§Lecture 6 Sequences and Series

Definition 1 By a sequence in a set A, we mean a mapping f : N −→ A.

It is customary to denote a sequence f by {sn } where, sn := f (n). A
sequence {zn } of (complex) numbers is said to be convergent to the limit
w if for every  > 0 there exists an integer n0 such that for all n ≥ n0 , we
have,
|zn − w| < .

It is easily seen that the limit of a sequence in A ⊂ C if it exists, is

unique. But it need not belong to the set A.
We use the following two notations

lim z
n−→∞ n
:= w; OR zn −→ w,

to represent the limit of the sequence {zn }. If a sequence is not convergent

then it is said to be divergent. If {an }, {bn } are convergent sequences of
real or complex numbers, then the sequences {an + bn } and {zan } are also
so with limits given by

lim (an + bn ) = n−→∞

n−→∞
lim an + n−→∞
lim bn ; lim zan = z n−→∞
n−→∞
lim an .

It is fairly obvious that if zn = an + ıbn where, an , bn ∈ R, then zn −→ w

iff an −→ <(w) and bn −→ =(w). The following criterion for convergence
of sequences of real (complex) is one of the immediate consequences of
the construction of real numbers.

Theorem 1 (Cauchy ) A sequence {zn } of numbers is convergent iff

for every  > 0 there exists n0 such that for all n, m ≥ n0 , we have
|zn − zm | < .

Theorem 2 Let f : X −→ C be any function defined on a subset of C.

For any z ∈ X, f is continuous at z iff for every sequence zn −→ z, we
have, f (zn ) −→ f (z).

Definition 2 By a series of real (complex) numbers we mean an infinite

sum:
X
zn := z0 + zn + · · · + zn + · · ·
n

1
Of course, it is possible that there are only finitely many non zero terms
here. The sequence of partial sums associated to the above series is defined
n
X
to be sn := zk . We say the series is convergent if the associated sequence
k=1
{sn } of partial sums is convergent. In that case, if s is the limit of this
sequence, then we say s is the sum of the series and write
X
zn := s.
n

Once again it is immediate that if n zn and n wn are convergent series

P P

then for any complex number λ, we have, n λzn and n (zn + wn ) are
P P

convergent and
X X X X X
λzn = λ zn ; (zn + wn ) = zn + wn .
n n n n n

Cauchy’s convergence criterion can be applied to series also. This

yields:

Theorem 3 A series n zn of real or complex numbers is convergent iff

P

for every  > 0, there exists n0 such that for all n ≥ n0 and for all p ≥ 0,
we have,
|zn + zn+1 + · · · + zn+p | < .

Indeed, all notions and results that we have for sequences have cor-
responding notions and results for series also, via the sequence of partial
sums of the series. Thus, once a result is established for a sequence, the
corresponding result is available for series as well and vice versa, without
specifically mentioning it.
It follows that if a series is convergent, then its nth term zn tends to
0. However, this is not a sufficient condition for convergence of the series,
X 1
as illustrated by the series .
n n

Definition 3 A series n zn is said to be absolutely convergent if the

P

series n |zn | is convergent.

P

Again, it is easily seen that an absolutely convergent series is conver-

gent, whereas the converse is not true as seen with the standard example
1
(−1)n . The notion of absolute convergence plays a very important role
X

n n
2
throughout the study of convergence of series. As an illustration we shall
obtain the following useful result about the convergence of the product
series.

Definition 4 Given two series n an , n bn the (Cauchy) product of these

P P

two series is defined to be n cn where cn = nk=0 ak bn−k .

P P

Theorem 4 If n an , n bn are two absolutely convergent series then their

P P

Cauchy-product series is absolutely convergent and its sum is equal to the

product of the sums of the two series:
! !
X X X
cn = an bn . (1)
n n n

Proof: [not done in the class] We begin with the remark that if both the
series are of non negative real numbers, then the assertion of the theorem
is obvious. We shall use this in what follows.
Consider the remainder after (n − 1) terms of the corresponding
absolute series:
X X
Rn = |ak |; Tn = |bk |.
k≥n k≥n
Clearly,
X X X
|cn | ≤ |ak ||bl | = R0 T0 .
n≥0 k≥0 l≥0
Therefore the series n cn is absolutely convergent. Further,
P

  
X X X
ck −  ak   bk  ≤ R0 Tn+1 + T0 Rn+1 ,
k≤2n k≤n k≤n

since the terms that remain on the LHS after cancellation are of the form
ak bl where either k ≥ n+1 or l ≥ n+1. Upon taking the limit as n −→ ∞,
we obtain (1). ♠
An important property of an absolutely convergent series is:

Theorem 5 Let n zn be an absolutely convergent series. Then every

P

rearrangement n zσn of the series is also absolutely convergent.

P

Recall that a rearrangement n zσn of n zn is obtained by taking a

P P

permutation σ : N −→ N.
Lecture 7

3
 §Uniform Convergence of Sequ. of functions
We shall now consider the study of a family of sequences, viz., let
{fn } be a sequence of functions taking complex values defined on some
set A. Then to each x ∈ A, we get a sequence {fn (x)}. If each of these
sequences is convergent, then we get a function

f (x) = n→∞
lim fn (x).

Definition 5 Let {fn } be sequence of complex valued functions on a set

A. We say that it is uniformly convergent on A to a function f (x) if for
every  > 0 there exists n0 , independent of x ∈ A such that for all n ≥ n0 ,
we have, |fn (x) − f (x)| < , for all x ∈ A.

Remark 1 Observe that if {fn } is uniformly convergent, then for each

x ∈ A, we have, fn (x) −→ f (x). This is called point-wise convergence
of the sequence of functions. As seen in the example below, point-wise
convergence does not imply uniform convergence. However, it is fairly
easy to see that this is so if A is a finite set. Thus the interesting case
of uniform convergence occurs only when A itself is an infinite set. The
terminology is also adopted in an obvious way for series of functions via
the associated sequences. As in the case of ordinary convergence, we have
Cauchy’s criterion here also.

Theorem 6 A sequence of complex valued functions {fn } is uniformly

convergent iff it is uniformly Cauchy i.e., given  > 0 there exists n0 ,
such that for all n ≥ n0 and for all x ∈ A, we have,

|fn+p (x) − fn (x)| < .

Example 1 A simple example of a sequence which is point-wise conver-

gent but not uniformly convergent is fn : (0, ∞) → R given by fn (x) =
1/nx. It is uniformly convergent in [α, ∞) for all α > 0 but not so in
(0, α).

Example 2 The most useful series is the geometric series

1 + z + z2 + · · ·

4
The sequence of partial sums is given by
1 − zn
1 + z + · · · + z n−1 = .
1−z
As z n −→ 0 if |z| < 1, it follows that the geometric series point-wise
converges to 1/(1 − z) for all |z| < 1. In fact, if we take 0 < r < 1, then
in the disc Br (0), the series is uniformly convergent. For, given  > 0,
choose n0 such that rn0 < (1 − r). Then for all |z| < r and n ≥ n0 , we
have,
1 − zn 1 zn |z n0 |
− = ≤ <
1−z 1−z 1−z 1 − |z|

Theorem 7 Weierstrass1 M-test: Let n an be a convergent series of

P

positive terms. Suppose there exists M > 0 and an integer N such that
|fn (x)| < M an for all n ≥ N and for all x ∈ A. Then n fn is uniformly
P

and absolutely convergent in A.

Proof: Given  > 0 choose n0 > N such that an + an+1 + · · · + an+p <
/M, for all n ≥ n0 . This is possible by Cauchy’s criterion, since n an is
P

convergent. Then it follows that

|fn (x)| + · · · + |fn+p (x)| ≤ M (an + · · · an+p ) < ,

for all n ≥ n0 and for all x ∈ A. Again, by Cauchy’s criterion, this means
that fn is uniformly and absolutely convergent. ♠
P

Remark 2 The series n an in the above theorem is called a ‘majorant’

P

for the series n fn .

P

As an illustration of the importance of uniform convergence, we shall

prove:

Theorem 8 Let {fn } be a sequence of continuous functions defined and

uniformly convergent on a subset A of R or C. Then the limit function
f (x) = n−→∞
lim fn (x) is continuous on A.
1
Karl Weierstrass (1815-1897) a German mathematician is well known for his perfect rigor. He
clarified any remaining ambiguities in the notion of a function, of derivatives, of minimum etc..

5
Proof: [not done in the classLet x ∈ A be any. In order to prove the
continuity of f at x, given  > 0 we should find δ > 0 such that for all
y ∈ A with |y − x| < δ, we have, |(f (y) − f (x)| < . So, by the uniform
convergence, first we get n0 such that |fn0 (y) − f (y)| < /3 for all y ∈ A.
Since fn0 is continuous at x, we also get δ > 0 such that for all y ∈ A
with |y − x| < δ, we have |fn0 (y) − fn0 (x)| < /3. Now, using triangle
inequality, we get,

|f (y) − f (x)| ≤ |f (y) − fn0 (y)| + |fn0 (y) − fn0 (x)| + |fn0 (x) − f (x)| < ,

whenever y ∈ A is such that |y − x| < δ. ♠]

§Power series

K = R, or C.

Definition 6 By a formal power series in one variable t over K, we mean

a sum of the form
∞
an tn , an ∈ K.
X

n=0
Observe that when at most a finite number of an are non zero the
above sum gives a polynomial. Thus, all polynomials in t are power series
in t. We can add two power series, by ‘term-by-term’ addition and we can
also multiply them by scalars, just like polynomials, viz.,

an tn + bn tn := (an + bn )tn ; α( an tn ) := αan tn .

X X X X X

n n n n n

Let K[[t]] denote the set of all formal power series in t over K. The set
of polynomials is contained in K[[t]] and is closed under each of these
operations.

Definition 7 A formal power series P (t) is said to be convergent if there

exists a non zero number z (real or complex) such that the series of
complex numbers n an z n is convergent.
P

We shall now consider a few theorems, which are attributed to Cauchy-

Hadamard2 and Abel3 , are most fundamental in the theory of power series.
2
Jacques Hadamard(1865-1963) was a French Mathematician who was the most influential mathe-
matician of his days, worked several areas of mathematics such as complex analysis, analytic number
theory, partial differential equations, hydrodynamics and logic.
3
Niels Henrik Abel (1802-1829) was a Norwegian, who died young under deprivation. At the age of

6
 It may be recalled that the limsup of a sequence of real numbers {bn }
is defined to be

lim sup{bn } := lim

n
(sup{bn , bn+1 , . . .}).
n

This is also the same as the least upper bound of the set of limits of all
convergent subsequences of {bn }.

n
Theorem 9 (Cauchy-Hadamard) Let P = n≥0 an t be a power se-
P

ries over C. Then

(a) there exists a number 0 ≤ R ≤ ∞ such that for all 0 < r < R, the
series P (z) is absolutely and uniformly convergent in |z| ≤ r and for all
|z| > R the series is divergent.
1 q
(b) = lim sup n |an |, with the convention (0)−1 = ∞; (∞)−1 = 0.
R

Proof: Indeed, all that we have to prove is to take R as given by (b)

and show that it satisfies (a). So, let 0 < r < R. Choose r < s <
R. Then 1/s > 1/R and hence by the definition of limsup, we must
q
have n0 such that for all n ≥ n0 , n |an | < 1/s. Therefore, for all |z| ≤
r, |an z n | < (r/s)n , n ≥ n0 . Since r/s < 1, by Weierstrass majorant
criterion, (theorem 7), it follows that P (z) is absolutely and uniformly
convergent. On the other hand, suppose |z| > R. We fix s such that |z| >
s > R. Then 1/s < 1/R, hence, by the property of limsup, there exist
q
infinitely many nj , for which nj |anj | > 1/s. This means that |anj z nj | >
(R/s)nj . It follows that the sequence {an z n } is unbounded and hence the
series n an z n is divergent. ♠
P

Remark 3 Observe that if P (z) is convergent for some z, then from the
last part of (a), the radius of convergence of P is at least |z|.

Definition 8 The number R obtained in the above theorem is called the

radius of convergence of P (t). The second part of the theorem gives you
the formula for it. This is called the Cauchy-Hadamard formula. Thus the
21, he proved the impossibility of solving a general quintic by radicals. He did not get any recognition
during his life time for his now famous works on convergence, on so called abelian integrals, and on
elliptic functions.

7
collection of all points at which a given power series converges consists
of an open disc centered at the origin and perhaps some points on the
boundary of the disc. Observe that the theorem does not say anything
about the convergence of the series at points on the boundary |z| = R.
X n X tn X tn
Example 3 The series t , , all have radius of convergence
n n n n n2
1. The first one is not convergent at any point of the boundary of the disc
of convergence |z| = 1. The second is convergent at all the points of the
boundary except at z = 1 and the last one is convergent at all the point
of the boundary. These examples clearly illustrate that the boundary
behavior of a power series needs to be studied more carefully.

Definition 9 Given a power series P (t) = n≥0 an tn , the derived series

P

P 0 (t) is defined by taking term-by-term differentiation: P 0 (t) = n≥1 nan tn−1 .

P

an n+1
The series n≥0 n+1 t is called the integrated series.
P

As an application of Cauchy-Hadamard formula, we derive:

Theorem 10 Any given power series, its derived series and its integrated
series all have the same radius of convergence.

Proof: Let the radius of convergence of P (t) = n an tn , and P 0 (t) be r, r0

P

respectively. It is enough to prove that r = r0 .

We will first show that r ≥ r0 . For this we may assume without loss
of generality that r0 > 0. Let 0 < r1 < r0 . Then
   

|an |r1n ≤ r1  n|an |r1n−1 | < r0  n|an |r0n−1 | < ∞.

X X X

n≥1 n≥1 n≥1

It follows that r ≥ r1 . Since this is true for all 0 < r1 < r0 this means
r ≥ r0.
Now to show that r ≤ r0 , we can assume that r > 0 and let 0 < r1 < r.
Choose r2 such that r1 < r2 < r. Then for each n ≥ 1
n r1 n n M n
!
n−1
nr1 ≤ r2 ≤ r2
r1 r2 r1
r1 k
where M = k≥1 k r2 < ∞, since the radius of convergence of ktk is 1.
P P
k
Therefore,
M
n|an |r1n−1 ≤ |an |r2n < ∞.
X X

n≥1 r1 n≥1

8
We conclude that r0 > r1 and since this holds for all r1 < r, it follows that
r0 ≥ r. ♠

9
Lecture 8

Remark 4 A power series with radius of convergence 0 is useless for

us, for, it only defines a function at a point. A power series P (t) with a
positive radius of convergence R defines a continuous function z 7→ p(z)
P

in the disc of convergence BR (0), by theorem 8. Also, by shifting the

origin, we can even get continuous functions defined in BR (z0 ), viz., by
substituting t = z − z0 . One expects that functions which agree with a
convergent power series in a small neighborhood of every point will have
properties akin to those of polynomials. Before taking up this study, let
us make a definition.

Definition 10 Let f : Ω −→ C be a function. We say f is analytic at

z0 ∈ Ω if there exists r > 0 and a power series P of radius of convergence
≥ r such that f (z) = P (z − z0 ) for all z ∈ Br (z0 ). We say f is analytic in
Ω if it is so at each point z0 ∈ Ω.

Remark 5 As seen above it follows that an analytic function is contin-

uous. On the other hand, what is not obvious is that if P (t) is a for-
mal power series with positive radius of convergence r, then the function
f (z) = P (z − z0 ) is analytic in |z − z0 | < r. (Observe that the definition
only says that it is analytic at z0 .) This fact can be directly proved by
algebraic methods. However, we need not do this, as the result will follow
easily from Taylor’s theorem which we shall prove in due course.
The next theorem tells you that analytic functions are holomorphic.

Theorem 11 Abel: Let n≥0 an tn be a power series of radius of conver-

P

gence R > 0. Then the function defined by

an (z − z0 )n
X
f (z) =
n

is complex differentiable in Br (z0 ). Moreover the derivative of f is given

by the derived series

f 0 (z) = nan (z − z0 )n−1

X

n≥1

inside |z − z0 | < R.

10
Proof: Without loss of generality, we may assume that z0 = 0. We already
know that the derived series is convergent in BR (0) and hence defines a
continuous function g on it. We have to show that this function g is the
derivative of f at each point of BR (0). So, fix a point z ∈ BR (0). Let
|z| < r < R and let 0 6= |h| ≤ r − |z|. Consider the difference quotient
f (z + h) − f (z) X
− g(z) = un (h) (2)
h n≥1

an [(z + h)n − z n ]
where, we have put un (h) := − nan z n−1 . Upon simplifi-
h
cation we get

un (h) = an [(z + h)n−1 + (z + h)n−2 z + · · · + (z + h)z n−2 + z n−1 − nz n−1 ]. (3)

We must show that given  > 0, there exists δ > 0 such that for all
0 < |h| < δ, we have,
f (z + h) − f (z)
− g(z) < . (4)
h
Since |z| < r and |z + h| < r, using (3), it follows that

|un (h)| ≤ 2n|an |rn−1 . (5)

Since the derived series has radius of convergence R > r, it follows that
we can find n0 such that

2|an |nrn−1 < /2.

X
(6)
n≥n0

On the other hand, again using (3), each un (h) is a polynomial in h which
vanishes at h = 0. Therefore so does the finite sum n<n0 un (h) . Hence
P

by continuity, there exists δ 0 such that |h| < δ 0 implies that

2|an |nrn−1 < /2.

X
(7)
0<n<n0

Taking δ = min{δ 0 , r − |z|} and combining (6) and (7) yields (4). ♠

Remark 6
(i) It follows that an analytic function is repeatedly differentiable, with
its nth derivatives at z0 given by n!an . As a consequence, we know that,

11
at each point the power series representing the function is unique.
(ii) It is fairly obvious that the sum of any two analytic functions is
again an analytic function. The corresponding statement about power
series is that the sum of two formal power series is convergent with radius
of convergence at least the minimum of the two radii of convergence.
Similarly, the product of two analytic functions is also analytic. The
identity function written f (z) = z is clearly analytic in the entire plane
(take P (t) = z0 + t to see that f is analytic at z0 ). Starting from this
and using the above two observations we can deduce that any polynomial
function is analytic throughout the plane.
(iii) Later on we shall prove that every holomorphic function is analytic,
thus completing the picture. In particular, this will then prove that a
function which is complex differentiable once in an open set is complex
differentiable infinitely many times.
Of course, such a result is far from being true in the real differentiable
case. Examples such as g(x) = xn |x| illustrate the existence of functions
which are n-times real differentiable but not (n+1)-times. Also, there are
functions which are C ∞ but not real analytic. (Take f (x) = 0 for x ≤ 0
2
and = e−1/x for x > 0.)
For the time being, we are satisfied with getting plenty of examples
of holomorphic functions via analytic functions. That is the topic that
we are going to take up now.

12