LECTURE-16

VED V. DATAR∗

Meromorphic functions
A function on a domain Ω is called meromorphic, if there exists a sequence
of points p1 , p2 , · · · with no limit point in Ω such that if we denote Ω∗ =
Ω \ {p1 , · · · }
• f : Ω∗ → C is holomorphic.
• f has poles at p1 , p2 · · · .
We denote the collection of meromorphic functions on Ω by M(Ω). We
have the following observation, whose proof we leave as an exercise.
Proposition 0.1. The class of meromorphic function forms a field over C.
That is, given any meromorphic functions f, g, h ∈ M(Ω), we have that
(1) f ± g ∈ M(Ω),
(2) f g ∈ M(Ω),
(3) f (g + h) = f g + f h.
(4) f ± 0 = f, f · 1 = f ,
(5) 1/f ∈ M.
Recall that if a holomorphic function has finitely many roots, then it can
be “factored” as a product of a polynomial and a no-where vanishing holo-
morphic function. Something similar holds true for meromorphic functions.
Proposition 0.2. Let f ∈ M(Ω) such that f has only finitely many poles
{p1 , · · · , pn } with orders {m1 , · · · , mn }. Then there exist holomorphic func-
tions g, h ∈ O(Ω) such that for all z ∈ Ω \ {p1 , · · · , pn },
g(z)
f (z) = .
h(z)
Moreover, we can choose g and h such that f (z) and g(z) have the exact same
roots with same multiplicities, while h(z) has zeroes precisely at p1 , · · · , pn
with multiplicities exactly m1 , · · · , mn .
Proof. We define g : Ω \ {p1 , · · · , pm } by
 
g(z) = Πnk=1 (z − pk )mk f (z).
This is clearly a holomorphic function. Moreover, since f (z) has a pole of
order mk at pk , g(z) is bounded in a neighbourhood of pk . Thus, by the
Date: 24 August 2016.
1
theorem on removable singularities, g(z) can be extended as a holomorphic
function on Ω. The theorem is then proved with
h(z) = (z − p1 )m1 · · · (z − pn )mn .

Remark 0.1. The same is true even if the meromorphic function has in-
finite number of poles. This is a consequence of Weierstrass’ factorization
theorem. We will prove this theorem for the special case when Ω = C. For a
general open set the proof requires the use of Runge’s approximation theorem.

Partial fraction decomposition of meromorphic functions on C.

Recall that if f has a pole of order m at p, then the Laurent series ex-
pansion can be written as
 1 
f (z) = Qp f + Hp f (z),
z−p
where Hp f is holomorphic near p, and Qp f (w) is a polynomial
Qp f (w) = a−m wm + · · · a−1 w,
where for each n = 1, 2, · · · , m and each ε << 1, we have
Z
1
a−n = f (z)(z − p)n−1 .
2πi |z−p|=ε
The difference f (z) − Hp f (z) is called the principal part of f (z) at p. We
then have the following fundamental theorem.
Theorem 0.1 (Mittag-Leffler). Let {pk } be a discrete set of points in Ω, and
for each k, let Qk (w) be a polynomial without a constant term. There there
exists a f ∈ M(Ω) with poles at pn and holomorphic everywhere else, with
principle part at pk given by Qk (1/(z−pk )). Moreover, all such meromorphic
functions are of the form
X  1  
f (z) = Qk − qk (z) + H(z),
z − pk
k
where each qk (z) and H(z) are holomorphic functions on Ω. Furthermore:
(1) If {pk } is a finite sequence, then one could take qk ≡ 0.
(2) If Ω = C, and |pk | → ∞, then one could take each qk to be a poly-
nomial.
We will prove parts (1) and (2) in the next lecture. For a general Ω and
infinitely many poles, the proof require’s Runge’s theorem, and an outline
will be provided in the appendix to the next lecture.
P
Remark 0.2. Note that if k gk is a compactly convergent series on Ω (for
P
instance, take a power series), then q̃k = qk + gk , and H̃ = H + k gk will
give another representation for the function f (z), and hence qk and H are
by no means unique.
2
Remark 0.3. Note that given a meromorphic function, the theorem does
not say whether that particular function has a partial fraction decomposition
(unlike say for rational functions, as we will see in the next section, or
more generally meromorphic functions with only a finite number of poles).
In other words, we are not claiming that the converse holds (even when
Ω = C). It does turn out that the converse holds under some additional
conditions on the distribution of poles. But in particular examples, one can
get away by a more hands-on approach. We will see a beautiful illustration
of this below.
Example 0.1. Consider the meromorphic function f (z) = π 2 / sin2 πz which
is a meromorphic function with poles at integers. Near zero,
π2 π2 1 1 z2 2
= = = 1 + + · · · ,
sin2 πz 2 2 z2 6
   
πz + O(z 3 ) z 2 1 − z 2 /6 + · · ·
and so the principal part of f (z) is given by 1/z 2 . Using the identity
sin2 (π(z − n)) = sin2 πz, it is easy to see that the principal parts at z = n
are given by (z − n)−2 . Now consider the series
∞
X 1
.
n=−∞
(z − n)2
This converges uniformly on all compact subsets of C \ Z, and hence repre-
sents a meromorphic function on C with poles of order two at all integers.
Moreover the principle parts at each pole z = n is given by Qn ((z − n)−1 ),
where Qp (w) = w2 . It is then easy to see that the difference
∞
π2 X 1
H(z) := −
sin2 πz n=−∞ (z − n)2
extends to an entire function.

Claim. H ≡ 0.
Proof. Note that the series and the function on the left are both periodic
with period 1, and hence so is H(z). That is, H(z + 1) = H(z) for all z ∈ C.
Also by Euler’s identity, if z = x + iy, then
eiπz − e−iπz
sin πz = = sin(πx) cosh(πy) + sinh(πy) cos(πx),
2i
and so
| sin πz|2 = cosh2 (πy) − cos2 (πx) ≥ cosh2 (πy) − 1 → ∞
uniformly as |y| → ∞. As a consequence π 2 / sin2 πz converges uniformly to
zero as |y| → ∞. But the infinite series also shares this property. Indeed,
since the series converges uniformly on |y| ≥ 1, we can take pointwise limit,
and clearly each (z − n)−2 → 0 uniformly as |y| → ∞. The upshot is that
H(z) converges uniformly to zero as |y| → 0. In particular, H(z) is bounded
3
on the strip {z ∈ C | 0 ≤ Re(z) ≤ 1}. But then since H is periodic with
period one, this means that H is a bounded entire function, and hence a
constant by Liouville. But since limy→0 H(iy) = 0, we can conclude that
H ≡ 0. 

Assuming this, we get the identity

∞
π2 X 1
2 =
sin πz n=−∞ (z − n)2

for all z ∈
/ Z. Plugging in z = 1/2, we obtain the identity
∞ ∞
π2 X 1 X 1
= 2
= 2 .
4 n=−∞
(2n − 1) (2n − 1)2
n=1
P∞ −2 .
Now let, S = m=0 m Then we have
∞ ∞
X 1 X 1
S= 2
+
(2m − 1) (2n)2
n=1 n=1
π2S
= . +
8 4
Solving for S, we get the beautiful identity
∞
X 1 π2
= .
m2 6
m=1

Proving this identity was the so-called Basel problem, first “solved” by Euler.
But his “proof” would not pass our modern day standards of rigour. Euler
used a “facotrization” for sine, but a rigorous development of the theory of
infinite product factorizations of entire functions had to wait till Weierstrass
came along many decades later. Nevertheless, Euler’s insights were of course
crucial in all subsequent developments.

Meromorphic functions on the extended complex plane

It is often useful to think of z = ∞ on the same footing as other points
in the complex plane, and to define the extended complex plane
Ĉ = C ∪ ∞.
We can then think of meromorphic functions f : Ω \ {p1 , · · · , pj , · · · } → C,
as functions f : Ω → Ĉ, by defining
f (pj ) = ∞
for all the poles pj s. Similarly, in studying meromorphic functions on C, it
is also useful to consider the extension of the function themselves to Ĉ. We
say that z = ∞ is a pole of order m (resp. removable or essential singularity)
4
if z = 0 is a pole of order m (resp. removable or essential singularity) for
the function
fˆ(z) = f (1/z).
Similarly we can also define a zero of order m at infinity. We then say that a
meromorphic function on C is meromorphic on the extended plane, if it does
not have an essential singularity at z = ∞. It turns out that meromorphic
functions on Ĉ can be classified. Recall that a rational function on C is a
function of the form
P (z)
R(z) =
Q(z)
where both P (z) and Q(z) are polynomials.
Example 0.2. (1) A polynomial P (z) = an z n +· · ·+a0 with an 6= 0 has
a pole of order n at infinity. In fact, conversely, ever entire function
p(z) with a pole of order n at infinity is a polynomial of degree n.
This follows from the Cacuhy estimates.
(2) The function ez has an essential singularity at infinity.
(3) A rational function has a pole or removable singularity at infinity.
In fact a rational function R(z) = P (z)/Q(z) as above has
• a pole of order deg P − deg Q at infinity if deg(P ) > deg(Q),
• a removable singularity at infinity if deg(P ) ≤ deg(Q),
• a zero of order deg(Q) − deg(P ) if deg(P ) < deg(Q).
Theorem 0.2. The only meromorphic functions on Ĉ are rational func-
tions.
Proof. Let F : Ĉ → Ĉ be a meromorphic function.

Claim-1. F has only finitely many poles {p1 , · · · , pn } in the complex plane
C.
To see this, note that F (1/z) has either a pole or zero at z = 0. In either
case there is a small neighborhood |z| < ε which has no other pole. Which
is the same as saying that F has no finite pole in |z| > 1/ε. But |z| ≤ 1/ε
is compact, and since all poles are isolated, this shows that there are only
finitely many poles. Now, corresponding to each of the poles pk ∈ C there
exists a polynomial Qk (see Remark 0.2 in Lecture-20) such that
 1 
F (z) = Qk + Hk (z),
z − pk
where Gk is holomorphic on a whole neighborhood around pk (including at
the point pk ). Similarly if |z| > R, we can write
1
F (z) = Q∞ (z) + H∞ ,
z
where as before, H∞ (z) is holomorphic in a neighborhood of z = 0.

5
Claim-2. The function
n
X  1 
G(z) = F (z) − Q∞ (z) − Qk
z − pk
k=1
is an entire and bounded function.
Assuming the claim, by Liouville’s theorem, G(z) is a constant, and hence
F (z) must be rational, and the theorem is proved. To prove the claim, first
note that clearly, G(z) is holomorphic away from {p1 , · · · , pn }. At some
z = pk , Qj (1/z − pj ) is holomorphic for all j 6= k. On the other hand, near
pk ,
 1 
F (z) − Qk = Hk (z)
z − pk
which is holomorphic. This shows that G(z) is entire. As a consequence, to
show boundedness, we only need to show boundedness on |z| > R for some
large R. To see, first observe that since Qk are polynomials,
 1 
lim Qk = 0.
z→∞ z − pk
Hence it is enough to show that F (z) − Q∞ (z) is bounded near infinity. But
this follows immediately from noting that
1 1
H∞ (z) = F − Q∞
z z
is holomorphic near z = 0 and hence is bounded on |z| < ε for some ε > 0.
In particular F (z) − Q∞ (z) is bounded on |z| > 1/ε. This proves the claim,
and hence completes the proof of the theorem.

Remark 0.4. A meromorphic function f ∈ M(Ĉ) gives rise to a holomor-
phic map F : P! → P1 . Conversely, given any map F : P1 → P1 , one gets
a meromorphic map from Ĉ → C with poles at F −1 ([0, 1]). So the theo-
rem can be reformulated in the following way - all holomorphic maps from
F : P1 → P1 are given by rational functions of two variables, where the
numerator and denominator are homogenous polynomials.
A simple consequence of the proof is the following theorem on partial
fraction decomposition that we take for granted as an important tool in
integration theory, but never see the proof of.
Corollary 0.1. For any rational function R(z) = P (z)/Q(z) has a partial
fraction decomposition of the form
n
X  1 
R(z) = Q∞ (z) + Qk ,
z − pk
k=1
where pk is a root of Q(z) of order mk , Qk is a polynomial of degree mk ,
and deg Q∞ = deg P − deg Q if this number is non-negative. Else we have
that Q∞ ≡ 0.
6
∗ Department of Mathematics, Indian Institute of Science
Email address: vvdatar@iisc.ac.in