Understanding Analysis (Abbott) Solutions

Manual

Trajan and Varun

Contents

1 Solutions 2

2 Solutions 3

1
Chapter 1

Solutions

We found this chapter boring...

2
Chapter 2

Solutions

2.2 Solutions
2.2.1
Definition: A sequence (xn ) verconges to x if there exists an  > 0 such that for
all N ∈ N it is true that n ≥ N implies |xn −x| < . An example of a vercongent
sequence is the sequence an = 1/2n which verconges to 0. To prove this, pick
 = 1. Then for every N ∈ N, |xn − 0| < 1, since the sequence is bounded above
by 1 and below by 0.
An example of a vercongent sequence that diverges is the harmonic sequence.
Let an = 1/(n + 1). We claim an verconges to 0. Let  = 1. Then

|xn − x| = |1/(n + 1) − 0| < 1

for all n ∈ N.
A sequence can verconge to two different values. Consider the above sequence.
It verconges to 1 and 0 since

|xn − 1| < 1 and |xn − 0| < 1

for all n ∈ N.
What is really being described here is a bounded sequence. These sequences are
bounded, so you can declare the sequence to be vercongent to any x, so long as
|xn − x| is less than the bound.

2.2.2 a) Let  > 0. We wish to find N in terms of  such that ∀n ≥ N ,

2n+1
| 5n+4 − 25 | < 

3−20
Let N > 25 . Then
3 − 20
n>
25

3
CHAPTER 2. SOLUTIONS 4

→ 3 < 25n + 20

3
→ <
25n + 20
 
 −3  3
Now notice that  25n+20 = 25n+20 .
 
 −3 
 25n + 20  < 
 

2n+1 2 −3
Now, notice that 5n+4 − 5 = 25n+20 , so
 
 2n + 1 2 
 −
 5n + 4 5  < 


.
Thus, by the definition of convergence of a sequence, the sequence converges
to 52 . 

 2 
2
b) Let N > . We wish to show that for n ≥ N that  n2n
3 +3  < . Now
 
notice that

2n2 2n2 2
3
< 3
=
n +3 n n
Thus we have that
 2n2   2 
   
2
 n3 + 3  <  n  = n
   

However since n ≥ N > 2 , we have that 2

n < . Thus we have shown that
the sequence converges to 0.

1
c) Let  be an arbitrary positive number. Choose N such that N > 3 . Let
n ≥ N . Then
1 1
n > 3 =⇒ √ <
 3
n
Note that
sin(n2 ) 1
√ ≤ √ <
3
n 3
n
2
So it follows that |an − 0| <  and thus lim sin(n
√3 n
)
=0

2.2.3 a) Show there exists a college where every student is shorter than seven
feet tall.

b) Show that there exists a college where every professor gives at least one
student a grade lower than a B.
CHAPTER 2. SOLUTIONS 5

c) Show that at every college, there exists a student shorter than six feet tall.

2.2.4 a) Consider the divergent sequence (1, −1, 1, −1, 1, −1, ...). It cannot
converge because the limit definition is not satisfied for  < 2.

b) This is impossible. Assume that there does exist such a sequence an with
an infinite number of ones and that it converges to a limit, L 6= 1. Then ∀ > 0
there exists N such that ∀n ≥ N , |an − L| < .
Let  = |1−L|
2 . Now since there are an infinite number of ones in the sequence,
there exists a term an where n ≥ N such that an = 1. Since the sequence
converges, this means that |1 − L| < |1−L|
2 . However, this is clearly false and
we arrive at a contradiction.

(Remark: Note that a) and b) imply that if a sequence has an infinite number
of ones, it either converges to 1 or diverges).

c) Consider the sequence (0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, ...). This sequence is

constructed by having a block of ones ”sandwiched” by zeros on either side
where the number on ones in each block progressively increases. This is diver-
gent with divergence failing for  < 1.

2.2.5 a) an = [[5/n]] where [[x]] denotes the greatest integer less than or equal
to x. We claim that lim an = 0. Let  > 0 be arbitrary. Let N = 5. Let n ≥ N .
Then, n > 5 implies [[5/n]] = 0, and hence

|an − 0| = |0 − 0| < 

for n > 5 so lim an = 0.

b) an = [[(12 + 4n)/3n]]. Note we can rewrite this as [[12/3n + 4n/3n]] =

[[4/n + 4/3]]. We claim lim an = 1. Let  > 0 be arbitrary. Take N = 24. Let
n ≥ N . Then n > 24 implies

an = [[4/n + 4/3]] < [[1/6 + 4/3]] = 2

Since an only takes integer values, for n > 24, an = 1, and thus

|an − 1| = 0 < 

and so lim an = 1.
2.2.6
Theorem 2.2.7 (Uniqueness of Limits): The limit of a sequence, when it exists,
must be unique.

Proof: Let  > 0. Assume that a 6= b then there exists N1 such that ∀n ≥ N1 ,
|an − a| < . Also since an converges to b, there exists N2 such that ∀n ≥ N2 ,
|an − b| < .
CHAPTER 2. SOLUTIONS 6

|a−b|
Let N = max(N1 , N2 ), n ≥ N , and  = 2 . Then |a − b| = |a − an + an − b|.
By the Triangle Inequality,

|a − an + an − b| ≤ |a − an | + |an + b| < 2 = |a − b|
However, |a − b| < |a − b| is a contradiction and thus a = b, showing that
limits of sequences are unique.

2.2.7 a) The sequence (−1)n is frequently in the set {1}.

b) Eventually is the stronger definition. Eventually implies frequently.
c) A sequence (an ) converges to a if, given any -neighborhood V (a) of a, the
sequence (an ) is eventually in V (a).
d) A sequence with an infinite number of terms equal to 2, is not necessarily
eventually in the interval (1.9,2.1), but it is frequently in the interval (1.9,2.1).

2.2.8 a) It is zero heavy with M = 2

b) It must contain infinite zeros. Assume that a zero heavy sequence contains
only a finite number of zeros. Then there exists m such that for all n ≥ m we
have that xn 6= 0. Let N in the definition of zero heavy be greater than m, then
we see that the definition cannot be satisfied because there are no more zero
terms. Thus we arrive at a contradiction.
c) Consider the sequence (0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, ...) as seen before in
2.2.4 c). Notice that the spacing of the zeros gets larger and larger and so there
exists no M that satisfied the definition for all N ∈ N.

2.3 Solutions
2.3.1 a) We are given (xn ) → 0, so we can make |xn − 0| as small (or large) as
√
we want. In particular, we choose N such that |xn | < | xn |, whenever n ≥ N .
To see that this N indeed works, observe that for all n ≥ N ,

√ |xn |  √
| xn | = √ < √ | xn | = 
| xn | | xn |
√
so ( xn ) → 0.

b) We are given (xn ) → x, so we can make |xn − x| as small or large as we

want. We choose N such that
√ √
|xn − x| < | xn + x|

whenever n ≥ N . To see that this N works, notice that for all n ≥ N ,

√ √ |xn − x|  √ √
| xn − x| = √ √ < √ √ | xn + x| = 
| xn + x| | xn + x|
CHAPTER 2. SOLUTIONS 7

√ √
Therefore, ( xn ) → x.

2.3.3 Squeeze Theorem. Show that if xn ≤ yn ≤ zn for all n ∈ N, and if

lim xn = lim zn = l, then lim yn = l.
Proof:We know that lim xn = lim zn = l. So there exists N1 and N2 such
that for n ≥ N1 ,
|xn − l| < 
and for n ≥ N2 ,
|zn − l| < 
Let N = max {N1 , N2 }. Since xn ≤ yn ≤ zn , xn − l ≤ yn − l ≤ zn − l. Then let
n ≥ N . We have
− < xn − l ≤ yn − l ≤ zn − l < 
− < yn − l < 
and so (yn ) → l as desired.

2.3.2 a) Since xn is converges to 2, we know that there exists N such that

∀n ≥ N we have that |xn − 2| < 32 . In a moment, we will see the reasoning for
the 32 coefficient.

   
 2xn − 1   2xn − 4  2
 − 1=  = |xn − 2| < 2 · 3  = 
 3   3  3 3 2
Thus we have proven the limit.
2.3.2 b) Once again, we use the fact that xn converges. We know there exists
N1 such that ∀n ≥ N1 , |xn − 2| < 2.
 
 1 1  |2 − xn |
 xn − 2  = |2xn |


Now, we know that the |2 − xn | in the numerator can be made arbitrarily

small because of the convergence of xn . But to make sure the denominator is
not unbounded, we wish to show a lower bound on the quantity |xn |.

Since xn is convergent, we know there exists N2 such that |xn − 2| < 1. Notice
that

2 = |xn + 2 − xn | ≤ |xn | + |xn − 2|

by the Triangle Inequality. Therefore 2 − |xn − 2| ≤ xn . Furthermore using
the previous inequality that |xn −2| < 1, 2−|xn −2| > 1. Putting these together,
1 < 2 − |xn − 2| ≤ |xn |, which gives us the bound that |xn | > 1. Therefore,
|2xn | > 2. From this, we can see that for n ≥ max(N1 , N2 ),

|2 − xn | 1 1
< |2 − xn | < · 2 = 
|2xn | 2 2
CHAPTER 2. SOLUTIONS 8

Thus we have proven the limit.

2.3.4 a) To use the Algebraic Limit Theorem we must first ensure that the
limit of the denominator is nonzero.

lim 1 + 3an − 4a2n = 1 + 3(0) − 4(0)(0) = 1 6= 0

Moving on to the numerator, we see that

lim 1 + 2an = 1 + 2(0) = 1

Puting these two pieces together,
1 + 2an 1
lim = =1
1 + 3an − 4a2n 1

b) Notice that

(an + 2)2 − 4 a2 + 4an

= n = an + 4
an an
Now by the Algebraic Limit Theorem, we have

lim an + 4 = 4

c) Notice that
2 2+3an
an +3 an 2 + 3an
1 = 1+5an =
an +5 an
1 + 5an
So now our limit is equivalent to evaluating
2 + 3an
lim
1 + 5an
We can use the Algebraic Limit Theorem since lim 1+5an = 1+5(0) = 1 6= 0
and so

2 + 3an 2 + 3(0) 2
lim = = =2
1 + 5an 1 + 5(0) 1

2.3.6 We shall first write the expression in a more convenient form by mul-
tiplying by the conjugate.

√ √
p (n − n2 + 2n)(n + n2 + 2n) −2n −2n
n− n2 + 2n = √ = √ =  √ 
2
n + n + 2n 2
n + n + 2n 2
n 1 + n n+2n
CHAPTER 2. SOLUTIONS 9

−2
= √
n2 +2n
1+ n
Now, we can invoke the Algebraic Limit Theorem, as long as the denominator
is non-zero. We shall calculate that limit.
√ r r
n2 + 2n n2 + 2n 2
lim 1 + = lim 1 + lim = 1 + lim 1 +
n n2 n
Now using exercise 2.3.1 and that n1 → 0, we have that

s
√
r  
2 2
lim 1 + = lim 1 + = 1=1
n n
√
n2 +2n
and so lim 1 + n = 2. Therefore,
p −2 −2
lim(n − n2 + 2n) = lim √ = = −1
n2 +2n 2
1+ n
.

2.3.8 a) We can write p as

l
X
p(t) = ak tk
k=0
.
Thus,
l
X
lim p(xn ) = lim ak xkn
k=0

From here on, we will be invoking the Algebraic Limit Theorem without
explicitly mentioning it
l
X l
X l
X l
X
lim ak xkn = lim ak xkn = ak lim xkn = ak xk = p(x)
k=0 k=0 k=0 k=0

Thus we have shown the desired limit.

2.3.8 b) Consider the sequence xn = n1 that converges to 0 and the piecewise
function
(
1 if x = 0
f (t) =
0 if x 6= 0
The sequence {f (xn )} = (0, 0, 0, ...) which converges to 0. However, f (0) =
1.
CHAPTER 2. SOLUTIONS 10

2.3.10 a) Consider the sequences an = (−1)n and bn = (−1)( n + 1). No-

tice that lim an − bn = lim 0 = 0. However, neither an nor bn converges.
b) Let  > 0. Since bn → b, there exists N such that ∀n ≥ N , we have that
|bn − b| < . By the Reverse Triangle inequality, we have that

||bn | − |b|| ≤ |bn − b| < 

Thus by the definition of a limit, |bn | → |b|.
c) It seems like textbooks meant to say an → a NOT an → 0. The following
solution will assume this correction.
Let  > 0. Since an → a, we know there exists N1 such that |an − a| < 2 .
Similarly, there exists N2 such that |bn − an | < 2 . Now, let N = max(N1 , N2 )
and n ≥ N , then

 
|bn − a| = |bn − an + an − a| ≤ |bn − an | + |an − a| < + =
2 2
Thus, we have proven the limit.
d) Let  > 0. Since an → 0, there exists N such that ∀n ≥ N , we have that
|an | < . Let n ≥ N , then

|bn − b| ≤ an ≤ |an | < 

Thus, we have proven the limit.

2.3.12 a) This problem seems to already assume that B is a nonempty subset

of R. Thus since B has an upper bound, sup B exists. Let b = sup B.
Assume that a is not an upper bound of B, so a < b. Since {an } → a, we know
there exists N such that ∀n ≥ N , we can have that |an − a| < |b−a| 2 by the
definition of a limit. However, this implies that there is am where m ≥ N that
is less than b. This contradicts that every an is an upper bound. Thus, our
assumption that a is not an upper bound is incorrect and by contradiction we
have that a must be an upper bound.
b) Assume that a ∈ (0, 1). Let γ = min(a, 1 − a). By definition of a limit, there
exists N such that ∀n ≥ N , we have that |an − a| < γ2 . However, this implies
there exists am with m ≥ N such that am ∈ (0, 1). This is a contradiction.
Thus, a ∈ / (0,
p1).
c) Let a = (2). We will proceed to construct a sequence implicitly that con-
verges to a.
Let n = n1 . Because the rationals are dense over R, there exists r such that
√ √
2 − n < r < 2. For each n, choose a corresponding rational number an .
This process will produce
√ a sequence {an }. We must now show that this se-
quence converges to 2. Let  > 0. First, notice that by construction,
√ 1
|an − 2| < n =
n
CHAPTER 2. SOLUTIONS 11
p
Let n > 1 , then n1 < . Thus, we have that |an − (2)| <  and by defini-
p
tion, {an } converges to (2). Notice that every an was rational but the limit
is irrational.

2.4 Solutions
2.4.2 a) The problem is, we have yet to establish that the sequence even con-
verges. Using y1 to then recursively find further values in the sequence, we see
that y1 = 1, y2 = 2, y3 = 1. Notice that y3 = y1 . Thus, the sequence will
simply alternate between the values of 1 and 2.

Remark: For further information on when a recursive system such as the one
above has cyclic behavior and so on, look into discrete dynamical systems.

2.4.2 b) To see if the strategy will even work, we must first establish that
the sequence converges. Only then will the assumption that lim yn = lim yn+1
actually hold. To do this, we shall invoke the Monotone Convergence Theorem
by first showing that the sequence is monotone increasing and bounded above.
First, we shall show that {yn } is monotone increasing by induction. Note that
y1 = 1 and y2 = 3 − 11 = 2. It is obvious that y1 ≤ y2 . This is our base case.
For the induction step, we wish to show that yn+1 ≤ yn+2 . We shall start from
the induction hypothesis, yn ≤ yn+1 .

yn ≤ yn+1
−yn ≥ −yn+1
1 1
− ≤−
yn yn+1
1 1
3− ≤3− → yn+1 ≤ yn+2
yn yn+1
Thus by induction, we see that the sequence is monotone increasing.
Now, we will show that yn ≥ 1 which will later be used to show that yn ≤ 3.
Again, we shall use induction. First, we see that y1 = 1 ≥ 1. Next, notice that

yn ≥1
1
≤1
yn
1
− ≥ −1
yn
1
3− ≥ 2 ≥ 1 → yn+1 ≥ 1
yn
CHAPTER 2. SOLUTIONS 12

Thus, by induction, we have that yn ≥ 1. Now to show that yn ≤ 3. We

have that y1 ≤ 3. Notice that

yn ≥ 1
−yn ≥ −1
1
− ≤ −1
yn
1
3− ≤ 2 ≤ 3 → yn+1 ≤ 3
yn
By induction, we have that yn ≤ 3. Thus by the Monotone Convergence
Theorem, we have that yn converges. Since we have shown that the sequence
converges to some limit, y, we can apply the strategy and we get the equation
√
y = 3 − y1 , which when solved using the quadratic formula gives us y = 3±2 5 .
Since all √
the terms are between 1 and 3, we see that the only reasonable value
is y = 3+2 5 .

2.4.4 a) In the following paragraph, Archimadean Property will be used to refer

to the first part of the Archimadean Property

Assume the Archimedean Property does not hold. Let x ∈ R and x > 0. Con-
sider the sequence an = nx. By our assumption that the Archimadean Property
does not hold, we have that for n ∈ B, n ≤ x. This implies that nx ≤ x2 and
therefore our sequence is bounded. Furthermore, an+1 − an = (n + 1)x − nx =
x > 0 and so the sequence is monotone increasing. Therefore by the Monotone
Convergence Theorem, we have that the sequence converges to some limit, L.
Therefore, there exists N such that ∀n ≥ N , we have that |nx − L| < x. Now
since the sequence is monotone increasing, we must have that L ≥ nx ∀n ∈ N.
Therefore, |nx − L| = L − nx < x → L < (n + 1)x. However, this contradicts
the previous statement that L ≥ nx ∀n ∈ N. Therefore, our assumption that
the Archimadean Property does not hold is false.

The second part of the Archimadean Property follows by letting x ∈ R. We

have just shown that the first part of the Archimadean Property holds, so there
exists n such that n > x1 which can then be rearranged to give x > n1 , as desired.
b) So we have that In = [an , bn ] and In ⊆ In+1 . Let us look at the sequence
of an , the lower bounds of each interval. Now notice that each bn serves as
an upper bound and also that an ≤ an+1 , thus by the Monotone Convergence
Theorem, we have that the sequence of an converges to some limit, a. Now
notice that since the an are monotone increasing, we have that an ≤ a and also
since each bn is an upper bound, we have that an ≤ a ≤ bn , showing that a ∈ In
for each n ∈ N. Thus, we have proven the Nested Interval Property.

2.4.6 a) Notice that

CHAPTER 2. SOLUTIONS 13

(x − y)2 = x2 + y 2 − 2xy = (x + y)2 − 4xy ≥ 0

(x + y)2
≥ xy
4
x+y √
≥ xy
2
Which is the desired result.
b) We already see from the part (a) that xn+1 ≤ yn+1 for all n ∈ N. Thus,
xn + yn 2yn
xn ≤ yn = ≤ = yn
2 2
Also, we have that
√ p
xn+1 = xn yn ≥ x2n = xn
Thus from these inequalities, we have that {xn } is monotone increasing and
{yn } is monotone decreasing. Also from the first inequality, we see that each xn
is a lower bound for each yn and vice versa. Thus by the Monotone Convergence
Theorem, both sequences must converge. Let lim xn = x and lim yn = y. Then
we can apply the strategy as seen in 2.4.2,

√ xn + yn
lim xn+1 = lim xn yn lim yn+1 = lim
2
√ x+y
x= xy y=
2
x2 = xy 2y = x + y
x=y x=y

Thus we have shown the desired result.

2.4.8 a) Remark: sk will denote the partial sum of the first k terms in this
question

First let’s find the explicit formula for the partial sums.
k
X 1 1 1 1
sk = n
= + + ... + k
n=1
2 2 4 2
 
1 1 1 1 1
= + + + ... + k−1
2 2 2 4 2
 
1 1 1
= + sk + k
2 2 2

So from this, we have that sk = 12 + 21 sk + 21k . Solving for sk , we find that

sk = 1 + 21k . Now notice that lim sk = 1, and thus the series converges.
CHAPTER 2. SOLUTIONS 14

(Proof of the limit has been omitted but it boils down to showing that for every
 > 0, there exists N such that ∀k, 21k < . This is easily shown to be true.)
b) Once again, we shall find an explicit formula for sk . This time, notice that
1 1 1
n(n+1) = n − n+1 . (This can be derived through partial fraction decomposition).
Thus,

k k  
X 1 X 1 1
= −
n=1
n(n + 1) n=1 n n + 1
     
1 1 1 1 1 1
= − + − + ... + −
1 2 2 3 k k+1
     
1 1 1 1 1 1 1 1
= + − + + − + + ... + − + −
1 2 2 3 3 k k k+1
1
=1−
k+1
We see that lim sk = 1 and thus the series converges to 1.
(Proof of the limit has once again been omitted, but it comes down to showing
1
that for every  > 0, there exists N such that ∀k ≥ N , k+1 < . This is easily
shown to be true.)
c) To
find an explicit formula for sk , we will use the property of logarithms that
log ab = log a − log b.

k   k
X n+1 X
sk = log = (log n + 1 − log n)
n=1
n n=1
= (log 2 − log 1) + (log 3 − log 2) + ... + (log(k + 1) − log k)
= − log 1 + (log 2 − log 2) + (log 3 − log 3) + ... + (log k − log k) + log(k + 1)
= log(k + 1) − log 1 = log(k + 1)
So we have that sk = log(k + 1). However, this limit does not exist since
log(k + 1) is unbounded.

2.4.10 a) For an let’s look at the partial products, pk .

k   Y k
Y 1 n+1
pk = 1+ =
n=1
n n=1
n
2 3 4 k+1
= · · · ... ·
1 2 3 k
=k+1
Clearly, the partial products of these infinite product do not converge to any
finite limit.
Now, let’s discuss the an = n12 case.
CHAPTER 2. SOLUTIONS 15

k  k
n2 + 1

Y 1 Y
pk = 1+ 2 =
n=1
n n=1
n2
Now let’s look at specific partial products,

p1 = 2
5
p2 = = 2.5
2
25
p3 = ≈ 2.777...
9
p4 ≈ 2.9513
.
.
.
p100 ≈ 3.6396
.
.
.
p900 ≈ 3.6719979143704307
p901 ≈ 3.6720024376438865
p902 ≈ 3.67200695089903
p903 ≈ 3.6720114541691076

So basically it appears that the sequence of partial products at first seems

to be growing rapidly, but much farther out in the sequence, it seems to be
still increasing but not as rapidly and it appears to be approaching some value
around 3.67. However, numerical evidence is not enough. In part (b) of this
problem, we will properly prove that this infinite product has a finite value.
b) Note that in both the infinite product and the infinite sum, if any values of
an = 0, then the result is not affected (because the 1 + an in the product will
evaluate to 1, and the term will simply be 0 in the sum. In either case, the
final value will not be affected). Thus, we can assume that our sequence of an
is strictly positive.
Let sk denote the partial sum in the infinite series and pk be the partial product
in the infinite product. We shall start with the backwards direction. So we shall
assume that the infinite sum converges. Using the inequality given to us that
1 + x ≤ 3x for positive x, we have that

pk = (1 + a1 )(1 + a2 )...(1 + ak ) ≤ 3a1 · 3a2 · ... · 3ak = 3a1 +a2 +...+ak = 3sk

Now since the partial sums converge, we have that the sequence of pk must
CHAPTER 2. SOLUTIONS 16

converge. Thus, the infinite product converges.

Now for the forward direction. Notice that

ak = a1 + a2 + ... + ak ≤ 1 + a1 + a2 + ... + ak ≤ (1 + a1 )(1 + a2 )...(1 + ak ) = pk

By a similar argument as was used in the backwards direction, we have that

since the sequence of pk converge, the sequence of ak must converge as well.

Section 2.5
2.5.2 a) Consider the subsequence (x2 , x3 , x4 , ...). Since this converges, we can
see that by simply adding on the first term x1 to this subsequence will not affect
the limit. Thus, the full sequence also converges to this same limit.
b) Assume the contrary: that (xn ) contains a divergent subsequence but xn
converges. This is contradiction because every subsequence must converge to
the limit of xn . Thus if there is a divergent subsequence, then xn diverges.
c) Since the sequence is bounded, there exists a convergent subsequence that
converges to some limit c. Now if any other subsequence also converged to c,
this would imply that xn would converge. However, this cannot be so there
must be another subsequence that converges to a different limit.
d) Assume that the sequence is strictly monotone increasing. Let (xnk ) be a
convergent subsequence that converges to L. We will show that xn also con-
verges to L. To do this, we must show that for every  > 0 there exists N such
that ∀n ≥ N , we have that |xn − L| < .
Now since the subsequence converges to L, there exists N such that ∀nk ≥ N ,
we have that |xnk − L| < . Now since the sequence is monotone increasing,
there exists some n ≥ N such that

xnk ≤ xn ≤ xnk+1
subtracting L and using that |xnk − L| <  → − < xnk − L < 0, we find
that

− < xnk − L ≤ xn − L ≤ xnk+1 − L < 0

From this, we see that |xn −L| < , as desired. A proof for strictly monotone
decreasing functions follows the same lines except that we have

xnk ≥ xn ≥ xnk+1
which in turn gives us

 > xnk − L ≥ xn − L ≥ xnk+1 − L > 0

From this, we also have that |xn − L| < . The last case is where all the
terms are equal, which is trivially true.
CHAPTER 2. SOLUTIONS 17

2.5.4

2.5.6 We shall prove that the sequence converges to 1. We will split this into
two cases: b ≥ 1 and b ≤ 1.
1
Case 1: If b ≥ 1, then we see that b n ≥ 1. As a proof of this, assume that this
1
were not so, then b n ≤ 1. Raising this to the nth power, we see that b ≤ 1.
However this is a contradiction.
Next, we shall show that the sequence is monotone decreasing by induction.
First notice that b ≤ b2 and after taking the square root of both sides, we see
1 1 1
that b 2 ≤ b. This is our base case. Now, we wish to show that b n+1 ≤ b n . By
1 1
our induction hypothesis, we have that b n ≤ b n−1 . From this, we see that
1 1 1
b · b n−1 ≥ b · b n = b1+ n ≥ b
1 n+1
Thus we see that b ≤ b · b n = b n . Taking the n + 1 root of both sides, we
1 1
have that b n+1 ≤ b n . Thus by the Monotone Convergence Theorem, we have
that the sequence converges for b ≥ 1.
1
Case 2: If 0 ≤ b ≤ 1, then by a similar argument for b ≥ 1, we see that b n ≤ 1.
1 1
Furthermore, we also see by induction that b n+1 ≥ b n and so the sequence is
increasing. Therefore the sequence also converges for b ≤ 1 by the Monotone
Convergence Theorem.
So we see that the sequence converges to some limit, l. And so we can see  1ev-

ery subsequence must also converge to l. In particular, the subsequence b 2n
1 1 √
converges to l. So we have that lim b n = lim b 2n which implies that l = l (we
have used exercise 2.3.1 in doing this) and thus we see that l = 1.

2.5.8 a)
(1, 2, 3, 4, 5, 6, ...) has 0 peaks.
(1, − 12 , − 13 , − 14 , ...) has 1 peak.
(2, 1, − 12 , − 13 , ...) has 2 peaks.
(1, −1, 1, −1, ...) has infinitely many peaks.
b) A sequence has either finitely many peaks or infinitely many peaks.
Case 1: Assume a sequence xn has infinitely many peaks. Then there exists
n0 < n1 < ... such that an0 ≥ an1 ≥ an2 ≥ .... This implies that the sequence
of ank is monotone decreasing.
Case 2: Assume the sequence has finitely many peaks. Let aN be the last peak.
Let n0 = N + 1. Since n0 is not a peak (since an was the last peak), there exists
n1 > n0 such that an1 ≥ an0 Similarly, there exists n2 > n1 such that an2 ≥ an1 .
Continuing this process, there exists nk+1 > nk such that ank+1 ≥ ank . There-
fore, we see that we have n0 < n1 < n2 < ... and an0 ≤ an1 ≤ an2 ≤ ... which is
a monotone increasing sequence.
Thus we have shown that every sequence has a monotone subsequence.

Now a proof of Bolzano-Weierstrass follows when xn is a bounded sequence.

Then we have that there is a monotone subsequence by the above proof. Since
CHAPTER 2. SOLUTIONS 18

xn is bounded, the monotone subsequence of xn is also bounded and thus by the

Montone Convergence Theorem, we have that this subsequence must converge.
Thus, we see that xn contains a convergent subsequence.

2.6
n
2.6.2 a) The sequence { (−1) n is convergent and therefore Cauchy and is clearly
not monotone.
b) This cannot happen. Since Cauchy sequences converge, every subsequence
must also converge and therefore every subsequence is bounded.
c) This cannot happen. Assume that this were possible. Let {xn } be a mono-
tone increasing sequence and assume there exists a convergent (and therefore
Cauchy) subsequence, {xnk } that converges to a limit L. Since {xn } is mono-
tone increasing and is divergent, there exists x` > L. This means that all the
nk < `. However, this would imply that there are only a finite number of terms
in the subsequence. This is contradiction.
A similar argument applies to monotone decreasing sequences by negating each
term, resulting in a montone increasing sequence.
d) Consider the sequence (0, 1, 0, 2, 0, 3, 0, 4, ...) then we can see that the se-
quence is clearly unbounded, however every odd term creates the subsequence
(0, 0, 0, ...) which is clearly Cauchy.

2.6.4 a) Let  > 0. Since an and bn are Cauchy, there exists N1 such that
for n, m ≥ N1 , we have that |an − am | < 2 and similarly there exists N2 such
that for n, m ≥ N2 , we have that |bn − bm | < 2 . Now, let N = max(N1 , N2 ),
then we see that

|cn − cm | = ||an − am | − |am − bm || ≤ |an − am − bn + bm |

≤ |an − am | + |bn − bm |
 
< + =
2 2
Thus we have shown that cn is Cauchy.
b) It is not necessarily true that cn is Cauchy. Let an = 1, which is clearly
a Cauchy sequence. Then cn = (−1)n , which is clearly not Cauchy (consider
 < 2).
n
c) It is not necessarily true that cn is Cauchy. Let an = (−1)n . Since an
converges, it is Cauchy. However, we then have that
(
−1 if n is odd
cn = [[an ]] =
0 if n is even
Clearly, cn is not Cauchy.
CHAPTER 2. SOLUTIONS 19

(−1)n 2
2.6.6 a) Let an = n . Let  > 0 and N >  then
1 1  
|an − am | ≤ |an | + |am | = + < + =
n m 2 2
And so we see that − < an − am <  which gives us am > an − . So this
sequence is quasi-increasing.
b)
c)

2.7