You are on page 1of 30

Real Analysis∗

M.T.Nair

Contents

1 The set R of real numbers 2


1.1 What is lacking in Q? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Field, ordered field, and R . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Some properties of real numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Decimal expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Sequences in R and Completeness of R . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5.1 Convergence of sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5.2 Cauchy sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.5.3 Completeness property of R . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2 Metric spaces 19
2.1 Definition and examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2 Basic notions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3 Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.4 Complete metric spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

∗ The course F214: Elementary Real Analysis, August-December 2022 at BITS Pilani, Goa Campus.

1
1 The set R of real numbers

The set R of real numbers is an extension of the set Q of rational numbers with an additional crucial
property that Q does not have, namely the leat upper bound property.

1.1 What is lacking in Q?

Definition 1.1. A subset A of Q is said to be bounded above if there exists α ∈ Q such that x ≤ α
for every x ∈ A, and such α ∈ Q is called an upper bound of A. ♢

Note that, if A is a subset of Q, then A need not have an upper bound. Also, A can have
more than one upper bounds. For example, the set N ⊆ Q does not have an upper bound, whereas
{x ∈ Q : 0 ≤ x < 1} has infinitely many upper bounds.
Definition 1.2. Let A ⊆ Q. A number β ∈ Q is called a least upper bounded (lub) of A if β is
an upper bound, and it is the least among all the upper bounds of A, that is, if α is any upper bound
of A then β ≤ α. ♢

Example 1.3. For the set A = {x ∈ Q : 0 ≤ x < 1}, β = 1 is the least upper bound. ♢
Exercise 1.4. If A ⊆ Q has a least upper bound, then it is unique. ♢
Exercise 1.5. If A ⊆ Z is bounded above then it has a least upper bound in Q. ♢

ˆ Every subset of Q which is bounded above need not have a least upper bound.

We show that the set


A = {x ∈ Q : x > 0, x2 < 2}
does not have a least upper bound. For this, consider

B = {x ∈ Q : x > 0, x2 > 2}.

Clearly, A is bounded above (Is it clear?), and every upper bound of A in Q lies in B. We show that

(1) x ∈ A ⇒ ∃ y ∈ A such that x < y, and

(2) x ∈ B ⇒ ∃ y ∈ B such that y < x.

Once (1) and (2) are shown, then that we can show that A does not have a lub. The proof of this
statement is as follows:

Suppose A has a lub. Since α2 ̸= 2 (as α ∈ Q), either α ∈ A or α ∈ B. If α ∈ A, then by


(1) above α cannot be an upper bound. If α ∈ B, then α cannot be the lub. Hence, our
assumption that A has a lub is wrong.

Proof of (1) and (2). Let x ∈ Q with x > 0. Define

2 − x2
y =x+ .
2+x

2
Then y ∈ Q and
2(2 − x2 )
2 − y2 = .
(2 + x)2
From this we obtain (1) and (2).
Remark 1.6. One may wonder how one can guess of such a y which would work. Well, it is obtained
by some general procedure: We have to find some y ∈ A with y > x. So, we have to look for a y of
the form y = x + α for some rational α > 0. Thus, for an arbitrary α ∈ Q with α > 0, write y = x + α
and fix the value of α appropriately so that y ∈ A. For this note that
y 2 = x2 + 2xα + α2 .
Hence,
2 − x2
y 2 < 2 ⇐⇒ x2 + α(2x + α) < 2 ⇐⇒ α < .
2x + α
2−x2
Many choices of α would satisfy the last relation above. For example, α = 2x is a choice. The
2
choice α = 2−x
x+2 also works, since

2 − x2 2 − x2
< ⇐⇒ 2x + α < x + 2 ⇐⇒ x + α < 2
x+2 2x + α
2 − x2 2x + 2
⇐⇒ x+ < 2 ⇐⇒ <2
x+2 x+2
which is obviously true. ♢

Analogously, we can define the notion of lower bound and greatest lower bound of subsets of Q
and show that every set which is bounded below need not have a greatest lower bound. To make the
exposition complete let us define these notions.
Definition 1.7. A subset A of Q is said to be bounded below if there exists α ∈ Q such that α ≤ x
for every x ∈ A, and such α ∈ Q is called a lower bound of A. ♢

Note that, if A is a subset of Q, then A need not have a lower bound. Also, A can have more
than one lower bounds. For example, the set {n ∈ Z : n ≤ 0} does not have a lower bound, whereas
{x ∈ Q : 0 ≤ x < 1} has infinitely many lower bounds.
Definition 1.8. Let A ⊆ Q. A number β ∈ Q is called a greatest lower bounded (glb) of A if
β is a lower bound, and it is the greatest among all the upper bounds of A, that is, if α is any lower
bound of A then α ≤ β. ♢
Example 1.9. For the set B = {x ∈ Q : 0 ≤ x < 1}, β = 0 is the glb. ♢
Exercise 1.10. If A ⊆ Q has a glb, then it is unique. ♢
Exercise 1.11. If A ⊆ Z is bounded below then it has a glb in Q. ♢

ˆ Every subset of Q which is bounded below need not have a greatest lower bound.

As earlier, consider the sets


A = {x ∈ Q : x > 0, x2 < 2}, B = {x ∈ Q : x > 0, x2 > 2}.
Observe that B is bounded below and every lower bound of B in Q lies in A (Did you observe?). We
already know that:

3
(1) x ∈ A ⇒ ∃ y ∈ A such that x < y, and
(2) x ∈ B ⇒ ∃ y ∈ B such that y < x.

From (1) and (2), it follows (Does it?) that B does not have a glb.

1.2 Field, ordered field, and R


ˆ The operations of addition (x, y) 7→ x + y, and multiplication (x, y) 7→ xy on Q have the
properties listed in the following definition of a field. Further, Q is an ordered field with respect
to the order relation < in the sense that
1. ∀ x, y, z ∈ Q, x < y and y < z imply x < z,
2. ∀ x, y ∈ Q, one and only one of the following holds:

x < y, y < x, x = y,

3. ∀ x, y ∈ Q, x < y ∈ ⇒ x + z < y + z for all z ∈ Q,


4. ∀ x, y ∈ Q, x < y ∈ ⇒ xz < yz for all z ∈ Q with z > 0.

ˆ The set R of real numbers is an ordered field which is an extension of the ordered field Q such
that every bounded subset of R has a least upper bound.

Definition 1.12. (Field) A set F together with two binary operators, namely, addition:

(x, y) 7→ x + y

and multiplication:
(x, y) 7→ xy
which are functions from F × F to F, is called a field if they satisfy the following properties:

(A1) x+y =y+x ∀ x, y ∈ F (Commutative property of addition)


(A2) (x + y) + z = x + (y + z) ∀ x, y, z ∈ F (Associative property of addition)
(A3) ∃ θ ∈ F such that x + θ = x ∀x ∈ F (Existence of additive identity)
(A4) ∀ x ∈ F, ∃ x̃ ∈ F such that x + x̃ = θ (Existence of additive inverse)
(M1) xy = yx ∀ x, y ∈ F (Commutative property of multiplication)
(M2) (xy)z = x(yz) ∀ x, y, z ∈ F (Associative property multiplication)

(M3) ∃ θ̂ ∈ F such that xθ̂ = x ∀x ∈ F (Existence of multiplicative identity)

(M4) ∀ x ∈ F with x ̸= θ, ∃ x̂ ∈ F such that xx̂ = θ̂ (Existence of multiplicative inverse)


(D) x(y + z) = xz + yz ∀ x, y, z ∈ F (Distributive property)

The following can be proved easily (Exercise):

4
ˆ The additive identity and multiplicative identity are unique. That is, there exists a unique θ ∈ F
such that x + θ = x for all x ∈ F and there exists a unique θ̂ ∈ F such that xθ̂ = x for all x ∈ F.
ˆ The additive inverse and the multiplicative inverse are unique. That is, for each x ∈ F, there
exists a unique x̃ ∈ F such that x + x̃ = θ, and for each x ∈ F with x ̸= θ, there exists a unique
x̂ ∈ F such that xx̂ = θ̂.

The additive identity θ and the multiplicative identity θ̂ are usually denoted by 0 and 1, respectively.
Also, the additive inverse of x ∈ F is denoted by −x and the multiplicative inverse of a x ̸= 0 in F is
denoted by x−1 .
Example 1.13. The set Q of rational numbers is a field with respect to the usual addition and
multiplication. ♢
Exercise 1.14. For a n ∈ N, let Zn := {1, 2, . . . , n − 1} and for x, y ∈ Zn , let ⊕ and ⊗ be defined as
follows:
x ⊕ y = (x + y)(modn), x ⊗ y = xy(modn).
Then Zn with addition ⊕ and multiplication ⊗ is a field if and only if n is a prime number. ♢
Definition 1.15. Let S be a nonempty set. Then any subset of S × S is called a relation on S. ♢

If R is a relation on S and if (x, y) ∈ R, then we shall write xRy.


Definition 1.16. A relation R on a set S called an order relation if the following properties hold:

1. ∀ x, y, z ∈ S, xRy and yRz imply xRz.


2. ∀ x, y ∈ S, one and only one of the following holds:

xRy, yRx, x = y.

A set with an order relation is called an ordered set. ♢


Example 1.17. The sets N, Z, Q are ordered sets. ♢

An order relation is usually denoted by the symbol <. For x, y ∈ S, we write x ≤ y if either x < y
or x = y. Also, we write x ≥ y for y ≤ x and x > y for y < x.
Definition 1.18. A field F together with an order relation < is called an ordered field if x < y
implies

1. x + z < y + z ∀ z ∈ F,
2. xz < yz ∀ z > 0. ♢
Example 1.19. The filed Q of rational numbers is an ordered field. ♢
Definition 1.20. Let S be a set with an order relation <, and A ⊆ S.

1. x ∈ S is called an upper bound of A if a ≤ x for every a ∈ A.

2. A is said to be bounded above if an upper bound exists for A.

5
3. An element x ∈ S is called a least upper bound (lub) of A if x an upper bound and if y is
any upper bound of A, then x ≤ y; equivalently, u ≤ x for every u ∈ A, and if v < x, then there
exists w ∈ A such that v < w ≤ x.
4. x ∈ S is called a lower bound of A if x ≤ a for every a ∈ A.
5. A is said to be bounded below if a lower bound exists for A.
6. An element x ∈ S is called a greatest lower bound (glb) of A if x a lower bound and if y is
any lower bound of A, then y ≤ x; equivalently, x ≤ u for every u ∈ A, and if x < v, then there
exists w ∈ A such that x ≤ w < v. ♢
Exercise 1.21. Least upper bound (resp. greatest lower bound), if exists, is unique. ♢
Exercise 1.22. Let F be an ordered field. Then every subset which is bounded above has a lub if
and only if every subset which is bounded below has a glb. ♢

ˆ Least upper bound of A, if exists, is also called the supremum of A and is denoted by

sup A.

ˆ Greatest lower bound of A, if exists, is also called the infimum of A and is denoted by

inf A.

Definition 1.23. An ordered field F is said to have the least upper bound property if every
subset of F which is bounded above has a least upper bound. ♢
Definition 1.24. The set R of real numbers is an ordered field which is an extension of the ordered
field Q such thatit has least upper bound property. ♢

ˆ In the above definition, by an extension we mean that Q ⊂ R and the operations of addition,
multiplication and order in R matches with that of Q for elements of Q.

One may very well ask whether such an ordered field R exists?

We shall prove this fact in the Appendix to this section.

ˆ It is also true that such an ordered filed R is unique in the sense that any two such fields are
isomorphic. That is, if F is an ordered field with lub property, then then there exists a bijective
map T : R → F which preserves addition, multiplication and order.

We shall denote the operations of addition, multiplication and order relation on R by by the same
notations as in Q, that is,
x + y, xy, x < y,
respectively, for x, y ∈ R.
Remark 1.25. Let R be a relationn on S. Then R is called a partial order if the following are
satisfied:

1. (Reflexive): xRx for every x ∈ S.

6
2. (Antisymmetric): For x, y ∈ S, xRy and yRx imply x = y.
3. (Transitive): For x, y, z ∈ S, xRy and yRz imply xRz

A partial order R on S is called a total order or linear order if for every x, y ∈ S, either xRy or
yRy.
A partial order R on S is usually denoted by ⪯. If ⪯ ias a partial order, then for x, y ∈ S, we
write x ≺ y for x ⪯ y and x ̸= y. Suppose ⪯ is a total order on S. Then, it can be shown (Exercise)
that:
Exercise 1.26. Show that ⪯ is a total order iff ≺ is an order rlation. ♢

1.3 Some properties of real numbers

Thorem 1.27. (Archimedean property) If a > 0, b ∈ R, then there exists n ∈ N such that na > b.

Proof. Let a > 0, b ∈ R. Suppose there is no n ∈ N such that na > b. Then na ≤ b for all n ∈ N.
That is, S := {na : n ∈ N} is bounded above by b. Let β := sup S. Then β − a < β so that there
exists n ∈ N such that β − a < na. Thus, β < (n + 1)a, which contradicts the definition of β.
Thorem 1.28. (Densenes of Q) If a, b ∈ R with a < b, then there exists r ∈ Q such that a < r < b.

Proof. Let a, b ∈ R with a < b. Then b − a > 0. By the Archimedean property, there exists n ∈ N
such that n(b − a) > 1, that is, nb > na + 1. Now, by Archemedean property, there exist m1 , m2 ∈ N
such that m1 > na and m2 > −na so that

−m2 < na < m1 .

Hence, there exists m ∈ Z such that na < m < nb (Why?). Hence, there exists m ∈ Z such that

m − 1 ≤ na < m.

Thus,
na < m ≤ na + 1 < nb,
m
so that, using the fact that n > 0, we have a < n < b.

ˆ There exists a ∈ R such that x2 = 2.

To see this, consider the set A := {x ∈ Q+ : x2 ≤ 2}. Since A is bounded above, it has a lub, say
a ∈ R. Then, it follows (How?) that a2 = 2.
Definition 1.29. The elements of R \ Q are called irrational numbers. ♢

ˆ (Denseness of irrational numbers) If a, b ∈ R with a < b, then there exists x ∈ R \ Q such


that a < x < b.

√ √
To see this, let a, b ∈ R be such√that a < b. √
Then a/ 2 < b/ 2. hence,
√ by the denseness
√ of Q in
R, there exists r ∈ Q such that a/ 2 < r < b/ 2. This implies a < 2r < b. Note that 2r is an
irrational number.

7
Thorem 1.30. (nth root) For every a ∈ R with a > 0 and n ∈ N, there exists b ∈ R with b > 0
such that bn = a.

Proof. Let a ∈ R with a > 0 and n ∈ N. Let

S := {x ∈ R : x > 0 and xn < a}.

Then S is nonempty and bounded above (Exercise). Hence sup(S) exists in R. Let b := sup(S). It is
enough to show that (Why?)

(i) bn < a is not possible, and


(ii) bn > a is not possible.

Proof of (i): Suppose bn < a. Then for every ε > 0, we have

(b + ε)n − bn = ε[(b + ε)n−1 + (b + ε)n−2 b + · · · + (b + ε)bn−2 + bn−1 ]


≤ εn(b + ε)n−1 < nε(b + 1)n−1 if 0 < ε < 1.

Thus,
a − bn
(b + ε)n − bn < a − bn if 0<ε<1 and ε < .
n(b + 1)n−1
So, let
n a − bn o
0 < ε < min 1, .
n(b + 1)n−1
Then we have
(b + ε)n < a,
and there by b + ε ∈ S, contradicting the fact that b is an upper bound of S.
Proof of (ii): Suppose bn > a. Then for 0 < ε < b, we have

bn − (b − ε)n = ε[bn−1 + bn−2 (b − ε) + · · · + b(b − ε)n−2 + (b − ε)n−1 ]


≤ εnbn−1 .

Thus,
bn − a
bn − (b − ε)n < bn − a if 0<ε<b and ε < .
nbn−1
So, let
n bn − a o
0 < ε < min b, .
nbn−1
Then we have
a < (b − ε)n ,
and there by b − ε ̸∈ S, so that b − ε is an upper bound of S, contradicting the fact that b is the lub
of S.

ˆ Let a > 0 and n ∈ N. Then there exists b > 0 such that



n a if n is even,
(−b) =
−a if n is odd.

8
1.4 Decimal expansion

If m is a non-negative integer and m1 , m2 , . . . , mk ∈ {0, 1, . . . , 9}, then we denote


m1 m2 mk
m.m1 m2 · · · mk := m + + 2 + ··· k.
10 10 10
Definition 1.31. Let x be a positive real number. Then we say that

m.m1 m2 m3 · · ·

with mj ∈ {0, 1, . . . , 9} is a decimal expansion of x if

1
m.m1 m2 · · · mk ≤ x ≤ m.m1 m2 · · · mk + ∀ k ∈ N.
10k+1

Thorem 1.32. Every non-negative real number has a decimal expansion.

Proof. Let x be a non-negative real number. Let m ∈ N be the largest integer such that m ≤ x. If
m = x, then m.00 . . . is a decimal expansion of x. Suppoe m < x. Then let m1 be the largest integer
such that m + m10 ≤ x, that is,
1

n
m1 := sup{n ∈ Z : m + ≤ x}.
10
If m + m10 = x, then x = m.m1 00 · · · is a decimal expansion of x. Otherwise, let m2 be the largest
1

integer such that m + m m2


10 + 102 ≤ x, that is,
1

m1 n
m2 := sup{n ∈ Z : m + + ≤ x}.
10 10
If this procedure stops at a k-th step, then x = m.m1 m2 · · · mk 000 is a decimal expansion of x. It the
procedure does not stop, then x = m1 m2 m3 · · · is a decimal expansion of x.

1.5 Sequences in R and Completeness of R

Definition 1.33. Let S be any nonempty set. By a sequence in S we mean a function f : N → S,


and it is usually denoted by (xn ), where xn = f (n) for n ∈ N. ♢

In this section we shall consider the sequences in R.

1.5.1 Convergence of sequences

Definition 1.34. Let (xn ) be a sequence of real numbers. We say that (xn ) converges to x ∈ R if
for every ε > 0, there exists n0 ∈ N such that |xn − x| < ε for all n ≥ n0 , and in that case, we write

xn → x as n → ∞ or xn → x or lim xn = x.
n→∞

9
Example 1.35. Let xn = 1/n for n ∈ N. Then (xn ) converges to 0. To see this, note that, given any
ε > 0, we have
1 1
< ε ⇐⇒ n > .
n ε
Hence, we may take n0 = [1/ε] or any integer greater than [1/ε]. ♢
Thorem 1.36. The limit of a convergent sequence is unique.

Proof. Suppose xn → a and xn → b. Then, since

|a − b| ≤ |a − xn | + |xn − b| ∀ n ∈ N,

it follows that a = b.
Definition 1.37. A sequence is said to be a divergent sequence if it is not convergent. ♢
Example 1.38. The sequence ((−1)n ) is divergent. We show this by contradiction. Suppose (−1)n
converges to some x ∈ R. Let ε > 0 be any positive real number. Then there exits nε ∈ N such
that |x − (−1)n | < ε for all n ≥ nε . Hence, we have |x − (−1)2n | < ε and |x − (−1)2n+1 | < ε for all
n ≥ nε . That is, |x − 1| < ε and |x + 1| < ε for all n ≥ nε . Hence, taking 0 < ε ≤ 1 we obtain that
2 = |1 − (−1)| ≤ |1 − x| + |x + 1| < 2ε ≤ 2, which is a contradiction, ♢
Example 1.39. The sequence (n2 ) is divergent. We show this by contradiction. Suppose n2 → x for
some x ∈ R. Let ε > 0 be any positive real number. Then there exits N ∈ N such that |n2 − x| < ε
for all n ≥ N . This implies that

n2 ≤ |n2 − x| + |x| ∀ n ≥ N.

This is not possible. ♢

A simple observation on convergence is the following:


Thorem 1.40. Let (an ) and bn ) are sequences of real numbers such that 0 ≤ an ≤ bn for all n ∈ N.
If bn → 0, then an → 0.

Proof. Let ε > 0 be given. Since bn → 0, there exists N ∈ N such that

bn = |bn | = |bn − 0| < ε ∀ n ≥ N.

Since 0 ≤ an ≤ bn for all n ∈ N, it follows that

|an − 0| = an < bn < ε ∀ n ≥ N.

Hence, an → 0.

The following example illustrate the above theorem.


Example 1.41. Consider te sequence (e−n ). We know that n ≤ en so that e−n ≤ n1 for all n ∈ N.
Also, we know that n1 → 0 as n → ∞. Hence, by Theorem 1.40, it follows that e−n → 0 as n → ∞. ♢
1
Example 1.42. Let 0 < a < 1. Then an → 0. To see this, let us write a as a = 1+b . Then b > 0 and

1 1 1
an = ≤ ≤ .
(1 + b)n 1 + nb nb
From this we obtain that an → 0. ♢

10
ˆ Let (xn ) be a sequence in R and x ∈ R.

1. xn → x iff for every ε > 0 and for every α > 0, there exists N ∈ N such that |xn − x| ≤ αε
for all n ≥ N .
2. xn → x iff for every ε > 0, there exists N ∈ N such that xn ∈ (x − ε, x + ε) for all n ≥ N .

Exercise 1.43. Prove the following:

(1) Let (an ) and (bn ) are sequences of real numbers such that an ≤ bn for all n ∈ N. If an → a and
bn → b, then a ≤ b.
(2) Let (an ), (bn ) and (cn ) are sequences of real numbers such that an ≤ bn ≤ cn for all n ∈ N. If
there exists x ∈ R such that an → x and cn → x, then bn → x. ♢

From Exercise 1.43(2) we can derive the following (How??).

Thorem 1.44. Suppose xn → x and yn → y. Prove the following:

(i) xn + yn → x + y and xn yn → xy.


(ii) If y ̸= 0, then there exists N ∈ N such that yn ̸= 0 for all n ≥ N and xn /yn → x/y.

Proof.

ˆ Suppose (xn ) and (an ) are sequences in R such that |xn − x| ≤ an for all n ∈ N and an → 0.
Then xn → x.

Let us derive an interesting observation from the above.

Golden ratio: Recall that the Hemachandra-Fibonacchi sequence (an ) is defined by

an+2 = an+1 + an with a1 = 1 = a2 . (1)


an+2
Define b1 = 1 and bn+1 := an+1 for n = 2, 3, . . .. Thus, from (1), we have

1
bn+1 = 1 + , n ∈ N.
bn
Thus, if there exists b ∈ R such that bn → b, then, we must have
1
b=1+ . (2)
b
Clearly, √
1 2 1+ 5
b=1+ ⇐⇒ b − b − 1 = 0 ⇐⇒ b = .
b 2
Now, we show that (bn ) converges to √
1+ 5
b= ,
2
which is known as the golden ratio.

11
From (1) and (2), we have
1 1 b − bn
bn+1 − b = − = .
bn b bbn
Since bn ≥ 1 and b ≥ 3/2, it follows that
2
|bn+1 − b| ≤ |bn − a|.
3
From this we obtain  2 n
|bn+1 − b| ≤ |b1 − a| ∀ n ∈ N.
3
Since ( 32 )n → 0, it follows that bn → b.
Exercise 1.45. Let (xn ) be a sequence in R and x ∈ R. Prove the following:

1. If there exists r such that 0 < r < 1 such that |xn+1 − x| ≤ r|xn − x| for all n ∈ N, then xn → x.
2. Give an example with |xn+1 − x| < |xn − x| for all n ∈ N, but xn ̸→ x. ♢

Thorem 1.46. Let (xn ) be a sequence in R and x ∈ R. Then xn → x iff x2n → x and x2n+1 → x.

Proof. Suppose xn → x. We show x2n → x and x2n+1 → x. For this, let ε > 0 be given. Since
xn → x, there exists N ∈ N such that |xn − x| < ε for all n ≥ N . This implies that |x2n − x| < ε and
|x2n+1 − x| < ε for all n ≥ N . Hence, x2n → x and x2n+1 → x.
Conversely, suppose x2n → x and x2n+1 → x. We have to show that xn → x. For this, let ε > 0
be given. Since x2n → x and x2n+1 → x, there exist N1 , N2 ∈ N such that

|x2n − x| < ε ∀ n ≥ N1 and |x2n+1 − x| < ε ∀ n ≥ N2 .

Thus, if k = max{N1 , N2 }, then we have

|x2n − x| < ε and |x2n+1 − x| < ε ∀ n ≥ k,

so that |xn − x| < ε for all n ≥ 2k. This completes the proof.
Exercise 1.47. Let (xn ) be a sequence in R and x ∈ R. Prove: xn → x iff x3n → x, x3n+1 → x and
x3n+2 → x. More generally, xn → x iff for each k ∈ N, xkn+j → x for j = 0, 1, . . . , k. ♢
Exercise 1.48. From the above observation and Theorem 1.46, deduce that ((−1)n ) diverges. ♢
Exercise 1.49. For r ∈ R, let xn := 1 + r + · · · + rn . Show that (xn ) converges iff |r| < 1, and in
that case the limit is 1/(1 − r). ♢

Definition 1.50. A sequence (xn ) in R is said to be bounded if there exists M > 0 such that
|xn | ≤ M for all n ∈ N. ♢

Let (xn ) be a sequence of real numbers. Then observe the following:

ˆ (xn ) in R is bounded iff there exists M > 0 such that the set S := {xn : n ∈ N} is bounded.

ˆ (xn ) is bounded iff there exists M > 0 and N ∈ N satisfying |xn | ≤ M for all n ≥ N .

Thorem 1.51. Every convergent sequence is bounded. The converse is not true.

12
Proof. Suppose (xn ) be a convergent sequence with xn → x. Then, there exists n0 ∈ N such that
|xn − x| ≤ 1 for all n ≥ n0 . Hence,

|xn | ≤ |xn − x| + |x| ≤ 1 + |x| ∀ n ≥ n0 .

To see that the converse is not true, consider the sequence (−1)n ). It is bounded, but does not
converge.

1.5.2 Cauchy sequences

Suppose (xn ) is a sequence in a subset S of R. It can happen that (xn ) converges in R, but the limit
need not be in S. For example, the sequence (1/n) is in the interval (0, 1] := {x ∈ R : 0 < x ≤ 1}. It
converges in R to 0, but 0 ̸∈ (0, 1]. However, for a sequence (xn ) in S ⊂ R, if it converges in R, then
it has the property that for any two m, n, the terms xn and xm come closer and closer when n and m
become larger and larger. Such a sequence is called a Cauchy sequence.
Definition 1.52. A sequence (xn ) in R is said to be a Cauchy sequence if for every ε > 0, there
exists a positive integer n0 such that |xn − xm | < ε for all n, m ≥ n0 . ♢
Thorem 1.53. Every convergent sequence is a Cauchy sequence.

Proof. Suppose xn → x and ε > 0. Let n0 ∈ N be such that |xn − xm |⟨ε for all n, m ≥ n0 . Then

|xn − xm | ≤ |xn − x| + |x − xm | ≤ 2ε ∀ n, m ≥ n0 .

Exercise 1.54. Every Cauchy sequence is bounded. ♢

ˆ Let (xn ) be a sequence in R and x ∈ R. Then (xn ) is a Cauchy sequence in R iff for every ε > 0
and for every α > 0, there exists N ∈ N such that |xn − xm | ≤ αε for all n, m ≥ N .

In fact, if we have the following important property of R, which is called the completeness
property of R.

1.5.3 Completeness property of R

Definition 1.55. Let S be a subset of R and (xn ) be a sequence in S. We say that (xn ) converges
in S if there exits x ∈ S such that xn → x. ♢

We have already observed that a sequence (xn ) in a subset S of R may converge in R, but need
not converge in S. However, (xn ) is a Cauchy sequence in S.
Thorem 1.56. Suppose (xn ) is a sequence in [a, b] which converges to x ∈ R. Then x ∈ [a, b].

Proof. If x ̸∈ [a, b], then there exists ε > 0 such that (x−ε, x+ε)∩[a, b] = ∅. Since xn → x, there exists
N ∈ N such that xn ∈ (x−ε, x+ε) for all n ≥ N . This is a contradiction to (x−ε, x+ε)∩[a, b] = ∅.

Definition 1.57. A subset S of R is said to be complete if every Cauchy sequence in S converges


in S. ♢

13
We have already proved that every closed interval of the form [a, b] is complete.
Exercise 1.58. Intervals of the forms (a, b), (a, b], [a, b) are not complete. ♢

For a, b ∈ R, we denote
(a, ∞) := {x ∈ R : x > a},
[a, ∞) := {x ∈ R : x ≥ a},
(−∞, b) := {x ∈ R : x < b},
(∞, b] := {x ∈ R : x ≥ b}.
Exercise 1.59. Intervals of the forms (−∞, b) and (a, ∞) are not complete. ♢

Question: Is [a, b] complete?


We can answer this question affirmatively if we show that R is complete. In fact, it is true!
Thorem 1.60. Every Cauchy sequence in R converges.

Proof. Suppose (xn ) be a Cauchy sequence in R. Then, for every ε > 0, there exists nε ∈ N such that
|xn − xm | < ε for all n, m ≥ nε . In particular,
|xn − xnε | < ε ∀ n ≥ nε .
Hence, for any ε1 , ε2 > 0, we have
xnε1 − ε1 < xn < xnε2 + ε2 ∀ n ≥ max{ε1 , ε2 }. (1)
This shows that the set S := {xnε − ε : ε > 0} is bounded above. Let x := sup(S). Then we have
xnε − ε ≤ a < xnε + ε. (2)
From (1) and (2), we obtain |xn − x| ≤ 2ε for all n ≥ nε . From this, it follows that xn → x.
Example 1.61. The intervals [a, b], [a, ∞), (−∞b] are complete. This can be shown by using the
ideas used in the proof of Theorem 1.56. ♢
Exercise 1.62. Write details required in the above example. ♢
Exercise 1.63. Show that that sets Q and R \ Q are incomplete. ♢
Exercise 1.64. Let S ⊂ R and Se be the set S together with limits of all convergent sequences in S.
Then S is complete. ♢
Exercise 1.65. Show that the set {0} ∪ { n1 : n ∈ N} is complete. ♢
Thorem 1.66. (Nested interval theorem) If (In ) is a decreasing sequence of closed and bounded
intervals, then ∩∞ ∞
n=1 In is nonlempty. Further, if ℓ(In ) → 0, then ∩n=1 In is a singleton set.

Proof. Let In = [an , bn ], n ∈ N. Since In ⊇ In+1 for all n ∈ N, an ≤ an+1 ≤ bn+1 ≤ bn for all n ∈ N.
In particular, (an ) is bounded above by b1 and (bn ) is bounded below by a1 . Let
a := sup an , b = inf bn .
Then an ≤ an+1 ≤ a ≤ b ≤ bn+1 ≤ bn for all n ∈ N (Why a ≤ b?). Thus, a, b ∈ ∩∞
n=1 In . Clearly, if
ℓ(In ) → 0, then bn − an → 0 so that a = b, and an → a, bn → b.
Exercise 1.67. If E ⊆ R is a bounded infinite set, then E contains a convergent sequence with
distinct entries. ♢
Exercise 1.68. If (xn ) is monotonically increasing1 and bounded above, and if x := sup{xn : n ∈ N},
1 That is, xn ≤ xn+1 for all n ∈ N.

14
then (xn ) converges to x. ♢
Definition 1.69. A sequence (yn ) is called a subsequence of a sequence (xn ) if there exists a strictly
increasing sequence (kn ) of positive integers such that yn = xkn for all n ∈ N. ♢
Example 1.70. 1. The sequence (1/2n ) is a subsequence of (1/n).
2. The sequence (n4 ) is a subsequence of (n2 ). ♢
Example 1.71. The sequences (x2n ) and (x2n+1 ) are subsequences of (xn ). ♢
Thorem 1.72. A sequence (xn ) in R converges to x iff all its subsequences converges to x.

Proof. Suppose xn → x and (xkn ) be a subsequence of (xn ). Let ε > 0 be given. Since xn → x, there
exists N ∈ N such that
|xkn − x| < ε ∀ n ≥ N.

Conversely, if every subsequence of (xn ) converges to x, then since (xn ) is a subsequence of itself,
we obtain that xn → x.
Exercise 1.73. Let (xn ) be a sequence in R. Prove the following:

1. If (xn ) is a Cauchy sequence having a subsequence which converges to x, then (xn ) also converges
to x.
2. (xn ) converges to x iff every subsequence of (xn ) has a subsequence which converges.


Definition 1.74. Given a sequence (xn ) of real numbers, the expression of the form

X
x1 + x2 + · · · or xn
n=1

is called a series based on the sequence (xn ), and for each n ∈ N,

sn := x1 + x2 + · · · + xn
P∞
is called the n-th partial sum of of the series n=1 xn . ♢

X
Definition 1.75. We say that the series xn converges to s if the sequence (sn ) of partial sums
n=1

X ∞
X
of xn converges to s, and in that case we write xn = s.
n=1 n=1

If a series does not converge, then we say that it is a divergent series. ♢



X
Exercise 1.76. If xn converges then xn → 0.
n=1
Hint: (sn ) converges ⇒ it is a Cauchy sequence. ♢

X 1
Exercise 1.77. Show that the series converges to 1. ♢
n=1
2n

15
P∞ n
Exercise 1.78. The series diverges. Why?
n=1 n+1 ♢
P∞ n
P∞ n 1.79. Show that, for r ∈ R, the series n=1 r converges to iff |r| < 1, and in that case
Exercise
n=1 r = 1/(1 − r). ♢
P∞ P∞
Exercise 1.80. Suppose xn ≥ yn ≥ 0 for all n ∈ N. Prove that if n=1 xn converges, then n=1 yn
converges.
Hint: You may use the following two facts:

1. If (an ) is motonically increasing (i.e., an ≤ an+1 for all n ∈ N) and bounded above, then (an )
converges.
P∞
2. If an ≥ 0 for all n ∈ N, then the sequence (sn ) of partial sums of the series n=1 an is
monotonically increasing.


P∞ P∞
Exercise 1.81. Let xn ≥ yn ≥ 0 for all n ∈ N. Prove that if n=1 yn diverges, then n=1 xn
diverges.
Hint: Use the previous exercise. ♢
P∞ 1
Exercise 1.82. The series n=1 n2 converges.
Hint: Verify that s2n ≤ 1 + 1/2 + 1/22 + · · · + 1/2n−1 + 1/22n ≤ 2, and use the fact that the sequence
of its partial sums is monotonically increasing. ♢
P∞ 1
Exercise 1.83. If p ≤P1, then n=1 np diverges.

Hint: Show first that n=1 n1 diverges, by showing that s2n ≥ 1 + n2 . Then use one of the previous
exercises. ♢
xn+1
Exercise 1.84. (D’Alembert’s Ratio test) Let xn > 0 for all n ∈ N and limn→∞ xn = ℓ. Prove:

P∞
1. If ℓ < 1, then n=1 xn converges.
P∞
2. If ℓ > 1, then n=1 xn diverges.
P∞
3. If ℓ = 1, then n=1 xn can either converge or diverge.
P∞ n
Hint: (1) Choose ε > 0 such that ℓ + ε < 1 and use the convergence
P∞ of n=1n(ℓ + ε) .
(2) Choose ε > 0 such that ℓ − ε > 1 and use the divergence of n=1 (ℓ − ε) . ♢
1/n
Exercise 1.85. (Cauchy’s Root test) Let xn > 0 for all n ∈ N and limn→∞ xn = ℓ. Prove:
P∞
1. If ℓ < 1, then n=1 xn converges.
P∞
2. If ℓ > 1, then n=1 xn diverges.
P∞
3. If ℓ = 1, then n=1 xn can either converge or diverge.

Hint: Use analogous arguments as in D’Alembert’s atio test. ♢

16
Appendix

Thorem 1.86. There exists an ordered filed F with lub-property and an injective map T : Q → F such
that R(T ) is a subfield of F, satisfying

T (x + y) = T (x) + T (y) ∀ x, y ∈ Q,

T (xy) = T (x)T (y) ∀ x, y ∈ Q.

Note that, for every x ∈ Q,


T (x) = T (x.1) = T (x)T (1)

so that T (1) = 1 , the multiplicative identity in R(T ). Hence, if x ̸= 0 in Q, then

1∗ = T (1) = T (xx−1 ) = T (x)T (x−1 ),

so that T (x) ̸= 0 and []T (x)]−1 = T (x−1 ).



Let us describe the construction of R: As a motivation for the construction, let us understand 2
with the help of rational numbers. For this look at the sets:

A := {x ∈ Q : x < 0 or x2 < 2} and B = {x ∈ Q : x > 0 and x2 ≥ 2}.

Then we see that A and B are nonempty subsets of Q whose union is Q,√having empty intersection,
and that a < b for every a ∈ A and√ b ∈ B. In other words, assuming that 2 is known, A contains
√ all
the
√ rational numbers less than 2 and B contains all the rational numbers greater than 2. Thus,
2 is identified by the pair (A, B). Taking a cue from here, we define the following:
Definition 1.87. By a (Dedikind) cut in Q we mean an ordered pair (A, B) of non-empty subsets
of Q such that

1. A ∪ B = Q,
2. A ∩ B = ∅,
3. x < y for every (x, y) ∈ A × B, and
4. A does not contain a largest element.

ˆ If r ∈ Q, then taking Ar = {x ∈ Q : x < r} and Br = Q \ Ar , we see that (Ar ,r B) is a cut. We


shall denote such cuts by r∗ .

Let R be the set of all cuts in Q.

ˆ If α := (A, B) and β := (C, D) are cuts, then we write

α ≤ β ⇐⇒ A ⊆ C,

and
α < β ⇐⇒ A ⊊ C.

ˆ Let α, β ∈ R. Then we write α ≥ β iff β ≤ α, and we write α > β iff β < α.

17
ˆ A cut α ∈ R is said to be positive cut if α > 0∗ , and is said to be a negative cut if α < 0∗ .

ˆ A cut α ∈ R is said to be non-negative if α ≥ 0∗ .

Let C be a subcollection of R.

ˆ C is said to be bounded above if there exists β ∈ R such that α ≤ β for all α ∈ C, and in that
case β is called an upper bound of C.
ˆ An upper bound β0 ∈ R of C is said to be a least upper bound (lub) of C if β0 ≤ β for every
upper bound β of C.

Thorem 1.88. Every subset of R which is bounded above has a least upper bound.

Proof. Let A ⊆ R be bounded above. Let A0 be the set of all x ∈ Q such that x ∈ A for some cut
(A, B) ∈ A, and let B0 := Q \ A0 . Then we see that β0 := (A0 , B0 ) is a cut and β0 is an upper bound
for A. Suppose β := (C, D) be any upper bound for A. Then A ⊆ C for every cut (A, B) ∈ A. Then,
we have A0 ⊆ C so that β0 ≤ β.

ˆ Given a cut α := (A, B), we define

−α := {x ∈ Q : x = −b for some b ∈ Bwhich is not the smallest element in B}.

ˆ It can be seen that, for α ∈ R, α > 0∗ iff −α < 0∗ .

ˆ For α := (A, B), β := (C, D) ∈ R, define

α + β := {r ∈ Q : r = x + y fo some x ∈ A, y ∈ C}.

It can be verified that

ˆ α + 0∗ = α and α + (−α) = 0∗ for every α ∈ R.

ˆ If α := (A, B) and β := (C, D) are positive elements in R, then we define αβ = (E, F ), wehre

E := {r ∈ Q : r = xy for some x ∈ A, y ∈ C with x > 0∗ , y > 0∗ }, F = Q \ F.

ˆ It can be verified that α + β = β + α and αβ = βα for every α, β ∈ R.

ˆ For α, β ∈ R, if α > 0∗ and β < 0∗ , then we define

αβ := −[α(−β)].

ˆ For α, β ∈ R, if α < 0∗ and β < 0∗ , then we define

αβ := (−α)(−β).
Thorem 1.89. The set R is an ordered field with respect to the operations of addition, multiplication
and order defined above, which contains Q as a subfield.

18
2 Metric spaces

2.1 Definition and examples

Definition 2.1. Let X be a non-emplty set. A function d : X × X → R is said to be a metric on X


if it satisfies the following properties:

1. d(x, y) ≥ 0 ∀ x, y ∈ X;
2. x, y ∈ X d(x, y) = 0 ⇒ x = y;
3. x, y ∈ X ⇒ d(x, y) = d(y, x);

4. x, y, z ∈ X ⇒ (d(x, y) ≤ d(x, z) + d(z, x).

A set together with a metric on it is called a metric space. ♢

Example 2.2. (x, y) 7→ d(x, y) := |x − y| is a metric on R. ♢



0 if x = y,
Example 2.3. Let X be a non-empty set. For x, y ∈ X, let d(x, y) = Then d is
1 ̸ y.
if x =
metric on X, called the discrete metric. ♢
Example 2.4. For x = (x1 , . . . , xn )) and y = (y1 , . . . , yn ) in Rn , let
n
X
d(x, y) := ˜ y) := max{|xk − yk | : k = 1, . . . , k}.
|xk − yk | and d(x,
k=1

Then d and d˜ are metrics on Rn . ♢


Example 2.5. For z = (z1 , . . . , zn )) and w = (w1 , . . . , wn ) in Cn , let
n
X
d(z, w) := ˜ w) := max{|zk − wk | : k = 1, . . . , k}.
|zk − wk | and d(z,
k=1

Then d and d˜ are metrics on Cn . ♢


Example 2.6. Let X be either Rn or Cn for some n ∈ N. For x = (x1 , . . . , xn )) and y = (y1 , . . . , yn )
in X, let
Xn  21
d(x, y) := |xk − yk |2 .
k=1

We show that d is a metric on X. We note that, all cconditions, except, the triangle inequality, can
be verified easily. For proving the triangle inequality, we use

Thorem 2.7. (Cauchy-Schwarz inequality (CSE)) For x = (x1 , . . . , xn )) and y = (y1 , . . . , yn )


in X, which is Rn or Cn ,
n
X n
X n
 21  X  12
|xk yk | ≤ |xk |2 |yk |2 . (CSE)
k=1 k=1 k=1

19
Proof. For x = (x1 , . . . , xn ) ∈ X, let
n
X  21
∥x∥ := |xk |2 , (2.1)
k=1

called the norm of x. Clearly, if either ∥x∥ = 0 or ∥y∥ = 0, then (CSE) is satisfied trivially. So,
assume that ∥x∥ =
̸ 0 and ∥y∥ =
̸ 0. Now, using the relation
1 2
ab ≤ (a + b2 ) ∀ a, b ∈ R,
2
we obtain
|xk | |yk 1  |xk |2 |yk |2 
≤ + , k = 1, . . . , n.
∥x∥ ∥y∥ 2 ∥x∥2 ∥y∥2
Summing the above expressions, we have
n n n
X |xk | |yk 1  X |xk |2 X |yk |2 
≤ + = 1.
∥x∥ ∥y∥ 2 ∥x∥2 ∥y∥2
k=1 k=1 k=1

Thus, (CSE) is proved.

Form CSE, we deduce


Thorem 2.8. (Triangle inequality) For x = (x1 , . . . , xn )) and y = (y1 , . . . , yn ) in X, which is Rn
or Cn ,
X n  12 n
X  12  X n  21
|xk + yk | ≤ |xk |2 + |yk |2 . (TE)
k=1 k=1 k=1

Proof. We observe that For x = (x1 , . . . , xn )) and y = (y1 , . . . , yn ) in X


n
X n
X
|xk + yk |2 ≤ (|xk | + |yk |)|xk + yk |
k=1 k=1
Xn n
X
= |xk ||xk + yk | + |yk ||xk + yk |.
k=1 k=1

Hence, by CSE and the notation in (2.1),

∥x + y∥2 ≤ (∥x∥ + ∥y∥)∥x + y∥.

If ∥x + y∥ = 0, then (TE) is automatically satisfied. If ∥x + y∥ ̸= 0, then canceling out ∥x + y∥ from


the above inequality, we obtain ∥x + y∥ ≤ ∥x∥ + ∥y∥, which is nothing but the (TE).

Now, using the notation (2.1), we have the (TE)

∥x + y∥ ≤ ∥x∥ + ∥y∥ ∀ x, y ∈ X.

Hence, for any x, y, z ∈ X, we have

d(x, y) = ∥x − y∥ = ∥(x − z) + (z − y)∥ ≤ ∥x − z∥ + ∥z − y∥ = d(x, z) + d(z, y).

This proves triangle inequality. ♢

20
Example 2.9. Let X := C[a, b] be the set of all real valued continuous functions defined on [a, b].
Recall that every x ∈ C[a, b] is a bounded function and it attains a maximum at some point in [a, b],
Rb
and the Riemann integral a |x(t)|dt is well-defined.
For x, y ∈ C[a, b], let
Z b
d(x, y) := max |x(t) − y(t)| and d1 (x, y) := |x(t) − y(t)|dt.
a≤t≤b a

It can be seen that d and d1 are metrics on C[a, b] (Verify!)2 . ♢


Example 2.10. Let X := C[a, b] be the set of all real valued continuous functions defiend on [a, b].
Rb
Recall that every x ∈ C[a, b], the Riemann integral a |x(t)|2 dt is well-defined. For x, y ∈ C[a, b], let
Z b 1/2
d(x, y) := |x(t) − y(t)|2 dt .
a

We show that d is a metric on C[a, b]. It can be seen that all the conditions, except the triangle
inequality, follow easily. In order to prove the triangle inequality, we make use of the following two
theorems.
Thorem 2.11. (Cauchy-Schwarz inequality in C[a, b] (CSE)) For x, y ∈ C[a, b],
Z b Z b  21  Z b  21
|(x(t)y(t)| dt ≤ |x(t)|2 |dt |y(t)|2 dt . (CSE)
a a a

Proof. For x ∈ C[a, b], let


Z b  21
∥x∥ := |x(t)|2 dt , (2.2)
a

called the norm of x. We know (do you ??) that if either ∥x∥ = 0 or ∥y∥ = 0, then (CSE) is satisfied
trivially. So, assume that ∥x∥ =
̸ 0 and ∥y∥ =
̸ 0. Now, using the relation
1 2
ab ≤ (a + b2 ) ∀ a, b ∈ R,
2
we obtain
|x(t)| |y(t) 1  |x(t)|2 |y(t)|2 
≤ 2
+ ∀ t ∈ [a, b].
∥x∥ ∥y∥ 2 ∥x∥ ∥y∥2
Taking integral on both sides of the above, we have
Z b
1  b |x(t)|2
Z b
|x(t)| |y(t) |y(t)|2 
Z
dt ≤ dt + dt = 1.
a ∥x∥ ∥y∥ 2 a ∥x∥2 a ∥y∥
2

Thus, (CSE) is proved.


Thorem 2.12. (Triangle inequality in C[a, b] (TE)) For x, y ∈ C[a, b], and using the notation in
(2.2),
∥x + y∥ ≤ ∥x∥ + ∥y∥.

From the above we obtain the triangle inequality for d. ♢


2 To verify d (x, y) = 0 implies x = y, you may have to use the following result: If f : [a, b] → R is a continuous
1
function and t0 ∈ [a, b] is such that f (t0 ) ̸= 0, then there exists an interval I ⊆ [a, b] with distinct end points such that
t0 ∈ I and |f (t)| > |f (t0 )|/2 for all t ∈ I.

21
Example 2.13. For a non-empty subst S, let B(S) be the set of all real valued bounded functions
defined on S. Thus,
x ∈ B(S) ⇐⇒ ∃ M > 0 such that |x(s)| ≤ M ∀ s ∈ S.
If x ∈ B(S), sups∈S |x(s)| < ∞. Note that, if x, y ∈ B(S), the x − y, defined by
(x − y)(s) := x(s) − y(s), s ∈ S,
belongs to B(S). Now, for x, y in B(S), let
d(x, y) := sup{|x(s) − y(s)| : s ∈ S}.
Then d is a metric on B(S).
Taking S = N, we see that B(N) is the set of all bounded sequences. This metrc space B(N) is
usually denoted by ℓ∞ (N). ♢
1 1
P∞
Example 2.14. (Space ℓP(N) Let ℓ (N) be the set of all real sequences (an ) such that n=1 |an |
n
converges, that is, if sn := k=1 |ak |, then (sn ) converges. We see that, if x = (xn ) and y = (yn ) are
1 1
in ℓ (N), then x + y ∈ ℓ (N). Define
X
d(x, y) := |xn − yn |.
n=1

Then d is a metric on ℓ1 (N). ♢


2 1
P∞
Example 2.15. (Space ℓP (N) Let ℓ (N) be the set of all real sequences (an ) such that n=1 |an |2
n
converges, that is, if sn := k=1 |ak |2 , then (sn ) converges. For x = (xn ) and y = (yn ), let
X  21
d(x, y) := |xn − yn |2 .
n=1

We show that d is a metric on ℓ2 (N). All conditions, except the triangle inequality follow easily. To
prove the triangle inequality, let

X 1/2
∥x∥2 := |an |2 , x ∈ ℓ2 (N).
n=1

We first show that


∥x + y∥ ≤ ∥x∥ + ∥y∥ ∀ x, y ∈ ℓ2 (N).
So, let x = (xn ) and y = (yn ) be in ℓ2 (N). From triangle inequality in Rk , we have
k
X  21 k
X  12 k
X  12
|xj + yj |2 ≤ |xj |2 + |yj |2 ≤ ∥x∥2 + ∥y∥2 .
j=1 j=1 j=1
Pk
Hence, j=1 |xj + yj |2 ≤ (∥x∥2 + ∥y∥2 )2 . From this, letting k → ∞, we obtain,

X
|xj + yj |2 ≤ (∥x∥2 + ∥y∥2 )2 .
j=1

That is ∥x + y∥2 ≤ (∥x∥ + ∥y∥)2 , so that we obtain ∥x + y∥ ≤ ∥x∥ + ∥y∥. From this, we obtain the
triangle inequality, since
∥x − y∥ = ∥(x − z) + (z − y)∥ ≤ ∥x − z∥ + ∥z − y∥ ∀ x, y, z ∈ ℓ2 (N).
P 1
2 2
Thus, (x, y) 7→ d(x, y) := n=1 |xn − yn | is a metric on ℓ2 (N). ♢

22
Thorem 2.16. If S is a subset of a metric space X with metric d, then (x, y) 7→ d(x, y) defines a
metric on S.
Definition 2.17. If (X, d) is a metric space and S ⊆ X, then (S, d) is called a sub-metric space or
simple a subspace of (X, d). ♢
Example 2.18. 1. The metric space C[a, b] with sup-metric is a subspace of B[a, b].
2. Note that ℓ1 (N) is a subset of ℓ∞ (N). Hence, ℓ1 (N) is also a subspace of ℓ∞ (N) with respect to
the sup-metric. Similarly, ℓ2 (N) is a subset of ℓ∞ (N) so that ℓ2 (N) is also a subspace of ℓ∞ (N)
with respect to the sup-metric. ♢
Example 2.19. Let c00 (N) be the set of all real sequences each having only a finite number of non-zero
entries. For x = (xn ), y := (yn ) in c00 (N), and define

X
d1 (x, y) := |xn − yn |,
n=1
X∞  21
d2 (x, y) := |xn − yn |2 ,
n=1
d∞ (x, y) := sup{|xn − yn | : n ∈ N}.

Then d1 , d2 , d∞ are metrics on c00 (N). ♢

2.2 Basic notions

Definition 2.20. Let (X, d) be a metric space. For x ∈ X and r > 0, the set

B(x, r) := {y ∈ X : d(y, x) < r}

is called an open ball with centre x and radius r > 0. ♢


Example 2.21. Here a re a few examples of open balls:

1. Let X = R with usual metric. Every open interval (a, b) is an open ball with centre x0 as its
mid point, that is, x0 = a+b
2 and radius r := (b − a)/2.

2. Let X = C with usual metric. Then the unit circle {z ∈ C : |z| < 1} is an open ball with centre
0 and radius 1. ♢
Definition 2.22. Let (X, d) be a metric space and A ⊆ X.

1. A is called an open set if for every x ∈ A, there exists r > 0 such that B(x, r) ⊆ A.
2. A is called a closed set if Ac := X ⊆ A is an open set.
Example 2.23. Let X = R with usual metric. Then every open interval is an open set, and every
closed interval is a closed set. However, semiclosed interval of the forms [a, b) and (a, b] are neither
closed nor closed. ♢
Example 2.24. Let X = R with usual metric. Then Z is closed but not open in X. ♢
Example 2.25. If X is with discrete metric, then every singleton set is open as well as closed, and
every subset of X is open and closed. ♢

23
Example 2.26. In any metric space X, every finite subset is closed. ♢

ˆ Union of open sets is open: To see this, suppose G is a family of open sets, and A be the union
of sets in G. That is,
A = {x ∈ X : x ∈ G for slome G ∈ G}.
Let x ∈ A, and let G ∈ G such that x ∈ G. Then, since G is open, there exists r > 0 such that
B(x, r) ⊆ G. Hence, B(x, r) ⊆ A. This is true for each x ∈ A. Hence, A is an open set.

Example 2.27. Let X = N with usual metric, that is, d(n, m) = |n − m| for all n, m ∈ N. Then for
0 < r < 1,
B(n, r) = {m ∈ N : |m − n| < r} = {n}.
Hence, every subset of N is an open set. Hence, every subset of N is closed as well. ♢
Definition 2.28. Let (X, d) be a metric space, A ⊆ X and x ∈ X.

1. x is called an interior point of A if B(x, r) ⊆ A for some r > 0.

2. x is called a closure point of A if B(x, r) ∩ A ̸= ∅ for every r > 0.


3. x is called a limit point or accumulation pointof A if for every r > 0, B(x, r) ∩ (A \ {x}) ̸= ∅.
4. x is called a boundary point of A if for every r > 0, B(x, r) ∩ A ̸= ∅ and B(x, r) ∩ Ac ̸= ∅. ♢

ˆ Let (X, d) be a metric space and A ⊆ X. Then A is open iff every point in A is an interior point
of A.

Thorem 2.29. Let (X, d) be a metric space, A ⊆ X. Then the following are equivalent:

(i) A is a closed set.

(ii) A contains all its closure points.


(iii) A contains all its limit points.
(iv) A contains all its boundary points.

Proof. (i) ⇒ (ii): Assume (i), that is, A is a closed set. To show that it contains all its closure points.
Let x be a closure point of A. If x ̸∈ A, and since Ac is open, there exists r > 0 such that B(x, r) ⊆ Ac
so that B(x, r) ∩ A = ∅. This contradicts the fact that x is a closure point of A.
(ii) ⇒ (iii): Assume (ii), that is, A contains all its closure points. Let x be a limit point of A.
Since every limit point is a closure point, (iii) follows.
(iii) ⇒ (iv): Assume (iii), that is, A contains all its limit points. Let x be a boundary point of
A. Then for every r > 0, B(x, r) ∩ A ̸= ∅. Hence, if x ̸∈ A, then for every r > 0, B(x, r) contains
some point of A other than x so that x is limit point of A. Then, by (iii) x ∈ A. This leads to a
contradiction.
(iv) ⇒ (i): Assume (iv), that is A contains all its boundary points. To show that Ac is open. Let
x ∈ Ac . We have to show that there exists r > 0 such that B(x, r) ⊆ Ac . Suppose this is not true.
Then every open ball centered at x contains some point of A, and in that case, x is a boundary point
of A, which implies by (iv), x ∈ A, which leads to a contradiction.

24
Definition 2.30. Let (X, d) be a metric space and A ⊆ X.

1. The set of all interior points of A is called the interior of A, and it is denoted by Ao or int(A)
or intX (A).
2. The set of all closure points of A is called the closure of A, and it is denoted by Ā or cl(A) or
clX (A).
3. The set of all boundary points of A is called the boundary of A, and it is denoted by ∂A or
bd(A). ♢
Exercise 2.31. Let A ⊆ X. Show:

1. A is open iff every point in A is an interior point of A.


2. A is closed iff A is the closure of A. iff A contains all its boundary points. ♢
Exercise 2.32. Let (X, d) be a metric space and A ⊆ X.

1. Ā = A ∪ ∂A

2. A is open iff A = Ao .
3. A is closed iff A = Ā. ♢
Exercise 2.33. Let (X, d) be a metric space.

1. Every open ball in X is an open set.


2. Union of every arbitrary collection of open sets in X is open in X.

3. Finite intersection of opens sets in X is open in X.


4. Inersection of every arbitrary collection of closed sets in X is closed in X.
5. Finite union of of closed sets in X is closed in X. ♢
Definition 2.34. Let (X, d) be a metric space. The collection Td of all open subsets of X is called
the topology on X induced by the metric d. ♢
Definition 2.35. Let (X, d) be a metric space. A set S ⊆ X is said to be dense in X if S = X. ♢
Example 2.36. The set Q of rational numbers is dense in R with respect to the usual metric; so too
the set R \ Q of all irrational numbers. ♢
Example 2.37. Let X = [0, 1] with usual metric. Then the sets (0, 1], [0, 1), (0, 1) and [a, 1] ∩ Q are
dense in X. ♢

2.3 Convergence

Definition 2.38. Let (X, d) be a metric space and (xn ) be a sequene in X. We say that (xn )
converges to x ∈ X if d(xn , x) → 0 as n → ∞ and in that case we write

xn → x as n → ∞ or xn → x or lim xn = x. ♢
n→∞

25
Exercise 2.39. Let (X, d) be a metric space and (xn ) be a sequene in X. Then the following are
equivalent:

1. xn → x.
2. For every r > 0, there exists N ∈ N such that xn ∈ B(x, r) for all n ≥ N .
3. For every open set U containing x, there exists N ∈ N such that xn ∈ U for all n ≥ N . ♢

Exercise 2.40. Let X be with discrete metric.

1. Every subset of X is open and closed.

2. A sequence in X is convergent iff it is eventually constant.


3. A sequence in X is a Cauchy sequence iff it is eventually constant. ♢
Exercise 2.41. Let X be with discrete metric, (xn ) be a sequence in X and x ∈ X.

1. xn → x iff every subsequenece of (xn ) converges to x.


2. If (xn ) is a sequence, then xn → x iff (xn ) has a subsequence which converges to x.

3. A subset A of X is closed iff for every (xn ) in A, if xn → x for some x ∈ X, then x ∈ A. ♢


Exercise 2.42. Let (X, d) be a metric space and A ⊆ X. Then A dense in X iff for every x ∈ X,
there exists (xn ) in A such that xn → x. ♢
Definition 2.43. Suppose d and ρ are metrics on a set X.

1. The metric d is said to be weaker than ρ if there exists some c > 0 such that

d(x, y) ≤ cρ(x, y) ∀ x, y ∈ X,

and in that case ρ is said to be stronger than d.


2. The metrics d and ρ are said to be equivalent if there exists c1 > 0 and c2 > 0 that

c1 d(x, y) ≤ ρ(x, y) ≤ c2 d(x, y) ∀ x, y ∈ X. ♢


Exercise 2.44. Suppose d and ρ are equivalent metrics on a set X. Show that the topologies induced
by d and ρ are the same, that is, Td = Tρ . ♢

2.4 Complete metric spaces

Definition 2.45. Let (X, d) be a metric space. A sequence (xn ) in X said to be a Cauchy sequence
if for every ε > 0, there exists N ∈ N such that

d(xn , xm ) < ε ∀ n, m ≥ N. ♢

Definition 2.46. A metric space X is said to be a complete metric space if every Cauchy sequence
in X converes to some point in X. ♢

Example 2.47. 1. R with d(x, y) = |x − y| is complete.

26
2. Q with (x, y) 7→ |x − y| is not complete.
3. R2 is complete with the metrics
˜ y) := max{|x1 − y1 |, |x2 − y2 |}.
d(x, y) := |x1 − y1 | + |x2 − y2 | and d(x,

4. Any non-empty set with discrete metric is complete. ♢


Thorem 2.48. 1. Every closed subset of a complete metric space is complete.
2. Every complete subset of a metric space is closed.

Proof. Let (X, d) be a metric space and A ⊆ X.


1. Suppose X is complete and A is closed in X. Let (xn ) be a Cauchy sequence in A. Since (xn )
is also a Cauchy sequence in X, and since X is complete, there exists x ∈ X such that xn → x. Since
A is closed, it follows that x ∈ A. Thus, A is complete.
2. Suppose A is a complete subset of X. Let (xn ) be any sequence sequence in A which converges
to some x ∈ X. Since xn → x, (xn ) is a Cauchy sequence in A. Since A is complete, there exists
y ∈ A such that xn → y. Then we have x = y ∈ A. Thus, A is closed.
Corollary 2.49. Suppose X1 , X2 are metric spaces with metrics d1 , d2 , respectively, and X1 ⊆ X2
and d2|X1 ×X1 = d1 . If X1 is not closed in X2 , then X1 is not complete.

Example 2.50. For any set Ω ̸= ∅, B(Ω) is complete with respect to the sup-metric

d(f, g) := sup |f (s) − g(s)|, f, g ∈ B(Ω).


s∈Ω

Let (fn ) be a Cauchy sequence in B(Ω). Then for each s ∈ Ω,

|fn (s) − fm (s)| ≤ d(fn , fm ) ∀ n, m ∈ N. (1)

Hence, for each s ∈ Ω, (fn (s)) is a Cauchy sequence in R. Define

f (s) := lim fn (s), s ∈ Ω. (2)


n→∞

We show that f ∈ B(Ω) and d(fn , f ) → 0. Let ε > 0 be given. Since (fn ) is a Cauchy sequence,
there exists N ∈ N such that
d(fn , fm ) < ε ∀ n, m ≥ N.
Hence, by (1), for each s ∈ Ω,

|fn (s) − fm (s)| ≤ ε ∀ n, m ≥ N.

From this, letting m → ∞,


|fn (s) − f (s)| ≤ ε ∀ n ≥ N.
This is true for all s ∈ Ω. Hence,

|f (s)| ≤ |f (s) − fN (s)| + |fN (s)| ≤ ε + ∥fN ∥∞ ∀n ≥ N ∀s ∈ Ω

so that f ∈ B(Ω) and


d(fn , f ) = sup |fn (s) − f (s)| ≤ ε ∀ n ≥ N.
s∈Ω

This shows that d(fn , f ) → 0. ♢

27
Remark 2.51. We may observe that B(Ω) is a vector space. Further, we have the following:

ˆ If Ω = {1, . . . , k}, then B(Ω) is isomorphic with Rk . In fact, the map T : B(Ω) → Rk defined by

T f = (f (1), f (2), . . . , f (k)), f ∈ B(Ω),

is a bijective linear transformation. In this case B(Ω) is a finite dimensional vector space with
basis {f1 , . . . , fk } with 
1 if i = j,
fi (j) = δij =
0 if i ̸= j
form a basis of B(Ω).
ˆ If Ω = N, then B(Ω) is isomorphic with the set ℓ∞ (N) of all bounded sequences. In fact, the
map T : B(Ω) → Rk defined by

T f = (f (1), f (2), . . .), f ∈ B(Ω),

is a bijective linear transformation. In this case B(Ω) is a infinite dimensional vector space,
since {f1 , f2 , . . .} with 
1 if i = j,
fi (j) = δij =
0 if i ̸= j
is an infinite linearly independent set.
ˆ If Ω is any set, then for each t ∈ Ω, we may define ft : Ω → R by

1 if s = t,
ft (s) =
0 if s ̸= t.

Then, {ft : t ∈ [a, b]} is a linearly independent set.


Example 2.52. The set C[a, b] is closed in B[a, b] with respect to the sup-metric, and hence C[a, b]
is complete with respect to the sup-metric. ♢
Example 2.53. The metric space C[a, b] with metric
Z b
d1 (x, y) := |x(t) − y(t)| dt, x, y ∈ C[a, b]
a

is not complete. ♢
Definition 2.54. Let (X, d) and (Y, ρ) be metric spaces. A function φ : X → Y is said to be an
isometry from (X, d) to (Y, ρ) if

ρ(φ(x), φ(u)) = d(x, u) ∀ x, u ∈ X. ♢

ˆ Every isometry is injective.

We note that if φ : X → Y is a surjective isometry from (X, d) to (Y, ρ), then φ−1 : Y → X is also a
surjective isometry from (Y, d) to (X, ρ).
Definition 2.55. Metric spaces (X, d) and (Y, ρ) are said to me isometric if there exists a surjective
isometry between them. ♢

28
Definition 2.56. A metric space (Y, ρ) is said to be a completion of a metric space (X, d) if there
exists an isometry φ : X → Y such that R(φ) is dense in Y . ♢
Exercise 2.57. If (Y1 , ρ1 ) and (Y2 , ρ2 ) are completions of (X, d), then (Y1 , ρ1 ) and (Y2 , ρ2 ) are iso-
metric. ♢

Since closed subset of a compelte metric space is complete, to obtain a completion of a metric
space (X, d), it is enough to identify a complete metric space (Y, ρ) and an isometry T : X → Y , so
e := R(T ) is a completion of X.
that X
Thorem 2.58. Every metric space has a completion.

Proof. Let (X, d) be a metric space qn let x0 ∈ X. For each u ∈ X, let φu : X → R be defined by

φu (x) = d(x, u) − d(x, x0 ), x ∈ X.

Then we have
|φu (x)| = |d(x, u) − d(x, x0 )| ≤ d(u, x0 ), x ∈ X.
Hence, φu ∈ B(X). Define T : X → B(X) by

T (u) = φu , u ∈ X.

Note that

∥T (u) − T (v)∥∞ = sup |φu (x) − φv (x)| = sup |d(u, x) − d(v, x)| ≤ d(u, v)
x∈X x∈X

and
|d(u, u) − d(v, u)| = d(u, v).
Therefore,
∥T (u) − T (v)∥∞ = d(u, v) ∀ u, v ∈ X.
Thus, T is an isometry, so that Y := R(T ), which is a closed subset of the complete metric space
B(X), is a completion of X.

Let (X, d) be a metric space. Suppose it is not complete. Then there exists Cauchy sequence (xn )
in X which does not converge in X. We specify a metric space (Y, ρ) which is an extension of (X, d)
such that (xn ) converges in Y . Here, by an extension of (X, d) we mean a metric space (Y, ρ) such
that
X ⊆ Y and ρ(x, y) = d(x, y) for x, y ∈ X.

Let us denote the sequence (xn ) by Φ. Consider Y = X ∪ {Φ}. Define ρ : Y × Y → R by

if x, y ∈ X,

 d(x, y)
ρ(x, y) = lim d(x, xn ) if x ∈ X, y = Φ,
 n→∞
0 if x = y.

In the above, we used the fact that the sequence (d(x, xn )) converges for every x ∈ X. This is true,
because, for any n, m ∈ N, by triangle inequality of f ,

|d(x, xn ) − d(x, xm )| ≤ d(xn , xm )

29
so that (d(x, xn )) is a Cauchy sequence in R, and hence it converges. It can be shown that ρ is a
metric on Y . Now, we show that the sequence (xn ) in Y converges to Φ ∈ Y. For this, we observe that

ρ(xn , Φ) = lim d(xn , xm ) ∀ n ∈ N.


m→∞

Now, let ε > 0 be given, and let N ∈ N be such that d(xn , xm ) < ε for all n, m ∈ N. Hence, we have

ρ(xn , Φ) = lim d(xn , xm ) < ε ∀ n ≥ N.


m→∞

Thus, ρ(xn , Φ) → 0 as n → ∞. Thus, we have proved the following theorem.


Thorem 2.59. Let (X, d) be a metric space. If (xn ) is a Cauchy sequence in X, then there exists a
metric space (Y, ρ) (e.g., described as above) such that (xn ) converges in Y .

Now, we describe one of the standard methods of construction of a completion.


Let (X, d) be a metric space. Suppose X is not complete. Let X e be the family of all Cauchy
sequences in X. We observe that if x̃ := (xn ) and ỹ := (yn ) are in X , then

lim d(xn , yn ) exists.


n→∞

Indeed, since
d(xn , yn ) ≤ d(xn , xm ) + d(xm , ym ) + d(ym , yn ),
we have
|d(xn , yn ) − d(xm , ym )| ≤ d(xn , xm ) + d(ym , yn ),
so that (d(xn , yn )) is a Cauchy sequence in R and hence it converges.

We identify elements x̃ := (xn ) and ỹ := (yn ) in X


e if

lim d(xn , yn ) = 0.
n→∞

For x̃ := (xn ) and ỹ := (yn ) in X,


e define

ρ(x̃, ỹ) = lim d(xn , yn ).


n→∞

It can be shown that:

ˆ ρ is a metric on X and (X,


e ρ) is complete.

ˆ The map x 7→ (x, x, . . .) from X to X


e is an isometry.

Exercise 2.60. Let (X, d) be a metric space. Prove the following.

(i) If (xn ) is a Cauchy sequence in X, then for every x ∈ X, the sequence (d(xn , x)) converges.
(ii) If (xn ) and (yn ) are Cauchy sequences, then the sequence (d(xn , yn )) converges. ♢

30

You might also like