Math 320, Real Analysis I

Solutions to Homework 2 Problems

Exercise 2.2.1. Verify, using the definition of convergence of a sequence, that the following sequences converge
to the proposed limit.
(a) lim

1
= 0.
6n2 + 1



1
Proof. Let > 0 be arbitrary. Set = min ,
, so 0 < < 1 and . Then
2
r
1
1
is a positive real number, so
is as well. Therefore, by the Archimedean
6
6
r
1
Property, there is a natural number N N such that
< N . For this N , we have
6


 1

1
1
1
1
1


=
=
=
 6n2 + 1 0 = 6n2 + 1 6N 2 + 1 <  1 
1
1/
+
6
+1

6
whenever n N . Therefore, as > 0 is arbitrary, we conclude that lim

(b) lim

1
= 0.
+1

6n2

3n + 1
3
= .
2n + 5
2



1 13
Proof. Let > 0 be arbitrary. Then
10 is a real number, so there is a natural
 4 
1 13
10 by the Archimedean Property. For this N ,
number N N such that N >
4
we have

 
 

 3n + 1 3   6n + 2
6n + 15   13 
13
13
13




=
=
 2n + 5 2  =  4n + 10 4n + 10  =  4n + 10  4N + 10 <  1  13
13/
4
10 + 10
4
whenever n N . Therefore, as > 0 is arbitrary, we conclude that lim

(c) lim

3n + 1
3
= .
2n + 5
2

2
= 0.
n+3

4
Proof. Let > 0 be arbitrary. Then 2 is a real number, so there is a natural number

4
N N such that N > 2 by the Archimedean Property. For this N , we have



 2

2
2
2
2

= 2

< < s  =
=
 n+3

2/
n+3
N +3
N
4
2
whenever n N . Therefore, as > 0 is arbitrary, we conclude that lim

2
= 0.
n+3

Math 320, Real Analysis I

Solutions to Homework 2 Problems

Exercise 2.3.7. (a) Let (an ) be a bounded (not necessarily convergent) sequence, and assume lim bn = 0.
Show that lim(an bn ) = 0.
Proof. Let (an ) and (bn ) be sequences of real numbers such that (an ) is bounded and
(bn ) 0. We claim that the sequence (an bn ) also converges to zero. Thus, we begin
by letting > 0 be arbitrary. As (an ) is bounded, there is a real number M > 0 such
that |an | M for all n N by the definition of a bounded sequence. Since both > 0

and M > 0, we have

> 0 as well. We now invoke the definition of (bn ) 0, for our
M

positive value
above, to obtain a natural number N N such that |bn 0| <
for
M
M
all n N . Now consider
h i
|an bn 0| = |an bn | = |an | |bn | < [M ]
=
M
for every n N . Hence, as > 0 is arbitrary, we have (an bn ) 0 as claimed.
Question: Why are we not allowed to use the Algebraic Limit Theorem to prove this?
Answer: We are not allowed to use the Algebraic Limit Theorem in this proof since
this Theorem assumes both sequences (an ) and (bn ) converge as part of its hypothesis.
Since we do NOT assume the sequence (an ) is convergent, but only that it is bounded,
we have not satisfied the hypotheses of the Algebraic Limit Theorem, so it cannot be
used.
(b) Can we conclude anything about the convergence of (an bn ) if we assume that (bn ) converges to some nonzero limit b?
Answer: No!
To be a touch more precise, in the event that (bn ) converges to some nonzero limit b, we
may conclude that
the sequence (an bn ) converges

if and only if

the sequence (an ) converges.

For, if (an ) converges, the Algebraic Limit Theorem implies that the product (an bn )
of these convergent sequences is also convergent. On the other hand, if we know that
(an bn ) converges
to c, again it is the Algebraic Limit Theorem that implies the quo
an bn
tient
= (an ) of these convergent sequences also converges to c/b, since we are
bn
assuming that b = lim bn is not zero.
Therefore, only knowing that (an ) is a bounded sequence, we cannot conclude that (an bn )
converges without also concluding that the sequence (an ) converges, but we only assume
it is bounded and not necessarily convergent.
(c) Use Part (a) to prove Theorem 2.3.3, Part (iii), for the case when a = 0.
Proof. Let (an ) and (bn ) be convergent sequences and assume (an ) a = 0 while
(bn ) b. We are asked to prove that (an bn ) 0 b = 0 in this case. As (bn ) is assumed
to converge, we know that (bn ) is a bounded sequence by Theorem 2.3.2. However, by
Part (a) of this Exercise, we have shown that whenever (xn ) is a bounded (not necessarily
convergent) sequence and (yn ) 0, the product (xn yn ) 0 as well. Therefore, since
(bn ) is a bounded sequence and (an ) 0, their product (an bn ) 0 as well. Hence
lim(an bn ) = ab even in the case when a = 0.

Math 320, Real Analysis I

Solutions to Homework 2 Problems

Exercise 2.3.12. Consider the doubly indexed array am,n =

(a1 ) Intuitively speaking, what should

m
.
m+n

lim am,n represent?

m,n

Solution: First of all, if it exists, lim am,n must be a real number. Intuitively speaking,
we would say that lim am,n = L if the numbers am,n get arbitrarily close to L as
m,n

m and n both get large. If I wished to be just a little more precise, I could say that
lim am,n = L if, for whatever positive error you provide that you say is small enough
m,n

to convince you, I can ensure all of the numbers am,n are within this distance from L so
long as I take both m and n to be sufficiently large.
(a2 ) Compute the iterated limits
lim lim am,n

n m

and

lim lim am,n .

m n

Solution: The first is




lim
lim am,n = lim
lim
n

m
m m + n


= lim ([1]) = [1].
n

However, the second iterated limit is

lim

lim am,n = lim

m
lim
n m + n


= lim ([0]) = [0].
m

As these are not equal to one another, we cannot use iterated limits to define the limit
of a doubly indexed sequence. (This limit is path-dependent.)
(b) Formulate a rigorous definition in the style of Definition 2.2.3 for the statement
lim am,n = L.

m,n

Definition: A doubly indexed array (am,n ) is said to converge to the real number L
if, for every positive real number > 0, there is a natural number N N such that
whenever both m N and n N it is true that
|am,n L| < .
In this case, we call the array (am,n ) convergent and refer to L as its limit, and we
write
lim am,n = L.
m,n

Exercise 2.4.5. Calculating Square Roots: Let x1 = 2, and define



1
2
xn+1 =
xn +
.
2
xn
(a1 ) Show that x2n is always greater than 2. That is, prove that x2n > 2 for all n N.
Proof. When asked to prove a result for all n N, the mathematical tool that should
come to mind first is the Principle of Mathematical Induction (PMI). This is as important

Math 320, Real Analysis I

Solutions to Homework 2 Problems

to our understanding of the natural numbers, N, as the Axiom of Completeness is to our

understanding of the set of real numbers, R. Now on to the proof.
We begin with the base case, in which we must establish the validity of the claim
x2n > 2 for n = 1. Now, when n = 1, we have x1 = 2, so x21 = 22 = 4, which is strictly
larger than 2. Hence the claim is true in our base case when n = 1.
Next, after verifying the base case, we move on to the inductive step of our proof
by induction. In this step, we assume that the claim is true when n = k, for some value
of k 1, and, using this, prove that it must then also be true when n = k + 1. By
doing so, we will have established a causal relationship between the current term and
the subsequent one. With this relationship in place coupled together with our knowledge
that the claim is true when n = 1, it follows that it must be true for all n, just as we
can deduce that every domino in a row will fall so long as we know theyre close enough
together and that the first one is knocked over. So lets return to the proof.
Lets now assume that, for some integer k 1, we have x2k > 2. Now
 xk+1 and,
 consider
1
2
in particular, its square. By definition of our sequence, xk+1 =
, so
xk +
2
xk
x2k+1

1
=
4

2
xk +
xk

2

1
=
4

x2k

4
+4+ 2
xk


=

x2k
1
+1+ 2
4
xk

We claim that, based on our assumption that x2k > 2, it follows that x2k+1 > 2 as well.
Yet x2k+1 > 2 is true if and only if x2k+1 2 > 0, so consider
x2k+1 2







2
x2k
x2k
1 2
1
1
1
4
2 2 1 x2k 2
=
+ 1 + 2 2 =
1+ 2 =
x 4+ 2 =
xk
=
>0
4
4
4 k
4
xk
4
xk
xk
xk
xk


since x2k > 2 implies this fraction is not zero. Thus we have shown that x2k > 2 implies
that x2k+1 > 2 as well, so we have completed the inductive step. Therefore, by the
Principle of Mathematical Induction, we conclude that the statement x2n > 2 is true for
all n N.
(a2 ) Use Part (a1 ) to prove that xn xn+1 0.
Proof. Let n N be arbitrary. Then consider


1
2
xn
1
xn
1
x2 2
xn xn+1 = xn
xn +
= xn

= n
,
2
xn
2
xn
2
xn
2xn
which is strictly greater than 0 since, as we have shown above in Part (a1 ), x2n > 2.
Therefore, xn xn+1 0 as claimed.
(a3 ) Conclude that lim xn =

2.

Proof. By Part (a2 ), for all n xn xn+1 . Hence (xn ) is a decreasing sequence, and is
therefore monotone. Furthermore, since it is decreasing, xn x1 = 2 for all n,so the
sequence (xn ) is bounded above by 2. We also note that
it is bounded below by 2, for
all of the xn s are positive and x2n > 2 implies xn > 2 for all n. Thus the sequence
(xn ) is monotone and bounded, so it converges by the Monotone Convergence Theorem.
Let L = lim(xn ). As in the proof of the Monotone Convergence Theorem, we see
that L = lim(xn ) is equal to the infimum of the set of terms of the sequence, L =

Math 320, Real Analysis I

Solutions to Homework 2 Problems

inf{x1 , x2 , x
2 is a lower bound forthis
3 , x4 , . . . , xn , . . . }. We have already shown that
set, so L 2. We now show that it is impossible for L to be strictly larger than 2.

Suppose, for sake of contradiction, that L > 2. Then L2 > 2, so there is a natural
1
L2 2
number N such that
<
by the Archimedean Property. This implies that
N
L
L
1
2+
< L2 . Setting = , by Lemma 1.3.7s infimum version, there must be a term
N
N
1
xk of our sequence such that L xk < L + . Now consider the next term,
N

xk+1

1
=
2

2
xk +
xk

1
L
L+
L2 +
+2
xk
1
1
L2 + (L2 )
N + =
N
=
+
<
<
=L
2
xk
2
L
2L
2L

so that xk+1 < L, which contradicts that L is a lower bound for

the set of terms,
{xn : n N}, of the sequence (xn ).Thus, our assumption that L > 2 must be false,
so we conclude that lim(xn ) = L = 2 as desired.
(b) Let c 0 be a real number. Modify the sequence (xn ) so that it converges to
Solution: Define the sequence xn recursively by x1 = c and


1
c
xn+1 =
xn +
2
xn

c.

for all n 1.
Claim: The sequence (xn ) converges to

c.

Proof. First of all, if c = 0, then x1 = 0 is already the square root of 0, and we claim
that xn = 0 for all n, in which case it is obvious that (xn ) 0. By design, x1 = 0, so
our base case is valid. Now suppose that xn = 0 and consider


1
0
1
1
xn+1 =
xn +
= xn = [0] = 0.
2
xn
2
2

Therefore, xn = 0 for every n, and (xn ) 0 = 0 as needed.

Now suppose 0 < c < 1. Then we claim that (xn ) is an increasing, bounded sequence,
and hence is convergent. To see that it is increasing, we note that x2n < c is always true
by induction (x1 = c and 0 < c < 1 implies x21 = c2 < c establishing our base case,
x2 c
and if x2n < c then x2n+1 c = n
< 0 implying that x2n+1 < c, too). Therefore,
2xn
c
x2 c
xn
xn xn+1 =

= n
< 0, which implies xn xn+1 for all n. Hence (xn ) is
2
2xn
2xn

an increasing sequence which is bounded below by 0 and above by c, so it converges.





1
[1]
1
1
xn +
=
[1] +
=1
If c = 1, then x1 = c = 1 and if xn = 1, then xn+1 =
2
xn
2
[1]

also, so every xn = 1 and clearly (xn ) 1 = 1.

Finally, if c > 1, then (xn ) is decreasing and bounded above by c and below by c.
Thus, regardless of the value of c > 0, we find that our sequence (xn ) converges. Call its
limit x. Then the values of xn+1 approach x as n , so the limit of the left hand side

Math 320, Real Analysis I

Solutions to Homework 2 Problems


c
xn +
, is x. Likewise, both of the xn s on the
of the recursion relation, xn+1
xn
1
c
right hand side approach x, so the right hand side is approaching
x+
. Therefore,
2
x
c
c
c
1
x+
= 2x = x +
= x =
= x2 = c
x=
2
x
x
x

which implies that x = c as claimed.

1
=
2

Exercise 2.4.6. Limit Superior: Let (an ) be a bounded sequence.

(a) Prove that the sequence (yn ), whose n-th term is defined by
yn = sup{ak : k n},
converges.
Proof. For each n N, define the set
An = {ak : k n} = {an , an+1 , an+2 , . . . }.
Then these sets form a nested sequence A1 A2 A3 An An+1 .
Therefore, for each n, since An+1 An , we have yn+1 = sup An+1 sup An = yn by
Exercise 1.3.4 from last weeks assignment. Therefore, (yn ) is a decreasing sequence,
and hence is monotone. Furthermore, as we are assuming that the sequence (an ) is
bounded, we claim that the sequence (yn ) is likewise bounded. For, if M > 0 is a real
number such that M an M for all n N, then M is an upper bound for the set
A1 , and thus for all of its subsets An . Therefore, by definition of supremum, it follows
that yn = sup An M for all n. Also, by Lemma 1.3.7, for each n there is a k n
1
(i.e., an ak An ) for which yn < ak yn . Since M am for all m, it follows
n
that M ak yn , so each yn M . Hence the sequence (yn ) is bounded below as
well as above, so it is bounded. Therefore, by the Monotone Convergence Theorem, the
sequence (yn ) converges, for it is both monotone (decreasing) and bounded.
(b1 ) The limit superior of (an ), or lim sup an , is defined by
lim sup an = lim yn ,
where (yn ) is the sequence from Part (a) of this exercise.
Provide a reasonable definition for the limit inferior, lim inf an , of the sequence (an ).
Definition: The limit inferior of (an ), denoted lim inf an , is defined by
lim inf an = lim zn ,
where (zn ) is the sequence whose n-th term is given by
zn = inf{ak : k n} = inf An .

Math 320, Real Analysis I

Solutions to Homework 2 Problems

(b2 ) Briefly explain why lim inf an always exists for any bounded sequence.
Explanation: Since (an ) is a bounded sequence, it follows as before that the sequence
of infima, (zn ), is also bounded. Additionally, since An+1 An for each n, we must have
zn+1 = inf An+1 inf An = zn , so that (zn ) is also an increasing sequence. Therefore,
by the Monotone Convergence Theorem, (zn ) is convergent, so lim inf an = lim zn exists.
(c1 ) Prove that lim inf an lim sup an for every bounded sequence.
Proof. For each n, zn = inf An is a lower bound and yn = sup An is an upper bound for
the set An = {an , an+1 , an+2 , . . . } by definition of infimum and supremum. In particular,
by definition of upper and lower bounds, we must have
zn an yn ,
for each n, since an An . Thus, for all n, zn yn . Hence, by the Order Limit Theorem,
Part (ii), lim zn lim yn , so lim inf an = lim zn lim yn = lim sup an .
(c2 ) Give an example of a sequence (an ) for which the inequality lim inf an < lim sup an is
strict.
Example: Consider the bounded sequence (an ) whose n-th term is given by an = (1)n .
That is, (an ) is the sequence (1, 1, 1, 1, 1, 1, 1, 1, . . . ) which continues to alternate
between the values 1 and 1. Then, for each n, the set An contains both the numbers
1 and 1, so zn = inf An = 1 while yn = sup An = 1 for every n. Therefore,
lim inf(1)n = lim(1) = 1

and

lim sup(1)n = lim(+1) = +1

so that in this case, we have lim inf an < lim sup an .

(d) Show that lim inf an = lim sup an if and only if lim an exists.
Note: In this case, all three share the same value.
Proof. Let (an ) be a bounded sequence, so that both the limit superior and the limit
inferior exist.
(=): Assume lim inf an = lim sup an = `. As we have already shown in Part (c1 ), we
have
zn an yn
for all n N. Therefore, by the Squeeze Theorem (Exercise 2.3.3), lim an = ` also, since
lim zn = lim inf an and lim yn = lim sup an are equal. Therefore, lim an exists.
(=): Suppose lim an = a exists. Let > 0 be arbitrary. Then, by the topological
definition of convergence, there is a natural number N N such that
an V (a) = (a , a + )
whenever n N . In particular, this implies that the set AN = {ak : k N } =
{aN , aN +1 , aN +2 , . . . } is a subset of the -neighborhood V (a) of a. Therefore, as the
sets An are nested, it follows that for every n N , the set An V (a), so that
a = inf V (a) inf An = zn yn = sup An sup V (a) = a +
for every n N . From this, we conclude that for all n N , zn V (a) and yn V (a),
so (zn ) a and (yn ) a by the topological definition of convergence, since our > 0
is arbitrary. Hence, lim inf an = lim zn = a = lim yn = lim sup an , so lim inf an =
lim sup an .