MAL 100: Calculus

Lecture Notes
1 Sequences of real numbers
1.1 Real number system
We are familiar with natural numbers and to some extent the rational numbers. While
nding roots of algebraic equations we see that rational numbers are not enough to rep-
resent roots which are not rational numbers. For example draw the graph of y = x
2
2.
We see that it cross the x-axis twice. The roots are such that their square is 2, but they
cannot be rational numbers according to the following theorem.
Theorem 1.1.1. Suppose that a
0
, a
1
, ...., a
n
(n 1) are integers such that a
0
= 0, a
n
= 0
and that r satises the equation
a
n
x
n
+ a
n1
x
n1
+ .... + a
1
x + a
0
= 0.
If r =
p
q
where p, q are integers with no common factors and q = 0. Then q divides a
n
and p divides a
0
.
This theorem tells us that only rational candidates for solutions of the above equation
have the form
p
q
where p divides a
0
and q divides a
n
.
Proof: Since
p
q
satises the equation, we have
a
n
p
n
+ a
n1
p
n1
q + ... + a
0
q
n
= 0
i.e., a
n
p
n
= q(a
n1
p
n1
+... +a
0
q
n1
). This means q divides a
n
as p, q have no common
factors. On the other hand we can also write
a
0
q
n
= p(a
n
p
n1
+ a
n1
p
n2
+ ... + a
1
q
n1
).
Thus p divides a
0
. ///
Now we see that the possible rational roots of x
2
2 = 0 are 1, 2. But it is easy to
check that 1, 2 does not satisfy x
2
2 = 0. So the roots of x
2
2 = 0 are not rational
numbers. This means the set of rational numbers has gaps. So the natural question
to ask is: Can we have a number system without these gaps? The answer is yes and
the complete number system with out these gaps is the real line R. We will not look
1
into the development of R as it is not easy to dene the real numbers. We assume that
there is a set R, whose elements are called real numbers and R is closed with respect to
addition and multiplication. That is, given any a, b R, the sum a + b and product ab
also represent real numbers. Moreover, R has an order structure and has no gaps in
the sense that it satises the Completeness Axiom(see below).
Let S be a non-empty subset of R. If S contains a largest element s
0
, then we call s
0
the maximum of S. If S contains a smallest element s
0
, then we call s
0
the minimum of
S. If S is bounded above and S has least upper bound, then we call it the supremum
of S. If S is bounded below and S has greatest lower bound, then we call it as inmum of S.
Note that any bounded subset of Natural numbers has maximum and minimum.
Completeness Axiom: Every nonempty subset S of R that is bounded above has a
least upper bound. In other words, sup S exists and is a real number.
The completeness axiom does not hold for Q. That is, every non-empty subset of Q
that is bounded above by a rational number need not have rational least upper bound.
For example {r Q : 0 r, r
2
2}.
Archimedean property: For each x R, there exists a natural number N = N(x)
such that x < N.
Proof: Assume by contradiction that this is not true. Then there is no N N such that
x < N. i.e., x is an upper bound for N. Then, let u be the smallest such bound of N.
i.e., u = sup IN. Then u 1 N and u 1 < u. So m N such that u 1 < m but
m + 1 N and u < m + 1. This contradicts u = sup N.
Now it is easy to see the following corollary
Corollary: Let S = {
1
n
: n N}. Then w = inf S = 0.
Proof: We note that S is bounded below. Let > 0 be an arbitrary positive real number.
By above Archimedean property, there exists n N such that n >
1

. Then we have,
0 w
1
n
< .
Since is arbitrary, we have w = 0. (why?)
Corollary: If y > 0 be a real number, then there exists n = n(y) N such that
n 1 y < n.
Finally, we have the following density theorem
Theorem 1.1.2. Let x, y are real numbers such that x < y. Then there exists a rational
2
number q such that x < q < y.
Proof: W.l.g. assume that x > 0. Now let n N be such that y x >
1
n
. Now
consider the set
S = {m N :
m
n
> x}.
Then S is non-empty (by Archimedean property). By well -ordering of N, S has minimal
element say m
0
. Then x <
m
0
n
. By the minimality of m
0
, we see that
m
0
1
n
x. Then,
m
0
n
x +
1
n
< x + (y x) = y.
Therefore,
x <
m
0
n
< y.
1.2 Sequences and their limit
Denition 1.2.1. A sequence of real numbers is a function from the set of positive integers
to R.
Notation. It is customary to denote a sequence as {a
n
}

n=1
.
Examples 1.2.2. (i) {c}

n=1
, c R, (ii) {
(1)
n+1
n
}

n=1
, (iii) {
n1
n
}

n=1
and (iv) {

n}

n=1
.
Denition 1.2.3. A sequence {a
n
}

n=1
converges to limit L if for every > 0 (given)
there exists a positive integer N such that n N = |a
n
L| < .
Notation. L = lim
n
a
n
or a
n
L.
Examples 1.2.4.
(i) It is clear that the constant sequence {c}

n=1
, c R, has c as its limit.
(ii) Show that lim
n
1
n
= 0.
Solution. Let > 0 be given. In order to show that 1/n approaches 0, we must
show that there exists an integer N N such that for all n N,

1
n
0

=
1
n
< .
But 1/n < n > 1/. Thus, if we choose N N such that N > 1/, then for all
n N, 1/n < .
3
(iii) Consider the sequence {(1)
n+1
}

n=1
. It is intuitively clear that this sequence does
not have a limit or it does not approach to any real number. We now prove this
by denition. Assume to the contrary, that there exists an L R such that the
sequence {(1)
n+1
}

n=1
converges to L. Then for =
1
2
, there exists an N N such
that
|(1)
n+1
L| <
1
2
, n N. (1.1)
For n even, (1.1) says
| 1 L| <
1
2
, n N. (1.2)
while for n odd, (1.1) says
|1 L| <
1
2
, n N. (1.3)
which is a contradiction as 2 = |1 + 1| |1 L| +|1 + L| < 1.
Lemma 1.2.5. If {a
n
}

1
is a sequence and if both lim
n
a
n
= L and lim
n
a
n
= M holds,
then L = M.
Proof. Suppose that L = M. Then |L M| > 0. Let =
|LM|
2
. As lim
n
a
n
= L, there
exists N
1
N such that |a
n
L| < for all n N
1
. Also as lim
n
a
n
= M,there exists
N
2
N such that |a
n
M| < for all n N
2
. Let N = max{N
1
, N
2
}. Then for all
n N, |a
n
L| < and |a
n
M| < . Thus |LM| |a
n
L| +|a
n
M| < 2 = |LM|,
which is a contradiction. ///
Theorem 1.2.6 (Sandwich theorem for sequences). Let {a
n
}, {b
n
} and {c
n
} be three
sequences such that a
n
b
n
c
n
for all n N. If lim
n
a
n
= L and lim
n
c
n
= L, then
lim
n
b
n
= L.
Proof. Let > 0 be given. As lim
n
a
n
= L, there exists N
1
N such that
n N
1
= |a
n
L| < . (1.4)
Similarly as lim
n
c
n
= L, there exists N
2
N
n N
2
= |c
n
L| < . (1.5)
Let N = max{N
1
, N
2
}. Then, L < a
n
(from (1.4)) and c
n
L+ ( from (1.5)). Thus
L < a
n
b
n
c
n
L + .
Thus |b
n
L| < for all n N. Hence the proof. ///
4
Examples 1.2.7.
(i) Consider the sequence
_
cos n
n
_

n=1
. Then
1
n

cos n
n

1
n
. Hence by Sandwich
theorem lim
n
cos n
n
= 0.
(ii) As 0
1
2
n

1
n
and
1
n
0 as n ,
1
2
n
also converges to 0 by Sandwich theorem.
(iii) (1)
n
1
n
0 as
1
n
(1)
n
1
n

1
n
for all n 1.
(iv) If b > 0, then lim
n
n

b = 1.
Solution. First assume that b > 1. Let a
n
= b
1
n
1. As b > 1, a
n
> 0 for all n N.
Further,
b = (1 + a
n
)
n
=
n

k=0
n
C
k
a
k
n
1 + na
n
.
Then 0 a
n

b 1
n
. Thus a
n
0, i.e., b
1
n
1 as n .
Now if b < 1, then take c =
1
b
and it is easy to show the result. ///
Examples 1.2.8.
(i) lim
n
n

n = 1.
(ii) If x > 0 then lim
n
n
x
(1+x)
n
= 0.
(iii) If p > 0, then lim
n
log(n)
n
p
= 0.
Solution. (i) Let a
n
= n
1
n
1. Then 0 a
n
1 for all n N. Further,
n = (1 + a
n
)
n
=
n

k=0
n
C
k
a
k
n

n
C
2
a
2
n
=
n(n 1)
2
a
2
n
.
Thus 0 a
n

_
2
(n1)
(n 2). As
_
2
(n1)
0 as n , by Sandwich theorem,
a
n
0, i.e., n
1
n
1 as n .
(ii) Let k be an integer such that k > x, k > 0. Then for n > 2k,
(1 + p)
n
>
n
C
k
p
k
=
n!
k!(n k)!
p
k
=
p
k
k!
k

i=1
[n i + 1] >
n
k
2
k
p
k
k!
.
Hence,
0 <
n
x
(1 + x)
n
<
2
k
k!
x
k
n
xk
(n > 2k).
5
As x k < 0, n
xk
0. Thus
n
x
(1 + x)
n
0 as n .
(iii) For any n N there exists m N such that m n
p
< (m + 1) or equivalently
m
1
p
n < (m + 1)
1
p
. Let > 0. Since n
1
n
1 as n , there exists N N such that
n
1
n
(e

, e

), n N (or)
log n
n
(, ), n N. That is
log n
n
0 as n . This
implies that
lim
m
1
p
log(m + 1)
m
= 0.
As
1
n
p
<
log n
n
p
<
1
p
log(m + 1)
m
. Now the conclusion follows from Sandwich theorem.
Denition 1.2.9. (Subsequence): Let {a
n
} be a sequence and {n
1
, n
2
, ...} be a sequence
of positive integers such that i > j implies n
i
> n
j
. Then the sequence {a
n
i
}

i=1
is called
a subsequence of {a
n
}.
Theorem 1.2.10. If the sequence of real numbers {a
n
}

1
, is convergent to L, then any
subsequence of {a
n
} is also convergent to L.
Proof. Let {n
i
}

i=1
be a sequence of positive integers such that {a
n
i
}

i=1
forms a subse-
quence of {a
n
}. Let > 0 be given. As {a
n
} converges to L, there exists N N such
that
|a
n
L| < , n N.
Choose M N such that n
i
N for i M. Then
|a
n
i
L| < , i M.
Hence the proof. ///
Denition 1.2.11. (Bounded sequence): A sequence {a
n
} is said to be bounded above,
if there exists M R such that a
n
M for all n N. Similarly, we say that a sequence
{a
n
} is bounded below, if there exists N R such that a
n
N for all n N. Thus a
sequence {a
n
} is said to be bounded if it is both bounded above and below.
Lemma 1.2.12. Every convergent sequence is bounded.
Proof. Let {a
n
} be a convergent sequence and L = lim
n
a
n
. Let = 1. Then there exists
N N such that |a
n
L| < 1 for all n N. Further,
|a
n
| = |a
n
L + L| |a
n
L| + L < 1 + L, n N.
Let M = max{|a
1
|, |a
2
|, ..., |a
n1
|, 1 + |L|}. Then |a
n
| M for all n N. Hence {a
n
} is
bounded. ///
6
1.3 Operations on convergent sequences
Theorem 1.3.1. Let {a
n
}

1
and {b
n
}

1
be two sequences such that lim
n
a
n
= L and
lim
n
b
n
= M. Then
(i) lim
n
(a
n
+ b
n
) = L + M.
(ii) lim
n
(ca
n
) = cL, c R..
(iii) lim
n
(a
n
b
n
) = LM.
(iv) lim
n
_
a
n
b
n
_
=
L
M
if M = 0.
Proof. (i) Let > 0. Since a
n
converges to L there exists N
1
N such that
|a
n
L| < /2 n N
1
.
Also, as b
n
converges to M there exists N
2
N such that
|b
n
M| < /2 n N
2
.
Thus
|(a
n
+ b
n
) (L + M)| |a
n
L| +|b
N
M| < n N = max{N
1
.N
2
}.
(ii) Easy to prove. Hence left as an exercise to the students.
(iii) Let > 0. Since a
n
is a convergent sequence, it is bounded by M
1
(say). Also as a
n
converges to L there exists N
1
N such that
|a
n
L| < /2M n N
1
.
Similarly as b
n
converges to M there exists N
2
N such that
|b
n
M| < /2M
1
n N
2
.
Let N = max{N
1
, N
2
}. Then
|a
n
b
n
LM| = |a
n
b
n
a
n
M + a
n
M LM| |a
n
(b
n
M)| +|M(a
n
L)|
= |a
n
||b
n
M| + M|a
n
L| < /2 + /2 =
7
(iv) In order to prove this, it is enough to prove that if lim
n
a
n
= L, L = 0, then
lim
n
1/a
n
= 1/L. Without loss of generality, let us assume that M > 0. Let
> 0 be given. As {a
n
} forms a convergent sequence, it is bounded. Choose N
1
N
such that a
n
> L/2 for all n N
1
. Also, as a
n
converges to L, there exists N
2
N
such that |a
n
L| < L
2
/2 for all n N
2
. Let N = max{N
1
, N
2
}. Then
n N =

1
a
n

1
L

=
|a
n
L|
|a
n
L|
<
2
L
2
L
2

2
= . ///
Examples 1.3.2.
(i) Consider the sequence
_
5
n
2
_

1
. Then lim
n
5
n
2
= lim
n
5
1
n

1
n
= 5 lim
n
1
n
lim
n
1
n
=
5 0 0 = 0.
(ii) Consider the sequence
_
3n
2
6n
5n
2
+ 4
_

1
. Notice that
3n
2
6n
5n
2
+ 4
=
3 6/n
5 + 4/n
3/5.
(iii) Similarly one can show that lim
n
_
n 1
n
_
= 1.
1.4 Divergent sequence and Monotone sequences
Denition 1.4.1. Let {a
n
}be a sequence of real numbers. We say that a
n
approaches
innity or diverges to innity, if for any real number M > 0, there is a positive integer
N such that
n N = a
n
M.
If a
n
approaches innity, then we write a
n
as n .
A similar denition is given for the sequences diverging to . In this case we
write a
n
as n .
Examples 1.4.2.
(i) The sequence {log(1/n)}

1
diverges to . In order to prove this, for any M > 0,
we must produce a N N such that
log(1/n) < M, n N.
But this is equivalent to saying that n > e
M
, n N. Choose N e
M
. Then, for
this choice of N,
log(1/n) < M, n N.
Thus {log(1/n)}

1
diverges to .
8
(ii) Consider the sequence {(1)
n+1
n}

1
. Although this is not a convergent sequence, it
also does not approach to or .
(iii) The sequence (1)
n
is also an example of the previous type.
Denition 1.4.3. If a sequence {a
n
}does not converge to a value in R and also does not
diverge to or , we say that {a
n
}oscillates.
Lemma 1.4.4. Let {a
n
}and {b
n
} be two sequences.
(i) If {a
n
}and {b
n
} both diverges to , then the sequences {a
n
+ b
n
} and {a
n
b
n
} also
diverges to .
(ii) If {a
n
}diverges to and {b
n
} converges then {a
n
+ b
n
} diverges to .
Example 1.4.5. Consider the sequence {

n + 1

n}

n=1
. We know that

n + 1 and

n both converges to . But the sequence {

n + 1

n}

n=1
converges to 0. To see this,
notice that, for a given > 0,

n + 1

n < if and only if 1 <

2
+ 2

n. Thus, if N
is such that N >
1
4
2
, then for all n N,

n + 1

n < . Thus

n + 1

n converges
to 0. This example shows that the sequence formed by taking dierence of two diverging
sequences may converge.
Denition 1.4.6. A sequence {a
n
} of real numbers is called a nondecreasing sequence if
a
n
a
n+1
for all n N and {a
n
} is called a nonincreasing sequence if a
n
a
n+1
for all
n N. A sequence that is nondecreasing or nonincreasing is called a monotone sequence.
Examples 1.4.7.
(i) The sequences {1 1/n}, {n
3
} are nondecreasing sequences.
(ii) The sequences {1/n}, {1/n
2
} are nonincreasing sequences.
(iii) The sequences {(1)
n
}, {cos(
n
3
)}, {(1)
n
n}, {
(1)
n
n
} and {n
1/n
} are not monotonic
sequences.
Lemma 1.4.8. (i) A nondecreasing sequence which is not bounded above diverges to
.
(ii) A nonincreasing sequence which is not bounded below diverges to .
Example 1.4.9. If b > 1, then the sequence {b
n
}

1
diverges to .
Theorem 1.4.10.
(i) A nondecreasing sequence which is bounded above is convergent.
9
(ii) A nonincreasing sequence which is bounded below is convergent.
Proof. (i) Let {a
n
}be a nondecreasing, bounded above sequence and a = sup
nN
a
n
. Since
the sequence is bounded, a R. We claim that a is the limit point of the sequence {a
n
}.
Indeed, let > 0 be given. Since a is not an upper bound for {a
n
}, there exists N N
such that a
N
> a . As the sequence is nondecreasing, we have a < a
N
a
n
for all
n N. Also it is clear that a
n
a for all n N. Thus,
a a
n
a + , n N.
Hence the proof.
The proof of (ii) is similar to (i) and is left as an exercise to the students. ///
Examples 1.4.11.
(i) If 0 < b < 1, then the sequence {b
n
}

1
converges to 0.
Solution. We may write b
n+1
= b
n
b < b
n
. Hence {b
n
} is nonincreasing. Since
b
n
> 0 for all n N, the sequence {b
n
} is bounded below. Hence, by the above
theorem, {b
n
} converges. Let L = lim
n
b
n
. Further, lim
n
b
n+1
= lim
n
b b
n
=
b lim
n
b
n
= b L. Thus the sequence {b
n+1
} converges to b L. On the other hand,
{b
n+1
} is a subsequence of {b
n
}. Hence L = b L which implies L = 0 as b = 1.
(ii) The sequence {(1 + 1/n)
n
}

1
is convergent.
Solution. Let a
n
= (1 + 1/n)
n
=
n

k=0
_
n
k
__
1
n
_
k
. For k = 1, 2, ..., n, the (k + 1)
th
term in the expansion is
n(n 1)(n 2) (n k + 1)
1 2 k
1
n
k
=
1
k!
_
1
1
n
__
1
2
n
_

_
1
k 1
n
_
. (1.6)
Similarly, if we expand a
n+1
, then we obtain (n +2) terms in the expansion and for
k = 1, 2, 3, ..., the (k + 1)
th
term is
1
k!
_
1
1
n + 1
__
1
2
n + 1
_

_
1
k 1
n + 1
_
. (1.7)
It is clear that (1.7) is greater than or equal to (1.6) and hence a
n
a
n+1
which
implies that {a
n
}is nondecreasing. Further,
a
n
=(1 + 1/n)
n
=
n

k=0
_
n
k
__
1
n
_
k
< 2 +
n

k=2
1
k!
< 1 + e < 3.
10
Thus {a
n
}is a bounded monotone sequence and hence convergent.
Theorem 1.4.12. Every sequence has a monotone subsequence.
Proof. Pick x
N
1
such that x
n
x
N
1
for all n > N
1
. We call such x
N
as peak. If we
are able to pick innitely many x

N
i
s, then {x
N
i
} is decreasing and we are done. If there
are only nitely many x

N
s and let x
n
1
be the last peak. Then we can choose n
2
such
that x
n
2
x
n
1
. Again x
n
2
is not a peak. So we can choose x
n
3
such that x
n
3
x
n
2
.
Proceeding this way, we get a non-decreasing sub-sequence.
The following theorem is Bolzano-Weierstrass theorem. Proof is a consequence of Theorem1.4.12
Theorem 1.4.13. Every bounded sequence has a convergent subsequence.
Theorem 1.4.14. Nested Interval theorem: Let I
n
= [a
n
, b
n
], n 1 be non-empty closed,
bounded intervals such that
I
1
I
2
I
3
... I
n
I
n+1
...
and lim
n
(b
n
a
n
) = 0. Then

n=1
I
n
contains precisely one point
Proof. Since {a
n
}, {b
n
} [a
1
, b
1
], {a
n
}, {b
n
} are bounded sequences. By Bolzano-Weierstrass
theorem, there exists sub sequences a
n
k
, b
n
k
and a, b such that a
n
k
a, b
n
k
b. Since a
n
is increasing a
1
< a
2
< ...... a and b
1
> b
2
> .... b. It is easy to see that a b. Also
since 0 = lima
n
b
n
= a b, we have a = b.
It is easy to show that there is no other point in

n=1
I
n
.
1.5 Cauchy sequence
Denition 1.5.1. A sequence {a
n
}is called a Cauchy sequence if for any given > 0,
there exists N N such that n, m N = |a
n
a
m
| < .
Example 1.5.2. Let {a
n
}be a sequence such that {a
n
}converges to L (say). Let > 0 be
given. Then there exists N N such that
|a
n
L| <

2
n N.
Thus if n, m N, we have
|a
n
a
m
| |a
n
L| +|a
m
L| <

2
+

2
= .
Thus {a
n
}is Cauchy.
11
Lemma 1.5.3. If {a
n
}is a Cauchy sequence, then {a
n
}is bounded.
Proof. Since {a
n
}forms a Cauchy sequence, for = 1 there exists N N such that
|a
n
a
m
| < 1, n, m N.
In particular,
|a
n
a
N
| < 1, n N.
Hence if n N, then
|a
n
| |a
n
a
N
| +|a
N
| < 1 +|a
N
|, n N.
Let M = max{|a
1
|, |a
2
|, ..., |a
N1
|, 1 + |a
N
|}. Then |a
n
| M for all n N. Hence {a
n
}is
bounded. ///
Theorem 1.5.4. If {a
n
}is a Cauchy sequence, then {a
n
}is convergent.
Proof. Let a
n
k
be a monotone subsequence of the Cauchy sequence {a
n
}. Then a
n
k
is a
bounded, monotone subsequence. Hence {a
n
k
} converges to L(say). Now we claim that
the sequence {a
n
} itself converges to L. Let > 0. Choose N
1
, N
2
such that
n, n
k
N
1
= |a
n
a
n
k
| < /2
n
k
N
2
= |a
n
k
a| < /2.
Then
n, n
k
max{N, N
1
} = |a
n
a| |a
n
a
n
k
| +a
n
k
a| < .
Hence the claim. ///
Therefore, we have the following Criterion:
Cauchys Criterion for convergence: A sequence {a
n
} converges if and only if for
every > 0, there exists N such that
|a
n
a
m
| < m, n N.
Problem: Let {a
n
}be dened as a
1
= 1, a
n+1
= 1 +
1
an
. The show that {a
n
}is Cauchy.
Solution: Note that a
n
> 1 and a
n
a
n1
= a
n1
+ 1 > 2. Then
|a
n+1
a
n
| = |
a
n1
a
n
a
n
a
n1
|
1
2
|a
n
a
n1
|
1
2
n1
|a
2
a
1
|, n 2.
12
Hence
|a
m
a
n
| |a
m
a
m1
| +|a
m1
a
m2
| + .... +|a
n+1
a
n
| |a
2
a
1
|

n1
1
, =
1
2
So given, > 0, we can choose N such that
1
2
N1
<

2
.
Indeed the following holds,
Theorem 1.5.5. Let {a
n
} be a sequence such that |a
n+1
a
n
| < |a
n
a
n1
| for all
n N for some N and 0 < < 1. Then {a
n
} is a Cauchy sequence.
Theorem 1.5.6. For any sequence {a
n
} with a
n
> 0
lim
n
a
1/n
n
= lim
n
a
n+1
a
n
provided the limit on the right side exists.
Proof. Let > 0 be arbitrary. Suppose the second limit exists (say l), then there exists
N N such that
l <
a
n+1
a
n
< l + , n N.
Taking n = N, N + 1, ...., m1 and multiplying we get
(l )
mN
<
a
m
a
N
< (l + )
mN
, m N + 1
equivalently,
(l )
1N/m
a
1/m
N
< (a
m
)
1/m
< (l + )
1N/m
, m N + 1.
Now the result follows from the fact that lim
p
(l )
1N/m
a
1/m
N
= l . ///
Corollary: If a
n
> 0 and lim
n
a
n+1
a
n
= l < 1, then lim
n
a
n
= 0
Corollary: If a
n
> 0 and lim
n
a
n+1
a
n
= l > 1, then a
n
.
Problems: (i) lima
1/n
= 1, if a > 0.
(ii) limn

x
n
= 0, if |x| < 1 and IR.
Solution:
(i) Take a
n
= a, then lim
n
a
n+1
an
= 1.
(ii) If x = 0, take a
n
= n

x
n
, then lim
n
a
n+1
an
= lim(1 +
1
n
)

|x| = |x|.
13
1.6 Limit superior and limit inferior
Denition 1.6.1. Let {a
n
}be a bounded sequence. Then limit superior of the sequence
{a
n
}, denoted by limsup
n
a
n
, is dened as
limsup
n
a
n
:= inf
kN
sup
nk
a
n
.
Similarly limit inferior of the sequence {a
n
}, denoted by liminf
n
a
n
, is dened as
liminf
n
a
n
:= sup
kN
inf
nk
a
n
.
Example 1.6.2. (i) Consider the sequence {a
n
} = {0, 1, 0, 1, .....}. Then
n
= sup{a
m
, m
n} = 1 and
n
= inf{a
m
, m n} = 0. Therefore, liminf a
n
= 0, limsup a
n
= 1.
(ii) Consider the sequence {a
n
} = {
1
2
,
2
3
,
1
3
,
3
4
, ....}. Then for large k
1 sup{a
m
, m k} lim
k 1
k
0 < inf{a
m
, m k} lim
1
k
Then by sandwich theorem, we see that limsup a
n
= 1 and liminf a
n
= 0.
Lemma 1.6.3.
(i) If {a
n
}is a bounded sequence, then limsup
n
a
n
liminf
n
a
n
.
(ii) If {a
n
}and {b
n
} are bounded sequences of real numbers and if a
n
b
n
for all n N,
then
limsup
n
a
n
limsup
n
b
n
and
liminf
n
a
n
liminf
n
b
n
.
(iii) Let {a
n
}and {b
n
} are bounded sequences of real numbers. Then
limsup
n
(a
n
+ b
n
) limsup
n
a
n
+ limsup
n
b
n
and
liminf
n
(a
n
+ b
n
) liminf
n
a
n
+ liminf
n
b
n
.
14
Example 1.6.4. Consider the sequences {(1)
n
} and {(1)
n+1
}. Here a
n
= (1)
n
and
b
n
= (1)
n+1
. Also limsup
n
a
n
= limsup
n
b
n
= 1. But a
n
+b
n
= 0 for all n N and hence
limsup
n
(a
n
+ b
n
) = 0. Thus a strict inequality may hold in (iii) the above Lemma.
Theorem 1.6.5. If {a
n
}is a convergent sequence, then
liminf
n
a
n
= lim
n
a
n
= limsup
n
a
n
.
Proof. Let L = lim
n
a
n
. Then given > 0 there exists N N such that
|a
n
L| < , n N.
Equivalently L < a
n
< L+, for all n N. Thus, if n N, L+ is an upper bound
for the set {a
k
|k N}. If
k
:= sup{a
k
|k n}, then we note that L <
N
L+ and

N+1
< L + , ....,
n
< L + for all n N (As
n
is decreasing). Also a
n
> L , n
N =
n
L, n N. Therefore, lim
n
= L. Hence limsup a
n
= L. Similarly, one
can prove the liminf a
n
= L.
///
Theorem 1.6.6. If {a
n
}is a bounded sequence and if limsup
n
a
n
= liminf
n
a
n
= L, L R,
then {a
n
}is a convergent sequence.
Proof. Notice that
limsup
n
a
n
= lim
n
(sup{a
k
|k n})
and
liminf
n
a
n
= lim
n
(inf{a
k
|k n}).
Given that L = lim sup
n
a
n
. Thus for > 0, there exists N
1
N such that
| sup{a
n
, a
n+1
, ...} L| < , n N
1
.
This implies
a
n
< L + , n N
1
(1.8)
Similarly there exists N
2
N such that
| inf{a
n
, a
n+1
, ...} L| < , n N
2
.
This implies
L < a
n
, n N
2
(1.9)
15
Let N = max{N
1
, N
2
}. Then from (1.8)and (1.9) we get
|a
n
L| < , n N.
Thus the sequence {a
n
}converges. ///
Examples 1.6.7. lim
n
_
1 +
1
n
_
n
= e. Assume that e = lim
n

n
k=0
1
k!
.
Solution. Let a
n
=
n

k=0
1
k!
and b
n
=
_
1 +
1
n
_
n
. Now,
b
n
=
n

k=0
n
C
k
_
1
n
_
k
= 2 +
n

k=2
1
k!
k1

i=1
_
1
i
n
_
a
n
.(see (1.6)
This implies
limsup
n
b
n
limsup
n
a
n
= e.
Further, if n m, then
b
n
=
_
1 +
1
n
_
n
=
n

k=0
n
C
k
_
1
n
_
k

k=0
n
C
k
_
1
n
_
k
= 2 +
m

k=2
1
k!
k1

i=1
_
1
i
n
_
.
Keeping m xed and letting n , we get
liminf
n
b
n

m

k=0
1
k!
which implies a
n
liminf
n
b
n
. Hence
e = liminf
n
a
n
liminf
n
b
n
.
Finally we have the following more precise version of theorem 1.6.6
Theorem 1.6.8. Let {a
n
} be any sequence of nonzero real numbers. Then we have
liminf

a
n+1
a
n

liminf |a
n
|
1/n
limsup |a
n
|
1/n
limsup

a
n+1
a
n

.
Proof. The inequality in the middle is trivial. Now we show the right end inequality. Let
L = limsup |
a
n+1
an
|. W.l.g assume L < . Let > 0. Then there exists N N such that

a
n+1
a
n

< L + n N.
16
Then for any n > N, we can write
|a
n
| =

a
n
a
n1

a
n1
a
n1

...

a
N+1
a
N

|a
N
|
< (L + )
nN
|a
N
|
= (L + )
n
((L + )
N
|a
N
|).
Now taking a = ((L )
N
|a
N
|), we have, |a
n
|
1/n
< (L + )a
1/n
for n > N. Since
lim
n
a
1/n
= 1, we conclude that limsup |a
n
|
1/n
(L + ). Since is arbitrary, we get
the result. Similarly, we can prove the rst inequality.
References
[1] Methods of Real Analysis, Chapter 2, R. Goldberg .
[2] Elementary Analysis: The Theory of Calculus, K. A. Ross.
[3] Calculus, G. B. Thomas and R. L. Finney, Pearson .
[4] Calculus, James Stewart, Brooks/Cole Cengage Learning.
17