Sei sulla pagina 1di 19

Chapter 3

Sequences

Both the main elements of calculus (differentiation and integration) require the
notion of a limit. Sequences will play a central role when we work with limits.
Definition 3.1. A Sequence is a function whose domain is the natural numbers.
We sometimes denote a sequence as {f (n)}, where f (n) is a function on the
natural numbers, and we do not specify the range.
Example 3.2. Let
1
 
n n
f (n) = .
n2 (−1)n
Then f is a function from the natural numbers into 2 × 2 matrices, and {f (n)} is
a sequence of 2 × 2 matrices.
Example 3.3. Let
f (n) = xn
on [0, 1]. Then f is a function from the natural numbers to functions on [0, 1], and
{f (n)} is sequence of functions.
Example 3.4. Let
1
f (n) =
.
n
Then f is a function from the natural numbers to the real numbers, and {f (n)} is
a sequence of numbers.
Before proceeding to define the limit of a sequence, try to compute the following
limits. If you find the limit, ask yourself how you did it and what does the limit
represent. Imagine explaining the process to a freshman.
n+1 3. lim (ln(n + 1) − ln n)
1. lim n→∞
n→∞ 3n + 2
sin n
2. lim 4. lim (1 + n)1/n
n→∞ n n→∞

Answers: 1. 1/3; 2. 0; 3. 0; 4. 1.

17
18 3. Sequences

3.1. Convergence
For the rest of this chapter we will only consider sequences of real numbers, typically
denoted {an }.
Definition 3.5. A sequence of real numbers, {an }, converges to a real number, a,
if and only if, for each ǫ > 0, there exists a natural number N , such that, for all
n ≥ N , |an − a| < ǫ.

In plain language the definition says an is close to a for large n. We note that
implementing the definition requires a candidate for the limit. Later you will see
that we can establish the existence of a limit without having to actually know the
limit.
1
Example 3.6. Prove lim = 0.
n→∞ n

Proof. Let ǫ > 0. By the Archimedean Property we can find a natural number N
so that 1ǫ < N , or N1 < ǫ. Thus for all n ≥ N
1 1
≤ < ǫ,
n N
or
1
− 0 < ǫ,
n
for n ≥ N . 

Sometimes a limit does not exist. We need methods to deal with limits not
existing. One possibility would be to negate the definition.
Lemma 3.7. A sequence {an } does not converge to the number a if there exists a
ǫ0 > 0 such that, for all N ∈ N, there exists a n0 ≥ N with |an0 − a| ≥ ǫ0 .

Proof. In logic notation the definition of convergence of a sequence is


(∀ǫ > 0)(∃N ∈ N)(∀n ≥ N )(|an − a| < ǫ)
or
(∀ǫ > 0)(∃N ∈ N)(∀n ∈ N)(n ≥ N → |an − a| < ǫ).
Negating the two, we find
(∃ǫ0 > 0)(∀N ∈ N)(∃n0 ≥ N )(|an − a| ≥ ǫ)
or
(∃ǫ0 > 0)(∀N ∈ N)(∃n0 ∈ N)(n0 ≥ N and |an − a| ≥ ǫ).

n
Example 3.8. Prove the sequence {(−1) } does not have a limit.

Proof. The difficulty in applying the lemma is that we have to show there is no
limit for each real number a. Suppose a = 1. Set ǫ0 = 1. Then, for all N ∈ N,
one of |aN − 1| or |aN +1 − 1| is 2, and for all N ∈ N there exists a n0 ≥ N with
|an0 − 1| ≥ ǫ0 . A similar argument hold if we choose a = −1. Finally suppose a
is not equal to 1 or −1. Let ǫ0 = min{|a − 1|, |a + 1|}. Then given any N ∈ N we
have |an − a| ≥ ǫ0 for all n ∈ N. 
3.2. Some Topology 19

Here is another proof which uses contradiction.

Proof. Suppose the limit did exist. Then, for ǫ = 1/2, there exists a N ∈ N such
that |an − a| < 1/2 for n ≥ N . But
1 1
2 = |1 + 1| = |(1 − a) + (a + 1)| ≤ |1 − a| + |a + 1| = |1 − a| + | − 1 − a| < + = 1
2 2
a contradiction. 
Theorem 3.9. The limit of a sequence, if it exists, is unique.

Proof. Suppose there were two limits, say a and b. Given any ǫ > 0 there exists
an N1 and N2 such that |an − a| < ǫ/2 and |an − b| < ǫ/2 for all n ≥ max{N1 , N2 }.
Then
|a − b| = |a − an + an − b| ≤ |an − a| + |an − b| < ǫ,
for all ǫ > 0. By trichotomy, either |a − b| = 0 for |a − b| > 0. If the latter
holds, set ǫ = |a − b|/2. Then on the one hand |a − b| < ǫ, while on the other,
|a − b| > |a − b|/2 = ǫ - a contradiction. Thus |a − b| = 0 or a = b. 

3.2. Some Topology


A large part of mathematics consists of recognizing and generalizing patterns. If we
look carefully at the definition of convergence for sequences we see that we really
do not need the field axioms. All we require is the notion of closeness. We can have
that concept without the use of addition or subtraction (or even absolute value).
Definition 3.10. A subset O of R is called open in R if, for each point x ∈ O,
there is a r > 0 such that all points y in R satisfying |x − y| < r also belong to the
set O.
Example 3.11. The set O = {x | 0 < x < 1} is open in R.
Example 3.12. The set O = {x | 0 < x ≤ 1} is not open in R.

We state without proof the following theorems.


Theorem 3.13. Open Set Properties.
(a) The empty set and R are open in R.
(b) The intersection of any two open sets is open in R.
(c) The union of any collection of open sets is open in R.
Definition 3.14. (a) A subset of R is called closed in R if its complement is open
in R.
(b) A neighborhood of x ∈ R is any set containing an open set containing x.
(c) A point x ∈ R is called an boundary point of a set S ⊂ R if every neighborhood
of x contains a point in S and a point not in S.
(d) A point x ∈ R is called an interior point of a set S ⊂ R if there exists a
neighborhood of x contained in S.
(e) A point x ∈ R is called an exterior point of a set S ⊂ R if there is a neighbor-
hood of x which is entirely contained in complement of S.
20 3. Sequences

Example 3.15. The set O = {x | 0 ≤ x ≤ 1} is closed in R.


Example 3.16. Consider the set O = {x | 0 ≤ x ≤ 1}. The boundary points are
x = 0 and x = 1. Moreover x = 1/2 is an interior point.

Again without proof we have


Theorem 3.17. Let O ⊂ R. The following statements are equivalent
(a) O is open in R;
(b) every point in O is an interior point of O.
(c) O is a neighborhood of each of its points.
Theorem 3.18. A set is closed if and only if it contains all of its boundary points.

Theorem 3.19. A subset of R is open if and only if it is the union of a countable


collection of open intervals.

Finally we can recast the definition 3.5. The advantage of doing so it that we
do not need the field axioms to define an open set, and we could, if we had more
time, generalize the notion of convergence to set with less structure than the ones
we consider here.
A sequence {an } eventually has a certain property if it has that property for
all n ≥ N for some N ∈ N.
Definition 3.20. A sequence of real numbers {an } converges to a number a if, for
each neighborhood of a, an is eventually in that neighborhood.

3.3. Algebra of Limits


In this section we establish some familiar properties of sequences.
Theorem 3.21. Suppose {an } and {bn } are convergent sequence, with limits a and
b respectively. Then
(a) For all α ∈ R, lim αan = αa.
n→∞
(b) lim (an + bn ) = a + b.
n→∞
(c) lim an bn = ab.
n→∞
an a
(d) If bn 6= 0 for all n ∈ N and b 6= 0, then lim = .
n→∞ bn b

Proof. We will prove (c) and (d) and leave the other two as exercises. We are
given the following. Given any ǫ′ > 0 there exists N1 , N2 , natural numbers, such
that |an − a| < ǫ′ and |bn − b| < ǫ′ for all n ≥ max{N1 , N2 }. We calculate
|an bn − ab| = |(an − a)bn + a(bn − b)| ≤ |bn ||an − a| + |a||bn − b|.
We know |an − a| and |bn − b| are small. We have to worry about |bn | becoming
large. Let us suppose for the moment that |bn | ≤ M for all n ∈ N and some M > 0.
Then
|an bn − ab| ≤ M |an − a| + |a||bn − b|.
3.3. Algebra of Limits 21

Let ǫ > 0 be given. In one case we choose ǫ′ = ǫ/2M , and in the other ǫ′ = ǫ/2a.
Then
ǫ ǫ
|an bn − ab| ≤ M |an − a| + |a||bn − b| < + = ǫ
2 2
for all n ≥ max{N1 , N2 }. We show convergent sequencs are bounded in the next
lemma.
The proof of (d) is similar. Indeed,

an a ban − abn

bn − =
b bbn

ban − ba + ba − abn
=
bbn
|an − a| |a||bn − b|
≤ + .
|bn | |b||bn |
Again we must worry about 1/|bn | becoming large. Let us suppose for the moment
that 1/|bn| ≤ M for n ≥ N3 , for some N3 ∈ N and for some M . Then, as in Part
(c), let ǫ > 0 be given. We choose ǫ′ = ǫ/(2M ) for the bound on |an − a| and
ǫ′ = |b|ǫ/(2|a|) for the bound on |bn − b|. For n ≥ max{N1 , N2 , N3 }, we have

an a ǫ ǫ
bn − b < 2 + 2 .

We have two bounds to establish.


Lemma 3.22. Suppose {bn } converges to b. Then a M > 0 exists so that |bn | ≤ M
for all n ∈ N.

Proof. Let ǫ = 1 in Definition 3.5. Then a natural number N exists so that


|bn − b| < 1 for all n ≥ N . The triangle inequality implies, for such n,

|bn | − |b| ≤ |bn | − |b| ≤ |bn − b|.
That is, |bn | ≤ 1 + |b| for n ≥ N . Set M = max{|b1 |, |b2 |, . . . , |bN |, 1 + |b|}, and we
find |bn | ≤ M for n ∈ N. 
Lemma 3.23. Suppose {bn } converges to b with bn =6 0 for all n ∈ N and b 6= 0.
1
Then a N ∈ N and a M > 0 exists so that ≤ M for all n ≥ N .
|bn |

Proof. Here we have to worry about |bn | becoming small. Let ǫ = |b|/2 in Def-
inition 3.5. A N ∈ N exists such that |bn − b| < |b|/2 for n ≥ N . The triangle
inequality shows
|b|
|bn | = |bn − b + b| ≥ |b| − |bn − b| >
2
for all n ≥ N . Set M = 2/|b| and the result follows. 
Theorem 3.24. (Squeeze Theorem) Let {xn },{yn }, and {zn } be three sequences
of real numbers such that xn ≤ yn ≤ zn for all n. Suppose further that {xn } and
{zn } converge to x. Then {yn } converges x.
22 3. Sequences

Proof. A good place to start would be to write down what we know: Given any
ǫ > there exists Ni ∈ N, i = 1, 2, such that |xn − x| < ǫ for all n ≥ N1 and
|zn − x| < ǫ for all n ≥ N2 . The above inequalities imply
−ǫ < xn − x ≤ yn − x ≤ zn − x < ǫ
for n ≥ max{N1 , N2 }. That is, |yn − x| < ǫ for such n. 

Corollary 3.25. Given any ξ ∈ Qc , there exists sequence of rational numbers {rn }
which converge to ξ. That is, irrational numbers can by approximated with rational
numbers.

Proof. Clearly ξ < ξ + n1 . By the density of rationals there exists rn such that
ξ < rn < ξ + n1 . The squeeze theorem implies {rn } converges to ξ. 

While the supremum of a set may Not be in the set, it can however be approx-
imated by elements in the set.
Corollary 3.26. Let E ⊂ R be non empty and bounded from above. Then a
sequence {xn } in E such that xn → α = sup E as n → ∞.

Proof. The set E has a supremum, α, by the completeness axiom. Since α − n1


is no longer an upper bound there exists wn ∈ E with α − n1 < wn < α (Theorem
2.11). The squeeze theorem shows {wn } converges to α. 

3.4. Monotone Convergence Theorem


We will develop two techniques to show a sequence converges without having to find
the limit and apply Definition 3.5. The application of the Monotone Convergence
Theorem is one of the techniques. It is equivalent to the completeness axiom.
Theorem 3.27. (Monotone Convergence Theorem) Suppose {an } is a monotone
increasing, bounded sequence of real numbers. That is, an ≤ an+1 and |an | ≤ M
some M and for all n ∈ N. Then {an } has a limit.

Proof. To apply the definition for convergence we need a candidate for the limit.
Set E = {a1 , a2 , . . .}. The E is non empty and bounded. By the completeness
axiom E has a supremum. Set a = sup E. Let ǫ > 0. Then a − ǫ is no longer an
upper bound for E. Hence a aN exist such that a − ǫ < aN ≤ a. Since the sequence
is increasing a − ǫ < an ≤ a for all n ≥ N . That is 0 ≤ a − an < ǫ or |a − an | < ǫ
for all n ≥ N . 

Example 3.28. Consider the sequence


1 1 1
an = + + ···+ .
n+1 n+2 2n
The sequence is bounded since
1 1 1 n
|an | ≤ + + ··· + = ≤ 1.
n+1 n+1 n+1 n+1
3.4. Monotone Convergence Theorem 23

It is also monotone since


 
1 1 1 1 1
an+1 − an = + ...+ + +
n+2 n+3 2n 2n + 1 2n + 2
 
1 1 1
− + + ···+
n+1 n+2 2n
1 1 2
= − =
2n + 1 2(n + 1) (n + 1)(2n + 1)
≥ 0
By the monotone convergence theorem, the sequence {an } has a limit (in fact, we
will see later that lim an = ln 2).
n→∞
Example 3.29. Suppose 0 < a < 1. Prove lim an = 0.
n→∞

n
Proof. Set an = a . The sequence is a monotone decreasing sequence in this case.
The monotone convergence theorem still applies (see exercise 3.20. We need to
show the sequence is bounded from below. This is easy since an = an = |a|n ≥ 0.
We show the sequence is monotone decreasing. That is, an+1 ≤ an . We do this
by induction. Since a < 1, a2 < a. This is the base step. Suppose an < an−1 .
That is, an < an−1 . Multiplying by a, we find an+1 < an . That is, an+1 ≤ an .
By induction an+1 ≤ an for all n ∈ N, and the sequence is decreasing. By the
monotone convergence theorem, {an } has a limit.
To find the limit, we apply the algebra of limits (we had to know the limit
existed to do this!). Set lim an = L. Then
n→∞

L = lim an+1 = lim aan = a lim an = aL.


n→∞ n→∞ n→∞
That is, L(1 − a) = 0. This implies L = 0. 

Example 3.30. Suppose the sequence {xn } is defined recursively by xn+1 = 2 −


1/xn with x1 = 2. Prove that {xn } has a limit and find the limit.

Proof. Notice 1 ≤ xn ≤ 2 for all n ∈ N. Indeed, the inequality holds for n = 1.


Suppose 1 ≤ xn−1 ≤ 2. Then xn = 2 − 1/xn−1 , and under the assumption on xn−1 ,
1 ≤ xn ≤ 2. By induction, the inequality holds for all n.
The sequence is also decreasing. Indeed, for all n
(xn − 1)2
 
1
xn+1 − xn = 2 − − xn = − <0.
xn xn
By the monotone convergence theorem the sequence has a limit. Moreover, by the
algebra of limits, the limit satisfies x = 2 − 1/x, or x = 1.


Example 3.31. Find a rational sequence converging to 2.

Proof. We know by Corollary 3.25 such a sequence exists. Newton’s method for
finding zeros of a differentiable function is
f (xn )
xn+1 = xn − ′ .
f (xn )
24 3. Sequences

We do not need to study Newton’s method-we just use it to find a candidate for
the sequence. Applying the method to the function f (x) = x2 − 2, we find
 
1 2
(3.1) xn+1 = xn + .
2 xn
We arbitrary choose x0 = 2. Then x1 = 3/2, x2 = 17/12, x3 = 577/408 .... You √
can check that the sequence of rational numbers seems to converge (rapidly) to 2.
To prove the convergence we apply the monotone √ convergence theorem. We
first show the sequence {xn }∞0 is bounded. Clearly, 2 ≤ x0 = 32 ≤ 2. Suppose

2 ≤ xn ≤ 2. Then, using (3.1),
   
1 2 1 2
√ min x + ≤ x n+1 ≤ √ max x + .
2≤x≤2 2 x 2≤x≤2 2 x
√ √
The function g(x) = 21 x + x2 has a minimum of 2 at x = 2. The right side is


bounded by
  √
1 2 x 1 1 2+ 2
√max x+ ≤ √max + √max ≤1+ √ ≤ ≤2.
2≤x≤2 2 x 2≤x≤2 2 2≤x≤2 x 2 2

By induction 2 ≤ xn ≤ 2 for all n ∈ N.
Next we prove {xn } is monotone decreasing. Clearly x1 − x0 = −1/2 ≤ 0.
Suppose xn − xn−1 ≤ 0. Again using (3.1),
 
1 1
xn+1 − xn = (xn − xn−1 ) − .
2 xn xn−1

Since xn ≥ 2 for all n ∈ N,
1 1 1 1
− ≥ −√ √ ≥0.
2 xn xn−1 2 2 2
Hence, xn+1 − xn ≤ 0, and, by induction, xn − xn−1 ≤ 0 for all n ∈ N.
The monotone convergence applies, and {xn } has a limit. Suppose the limit is
L. Then, using the algebra of limits,
   
1 2 1 2
L = lim xn+1 = lim xn + = L+ .
n→∞ n→∞ 2 xn 2 L

Solving for L, we find L = 2.


We have one final application of the monotone convergence theorem. As you


will see, we need to know when an infinite intersection of intervals is non empty.
Example 3.32. Set In = (0, 1/n) = {x | 0 < x < 1/n}. The set {In } is a sequence
of intervals. We will be interested in the infinite intersection

\
E := = {x | x ∈ In , ∀n ∈ N}.
n=1

In this case note E = ∅.


3.5. Subsequences and The Bolzano-Weierstrass Theorem 25

Example 3.33. Set In = (n, ∞) = {x | n < x}. Again the set {In } is a sequence
of intervals. Here

\
E := = {x | x ∈ In , ∀n ∈ N} = ∅.
n=1

Definition 3.34. A sequence of intervals {In } is called nested if I1 ⊃ I2 ⊃ I3 ⊃ . . ..

Here are sufficient conditions for the intersection to be non empty.


Theorem 3.35. (Nested-Cell Theorem) Let {In } be a sequence of nested, closed,

\
bounded intervals. Then = {x | x ∈ In , ∀n ∈ N} is non empty.
n=1

Proof. By assumption In = [an , bn ] with a1 ≤ a2 ≤ . . . and b1 ≥ b2 ≥ . . .. For any


natural numbers n and k
ak ≤ an ≤ b n for k<n
ak ≤ bk ≤ bn for k ≥ n.
The two inequalities imply ak ≤ bn for all k ∈ N and any n. Thus the monotone
increasing sequence {an } is bounded by bn for any n. By the monotone convergence
theorem lim an = a := sup{an }. The same two inequalities above now imply that
n→∞
a is a lower bound for {bn }. Since {bn } is monotone decreasing and bounded below,
the monotone convergence theorem implies lim bn = b := inf{bn }. Moreover
n→∞
a ≤T
b. Finally, let x ∈ [a, b]. Then ai ≤ a ≤ x ≤ b ≤ bi for all i ∈ N. That is
x∈ ∞n=1 In . 

3.5. Subsequences and The Bolzano-Weierstrass Theorem


A subsequence of a sequence {an } is a subset of {an } which is ordered. More
specifically,
Definition 3.36. Let {an } be a sequence and {nk } be any sequence of natural
numbers such that n1 < n2 < n3 < . . .. The sequence {ank } is called a subsequence
of {an }.
n
Example 3.37. The sequence with an = 1+(−1) 2 does not converge. However, if
nk = 2k, {a2k } = {0, 0, 0, . . .}. Similarly, nk = 2k − 1 produces the subsequence
{a2k−1 } = {1, 1, 1, . . .}.
Theorem 3.38. A sequence converges to a if and only if each of it subsequences
converges to a. In fact, if every subsequence converges, they all converge to the
same limit.

Proof. If all subsequences converge, the sequence converges since the sequence is
a subsequence of itself.
Conversely, suppose the sequence converges to a and {ank } is a subsequence.
Let ǫ > 0. There exists a natural number N such that |an − a| < ǫ for all n ≥ N .
By the construction of the {nk } we have k ≤ nk for all k ∈ N. Thus, for k ≥ N ,
nk ≥ N . This implies |ank − a| < ǫ for all k ≥ N , and the subsequence {ank }
converges to a. 
26 3. Sequences

A restatement of part of the theorem gives


Corollary 3.39. If {an } converges, all subsequence converge to the same limit.

This gives us another technique for showing a sequence does not converge. We
need only find two convergent subsequences with distinct limits.
Example 3.40. Show the sequence an = (−1)n does not converge using three
different methods.

Proof. We have already shown non convergence using two difference methods in
Example 3.8. Note that {a2k } and {a2k−1 } converge to 1 and −1 respectively. Thus
{an } does not converge. 

The importance of the following theorem cannot be overstated.


Theorem 3.41. (Bolzano-Weierstrass) Every bounded sequence of real numbers
has a convergent subsequence.

Proof. Let us denote the sequence {xn }. We apply the Nest Cell Theorem,
Theorem 3.35. We will construct inductively a sequence of intervals such that
I1 ⊃ I2 ⊃ I3 ⊃ . . ..
Since the sequence is bounded, a ≤ xn ≤ b for all n ∈ N and some a and b. Set
I1 = [a, b]. Choose a xn1 ∈ I1 . We define |I1 | = b − a. Consider
   
′ ′′ a+b a+b
I1 = I ∪ I = a, ∪ ,b .
2 2
One of I ′ or I ′′ contains infinitely many of the xn . Call it I2 . Then |I2 | = (b − a)/2.
Choose xn2 ∈ I2 with n2 > n1 . Note that I2 ⊃ I1 . This is the base step.
Suppose I1 , . . . , Ik−1 have been chosen so that I1 ⊃ I2 ⊃ I3 ⊃ . . . Ik−1 with
b−a
|Ik−1 | = ,
2k−2
and xn1 ∈ I1 , xn2 ∈ I2 , . . ., xnk−1 ∈ Ik−1 , with n1 < n2 < . . . < nk−1 . By
construction Ik−1 contains infinitely many of the xn . Bisect it as before,
I1 = I ′ ∪ I ′′
with one of the I ′ or I ′′ containing infinitely many of the xn . Call the interval Ik .
Choose xnk ∈ Ik ⊂ Ik−1 with nk > nk−1 . Note that
b−a
|Ik | = .
2k−1
By induction this defines Ik and xnk for all k ∈ N.
The {Ik } satisfy the hypothesis of the Nested-Cell Theorem. Thus their inter-
\∞
section is non empty and there is an x ∈ In . We need to show the subsequence
n=1
{xnk } converges to x as k → ∞. Let ǫ > 0. By Example 3.29 there is a N such
that
b−a

2k
3.6. Cauchy Sequences 27

for all k ≥ N . Since xnk ∈ Ik and x ∈ Ik


b−x
|xnk − x| ≤ |Ik | = <ǫ
2k−1
(see HW 2.4) for all k ≥ N . 

There is a nice geometric interpretation of the Bolzano-Weierstrass Theorem.


Definition 3.42. Let E ⊂ R. A point a ∈ R is called a cluster point or accumu-
lation point of E if E ∩ (a − r, a + r) contains infinitely many points for all r > 0.
That is, a is a cluster point of E if every neighborhood of a contains a point in E
distinct from a.
Example 3.43. If E = (0, 1), the cluster points are [0, 1]. Note that a cluster
point need not be in E.
Example 3.44. If E = Q, the cluster points are all of R, since Q is dense in R.
Example 3.45. The set E = {1, 2, 3} has no cluster points.
Example 3.46. Suppose {xn } is a convergent sequence of real numbers converging
to x ∈ R. Then x is a cluster point of the set {x1 , x2 , . . .}.

We may restate the Bolzano-Weierstrass as follows.


Theorem 3.47. (Bolzano-Weierstrass for Sets) Every bounded infinite subset of
R has an cluster point.

Proof. We will call the infinite set E. We first construct a sequence of distinct
points in E. Since E is non empty, a x1 ∈ E exists. Consider the set E/{x1 }.
It too is non empty. Thus a x2 exists in E/{x1 }. Suppose the distinct points
x1 , x2 , . . . , xk−1 have been similarly chosen. Then the set E/{x1 , x2 , . . . , xk−1 } is
non empty, and there exists a xk ∈ E/{x1 , x2 , . . . , xk−1 }. By induction we have
constructed a sequence {xn } of distinct point in E.
The sequence {xn } is bounded since E is bounded. By the Bolzano-Weierstrass
theorem, a subsequence of {xn }, {xnk } exists and converges to some x ∈ R. Given
any neighborhood of x, the sequence {xn } is eventually in that neighborhood. Thus
every neighborhood of x contains a point in E, and x is a cluster point of E. 

3.6. Cauchy Sequences


There is yet another way to show a sequence has a limit without finding the limit
(the monotone convergence theorem was the other).
Definition 3.48. A sequence {xn } is called Cauchy (pronounced Co She) if for all
ǫ > 0 there exists a N ∈ N such that |xn − xm | < ǫ for all n, m ≥ N .
Theorem 3.49. (Cauchy) A sequence {xn } converges if and only if it is Cauchy.

Proof. Suppose {xn } is converges. Let ǫ > 0. Then there exists N ∈ N such that
|xn − x| < ǫ/2 for all n ≥ N . Then
ǫ ǫ
|xm − xn | = |xm − x + x − xn | ≤ |xm − x| + |xn − x| < + = ǫ
2 2
28 3. Sequences

for all n, m ≥ N . That is, {xn } is Cauchy.


Suppose {xn } is Cauchy. We need a candidate for the limit. We apply the
Bolzano-Weierstrass theorem to find the candidate. To do so we need to show the
sequence is bounded. We copy the proof of Lemma 3.22 for this. Let ǫ = 1 in
the definition of a Cauchy sequence. Then a natural number N exists such that
|xm − xn | < 1 for all m, n ≥ N . That is,
|xm | = |xm − xN + xN | ≤ |xm − xN | + |xN | < 1 + |xN |.
Thus |xm | ≤ max{|x1 |, |x2 |, . . . , |xN −1 |, 1 + |xN |} for all m ∈ N.
The Bolzano-Weierstrass theorem applies and there exists a subsequence {xnk }
converging to some x ∈ R. We need to show the whole sequence converges to x.
Let ǫ > 0. There exists N1 and N2 such that
ǫ
|xm − xn | < ∀m, n ≥ N1
2
ǫ
|xnk − x| <∀k ≥ N2 .
2
Set N = max{N1 , N2 }, and fix K ≥ N . Then nK ≥ N , and
|xn − x| = |xn − xnK + xnK − x|
≤ |xn − xnK | + |xnK − x|
ǫ ǫ
< + = ǫ,
2 2
for all n ≥ N . 

Example 3.50. Consider the sequence given recursively by


xn−2 + xn−1
xn =
2
with x1 = 1 and x2 = 2. One can check by writing out the first few terms that
the sequence is not monotone. Indeed, the sequence oscillates. We could try to
obtain a candidate for the limit by assuming it exists and use the algebra of limits.
If lim xn = L, taking the limit of xn above, we would find L = (L + L)/2, which
n→∞
gives us no useful information. Our only hope is to show the sequence is Cauchy
(if we believe it converges).
Note that

xn−1 + xn 1
|xn − xn+1 | = xn − = 2 |xn − xn−1 |.
2
Homework Exercise 3.29 shows that the sequence {xn } is Cauchy. Moreover,
Cauchy’s theorem shows that the sequence has a limit. We need only find the
limit.
Recall that, if the sequence converges, all subsequences converge to the same
limit (Corollary 3.39). We will find the limit of the subsequence {x2n+1 }. We claim
that
1
x2n − x2n−1 = n−1 .
4
3.6. Cauchy Sequences 29

Indeed, it clearly holds for n = 1. Suppose it hold at level n. Then


1
x2(n+1) − x2(n+1)−1 = x2n+2 − x2n+1 = (x2n + x2n+1 ) − x2n+1
2  
1 1 1
= (x2n − x2n+1 ) = x2n − (x2n−1 + x2n )
2 2 2
1 1
= (x2n − x2n−1 ) = n ,
4 4
and the result follows by induction.
1
We have discovered that x2n = x2n−1 + 4n−1 . Thus
x1 = 1
1 1
x3 = (1 + 2) = 1 +
2 2 
1 1 1 1 11
x5 = (x4 + x3 ) = x3 + 1 + x3 = 1 + +
2 2 4 2 24
 
1 1 1 1 11 1 1
x7 = (x6 + x5 ) = x5 + 2 + x5 = 1 + + +
2 2 4 2 2 4 2 42
 
1 1 1
= 1+ 1+ + 2
2 4 4
 
1 1 1 1
x2n+1 = 1+ 1 + + 2 + · · · + n−1
2 4 4 4
1 1 − 41n
= 1+ .
2 1 − 41
Taking the limit, we find
5
lim xn = lim x2n+1 = .
n→∞ n→∞ 3
30 3. Sequences

3.7. Review
• Techniques to show limit converges
– Apply the definition of a limit
– Algebra of limits
– Show sequence monotone and bounded
– Show sequence Cauchy
• Techniques to show no limit
– Negate the definition of a limit
– Show sequence not bounded
– Show sequence not Cauchy
– Find a contradiction
– More than one convergent subsequence with different limits
• Be able to state (and negate) the definitions of
– Supremum and infimum
– completeness axiom
– lim xn = x.
n→∞
– a sequence and a Cauchy (pronounced Coe She) Sequence
– lim f (x) = L.
x→a
– Continuity at a point and a continuous function.
• Be able to state and prove
– Archimedean principle (any version you want)
– Density of rationals (just the case 0 < x, i.e. prove the existence of r ∈ Q
s.t. 0 < r < x.)
– Monotone convergence Theorem
– Extreme-Value theorem
• Be able to state and apply
– Bolzano-Weierstrass Theorem
– Cauchy’s theorem
– Intermediate-value theorem
• The minimal level of proficiency means you can solve, quickly, problems like
– sup[a, b) = b.
– |a| < ǫ for all ǫ > 0 implies a = 0.
– If {an } and {bn } converge to a and b respectively and an ≤ bn for n ≥ 1.
Prove that a ≤ b.
– 1/n → 0 as n → ∞.
– xn → x as n → ∞. Then ∃M > 0, such that |xn | ≤ M for all n ≥ 1.
– If 0 < xn < 1 and xn converges to x, where must x live?
– Show that n does not converge.
– If {an } and {bn } converge to a and b respectively, then an bn converges
to ab as n → ∞.
– Prove that
n3 − 12n + 1
lim 4 sin(n3 )
n→∞ n + 15n2 + 5
has a limit. Give the details.
– Look over assigned homework problems!
3.7. Review 31

The Completeness Axiom


@
? @
MTC R
@
Archimedean Principle
? A
Nested Cell AU
Denisty of Q
?
Bolzano-Weierstrass
HH
j
H
Cauchy Thm

Summary of theorems so far. Try to understand the idea behind the proof of each theorem.
32 3. Sequences

3.8. Old First Test

1. (20 pts) Provide examples of sequences of real numbers. NO proof is necessary.

(a) An unbounded sequence with a convergent subsequence.


(b) A bounded sequence with no convergent subsequence.
(c) Suppose (xn )n∈N converges and (yn )n∈N is bounded. Show (xn yn )n∈N may
not converge.
(d) An example of a Cauchy sequence.
(e) An monotone sequence converging to one.

2. (20 pts) Answer the following.

(a) State the Bolzano-Weierstrass Theorem.


(b) Suppose E is a nonempty subset of R. What does it mean for E to be bounded
above? AND define the supremum of E.
(c) What does is mean for a sequence not to be Cauchy? That is, negate the
definition of a Cauchy sequence.
(d) State the completeness axiom.

3.
(a)(12 pts) State and prove the Monotone Convergence Theorem for monotone
increasing sequences.

(b) (6 pts) Apply the Monotone Convergence Theorem to show the sequence
2xn + 3
xn+1 = , with x1 = 1 converges.
4
(c) (2 pts) Find the limit.

4.
(a)(15 pts) Suppose a sequence (xn )n∈N satisfies xn > 0 and converges to x ∈ R.
Prove x ≥ 0.

(b) (5 pts) Must it be the case that x > 0? Prove it or give a counterexample.

5. (20 pts) Show that a bounded sequence in R that does not converge has more
than one subsequential limit. In particular, show that a nonconvergent bounded
sequence has two subsequences each with different limits.
3.9. Homework 33

3.9. Homework
Exercise 3.1. Give example of two sequences that do not converge but whose
(a) sum converges
(b) product converges
(c) quotient converges
Exercise 3.2. An example of a convergent sequence {xn } ⊂ (0, 1], but whose limit
is not in (0, 1].
Exercise 3.3. Use the Definition 3.5 to establish the following limits.
1
(a) lim 2 = 0
n→∞ n
2n
(b) lim =2
n→∞ n + 3
2n + 4 2
(c) lim =
n→∞ 5n + 1 5
Exercise 3.4. If a convergent sequence is rearranged, prove that the new sequence
has the same limit.
Exercise 3.5. Let 0 < xn < 4 for all n ∈ N. Show that the sequence {xn } has a
convergent subsequence. In what interval will the limit of this subsequence lie?
Exercise 3.6. Suppose an → a > 0 as n → ∞ and that an > 0 for all n ∈ N.
Show that a m > 0 exists so that an ≥ m for all n ∈ N.
Exercise 3.7. (Cesàro Summable) If the sequence {xn } converges to x, then the
sequence {σn } also converges to x. Here
1
(x1 + x2 + · · · + xn ).
σn =
n
Exercise 3.8. Show that the sequence {sin n} does not converge.
 1
n n odd
Exercise 3.9. Let an = . Does the limit exist? Proof required.
1 n even
Exercise 3.10. Suppose {xn } converges to x and satisfies 2 < xn < 3 for all n ∈ N.
Show that the limit satisfies 2 ≤ x ≤ 3.
Exercise 3.11. Let {an } be a sequence of real numbers converging to a ∈ R.
(a) Prove lim |an | = |a|.
n→∞
(b) If lim |an | = 0, then lim an = 0.
n→∞ n→∞
(c) Show by example that lim |an | may exist, while {an } may not converge.
n→∞
√ √
Exercise 3.12. Let xn ≥ 0 for all n ∈ N with xn → x as n → ∞. Prove xn → x
as n → ∞.
Exercise 3.13. Suppose {xn } is a convergent sequence of integers. Show that the
sequence is eventually constant.
34 3. Sequences

an+1
Exercise 3.14. Let {an } be a sequence of positive real numbers with lim =
n→∞ an
L. If L < 1, show that lim an = 0.
n→∞

Exercise 3.15. Suppose {xn } is a bounded sequence (not necessarily convergent),


and {yn } be a sequence converging to zero. Prove that {xn yn } converges to zero.
Show by example that the result is not true if either hypothesis on {xn } or {yn } is
dropped.
Exercise 3.16. Establish, by using any definition or theorem, the limits of the
following sequences.
   
sin n 2n + 4
(a) (c)
n 5n + 1
( 2 ) √ √
1 (d) n− n+1
(b) π+
n

Exercise 3.17. Prove Theorem 3.18.


Exercise 3.18. Show that every bounded, non empty closed subset of R has a
maximum. (Hint: show the supremum of a set is a boundary point of the set).
Exercise 3.19. What are the cluster points of the irrational numbers?
Exercise 3.20. Prove the monotone convergence theorem for a decreasing, bounded
below sequence.
Exercise 3.21. Suppose 0 < a. Prove that the sequence {a1/n } converges to one.
Exercise 3.22. Show the following recursively defined sequences are convergent
and find the limit
(a) Let x1 = 1 and xn = 14 (2xn−1 + 3).
(b) Let x1 = 3 and xn = 2 − 1/xn .
√ √
(c) Let x1 = 2 and xn = 2 + xn−1 .
Exercise 3.23. Give an example of a sequence whose set of convergent subse-
quences converge to {1, 1/2, 1/3, 1/4 . . .} ∪ {0}.
Exercise 3.24. Give an example of a sequence with no convergent subsequences.
Exercise 3.25. Give an example of two sequences {xn } and {yn } satisfy {xn } ⊂
{yn } as sets, but with {xn } not a subsequence of {yn }.
Exercise 3.26. The next two statements are equivalent - they are contrapositives.
Prove one of them.
(a) Suppose {xn } is a bounded sequence. If all its convergent subsequences have
the same limit, then the sequence is convergent.
(b) Show that a nonconvergent bounded sequence has two convergent subse-
quences each with distinct limits.
Exercise 3.27. Suppose {xn } is unbounded. Show there exists a subsequence
{xnk } such that 1/xnk → 0 as k → ∞.
3.9. Homework 35

Exercise 3.28. An example of a Cauchy sequence with {xn } ⊂ (0, 1], but whose
limit is not in (0, 1].
Exercise 3.29. Let {xn } be a sequence and 0 < r < 1. Suppose that
|xn+1 − xn | ≤ r|xn − xn−1 |
for n ≥ 2. Show that the sequence is Cauchy.
Exercise 3.30. Use Exercise 3.29 to show the sequence xn = 1/(2 + xn−1 ) for
n ≥ 2 and x1 > 0 is Cauchy. Find the limit.
Exercise 3.31. (a) Negate the definition of a Cauchy sequence.
(b) Show that the sequence
n
X 1
xn =
k
k=1
is not Cauchy.
(c) Show that |xn+1 − xn | → 0 as n → ∞ (compare this to the definition of a
Cauchy sequence).

Potrebbero piacerti anche