2 Series

Given a sequence (an ) of real numbers, we can forma new sequence (sn ) with
n
X
s n = a1 + · · · + an = ak .
k=1

Definition 21. Given a sequence (an ), an expression of the form ∞

P
k=1 aP
k is called a series,
∞
and sn is called the n-th partial sum of the series.
P∞We say that the series k=1 ak converges
to s ∈P R if sn → s, and in that case we write k=1 ak = s, and s is called the sum of the
series ∞ k=1 ak .
If (sn ) diverges, then we say that the series ∞
P
k=1 ak diverges. C

Here is an important property.

Theorem 22. If ∞
P
n=1 an converges, then an → 0.

Proof. Suppose sn → `. Then an+1 = sn+1 − sn → ` − ` = 0.

2.1 Some typical examples

1 1
Example 23. (1) The series + 2 + · · · converges to 1/9.
10 10
(2) For a ∈ R, consider the series ∞ n
P
n=1 a .
2 n
If a = 1, then sn = 1 + a + a + · · · + a = n + 1 → ∞. Then Suppose a 6= 1. Then

1 − an+1 1 an+1
1 + a + a2 + · · · + an = = − .
1−a 1−a 1−a
If |a| < 1, then |an+1 | = |a|n+1 → 0, and hence, in this case,

1 an+1 1
sn = − → .
1−a 1−a 1−a
P∞
If |a| ≥ 1, then |an+1 | = |a|n+1 ≥ 1 for all n ∈ N, and hence, an 6→ 0, so that n=1 an
diverges. Thus, we have proved that

1 + a + a2 + · · · converges iff |a| < 1 .

1 2 n
(3) The series + + ··· + + · · · diverges to ∞.
2 3 n+1
n
Note that for any n ∈ N, n+1
≥ 1/2. Hence, sn ≥ n/2 for all n ∈ N. Hence, (sn ) diverges
to ∞.

12
 1 1 1 1
(4) The series + + + · · · + + · · · diverges to ∞
1 2 3 n

Note that
1 1 1 1
s2n := + + + ··· + n
1 2 3 2
can be written as
1 1 1 1 1 1 1 1 1  1 1
+ + + + + ··· + + + + ··· + + n−1 + ··· + n .
1 2 3 4 5 8 9 10 16 2 +1 2
Note that
1 1
+ ≥ 1,
1 2
1 1 2
+ ≥ ,
3 4 22
1 1 1 23
+ + ··· + ≥ 4,
9 10 16 2
 1 1 2n−1
+ ··· + n ≥ ,
2n−1 + 1 2 2n
so that
2 22 23 2n−1 n−1 n+1
s2n ≥ 1 + 2
+ 3
+ 4
+ · · · + n
=1+ = .
2 2 2 2 2 2
Thus,
n+1
s2n ≥ ∀ n ∈ N.
2
This shows that (sn ) is unbounded, and hence it is divergent. In fact, it diverges to ∞.

1 1 1 1
(5) The series + 2 + 2 + · · · + 2 + · · · converges.
1 2 3 n

Clearly, the sequence (sn ) of its partial sums is monotonically incresing. Hence, it is
enough to show that it is bounded above. Note that
1 1 2 1
2
+ 2 ≤ 2 = ,
2 3 2 2
1 1 1 1 4 1
2
+ 2 + 2 + 2 ≤ 2 = 2,
4 5 6 7 4 2
n−1
1 1 1 2 1
n−1 2
+ n−1 2
+ · · · + n−1 2
≤ n−1 2
= n−1 .
(2 ) (2 + 1) (2 − 1) (2 ) 2
Hence sn ≤ 2 for all n ∈ N. Thus, (sn ) is monotonically increasing and bounded above, and
hence, it converges.

1 1 1 1
(6) For every p ≤ 1, the series + p + p + · · · + p + · · · diverges to ∞.
1 2 3 n

13
 (p) (p) (1) (1)
Let sn be its n-th partial sum. Since sn ≥ sn for all n ∈ N and (sn ) is unbounded,
(p)
(sn ) is also unbounded, and hence diverges to ∞.

1 1 1 1
(7) For every p ≥ 2, the series + p + p + · · · + p + · · · converges.
1 2 3 n
(p) (p) (2) (1)
Let sn be its n-th partial sum. Note that sn ≤ sn for all n ∈ N and (sn ) is bounded.
(p) (p)
Since (sn ) is monotonically increasing, it follows that (sn ) monotonically increasing and
bounded, and hence it converges.

1 1 1 1
(8) For every p > 1, the series + p + p + · · · + p + · · · converges.
1 2 3 n

Note that
Z n+1 Z n+1 Z n+1
1 1 dx 1 1
= dx ≤ ≤ dx = p
(n + 1)p n (n + 1)p n xp n n p n
⇒
k k Z n+1 k
X 1 X dx X 1
≤ ≤ .
n=1
(n + 1)p n=1 n xp n=1
np
But,
k Z n+1 Z k+1  −p+1 x=k+1  
X dx dx x 1 1 1
= = = 1− ≤ .
n=1 n xp 1 x p −p + 1 x=1 p−1 (k + 1) p−1 p−1

Hence,
k+1 k k+1 Z n+1
X 1 X 1 X dx 1
p
= p
≤ p
≤ .
n=2
n n=1
(n + 1) n=1 n x p − 1
Thus, the sequence of partial sums,which is already monotonically increasing, is bounded as
well. Hence, the given series converges. Thus, we have
∞
X 1
For p ∈ R, the series converges ⇐⇒ p > 1.
n=1
np

1 1 1 1
(9) The series 1+ + + + ··· + + · · · converges.
1! 2! 3! n!

Since
1 1
= n ∀ n ≥ 2.
n! 2
Hence,
1 1 1 1 1 1 1 1
+ + + ··· +
sn = 1 + ≤ 1 + + + 2 + · · · + n ≤ 3.
1! 2! 3! n! 1 2 2 2
Since (sn ) is monotonically increasing and bounded above, it converges.

14
 ∞
X 1
• The sum of the series is denoted by e.
n=0
n!
 1 n
• It can be shown that lim 1+ exists and it is equal to e (See the book2 ).
n→∞ n

2.2 Some tests for convergence

2.2.1 Integral test
The following theorem can be proved using the techniques adopted in Example 23 (8).

Theorem 24. (Integral test) If f : [1, ∞) → [0, ∞) is monotonically decreasing, then

∞
X Z n 
f (n) converges ⇐⇒ f (x)dx converges.
n=1 1

2.2.2 Comparison test

Theorem 25. (Comparison test) Suppose (an ) and (bn ) are sequences of non-negative
terms. Suppose there exists k ∈ N such that an ≤ bn for every n ≥ k.

1. If ∞
P P∞
n=1 bn converges, then n=1 an converges.

2. If ∞
P P∞
n=1 an diverges, then n=1 bn diverges.

Proof. Hint: Use monotone convergence theorem, by observing that (i) the sequence of
partial sums is monotonically increasing and (ii) every convergent sequence is bounded.

Theorem 26. (Limit comparison test) Suppose (an ) and (bn ) are sequences of positive
an
terms. Suppose lim = `.
n→∞ bn

1. If 0 < ` < ∞, then ∞

P P∞
n=1 an converges ⇐⇒ n=1 bn converges.

2. If ` = 0, then ∞
P P∞
n=1 bn converges ⇒ n=1 an converges.

3. If ` = ∞, then ∞
P P∞
n=1 an converges ⇒ n=1 bn converges.

Proof. Let ε > 0. Then there exists N ∈ N such that

an
`−ε< < ` + ε ∀ n ≥ N.
bn
(1) Suppose α > 0 and 0 < ε < α/2. Then

` an 3`
< < ∀ n ≥ N.
2 bn 2
2
See Example 1.25, page 30 in Calculus of One Variable by M.T. Nair

15
Hence
` 3`
bn < an < bn ∀ n ≥ N.
2 2
Therefore, the result follows from Theorem 25.
(2) Suppose ` = 0. Then
an
0< < ε ∀n ≥ N
bn
so that
0 < an < εbn ∀ n ≥ N.
Therefore, the result follows from Theorem 25.
∞
X 2n + n 2n +n
Example 27. Consider the series . Clearly, 3n −n
> 0 for all n ∈ N. Also,
n=1
3n −n

2n + n  2 n  1 + n/2n 
= .
3n − n 3 1 − n/3n
n 1+n/2n
Note that (verify) 1+n/2
1−n/3n
→ 1 as n → ∞. Hence, there exists k ∈ N such that 1−n/3n
< 3
2
for all n ≥ k. Therefore,

2n + n  2 n  1 + n/2n  3  2 n
= ≤ ∀ n ≥ k.
3n − n 3 1 − n/3n 2 3
P∞  2  n P∞ 2n +n
Since n=1 3
converges, it follows that n=1 3n −n also converges.

2.2.3 Ratio tests

Theorem 28. Suppose (an ) is a sequence of positive terms.
an+1
(i) Suppose
P∞ there exists k ∈ N and 0 < c < 1 such that an
≤ c for all n ≥ k. Then
n=1 an converges.
an+1 P∞
(ii) Suppose there exists k ∈ N such that an
≥ 1 for all n ≥ k. Then n=1 an diverges.

n+1−k
P∞ n
Proof.
P∞ (i) a n+1 ≤ c a k for all n ≥ k. Since n=k c converges, by comparison test,
n=k an also converges.
(ii) an+1 ≥ ak > 0 for all n ≥ k. Hence, an 6→ 0. Therefore, ∞
P
n=k an diverges.

Theorem 29. (d’Alembert’s atio test) Suppose (an ) is a sequence of positive terms, a and
an+1 n+1

lim = `, where either ` is a non-negative real number or ` = ∞, which means
n→∞ an an
diverges to ∞.
P∞
(i) If ` < 1, then n=1 an converges.
P∞
(ii) If ` > 1, then n=1 an diverges.

16
  
an+1
If an
diverges to ∞, then the conclusion in (ii) holds.

Proof. Suppose 0 ≤ ` < ∞. Let ε > 0 and let N ∈ N be such that ` − ε < an+1 an
< ` + ε for
all n ≥ N .
(i) Tale ε > 0 such that ` + ε < 1 and apply Theorem 28 (1).
(ii) Tale ε > 0 such that ` − ε ≥ 1 and apply Theorem 28 (2).
an+1
P∞an ≥ 1 for all n ≥ N . Hence, an+1 ≥ aN > 0
If ` = ∞, then there exists N ∈ N such that
for all n ≥ N so that an 6→ 0; consequently, n=1 an diverges to ∞.

2.2.4 Root test

Theorem 30. (Cauchy’s root test) Suppose (an ) is a sequence of positive terms,  a and
1/n n+1
lim an = `, where either ` is a non-negative real number or ` = ∞, which means
n→∞ an
diverges to ∞.
P∞
(i) If ` < 1, then n=1 an converges.
P∞
(ii) If ` > 1, then n=1 an diverges.

Proof. Use similar techniques as in the proof of d’Alembert’s ratio test.

Exercise 31. If ` =P1, then d’Alembert’s ratio test and Cauchy’s ratio test fail. One may
∞ 1 P∞ 1
consider the series n=1 n and n=1 n . C
∞
X xn
Example 32. For any x ∈ R, the series converges. This follows from ratio test, since
n=0
n!

xn+1 /(n + 1)! x

n
= → 0.
x /n! n+1
∞
X xn
The limit of the above series is denoted by exp(x) or by the notation ex .
n=1
n!

It can be shown that

• exp(x + y) = exp(x) exp(y);

• The function f (x) := exp(x) is strictly monotonically increasing, that is, x < y implies
exp(x) < exp(y).

• The function f (x) := exp(x) is continuous on R and it is one-one and onto [0, ∞).
an+1
Theorem 33. Let (an ) be a sequence of positive real numbers such that lim = `. If
n→∞ an
0 < ` < ∞, then lim a1/n
n = `.
n→∞

17
Proof. Let ε > 0 be given and let k ∈ N be such that
an+1
`−ε< < ` + ε ∀ n ≥ k,
an
that is,
(` − ε)an < an+1 < (` + ε)an ∀ n ≥ k.
Taking ε small enough such that ` − ε > 0, we obtain

(` − ε)j ak < ak+j < (` + ε)j ak ∀ j ∈ N.

Writing k + j = n, we have

(` − ε)n−k ak < an < (` + ε)n−k ak ∀ n > k,

that is,
ak n n ak
(` − ε) < an < (` + ε) ∀ n > k,
(` − ε)k (` − ε)k
so that h ak i1/n 1/n
h a
k
i1/n
(` − ε) < an < (` + ε) ∀ n > k,
(` − ε)k (` − ε)k
h a i1/n
k
Recall that for any x > 0, lim x1/n = 1. Hence, lim = 1. Let k1 ∈ N be
n→∞ n→∞ (` − ε)k
k1 ≥ k and
h a i1/n
k
1−ε< < 1 + ε ∀ n ≥ k1 .
(` − ε)k
Thus,
(1 − ε)(` − ε) < a1/n
n < (` + ε)(1 + ε) ∀ n ≥ k1 ,
that is,
` − ε(` + 1 + ε) < a1/n
n < ` + ε(` + 1 + ε) ∀ n ≥ k1 .
Hence, lim a1/n
n exists and it is equal to `.
n→∞

It can happen that root test can be applied whereas ratio test many not be applied.

Example 34. Consider the series

1 1 1 1 1 1 1 1 1 1
+ + + + + + · · · + + + + + ··· .
22 2 24 23 26 25 2n+1 2n 2n+3 2n+2
Hence, in this case,
a2n a2n+1 1
= 2 and = ,
a2n−1 a2n 2
1 1 1
whereas a1/n
n ∈ { 1+ 1 , } so that lim an1/2 = .
2 n 2 n→∞ 2
Thus, ratio test cannot be applied, whereas root test shows that the series converges.

18
2.3 Absolutely convergent series and alternating series
P∞
Theorem 35. Let (an ) be a sequence of real numbers such that n+1 |an | converges. Then
P∞
n+1 an converges.

For its proof we shall make use of the following result, whose proof we omit:
Theorem 36. (Cauchy’s criterion of convergence of sequences) Let (an ) be a sequence
of real numbers. Then (an ) converges iff for every ε > 0, there exists N ∈ N such that
|an − am | < ε ∀ n, m ≥ N.
Definition 37. A sequence (an ) of real numbers is called a Cauchy sequence if for every
ε > 0, there exists N ∈ N such that
|an − am | < ε ∀ n, m ≥ N. C
Thus, Theorem 36 says that

(an ) converges ⇐⇒ (an ) is a Cauchy sequence.

It is to be observed that:

Every convergent sequence is a Cauchy sequence.

Indeed, if an → a, then for any n, m ∈ N,

|an − am | ≤ |an − a| + |a − am |.
For any give ε > 0, let N ∈ N be such that |an − a| < ε/2 for all n ≥ N . Then from the
above,
ε ε
|an − am | ≤ |an − a| + |a − am | < + = ε ∀ n, m ≥ N.
2 2
Definition 38. A series n+1 an is said to be absolutely convergent if ∞
P∞ P
n+1 |an | con-
verges. C
Thus, Theorem 35 says that:

Every absolutely convergent series is convergent.

P∞ P∞
Proof of Theorem 35. Let sn and s0n be the n-th partial sums of n=1 an and n=1 |an |,
respectively. Then for n > m, we have
|sn − sm | ≤ s0n − s0m . (∗)
P∞
Since n=1 |an | assumed to be convergent, (s0n ) is a Cauchy sequence, and by the relation
(∗), (sn ) is also a Cauchy sequence. Hence, by Theorem 36, (sn ) converges.

19
 ∞
X xn
Example 39. For any x ∈ R, the series is absolutely convergent, and hence converges
n=1
n!
as well.

Definition 40. The function exp : R → R defined by

∞
X xn
exp(x) = , x ∈ R,
n=1
n!

is called the exponential function. C

It can be shown that3

exp(x + y) = exp(x) exp(y) ∀ x, y ∈ R.

Note that exp(x) > 0 for x > 0, and exp(0) = 1. Hence,

exp(x) exp(−x) = exp(0) = 1

so that
1
exp(x) 6= 0 and exp(−x) = ∀ x ∈ R.
exp(x)
In particular, exp(x) > 0 for every x ∈ R. It can also be proved that4

the function x 7→ exp(x) is continuous, strictly increasing, exp(x) → ∞ as x → ∞

and exp(x) → 0 as x → −∞.

Hence, for every y > 0, there exists a unique x ∈ R such that

exp(x) = y.

This x is called the logarithm of y, denoted by log(y).

Definition 41. A series of the form ∞ n+1

P
n=1 (−1) an , where an > 0 for all n ∈ N, is called
an alternating series. C
∞
X (−1)n
By Theorem 35, for any p > 1, the series is absolutely convergent, and hence
np
n=1
converges as well. Does the above series converge for 0 < p ≤ 1? The answer is in affirmative,
due to the following theorem.

Theorem 42. (Leibnitz test for convergence of alternating series) P∞If (an ) is a strictly
n+1
decreasing sequence of positive real numbers such that an → 0, then n=1 (−1) an con-
verges.
3
See page 110, Theorem 2.29, in Calculus of One variable, Anne Books, Pvt. Ltd, 2014, by M.T. Nair.
4
See page 110, Theorem 2.29, in Calculus of One variable, Anne Books, Pvt. Ltd, 2014, by M.T. Nair.

20
Proof. Let (an ) be a strictly decreasing sequence
P∞ of positive real numbers such that an → 0.
n
Let sn be the n-th partial sum of the series n=1 (−1) an . Then for every n ∈ N,
s2n = a1 − a2 + a3 − a4 + · · · + a2n−1 − a2n
= (a1 − a2 ) + (a3 − a4 ) + · · · + (a2n−1 − a2n )
= a1 − (a2 − a3 ) − (a4 − a5 ) · · · − (a2n−2 − a2n−1 ) − a2n .
From the above it follows that (s2n is a monotonically increasing and bounded above by a1 .
Hence, it converges, say a2n → s. Then we have
s2n+1 = s2n + a2n+1 → s + 0 = s.
Now, let ε > 0 be given and let N1 , N2 ∈ N be such that
|s2n − s| < ε ∀n ≥ N1 and |s2n+1 − s| < ε ∀ n ≥ N2 .

∞
X (−1)n
By the above theorem, the series is convergent. In fact:
n=1
n
∞
X (−1)n
For any p > 0, converges.
n=1
np
Remark 43. Does the series
1 1 1 1 1 1 1
1+ − − + + − − + ···
2 3 4 5 6 7 8
∞
X
converge? Note that the above series can be written as (an + bn ), where
n=1

(−1)n+1 (−1)n+1
an = , bn = .
2n − 1 2n
∞ ∞ ∞ h
X (−1)n+1 X (−1)n+1 X (−1)n+1 (−1)n+1 i
Since and converge, it follows that + also
n=1
2n − 1 n=1
2n n=1
2n − 1 2n
converges. C

2.4 Rearrangements
Suppose, for a given sequence (an ) of real numbers, the series ∞
P
n=1 an converges. If P
(bn ) is a
sequence obtained from (an ) by rearranging its terms, then is it true that the series ∞ n=1 bn
is also convergent?
In other words, ifP ∞
P
n=1 an converges and if σ : N → N is a bijective function, then is it
true that the series ∞ n=1 aσ(n) is also convergent?
The answer is in affirmative if the terms of the series are all positive (Exercise). In fact,
in this case, the sums of all the rearranged series are the same.
From the above discussion, it follows that:

21
 ∞
X ∞
X
If an is absolutely convergent, then an = s and all the
n=1 n=1
rearranged series are convergent.
However, if the series is conditionally convergent, then different rearrangements of the series
can have different sums.
Example 44. Consider the series
1 1 1 1 1 1 1
1 − + − + − + − + ··· . (1)
2 3 4 5 6 7 8
We know that this series converge. Now consider the series
1 1 1 1 1 1 1 1
1 − − + − − + ··· + − − + ··· . (2)
2 4 3 6 8 2n − 1 4n − 2 4n
If sn and s0n are the n-th partial sums of the series in (1) and (2), respectively, and if s is
the sum of the series in (1), then it can be shown that5
s2n
s03n = ∀ n ∈ N,
2
so that
s2n s
s03n = → ,
2 2
0 0 1 s
s3n+1 = s3n + → ,
2n + 1 2
1 1 s
s03n+2 = s03n + − → .
2n + 1 4n − 2 2
0
Therefore (why?) sn → s/2. Thus, the series in (1) and (2) have different sums.

Question: If ∞
P P∞
n=1 an is convergent, then is it true that every series obtained from n=1 an
by rearranging the terms also convergent?
Not necessarily. Look at the following example.
Example 45. Consider the convergent alternating series
1 1 1 1 1 1 1
1 − + − + − + − + ··· .
2 3 4 5 6 7 8
Consider the following rearranged series:
 1 1 1 1 1 1 1
1+ − + + + + −
3 2 5 7 9 11 4
1 1 1 1 1 1 1
+ + + + + + −
13 15 17 19 21 23 6
..
.
 1 1 1
+ + ··· + −
2(n − 1)n − 1 2n(n + 1) − 1 2n
+
..
.

5
See page 59, Example 153 in Calculus of One variable, Anne Books, Pvt. Ltd, 2014, by M.T. Nair.

22
If sn is the n-th partial sum of the above series, then we see that
 1 1
sn(n+2) = 1 + −
3 2
1 1 1 1 1
+ + + + −
5 7 9 11 4
1 1 1 1 1 1 1
+ + + + + + −
13 15 17 19 21 23 6
..
.
 1 1 1
+ + ··· + − .
2(n − 1)n − 1 2n(n + 1) − 1 2n

Note that 2n(n + 1) − 1 = [2(n − 1)n − 1] + 4n so that there are 2n number of terms in the
1 1
last bracket and, since 2n(n+1)−1 ≥ 2n(n+1) we have
 1 1  2n 1
+ ··· + ≥ = .
2(n − 1)n − 1 2n(n + 1) − 1 2n(n + 1) n+1

Thus,
1 1 1  1 1 1
sn(n+2) ≥ + ··· +
+ − + + ··· +
2 3 n+1 2 4 2n
1 1 1  11 1 1
= + + ··· + − + + ··· +
2 3 n+1 2 1 2 n
1 1  1 1
= + ··· + + −
2 n n+1 2

Hence, sn(n+2) → ∞ as n → ∞, so that the rearranged series diverges to ∞.

P∞
Question:
P∞ If n=1 an is absolutely convergent, then is it true that all the series obtained
from n=1 an with terms rearranged converge to the same sum?

23