Sequence and Series of Numbers

M.T.Nair (IIT Madras)

July 23, 2014
Contents
1 Sequence of numbers 2
1.1 Convergence and divergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Some tests for convergence and divergence . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Monotonic sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.4 Subsequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.5 Further examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.6 Cauchy criterion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2 Series of Real Numbers 25
2.1 Convergence and divergence of series . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.2 Some tests for convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.3 Alternating series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.4 Absolute convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

Taken from the forthcoming book Calculus of One Variable by M.Thamban Nair
1
1 Sequence of numbers
Suppose for each positive integer n, we are given a real number a
n
. Then, the list of numbers
a
1
, a
2
, . . . , a
n
, . . .
is called a sequence. A more precise denition of a sequence is the following:
Denition 1.1. A sequence of real numbers is a function from the set N of natural numbers to the
set R of real numbers.
If f : N R is a sequence, and if a
n
= f(n) for n N, then we write the sequence f as
(a
1
, a
2
, . . .) or (a
n
) or {a
n
},
and the term a
n
is called the n
th
term of the sequence (a
n
). In specic cases one may also write a
sequence as
(a
1
, a
2
, . . . , a
n
, . . .) instead of (a
1
, a
2
, . . .).
Remark 1.2. A sequence (a
1
, a
2
, . . .) is dierent from the set {a
n
: n N}. For instance, a number
may be repeated in a sequence (a
n
), but it need not be written repeatedly in the set {a
n
: n N}. As
an example, (1,
1
2
, 1,
1
3
, . . . , 1,
1
n
, . . .) is a sequence (a
n
) with a
2n1
= 1 and a
2n
= 1/(n + 1) for each
n N, whereas the set {a
n
: n N} is same as the set {1/n : n N}.
Example 1.3. Let us consider a few examples of sequences:
(i) (a
n
) with a
n
= 1 for all n N.
(ii) (a
n
) with a
n
= n for all n N.
(iii) (a
n
) with a
n
= 1/n for all n N.
(iv) (a
n
) with a
n
= n/(n + 1) for all n N.
(v) (a
n
) with a
n
= (1)
n
for all n N the sequence takes values 1 and 1 alternately.
1.1 Convergence and divergence
In certain sequences the n
th
term comes closer and closer to a particular number as n becomes larger
and larger. For example, in the sequence (
1
n
), the n
th
term comes closer and closer to 0, whereas in
(
n
n+1
), the n
th
term comes closer and closer to 1 as n becomes larger and larger. If you look at the
sequence ((1)
n
), the terms oscillate between 1 and 1 as n varies, whereas in (n
2
) the terms become
larger and larger.
2
What do we mean by the statement a
n
comes closer and closer to a number a as n
becomes larger and larger?
Saying dierently: a
n
can be made arbitrarily close to a by taking n large enough.
Denition 1.4. A sequence (a
n
) of real numbers is said to converge to a real number a if for every
> 0, there exists a positive integer N (in general depending on ) such that
|a
n
a| < n N.
A sequence which converges is called a convergent sequence, and a sequence which does not converge
is called a divergent sequence.
Suppose (a
n
) converges to a. Then as per the above denition, taking = 10
k
for some k N,
there exists some N
k
N such that
|a
n
a| < 10
k
n N
k
.
In other words:
If n N
k
and if a
n
and a are expressed in decimal expansion, then
the rst k decimal places of a
n
and a are the same.
If (a
n
) converges to a, then we write
a
n
a as n
or, in short,
a
n
a,
which we may read as a
n
tends to a as n tends to innity.
If (a
n
) does not converge to a, then we write a
n
a.
It is easily seen that (verify)
a
n
a |a
n
a| 0 |a
n
| |a|.
Remark 1.5. We must keep in mind that the symbol is not a number; it is only a notation used
in the context of describing some properties of real numbers, such as in Denition 1.4.
Note that, in the denition of convergence, dierent can result in dierent N, i.e., the number
N may vary as varies. This fact is illustrated in the following examples.
3
Example 1.6. Consider the sequence (a
n
) with a
n
= 1/n for n N. We expect that a
n
0 as
n . Let us prove this: So, let > 0 be given. Then
1
n
< n >
1

.
Thus, taking a positive integer N > 1/, we obtain |a
n
0| < for all n N. Similarly, for any
> 0,

(1)
n
n
0

=
1
n
< n N,
where N is a positive integer such that N > 1/. Thus we have shown that
1
n
0 and
(1)
n
n
0 as n .
Note that the positive integer N in both the examples depend on the chosen.
Example 1.7. Consider the sequence (a
n
) with a
n
= n/(n + 1) for n N. We claim that a
n
1 as
n . Note that
|a
n
1| =
1
n + 1
so that, for > 0,
|a
n
1| < n + 1 >
1

.
Thus, taking a positive integer N with N 1/, we obtain |a
n
1| < for all n N. Note that,
n 100 = |a
n
1| <
1
100
,
n 1000 = |a
n
1| <
1
1000
.
More generally, for a given k N,
|a
n
1| <
1
10
k
n 10
k
.
It is also clear that |a
n
1| 1/10
k
if n < 10
k
.
Example 1.8. For n N, let
a
n
=
9
10
+
9
10
2
+ +
9
10
n
,
i.e., a
n
= 0.99 9 with 9 repeated n times. You may guess that a
n
1 as n . This is true:
Note that
9
10
+
9
10
2
+ +
9
10
n
=
9
10
_
1 +
1
10
+ +
1
10
n1
_
=
9
10
_
1
1
10
n
1
1
10
_
= 1
1
10
n
.
Hence, |a
n
1| = 1/10
n
so that
|a
n
1| < 10
n
>
1

.
Let N N be such that 10
N
> 1/. Then, we have |a
n
1| < for all n N. Therefore, a
n
1 as
n .
4
Example 1.9. For n N, let
a
n
=
3
10
+
3
10
2
+ +
3
10
n
,
i.e., a
n
= 0.33 3 with 3 repeated n times. Note that
3
10
+
3
10
2
+ +
3
10
n
=
3
10
_
1
1
10
n
1
1
10
_
=
1
3
_
1
1
10
n
_
.
Thus, |a
n
1/3| = 1/(3 10
n
) so that, for any given > 0, we can choose a positive integer N
(depending on ) such |a
n
1/3| < for all n N. Therefore, a
n
1/3 as n .
In the examples above we have used an important property of real numbers, known as the
Archimedean Property :
Archimedean Property : Given any x R, there exists n N such that n > x.
We shall use this property throughout the text without mentioning it explicitly.
Let us show the convergence of a few more sequences which we encounter very often.
Exercise 1.10. Using the denition of convergence, show that
1
n
2
0 and
1
2
n
0.
It can happen that for a convergent sequence (a
n
), same N can work for dierent s. To
illustrate this point, look at the following examples.
Example 1.11. Let a
n
= 1 for all n N. We can easily guess that (a
n
) converges to 1. Note that
for any > 0,
|a
n
1| < n 1.
Thus, N = 1 works for any > 0.
Example 1.12. Let a
n
= n for all n {1, 2, . . . , 99} and a
n
= 1 for all n 100. You must have
guessed that, (a
n
) converges to 1. Yes; it is true: For any given > 0,
|a
n
1| < n 100.
Thus, N = 100 works for any > 0.
Denition 1.13. (i) A sequence is said to be a constant sequence if all its terms are the same,
i.e., a
n
= a
1
for all n N.
(ii) A sequence is said to be an eventually constant sequence if all its terms after a particular
stage are the same, i.e., there exists k N such that a
n
= a
k
for all n k.
Remark 1.14. We may observe that, for > 0,
|a
n
a| < a
n
(a , a + ).
5
Thus, (a
n
) converges to a R if and only if for every > 0, there exists a positive integer N such
that
a
n
(a , a + ) n N.
In other words, a
n
a as n if and only if for every > 0, a
n
belongs to the open interval
(a , a + ) for all n after some nite stage, and this nite stage may vary according as varies.
The above remark leads to the following (verify):
1. a
n
a if and only if for every open interval I containing a, there exists a positive integer N
(depending on I) such that a
n
I for all n N.
2. a
n
a if and only if there is an open interval I containing a such that innitely may a
n
s are
not in I.
Now, let us observe an important property of convergent sequences.
THEOREM 1.15. If (a
n
) converges to a, then (a
n
) cannot converge to b with b = a.
Proof. Suppose a
n
a and a
n
b as n , and suppose that b = a. Let I
1
and I
2
be disjoint
open intervals such that a I
1
and b I
2
. Then, there exist positive integers N
1
and N
2
such that
a
n
I
1
n N
1
and a
n
I
2
n N
2
.
Thus, a
n
I
1
I
2
for all n N := max{N
1
, N
2
}; a contradiction, since I
1
I
2
= .
Thus, we can dene the concept of the limit of a convergent sequence.
Denition 1.16. If (a
n
) converges to a, then a is called the limit of (a
n
), and we write
lim
n
a
n
= a.

Exercise 1.17. a
n
a if and only if a
k+n
a as n for any k N.
Now, let us give a few examples of divergent sequences.
Example 1.18. Let a
n
= (1)
n+1
for n N. The terms of the sequence change alternately between
1 and 1. So, one may guess that it cannot converge to any number. To see this, we observe that, for
any a R, if we take 0 < < 1/2, either 1 or 1 will be outside the interval (a , a + ); in other
words, innitely many of the terms of the sequence (a
n
) lie outside (a , a + ).
Example 1.19. Let (a
n
) be dened by a
2n1
= 1/n and a
2n
= 1. If we take a positive number less
than 1, then, for any a R, innitely many of the terms of the sequence (a
n
) lie outside (a, a+).
For instance, let us take = 1/2. If a 1/2, then innitely many of a
2n1
lie outside (a , a + ),
and if a < 1/2, then every a
2n
lie outside (a , a + ) for all n N. Thus, a
n
a for any a R.
6
Example 1.20. Let a
n
= n for n N. In this case, for any given > 0 and for any a R, innitely
many of the terms of the sequence (a
n
) lie outside (a , a + ). Similar is the case if a
n
= n
2
or
a
n
= n
2
/(n + 1) (verify). Thus, a
n
a for any a R.
Denition 1.21. A sequence (a
n
) is said to be an alternating sequence if a
n
changes sign alter-
nately, that is, a
n
a
n+1
< 0 for every n N.
An alternating sequence may converge or diverge. We have seen that the sequence ((1)
n
) diverges,
whereas ((1)
n
/n) converges to 0.
A divergent sequence may have some specic properties such as the ones dened below.
Denition 1.22. Consider a sequence (a
n
).
(1) We say that (a
n
) diverges to innity if for every M > 0, there exists N N such that
a
n
> M n N,
and write this fact as a
n
.
(2) We say that (a
n
) diverges to minus innity if for every M > 0, there exists N N such that
a
n
< M n N
and write this fact as a
n
.

Example 1.23. (i) The sequence (a

n
) with a
n
= n
2
/(n + 1) diverges to innity: Note that
n
2
n + 1

n
2
n N.
(ii) The sequence (a
n
) with a
n
= 2
n
diverges to innity: Note that 2
n
n for all n N.
Observe the following facts:
1. If a
n
or a
n
as n , then (a
n
) is a divergent sequence.
2. a
n
if and only if a
n
.
3. If (a
n
) and (b
n
) are such that a
n
b
n
for all n N, then
(a) a
n
as n implies b
n
as n ,
(b) b
n
as n implies a
n
as n .
7
Exercise 1.24. Prove the above statements.
The following theorem is useful for showing convergence of certain sequences using the convergence
of some other sequences.
THEOREM 1.25. Suppose a
n
a and b
n
b as n . Then the following results hold.
(i) a
n
+ b
n
a + b as n ,
(ii) c a
n
c a as n for any real number c,
(iii) If a
n
b
n
for all n N, then a b.
(iv) (Sandwich theorem) If a
n
c
n
b
n
for all n N, and if a = b, then c
n
a as n .
Proof. Let > 0 be given.
(i) Note that, for every n N,
|(a
n
+ b
n
) (a + b)| = |(a
n
a) + (b
n
b)|
|a
n
a| +|b
n
b|.
Since a
n
a and b
n
b, the above inequality suggests that we may take
1
= /2, and consider
N
1
, N
2
N such that
|a
n
a| <
1
n N
1
and |b
n
b| <
1
n N
2
so that
|(a
n
+ b
n
) (a + b)| |a
n
a| +|b
n
b| < 2
1
=
for all n N := max{N
1
, N
2
}.
(ii) Note that
|ca
n
ca| = |c| |a
n
a| n N.
If c = 0, then (ca
n
) is a constant sequence with every term 0, and hence ca
n
0. Next, suppose
c = 0. Then we have
|ca
n
ca| < |a
n
a| <

|c|
.
Since a
n
a, there exists N N such that
|a
n
a| <

|c|
n N.
Thus,
|ca
n
ca| < n N.
8
(iii) Suppose a
n
b
n
for all n N. Let N
1
, N
2
N be such that
|a
n
a| < n N
1
and |b
n
b| < n N
2
.
Hence,
a < a
n
< a + and b < b
n
< b + n N = max{N
1
, N
2
}.
In particular,
a < a
n
b
n
< b + n N.
Thus, we obtain a < b + 2. Since this is true for every > 0, we have a b.
(iv) Suppose a
n
c
n
b
n
for all n N and a = b. We have to show that c
n
a. Note that
c
n
a
n
b
n
a
n
= (b
n
a) + (a a
n
) n N.
Hence,
0 |c
n
a
n
| |b
n
a| +|a a
n
| n N.
Therefore, from the fact that a
n
a and b
n
a and the result in (iii), we obtain |c
n
a
n
| 0 so
that by (i), c
n
= a
n
+ (c
n
a
n
) a.
Suppose a
n
a and b
n
b as n . We shall prove:
1. a
n
b
n
ab and
2. a
n
/b
n
a/b whenever b
n
= 0 for all n N and b = 0.
The following theorem together with Theorem 1.25 can be used for inferring convergence or diver-
gence of certain sequences.
THEOREM 1.26. The following results hold.
(i) If a > 1, then a
n
.
(ii) If 0 < a < 1, then a
n
0.
Proof. (i) Suppose a > 1. Writing a = 1 + r with r > 0, we have
a
n
= (1 + r)
n
= 1 + nr +
n(n 1)
2!
r
2
+ + r
n
for all n N. Hence, a
n
rn for all n N. Since rn , we have a
n
.
(ii) Let 0 < a < 1 and > 0 be given. Then b = 1/a satises b > 1 so that b
n
. Hence there
exists N N such that b
n
> 1/ for all n N. But, b
n
> 1/ if and only if a
n
< . Hence,
0 < a
n
< n N.
Thus, a
n
0.
9
Example 1.27. For 0 < r < 1 and n N, let a
n
= 1 + r + r
2
+ + r
n
. Then
a
n
=
1 r
n+1
1 r

1
1 r
as n .
Exercise 1.28. Show that |a| < 1 implies a
n
0.
Exercise 1.29. For a 0, prove the following.
(i) a
n
0 a < 1.
(ii) a
n
a > 1.
Exercise 1.30. If a
n
a and a = 0, then show that there exists k N such that a
n
= 0 for all
n k.
Exercise 1.31. Show that
a
n
:=
_
1 +
1
n
_
1/n
1.
[Hint: Observe that 1 a
n
(1 + 1/n) for all n N.]
1.2 Some tests for convergence and divergence
THEOREM 1.32. (Ratio test) Suppose a
n
> 0 for all n N such that lim
n
a
n+1
a
n
= for some
0. Then the following hold.
(i) If < 1, then a
n
0.
(ii) If > 1, then a
n
.
Proof. (i) Suppose < 1. Let q be such that < q < 1. Then there exists N N be such that
a
n+1
a
n
q n N.
Hence,
0 a
n
q
nN
a
N
n N.
By Theorem 1.26, q
nN
0 as n . Hence, by Sandwich theorem, a
n
0 as n .
(ii) Suppose > 1. Let q be such that 1 < q < . Then there exists N N be such that
a
n+1
a
n
> q n N.
Hence,
a
n
q
nN
a
N
n N.
By Theorem 1.26, q
nN
as n . Hence, a
n
.
10
Example 1.33. Let 0 < a < 1. Then na
n
0 as n . To see this, let a
n
:= na
n
for n N.
Then we have
a
n+1
a
n
=
(n + 1)a
n+1
na
n
=
(n + 1)a
n
n N.
Hence, lim
n
a
n+1
a
n
= a < 1. Thus, by Theorem 1.32, na
n
0.
Similarly, it can be shown (give details) that, for any k N, n
k
a
n
0 as n .
Remark 1.34. The converse of the results (i) and (ii) in Theorem 1.32 does not hold. To see this
rst consider the following examples:
(i) Consider (a
n
) with a
n
= 1/n. Then a
n
0, but a
n+1
/a
n
1.
(ii) Consider (a
n
) with a
n
= n. Then a
n
, but a
n+1
/a
n
1.
In fact, we shall see in the next section that the condition < 1 in Theorem 1.32 is too strong, in
the sense that, not only we have the convergence of (a
n
) to 0, but also we can show the convergence
of the sequence (s
n
), where s
n
= a
1
+ a
2
+ + a
n
. Note that if s
n
s for some s R, then
a
n
= s
n
s
n1
0.
The following theorem gives a sucient condition for certain number to be a limit of a given
sequence.
THEOREM 1.35. Let (a
n
) be a sequence such that
|a
n+1
a| r|a
n
a| n N
for some a R and for some r with 0 < r < 1. Then a
n
a.
Proof. For each n N, we have
|a
n+1
a| r|a
n
a| r
n
|a
1
a|.
Since 0 < r < 1, by Theorem 1.26, r
n
0 so that by Sandwich theorem, a
n
a.
Example 1.36. Let a sequence (a
n
) be dened iteratively as follows :
a
1
= 1, a
n+1
=
2 + a
n
1 + a
n
, n = 1, 2, . . . .
Let us assume for a moment that (a
n
) converges to a R. Then we obtain
a =
2 + a
1 + a
,
i.e., a(1 + a) = 2 + a, i.e., a
2
= 2. Thus, if (a
n
) converges, then the limit must be a =

2. Now, we
prove that (a
n
) actually converges to a :=

2. Note that
a
n+1
a =
2 + a
n
1 + a
n

2 + a
1 + a
=
(2 + a
n
)(1 + a) (1 + a
n
)(2 + a)
(1 + a
n
)(1 + a)
=
a a
n
(1 + a
n
)(1 + a)
.
11
Note that a
n
1 for all n N and a 1. Hence, we obtain
|a
n+1
a|
|a a
n
|
(1 + a
n
)(1 + a)

|a
n
a|
4
n N.
Therefore, by Theorem 1.35, a
n
a =

2.
Remark 1.37. In view of Theorem 1.35, one may ask the following question:
If (a
n
) is such that |a
n+1
a| < |a
n
a| for all n N for some a R, then does (a
n
)
converge to a?
Not necessarily! To see this, consider the sequence (a
n
) with
a
n
=
n + 1
n
, n N.
Since
n + 2
n + 1
<
n + 1
n
n N,
taking a = 0, we have |a
n+1
a| < |a
n
a| for all n N. But (a
n
) does not converge to 0. In fact,
a
n
1.
One may also enquire if the condition on (a
n
) in Theorem 1.35 can be replaced by
|a
n+2
a
n+1
| r|a
n+1
a
n
| n N
for some r with 0 < r < 1. In this case, the answer is in the armative. We shall present its proof at
the end of the present section after introducing the concept of a Cauchy sequence.
Exercise 1.38. Let a
n+1
= 1 +
1
a
n
for all n N. Prove the following:
(i) a
n
a a =
1 +

5
2
.
(ii) a =
1 +

5
2
satises a = 1 + 1/a.
(iii) a
n+1
a| r|a
n
a| for some r with 0 < r < 1.
Remark 1.39. The terms of the sequence (a
n
) in the above exercise
1
1
,
2
1
,
3
2
,
5
3
,
8
5
,
13
8
, . . . .
These terms are ratios of consecutive terms of the
1, 1, 2, 3, 5, 8, 13, . . .
which is the well-known Fibonacci sequence
1
. The number
1 +

5
2
,
1
Named after the Italian mathematician Leonardo Fibonacci (1170 1250).
12
which is the limit of the sequence (a
n
), is called the golden ratio
2
. The golden ratio is approximately
equal to 1.618033 which is correct to (1 +

5)/2 up to six decimal places, i.e.,

1 +

5
2
1.618033

<
1
10
6
.
In fact, the number (1 +

5)/2 is the ratio of the sides and with < of a rectangle such that
when the square of side is cut out of the rectangle, the remaining rectangle has the same ratio as
its sides, i.e,

=


.
Indeed, if = /, then the above relation gives = 1/(1), i.e., = 1+1/ so that = (1+

5)/2.
Such a rectangle is called a golden rectangle.
Next we prove a necessary condition for the convergence of a sequence.
THEOREM 1.40. (Boundedness test) If (a
n
) converges, then there exists M > 0 such that
|a
n
| M n N.
Proof. Suppose a
n
a. Note that, for all n N,
|a
n
| = |(a
n
a) + a| |a
n
a| +|a|.
Since |a
n
a| 0, there exists N N such that |a
n
a| 1 for all n N. Then we have
|a
n
| 1 +|a| n N.
Thus, taking M = max{1 +|a|, |a
1
|, |a
2
|, . . . , |a
N
|}, we have |a
n
| M for all n N.
The converse of Theorem 1.40 is not true. Note that the sequence (a
n
) with a
n
= (1)
n
satises
|a
n
| = 1 for all n N, but it is not convergent.
For later use we introduce the following denition.
Denition 1.41. A sequence (a
n
) is said to be
(1) bounded above if there exists a real number M such that a
n
M for all n N;
(2) bounded below if there exists a real number M

such that a
n
M

for all n N.
A sequence which is bounded above and bounded below is said to be a bounded sequence.
A sequence which is not bounded is called an unbounded sequence.
2
It is known from antiquity that constructions based on approximate values of golden ratio are aesthetically pleasing.
13
We observe the following:
1. A sequence (a
n
) is bounded if and only if there exists M > 0 such that
|a
n
| M n N.
2. A sequence (a
n
) is unbounded if and only if it is either not bounded above or not bounded
below.
3. If (a
n
) diverges to , then it is not bounded above, and if it diverges to , then it is not
bounded below.
Thus, according to Theorem 1.40:
Every convergent sequence is bounded.
Example 1.42. Let us look at a few examples of bounded and unbounded sequences (verify).
(i) (a
n
) with a
n
= (1)
n
is a bounded sequence.
(ii) (a
n
) with a
n
= (1)
n
n is neither bounded above nor bounded below, and it neither diverges
to + nor diverges to .
(iii) (a
n
) with a
n
= n is bounded above, but not bounded below, and it diverges to .
(iv) (a
n
) with a
n
= n is bounded below, but not bounded above, and it diverges to +.
Theorem 1.40 can be used to show that certain sequence is not convergent, as in the following
example.
Example 1.43. For n N, let
a
n
= 1 +
1
2
+
1
3
+ . . . +
1
n
.
Then (a
n
) diverges: To see this, observe that
a
2
n = 1 +
1
2
+
1
3
+ . . . +
1
2
n
= 1 +
1
2
+
_
1
3
+
1
4
_
+
_
1
5
+
1
6
+
1
7
+
1
8
_
+
. . . +
_
1
2
n1
+ 1
+ . . . +
1
2
n
_
1 +
n
2
.
Hence, (a
n
) is not a bounded sequence, so that it diverges.
14
Theorem 1.40 will be used for proving the following.
THEOREM 1.44. Suppose a
n
a and b
n
b. Then we have the following.
(i) a
n
b
n
ab as n .
(ii) If b
n
= 0 for all n N and b = 0, then
1
b
n

1
b
and
a
n
b
n

a
b
.
Proof. (i) Note that, for every n N,
|a
n
b
n
ab| = |a
n
(b
n
b) + (a
n
a)b|
|a
n
| |b
n
b| +|a
n
a| |b|.
By Theorem 1.40, (a
n
) is bounded, say |a
n
| M for all n N. Hence, we have
0 |a
n
b
n
ab| M |b
n
b| +|b||a
n
a|.
Since a
n
a and b
n
b, by Theorem 1.25, we obtain a
n
b
n
ab as n .
(ii) Suppose b
n
= 0 for all n N and b = 0. Note that, for every n N,

1
b
n

1
b

=
|b
n
b|
|b
n
| |b|
and
|b
n
| = |b + (b
n
b)| |b| |b
n
b|
Since b
n
b and b = 0, there exists N N be such that
|b
n
b| < |b|/2 n N.
Then,
|b
n
| = |b + (b
n
b)| |b| |b|/2 = |b|/2 n N.
Thus, we obtain

1
b
n

1
b

|b
n
b|
(|b|/2) |b|
=
2
|b|
2
|b
n
b| n N.
Hence, by Theorem 1.25, we obtain 1/b
n
1/b as n . Now, using (i), we also obtain a
n
/b
n
a/b
as n .
Exercise 1.45. Prove the following.
(i) If (a
n
) is not bounded above, then there exists a sequence (k
n
) of natural numbers such that
k
n
k
n+1
for all n N and a
k
n
+ as n .
(ii) If (a
n
) is not bounded below, then there exists sequence (k
n
) of natural numbers such that
k
n
k
n+1
for all n N and a
k
n
as n .
15
Exercise 1.46. Let (a
n
) and (b
n
) are such that a
n
b
n
for all n N. Prove the following:
(i) If (b
n
) is bounded above, then (a
n
) bounded above.
(ii) If (a
n
) is bounded below, then (b
n
) bounded below.
1.3 Monotonic sequences
We have seen that a bounded sequence need not converge. However, we shall show that, if the terms
of a bounded sequence either increases or decreases, then the sequence does converge.
Denition 1.47. A sequence (a
n
) is said to be a
(1) monotonically increasing sequence if a
n
a
n+1
for all n N;
(2) monotonically decreasing sequence if a
n
a
n+1
for all n N;
(3) monotonic sequence if it is either monotonically increasing or monotonically decreasing.
If strict inequality occur in (1) (resp. (2)), then we say that the sequence is strictly increasing
(resp. strictly decreasing).
A monotonically increasing (respectively, a monotonically decreasing) sequence is also called an
increasing (respectively, a decreasing) sequence.
We may observe that:
1. A sequence (a
n
) is monotonically increasing and bounded above if and only if (a
n
) is mono-
tonically decreasing and bounded below.
2. Every monotonically increasing sequence is bounded below, and every monotonically decreasing
sequence is bounded above.
Example 1.48. The following statements can be easily veried (verify):
(i) (a
n
) with a
n
= n/(n + 1) is monotonically increasing.
(ii) (a
n
) with a
n
= (n + 1)/n is monotonically decreasing.
(iii) (a
n
) with a
n
= (1)
n
n/(n+1) is neither monotonically increasing nor monotonically decreas-
ing.
Exercise 1.49. Let (a
n
) be a monotonic sequence. Prove the following.
16
(i) Suppose (a
n
) converges. Then it is bounded above by the limit if it is increasing, and bounded
below by the limit if it is decreasing.
(ii) Suppose (a
n
) diverges. Then it diverges to either innity or minus innity depending on
whether it is increasing or decreasing.
A convergent sequence need not be monotonically increasing or monotonically decreasing.
However, we have the following theorem.
THEOREM 1.50. (Monotone convergence theorem) Every bounded monotonic sequence is
convergent.
Exercise 1.51. (i) A monotonically increasing sequence either converges or diverges to .
(ii) A monotonically decreasing sequence either converges or diverges to .
In the following examples, Theorem 1.50 has been used for showing convergence of certain se-
quences.
Example 1.52. We have already seen that if 0 < a < 1, then a
n
0 as n (See Theorem 1.26).
This can also be seen by making use of Theorem 1.50, as follows: Let x
n
= a
n
. Then
0 x
n+1
= a
n+1
= ax
n
x
n
n N.
Thus, (x
n
) is monotonically decreasing and bounded below. Hence (x
n
) converges to some x R.
Then x
n+1
= ax
n
ax. Therefore, x = ax so that x = 0.
Example 1.53. For n N, let
a
n
= 1 +
1
2
2
+
1
3
2
+ +
1
n
2
.
Clearly, (a
n
) is monotonically increasing. Now, we show that it is bounded above, so that by Theorem
1.50, it is convergent: We note that
1
2
2
+
1
3
2

2
2
2
=
1
2
,
1
4
2
+
1
5
2
+
1
6
2
+
1
7
2

4
4
2
=
1
4
,
and more generally,
1
(2
n1
)
2
+
1
(2
n1
+ 1)
2
+ +
1
(2
n
1)
2

2
n1
(2
n1
)
2
=
1
2
n1
.
Hence,
a
2
n = 1 +
1
2
2
+
1
3
2
+ +
1
(2
n
)
2
1 +
1
2
+
1
2
2
+ +
1
2
n1
+
1
2
2n
= 2
_
1
1
2
n
_
+
1
2
2n
2.
17
Since a
n
a
2
n for all n N, we obtain that (a
n
) is monotonically increasing and also bounded above,
and hence, by Theorem 1.50, it converges. We shall see in the nal chapter of this book that the
number to which this (a
n
) converges is
2
/6.
Example 1.54. Given a sequence (a
n
) with a
n
{0, 1, 2, 3, 4, 5, 6, 7, 8, 9}, consider the sequence (b
n
)
with
b
n
=
a
1
10
+
a
2
10
2
+ +
a
n
10
n
for n N. Does (b
n
) converge? By our knowledge in decimal expansion of a number, we can believe
that (b
n
) would converge to a number in the interval [0, 1]. Let us prove this:
Note that b
n
b
n+1
for all n N. Therefore, to assert the convergence of (b
n
), by Theorem 1.50,
it is enough to show that (b
n
) is bounded above. For this end, we observe that
b
n
=
a
1
10
+
a
2
10
2
+ +
a
n
10
n

9
10
+
9
10
2
+ +
9
10
n
=
9
10
_
1
1
10
n
1
1
10
_
= 1
1
10
n
.
Thus, (b
n
) is bounded above by 1. Therefore, (b
n
) converges to a number in [0, 1].
Example 1.55. (The number e) Consider the sequences (a
n
) and (b
n
) dened by
a
n
=
_
1 +
1
n
_
n
, b
n
= 1 +
1
1!
+
1
2!
+
1
3!
+ +
1
n!
.
We show that both (a
n
) and (b
n
) are monotonically increasing and bounded above. Hence, by Theorem
1.50, they converge. Further, we shall show that they have the same limit. For this we show that
(i) b
n
b
n+1
3 for all n N,
(ii) a
n
b
n
for all n N,
(iii) a
n
a
n+1
3 for all n N.
Clearly, b
n
b
n+1
for all n N. Also,
b
n
1 + 1 +
1
2
+
1
2
2
+ +
1
2
n1
< 3.
Thus, (i) is proved. Next,
a
n
= 1 + 1 +
1
2!
n(n 1)
n
2
+ +
1
n!
n(n 1) . . . 2.1
n
n
1 + 1 +
1
2!
+
1
3!
+ +
1
n!
= b
n
18
and
a
n
=
_
1 +
1
n
_
n
= 1 + n.
1
n
+
n(n 1)
2!
1
n
2
+ +
n(n 1) . . . 2.1
n!
1
n
n
= 1 + 1 +
1
2!
_
1
1
n
_
+
1
3!
_
1
1
n
__
1
2
n
_
+
+
1
n!
_
1
1
n
_

_
1
n 1
n
_
a
n+1
.
Thus, the proofs of (i)-(iii) are over. From, (i)-(iii), we see that both (a
n
) and (b
n
) are monotonically
increasing bounded above. Hence, by Theorem 1.50, both (a
n
) and (b
n
) converge. Let a and b their
limits. We show that a = b.
We have already observed that a
n
b
n
. Hence, taking limits, we obtain a b. Notice that
a
n
= 1 + 1 +
1
2!
_
1
1
n
_
+
1
3!
_
1
1
n
__
1
2
n
_
+
. . . +
1
n!
_
1
1
n
_
. . .
_
1
n 1
n
_
.
Hence, for m, n with m n, we have
a
n
1 + 1 +
1
2!
_
1
1
n
_
+
1
3!
_
1
1
n
__
1
2
n
_
+
+
1
m!
_
1
1
n
_
. . .
_
1
m1
n
_
.
Taking limit as n , we get (cf. Theorem 1.25 (c))
a 1 +
1
1!
+
1
2!
+
1
3!
+ +
1
m!
= b
m
.
Now, taking limit as m , we get a b. Thus we have proved a = b.
The common limit of the two sequence (a
n
) and (b
n
) above is denoted by the letter e, after the
great mathematician Euler
3
.
Further examples are given in Subsection 1.5.
1.4 Subsequences
Denition 1.56. A subsequence of a sequence (a
n
) is a sequence of the form (a
k
n
), where (k
n
) is
a strictly increasing sequence of positive integers.
3
Leonhard Euler (15 April 1707 18 September 1783) was a Swiss mathematician and physicist. He made important
discoveries in various elds in mathematics. He introduced many modern mathematical terminology and notation,
including the notion of a mathematical function (Curtsey to Wikipedia).
19
A sequence (b
n
) is a subsequence of a sequence (a
n
) if and only if there is a strictly
increasing sequence (k
n
) of positive integers such that b
n
= a
k
n
for all n N.
For example, given a sequence (a
n
), the sequences
(a
2n
), (a
2n+1
), (a
n
2), (a
2
n)
are some of the subsequences of (a
n
). As concrete examples,
(1) (
1
2n
), (
1
2n+1
), (
1
n
2
) and (
1
2
n
) are subsequences of (
1
n
);
(2) (
1
n
) and (
n
n+1
) are subsequences of (1,
1
2
,
1
2
,
2
3
,
1
3
,
3
4
, . . .).
(3) (2n), (2n + 1), (n
2
) and (2
n
) are subsequences of the sequence (a
n
) with a
n
= n for all n N.
THEOREM 1.57. If a sequence (a
n
) converges to a, then all its subsequences converge to the same
limit a.
Proof. Suppose a
n
a. Consider a subsequence (a
k
n
) of (a
n
). Let > 0 be given. Since a
n
a,
there exists N N such that
|a
n
a| < n N.
In particular, since k
N
N,
|a
n
a| < n {k
N
, k
N+1
, . . .}.
Thus,
|a
k
n
a| < n N.
Hence, a
k
n
a.
What about the converse of the above theorem? Obviously, if all subsequences of (a
n
) converge,
then (a
n
) also has to converge, since (a
n
) is a subsequence of itself. Thus, we have proved:
(a
n
) converges to a i every subsequence of (a
n
) converges to a.
We know that a divergent sequence can have convergent subsequences. But, this cannot happen
for monotonic sequences.
If (a
n
) is a monotonic sequence having at least one convergent
subsequence, say with limit a, then (a
n
) itself converge to a.
20
Exercise 1.58. Prove the above statement.
We have seen in Theorem 1.40 that every convergent sequence is bounded, but a bounded se-
quence need not be convergent. However, we have the following theorem, which is called the Bolzano-
Weierstrass theorem.
THEOREM 1.59. (Bolzano-Weierstrass theorem
4
). Every bounded sequence of real numbers
has at least one convergent subsequence.
1.5 Further examples
Example 1.60. Let a sequence (a
n
) be dened iteratively as follows: a
1
= 1 and
a
n+1
=
2a
n
+ 3
4
n N. We show that (a
n
) is monotonically increasing and bounded above.
Note that
a
n+1
=
2a
n
+ 3
4
=
a
n
2
+
3
4
a
n
a
n

3
2
.
Thus it is enough to show that a
n
3/2 for all n N.
Clearly, a
1
3/2. If a
n
3/2, then a
n+1
= a
n
/2 +3/4 < 3/4 +3/4 = 3/2. Thus, we have proved
that a
n
3/2 for all n N. Hence, by Theorem 1.50, (a
n
) converges. Let its limit be a. Then taking
limit on both sides of a
n+1
=
2a
n
+3
4
we have
a =
2a + 3
4
i.e., 4a = 2a + 3 so that a =
3
2
.
Another solution: Since a := 3/2 satises
a =
2a + 3
4
,
we obtain
a
n+1
a =
2a
n
+ 3
4

2a + 3
4
=
1
2
(a
n
a).
Thus, by Theorem 1.35, a
n
a = 3/2.
Example 1.61. Given an arbitrary positive real number a
0
, let a sequence (a
n
) be dened as follows
: a
1
= 2 and
a
n+1
=
1
2
_
a
n
+
2
a
n
_
for n N. Note that, if the sequence converges, then its limit a 0, and then
a =
1
2
_
a +
2
a
_
4
Bernard Bolzano (October 5, 1781 - December 18, 1848), was a Bohemian mathematician, logician, philosopher,
theologian and Catholic priest of Italian Origins, and Karl Weierstrass (October 31, 1815 February 19, 1897) was a
German mathematician who is often cited as the father of modern analysis.
21
so that a =

2.
Since a
1
= 2, one may try to show that (a
n
) is monotonically decreasing and bounded below.
Note that
a
n+1
:=
1
2
_
a
n
+
2
a
n
_
a
n
a
2
n
2, i.e., a
n

2,
and
a
n+1
:=
1
2
_
a
n
+
2
a
n
_

2 a
2
n
2

2a
n
+ 2 0
(a
n

2)
2
0.
Hence, a
n+1

2 for all n N so that a

n+1
a
n
for all n 2. Hence, (a
n
) is monotonically
decreasing and bounded below, so that by Theorem 1.50, (a
n
) converges and hence its limit is

2.
Example 1.62. The sequence (n
1/n
) converges and the limit is 1:
For each n N, since n
1/n
1, there exists r
n
0 such that
n
1/n
= 1 + r
n
.
Then we have
n = (1 + r
n
)
n

n(n 1)
2
r
2
n
,
so that
r
2
n

2
n 1
n 2.
Since 2/(n 1) 0, by Theorem 1.25(c), r
n
0. Hence n
1/n
1.
Example 1.63. For any a > 0, (a
1/n
) converges to 1:
If a > 1, then we can write a
1/n
= 1 + r
n
for some sequence (r
n
) of positive reals. Then we have
a = (1 + r
n
)
n
nr
n
so that r
n
a/n.
Since a/n 0, by Theorem 1.25(c), r
n
0 and a
1/n
= 1 + r
n
1.
In case 0 < a < 1, then 1/a > 1. Hence, by the rst part,
1/a
1/n
= (1/a)
1/n
1,
so that by Theorem 1.44 (ii), a
n
1.
Example 1.64. Let (a
n
) be a bounded sequence of non-negative real numbers. Then (1+a
n
)
1/n
1
as n :
Let M > 0 be such that 0 a
n
M for all n N. Then,
1 (1 + a
n
)
1/n
(1 + M)
1/n
n N.
Using the result in Example 1.63, (1 + M)
1/n
1. Hence, by the Sandwich theorem (Theorem 1.25
(iv)), (1 + a
n
)
1/n
1.
22
Example 1.65. Let (a
n
) be a sequence of nonnegative terms such that 1 a
n
n for all n N.
Then (1 + a
n
)
1/n
1: Note that
1 (1 + a
n
)
1/n
(1 + n)
1/n
(2n)
1/n
= 2
1/n
n
1/n
n N.
By the results in Examples 1.62 and 1.63, 2
1/n
1 and n
1/n
1. Hence, by the Sandwich theorem
(Theorem 1.25 (iv)), (1 + a
n
)
1/n
1.
Example 1.66. Let a
n
= (n!)
1/n
2
, n N. Then a
n
1 as n . We give two proofs for this.
(i) Note that, for every n N,
1 (n!)
1/n
2
(n
n
)
1/n
2
= n
1/n
.
Since n
1/n
1, by the Sandwich theorem (Theorem 1.25 (iv)), we have (n!)
1/n
2
1.
(ii) By GM-AM inequality, for n N,
(n!)
1/n
= (1.2. . . . .n)
1/n

1 + 2 + . . . + n
n
=
n + 1
2
n.
Thus, 1 (n!)
1/n
2
n
1/n
. Since n
1/n
1, by the Sandwich theorem (Theorem 1.25 (iv)), we have
(n!)
1/n
2
1.
What about the convergence of the sequence (a
n
) with a
n
= (n!)
1/n
?
Example 1.67. Let a
n
= (n!)
1/n
, n N. We show that a
n
1. In fact, we show that (a
n
) is
unbounded.
Note that, for any k, n N with n k,
(n!)
1/n
(k!)
1/n
(k
nk
)
1/n
= (k!)
1/n
k
1k/n
= k
_
k!
k
k
_
1/n
.
Since for any xed k, k
_
k!
k
k
_
1/n
k as n , we can conclude that (n!)
1/n
1. Now, for k N,
let n
k
N be such that n
k
k and
_
k!
k
k
_
1/n

1
2
n n
k
.
Thus,
(n!)
1/n

k
2
n n
k
.
Therefore, the sequence
_
(n!)
1/n
_
is unbounded. In fact (n!)
1/n
as n .
Remark 1.68. Suppose for each k N, a
(k)
n
0, b
(k)
n
1 as n , and also a
(n)
n
0, b
(n)
n
1
as n . In view of Theorems 1.25 and 1.44, one may think that
a
(1)
n
+ a
(2)
n
+ + a
(n)
n
0 as n
23
and
b
(1)
n
b
(2)
n
b
(n)
n
1 as n .
Unfortunately, that is not the case. To see this consider
a
(k)
n
=
k
n
2
, b
(k)
n
= k
1/n
k, n N.
Then, for each k N, we have
a
(k)
n
0, b
(k)
n
1 as n .
Also
a
(n)
n
0, b
(n)
n
1 as n .
But,
a
(1)
n
+ a
(2)
n
+ + a
(n)
n
=
1
n
2
+
2
n
2
+ +
n
n
2
=
n + 1
2n

1
2
as n
and from Example 1.67,
b
(1)
n
b
(2)
n
b
(n)
n
= 1
1/n
2
1/n
n
1/n
= (n!)
1/n
1 as n .
1.6 Cauchy criterion
Let us observe the following necessary condition for convergent sequences.
THEOREM 1.69. If (a
n
) converges, then for every > 0, there exists N N such that
|a
n
a
m
| < n, m N.
Proof. Suppose a
n
a as n , and let > 0 be given. Then we know that there exists N N
such that |a
n
a| < /2 for all n N. Hence, we have
|a
n
a
m
| |a
n
a| +|a a
m
| < n, m N.
This completes the proof.
Is the condition in Theorem 1.69 sucient as well? The answer is in the armative, which is the
so called Cauchy criterion
5
.
Denition 1.70. A a sequence (a
n
) is said to be a Cauchy sequence if for every > 0, there exists
N N such that
|a
n
a
m
| < n, m N.

5
Augustin-Louis Cauchy (21 August 1789 23 May 1857) was a French mathematician who made many contributions
to calculus, specically in terms of its rigorous foundation.
24
THEOREM 1.71. (Cauchy criterion) Every Cauchy sequence of real numbers converges.
Exercise 1.72. Let s
n
= 1 +
1
2
+ . . . +
1
n
for n N. Show that |s
2n
s
n
|
1
2
. Conclude that (s
n
) is
a divergent sequence.
Exercise 1.73. Let (a
n
) be a sequence of real numbers. Suppose there exists a positive real number
r < 1 such that
|a
n+2
a
n+1
| r|a
n+1
a
n
| n N.
Show that (a
n
) is a Cauchy sequence. If a = lim
n
a
n
, then show that
|a a
n
|
r
n1
1 r
|a
2
a
1
| n N.

2 Series of Real Numbers

Denition 2.1. A series of real numbers is an expression of the form
a
1
+ a
2
+ a
3
+ . . . ,
or more compactly

n=1
a
n
,
where (a
n
) is a sequence of real numbers. The number a
n
is called the n
th
term of the series and the
sum of the rst n terms of (a
n
), that is,
s
n
:= a
1
+ + a
n
,
is called the n
th
partial sum of the series.
Remark 2.2. Some authors dene a series as the sequence (s
n
), where s
n
is n
th
partial sum of
another sequence (a
n
).
Example 2.3. Following are some examples of series:
(i)
3
10
+
3
10
2
+ ,
(ii) 1 +
1
2
+
1
3
+ ,
(iii) 1 +
1
2
2
+
1
3
2
+
Note that in (i), (ii), (iii) above the partial sums are
n

k=1
3
10
k
,
n

k=1
1
k
,
n

k=1
1
k
2
,
respectively.
25
2.1 Convergence and divergence of series
Denition 2.4. A series

n=1
a
n
is said to be a convergent series if the corresponding sequence
(s
n
) of partial sums converge. If s
n
s, then we say that the series

n=1
a
n
converges to s, and
we write this fact as
s =

n=1
a
n
.
A series which does not converge is called a divergent series.
Observe the following:
Suppose a
n
0 for all n N. Then the sequence (s
n
) of partial sums of the series

n=1
a
n
is monotonically increasing. Hence, in this case, either (s
n
) converges or s
n
.
Exercise 2.5. Show that the geometric series
1 + q + q
2
+
for q R converges if and only if |q| < 1.
THEOREM 2.6. (A necessary condition) If

n=1
a
n
converges, then a
n
0 as n . The
converse is not true.
Proof. Suppose the series

n=1
a
n
converges to s, that is, s
n
s, where s
n
is the n
th
partial sum of
the series. Since a
n
= s
n
s
n1
for n = 2, 3, . . ., we have
a
n
= s
n
s
n1
0 as n .
Thus, a
n
0.
To see that the converse does not hold, consider the series

n=1
1/n. In this case, a
n
= 1/n 0,
but the series diverges.
Exercise 2.7. Prove the following.
(i) The series

n=1
n
n+1
diverges.
(ii) If a
n+1
> a
n
> 0 for all n N, then

n=1
a
n
diverges.
The following theorem shows that if we remove the rst few terms from a convergent (respectively,
divergent) series, then the resulting series remain convergent (respectively, divergent).
THEOREM 2.8. For each k N, the series

n=1
a
n
converges if and only if the series

n=1
a
k+n
converges, and in that case

n=1
a
n
= (a
1
+ + a
k
) +

n=1
a
k+n
.
26
The above theorem shows that if

n=1
b
n
is obtained from

n=1
a
n
by omitting or adding a nite
number of terms, then

n=1
a
n
converges

n=1
b
n
converges.
The following theorem is an immediate corollary to Theorem 2.8.
THEOREM 2.9. Suppose (a
n
) and (b
n
) are sequences such that for some k N, a
n
= b
n
for all
n k. Then

n=1
a
n
converges if and only if

n=1
b
n
converges.
THEOREM 2.10. Suppose

n=1
a
n
converges to s and

n=1
b
n
converges to s

. Then for any

, R,

n=1
(a
n
+ b
n
) converges to s + s

.
In particular,

n=1
(a
n
+ b
n
) converges to s + s

and

n=1
c a
n
converges to cs for any c R.
Proof. Let s
(1)
n
, s
(2)
n
and s
n
be the n
th
partial sums of the series

n=1
a
n
,

n=1
b
n
and

n=1
(a
n
+
b
n
) respectively. Then we obtain
s
n
=
n

k=1
(a
k
+ b
k
) =
n

k=1
a
k
+
n

k=1
b
k
= s
(1)
n
+ s
(2)
n
n N.
Since s
(1)
n
s and s
(2)
n
s

, we obtain s
n
= s
(1)
n
+ s
(2)
n
s + s

.
Exercise 2.11. As proof of Theorem 2.10, suppose we write

n=1
(a
n
+ b
n
) =

n=1
a
n
+

n=1
b
n
= s + s

.
What is wrong with it?
2.2 Some tests for convergence
THEOREM 2.12. (Comparison test) Let 0 a
n
b
n
for all n N.
(i) If

n=1
b
n
converges, then

n=1
a
n
converges.
(ii) If

n=1
a
n
diverges, then

n=1
b
n
diverges.
Proof. Suppose s
n
and s

n
are the n
th
partial sums of the series

n=1
a
n
and

n=1
b
n
respectively.
By the assumption, we have 0 s
n
s

n
for all n N, and both (s
n
) and (s

n
) are monotonically
increasing.
27
(i) Suppose

n=1
b
n
converges, that is, (s

n
) converges. Then, (s

n
) is bounded. Hence, by the
relation 0 s
n
s

n
for all n N, (s
n
) is bounded as well as monotonically increasing. Therefore, by
Theorem 1.50, (s
n
) converges.
(ii) This part is a restatement of (i).
Exercise 2.13. Suppose (a
n
) and (b
n
) are sequences of positive terms. Prove the following.
(i) Suppose := lim
n
a
n
b
n
exists.
(a) If > 0, then

n=1
b
n
converges

n=1
a
n
converges.
(b) If = 0, then

n=1
b
n
converges =

n=1
a
n
converges.
(ii) Suppose lim
n
a
n
b
n
= . Then

n=1
a
n
converges =

n=1
b
n
converges.

Example 2.14. We have seen that the sequence (s

n
) with s
n
=

n
k=1
1
k!
converges. This also follows
from comparison test, since
1
n!

1
2
n1
n N
and

n=1
1
2
n1
converges.
Example 2.15. (i) Since
1

n

1
n
for all n N, and since the series

n=1
1
n
diverges, by comparison
test, the series

n=1
1

n
also diverges. More generally, let p 1. Since
1
n
p

1
n
n N,
by comparison test,

n=1
1
n
p
diverges for p 1.
(ii) Let p 2. We know that

k=1
1
n
2
converges (Example 1.53). Since
1
n
p

1
n
2
for all n N,
by comparison test, the series

n=1
1
n
p
converges for p 2.
What about the convergence of

n=1
1
n
p
for any p > 1? It is dealt in the next example.
Example 2.16. In this example, we shall use the familiarity of the integral to the extent that the
reader knows the result:
_
b
a
x
k
dx =
_
x
k+1
k + 1
_
b
a
=
b
k+1
a
k+1
k + 1
for k = 1.
28
Consider the series

n=1
1
n
p
for p > 1. To discuss this general case, consider the function
f(x) := 1/x
p
, x 1.
Note that, for each k N, we have
k 1 x k =
1
k
p

1
x
p
=
1
k
p

_
k
k1
dx
x
p
.
Hence,
n

k=2
1
k
p

n

k=2
_
k
k1
dx
x
p
=
_
n
1
dx
x
p
=
n
1p
1
1 p

1
p 1
.
Thus,
s
n
:=
n

k=1
1
k
p

1
p 1
+ 1 =
p
p 1
.
Hence, (s
n
) is monotonically increasing and bounded above, and hence (s
n
) converges. Thus,

n=1
1
n
p
converges for p > 1.
A more general result on convergence of series in terms of integrals, known as integral test, will be
considered later.
Another consequence of the comparison test is the following.
THEOREM 2.17. Let (a
n
) be a sequence of positive numbers. If there exists r with 0 < r < 1 and
a positive integer N such that either
a
n+1
a
n
r n N or a
1/n
n
r n N,
then the series

n=1
a
n
converges.
Proof. Let 0 < r < 1. First, let N N be such that
a
n+1
a
n
r n N.
Then a
n+1
ra
n
for all n N so that
a
n+1
r
nN+1
a
N
=
_
a
N
r
N1
_
r
n
n N.
Since

n=1
r
n
converges, the comparison test shows that, the series

n=1
a
n
converges.
Next, let N N be such that
a
1/n
n
r n N.
Then a
n
r
n
for all n N. Again, since

n=1
r
n
converges, by comparison test,

n=1
a
n
converges.
29
Sometimes it may be easier to nd the limit of either
_
a
n+1
a
n
_
or (a
1/n
n
) rather than nding an
upper bound. In such cases, the following two tests due to dAlembert
6
and Cauchy, respectively, are
useful to decide the convergence of a series.
THEOREM 2.18. (DAlemberts ratio test) Suppose (a
n
) is a sequence of positive terms such that
lim
n
a
n+1
a
n
= exists.
(i) If < 1, then the series

n=1
a
n
converges.
(ii) If > 1, then a
n
0; in particular,

n=1
a
n
diverges.
Proof. (i) Suppose < 1. Let r be such that < r < 1. Then there exists N N such that
a
n+1
a
n
r n N.
Hence, by Theorem 2.17, the series

n=1
a
n
converges.
(ii) Let > 1. Then there exists N N such that
a
n+1
a
n
> 1 n N.
Hence, a
n+1
> a
n
for all n N. Therefore, a
n
0.
THEOREM 2.19. (Cauchys root test) Suppose (a
n
) is a sequence of positive terms such that
lim
n
a
n
1/n
= exists.
(i) If < 1, then the series

n=1
a
n
converges.
(ii) If > 1, then a
n
0; in particular,

n=1
a
n
diverges.
Proof. (i) Suppose < 1. Let r be such that < r < 1. Then there exists N N such that
a
n
1/n
r n N.
Hence, by Theorem 2.17, the series

n=1
a
n
converges.
(ii) Let > 1. Then there exists N N such that
a
n
1/n
> 1 n N.
Hence, a
n
1 for all n N. Therefore, a
n
0.
6
Jean-Baptiste le Rond dAlembert (16 November 1717 29 October 1783) was a French mathematician, physicist
and philosopher. A particular method of solution of wave equation is named after him - curtsey Wikipedia.
30
Remark 2.20. We remark that both dAlemberts ratio test and Cauchys root test are silent for the
case = 1. In this case, we cannot assert either way. For example, we know that the series

n=1
1
n
diverges whereas

n=1
1
n
2
, and in both the cases, we have
a
n+1
a
n
1, a
1/n
n
1.
However, for such cases, we may be able to infer the convergence or divergence by some other means.

Example 2.21. Let us test the convergence of the series

n=1
n
2
2
n
and

n=1
n!
2
n
.
(i)

n=1
n
2
2
n
: In this case a
n
=
n
2
2
n
so that
a
n+1
a
n
=
(n + 1)
2
/2
n+1
n
2
/2
n
=
1
2
_
n + 1
n
_
2

1
2
.
Hence, by dAlemberts ratio test, the series converges.
(ii)

n=1
n!
2
n
: In this case a
n
=
n!
2
n
so that
a
n+1
a
n
=
(n + 1)!/2
n+1
n!/2
n
=
n + 1
2
.
Hence, by dAlemberts ratio test, the series diverges.
Example 2.22. For x 0, the series

n=1
x
n
n!
converges:
Clearly the series converges if x = 0. For x > 00, let a
n
=
x
n
n!
. Then we have
a
n+1
a
n
=
x
n + 1
0.
Hence, by dAlemberts ratio test, the series converges. We shall also show that the series

n=1
x
n
n!
converges for any x R.
Example 2.23. The series

n=1
_
n
2n + 1
_
n
converges: In this case, we have
a
n
1/n
=
n
2n + 1

1
2
< 1.
31
Hence, by Cauchys root test, the series converges.
The convergence of the above series can also be proved using comparison test, since
a
n
=
_
n
2n + 1
_
n
=
_
1
2 + 1/n
_
n

1
2
n
for all n N and

n=1
1
2
n
converges.
Example 2.24. Consider the series

n=1
1
n(n + 1)
. In this case, we have
lim
n
a
n+1
a
n
= 1 = lim
n
a
n
1/n
.
Hence, we are not in a position to apply ratio test and root test. However,
s
n
:=
n

k=1
1
k(k + 1)
=
n

k=1
_
1
k

1
k + 1
_
= 1
1
n + 1
1.
Thus, the series converges to 1.
Exercise 2.25. Assert the convergence of the series

n=1
a
n
in Example 2.21(i) and Example 2.23
by showing that
a
n+1
a
n

8
9
n 3 and a
1/n
n

1
2
n N,
respectively.
Exercise 2.26. Assert the divergence of the series

n=1
a
n
in Example 2.21(ii) by showing either
a
n+1
a
n
1 n N or a
n
.
Exercise 2.27. Show the convergence of the series

n=1
1
n
2
by comparing it with

n=1
1
n(n+1)
.
Example 2.28. Consider the series

n=1
a
n
with
a
2n1
=
1
6
n1
, a
2n
=
1
3 6
n1
for n N. Note that
a
2n
a
2n1
=
1
3
,
a
2n+1
a
2n
=
1
2
for all n N so that lim
n
a
n+1
a
n
does not exist. However,
a
n+1
a
n
1/2 for all n N so that by Theorem
2.17(i), the series

n=1
a
n
converges.
32
Example 2.29. Consider the series

n=1
a
n
with
a
n
=
_

_
_
n
2n + 1
_
n
, n odd,
1
3
n
, n even.
Note that
lim
n
a
1/(2n+1)
2n+1
= lim
n
2n + 1
4n + 3
=
1
2
, lim
n
a
1/2n
2n
=
1
3
so that lim
n
a
1/n
n
does not exist. However, a
1/n
n
1/2 for all n N. Hence, by Theorem 2.17(i), the
series

n=1
a
n
converges.
2.3 Alternating series
In the last subsection we have described some tests for asserting the convergence or divergence of
series of non-negative terms. In this subsection we provide a sucient condition for convergence of
those series with alternatively positive and negative terms.
Denition 2.30. A series of the form

n=1
(1)
n+1
u
n
, where (u
n
) is a sequence of positive terms,
is called an alternating series.
Thus,
1
1
2
+
1
3

1
4
+
1
5
+ +
(1)
n+1
n
+
is an alternating series, whereas the series
1 +
1
2

1
3

1
4
+ +
1
4n 3
+
1
4n 2

1
4n 1

1
4n
+
is not an alternating series. As you can see, the above series, though not an alternating series, is of
the form

n=1
[(1)
n+1
u
n
+ (1)
n+1
v
n
]
with
u
n
=
1
2n 1
, v
n
=
1
2n
.
Note that

n=1
(1)
n+1
u
n
and

n=1
(1)
n+1
v
n
are alternating series. Thus, once we know the
convergence of both the series

n=1
(1)
n+1
u
n
and

n=1
(1)
n+1
v
n
, we can assert the convergence
of the original series

n=1
[(1)
n+1
u
n
+ (1)
n+1
v
n
].
In view of the above discussion, it is important to have a sucient condition for the convergence
of alternating series. Following theorem due to Leibnitz
7
provides such a condition.
7
Gottfried Wilhelm von Leibnitz (July 1, 1646 November 14, 1716) was a German mathematician and philosopher.
He developed the innitesimal calculus independently of Isaac Newton - curtsey Wikipedia.
33
THEOREM 2.31. (Leibnitzs theorem) Suppose (u
n
) is a sequence of positive terms such that
u
n
u
n+1
for all n N and u
n
0 as n . Then the alternating series

n=1
(1)
n+1
u
n
converges, and in that case,
|s s
n
| u
n+1
n N,
where
s
n
=
n

j=1
(1)
j+1
u
j
and s =

n=1
(1)
n+1
u
n
,
Proof. We observe that
s
2n+1
= s
2n
+ u
2n+1
n N.
Since u
n
0 as n , it is enough to show that (s
2n
) converges, and in that case, both (s
2n
) and
(s
2n+1
) converge to the same limit. Note that, for every n N,
s
2n+2
= s
2n
+ (u
2n+1
u
2n+2
) s
2n
and
s
2n
= u
1
(u
2
u
3
) . . . (u
2n2
u
2n1
) u
2n
u
1
.
Hence, (s
2n
) is monotonically increasing and bounded above. Therefore (s
2n
) converges.
To obtain the remaining part, note also that, for n N,
s
2n+1
= s
2n1
(u
2n
u
2n+1
) s
2n1
,
so that {s
2n1
} is a monotonically decreasing sequence. Thus,
s
2n1
= s
2n
+ u
2n
s + u
2n
, s s
2n+1
= s
2n
+ u
2n+1
,
s s
2n1
, s
2n
s.
Hence,
s
2n1
s u
2n
, s s
2n
u
2n+1
,
Consequently,
|s s
n
| u
n+1
n N.
This completes the proof.
COROLLARY 2.32. Suppose (a
n
) is a sequence of positive terms such that a
n
a
n+1
for all n N
and a
n
0 as n . Then the series
a
1
+ a
2
a
3
a
4
+ + a
4n3
+ a
4n2
a
4n1
a
4n
+
converges.
34
Proof. Write the given series as

n=1
[(1)
n+1
u
n
+ (1)
n+1
v
n
],
where (u
n
) and (v
n
) are decreasing sequences of positive terms such that u
n
0 and v
n
0, and
then apply Theorem 2.31.
Example 2.33. By Theorem 2.31,
1
1
2
+
1
3

1
4
+ +
(1)
n+1
n
+ ,
1
1
3
+
1
5

1
7
+ +
(1)
n+1
2n 1
+ ,
1
2

1
4
+
1
6

1
8
+ +
(1)
n+1
2n
+ ,
are convergent series. Using Theorem 2.31, it can also be shown that the series
1 +
1
2

1
3

1
4
+ +
(1)
n+1
2n 1
+
(1)
n+1
2n
+
is also a convergent series.
Remark 2.34. The convergence of the series
1
1
3
+
1
5

1
7
+ +
(1)
n+1
2n 1
+ ,
which is generally known as Leibnitz-Gregory series, was known to Indian mathematicians Madhava
and Nilaka ntha
8
as early as in 15-th century. In fact, it was known that the above series converges to

4
. We shall prove this in a later section.
Remark 2.35. The relation |s s
n
| u
n+1
in Theorem 2.31 shows the rate of convergence of the
partial sums to the sum of the series. In particular, for a given k N, if n is large enough such that
u
n
< 1/10
k
, then rst k decimal places of s
n
and s are the same.
The following example shows that the condition u
n
u
n+1
for all n N in Theorem 2.31 is not
redundant, that is, the conclusion in the theorem need not hold if the above condition is dropped.
Example 2.36. Consider the series
1

3 1

3 + 1
+
1

4 1

4 + 1
+ +
1

n 1

1

n + 1
+
8
The series

n=1
(1)
n+1
2n1
appear in the work of a Kerala mathematician Madhava around 1425 which was presented
later in the year around 1550 by another Kerala mathematician Nilaka ntha. The discovery of the above series is normally
attributed to Leibnitz and James Gregory after nearly 300 years of its discovery.
35
which is of the form

n=1
(1)
n+1
u
n
with
u
2n1
=
1

n + 2 1
, u
2n
=
1

n + 2 + 1
.
Note that, in this case, we have u
n
0 for all n N and u
n
0. However, the series diverges. To
see this, we observe that
s
2n
=
n

k=1
_
1

n + 2 1

n + 2 + 1
_
=
n

k=1
2
n + 1
.
Thus, (s
n
) has a subsequence which diverges. Therefore, (s
n
) diverges. Note that (u
n
) is not a
decreasing sequence.
2.4 Absolute convergence
We have seen in Example 1.55 that the series

n=1
|x|
n
n!
converges for every x R. What about the
convergence of the series

n=1
x
n
n!
?
Denition 2.37. Let (a
n
) be a sequence of real numbers. Then the series

n=1
a
n
is said to be
(1) absolutely convergent, if

n=1
|a
n
| is convergent,
(2) conditionally convergent, if it converges, but not absolutely.

Example 2.38. (i) the series

n=1
(1)
n
n
2
,

n=1
(1)
n+1
n!
,

n=1
sin n
n
2
are absolutely convergent, so also the series

n=1
a
n
n!
for any a R.
(ii) The series

n=1
(1)
n
n
,

n=1
(1)
n
2n 1
,

n=1
(1)
n
2n
are not absolutely convergent. They are convergent, by Leibnitz theorem. Thus, these series are
conditionally convergent.
THEOREM 2.39. Every absolutely convergent series is convergent. The converse does not hold.
36
Proof. Suppose

n=1
a
n
is an absolutely convergent series. Let s
n
and s

n
be the n
th
partial sums of
the series

n=1
a
n
and

n=1
|a
n
| respectively. Then, for n > m, we have
|s
n
s
m
| =

j=m+1
a
n

j=m+1
|a
n
| = |s

n
s

m
|.
Since, (s

n
) converges, it is a Cauchy sequence. Hence, from the above relation it follows that (s
n
) is
also a Cauchy sequence. Therefore, by the Cauchy criterion, it converges.
Recall that the series

n=1
1
n
does not converge. However, by Theorem 2.31,

n=1
(1)
n+1
n
is
convergent. Thus, the converse of the rst statement in the theorem does not hold.
In view of Theorem 2.39, the series

n=1
x
n
n!
converges for each x R. We shall prove in the next
chapter that the value of the above sum is same as
lim
n
_
1 +
x
n
_
n
,
and it is generally denoted by exp(x) or e
x
.
In view of Theorem 2.39 we can assert the following.
Let (a
n
) be a sequence of nonzero numbers
(i) If lim
n

a
n+1
a
n

= exists and < 1,

then

n=1
a
n
x
n
converges absolutely.
(ii) If lim
n
|a
n
|
1/n
= exists and < 1,
then

n=1
a
n
x
n
converges absolutely.
Given a convergent series

n=1
a
n
of real numbers, consider another series

n=1
b
n
obtained by
rearrangement of the terms of

n=1
a
n
, that is, b
n
= a
(n)
, n N, where is a bijection on N. A
natural question one would like to ask is:
Does

n=1
b
n
converge to the same limit as that of

n=1
a
n
?
Let us look at the following example.
Example 2.40. Consider the convergent series

n=1
(1)
n+1
n
. At the moment we do not know to what
number it converges. However, let us see if a rearrangement of it converges to a dierent number. For
this, let us consider the following rearrangement:
1
1
2

1
4
+
1
3

1
6

1
8
+ +
1
2n 1

1
4n 2

1
4n
+ .
37
Thus, if a
n
=
(1)
n+1
n
for all n N, the rearranged series is written as

n=1
b
n
, with
b
3n2
=
1
2n 1
, b
3n1
=
1
4n 2
, b
3n
=
1
4n
for n N. Let s
n
and s

n
be the n
th
partial sums of the series

n=1
a
n
and

n=1
b
n
respectively.
Then we have
s

3n
= 1
1
2

1
4
+
1
3

1
6

1
8
+ +
1
2n 1

1
4n 2

1
4n
=
_
1
1
2

1
4
_
+
_
1
3

1
6

1
8
_
+ +
_
1
2n 1

1
4n 2

1
4n
_
=
_
1
2

1
4
_
+
_
1
6

1
8
_
+ +
_
1
4n 2

1
4n
_
=
1
2
_
_
1
1
2
_
+
_
1
3

1
4
_
+ +
_
1
2n 1

1
2n
_
_
=
1
2
_
1
1
2
+
1
3

1
4
+ +
1
2n 1

1
2n
_
=
1
2
s
2n
.
Also, we have
s

3n+1
= s

3n
+
1
2n + 1
, s

3n+2
= s

3n
+
1
2n + 1

1
4n + 2
.
We know that (s
n
) converges. Let lim
n
s
n
= s. Since, a
n
0 as n , we obtain
lim
n
s

3n
=
s
2
, lim
n
s

3n+1
=
s
2
, lim
n
s

3n+2
=
s
2
.
Hence, we can infer that s

n
s/2 as n . Thus, a rearrangement of a convergent alternating
series need not converge to the same limit.
Not only that. Look at the following astonishing result!
THEOREM 2.41. Suppose

n=1
a
n
is a conditionally convergent series. Then, for each s R,
there exists a sequence obtained by the rearrangement of the terms of

n=1
a
n
such that the rearranged
series converges to s.
Such situation would not occur if the given series is absolutely convergent. More precisely,
THEOREM 2.42. Suppose

n=1
a
n
is an absolutely convergent series, say to s. Then every series
obtained by rearranging the terms of

n=1
a
n
converge to the same number s.
The proofs of the above two theorems (Theorems 2.41 and 2.42) are quite involved, and hence we
omit the proofs. Interested readers may refer Delninger [?] for the proofs.
38