LECTURE-3

VED V. DATAR∗

1. Power series
A power series centered at z0 ∈ C is an expansion of the form
∞
X
an (z − z0 )n ,
n=0

where an , z ∈ C. If an and z are restricted to be real numbers, this is the

usual power series that you are already familiar with. A priori it is only a
formal expression. But for certain values of z, lying in the so called disc of
convergence, this series actually converges, and the power series represents
a function of z. Before we discuss this fundamental theorem of power series,
let us review some basic facts about complex series, and series of complex
valued functions.

1.1. Infinite series of complex numbers: A recap. A series is an infi-

nite sum of the form
∞
X
an ,
n=0
where an ∈ C for all n. We say that the series converges to S, and write
∞
X
an = S,
n=0

if the sequence of partial sums

N
X
SN = an
n=0

converges to S. We need the following basic fact.

Proposition 1.1. If ∞
P P∞
n=0 |an | converges then n=0 an converges.

Proof. We will show that SN forms a Cauchy sequence if ∞

P
n=0 |an | con-
verges. Then the theorem will follow from the completeness of C. If we let

Date: 24 August 2016.

1
TN = N
P
n=0 |an |, then {TN } is a Cauchy sequence by the hypothesis. Then
by triangle inequality
M
X M
X
|SN − SM | = an ≤ |an | = |TM − TN |.
n=N +1 n=N +1
Now, given ε > 0, there exists a K > 0 such that for all N, M > K,
|TM − TN | < ε. But then |SM − SN | ≤ ε, and so {SN } is also Cauchy. 
P P
We say that an converges absolutely if |an | converges. Next sup-
pose fn : Ω → C are complex functions, we say that ∞
P
n=0 fn (z) converges
uniformly if the corresponding sequence of partial sums
N
X
SN (z) = fn (z)
n=0
converges uniformly.
Proposition 1.2 (Weierstrass’ M-Test). Suppose fn : Ω → C is a sequence
of complex functions, and {Mn } is a sequence of positive real numbers such
that
• P
|fn (z)| ≤ Mn for all n and all z ∈ Ω.
∞
• Mn converges.
P∞n=0
Then n=0 fn (z) converges uniformly.
Proof. Like before, this time we show that the sequence of partial sums
{SN (z)} is uniformly Cauchy. But again by triangle inequality if TN denotes
the N th partial sum of
P
Mn , then
M
X M
X
|SN (z) − SM (z)| ≤ |fn (z)| ≤ Mn = |TM − TN |,
n=N +1 n=N +1
for all z ∈ Ω. Since the right side does no depend on z, given ε > 0, one can
make |SN (z) − SM (z)| < ε by choosing N, M > K where K can be chosen
independent of z. 
1.2. Convergence of power series. By convergence of the power series,
we mean the following. Consider the truncations of the power series at the
N th term, also called the N th partial sum -
N
X
sN (z) = an (z − z0 )n .
n=0
We say that the power series converges (unifomrly) if the sequence of func-
tions {sN (z)} converges (uniformly). We say that the series converges ab-
solutely if the sequence of functions
N
X
|an ||z|n
n=0
2
converges. It is well known, and not difficult to see, that absolute conver-
gence implies convergence. The fundamental fact is the following.
Theorem 1.1 (Fundamental theorem of power series). There exists a 0 ≤
R ≤ ∞ such that
• If |z − z0 | < R, the series
∞
X
an (z − z0 )n
n=0

converges absolutely.

• For any compact set K ⊂ DR (z0 ), the absolute convergence is actu-

ally uniform.
• If |z − z0 | > R, then the series diverges.
Moreover, R can be computed using the Cauchy-Hadamard formula:
1
R= .
lim sup |an |1/n
The number R is called the radius of convergence, and the domain DR =
{z | |z −z0 | < R} is called the disc of convergence. Recall that for a sequence
{bn } of real numbers,
L = lim sup bn
if the following two conditions hold
• For all ε > 0, and all N > 0, there exists n > N such that
bn > L − ε.
• For all ε > 0, there exists an N such that for all n > N ,
bn < L + ε.
Remark 1.1. It is not difficult to show that if
an+1
lim
n→∞ an
exists, then it is equal to 1/R. On many occasions the limiting ratio is easier
to calculate.
Remark 1.2. Suppose 0 < R < ∞ is the radius of convergence of the above
power series. The by the theorem, the series converges on the open disc
|z − z0 | < R. The behavior of the series at points on boundary however is
subtle as the examples below indicate.
Example 1.1. The power series
∞
X
zn
n=0
3
has radius of convergence 1. In fact it easy to see that on |z| < 1,
∞
1 X
= zn.
1−z
n=0

Observe that on the unit disc |z n | = |z|n = 1, and so by the divergence test,
the series cannot converge at any boundary point. On the other hand, the
left hand side is defined and holomorphic at all points z 6= 1 even though
the power series is only defined inside the unit disc. We then say that the
holomorphic function 1/1 − z is an analytic continuation of the power series
to the domain C \ {z = 1}. We will say more about analytic continuation
towards the end of the course. We remark that the following misleading
formula often appears in popular culture (most notably in a video on the
youtube channel - Numberphile, an otherwise decent math channel), many
times accompanied with a quote with the effect that “Oh look - math is mag-
ical!”:
1
1 − 1 + 1 − 1..... = .
2
The formula as stated is of course junk since the left hand side is clearly
a divergent series. But there are ways of interpreting the left hand side.
For instance, the left hand side is in fact Cesaro summable, which is a
generalization of usual infinite summation in that a convergent series is also
Cesaro summable and the Cesaro sum equals the sum of the series. In this
case, the Cesaro sum does turn out to be 1/2. A more fundamental way
(at least in my opinion) ofP∞ interpreting the left hand side, as precisely the
n
analytic continuation of n=0 z to z = −1. Then as remarked above,
this analytic continuation is given by 1/(1 − z) which of course equals 1/2 at
z = −1. Another example of such misleading propogation of math, especially
in India, is that Ramanujan proved the “miraculous” identity that
−1
1 + 2 + 3 + ··· = .
12
We’ll see later in the course that the left hand
P side−sshould in fact be replaced
by the analytic continuation of the series ∞ n=1 n to s = −1. The infinite
seres is a priori only defined on the region Re(s) > 1, but can be analytically
continued to C \ {1}, and this if of course the famous ζ(s) of Riemann..
We’ll then compute that ζ(−1) = −1/12!
Example 1.2. Consider the power series
∞
X zn
.
n
n=0

Again, it is easy to see that the radius of convergence is 1. At z = 1 this is

the usual harmonic series, and is divergent. It turns out in fact, that this
series converges for all other points on |z| = 1, z 6= 1. This follows from
the following test due to Abel, which we state without proof.
4
Lemma 1.1 (Abel’s test). Consider the power series
∞
X
an z n .
n=0
Suppose
• an ∈ R, an ≥ 0.
• {an } is a decreasing sequence such that limn→0 an = 0.
Then the power series converges on |z| = 1 except possibly at z = 1.
Clearly the series in the example above satisfies all the hypothesis, and
hence is convergent at all points on |z| = 1 except at z = 1.
Example 1.3. Next, consider the power series
∞
X zn
.
n2
n=0
Again, the radius of convergence is 1, and again by Abel’s test the power
series is convergent on |z| = 1 except possibly at z = 1. But at z = 1,
the series is clearly convergent, for instance by the integral test. So in this
example the power series is convergent on the entire boundary.
Example 1.4. Finally consider the power series
∞
X zn
.
n!
n=0
To find the radius of convergence, we use the ratio test. Denoting an = 1/n!,
we see that
an+1 n! 1
lim = lim = lim = 0.
n→∞ an n→∞ (n + 1)! n→∞ n + 1

So R = ∞. Notice that if z is a real number then this is the usual Taylor

expansion of the exponential function. Inspired by this, we define the complex
exponential function, exp(z) : C → C by
∞
X zn
exp(z) = ez = .
n!
n=0
We will study this function in more detail in the next lecture.
1.3. Holomorphicity of power series. From the previous theorem, since
the convergence is uniform on comapct subsets of the disc of convergence, it
is clear that a power series represents a continuous function. In fact, much
more is true.
Theorem 1.2. Consider the function defined by
∞
X
f (z) = an (z − z0 )n ,
n=0
5
on the disc of convergence DR = {z | |z − z0 | < R} where 0 < R ≤ ∞. Then
f (z) is holomorphic on DR with
∞
X
0
f (z) = nan (z − z0 )n .
n=1

Proof. Without loss of generality we can assume

P z0 = 0. Firstly, observe
that since limn→∞ n1/n = 1, the power series ∞ n=1 na n (z − z 0 ) n also has

radius of convergence R. To prove the theorem, we need to show that for

any p ∈ DR (0),
∞
f (p + h) − f (p) X
lim = nan pn .
h→0 h
n=1
Or equivalently, given any ε > 0, we need to find a δ > 0 such that
∞
f (p + h) − f (p) X
(1.1) |h| < δ =⇒ − nan pn < ε
h
n=1

Let us denote by
N
X ∞
X
SN (z) = an z n , EN (z) = an z n ,
n=0 n=N +1

the N th partial sum, and the N th error term respectively, so that

f (z) = SN (z) + EN (z).
Then since the partial sums are polynomials, they are holomorphic, and in
fact
∞
X
0
(1.2) lim SN (z) = nan z n ,
N →∞
n=1

where the convergence is uniform on compact subsets of DR (0) Now suppose

|p| < r < R, then for any N , we can break the difference that we need to
estimate into three parts -
∞
f (p + h) − f (p) X SN (p + h) − SN (p) 0
− nan pn = − SN (p)
h h
n=1
∞
0
X EN (p + h) − EN (p)
+ |SN (p) − nan pn | +
h
n=1

Since SN , being a polynomial, is holomorphic, there exists a δ > 0 so that

for |h| < δ, the first term is smaller than ε/3. Similarly, by equation (1.2),
the second term can be made smaller than ε/3 by choosing N big enough.
So all that remains is to control the error term. Using the factorization
6
an − bn = (a − b)(an−1 + an−2 b + · · · + bn−1 ), we have
∞
EN (p + h) − EN (p) X (p + h)n − pn
≤ |an |
h h
n=N +1
X∞
≤ |an ||((p + h)n−1 + · · · + pn−1 )|
n=N +1

But if |h| < δ for sufficiently small δ (in particular if δ < r − |p|), then
|p + h| ≤ |p| + |h| < r, and so
∞
EN (p + h) − EN (p) X
≤ |an |nrn−1 .
h
n=N +1

n|an |rn−1 which converges for r < R,

P
But this is the tail of the series
so we can also make this term smaller than ε/3 by choosing N big enough.
This shows that we can find δ small enough so that (1.1) is satisfied.


Notice that the derivative is again a power series with the same radius of
convergence. So applying the above theorem inductively we obtain -
Corollary 1.1. A power series f (z) = ∞ n
P
n=0 an (z−z0 ) is infinitely complex
differentiable in it’s disc of convergence. Moreover, the derivatives can be
computed by successive term-wise differentiation:
∞
X
f (k) (z) = n(n − 1) · · · (n − k + 1)an (z − z0 )n−k .
n=k

In particular, the coefficients of the power series are given by

f (n) (z0 )
an = .
n!
We say that a function f : Ω → C is analytic if for every p ∈ Ω, there
exists an r = r(p) > 0 and a sequence of numbers {an = an (p)} such that
∞
X
f (z) = an (z − p)n
n=0

for every z ∈ Dr (p). A priori, it is not quite clear that if a function is

represented by a power series expansion on a disc of convergence DR (p)
then it is automatically analytic. For instance, it is not clear if there should
be a power series expansion around any other point q ∈ DR (p). The next
proposition answers this question in the afirmative.
P∞ n
Proposition 1.3. Let f (z) = n=0 an (z − z0 ) be a power series with
radius of convergence R. Then f is analytic on DR (z0 ). In fact, for every
7
p ∈ DR (z0 ), and every z ∈ Dr (p) where r := R − |p − z0 |, we have
∞
X f (n) (p)
f (z) = (z − p)n .
n!
n=0

Proof. We use the binomial expansion. As before, without loss of generality,

we assume that z0 = 0. Writing z = z − p + p, and applying the binomial
theorem, we see that on |z − p| < R − |p| (since |z| < R), we have
∞
X
f (z) = an (z − p + p)n
n=0
∞ Xn  
X n n−k
= an p (z − p)k
k
n=0 k=0
∞
X
= bk (z − p)k ,
k=0
where
∞  
X n f (k) (p)
bk = an pn−k =
k k!
n=k
by Corollary 1.1. 
Remark 1.3. In some cases the power series on the right in the conclusion
of Proposition 1.3 might have a larger radius of convergence than R − |q − p|.
In such cases, the new power will define an analytic continuation of f . For
instance, let us consider the power series
∞
1 X
= zn
1−z
n=0
on |z| < 1 with p = 0. Let q = −1/2. Then if |z + 1/2| < 1 − 1/2 = 1/2, we
have
∞
1 1 2 1 2 X 2n  1 n
= = · = z + .
1−z 3/2 − (z + 1/2) 3 1 − 2(z + 1/2)/3 3 3n 2
n=0
On the other hand, it can be easily seen that the power series on the right has
a radius of convergence R = 3/2, and hence defines an extension of the orig-
inal power series in the new region |z + 1/2| < 3/2. This was Weierstrass’
method of analytically continuing holomorphic functions.
Corollary 1.2. [Principle of analytic continuation for power series] Let f
be an analytic function on a connected open set Ω. If there exists a point
p ∈ Ω such that f (n) (p) = 0 for all n ∈ N, f ≡ 0 on all of Ω. In particular,
the conclusion is true if there is an open set U ⊂ Ω such that f ≡ 0.
U

Proof. Let S = {z ∈ Ω | f (n) (z) = 0 for all n = 0, 1, 2, · · · }. Then by the

continuity of f , S is closed in Ω (ie. all limit points of S in Ω, are also
8
contained in S). Also, S is non-empty by the hypothesis. Now suppose
q ∈ S. Since the function is analytic, there is an open disc Dr (q) on which
∞
X f (n) (q)
f (z) = (z − q)n = 0,
n!
n=0
since q ∈ S. But then Dr (q) ⊂ S, and so S is open in Ω. But then since Ω
is connected, this forces S = Ω, and in particular, f ≡ 0 on Ω. 

Appendix: Multiplication and composition of power series

Given two power series
∞
X X
an z n , bn z n ,
n=0 n=0
their product can be defined, at least formally, in the following way-
X∞ X
an z n · bn z n = (a0 + a1 z + a2 z 2 + · · · )(b0 + b1 z + b2 z 2 + · · · )
n=0 n=0
= a0 b0 + (a0 b1 + a1 b0 )z + (a0 b2 + a1 b1 + a2 b0 )z 2 + · · ·
∞
X
= cn z n ,
n=0
where cn is given by
n
X
cn = ak bn−k .
k=0
This product goes by the name of Cauchy product. The main theorem, which
we state without proof, is the following.
Theorem 1.3. If the radius of convergence of two series centered at z0
is R1 and R2 respectively, then their product power series has a radius of
convergence that is at least min (R1 , R2 ).
Proof of a slightly general version, applicable for any infinite series, can
be found on the wiki article https://en.wikipedia.org/wiki/Cauchy_
product#Convergence_and_Mertens.27_theorem.