5 Sequences

Let (X, d) be a metric space. A sequence in X is a function from N to X.

If : N X is a sequence we write
(x
n
), (x
n
)
nN
or (x
0
, x
1
, . . .)
for , where x
n
= (n) is the nth term of the sequence = (x
0
, x
1
, . . .).
Sequences in R are called real sequences.
Denition 5.1. A sequence (x
n
) in X converges to x if for every neighbor-
hood U of x there exists N N such that
x
n
U for all n N.
In this case we write
x
n
x or lim
n
x
n
= x or lim
n
x
n
= x.
A sequence (x
n
) which does not converge is called divergent.
Proposition 5.2. The following are equivalent.
(i) lim
n
x
n
= x.
(ii) For every > 0, there is N N such that x
n
B(x, ) for all n N.
(iii) For every > 0, there is N N such that d(x
n
, x) < for all n N
Proposition 5.3. (i) (Uniqueness of the limit) A sequence cannot have
more than one limit.
(ii) (Boundedness) If (x
n
) converges to x, then there is r > 0 such that
x
n
B
r
(x).
Proof of (ii). Assume that (x
n
) converges to x. By (iii) of the previous
proposition with = 1, there is N so that x
n
B
1
(x) for all n N. Now
take r be any number bigger or equal to d(x
1
, x), . . . , d(x
N1
, x) and 1. Then
x
n
B
r
(x) for all n, as claimed.
Denition 5.4. Let = (x
n
) be a sequence in X and let : N N be a
strictly increasing function. Then : N X is called a subsequence of
= (x
n
).
We write (x
n
k
) for the subsequece where (k) = n
k
. Since is
strictly increasing n
1
< n
2
< n
3
< . . ..
20
Proposition 5.5. If (x
n
) is a convergent sequence with limit x, then every
subsequence (x
n
k
) of (x
n
) converges to x.
Proof. Let (x
n
k
) be a subsequence of (x
n
). Take a neighborhood U of x.
Since limx
n
= x, there is N N such that x
n
U for all n N. From the
denition of a subsequence, n
k
k for all k N. So, n
k
N for all k N.
Thus x
n
k
U for all k N. Hence (x
n
k
) converges to x as claimed.
Denition 5.6. A sequence (x
n
) in a metric space (X, d) is said to be
Cauchy if for every > 0, there exists N such that
d(x
n
, x
m
) for all n, m N.
Proposition 5.7. Let (X, d) be a metric space.
(i) Every Cauchy sequence is bounded.
(ii) Every convergent sequence in X is Cauchy.
Proof of (ii). Assume that (x
n
) converges to x. Then, given > 0, there
is N such that
d(x
n
, x) < /2 for all n N..
Hence
d(x
n
, x
m
) d(x
n
, x) +d(x, x
m
) < /2 +/2 =
for all n, m N. So, (x
n
) is a Cauchy sequence as claimed.
5.1 Real sequences
Proposition 5.8. Let (x
n
) and (y
n
) be convergent sequences in R and let
R. Then
(i) The sequence (x
n
+y
n
) converges and lim(x
n
+y
n
) = limx
n
+limy
n
.
(ii) The sequence (x
n
) converges and lim(x
n
) = limx
n
.
Proof of (i). Take > 0. Since (x
n
) and (y
n
) converge to x and y, respec-
tively, there are N
1
and N
2
such that
|x
n
x| < /2 for n N
1
and |y
n
y| < /2 for n N
2
.
Consequently,
|(x
n
+y
n
) (x +y)| = |(x
n
x) + (y
n
y)| |x
n
x| +|y
n
y|
/2 +/2 =
for all n N = max{N
1
, N
2
}. Hence the sequence (x
n
+y
n
) converges and
lim(x
n
+y
n
) = limx
n
+ limy
n
, as claimed.
21
Note that above proposition holds true if instead of real sequences one
considers sequences in a normed space (X, ). We call a real sequence (x
n
)
a null sequence if it converges to 0. That is, for every > 0, there is N N
such that |x
n
| < for all n N.
Proposition 5.9. Let (x
n
) and (y
n
) be sequences in R.
(i) If (x
n
) is a null sequence and (y
n
) is a bounded sequence, then the
sequence (x
n
y
n
) is a null sequence.
(ii) If (x
n
) and (y
n
) are convergent sequences, then the sequence (x
n
y
n
)
converges and lim(x
n
y
n
) = limx
n
limy
n
.
(iii) If (x
n
) converges to x and x = 0, then almost all terms of (x
n
) are
nonzero and the sequence (1/x
n
) converges to 1/x.
Proof of (iI). Suppose that limx
n
= x and limy
n
= y. By (ii) of Proposi-
tion 5.3, |x
n
| A and |y
n
| B for all n. Moreover, given > 0, there are
N
1
and N
2
so that
|x
n
x| < /2(A+|y|) for n N
1
and |y
n
y| < /2(A+|y|) for n N
2
.
Consequently,
|(x
n
y
n
) (xy)| = |x
n
(y
n
y) +y(x
n
x)| |x
n
| |y
n
y| +|y| |x
n
x|
A|y
n
y| +|y| |x
n
x| A

2(A +|y|)
+|y|

2(A +|y|)
=
for all n N = max{N
1
, N
2
}. Hence the sequence (x
n
y
n
) converges and
lim(x
n
y
n
) = limx
n
limy
n
, as claimed.
Proposition 5.10. (i) Let (x
n
) and (y
n
) be convergent sequences in R
and x
n
y
n
for innitely many n. Then limx
n
limy
n
.
(ii) Let (x
n
) and (y
n
) and (z
n
) be sequences in R and x
n
y
n
z
n
for
almost all n. If limx
n
= limz
n
= a, then (y
n
) converges to a.
Proof of (ii). Take > 0. Then here are N
1
and N
2
so that
|x
n
a| < for n N
1
and |z
n
a| < for n N
2
.
Hence
a < x
n
< a + for n N
1
and a < z
n
< a + for n N
2
22
and since x
n
y
n
z
n
,
a < x
n
y
n
z
n
< a +
for all n N = max{N
1
, N
2
}. This means that |y
n
a| < for all n
N.
Denition 5.11. A sequence (x
n
) in R is increasing, if x
n
x
n+1
for all
n N, and decreasing if x
n+1
x
n
for all n N. If (x
n
) is either increasing
or decreasing, then it is called monotone.
Proposition 5.12. Every increasing (or decreasing) bounded sequence (x
n
)
in R converges,and
x
n
sup{x
n
| n N} ( or x
n
inf{x
n
| n N}).
Proof. Suppose that (x
n
) is increasing and Set a = sup{x
n
| n N} R. By
the characterization of the supremum,
(a) x
n
a for all n.
(b) given , there is N such that a < x
N
.
Since (x
n
) is increasing,
a < x
N
x
n
a < a +
for all n N. Hence (x
n
) converges to x as claimed.
Example 5.13. Let a > 0 and > 0. Dene a sequence (x
n
) by setting x
0
=
and
x
n+1
=
1
2

x
n
+
a
x
n

. (1)
We claim that the sequence (x
n
) converges to

a. To see this we shall show
that (x
n
) is bounded below by

a and it is decreasing. Consider the function
f(x) =
1
2

x +
a
x

for x > 0. Dierentiating one nds that f

(x) =
1
2

1
a
x
2

< 0
for x (0,

a) and f

(x) > 0 for x (

a, ). Hence f takes its minimum at

a =

a and f(x) f(

a) =

a. So, x
n


a for all n 1. Next
x
n
x
n+1
= x
n

1
2

x
n
+
a
x
n

=
1
2

x
n

a
x
n

=
x
2
n
a
2x
n
0
since x
2
n
a. This means that x
n
x
n+1
for all n 1 so that (x
n
) is decreasing.
Then the above proposition implies that x
n
x where x

a > 0 Then also
x
n+1
x. So, taking limits in (1), we get
x =
1
2

x +
a
x

and solving for x one nds x =

a.
23
Proposition 5.14. Every sequence contains a monotone subsequence.
Proof. Consider a sequence (x
n
). We call the m-th term x
m
a peak if
x
m
x
n
for all n m. There are two cases to consider.
Case 1. There are innitely many peaks. List them according the in-
creasing subscripts: x
n
1
, x
n
2
, . . .. Then, since each of them is a peak, we
have
x
n
1
x
n
2
x
n
3
. . . .
Hence (x
n
k
) is a decreasing subsequence of (x
n
).
Case 2. There are nitely many peaks. List them according to the increas-
ing x
m
1
, x
m
2
, . . . , x
m
l
. Then set n
1
= m
l
+ 1. The n
1
terms x
n
1
is not a
peak and so there is an index n
2
> n
1
such that x
n
1
< x
n
2
. The term x
n
2
is
also not a peak and there is n
3
> n
2
such that x
n
2
< x
n
3
. Continuing this
way we obtain a strictly increasing subsequence (x
n
k
) of (x
n
).
5.1.1 Innite limits
Denition 5.15. A sequence of real numbers (x
n
) converges to , in sym-
bols x
n
or limx
n
= , if for every M > 0 there exists N such that
x
n
M for all n N.
Similarly, (x
n
) converges to , in symbols x
n
or limx
n
= if
for every M < 0 , there is there exists N such that
x
n
M for all n N.
5.1.2 Some Special Sequences
(1) Let a R. Then
a
n
0 if |a| < 1
a
n
diverges if |a| > 1
Suppose that |a| < 1. Then |a| =
1
1+h
for some h > 0. By the binomial theorem,
(1 +h)
n
=
n

k=0

n
k

h
k
nh
so that
0 |a
n
| = |a|
n
=
1
(1 +h)
n

1
nh
.
24
Since 1/nh 0, a
n
0 by the comparison test. Assume that |a| > 1. Then
|a| = 1 +h for some h and then
|a
n
| = |a
n
| = (1 +h)
n
nh,
showing that (a
n
) is unbounded.
(2) Let k N and let a R be such that |a| > 1. Then
lim
n
n
k
a
n
= 0.
To see this write a = 1 +h where h > 0 and apply the binomial theorem,
a
n
= (1 +h)
n
=
n

m=0

n
m

h
m

n
k + 1

h
k+1

n(n 1) (n k)
(k + 1)!
h
k+1
.
So,
0
n
k
a
k

(k + 1)!n
k
h
k+1
n(n 1) (n k)
and the claimed follows using the comparison since the right hand side converges
to 0 as n .
(3) For all a R,
lim
n
a
n
n!
= 0.
(4) lim
n
n

n = 1.
To see this write
n

n = 1 +a
n
with a
n
0 (since it is easy to see that
n

n 1. So,
(1 +a
n
)
n
= n.
By the binomial theorem,
n = (1 +a
n
)
n
= 1 +

n
1

a
n
+

n
2

a
2
n
+. . . 1 +

n
2

a
2
n
= 1 +
n(n 1)
2
a
2
n
.
From this we get
0 a
2
n

2
n
, i.e., 0
n

n 1 = a
n

2/n
which by comparison implies that
n

n 1.
(5) For all a > 0, lim
n
n

a = 1.
If a = 1, then there is nothing to prove. If a > 1, then
1
n

a
n

n
which implies that (
n

a) converges to 1. So, assume that 0 < a < 1, then 1/a > 1

so that
n

1/a 1. Since
n

a = 1/
n

1/a, the claim follows.

25
5.1.3 The limit superior and limit inferior
Let (x
n
) be a bounded sequence. For each n N, let
y
n
= sup
mn
x
m
= sup{x
n
, x
n+1
, . . .}.
Then {y
n
} is decreasing, and it is bounded since (x
n
) is bounded. By Proposition
5.12, (y
n
) converges. The limit is called upper limit of (x
n
), and is denoted by
limsup x
n
. Similarly, let z
n
= inf
mn
x
m
= inf{x
n
, x
n+1
, . . .}. Then (z
n
) is in-
creasing, and it is bounded since (x
n
) is bounded. The limit of (z
n
) is called lower
limit of (x
n
), and is denoted by liminf x
n
. If (x
n
) is not bounded from above, then
its upper limit is equal to and if (x
n
) is not bounded from below, then its lower
limit is equal to . Summarizing
limsup x
n
= limx
n
= inf
nm
sup
kn
x
k
= lim
n
sup
kn
x
k
liminf x
n
= limx
n
= sup
nm
inf
kn
x
k
= lim
n
inf
kn
x
k
.
The basic properties of the upper and the lower limits are listed in the following
proposition:
Proposition 5.16. If (x
n
) and (y
n
) are sequences of real numbers, then:
(a) limsup(x
n
) = liminf x
n
and liminf(x
n
) = limsupx
n
.
(c) limsup(ax
n
) = a limsupx
n
and liminf(ax
n
) = a liminf x
n
for any a > 0.
(d) limsup(x
n
+y
n
) limsup x
n
+limsupy
n
and liminf x
n
+liminf y
n
liminf(x
n
+
y
n
).
(e) liminf x
n
limsup x
n
, with equality if and only if (x
n
) converges. In this
case limsup x
n
= limx
n
.
Theorem 5.17 (Bolzano-Weierstrass Theorem).
Let (x
n
) be a bounded sequence in R. Then (x
n
) has convergent subsequence.
Proof. By Proposition 5.14, (x
n
) has a monotone subsequence (x
n
k
). Since (x
n
k
)
is bounded, Proposition 5.12 implies that is converges.
Theorem 5.18 (Cauchys criterion). A sequence (x
n
) of real numbers converges
if and only if (x
n
) is a Cauchy sequence.
Proof. We already know that a convergent sequence is Cauchy (Proposition 5.7).
Hence we have to prove that a Cauchy sequence (x
n
) converges to some x R. By
Proposition 5.7, (x
n
) is bounded so that by Theorem 5.17, it contains a convergent
subsequence, say x
n
k
x. We claim that x
n
x. To see this, take > 0. Then
there is K such that
|x
n
k
x| < /2 for all k K.
26
Moreover, since (x
n
) is Cauchy, there is N such that
|x
n
x
m
| < /2 for all n, m N.
Take k such that n
k
N and let n N. Then
|x
n
x| |x
n
x
n
k
| +|x
n
k
x| < /2 +/2 = .
Hence (x
n
) converges to x, as claimed.
Example 5.19. Dene (x
n
) by setting x
1
= 1, x
2
= 2, and
x
n
=
1
2
(x
n1
+x
n2
) .
We claim that the sequence converges to x = 5/3. To see this we shall show that
(x
n
) is Cauchy and that the subsequence (x
2n+1
) converges to x = 5/3. First note
that 1 x
n
2 for all n. Next
|x
n
x
n+1
| =
1
2
n1
.
Indeed, for n = 1, |x
1
x
2
| = 1 =
1
2
0
. Assuming that |x
n
x
n+1
| =
1
2
n1
, we show
that |x
n+1
x
n+2
| =
1
2
n
. We have
|x
n+1
x
n+2
| =

x
n+1

1
2
(x
n+1
x
n
)

=
1
2
|x
n
x
n+1
| =
1
2
n
,
as claimed. Consequently, if m > n,
|x
n
x
m
| |x
n
x
n+1
| +|x
n+1
x
n+2
| +. . . +|x
m1
x
m
|

1
2
n1
+
1
2
n
+. . . +
1
2
m2
=
1
2
n1

1 +
1
2
+. . . +
1
2
mn1

<
1
2
n2
and since 1/2
n2
0, it follows that indeed (x
n
) is Cauchy. Now to nd the limit
of x we rst show that x
2n1
< x
2n
. This holds true for n = 1. Assuming that
x
2n1
< x
2n
, we show that x
2n+1
< x
2n+2
. From the assumption if follows that
x
2n1
<
1
2
(x
2n
+x
2n1
) < x
2n
.
So, x
2n1
< x
2n+1
and
x
2n+2
=
1
2
(x
2n+1
+x
2n
) >
1
2
(x
2n1
+x
2n
) = x
2n+1
,
as claimed. From this and |x
2n+1
x
n+2
| = 1/2
2n
, it follows that
x
2n+2
= x
2n+1
+
1
2
2n
.
27
Now use induction to prove that
x
2n+1
= 1 +
1
2
+
1
2
3
+. . . +
1
2
2n1
.
For n = 1 this is clear. Assuming the above equality we show that x
2n+3
=
1 +
1
2
+
1
2
3
+. . . +
1
2
2n+1
. We have
x
2n+3
=
1
2
(x
2n+2
+x
2n+1
) =
1
2

1
2
2n
+ 2x
2n+1

= 1 +
1
2
+
1
2
3
+. . . +
1
2
2n1
+
1
2
2n+1
,
as claimed. Now, note that
x
2n+1
= 1 +
1
2
+
1
2
3
+. . . +
1
2
2n1
= 1 +
1
2

1 +
1
2
2
+
1
2
4
+. . . +
1
2
2n2

= 1 +
1
2

1 +
1
4
+
1
4
2
+. . . +
1
4
n1

= 1 +
2
3

1
1
4
n

5
3
,
where we have used the formula

n
k=0
a
k
=
1a
n+1
1a
provided that a = 1.
28