70 Principles of Real Analysis

CHAPTER 4

Infinite Series

4.1 INTRODUCTION
In this chapter we shall discuss the techniques of testing the behaviour of infinite series as regards
convergence. The most important technique for series, all of whose terms are of the same sign (all positive
or all negative), is to compare the given series with another suitably chosen series with known behaviour.
So, first of all, comparison tests are discussed, and then some special tests for convergence are considered.
Leibnitz’s test for alternating series, Abel’s and Dirichlet’s tests for arbitrary term series, and Dirichlet’s
and Riemann’s theorems on rearrangement of terms will be discussed in detail towards the end.
4.1.1 A series is the sum of the terms of a sequence. Thus if u1, u2, u3, ... is a sequence then the sum
∞
u1 + u2 + u3 + ... of all the terms is called an infinite series and is denoted by ∑ un or simply by ∑ un .
n =1
Evidently we cannot just add up all the infinite number of terms of the series in the ordinary way and
in fact it is not obvious that this kind of sum has any meaning. We thus start by associating with the given
series, a sequence {Sn}, where Sn denotes the sum of the first n terms of the series. Thus
S n = u1 + u2 + ... + un ,
The sequence {Sn} is called the sequence of partial sums of the series and the partial sums, S1 = u1,
S2 = u1 + u2, S3 = u1 + u2 + u3, and so on, may be regarded as approximations to the full infinite sum
∞
∑ un of the series. If the sequence {Sn} of partial sums converges, then the series is regarded as
n =1

convergent, and lim Sn is said to be the sum of the series. If however, {Sn} does not tend to a limit, we
must take it that the sum of the infinite series does not exist. We express this fact by saying that the series
does not converge. In fact an infinite series is said to converge, diverge or oscillate according as its
sequence of partial sums {Sn} converges, diverges or oscillates.
4.1.2 A Necessary Condition for Convergence
Copyright © 2004. New Academic Science. All rights reserved.

Theorem 4.1. A necessary condition for convergence of an infinite series ∑ un is that lim un = 0.
n ®¥

Let Sn = u1 + u2 + ... + un so that {Sn} is the sequence of partial sums.

Since the series converges, therefore, the sequence {Sn} also converges.
Consequently
b g
lim Sn = s say
n →∞

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 71

Now,
un = Sn − Sn −1 , n > 1

∴ lim un = lim Sn − lim Sn −1

n →∞ n→∞ n →∞

=s–s=0
Hence for a convergent series
lim un = 0
In other words, a series cannot converge if its nth term does not tend to zero,

Notes 1. It must be clearly understood that lim un = 0 does not prove that a series in convergent, for there exist series
which do not converge even though lim un = 0. See Example 4.2 below.
2. However, lim un ≠ 0 proves that the series does not converge, see Example 4.1 below:

Example 4.1. Show that the series

1 2 3
+ + + ... is not convergent.
2 3 4
n
Solution. Here un =
n+1
n
∴ lim un = lim =1≠ 0
n+1
Since lim un ≠ 0, therefore, the series is not convergent.

4.1.3 Cauchy’s General Principle of Convergence for Series

∞
A necessary and sufficient condition for the convergence of an infinite series ∑ un is that the sequence
n =1

of its partial sums {Sn} is convergent.

Therefore, a test for convergence of infinite series may be derived from our knowledge of sequences.

Theorem 4.2. A series ∑ un converges iff to each ε > 0, there exists a positive integer m such that

un + 1 + un + 2 + ... + un + p < A , " n ³ m and p ³ 1.

By Cauchy’s General Principle of Convergence (for sequences), the sequence {Sn}of partial sums
of ∑ un converges iff to each ε > 0 ∃ a positive integer m, such that

Sn + p − Sn < ε , ∀ n ≥ m and p ≥ 1
Copyright © 2004. New Academic Science. All rights reserved.

or

un + 1 + un + 2 + ... + un + p < ε , ∀ n ≥ m and p ≥ 1

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 72 Principles of Real Analysis

1
Example 4.2. Show that the series ∑ does not converge.
n
Solution. Suppose, if possible, the series converges.

 1
Therefore, for any given ε  say,  , ∃ a positive integer m, such that
 4

1 1 1
+ + ... + < ε , ∀ n ≥ m and p ≥ 1
n+1 n+2 n+ p
In particular, if n = m and p = m, we get
1 1 1 1 1 1
+ + ... + = + + ... +
n+1 n+2 n+ p m+1 m+2 2m
1 1
> m⋅ = >ε.
2m 2
Thus, there is a contradiction. Hence, the given series does not converge.
1
It may also be seen that lim un = lim = 0 even though the series does not converge.
n
4.1.4 Some Preliminary Theorems
Theorem 4.3. If å un = u, then å cun = cu , where c is any number independent of n.
The result follows at once from the identity
n n
∑ cur = c ∑ ur
r =1 r =1

on making n tends to infinity.

¥ ¥ ¥
Theorem 4.4. If å un = u, then å un = u + u0 , and å un = u - u1 .
n =1 n=0 n=2

n n
Let Sn′ = ∑ ur and Sn = ∑ ur .
r=0 r =1
Clearly
S n′ = u0 + S n
∴ By letting n tends to infinity
∞
∑ un = u0 + u
Copyright © 2004. New Academic Science. All rights reserved.

n=0

The proof of the second part is similar.

A slight modification and extension enables us to conclude that the insertion or removal of any finite
number of terms from a convergent series does not affect its convergence. Of course, the sums of the
various series are related in the expected way.

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 73

∞ ∞ ∞ ∞
It is also clear that if the series ∑ un is divergent, the changed series ∑ cun , ∑ un or ∑ un is
n =1 n =1 n=0 n=2
also divergent.
Hence, the behaviour of a series as regards convergence is not altered by
(i) the alteration, addition or omission of a finite number of terms; or
(ii) multiplication of all the terms by a finite number other than zero.
Theorem 4.5. (Convergent series may be added or subtracted term by term). If ∑ un = u and ∑ vn = v ,
then ∑ wn = u ± v , where wn = un ± vn , for all n.
The result follows from the identity
n n
b g n
∑ wr = ∑ ur ± v r = ∑ ur ± ∑ v r
r =1 r =1 r=1
n

r =1

by making n tend to infinity.

The same proof shows that
(i) if any two of the three series are convergent, the third is also convergent,
(ii) if one of the series is divergent and another convergent then the third is necessarily divergent,
but
(iii) if two of the series are divergent, no conclusion can be drawn about the behaviour of the
third, which may converge or diverge.
Theorem 4.6. If a series ∑ un converges to the sum u then so does any series obtained from ∑ un by
grouping the terms in brackets without altering the order of the terms.
Suppose that the series derived from ∑ un by the insertion of brackets is ∑ vn and let σ r denote the
rth partial sum of the series ∑ vn . Suppose that σ r contains nr terms of the given series then, since the
order of the terms is unaltered, σ r = Snr .
Also as r → ∞ , nr → ∞.
Since the given series converges to u, the sequence {Sn} of its partial sums also converges to u.
Hence, as r → ∞ or nr → ∞ , snr → u, and hence σ r → u.

Remarks 1. Converse of the theorem is not always true.

For example, the series (1 – 1) + (1 – 1) + ... is convergent, whereas the series 1 – 1 + 1 – 1 + ..., obtained
by removing the brackets is not.
Hence, in convergent series brackets may be inserted at will without affecting convergence but may not be
removed.
In the case of convergent positive term series, or the absolutely convergent series, however, it will be
Copyright © 2004. New Academic Science. All rights reserved.

shown later that the brackets may be inserted or removed without affecting convergence.
2. The theorem may as well be proved by the following alternative argument.
Since the given series converges to u, its sequence {Sn} of partial sums will also converge to u. Therefore,
the sequence {σ n } of partial sums of ∑ vn , being a subsequence of {Sn} will also converge to u.
3. The theorem may be restated, “A series obtained from a given convergent series by a grouping of terms
converges to the same limit”.
By grouping we simply mean the placing of brackets or associating the terms of the series without
changing the order of the terms.

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 74 Principles of Real Analysis

4.2 POSITIVE TERM SERIES

Series with non-negative terms are the simplest and the most important type of series one comes across.
The simplicity arises mainly from the fact that the sequence of its partial sums is monotonic increasing.
Let ∑ un be an infinite series of positive terms and {Sn}, the sequence of its partial sums, so that
S n = u1 + u2 + ... + un ≥ 0, ∀ n
∴ Sn − Sn − 1 = un ≥ 0
Þ Sn ³ Sn - 1 ,
Thus the sequence {Sn} of partial sums of a series of positive terms is a monotonic increasing
sequence.
Since a monotonic increasing sequence can either converge, or diverge to ∞, but cannot oscillate,
therefore, there are only two possibilities for a positive term series—it may either converge, or diverge
to + ∞.
Theorem 4.7. A positive term series converges iff the sequence of its partial sums is bounded above.
We know that the sequence of partial sums of a positive term series is a monotonic increasing
sequence and a monotonic increasing sequence converges iff it is bounded above. Therefore, it follows
that a positive term series converges iff its sequence of partial sums is bounded above.

Remarks 1. The sequence of partial sums of a series with negative terms can be shown to be monotonic decreasing and
hence a series with negative terms converges iff the sequence of its partial sums is bounded below.
2. It may similarly be seen that a series of negative terms can either converge, or diverge to – ∞.
3. A series of positive terms can either converge, or diverge to + ∞. But a series with arbitrary terms can have
five possible behaviours depending upon the behaviour of the sequence of its partial sums.

4.2.1 Geometric Series

The positive term geometric series 1 + r + r2 + ..., converges for r < 1, and diverges to + ∞ for r ≥ 1.

Case I. 0 ≤ r < 1.
Let {Sn} be the sequence of its partial sums, so that
Sn = 1 + r + r2 + ... + rn – 1
i.e.,
1 − rn 1 rn 1
Sn =
= − ≤ , ∀ n,
1−r 1− r 1− r 1− r
so that {Sn} is bounded above and hence the series converges.
Copyright © 2004. New Academic Science. All rights reserved.

Case II. When r = 1, Sn = n, so that {Sn} is not bounded above and hence the series diverges to + ∞.
Case III. When r > 1, every term of Sn after the first is greater than 1, so that
S n > n, ∀ n .
Therefore, the sequence {Sn} is not bounded above and consequently the given series diverges
to + ∞.
Hence, the given series converges if r < 1 and diverges if r ≥ 1.

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 75

4.2.2 A Comparison Series

As mentioned earlier, an important technique for testing the convergence of a series is to compare the
given series with a suitably selected series with known behaviour. We now discuss one such series
which is most frequently used for such a purpose.
1
Theorem 4.8. A positive term series Σ
is convergent if and only if p > 1.
np
Let Sn denote the sum of the first n terms of the given series.
Case I. When p > 1.
Now
1
=1
1p
1 1 2 1
p
+ p < p
= p −1
2 3 2 2
1 1 1 1 4 1 FG 1 IJ 2

4 p
+ p + p + p <
5 6 7 4 p
= p−1
4
=
H2 Kp −1

1 1 1 8 1
=G
F 1 IJ 3

8 p
+ p + ... + p <
9
... ... ... ... ...
15 8 p
= p −1
8
... ...
H2 K
... ...
p −1
...
... ... ... ... ... ... ... ... ... ...

1 1 1 2n + 1 − 2n 1 F 1 IJ
=G
n

d2 i d2
n p
+
n
+1 i p
+ ... +
d2 n +1
−1 i p
<
d2 i n p
=
d2 i
n p −1 H2 K
p −1

Adding,

1 FG 1 IJ 2
FG 1 IJ
n
S2 n + 1 − 1 < 1 +
2p −1
+
H2 K p −1
+ ... +
H2 p −1 K
F 1 IJ n +1 LM1 − FG 1 IJ OP
n+1

1−G
MN H 2 K PQ
2p−1
=
H2 K p −1
=
p−1

1−
1 2 p− 1 − 1
p −1
2
p −1
2
, ∀n < p −1
2 −1
Copyright © 2004. New Academic Science. All rights reserved.

We know that when n is any positive integer,

2n + 1 − 1 > 2 n > n

2p −1
∴ Sn < S2 n < S2 n + 1 − 1 < .
2p −1 − 1

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 76 Principles of Real Analysis

2p −1
Since for a given p, is a fixed number, therefore, the sequence {Sn} of partial sums of the
p−1
2 −1
given positive term series is bounded above and hence the series converges for p > 1.
Case II. When p ≤ 1.
We know, if n is any positive integer and p ≤ 1, then
1 1
np ≤ n ⇒ ≥
np n
1 1 1
∴ p
1+
≥1+ >
2 2 2
1 1 1 1 2 1
+ ≥ + > =
3p 4 p 3 4 4 2
1 1 1 1 1 1 1 1 4 1
p
+ p + p + p ≥ + + + > =
5 6 7 8 5 6 7 8 8 2
1 1 1 1 1 1 8 1
p
+ p + ... + p ≥ + + ... + > =
9 10 16 9 10 16 16 2
... ... ... ... ... ... ... ... ...
... ... ... ... ... ... ... ... ...

1 1 2 1 1 1 2 m − 2m − 1 1
+ + ... + ≥ + + ... + > =
d2 m−1
+1 i d2
p
m −1
+2 i p
d2 i
m
p
2m − 1 + 1 2m − 1 + 2 2m 2m 2

Adding,
m
. S2 m >
2
We shall now show that {Sn} is not bounded above.
If G be any number, however large, then ∃ m ∈ N such that
m
>G
2
Let n > 2m,
∴ Sn > S2m > G
Thus, the sequence {Sn} of partial sums of the given series is not bounded above, and hence the
series diverges for p ≤ 1.
Copyright © 2004. New Academic Science. All rights reserved.

1
Thus, the given series ∑ converges iff p > 1.
np

4.3 COMPARISON TESTS FOR POSITIVE TERM SERIES

Two types of comparison tests shall now be discussed. In the first type, the general term of one series
will be compared with the general term of the second series. In the second type, the ratio of two consecutive

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 77

terms of one series will be compared to the ratio of the corresponding consecutive terms of the second
series.
4.3.1 Comparison Test (First type)
I. If ∑ un and ∑ vn are two positive term series, and k ¹ 0, a fixed positive real number (independent
of n) and there exists a positive integer m such that un ≤ kvn , ∀ n ≥ m, then
(i) ∑ un is convergent, if ∑ vn is convergent, and
(ii) ∑ vn is divergent, if ∑ un is divergent.
Let n ≥ m, Sn = u1 + u2 + ... + un, and tn = v1 + v2 + ... + vn.
Now " n ³ m, we have
Sn − Sm = um + 1 + um + 2 + ... + un

d i b
≤ k vm + 1 + vm + 2 + ... + vn = k t n – t m g
or b
Sn ≤ kt n + Sm − kt m g
⇒ S n £ kt n + h ...(1)
where h = Sm – ktm, is a finite quantity.
(i) If ∑ vn is convergent, then the sequence {tn} of its partial sums is bounded above, so that ∃
a number B such that
tn ≤ B, ∀ n .
So from (1), we get
Sn ≤ kB + h, for all n ≥ m,
⇒ the sequence {Sn} is bounded above.
⇒ å un is convergent.
(ii) If ∑ un is divergent, then the sequence {Sn} of its partial sums is not bounded above, so that
if G be any number, however, large, ∃ a positive integer m0 such that
Sn > G " n ³ m0 .
Thus from (1), ∀ n ≥ max ( m, m0 ),
1
tn ≥ ( G − h), k ≠ 0
k
⇒ the sequence {tn} is unbounded
⇒ åvn is divergent.
Copyright © 2004. New Academic Science. All rights reserved.

FG u IJ = l , where l is a
Hv K
n
II. Limit Form. If å un and å vn are two positive term series such that lim
n →∞
n

non-zero finite number, then the two series converge or diverge together.
Evidently l > 0.
Let ε be a positive number such that l − ε > 0.

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 78 Principles of Real Analysis

un
Since lim = l , therefore ∃ a positive integer m such that
n→∞ vn
un
- l < A,
vn "n³m

un
⇒ l-A< < 1 + A, "n³m
vn
⇒ bl - A g v n b g
< un < 1 + A v n , "n³m ...(1)
Now, if ∑ vn is convergent, then from (1)
un < (1 + A ) vn , "n³m
so that by Test I, ∑ un is convergent.
Again, if ∑ vn is divergent, then from (1)
b
un > 1 - A vn , g "n³m
so that by Test I, ∑ un is divergent.
Similarly, we may show that ∑ vn converges or diverges with ∑ un . Hence, the two series behave
alike.
4.3.2 Comparison Test (Second type)
III. If ∑ un and ∑ vn are two positive term series and ∃ a positive integer m such that
un v
≥ n , ∀ n ≥ m,
un + 1 v n + 1
then (i) ∑ un is convergent, if ∑ vn is convergent, and (ii) ∑ vn is divergent, if ∑ un is divergent.
Let m be a positive integer and Sn = u1 + u2 + ... + un
and tn = v1 + v2 + ... + vn .
For n ≥ m, we have
um u um + 1 un − 1 v vm + 1 vn − 1 v m
= m ⋅ ... ≥ m ⋅ ... =
un um + 1 um + 2 un vm + 1 vm + 2 vn vn
um
⇒ un £ vn .
vm
Since m is a fixed positive integer, therefore um/vm is fixed number, say k. Thus ∀ n ≥ m, we have
Copyright © 2004. New Academic Science. All rights reserved.

un ≤ kv n
Hence by Test I, ∑ un converges if ∑ vn converges and ∑ vn diverges if ∑ un diverges.

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 79

Notes 1. For practical purposes, Test II is very useful and can be easily applied.
2. For a successful application of the comparison test, we first make an estimate of the magnitude of the general
term un of the given series, and then select the auxiliary series ∑ v n of such a magnitude that
lim( un /vn ) = l ≠ 0, ∞, or in other words un ~ vn . Thus for large values of n,
nr
n 3 + 1 ± ~ ± n 3/2 , ~ ± nr - s
(1 + n ) s
1 1
sin ±~±
n n
3. It will help to remember that for large n, e an >> n b >> ( log n) c , where a, b, c are positive numbers.

4.3.3 Solved Examples

1 1 1
Example 4.3. Show that the series 1 + + + +... is convergent.
2! 3! 4!
Solution. We have
1 1
=
2! 2
1 1
< 2
3! 2
1 1
< 3
4! 2
... ...
... ...
1 1
< n −1
n! 2
1 1 1 1 1
∴ 1 + + + ... + + ... < 1 + + 2 + ...
2! 3! n! 2 2
Thus each term of the given series after the second is less than the corresponding term of the
convergent geometric series
1 1
1 + + 2 + ...
2 2
Thus by Test I, the given series converges.
Example 4.4. Show that the series
1 1 1
Copyright © 2004. New Academic Science. All rights reserved.

+ + ... + + ...
blog 2g blog 3g
p p
blog ng p

diverges for p > 0.

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 80 Principles of Real Analysis

Solution. Since lim

blog ng p

= 0, therefore
n →∞ n
blog ng p
< n, ∀ n > 1
1 1
> , "n>1
∴
blog ng p
n
Let us compare the given series with the divergent series
1 1 1
+ + + ...
2 3 4
Since each term of the given series is greater than the corresponding term of the divergent series,
therefore, the given series diverges.
Example 4.5. Show that the series
1⋅ 2 3⋅ 4 5⋅ 6
2 2
+ 2 2 + 2 2 + ..., is convergent
3 ⋅4 5 ⋅6 7 ⋅8
Solution. Let us denote the given series by ∑ un , where
b2n − 1g b2ng , FG ~ 1 IJ .
b2n + 1g b2n + 2g H n K
un = 2 2 2

Let us compare it with the convergent series ∑ vn ,

where vn = 1/n2.
un ( 2 + 1/n) (2 + 2/n)
Now lim
n ®¥ vn
= lim
n ®¥ b
2 + 3/n 2 + 4 /ngb =1
g
Thus, the two series converge or diverge together.
Since, ∑ vn converges, therefore ∑ un also converges.
Example 4.6. Investigate the behaviour of the series whose nth term is sin 1/n.
Solution. Let un = sin 1/n, and vn = 1/n.
Now
un sin 1/n
lim = lim =1
n →∞ vn n→∞ 1/n
Therefore, the two series behave alike.
Since, ∑ vn diverges, therefore ∑ sin 1/n also diverges.
Copyright © 2004. New Academic Science. All rights reserved.

Example 4.7. Test for convergence of the series whose nth term is
{(n3 + 1)1/3 – n}

Solution. Let un = n 3 + 1 d i −n
1 /3

|RF 1 I
= n SG1 + J
1/3
|UV
|TH n K 3
−1
W|

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 81

RS 1 UV 1 FG ~ 1 IJ
=n
T 3n 3
+ ... =
W 3n 2
+ ...
H nK 2

1
Let ∑ v n = ∑
n2
Now
un 1
lim =
n →∞ vn 3
Therefore, the two series converge and diverge together.
Since, ∑ vn converges, therefore, the given series also converges.
1
Example 4.8. Test the convergence of the series ∑ 1 + 1 /n
.
n
1 1
Solution. Let un = ∑ and vn =
n1 + 1/n n

un 1
Now = lim 1/ n = 1 lim
vn n →∞ n n →∞

Hence, the two series ∑ un and ∑ vn behave alike.

1
Since, ∑ vn is divergent, therefore ∑ 1 + 1 /n
is also divergent.
n

EXERCISE

1 3 5 7
1. + + + + ...
1⋅ 2 ⋅ 3 2 ⋅ 3⋅ 4 3 ⋅ 4 ⋅ 5 4 ⋅ 5 ⋅ 6
1 1 1
2. + + + ...
1⋅ 2 2⋅3 3⋅4
1 3 5 7
3. + + + + ...
4 ⋅ 6 6 ⋅ 8 8 ⋅ 10 10 ⋅ 12
n+1
4. ∑
np
1
5. ∑
n + n+1
Copyright © 2004. New Academic Science. All rights reserved.

6. ∑ FH n4 + 1 − n4 − 1 IK [Hint: Rationalize]

1 1
7. (i) ∑ sin , (ii) ∑ cos
n2 n

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 82 Principles of Real Analysis

1 1
8. ∑ tan
n n
R| n +1 − n −1 U|
9. ∑ S| V|
T n W
1 1 1
10. Show that the series 1 + 2
+ 3 + 4 + ... converges.
2 3 4
[Hint: nn > 2n, for n > 2. Compare with the convergent geometric series ∑1/2 n ]

ANSWERS
1. Convergent, 2. Divergent, 3. Convergent, 4. Convergent for p > 2, 5. Divergent, 6. Convergent,
7. (i) Convergent, (ii) Divergent, 8. Convergent, 9. Convergent.

4.4 CAUCHY’S ROOT TEST

If ∑ un is a positive term series, such that lim ( un )1/n = l , then the series
n →∞

(i) converges, if l < 1,

(ii) diverges, if l > 1, and
(iii) the test fails to give any definite information, if l = 1.
Case I. l < 1.
Let us select a positive number ε , such that l + ε < 1.
Let l + ε = α < 1.
Since, lim ( un )1/n = l , therefore ∃ a positive integer m such that
n →∞

bu g
n
1/n
- l < A, "n³m

⇒ l - A < un b g 1/n
< l + A, "n³m

⇒ bl - A g < u < bl + A g
n
n
n
= =n, "n³m

⇒ un < = n , " n ³ m.

But since ∑ α n is a convergent geometric series (common ratio α < 1), therefore, by comparison
Copyright © 2004. New Academic Science. All rights reserved.

test, the series ∑ un converges.

Case II. l > 1.
Let us select a positive number ε such that l - A > 1
Let l − ε = β > 1.

Since, lim ( un ) 1/n = l , therefore ∃ a positive integer m1 such that

n→∞

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 83

b g < l + A,
l - A < un
1/n
" n ³ m1

Þ bl - A g < u < bl - A g ,
n n
n " n ³ m1

Þ u > bl - A g = > ,
n n
n " n ³ m1 .
n
But since ∑ β is a divergent geometric series (common ratio β > 1), therefore by comparison
test, the series ∑ un diverges.

Note. The test fails to give any definite information for l = 1.

Consider the two series ∑ (1/n) and Σ (1/n 2 ).

∑ (1/n) diverges, while lim (1/n )1/n = 1, and ∑ (1/n 2 ), converges, while lim (1/n 2 )1/n = 1.
n →∞ n →∞

F 1 IJ
Test for convergence of the series whose general term is G 1 +
− n 3/ 2
Example 4.9.
H nK .

1
Solution. Let un = , then
e1 + 1/ n j
n 3 /2

lim unb g 1/n

= lim
1
=
1
< 1.
n →∞ n→∞
e1 + 1/ n j n e

Hence, the series converges.

4.5 D’ALEMBERT’S RATIO TEST

un + 1
If ∑ un is a positive term series, such that lim = l , then the series
n →∞ un
(i) converges, if l < 1,
(ii) diverges, if l > 1, and
(iii) the test fails, if l = 1.
Case I. 0 < l < 1.
Let us select a positive number ε , such that l + ε < 1.
Let l + ε = α < 1, α ≠ 0.
Copyright © 2004. New Academic Science. All rights reserved.

un + 1
Since, lim = l , therefore ∃ a positive integer m such that
n →∞ un

un + 1
− l < ε, ∀ n ≥ m
un

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 84 Principles of Real Analysis

un + 1
⇒ l-A< < l + A, "n³m
un
un + 1
⇒ < l + A = =, "n³m
un
For n ≥ m,
un um + 1 um + 2 u
= ⋅ ... n < α n − m
um um um +1 un − 1

FG u IJ = ,
H= K
m n
⇒ un < m " n ³ m, = < 1.

Since m is a fixed integer, therefore G

F u IJ is a fixed finite number, say k.
Hα K
m
m

Thus, ∀ n ≥ m, we have
un < kα n
But since ∑ α n is a convergent geometric series (common ratio, = < 1), therefore by comparison
test ∑ un converges.
Case II. l > 1.
Let us select a positive number ε , such that l − ε > 1.
Let l − ε = β > 1.
un + 1
Since, lim = l , therefore ∃ a positive integer m1 such that
n →∞ un

un + 1
l-A< < l + A, " n ³ m1
un
un + 1
⇒ > l - A = >, " n ³ m1
un
Now, for n ≥ m1 ,
un um + 1 um + 2 u
= 1 × 1 ... n ³ > n - m1
um1 um1 um1 + 1 un - 1

um1
Copyright © 2004. New Academic Science. All rights reserved.

⇒ >n,
un ³ " n ³ m1
> m1
m
Since m1 is a fixed integer, therefore um1 /β 1 is a fixed finite number, say k1.
Thus, for n ≥ m1 , we have
un > k1> n .

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 85

n
But ∑ β is a divergent geometric series (common ratio, > > 1); therefore by comparison test,
∑ un diverges.

Note. The test fails for l = 1 in the sense that it fails to give any definite information.
For example, consider the two series ∑(1/n) and ∑ (1/n 2 )

∑
1 un + 1
= lim
n 1 u +1
= 1, and ∑ 2 converges, while lim n = lim
n FG IJ 2

n
diverges, while lim
n →∞ u
n
n →∞ n + 1 n n →∞ u n
n →∞ n +1 H K = 1.

n2 − 1 n
Example 4.10. Test for convergence of the series ∑ x , x > 0.
n2 + 1

n2 − 1 n
Solution. Let un = x .
n2 + 1

lim
un + 1
= lim 2 ⋅ ⋅ n =x
b
n2 + 1 n + 1 − 1 x n + 1 g 2

∴ n →∞ u
n
n →∞ n − 1 2
n+1 +1 x b g
Hence by D’Alembert’s Ratio Test the series converges if x < 1 and diverges if x > 1.
The test fails to give any information when x = 1.
n2 − 1
When x = 1, un = , and lim un = 1 ≠ 0
n2 + 1 n →∞

⇒ The series is divergent.

Hence, the series converges if x < 1 and diverges if x ≥ 1.

4.6 RAABE’S TEST

Fu I
If ∑ un is a positive term series, such that lim n
n →∞
GH u
n

n +1
JK
− 1 = 1, then the series

(i) converges, if l >1,

(ii) diverges, if l < 1, and
(iii) the test fails, if l = 1.
Case I. l > 1.
Let us select a positive number ε , such that l − ε > 1.
Copyright © 2004. New Academic Science. All rights reserved.

Let l − ε = α > 1.

Fu I
Since lim n
n →∞
GH u
n

n +1
JK
− 1 = l , therefore ∃ a positive integer m such that for all n ≥ m,

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 86 Principles of Real Analysis

Fu I
l−ε<n GH u n

n +1
JK
−1 <l +ε

Fu I
⇒ n GH u n +1
n
JK
-1 >l -A ==

⇒ nun - nun + 1 > = un + 1

⇒ b g
nun - n + 1 un + 1 > = - 1 un + 1 , = - 1 > 0 b g
Putting n = m, m + 1, m + 2, ..., n – 1, and adding, we get
b gd
mum − nun > α − 1 um + 1 + um + 2 + ... + un i
n
= (α − 1) ( Sn − Sm ) , where Sn = ∑u r
r =1

⇒ b= - 1g b S n g
- S m < mum , "n³m
m
⇒ Sn < Sm + um , " n ³ m,
= -1

m
Since m is a fixed integer, therefore Sm + um is a fixed finite number.
α −1
Thus, the sequence {Sn} of partial sums of the given series is bounded above and hence the series
∑ un is convergent.
Case II. l < 1.
Let us select a positive number ε , such that l + A < 1.

Fu I
Since lim n
n →∞
GH u
n

n +1
JK
− 1 = l , therefore ∃ a positive integer m such that for all n ≥ m,

Fu I
l−ε<n GH u n

n +1
JK
−1 <l +ε <1

un n +1
⇒ < , "n³m
un + 1 n

If vn = 1/n, the series ∑ vn is divergent, and,

Copyright © 2004. New Academic Science. All rights reserved.

un v
< n
un + 1 v n + 1
Hence by comparison test the series ∑ un diverges.

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 87

1 1
Notes 1. The test fails to give any definite information for l = 1. Consider the two series ∑ and ∑ .
n b
n log n g 2

The former is divergent, while the later is convergent but

Fu I
lim n
n →∞
GH u
n+1
n
JK
− 1 = 1 for both.

2. Raabe’s Test is stronger that D’Alembert’s Ratio Test and may succeed where Ratio Test fails.

Example 4.11. Test for convergence of the series

α 1+α 1+α 2 +α b gb g
+
β 1+ β
+
1+ β 2 + β
+ ...
b gb g
Solution. Here
b1 + α g b2 + α g b3 + α g ... bn − 1 + α g
un =
b1 + β g b2 + β g b3 + β g ... bn − 1 + β g
un + 1 n+α
lim = lim =1
un n →∞ n→∞ n+β
Hence, the Ratio Test fails.
Again
F n+β I FG IJ
lim n
n →∞
un
un + 1
GH
− 1 = lim n
n →∞ n+α
JK
− 1 = lim
β −α
n →∞ 1 + α /n H
=β −α
K
Thus by Raabe’s Test, the series converges if β − α > 1 or β > α + 1, and diverges if β < α + 1.
The test fails for β = α + 1.
But for β = α + 1, the series becomes
α 1+ α 1+α 1+ α
+ + + ... = ∑
α +1 2+α 3+α n+α

which diverges, by comparison with ∑1/n.

Example 4.12. Show that the series
3⋅ 6 ⋅ 9 ... 3n
xn , x > 0
∑
7 ⋅10 ⋅13 ... 3n + 4 b g
converges for x ≤ 1, and diverges for x > 1.
Copyright © 2004. New Academic Science. All rights reserved.

Solution. Now

un
= xn
3⋅ 6 ⋅ 9 ... 3n 7 ⋅10 ⋅13 ... 3n + 4 3n + 7 1 b gb g
un + 1 7 ⋅10 ... 3n + 4b ⋅
g
3⋅ 6 ⋅ 9 ... 3n 3n + 3 x n+1
b g

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 88 Principles of Real Analysis

3n + 7 1
= ⋅
3n + 3 x

un + 1
∴ lim =x.
n →∞ un
By Ratio Test, the series converges for x < 1 and diverges for x > 1. The test fails for x = 1.
But for x = 1,
un 3n + 7
= .
un + 1 3n + 3

Fu I
∴ lim n
n →∞
GH u n +1
n
JK
− 1 = lim
n→∞
4n 4
= >1
3n + 3 3

∴ By Raabe’s Test, the series converges.

Thus the series converges for 0 < x ≤ 1 and diverges for x > 1.

EXERCISE
Test the behaviour of the following series:
1 1 1 1
1. + + + + ...
2 2 ⋅ 22 3 ⋅ 23 4 ⋅ 24

FG 1IJ + FG 1⋅2 IJ + FG 1⋅2 ⋅3 IJ

2 2 2

2.
H 3K H 3⋅ 5 K H 3⋅ 5 ⋅ 7 K + ...

x x2 x3 x4
3. + + + + ...,
1 2 3 4

x2 x3 x4 x5
4. + + + + ...,
2 1 3 2 4 3 5 4

x2 x4 x6
5. 1 + + + + ...
2p 4p 6p

12 ⋅ 2 2 2 2 ⋅ 32 32 ⋅ 42 4 2 ⋅52
6. + + + + ...
1! 2! 3! 4!
F2 I + F3 I + F4 I
Copyright © 2004. New Academic Science. All rights reserved.

2 −1 3 −2 4 −3

7. GH 1
2
−
2
1 JK GH 3 3
−
3
2 JK GH 3 4
−
4
3 JK + ...

1 1 1 1
8. + + 3 + 4 + ...
2 32 2 3

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 89

Test for convergence each of the following series whose nth terms are given:
n
9. ,r>0
nn
1⋅ 2 ⋅ 3 ... n
b
10. 7 ⋅ 10 ... 3n + 4
g
n
r
11. , r >0
nn
n xn
12. , x>0
n2 + 1

13.
b
1⋅ 3⋅5 ... 2n − 1 g⋅1
2 ⋅ 4 ⋅ 6 ... 2n n
b g⋅x
1⋅ 3⋅ 5 ... 4n − 3 2n

2 ⋅ 4 ⋅ 6 ... b4n − 2g 4n
14.

n −1 n
15. x , x>0
n3 + 1
b
2 × 4 × 6 ... 2 n + 2 g
b gx
n-1
16.
3× 5 × 7 ... 2n + 3

ANSWERS
1. Converges 2. Converges
3. Converges for x < 1, diverges for x≥1 4. Converges for x ≤ 1, diverges for x > 1
5. Converges for | x | < 1, diverges for | x | > 1; for x = 1, converges for p > 1, diverges for p ≤1
6. Convergent 7. Convergent
8. Convergent 9. Convergent
10. Convergent 11. Convergent
12. Convergent for x < 1, divergent for x ≥ 1 13. Convergent

14. Convergent for x £ 1, divergent for x > 1 15. Convergent for x < 1, divergent for x ≥ 1

16. Convergent for x < 1, divergent for x ≥ 1.

4.7 LOGARITHMIC TEST

Copyright © 2004. New Academic Science. All rights reserved.

If ∑ un is positive term series such that,

F I
n ®¥
GH
lim n log
un
JK
un + 1
= l,

then the series converges for l > 1, and diverges for l < 1.

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 90 Principles of Real Analysis

First, let l > 1.

Let us select ε > 0, such that l - A < 1.
Let l − ε = α > 1.

F I = l, therefore ∃ an m, such that

Since lim n log
n →∞
GH un
un + 1
JK
un
l - A < n log < l + A, " n ³ m
un + 1
un
⇒ n log > =, " n ³ m
un + 1
un
⇒ > e= /n , " n ³ m
un + 1
Now since {(1 + 1/n)n} is a monotonic increasing sequence converging to e, therefore

FG1 + 1 IJ n

H nK ≤ e, ∀ n

so that, we get
un FG
1 IJ = bn + 1g
α α
vn
un + 1
> 1+
H
n K n α
=
vn + 1
, ∀n≥m

where vn = 1/nα .
But since for α > 1, ∑ vn converges, therefore by comparison test, ∑ un also converges.
We may similarly show that for l < 1, the series ∑ un diverges.
Example 4.13. Test for convergence the series
x 2 2 x 2 33 x 3
1+ + + + ..., for x > 0
1! 2! 3!
nn x n
Solution. Ignoring the first term, un = .
n!

un + 1 FG n + 1IJ n
Copyright © 2004. New Academic Science. All rights reserved.

Now lim
n →∞ un
= lim
n→∞ H n K x = ex

By Ratio Test, the series converges for ex < 1 or x < 1/e, and diverges for x > 1/e.

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 91

For x = 1/e,
un n FG IJ n

un + 1
=
n+1 H K ⋅e

Now
F I LM FG IJ OP
GH
lim n log
n →∞
un
un + 1 n
JK
= lim n 1 − n log 1 +
→∞ N
1
n
.
H KQ
L F1 1
= lim n M1 − n G −
1 IJ OP = 1 < 1
n →∞ N H n 2n 2
+
3n 3
− ...
KQ 2
Therefore by logarithmic test, the series diverges.
Hence, the series converges for x < 1/e, and diverges for x ≥ 1/e.

Note. Logarithmic test is generally more helpful in situations like those above, where the presence of a number like e
in un/un + 1 makes the application of Raabe’s Test difficult.

4.8 INTEGRAL TEST

Improper Integral. As preparatory to the introduction of Cauchy’s Integral Test, it will help to remember

that the infinite integral za

∞
u( x ) dx is said to converge if t ( x ) =

x → ∞; otherwise the integral is said to diverge.

za
x
u( x ) dx tends to a finite limit as

Further, if u( x) ≥ 0 for all x > a, it can be shown geometrically or otherwise that the integral

z t
u( x ) dx is a monotonic increasing function of t, so that the improper integral

z
a
∞
u( x ) dx , where u ( x ) ≥ 0, ∀ x > a
a
converges iff it is bounded above i.e., ∃ a positive number k such that

z a
t
u( t ) dt ≤ k , ∀ t ≥ a

4.8.1 Cauchy’s Integral Test

If u is a non-negative monotonic decreasing integrable function such that u(n) = un, for all positive

integral values of n, then the series ∑ un and

∞

z ∞
u( x ) dx converge or diverge together.
Copyright © 2004. New Academic Science. All rights reserved.

n =1 1

As u is monotonic decreasing, we have

bg bg b
u n ≥ u x ≥ u n + 1 , whenever n ≤ x ≤ n + 1 g
Also, since u is non-negative and integrable,

z n
n +1
u(n) dx ³ zn
n+1
u( x ) dx ³ z
n
n+1
u( n + 1) dx

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 92 Principles of Real Analysis

⇒ bg
un ≥ z
n
n +1
u( x ) dx ≥ u n + 1 b g
or un ≥ zn
n +1
u( x ) dx ≥ un + 1 . ...(1)

Let us write Sn = u1 + u2 + ... + un, and I n =

and adding, we get
z1
n
u( x ) dx , and putting n = 1, 2, ..., (n – 1) successively,

S n − un ≥ I n ≥ S n − u1
⇒ 0 < un £ Sn - I n £ u1 ...(2)
Let us consider the sequence {(Sn – In)}.
bS − I g − dS
n n n −1 − i
I n − 1 = Sn − Sn − 1 − I n − I n − 1 d i
= un − z
n −1
n
u( x ) dx

≤0 [using (1)].
Therefore, the sequence {(Sn – In)} is monotonic decreasing, bounded by 0 and u1.
Hence, the sequence converges and has a limit such that
b
0 ≤ lim Sn − I n ≤ u1 g ...(3)

Thus the series ∑ un converges or diverges with the integral

the series differs from the integral by less than u1; if divergent, the limit of (Sn – In) still exists and lies
z ∞

1
u( x ) dx; if convergent, the sum of

between 0 and u1.

Example 4.14. Show that the series ∑(1/n p ) converges, if p > 1, and diverges if p ≤ 1.
Solution. Let u(x) = 1/xp, so that for x ≥ 1, the function u is a non-negative monotonic decreasing
integrable function such that

un = u n = bg 1
np
, ∀ n ∈N

By Integral Test, ∑ un and

∞

n =1 z
1
∞
u( x ) dx converge or diverge together.

Let us now test the convergence of the infinite integral.

 1
( x1− p − 1), if p ≠ 1
Copyright © 2004. New Academic Science. All rights reserved.

x x 1 
∵ ∫1 u ( x)dx = ∫1 xp
dx =  1 − p
log x, if p = 1

 1 if p > 1
∞ x  ,
∴ ∫1 u ( x )dx = lim ∫
x →∞ 1
u ( x) dx =  p − 1
 ∞, if 0 < p ≤ 1


Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 93

Thus z1
∞
u( x ) dx converges, if p > 1, and diverges, if 0 < p ≤ 1.
Hence, the infinite series ∑(1/n p ) converges if p > 1, and diverges if 0 < p ≤ 1.
But, when p < 0, the series ∑(1/n p ) diverges for then the nth term n–p does not tend to zero as
n → ∞.
Hence the series ∑(1/n p ) converges when p > 1, and diverges when p ≤ 1.
∞ 1
Example 4.15. The series ∑ , p > 0, converges for p > 1 and diverges for p ≤ 1.
n=2 b
n log n g p

1
Solution. Let u( x ) = , so that for x ≥ 2 , the function u is a non-negative monotonic decreasing
b
x log x g p

integrable function, such that

un = u n = b g nb log1 ng p
, ∀ p > 0, n ∈ N

By Integral Test ∑ un and

n=2
∞

Let us now test the convergence of the infinite integral.

z
2
∞
u( x ) dx converge or diverge together.

3 z X

2
u( x ) dx = zX

2
b
x log x
1
g p
dx , p>0

R| blog X g − blog 2g 1− p 1− p

=S , if p ≠ 1,
|Tlog log X1−−logp log 2, if p = 1.

R| blog 2g
z z
1− p

u( x ) dx = S p − 1
∞ X
u( x ) dx = lim , if p > 1,
∴
2 X →∞ 2
|T ∞ , if 0 < p ≤ 1.

Thus z2
∞
u( x ) dx converges if p > 1, and diverges if 0 < p ≤ 1.

∞ 1
Hence, the series ∑ , p > 0, converges if p > 1, and diverges if p ≤ 1.
b g p
Copyright © 2004. New Academic Science. All rights reserved.

n=2 n log n

4.9 GAUSS’S TEST

If ∑ un is a positive term series such that,
un β γ
= α + + np ,
un + 1 n n

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 94 Principles of Real Analysis

where α > 0, p > 1, and {γ n } is a bounded sequence, then

(i) for α ≠ 1, Σ u n converges if α > 1, and diverges if α < 1, whatever β may be,

(ii) for α = 1, Σ u n converges if β > 1 and diverges if β ≤ 1.

Proof: (i) When α ≠ 1,
un
lim =α
n →∞ un + 1
Hence by Ratio Test, the series converges if α > 1, and diverges if α < 1.
(ii) When α = 1,
Fu I
lim n
n →∞
GH u
n+1
n
JK
−1 = β

Hence by Raabe’s Test, the series converges if β > 1, and diverges if β < 1.
For β = 1, we have
un 1 γ
= 1 + + np , p>1
un + 1 n n
1
Let us compare the given series with the divergent series ∑ vn where vn = .
n log n
Now
un v 1 γ
− n = 1 + + np −
n + 1 log n + 1 b g b g
un + 1 vn + 1 n n n log n

= −
LM b g OP
γ n n + 1 log n + 1
−1
np n Nlog n Q
1 L F 1 I n OP
γ − bn + 1g log G1 + J ⋅
p −1
=
n N
Mp n
H n K log n Q
F 1I L F 1I F 1I O
lim bn + 1g log G 1 + J = lim Mlog G1 + J + log G 1 + J P = 1
n
Since
n=∞ H nK n →∞
MN H n K H n K PQ
np −1
Copyright © 2004. New Academic Science. All rights reserved.

and nlim = ∞ , p > 1, and {γ n } is bounded, therefore, for sufficiently large values of n,
=∞ log n

b g FGH
γ n − n + 1 log 1 +
IJ
1 np−1
K
⋅
n log n
remains negative.

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 95

Thus ∃ a positive integer m such that

b g FGH
γ n − n + 1 log 1 + ⋅
IJ
1 np −1
K
n log n
< 0, ∀n≥m

un v
⇒ − n < 0, ∀n≥m
un + 1 v n + 1

un vn ,
⇒ < ∀n≥m
un + 1 vn + 1

Since ∑ vn is divergent, therefore by Comparison Test, the series ∑ un is also divergent.

Remarks 1. Gauss’s Test is very useful and may be used after the failure of Raabe’s Test or directly without recourse
to other tests.
2. If

un β γ δ
= α + + 2 + 3 + ... ...(1)
un + 1 n n n
where α , β , γ ... are independent of n, then we can write

un β γ
= α + + 2n
un + 1 n n
where γ n = γ + δ /n + ... , so that lim γ n = γ , i.e., {γ n } is a bounded sequence.
Thus for the application of Gauss’s Test we may expand un /un + 1 in powers of 1/n as in (1).

Example 4.16. Test the convergence of the series

2 2 2 2 ⋅ 4 2 2 2 ⋅ 4 2 ⋅ 62
+ 2 2 + 2 2 2 + ...
32 3 ⋅5 3 ⋅5 ⋅7
Solution. Now

un =
2 2 ⋅ 4 2 ⋅ 62 ... 2 nb g 2

b
32 ⋅52 ⋅ 7 2 ... 2n + 1 g 2

un b g = FG1 + 3 IJ FG1 + 1 IJ
2n + 3
2 2 −2

b g H 2n K H n K
∴ = 2
un + 1 2n + 2

F 3 9 IJ FG1 − 2 + 3 − ...IJ
= G1 + +
Copyright © 2004. New Academic Science. All rights reserved.

H n 4n K H n n K 2 2

1 3 1
=1+ − 2 + ... higher powers of ,
n 4n n
so that α = 1 and β = 1.
Hence by Gauss’s Test, the series diverges.

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 96 Principles of Real Analysis

Example 4.17. Test for convergence of the series

∑
b
12 ⋅ 32 ... 2n − 1 g 2

xn − 1 , x > 0.
2 2
2 ⋅ 4 ... 2n b g 2

Solution. Here

u
lim n = lim
2n + 2 1 1
⋅ =
b g 2

n →∞ u
n+1
n→∞ 2 + 1
n
2
x x b g
Hence by Ratio Test, the series converges if x < 1, and diverges if x > 1
Now for x = 1,
un
=
2n + 2 b g 2

un + 1 2n + 1
2
b g
Fu I 4 n 2 + 3n
∴ lim n
n →∞
GH u n

n +1
JK
− 1 = lim
n →∞
b2n + 1g 2
=1

Hence Raabe’s Test fails.

Let us now apply Gauss’s Test.

un FG IJ FG1 + 1 IJ
1
2 −2

un + 1 H K H 2n K
= 1+
n

F 2 1 IJ FG1 − 1 + 3
= G1 + +
IJ
H n n K H n 4n 2 2
+ ...
K
1 1 1
− =1+ + ... higher powers of
n 4n 2 n
so that by Gauss’s Test, the series diverges.
Hence, the series converges for x < 1, and diverges for x ≥ 1.

Note. We could get the result directly by Gauss’s Test, for

un FG
1 2n + 2 IJ 2
1 1/ x 1/4 x
un + 1
=
H
x 2n + 1 K =
x
+
n
− 2 + ...
n

where α = 1/x , β = 1/x .

Copyright © 2004. New Academic Science. All rights reserved.

Example 4.18. Test for convergence of the hypergeometric series

α ⋅β b g b g
α α +1 β β +1 2 α α +1 α +2 β β +1 β +2 3 b gb g b gb g
1+
1⋅γ
x+
b g
1⋅ 2γ γ + 1
x +
1⋅ 2 ⋅ 3 γ γ + 1 γ + 2
x + ...
b gb g
for all positive values of x ; α , β , γ being all positive.

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 97

Solution. It is a positive term series.

Ignoring the first term, which does not affect the behaviour of the series, we have
b g b g b
α α + 1 ... α + n − 1 β β + 1 ... β + n − 1 n g b g
un =
1⋅ 2 ⋅ 3 ... n b
γ γ + 1 ... γ + n − 1
x
g b g
so that

un b
n+1 γ +n 1 1 gb g
lim
n →∞
= lim
b =
un + 1 n → ∞ α + n β + n x x gb g
Hence, by Ratio Test, the series converges, if x < 1, and diverges, if x > 1.
For x = 1, we have
FG1 + 1 IJ FG1 + γ IJ FG1 + γ + 1 + γ IJ
un bn + 1gbγ + ng = H n K H n K = H n n K 2

un + 1
=
bα + ngbβ + ng FG1 + α IJ FG1 + β IJ FG1 + α + β + αβ IJ
H nK H nK H n n K 2

F γ + 1 + γ IJ LM1 − FG α + β + αβ IJ + FG α + β + αβ IJ + ...OP
= G1 +
2

H n n K MN H n n K H n n K PQ
2 2 2

1 + γ − α − β bα + β − γ gbα + β − 1g − αβ
=1+ + + ...
n n2
Hence by Gauss’s Test the series converges if 1 + γ − α − β > 1 or γ > α + β and diverges if
1 + γ − α − β ≤ 1 or γ ≤ α + β .
Thus for positive values of α , β , γ and x,
(i) for x < 1, the series converges,
(ii) for x > 1, the series diverges, and
(iii) for x = 1, the series converges if γ > (α + β ) and diverges if γ ≤ (α + β ).

EXERCISE
Test the convergence of the series:

2 x 32 ⋅ x 2 43 ⋅ x 3
1. 1 + + + + ...
2! 3! 4!
Copyright © 2004. New Academic Science. All rights reserved.

x 1 x3 1× 3 x 5 1× 3× 5 x 7
2. + + + + ...
1 2 3 2 ×4 5 2 ×4 × 6 7

3.
a+x
+
b
a + 2x g + ba + 3xg
2 3

+ ..., for x > 0.

1! 2! 3!
12 12 ⋅ 32 12 ⋅ 32 ⋅ 52
4. 2
+ 2 2 + 2 2 2 + ...
2 2 ⋅4 2 ⋅4 ⋅6

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 98 Principles of Real Analysis

5. Apply Cauchy’s Integral Test to test the convergence of the series.

∞ 1 ∞ 1
(i) ∑ 2
, (ii) ∑ 2
n=1 n +n n=1 n +1
¥ 1
6. Test the convergence of å
n=3 b
n log n log log n g p

Test the convergence of the series whose nth term is

bn!g x
b2ng!
n
7.

n!
8. xn
bn + 1g n

b
α α α +1 g
α α +1 α +2 b gb g
9. Prove that 1 + +
b
β β β +1
+
g
β β +1 β + 2b gb g
+ ..., where x, are positive, converges if β > α + 1, and

diverges if β ≤ α + 1.

ANSWERS
1. Converges if x < 1/e, diverges if x ≥ 1/e .
2. Converges if x £ 1, diverges if | x | > 1.
3. Converges for x < 1/e, diverges for x ≥ 1/e .
4. Diverges
5. (i) Convergent (ii) Convergent.
6. Converges for p > 1, diverges for p ≤ 1.
7. Converges for x < 4, diverges for x ≥ 4.
8. Converges for 0 < x < e, diverges for x ≥ e .

4.10 SERIES WITH ARBITRARY TERMS

So far we have considered series with positive terms only. We shall now discuss series with terms having
any sign whatsoever.
4.10.1 Alternating Series
A series whose terms are alternatively positive and negative, e.g.,
1 1 1
1- + - + ...
Copyright © 2004. New Academic Science. All rights reserved.

2 3 4
is called an alternating series.
Leibnitz Test. If the alternating series
b
u1 − u2 + u3 − u4 + ..., un > 0, ∀ ng
is such that

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 99

(i) un + 1 ≤ un , ∀ n , and

(ii) lim un = 0, then the series converges.

n ®¥

Let Sn = u1 – u2 + u3 – u4 + ... + (–1)n – 1 un

Now for all n,
S 2 n + 2 − S 2 n = u 2 n + 1 − u2 n + 2 ≥ 0
⇒ S2 n + 2 ³ S2 n
⇒ {S2 n } is a monotonic increasing sequence.
Again
S2n = u1 − u2 + u3 − ... + u2 n − 1 − u2n
= u1 − ( u2 − u3 ) − ( u4 − u5 ) − ... − (u2 n − 2 − u2 n − 1 ) − u2 n
But since un + 1 £ un , for all n, therefore, each bracket on the right is positive and hence
S2 n < u1 , ∀ n
Thus, the monotonic increasing sequence {S2n} is bounded above and is consequently convergent.
Let lim S2 n = S .
n →∞

We shall now show that the sequence {S2n + 1} also converges to the same limit S.
Now
S2n + 1 = S2n + u2 n + 1

∴ lim S2 n + 1 = lim S2 n + lim u2 n + 1

n →∞ n→∞ n →∞

But by condition (ii),

lim u2n + 1 = 0
n →∞

∴ lim S2n + 1 = lim S2n = S

n →∞ n →∞

Thus, the sequences {S2n} and {S2n + 1} both converge to the same limit S. We shall now show that
the sequence {Sn} also converges to S.
Let ε > 0 be given.
Since the sequences {S2n} and {S2n + 1} both converge to S, therefore ∃ positive integers m1, m2,
respectively, such that
Copyright © 2004. New Academic Science. All rights reserved.

S2 n − S < ε , ∀ n ≥ m1 ...(1)
and
S2 n + 1 − S < ε , ∀ n ≥ m2 ...(2)

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 100 Principles of Real Analysis

Thus from (1) and (2) we have

Sn − S < ε , ∀ n ≥ max 2m1 , 2m2 + 1 b g
⇒ {S n } converges to S

⇒ the series ∑ (−1) n − 1 un converges.

1 1 1 1
Example 4.19. Show that the series p − p + p − p + ... converges for p > 0.
1 2 3 4
Solution. Let un = 1/np.
Here
un + 1 ≤ un ; ∀ n
and
1
lim un = lim =0
n →∞ n →∞ np
( −1) n − 1
Hence by Leibnitz test, the alternating series ∑ converges.
np

4.10.2 Absolute Convergence

A series ∑ un is said to be absolutely convergent if the series ∑ un obtained on taking every term of the
given series with a positive sign is convergent, i.e., if the series ∑ un is convergent.

Conditional Convergence. A series which is convergent but is not absolutely convergent is called a
conditionally convergent series.
ILLUSTRATIONS
1. The series
1 1 1
1− 3
+ 3 − 3 + ... ,
2 3 4
is absolutely convergent because the series
1 1 1
1 + 3 + 3 + 3 + ... ,
2 3 4
obtained on taking every term of the given series with a positive sign, is convergent.
2. The series
Copyright © 2004. New Academic Science. All rights reserved.

1 1 1
1 - + - + ...
2 3 4
is convergent by Leibnitz test, but the series
1 1 1
1+ + + + ...
2 3 4
obtained on taking every term with a positive sign is divergent. Thus the given series is
conditionally convergent.

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 101

Theorem 4.9. Every absolutely convergent series is convergent.

Let ∑ un be absolutely convergent, so that ∑ un is convergent.
\ For any ε > 0, by Cauchy’s General Principle of convergence, ∃ a positive integer m such that

un + 1 + un + 2 + ... + un + p < ε , ∀n≥m∧ p≥1

Also for all n and p > 1,

un + 1 + un + 2 + ... + un + p ≤ un + 1 + un + 2 + ... + un + p < ε , ∀ n ≥ m ∧ p ≥ 1.

Hence, by Cauchy’s General Principle of convergence the series ∑ un converges.

Aliter. Let ∑ un be absolutely convergent, so that ∑ un is convergent.

Let an =
RSu , if u ≥ 0
n n
and bn =
RS −u , if u < 0
n n

T 0 , if u < 0
n T0 , if u ≥ 0 .
n

Then clearly,
an ≥ 0 , bn ≥ 0 , ...(i)
un = an − bn ...(ii)
and
un = an + bn ...(iii)
From (i) and (iii) it follows that
an £ un , bn £ un
Since ∑ un is convergent, therefore, by Comparison Test, both ∑ an and ∑ bn are convergent.

Hence, by Theorem 4.5, ∑ ( an − bn ) is convergent.

\ From (ii), it follows that ∑ un is convergent.

Remarks 1. The divergence of ∑ un does not imply the divergence of ∑ un .

(-1) n - 1
For example, if un = , we have seen above that ∑ un is divergent, whereas ∑ un is convergent.
n
2. The very great significance of the concept of Absolute Convergence is that the convergence of absolutely
convergent series is much more easy to recognise than that of conditionally convergent series—usually by
comparison with series of positive terms. In fact all the tests for positive term series become available for
Copyright © 2004. New Academic Science. All rights reserved.

the purpose. But this significance becomes all the more visible in the discussion of rearrangement of
series—so much so that we may operate on absolutely convergent series, precisely as we operate on sums
of a finite number of terms, whereas in the case of conditionally convergent series this in general is not
possible.

∞ sin nx
Example 4.20. Show that for any fixed value of x, the series ∑ is convergent.
n =1 n2

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 102 Principles of Real Analysis

sin nx sin nx
Solution. Let un = 2
, so that un =
n n2
sin nx 1 1
Now 2
≤ 2 , ∀ n and ∑ 2 converges.
n n n

sin nx
Hence, by Comparison Test, the series ∑ converges.
n2

sin nx
Since every absolutely convergent series is convergent, therefore Σ is convergent.
n2
Example 4.21. Show that the series
x2 x3
x+ + + ...
2! 3!
converges absolutely for all values of x.

xn
Solution. Let un = .
n!

un n+1
Now lim = lim → ∞ except when x = 0.
n →∞ un + 1 n→∞ x

So by Ratio Test, the series converges absolutely for all x except possibly zero.
But for x = 0 the series evidently converges absolutely.
Hence, the series converges absolutely for all values of x.

Note. Since for a convergent series ∑ un , lim un = 0 ,

n→∞

xn
∴ lim = 0 a useful result.
n →∞ n!

Example 4.22. Show that

b g b
m m − 1 ... m − n + 1 gx
bn − 1g!
n
lim = 0,
n →∞

where | x | < 1 and m is any real number.

Solution. Consider the series ∑ un , where
Copyright © 2004. New Academic Science. All rights reserved.

b g b
m m − 1 ... m − n + 1 gx n
un =
bn − 1g!

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 Infinite Series 103

Now
un n 1
lim = lim ⋅
n →∞ un + 1 n →∞ m−n x

1 1 1
= lim ⋅ =
m n →∞ x x
−1
n
Hence, the series ∑ un converges absolutely for x < 1
⇒ the series ∑ un converges for x < 1

Þ lim un = 0 , for x < 1

n ®¥

b g b
m m − 1 ... m − n + 1 gx
bn − 1g!
n
i.e., lim = 0 , if x < 1
n →∞

Note. The results of Example 4.21 and 4.22 are very useful.

Also see Examples 3.9 and 3.10 of Ch. 3.

EXERCISE
1. Show that the following series are convergent:
1 1 1
(i) 1 − + − + ...
2 3 4
1 1 1 1
(ii) − + 2 − 2 + ...
12 2 2 3 4
1 1 1
(iii) 1 − + − + ...
2 2 3 3 4 4
log 2 log 3 log 4
(iv) − 2 + − ...
22 3 42
2. Prove that the following series are absolutely convergent:

x x2
(i) 1 + + + ...
1! 2!
x2 x4 x6
Copyright © 2004. New Academic Science. All rights reserved.

(ii) 1 − + − + ...
2! 4! 6!
x 3 x5
(iii) x − + − ...
3! 5!

nr
3. Show that lim =0
n ®¥ xn

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.
 104 Principles of Real Analysis

4. Show that the following series are conditionally convergent:

(i) ∑
b−1g n +1

(ii) ∑
b−1g n +1

3n − 2
( −1) n + 1
5. Show that the series ∑ is absolutely convergent for p > 1, but conditionally convergent for 0 < p ≤ 1.
np
6. Show that the following series are absolutely convergent:

R U
(i) b g |S| n1 + bn +1 1g |V|
∑ −1
n −1

T 2
W 2

R| 1 U
(ii) ∑ b −1g S| n + bn +11g |V|
n−1

T 5 /2
W 5 /2

(iii) b g
∑ −1
n n+2
2n + 5

F 1 + b−1g I is divergent.
n −1
7. Show that the series ∑ GH n n JK
1 1 1
8. Show that the series 1 − + − + ... converges.
3⋅ 4 5⋅ 4 2 7 ⋅43
Copyright © 2004. New Academic Science. All rights reserved.

Principles of real analysis. (2004). New Academic Science.

Created from inflibnet-ebooks on 2022-06-02 15:19:33.