Supplement For Measure, Integration & Real Analysis: Sheldon Axler

Download as pdf or txt
Download as pdf or txt
You are on page 1of 44

Supplement for

Measure, Integration & Real Analysis


Sheldon Axler

This supplement for Measure, Integration & Real Analysis should refresh your
understanding of standard definitions, notation, and results from undergraduate real
analysis. You will then be ready to read Measure, Integration & Real Analysis.
The set R of real numbers, with the usual operations of addition and multiplication
and the usual ordering, is a complete ordered field. This supplement begins by
explaining the meaning of the last three words of the previous sentence. Because you
have used the ordered field properties of R since childhood, in this supplement we
emphasize the deep properties that follow from completeness.
Section B of this supplement presents a construction of the real numbers using
Dedekind cuts (this is the only section of this supplement not used in Measure,
Integration & Real Analysis, so you can choose to skip this section).
Section C focuses on the crucial concepts of supremum and infimum. The basic
properties of open and closed subsets of Rn are discussed in Section D.
In Section E of this supplement we prove the Bolzano–Weierstrass Theorem,
which is then used as a key tool for results concerning uniform continuity and
maxima/minima of continuous functions on closed bounded subsets of Rn.

Nineteenth-century painting of Cicero discovering the tomb of Archimedes.


In Section C we see how the Archimedean Property of the real
numbers follows from the completeness property.

Supplement for Measure, Integration & Real Analysis by Sheldon Axler


is licensed under a Creative Commons Attribution-NonCommercial 4.0 1
International License. 30 November 2019. ©Sheldon Axler
2 The Real Numbers and Rn

A Complete Ordered Fields


Fields
The algebraic structure of a field captures the arithmetic properties we expect of the
real numbers. In the definition of a field below, an operation of addition on a set F is
a function that assigns an element of F denoted a + b to each ordered pair ( a, b) of
elements of F. An operation of multiplication on a set F is a function that assigns an
element of F denoted ab or a · b to each ordered pair ( a, b) of elements of F.

0.1 Definition field

A field is a set F along with operations of addition and multiplication on F that


have the following properties:
• commutativity: a + b = b + a and ab = ba for all a, b ∈ F;
• associativity: ( a + b) + c = a + (b + c) and ( ab)c = a(bc) for all
a, b, c ∈ F;
• distributive property: a(b + c) = ab + ac for all a, b, c ∈ F;
• additive identity: there exists an element 0 ∈ F such that a + 0 = a for all
a ∈ F;
• additive inverse: for each a ∈ F, there exists an element − a ∈ F such that
a + (− a) = 0;
• multiplicative identity: there exists an element 1 ∈ F such that 1 6= 0 and
a1 = a for all a ∈ F;
• multiplicative inverse: for each a ∈ F with a 6= 0, there exists an element
a−1 ∈ F such that aa−1 = 1.

The set Q of rational numbers, with the usual operations of addition and multi-
plication, is a field. As another example, the set {0, 1}, with the usual operations of
addition and multiplication except that 1 + 1 is defined to be 0, is a field.
Because 0 is the additive identity in a field, the result below connects addition and
multiplication. The only field property that connects addition and multiplication is
the distributive property. Thus the proof of the result below must use the distributive
property. With that hint, the main idea of the proof (writing 0 as 0 + 0 and then using
the distributive property) becomes easier to discover.

0.2 a0 = 0

Suppose F is a field. Then a0 = 0 for every a ∈ F.

Proof Suppose a ∈ F. Then


a0 = a(0 + 0)
= a0 + a0.
Now add −( a0) to each side of the equation above, getting 0 = a0, as desired.
Section A Complete Ordered Fields 3

The familiar properties of arithmetic all follow easily from the field properties
listed in the definition of a field. For example, here are a few properties of the additive
inverse.

0.3 properties of the additive inverse in a field

Suppose F is a field. Then


(a) for each a ∈ F, the additive inverse of a is unique (thus the notation − a
makes sense);
(b) −(− a) = a for each a ∈ F;
(c) (−1) a = − a for each a ∈ F.

You should be able to write down analogous properties for the multiplicative
inverse in a field. The proofs of the easy and familiar field properties are not provided
here because we need to get to other topics. However, you may benefit by finding
your own proof of the result above and other basic results about the arithmetic of
fields.
Subtraction and division are defined in a field using the appropriate inverse, as
follows.

0.4 Definition subtraction; division

Suppose F is a field and a, b ∈ F.


• The difference a − b is defined by the equation

a − b = a + (−b).

a
• If b 6= 0, then the quotient b (which is also denoted by a/b and by a ÷ b) is
defined by the equation
a
= ab−1 .
b

Ordered Fields
The usual arithmetic properties that we expect of the real numbers follow from the
definition of a field. However, more structure is needed to generate meaning for the
the order relationship a < b that we expect of the real numbers. The easiest way to
get at order properties in a field comes from designating a subset to be thought of as
the positive numbers. Then the ordering a < b can be defined to mean that b − a is
positive.
As motivation for the following definition, think of the properties we expect of the
positive numbers as a subset of the real numbers: every real number is either positive
or 0 or its additive inverse is positive; a real number and its additive inverse cannot
both be positive; the sum and product of two positive numbers are both positive
numbers.
4 The Real Numbers and Rn

In the following definition, the symbol P should remind you of the positive
numbers.

0.5 Definition ordered field; positive

An ordered field is a field F along with a subset P of F, called the positive subset,
with the following properties:
• if a ∈ F, then a ∈ P or a = 0 or − a ∈ P;
• if a ∈ P, then − a ∈
/ P;
• if a, b ∈ P, then a + b ∈ P and ab ∈ P.

For example, the field Q of rational numbers, with the usual operations of addition
and multiplication and with P denoting the usual set of positive rational numbers, is
an ordered field.
Because you are already familiar with the properties of the positive numbers, the
statements and easy proofs of results concerning the positive subset are left to you as
exercises. The following result and its proof give an example of how to work with
the definition of an ordered field.

0.6 the positive subset is closed under multiplicative inverse

Suppose F is an ordered field with positive subset P. Then

(a) 1 ∈ P;
(b) a−1 ∈ P for every a ∈ P.

Proof To prove (a), note that the definition of an ordered field implies that either
1 ∈ P or −1 ∈ P. If 1 ∈ P, then we are done. If −1 ∈ P, then 1 = (−1)(−1) ∈ P
(because P is closed under multiplication). Either way, we conclude that 1 ∈ P.
To prove (b), suppose a ∈ P. If we had − a−1 ∈ P, then we would have
−1 = (− a−1 ) a ∈ P (because P is closed under multiplication), which contradicts
the first bullet point. Thus a−1 ∈ P, as desired.

Now we use the positive subset of an ordered field to define the order relations.

0.7 Definition less than; greater than

Suppose F is an ordered field with positive subset P. Suppose a, b ∈ F. Then


• a < b is defined to mean b − a ∈ P;
• a ≤ b is defined to mean a < b or a = b;
• a > b is defined to mean b < a;
• a ≥ b is defined to mean a > b or a = b.
Section A Complete Ordered Fields 5

An important but easy result is proved


Notice that the statement 0 < b is
below to give you an indication of the pat-
equivalent to the statement b ∈ P.
tern of proofs involving order properties.
However, the other statements and proofs of the simple ordering properties are left
to you as exercises because you already have many years of experience with the
appropriate properties, and we need to get to other topics.

0.8 transitivity

Suppose F is an ordered field and a, b, c ∈ F. If a < b and b < c, then a < c.

Proof Suppose a < b and b < c. Hence c − b ∈ P and b − a ∈ P. Because P is


closed under addition, we conclude that

c − a = (c − b) + (b − a) ∈ P.

Thus a < c, as desired.

The familiar concept of absolute value can be defined on an ordered field, as


follows.

0.9 Definition absolute value

Suppose F is an ordered field and b ∈ F. The absolute value of b, denoted |b|, is


defined by (
b if b ≥ 0,
|b| =
−b if b < 0.

The observation that b ≤ |b| and −b ≤ |b| for every b in an ordered field F
provides the key to the proof of our next result.

0.10 | a + b| ≤ | a| + |b|

Suppose F is an ordered field and a, b ∈ F. Then

| a + b | ≤ | a | + | b |.

Proof First suppose a + b ≥ 0. In that case, we have

| a + b | = a + b ≤ | a | + | b |.

Now suppose a + b < 0. In that case, we have

| a + b| = −( a + b) = (− a) + (−b) ≤ | a| + |b|,

completing the proof.


6 The Real Numbers and Rn

The next result allows us to think of Q, the ordered field of rational numbers
with the usual operations of addition and multiplication and the usual ordering, as
contained in each ordered field.

0.11 every ordered field contains Q

Suppose F is an ordered field. Define ϕ : Q → F as follows: ϕ(0) = 0, and for


m and n positive integers with no common integer factors bigger than 1, let

ϕ( m
n ) = (1 + · · · + 1)(1 + · · · + 1)−1
| {z } | {z }
m times n times

and let
ϕ(− m · · + 1})−1 ,

(−1) + · · · + (−1) (1| + ·{z
n) = | {z }
m times n times

where each 1 above is the multiplicative identity in F. Then ϕ is a one-to-one


function that preserves all the ordered field properties.

Proof
The properties of an ordered field
To show that ϕ is a one-to-one func-
imply that 1 + · · · + 1 > 0. In
tion, suppose first that | {z }
n times
ϕ( m
p particular, 1 + · · · + 1 6= 0, and
n ) = ϕ ( q ), | {z }
n times
where m, n, p, q are positive integers and thus the multiplicative inverses
both fractions are in reduced form. The above make sense in F.
equality above implies that

(1| + ·{z
· · + 1})(1| + ·{z
· · + 1}) = (1| + ·{z
· · + 1})(1| + ·{z
· · + 1}).
m times q times p times n times

Repeated applications of the distributive property show that both sides of the equation
above are a sum, with 1 appearing mq times on the left side and pn times on the right
side. Thus
mq = pn
(because otherwise, after adding the additive inverse of the shorter side to both sides
of the equation above, we would have a sum of 1’s equaling 0, which would violate
p
the order properties). Thus m n = q , which shows that the restriction of ϕ to the
positive rational numbers is a one-to-one function.
Using the ideas of the paragraph above, the reader should be able to show that ϕ
is a one-to-one function on all of Q. The reader should also be able to verify that ϕ
preserves all the ordered field properties. In other words, ϕ( a + b) = ϕ( a) + ϕ(b),
ϕ( ab) = ϕ( a) ϕ(b), ϕ(− a) = − ϕ( a), ϕ( a−1 ) = ϕ( a)−1 , and ϕ( a) > 0 if and
only if a > 0 for all a, b ∈ Q (with a 6= 0 for the multiplicative inverse condition).

The result above means that we can identify a ∈ Q with ϕ( a) ∈ F. Thus from
now on we think of Q as a subset of each ordered field.
Section A Complete Ordered Fields 7

Completeness
The Pythagorean Theorem implies that a right trian-
gle with two legs of length 1 has a hypotenuse whose
length c satisfies the equation c2 = 2. The ancient
Greeks discovered that the rational numbers are not
rich enough to have such a number, as shown by the
next result.
An isoceles right triangle,
with c2 = 2.

0.12 no rational number has a square equal to 2

There does not exist a rational number whose square is 2.

Proof Suppose there exist integers m and n such that


 m 2
= 2.
n
By canceling common factors, we can choose m and
n to have no common integer factors greater than 1.
In other words, we can assume that m n is a fraction in
reduced form.
The equation above is equivalent to the equation
m2 = 2n2 .
Thus m2 is even. Hence m is even. Thus m = 2k for
some integer k. Substituting 2k for m in the equation
above gives Pythagoras explaining his
4k2 = 2n2 , work (from The School of
or equivalently Athens, painted by Raphael
2k2 = n2 . around 1510).

Thus n2 is even. Hence n is even.


We have now shown that both m and
“When you have excluded the
n are even, contradicting our choice of
impossible, whatever remains,
m and n as having no common integer
however improbable, must be the
factors greater than 1.
truth.”
This contradiction means our original
—S HERLOCK H OLMES
assumption that there is a rational num-
ber whose square equals 2 was incorrect,
completing the proof.

Intuitively, we expect that the length of any line segment (including the hypotenuse
of a right triangle with two legs of length 1) should√ be a real number. Thus the result
above tells us that there should be a real number 2 that is not rational. The rational
numbers Q, with the usual operations of addition and multiplication, form an ordered
field. Hence we see that we need more than the properties of an ordered field to
describe our notion of the real numbers.
8 The Real Numbers and Rn

Experience has shown that the best way to capture the expected properties of the
real numbers comes through consideration of upper bounds.

0.13 Definition upper bound

Suppose F is an ordered field and A ⊂ F. An element b of F is called an upper


bound of A if a ≤ b for every a ∈ A.

0.14 Example upper bounds


If we work in the ordered field Q of rational numbers and

A1 = { a ∈ Q : a ≤ 3} and A2 = { a ∈ Q : a < 3},

then every rational number b with b ≥ 3 is an upper bound of A1 and of A2 .

Some subsets of Q do not have an upper bound.

0.15 Example no upper bound


If we work in the ordered field Q of rational numbers and A is the set of even
integers, then A does not have an upper bound.

Now we come to a crucial definition.

0.16 Definition least upper bound

Suppose F is an ordered field and A ⊂ F. An element b of F is called a least


upper bound of A if both the following conditions hold:
• b is an upper bound of A;
• b ≤ c for every upper bound c of A.

In other words, a least upper bound of a set is an upper bound that is less than or
equal to all the other upper bounds of the set.
If b and d are both least upper bounds of a subset A of an ordered field F, then
b ≤ d and d ≤ b (by the second bullet point above), and hence b = d. In other
words, a least upper bound of a set, if it exists, is unique.

0.17 Example least upper bounds


If we work in the ordered field Q of rational numbers and

A1 = { a ∈ Q : a ≤ 3} and A2 = { a ∈ Q : a < 3},

then 3 is the least upper bound of A1 and 3 is the least upper bound of A2 . Note that
the least upper bound 3 of A1 is an element of A1 , but the least upper bound 3 of A2
is not an element of A2 .
Section A Complete Ordered Fields 9

The next example shows that a nonempty subset of an ordered field can have an
upper bound without having a least upper bound.

0.18 Example { a ∈ Q : a2 < 2} does not have a least upper bound in Q


Suppose b ∈ Q. We want to show that
Intuitively, the least upper bound
√ of
b is not a least upper bound of { a ∈ Q :
{ a ∈ Q : a2 < 2} should be 2,
a2 < 2}. We know from 0.12 that b2 6= 2.
but there is no such number in Q.
Thus b2 < 2 or b2 > 2.
First consider the case where b2 < 2.
If we can find a positive rational number The idea here is that if b2 < 2, then
δ such that (b + δ)2 < 2, then b is not an we can find a number slightly bigger
upper bound of { a ∈ Q : a2 < 2}. Take than b in { a ∈ Q : a2 < 2}.

2 − b2
δ= .
5
Because b < 2 and 0 < δ < 1, we have 2b + δ < 5 and

(b + δ)2 = b2 + (2b + δ)δ


< b2 + 5δ
= 2.

Thus b is not an upper bound of { a ∈ Q : a2 < 2}.


Now consider the case where b > 0
and b2 > 2. If we can find a rational The idea here is that if b2 > 2, then
number δ such that 0 < δ < b and we can find a number slightly
2
(b − δ) > 2, then b − δ is an upper smaller than b that is an upper
bound of { a ∈ Q : a2 < 2}, which bound of { a ∈ Q : a2 < 2}.
implies that b is not a least upper bound
of { a ∈ Q : a2 < 2}. Take
b2 − 2
δ= .
2b
Then 0 < δ < b and
(b − δ)2 = b2 − 2bδ + δ2
> b2 − 2bδ
= 2.
Thus b is not a least upper bound of { a ∈ Q : a2 < 2}, which completes the
explanation of why { a ∈ Q : a2 < 2} does not have a least upper bound in Q.

The last example indicates that the absence of a least upper bound prevents the field
of rational numbers from having a square root of 2, motivating the next definition.

0.19 Definition complete ordered field

An ordered field F is called complete if every nonempty subset of F that has an


upper bound has a least upper bound.
10 The Real Numbers and Rn

Example 0.18 shows that Q is not a complete ordered field. Our intuitive notion
of the real line as having no holes indicates that the field of real numbers should be a
complete ordered field. Thus we want to add completeness to the list of properties
that the field of real numbers should possess.
Experience shows that no additional properties beyond being a complete ordered
field are needed to prove all known properties of the real numbers. Thus we define
the real numbers (which we have not previously defined) to be a complete ordered
field.
Here is the formal definition, which is followed by a discussion of philosophical
issues that this definition raises.

0.20 Definition R, the field of real numbers

• The symbol R denotes a complete ordered field.


• R is called the field of real numbers.

The definition of R that we have just given raises the following three philosophical
issues:

• Definitions or axioms? We have defined R to be a complete ordered field. For


the rest of this book, we will prove results about R based on this definition. An
alternative approach, in the spirit of the ancient Greeks, would be to list axioms
(such as that every nonempty set with an upper bound has a least upper bound)
that are assumed to be self-evident for the real line (which would not be defined
in this approach).
The more modern viewpoint taken here dispenses with axioms about R and
instead concentrates on proving results about a certain structure—complete
ordered fields—that is independent of any supposedly self-evident properties.
• Existence? We will be proving theorems about R, which means theorems about
complete ordered fields. Those theorems would be meaningless if there do not
exist any complete ordered fields. Thus in the next section we will construct a
complete ordered field.
• Uniqueness? The question of uniqueness is far less important than the question
of existence. If there exist many different complete ordered fields, then all the
theorems that we prove will apply to all those different complete ordered fields,
which is a fine situation. However, you may be interested to know that there is
essentially only one complete ordered field.
Here essentially only one means that any two complete ordered fields differ only
in the names of their elements. Specifically, Exercise 16 in Section C shows that
there is a one-to-one function from any complete ordered field onto any other
complete ordered field that preserves all the properties of a complete ordered
field.
Section A Complete Ordered Fields 11

EXERCISES A
1 Prove 0.3.
2 State and prove properties for the multiplicative inverse in a field that are
analogous to the properties in 0.3.
3 Suppose F is a field and a, b ∈ F. Prove that (− a)(−b) = ab.
4 Suppose F is a field and a, b, c ∈ F, with b 6= 0 and c 6= 0. Prove that
ac a
= .
bc b
5 Suppose F is a field and a, b, c, d ∈ F, with b 6= 0 and d 6= 0. Prove that
a c ad − bc
− = .
b d bd
6 Suppose F is a field and a, b, c, d ∈ F, with b 6= 0, c 6= 0, and d 6= 0. Prove that
a c ad
÷ = .
b d bc
7 Suppose F is an ordered field and a, b, c, d ∈ F. Prove that if a < b and c ≤ d,
then a + c < b + d.
8 Suppose F is an ordered field and a, b, c, d ∈ F. Prove that if 0 ≤ a < b and
0 < c ≤ d, then ac < bd.
9 Suppose F is an ordered field and a, b ∈ F. Prove that if a < b and ab > 0, then
a −1 > b −1 .
10 Prove that if a and b are elements of an ordered field, then | ab| = | a||b|.

11 Prove that if a and b are elements of an ordered field, then | a| − |b| ≤ | a − b|.
12 Prove that every ordered field has at most one positive element whose square
equals 2 (where 2 is defined to be 1 + 1).
13 Suppose F is an ordered field. Prove that there does not exist i ∈ F such
that i2 = −1. (Thus the set of complex numbers, with its usual operation of
multiplication, cannot be made into an ordered field.)
14 Suppose F is the field of rational functions with coefficients in R. This means
p
that an element of F has the form q , where p and q are polynomials with real
p
coefficients and q is not the 0 polynomial. Rational functions q and rs are
declared to be equal if ps = rq, and addition and multiplication are defined in F
as you would naturally assume.
(a) Let P denote the subset F consisting of rational functions that can be written
p
in the form q , where the highest order terms of p and q both have positive
coefficients. Show that F is an ordered field with this definition of P.
(b) Show that F, with P defined as above, is not a complete ordered field.
12 The Real Numbers and Rn

B Construction of the Real Numbers:


Dedekind Cuts
You may intuitively think of a real number as a point on the real line (whatever that
is) or as an integer followed by a decimal point followed by an infinite string of
digits. Neither of those intuitive notions provides enough structure to easily verify
the properties of a complete ordered field. Even defining the sum and the product
of two real numbers can be difficult with those two approaches. Furthermore, these
approaches make it hard to give a rigorous verification of the completeness property.
We will construct the real numbers us-
In 1858 Richard Dedekind
ing what are called Dedekind cuts. This
(1831–1916) invented the simple but
clever construction uses only rational
rigorous construction of the real
numbers, with no prior knowledge as-
numbers that have been named after
sumed about real numbers.
him.

0.21 Definition Dedekind cut; D

A Dedekind cut is a nonempty subset D of Q with the following properties:


• D 6= Q;
• D contains all rational numbers less than any element of D; in other words,
if b ∈ D, then a ∈ D for all a ∈ Q with a < b;
• D does not contain a largest element.
The set of all Dedekind cuts is denoted by D .

The last bullet point above implies, for


Intuitively, a Dedekind cut is the set
example, that { a ∈ Q : a ≤ 3} is not a
of all rational numbers to the left of
Dedekind cut because it contains a largest
some point on the real line.
element (namely, 3). Note that all the
following examples of Dedekind cuts are defined only in terms of rational numbers.

0.22 Example Dedekind cuts


Each of the following sets is a Dedekind cut:
• { a ∈ Q : a < 3};
• { a ∈ Q : a < 0 or a2 < 2};
1 1 1
• {a ∈ Q : a < 1 + 1! + 2! +···+ n! for some positive integer n}.
The set in the first bullet point above consists of all rational numbers less than 3. If
we had already defined the real numbers, then we could √ say that the set in the second
bullet point consists of all rational numbers less than 2 and the set in the third bullet
point consists of all rational numbers less than e. However, the Dedekind cuts defined
above make sense even if we know nothing about the real numbers.
Section B Construction of the Real Numbers: Dedekind Cuts 13

As we will see, the set D of all


Dedekind defined his cuts slightly
Dedekind cuts can be given the structure
differently than is done here.
of a complete ordered field. Intuitively,
Specifically, Dedekind defined his
we identify a Dedekind cut with the real
cuts as ordered pairs of sets of
number that would be its right endpoint
rational numbers, with each number
on the real line if that made sense. In other
in the first set less than each number
words, think about the Dedekind cut in the
in the second set, and with the union
first bullet point in the previous example
of the two sets equaling Q. The
as identified with 3, the Dedekind cut in
approach taken here is a bit simpler,
the
√ second bullet point as identified with while still using Dedekind’s basic
2, and the Dedekind cut in the third bul-
idea.
let point as identified with e.
To give the set D of all Dedekind cuts the structure of a field, first we define
addition on D , along with the additive identity and additive inverses. You should
think about why these are the right definitions.

0.23 Definition sum of two Dedekind cuts

• Suppose C and D are Dedekind cuts. Then C + D is defined by

C + D = {c + d : c ∈ C, d ∈ D }.

• The Dedekind cut 0̃ is defined by

0̃ = { a ∈ Q : a < 0}.

• Suppose D is a Dedekind cut. Then − D is defined by

− D = { a ∈ Q : a < −b for some b ∈ Q with b ∈


/ D }.

The next result states that addition is well defined, that addition is commutative,
that addition is associative, that 0̃ is the additive identity, and that − D is that additive
inverse of D for each Dedekind cut D. These properties are part of what is needed to
make D into a field.

0.24 addition of Dedekind cuts

(a) If C and D are Dedekind cuts, then C + D is a Dedekind cut.


(b) C + D = D + C for all Dedekind cuts C, D.
(c) ( B + C ) + D = B + (C + D ) for all Dedekind cuts B, C, D.
(d) D + 0̃ = D for every Dedekind cut D.
(e) D + (− D ) = 0̃ for every Dedekind cut D.

The proof of the result above is left as an exercise.


14 The Real Numbers and Rn

The next step in making the set of all


The fourth bullet point in the
Dedekind cuts D into a field requires that
previous result states that 0̃ is the
we define a multiplication. A definition
additive identity for the set D of all
analogous to the definition for the sum
Dedekind cuts. Usually we call the
of two Dedekind cuts would define the
additive identity 0, but here we use
product of two Dedekind cuts C and D
the notation 0̃ to avoid confusion
to be the set of rational numbers less than
with the rational number 0.
or equal to cd for some c ∈ C, d ∈ D.
However, that definition does not work because each Dedekind cut contains negative
numbers with arbitrarily large absolute values. If we consider only Dedekind cuts C
and D that each contain at least one positive number, then clearly we should define

CD = { a ∈ Q : a ≤ cd for some c ∈ C, d ∈ D with c > 0, d > 0},

with similarly appropriate definitions for the other three cases concerning C and D.
For details, see the instruction before Exercise 4 in this section. That exercise ends
with the conclusion that D (with the operations of addition and multiplication as
defined) is a field.
Now that we have made D into a field, we want to make it into an ordered
field. Thus we must define the positive subset of D , which is done in the next
definition. To motivate this definition, think of the intuitive notion of a Dedekind cut
as corresponding to what should be its right endpoint.

0.25 Definition positive Dedekind cut

A Dedekind cut D is called positive if b > 0 for some b ∈ D.

Exercise 5 asks you to verify that the definition above satisfies the requirements
for the positive subset of a field (see 0.5). In other words, the definition above makes
D into an ordered field.
Now that D is an ordered field, the positive subset of D defines the meaning of
inequalities in the usual way (see 0.7). For example, C ≤ D means that D + (−C )
is positive or C = D. The next result shows that this ordering of D has a particularly
nice interpretation. Again, the proof is left as an exercise.

0.26 ordering of Dedekind cuts

Suppose C and D are Dedekind cuts. Then C ≤ D if and only if C ⊂ D.

Now we are ready to prove the main point about what we have been doing with
Dedekind cuts. Specifically, we will prove that the ordered field D of Dedekind cuts
is complete. The clean, easy proof of this result should be attributed to the cleverness
of the definition of Dedekind cuts.

0.27 completeness of D

The ordered field D is complete.


Section B Construction of the Real Numbers: Dedekind Cuts 15

Proof Suppose A is a nonempty subset of D that has an upper bound. We must


show that A has a least upper bound.
Let
B = {b ∈ Q : b ∈ D for some D ∈ A}.
In other words, B is the union of all the Dedekind cuts in A.
We will show that B is a least upper bound of A. For this to even make sense, we
must first verify that B is a Dedekind cut. The following bullet points provide that
verification:
• Clearly B is nonempty, because A is nonempty.
• To show that B 6= Q, we use the hypothesis that A has an upper bound. Because
that upper bound is a Dedekind cut and thus is not all of Q, there exists a ∈ Q
not in that upper bound. Thus for every D ∈ A, we see that a ∈/ D. Thus a ∈ / B.
Hence B 6= Q.
• The definition of B clearly implies that if b ∈ B, then a ∈ B for all a ∈ Q with
a < b.
• To show that B has no largest element, suppose b ∈ B. Then b ∈ D for some
D ∈ A. Because D is a Dedekind cut, b is not a largest element of D. Because
D ⊂ B, this implies that b is not a largest element of B. Thus B has no largest
element, completing the verification that B is a Dedekind cut.
Now it makes sense to show that B is a least upper bound of A. As usual, this
least-upper-bound proof consists of two parts, first showing that B is an upper bound
of A and then showing that B is less than or equal to all other upper bounds of A:
• Obviously D ⊂ B for every D ∈ A, and thus D ≤ B for every D ∈ A (by
0.26). Hence B is an upper bound of A.
• To show that B is a least upper bound of A, suppose C ∈ D is an upper bound
of A. Thus D ≤ C for every D ∈ A, which can be restated (by 0.26) as D ⊂ C
for every D ∈ A. Because B is the union of all D ∈ A, this implies that B ⊂ C.
In other words, B ≤ C. Hence B is a least upper bound of A, completing the
proof.

What is a real number? The result


Recall that R is defined to be a
above implies that we could think of R as
complete ordered field. The
D , which would mean that a real number
existence of a complete ordered field,
is a Dedekind cut. Although that view-
as proved in the previous result,
point is technically correct, you may find
means that the theorems proved in
it more useful to think of a real number in-
the rest of this book are not
tuitively as a point on the real line or as an
meaningless.
element of an abstract complete ordered
field.

A number x on the real line.


16 The Real Numbers and Rn

EXERCISES B
1 Prove 0.24 (the addition properties on the set of Dedekind cuts).
2 Prove that a Dedekind cut D is positive if and only if 0 ∈ D.
3 Prove 0.26. In other words, show that if C and D are Dedekind cuts, then
C ≤ D if and only if C is a subset of D.

For a Dedekind cut D, define the set difference Q \ D by


Q \ D = {r ∈ Q : r ∈
/ D}
and define D + and D − by
D + = { d ∈ D : d > 0} and D − = {r ∈ Q \ D : r ≤ 0}.
Think of the condition D + 6= ∅ as equivalent to D > 0. Now define the product
CD of two Dedekind cuts C and D as follows:
CD =
{cd : c ∈ C + , d ∈ D + } ∪ {q ∈ Q : q ≤ 0} if C + 6= ∅, D + 6= ∅;
{cr : c ∈ C, r ∈ Q \ D } if C + = ∅, D + 6= ∅;
{rd : r ∈ Q \ C, d ∈ D } if C + 6= ∅, D + = ∅;
{ a ∈ Q : a < rs for some r ∈ C − , s ∈ D − } if C + = ∅, D + = ∅.
You should think about why the definition above works to capture your intuitive
expectations for multiplication of Dedekind cuts.
4 (a) Prove that if C and D are Dedekind cuts, then CD (as defined above) is a
Dedekind cut.
(b) Let 1̃ = { a ∈ Q : a < 1}. Show that 1̃ is a multiplicative identity for the
set D of all Dedekind cuts.
(c) For D a Dedekind cut with D 6= 0̃, find a formula for a Dedekind cut D −1
such that DD −1 = 1̃.
(d) Prove that with the addition and multiplication we have defined, the set D
of all Dedekind cuts is a field.
5 Show that the definition of the positive subset of D as given in 0.25 satisfies the
requirements for an ordered field (see 0.5). In other words, show the following:
(a) If D is a Dedekind cut, then D is positive or D = 0̃ or − D is positive.
(b) If D is a positive Dedekind cut, then − D is not a positive Dedekind cut.
(c) If C and D are positive Dedekind cuts, then C + D and CD are positive.

6 Because D is an ordered field, the absolute value is defined on D (see 0.9).


Prove that | D | = D ∪ (− D ) for every Dedekind cut D.
Section C Supremum and Infimum 17

C Supremum and Infimum


Archimedean Property
Let Z denote the set of integers and Z+
denote the set of positive integers. Thus

Z+ ⊂ Z ⊂ Q ⊂ R,

where the last inclusion comes from the


result 0.11.
The Archimedean Property states that
if t is a real number, then there is a posi-
tive integer larger than t. This result surely
does not surprise you. What may surprise
you, however, is that this result cannot be
proved using only the properties of an or-
dered field. Completeness must be used in
every proof of the Archimedean Property
because there are ordered fields in which
this result fails (for an example, see Exer-
cise 9 in this section).
The death of Archimedes, as depicted
in a seventeenth-century painting.

0.28 Archimedean Property

Suppose t ∈ R. Then there is a positive integer n such that t < n.

Proof Suppose there does not exist a positive integer n such that t < n. This implies
that t is an upper bound of Z+. Because R is complete, this implies that Z+ has a
least upper bound, which we will call b.
Now b − 1 is not an upper bound of Z+ (because b is the least upper bound of
Z+ ). Thus there exists m ∈ Z+ such that b − 1 < m. Thus b < m + 1. Because
m + 1 ∈ Z+, this contradicts the property that b is an upper bound of Z+. This
contradiction completes the proof.

The next result gives a useful restatement of the Archimedean Property.

0.29 Archimedean Property

1
Suppose ε ∈ R and ε > 0. Then there is a positive integer n such that n < ε.

Proof In the first version of the Archimedean Property (0.28), let t = 1ε .

Now we have an important consequence of the Archimedean Property.


18 The Real Numbers and Rn

0.30 rational number between every two distinct real numbers

Suppose a, b ∈ R, with a < b. Then there exists a rational number c such that
a < c < b.

Proof First suppose a ≥ 0. By the Archimedean Property (0.29), there is a positive


integer n such that
1
< b − a.
n
Let n mo
A= m∈Z:a< .
n
By the Archimedean Property (0.28), there is a positive integer m such that an < m.
Thus A is a nonempty set of positive integers. Hence A has a smallest element,
which we will call M. Because M ∈ A, we have a < M n.
Now M − 1 ∈ / A (because M is the smallest element of A). Thus Mn−1 ≤ a,
which implies that
M 1
≤ a+
n n
< b.
Hence taking c = M n completes the proof in the case where a ≥ 0.
Now suppose a < 0. If b > 0, then take c = 0. If b ≤ 0, then apply the previous
case to find a rational number d such that −b < d < − a, then take c = −d.

Greatest Lower Bound


The concepts of upper bound and least upper bound played a key role in our devel-
opment of the notion of a complete ordered field. Now we make the situation more
symmetric by introducing the concepts of lower bound and greatest lower bound.

0.31 Definition lower bound

Suppose A ⊂ R. A number b ∈ R is called a lower bound of A if b ≤ a for


every a ∈ A.

0.32 Definition greatest lower bound

Suppose A ⊂ R. A number b ∈ R is called a greatest lower bound of A if both


the following conditions hold:
• b is a lower bound of A;
• b ≥ c for every lower bound c of A.

If a subset of R has a greatest lower bound, then the subset has a unique greatest
lower bound. The uniqueness follows from the same reasoning as for the uniqueness
of the least upper bound (see the comment after 0.17).
Section C Supremum and Infimum 19

0.33 Example greatest lower bounds


If
A1 = { a ∈ R : 3 < a < 5} and A2 = { a ∈ R : 3 ≤ a ≤ 5},
then every real number b with b ≤ 3 is a lower bound of A1 and of A2 . Thus 3 is the
greatest lower bound of both A1 and A2 .

The completeness property of the real numbers tells us that every nonempty
subset of R with an upper bound has a least upper bound. The result below is the
corresponding statement for lower bounds. Note that the upper bound property is
part of the definition of R, but the lower bound property below is a theorem.

0.34 existence of greatest lower bound

Every nonempty subset of R that has a lower bound has a greatest lower bound.

Proof Suppose A is a nonempty subset of R that has a lower bound b. Let


− A = {− a : a ∈ A}.
Then −b is an upper bound of − A, as you should verify. The completeness of R
implies that − A has a least upper bound, which we will call t. Now −t is a greatest
lower bound of A, as you should verify.

The terminology defined below has wide usage in many areas of mathematics.

0.35 Definition supremum and infimum

Suppose A ⊂ R. The supremum of A, denoted sup A, is defined as follows:



the least upper bound of A if A has an upper bound and A 6= ∅,

sup A = ∞ if A does not have an upper bound,
−∞ if A = ∅.

The infimum of A, denoted inf A is defined as follows:



the greatest lower bound of A if A has a lower bound and A 6= ∅,

inf A = −∞ if A does not have a lower bound,
∞ if A = ∅.

The term supremum, which comes from the same Latin root as the word superior,
should help remind you that sup A is trying to be the largest number in A (if
sup A ∈ A, then sup A is the largest number in A). Similarly, the term infimum,
which comes from the same Latin root as the word inferior, should help remind you
that inf A is trying to be the smallest number in A (if inf A ∈ A, then inf A is the
smallest number in A).
20 The Real Numbers and Rn

0.36 Example infimum and supremum

• If A1 = { a ∈ R : 3 < a < 5} and A2 = { a ∈ R : 3 ≤ a ≤ 5}, then

inf A1 = inf A2 = 3 and sup A1 = sup A2 = 5.

1
• If A = {1 − n : n ∈ Z+} = {0, 12 , 23 , 34 , . . . }, then inf A = 0 and sup A = 1.

The symbols ∞ and −∞ that appear in the definitions of supremum and infimum
do not represent real numbers. The equation sup A = ∞ is simply an abbreviation for
the statement A does not have an upper bound. Similarly, the equation inf A = −∞
is an abbreviation for the statement A does not have a lower bound.

Irrational Numbers
The completeness property of the real numbers implies the existence of a real number
whose square is 2.


0.37 existence of 2

There is a positive real number whose square equals 2.

Proof Let
b = sup{ a ∈ R : a2 < 2}.
The set { a ∈ R : a2 < 2} has an upper bound (for example, 2 is an upper bound)
and thus b as defined above is a real number.
If b2 < 2, then we can find a number slightly bigger than b in { a ∈ R : a2 < 2}
(see the second paragraph of Example 0.18 for this calculation), which contradicts
the property that b is an upper bound of { a ∈ R : a2 < 2}.
If b2 > 2, then we can find a number slightly smaller than b that is an upper bound
of { a ∈ R : a2 < 2} (see the third paragraph of Example 0.18 for this calculation),
which contradicts the property that b is the least upper bound of { a ∈ R : a2 < 2}.
The two previous paragraphs imply that b2 = 2, as desired.

The real number


√ b produced by the proof above is called the square root of 2 and
is denoted by 2.

0.38 Definition irrational number

A real number is called irrational if it is not rational.



For example, 0.12 shows that 2 is irrational. Other well-known irrational
numbers include e, π, and ln 2.
We showed previously that there is a rational number between every two distinct
real numbers (see 0.30). Now we can show that there is also an irrational number
between every two distinct real numbers.
Section C Supremum and Infimum 21

0.39 irrational number between every two distinct real numbers

Suppose a, b ∈ R, with a < b. Then there exists an irrational number c such that
a < c < b.

Proof By the Archimedean Property


Are the numbers
(0.29), there is a positive integer n such
that n1 < b − a. Let e + π, eπ,
π
 √ e
 a + 2 if a is rational,
c= 2n rational or irrational? No one has
a + 1 if a is irrational. been able to answer this question.
n
However, almost all mathematicians
Then c is irrational and a < c < b. suspect that all three of these
numbers are irrational.

Intervals
We will find it useful sometimes to consider a set (not a field) consisting of R and
two additional elements called ∞ and −∞. We define an ordering on R ∪ {∞, −∞}
to behave exactly as you expect from the names of the two additional symbols.

0.40 Definition ordering on R ∪ {∞, −∞}

• The ordering < on R is extended to R ∪ {∞, −∞} as follows:


a < ∞ for all a ∈ R ∪ {−∞};
−∞ < a for all a ∈ R ∪ {∞}.
• For a, b ∈ R ∪ {∞, −∞},
the notation a ≤ b means that a < b or a = b;
the notation a > b means that b < a;
the notation a ≥ b means that a > b or a = b.

The notation defined below is probably already familiar to you.

0.41 Definition interval notation

Suppose a, b ∈ R ∪ {∞, −∞}. Then


• ( a, b) = {t ∈ R : a < t < b};
• [ a, b] = {t ∈ R ∪ {∞, −∞} : a ≤ t ≤ b};
• ( a, b] = {t ∈ R ∪ {∞} : a < t ≤ b};
• [ a, b) = {t ∈ R ∪ {−∞} : a ≤ t < b}.
≈ 7π The Real Numbers and Rn

If a > b, then all four of the sets listed above are the empty set. If a = b, then
[ a, b] is the set { a} containing only one element and the other three sets listed above
are the empty set.
The definition above implies that (−∞, ∞) equals R and that (0, ∞) is the set
of positive numbers. Also note that [−∞, ∞] = R ∪ {∞, −∞} and that [0, ∞] =
[0, ∞) ∪ {∞}; thus neither [−∞, ∞] nor [0, ∞] is a subset of R.

0.42 Definition interval

• A subset of [−∞, ∞] is called an interval if it contains all numbers that are


between pairs of its elements.
• In other words, a set I ⊂ [−∞, ∞] is called an interval if c, d ∈ I implies
(c, d) ⊂ I.

The next result gives a complete description of all intervals of [−∞, ∞].

0.43 description of intervals

Suppose I ⊂ [−∞, ∞] is an interval. Then I is one of the following sets for some
a, b ∈ [−∞, ∞]:
( a, b), [ a, b], ( a, b], [ a, b).

Proof Let a = inf I and let b = sup I. Suppose s ∈ I. Then a ≤ s because a is a


lower bound of I. Similarly, s ≤ b. Thus s ∈ [ a, b]. We have shown that I ⊂ [ a, b].
Now suppose t ∈ ( a, b). Because a < t and a is the greatest lower bound of I,
the number t is not a lower bound of I. Thus there exists c ∈ I such that c < t.
Similarly, because t < b, there exists d ∈ I such that t < d. Hence t ∈ (c, d).
Because c, d ∈ I and I is an interval, we can conclude that t ∈ I. Hence we have
shown that ( a, b) ⊂ I.
We now know that
( a, b) ⊂ I ⊂ [ a, b].
This implies that I is ( a, b), [ a, b], ( a, b] or [ a, b).

EXERCISES C
1
1 Suppose b ∈ R and |b| < n for every positive integer n. Prove that b = 0.

2 Suppose A ⊂ B ⊂ R. Show that

inf A ≥ inf B and sup A ≤ sup B.

3 Explain why it makes no sense to inquire about whether your current height as
measured in meters is a rational number or an irrational number.
Section C Supremum and Infimum 23

4 Is the speed of light in a vacuum, as measured in meters per second, a rational


number or an irrational number?
5 Prove or give a counterexample: Suppose a1 , a2 , a3 , . . . is a sequence of rational
numbers such that √
sup{ a1 , a2 , a3 , . . .} = 2.
Then √
sup{ an , an+1 , an+2 , . . .} = 2
for every positive integer n.

For A and B nonempty subsets of R, define A + B by

A + B = { a + b : a ∈ A, b ∈ B}.
Define arithmetic with ∞ and −∞ as you would expect. For example, s + ∞ = ∞
for all s ∈ (−∞, ∞] and −∞ + t = −∞ for all t ∈ [−∞, ∞). Note, however,
that ∞ + (−∞) should remain undefined.
6 Prove that if A and B are nonempty subsets of R, then

sup( A + B) = sup A + sup B

and
inf( A + B) = inf A + inf B.

An expression such as sup f means sup{ f ( x) : x ∈ X }.


X
Similarly, inf f ( x) means inf{ f ( x) : x ∈ X }.
X

7 Suppose X is a nonempty set and f , g : X → R are functions.


(a) Prove that
sup( f + g) ≤ sup f + sup g.
X X X

(b) Give an example to show that the inequality above can be a strict inequality.

8 Suppose X is a nonempty set and f , g : X → R are functions.


(a) Prove that
inf( f + g) ≥ inf f + inf g.
X X X

(b) Give an example to show that the inequality above can be a strict inequality.

9 Prove that the ordered field of rational functions with coefficients in R (see
Exercise 14 in Section A for the definition of this ordered field) does not satisfy
the Archimedean Property.
24 The Real Numbers and Rn

10 (a) Suppose a1 , a2 , . . . is a sequence in R. Prove that

inf{ am , am+1 , . . .} ≤ sup{ an , an+1 , . . .}

for all positive integers m, n.


(b) Describe all sequences a1 , a2 , . . . in R such that the inequality above is an
equality for some positive integers m, n.
11 Suppose I ⊂ [−∞, ∞] is an interval. Prove that the number of elements in
I ∩ Q is 0 or 1 or is not finite.
12 Suppose A is a subset of Q that contains all the rational numbers that are
between pairs of its elements. Prove that there exists an interval I ⊂ R such
that A = I ∩ Q.
13 Prove or give a counterexample: The intersection of every collection of intervals
is an interval.
14 Prove or give a counterexample: The union of every collection of intervals with
a nonempty intersection is an interval.
15 Suppose I ⊂ R is an interval containing more than one number. Prove that

inf( I ∩ Q) = inf I and sup( I ∩ Q) = sup I.

16 Suppose R1 and R2 are complete ordered fields. Let ϕ1 : Q → R1 and


ϕ2 : Q → R2 be the functions that allow us to think of Q as a subset of
R1 and as a subset of R2 (see 0.11). Define ψ : R1 → R2 as follows: for
a ∈ R1 , let ψ( a) be the least upper bound in R2 of

{ ϕ2 (q) : q ∈ Q and ϕ1 (q) ≤ a}.

(a) Show that ψ is a well-defined, one-to-one function from R1 onto R2 .


(b) Show that ψ(0) = 0 and ψ(1) = 1.
(c) Show that ψ( a + b) = ψ( a) + ψ(b) for all a, b ∈ R1 .
(d) Show that ψ(− a) = −ψ( a) for all a ∈ R1 .
(e) Show that ψ( ab) = ψ( a)ψ(b) for all a, b ∈ R1 .
 −1
(f) Show that ψ( a−1 ) = ψ( a) for all a ∈ R1 with a 6= 0.
(g) Suppose a ∈ R1 . Show that a > 0 if and only if ψ( a) > 0.
[Items (a)–(g) above show that R1 and R2 are essentially the same as complete
ordered fields, with ψ providing the relabeling.]
Section D Open and Closed Subsets of Rn 25

D Open and Closed Subsets of Rn


Limits in Rn
Throughout the next two sections, assume that m and n are positive integers. Thus, for
example, 0.48 should include the hypothesis that n is a positive integer, but theorems
and definitions become easier to state without explicitly repeating this hypothesis.
We identify R1 with R, the real line. The set R2 , which you can think of as a
plane, is the set of all ordered pairs of real numbers. The set R3 , which you can
think of as ordinary space, is the set of all ordered triples of real numbers. The next
definition gives the obvious generalization to higher dimensions.

0.44 Definition Rn

Rn is the set of all ordered n-tuples of real numbers:

Rn = {( x1 , . . . , xn ) : x1 , . . . , xn ∈ R}.

Now we generalize to the setting of Rn the standard Euclidean distance from a


point in R2 or R3 to the origin. We also introduce another measurement of distance
that is a bit easier to manipulate.

0.45 Definition k·k ; k·k∞

For ( x1 , . . . , xn ) ∈ Rn , let
p
k( x1 , . . . , xn )k = x1 2 + · · · + x n 2

and
k( x1 , . . . , xn )k∞ = max{| x1 |, . . . , | xn |}.

The reason for using the subscript ∞ here will become clear when you get to
L p -spaces in Chapter 7 of Measure, Integration & Real Analysis. For now, note that
the triangle inequality

k x + yk∞ ≤ k x k∞ + kyk∞ for all x, y ∈ Rn

is an easy consequence of the definition of k·k∞ . The triangle inequality also holds
for k·k but its proof when n ≥ 3 is far from obvious; Measure, Integration & Real
Analysis contains two nice proofs of the triangle inequality for k·k (see 7.14 with
p = 2 and 8.15). Meanwhile, here we will use k·k∞ for simpler proofs. Note that if
n = 1, then k·k and k·k∞ both equal the absolute value |·|.
Now we are ready to define what it means for a sequence of elements of Rn to
have a limit. The intuition concerning limits is that if we go far enough out in a
sequence, then all the terms beyond that will be as close as we wish to the limit. You
should have seen limits in previous courses. Thus some key properties of limits are
left as exercises.
26 The Real Numbers and Rn

0.46 Definition limit

Suppose a1 , a2 , . . . ∈ Rn and L ∈ Rn. Then L is called a limit of the sequence


a1 , a2 , . . . and we write
lim ak = L
k→∞

if for every ε > 0, there exists m ∈ Z+ such that

k ak − Lk∞ < ε

for all integers k ≥ m.

The definition above implies that

lim ak = L if and only if lim k ak − Lk∞ = 0.


k→∞ k→∞

Because √
k x k∞ ≤ k x k ≤ nk x k∞ for all x ∈ Rn ,
we see that
lim ak = L if and only if lim k ak − Lk = 0.
k→∞ k→∞
We will need the following useful terminology.

0.47 Definition converge; convergent

A sequence in Rn is said to converge and to be a convergent sequence if it has a


limit.

The next result states that a sequence of elements of Rn converges if and only if it
converges coordinatewise. Thus questions about convergence of sequences in Rn can
often be reduced to questions about convergence of sequences in R. The proof of this
next result is left to the reader.

0.48 coordinatewise limits

Suppose a1 , a2 , . . . ∈ Rn and L ∈ Rn. For k ∈ Z+, let

( ak,1 , . . . , ak,n ) = ak ,

and let ( L1 , . . . , Ln ) = L. Then lim ak = L if and only if


k→∞

lim ak,j = L j
k→∞

for each j ∈ {1, . . . , n}.

You should show that each sequence in Rn has at most one limit. Thus the phrase
a limit in 0.46 can be replaced by the limit.
Section D Open and Closed Subsets of Rn 27

Open Subsets of Rn
If n = 3, then the open cube B( x, δ) defined below is the usual cube in R3 centered
at x with sides of length 2δ.

0.49 Definition open cube

For x ∈ Rn and δ > 0, the open cube B( x, δ) is defined by

B( x, δ) = {y ∈ Rn : ky − x k∞ < δ}.

As a test that you are comfortable with these concepts, be sure that you can verify
the following implication:

0.50 y ∈ B( x, δ) =⇒ B(y, δ − ky − x k∞ ) ⊂ B( x, δ).

In the next definition, we allow the endpoints of an open interval to be ±∞.

0.51 Definition open interval

A subset of R of the form ( a, b) for some a, b ∈ [−∞, ∞] is called an open


interval.

Note that if n = 1 and x ∈ R, then B( x, δ) is the open interval ( x − δ, x + δ).


Two equivalent definitions of open subsets of Rn are given below. The definition in
the first bullet point is more useful when generalizing to metric spaces. The definition
in the second bullet point is more useful when considering bases of topological
spaces.

0.52 Definition open subset of Rn

• A subset G of Rn is called open if for every x ∈ G, there exists δ > 0 such


that B( x, δ) ⊂ G.
• Equivalently, a subset G of Rn is called open if every element of G is
contained in an open cube that is contained in G.

Make sure you take the time to understand why the definitions given by the two bullet
points above are equivalent (you will need to use 0.50).
Open sets could have been defined using the open balls {y ∈ Rn : ky − x k < δ}
instead of the open cubes B( x, δ). These two possible approaches are equivalent
because every open cube contains an open ball with the same center, and every open
ball contains an open cube with the same center. Specifically, if x ∈ Rn and δ > 0
then

{y ∈ Rn : ky − x k < δ} ⊂ B( x, δ) ⊂ {y ∈ Rn : ky − x k < nδ},

as you should verify.


28 The Real Numbers and Rn

The union ∞
S
k=1 Ek of a sequence E1 , E2 , . . . of subsets of T
a set S is the set of
elements of S that are in at least one of the Ek . The intersection ∞ k=1 Ek is the set of
elements of S that are in all the Ek .
More generally, we can consider unions and intersections that are not indexed by
the positive integers.

0.53 Definition union and intersection

Suppose A is a collection of subsets of some set S.


[
• The union of the collection A, denoted E, is defined by
E∈A
[
E = { x ∈ S : x ∈ E for some E ∈ A}.
E∈A

\
• The intersection of the collection A, denoted E, is defined by
E∈A
\
E = { x ∈ S : x ∈ E for every E ∈ A}.
E∈A

0.54 Example union and intersection


∞  ∞
1 1
− 1k , 1k = {0}
[  \ 
k,1 − k = (0, 1) and
k =1 k =1

0.55 union and intersection of open sets

(a) The union of every collection of open subsets of Rn is an open subset of Rn.

(b) The intersection of every finite collection of open subsets of Rn is an open


subset of Rn.

Proof The proof of (a) is left to the reader.


To prove (b), suppose G1 , . . . , Gm are open subsets of Rn and x ∈ G1 ∩ · · · ∩ Gm .
Thus x ∈ Gj for each j = 1, . . . , m. Because each Gj is open, there exist positive
numbers δ1 , . . . , δm such that B( x, δj ) ⊂ Gj for each j = 1, . . . , m. Let

δ = min{δ1 , . . . , δm }.

Then δ > 0 and B( x, δ) ⊂ G1 ∩ · · · ∩ Gm . Thus G1 ∩ · · · ∩ Gm is an open subset


of Rn, completing the proof.

The conclusion of 0.55(b) cannot be strengthened to infinite intersections because


0.54 gives an example of a sequence of open subsets of R whose intersection is the
set {0}, which is not open.
Section D Open and Closed Subsets of Rn 29

The next definition helps us distinguish some sets as having the same number of
elements as Z+.

0.56 Definition countable; uncountable

• A set C is called countable if C = ∅ or if C = {c1 , c2 , . . .} for some


sequence c1 , c2 , . . . of elements of C.
• A set is called uncountable if it is not countable.

The following two points follow easily from the definition of a countable set.
• Every finite set is countable. This holds because if C = {c1 , c2 , . . . , cn }, then
C = {c1 , c2 , . . . , cn , cn , cn , . . . } (repetitions do not matter for sets).
• If C is an infinite countable set, then C can be written in the form {b1 , b2 , . . .}
where b1 , b2 , . . . are all distinct. This holds because we can delete any terms in
the sequence c1 , c2 , . . . that appear earlier in the sequence.
We will use the next result to prove our description of open subsets of R (0.59).

0.57 Q is countable

The set of rational numbers is countable.

Proof At step 1, start with the list −1, 0, 1. At step n, adjoin to the list in increasing
order the rational numbers in the interval [−n, n] that can be written in the form m n
for some integer m. Thus halfway through step 3, the list is as follows:
−1, 0, 1, −2, − 32 , −1, − 12 , 0, 21 , 1, 32 , 2, −3, − 83 , − 37 , −2, − 53 , − 34 , −1, − 23 , − 13 , 0.
Continue in this fashion to produce a sequence that contains each rational number,
completing the proof.
Deleting the entries in the list that already appear earlier in the list (shown above
in red) produces a sequence that contains each rational number exactly once.

As you probably know, R is uncountable (see 2.17 in Measure, Integration & Real
Analysis for a proof that may be new to you). Similarly, the set of irrational numbers
is uncountable. Thus there are more irrational numbers than rational numbers.
The following terminology will be useful.

0.58 Definition disjoint

A sequence E1 , E2 , . . . of sets is called disjoint if Ej ∩ Ek = ∅ whenever j 6= k.

The next result gives a complete description of the open subsets of R. As an


example of this result, the open set consisting of all real numbers that are not integers
equals the union of the disjoint sequence of open intervals
(0, 1), (−1, 0), (1, 2), (−2, −1), (2, 3), (−3, −2), . . . .
30 The Real Numbers and Rn

In the result below, some (perhaps infinitely many) of the open intervals may be
the empty set.

0.59 open subset of R is countable disjoint union of open intervals

A subset of R is open if and only if it is the union of a disjoint sequence of open


intervals.

Proof One direction of this result is easy: The union of every sequence (disjoint or
not) of open intervals is open [by 0.55(a)].
To prove the other direction, suppose G is an open subset of R. For each t ∈ G,
let Gt be the union of all the open intervals contained in G that contain t. A moment’s
thought shows that Gt is the largest open interval contained in G that contains t.
If s, t ∈ G and Gs ∩ Gt 6= ∅, then Gs = Gt (because otherwise Gs ∪ Gt would
be an open interval strictly larger than at least one of Gs and Gt and containing both
s and t). In other words, any two intervals in the collection of intervals { Gt : t ∈ G }
are either disjoint or equal to each other.
Because t ∈ Gt for each t ∈ G, we see that the union of the collection of intervals
{ Gt : t ∈ G } is G.
Let r1 , r2 , . . . be a sequence of rational numbers that includes every rational
number (such a sequence exists by 0.57). Define a sequence of open intervals
I1 , I2 , . . . as follows:

∅
 if rk ∈/ G,
Ik = ∅ if rk ∈ Ij for some j < k,

Grk if rk ∈ G and rk ∈ / Ij for all j < k.

If t ∈ G, then the open interval Gt contains a rational number (by 0.30) and thus
Gt = Ik for some positive integer k. Thus G is the union of the disjoint sequence of
open intervals I1 , I2 , . . . .

Closed Subsets of Rn

0.60 Definition set difference; complement

• If S and A are sets, then the set difference S \ A is defined to be the set of
elements of S that are not in A. In other words, S \ A = {s ∈ S : s ∈ / A }.
• If A ⊂ S, then S \ A is called the complement of A in S.

For example, the complement in R of the interval (−∞, 5) is [5, ∞).

0.61 Definition closed subset of Rn

A subset of Rn is called closed if its complement in Rn is open.


Section D Open and Closed Subsets of Rn 31

For example, the interval [1, 4] is a closed subset of R because its complement in
R is the open set (−∞, 1) ∪ (4, ∞).
Unlike doors, a subset of Rn need not be either open or closed. For example, the
interval (3, 7] is neither an open nor a closed subset of R.
Closed sets can be more complicated than open sets. There exist closed subsets of
R that are not the union of a sequence of intervals (an example follows from 2.76
and 2.80 of Measure, Integration & Real Analysis).
The following characterization of closed sets will frequently be useful.

0.62 characterization of closed sets

A subset of Rn is closed if and only if it contains the limit of every convergent


sequence of elements of the set.

Proof We will prove the contrapositive in both directions.


First suppose A is a subset of Rn such that some convergent sequence a1 , a2 , . . .
of elements of A has a limit L that is not in A. Because L = limk→∞ ak , for each
δ > 0 there exists k ∈ Z+ such that k L − ak k∞ < δ. Thus L ∈ Rn \ A and

B( L, δ) 6⊂ Rn \ A

for every δ > 0. Hence Rn \ A is not an open subset of Rn. Thus A is not a closed
subset of Rn, completing the proof in one direction.
To prove the other direction, now suppose A is a subset of Rn that is not closed.
Thus Rn \ A is not open. Hence there exists L ∈ Rn \ A such that

B( L, 1k ) 6⊂ Rn \ A

for every k ∈ Z+. Thus for each k ∈ Z+, there exists ak ∈ A such that

k L − ak k∞ < 1k .
The inequality above implies that the sequence a1 , a2 , . . . of elements of A has limit
L. Thus there exists a convergent sequence of elements of A whose limit is not in A,
completing the proof in the other direction.

The next result can be stated without


Augustus De Morgan (1806–1871)
symbols: The complement of a union is
became the first professor of
the intersection of the complements, and
mathematics at University College
the complement of an intersection is the
London when he was 22 years old.
union of the complements.

0.63 De Morgan’s Laws

Suppose A is a collection of subsets of some set X. Then


[ \ \ [
X\ E= ( X \ E) and X \ E= ( X \ E ).
E∈A E∈A E∈A E∈A
32 The Real Numbers and Rn

An element x ∈ X is not in E∈A E if and only if x is not in E for every


S
Proof
E ∈ A. Thus the first equality above holds.
An element x ∈ X is not in E∈A E if and only if x is not in E for some E ∈ A.
T

Thus the second equality above holds.

0.64 union and intersection of closed sets

(a) The intersection of every collection of closed subsets of Rn is a closed subset


of Rn.
(b) The union of every finite collection of closed subsets of Rn is a closed subset
of Rn.

Proof This result follows immediately from De Morgan’s Laws (0.63), the definition
of a closed set, and 0.55.

The conclusion in 0.64(b) cannot be strengthened to infinite unions because 0.54


gives an example of a sequence of closed subsets of R whose union is the interval
(0, 1), which is not closed.
The empty set ∅ and the whole space Rn are both subsets of Rn that are both
open and closed. The next result says that there are no other such sets.

0.65 sets that are both open and closed

The only subsets of Rn that are both open and closed are ∅ and Rn.

Proof Suppose A is a subset of Rn that is both open and closed. Suppose A 6= ∅


and A 6= Rn (this will be a proof by contradiction). Thus there exist a ∈ A and
b ∈ Rn \ A. Let
T = {t ∈ [0, 1] : (1 − t) a + tb ∈ A}
and let
s = sup T.
The set T is nonempty because 0 ∈ T; thus s ∈ [0, 1]. Actually s > 0 because A is
open and thus (1 − t) a + tb ∈ A for all t ∈ [0, δ) for some sufficiently small positive
number δ. Similarly, s < 1 because Rn \ A is open and thus (1 − t) a + tb ∈ Rn \ A
for all t ∈ (1 − δ, 1] for some sufficiently small positive number δ. Hence s ∈ (0, 1).
Let
c = (1 − s) a + sb.
If c ∈ A, then because A is open, T contains numbers slightly larger than s, which
contradicts the definition of s as an upper bound of T.
If c ∈ Rn \ A, then because R \ A is open, T contains no numbers close to s,
which contradicts the definition of s as the least upper bound of T.
Thus we arrive at a contradiction whether c ∈ A or c ∈ Rn \ A, completing the
proof.
Section D Open and Closed Subsets of Rn 33

EXERCISES D
1 Suppose a1 , a2 , . . . and c1 , c2 , . . . are convergent sequences in Rn. Prove that

lim ( ak + ck ) = lim ak + lim ck .


k→∞ k→∞ k→∞

2 Suppose a1 , a2 , . . . and c1 , c2 , . . . are convergent sequences in R. Prove that

lim ( ak ck ) = ( lim ak )( lim ck ).


k→∞ k→∞ k→∞

3 Prove or give a counterexample: If a1 , a2 , . . . and c1 , c2 , . . . are convergent


sequences in R and ck 6= 0 for each k ∈ Z+, then

lim ak
ak
lim = k→∞ .
k→∞ ck lim ck
k→∞

4 Suppose a ∈ R. Prove that there exists a sequence a1 , a2 , . . . of rational numbers


such that a = limk→∞ ak .
5 Suppose a ∈ R. Prove that there exists a sequence a1 , a2 , . . . of irrational
numbers such that a = limk→∞ ak .
6 Prove that the union of a sequence of countable sets is a countable set.
7 Suppose X is a set and w : X → [0, ∞) is a function such that
n n o
sup ∑ w(xk ) : n ∈ Z+ and x1 , . . . , xn are distinct elements of X < ∞.
k =1

Prove that { x ∈ X : w( x ) > 0} is a countable set.


8 Suppose G is an open subset of R. Prove that inf G ∈
/ G and sup G ∈
/ G.
9 Suppose F is a nonempty closed set of positive numbers. Prove that inf F ∈ F.
10 Suppose G is an open subset of Rn and F is a closed subset of Rn. Prove that
the set difference G \ F is an open subset of Rn.
11 Suppose F is a closed subset of Rn and G is an open subset of Rn. Prove that
the set difference F \ G is a closed subset of Rn.
12 Suppose a1 , a2 , . . . is a convergent sequence in Rn with limit L. Prove that
{ L} ∪ { a1 , a2 , . . .} is a closed subset of Rn.
13 Suppose A is a subset of R that does not contain 0. Let A−1 = { a−1 : a ∈ A}.
(a) Prove that A is open if and only if A−1 is open.
(b) Give an example of a closed subset A of R that does not contain 0 such that
A−1 is not a closed subset of R.
34 The Real Numbers and Rn

14 Prove or give a counterexample: If G is an open subset of R, then { a2 : a ∈ G }


is open.
15 Prove or give a counterexample: If F is a closed subset of R, then { a2 : a ∈ F }
is closed.
16 Prove or give a counterexample: If F is a subset of R such that { a2 : a ∈ F } is
closed, then F is closed.
17 Prove that a subset F of R is closed if and only if F ∩ [−n, n] is closed for every
n ∈ Z+.
18 Suppose F is a closed subset of R and Q ⊂ F. Prove that F = R.
19 Suppose F is a closed subset of R and (R \ Q) ⊂ F. Prove that F = R.
20 Prove that

{ a ∈ Rn : k a − bk∞ ≤ δ} and { a ∈ Rn : k a − bk ≤ δ}

are both closed subsets of Rn for every b ∈ Rn and every δ > 0.


21 Prove that every closed subset of R is the intersection of some sequence of open
subsets of R, each of which has the form (−∞, a) ∪ (b, ∞) for some a, b ∈ R.
22 Suppose X is a set and A, E are subsets of X. Show that
 
A ∩ E = X \ ( X \ A) ∪ ( X \ E) and A ∪ E = X \ ( X \ A) ∩ ( X \ E) .

23 Suppose F1 and F2 are disjoint closed subsets of R such that F1 ∪ F2 is an


interval. Prove that F1 = ∅ or F2 = ∅.
24 Suppose G1 , G2 , . . . is a disjoint sequence of open sets whose union is an interval.
Prove that Gk = ∅ for all k ∈ Z+ with at most one exception.
25 Construct a one-to-one function from R onto R \ Q.
26 Prove that if E is a subset of Rn and G is an open subset of Rn, then E + G
(which is defined to be { x + y : x ∈ E, y ∈ G }) is an open subset of Rn.
Section E Sequences and Continuity 35

E Sequences and Continuity


Bolzano–Weierstrass Theorem
We now define some types of sequences that will turn out to be especially useful.

0.66 Definition increasing; decreasing; monotone

A sequence a1 , a2 , . . . of real numbers is called


• increasing if ak ≤ ak+1 for every k ∈ Z+ ;
• decreasing if ak ≥ ak+1 for every k ∈ Z+ ;
• monotone if it is either increasing or decreasing.

0.67 Example monotone sequences

• The sequence 2, 4, 6, 8, . . . (here the kth term is 2k) is increasing.


• The sequence 1, 12 , 13 , 14 , . . . (here the kth term is 1k ) is decreasing.
• Both of the two previous sequences are monotone.
• The sequence
1, 12 , 3, 14 , 5, 61 , . . .
(here the kth term is k if k is odd and 1k if k is even) is neither increasing nor
decreasing; thus this sequence is not monotone.

The next definition should not be a surprise.

0.68 Definition bounded

• A set A ⊂ Rn is called bounded if sup{k ak∞ : a ∈ A} < ∞.


• A function into Rn is called bounded if its range is a bounded subset of Rn.
• As a special case of the previous bullet point, a sequence a1 , a2 , . . . of
elements of Rn is called bounded if sup{k ak k∞ : k ∈ Z+ } < ∞.

To test that you are comfortable with this terminology, make sure that you can
show that a set A ⊂ R is bounded if and only if A has an upper bound and A has a
lower bound.
As you should verify, if a sequence
Note the crucial role that the
of real numbers converges, then it is
completeness of the field of real
bounded. The next result states that the
numbers plays in the next proof.
converse is true for monotone sequences.
36 The Real Numbers and Rn

0.69 bounded monotone sequences converge

Every bounded monotone sequence of real numbers converges.

Proof Suppose a1 , a2 , . . . is a bounded monotone sequence of real numbers.


First we consider the case where a1 , a2 , . . . is an increasing sequence. Let
L = sup{ a1 , a2 , . . . }.
Suppose ε > 0. Then L − ε is not an upper bound of the sequence. Thus there exists
a positive integer m such that L − ε < am . Because the sequence is increasing, we
conclude that L − ε < ak for all integers k ≥ m. Thus | L − ak | = L − ak < ε for all
integers k ≥ m. Hence limk→∞ ak = L, completing the proof in this case.
Now consider the case where a1 , a2 , . . . is a bounded decreasing sequence. Then
the sequence − a1 , − a2 , . . . is a bounded increasing sequence, which converges by
the first case that we considered. Multiplying by −1, we conclude that a1 , a2 , . . .
converges, completing the proof.

0.70 Definition subsequence

A subsequence of a sequence a1 , a2 , a3 , . . . is a sequence of the form


ak1 , ak2 , ak3 , . . . ,
where k1 , k2 , k3 , . . . are positive integers with k1 < k2 < k3 < · · · .

0.71 Example subsequence


The sequence 2, 5, 8, 11, . . . (here the kth term is 3k − 1) has 2, 11, 26, 47, . . . as a
subsequence (here k j = j2 ).

The next result provides us with a powerful tool.

0.72 monotone subsequence

Every sequence of real numbers has a monotone subsequence.

Proof Suppose a1 , a2 , . . . is a sequence of real numbers. For m ∈ Z+, we say that


m is a peak of this sequence if am ≥ ak for all k > m.
First consider the case where the sequence a1 , a2 , . . . has an infinite number of
peaks. Thus there exist positive integers m1 < m2 < m3 < · · · such that m j is a
peak of this sequence for each j ∈ Z+. Because m j is a peak, we have am j ≥ am j+1
for each j ∈ Z+. Thus the subsequence am1 , am2 , am3 , . . . is decreasing.
Now consider the case where the sequence a1 , a2 , . . . has only a finite number
of peaks. Thus there exists m1 ∈ Z+ such that k is not a peak for every k ≥ m1 .
Because m1 is not a peak, there exists m2 > m1 such that am1 < am2 . Because
m2 is not a peak, there exists m3 > m2 such that am2 < am3 , and so on. Thus the
subsequence am1 , am2 , am3 , . . . is increasing.
Section E Sequences and Continuity 37

Now we come to a major theorem that has multiple important consequences.


Because of the results we have already proved about monotone sequences, the proof
below is quite short.

0.73 Bolzano–Weierstrass Theorem

Every bounded sequence in Rn has a convergent subsequence.

Proof Suppose b1 , b2 , . . . is a bounded sequence in Rn .


For n = 1, the desired result follows immediately from combining two results:
every sequence of real numbers has a monotone subsequence (0.72), and every
bounded monotone sequence converges (0.69).
For n > 1, first take a convergent subsequence of the sequence of first coordinates
of b1 , b2 , . . .. Then take a convergent subsequence of second coordinates of that
subsequence. Continue this process until taking a convergent subsequence of nth
coordinates. The result on coordinatewise limits (0.48) now gives a convergent
subsequence of b1 , b2 , . . ., as desired.

Plaque honoring Bernard Bolzano (1781–1848) in his native city Prague.


Bolzano proved the result above in 1817. However, this work did not become widely
known in the international mathematical community until the work of
Karl Weierstrass (1815–1897) about a half-century later.
CC-BY-SA Matěj Bat’ha

The next result is called a characterization of closed bounded sets because the
converse, although less important, is also true (see Exercise 3 in this section).

0.74 characterization of closed bounded sets

Suppose F is a closed bounded subset of Rn. Then every sequence of elements of


F has a subsequence that converges to an element of F.

Proof Consider a sequence of elements of F. Because F is a bounded set, this


sequence is bounded and thus has a convergent subsequence (by the Bolzano–
Weierstrass Theorem, 0.73). Because F is closed, the limit of this convergent
subsequence is in F (by 0.62).
38 The Real Numbers and Rn

Continuity and Uniform Continuity


Although Isaac Newton (1643–1727) and Gottfried Leibniz (1646–1716) invented
calculus in the seventeenth century, a rigorous definition of continuity did not arise
until the nineteenth century.

0.75 Definition continuity

Suppose A ⊂ Rm and f : A → Rn is a function.


• For b ∈ A, the function f is called continuous at b if for every ε > 0, there
exists δ > 0 such that
k f ( a) − f (b)k∞ < ε
for all a ∈ A with k a − bk∞ < δ.
• The function f is called continuous if f is continuous at b for every b ∈ A.

In the first bullet point above, continuity is defined at an element of the domain.
The second bullet point above establishes the convention that simply calling a function
continuous means that the function is continuous at every element of its domain.
The next result, whose proof is left as an exercise, allows us to think about
continuity in terms of limits of sequences.

0.76 continuity via sequences

Suppose A ⊂ Rm and f : A → Rn is a function. Suppose b ∈ A. Then f is


continuous at b if and only if

lim f (bk ) = f (b)


k→∞

for every sequence b1 , b2 , . . . in A such that lim bk = b.


k→∞

The concept of uniform continuity also evolved in the nineteenth century.

0.77 Definition uniform continuity

Suppose A ⊂ Rm . A function f : A → Rn is called uniformly continuous if for


every ε > 0, there exists δ > 0 such that

k f ( a) − f (b)k∞ < ε

for all a, b ∈ A with k a − bk∞ < δ.

The symbols appearing in the definition of continuity and in the definition of


uniform continuity are the same, but pay careful attention to the different order of
the quantifiers. In the definition above of continuity, δ can depend upon b, but in the
definition of uniform continuity, δ cannot depend upon b.
Section E Sequences and Continuity 39

Clearly every uniformly continuous function is continuous, but the converse is not
true, as shown by the following example.

0.78 Example a continuous function that is not uniformly continuous


Define f : R → R by f ( x ) = x2 . Then
f (n + n1 ) − f (n) = 2 + 1
n2
>2
for every n ∈ Z+. The inequality above implies that f is not uniformly continuous.

The following remarkable result states that for functions whose domain is a closed
bounded subset of Rm, continuity implies uniform continuity. This result plays a
crucial role in showing that continuous functions are Riemann integrable (see 1.11 in
Measure, Integration & Real Analysis).

0.79 continuity implies uniform continuity on closed bounded sets

Every continuous Rn -valued function on each closed bounded subset of Rm is


uniformly continuous.

Proof Suppose F is a closed bounded subset of Rm and g : F → Rn is continuous.


We want to show that g is uniformly continuous.
Suppose g is not uniformly continuous. Then there exists ε > 0 such that for each
k ∈ Z+, there exist ak , bk ∈ F with
1
k a k − bk k ∞ < k and k g( ak ) − g(bk )k∞ ≥ ε.
Because F is bounded, the sequence a1 , a2 , . . . is bounded. Thus by the Bolzano–
Weierstrass Theorem (0.73), some subsequence ak1 , ak2 , . . . converges to some limit
a. Because F is closed, we have a ∈ F (by 0.62).
Now
k a − bk j k∞ = k( a − ak j ) + ( ak j − bk j )k∞
≤ k a − a k j k ∞ + k a k j − bk j k ∞
1
< k a − ak j k∞ + kj ,

which implies that lim j→∞ bk j = a.


Because g is continuous at a and lim j→∞ ak j = a and lim j→∞ bk j = a, we
conclude that
lim g( ak j ) = g( a) and lim g(bk j ) = g( a).
j→∞ j→∞

Thus 
lim g( ak j ) − g(bk j ) = 0.
j→∞

The equation above contradicts the inequality k g( ak ) − g(bk )k∞ ≥ ε, which holds
for all k ∈ Z+. This contradiction means that our assumption that g is not uniformly
continuous is false, completing the proof.
40 The Real Numbers and Rn

Max and Min on Closed Bounded Subsets of Rn


If f : (2, 3) → R is the function defined by f ( x ) = x2 , then

sup{ f ( x ) : x ∈ (2, 3)} = 9 and inf{ f ( x ) : x ∈ (2, 3)} = 4.

However, there is no x in the domain of f such that f ( x ) = 9 or f ( x ) = 4. In


other words, this function f attains neither a maximum value nor a minimum value.
The next result shows that such behavior cannot happen for continuous functions on
closed bounded subsets of Rm.
In the following proof, the hypothesis that the domain is a closed bounded subset
of Rm is used in conjunction with the Bolzano–Weierstrass Theorem. Specifically,
because the domain is bounded, every sequence in the domain has a convergent
subsequence. Furthermore, because the domain is closed, the limit of any such
convergent subsequence is in the domain (by 0.62).

0.80 maximum and minimum attained on closed bounded sets

Every continuous real-valued function on each closed bounded subset of Rm


attains its maximum and minimum.

Proof Suppose F is a nonempty closed bounded subset of Rm and g : F → R is a


continuous function.
If sup{ g( x ) : x ∈ F } = ∞, then there exists a sequence a1 , a2 , . . . in F such
that g( ak ) > k for each k ∈ Z+. By the Bolzano–Weierstrass Theorem (0.73), some
subsequence ak1 , ak2 , . . . converges to some a ∈ F. Because g is continuous at a, this
implies that lim j→∞ g( ak j ) = g( a), which is impossible because g( ak j ) > k j for
each j ∈ Z+. This contradiction proves that sup{ g( x ) : x ∈ F } < ∞.
Now let a1 , a2 , . . . be a sequence in F such that

lim g( ak ) = sup{ g( x ) : x ∈ F }.
k→∞

By the Bolzano–Weierstrass Theorem (0.73), some subsequence of a1 , a2 , . . . con-


verges to some a ∈ F. Because g is continuous at a, the equation above implies
that
g( a) = sup{ g( x ) : x ∈ F }.
Thus g attains its maximum on F, as desired.
To prove that g also attains its minimum on F, use similar ideas or apply the result
about a maximum to the function − g.

0.81 Definition image

Suppose S, T are sets and g : S → T is a function. If A ⊂ S, then the image of


A under g, denoted g( A), is the subset of T defined by

g ( A ) = { g ( x ) : x ∈ A }.
Section E Sequences and Continuity 41

For example, if g : R → R is the function defined by g( x ) = cos x, then


g([ π2 , π ]) = [−1, 0].
In the next proof, the Bolzano–Weierstrass Theorem will again play a key role.

0.82 continuous image of a closed bounded set is closed and bounded

Suppose F is a closed bounded subset of Rm and g : F → Rn is continuous. Then


g( F ) is a closed bounded subset of Rn.

Proof By 0.80, g( F ) is bounded.


To prove that g( F ) is closed, suppose g( a1 ), g( a2 ), . . . is a convergent sequence,
where each ak ∈ F. Let t = limk→∞ g( ak ). By the Bolzano–Weierstrass Theorem
(0.73), some subsequence of a1 , a2 , . . . converges to some a ∈ F. Because g is
continuous at a, this implies that t = g( a). Thus t ∈ g( F ). By 0.62, this implies that
g( F ) is closed, as desired.

EXERCISES E
1 Prove that every convergent sequence of elements of Rn is bounded.
2 Prove that a sequence of elements of Rn converges if and only if every subse-
quence of the sequence converges.
3 Prove the converse of 0.74. Specifically, prove that if F is a subset of Rn with the
property that every sequence of elements of F has a subsequence that converges
to an element of F, then F is closed and bounded.
4 Define f : R → R as follows:

0 if a is irrational,

f ( a) = 1 if a is rational and n is the smallest positive integer
n
such that a = m

n for some integer m.

At which numbers in R is f continuous?


5 Prove 0.76, which characterizes continuity via sequences.
6 Show that the function f : (0, ∞) → R defined by f ( x ) = 1
x is not uniformly
continuous.
7 Suppose p ∈ (0, ∞). Show that the function f : R → R defined by f ( x ) = | x | p
is uniformly continuous if and only if p ∈ (0, 1].
8 Prove or give a counterexample: If f : R → R is a bounded continuous function,
then f is uniformly continuous.
9 Prove or give a counterexample: If f : (0, 1) → R is a bounded continuous
function, then f is uniformly continuous.
42 The Real Numbers and Rn

10 Prove that if A is a bounded subset of Rm and f : A → R is uniformly continu-


ous, then f is a bounded function.
11 Prove that if f : R → R is differentiable everywhere and f 0 is a bounded
function on R, then f is uniformly continuous.
12 Give an example of a uniformly continuous function f : [−1, 1] → R such that
f is differentiable at every element of [−1, 1] but f 0 is not a bounded function
on [−1, 1].
13 Prove or give a counterexample: If f : Rm → Rn is continuous and
1
k f ( x )k <
kxk
for all x ∈ Rm with k x k > 1, then f is uniformly continuous.
14 Prove or give a counterexample: The sum of two uniformly continuous functions
from Rm to Rn is uniformly continuous.
15 Prove or give a counterexample: The product of two uniformly continuous
functions from R to R is uniformly continuous.
16 Prove or give a counterexample: If f : R → (0, ∞) is uniformly continuous,
then the function 1f is uniformly continuous on R.

17 Prove or give a counterexample: If f , g : R → R are uniformly continuous


functions, then the composition f ◦ g : R → R is uniformly continuous.

If S, T are sets and f : S → T is a function and A ⊂ T, then the set f −1 ( A) is


defined by
f −1 ( A ) = { s ∈ S : f ( s ) ∈ A }.

18 Suppose h : Rm → Rn is a function. Prove that h is continuous if and only if


h−1 ( G ) is an open subset of Rm for every open subset G of Rn.
19 Suppose h : Rm → Rn is a function. Prove that h is continuous if and only if
h−1 ( F ) is a closed subset of Rm for every closed subset F of Rn.
20 Give an example of a decreasingTsequence G1 ⊃ G2 ⊃ · · · of nonempty open
bounded subsets of R such that ∞k=1 Gk = ∅.

21 Give an example of a decreasing sequence F1 ⊃ F2 ⊃ · · · of nonempty closed


subsets of R such that ∞ ∅.
T
k =1 Fk =
22 Suppose F1 ⊃ F2 ⊃ · · · isTa decreasing sequence of nonempty closed bounded
subsets of Rn. Prove that ∞k=1 Fk 6 = ∅.

23 Prove that every continuous real-valued function on each closed subset of R can
be extended to a continuous real-valued function on R. More precisely, prove
that if F is a closed subset of R and g : F → R is continuous, then there exists a
continuous function h : R → R such that g( x ) = h( x ) for all x ∈ F.
Section E Sequences and Continuity 43

24 Prove or give a counterexample: If G is a bounded open subset of R and


h : G → R is continuous, then h( G ) is an open subset of R.
25 Prove or give a counterexample: If F is a closed subset of R and h : F → R is
continuous, then h( F ) is a closed subset of R.
26 Suppose F is a subset of Rn such that every continuous real-valued function on
F attains a maximum. Prove that F is closed and bounded.
27 Suppose f : R → R is an increasing function [meaning that a < b implies
f ( a) ≤ f (b)]. Prove that there exists a countable set A ⊂ R such that f is
continuous at each element of R \ A.
28 Suppose a1 , a2 , . . . is a sequence of real numbers. For each k ∈ Z+, define
gk : R → R by (
0 if ak ≥ x,
gk ( x ) =
1 if ak < x.
Define f : R → R by

gk ( x )
f (x) = ∑ 2k
.
k =1
Suppose x ∈ R. Prove that f is continuous at x if and only if x ∈
/ { a1 , a2 , . . . }.
29 Prove that the continuous image of an interval is an interval. In other words,
prove that if f : [ a, b] → R is continuous and t is between f ( a) and f (b), then
there exists c ∈ [ a, b] such that f (c) = t.
[This result is called the Intermediate Value Theorem.]
30 Prove that every continuous function from R to R \ Q is a constant function.
31 Prove that every polynomial with odd degree has a real zero. In other words,
prove that if p : R → R is a polynomial with odd degree, then there exists
b ∈ R such that p(b) = 0.

A sequence a1 , a2 , . . . of elements of Rn is called a Cauchy sequence if for every


ε > 0, there exists m ∈ Z+ such that | a j − ak | < ε for all integers j and k
greater than m.
32 Prove that every convergent sequence of elements of Rn is a Cauchy sequence.
33 (a) Prove that every Cauchy sequence of elements of Rn is bounded.
(b) Prove that if some subsequence of a Cauchy sequence of elements of Rn
converges to some L ∈ Rn, then the Cauchy sequence has limit L.
(c) Prove that every Cauchy sequence of elements of Rn converges.
34 Prove that if F1 is a closed subset of Rn and F2 is a closed bounded subset of
Rn, then F1 + F2 (which is defined to be { x + y : x ∈ F1 , y ∈ F2 }) is closed.
35 Give an example of two closed subsets of R whose sum is not closed.
Photo Credits

• page 1: painting by Paul Barbotti in 1853; public domain image from Wikipedia;
verified on 10 July 2019 at
https://commons.wikimedia.org/wiki/File:Cicero_discovering_tomb_of_Archimedes_(_Paolo_Barbotti_).jpeg

• page 7: The School of Athens (detail) by Raphael; public domain image from
Wikipedia; verified on 10 July 2019 at
https://commons.wikimedia.org/wiki/File:%22The_School_of_Athens%22_by_Raffaello_Sanzio_da_Urbino.jpg

• page 17: painting by Salvator Rosa (1615–1673); public domain image from
Wikipedia; verified on 10 July 2019 at
https://commons.wikimedia.org/wiki/File:Morte_di_Archimede.JPG
• page 37: photo by Matěj Bat’ha; Creative Commons Attribution-Share Alike 2.5
Generic license; verified on 10 July 2019 at
https://commons.wikimedia.org/wiki/File:Pametni_cedule_Bernard_Bolzano.jpg

This Supplement (except for the images) is licensed under the terms of the
Creative Commons Attribution-NonCommercial 4.0 International License
(http://creativecommons.org/licenses/by-nc/4.0/), which permits any non-
commercial use, sharing, adaptation, distribution and reproduction in any
medium or format, as long as you give appropriate credit to the original
author and the source, provide a link to the Creative Commons license, and
indicate if changes were made.
The copyright status of the each image in this Supplement is indicated in
the corresponding photo credit entry above.

44

You might also like