POWER SERIES

1 Introduction
A power series is an expression of the form
∞
∑
an (z − c)n
n=0

where a0 , a1 , a2 , . . . ; c are given complex numbers and z is a complex variable. We say this series is centered at c.
We call the complex numbers an the coefficients of the series. The main question relating to power series is for
∑∞
which complex numbers z does the series of complex terms an (z − c)n converge? In other words, if we define
n=0
for n = 0, 1, 2, . . . the polynomials

sn (z) = a0 + a1 (z − c) + · · · + an (z − c)n ,

the question can be formulated as follows: for what complex numbers z does there exist a limit lim sn (z)?
n→∞

Polynomials are a very particular case of power series; a polynomial is a power series, usually shown as centered
at 0, having all but a finite number of coefficients equal to 0. The answer to the question of where does the series
converge is in the case of a polynomial, everywhere. For all z ∈ C.

2 Generalities
Power series are series of functions; their partial sums are also functions, thus results about convergence of sequences
of functions are relevant. I’ll begin recalling some general results about sequences (and series) of functions. I am
assuming the functions are complex valued functions of a complex variable, but the results are valid in a much
more general context, as I point out.

Definition 1 Let S be a subset of C and let fn : S → C for n = 1, 2, 3, . . .. The sequence {fn } converges
uniformly to a function f on S iff for every ϵ > 0 there is N such that

|fn (z) − f (z)| < ϵ

for all z ∈ S whenever n ≥ N .

We say a sequence {fn } of functions defined on a set S with values in C converges uniformly on S iff there
is a function f : S → C such that the sequence converges uniformly to f . In this case lim fn (z) = f (z0 ) for all
n→∞
z ∈ S. The function f , the uniform limit of the sequence {fn }, is unique.

(This definition makes sense for functions defined on an arbitrary set S taking values in a metric space; one simply
has to replace |fn (z) − f (z)| by d(fn (z), f (z)).)

∑
Definition 2 Assume S ⊂ C, an : S → C for n = 0, 1, 2, . . .. We say the series n=0 an converges uniformly
∑
n
on S iff the sequence of partial sums {sn }, where sn (z) = ak (z) for z ∈ S, n = 0, 1, 2, . . . ,, converges uniformly
k=0
on S.

One important result (valid in the general context of metric or even uniform, spaces) is:

Theorem 1 Let S be a subset of C and let fn : S → C be continuous for n = 0, 1, 2, . . .. Assume {fn } converges
uniformly to f on S. Then f is continuous on S.
2 GENERALITIES 2

I’ll refer to any introductory analysis, or metric spaces, textbook for the simple proof.

The following definition makes sense for a sequence of functions defined on an arbitrary set S taking values in
a metric space; mutatis-mutandis. But, I stick to C.
Definition 3 Let S be a subset of C and let fn : S → C for n = 1, 2, 3, . . .. We say the sequence {fn } is uniformly
Cauchy iff for each ϵ > 0 there is N such that |fn (z) − fm (z)| < ϵ for all z ∈ S whenever n, m ≥ N .

Because C is complete, we have the following useful convergence result (valid actually for functions defined on
an arbitrary set S, taking values in a complete metric space). I state it without proof.
Theorem 2 Let S be a subset of C and let fn : S → C for n = 1, 2, 3, . . .. The sequence {fn } converges uniformly
on S if and only if it is uniformly Cauchy.

Another basic result, but very useful in proving uniform convergence, is the so called M -test of Weierstrass (M
for majorant).
Theorem 3 Let S be a subset of ∑ C and let fn : S → C for n = 0, 1,∑2, . . .. Assume also that |fn (z)| ≤ an for
∞ ∞
z ∈ S, n = 0, 1, 2 . . .. If the series n=0 an converges, then the series n=0 fn converges uniformly in S.
Proof. For z ∈ S, n = 0, 1, . . ., let
∑
n
Fn (z) = fk (z).
k=0

∑∞ {fn } converges uniformly on S.

We have to prove that the sequence of functions ∑∞
Let ϵ > 0 be given. Because the series n=0 an converges, there is N ∈ N such that n=N +1 an < ϵ. Assume
now n > m ≥ N . Then, for every z ∈ S,
 n 
 ∑  ∑n
 
|Fn (z) − Fm (z)| =  fk (z) ≤ |fk (z)|
 
k=m+1 k=m+1
∑n ∞
∑
= ≤ fk ≤ ak < ϵ.
k+1 k=N +1

Since the roles of m, n can be interchanged, the same estimate is valid if m > n ≥ N ; it is also valid if m = n
because then it reduces to 0 < ϵ.
We proved: Given ϵ > 0, there exists N such that |Fn (z) − Fm (z)| < ϵ for all z ∈ S whenever m, n ≥ N . This
means that the sequence {Fn } is uniformly Cauchy on S, hence converges uniformly.

Another useful result is the so called root test for convergence of series of non-negative terms:
1/n
Theorem 4 Let an ≥ 0 for n = 0, 1, 2, . . .. Let λ = lim supn→∞ an .
∑∞
1. If λ < 1, then the series n=0 an converges.
∑∞
2. If λ > 1, then the series n=0 an diverges.
Proof. Assume first λ < 1. Let σ be such that λ < σ < 1, for example let σ = (λ + 1)/2. By the definition of
1/n ∑∞
limsup, there is N ∈ N such that a∑ n < σ, hence an < σ n if n ≥ N . Since σ < 1 the geometric series n=0 σ n
∞
converges; by comparison, so does n=0 an .
Assume next that λ > 1. Then there exists a subsequence {ank k }∞
1/n 1/n 1/nk
k=0 of {an } such that ank > 1 for
k = 1,∑
2, . . .. But then ank > 1 for k = 1, 2, . . . and limn→∞ an either does not exist or is not 0. In either case, the
∞
series n=0 an diverges.

1/n
Exercise 1 Show that if lim supn→∞ an > 1, then there is a subsequence {ank } of {an } such that limk→∞ ank =
∞.
3 POWER SERIES 3

3 Power Series
Given a power series
∞
∑
an (z − c)n
n=0

as mentioned above, ∑∞it is of interest is to determine the set of z for which the series converges. But it is obvious
that if the series n=0 an z n centered at 0 converges for all z in a set S, then the series with the same coefficients
a0 , a1 , . . . centered at c converges for z ∈ c + S, where c + S = {c + z : z ∈ S}. Thus it suffices to consider series
centered at 0. Everything we say about a series centered at 0 can be immediately translated to a series centered at
c.
Another obvious observation is that the series
∞
∑
an z n
n=0
∑∞
always converges for z = 0, namely to a0 . (This translates, for a series n=0 an (z − c)n centered at c that it
always converges for z = c, namely to a0 .) If a power series only converges for z equal to the center, it is said to
be divergent. Otherwise it is said to be convergent (somewhere).

The main theorem about general power series is the existence of a (extended) number r ∈ [0, ∞] such that the
series (centered at 0) converges for all z with |z| < r and diverges for all z with |z| > r. There even is a formula
for r. The precise statement is as follows.
Theorem 5 Let
1
r= ,
lim supn→∞ |an |1/n

∑∞ we ninterpret r = 0 if lim supn→∞ |an | = ∞; r = ∞ if lim supn→∞ |an |1/n = 0. The power series
1/n
where
n=0 an z converges absolutely for all z ∈ C with |z| < r and diverges for all z ∈ C with |z| > r. In particular, if
r = 0 it diverges for all z ̸= 0, if r = ∞ it converges for all z ∈ C. Moreover (and this condition is only significant
if 0 < r ≤ ∞) the series converges uniformly on any closed disc D̄ρ (0) = {z ∈ C : |z| ≤ ρ} with 0 < ρ < r.
Note on notation. In the sequel, if a ∈ C, r > 0, we denote by Dr (a) the open disc of center a, radius r;

Dr (a) = {z ∈ C : |z − a| < r}.

We denote by D̄r (a) its closure {z ∈ C : |z − a| ≤ r}.

Proof. Let r be defined as in the statement of the∑theorem and let ρ be a real number satisfying 0 ≤ ρ < r (so
∞
implicitly, r > 0). The series of non-negative terms n=0 |an |ρn converges by the root test. In fact,

1/n 1
lim sup (|an |ρn ) = lim sup |an |1/n ρ = · ρ < 1.
n→∞ n→∞ r
∑∞
This has two immediate consequences. First, if |z| < r, we can take ρ = |z| and see that the series n=0 an z n
converges absolutely. Second, if we consider the disc D̄ρ (0), we have that

|an z n | ≤ |an |ρn

for n = 0, 1, 2, . . . and all z ∈ D̄ρ (0). By Theorem 3, the series converges uniformly in D̄ρ (0). On the other hand, if
|z| > r we proceed as in the proof of the root test (Theorem 4). We have then |z| lim supn→∞ |an |1/n > 1; and thus
there is a subsequence satisfying |z||ank |1/nk > 1, hence |ank z nk | > 1 for k = 1, 2, . . .. It follows that the sequence
{an z n } cannot converge to 0, hence the series must diverge.

Remark. If 0 < r < ∞, the theorem gives no information about the convergence behavior of the power series for
|z| = r. In fact, almost anything can happen.
3 POWER SERIES 4

∑∞
For the rest of this section we assume given a power series n=0 an z n of radius r > 0; we set D = Dr (0) =
{z ∈ C : |z| < r} (D = C if r = ∞) and define f : D → C by
∞
∑
f (z) = an z n
n=0

if z ∈ D. We know that f is continuous. In fact, we proved that the series converges uniformly on every closed
disc centered at 0 included (with its boundary) in D thus, by Theorem 1, the restriction of f to all such discs
is continuous. If z ∈ D it will be contained in the interior of such a closed disc and continuity of f on the disc
translates to continuity at z. We want to prove a bit more.

Theorem 6 The function f defined above is holomorphic in D and

∞
∑ ∞
∑
f ′ (z) = nan z n−1 = (n + 1)an+1 z n
n=1 n=0

for all z ∈ D.

Remark. There are easy and hard ways of proving this theorem. If one can bear to do complex analysis without
too many examples, and postpones dealing with power series until after one has Cauchy’s theorem and formula,
then there is a very easy, one line proof. But it may be best to have good examples right now, so we go for the
hard stuff. However, to minimize the mess a bit, we state a few auxiliary lemmas.

Lemma 7 If {an }, {bn } are sequences of real numbers, if lim supn→∞ an = α ∈ [−∞, ∞], and if {bn } converges
to a positive number β, then lim supn→∞ an bn = αβ

The proof of this lemma is left as an exercise.

∞
∑ ∞
∑
Lemma 8 Let an z n be a power series of radius of convergence r ∈ [0, ∞]. Then the series (n + 1)an+1 z n
n=0 n=0
also has radius of convergence r.
∞
∑
Proof. The radius of convergence of the series in question is exactly the same as that of the series nan z n .
n=1
∞
∑
In fact, both series converge for exactly the same numbers z. To see this, suppose first z is such that nan z n
n=1
∞
∑
converges. If z = 0, then (of course!) (n + 1)an+1 z converges, so assume z ̸= 0. Then the convergence of
n

n=0
∞
∑
nan z n implies the convergence of
n=1

∞
∑ ∞
∑ ∞
1∑
(n + 1)an+1 z n = nan z n−1 = nan z n .
n=0 n=1
z n=1
∞
∑
Conversely, assume that the series (n + 1)an+1 z n converges. Then so does
n=0

∞
∑ ∞
∑ ∞
∑
z (n + 1)an+1 z n = (n + 1)an+1 z n+1 = nan z n .
n=0 n=0 n=1

The radius of convergence of the first series can thus be calculated by applying the formula to the second series,
and we get that (by Lemma 7)
1/n 1 1
1/radius = lim sup (n|an |) = lim n1/n lim sup |an |1/n = 1 · = .
n→∞ n→∞ n→∞ r r
3 POWER SERIES 5

The process can be repeated; replacing an by (n + 1)an+1 we see that the series
∞
∑ ∞
∑
(n + 1)(n + 2)an+2 z n = n(n − 1)an z n−2
n=0 n=2

also has radius r. Etc.

Lemma 9 Let z, h ∈ C, h ̸= 0, let n ∈ Z, n ≥ 0. Then

 
 (z + h)n − z n 
 − nz n−1 
 ≤ n(n − 1)|h|(|z| + |h|)
n−2
 .
h

Proof. If n = 0 or n = 1, then both sides of the inequality to be proved are 0; the inequality holds. Assume thus
that n ≥ 2. Then,
   
   ∑n ( ) 
 (z + h)n − z n   1 n 
 − nz n−1    j n−j
−z n
− nz n−1 
 h  = h j
z h 
 j=0 
   
 ∑( n ) 
 1 n−2 
=   j n−j
z h + nz n−1
h+z −zn n
− nz n−1 
h j 
 j=0 
   
n−2 ( )  n−2 ( ) 
∑ n  ∑ n 
 j n−j−1   j n−j−2 
= 
j
z h  = |h|  j
z h 
 j=0   j=0 

If we use that ( ) ( ) ( )
n n(n − 1) n−2 n−2
= ≤ n(n − 1) ,
j (n − j)(n − j − 1) j j
we can continue this chain of equalities with the following estimates
 
  n−2
∑( n ) 
 (z + h)n − z n   
 − nz n−1 
= |h|  j n−j−2 
z h
 h  j 
 j=0 
∑( n )
n−2
≤ |h| |z|j |h|n−2−j
j
j=0

∑(
n−2
n−2
)
≤ n(n − 1)|h| |z|j |h|n−2−j = n(n − 1)|h|(|z| + |h|)n−2 .
j
j=0

We are ready to prove Theorem 6.

∞
∑
Proof of Theorem 6. Let z ∈ D, thus |z| < r. Select ρ so that |z| < ρ < r. By Lemma 8, the series n(n −
n=2
1)|an |ρn−2 converges, set
∞
∑
A= n(n − 1)|an |ρn−2 .
n=2
∞
∑
By the same lemma, the series nan z n−1 also converges. Assume now h ∈ C and |h| < ρ − |z|. Then z + h ∈ D
n=1
3 POWER SERIES 6

(in fact, |z + h| < ρ < r) and

  ∞ )
 f (z + h) − f (z) ∑ ∞   ∑ ( (z + h)n − z n
   
 − nan z n−1  =  an − nz n−1 
 h   h 
n=1 n=2
∑ ∞  
 (z + h)n − z n 
≤ |an |  − nz n−1 
n=2
h
∑
≤ |h| n(n − 1)|an |(|z| + |h|)n−2
n=2

the last estimate being a consequence of Lemma 9. If we use that |z| + |h| < ρ and the definition of A, we proved
that  
 f (z + h) − f (z) ∑ ∞ 
 n−1 
 − nan z  ≤ A|h|
 h 
n=1

if |h| < ρ − |z|. The theorem follows.

Exercise 2 Compute the radius of convergence of the following power series

∞
∑
1. 3n z n .
n=0

∑∞
1 n
2. z .
n=0
n!
∞
∑
3. n!z n
n=0

∞
∑
Exercise 3 Consider the power series an z n where an ∈ Z for all n. Prove: The radius of convergence is at
n=0
most 1.

Exercise 4 Let an > 0 for all n ∈ N. Prove, accept or look it up:

an+1 √ √ an+1
lim inf ≤ lim inf n an ≤ lim sup n an ≤ lim sup .
n→∞ an n→∞ n→∞ n→∞ an
∞
∑
Conclude from this that given a power series an z n , if r = limn→∞ |an |/|an+1 | exists, then r is the radius of
n=0
convergence of the series.

(The formula
|an |
r = lim
n→∞ |an+1 |
is sometimes easier to use to determine the radius of convergence. Of course, it is only applicable when the limit
in question exists.)
Exercise 5 Let {an } be the Fibonacci sequence; c0 = 0, c1 = 1, cn = cn−1 + cn−2 for n ≥ 2.
∞
∑
1. Determine the radius of convergence of the series cn z n .
n=0

2. Show that in its disc of convergence, the sum of the series is a rational function. That is, with {cn } the
∞
∑
Fibonacci sequence, express cn z n in the form p(z)/q(z) where p, q are polynomial functions.
n=0
3 POWER SERIES 7

As a final result of this section, we get a first theorem on uniqueness of a power series expansion. Could two
different powers series converge to the same function? If by converge we mean converge at least for all z in some
open disc around the center, the answer is NO. That is because we can recover all the coefficients from the sum;
equal sums imply equal coefficients.

Theorem 10 Let 0 < r ≤ ∞ and let f : Dr (0) → C satisfy

∞
∑
f (z) = an z n
n=0

for all z ∈ Dr (0). Then f has complex derivatives of all orders in Dr (0) and
1 (n)
an = f (0) for n = 0, 1, 2, . . . .
n!
Proof. Exercise.