Sequences of Real Numbers
Sequences of Real Numbers
It describes in detail various properties that a sequence may possess, the most important
of which is convergence. Roughly speaking, a sequence is convergent, or tends to a limit, if
the numbers, or terms, in the sequence approach arbitrarily close to a unique real number,
which is called the limit of the sequence. For example, we shall see that the sequence
1 1 1 1
1, , , , , . . .
2 2 4 5
is convergent with limit 0. On the other hand, the terms of the sequence
0, 1, 0, 1, 0, 1, 0, 1, . . .
do not approach arbitrarily close to any unique real number, and so this sequence is not
convergent. Likewise, the sequence
1, 2, 4, 8, 16, 32, . . .
is not convergent.
A sequence which is not convergent is called divergent. The sequence
0, 1, 0, 1, 0, 1, 0, 1, . . .
1, 2, 4, 8, 16, 32, . . .
is unbounded ; its terms become arbitrarily large and positive, and we say that it tends to
infinity.
Intuitively, it seems plausible that some sequences are convergent, whereas others are
not. The importance of convergence of sequences can be evident in many mathematical
procedures already developed. For instance, The Newton iterative procedure for solving
the equation f (x) = 0 heavily depends on the concept of convergence of sequences. For
example, to solve an equation f (x) = 0, using the Newton method, we proceed as follows:
We introduce a new function
f (x)
F (x) = x − ′
f (x)
We begin with a guess at a solution of f (x) = 0, say x1 and compute x2 = F (x1 ) in the hope
that x2 is closer to a solution than x1 was. The process is repeated so that
If |xn − xn+1 | is less that the desired accuracy, let xn+1 serve as the final approximation of
the solution of f (x) = 0. For example, we can solve the equation f (x) = x2 − 2 = 0 using
the Newton method as follows. Make an initial guess x1 = 1. The iterative process is given
by
f (xn )
xn+1 = xn − ′
f (xn )
n xn
1 1.000000
2 1.500000
3 1.416667
4 1.414216
5 1.414214
6 1.414214
√
The true solution to f (x) = x2 − 2 = 0 is x = 2 ≈ 1.414214 to 6 d.p. We can see from the
iteration procedure that after only 4 iterations, the procedure gives a value differing from
the true value by 0.000002 i.e.
√
|xn − 2| < 0.000002 f or n ≥ 4
Expressed more formally, if we are given a positive ϵ( we call it epsilon to suggest that it
measures an error) no matter how small, can we find a √ stage in this procedure so that the
value computed and all subsequent values are closer to 2 within ϵ?. In symbols, is there
an integer N (which will depend on ϵ) that is large enough so that
√
|xn − 2| < ϵ ∀n ≥ N ?
If this is true√
then this sequence has a remarkable property. We shall say that the sequence
converges to 2 and write √
lim xn = 2
n→∞
This is the beginning of the theory of convergence that is central to analysis. These ideas
lead to an ideal world of infinite precision, where sequences are precise tools in discussing
real numbers. All of analysis, both pure and applied, rests on an understanding of limits.
What you learn in this chapter will offer a foundation for all the rest that you will have to
learn later.
1.2 Sequences
Definition 1.1. A sequence is simply a list of real numbers a1 , a2 , a3 , . . . , an , . . .
There is a first element in the list, a second element, a third element, and so on continuing
in an order forever. Each of the numbers in the sequence is called a term. The symbol n is
called the index variable for the sequence. The real number an is called the nth −term of the
sequence. We use the notation (an )∞ n=1 or simply (an ) to denote this sequence.
1.3 Examples
∞
1. ( 21 , 23 , 34 , 54 , . . .) = n
n+1 n=1
(1, 3, 5, 7, 9, 11, . . .)
and the nth term of this sequence is described by the explicit formula
an = 2n − 1
The explicit formula can be obtained from the listed sequence as follows:
a1 = 1
a2 = 1 + 2 = 1 + (2 × 1)
a3 = 1 + 2 + 2 = 1 + (2 × 2)
a4 = 1 + 2 + 2 + 2 = 1 + (2 × 3)
a5 = 1 + 2 + 2 + 2 + 2 = 1 + (2 × 4)
..
.
an = 1 + |2 + 2 + 2{z+ · · · + 2} = 1 + 2(n − 1) = 2n − 1
n−1 times
pn 2n
This implies that an = qn
= 2n−1
and hence
∞
2n
(an ) =
2n − 1 n=1
1. Determine the nth term for the sequence whose first five terms are
− 21 , 28 , − 26 , 80 , − 242
6 24 120
,...
2. Find a sequence (an ) whose first four terms are
1 4 9 16
, , , ,...
1 3 7 15
Sequences can also be defined using recurrence relation i.e. one term (or more) of the
sequence is given explicitly and subsequent terms are defined in terms of the earlier terms
in the sequence. For example, the sequence 1, 3, 5, 7, 9, . . . can be defined recursively as
a1 = 1, an = an−1 + 2 ∀n = 2, 3, . . .
a1 = 2, an = 2an−1 ∀n ≥ 2
1.5 Exercise
1. Write down the first 5 terms of each of the following sequences.
3n
i. an = 3n ii. an = n!
iii. an = (− 14 ) iv. an = sin nπ
2
v. an = 2n
n+3
vi. a1 = 6, an+1 =
1 2
a
3 n
2. Write the next two apparent terms of the sequence. Describe the pattern you used to
find the terms.
i. 2, 5, 8, 11, . . .
ii. 5, 10, 20, 40, . . .
iii. 3, − 32 , 34 , − 38 , . . .
7
iv. 2
, 4, 92 , 5, . . .
3. For each of the following sequences, find an explicit formula for the nth term of the
sequence
i) 51 , −1 1 −1
,
7 9 11
, , . . .
1 3 7 15
ii) 2 , 4 , 8 , 16 , . . .
iii. (3, 5, 7, 9, 11, . . .)
4. For the recursively defined sequence a1 = −4 and an = an−1 + 6, for n ≥ 2, find its
explicit formula.
5. A government program that currently costs taxpayers Ksh. 4.5 billion per year is cut
back by 20% per year.
i. Write an expression for the amount budgeted for this program after n years. Ans:
An = 0.8n−1 (4.5b)
ii. Compute the budgets for the first four years.
√ p √ √
q p
6. Consider the sequence 2, 2 + 2, 2 + 2 + 2, . . .
i) 1 + 3n = 4, 7, 10, 13, . . .
1 n
The terms 1 − 2
→ 1 as n → ∞
iii) (−1)n = −1, 1, −1, 1, . . .
(−1)n
The terms alternate for this sequence as well, but n
→ 0 as n → ∞.
Remark 1.2. From these examples, we see several possibilities for the behavior of the terms
of a sequence as n → ∞ .
In two of the sequences, the terms approach a finite number as n → ∞. In the other two
sequences, the terms do not.
If the terms of a sequence approach a finite number L as n → ∞, we say that the
sequence is a convergent sequence and the real number L is the limit of the sequence. We
give an informal definition below.
Definition 1.3. Given a sequence (an ), if the terms an become arbitrarily close to a finite
number L as n becomes sufficiently large, we say (an ) is a convergent sequence and L is the
limit of the sequence. In this case, we write
lim an = L (1.6.1)
n→∞
In this case, we say the sequence (an ) is a convergent sequence. If a sequence does not
converge, it is a divergent sequence, and we say the limit does not exist or is infinite.
The positive number ϵ is the assigned tolerance demanded. Note that the smaller ϵ is,
the larger we should expect N to be.
Remark 1.5. It is clear from the definition that convergence or divergence of a sequence is
about the behaviour of the ’Tail-end’ of a sequence. Therefore, altering a finite number of
terms of a sequence does not affect its convergence or divergence.
2n + 1 2
− < ϵ, whenever n > N
3n + 5 3
The idea is to look for an N (ϵ) that will satisfy the convergence condition |an − L| <
ϵ, ∀n ≥ N.
(−1)n
2. To show that limn→∞ n
= 0, let ϵ > 0 and choose an integer N > 1ϵ . Then
(−1)n 1
− 0 = < ϵ ∀n ≥ N.
n n
1
3. Show that limn→∞ n
= 0.
1
− 0 < ϵ ∀n ≥ N
n
1
By the Archimedean property, there is an N ∈ N such that 0 < N
< ϵ.
1 1 1
−0 = < <ϵ
n n N
1
i.e. limn→∞ n
= 0.
4. The sequence ((−1)n ) oscillates between -1 and 1 and so cannot converge. That is,
suppose (−1)n → a for some a ∈ R.
5. Show that a sequence (an ) converges to zero if and only if the sequence (|an |) converges
to zero.
Solution Assume that the sequence (an ) converges to zero. Then, given ϵ > 0, there
exists a natural number N (ϵ) such that
Now, ∀n ≥ N , we have
||an | − 0| = |an | < ϵ
that is the sequence (|an |) converges to zero. Conversely, assume that the sequence
(|xn |) converges to zero. i.e. given ϵ > 0, ∃N (ϵ) such that
Proof. Let ϵ > 0 be given. Since we know that an → l1 , then by definition, there is some
N1 ∈ N such that
ϵ
|an − l1 | < ∀n > N1
2
Similarly, we know that an → l2 , and by definition, there is some N2 ∈ N such that
ϵ
|an − l2 | < ∀n > N2
2
Let N = max{N1 , N2 }. Then ∀n ≥ N we have that
n−1
b. limn→∞ n+1
=1
√ √
c. limn→∞ n+1− n =0
Theorem 1.8. Given sequences (an ) and (bn ) and any real number c, if there exist constants
M and L such that limn→∞ (an ) = M and limn→∞ (bn ) = L, then
i. limn→∞ c = c
an limn→∞ (an ) M
v. limn→∞ bn
= limn→∞ (bn )
= L
provided L ̸= 0 and for each bn ̸= 0.
Proof. We prove only part (iii), the rest left to the reader.
ϵ
Let ϵ > 0 be given. Then ∃N1 and N2 in N such that |an − M | < 2
∀n ≥ N1 and
|bn − L| < 2ϵ ∀n ≥ N2 .
Let N = max{N1 , N2 }.Then for all n ≥ N we have,
3n2 −4
Solution Let an = 2n2 +1
. We can rewrite an as
2
an = 3−4/n
2+1/n2
. Note that −4/n2 → 0 and 1/n2 → 0 so that 3−4/n2 → 3 and 2+1/n2 → 2
as n → ∞. It follows that an → 3/2 as n → ∞. Therefore, limn→∞ an = 3/2, and
hence (an ) is convergent
3 2
7n −5n +3n−9
2. (an ) = 3n 3 +4n2 −8n+2
5. (an ) = ((−1)n )
Solution: an = (−1)n , (an ) oscillates between -1 and 1 hence limit does not exist. (an )
diverges.
(n+1)!
6. (an ) = n!
Solution: an = (n+1)!
n!
= (n+1)n!
n!
= n + 1 → ∞ as n → ∞. Hence, the limit does not
exist and (an ) therefore diverges.
(n+1)!
7. (an ) = (n+2)!
(n+1)! (n+1)! 1
Solution: an = (n+2)!
= (n+2)(n+1)!
= n+2
→ 0 as n → ∞. Therefore, limit exist and
(an ) converges.
n
8. (an ) = 2n2 .
2n
Solution: an = n2
, apply L’Hopital’s Rule to find limit.
n n 2n (ln 2)2
limn→∞ 2n2 = limn→∞ 2 2nln 2 = limn→∞ 2
→ ∞ as n → ∞. The limit does not
exist and (an ) diverges.
9. (an ) = ((1 + 4/n)n )
Solution: an = (1+4/n)n . Use the fact that limx→0 (1+x)1/x = e or limx→∞ (1+1/x)x =
e.
2n+4 5n+1
Solution: an = 2n−2
and rewrite an as (1 + α)1/α .
2n+4
2n−2
= 2n−2+6
2n−2
6
= 1 + 2n−2 6
and let α be 2n−2 . This implies that n = 3/α + 1 and as
n → ∞, α → 0. Thus,
2n+4 5n+1
2n−2
= (1 + α)5(1+3/α)+1 = (1 + α)15/α+6 → e15 as α → 0. It converges.
2
11. (an ) = ln(n
n
)
ln(n2 )
Solution: an = n
apply L’Hopital’s Rule.
ln(n2 )
lim an = lim
n→∞ n→∞
n
2n
= lim ÷1
n→∞ n2
2
= lim = 0. It converges.
n→∞ n
2n3 +5
12. (an ) = 3n2 +1
3 2+5/n 3
2n +5
Solution: limn→∞ an = limn→∞ 3n 2 +1 = limn→∞ 3/n+1/n3 . The numerator goes to 2
and the denominator goes to 0. This means (an ) will increase without bound, i.e.
limn→∞ an = ∞. We therefore conclude that the sequence (an ) diverges.
9n2 −n−3
b. (xn ) = 2n3 +4n2 +n−4
5n3 +6
c. (xn ) = 4n2 +n+11
4 n
d. (xn ) = 1+ n
2n
e. (xn ) = n2
Theorem 1.10. Consider a sequence (xn ) and suppose there exists a real number L such
that the sequence (xn ) converges to L . Suppose f is a continuous function at L. Then there
exists an integer N such that f is defined at all values xn for n ≥ N , and the sequence
f (xn ) converges to f (L).
Proof. Let ϵ > 0. Since f is continuous at L, there exists δ > 0 such that |f (x) − f (L)| < ϵ if
|x − L| < δ. Since the sequence (xn ) converges to L, there exists an N such that |xn − L| < δ
for all n ≥ N. Therefore, for all n ≥ N , |xn − L| < δ, which implies |f (xn ) − f (L)| < ϵ. We
conclude that the sequence (f (xn )) converges to f (L).
Examples
3
13. Determine whether the sequence cos n2
converges. If it converges, find its limit.
q
2n+1
14. (an ) = 3n+5
. Determine whether this sequence converges or not.
q q q
2n+1
Solution: limn→∞ 3n+5
= limn→∞ 2n+1
3n+5
= 2
3
. The sequence converges with limit
q
2
3
.
q
2n+1
exercise 1.11. Determine if the sequence 3n+5
converges. If it converges, find its limit.
Theorem 1.12. Suppose that (an ) and (bn ) are convergent sequences such that limn→∞ an =
limn→∞ bn and that an ≤ cn ≤ bn for all n. Then (cn ) is also convergent and limn→∞ cn =
limn→∞ an = limn→∞ bn
Proof. Let L be the limit of the two sequences (an ) and (bn ). Choose N1 so that |an − L| < ϵ
if n > N1 , and also choose N2 so that |bn − L| < ϵ if n > N2 .
Set N = max N1 , N2 . Note that
an − L ≤ cn − L ≤ bn − L ∀n
and so
−ϵ < an − L ≤ cn − L ≤ bn − L < ϵ if n > N
From this, we see that −ϵ < cn − L < ϵ or |cn − L| < ϵ. This implies that cn → L as
n → ∞.
Examples:
cos n
15. Use the Squeeze Theorem to find limn→∞ n2
Theorem 1.13. Suppose that (an ) is a convergent sequence. Then the sequence (|an |) is
also a convergent sequence and
lim |an | = | lim an |
n→∞ n→∞
This implies that ||an | − |A|| < ϵ ∀n > N . Hence, limn→∞ |an | = |A| as required.
Thus, a sequence (an ) is bounded if there is a positive real number M such that
|an | ≤ M, ∀n ∈ N
Examples
1
is bounded since 0 < n1 ≤ 1∀n ∈ N
17. The sequence n
19. The sequence ((−1)n n) is not bounded above and it is not bounded below.
20. The sequence ((−1)n ) is bounded above by 1 and bounded below by -1.
Note that the sequence (an ) = ((−1)n ) = (−1, 1, −1, 1, −1, . . .) is bounded above by 1 and
bounded below by -1 but oscillates between -1 and 1 and so cannot converge i.e. limit does
not exist. This shows that a bounded sequence is not necessarily convergent. However, a
convergent sequence is necessarily bounded as shown below.
Proof. Let (an ) be a sequence of real numbers which converges to say L. Then, for every
ϵ > 0, there is an N so that
In particular we could take just one value of ϵ, say ϵ = 1, and find a number N so that
|an | ≤ K ∀n
as required.
Remark 1.16. The converse of above theorem is not necessarily true. That is, there are
sequences which are bounded but do not converge. One such example is the sequence ((−1)n )
which is bounded below by -1 and above by 1 but does not converge since the limit does
not exist. We shall however see later that every bounded sequence which is monotone will
always converge.
From the above Theorem, we obtain the following test for divergence of a sequence.
Examples:
22. The sequence (2n) and ((−1)n n) are both unbounded, so they are both divergent.
Solution: The easiest proof is to show that it is unbounded and hence by the corollary
above it diverges.
√
4. 1 − 21n
5. (1 + (−1)n ) 6. ( n n)
• Thus, if (an ) is increasing then limn→∞ an = sup(an ) and if (an ) is decreasing then
limn→∞ an = inf(an ).
Examples of monotonic sequences 1. The sequences (1, 1, 2, 3, 5, . . .), (1, 2, 3, 4, . . . , n, . . .), (1, 2, 2, 3, 3, 3,
are increasing
2. The sequences (1, 1/2, 1/3, . . . , 1/n, . . .), (3, 1, 0, 0, −3, −7, . . .) are decreasing
3. The sequence (n2 ) is strictly increasing
4. The sequence (−n) is strictly decreasing.
5. The sequences ((−1)n n), ((−1)n+1 ) are not monotone.
3n 3x
6. The sequence (an ) = n+2 is monotone. Letting an = f (x) = x+2 , we see that
′ 6
f (x) = (x+2)2 > 0 ∀x = 1, 2, 3, . . . and by the first derivative test for increasing/decreasing
functions, f (x) is increasing and consequently, (an ) is increasing and hence monotone.
Up to now, we know that if a sequence converges, then there is some way of finding its
limit explicitly. However, it is sometimes possible to prove that a sequence is convergent,
even though we do not know its limit. This situation occurs with monotonic sequences that
are bounded.
Proof. We know that if a sequence converges then it is bounded. To prove the converse, let
(an ) be a bounded increasing sequence. Since (an ) is bounded above, the set {an : n ∈ N}
has a least upper bound (supremum), L say.
We claim that limn→∞ an = L. Let ϵ > 0 be given. By the characterization of supremum,
there is an integer N such that aN > L − ϵ. Since (an ) is increasing, an ≥ aN ∀n > N
and so an > L − ϵ ∀n > N . Thus, |an − L| = L − an < ϵ ∀n > N . This implies that
|an − L| < ϵ ∀n > N .
The proof for the case when the sequence (an ) is decreasing is similar.
3n
The MCT tells us that a sequence such as n+2 which is increasing and bounded above
by 3 must be convergent.
MCT is often used when we suspect that a sequence is convergent, but we cannot find
the actual limit.
Solution We show that the sequence is monotone and bounded. Its convergence will
then follow from the Monotone Convergence Theorem above.
Monotonicity
Consider fx = x+1x
for all x ∈ [1, ∞). Then f ′ (x) = − x12 < 0 for all x ∈ [1, ∞). Thus,
f is decreasing on [1, ∞). Therefore, f (n) > f (n + 1) i.e. n+1
n
> n+2
n+1
for all n ∈ N
Boundedness n+1n
= 1 + n1 > 1∀n ∈ N. Thus, the sequence is bounded below below by
1. Thus, the sequence n+1
n
is convergent by the MCT.
25. Show that the sequence √1n is convergent.
Solution Note that √1n > 0∀n ∈ N and hence the sequence is bounded from below. We
have to show that the sequence is monotone decreasing.
Consider f (x) = √1x for all x ∈ [1, ∞). Then f ′ (x) = − 2x13/2 < 0 for all x ∈ [1, ∞).
1
Thus, f is decreasing on [1, ∞). Therefore, √n+1 < √1n . So the sequence is monotone
decreasing and bounded from below (hence bounded). Convergence of √1n follows
from MCT.
Solution: The sine function is bounded above by 1 and bounded below by -1 but is
not monotone and hence does not converge by the MCT.
n
exercise 1.22. 1. Show that 1 + n1 is a convergent sequence, and hence find its limit.
n
2. Show that the sequence 4n! is convergent.
3. Give an example of a sequence that is
a) bounded but not convergent
b) Convergent but not bounded
c) monotonic but not bounded
d) monotonic and bounded but not convergent
e) unbounded but not monotonic
f) divergent to ∞ but not monotonic
g) convergent but not monotonic.
1.15 Subsequences
The sequence (1, −1, 2, −2, 3, −3, 4, −4, 5, −5, . . .) appears to contain within itself the two
sequences
(1, 2, 3, 4, 5, . . .) and
(−1, −2, −3, −4, −5, . . .)
In order to have a language to express this, we introduce the term subsequence. We say
that the latter two sequences are subsequences of the first sequence.
A subsequence of (an ) is just a sequence of terms from (an ) with one requirement that each
term in the sequence is moving further along in the original sequence. For example, from the
sequence (1, −1, 2, −2, 3, −3, 4, −4, 5, −5, . . .) we can pick the subsequence (1, 2, 3, 4, 5, . . .)
and (−1, −2, −3, −4, −5, . . .).
Definition 1.23. Let (an ) be a sequence of real numbers and let (nk ) be a sequence of
natural numbers such that n1 < n2 < n3 < . . . . Then (an1 , an2 , an3 , . . .) is a subsequence of
(an ) and is denoted by (ank ) i.e. (ank ) = (asn1 , an2 , an3 , . . . , ank , . . .)
Example:
1. We can consider 1, 2, 3, 4, 5, . . . to be a subsequence of sequence (1, −1, 2, −2, 3, −3, 4, −4, 5, −5, . . .)
because it contains the first, third, fifth, e.t.c. terms of the original sequence. Here
n1 = 1, n2 = 3, n3 = 5, . . . and pick from (1, −1, 2, −2, 3, −3, 4, −4, 5, −5, . . .) elements
a1 , a3 , a5 , a7 , . . ..
2. Let (an ) be the sequence (1, 2, 1/2, 3, 1/3, . . .). Then (1, 1/2, 1/3, . . .) and (1, 2, 1/3, . . .)
are subsequences of (an ) while (1/2, 1, 1/3, 3, . . .) is not a subsequence of (an ).
Theorem 1.24. Let (an ) be a sequence which converges to a. Then any subsequence of (an )
converges to a.
Proof. Let (ank ) be a subsequence of (an ) and let ϵ > 0 be given. Then ∃ an N ∈ N such
that |an − a| < ϵ for each n ≥ N. Thus, when k ≥ N , we have that nk ≥ k ≥ N and so
|ank − a| < ϵ ∀k ≥ N . Thus, limk→∞ ank = a.
From this result, we note that if a sequence (an ) has two convergent subsequences whose
limits are not equal then the sequence (an ) is divergent.
We now state the Bolzano-Weierstrass Theorem which guarantees existence of a conver-
gent subsequence for any bounded sequence.
Theorem 1.25. Every bounded sequence (an ) of real numbers contains a convergent subse-
quence.
lim |an − am | = 0.
n,m→∞
Examples
n+1
1. Show that the sequence (an ), where an = n
, is a Cauchy sequence.
Solution: For all n, m ∈ N
n+1 m+1
|an − am | = −
n m
mn + m − nm − n
=
mn
m−n m+n
= ≤
nm nm
m+n 2m
Therefore, if m ≥ n, then |an − am | ≤ mn
≤ nm
= n2 .
1
Let ϵ > 0 be given. Then there is an N ∈ N such that N
< 2ϵ . Thus, for all n ≥ N, we
have
|an − am | = n+1 − m+1 < n2 < N2 < ϵ
n m
1 1 1
3. Determine whether the sequence (xn ) where xn = 1 + 5
+ 52
+ ··· + 5n
is a Cauchy
sequence.
Solution: Take m > n
1 1 1 1
|xm − xn | = + 2 1 − + 2 1
5 5 + · · · + 5m 5 5 + ··· + 5n
1 1 1
= n+1
+ n+2 + · · · + m
5 5 5
1 1 1 1
= 1 + + 2 + · · · + m−n−1
5n+1 5 5 5
1 1 1
≤ n+1
1 + + 2 + · · · G − series
5 5 5
1 1 1
= n+1
= <ϵ
5 1 − 1/5 4 · 5n
This implies 5n > 4ϵ1 and hence n > log5 4ϵ1 . Thus, given ϵ > 0 there exists N (ϵ) = log5 1
4ϵ
such that |xm − xn | < ϵ ∀n, m > N . Therefore, (xn ) is a Cauchy sequence.
n2 −1
exercise 1.27. Prove that n2
is Cauchy using directly the definition of Cauchy se-
quences.
Some Properties of Cauchy Sequences
Theorem 1.28. Every convergent sequence (xn ) is a Cauchy sequence.
Proof. Assume that (xn ) converges to x. Then, given any ϵ > 0, there exists an N ∈ N such
that
|xn − x| < ϵ/2 ∀n ≥ N
Now for all n, m ≥ N we see that
|xn − xm | = |(xn − x) + (x − xm )|
≤ |xn − x| + |x − xm |
< ϵ/2 + ϵ/2 = ϵ
Theorem 1.30. A sequence (xn ) is convergent if and only if it is a Cauchy sequence i.e.
for each ϵ > 0 there exists an integer N with the property that
Proof. Let (xn ) be a convergent sequence. Then it follows that (xn ) is Cauchy. Conversely,
let (xn ) be a Cauchy sequence. We have to show that (xn ) is convergent to some real number.
Since (xn ) is Cauchy, we have that (xn ) is bounded. Therefore, by the Bolzano-Weierstrass
Theorem, there is a subsequence (xnk ) that converges to some real number x∗ .
We claim that (xn ) converges to x∗ .
Let ϵ > 0 be given. Then there exists natural numbers N1 and N2 such that
|xn − xm | < ϵ/2 ∀n, m ≥ N1 and |xnk − x∗ | < ϵ/2 ∀ k ≥ N2
Let N = max{N1 , N2 }. Then ∀n ≥ N, we have
|xn − x∗ | ≤ |xn − xnk | + |xnk − x∗ | < ϵ/2 + ϵ/2 = ϵ
Therefore, xn → x∗ as n → ∞ verifying the claim.
Examples:
(−1)n
1. Use Cauchy’s criterion to show that the sequence converges. n
n
Solution: We must show that the sequence (−1) n
is Cauchy. To that end, let ϵ > 0
(−1)n
and xn = n
. Then, ∀n, m ∈ N with m ≥ n,
(−1)n (−1)m 1 1 1 1 2
|xn − xm | = n
− m
≤ n
+ m
≤ n
+ n
= n
yn = sup{xk : k ≥ n} zn = inf{xk : k ≥ n}
As n increases, the sup and inf are taken over smaller sets, so (yn ) is monotone decreasing
and (zn ) is monotone increasing.
The limit of these sequences are the lim sup and lim inf, respectively, of the original sequence.
Examples:
1. Consider the bounded, increasing sequence (0, 1/2, 2/3, 3/4, . . .), xn = 1 − 1/n
defining yn and zn as above, we have
In this case the original sequence does not converge since lim supn→∞ xn ̸= lim inf n→∞ xn .
Alternatively, one can find lim sup and lim inf of (xn ) = ((−1)n+1 ) as follows:
Consider the sequence (xn ) where for each n ∈ N, xn = (−1)n . Then
To study the ’limit’ of this sequence, consider some subsequences of the original sequence.
After considering all subsequences of (xn ), we have a set of collection of limit of subsequences
L i.e.
L = {−1, 1}
We define limit superior and limit inferior of (xn ) to be
4. Consider the unbounded, non-monotone sequence (xn ) = (1, −2, 3, −4, 5, . . .). Find lim sup
and lim inf
Solution: We have two subsequences
and L = {∞, −∞} and lim supn→∞ xn = sup L = ∞, lim inf n→∞ xn = inf L = −∞.
Thus, the sequence does not converge.
5. Consider the sequence (1, 1/2, 3, 1/4, 5, . . .). Find lim sup and lim inf
Solution: We have two subsequences
and L = {∞, 0} and lim supn→∞ xn = sup L = ∞, lim inf n→∞ xn = inf L = 0.
Thus, the sequence does not converge.
n+1
6. Consider the sequence (2, −3/2, 4/3, −5/4, . . .), xn = (−1)n+1 n
. Find lim sup
and lim inf
Solution: We have two subsequences
and L = {1, −1} and lim supn→∞ xn = sup L = 1, lim inf n→∞ xn = inf L = −1.
Thus, the sequence does not converge.
exercise 1.33. For each of the following, find the limit superior and limit inferior of the
sequence
(−1)n
1. (xn ) where xn = n+1
n
2. (xn ) where xn = 2(−1)n + n+1
3. (an ) where 1
2 n+1 if n is odd
an =
1 if n is even