Solutions Manual for:

Understanding Analysis, Second Edition

Stephen Abbott
Middlebury College

June 25, 2015

Author’s note

What began as a desire to sketch out a simple answer key for the problems
in Understanding Analysis inevitably evolved into something a bit more ambi-
tious. As I was generating solutions for the nearly 200 odd-numbered exercises
in the text, I found myself adding regular commentary on common pitfalls and
strategies that frequently arise. My sense is that this manual should be a use-
ful supplement to instructors teaching a course or to individuals engaged in an
independent study. As with the textbook itself, I tried to write with the in-
troductory student firmly in mind. In my teaching of analysis, I have come to
understand the strong correlation between how students learn analysis and how
they write it. A final goal I have for these notes is to illustrate by example how
the form and grammar of a written argument are intimately connected to the
clarity of a proof and, ultimately, to its validity.
The decision to include only the odd-numbered exercises was a compromise
between those who view access to the solutions as integral to their educational
needs, and those who strongly prefer that no solutions be available because of the
potential for misuse. The total number of exercises was significantly increased
for the second edition, and almost every even-numbered problem (in the regular
sections of the text) is one that did not appear in the first edition. My hope is
that this arrangement will provide ample resources to meet the distinct needs
of these different audiences.
I would like to thank former students Carrick Detweiller, Katherine Ott,
Yared Gurmu, and Yuqiu Jiang for their considerable help with a preliminary
draft. I would also like to thank the readers of Understanding Analysis for the
many comments I have received about the text. Especially appreciated are the
constructive suggestions as well as the pointers to errors, and I welcome more
of the same.

Middlebury, Vermont Stephen Abbott

May 2015

v
vi Author’s note
Contents

Author’s note v

1 The Real Numbers √ 1

1.1 Discussion: The Irrationality of 2 . . . . . . . . . . . . . . . . . 1
1.2 Some Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 The Axiom of Completeness . . . . . . . . . . . . . . . . . . . . . 4
1.4 Consequences of Completeness . . . . . . . . . . . . . . . . . . . 6
1.5 Cardinality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 Cantor’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2 Sequences and Series 15

2.1 Discussion: Rearrangements of Infinite Series . . . . . . . . . . . 15
2.2 The Limit of a Sequence . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 The Algebraic and Order Limit Theorems . . . . . . . . . . . . . 16
2.4 The Monotone Convergence Theorem and a First Look at
Infinite Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.5 Subsequences and the Bolzano–Weierstrass Theorem . . . . . . . 24
2.6 The Cauchy Criterion . . . . . . . . . . . . . . . . . . . . . . . . 26
2.7 Properties of Infinite Series . . . . . . . . . . . . . . . . . . . . . 28
2.8 Double Summations and Products of Infinite Series . . . . . . . . 31

3 Basic Topology of R 35
3.1 Discussion: The Cantor Set . . . . . . . . . . . . . . . . . . . . . 35
3.2 Open and Closed Sets . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3 Compact Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.4 Perfect Sets and Connected Sets . . . . . . . . . . . . . . . . . . 41
3.5 Baire’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

4 Functional Limits and Continuity 45

4.1 Discussion: Examples of Dirichlet and Thomae . . . . . . . . . . 45
4.2 Functional Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.3 Continuous Functions . . . . . . . . . . . . . . . . . . . . . . . . 48
4.4 Continuous Functions on Compact Sets . . . . . . . . . . . . . . 52
4.5 The Intermediate Value Theorem . . . . . . . . . . . . . . . . . . 55

vii
viii Contents

4.6 Sets of Discontinuity . . . . . . . . . . . . . . . . . . . . . . . . . 56

5 The Derivative 59
5.1 Discussion: Are Derivatives Continuous? . . . . . . . . . . . . . . 59
5.2 Derivatives and the Intermediate Value Property . . . . . . . . . 59
5.3 The Mean Value Theorems . . . . . . . . . . . . . . . . . . . . . 63
5.4 A Continuous Nowhere-Differentiable Function . . . . . . . . . . 65

6 Sequences and Series of Functions 69

6.1 Discussion: The Power of Power Series . . . . . . . . . . . . . . . 69
6.2 Uniform Convergence of a Sequence of Functions . . . . . . . . . 69
6.3 Uniform Convergence and Differentiation . . . . . . . . . . . . . 74
6.4 Series of Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.5 Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6.6 Taylor Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.7 The Weierstrass Approximation Theorem . . . . . . . . . . . . . 84

7 The Riemann Integral 87

7.1 Discussion: How Should Integration be Defined? . . . . . . . . . 87
7.2 The Definition of the Riemann Integral . . . . . . . . . . . . . . . 87
7.3 Integrating Functions with Discontinuities . . . . . . . . . . . . . 90
7.4 Properties of the Integral . . . . . . . . . . . . . . . . . . . . . . 93
7.5 The Fundamental Theorem of Calculus . . . . . . . . . . . . . . . 95
7.6 Lebesgue’s Criterion for Riemann Integrability . . . . . . . . . . 98

8 Additional Topics 103

8.1 The Generalized Riemann Integral . . . . . . . . . . . . . . . . . 103
8.2 Metric Spaces and the Baire Category Theorem . . . . . . . . . . 105
8.3 Euler’s Sum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
8.4 Inventing the Factorial Function . . . . . . . . . . . . . . . . . . 113
8.5 Fourier Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
8.6 A Construction of R From Q . . . . . . . . . . . . . . . . . . . . 123
Chapter 1

The Real Numbers

√
1.1 Discussion: The Irrationality of 2
1.2 Some Preliminaries
Exercise 1.2.1. (a) Assume, for contradiction, that there exist integers p and
q satisfying
( )2
p
(1) = 3.
q

Let us also assume that p and q have no common factor. Now, equation (1)
implies

(2) p2 = 3q 2 .

From this, we can see that p2 is a multiple of 3 and hence p must also be
a multiple of 3. This allows us to write p = 3r, where r is an integer. After
substituting 3r for p in equation (2), we get (3r)2 = 3q 2 , which can be simplified
to 3r2 = q 2 . This implies q 2 is a multiple of 3 and hence q is also a multiple of
3. Thus we have shown p and q have a common factor, namely 3, when they
were originally assumed to have no common
√ factor.
A similar argument will work for 6 as well because we get p2 = 6q 2 which
implies p is a multiple of 2 and 3. After making √ the necessary substitutions, we
can conclude q is a multiple of 6, and therefore 6 must be irrational.
(b) In this case, the fact that p2 is a multiple of 4 does not imply p is also a
multiple of 4. Thus, our proof breaks down at this point.
Exercise 1.2.2.
Exercise 1.2.3. (a) False, as seen in Example 1.2.2.
(b) True. This will follow from upcoming results about compactness in
Chapter 3.

1
2 Chapter 1. The Real Numbers

(c) False. Consider sets A = {1, 2, 3}, B = {3, 6, 7} and C = {5}. Note that
A ∩ (B ∪ C) = {3} is not equal to (A ∩ B) ∪ C = {3, 5}.
(d) True.
(e) True.

Exercise 1.2.4.

Exercise 1.2.5. (a) If x ∈ (A ∩ B)c then x ∈ / (A ∩ B). But this implies x ∈/A
or x ∈/ B. From this we know x ∈ Ac or x ∈ B c . Thus, x ∈ Ac ∪ B c by the
definition of union.
(b) To show Ac ∪ B c ⊆ (A ∩ B)c , let x ∈ Ac ∪ B c and show x ∈ (A ∩ B)c .
So, if x ∈ Ac ∪ B c then x ∈ Ac or x ∈ B c . From this, we know that x ∈ / A or
x∈/ B, which implies x ∈ / (A ∩ B). This means x ∈ (A ∩ B)c , which is precisely
what we wanted to show.
(c) In order to prove (A ∪ B)c = Ac ∩ B c we have to show,

(1) (A ∪ B)c ⊆ Ac ∩ B c and,

(2) Ac ∩ B c ⊆ (A ∪ B)c .

To demonstrate part (1) take x ∈ (A ∪ B)c and show that x ∈ (Ac ∩ B c ). So,
if x ∈ (A ∪ B)c then x ∈ / (A ∪ B). From this, we know that x ∈A / and x ∈ / B
which implies x ∈ Ac and x ∈ B c . This means x ∈ (Ac ∩ B c ).
Similarly, part (2) can be shown by taking x ∈ (Ac ∩ B c ) and showing that
x ∈ (A ∪ B)c . So, if x ∈ (Ac ∩ B c ) then x ∈ Ac and x ∈ B c . From this, we know
that x ∈/ A and x ∈ / B which implies x ∈ / (A ∪ B). This means x ∈ (A ∪ B)c .
Since we have shown inclusion both ways, we conclude that (A ∪ B)c = Ac ∩ B c .

Exercise 1.2.6.

Exercise 1.2.7. (a) f (A) = [0, 4] and f (B) = [1, 16]. In this case, f (A ∩ B) =
f (A) ∩ f (B) = [1, 4] and f (A ∪ B) = f (A) ∪ f (B) = [0, 16].
(b) Take A = [0, 2] and B = [−2, 0] and note that f (A ∩ B) = {0} but
f (A) ∩ f (B) = [0, 4].
(c) We have to show y ∈ g(A ∩ B) implies y ∈ g(A) ∩ g(B). If y ∈ g(A ∩ B)
then there exists an x ∈ A ∩ B with g(x) = y. But this means x ∈ A and x ∈ B
and hence g(x) ∈ g(A) and g(x) ∈ g(B). Therefore, g(x) = y ∈ g(A) ∩ g(B).
(d) Our claim is g(A ∪ B) = g(A) ∪ g(B). In order to prove it, we have to
show,

(1) g(A ∪ B) ⊆ g(A) ∪ g(B) and,

(2) g(A) ∪ g(B) ⊆ g(A ∪ B).

To demonstrate part (1), we let y ∈ g(A ∪ B) and show y ∈ g(A) ∪ g(B). If

y ∈ g(A ∪ B) then there exists x ∈ A ∪ B with g(x) = y. But this means
1.2. Some Preliminaries 3

x ∈ A or x ∈ B, and hence g(x) ∈ g(A) or g(x) ∈ g(B). Therefore, g(x) = y ∈

g(A) ∪ g(B).
To demonstrate the reverse inclusion, we let y ∈ g(A) ∪ g(B) and show
y ∈ g(A ∪ B). If y ∈ g(A) ∪ g(B) then y ∈ g(A) or y ∈ g(B). This means we
have an x ∈ A or x ∈ B such that g(x) = y. This implies, x ∈ A ∪ B, and
hence g(x) ∈ g(A ∪ B). Since we have shown parts (1) and (2), we can conclude
g(A ∪ B) = g(A) ∪ g(B).
Exercise 1.2.8.
Exercise 1.2.9. (a) f −1 (A) = [−2, 2] and f −1 (B) = [−1, 1]. In this case,
f −1 (A ∩ B) = f −1 (A) ∩ f −1 (B) = [−1, 1] and f −1 (A ∪ B) = f −1 (A) ∪ f −1 (B) =
[−2, 2].
(b) In order to prove g −1 (A ∩ B) = g −1 (A) ∩ g −1 (B), we have to show,

(1) g −1 (A ∩ B) ⊆ g −1 (A) ∩ g −1 (B) and,

(2) g −1 (A) ∩ g −1 (B) ⊆ g −1 (A ∩ B).

To demonstrate part (1), we let x ∈ g −1 (A ∩ B) and show x ∈ g −1 (A) ∩ g −1 (B).

So, if x ∈ g −1 (A ∩ B) then g(x) ∈ (A ∩ B). But this means g(x) ∈ A and
g(x) ∈ B, and hence g(x) ∈ A ∩ B. This implies, x ∈ g −1 (A) ∩ g −1 (B).
To demonstrate the reverse inclusion, we let x ∈ g −1 (A) ∩ g −1 (B) and show
x ∈ g −1 (A ∩ B). So, if x ∈ g −1 (A) ∩ g −1 (B) then x ∈ g −1 (A) and x ∈ g −1 (B).
This implies g(x) ∈ A and g(x) ∈ B, and hence g(x) ∈ A ∩ B. This means,
x ∈ g −1 (A ∩ B).
Similarly, in order to prove g −1 (A ∪ B) = g −1 (A) ∪ g −1 (B), we have to show,

(1) g −1 (A ∪ B) ⊆ g −1 (A) ∪ g −1 (B) and,

(2) g −1 (A) ∪ g −1 (B) ⊆ g −1 (A ∪ B).

To demonstrate part (1), we let x ∈ g −1 (A ∪ B) and show x ∈ g −1 (A) ∪ g −1 (B).

So, if x ∈ g −1 (A ∪ B) then g(x) ∈ (A ∪ B). But this means g(x) ∈ A or
g(x) ∈ B, which implies x ∈ g −1 (A) or x ∈ g −1 (B). From this we know
x ∈ g −1 (A) ∪ g −1 (B).
To demonstrate the reverse inclusion, we let x ∈ g −1 (A) ∪ g −1 (B) and show
x ∈ g −1 (A ∪ B). So, if x ∈ g −1 (A) ∩ g −1 (B) then x ∈ g −1 (A) or x ∈ g −1 (B).
This implies g(x) ∈ A or g(x) ∈ B, and hence g(x) ∈ A ∪ B. This means,
x ∈ g −1 (A ∪ B).
Exercise 1.2.10.
Exercise 1.2.11. (a) There exist two real numbers a and b satisfying a < b
such that for all n ∈ N we have a + 1/n ≥ b.
(b) For all real numbers x > 0, there exists n ∈ N satisfying x ≥ 1/n.
(c) There exist two distinct rational numbers with the property that every
number in between them is irrational.
4 Chapter 1. The Real Numbers

Exercise 1.2.12.

Exercise 1.2.13. (a) From Exercise 1.2.5 we know (A1 ∪ A2 )c = Ac1 ∩ Ac2 which
proves the base case. Now we want to show that
c
if we have (A1 ∪ A2 ∪ · · · ∪ An ) = Ac1 ∩ Ac2 ∩ · · · ∩ Acn , then it follows that

(A1 ∪ A2 ∪ · · · ∪ An+1 )c = Ac1 ∩ Ac2 ∩ · · · ∩ Acn+1 .

Since the union of sets obey the associative law,

(A1 ∪ A2 ∪ · · · ∪ An+1 )c = ((A1 ∪ A2 ∪ · · · ∪ An ) ∪ An+1 )c

which is equal to
(A1 ∪ A2 ∪ · · · ∪ An )c ∩ Acn+1 .
Now from our induction hypothesis we know that

(A1 ∪ A2 ∪ · · · ∪ An )c = Ac1 ∩ Ac2 ∩ · · · ∩ Acn

which implies that

(A1 ∪ A2 ∪ · · · ∪ An )c ∩ Acn+1 = Ac1 ∩ Ac2 ∩ · · · ∩ Acn ∩ Acn+1 .

By induction, the claim is proved for all n ∈ N.

(b) Example 1.2.2 illustrates
∪∞ thisc phenomenon.
∩∞
(c) In order to prove ( n=1 An ) = n=1 Acn we have to show,
( ∞
)c ∞
∪ ∩
(1) An ⊆ Acn and,
n=1 n=1

∞
( ∞
)c
∩ ∪
(2) Acn ⊆ An .
n=1 n=1
∪∞ c ∩∞
To demonstrate
∪∞ part (1), we let x ∈ ( n=1 An ) and show x ∈ n=1 Acn . So, if
c
x ∈ ( n=1 An ) then x ∈ / An for all n ∈ N. This implies
∩∞ x iscin the complement
of each An and by the definition of intersection x ∈ n=1 ∩∞An . c
∪∞ To demonstrate the reverse
∩∞ inclusion, we let x ∈ n=1 An and show x ∈
c
( n=1 An ) . So, if x ∈ n=1 Acn then x∪∈ Acn for all n ∈ N which means
∞
x∈ ∪n∞for all cn ∈ N. This implies x ∈
/ A / ( n=1 An ) and we can now conclude
x ∈ ( n=1 An ) .

1.3 The Axiom of Completeness

Exercise 1.3.1. (a) A real number i is the greatest lower bound, or the infimum,
for a set A ⊆ R if it meets the following two criteria:
1.3. The Axiom of Completeness 5

(i) i is a lower bound for A; i.e., i ≤ a for all a ∈ A, and

(ii) if l is any lower bound for A, then l ≤ i.

(b) Lemma: Assume i ∈ R is a lower bound for a set A ⊆ R. Then, i = inf A

if and only if, for every choice of ϵ > 0, there exists an element a ∈ A satisfying
i + ϵ > a.
(i) To prove this in the forward direction, assume i = inf A and consider
i + ϵ, where ϵ > 0 has been arbitrarily chosen. Because i + ϵ > i, statement
(ii) implies i + ϵ is not a lower bound for A. Since this is the case, there must
be some element a ∈ A for which i + ϵ > a because otherwise i + ϵ would be a
lower bound.
(ii) For the backward direction, assume i is a lower bound with the property
that no matter how ϵ > 0 is chosen, i + ϵ is no longer a lower bound for A. This
implies that if l is any number greater than i then l is no longer a lower bound
for A. Because any number greater than i cannot be a lower bound, it follows
that if l is some other lower bound for A, then l ≤ i. This completes the proof
of the lemma.

Exercise 1.3.2.
Exercise 1.3.3. (a) Because A is bounded below, B is not empty. Also, for
all a ∈ A and b ∈ B, we have b ≤ a. The first thing this tells us is that B is
bounded above and thus α = sup B exists by the Axiom of Completeness. It
remains to show that α = inf A. The second thing we see is that every element
of A is an upper bound for B. By part (ii) of the definition of supremum, α ≤ a
for all a ∈ A and we conclude that α is a lower bound for A.
Is it the greatest lower bound? Sure it is. If l is an arbitrary lower bound
for A then l ∈ B, and part (i) of the definition of supremum implies l ≤ α. This
completes the proof.
(b) We do not need to assume that greatest lower bounds exist as part of
the Axiom of Completeness because we now have a proof that they exist. By
demonstrating that the infimum of a set A is always equal to the supremum
of a different set, we can use the existence of least upper bounds to assert the
existence of greatest lower bounds.
Another way to achieve the same goal is to consider the set −A = {−a : a ∈
A}. If A is bounded below it follows that −A is bounded above and it is not
too hard to prove inf A = sup(−A).
Exercise 1.3.4.

Exercise 1.3.5. (a) In the case c = 0, cA = {0} and without too much difficulty
we can argue that sup(cA) = 0 = c sup A. So let’s focus on the case where c > 0.
Observe that c sup A is an upper bound for cA. Now, we have to show if d is
any upper bound for cA, then c sup A ≤ d. We know ca ≤ d for all a ∈ A, and
thus a ≤ d/c for all a ∈ A. This means d/c is an upper bound for A, and by
Definition 1.3.2 sup A ≤ d/c. But this implies c sup A ≤ c(d/c) = d, which is
precisely what we wanted to show.
6 Chapter 1. The Real Numbers

(b) Assuming the set A is bounded below, we claim sup(cA) = c inf A for the
case c < 0. In order to prove our claim we first show c inf A is an upper bound
for cA. Since inf A ≤ a for all a ∈ A, we multiply both sides of the equation
to get c inf A ≥ ca for all a ∈ A. This shows that c inf A is an upper bound for
cA. Now, we have to show if d is any upper bound for cA, then c inf A ≤ d. We
know ca ≤ d for all a ∈ A, and thus d/c ≤ a for all a ∈ A. This means d/c
is a lower bound for A and from Exercise 1.3.1, d/c ≤ inf A. But this implies
c inf A ≤ c(d/c) ≤ d, which is precisely what we wanted to show.
Exercise 1.3.6.
Exercise 1.3.7. Since a is an upper bound for A, we just need to verify the
second part of the definition of supremum and show that if d is any upper bound
then a ≤ d. By the definition of upper bound a ≤ d because a is an element of
A. Hence, by Definition 1.3.2, a is the supremum of A.
Exercise 1.3.8.
Exercise 1.3.9. (a) Set ϵ = sup B − sup A > 0. By Lemma 1.3.8, there exists
an element b ∈ B satisfying sup B − ϵ < b, which implies sup A < b. Because
sup A is an upper bound for A, then b is as well.
(b) Take A = [0, 1] and B = (0, 1).
Exercise 1.3.10.
Exercise 1.3.11. (a) True. Observe that all elements of B are contained in
A and hence sup A ≥ b for all b ∈ B. By Definition 1.3.2 part (ii), sup B is
less than or equal to any other upper bounds of B. Because sup A is an upper
bound for B, it follows that sup B ≤ sup A.
(b) True. Let c = (sup A + inf B)/2 from which it follows that
a ≤ sup A < c < inf B ≤ b.
(c) False. Consider, the open sets A = (d, c) and B = (c, f ). Then a < c < b
for every a ∈ A and b ∈ B, but sup A = c = inf B.

1.4 Consequences of Completeness

Exercise 1.4.1. (a) We have to show if a, b ∈ Q, then ab and a + b are elements
of Q. By definition, Q = {p/q : p, q ∈ Z, q ̸= 0}. So take a = p/q and b = c/d
where p, q, c, d ∈ Z and q, d ̸= 0. Then, ab = qd
pc
where pc, qd ∈ Z because Z is
closed under multiplication. This implies ab ∈ Q. To see that a + b is rational,
write
p c pd + qc
+ = ,
q d qd
and observe that both pd + qc and qd are integers with qd ̸= 0.
(b) Assume, for contradiction, that a + t ∈ Q. Then t = (t + a) − a is the
difference of two rational numbers, and by part (a) t must be rational as well.
This contradiction implies a + t ∈ I.
1.5. Cardinality 7

Likewise, if we assume at ∈ Q, then t = (at)(1/a) would again be rational

by the result in (a). This implies at ∈ I.
(c) The set of irrationals is not closed under addition and multiplication.
Given two irrationals
√ s and t, √
s + t can be either irrational or rational. For
instance, if s = √2 and t = −√ 2, then s + t =√0 which √ is an√element of Q.
However, if s = 2 and t = 2 2 then s + t = 2 + 2 2 = 3 3 which √ is an
element√ of I. Similarly, st can be either irrational or rational. If s = √2 and
t =√ − 2, then st√= √ −1 which
√ is a rational number. However,if s = 2 and
t = 3 then st = 2 3 = 6 which is an irrational number.
Exercise 1.4.2.
∩∞
Exercise 1.4.3. Let x ∈ R be arbitrary. To prove n=1 (0, 1/n) = ∅ it is
enough to show that x ∈/ (0, 1/n) for some n ∈ N. If x ≤ 0 then we can take
n = 1 and observe x ∈/ (0, 1). If x > 0 then by Theorem 1.4.2 we know there
∩∞
exists an n0 ∈ N such that 1/n0 < x. This implies x ∈/ n=1 (0, 1/n), and our
proof is complete.
Exercise 1.4.4.
Exercise 1.4.5. We have to show the existence of an irrational number between
any√two real numbers
√ a and b. By applying Theorem 1.4.3 on √ the real numbers
√
a− 2 and b− 2 we can √ find a rational number r satisfying a− 2<r< √ b− 2.
This implies a < r + 2 < b. From Exercise 1.4.1(b) we know r + 2 is an
irrational number between a and b.
Exercise 1.4.6.
Exercise 1.4.7. Now, we need to pick n0 large enough so that

1 α2 − 2 2α
< or < α2 − 2.
n0 2α n0
With this choice of n0 , we have

(α − 1/n0 )2 > α2 − 2α/n0 = α2 − (α2 − 2) = 2.

This means (α − 1/n0 ) is an upper bound for T . But (α − 1/n0 ) < α and
α = sup T is supposed to be the least upper bound. This contradiction means
that the case α2 > 2 can be ruled out. Because√we have already ruled out
α2 < 2, we are left with α2 = 2 which implies α = 2 exists in R.
Exercise 1.4.8.

1.5 Cardinality
Exercise 1.5.1. Next let n2 = min{n ∈ N : f (n) ∈ A\{f (n1 )}} and set
g(2) = f (n2 ). In general, assume we have defined g(k) for k < m, and let
g(m) = f (nm ) where nm = min{n ∈ N : f (n) ∈ A\{f (n1 ) . . . f (nk−1 )}}.
8 Chapter 1. The Real Numbers

To show that g : N → A is 1–1, observe that m ̸= m′ implies nm ̸= nm′ and

it follows that f (nm ) = g(m) ̸= g(m′ ) = f (nm′ ) because f is assumed to be 1–1.
To show that g is onto, let a ∈ A be arbitrary. Because f is onto, a = f (n′ ) for
some n′ ∈ N. This means n′ ∈ {n : f (n) ∈ A} and as we inductively remove the
minimal element, n′ must eventually be the minimum by at least the n′ − 1st
step.
Exercise 1.5.2.
Exercise 1.5.3. (a) Because A1 is countable, there exists a 1–1 and onto func-
tion f : N → A1 .
If B2 = ∅, then A1 ∪ A2 = A1 which we already know to be countable.
If B2 = {b1 , b2 , . . . , bm } has m elements then define h : A1 ∪ B2 via
{
bn if n ≤ m
h(n) =
f (n − m) if n > m.
The fact that h is a 1–1 and onto follows immediately from the same properties
of f .
If B2 is infinite, then by Theorem 1.5.7 it is countable, and so there exists a
1–1 onto function g : N → B2 . In this case we define h : A1 ∪ B2 by
{
f ((n + 1)/2) if n is odd
h(n) =
g(n/2) if n is even.
Again, the proof that h is 1–1 and onto is derived directly from the fact that f
and g are both bijections. Graphically, the correspondence takes the form

N: 1 2 3 4 5 6 ···
↕ ↕ ↕ ↕ ↕ ↕
A1 ∪ B2 : a1 b1 a2 b2 a3 b3 ···

To prove the more general statement in Theorem 1.5.8, we may use induction.
We have just seen that the result holds for two countable sets. Now let’s assume
that the union of m countable sets is countable, and show that the union of m+1
countable sets is countable.
Given m + 1 countable sets A1 , A2 , . . . , Am+1 , we can write

A1 ∪ A2 ∪ · · · ∪ Am+1 = (A1 ∪ A2 ∪ · · · ∪ Am ) ∪ Am+1 .

Then Cm = A1 ∪ · · · ∪ Am is countable by the induction hypothesis, and Cm ∪

An+1 is just the union of two countable sets which we know to be countable.
This completes the proof.
(b) Induction cannot be used when we have an infinite number of sets. It
can only be used to prove facts that hold true for each value of n ∈ N. See the
discussion in Exercise 1.2.13 for more on this.
1.5. Cardinality 9

(c) Let’s first consider the case where the sets {A∪
n } are disjoint. In order to
∞
achieve 1-1 correspondence between the set N and n=1 An , we first label the
elements in each countable set An as

An = {an1 , an2 , an3 , . . .}.

∪∞
Now arrange the elements of n=1 An in an array similar to the one for N given
in the exercise:
A1 = a11 a12 a13 a14 a15 ···
A2 = a21 a22 a23 a24 ···
A3 = a31 a32 a33 ···
A4 = a41 a42 ···
A5 = a51 · · ·
..
.
∪∞
This establishes a 1–1 and onto mapping g : N → n=1 An where g(n) corre-
sponds to the element ajk where (j, k) is the row and column location of n in
the array for N given in the exercise.
If the sets {An } are not disjoint then our mapping may not be 1–1. In this
case we could again replace An with Bn = An \{A1 ∪ · · · ∪ An−1 }. Another
approach∪is to use the previous argument to establish a 1–1 correspondence
∞
between n=1 An and an infinite subset of N, and then appeal to Theorem
1.5.7.

Exercise 1.5.4.

Exercise 1.5.5. (a) The identity function f (a) = a for all a ∈ A shows that
A ∼ A.
(b) Since A ∼ B we know there is 1-1, onto function from A onto B. This
means we can define another function g : B → A that is also 1-1 and onto. More
specifically, if f : A → B is 1-1 and onto then f −1 : B → A exists and is also
1-1 and onto.
(c) We will show there exists a 1-1, onto function h : A → C. Because
A ∼ B, there exists g : A → B that is 1–1 and onto. Likewise, B ∼ C implies
that there exists f : B → C that is also 1-1 and onto. So let’s define h : A → C
by the composition h = f ◦ g.
In order to show f ◦ g is 1-1, take a1 , a2 ∈ A where a1 ̸= a2 and show
f (g(a1 )) ̸= f (g(a2 )). Well, a1 ̸= a2 implies that g(a1 ) ̸= g(a2 ) because g is 1–1.
And g(a1 ) ̸= g(a2 ) implies that f (g(a1 )) ̸= (f (g(a − 2)) because f is 1–1. This
shows f ◦ g is 1–1.
In order to show f ◦ g is onto, we take c ∈ C and show that there exists an
a ∈ A with f (g(a)) = c. If c ∈ C then there exists b ∈ B such that f (b) = c
because f is onto. But for this same b ∈ B we have an a ∈ A such that g(a) = b
since g is onto. This implies f (b) = f (g(a)) = c and therefore f ◦ g is onto.

Exercise 1.5.6.
10 Chapter 1. The Real Numbers

Exercise 1.5.7. (a). The function f (x) = (x, 13 ) is 1–1 from (0, 1) to S.
(b) Given (x, y) ∈ S, let’s write x and y in their decimal expansions

x = .x1 x2 x3 . . . and y = .y1 y2 y3 . . .

where we make the convention that we always use the terminating form (or
repeated 0s) over the repeating 9s form when the situation arises.
Now define f : S → (0, 1) by

f (x, y) = .x1 y1 x2 y2 x3 y3 . . .

In order to show f is 1–1, assume we have two distinct points (x, y) ̸= (w, z)
from S. Then it must be that either x ̸= w or y ̸= z, and this implies that in
at least one decimal place we have xi ̸= wi or yi ̸= zi . But this is enough to
conclude f (x, y) ̸= f (w, z).
The function f is not onto. For instance the point t = .555959595 . . . is
not in the range of f because the ordered pair (x, y) with x = .555 . . . and
y = .5999 . . . would not be allowed due to our convention of using terminating
decimals instead of repeated 9s.
Exercise 1.5.8.
√ √
Exercise 1.5.9. (a) 2 is a√root √ of the polynomial x2 − 2, 3 2 is a root of
the polynomials x3 − 2, and 3 + 2 is a root of x4 − 10x2 + 1. Since all of
these numbers are roots of polynomials with integer coefficients, they are all
algebraic.
(b) Fix n, m ∈ N. The set of polynomials of the form

an xn + an−1 xn−1 + · · · + a1 x + a0

satisfying |an | + |an−1 | + · · · + |a0 | ≤ m is finite because there are only a finite
number of choices for each of the coefficients (given that they must be integers.)
If we let Anm be the set of all the roots of polynomials of this form, then because
each one of these polynomials has at most n roots, the set Anm is finite. Thus
An , the set of algebraic numbers obtained as roots of any polynomial (with
integer coefficients) of degree n, can be written as a countable union of finite
sets
∪∞
An = Anm .
m=1

It follows that An is countable. ∪∞

(c) If A is the set of all algebraic numbers, then A = n=1 An . Because each
An is countable, we may use Theorem 1.5.8 to conclude that A is countable as
well.
If T is the set transcendental numbers, then A ∪ T = R. Now if T were
countable, then R = A ∪ T would also be countable. But this is a contradiction
because we know R is uncountable, and hence the collection of transcendental
numbers must also be uncountable.
1.6. Cantor’s Theorem 11

Exercise 1.5.10.

Exercise 1.5.11. (a) For all x ∈ A′ , there exists a unique y ∈ B ′ satisfying

g(y) = x. This means that there is a well-defined inverse function g −1 (x) = y
that maps A′ onto B ′ . Setting
{
f (x) if x ∈ A
h(x) =
g −1 (x) if x ∈ A′

gives the desired 1–1, onto function from X to Y .

(b) To see that the sets A1 , A2 , A3 , . . . are pairwise disjoint, note that A1 ∩
An = ∅ for all n ≥ 2 because A1 = X\g(Y ) and An ⊆ g(Y ) for all n ≥ 2.
In the general case of An ∩ Am where 1 < n < m, note that if x ∈ An ∩ Am
then f −1 (g −1 (x)) ∈ An−1 ∩ Am−1 . Continuing in this way, we can show A1 ∩
Am−n+1 is not empty, which is a contradiction. Thus An ∩ Am = ∅. Just to be
clear, the disjointness of the An sets is not crucial to the overall proof, but it
does help paint a clearer picture of how the sets A and A′ are constructed.
(c) This is very straightforward. Each x ∈ A comes from some An and so
f (x) ∈ f (An ) ⊆ B. Likewise, each y ∈ B is an element of some f (An ) and thus
y = f (x) for some x ∈ An ⊆ A. Thus f : A → B is onto.
(d) Let y ∈ B ′ . Then y ∈
/ f (An )) for all n, and because g is 1–1, g(y) ∈ / An+1
/ A1 either and so g maps B ′ into A′ .
for all n. Clearly, g(y) ∈
To see that g maps B ′ onto A′ , let x ∈ A′ be arbitrary. Because A′ ⊆ g(Y ),
there exists y ∈ Y with g(y) = x. Could y be an element of B? No, because
g(B) ⊆ A. So y ∈ B ′ and we have g : B ′ → A′ is onto.

1.6 Cantor’s Theorem

Exercise 1.6.1. The function f (x) = (x − 1/2)/(x − x2 ) is a 1–1, onto mapping
from (0, 1) to R. This shows (0, 1) ∼ R, and the result follows using the ideas
in Exercise 1.5.5.

Exercise 1.6.2.

Exercise 1.6.3. (a) If we imitate the proof to try and show that Q is uncount-
able, we can construct a real number x in the same way. This x will again
fail to be in the range of our function f , but there is no reason to expect x to
be rational. The decimal expansions for rational numbers either terminate or
repeat, and this will not be true of the constructed x.
(b) By using the digits 2 and 3 in our definition of bn we eliminate the
possibility that the point x = .b1 b2 b3 . . . has some other possible decimal repre-
sentation (and thus it cannot exist somewhere in the range of f in a different
form.)

Exercise 1.6.4.

Exercise 1.6.5. (a) P (A) = {∅, {a}, {b}, {c}, {a, b}, {a, c}, {b, c}, {a, b, c}}.
12 Chapter 1. The Real Numbers

(b) First, list the n elements of A. To construct a subset of A, we consider

each element and associate either a ‘Y’ if we decide to include it in our subset
or an ‘N’ if we decide not to include it. Thus, to each subset of A there is an
associated sequence of length n of Y s and N s. This correspondence is 1–1, and
the proof is done by observing there are 2n such sequences.
A second way to prove this result is with induction.

Exercise 1.6.6.

Exercise 1.6.7. This solution depends on the mappings chosen in the previous
exercise. The key point is that no matter how this is done, the resulting set B
should not be in the range of the particular 1–1 mapping used to create it.

Exercise 1.6.8.

Exercise 1.6.9. It is unlikely that there is a reasonably simple way to explicitly

define a 1–1, onto mapping from P (N) to R. A more fruitful strategy is to make
use of the ideas in Exercise 1.5.5 and 1.5.11. In particular, we have seen earlier
in this section that R ∼ (0, 1). It is also true that P (N) ∼ S where S is the
set of all sequences consisting of 0s and 1s from Exercise 1.6.4. To see why,
let A ∈ P (N) be an arbitrary subset of N. Corresponding to this set A is the
sequence (an ) where an = 1 if n ∈ A and an = 0 otherwise. It is straightforward
to show that this correspondence is both 1–1 and onto, and thus P (N) ∼ S.
With a nod to Exercise 1.5.5, we can conclude that P (N) ∼ R if we can
demonstrate that S ∼ (0, 1). Proving this latter fact is easier, but it is still
not easy by any means. One way to avoid some technical details, is to use the
Schröder–Bernstein Theorem (Exercise 1.5.11). Rather than finding a 1–1, onto
function, the punchline of this result is that we will be done if we can find two
1–1 functions, one mapping (0, 1) into S, and the other mapping S into (0, 1).
There are a number of creative ways to produce each of these functions.
Let’s focus first on mapping (0, 1) into S. A fairly natural idea is to think
in terms of binary representations. Given x ∈ (0, 1) let’s inductively define a
sequence (xn ) in the following way. First, bisect (0, 1) into the two parts (0, 1/2)
and [1/2, 1). Then set x1 = 0 if x is in the left half, and x1 = 1 if x is in the
right half. Now let I be whichever of these two intervals contains x, and bisect
it using the same convention of including the midpoint in the right half. As
before we set x2 = 0 if x is in the left half of I and x2 = 1 if x is in the right
half. Continuing this process inductively, we get a sequence (xn ) ∈ S that is
uniquely determined by the given x ∈ (0, 1), and thus the mapping is 1–1.
It may seem like this mapping is onto S but it falls just short. Because of
our convention about including the midpoint in the right half of each interval,
we never get a sequence that is eventually all 1s, nor do we get the sequence of
all 0s. This is fixable. The collection of all sequences in S that are NOT in the
range of this mapping form a countable set, and it is not too hard to show that
the cardinality of S with a countable set removed is the same as the cardinality
of S. The other option is to use the Schröder–Bernstein Theorem mentioned
previously. Having found a 1–1 function from (0, 1) into S, we just need to
1.6. Cantor’s Theorem 13

produce a 1–1 function that goes the other direction. An example of such a
function would be the one that takes (xn ) ∈ S and maps it to the real number
with decimal expansion .x1 x2 x3 x4 . . .. Because the only decimal expansions
that aren’t unique involve 9s, we can be confident that this mapping is 1–1.
The Schröder–Bernstein Theorem now implies S ∼ (0, 1), and it follows that
P (N) ∼ R.

Exercise 1.6.10.
14 Chapter 1. The Real Numbers
Chapter 2

Sequences and Series

2.1 Discussion: Rearrangements of Infinite Se-

ries

2.2 The Limit of a Sequence

Exercise 2.2.1. Consider the sequence (− 12 , 12 , − 12 , 21 · · · ) This sequence ver-
conges to x = 0. To see this, note that we only have to produce a single ϵ > 0
where the prescribed condition follows, and in this case we can take ϵ = 2. This
ϵ works because for all N ∈ N, it is true that n ≥ N implies |xn − 12 | < 1.
This is also an example of a vercongent sequence that is divergent. Notice
that the “limit” x = 0 is not unique. We could also show this same sequence
verconges to x = 1 by choosing ϵ = 3.
In general, a vercongent sequence is a bounded sequence. By a bounded
sequence, we mean that there exists an M ≥ 0 satisfying |xn | ≤ M for all n ∈ N.
In this case we can always take x = 0 and ϵ = M + 1. Then |xn − x| = |xn | < ϵ,
and the sequence (xn ) verconges to 0.

Exercise 2.2.2.

Exercise 2.2.3. a) There exists at least one college in the United States where
all students are less than seven feet tall.
b) There exists a college in the United States where all professors gave at
least one student a grade of C or less.
c) At every college in the United States, there is a student less than six feet
tall.

Exercise 2.2.4.

Exercise 2.2.5. a) The limit of (an ) is zero. To show this let ϵ > 0 be arbitrary.
We must show that there exists an N ∈ N such that n ≥ N implies |[[5/n]]−0| <

15
16 Chapter 2. Sequences and Series

ϵ. Well, pick N = 6. If n ≥ N we then have;

[[ ]]
5
− 0 = |0 − 0| < ϵ,
n
because [[5/n]] = 0 for all n > 5.
b) Here the limit of an is 1. Let ϵ > 0 be arbitrary. By picking N = 7 we
have that for n ≥ N ,
[[ ]]
12 + 4n
− 1 = |1 − 1| < ϵ,
3n
because [[(12 + 4n)/3n]] = 1 for all n ≥ 7.
In these exercises, the choice of N does not depend on ϵ in the usual way. In
exercise (b) for instance, setting N = 7 is a suitable response for every choice
of ϵ > 0. Thus, this is a rare example where a smaller ϵ > 0 does not require a
larger N in response.
Exercise 2.2.6.
Exercise 2.2.7. (a) The sequence (−1)n is frequently in the set 1.
(b) Definition (i) is stronger. “Frequently” does not imply “eventually”, but
“eventually” implies “frequently”.
(c) A sequence (an ) converges to a real number a if, given any ϵ-neighborhood
Vϵ (a) of a, (an ) is eventually in Vϵ (a).
(d) Suppose an infinite number of terms of a sequence (xn ) are equal to 2,
then (xn ) is frequently in the interval (1.9, 2.1). However, (xn ) is not necessarily
eventually in the interval (1.9, 2.1). Consider the sequence (2, 0, 2, 0, 2, · · · ), for
instance.
Exercise 2.2.8.

2.3 The Algebraic and Order Limit Theorems

Exercise 2.3.1. (a) Let ϵ > 0 be arbitrary. We must find an N such that
√
n ≥ N implies | xn − 0| < ϵ. Because (xn ) → 0, there exists
√ N ∈ N such that
n ≥ N implies |xn − 0| = xn < ϵ . Using this N , we have (xn ) < ϵ , which
2 2 2
√
gives | xn − 0| < ϵ for all n ≥ N , as desired.
(b) Part (a) handles the case x = 0, so we may assume x >√0. Let ϵ > 0.
√
This time we must find an N such that n ≥ N implies | xn − x| < ϵ, for all
n ≥ N . Well,
(√ √ )
√ √ √ √ xn + x
| xn − x| = | xn − x| √ √
xn + x
|xn − x|
= √ √
xn + x
|xn − x|
≤ √
x