1

Sequen e

Informally, a sequen e is a set with an order in the sense that there is a rst element, se ond
and so on. In other words to ea h positive integer 1,2,3, . . . , we asso iate a real number. In
the sequel, we will onsider only sequen es of real numbers.
De nition A fun tion f : f1; 2; 3; : : :g ! IR is alled a sequen e of real numbers. We write
f (n)= xn and then the sequen e is denoted by x1 ; x2 ; : : :, or simply by fxn g. We all xn the
nth element of the sequen e or the value of the sequen e at n.

Some examples of sequen es:

1. fng = 1; 2; 3; : : :
2. f g = 1; ; ; : : :
1
n
1
2
1
3

3. f n g = 1; ; ; : : :
( 1)
n
1
2 3
1

4. f1 g = 0; 1 ; 1 ; : : :
1
n
1
2
1
3

5. f1 + n g = 1:1; 1:01; 1:001; : : :

1
10

6. f( 1) g = 1; +1; 1; +1; : : :
n

7. f5g = 5; 5; 5; : : :
8. f1+( n
g = 0; 1 ; 0 ; 1 ; 0 ; : : :
2
1)

Before giving the formal de nition of onvergen e of a sequen e, let us take a look at the
behaviour of the sequen es in the above examples.
The elements of the sequen es f n1 g, f1 1=ng and f1 + 1=10ng seem to \approa h" a
single point as n in reases. In these sequen es the values are either in reasing or de reasing
as n in reases, but they \eventually approa h" a single point. Though the elements of the
sequen e f( 1)n =ng os illate, they \eventually approa h" the single point 0. Similarly for
the sequen e 5; 5; : : :. The ommon feature of these sequen es is that the terms of ea h
sequen e \a umulate" at only one point. On the other hand, values of the sequen e fng
be ome larger and larger and do not a umulate anywhere. The elements of the sequen e
f( 1)ng os illate between two di erent points 1 and 1; i.e., the elements of the sequen e
ome lose to 1 and 1 \frequently" as n in reases.
2.1 Convergen e of a Sequen e
Let us distinguish sequen es whose elements approa h a single point as n in reases (in this
ase we say that they onverge) from those sequen es whose elements do not. Geometri ally it
is lear that if the elements of the sequen e fxn g ome eventually inside every neighbourhood
(x0 ; x0 + );  > 0, of x0 , then fxn g approa hes x0 .
Let us now state the formal de nition of onvergen e.
We say that a sequen e fxn g onverges if there exist x 2 IR su h that for every  > 0,
there exists a positive integer N (depending on ) su h that jxn xj <  for all n  N . It
 2

an be easily veri ed that if su h a number x exists then it is unique. In this ase, we say
that the sequen e fxn g onverges to x and we all x the limit of the sequen e fxn g.
If x is the limit of fxn g, we write lim xn = x or xn ! x.
n!1

Let us show that the sequen e f g in Example (ii) has limit equal to 0.
1
n
For arbitrary
 > 0, the inequality

jx j = n1 < 
n

is true for all n > 1 and hen e for all n > N where N is any natural number su h that
N >  . Thus for any  > 0, there is a natural number N su h that jxn j <  for every n  N .
1

Similarly, we an prove that the sequen e in Example (iii) onverges to 0.

The sequen e in Example (iv) onverges to 1, be ause in this ase
1
j1 xn j = j1 n
n
j = n1  
for all n > N where N is any natural number greater than 1 .
Arithmeti Operations on Sequen es
Sin e sequen es of real numbers are real valued fun tions, the de nition of sum, di eren e,
produ t and quotient of sequen es follows from the de nition of the respe tive operations
of real numbers. For example, if fxn g and fyn g are given then fxn g + fyng is the sequen e
fxn + yng.
Theorem 2.1 Suppose xn ! x and yn ! y . Then

1. xn + yn ! x + y
2. ! xy
xn y n

3. xn ! if y 6= 0 and y 6= 0 8 n.
x
yn y n

Proof: Exer ise.

2.3 Monotone Sequen es
We now give a simple riterion for the onvergen e of a sequen e (without having any knowl-
edge of its limit).
Suppose a sequen e fx g onverges to x. Then, for  = 1, there exist N su h that jx
n n
xj  1 8 n  N . This implies jx j  jxj+1 8 n  N . If we let M = maxfjx j; jx j; : : : ; jx jg,
n 1 2 N 1
then jx j  M + jxj + 1 8 n. Hen e fx g is a bounded sequen e.
n n

Every onvergent sequen e is a bounded sequen e. However, bounded sequen es may not
onverge. We now nd a ondition on a bounded sequen e whi h ensures the onvergen e of
the sequen e.
2.2 De nition We say that a sequen e fx g is in reasing if x  x
n n n+1 8 n and stri tly
in reasing if xn < xn+1 8 n. Similarly, we de ne de reasing and stri tly de reasing sequen es.
 3

The following result is an appli ation of the least upper bound property of the real number
system.
Theorem 2.2 Suppose fxn g is a bounded and in reasing sequen e. Then the least upper
bound of the set fxn : n 2 N g is the limit of fxn g.
Proof: Suppose sup xn = M . For given  > 0, there exists n0 su h that M   x . Sin e
n0
n
fx g is in reasing, we have x  x 8 n  n .
n n0 n 0 This implies that M   x  M 
n
M +  8 n  n . That is x ! M . 
0 n

Exer ise: Prove that a bounded de reasing sequen e onverges to

inf fxn : n 2 N g.
We will now give appli ations of the previous theorem.
The following theorem, alled nested interval theorem, is a famous result about the set
of real numbers.
Theorem 2.3 For ea h n, let In = [an ; bn ℄ be a (nonempty) bounded interval of real numbers
su h that
I1  I 


 I  I
2 n n+1 



1
T
and lim (bn an ) = 0. Then In ontains only one point.
n!1 n=1

Proof: Note that the sequen es (an ) and (bn ) are respe tively in reasing and de reasing
sequen es; moreover both are bounded. Hen e both onverge, say an ! a and bn ! b. Then
an  a and b  bn 8 n 2 N . Sin e b a = lim(bn an ) = 0. Thus a = b. Sin e an  bn 8 n;
1
T 1
T
this implies that a 2 In . Clearly if x 6= a then x does not belong to In .
n=1 n=1


Example: The sequen e f 1+ 1 n
n
g is onvergent.

n
Proof: Let xn = 1 + n1 . By the binomial theorem
1 n(n 1) 1 n(n 1)


1 1
= 1+n
+
+




1
2 1


2


n nn
xn
2
n n
1 1 1 1 2
= 1 + 1 + (1 ) + (1 )(1 )
2! n 3! n n
1 1 1
+


+ (1 )


(1
n
)
n! n n

 1 + 1 + 2!1 (1 n +1 1 ) + 3!1 (1 n +1 1 )(1 n +2 1 )

1 1 2
+


+ )


(1
n
(1 )(1 )
(n + 1)! n+1 n+1 n+1
= xn+1
Therefore fxn g is an in reasing sequen e. Moreover,
1 1 1
xn  + +:::+
2! 3!
2+
n!

 2 + 2 + 22 + : : : + 2n1 1  3:
1 1

Therefore, fxn g is bounded and in reasing. Hen e by the previous theorem fxn g onverges.
 4

It is lear that the limit lies between 2 and 3. This limit is alled the number e. Thus
lim (1 + n1 )n = e.
n!1

2.4 Cau hy Sequen es

We have seen one riterion for proving that a sequen e onverges without knowing its limit.
We will now present another riterion.
Suppose that a sequen e fxn g onverges to x. Then for  > 0, there exists an N su h
that jxn xj < =2 8 n  N . Hen e for n; m  N we have jxn xm j = jxn x + x xm j
jxn xj + jx xm j < . Thus we arrive at the following on lusion.
If a sequen e fxn g onverges then it satis es the Cau hy's ondition: for  > 0, there
exist N su h that jxn xm j <  8 n; m  N .
If a sequen e fx g satis
n es the Cau hy's ondition then we say that fx g is a Cau hy
n
sequen e.
If a sequen e onverges then the elements of the sequen e get lose to the limit as n
in reases. In ase of Cau hy sequen e the elements get lose to ea h other as m; n in reases.
Exer ise : Verify that every Cau hy sequen e is a bounded sequen e.
Thus a Cau hy sequen e is a bounded sequen e with some additional property. Moreover,
intuitively it seems as if it onverges. So it is natural to ask whether every Cau hy sequen e
does onverge.
Theorem 2.4 Every Cau hy sequen e is onvergent.
Proof: Sin e every Cau hy sequen e is bounded, for an arbitrary n, we de ne
an = inf xk and bn = sup xk :
k>n k>n

Then the intervals [an ; bn ℄ form a nested family:

[an+1 ; bn+1 ℄  [an ; bn ℄:
Sin e fxn g is a Cau hy sequen e, bn an ! 0 as n ! 1. Hen e by the nested interval
1
T
theorem, there exists x su h that x 2 [an ; bn ℄: This x is the limit of fxn g (verify !).
n=1

2.5 Subsequen es
Let fx g be a sequen e and let fn g be any sequen e of positive integers su h that n <
n k 1
The sequen e fx k g is alled a subsequen e. Note that here k varies from 1
n2 < n3 < : : :. n
to 1.
A subsequen e is formed by deleting some of the elements of the sequen e and retaining
the remaining with the same order.
For example, f k12 g and f 21k g (k varies from 1 to 1) are subsequen es of f n1 g, where
nk = k 2 and nk = 2k . Sequen es (1; 1; 1; : : :) and (0; 0; 0; : : :) are both subsequen es of
(1; 0; 1; 0; : : :). We also note that every sequen e is a subsequen e of itself.
The following three results are easy to verify so they are left as exer ise.
 5

Theorem 2.5 Suppose xn ! x. Then every subsequen e of fxn g also onverges to x.

Exer ise : If fxn g is a Cau hy sequen e of real numbers whi h has a subsequen e onverging
to x, show that fxn g itself onverges to x.
From the previous examples it is lear that a given sequen e may have onvergent sub-
sequen e though the sequen e itself is not onvergent. We also note that if all onvergent
subsequen es of a sequen e onverge to the same limit even then the original sequen e may
not onverge. At the same time, we have the following result.
Theorem 2.6 If every subsequen e of a sequen e fxn g has a onvergent subsequen e whi h
onverges to the same limit then the original sequen e onverges.
Proof : Exer ise.
We say that x is a luster point of fxn g if there is a subsequen e fxnk g of fxn g su h that
xn k ! x.
Theorem 2.7 An element x is a luster point of a sequen e fxn g if and only if for every
> 0, there exist in nitely many integers n su h that jxn j
x < .

If x is a luster point of a sequen e fxn g then the elements of the sequen e fxn g ome
frequently inside every neighbourhood (x ; x + ),  > 0; i.e. 8 positive integer n there
exists m > n su h that xm 2 (x ; x + ).
The following theorem, often alled the Bolzano Weierstrass theorem, gives an existen e
riterion for luster points.
Theorem 2.8 Every bounded sequen e in IR has a onvergent subsequen e.
Proof: (Sket h). Let fxn g be a bounded sequen e su h that the set fx1 ; x2 ;


g  [a; b℄.
Divide this interval into two equal parts. Let I1 be that interval whi h ontains an in nite
number of elements of fxn g. Let xn1 be one of the elements belonging to the interval I1 .
Divide I1 into two equal parts and let I2 be that interval whi h ontains an in nite number
of elements. Choose a point xn2 in I2 su h that n1 > n2 : Keep dividing the intervals Ik , to
T1
generate Ik 's and xnk 's. By nested interval theorem Ik = fxg, for some x 2 [a; b℄: Then
k =1
the subsequen e fx k g
n onverges to x. 
Let S be an in nite set of numbers, then x is alled a luster point of S if for every  > 0,
there exists an in nite number of elements y 2 S su h that jx y j < .
Corollary 2.1 If S is a bounded and in nite subset of IR then S has a luster point.
If fxn g is a onvergent sequen e in IR and if it onverges to x, then we an roughly say that
the size of the element xn is \almost" x when n is large. Many times we deal with sequen es
whi h do not onverge. Informally, the limit supremum of a sequen e fxn g is a measure of
how big xn an be when n is large. Similarly, limit in mum is a measure of how small xn
an be when n is large. Before de ning limit in mum and limit supremum formally, let us
onsider an example.
Consider the sequen e,
1 1 1 1
fx g = 1;
n 1; 1;
2
; 2; 1 + ;
2 3
; 3; 1 +
3
; :::
 6

This sequen e is not onvergent and it is not even bounded. We guess that the limit supre-
mum of (xn ) is 1. However, 1 is not the supremum of the set fx1 ; x2 ; : : :g, be ause there are
elements whi h lie to the right of 1. But if we look slightly right side of 1, i.e., if we take 1+ ,
 > 0, then ultimately all elements of the sequen e lie to the left of 1 + . So we guess that if
x is the limit supremum of the sequen e fxn g, then x satis es the following ondition: \for
every  > 0, 9 N su h that xn  x +  8 n  N ." If an element x satis es this ondition then
x need not be the limit supremum. For example, onsider the above mentioned example,
then any x  2 satis es this ondition. But any element x whi h is greater than or equal
to 2 is totally \away" from the sequen e. So we require additional onditions to de ne the
limit supremum.
Let us now state the formal de nition of limit supremum.
De nition 2.3 Let fxng be a sequen e of real numbers. Suppose there is a real number x
satisfying the following onditions:

1. for every > 0, there exists N su h that xn  x + ; 8 n  N

2. for given > 0 and m > 0, there exists an integer n>m su h that xn x .

It is lear that there an't be more than one x satisfying both onditions (i) and (ii). If
an element x satis es both (i) and (ii), then we say that x is the limit supremum of fxn g
and write x = lim sup xn .
n!1

If the sequen e (xn ) is not bounded above, then we de ne lim sup xn = 1.

n!1

If the sequen e is bounded above but not bounded below and if fx g has no
n nite limit
supremum, then we de ne lim sup xn = 1.
n!1

Similarly we an de ne limit in mum of a sequen e (xn ), whi h will be denoted by

lim inf xn . One an also verify that x is the limit in mum of xn if and only if x =
n!1
lim sup( xn ).
n!1

We note that every sequen e has a limit supremum and limit in mum in the extended
real number system.

Examples


(i) an = ( 1)n 1 + n1
lim inf an = 1 lim sup an = 1
n!1 n!1
(ii) an = ( 1)n lim inf an = 1 lim sup an = 1
n!1 n!1
(iii) an = ( 1)n
n lim inf an = 1 lim sup an = 1
n!1 n!1

The proof of the following theorem is easy, hen e it is left as an exer ise.
Theorem 2.9
Let (xn ) be a sequen e of real numbers. Then we have
 7

1. lim inf xn  lim sup xn :

n!1 n!1

2. fx g
n onverges to x if lim sup xn = lim inf xn = x:
n!1
n!1

3. If fyng is sequen e, su h that xn y . Then lim inf xn  lim!1inf y and lim sup xn 
n!1
n n
n n!1
lim sup yn :
n!1