1
at z = 0 also has non-negative real numbers as coefficients. It implies that the
maximal modulus of h(z) for |z| ≤ 1 is at h(1) = 1. 

Given f ∈ A(U ) and a sequence {zk } ⊂ U , we say that {zk } is a list of zeros
of f (counting with multiplicity) if each zk is a zero of f and each zero w of f
occurs ordw (f ) times in this sequence. Now we can define an infinite product
called Weierstrass product to answer the previous question.
Theorem 1.95. If {zk } is a sequence of non-zero complex numbers and {pk } is a
sequence of nonnegative integers such that
∞ pk +1
X z
<∞
zk
k=1

for all z, then the Weierstrass product

∞
Y
f (z) = Epk (z/zk )
k=1

converges uniformly on compact subsets of C to an entire function that has {zk } as

a list of zeroes (counting with multiplicity).
Proof. By the previous theorem, we have
z pk +1
|Epk (z/zk ) − 1| ≤
zk
if |z| ≤ |zk |. The assumption implies that for given z, |z| ≤ |zk | holds for all
P∞ large k. Now we can apply Theorem 1.88 with ak (z) = Epk (z/zk ) − 1,
sufficiently
since k=1 |ak (z)| converges uniformly on compact subsets of C. 

Theorem 1.96 (The Weierstrass Theorem). If {zk } is a sequence of complex num-

bers converging to infinity, then there exists an entire function with {zk } as a list
of its zeros (counting with multiplicity).
32

Proof. Without loss of generality, suppose exactly m terms in the sequence are 0.
We may arrange them to be the first m terms. Then {zk }∞ k=m+1 is a sequence of
nonzero complex numbers. Given any R > 0, since zk → ∞, there is K > m such
that |zk | > 2R for all k ≥ K. Then the series
X z k
zk
k=m+1

converges uniformly on |z| ≤ R by comparing with the geometric series. Hence

the hypotheses of the previous theorem hold if we set pk = k − 1 for each k. The
resulting Weierstrass product
∞
Y
Epk (z/zk )
k=m+1

converges uniformly on compact subsets of C to an entire function which has

{zk }∞
k=m+1 as a list of its zeros. Finally we can set
∞
Y
f (z) = z m Epk (z/zk )
k=m+1

which has {zk }∞

k=1 as a list of its zeros. 
Example 1.97. Let us construct an entire function which has a zero of order k at
each positive integer k. The corresponding sequence of zeros is
{zk } = 1, 2, 2, 3, 3, 3, · · ·
Then we have
∞ ∞ ∞
X 1 X n X 1
= =
|zk |p+1 n=1
n p+1
n=1
n p
k=1
which converges say for p = 2. Hence the Weierstrass product for the sequence
{zk } and {pk = 2} gives
∞ ∞  n
Y Y 2 2
E2 (z/zk ) = (1 − z/n)ez/n+z /(2n ) .
k=1 n=1
By the previous theorem, this infinite product converges to an entire function with
the desired zeros.
In the opposite direction, we have the following factorization result.
Theorem 1.98 (Weierstrass Factorization). Let f be an entire function, not iden-
tically zero. Let m be the zero order of f at 0, and let {zk } be a list of the non-zero
zeros of f (counting with multiplicity). Then there exist non-negative integers {pk }
and an entire function h such that
∞
Y
f (z) = eh(z) z m Epk (z/zk ).
k=1

Proof. Choose {pk } such that the hypotheses of the previous theorem hold (say,
taking pk = k − 1 always works). Then the resulting function
∞
Y
g(z) = z m Epk (z/zk )
k=1
 33

converges uniformly on compact sets and has the same zeros as f with the same
multiplicities. Therefore, f g −1 is an entire function with no zeros, hence f g −1 = eh
for some entire function h. 

Example 1.99. Consider f (z) = sin(πz). It has a simple zero at each integer and
no other zeros. Since
∞
X 1
< ∞,
k2
k=1
we can take pk = 1 for all k. Then the above theorem implies that
Y Y
sin(πz) = eh(z) z E1 (z/k) = eh(z) z (1 − z/k)ez/k .
k6=0 k6=0

Note that
(1 − z/k)ez/k (1 + z/k)e−z/k = 1 − z 2 /k 2 .
Hence we can write
∞
Y
sin(πz) = eh(z) z (1 − z 2 /k 2 ).
k=1
Indeed with some more work one can show that
Y∞
sin(πz) = πz (1 − z 2 /k 2 ).
k=1

Below we introduce some variations of the Weierstrass Theorem. First we replace

C by an arbitrary non-empty, proper open subset of S 2 .
Theorem 1.100 (General Weierstrass Theorem). Let U be a non-empty, proper
open subset of S 2 . If {zk } is any sequence of points of U with no limit points in U ,
then there is an analytic function f on U such that f has {zk } as a list of its zeros
(counting with multiplicity).
Proof. Using Möbius transformation, we may assume that ∞ ∈ U (and not equal
to any zk ), hence S 2 \ U is a compact subset of C. Since {zk } has no limit point in
U , the distance between zk and S 2 \ U must approach 0 as k → ∞. Hence we can
choose a sequence {wk } of points in S 2 \ U such that limk→∞ |zk − wk | = 0. Define
∞  
Y zk − wk
f (z) = Ek .
z − wk
k=1

Since limk→∞ |zk − wk | = 0, the series

∞
X zk − wk k+1

z − wk
k=1

converges uniformly on compact subsets of U , hence f is analytic on U and has

{zk } as a list of its zeros. 

Next we consider poles as well.

Theorem 1.101. If U is an open subset of C, then every meromorphic function
on U has the form f /g, where f and g are analytic on U and g is not identically
zero.
34

Proof. Let h ∈ M(U ) and let {zk } be a sequence of the poles of h, where each
zk appears as many times as its pole order. By the previous theorem, there exists
g ∈ A(U ) such that g has {zk } as a list of zeros. It implies that f = gh ∈ A(U ),
and h = f /g as desired. 

Theorem 1.102 (Mittag-Leffler Theorem). Let R > 0 or R = ∞, let S be a

discrete set of points of D(0, R) with no limit points in D(0, R), and let {hw : w ∈
S} be a set of polynomials with no constant terms. Then there exists a meromorphic
function f whose poles are exact at w for each w ∈ S with principal part given by
hw ((z − w)−1 ).

Proof. Choose an increasing sequence of radii {rn } with rn → R. Denote by S1 the

subset of S contained in D(0, r1 ), and for n > 1, let

Sn = {w ∈ S : rn−1 < |w| ≤ rn }.

Then for each n,

X
gn (z) = hw ((z − w)−1 )
w∈Sn

is a meromorphic function on C with a pole at w and with the required principal

part for each w ∈ Sn . We might hope to take the infinite sum of gn for all n, but
there is no reason why the sum converges on D(0, R). Nevertheless, we can modify
each gn without changing its poles and principal parts such that the modified series
converges.
More precisely, gn is analytic on an open set containing D(0, rn−1 ). Hence gn is
the uniform limit of its power series at 0 on D(0, rn−1 ). It follows that there is a
polynomial pn such that
|gn (z) − pn (z)| < 2−n
for all |z| ≤ rn−1 . Set f1 = g1 and fn = gn − pn . Then for each m, the series
∞
X
fn (z)
n=m+1

converges uniformly to an analytic function on D(0, rm ). It implies that

∞
X
f (z) = fn (z)
n=1

is defined as a meromorphic function on D(0, rm ) and has the required poles and
principal parts at the points of S contained in D(0, rm ). Since limm→∞ rm = R, we
thus conclude that f (z) is meromorphic on the entire D(0, R) and has the desired
poles and principal parts. 

2. Riemann Surfaces
In this section we carry out complex analysis on Riemann surfaces, introduce
fundamental results about Riemann surfaces, and interpret them from a modern
viewpoint.
 35

2.1. Preliminaries. We go over some of the previous concepts and results in the
context of Riemann surfaces.
Definition 2.1 (Riemann surface). A Riemann surface X is a one-dimensional
complex manifold, i.e., X is a real surface with a complex atlas of charts
{φi : Ui → Vi ⊂ C},
where {Ui } is an open covering of X, φi is a homeomorphism, and
φj ◦ φ−1
i : φi (Ui ∩ Uj ) → φj (Ui ∩ Uj )
is a conformal equivalence. We say that such a complex atlas of charts is a complex
structure on the underlying real surface.
Example 2.2. (a) Recall that P1 = C ∪ {∞}, homeomorphic to the real sphere.
Take
U1 = P1 \{∞} = C,
U2 = P1 \{0} = C∗ ∪ {∞}.
Define φ1 (z) = z, φ2 (z) = 1/z for z 6= ∞ and φ2 (∞) = 0. Then φ2 ◦ φ−1 1 :
C∗ → C∗ is given by z 7→ 1/z, which is a conformal equivalence. Therefore, P1
is a Riemann surface, called the Riemann sphere or the (complex) projective
line.
(b) Suppose ω1 , ω2 ∈ C are linearly independent over R. Define
Γ = Zω1 + Zω2 = {nω1 + mω2 | n, m ∈ Z},
called the lattice spanned by ω1 and ω2 . Define an equivalence relation z1 ∼ z2
on C if z1 − z2 ∈ Γ. Denote by E the set of equivalence classes, i.e., E = C/Γ.
Equip E with the quotient topology, i.e., U ⊂ E is open if and only if π −1 (U )
is open for the projection π : C → E. Then E is homeomorphic to a torus.
One can enable E a complex structure as follows. Let V ⊂ C be an open
subset such that no two points in V are equivalent with respect to Γ. Then
U = π(V ) is open in E, and let φ : U → V be the inverse of π, which forms a
complex chart on E.
Exercise 2.3. Prove that the above complex charts form a complex structure on
E.
Definition 2.4 (Holomorphic function). Let X be a Riemann surface. A function
f : X → C is called holomorphic, if for every chart φ : U → V ⊂ C on X the
function
f ◦ φ−1 : V → C
is holomorphic. The set of all holomorphic functions on X is denoted by O(X)
(new notation, previously we used H(X)).
Definition 2.5 (Holomorphic map). Let X and Y be Riemann surfaces. A con-
tinuous map f : X → Y is a called holomorphic, if for every chart φ : U1 → V1 ⊂ C
on X and ψ : U2 → V2 ⊂ C on Y with f (U1 ) ⊂ U2 the function
ψ ◦ f ◦ φ−1 : V1 → V2
is holomorphic. If f is biholomorphic, then X and Y are called isomorphic.
36

Definition 2.6 (Meromorphic function). Let X be a Riemann surface and Σ ⊂ X

a discrete set of points. Suppose f : X\Σ → C is a holomorphic function such that
lim |f (x)| = ∞
x→p

for every p ∈ Σ, then f is called meromorphic, and the points in Σ are poles of f .
The set of all meromorphic functions on X is denoted by M (X).
Remark 2.7. In a neighborhood of a pole p, let z be a suitable coordinate with
z(p) = 0. Then as we saw before the Laurent series expansion of f is
∞
X
f= ai z i
i=−k

for some k ∈ Z+ with a−k 6= 0. We say that k is the pole order of f at p.

Example 2.8. Let f = z n + an−1 z n−1 + · · · + a0 be a polynomial on C ⊂ P1 with
n ≥ 1. Then lim |f (z)| = ∞, hence f ∈ M (P1 ).
z→∞

Theorem 2.9. Suppose X is a Riemann surface and f ∈ M (X). For each pole p
of f , define f (p) = ∞. Then f : X → P1 is a holomorphic map.
Proof. Take the Laurent series expansion
∞
X
f (z) = ai z i = z −k g(z)
i=−k

locally at a pole p with a suitable coordinate z(p) = 0, where k is the pole order of
p and g(p) 6= 0. Then 1/f = z k (1/g) is holomorphic at p. 
Exercise 2.10. Prove that every holomorphic function on a compact Riemann
surface is constant. Hint: a non-constant holomorphic function is open.
Definition 2.11 (Covering, branch and ramification). Suppose X and Y are Rie-
mann surfaces and π : Y → X is a non-constant holomorphic map. Then π is called
a (possibly branched) covering.
A point y ∈ Y is called a ramification point of π, if there is no neighborhood
V of y such that π|V is injective. A point x ∈ X is called a branch point of π, if
π −1 (x) contains a ramification point. The map π is called unramified if it has no
branch point (i.e., no ramification point).
For p ∈ Y and q = π(p) ∈ X, by properties of non-constant holomorphic func-
tions, there exist neighborhoods U of p and V of q as well as suitable coordinates
x and y such that x(p) = y(q) = 0 and f |U (x) = xk for some k ∈ Z+ . If k = 1,
then p is not a ramification point of π. If k > 1, then p is a ramification point. We
say that the multiplicity of π at p is k, denoted by multp (π), and the ramification
order of π at p is k − 1, denoted by ordp (π). In particular, multp (π) − 1 = ordp (π).
Example 2.12. (a) Let π : C → C be π(z) = z k for an integer k ≥ 2. Then 0 is
the only ramification point.
(b) Consider the exponential map exp : C → C∗ by z 7→ ez . Then exp is unramified.
In particular, if V ⊂ C does not contain two points that differ by an integer
multiple of 2πi, then exp |V is injective.
(c) Let Γ be a lattice in C, and E = C/Γ the associated torus. Then the projection
C → E is unramified.
 37

Theorem 2.13 (Degree of covering maps). Let π : Y → X be a covering map of

two connected compact Riemann surfaces. Then for q ∈ X, the fiber cardinality
over q, counting with multiplicity, is independent of q. Namely,
X
d= multp (π)
p∈π −1 (q)

is constant for all q ∈ X, called the degree of π.

Proof. Since π is non-constant, π is open, π(Y ) is both open and closed in X,
hence π is onto. For q ∈ X, suppose π −1 (q) = {p1 , . . . , pn } as a set. At each
ki
Pn pi , suppose0 π is of type z 7→ z under a suitable local coordinate. Then
preimage
d = i=1 ki for all q in a neighborhood of q, and hence d is a local constant. Since
X is connected, d is a global constant. 
Exercise 2.14. Suppose X is a compact Riemann surface and f is a non-constant
meromorphic function on X. Then f has as many zeros as poles, counting with
multiplicity.
2.2. Sheaves. We introduce an important concept “sheaf” which has been used
extensively in modern mathematics.
Let X be a topological space. A sheaf F on X associates to each open set U
an abelian group F (U ), called the sections of F over U , along with a restriction
map rV,U : F (V ) → F (U ) for any open sets U ⊂ V (for a section σ ∈ F (V ), we
often write σ|U to denote rV,U (σ)), satisfying the following conditions:
(1) For any open sets U ⊂ V ⊂ W , rV,U ◦ rW,V = rW,U ;
(2) For a collection of open sets {Ui }i∈I and sections αi ∈ F (Ui ), if αi |Ui ∩Uj =
αj |Ui ∩Uj for any i, j ∈ I, then there exists a unique α ∈ F (∪i Ui ) such that α|Ui =
αi for any i.
Remark 2.15. If F satisfies (1) only, we call it a presheaf. One can perform
sheafification for a presheaf to make it become a sheaf. For many sheaves we
consider, the restriction maps are the obvious ones by restricting functions from
bigger subsets to smaller ones.
Exercise 2.16. Show that F (∅) consists of exactly one element.
Example 2.17. Let G be an abelian group. We have the sheaf of locally constant
functions G on a topological space X, where G(U ) is the group of locally constant
maps f : U → G on a non-empty open set U ⊂ X and G(∅) = 0.
Exercise 2.18. Show that for the sheaf G of locally constant functions, we have
G(U ) = G for any non-empty connected open set U .
Exercise 2.19. Suppose we define G(U ) = G as the set of constant functions on
a non-empty open set U with the natural restriction maps. If G contains at least
two elements and if X has two disjoint non-empty open subsets, show that G is a
sheaf.
Example 2.20. Let X be a complex manifold and U ⊂ X an open set.
(1) Sheaf O of holomorphic functions:
O(U ) = {holomorphic functions on U }.
The group law is given by addition.
38

(2) Sheaf O ∗ of everywhere nonzero holomorphic functions:

O ∗ (U ) = {holomorphic functions f on U : f (p) 6= 0 for any p ∈ U }.
The group law is given by multiplication.
(3) Sheaf M of meromorphic functions: strictly speaking, a meromorphic func-
tion is not a function, even we take ∞ into account. If X is compact, we cannot take
a meromorphic function as a quotient of two holomorphic functions, since any glob-
ally defined holomorphic function is a constant on X. Instead, we define f ∈ M (U )
as local quotients of holomorphic functions compatible with each other. Namely,
there exists an open covering {Ui } of U such that on each Ui , f is given by gi /hi
for some gi , hi ∈ O(Ui ) satisfying gi /hi = gj /hj , i.e. gi hj = gj hi ∈ O(Ui ∩ Uj ),
hence these local quotients can be glued together over U .
(4) Sheaf M ∗ of meromorphic functions not identically zero: this is defined
similarly as above and the group law is given by multiplication.
2.3. Maps between sheaves. Let E and F be two sheaves on a topological space
X. A map f : E → F is a collection of group homomorphisms
{fU : E (U ) → F (U )}
such that they commute with the restriction maps, i.e., for any open sets U ⊂ V
and σ ∈ E (V ) we have
fV (σ)|U = fU (σ|U ).
Define the sheaf of kernel ker(f ) as
ker(f )(U ) = {ker(fU : E (U ) → F (U ))}.
Exercise 2.21. Prove that in the above definition ker(f ) is a sheaf.
Example 2.22. Let X be a complex manifold. Define the exponential map
exp : O → O ∗
by exp(h) = e2πih for any open set U ⊂ X and section h ∈ O(U ). It is easy to see
that ker(exp) is the locally constant sheaf Z.
The sheaf of cokernel is harder to define. Naively, one would like to define
coker(f )(U ) = coker(fU : E (U ) → F (U )), but this is problematic. For instance,
consider the exponential map exp : O → O ∗ on the punctured plane C\{0}. The
section z ∈ O ∗ (C\{0}) is not in the image of f , hence it would define a section in
the cokernel. Nevertheless, restricted to any contractible open set U ⊂ C\{0}, z
lies in the image of f . Now cover C\{0} by contractible open sets. By the gluing
property of sheaves, z would be zero everywhere, leading to a contradiction.
Instead, we define a section of coker(f )(U ) to be a collection of sections σα ∈
F (Uα ) for an open covering {Uα } of U such that for all α, β we have
σα |Uα ∩Uβ − σβ |Uα ∩Uβ ∈ fUα ∩Uβ (E (Uα ∩ Uβ )).
Here the definition still depends on the choice of an open covering. To get rid of
this, we pass to the direct limit. More precisely, we further identify two collections
{(Uα , σα )} and {(Vβ , σβ } if for all p ∈ Uα ∩Vβ , there exists an open set W satisfying
p ∈ W ⊂ Uα ∩ Vβ such that
σα |W − σβ |W ∈ fW (E (W )).
This identification yields an equivalence relation and correspondingly we define
coker(f )(U ) as the group of equivalence classes of the above sections.
 39

Exercise 2.23. Prove that in the above definition coker(f ) is a sheaf.

If ker(f ) (resp. coker(f )) is the zero sheaf, we say that f is injective (resp.
surjective).
Consider the following sequence of maps between sheaves:
α β
0 −−−−→ E −−−−→ F −−−−→ G −−−−→ 0.
We say that it is a short exact sequence if E = ker(β) and G = coker(α). In this
case we also say that E is a subsheaf of F and G is the quotient sheaf F /E .
Example 2.24. Let X be a complex manifold. We have the exact exponential
sequence:
i exp
0 −−−−→ Z −−−−→ O −−−−→ O ∗ −−−−→ 0,
√
where i is the natural inclusion and exp(f ) = e2π −1f for f ∈ O(U ).
Exercise 2.25. Prove that the exponential sequence is exact.
Example 2.26. Let X be a complex manifold and Y ⊂ X a submanifold. Define
the ideal sheaf IY /X of Y in X (or simply IY if there is no confusion) by
IY (U ) = {holomorphic functions in U vanishing on Y ∩ U }.
We have the exact sequence:
i r
0 −−−−→ IY −−−−→ OX −−−−→ i∗ OY −−−−→ 0,
where i is the natural inclusion and r is defined by the natural restriction map.
Here i∗ OY is the extension of OY by zero outside Y , as a sheaf defined on X.
Exercise 2.27. Prove that the above sequence is exact.
2.4. Stalks and germs. Let F be a sheaf on a topological space X and p ∈ X a
point. Suppose U and V are two open subsets, both containing p, with two sections
α ∈ F (U ) and β ∈ F (V ). Define an equivalence relation α ∼ β, if there exists an
open subset W satisfying p ∈ W ⊂ U ∩ V such that α|W = β|W . Define the stalk
Fp as the union of all sections in open neighborhoods of p modulo this equivalence
relation. Namely, Fp is the direct limit
G 
Fp := lim F (U ) = F (U ) / ∼ .
−→
U 3p U 3p

Note that Fp is also a group, by adding representatives of two equivalence classes.

There is a group homomorphism rU : F (U ) → Fp mapping a section α ∈ F (U )
to its equivalence class. The image is called the germ of α.
Example 2.28 (Skyscraper sheaf). Let p ∈ X be a point on a topological space X.
Define the skyscraper sheaf F at p by F (U ) = {0} for p 6∈ U and F (U ) = A for
p ∈ U , where A is an abelian group. The restriction maps are either the identity
map A → A or the zero map. For q 6= p, the stalk Fq = {0}. At p, we have
Fp = A. Note that F can also be obtained by extending the constant sheaf A at
p by zero to X\{p}.
Exercise 2.29. Let X be a Riemann surface and p ∈ X a point. Let Ip be the
ideal sheaf of p in X parameterizing holomorphic functions vanishing at p. We have
the exact sequence
i r
0 −−−−→ Ip −−−−→ OX −−−−→ Op −−−−→ 0.
40

Show that the quotient sheaf Op is isomorphic to the skyscraper sheaf with stalk C
at p.
It is more convenient to verify injectivity and surjectivity for maps of sheaves by
using stalks.
Proposition 2.30. Let φ : E → F be a map for sheaves E and F on a topological
space X.
(1) φ is injective if and only if the induced map φp : Ep → Fp is injective for the
stalks at every point p.
(2) φ is surjective if and only if the induced map φp : Ep → Fp is surjective for
the stalks at every point p.
(3) φ is an isomorphism if and only if the induced map φp : Ep → Fp is an
isomorphism for the stalks at every point p.
Proof. The claim (3) follows from (1) and (2). Let us prove (1) only, and one can
find the proof of (2) in many standard textbooks.
Suppose φ is injective. Take a section σ ∈ E (U ) on an open subset U . If φ([σ]) =
0 ∈ Fp , there exists a smaller open subset V ⊂ U such that φV (σ) = 0 ∈ F (V ),
hence σ|V = 0 ∈ E (V ). Consequently the equivalence class [σ] = 0 ∈ Ep and we
conclude that φp is injective.
Conversely, suppose φp is injective for every point p. Take a section σ ∈ E (U ).
If φ(σ) = 0 ∈ F (U ), then for every point p ∈ U , [φ(σ)] = 0 ∈ Fp . Since φp is
injective, it implies that [σ] = 0 ∈ Ep i.e. there exists an open subset Up 3 p such
that σ|Up = 0 ∈ E (Up ). Applying the gluing property to the open covering {Up } of
U , we conclude that σ = 0 ∈ E (U ). 
Remark 2.31. The image of φ does not automatically form a sheaf. In general,
it is only a presheaf. If the sheafification of Im(φ) equals F , we say that φ is
surjective. In particular, it does not mean E (U ) → F (U ) is surjective for every
open set U . Sometimes one has to pass to a refined open covering in order to obtain
a surjection between sections.
Example 2.32. Consider the exponential map exp : O → O ∗ on the punctured
plane C\{0}. As a map of sheaves it is surjective, but the section z over C\{0}
does not have an inverse. Nevertheless, it does have an inverse over any contractible
open subset.
2.5. Sheaf cohomology. Let F be a sheaf on a topological space X. Take an
open covering U = {Uα } of X. Define the k-th cochain group
Y
C k (U , F ) := F (Uα0 ∩ · · · ∩ Uαk ).
α0 ,...,αk

An element σ of C (U , F ) consists of a section σα0 ,...,αk ∈ F (Uα0 ∩ · · · ∩ Uαk ) for

k

every Uα0 ∩ · · · ∩ Uαk , where α0 , . . . , αk are ordered.

Define a coboundary map δ : C k (U , F ) → C k+1 (U , F ) by
k+1
X
(δσ)α0 ,...,αk+1 = (−1)j σα0 ,...,α̂j ,...,αk+1 |Uα0 ∩···∩Uαk+1 .
j=0

Example 2.33. Consider U = {U1 , U2 , U3 } as an open covering of X. Take a

cochain element σ ∈ C 0 (U , F ), i.e. σ is a collection of a section σi ∈ F (Ui ) for
 41

every i. Then we have

(δσ)ij = (σj − σi )|Ui ∩Uj ∈ F (Ui ∩ Uj ).
Now take τ ∈ C 1 (U , F ), i.e. τ is a collection of a section τij ∈ F (Ui ∩ Uj ) for
every pair i, j. Then we have
(δτ )123 = (τ23 − τ13 + τ12 )|U1 ∩U2 ∩U3 ∈ F (U1 ∩ U2 ∩ U3 ).
A cochain σ ∈ C k (U , F ) is called a cocyle if δσ = 0. We say that σ is a
coboundary if there exists τ ∈ C k−1 (U , F ) such that δτ = σ.
Lemma 2.34. A coboundary is a cocycle, i.e. δ ◦ δ = 0.
Proof. Let us prove it for the above example. The same idea applies in general
with messier notation. Under the above setting, we have
((δ ◦ δ)σ)123 = (δσ)23 − (δσ)13 + (δσ)12
= (σ3 − σ2 ) − (σ3 − σ1 ) + (σ2 − σ1 )
= 0 ∈ F (U1 ∩ U2 ∩ U3 ).
Here we omit the restriction notation, since it is obvious. 
Exercise 2.35. Prove in full generality that δ ◦ δ = 0.
For the coboundary map δk : C k (U , F ) → C k+1 (U , F ), define the k-th coho-
mology group (respect to U ) by
ker(δk )
H k (U , F ) := .
Im(δk−1 )
This is well-defined due to the above lemma.
Example 2.36. For k = 0, we have H 0 (U , F ) = ker(δ0 ). Take an element {σi ∈
F (Ui )} in this group. Because it is a cocycle, it satisfies
σi |Ui ∩Uj = σj |Ui ∩Uj ∈ F (Ui ∩ Uj ).
By the gluing property of sheaves, there exists a global section σ ∈ F (X) such
that σ|Ui = σi . Conversely, if σ is a global section, then define σi = σ|Ui ∈ F (Ui ).
In this way we obtain a cocycle in C 1 (U , F ). From the discussion we see that
H 0 (U , F ) = F (X), which is independent of the choice of an open covering. Hence
H 0 (U , F ) is called the group of global sections of F and we often denote it by
H 0 (X, F ) or simply H 0 (F ).
In general, we would like to define cohomology independent of open coverings.
Take two open coverings U = {Uα }α∈I and V = {Vβ }β∈J . We say that U is a
refinement of V if for every Uα there exists a Vβ such that Uα ⊂ Vβ and we write it
as U < V . Then we also have an index map φ : I → J sending α to β. It induces
a map
ρφ : C k (V , F ) → C k (U , F )
given by
ρφ (σ)α0 ,...,αk = σφ(α0 ),...,φ(αp ) |Uα0 ∩···∩Uαk .
One checks that it commutes with the coboundary map δ, i.e. δ ◦ ρφ = ρφ ◦ δ.
Moreover, it induces a map
ρ : H k (V , F ) → H k (U , F ),
42

which is independent of the choice of φ. Finally, we define the k-th (Čech) coho-
mology group by passing to the direct limit:
H k (X, F ) := lim H k (U , F ).
−→
The definition involves direct limit, which is inconvenient to use in practice.
Nevertheless, we can simplify the situation once the open covering U is fine enough.
We say that U = {Ui }i∈I is acyclic respect to F , if for any k > 0 and i1 , . . . , il ∈ I
we have
H k (Ui1 ∩ · · · ∩ Uil , F ) = 0.
Theorem 2.37 (Leray’s Theorem). If the open covering U is acyclic respect to F ,
then H ∗ (U , F ) ∼
= H ∗ (X, F ).
Remark 2.38. In the context of complex manifolds, if Ui ’s are contractible, then
U is acyclic respect to the sheaves we will consider. While for algebraic varieties,
if Ui ’s are affine, then U is acyclic.
Example 2.39. Let us compute the cohomology of the structure sheaf O on P1 .
It is clear that H 0 (P1 , O) = C, since any global holomorphic function on P1 is
constant. For higher cohomology, use [X, Y ] to denote the homogeneous coordinates
of P1 . Take the standard open covering U = {[X, Y ] : X 6= 0} and V = {[X, Y ] :
Y 6= 0}. It is acyclic respect to the structure sheaf O (morally because U, V ∼ =C
is contractible). Let s = Y /X and t = X/Y as affine coordinates of U and V ,
respectively. Suppose h is an element in C 1 ({U, V }, O), i.e. h ∈ O(U ∩ V ). We can
write
X∞
h= ai si .
i=−∞
Now take
∞
X
f =− ai si ∈ O(U ),
i=0
−1
X −1
X
g= ai si = ai t−i ∈ O(V ).
i=−∞ i=−∞

Then we have (f, g) ∈ C 0 ({U, V }, O) and δ((f, g)) = g − f = h. It implies that

H 1 (P1 , O) = 0. All the other H k (P1 , O) = 0 for k > 1, since there are only two
open subsets in the covering.
Example 2.40. Let Ω denote the sheaf of holomorphic one-forms on a Riemann
surface, i.e. locally a section of Ω can be expressed as f (z)dz, where z is local
coordinate and f (z) a holomorphic function, satisfying the obvious change of coor-
dinates. Let us compute the cohomology of Ω on P1 . Take the above open covering.
Suppose ω is a global holomorphic one-form. Then on the open chart U , it can be
written as
X ∞ 
ai si ds.
i=0
Using the relation s = 1/t and ds = −dt/t2 , on V it can be expressed as
X∞ 
− ai t−i−2 dt,
i=0
 43

which is holomorphic if and only if ai = 0 for all i. Hence w is the zero one-form
and H 0 (P1 , Ω) = 0. Now take ω ∈ C 1 ({U, V }, Ω), i.e. ω ∈ Ω(U ∩ V ) = Ω(C∗ ), we
express it as
 X ∞ 
ω= ai ti dt.
i=−∞
Note that any α ∈ Ω(U ) and β ∈ Ω(V ) can be written as
X ∞ 
α= bi si ds,
i=0
∞
X 
β= ci ti dt.
i=0
Hence on U ∩ V we have
∞
X  ∞
X 
δ((α, β)) = β − α = − bi t−i−2 dt + ci ti dt.
i=0 i=0
−1
Note that only the term t is missing from the expression. We conclude that
H 1 (P1 , Ω) = {a−1 t−1 dt} ∼
= C.
Remark 2.41. In general, the rank of H 1 (X, O) ∼ = H 0 (X, Ω) (by Serre Duality)
is called the genus of a Riemann surface (or a complex algebraic curve) X.
Exercise 2.42. Let D = p1 + · · · + pn be a collection of n points in P1 . We say
that D is an effective divisor of degree n. Define the sheaf O(D) on P1 by
O(D)(U ) = {f ∈ M (U ) : f ∈ O(U \{p1 , . . . , pn }) with at worst a simple pole at each pi }.
Assume that the standard covering of P1 is acyclic respect to O(D). Use it to
calculate the cohomology groups H ∗ (P1 , O(D)).
As many other homology/cohomology theories, one can associate a long exact
sequence of cohomology to a short exact sequence. Suppose we have a short exact
sequence of sheaves
α β
0 −−−−→ E −−−−→ F −−−−→ G −−−−→ 0.
Then α and β induce maps
α : C k (U , E ) → C k (U , F ), β : C k (U , F ) → C k (U , G ).
Since the coboundary map δ is given by alternating sums of restrictions, α and
β commute with δ, hence they send a cocycle to cocycle and a coboundary to
coboundary. Consequently they induce maps for colomology
α∗ : H k (X, E ) → H k (X, F ), β∗ : H k (X, F ) → H k (X, G ).
Next we define the coboundary map
δ∗ : H k (X, G ) → H k+1 (X, E ).
For σ ∈ C k (U , G ) satisfying δσ = 0, after refining U (still denoted by U ) such
that there exists τ ∈ C k (U , F ) satisfying β(τ ) = σ, because β is surjective. Then
β(δτ ) = δ(β(τ )) = δσ = 0, hence after refining further there exists µ ∈ C k+1 (U , E )
satisfying α(µ) = δτ . Note that µ is a cocycle. It is because α(δµ) = δ(α(µ)) =
δδ(τ ) = 0 and α is injective, hence δµ = 0 and µ ∈ ker(δ). We thus take δ∗ σ :=
[µ] ∈ H k+1 (X, E ). One checks that this is independent of the choice of τ and µ.
44

We say that a sequence of maps

αn−1 α
· · · −−−−→ An−1 −−−−→ An −−−n−→ An+1 −−−−→ · · ·
is exact if Im(αn−1 ) = ker(αn ).
Proposition 2.43. The long sequence of cohomology associated to a short exact
sequence of sheaves is exact.
Proof. We prove it under an extra assumption that there exists an acyclic open
covering U such that for any U = Ui1 ∩ · · · ∩ Uik we have the short exact sequence:
0 → E (U ) → F (U ) → G (U ) → 0.
At least for sheaves considered in this course, this assumption always holds. It
further implies the following sequence is exact:
0 → C k (U , E ) → C k (U , F ) → C k (U , G ) → 0.
Let us prove that
β∗ δ
H k (U , F ) −−−−→ H k (U , G ) −−−∗−→ H k+1 (U , E )
is exact. The other cases are easier.
Consider τ ∈ Z k (U , F ). In the definition of δ∗ , take σ = β(τ ). Then there exists
µ ∈ C k (U , E ) such that α(µ) = δτ = 0. Then we have µ = 0 since α is injective.
Consequently δ∗ β∗ (τ ) = δ∗ (σ) = µ = 0, hence δ∗ β∗ = 0 and Im(β∗ ) ⊂ ker(δ∗ ).
Conversely, suppose δ∗ σ = 0 for σ ∈ Z k (U , G ). In the definition of δ∗ , it implies
that µ = 0 ∈ H k+1 (U , E ), hence there exists γ ∈ C k (U , E ) such that δγ = µ.
Since α(µ) = δτ , we have δτ = δα(γ) and τ − α(γ) ∈ Z k (U , F ) is a cocycle.
Moreover, β(τ − α(γ)) = β(τ ) = σ, hence β∗ (τ − α(γ)) = σ. We conclude that
ker(δ∗ ) ⊂ Im(β∗ ). 

Exercise 2.44. Prove in general the cohomology sequence is exact.

Example 2.45. Consider the short exact sequence
i r
0 −−−−→ Ip −−−−→ OP1 −−−−→ Op −−−−→ 0.
Its long exact sequence of cohomology is as follows:
0 → H 0 (Ip ) → H 0 (OP1 ) → H 0 (Op ) → H 1 (Ip ) → H 1 (OP1 ) → 0.
The last term is zero because p is a point so it does not have higher cohomology.
We have H 0 (OP1 ) = C because any global regular function on P1 is constant. Note
that H 0 (Ip ) = 0, because vanishing at p forces such a constant function to be
zero. Moreover we have seen that H 1 (OP1 ) = 0. Altogether it implies H 1 (Ip ) = 0,
because H 0 (OP1 ) → H 0 (Op ) is an isomorphism by evaluating at p.
Exercise 2.46. Let D be an effective divisor of degree n on P1 . We have the short
exact sequence
i r
0 −−−−→ ID −−−−→ OP1 −−−−→ OD −−−−→ 0.
Use the associated long exact sequence to calculate the cohomology H ∗ (P1 , I (D)).
 45

2.6. Holomorphic vector bundles. Let k be a positive integer. Consider π :

E → X a holomorphic map between complex manifolds, such that for every x ∈ X
the fiber Ex = π −1 (x) is isomorphic to Ck and there exists an open neighborhood
U of x along with an isomorphism
∼
φU : EU = π −1 (U ) −
→ U × Ck
mapping Ex to {x}×Ck which is a linear isomorphism between vector spaces. Then
E is called a holomorphic vector bundle of rank k on X and has a trivialization
{(U, φU )}. If E is of rank one, we say that E is a line bundle.
We give another characterization of vector bundles based on transition functions.
Suppose U = {Uα } is an open covering of X. Given holomorphic functions
gαβ : Uα ∩ Uβ → GL(Ck ),
we can construct a vector bundle E by gluing Uα × Ck together. More precisely,
E = t(Uα × Ck )/ ∼
as a complex manifold is defined by identifying (x, v) with (x, gαβ (v)) for x ∈ Uα ∩Uβ
and v ∈ Ck and E → X is given by projection to the bases Uα . Call {gαβ }
the transition functions of E. They have to satisfy the following compatibility
conditions:
gαβ (x) · gβα (x) = I, for all x ∈ Uα ∩ Uβ ,
gαβ (x) · gβγ (x) · gγα (x) = I, for all x ∈ Uα ∩ Uβ ∩ Uγ .
Exercise 2.47. Let E and F be two vector bundles on X of rank k and l, respec-
tively. Define the direct sum E ⊕ F , the tensor product E ⊗ F , the dual E ∗ , and the
wedge product ∧r E for r ≤ k. Calculate the ranks of these bundles and represent
their transition functions in terms of the transition functions of E and F .
A map between two vector bundles E and F on X is given by a holomorphic
map f : E → F such that f (Ex ) ⊂ Fx and fx = f |Ex : Ex → Fx is linear. Note
that if f (Ex ) has the same rank for every x, then ker(f ) and Im(f ) are naturally
subbundles of E and F , respectively. We say that E and F are isomorphic if fx is
a linear isomorphism for every x. A vector bundle is called trivial if it is isomorphic
to X × Ck .
Exercise 2.48. Give an example of a map between vector bundles f : E → F on
X such that the the image of f is not a vector bundle.
Exercise 2.49. Let L be a line bundle on X. Prove that L ⊗ L∗ is a trivial line
bundle.
Define a section σ as a holomorphic map σ : X → E such that σ(x) ∈ Ex for
every x ∈ X, i.e. π ◦ σ is identity. If σ(x) = 0 ∈ Ex , we say that σ is vanishing on
x.
Exercise 2.50. Let L be a line bundle on X. Prove that L is trivial if and only if
it possesses a nowhere vanishing section.
Example 2.51 (Holomorphic tangent bundles). Let X be an n-dimensional com-
plex manifold. Suppose φU : U → Cn are coordinate charts of X. Define the
(holomorphic) tangent bundle TX by setting TX = tTx with
∂ ∂ ∼ n
Tx = C{ ,..., }=C
∂z1 ∂zn
46

as well as transition functions gU V = J(φV φ−1 U ), where J denotes the Jacobian

matrix ( ∂w
∂zj
i
) for 1 ≤ i, j ≤ n. The dual bundle ∗
TX is called the cotangent bundle
∗
of X. The determinant det(TX ) is called the canonical line bundle of X.
Remark 2.52. Alternatively, one can define vector bundles on a topological space,
a differential manifold and an algebraic variety. The above definitions and prop-
erties go through word by word once replacing “holomorphic map” by “homomor-
phism”, “smooth map” or “regular map”.
2.7. Vector bundles and locally free sheaves. There is a one-to-one correspon-
dence between isomorphism classes of vector bundles of rank n and isomorphism
classes of locally free sheaves of rank n on a variety X. Here we briefly explain the
idea. The reader can refer to Hartshorne II 5, especially Ex. 5.18 for more details.
Let OX be the structure sheaf of a complex manifold X. Note that OX (U ) has
a ring structure (not only a group) for any open set U . A sheaf of OX -modules is a
sheaf F on X such that for each open set U , the group F (U ) is an OX (U )-module.
An OX -module F is called free if it is isomorphic to a direct sum of OX . It is called
locally free if there is an open covering U = {Uα } such that for each open subset
Uα , F |Uα is a free OX |U -module. The rank of F on U is the number of copies
of O in the summation. If X is connected, the rank of F does not vary with the
open subsets. In particular, a locally free sheaf of rank 1 is also called an invertible
sheaf.
Roughly speaking, if F is locally free of rank n, we can choose a set of n gener-
ators x1 , . . . , xn for the OX (U )-module F (U ). They span an n-dimensional vector
space C[x1 , . . . , xn ] over U . By changing to a different set of generators over an-
other open subset, one can write down the transition functions, hence it associates
to F a vector bundle structure. Conversely if F is a vector bundle on X, locally we
have F |U ∼ = U × Cn with x1 , . . . , xn a basis (i.e. n linearly independent sections)
of C over U . Then we can associate to F |U an OX (U )-module of rank n using
n

x1 , . . . , xn as generators.
Example 2.53. Let X ⊂ Pn be a submanifold and Y ⊂ X a hypersurface, i.e.
Y is cut out (transversely) by a hypersurface F in Pn with X. We have the short
exact sequence
0 → IY /X → OX → OY → 0.
The ideal sheaf IY /X is an invertible sheaf. Indeed, for an open subset U ⊂ X,
IY /X (U ) can be expressed as (F |U ) · OX (U ), hence is locally free of rank 1. The
sheaf OY (extended to X by zero) is not locally free. For U ∩ Y = ∅, OY (U ) = 0
and for U ∩ Y 6= ∅, OY (U ) is non-zero. Later we will see how to construct a line
bundle corresponding to IY /X .
2.8. Divisors. Let X be a complex manifold. Suppose Y ⊂ X is an irreducible
subspace of codimension one satisfying that for every p ∈ Y there exists an open
neighborhood U ⊂ X of p such that U ∩ Y is cut out by a holomorphic function
f . Then we say that Y is an irreducible divisor of X. We call f a local defining
equation for Y near p. A divisor D on X is a formal linear combination of irreducible
divisors:
X n
D= ai Yi ,
i=i
 47

where ai ∈ Z (or Q, R depending on the context). If ai ≥ 0 for all i, we say that

D is effective and denote it by D ≥ 0. The divisors on X form an additive group
Div(X).
Suppose f is a local defining equation of an irreducible divisor Y ⊂ X on an
open subset U ⊂ X. For another function g on X, locally we can write
g = fa · h
such that the regular function h is coprime with f in OX (U ). We say that a is
the vanishing order of g along Y ∩ U . Note that the vanishing order is locally a
constant, hence is independent of U . We use
ordY (g) = a
to denote the vanishing order of g along Y .
For two holomorphic functions g, h on X, we have
ordY (gh) = ordY (g) + ordY (h).
For a function f = g/h, we define
ordY (f ) = ordY (g) − ordY (h).
If ordY (f ) > 0, we say that f has a zero along Y . If ordY (f ) < 0, we say that f
has a pole along Y . We also define the divisor associated to f by
X
(f ) = ordY (f ),
Y
as well as the divisor of zeros
X
(f )0 = ordY (g)
Y
and the divisor of poles X
(f )∞ = ordY (h).
Y
They satisfy
(f ) = (f )0 − (f )∞ .
If D = (f ) is the associated divisor of a global meromorphic function f , D is called
a principal divisor.
Recall that M ∗ is the multiplicative sheaf of (not identically zero) meromorphic
functions and O ∗ the multiplicative sheaf of nowhere vanishing regular functions,
which is a subsheaf of M ∗ .
Proposition 2.54. Div(X) ∼
= H 0 (X, M ∗ /O ∗ ).
Proof. Suppose {fα } represents a global section of M ∗ /O ∗ with respect to an open
covering U = {Uα }. Associate to it a divisor Dα = (fα ) in Uα . We claim that
Dα = Dβ in Uα ∩ Uβ . This is due to
fα
∈ O ∗ (Uα ∩ Uβ ),
fβ
hence fα and fβ define the same divisor. Consequently {Dα } defines a global
divisor. Moreover, if {fα } and {gα } define the same divisor, then fα /gα ∈ O ∗ (Uα ),
hence {fα } and {gα } represent the same section of M ∗ /O ∗ . This shows an injection
H 0 (X, M ∗ /O ∗ ) ,→ Div(X).
48

P
Conversely, suppose D = ai Yi is a divisor on X with ai ∈ Z and Yi effective.
We can choose an open covering U = {Uα } such that Yi is locally defined by
giα ∈ O(Uα ). Consider Y
fα = (giα )ai ∈ M ∗ (Uα ).
i
Then we have
fα Y  giα ai
= .
fβ i
giβ
Both giα and giβ cut out the same divisor Yi |Uα ∩Uβ in Uα ∩ Uβ , hence we conclude
that
giα fα
∈ O ∗ (Uα ∩ Uβ ), ∈ O ∗ (Uα ∩ Uβ ).
giβ fβ
Then {fα } defines a global section of M ∗ /O ∗ . Finally if D determines the zero
section of M ∗ /O ∗ (which is 1 since the group structure is multiplicative), it means
locally fα ∈ O ∗ (Uα ) (after refining the open covering). Then it does not have zeros
or poles, hence D|Uα = 0 for each Uα and D is globally zero. This shows the other
injection
Div(X) ,→ H 0 (X, M ∗ /O ∗ ).

2.9. Line bundles. Recall that a line bundle L on X is a vector bundle of rank
1. Equivalently, it is a locally free sheaf of rank 1. Define the Picard group Pic(X)
parameterizing isomorphism classes of line bundles on X. The group law is given
by tensor product. We can interpret Pic(X) as a cohomology group.
Proposition 2.55. There is a one-to-one correspondence between the isomorphism
classes of line bundles on X and H 1 (X, O ∗ ), i.e.
Pic(X) ∼= H 1 (X, O ∗ ).
Proof. Take an open covering U = {Uα } of X with respect to the trivialization of
a line bundle L. The transition function
gαβ : (Uα ∩ Uβ ) × C → (Uα ∩ Uβ ) × C
can be regarded as a section of O ∗ (Uα ∩ Uβ ), satisfying
gαβ · gβα = 1,
gαβ · gβγ · gγα = 1.
Therefore, {gαβ } is a cocycle in C 1 (U , O ∗ ), hence represents a cohomology class in
H 1 (X, O ∗ ).
Suppose M is another line bundle with transition functions {hαβ }. If M and
L are isomorphic, then L ⊗ M ∗ is trivial, i.e. {gαβ /hαβ } are transition functions
of L ⊗ M ∗ , which has a nowhere vanishing section σ. Suppose on Uα we have
σα : Uα → C∗ as the restriction of σ. Then on Uα ∩ Uβ we have
gαβ
· σα = σβ .
hαβ
Therefore we conclude that
gαβ σβ
= ∈ δC 0 (U , O ∗ ).
hαβ σα

 49

Now we describe another important correspondence between line bundles and

divisors. Suppose D is a divisor on X with local defining equations {fα } such that
fα ∈ M ∗ (Uα ). Define
fβ
gαβ = .
fα
Then we have gαβ ∈ O ∗ (Uα ∩Uβ ). Moreover, {gαβ } satisfy the assumptions imposed
to transition functions, hence they define a line bundle, denoted by L = [D] or
L = OX (D). We have a group homomorphism
Div(X) → Pic(X)
induced by
D + D0 7→ [D] ⊗ [D0 ].
We say that D and D0 are linearly equivalent, if [D] and [D0 ] are isomorphic line
bundles. We denote linear equivalence by
D ∼ D0 .
The following result says that the kernel of the above map consists of principal
divisors. In other words, two divisors D ∼ D0 if and only if D − D0 is a principal
divisor.
Proposition 2.56. The associated line bundle [D] is trivial if and only if D is a
principal divisor, i.e. D = (f ) for some f ∈ M ∗ (X).
Proof. Suppose D = (f ) is the associated divisor of a meromorphic function f on
X. Then D has local defining equations {fα = f |Uα }. The transition functions
associated to [D] are all equal to 1, hence [D] is a trivial line bundle. Conversely,
suppose [D] is trivial. Then it has a nowhere vanishing section σ whose restriction
to Uα is denoted by σα . The transition functions gαβ = fβ /fα defined above satisfy
gαβ · σα = σβ ,
hence we have
fα fβ
= ∈ M ∗ (Uα ∩ Uβ ).
σα σβ
We can glue {fα /σα } to form a global function f ∈ M ∗ (X). Since σ is nowhere
vanishing, we obtain that (f ) = D. 

Let us summarize, using the short exact sequence

0 → O ∗ → M ∗ → M ∗ /O ∗ → 0.
Recall that
Div(X) ∼
= H 0 (X, M ∗ /O ∗ ), Pic(X) ∼
= H 1 (X, O ∗ ).
Then we have the long exact sequence
(·) [·]
0 → H 0 (X, O ∗ ) → H 0 (X, M ∗ ) −→ Div(X) −→ Pic(X) → · · ·
which encodes all the information in the above discussions.
50

2.10. Sections of a line bundle. Let L be a line bundle on X with transition

functions {gαβ }. A holomorphic section s of L has restriction sα ∈ O(Uα ), satisfying
gαβ sα = sβ .
Conversely, a collection {sα ∈ O(Uα )} such that sβ /sα = gαβ determines a section
of L.
Similarly, we define a meromorphic section s to be a collection
{sα ∈ M (Uα )}
such that gαβ sα = sβ . Suppose t 6≡ 0 is another meromorphic section with collection
{tα }. We have
sβ sα
= ,
tβ tα
hence the quotient s/t is a global meromorphic function. Conversely, if f is a global
meromorphic function, then {f · sα } defines another meromorphic section of L.
For a meromorphic section s 6≡ 0, consider the divisor (sα ) associated to the
local section sα in Uα . Since
sβ
= gαβ ∈ O ∗ (Uα ∩ Uβ ),
sα
{(sα )} form a global divisor (s) on X. Conversely, suppose D is a divisor. Con-
sider the construction of the associated line bundle [D]. Suppose the local defin-
ing equations of D are given by {sα }. Then the transition functions of [D] are
{gαβ = sβ /sα } and consequently the collection {sα } gives rise to a meromorphic
section of [D]. Note that for a section s of L, the divisor (s) is effective if and only
if s is a holomorphic section. We thus obtain the following result.
Proposition 2.57. For any section s of L, we have L ∼ = [(s)]. A line bundle L is
associated to a divisor D if and only if it has a meromorphic section s such that
(s) = D. In particular, L has a holomorphic section if and only if it is associated
to an effective divisor.
Now we treat a line bundle as a locally free sheaf of rank 1 and reinterpret the
above correspondence. Let D be a divisor on X. Define a sheaf OX (D) or simply
O(D) by
O(D)(U ) = {f ∈ M (U ) : (f ) + D|U ≥ 0}.
It has a vector space structure since (f ) + D|U ≥ 0 and (g) + D|U ≥ 0 implies that
(af + bg) + D|U ≥ 0 for any a, b in the base field.
Proposition 2.58. The space of holomorphic sections of [D] can be identified with
H 0 (X, O(D)).
Proof. A global section s ∈ H 0 (X, O(D)) is a meromorphic function satisfying
(s) + D ≥ 0.
Suppose D is locally defined by {fα }. The associated line bundle [D] has transition
functions
fβ
{gαβ = }.
fα
Then the collection {s · fα } defines a section σ of [D]. Since
(s) + (fα ) ≥ 0
 51

in every Uα , σ is a holomorphic section of [D]. Moreover, the associated divisor

D0 = (s) + D of the section is linearly equivalent to D, since D0 − D = (s) is
principal.
Conversely, given a holomorphic section σ of [D], i.e. a collection {hα ∈ O(Uα )}
such that
hβ fβ
= gαβ = .
hα fα
Then {hα /fα } defines a global meromorphic function g. Since (hα ) ≥ 0 in every
Uα , we have
(g|Uα ) + (fα ) = (hα ) ≥ 0,
hence (g) + D ≥ 0 globally on X and g ∈ H 0 (X, O(D)). 

Remark 2.59. Replacing X by any open subset U , the proposition implies that
the sheaf O(D) can be regarded as gathering local holomorphic sections of the line
bundle [D]. If D ∼ D0 , i.e. D0 − D = (f ) for a global meromorphic function f ,
then for any g ∈ O(D0 )(U ), we have

0 ≤ (g) + D0 = (g) + (f ) + (D) = (f g) + (D)

restricted to U . So we obtain an isomorphism

·f
O(D0 )(U ) −→ O(D)(U )

for any open subset U , compatible with the sheaf restriction maps. In this sense,
the sheaf O(D) and the line bundle [D] have a one-to-one correspondence up to
isomorphism and linear equivalence, respectively, assuming that every line bundle
can be associated to a divisor.

Let |D| be the set of effective divisors that are linearly equivalent to D. We call
|D| the linear system associated to D.

Proposition 2.60. Let X be compact and D a divisor on X. Then we have

PH 0 (X, O(D)) = |D|,

i.e. an effective divisor in |D| and a holomorphic section of [D] (up to scalar)
determine each other.

Proof. For any D0 ∈ |D|, by definition D0 −D = (f ) is principal for some f ∈ M (X),

hence (f ) + D = D0 ≥ 0 and f ∈ H 0 (X, O(D)). Since X is compact, if g is another
function such that D0 − D = (g), then (f /g) = 0, i.e. f /g is holomorphic, hence it
is a constant.
Conversely, any f ∈ H 0 (X, O(D)) defines an effective divisor D0 = (f ) + D. If
(f ) + D = (g) + D, then (f /g) = 0 and f /g is a constant, since X is compact. 

1 1
P
Exercise 2.61. Let D = i pi be a divisor on P with ai ∈ Z and pi ∈ P . Define
aP
the degree of D by deg(D) = ai .
(1) Prove that D ∼ D0 if and only if deg(D) = deg(D0 ).
(2) Calculate the cohomology H ∗ (P1 , O(D)) in terms of deg(D).
52

2.11. The Riemann-Roch formula. Let X be a compact connected Riemann

surface (or smooth complex curve). Define its arithmetic genus by
g := h1 (OX ) = dimC H 1 (OX ).
Theorem 2.62 (Riemann-Roch Formula). Let D be a divisor on X and O(D) the
associated line bundle (or invertible sheaf ). Then we have
h0 (O(D)) − h1 (O(D)) = 1 − g + deg(D).
Remark 2.63. Define the (holomorphic) Euler characteristic of a sheaf F by
X
χ(F ) := (−1)i hi (F ).
i≥0

Then the Riemann-Roch formula can be written as

χ(O(D)) − χ(OX ) = deg(D).
Proof. Let us first prove it for effective divisors of degree ≥ 0. Do induction on n.
The formula obviously holds for OX . Suppose it is true for deg(D) < n. Consider
D = p + D0 with D0 an effective divisor of degree n − 1. We have the short exact
sequence
0 → O(D0 ) → O(D) → Cp → 0,
where Cp is the skyscraper sheaf with one-dimensional stalk supported at p. The
exactness can be easily checked. The map O(D0 ) → O(D) is an inclusion, since
(f ) + D0 ≥ 0
implies that
(f ) + D = (f ) + D0 + p ≥ 0.
The quotient corresponds to germs of functions f at p such that
(f )|U + D0 |U = −p
in arbitrarily small neighborhoods U of p. In other words, if ordp (D0 ) = m ≥ 0, we
can write
f = z −m−1 h(z),
where h ∈ O ∗ (U ). So the quotient sheaf is given by C · {z −m−1 } ∼
= C supported at
p. Since the associated cohomology sequence is long exact, we have
χ(O(D)) = χ(O(D0 )) + 1 = 1 − g + (n − 1) + 1 = 1 − g + n.
In general, write a divisor D = D1 − D2 , where D1 and D2 are both effective
divisors of degree d1 and d2 , respectively, and d1 − d2 = deg(D). By the same
token, we have the short exact sequence
0 → O(D) → O(D1 ) → Cd2 → 0.
Then we obtain that
χ(O(D)) = χ(O(D1 )) − d2 = 1 − g + d1 − d2 = 1 − g + deg(D).

 53

Remark 2.64. Assuming the Serre duality

H 1 (O(D)) =∼ H 0 (K ⊗ O(−D)),
where K is the canonical line bundle of X, then we can rewrite the Riemann-Roch
formula as
h0 (L) − h0 (K ⊗ L∗ ) = 1 − g + deg(L),
where L is a line bundle on X. Note that K is a degree 2g − 2 line bundle (to be
discussed later). We conclude that
h0 (K) = g, h1 (K) = h0 (O) = 1.
It implies that the space of holomorphic one-forms on a genus g Riemann surface
is g-dimensional.
2.12. The Riemann-Hurwitz formula. Recall that a branched cover π : X → Y
between two (compact, connected) Riemann surfaces is a (surjective) holomorphic
map. For a general point q ∈ Y , π −1 (q) consists of d distinct points. Call d the
degree of π. Locally around p 7→ q, if the map is given by
x 7→ y = xm ,
where x, y are local coordinates of p, q, respectively, call m the vanishing order of
π at p and denote it by
ordp (π) = m.
If ordp (π) > 1, we say that p is a ramification point. If π −1 (q) contains a ramifica-
tion point, then q is called a branch point. Define the pullback
X
π ∗ (q) = ordp (π) · p ∈ Div(X).
p∈π −1 (q)

Note that π ∗ (q) is a degree d effective divisor on X.

Theorem 2.65 (Riemann-Hurwitz Formula). Let π : X → Y be a branched cover
of Riemann surfaces. Then we have
X
KX ∼ π ∗ KY + (ordp (π) − 1) · p,
p∈X

where KX and KY are canonical divisor classes of X and Y , respectively.

Proof. Take a one-form ω on Y locally expressed as f (w)dw around a point q =
π(p). Suppose the covering at p is given by
z 7→ w = z m ,
then we have
π ∗ (f (w)dw) = mf (z m )z m−1 dz.
Namely, the associated divisors satisfy the relation
(π ∗ ω)|U = (π ∗ (ω))|U + (ordp (π) − 1) · p
in a local neighborhood U of p. So globally it implies that
X
(π ∗ ω) = π ∗ (ω) + (ordp (π) − 1) · p.
p∈X
∗ ∗
Since π ω is a one-form on X, (π ω) is a canonical divisor of X and the claimed
formula follows. 
54

We can interpret the (numerical) Riemann-Hurwitz formula from a topological

viewpoint. Let χ(X) denote the topological Euler characteristic of X. If X is a
Riemann surface of genus g, take a triangulation of X and suppose the number of
k-dimensional edges is ck for k = 0, 1, 2. Then we have
χ(X) = c0 − c1 + c2 = 2 − 2g.
Proposition 2.66. Let π : X → Y be a degree d branched cover of two Riemann
surfaces. Then we have
X
χ(X) = d · χ(Y ) − (ordp (π) − 1).
p∈X

Proof. Take a triangulation of Y such that every branch point is a vertex. Pull it
back as a triangulation of X. Note that it pulls back a face to d faces, an edge to
d edges and a vertex v to |π −1 (v)| vertices. Note that if
k
X
−1
π (v) = mi pi
i=1

for distinct points pi , then |π −1 (v)| = k. In other words, we have

X
|π −1 (v)| = d − (ordp (π) − 1).
p∈π −1 (v)

Then the claimed formula follows right away. 

Corollary 2.67 (Numerical Riemann-Hurwitz). Let π : X → Y be a degree d

branched cover of two compact Riemann surfaces of genus g and h, respectively.
Then we have X
2g − 2 = d(2h − 2) + (ordp (π) − 1).
p∈X

In particular, if g < h, such branched covers do not exist.

Corollary 2.68. The canonical line bundle of a genus g Riemann surface X has
degree equal to 2g − 2.
Proof. Every Riemann surface X possesses a nontrivial meromorphic function, say
by the Riemann-Roch formula. It induces a branched cover π : X → P1 of some
degree d. By the Riemann-Hurwitz Formula we know
X
deg(KX ) = d(−2) + (ordp (π) − 1),
p∈X

since we have seen that deg(KP1 ) = −2. By the Numerical Riemann-Hurwitz we

have X
2 − 2g = 2d − (ordp (π) − 1).
p∈X

Then the claim follows immediately. 

Exercise 2.69. Let X be a compact Riemann surface of genus g. If X admits a

branched cover of degree 2 to P1 , we say that X is hyperelliptic. Prove that every
Riemann surface of g ≤ 2 is hyperelliptic.
 55

2.13. Genus formula of plane curves. In this section we consider a Riemann

surface as a complex (one-dimensional) curve. Suppose F (Z0 , Z1 , Z2 ) is a general
degree d homogeneous polynomial whose vanishing locus is a complex curve C ⊂ P2
in the complex projective plane. Since F is general, C is a (smooth) Riemann
surface. In other words, the singularities of C locate at the common zeros of F = 0
and ∂F/∂Zi = 0 for all i, which are empty for a general F . On the other hand, if
F is special, its vanishing locus C may be singular. For example, let F = Z02 − Z12 .
Then C is a union of two lines, hence has a (nodal) singularity at the intersection
of the two lines. If F = Z0 Z12 − Z23 . Then C has a (cuspidal) singularity at [1, 0, 0].
Theorem 2.70. In the above setting, the genus g of C is given by
(d − 1)(d − 2)
g= .
2
Proof. We give two proofs. The first one is more algebraic. Suppose C1 and C2 are
two plane curves of degree d, defined by F1 and F2 . Then F1 /F2 is a meromorphic
function on P2 , hence C1 and C2 are linearly equivalent. It follows that all degree d
plane curves are linearly equivalent. Hence it makes sense to use OP2 (d) to denote
the line bundle associated to a plane curve of degree d. In particular, O(1) is the
line bundle associated to a line L in P2 . The ideal sheaf of L has sections given by
holomorphic functions vanishing along L, hence it can be identified with O(−1),
the dual of O(1). Then we have the short exact sequence
0 → OP2 (−1) → OP2 → OL → 0.
Tensor it with OP2 (1 − m). We obtain that
0 → OP2 (−m) → OP2 (−(m − 1)) → OP2 (1 − m)|L → 0.
Since OP2 (1−m)|L is the line bundle associated to a degree 1−m divisor on L ∼
= P1 ,
we conclude that
χ(OP2 (−(m − 1))) − χ(OP2 (−m)) = χ(OP1 (1 − m)) = 2 − m,
where we apply the Riemann-Roch formula to P1 in the last equality. Then we
obtain that
d
X d(d − 3)
χ(OP2 ) − χ(OP2 (−d)) = (2 − m) = − .
m=1
2
Now by the exact sequence
0 → OP2 (−d) → OP2 → OC → 0,
we have
d(d − 3)
1 − g = χ(OC ) = − ,
2
hence the genus formula follows. Here we implicitly assumed that hi (OC ) = 0
for i ≥ 2. In general, for any line bundle L on a Riemann surface X, we have
hi (X, L) = 0 for i ≥ 2.
The other proof is an application of the Riemann-Hurwitz formula. Without loss
of generality, suppose o = [0, 0, 1] 6∈ C. Let L be the line Z2 = 0 and project C to
L from o, i.e.,
[Z0 , Z1 , Z2 ] 7→ [Z0 , Z1 ].
56

In affine coordinates x = Z1 /Z0 and y = Z2 /Z0 , this map is given by vertical

projection
(x, y) 7→ x,
i.e., we project C vertically to the x-axis. This yields a degree d branched cover
π:C→L∼ = P1 .
A point p is a ramification point of π if and only if there exists a vertical line tangent
to C at p, i.e., p is a common zero of F and ∂F/∂Z2 . Since F and ∂F/∂Z2 have
degree d and d − 1, respectively, they intersect at d(d − 1) points. By Riemann-
Hurwitz, we have
2g − 2 = d(−2) + d(d − 1),
hence the genus formula follows. In order to make sure that all the ramification
points are simple, we can choose a general projection direction such that it is
different from those given by the (finitely many) lines with higher tangency order
to C. 
Remark 2.71. In the first proof, indeed we did not use the smoothness of C. So the
(arithmetic) genus formula holds for an arbitrary plane curve, even if it is singular.
Similarly in the second proof, even if the projection has higher ramification points,
a detailed local study plus Riemann-Hurwitz can provide the same formula.
2.14. Base point free and very ample line bundles. Let L be a line bundle
on a compact complex manifold X. We say that L has a base point at p ∈ X if p
belongs to the vanishing locus of every holomorphic section of L. If the base locus
of L is empty, then L is called base point free.
For a base point free line bundle L, let σ0 , . . . , σn be a basis of the space H 0 (X, L)
of holomorphic sections. Locally around a point p ∈ X, consider σi as a holomorphic
function and associate to p the point
[σ0 (p), . . . , σn (p)] ∈ Pn .
This is well-defined, since if we take a different chart, then we get the same point
[gαβ σ0 (p), . . . , gαβ σn (p)] ∈ Pn
where {gαβ } are transition functions of L. Therefore, we obtain a holomorphic map
φ L : X → Pn .
Remark 2.72. We can give a more conceptual and coordinate free description of
φL . Since L is base point free, the space of holomorphic sections σ vanishing at p
forms a hyperplane Hp ⊂ H 0 (X, L) ∼ = Cn+1 . Then one can define φL (p) = [Hp ] ∈
n ∗
(P ) in the dual projective space parameterizing hyperplanes.
Proposition 2.73. In the above setting, there is a one-to-one correspondence be-
tween (the pullback of ) hyperplane sections of X and effective divisors in the linear
system |L|.
Proof. This is just a reformulation of the one-to-one correspondence
|L| = PH 0 (X, L),
which we proved before. In other words,
Pn an effective divisor in |L| uniquely deter-
mines a holomorphic section σ = i=0 ai σi up to scaling, which defines a hyper-
plane in PH 0 (X, L). 
 57

Example 2.74. If L = O, then H 0 (O) = C, hence φO maps X to a single point.

Example 2.75. Let X = P1 and L = O(2p) where p = [0, 1]. Then H 0 (P1 , L) is
3-dimensional and we can choose a basis by
Y Y2
1, , .
X X2
Recall that around p the sections of L are given by f · (X/Y )2 . Hence we obtain
that
φL ([X, Y ]) = [X 2 , XY, Y 2 ],
which is a smooth conic in P2 . The genus formula for a plane curve of degree two
also implies that the image has g = 0.
Exercise 2.76. A submanifold X ⊂ Pn is called non-degenerate if it is not con-
tained in any hyperplane. If X is isomorphic to P1 , we call it a smooth rational
curve. For any complex curve X in Pn , the intersection number of X with a general
hyperplane is called the degree of X.
(1) Show that any non-degenerate smooth rational curve in Pn has degree ≥ n.
(2) For d ≥ n ≥ 3, show that there exist non-degenerate smooth degree d rational
curves in Pn .
Example 2.77. Let E be a complex curve of genus one and L = O(2p) for a point
p ∈ E. By Riemann-Roch, h0 (E, L) = 2. Moreover, L is base point free. Otherwise
if q is a base point, then q has to be p and there exists another effective divisor
p + r ∈ |2p| such that p + r ∼ 2p. But this implies that r − p is principal and
E∼ = P1 , leading to a contradiction. Now φL : E → P1 is a branched cover of degree
two. Two points s and t lie in the same fiber of φL if and only if s + t ∼ 2p.
The above example indicates that φL is not always an embedding. We say that
L is very ample if φL is an embedding and that L is ample if L⊗m is very ample
for some m > 0.
Example 2.78. The line bundle O(d) is very ample on P1 if and only if d > 0.
The induced map φ embeds P1 into Pd as a degree d smooth rational curve, which
is called a rational normal curve.
Let us give a criterion for base point free and very ample line bundles.
Proposition 2.79. Let L be a line bundle on a Riemann surface X.
(1) L is base point free if and only if
h0 (X, L ⊗ O(−p)) = h0 (X, L) − 1
for any p ∈ X.
(2) L is very ample if and only L is base point free and for any p, q ∈ X (not
necessarily distinct)
h0 (X, L ⊗ O(−p − q)) = h0 (X, L ⊗ O(−p)) − 1 = h0 (X, L ⊗ O(−q)) − 1.
Proof. Treat L as a locally free sheaf of rank one. By the short exact sequence
0 → L ⊗ O(−p) → L → Cp → 0,
we have
h0 (X, L) − 1 ≤ h0 (X, L ⊗ O(−p)) ≤ h0 (X, L).
58

Then L has a base point at p if and only if all holomorphic sections of L vanish at
p, i.e., H 0 (X, L ⊗ O(−p)) = H 0 (X, L). This proves (1).
For (2), a very ample line bundle is necessarily base point free by definition.
If p 6= q ∈ X have the same image under φL , it is equivalent to saying that the
subspace of sections vanishing at p is the same as the subspace of sections vanishing
at q, which is further equivalent to
h0 (X, L ⊗ O(−p)) = h0 (X, L ⊗ O(−p − q)) = h0 (X, L ⊗ O(−q)).
Moreover, φL induces an injection restricted to the tangent space Tp (X) if and
only if there exists a hyperplane such that it cuts out X locally a simple point at
p, namely, if and only if there is a section vanishing at p with multiplicity one, i.e.,
h0 (X, L ⊗ O(−2p)) < h0 (X, L ⊗ O(−p)).
But we have
h0 (X, L ⊗ O(−2p)) ≥ h0 (X, L ⊗ O(−p)) − 1.
Hence (2) follows from combining the two cases. 

Remark 2.80. In (2), for p 6= q the condition geometrically means that the sections
of L separate any two points. When p = q, it says that the sections of L separate
tangent vectors at p.
Example 2.81. Let E be an elliptic curve, i.e., a Riemann surface of genus one.
Fix a point p ∈ E. Consider the morphism
τ : E → Pic0 (E)
by τ (q) = [q − p], where Pic0 (E) is the Picard group of isomorphism classes of line
bundles of degree zero. For q1 6= q2 , we have q1 − p 6∼ q2 − p, hence τ is injective.
For any line bundle L of degree zero, by Riemann-Roch h0 (L ⊗ O(p)) ≥ 1, hence
L ⊗ O(p) has a section vanishing at a single point q. It implies that L = [q − p],
hence τ is surjective. Therefore, we thus conclude that τ is an isomorphism. This
defines a group law on E with respect to p, i.e., q + r = s, where s ∈ E is the unique
point satisfying that
(q − p) + (r − p) ∼ s − p.
Now consider the linear system |3p| on E. Since
h0 (E, O(3p)) = 3, h0 (E, O(2p)) = 2, h0 (E, O(p)) = 1,
O(3p) is very ample. It induces an embedding of E into P2 as a plane cubic curve.
A line cuts out a divisor of degree three in E, say, q + r + s (not necessarily distinct)
if and only if
q + r + s ∼ 3p.
Note that the tangent line L of E at p is a flex line, i.e., the tangency multiplicity
(L · E)p = 3. Such p is called a flex point.
Exercise 2.82. Show that there are in total nine flex points in a smooth plane
cubic curve.
Let V ⊂ |L| be a linear subspace. We say that V is a linear series of L. The
linear system |L| is also called a complete linear series. The above definitions and
properties go through similarly for the induced map φV .
 59

Exercise 2.83. Write down a linear series of |O(3)| on P1 such that it maps P1
into P2 as a singular plane cubic curve. How many different types of such singular
plane cubics can you describe?
2.15. Canonical maps. Let K be the canonical line bundle on a Riemann surface
X. If X is P1 , deg(K) = −2 and K is not effective. If X is an elliptic curve,
then K ∼ = O and the induced map φK is onto a point. From now on, assume
that the genus of X satisfies g ≥ 2. We say that X is hyperelliptic if it admits a
degree 2 branched cover of P1 . Two points p, q ∈ X are called conjugate if they
have the same image in P1 . A ramification point of the double cover is called a
Weierstrass point of X, i.e., it is self conjugate. By Riemann-Hurwitz, a genus
g ≥ 2 hyperelliptic curve possesses 2g + 2 Weierstrass points.
Lemma 2.84. If X is a hyperelliptic curve of genus ≥ 2, then X admits a unique
double cover of P1 .
Proof. Otherwise suppose h0 (X, O(p + q)) = 2 and h0 (X, O(p + r)) = 2 for q 6=
r. Let L = OX (p + q + r). If h0 (X, L) = 3, since h0 (X, L(−x − y)) ≤ 1 and
h0 (X, L(−x)) ≤ 2 by degree reason, φL would map X as a plane cubic of genus
one, leading to a contradiction. Then we conclude that
H 0 (X, O(p + q)) = H 0 (X, O(p + r)) = H 0 (X, O(p + q + r)),
which implies that both q, r are base points of |p + q + r| and h0 (X, O(p)) = 2,
X∼= P1 , leading to a contradiction. 
Proposition 2.85. Let X be a curve of genus g ≥ 2. Then the canonical line
bundle K is base point free. The induced map
φK : X → Pg−1
is an embedding if and only if X is not hyperelliptic. If X is hyperelliptic, φK is a
double cover of a rational normal curve in Pg−1 .
Proof. First, let us show that K is base point free. For any point p ∈ X, by
Riemann-Roch we have
h0 (X, K ⊗ O(−p)) − h0 (X, O(p)) = 1 − g + (2g − 3),
h0 (X, K ⊗ O(−p)) = g − 1 = h0 (X, K) − 1.
Hence K satisfies the criterion of base point freeness.
Next, K fails to separate p, q (not necessarily distinct) if and only if
h0 (X, K ⊗ O(−p − q)) = h0 (X, K ⊗ O(−p)) = g − 1,
which is equivalent to, by Riemann-Roch again, that
h0 (X, O(p + q)) = 2.
In other words, the linear system |p + q| induces a double cover X → P1 .
Finally, if X is hyperelliptic of genus ≥ 2, it admits a unique double cover of P1 .
By the above analysis, two points p, q have the same image under the canonical
map if and only if h0 (X, O(p + q)) = 2, i.e., p, q are conjugate. Then the canonical
map is a double cover of a rational curve of degree deg(K)/2 = g − 1 in Pg−1 , i.e.,
a rational normal curve. A hyperplane section of φK (X) pulls back to X a divisor
g−1
X
(pi + qi ),
i=1
60

where pi , qi are conjugate or pi = qi a Weierstrass point. 

Remark 2.86. For a non-hyperelliptic curve X, φK is called the canonical embed-

ding of X and its image is called a canonical curve.
Example 2.87. Let X be a curve of genus two. Then h0 (X, K) = 2, hence X is
hyperelliptic and the double cover of P1 is induced by the canonical line bundle, as
we have seen.
Example 2.88. A non-hyperelliptic curve of genus three admits a canonical em-
bedding to P2 as a plane quartic. An effective canonical divisor corresponds to a
line section of the quartic. By the genus formula, any smooth plane quartic also has
genus equal to 3. Moreover, a smooth plane quartic X gives rise to a line bundle L
of degree 4 on X by restricting OP2 (1). By Riemann-Roch, h0 (X, K ⊗ L∗ ) ≥ 1, but
deg(K ⊗ L∗ ) = 0, hence L ∼ = K. So any plane quartic is a canonical embedding of
a non-hyperelliptic curve of genus three.
2.16. Dimension of linear systems. Let D = p1 + · · · + pd be an effective divisor
of degree d on a genus g complex curve X. Recall that the linear system |D| can be
identified with PH 0 (X, O(D)) parameterizing effective divisors linearly equivalent
to D. Suppose as a projective space
r = dim |D| = h0 (X, O(D)) − 1.
By Riemann-Roch and Serre Duality, we have
dim |K ⊗ O(−D)| = r + g − d − 1.
Note that |K⊗O(−D)| can be identified with the linear system of effective canonical
divisors that contain D. By the canonical map
φK : X → Pg−1 ,
it says that the space of hyperplanes of Pg−1 that contain the points φK (D) =
{φK (p1 ), . . . , φK (pd )} is (r + g − d − 1)-dimensional. In other words, the linear
subspace in Pg−1 spanned of φK (p1 ), . . . , φK (pd ) has dimension
(g − 2) − (r + g − d − 1) = (d − 1) − r.
Since we expect d points to span a (d−1)-dimensional linear subspace, geometrically
it says that φK (D) fails to impose
r = dim |D|
independent conditions. We summarize the discussion as the following geometric
version of the Riemann-Roch formula.
Theorem 2.89 (Geometric Riemann-Roch). In the above setting, let φK (D) be
the linear subspace in Pg−1 spanned by the image of D under the canonical map.
Then we have
dim |D| = deg(D) − 1 − dim φK (D).
Remark 2.90. Even if D contains points with multiplicity, the above formulation
still holds. Say, if D contains 2p, then 2p spans the tangent line at p. If D contains
3p, then 3p spans an osculating 2-plane at p etc.
 61

Example 2.91. Let us revisit the canonical embedding of a non-hyperelliptic curve

X of genus three in P2 . Consider D = p+q+r. Then h0 (O(D)) ≤ 2, i.e. dim |D| = 1
or 0. Note that dim |D| = 1 if and only if
dim φK (D) = 3 − 1 − 1 = 1
by Geometric Riemann-Roch, i.e., if and only if p, q and r are collinear in P2 .
Let us study in detail the dimension of a linear system.
Lemma 2.92. Let D be a divisor on a complex curve X. Then dim |D| ≥ k if
and only if for every k points p1 , . . . , pk ∈ X there exists an effective divisor in |D|
containing all of them.
Proof. First, suppose for every k points p1 , . . . , pk ∈ X there exists an effective di-
Pk
visor in |D| containing all of them. Since i=1 pi varies in a complex k-dimensional
family, then dim |D| ≥ k is obvious. Alternatively, we can prove it by induction.
Suppose it holds for ≤ k. Assume for every p1 , . . . , pk+1 , there exists D0 ∈ |D|
containing all of them. Then we conclude that dim |D − p| ≥ k for any p ∈ X.
Choose a point p not in the base locus of |D|. Consequently we have
dim |D| = dim |D − p| + 1 ≥ k + 1.
Conversely, suppose dim |D| ≥ k. Then we have
k
!!
X
h X, O D −
0
pi ≥ h0 (X, O(D)) − k ≥ 1.
i=1
It implies that there exists a meromorphic function f such that
k
X
(f ) + D − pi ≥ 0,
i=1

hence (f ) + D = D0 is an effective divisor in |D| containing p1 , . . . , pk . 

Corollary 2.93. For any two effective divisors D1 and D2 on X, we have
dim |D1 | + dim |D2 | ≤ dim |D1 + D2 |.
Proof. Suppose dim |Di | = ki for i = 1, 2. Take any k1 + k2 points
p1 , . . . , pk1 , q1 , . . . , qk2
in X. By the above lemma, there exist D10 ∈ |D1 | and D20 ∈ |D2 | such that D10
contains all the pi and D20 contains all the qj . Then D10 + D20 ∈ |D1 + D2 | contains
all the pi , qj , hence we obtain that
dim |D1 + D2 | ≥ k1 + k2
by using the lemma again. 
Note that if h0 (X, K ⊗ (−D)) = 0, then Riemann-Roch implies that that
h0 (X, O(D)) = 1 − g + deg(D).
Some subtlety may occur if
h0 (X, K ⊗ (−D)) > 0
and we call such a divisor D a special divisor and the associated linear system |D|
a special linear system. By Riemann-Roch, any divisor D with deg(D) > 2g − 2 is
62

non-special. By Geometric Riemann-Roch, D is non-special if and only if the linear

span of φK (D) is the entire space Pg−1 .
Theorem 2.94 (Clifford’s Theorem). Let D be an effective divisor such that deg(D) ≤
2g − 2 on a genus g complex curve X. Then we have
1
dim |D| ≤ · deg(D).
2
Proof. If D is non-special, we have
1
dim |D| = deg(D) − g < deg(D).
2
If D is special, then there exists an effective divisor D0 such that D + D0 ∼ K. By
the above lemma we have
dim |D| + dim |D0 | ≤ dim |K| = g − 1.
By Riemann-Roch and Serre Duality, we have
dim |D| − dim |D0 | = 1 − g + deg(D).
The desired inequality follows by combining the two relations. 
Remark 2.95. Indeed, the above equality holds only if D = 0, D = K or X is
hyperelliptic. If D = 0 or D = K, one easily checks that the equality holds. If X is
hyperelliptic, we can take D = p + q, where p, q are conjugate and dim |p + q| = 1.
To prove that these are the only possibilities, we need the uniform position theorem
regarding a general hyperplane section of a non-degenerate space curve, see [GH,
p. 249].
Exercise 2.96. Let X be a hyperelliptic curve of genus ≥ 2. For 0 < 2k ≤ g, find
an effective divisor D of degree 2k on X such that dim |D| = k. Classify all such
divisors up to linear equivalence.

References
[GH] Phillip Griffiths and Joe Harris, Principles of Algebraic Geometry.
[S] Wilhelm Schlag, A concise course in complex analysis and Riemann surfaces.
[T] Joseph Taylor, Complex Variables.