INDIAN INSTITUTE OF TECHNOLOGY ROORKEE

SEQUENCE AND SERIES OF REAL NUMBERS

UADAY SINGH

DEPARTMENT OF MATHEMATICS
 SEQUENCES
∞
Definition: A sequence 𝑆 = {𝑠𝑛 } 𝑛=1 of real numbers is a function 𝑓 from ℕ
(the set of natural numbers) to ℝ (the set of real numbers), that is, 𝑓: ℕ → ℝ.
∞
– We usually denote a sequence by {𝑠𝑛 } 𝑛=1 or {𝑠𝑛 }, where 𝑠𝑛 = 𝑓 𝑛 is the 𝑛𝑡ℎ
term of the sequence.
– We write 𝑠1 = 𝑓 1 , 𝑠2 = 𝑓 2 , … , 𝑠𝑛 = 𝑓(𝑛).
– The set {𝑠1 , 𝑠2 , … , 𝑠𝑛 } is called the range set of the sequence.
∞
– For example, 𝑓 𝑛 = 𝑛2 can be written as {𝑛2 } 𝑛=1 or {𝑛2 } or 𝑠𝑛 = 𝑛2 with the
range set {1, 4, 9, … , 𝑛2 , … }.
∞
– The sequence {(−1)} 𝑛=1 is the sequence −1, 1, −1, 1, … with the range set
{−1, 1}.

∞
−1 𝑛 , if 𝑛 is even
– For the sequence {𝑠𝑛 } 𝑛=1 defined by 𝑠𝑛 = ൝ 𝑛−1 , the range set of
, if 𝑛 is odd
2
the sequence is {0, 1, 2,3, … }.
– Sometimes a sequence is defined by recurrence relations also. For example,
𝑠1 = 1, 𝑠𝑛 = 2𝑠𝑛−1 + 1 for 𝑛 ≥ 2. The range of this sequence is {1,3,7,15,31, … }.

2
 CONVERGENCE OF A SEQUENCE
1 ∞ 1 1
• Consider the sequence 𝑛=1
whose terms are 1, , , …
𝑛 2 3
• This sequence has a property that its terms come closer and closer to 0 as
𝑛 increases, and we say that the sequence converges to 0.
• In general, if the terms of a sequence {𝑠𝑛 } come arbitrarily close to a real
number 𝑙 as 𝑛 increases, we say that the sequence {𝑠𝑛 } converges to 𝑙.
• This fact is written as lim 𝑠𝑛 = 𝑙 or 𝑠𝑛 → 𝑙 as 𝑛 → ∞.
𝑛→∞
3+𝑛 1
• The sequence converges to .
5+2𝑛 2
DEFINITION: A sequence of real numbers {𝑠𝑛 } converges to a real number 𝑙 if,
for any given 𝜖 > 0 there is a positive integer 𝑛0 such that
𝑠𝑛 − 𝑙 < 𝜖, (1)
for all 𝑛 ≥ 𝑛0 . The number 𝑙 is called the limit of the sequence {𝑠𝑛 }, and we
write lim 𝑠𝑛 = 𝑙.
𝑛→∞
– The condition (1) implies that 𝑙 − 𝜖 < 𝑠𝑛 < 𝑙 + 𝜖 for all 𝑛 ≥ 𝑛0 and 𝑠𝑛 − 𝑙 → ∞ as 𝑛 → ∞.
– This also mean that there are infinite number of terms of the sequence after 𝑠𝑛 0 which lie
in an arbitrary small interval containing 𝑙
3
 CONVERGENCE OF A SEQUENCE
• A sequence which is not convergent is called divergent.
3𝑛+4 3 2
EXAMPLE 1: Consider the sequence 𝑠𝑛 = = + .
2𝑛 2 𝑛
3 3 3 2 3 1 3 2 3
▪ We observe that 𝑠1 = + 2, 𝑠2 = + 1, 𝑠3 = + , 𝑠4 = + , 𝑠5 = + , … and 𝑠𝑛 →
2 2 2 3 2 2 2 5 2
as 𝑛 → ∞.
3
▪ It suggests that limit of the sequence is = 𝑙.
2
3𝑛+4 3 2 2
▪ Now let for any 𝜖 > 0, 𝑠𝑛 − 𝑙 = − = < 𝜖, if 𝑛 > .
2𝑛 2 𝑛 𝜖
2 3𝑛+4 3
▪ For any given 𝜖 > 0, we can choose any 𝑛0 > so that − < 𝜖 for 𝑛 ≥ 𝑛0 (existence
𝜖 2𝑛 2
of 𝑛0 ).
▪ It can be noted that 𝑛0 depends on the choice of 𝜖.
2
▪ For 𝜖 = 0.01, 𝑛0 can be any integer greater than = 200, i.e.,
0.01
3𝑛+4 3
− < 0.01 for 𝑛 > 200.
2𝑛 2
2 3𝑛+4 3
▪ For 𝜖 = 1, 𝑛0 can be any integer greater than = 2, i.e., − < 1 for 𝑛 > 2
1 2𝑛 2
3𝑛+4 3
▪ Hence, lim 𝑠𝑛 = lim = .
𝑛→∞ 𝑛→∞ 2𝑛 2
3𝑛+4 3
▪ The sequence is convergent and converges to .
2𝑛 2
4
CONVERGENCE OF A SEQUENCE
1+2𝑛
EXAMPLE 2: Consider the sequence 𝑠𝑛 = .
3+4𝑛
1
1+2𝑛 +2 1
▪ It can be shown that lim 𝑠𝑛 = lim = lim 𝑛3 = = 𝑙.
𝑛→∞ 𝑛→∞ 3+4𝑛 𝑛→∞ +4 2
𝑛
1+2𝑛 1 1 1 1
▪ Now let for any 𝜖 > 0, 𝑠𝑛 − 𝑙 = − = < < 𝜖, if 𝑛 > 𝑛0 > .
3+4𝑛 2 2(3+4𝑛) 8𝑛 8𝜖
1 1+2𝑛 1 2𝑛 2 2
▪ If one chose 𝑙 = , then 𝑠𝑛 − 𝑙 = − = = 3 > for all 𝑛 ≥ 1.
3 3+4𝑛 3 3+4𝑛 +4 7
𝑛
2 1+2𝑛 1 2
▪ We cannot find any 𝑛0 for 𝜖 = such that − < for 𝑛 > 𝑛0 .
7 3+4𝑛 3 7
1 1+2𝑛
▪ Therefore, cannot be a limit of the sequence .
3 3+4𝑛

𝑛
EXAMPLE 3: Show that for 0 < 𝑐 < 1, lim 𝑐 = 0.
𝑛→∞
𝜖
▪ Here 𝑐 𝑛 − 0 = 𝑐 𝑛
< 𝜖 iff 𝑐 𝑛
< 𝜖 or 𝑛 log 𝑐 < log 𝜖 or −𝑛𝛿 < 𝜖 or 𝑛 > − .
𝛿
log 𝜖
▪ We can choose 𝑛0 > so that 𝑐 𝑛 − 0 < 𝜖 for all 𝑛 > 𝑛0 .
log 𝑐

EXAMPLE 4: Show that sequence { −1 𝑛 } has no limit.

▪ Here the terms of the sequence oscillates from −1 and 1. It appears that there is no
number 𝑙 to which all the terms (after a particular term) are arbitrarily close.
5
CONVERGENCE OF A SEQUENCE
𝑛
▪ For if, lim −1 = 𝑙, then for 𝜖 = 1 > 0 there is some positive integer 𝑛0 such that
𝑛→∞
𝑛
−1 − 𝑙 < 1 for all 𝑛 ≥ 𝑛0 .
▪ Thus −1 − 𝑙 < 1 for all odd 𝑛 ≥ 𝑛0 and 1 − 𝑙 < 1 for all even 𝑛 ≥ 𝑛0 .
or −2 < 𝑙 < 0 and 0 < 𝑙 < 2, which is a contradiction.

THEOREM: The limit of a convergent sequence is unique.

Proof: Suppose to the contrary a convergent sequence {𝑠𝑛 } has two distinct
𝑙 −𝑙
limits 𝑙1 and 𝑙2 . If we choose 𝜖 = 2 1 > 0, then there exist positive integers
2
𝑛1 and 𝑛2 such that
𝑙2 −𝑙1 𝑙2 −𝑙1
𝑠𝑛 − 𝑙1 < for all 𝑛 ≥ 𝑛1 and 𝑠𝑛 − 𝑙2 < for all 𝑛 ≥ 𝑛2 .
2 2
If 𝑛0 = 𝑚𝑎𝑥{𝑛1 , 𝑛2 }, then we can write
𝑙2 −𝑙1 𝑙2 −𝑙1
𝑠𝑛 − 𝑙1 < and 𝑠𝑛 − 𝑙2 < for all 𝑛 ≥ 𝑛0 .
2 2
Therefore, 𝑙1 − 𝑙2 = 𝑙1 − 𝑠𝑛0 + 𝑠𝑛0 − 𝑙2 ≤ 𝑙1 − 𝑠𝑛0 + 𝑠𝑛0 − 𝑙2 < 𝑙2 − 𝑙1 ,
which is a not possible. Hence, 𝑙1 and 𝑙2 must be equal.

6
LIMIT THEOREMS FOR CONVERGENT SEQUENCES
Bounded Sequence: A sequence {𝑠𝑛 } is said to be bounded if there is a
positive constant 𝑀 such that 𝑠𝑛 ≤ 𝑀 for all 𝑛 ∈ ℕ.
▪ A sequence {𝑠𝑛 } is said to be unbounded if there for any positive constant 𝑀 there exists
𝑛 ∈ ℕ such that 𝑠𝑛 > 𝑀.
▪ The sequence {2 sin 𝑛 − 3 cos 𝑛} is bounded by 5 since
2 sin 𝑛 − 3 cos 𝑛 ≤ 2 sin 𝑛 + 3 cos 𝑛 ≤ 5 for all 𝑛 ∈ ℕ.
▪ The sequence {2𝑛 } is unbounded as 𝑠𝑛 = 2𝑛 > 𝑛 for all all 𝑛 ∈ ℕ
𝑛𝑛 , if 𝑛 ≤ 106
6
▪ The sequence {𝑠𝑛 } defined by 𝑠𝑛 = ൞ is bounded by 1010
1
, if 𝑛 > 106
𝑛

𝑛
, if 𝑛 is divisible by 106
106
▪ The sequence {𝑠𝑛 } defined by 𝑠𝑛 =
1
, otherwise
𝑛

is unbounded since for any 𝑀 > 0, we can choose a positive integer 𝑁 = 𝑀 + 1 106
such that 𝑠𝑁 = 𝑀 + 1 > 𝑀.

7
LIMIT THEOREMS FOR CONVERGENT SEQUENCES
THEOREM: A convergent sequence is bounded.
Proof: Let {𝑠𝑛 } be a sequence converging to 𝑙. For 𝜖 = 1, we can find some
positive integer 𝑛0 such that 𝑠𝑛 − 𝑙 < 𝜖 = 1 for all 𝑛 ≥ 𝑛0 .
⇒ 𝑠𝑛 − 𝑙 ≤ 𝑠𝑛 − 𝑙 < 1 for all 𝑛 ≥ 𝑛0 .
⇒ 𝑠𝑛 < 𝑙 + 1 for all 𝑛 ≥ 𝑛0 .
If 𝑀 = { 𝑙 + 1, 𝑠1 , 𝑠2 , … , 𝑠𝑛0−1 }, then 𝑠𝑛 < 𝑀 for all 𝑛 ∈ ℕ. Hence, {𝑠𝑛 } is a
bounded sequence.
The converse of the theorem is not true, that is, every bounded sequence is
not bounded. For example, the sequence { −1 𝑛 } is bounded but not
convergent.
THEOREM: Let {𝑠𝑛 } and {𝑠𝑛 } be two convergent sequence such that
lim 𝑠𝑛 = 𝑙 and lim 𝑡𝑛 = 𝑙′. Then
𝑛→∞ 𝑛→∞
(i) lim 𝑐𝑠𝑛 = 𝑐𝑙, 𝑐 ∈ ℝ (ii) 𝑙𝑖𝑚 𝑠𝑛 ± 𝑡𝑛 = 𝑙 ± 𝑙′ (iii) 𝑙𝑖𝑚 𝑠𝑛 𝑡𝑛 = 𝑙𝑙′
𝑛→∞ 𝑛→∞ 𝑛→∞
(iv) 𝑙𝑖𝑚 𝑠𝑛 /𝑡𝑛 = 𝑙/𝑙′, provided 𝑙′ ≠ 0 (v) lim 𝑐 + 𝑠𝑛 = 𝑐 + 𝑙, 𝑐 ∈ ℝ
𝑛→∞ 𝑛→∞
𝑝
(vi) lim 𝑠𝑛 = 𝑙𝑝 for 𝑝 ∈ ℕ.
𝑛→∞

8
CONVERGENCE OF SEQUENCE
Proof (iv): We can write
𝑠𝑛 𝑙 𝑠𝑛 𝑙′ −𝑡𝑛 𝑙 𝑠𝑛 𝑙′ −𝑙𝑙 ′ +𝑙𝑙′−𝑡𝑛 𝑙 𝑙 ′ 𝑠𝑛 −𝑙 + 𝑙 𝑡𝑛 −𝑙 ′ 𝑠𝑛 −𝑙 𝑙 𝑡𝑛 −𝑙′
− = = ≤ = + . (1)
𝑡𝑛 𝑙′ 𝑙′𝑡𝑛 𝑙′𝑡𝑛 𝑡𝑛 𝑙′ 𝑡𝑛 𝑡𝑛 𝑙 ′

Now, lim 𝑡𝑛 = 𝑙′ implies that for any 𝜖 > 0 and for 𝑙 ′ 2

> 0, there are positive integers
𝑛→∞
𝑛1 and 𝑛2 such that
2
𝜖 𝑙′ 𝑙′
𝑡𝑛 − 𝑙′ < 4( 𝑙 +1) whenever 𝑛 ≥ 𝑛1 and 𝑙′ − 𝑡𝑛 ≤ 𝑡𝑛 − 𝑙′ < for all 𝑛 ≥ 𝑛2
2
𝜖 𝑙′ 𝑙′
⇒ 𝑡𝑛 − 𝑙′ < and 𝑡𝑛 > for all 𝑛 ≥ 𝑛3 = max{𝑛1 , 𝑛2 }. (2)
2 2
Now, lim 𝑠𝑛 = 𝑙 implies that for 𝜖 > 0, there is a positive integer 𝑛4 such that
𝑛→∞
𝜖 𝑙′
𝑠𝑛 − 𝑙′ < whenever 𝑛 ≥ 𝑛4 . (3)
4
From (1), (2) and (3), we have for 𝜖 > 0, there is a positive integer 𝑛0 = max{𝑛3 , 𝑛4 }
such that
2
𝑠𝑛 𝑙 𝜖 𝑙′ 2 𝑙 𝜖 𝑙′ 2 𝜖 𝜖
− < + = 2 + 2 = 𝜖 for all 𝑛 ≥ 𝑛0 .
𝑡𝑛 𝑙′ 4 𝑙′ 𝑙 ′ 4( 𝑙 +1) 𝑙 ′
𝑠𝑛 𝑙
Hence, lim = 𝑙′ . (Rest of the proofs are H.W.)
𝑛→∞ 𝑡𝑛

9
 CONVERGENCE OF SEQUENCE
𝑠𝑛 6−𝑛 2
EXAMPLE 5: Consider 𝑢𝑛 = = . Both 𝑠𝑛 and 𝑡𝑛 are divergent, but lim 𝑢𝑛 =
𝑡𝑛 1+6𝑛 2 𝑛→∞
𝑠𝑛 6−𝑛 2 1
lim = lim = 36 .
𝑛→∞ 𝑡𝑛 𝑛→∞ 1+6𝑛 2

THEOREM: If {𝑥𝑛 } and {𝑦𝑛 } are two convergent sequences such that lim 𝑥𝑛 = 𝑎,
𝑛→∞
lim 𝑦𝑛 = 𝑏 and 𝑥𝑛 ≤ 𝑦𝑛 for all 𝑛 ∈ ℕ, then 𝑎 ≤ 𝑏.
𝑛→∞
𝑎−𝑏
Proof: Suppose 𝑎 > 𝑏. Then for 𝜖 = > 0, we have a positive integers 𝑛1 and 𝑛2
2
𝑎−𝑏 𝑎−𝑏
such that 𝑥𝑛 − 𝑎 < 2
or − 2
< 𝑥𝑛 − 𝑎 for all 𝑛 ≥ 𝑛1
𝑎−𝑏 𝑎−𝑏
and 𝑦𝑛 − 𝑏 < or 𝑦𝑛 − 𝑏 < 2 for all 𝑛 ≥ 𝑛2 .
2
Thus for 𝑛0 = max{𝑛1 , 𝑛2 }, we have
𝑎−𝑏 𝑎+𝑏 𝑎−𝑏
𝑦𝑛 < +𝑏 = =𝑎− < 𝑥𝑛 .
2 2 2
This contradicts the given hypothesis 𝑥𝑛 ≤ 𝑦𝑛 . Hence, 𝑎 ≤ 𝑏.
COROLLARY: If 𝑥𝑛 → 𝑐 (sequence {𝑥𝑛 } converges to 𝑐) and 𝑎 ≤ 𝑥𝑛 ≤ 𝑏, then
𝑎≤𝑐≤𝑏

10
CONVERGENCE OF SEQUENCE
THEOREM: If {𝑥𝑛 }, {𝑦𝑛 } and {𝑧𝑛 } are three convergent sequences such that lim 𝑥𝑛 = 𝑎 =
𝑛→∞
lim 𝑧𝑛 , and 𝑥𝑛 ≤ 𝑦𝑛 ≤ 𝑧𝑛 for all 𝑛 ∈ ℕ, then lim 𝑦𝑛 = 𝑎.
𝑛→∞ 𝑛→∞
Proof: For any 𝜖 > 0, we have a positive integers 𝑛1 and 𝑛2 such that
𝑥𝑛 − 𝑎 < 𝜖 for all 𝑛 ≥ 𝑛1 and 𝑧𝑛 − 𝑎 < 𝜖 for all 𝑛 ≥ 𝑛2 .
⇒ 𝑎 − 𝜖 < 𝑥𝑛 < 𝑎 + 𝜖 and 𝑎 − 𝜖 < 𝑦𝑛 < 𝑎 + 𝜖 for all 𝑛 ≥ 𝑛0 = max{𝑛1 , 𝑛2 }
⇒ 𝑎 − 𝜖 < 𝑥𝑛 ≤ 𝑦𝑛 ≤ 𝑧𝑛 < 𝑎 + 𝜖 for all 𝑛 ≥ 𝑛0
⇒ for any 𝜖 > 0 there is a positive integer 𝑛0 such that
𝑦𝑛 − 𝑎 < 𝜖 for all 𝑛 ≥ 𝑛0 ⇒ lim 𝑦𝑛 = 𝑎.
𝑛→∞
This theorem is also known as Sandwich Theorem.

H.W.: If lim 𝑠𝑛 = 0 and {𝑡𝑛 } is a bounded sequence, then lim 𝑠𝑛 𝑡𝑛 = 0.

𝑛→∞ 𝑛→∞
𝑛
sin 𝑛+cos 𝑛 2
EXAMPLE 6: (i) lim = 0 (ii) Find lim .
𝑛→∞ 𝑛 𝑛→∞ 𝑛
1
(i) Here 𝑠𝑛 = 𝑛 → 0 and 𝑡𝑛 = sin 𝑛 + cos 𝑛 so that |𝑡𝑛 | = |sin 𝑛 + cos 𝑛 | ≤ √2(bounded)
sin 𝑛+cos 𝑛
Therefore, lim 𝑠𝑛 𝑡𝑛 = lim = 0.
𝑛→∞ 𝑛→∞ 𝑛

11
 CONVERGENCE OF SEQUENCE
𝑛 𝑛 𝑛 𝑛
−1 2 1 1 2 1 1 1 1 1
2 2
(ii) We can write ≤ ≤ or − ≤ ≤ 2 , where lim − 𝑛 = lim =2 .
𝑛 𝑛 𝑛 2 𝑛 𝑛 𝑛→∞ 2 𝑛→∞ 2
𝑛
2 1
Hence by sandwich theorem, lim =2.
𝑛→∞ 𝑛
𝑠1 +𝑠2 +⋯+𝑠𝑛
Cauchy’s First Theorem on Limits: If lim 𝑠𝑛 = 𝑙, then lim = 𝑙.
𝑛→∞ 𝑛→∞ 𝑛
PROOF: Define a sequence {𝑡𝑛 } by 𝑡𝑛 = 𝑠𝑛 − 𝑙 or 𝑠𝑛 = 𝑡𝑛 + 𝑙 for all 𝑛 ∈ ℕ so that
lim 𝑡𝑛 = lim 𝑠𝑛 − 𝑙 = 𝑙 − 𝑙 = 0. Also
𝑛→∞ 𝑛→∞
𝑠1 +𝑠2 +⋯+𝑠𝑛 𝑡1 +𝑡2 +⋯+𝑡𝑛
𝑛
=𝑙+ 𝑛
. (1)
Now since lim 𝑡𝑛 = 0, for any 𝜖 > 0, there exists 𝑚 ∈ ℕ such that 𝑡𝑛 − 0 = 𝑡𝑛 < 𝜖 for all
𝑛→∞
𝑛 ≥ 𝑚. Also sequence {𝑡𝑛 } being bounded satisfies |𝑡𝑛 | < 𝐾(a positive constant) for all 𝑛 ∈ ℕ.
Therefore, for all 𝑛 ≥ 𝑚, we have
𝑡1 +𝑡2 +⋯+𝑡𝑛 𝑡1 +𝑡2 +⋯+𝑡𝑚 𝑡𝑚+1 +𝑡𝑚+2 +⋯+𝑡𝑛 𝑡1 + 𝑡2 +⋯+ 𝑡𝑚 𝑡𝑚+1 + 𝑡𝑚+2 +⋯+ 𝑡𝑛
= + ≤ +
𝑛 𝑛 𝑛 𝑛 𝑛
𝑚𝐾 𝑛−𝑚 𝜖 𝑚𝐾 𝜖
≤ 𝑛 + 2𝑛 < 𝑛 + 2 . (2)
𝑚𝐾 𝜖 2𝑚𝐾 2𝑚𝐾
Also < iff 𝑛 > . If we choose 𝑛0 = max 𝑚, , then from (2),
𝑛 2 𝜖 𝜖

12
 CONVERGENCE OF SEQUENCE
𝑡1 +𝑡2 +⋯+𝑡𝑛 𝜖 𝜖
< 2 + 2 = 𝜖 for all 𝑛 > 𝑛0 .
𝑛
𝑡1 +𝑡2 +⋯+𝑡𝑛 𝑠1 +𝑠2 +⋯+𝑠𝑛
⟹ lim ⟹ lim = 𝑙 [from (1)]. (Limit of AM)
𝑛⟶∞ 𝑛 𝑛→∞ 𝑛

NOTE: The converse of the above theorem is not true. For example, for 𝑠𝑛 = −1 𝑛 , we have
1
𝑠1 +𝑠2 +⋯+𝑠𝑛 − 𝑛 , if 𝑛 is odd
=൝
𝑛
0, if 𝑛 is even
𝑠1 +𝑠2 +⋯+𝑠𝑛
⟹ lim 𝑛
= 0. But lim 𝑠𝑛 does not exist.
𝑛→∞ 𝑛→∞

Cauchy’s Second Theorem on Limits: If {𝑠𝑛 } is a sequence of positive terms such that
lim 𝑠𝑛 = 𝑙, then lim 𝑠1 𝑠2 … 𝑠𝑛 1/𝑛 = 𝑙.
𝑛→∞ 𝑛→∞
Proof: Let us define a sequence {𝑡𝑛 } by 𝑡𝑛 = log 𝑠𝑛 for all 𝑛 ∈ ℕ. Then, lim 𝑡𝑛 = log 𝑙. By
𝑛→∞
Cauchy’s first theorem, we have
𝑡1 +𝑡2 +⋯+𝑡𝑛 log 𝑠1 +log 𝑠2 +⋯+log 𝑠𝑛 1/𝑛
lim = lim = lim log 𝑠1 𝑠2 … 𝑠𝑛 = log 𝑙.
𝑛⟶∞ 𝑛 𝑛⟶∞ 𝑛 𝑛→∞

⟹ lim 𝑠1 𝑠2 … 𝑠𝑛 1/𝑛 = 𝑙. (Limit of GM) (What about the converse?)

𝑛→∞

13
 CONVERGENCE OF SEQUENCE
𝑠𝑛+1
THEOREM: If {𝑠𝑛 } is a sequence of positive terms such that lim = 𝑙 , then
𝑛→∞ 𝑠𝑛
lim 𝑠𝑛 1/𝑛 = 𝑙.
𝑛→∞
𝑠𝑛
Proof: Let us define a sequence {𝑡𝑛 } by 𝑡1 = 𝑠1 and 𝑡𝑛 = 𝑠 for all 2 ≤ 𝑛 ∈ ℕ so that
𝑛−1
𝑠 𝑠 𝑠 𝑠𝑛 𝑠𝑛 𝑠𝑛+1
𝑡1 𝑡2 … 𝑡𝑛 = 𝑠1 𝑠2 𝑠3 𝑠4 … 𝑠 = 𝑠𝑛 and lim 𝑡𝑛 = lim = lim = 𝑙.
1 2 3 𝑛−1 𝑛→∞ 𝑛→∞ 𝑠𝑛−1 𝑛→∞ 𝑠𝑛
As 𝑠𝑛 ≥ 0 for all 𝑛 ∈ ℕ, by the Cauchy’s second theorem, we have
lim 𝑡1 𝑡2 … 𝑡𝑛 1/𝑛 = lim 𝑠𝑛 1/𝑛 = 𝑙. [What about converse]
𝑛→∞ 𝑛→∞
𝑛+1 𝑛 1 𝑛
EXAMPLE 7: Evaluate the following limits (ii) Let 𝑠𝑛 = so that lim 1 + = 𝑒 (why)
𝑛 𝑛→∞ 𝑛
1 1 1 1 By second theorem of Cauchy
(i) lim 1 + + + ⋯+
𝑛→∞ 𝑛 3 5 2𝑛−1 1
1/𝑛 2 𝑛 𝑛
2 3 2 4 3 𝑛+1 𝑛 1/𝑛
2 3 𝑛+1
(ii) lim
𝑛→∞ 1 2 3
…
𝑛
lim 𝑠1 𝑠2 … 𝑠𝑛 = lim
𝑛→∞ 1 2
…
𝑛
=𝑒
𝑛→∞
𝑛
(iii) lim
𝑛→∞ 𝑛! 1/𝑛 𝑛𝑛 𝑠𝑛+1 (𝑛+1)𝑛+1 𝑛! 1 𝑛
(iii) Let 𝑠𝑛 = so that = = 1+ and
1 1 𝑛! 𝑠𝑛 (𝑛+1)! 𝑛𝑛 𝑛
(i) Let 𝑠𝑛 = so that lim 𝑠𝑛 = lim = 0.
2𝑛−1 𝑛→∞ 𝑛→∞ 2𝑛−1 𝑠𝑛 1 𝑛 1/𝑛
By Cauchy’s first theorem lim = lim 1 + = 𝑒. Therefore, lim 𝑠𝑛 = 𝑒.
𝑛→∞ 𝑠𝑛−1 𝑛→∞ 𝑛 𝑛→∞
𝑠1 +𝑠2 +⋯+𝑠𝑛 1 1 1
lim = lim 1 + + ⋯+ =0
𝑛→∞ 𝑛 𝑛→∞ 𝑛 3 2𝑛−1
EXERCISE: Test 𝑛𝑐 𝑛 , 0 < 𝑐 < 1, for convergence.
14
MONOTONOCITY AND CONVERGENCE OF SEQUENCES

Monotonic Sequences: A sequence {𝑠𝑛 } is said to be

▪ increasing if 𝑠𝑛+1 ≥ 𝑠𝑛 for all 𝑛 ∈ ℕ
▪ strictly increasing 𝑠𝑛+1 > 𝑠𝑛 for all 𝑛 ∈ ℕ
▪ decreasing if 𝑠𝑛+1 ≤ 𝑠𝑛 for all 𝑛 ∈ ℕ
▪ strictly decreasing if 𝑠𝑛+1 < 𝑠𝑛 for all 𝑛 ∈ ℕ
▪ monotonic if {𝑠𝑛 } is either increasing or decreasing
EXAMPLES:
𝑛𝜋 1 1 1 1
1. Sequences and σ𝑛𝑘=1 = 1 + + + ⋯+ are increasing.
2 𝑘 2 3 𝑛
𝜋 −1 1 1 1
2. Sequences and σ𝑛𝑘=1 = −1 − − − ⋯ − are increasing.
2𝑛 𝑘 2 3 𝑛
𝑛𝜋 𝑛 −1 𝑘 1 1 −1 𝑛
3. Sequences sin and σ𝑘=1 = −1 + − + ⋯+ are neither
2 𝑘 2 3 𝑛
increasing nor decreasing.
𝜋
4. Constant sequences is both increasing and decreasing
2
THEOREM: A bounded and increasing sequence is convergent.
15
MONOTONOCITY AND CONVERGENCE OF SEQUENCES

Proof: Let {𝑠𝑛 } be a bounded and increasing sequence with range set 𝐴 = {𝑠𝑛 : 𝑛 ∈ ℕ}.
Then, there is a constant 𝑀 such that so that 𝑠𝑛 ≤ 𝑀 for all 𝑛 ∈ ℕ.
⟹ 𝐴 is a non-empty set of real numbers which is bounded above.
⟹ 𝐴 has a least upper bound, say, 𝛼 = 𝑙𝑢𝑏 𝐴, that is, 𝑠𝑛 ≤ 𝛼 for all 𝑛 ∈ ℕ (1)
⟹ for any 𝜖 > 0, there is an element 𝑎 ∈ 𝐴 such that 𝑎 > 𝛼 − 𝜖.
⟹ there exists an integer 𝑛0 such that 𝑠𝑛0 = 𝑎 > 𝛼 − 𝜖 and 𝑠𝑛0 ≤ 𝑠𝑛 for all 𝑛 ≥ 𝑛0 .
⟹ from (1), 𝛼 − 𝜖 < 𝑠𝑛 < 𝛼 + 𝜖 for all 𝑛 ≥ 𝑛0
⟹ |𝑠𝑛 − 𝛼| < 𝜖 for all 𝑛 ≥ 𝑛0
⟹ {𝑠𝑛 } in convergent and has the limit 𝛼, that is, sup {𝑠𝑛 }

Theorem: A bounded and decreasing sequence is convergent. (H.W.)

EXAMPLE 8: Consider the sequence 𝑥𝑛 = 0.11 … 𝑛 times, that is, 𝑥1 = 0.1, 𝑥2 =
0.11, 𝑥3 = 0.111,…
This sequence is increasing as 𝑥𝑛+1 − 𝑥𝑛 = 0.0000 … 1 > 0 and bounded by 0.12.
Therefore, {𝑥𝑛 } is convergent has a limit 𝑙, say. This sequence also has the property
that
16
MONOTONOCITY AND CONVERGENCE OF SEQUENCES

10𝑥2 − 𝑥1 = 1.1 − 0.1 = 1, 10𝑥3 − 𝑥2 = 1.11 − 0.11 = 1, … , 10𝑥𝑛+1 − 𝑥𝑛 = 1.

1
Therefore, 10 lim 𝑥𝑛+1 − lim 𝑥𝑛 = 1 or 10𝑙 − 𝑙 = 1 or 𝑙 = .
𝑛→∞ 𝑛→∞ 9

2𝑠𝑛 +3
EXAMPLE 9: Suppose the sequence {𝑠𝑛 } is defined by 𝑠1 = 1, 𝑠𝑛+1 = , 𝑛 ≥ 1.
4
2𝑠 +3 5 2𝑠𝑛−1 +3 2𝑠 +3
Here 𝑠2 = 1 = > 𝑠1 . Assume that 𝑠𝑛−1 < 𝑠𝑛 . This implies that < 𝑛 ,
4 4 4 4
that is, 𝑠𝑛 < 𝑠𝑛+1 for all 𝑛 ∈ ℕ.
⇒ {𝑠𝑛 } is an increasing sequence
5 2𝑠𝑛−1 +3 7
Also 𝑠1 < 𝑠2 = < 2. Assume 𝑠𝑛−1 < 2 ⇒ 2𝑠𝑛−1 + 3 < 4 + 3 ⇒ < < 2.
4 4 4
Hence, {𝑠𝑛 } is an increasing and bounded sequence, and therefore it is convergent. Let
2 lim 𝑠𝑛 +3 2𝑙+3 3
𝑛→∞
lim 𝑠𝑛 = 𝑙. Then, lim 𝑠𝑛+1 = ⇒𝑙= ⇒𝑙= .
𝑛→∞ 𝑛→∞ 4 4 2

EXERCISE: Discuss the convergence of the sequence defined by

3 1
𝑠1 = , 𝑠𝑛+1 = 2 − ,𝑛 ≥ 1.
2 𝑠𝑛

17
SUBSEQUENCES
∞
DEFINITION: If {𝑛𝑘 } 𝑘=1 is a strictly increasing sequence of positive integers, then
∞ ∞
the sequence {𝑠𝑛𝑘 } 𝑘=1 is called a sub-sequence of the sequence {𝑠𝑛 } 𝑛=1 .
∞ 𝑛𝜋 ∞
For example the sequence 𝑠𝑛 𝑛=1
= sin = {1,0, −1,0,1,0, −1,0,1, … } has the
2 𝑛=1
following subsequences.
∞ ∞ ∞ ∞
𝑠2𝑘 𝑘=1
= sin 𝑘𝜋 𝑘=1
= {0,0,0, … }, 𝑛𝑘 𝑘=1
= 2𝑘 𝑘=1
= {2,4,6, … }
∞ 2𝑘−1 𝜋 ∞ ∞ ∞
𝑠2𝑘−1 𝑘=1
= sin 𝑘=1
= {1, −1,1, −1, … }, 𝑛𝑘 𝑘=1
= 2𝑘 − 1 𝑘=1
= {1,3,5, … }
2
∞ 4𝑘+1 𝜋 ∞ ∞ ∞
𝑠4𝑘+1 𝑘=1
= sin 𝑘=1
= {1,1,1, … }, 𝑛𝑘 𝑘=1
= 4𝑘 + 1 𝑘=1
= {5,9,13, … }
2

THEOREM: If {𝑠𝑛 } converges to 𝑙, then every subsequence of {𝑠𝑛 } converges to 𝑙.

Proof: Let {𝑠𝑛𝑘 } be a subsequence of {𝑠𝑛 }. Since {𝑠𝑛 } converges to 𝑙, for any 𝜖 > 0,
there is 𝑛0 ∈ ℕ such that 𝑠𝑛 − 𝑙 < 𝜖 for all 𝑛 ≥ 𝑛0 .
For 𝑘 ≥ 𝑛0 , 𝑛𝑘 ≥ 𝑘 ≥ 𝑛0 so that 𝑠𝑛𝑘 − 𝑙 < 𝜖 for all 𝑘 ≥ 𝑛0 . Hence, lim 𝑠𝑛𝑘 = 𝑙.
𝑘→∞

The converse of this theorem is trivially true, since every sequence is a

subsequence of itself.
18
SUBSEQUENCES (NESTED INTERVAL THEOREM)

THEOREM: Let {𝐼𝑛 } be a sequence of closed intervals 𝐼𝑛 = [𝑎𝑛 , 𝑏𝑛 ] such that

𝐼𝑛+1 ⊂ 𝐼𝑛 for all 𝑛 ∈ ℕ. Then,
• The sequences {𝑎𝑛 } and {𝑏𝑛 } converges, and ‫∞ځ‬𝑛=1 𝐼𝑛 = [𝛼, 𝛽], where 𝛼 =
lim 𝑎𝑛 and 𝛽 = lim 𝑏𝑛 .
𝑛→∞ 𝑛→∞
• If lim (𝑎𝑛 − 𝑏𝑛 ) = 0, then 𝛼 = 𝛽.
𝑛→∞
Proof: H. W. (You can take help from Unit-1)
𝑛+1
EXAMPLE 10: If 𝐼𝑛 = 0, , then ‫∞ځ‬
𝑛=1 𝐼𝑛 = [0,1]. On the other hand, if 𝐼𝑛 =
𝑛
1 1
0, , then no point is common to each 𝐼𝑛 . For if, 𝑧 ∈ 𝐼𝑛 ⇒ 𝑧 ≤ . We can always
𝑛 𝑛
1
find 𝑚 ∈ ℕ such that < 𝑧, that is, 𝑧 ∉ 𝐼𝑚 . Therefore, closedness is important.
𝑚

We shall use this result to prove that every bounded sequence has a convergent
subsequence.

19
SUBSEQUENCES
BOLZANO-WEIERSTRASS THEOREM: Every bounded sequence has a convergent
subsequence.
Proof: Let {𝑠𝑛 } be a bounded sequence, i. e., for some constant 𝑀 > 0, 𝑠𝑛 < 𝑀 for
all 𝑛 ∈ ℕ. Let us construct a nested sequence of intervals. Define, 𝐼1 = 𝑎1 , 𝑏1 =
𝑎 +𝑏
[−𝑀, 𝑀]. Let 𝑐1 = 1 1 be the mid point of 𝐼1 (here 𝑐1 = 0).
2
– Since every sequence has infinitely many terms, there are infinitely many terms in of {𝑠𝑛 } in at
least one of the sub intervals [𝑎1 , 𝑐1 ] or [𝑐1 , 𝑏1 ]. Let us choose one and denote it by 𝐼2 = 𝑎2 , 𝑏2 .
𝑎 +𝑏
– Similarly, divide the interval 𝐼2 into two sub intervals 𝑎2 , 𝑐2 and 𝑐2 , 𝑏2 , where 𝑐2 = 2 2 . At
2
least one of them has infinitely many terms of {𝑠𝑛 }. Let us fixed it as 𝐼3 = 𝑎3 , 𝑏3 .
– Continuing in this way, we have a nested sequence of intervals {𝐼𝑘 }, where 𝐼𝑘 = 𝑎𝑘 , 𝑏𝑘 and
𝐼𝑘+1 ⊂ 𝐼𝑘 .
𝑏 −𝑎
– By nested interval theorem, the sequences {𝑎𝑛 } and {𝑏𝑛 } converges. Also 𝑏𝑘 − 𝑎𝑘 = 𝑘−1 𝑘−1 =
2
𝑏𝑘−2 −𝑎𝑘−2 𝑏1 −𝑎1 2𝑀 𝑀
=⋯= = = so that lim (𝑏𝑘 − 𝑎𝑘 ) = 0 ⇒ lim 𝑎𝑘 = lim 𝑏𝑘 = 𝑙, say. (1)
22 2𝑘−1 2𝑘 2𝑘−2 𝑘→∞ 𝑘→∞ 𝑘→∞

We now form a subsequence {𝑠𝑛𝑘 } of {𝑠𝑛 } such that 𝑠𝑛𝑘 ∈ 𝐼𝑘 for each 𝑘 ∈ ℕ.
– Let 𝑠𝑛1 be any term of the sequence {𝑠𝑛 }. Since {𝑠𝑛 : 𝑛 ∈ ℕ} ⊂ 𝐼1 , 𝑠𝑛1 ∈ 𝐼1 .
– Since 𝐼2 contains infinitely many terms of the sequence {𝑠𝑛 }, we can choose 𝑛2 > 𝑛1 such that
𝑠𝑛1 ∈ 𝐼2
– Continue by induction and assume that 𝑠𝑛𝑘 ∈ 𝐼𝑘 . Since 𝐼𝑘 contains infinitely many terms of the
sequence {𝑠𝑛 }, we can choose 𝑛𝑘+1 > 𝑛𝑘 such that 𝑠𝑛𝑘+1 ∈ 𝐼𝑘+1 .
20
SUBSEQUENCES (Limit point)

– Thus we have constructed a subsequence a sequence {𝑠𝑛𝑘 }, which is a subsequence of {𝑠𝑛 } and
𝑎𝑘 ≤ 𝑠𝑛𝑘 ≤ 𝑏𝑘 for each 𝑘 ∈ 𝐼𝑘 .
⇒ from (1) that 𝑎𝑘 ≤ 𝑠𝑛𝑘 ≤ 𝑏𝑘 so that lim 𝑎𝑘 = lim 𝑠𝑛𝑘 = lim 𝑏𝑘 = 𝑙.
𝑘→∞ 𝑘→∞ 𝑘→∞
⇒ {𝑠𝑛𝑘 } is convergent.

EXAMPLE: The sequence { −1 𝑛 } is bounded and has two convergent

subsequence {1} and {−1} .
LIMIT POINT OF A SEQUENCE: A real number 𝑝 is said to be a limit point (cluster
point) of a sequence {𝑠𝑛 } if there is a subsequence of {𝑠𝑛 } converging to 𝑝.
In other words, there are infinite number of terms of {𝑠𝑛 } in every neighbourhood
of 𝑝.
EXAMPLES:
– { −1 𝑛 } has two limit points 1 and −1.
𝑛𝜋
– 𝑛 sin has one limit point 0.
2
𝑛
– −1 𝑛 has no limit point.

THEOREM: A number 𝑙 is a limit point of a sequence {𝑠𝑛 } if and only if, for any 𝜖 >
0 and positive integer 𝑛0 , there exists 𝑛 ≥ 𝑛0 such that 𝑠𝑛 − 𝑙 < 𝜖.
(Proof is H.W.)
21
INFINITE LIMITS AND DIVERGENT SEQUENCES

• A sequence {𝑠𝑛 } diverges to +∞, denoted by lim 𝑠𝑛 = +∞ or 𝑠𝑛 → ∞ as 𝑛 → ∞, if

𝑛→∞
for any 𝑀 > 0 there exists a positive integer 𝑛0 such that 𝑠𝑛 ≥ 𝑀 for all 𝑛 ≥ 𝑛0 .
• Similarly, {𝑠𝑛 } diverges to +∞, denoted by lim 𝑠𝑛 = −∞ or 𝑠𝑛 → ∞ as 𝑛 → ∞, if for
𝑛→∞
any 𝑀 > 0 there exists a positive integer 𝑛0 such that 𝑠𝑛 ≤ −𝑀 for all 𝑛 ≥ 𝑛0 .
1 1
• The sequence 𝑠𝑛 = 𝑛 diverges to +∞. Here for any 𝑀 > 0, 𝑠𝑛 = 𝑛 > 𝑀 iff 𝑛 > 𝑀3 .
3 3

We can choose any positive integer greater than 𝑀 as 𝑛0 . Thus 𝑠𝑛 > 𝑀 for all 𝑛 ≥
𝑛0 . Hence, lim 𝑠𝑛 = +∞.
𝑛→∞
1 1
• The sequence 𝑠𝑛 = log diverges to −∞ since for any 𝑀 > 0, 𝑠𝑛 = log < −𝑀 iff
𝑛 𝑛
𝑀 𝑀
𝑛 > 𝑒 . We can choose any positive integer greater than 𝑒 as 𝑛0 . Thus 𝑠𝑛 < −𝑀
for all 𝑛 ≥ 𝑛0 . Hence, lim 𝑠𝑛 = −∞.
𝑛→∞
THEOREM: Suppose {𝑠𝑛 } is a positive sequence. Then lim 𝑠𝑛 = +∞ if and only if
𝑛→∞
1
lim 𝑠 = 0.
𝑛→∞ 𝑛
Proof: First suppose that 𝑠𝑛 > 0 for all 𝑛 and lim 𝑠𝑛 = +∞. Then for any 𝜖 > 0, there
𝑛→∞
1
is a positive integer 𝑛0 such that 𝑠𝑛 > for all 𝑛 ≥ 𝑛0 .
𝜖
1 1 1
⇒ −0 = < 𝜖 for all 𝑛 ≥ 𝑛0 ⇒ lim = 0.
𝑠𝑛 𝑠𝑛 𝑛→∞ 𝑠𝑛
22
INFINITE LIMITS AND DIVERGENT SEQUENCES
1
Conversely, suppose that 𝑠𝑛 > 0 for all 𝑛 and lim = 0. We take any 𝑀 > 0
𝑛→∞ 𝑠𝑛
1 1 1 1
so that for 𝜖 = > 0, there is a positive integer 𝑛0 such that −0 = <
𝑀 𝑠𝑛 𝑠𝑛 𝑀
or 𝑠𝑛 > 𝑀 for all 𝑛 ≥ 𝑛0 . This means lim 𝑠𝑛 = +∞.
𝑛→∞
𝑛2 +1
EXAMPLE 11: The sequence 𝑠𝑛 = is divergent. For any constant 𝑀 >
𝑛+100
𝑛2 +1 𝑛2 𝑛 𝑛
0, we have 𝑠𝑛 = > = for all 𝑛 ≥ 100 and 𝑠𝑛 > > 𝑀 iff 𝑠𝑛 = 𝑛 >
𝑛+100 2𝑛+𝑛 2
2𝑀. Thus for any 𝑀 > 0, 𝑠𝑛 > 𝑀 for all 𝑛 ≥ 𝑛0 = max{100, 2𝑀}. Therefore,
1 𝑛+100
lim 𝑠𝑛 = +∞. Also, lim = lim 2 = 0.
𝑛→∞ 𝑛→∞ 𝑠𝑛 𝑛→∞ 𝑛 +1

EXERCISE: If {𝑠𝑛 } is divergent and {𝑡𝑛 } is bounded, then show that {𝑠𝑛 + 𝑡𝑛 }
is divergent.

23
LIMIT POINT

THEOREM: A convergent sequence has a unique limit point as its limit. The
converse is not true. (Proof is H.W.)

LEMMA: If the sequence is bounded above and does not have any cluster point,
then for any constant 𝑄 (+ve or –ve), only a finite number of terms of {𝑠𝑛 } can be
grater than 𝑄.
Proof: If 𝑀 is an upper bound of {𝑠𝑛 }, then there are infinite number of terms of
the sequence {𝑠𝑛 } lying between 𝑄 and 𝑀. It means that there is a subsequence of
{𝑠𝑛 } , say, {𝑠𝑛𝑘 } which is bounded. This subsequence has a convergent
subsequence, say, {𝑠𝑛𝑝 } which is also a subsequence of {𝑠𝑛 }. Thus {𝑠𝑛 } has a
convergent subsequence, that is, {𝑠𝑛 } has a limit point. Which is not possible.
Hence, only a finite number of terms of {𝑠𝑛 } can be grater than 𝑄.

Now any sequence {𝑠𝑛 } must be in exactly one of the following three categories:
– {𝑠𝑛 } is not bounded above.
– 𝑠𝑛 is bounded above and has at least one cluster point.
– 𝑠𝑛 is bounded above and has no limit point.

24
 LIMIT SUPERIOR AND LIMIT INFERIOR

LIMIT SUPERIORS: Let {𝑠𝑛 } be a sequence. Then limit superior of {𝑠𝑛 } denoted by
lim sup 𝑠𝑛 is defined as follows.
𝑛→∞
• lim sup 𝑠𝑛 = +∞, if {𝑠𝑛 } is not bounded above.
𝑛→∞
• if {𝑠𝑛 } is bounded above and has at least one limit point, then the set 𝑇 of all
cluster points of {𝑠𝑛 } is non-empty and bounded above. Thus 𝑇 has a least
upper bound, say, 𝑙. In this case and in this case lim sup 𝑠𝑛 = 𝑙.
𝑛→∞
• if {𝑠𝑛 } is bounded above and has no limit point, then lim sup 𝑠𝑛 = −∞.
𝑛→∞
EXAMPLES:
𝑛
– lim sup −1 = 1, since 𝑇 = {1, −1}
𝑛→∞
𝑛𝜋
– lim sup sin = 1, since 𝑇 = {−1, 0, 1}
𝑛→∞ 2
1 1 1 1 1 1 1 1
– For the sequence 𝑠𝑛 = 1, 1, , 1, , , … , 1, , , , … , 𝑇 = 1, , , … ∪ {0}. Therefore,
2 2 3 2 3 4 2 3
lim sup 𝑠𝑛 = 1.
𝑛→∞
𝑛𝜋
– lim sup 𝑛 sin = +∞
𝑛→∞ 2

– lim sup −1 𝑛 𝑛 = +∞
𝑛→∞
25
LIMIT SUPERIOR AND LIMIT INFERIOR

THEOREM: If lim sup 𝑠𝑛 = 𝛼 ∈ ℝ and 𝑇 is the set of cluster points of {𝑠𝑛 }. Then,
𝑛→∞
𝛼 = 𝑙𝑢𝑏𝑇 is in 𝑇.
PROOF: Since 𝛼 = sup 𝑇, 𝑎 ≤ 𝛼 for any cluster point 𝑎 of the sequence {𝑠𝑛 }.
Assume that 𝛼 ∉ 𝑇, i.e., 𝛼 is not a cluster point of {𝑠𝑛 }.
⇒ there is an 𝜖 > 0 such that 𝑠𝑛 − 𝛼 ≥ 𝜖 for 𝑛 ≥ 𝑛0 (some positive integer)
⇒ there are only a finite number of terms of {𝑠𝑛 } in the interval ]𝛼 − 𝜖, 𝛼 + 𝜖[
⇒ no subsequence of {𝑠𝑛 } can converge to a point of the interval ]𝛼 − 𝜖, 𝛼 + 𝜖[
⇒ no point of the interval ]𝛼 − 𝜖, 𝛼 + 𝜖[ is a limit point of {𝑠𝑛 }, i.e., element of 𝑇
⇒ 𝛼 − 𝜖 is an upper bound of 𝑇, which contradicts the facts that 𝛼 = 𝑙𝑢𝑏𝑇
Hence, 𝛼 ∈ 𝑇.
Limit Inferior: The limit inferior of a bounded sequence {𝑠𝑛 }, denoted by
lim inf 𝑠𝑛 , is defined to be the inf 𝑇, where 𝑇 is the set of all limit/cluster points
𝑛→∞
of {𝑠𝑛 } .
– The lim inf 𝑠𝑛 = −∞, if {𝑠𝑛 } is bounded above and has no limit point.
𝑛→∞
– The lim inf 𝑠𝑛 = +∞, if {𝑠𝑛 } is not bounded above and has no limit point.
𝑛→∞
– In general, lim inf 𝑠𝑛 = − lim sup 𝑡𝑛 , where {𝑡𝑛 } = {−𝑠𝑛 }.
𝑛→∞ 𝑛→∞

26
LIMIT SUPERIOR AND LIMIT INFERIOR

EXAMPLES:
𝑛
– lim inf −1 = −1
𝑛→∞
𝑛𝜋 1 𝑛𝜋
– lim inf sin = −1 , 𝑇 = {±1, ± , 0}, lim sup sin =?
𝑛→∞ 4 2 𝑛→∞ 4

𝑛 1 𝑛 1
– lim inf −1 1+ = −1, lim sup −1 1+ =?
𝑛→∞ 𝑛 𝑛→∞ 𝑛

NOTE: If {𝑠𝑛 } is a bounded sequence, then lim inf 𝑠𝑛 ≤ lim sup 𝑠𝑛 .

𝑛→∞ 𝑛→∞
ALTERNATIVE DEFINITION LIMSUP AND LIMINF: Let {𝑠𝑛 } be a bounded
sequence. Define for 𝑛 ∈ ℕ,
𝑀𝑛 = sup{𝑠𝑘 : 𝑘 ≥ 𝑛} = sup{𝑠𝑛 , 𝑠𝑛+1 , … } and 𝑚𝑛 = inf{𝑠𝑘 : 𝑘 ≥ 𝑛}.
Both {𝑀𝑛 } and {𝑚𝑛 } being monotonic and bounded are convergent. We define,
lim sup 𝑠𝑛 = lim 𝑀𝑛 and lim inf 𝑠𝑛 = lim 𝑚𝑛 .
𝑛→∞ 𝑛→∞ 𝑛→∞ 𝑛→∞
If {𝑠𝑛 } is bounded above and diverges to minus infinity, then lim sup 𝑠𝑛 = −∞.
𝑛→∞
If {𝑠𝑛 } is not bounded above, then lim sup 𝑠𝑛 = lim 𝑀𝑛 = ∞.
𝑛→∞ 𝑛→∞
If {𝑠𝑛 } is bounded below and diverges to infinity lim inf 𝑠𝑛 = ∞.
𝑛→∞
If {𝑠𝑛 } is not bounded below, then lim inf 𝑠𝑛 = lim 𝑚𝑛 = −∞.
𝑛→∞ 𝑛→∞
27
LIMIT SUPERIOR AND LIMIT INFERIOR

THEOREM: If {𝑠𝑛 } is a bounded sequence, then there are subsequences

{𝑎𝑛𝑘 } and {𝑏𝑛𝑘 } such that lim sup 𝑠𝑛 = lim 𝑎𝑛𝑘 and lim inf 𝑠𝑛 = lim 𝑏𝑛𝑘 .
𝑛→∞ 𝑛→∞ 𝑛→∞ 𝑛→∞
PROOF: Since {𝑠𝑛 } is bounded, the lim sup 𝑠𝑛 = 𝛼 exists. Then by definition,
𝑛→∞
for each 𝑘 (large enough)
1 1
𝛼 − < sup{𝑠𝑛 : 𝑛 ≥ 𝑘} < 𝛼 +
𝑘 𝑘
1 1
⇒ there exists 𝑎𝑛𝑘 (some term of {𝑠𝑛 }) such that 𝛼 − < 𝑎𝑛𝑘 < 𝛼 + for each
𝑘 𝑘
𝑘
⇒ lim 𝑎𝑛𝑘 = 𝛼 = lim sup 𝑠𝑛 . Similarly, one can obtain {𝑏𝑛𝑘 }.
𝑛→∞ 𝑛→∞

REMARK: If {𝑠𝑛 } is a convergent sequence with lim 𝑠𝑛 = 𝑙 , then the

𝑛→∞
subsequences {𝑎𝑛𝑘 } and {𝑏𝑛𝑘 } converge to 𝑙. Therefore,
lim sup 𝑠𝑛 = lim 𝑠𝑛 = lim inf 𝑠𝑛 = 𝑙.
𝑛→∞ 𝑛→∞ 𝑛→∞
THEOREM: If {𝑠𝑛 } is a sequence lim sup 𝑠𝑛 = lim inf 𝑠𝑛 = 𝑙, then lim 𝑠𝑛 = 𝑙.
𝑛→∞ 𝑛→∞ 𝑛→∞
(Proof is H.W.)
28
CAUCHY SEQUENCES

DEFINITION: A sequence {𝑠𝑛 } is said to be a Cauchy sequence, if for any

𝜖 > 0, there is a positive integer 𝑛0 such that 𝑠𝑛 − 𝑠𝑚 < 𝜖 for all 𝑛, 𝑚 ≥ 𝑛0 .
𝑛+1 1
EXAMPLES: 1. Consider 𝑠𝑛 = = 1 + 𝑛. Here,
𝑛
1 1 1 1
𝑠𝑛 − 𝑠𝑚 = −𝑚 ≤𝑚+𝑛. (1)
𝑛
1 𝜖
We know that for any 𝜖 > 0, there exists a positive integer 𝑛0 such that 𝑛 < 2.
0
1 1 𝜖 𝜖
Therefore, for 𝑚, 𝑛 > 𝑛0 , we have 𝑚 + 𝑛 < 2 + 2 = 𝜖. Thus from (1), we have
1 1
𝑠𝑛 − 𝑠𝑚 = 𝑚 + 𝑛 < 𝜖 for all 𝑚, 𝑛 > 𝑛0 .
Hence, {𝑠𝑛 } is a Cauchy sequence.
1 1 1 1
2. Consider 𝑠𝑛 = 1 + 2 + 3 + ⋯ + 𝑛. Here 𝑠𝑛+1 − 𝑠𝑛 = 𝑛+1 → 0 as 𝑛 → 0.
But for 𝑛 ≥ 𝑚, we have
1 1 1 𝑛−𝑚
𝑠𝑛 − 𝑠𝑚 = + + ⋯+ ≥ .
𝑚+1 𝑚+2 𝑛 𝑛
2𝑚−𝑚 1 1
In particular, for 𝑛 = 2𝑚, 𝑠𝑛 − 𝑠𝑚 ≥ 2𝑚 = 2. Hence for any 𝑛0 , 𝑠2𝑛0 − 𝑠𝑛0 ≥ 2.
Which shows that {𝑠𝑛 } is not a Cauchy sequence.
29
CAUCHY SEQUENCES

THEOREM: A sequence is a Cauchy sequence, if and only if, it is convergent.

PROOF: First suppose that {𝑠𝑛 } is a convergent sequence with lim 𝑠𝑛 = 𝑙. Then for
𝑛→∞
𝜖
any 𝜖 > 0, there exists a positive integer 𝑛0 such that 𝑠𝑛 − 𝑙 < for all 𝑛 > 𝑛0 .
2
𝑠𝑛 − 𝑠𝑚 = 𝑠𝑛 − 𝑙 + 𝑙 − 𝑠𝑚 ≤ 𝑠𝑛 − 𝑙 + 𝑙 − 𝑠𝑚 < 𝜖 for all 𝑚, 𝑛 > 𝑛0 .
Hence, {𝑠𝑛 } is a Cauchy sequence.
Conversely, suppose that {𝑠𝑛 } is a Cauchy sequence. Then for any 𝜖 > 0, there exists
a positive integer 𝑛0 such that 𝑠𝑛 − 𝑠𝑚 < 𝜖 for all 𝑚, 𝑛 > 𝑛0 .
⇒ 𝑠𝑛 − 𝑠𝑛0 < 𝑠𝑛 − 𝑠𝑛0 < 𝜖 for all 𝑛 > 𝑛0 .
⇒ 𝑠𝑛 < 𝑠𝑛0 + 𝜖 for all 𝑛 > 𝑛0 .
If we choose 𝑀 = max{|𝑠1 |, |𝑠2 |, … , 𝑠𝑛0−1 , 𝑠𝑛0 + 𝜖}, then 𝑠𝑛 < 𝑀 for all 𝑛 ∈ ℕ.
Hence, {𝑠𝑛 } is a bounded sequence, and has a convergent subsequence, say, {𝑠𝑛𝑘 },
by the Wolzano-Weierstrass theorem. Let lim 𝑠𝑛𝑘 = 𝑎. We shall complete the proof by
𝑘→∞
proving that lim 𝑠𝑛 = 𝑎. Since {𝑠𝑛 } is a Cauchy sequence, for any 𝜖 > 0, there exists a
𝑛→∞
𝜖
positive integer 𝑛0 such that 𝑠𝑛 − 𝑠𝑚 < 2 for all 𝑚, 𝑛 > 𝑛0 .
Again the convergence of {𝑠𝑛𝑘 } implies that there is a [positive integer 𝑛𝑘′ > 𝑛0 such
𝜖
that 𝑠𝑛𝑘 ′ − 𝑎 < 2 .
30
CAUCHY SEQUENCES
𝜖 𝜖
⇒ 𝑠𝑛 − 𝑎 = 𝑠𝑛 − 𝑠𝑛𝑘 ′ + 𝑠𝑛𝑘 ′ − 𝑎 ≤ 𝑠𝑛 − 𝑠𝑛𝑘 ′ + 𝑠𝑛𝑘 ′ − 𝑎 < + = ϵ
2 2
for all 𝑛 > 𝑛0 . Hence, {𝑠𝑛 } is a convergent sequence and lim 𝑠𝑛 = 𝑎.
𝑛→∞
1
EXAMPLE 12: The sequence defined by 𝑠𝑛 = σ𝑛𝑘=1 𝑘 is not a Cauchy sequence.
Therefore it is not convergent.
1
EXAMPLE 13: Consider the sequence 𝑠𝑛 = σ𝑛𝑘=1 𝑘 2 . Here, for 𝑛 > 𝑚 , we have
1 1 1 1 1 1 1
𝑠𝑛 − 𝑠𝑚 = + + ⋯ + + < + + ⋯ +
𝑚+1 2 𝑚+2 2 𝑛−1 2 𝑛2 𝑚 𝑚+1 𝑚+1 𝑚+2 𝑛−1 𝑛
1 1 1 1 1 1 1 1 1
= 𝑚 − 𝑚+1 + 𝑚+1 − 𝑚+2 + ⋯ + 𝑛−1 − 𝑛 = 𝑚 − 𝑛 < 𝑚.
1
For any 𝜖 > 0, there exists a positive integer 𝑛0 such that 𝑛 < 𝜖. Therefore,
0
1
𝑠𝑛 − 𝑠𝑚 < 𝑛 < 𝜖 for all 𝑛, 𝑚 > 𝑛0 . Hence, {𝑠𝑛 } being a Cauchy sequence is
0
convergent.
EXERCISE:
1. If 𝑠𝑛 = 𝑛, then show that for any 𝜖 > 0 there exists a positive integer 𝑛0 such that 𝑠𝑛+1 − 𝑠𝑛 < 𝜖 for
all 𝑛 > 𝑛0 . Check whether {𝑠𝑛 } is a Cauchy sequence.
1
2. Using the Cauchy’s criterion test the sequence σ𝑛𝑘=1 for convergence.
𝑘!

31
INFINITE SERIES:
1 3 3 3 3
• Consider = 0.3333 … = + + + ⋯ = σ∞
𝑛=1 . (Infinite series)
3 10 102 103 10𝑛
• The middle part is a sum of infinite terms/numbers.
• It is meaningless to ask anyone or anything, even the fastest computer, to
add together infinitely many numbers.
• The notion of the sum of an infinite series thus requires a precise
definition/process.
1 3 3 3
In order to give meaning to the expression = + 2 + + ⋯, one can
3 10 10 103
proceed as follows:
– Form a sequence of the sum of first 𝑛 terms of the series, that is,
3 1 𝑛
3 3 3 3 3 1−
10 10
𝑆𝑛 = + + +⋯+ = 0.33 … 𝑛 times = σ𝑛𝑘=1 = 1 .
10 102 103 10𝑛 10𝑘 1−
10
1 1
– It can be shown that lim 𝑆𝑛 = , that is, the sequence {𝑆𝑛 } converges to .
𝑛→∞ 3 3
3 1
– It means that σ∞
𝑘=1 10𝑘 = 3 .

32
 INFINITE SERIES:
DEFINITION: Let σ∞ ∞
𝑛=1 𝑢𝑛 be an infinite series, i.e., σ𝑛=1 𝑢𝑛 = 𝑢1 + 𝑢2 + 𝑢3 + ⋯
so that 𝑆𝑛 = σ𝑛𝑘=1 𝑢𝑛 . The sequence {𝑆𝑛 } is called the sequence of the partial
sums of the series σ∞ 𝑛=1 𝑢𝑛 . We say that the series is convergent and has the
sum 𝑆, if {𝑆𝑛 } converges to 𝑆.
– A series which is not convergent is called divergent.
– The number 𝑢𝑛 is called the 𝑛𝑡ℎ term of the series.

EXAMPLES:
– The series σ∞ 𝑛
𝑛=1 1 = 1 + 1 + ⋯ has infinite sum since 𝑆𝑛 = σ𝑛=1 1 = 1 + 1 + ⋯ +
1 𝑛 times = 𝑛 → ∞ as 𝑛 → ∞.
– The series σ∞
𝑛=1 −1
𝑛
= −1 + 1 − 1 + 1 − ⋯ does not have a definite value. It oscillates
between −1 and 1.
1
– In case of σ∞
𝑘=1 , we have
𝑘 𝑘+1
1 1 1 1 1 1 1 1 1 1 1
𝑆𝑛 = σ𝑛𝑘=1 = σ𝑛𝑘=1 − = 1 − + − + ⋯+ − + − =1− → as 𝑛 →
𝑘 𝑘+1 𝑘 𝑘+1 2 2 3 𝑛−1 𝑛 𝑛 𝑛+1 𝑛+1
∞. Therefore the series is convergent and has sum 1.

GEOMETRIC SERIES: The series σ∞

𝑛=1 𝑎𝑟
𝑛−1
is convergent if and only if 𝑟 < 1.
(H.W.)

33
 INFINITE SERIES:
EXAMPLE: Test the series 0.9 + 0.09 + 0.009 + ⋯ for convergence.
9 9 9 9
Given series can be written as + + + ⋯ = σ∞
𝑘 , which is a geometric series with
10 102 103 10𝑘
9
9 1 10
first term and common ratio < 1. Hence, the series is convergent, and has sum 1 = 1.
10 10 1−
10

A Necessary Condition for Convergence: If σ∞

𝑘=1 𝑢𝑘 is convergent, then
lim 𝑢𝑛 = 0.
𝑛→∞
We can write 𝑆𝑛 − 𝑆𝑛−1 = σ𝑛𝑘=1 𝑢𝑘 − σ𝑛−1
𝑘=1 𝑢𝑘 = 𝑢𝑛 …(1)
Since the series is convergent, lim 𝑆𝑛 = lim 𝑆𝑛−1 = 𝑆. Therefore, from (1)
𝑛→∞ 𝑛→∞
lim 𝑢𝑛 = lim 𝑆𝑛 − lim 𝑆𝑛−1 = 0.
𝑛→∞ 𝑛→∞ 𝑛→∞

EXAMPLES:
𝑘 𝑛
– Consider the series σ∞
𝑘=1 log with 𝑢𝑛 = log . Here, lim 𝑢𝑛 = log 1 = 0, but
𝑘+1 𝑛+1 𝑛→∞
𝑘
𝑆𝑛 = σ𝑛𝑘=1 log = σ𝑛𝑘=1 log 𝑘 − log(𝑘 + 1) = log 1 − log 𝑛 + 1 = −log(𝑛 + 1) → −∞ as 𝑛 → ∞
𝑘+1
(series is divergent).
1
– The series σ∞
𝑘=1 𝑘 , being a geometric series with common ratio less than 1, is convergent. But,
2
1 𝑛
lim 𝑢𝑛 = lim =0
𝑛→∞ 𝑛→∞ 2

34
 INFINITE SERIES:
PROPERTIES: If σ∞
𝑘=1 𝑢 𝑘 = 𝑈, σ∞
𝑘=1 𝑣𝑘 = 𝑉 and 0 ≠ 𝑐 ∈ ℝ, then
– σ∞ ∞
𝑘=1(𝑢𝑘 +𝑣𝑘 ) = 𝑈 + 𝑉 and σ𝑘=1 𝑐𝑢𝑘 = 𝑐𝑈
– Addition or removal of finite number of terms to a series does not affect its convergence
and divergence. [Proofs are trivial. Hence H.W.]
CRITERION FOR CONVERGENCE:
– If σ∞ 𝑘=1 𝑢𝑘 converges, then for any 𝜖 > 0, there exists a positive integer 𝑛0 such that
∞
σ𝑘=𝑛 𝑢𝑘 < 𝜖 for all 𝑛 ≥ 𝑛0 .
– The series σ∞ 𝑘=1 𝑢𝑘 converges if and only if for any 𝜖 > 0, there exists a positive integer 𝑛0
such that 𝑢𝑚+1 + 𝑢𝑚+2 + ⋯ + 𝑢𝑛 < 𝜖 for all 𝑚, 𝑛 ≥ 𝑛0 .
Proof: Let 𝑆 = σ∞ 𝑛
𝑘=1 𝑢𝑘 and 𝑆𝑛 = σ𝑘=1 𝑢𝑘 so that sequence {𝑆𝑛 } converges to 𝑆. Then, for 𝜖 > 0
there exists a positive integer 𝑛0 such that
𝑆 − 𝑆𝑛−1 = σ∞ 𝑛−1 ∞
𝑘=1 𝑢𝑘 − σ𝑘=1 𝑢𝑘 = σ𝑘=𝑛 𝑢𝑘 < 𝜖 for all 𝑛 ≥ 𝑛0 .
Also, the series σ∞𝑘=1 𝑢𝑘 is convergent iff the sequence {𝑆𝑛 } is convergent iff it is a Cauchy
sequence. Therefore, the series σ∞
𝑘=1 𝑢𝑘 is convergent iff there is a positive integer 𝑛0 such that
σ𝑛𝑘=1 𝑢𝑘 − σ𝑚𝑘=1 𝑢𝑘 = 𝑢𝑚+1 + 𝑢𝑚+2 + ⋯ + 𝑢𝑛 = 𝑆𝑛 − 𝑆𝑚 < 𝜖 for all 𝑚, 𝑛 ≥ 𝑛0 .

SERIES OF POSITIVE TERMS: If terms of a series σ∞ 𝑘=1 𝑢𝑘 are positive (or non-
negative), then sequence of partial sums {𝑆𝑛 } is increasing since 𝑆𝑛+1 = σ𝑛+1 𝑘=1 𝑢𝑘 =
𝑛
σ𝑘=1 𝑢𝑘 + 𝑢𝑛+1 = 𝑆𝑛 + 𝑢𝑛+1 ≥ 𝑆𝑛 . It is sufficient for the series to be convergent that {𝑆𝑛 }
is bounded.
35
 INFINITE SERIES (Convergence Tests):
INTEGRAL TEST: Let σ∞ 𝑘=1 𝑢𝑘 be a series of positive terms, and suppose
there is a positive decreasing function 𝑓 defined on 1, ∞ with 𝑓(𝑛) = 𝑎𝑛 for all
∞
𝑛 ∈ ℕ. Then σ∞ 𝑢
𝑘=1 𝑘 is convergent if and only if ‫׬‬1
𝑓 𝑥 𝑑𝑥 < ∞.
𝑡
Proof: Define 𝑇 𝑡 = ‫׬‬1 𝑓 𝑥 𝑑𝑥 , 𝑡 ≥ 1 and let 𝑆𝑛 = σ𝑛𝑘=1 𝑢𝑘 so that 𝑆𝑛 ≤ 𝑆𝑛+1 .
➢ 𝑇 𝑡 is an increasing function since for 𝑡2 > 𝑡1 ,
𝑡 𝑡 𝑡
𝑇 𝑡2 − 𝑇 𝑡1 = ‫׬‬1 2 𝑓 𝑥 𝑑𝑥 − ‫׬‬1 1 𝑓 𝑥 𝑑𝑥 = ‫ 𝑡׬‬2 𝑓 𝑥 𝑑𝑥 ≥ 0 as 𝑓 𝑥 ≥ 0.
1

Now in any interval 𝑘 − 1, 𝑘 , 𝑘 = 2, 3, … we have

𝑘 𝑘 𝑘
𝑢𝑘 = 𝑓 𝑘 ≤ 𝑓 𝑥 ≤ 𝑓 𝑘 − 1 = 𝑢𝑘−1 so that ‫𝑘׬‬−1 𝑢𝑘 𝑑𝑥 ≤ ‫𝑘׬‬−1 𝑓(𝑥) 𝑑𝑥 ≤ ‫𝑘׬‬−1 𝑢𝑘−1 𝑑𝑥
𝑘
➢ 𝑢𝑘 ≤ ‫𝑘׬‬−1 𝑓 𝑥 𝑑𝑥 ≤ 𝑢𝑘−1 for 𝑘 = 2, 3, … (1)
∞
Suppose that ‫׬‬1 𝑓 𝑥 𝑑𝑥 converges and has value 𝛽. Now, from (1)
2 3 𝑛 𝑛 ∞
𝑆𝑛 − 𝑎1 = 𝑎2 + 𝑎3 + ⋯ + 𝑎𝑛 ≤ ‫׬‬1 𝑓 𝑥 𝑑𝑥 + ‫׬‬2 𝑓 𝑥 𝑑𝑥 + ⋯ + ‫𝑛׬‬−1 𝑓 𝑥 𝑑𝑥 = ‫׬‬1 𝑓 𝑥 𝑑𝑥 ≤ ‫׬‬1 𝑓 𝑥 𝑑𝑥 = 𝛽
➢ 𝑆𝑛 ≤ 𝑎1 + 𝛽, and {𝑆𝑛 } being increasing is convergent.
Conversely, suppose that σ∞
𝑘=1 𝑢𝑛 converges to 𝑆. Then, from (1)
𝑛 2 3 𝑛
‫׬‬1 𝑓 𝑥 𝑑𝑥 ≤ ‫׬‬1 𝑓 𝑥 𝑑𝑥 + ‫׬‬2 𝑓 𝑥 𝑑𝑥 + ⋯ + ‫𝑛׬‬−1 𝑓 𝑥 𝑑𝑥 = 𝑎1 + 𝑎2 + ⋯ + 𝑎𝑛 = 𝑆𝑛
∞ ∞
➢ ‫׬‬1 𝑓 𝑥 𝑑𝑥 ≤ lim 𝑆𝑛 = 𝑆. Hence, ‫׬‬1 𝑓 𝑥 𝑑𝑥 is finite.
𝑛→

36
INFINITE SERIES (Convergence Tests):

y = f(x)

𝑢1
𝑢2 𝑢1 𝑢8
𝑢8 𝑢2

∞ ∞
‫׬‬1 𝑓 𝑥 𝑑𝑥 ≥ σ∞ 𝑢 ,
𝑛=1 𝑛 1‫׬‬ 𝑓 𝑥 𝑑𝑥 ≤ σ ∞
𝑛=1 𝑢𝑛
(Integral Test)

37
 INFINITE SERIES (Convergence Tests):
1 log 𝑛
EXAMPLE 1: Discuss the convergence of σ∞
𝑛=1 and σ∞
𝑛=1 .
𝑛𝑝 𝑛2
∞, 𝑝 < 0
1
In first series, lim 𝑢𝑛 = lim 𝑝 = ቐ 1, 𝑝 = 0 .
𝑛→∞ 𝑛→∞ 𝑛
0, 𝑝 > 0
1
➢ σ∞
𝑛=1 is divergent for 𝑝 ≤ 0.
𝑛𝑝
1 1
For 𝑝 > 0, we have 𝑢𝑛 = 𝑝 is positive and decreasing. Let 𝑓 𝑥 = 𝑝 (𝑝 > 0),
𝑛 𝑥
which is positive and decreasing on [1, ∞). Also,
∞ ∞
∞, 0 < 𝑝 ≤ 1
‫׬‬1 𝑓 𝑥 𝑑𝑥 = ‫׬‬1 𝑥 −𝑝 𝑑𝑥 =ቐ 1
, 𝑝>1
𝑝−1
1
Hence the series σ∞
𝑛=1 is convergent for 𝑝 > 1 and divergent 𝑝 ≤ 1.
𝑛𝑝
log 𝑛 log 𝑥
In case of the series σ∞
𝑛=2 , let 𝑓 𝑥 = > 0 on 𝑥 ∈ [2, ∞).
𝑛2 𝑥2
−2log 𝑥 1 1−2 log 𝑥 log 𝑛
For 𝑥2 > 𝑥1 ∈ [1, ∞), we have 𝑓 ′ 𝑥 = 3 + 3= < 0 for all 𝑥 ≥ 2. Also 𝑢𝑛 = →
𝑥 𝑥 𝑥3 𝑛2
∞ ∞ log 𝑥 log 𝑥 ∞ ∞ 1 log 2 1
0 as 𝑛 → ∞. Then, ‫׬‬2 𝑓 𝑥 𝑑𝑥 = ‫׬‬2 𝑥 2 𝑑𝑥 = − 𝑥 | 2 + ‫׬‬2 𝑥 2 𝑑𝑥 = 2 + 2 < ∞.
Hence by the integral test the series is convergent.

38
 INFINITE SERIES (Convergence Tests):
COMPARISION TESTS:
1. Let 0 ≤ 𝑢𝑛 ≤ 𝑣𝑛 for all 𝑛 ∈ ℕ. Then
𝑛 𝑛
▪ σ∞ ∞
𝑛=1 𝑣𝑛 is convergent implies σ𝑛=1 𝑢𝑛 is convergent [Hint: σ𝑘=1 𝑢𝑛 ≤ σ𝑘=1 𝑣𝑛 , for each 𝑛]
▪ σ∞ ∞
𝑛=1 𝑢𝑛 is divergent implies σ𝑛=1 𝑣𝑛 is divergent
𝑢𝑛
2. Let 𝑢𝑛 , 𝑣𝑛 ≥ 0 for all 𝑛 ∈ ℕ and lim > 0. Then σ∞
𝑛=1 𝑣𝑛 converges if and only if
𝑛→∞ 𝑣𝑛
σ∞
𝑛=1 𝑢𝑛 converges.
𝑢 𝜆
Proof: Let lim 𝑣𝑛 = 𝜆 > 0. Then for 𝜖 = 2 > 0, there is a positive integer 𝑛0 such that
𝑛→∞ 𝑛
𝑢𝑛 𝜆 𝜆 𝑢𝑛 3 𝜆 3
−𝜆 < 2 for all 𝑛 ≥ 𝑛0 ⇒ 2 ≤ ≤ 𝜆 or 𝑣 ≤ 𝑢 ≤ 𝜆𝑣𝑛 for all 𝑛 ≥ 𝑛0 .
𝑣𝑛 𝑣𝑛 2 2 𝑛 𝑛 2
3 𝜆 𝑛
Thus we have two relations σ𝑛𝑘=𝑛0 𝑢𝑘 ≤ 𝜆 σ𝑛𝑘=𝑛0 𝑣𝑘 and σ𝑘=𝑛0 𝑣𝑘 ≤ σ𝑛𝑘=𝑛0 𝑢𝑘 . Hence,
2 2
convergence/divergence of σ∞𝑛=1 𝑢𝑛 implies the convergence/divergence of σ∞ 𝑛=1 𝑣𝑛 .
𝑛+1 𝑛+1
EXAMPLE 2: Test the series σ∞ ∞
𝑛=1 𝑛2𝑛 and σ𝑛=1 𝑛2 +𝑛+1 for convergence.
𝑛+1 1 1 1 1
• Let 𝑢𝑛 = = 1+ ≤2 for all 𝑛 ∈ ℕ. We know that the series σ∞
𝑘=1 , being a
𝑛2𝑛 𝑛 2𝑛 2𝑛 2𝑛
1
geometric series with common ratio < 1, is convergent. Therefore, by comparison test,
2
𝑛+1
the series σ∞
𝑛=1 is convergent.
𝑛2𝑛
1 1
𝑛+1 1 1+ 1 𝑢 1+
• Let 𝑢𝑛 = = 𝑛
and 𝑣𝑛 = so that lim 𝑛 = lim 1
𝑛
1 = 1 > 0. Hence by
𝑛2 +𝑛+1 𝑛 1+ + 12
1 𝑛 𝑛→∞ 𝑣𝑛 𝑛→∞ 1+ + 2
𝑛 𝑛 𝑛 𝑛
𝑛+1 1
comparison test, the series σ∞
𝑛=1 is divergent since ∞
σ𝑛=1 is divergent.
𝑛2 +𝑛+1 𝑛 39
 INFINITE SERIES (Convergence Tests):
CAUCHY ROOT TEST: Let σ∞
𝑘=1 𝑢𝑘 be a series of positive terms. Then the
1 1
series converges or diverges according as lim 𝑢𝑛 𝑛 < 1 or lim 𝑢𝑛 𝑛 > 1. The
𝑛→∞ 𝑛→∞
1
test fails if lim 𝑢𝑛 𝑛 = 1.
𝑛→∞
1
Proof: Let lim 𝑢𝑛 𝑛 = 𝑙. Then for any 𝜖 > 0, there is positive integer 𝑛0 such that
𝑛→∞
1/𝑛 1/𝑛 𝑛 𝑛
𝑢𝑛 − 𝑙 < 𝜖 or 𝑙 − 𝜖 < 𝑢𝑛 < 𝑙 + 𝜖 or 𝑙 − 𝜖 < 𝑢𝑛 < 𝑙 + 𝜖 ∀ 𝑛 ≥ 𝑛0 . (1)
For 𝑙 < 1, we can choose 𝜖 > 0 in (1) such that 𝑙 + 𝜖 < 1 so that
σ𝑛𝑘=𝑛0 𝑢𝑘 < σ𝑛𝑘=𝑛0 𝑙 + 𝜖 𝑘 ∀ 𝑛 ≥ 𝑛0 .

Since the series σ∞ 𝑘

𝑘=𝑛0 𝑙 + 𝜖 , being a geometric series with common ratio less than 1, is
convergent, the series σ∞
𝑘=1 𝑢𝑘 is convergent by comparison test.
For 𝑙 > 1, we can choose 𝜖 > 0 in (1) such that 𝑙 − 𝜖 > 1 so that
σ𝑛𝑘=𝑛0 𝑢𝑘 > σ𝑛𝑘=𝑛0 𝑙 − 𝜖 𝑘 ∀ 𝑛 ≥ 𝑛0 .

Since the series σ∞ 𝑘

𝑘=𝑛0 𝑙 − 𝜖 , being a geometric series with common ratio greater than 1, is
divergent, the series σ∞
𝑘=1 𝑢𝑘 is divergent by comparison test.
1
1 1 𝑛
For 𝑙 = 1, we have both the possibilities. The series σ∞
𝑘=1 𝑘 is divergent but lim = 1. On the
𝑛→∞ 𝑛
1
1 1 𝑛
other hand, the series σ∞
𝑘=1 𝑘 2 is convergent but lim 2 = 1.
𝑛→∞ 𝑛

40
 INFINITE SERIES (Convergence Tests):
1
REMARK: If 𝑢𝑛 > 0 and lim 𝑢𝑛 𝑛 < 1, then the series σ∞
𝑛=1 𝑢𝑛 is convergent and hence
𝑛→∞
1
lim 𝑢𝑛 = 0. Thus, 𝑢𝑛 > 0 and 𝑙𝑖𝑚 𝑢𝑛 𝑛 < 1 implies that 𝑙𝑖𝑚 𝑢𝑛 = 0.
𝑛→∞ 𝑛→∞ 𝑛→∞
−1 −2
22 2 33 3
EXAMPLE 3: Discus convergence of the series − + −2 +⋯
1 1 23
−𝑛 1 −1
(𝑛+1)𝑛+1 𝑛+1 (𝑛+1)𝑛+1 𝑛+1 −1
Here 𝑢𝑛 = − so that lim 𝑢𝑛 =𝑛 lim − = 𝑒−1 < 1.
𝑛𝑛+1 𝑛 𝑛→∞ 𝑛→∞ 𝑛𝑛+1 𝑛
Hence by the Cauchy root test the given series is convergent.
∞ 𝑛+1 𝑛 𝑛
EXAMPLE 4: Discus convergence of the series σ𝑛=1 𝑥 , 𝑥 > 0.
𝑛+2
1
𝑛+1
Here lim 𝑢𝑛 𝑛 = lim 𝑥 = 𝑥. Hence by the Cauchy root test the given series converges if 𝑥 < 1,
𝑛→∞ 𝑛→∞ 𝑛+2
𝑛+1 𝑛
diverges if 𝑥 > 1 and test fails if 𝑥 = 1. For 𝑥 = 1, 𝑢𝑛 = → 𝑒 −1 ≠ 0 as 𝑛 → ∞.
𝑛+2
Thus series is divergent for 𝑥 = 1.
𝑢𝑛+1
D’Alembert Ratio Test: Let σ∞
𝑛=1 𝑢𝑛 be a series of positive terms such that lim = 𝑙.
𝑛→∞ 𝑢𝑛
Then the series converges if 𝑙 < 1, diverges 𝑙 > 1 and test fails if 𝑙 = 1.
𝑢
Proof: Since 𝑢𝑛 is positive for each 𝑛, lim 𝑢𝑛+1 = 𝑙 ≥ 0. For any 𝜖 > 0, there is positive
𝑛→∞ 𝑛
𝑢𝑛+1
integer 𝑛0 such that 𝑙−𝜖 < < 𝑙 + 𝜖 for all 𝑛 ≥ 𝑚. (1)
𝑢𝑛

41
 INFINITE SERIES (Convergence Tests):
Putting 𝑛 = 𝑚, 𝑚 + 1, … , 𝑛 − 1 in (1) and multiplying all the 𝑛 − 𝑚 inequalities, we get
𝑛−𝑚 𝑢 𝑛−𝑚 𝑢𝑚 𝑛 𝑢𝑚 𝑛
𝑙−𝜖 < 𝑢𝑛 < 𝑙 + 𝜖 or 𝑙−𝜖 < 𝑢𝑛 < 𝑙+𝜖 ∀ 𝑛 ≥ 𝑚. (1)
𝑚 𝑙−𝜖 𝑚 𝑙+𝜖 𝑚
For 𝑙 < 1, we can choose 𝜖 > 0 in (1) such that 𝑙 + 𝜖 < 1 so that
𝑢𝑚
σ𝑛𝑘=𝑚 𝑢𝑘 < 𝑚
σ𝑛𝑘=𝑚 𝑙 + 𝜖 𝑘 ∀ 𝑛 ≥ 𝑚.
𝑙+𝜖

Since the series σ∞ 𝑘

𝑘=𝑚 𝑙 + 𝜖 , being a geometric series with common ratio less than 1,
is convergent, the series σ∞
𝑘=1 𝑢𝑘 is convergent by comparison test.
For 𝑙 > 1, we can choose 𝜖 > 0 in (1) such that 𝑙 − 𝜖 > 1 so that
𝑢
σ𝑛𝑘=𝑚 𝑢𝑘 > 𝑚 𝑚 σ𝑛𝑘=𝑚 𝑙 − 𝜖 𝑘 ∀ 𝑛 ≥ 𝑚.
𝑙−𝜖

Since the series σ∞ 𝑘

𝑘=𝑚 𝑙 − 𝜖 , being a geometric series with common ratio greater than
1, is divergent, the series σ∞
𝑘=1 𝑢𝑘 is divergent by comparison test.
1 𝑢𝑛+1
For 𝑙 = 1, we have both the possibilities. The series σ∞𝑘=1 𝑘 is divergent and lim =
𝑢 𝑛→∞ 𝑛
1 𝑢
1. On the other hand, the series σ∞
𝑘=1 𝑘 2 is convergent and lim 𝑢𝑛+1 = 1.
𝑛→∞ 𝑛
𝑢𝑛+1
REMARK: If 𝑢𝑛 > 0 and lim < 1, then the series σ∞
𝑛=1 𝑢𝑛 is convergent and hence
𝑛→∞ 𝑢𝑛
𝑢𝑛+1
lim 𝑢𝑛 = 0. Thus, 𝑢𝑛 > 0 and lim < 1 implies that 𝑙𝑖𝑚 𝑢𝑛 = 0.
𝑛→∞ 𝑛→∞ 𝑢𝑛 𝑛→∞

42
 INFINITE SERIES (Convergence Tests):
1
EXAMPLE 5: Discus convergence of the series σ∞
𝑛=1 , 𝑥 > 0.
𝑥 𝑛 +𝑥 −𝑛
0, 0 < 𝑥 < 1
1 𝑥𝑛 1 𝑥𝑛
Here 𝑢𝑛 = 𝑛 −𝑛 = 2𝑛 so that lim 𝑢𝑛 = lim 2𝑛 = ൞ 2 , 𝑥=1
𝑥 +𝑥 𝑥 +1 𝑛→∞ 𝑛→∞ 𝑥 +1
0, 𝑥>1
Therefore, the series is divergent for 𝑥 = 1. For other values of 𝑥, we have
𝑢𝑛+1 𝑥 𝑛+1 𝑥 2𝑛 +1
1
𝑥 1+ 2𝑛 1 𝑢𝑛+1 𝑥, 0 < 𝑥 < 1
𝑥
= 2(𝑛+1) = 2 1 , that is, lim = ൝1
𝑢𝑛 𝑥 +1 𝑥 𝑛 𝑥 + 2𝑛 𝑛→∞ 𝑢𝑛 < 1, 𝑥 > 1
𝑥 𝑥
Hence by the D’ Alembert ratio test is convergent for 𝑥 ≠ 1.
𝑥2 𝑥4
1
EXAMPLE 6: Discus convergence of the series + + + ⋯ , 𝑥 > 0.
3 2 2
4 3
𝑥 2(𝑛−1) 𝑢𝑛+1 𝑥 2𝑛 𝑛+1 𝑛
Here 𝑢𝑛 = so that lim = lim 2(𝑛−1) = 𝑥2.
𝑛+1 √𝑛 𝑛→∞ 𝑢𝑛 𝑛→∞ 𝑛+2 𝑛+1 𝑥
Hence by the D’ Alembert ratio test is convergent for 𝑥 2 < 1, divergent for 𝑥 2 >
1 and test fails for 𝑥 2 = 1.
1 1 𝑢𝑛 1
For 𝑥 2 = 1, 𝑢𝑛 = = 3 1 so that lim 3 = lim 1 = 1 > 0.
𝑛+1 √𝑛 2 1+𝑛 𝑛→∞ 𝑛2 𝑛→∞ 1+𝑛
𝑛
1
By comparison test given series is convergent since σ∞
𝑛=1 𝑛3/2 converges.
43
 INFINITE SERIES (Convergence Tests):
RABBE’S TEST: Let σ∞
𝑛=1 𝑢𝑛 be a series of positive terms such that
𝑢𝑛
lim 𝑛 − 1 = 𝑙. Then, series converges if 𝑙 > 1, series diverges if 𝑙 < 1
𝑛→∞ 𝑢𝑛+1
and test fails if 𝑙 = 1.
4.7….(3𝑛+1) 𝑛
EXAMPLE 7: Test the series σ∞
𝑛=1 𝑥 , 𝑥 > 0 for convergence.
1.2….𝑛
𝑢𝑛+1 4.7….(3𝑛+1)(3𝑛+4) 𝑛+1 1.2.3…𝑛 1 𝑥 3𝑛+4
Here, = 𝑥 = → 3𝑥 as 𝑛 → ∞.
𝑢𝑛 1.2….𝑛(𝑛+1) 4.7….(3𝑛+1) 𝑥 𝑛 𝑛+1
1
Hence by the D’Alembert ratio test the series is convergent if 3𝑥 < 1 or 𝑥 < and
3
1 1 1
divergent if 𝑥 > 3 . The test fails for 𝑥 = 3 and we shall use the Rabbe’s test. For 𝑥 = 3,
𝑢𝑛 3𝑛+3 −1 1
we have 𝑛 −1 =𝑛 −1 =𝑛 → − 3 as 𝑛 → ∞.
𝑢𝑛+1 3𝑛+4 3𝑛+4
1
By Rabbe’s test series is divergent for 𝑥 =3.

Note: From the application point of view the D’Alembert ratio test can be stated
as: a series σ∞
𝑛=1 𝑢𝑛 of positive terms converges or diverges according as
𝑢 𝑢
lim 𝑛 > 1 or lim 𝑛 < 1 and test fails if 𝑙 = 1.
𝑛→∞ 𝑢𝑛+1 𝑛→∞ 𝑢𝑛+1

44
 INFINITE SERIES (Convergence Tests):
LOGARITHMIC TEST: Let σ∞
𝑛=1 𝑢𝑛 be a series of positive terms such that
𝑢𝑛
lim 𝑛log = 𝑙. Then series converges if 𝑙 > 1, diverges if 𝑙 < 1 and test
𝑛→∞ 𝑢𝑛+1
fails if 𝑙 = 1.
De Morgan’s Test: Let σ∞
𝑛=1 𝑢𝑛 be a series of positive terms such that
𝑢𝑛
lim 𝑛 − 1 − 1 log𝑛 = 𝑙. Then series converges if 𝑙 > 1, diverges if 𝑙 <
𝑛→∞ 𝑢𝑛+1
1 and test fails if 𝑙 = 1.
𝑎 𝑎(𝑎+1) 𝑎(𝑎+1)(𝑎+2)
EXAMPLE 8: Test convergence of the series + + +
𝑏 𝑏(𝑏+1) 𝑏(𝑏+1)(𝑏+2)
⋯ , 𝑎, 𝑏 > 0.
𝑎 𝑎+1 𝑎+2 …(𝑎+𝑛−1) 𝑢𝑛 (𝑏+𝑛)
Here 𝑢𝑛 = so that = → 1 as 𝑛 → ∞. This means
𝑏 𝑏+1 𝑏+2 …(𝑏+𝑛−1) 𝑢𝑛+1 (𝑎+𝑛)
by the D’Alembert ratio test we can say nothing about convergence of the
𝑢𝑛 𝑏−𝑎
series. Now, lim 𝑛 − 1 = lim 𝑛 = 𝑏 − 𝑎. Hence by Rabbe’ s test
𝑛→∞ 𝑢𝑛+1 𝑛→∞ 𝑎+𝑛
the series is convergent for 𝑏 − 𝑎 > 1, divergent for 𝑏 − 𝑎 < 1 and test fails for
𝑢𝑛
𝑏 − 𝑎 = 1. Applying the De Morgan’s test we have lim 𝑛 − 1 − 1 log𝑛
𝑛→∞ 𝑢𝑛+1
45
 INFINITE SERIES (Convergence Tests):
𝑢𝑛 𝑛 −𝑎
lim 𝑛 − 1 − 1 log𝑛 = lim − 1 log 𝑛 = lim log 𝑛 = 0 < 1.
𝑛→∞ 𝑢𝑛+1 𝑛→∞ 𝑎+𝑛 𝑛→∞ 𝑎+𝑛
Hence the series is convergent for 𝑏 − 𝑎 = 1.
𝑥 22 𝑥 2 33 𝑥 3
EXAMPLE 9: Test convergence of the series 1 + + + + ⋯.
1! 2! 3!
𝑛𝑛 𝑥 𝑛 𝑢𝑛 1 1
If we leave the first term, 𝑢𝑛 = so that = 1 𝑛
→ as 𝑛 → ∞.
𝑛! 𝑢𝑛+1 1+ 𝑥 𝑒𝑥
𝑛
1
This means by the D’Alembert ratio test, the series is convergent for > 1 or
𝑒𝑥
1 1 1
𝑥 < , diverges for 𝑥 > and test fails for 𝑥 = .
𝑒 𝑒 𝑒
1
For 𝑥 = , we have by logarithmic test:
𝑒
𝑢𝑛 𝑒 1
lim 𝑛log = lim 𝑛log 1 𝑛
= lim 𝑛 − 𝑛log 1 + = ∞ > 1.
𝑛→∞ 𝑢𝑛+1 𝑛→∞ 1+𝑛 𝑛→∞ 𝑛

Hence the series is convergent.

Euler’s constant: The Euler’s constant, also called the Euler-Mascheroni
1
constant, is defined as the limiting difference of the series σ∞
𝑛=1 and log 𝑛, 𝑛
1 ∞ 1 1
that is, 𝛾 = lim σ𝑛𝑘=1 − log 𝑛 = ‫׬‬1 − 𝑑𝑥.
𝑛→∞ 𝑘 𝑥 [𝑥]

46
 Alternating/Mixed Series
Definition: A series of the form σ∞
𝑛=1(−1)
𝑛+1 𝑢 = 𝑢 − 𝑢 + 𝑢 + ⋯ is called
𝑛 1 2 3
an alternating series.
∞ −1 𝑛+1
Examples: σ𝑛=1 and σ∞ 𝑛
𝑛=1 −1 .
𝑛
Alternating Series Test: If 𝑢𝑛+1 ≤ 𝑢𝑛 for all 𝑛 ∈ ℕ and lim 𝑢𝑛 = 0, then the
𝑛→∞
alternating series σ∞
𝑛=1(−1)
𝑛+1 𝑢
𝑛 converges.
Proof: Let us define 𝑆𝑛 = σ∞
𝑛=1(−1)
𝑛+1 𝑢 so that for any natural number 𝑛,
𝑛
2𝑛+2 𝑢 2𝑛+3 𝑢
𝑆2𝑛+2 = 𝑆2𝑛 + −1 2𝑛+1 + −1 2𝑛+2 = 𝑆2𝑛 + 𝑢2𝑛+1 − 𝑢2𝑛+2
and 𝑆2𝑛+1 = 𝑆2𝑛−1 − 𝑢2𝑛 + 𝑢2𝑛+1 = 𝑆2𝑛−1 − 𝑢2𝑛 + 𝑢2𝑛+1 .
From the hypothesis, 𝑢2𝑛+1 − 𝑢2𝑛+2 ≥ 0 and 𝑢2𝑛+1 − 𝑢2𝑛 ≤ 0.
⇒ 𝑆2𝑛+2 ≥ 𝑆2𝑛 , 𝑆2𝑛+1 ≤ 𝑆2𝑛−1 and 𝑆2𝑛 = 𝑆2𝑛−1 − 𝑢2𝑛 , 𝑖. 𝑒. , 𝑆2𝑛 ≥ 𝑆2𝑛−1 for all 𝑛 ∈ ℕ.
⇒ 𝑆2 ≤ 𝑆2𝑛 ≤ 𝑆2𝑛−1 ≤ 𝑆1 for all 𝑛 ∈ ℕ.
Thus the sequences, {𝑆2𝑛−1 }is decreasing and bounded below by 𝑆2 , and {𝑆2𝑛 } is
increasing and bounded below by 𝑆1 . Hence both the subsequences are convergent.
Let lim 𝑆2𝑛−1 = 𝛼 and lim 𝑆2𝑛 = 𝛽 . Since, lim 𝑆2𝑛 = lim 𝑆2𝑛−1 − lim 𝑢2𝑛 =
𝑛→∞ 𝑛→∞ 𝑛→∞ 𝑛→∞ 𝑛→∞
∞
σ𝑛=1(−1)𝑛+1 𝑢𝑛
lim 𝑆2𝑛−1 , 𝛽 = 𝛼 = 𝑆, say, lim 𝑆𝑛 = 𝑆. Hence, converges.
𝑛→∞ 𝑛→∞

47
 Alternating/Mixed Series
Corollary: Under the hypothesis of the above theorem, if 𝑆 = σ∞
𝑛=1(−1)
𝑛+1 𝑢
𝑛
𝑛 𝑛+1
and 𝑆𝑛 = σ𝑘=1(−1) 𝑢𝑛 , then 𝑆 − 𝑆𝑛 ≤ 𝑢𝑛+1 for all 𝑛 ∈ ℕ.
Proof: We have, from the above theorem, that {𝑆2𝑛 } and {𝑆2𝑛−1 } both are monotonic
and convergent with lim 𝑆2𝑛 = lim 𝑆2𝑛−1 = 𝑆. Also 𝑆2𝑛 ≤ 𝑆 ≤ 𝑆2𝑛−1 for all 𝑛 ∈ ℕ implies
𝑛→∞ 𝑛→∞
that 0 ≤ 𝑆 − 𝑆2𝑛 ≤ 𝑆2𝑛+1 − 𝑆2𝑛 = 𝑢2𝑛+1 and 0 ≤ 𝑆2𝑛−1 − 𝑆 ≤ 𝑆2𝑛−1 − 𝑆2𝑛 = 𝑢2𝑛 .
⇒ 𝑆 − 𝑆𝑛 ≤ 𝑢𝑛+1 for all 𝑛 ∈ ℕ.

∞ −1 𝑛+1 1
Examples: 1. σ𝑛=1 is convergent as 𝑢𝑛 = 𝑛 decreases to 0.
𝑛
∞ −1 𝑛+1 𝑛 −1 𝑛+1 𝑛
2. σ𝑛=1 is divergent as convergent as lim 𝑢𝑛 ≠ 0. But
𝑛+1 𝑛→∞ 𝑛+1
cannot be decided by the alternating series test.
Theorem: Let σ∞ ∞
𝑛=1 𝑢𝑛 be a series of real numbers. If the series σ𝑛=1 |𝑢𝑛 | converges,
then so does the series σ∞ ∞ ∞
𝑛=1 𝑢𝑛 , and σ𝑛=1 𝑢𝑛 ≤ σ𝑛=1 |𝑢𝑛 |.
Proof: For any 𝑛 ∈ ℕ, we have 0 ≤ 𝑢𝑛 − 𝑢𝑛 ≤ 2 𝑢𝑛 . Since, σ∞ 𝑛=1 |𝑢𝑛 | is convergent,
by comparison test, σ∞ 𝑛=1(|𝑢𝑛 | − 𝑢𝑛 ) is convergent. Now,
σ𝑛𝑘=1 𝑢𝑘 = σ𝑛𝑘=1( 𝑢𝑘 − 𝑢𝑘 + 𝑢𝑘 ) = σ𝑛𝑘=1 𝑢𝑘 − σ𝑛𝑘=1( 𝑢𝑘 − 𝑢𝑘 ) for all 𝑛 ∈ ℕ.
⇒ σ∞ ∞ ∞ ∞
𝑘=1 𝑢𝑘 = σ𝑘=1( 𝑢𝑘 − 𝑢𝑘 + 𝑢𝑘 ) = σ𝑘=1 𝑢𝑘 − σ𝑘=1( 𝑢𝑘 − 𝑢𝑘 ), being difference of
two convergent series is convergent.
48
 Alternating/Mixed Series
Next for any 𝑚 ∈ ℕ, we have
𝑢1 + 𝑢2 + ⋯ + 𝑢𝑚 ≤ 𝑢1 + 𝑢2 + ⋯ + 𝑢𝑚
⇒ σ𝑚 𝑚 ∞ ∞
𝑘=1 𝑢𝑘 ≤ σ𝑘=1 𝑢𝑘 ⇒ σ𝑛=1 𝑢𝑛 ≤ σ𝑛=1 |𝑢𝑛 |.
sin 𝑛 sin 𝑛 1
Examples: 1. σ∞
𝑛=1 is a mixed convergent series. Here ≤ for all
𝑛2 𝑛2 𝑛2
sin 𝑛
𝑛 ∈ ℕ . By comparison test, the series σ∞
𝑛=1 is convergent. Hence,
𝑛2
sin 𝑛
σ∞
𝑛=1 is also convergent.
𝑛2
∞ −1 𝑛+1 −1 𝑛+1 1
2. σ𝑛=1 is convergent but σ∞
𝑛=1 = σ∞
𝑛=1 is divergent.
𝑛 𝑛 𝑛
The above examples lead to the following definitions.
• We say that the series σ∞ ∞
𝑛=1 𝑢𝑛 converges absolutely, if σ𝑛=1 |𝑢𝑛 |
converges.
• If σ∞ ∞ ∞
𝑛=1 𝑢𝑛 converges but σ𝑛=1 |𝑢𝑛 | diverges, then we say that σ𝑛=1 𝑢𝑛
converges conditionally.
−1 𝑛 𝑛 −1 𝑛+1
Exercise: The series σ∞𝑛=1 is absolutely convergent and the series σ∞
𝑛=1 is
𝑛3 +1 𝑛
conditionally convergent.
49
 Summation by parts
Theorem 1: Let 𝑎𝑛 ∞ ∞
𝑛=1 and 𝑏𝑛 𝑛=1 be two sequences of real numbers and
let 𝑠𝑛 = 𝑎1 + ⋯ + 𝑎𝑛 . Then, for each 𝑛 ∈ ℕ,
σ𝑛𝑘=1 𝑎𝑘 𝑏𝑘 = 𝑠𝑛 𝑏𝑛+1 − σ𝑛𝑘=1 𝑠𝑘 𝑏𝑘+1 − 𝑏𝑘 (1)
Proof: Define 𝑠0 = 0. Since 𝑎𝑘 = 𝑠𝑘 − 𝑠𝑘−1 , we have
σ𝑛𝑘=1 𝑎𝑘 𝑏𝑘 = σ𝑛𝑘=1 𝑏𝑘 𝑠𝑘 − 𝑠𝑘−1
= 𝑏1 𝑠1 − 𝑠0 + 𝑏2 𝑠2 − 𝑠1 + ⋯ + 𝑏𝑛−1 𝑠𝑛−1 − 𝑠𝑛−2 + 𝑏𝑛 𝑠𝑛 − 𝑠𝑛−1
= 𝑠1 𝑏1 − 𝑏2 + 𝑠2 𝑏2 − 𝑏3 + ⋯ + 𝑠𝑛−1 𝑏𝑛−1 − 𝑏𝑛 + 𝑠𝑛 𝑏𝑛
= − σ𝑛−1
𝑘=1 𝑠𝑘 𝑏𝑘+1 − 𝑏𝑘 − 𝑠𝑛 (𝑏𝑛+1 −𝑏𝑛 ) + 𝑠𝑛 𝑏𝑛+1
= − σ𝑛𝑘=1 𝑠𝑘 𝑏𝑘+1 − 𝑏𝑘 + 𝑠𝑛 𝑏𝑛+1 ,
which proves (1).
∞
If we introduce the notation ∆𝑎𝑘 = 𝑎𝑘+1 − 𝑎𝑘 for any sequence 𝑎𝑘 𝑘=1 , then
𝑎 −𝑎
∆𝑎𝑘 = 𝑘+1 𝑘, resembles a “derivative of 𝑎𝑘 with respect to 𝑘.”
𝑘+1 −𝑘
The formula (1) becomes σ𝑛𝑘=1 𝑏𝑘 ∆𝑠𝑘−1 = 𝑠𝑛 𝑏𝑛+1 − σ𝑛𝑘=1 𝑠𝑘 ∆𝑏𝑘 . 2
This resembles the formula
𝑑 𝑑
‫ 𝑐𝑑|𝑏𝑠 = 𝑠𝑑 𝑏 𝑐׬‬− ‫𝑏𝑑 𝑠 𝑐׬‬.
The formula (1) is thus sometimes called summation by parts.
50
 Consequences of summation by parts
A consequence of Theorem 1 is the following result called Abel’s Lemma. It,
in turn, yields a new test for convergence and, in a later section, a theorem
on summation of series.
Abel’s Lemma: If 𝑎𝑛 ∞ 𝑛=1 is a sequence of real numbers whose partial sums
𝑠𝑛 = σ𝑛𝑘=1 𝑎𝑘 satisfy 𝑚 ≤ 𝑠𝑛 ≤ 𝑀. For some 𝑚, 𝑀 ∈ ℝ, and if 𝑏𝑛 ∞ 𝑛=1 is a non-
increasing sequence of nonnegative numbers, then
𝑚𝑏1 ≤ σ𝑛𝑘=1 𝑎𝑘 𝑏𝑘 ≤ 𝑀𝑏1 𝑛 ∈ ℕ . 3
Proof: From (1) of Theorem 1, we have
σ𝑛𝑘=1 𝑎𝑘 𝑏𝑘 = σ𝑛𝑘=1 𝑠𝑘 𝑏𝑘 − 𝑏𝑘+1 + 𝑠𝑛 𝑏𝑛+1 .
Since, by hypothesis, 𝑏𝑘 − 𝑏𝑘+1 ≥ 0 and 𝑠𝑘 ≤ 𝑀, we have
σ𝑛𝑘=1 𝑎𝑘 𝑏𝑘 ≤ 𝑀 σ𝑛𝑘=1 𝑏𝑘 − 𝑏𝑘+1 + 𝑀𝑏𝑛+1
= 𝑀 𝑏1 − 𝑏2 + 𝑏2 − 𝑏3 + ⋯ + 𝑏𝑛 − 𝑏𝑛+1 + 𝑏𝑛+1 = 𝑀𝑏1 .
This proves the right hand inequality in (3). The left-hand inequality may be
proved similarly. Thus, in Abel’s Lemma, if σ𝑛𝑘=1 𝑎𝑘 ≤ 𝑀 for all 𝑛 then
σ𝑛𝑘=1 𝑎𝑘 𝑏𝑘 ≤ 𝑀𝑏1 . The following consequence of Abel’s Lemma is called the
Dirichlet’s test.
51
 Dirichlet’s Test
Theorem 2: Let 𝑎𝑛 ∞ 𝑛=1 be a sequence of real numbers whose partial sums
𝑠𝑛 = σ𝑛𝑘=1 𝑎𝑘 form a bounded sequence, and let 𝑏𝑛 ∞ 𝑛=1 be a non-increasing
sequence of non-negative numbers which converges to 0. Then σ∞ 𝑘=1 𝑎𝑘 𝑏𝑘
converges.
Proof: It is sufficient to prove that the partial sums of σ∞𝑘=1 𝑎𝑘 𝑏𝑘 form a Cauchy
sequence. That is, given 𝜖 > 0, we must find 𝑛0 ∈ ℕ such that
σ𝑛𝑘=𝑚 𝑎𝑘 𝑏𝑘 < 𝜖 𝑛 ≥ 𝑚 ≥ 𝑛0 .
Now, by hypothesis, there exists 𝑀 > 0 such that 𝑠𝑛 ≤ 𝑀 𝑛 ∈ ℕ . Hence, for
any 𝑚, 𝑛 ∈ ℕ,
σ𝑛𝑘=𝑚 𝑎𝑘 = 𝑠𝑛 − 𝑠𝑚−1 ≤ 𝑠𝑛 + 𝑠𝑚−1 ,
and so
σ𝑛𝑘=𝑚 𝑎𝑘 ≤ 2𝑀 𝑚, 𝑛 ∈ ℕ; 𝑚 ≤ 𝑛 .
By Abel’s Lemma (applied to 𝑎𝑘 ∞ ∞
𝑘=𝑚 and 𝑏𝑘 𝑘=𝑚 ) this implies
σ𝑛𝑘=𝑚 𝑎𝑘 𝑏𝑘 ≤ 2𝑀𝑏𝑚 𝑚, 𝑛 ∈ ℕ; 𝑚 ≤ 𝑛 . (4)
Now, since 𝑏𝑛 → 0 as 𝑛 → ∞, there is 𝑛0 ∈ 𝑁 such that
𝜀
𝑏𝑛 − 0 = 𝑏𝑛 ≤ ∀𝑛 ≥ 𝑛0 .
2𝑀
Hence from (4), σ𝑛𝑘=𝑚 𝑎𝑘 𝑏𝑘 ≤ 𝜀 ∀ 𝑛 ≥ 𝑚 > 𝑛0 , which is what we wished to
show.
52
 Application
sin 𝑛𝑥 1
Consider the series σ∞
𝑛=1 ,𝑥 ∈ ℝ. Let 𝑎𝑘 = sin 𝑘𝑥 and 𝑏𝑘 = . Note that
𝑛 𝑘
1
• 𝑏𝑘 = is non-negative, non-increasing and tends to 0 as 𝑘 → ∞.
𝑘
𝑥 2𝑛+1
cos −cos 𝑥
• 𝑠𝑛 = σ𝑛𝑘=1 sin 𝑘𝑥 = 2
𝑥
2
, 𝑥 ≠ 2𝑚𝜋.
2sin2
𝑥
• We see that if sin ≠ 0, then 𝑠𝑛 ≤ 1/| sin (𝑥Τ2) |
2
• Hence by the Dirichlet test the given series is convergent.

53