Real Analysis

Dr. B. Surendranath Reddy

&
Dr. Rupali S. Jain
Assistant Professor
School of Mathematical Sciences
Swami Ramanand Teerth Marathwada University, Nanded

Note: This notes is prepared based on the lectures delivered to our M.Sc.Students.
There may be some typographical errors or there may be some mistakes. You are
most welcome to send corrections or suggestions to [email protected]
Most of the content is based on the following text books:
Real Analysis, N.L. Carothers,Cambridge University Press A Basic Course in
Real Analysis, Ajit Kumar and S. Kumaresan, Taylor & Francis Group.

Contents
1 LUB Axiom and Archimedean property 4

1
2 Sequences 10

3 lim sup And lim inf of an 18

4 Countability 22

5 Metrics and Norms 27

6 Open sets and Closed sets 30

7 Continuity 51

8 Connectedness 59

9 Totally Bounded 65

10 Completion of a metric/norm space 78

11 Compactness 81

12 Uniform Continuity 87

13 Equivalent Metrics 91

14 Sequence of functions 98

15 The Weierstrass Theorem 105

16 Equicontinuity 110

17 Infinitely Differentiable Functions 114

18 The Stone-Weierstrass Theorem 115

2 Dr.B. Surendranath Reddy

19 Inequalities 118

3 Dr.B. Surendranath Reddy

1 LUB Axiom and Archimedean property

1 LUB Axiom and Archimedean property

Definition 1.1. Upper Bound
Let A be a non empty subset of R. Then an element α ∈ R is said to be an upper
bound of A if x ≤ α ∀ x ∈ A.

Remark 1.2. 1. α ∈ R is not an upper bound of A if ∃ x ∈ A such that α < x.

2. If α is an upper bound of A and β > α then β is also an upper bound of A.

Definition 1.3. Bounded above Set

Let A be a non empty subset of R then A is said to be bounded above if ∃ α ∈ R
such that α is an upper bound of A (i.e. ∃ α ∈ R such that x ≤ α ∀ x ∈ A).

Remark 1.4. φ 6= A ⊂ R is said to be not bounded above if given α ∈ R, ∃ x ∈ A

such that α < x i.e. ∀ α ∈ R ∃ x ∈ A

Definition 1.5. Least Upper Bound

Let φ 6= A ⊂ R then α ∈ R is said to be least upper bound of A or supremum of A if

1. α is an upper bound of A.

2. If β is an upper bound of A then α ≤ β.

and is denoted by α = lub A or α = sup A.

Remark 1.6. If an upper bound α ∈ R of A belong to A then α is called maximum

of A.

Lemma 1.7. If α and β are lub of A then α = β.

Proof. By consider α as an upper bound and β as a lub, we get β ≤ α.

Similarly, as β is an upper bound and α is a lub of A, we get α ≤ β.
⇒ α = β.

4 Dr.B. Surendranath Reddy

1 LUB Axiom and Archimedean property

Lemma 1.8. If α is the lub of A then for given  > 0, α −  is not an upper bound
of A.

Proof. Given that α = lub A. Since  > 0, α −  < α. Suppose α −  is an upper

bound of A. Since α is lub of A, α ≤ α −  which is a contradiction.
∴ α −  is not an upper bound.

Example 1.9. Let A = (0, 1) then lub = 1.

Proof. Clearly 1 is an upper bound for A.Let β be an upper bound of A.

β+1
We have to show that 1 ≤ β. Suppose not i.e. β < 1. Take x = 2
then 0 < x < 1.
⇒x∈A
β+1 β+1−2β
Now x − β = 2
−β = 2
1−β
x−β = 2
1−β
Since β < 1 ⇒ 2
>0
∴x−β >0
⇒ x > β, which is a contradiction to x ≤ β.
∴ 1 ≤ β ⇒ 1 is the lub of A.

Remark 1.10. 1. For an unbounded set A, we define sup A = +∞.

2. If A = ∅, then sup A = −∞ as every real number is an upper bound for A.

Definition 1.11. Lower Bound

Let A be a nonempty subset of R then an element β ∈ R is said to be a lower bound
if x ≥ β ∀ x ∈ A.

Definition 1.12. Bounded Below

Let A be a non empty subset of R then A is said to be bounded below if ∃ β ∈ R
such that β is a lower bound of A i.e. ∃ β ∈ R such that x ≥ β ∀x ∈ A

Example 1.13. Let A = (0, 1) = {x ∈ R|0 < x < 1} then 0 is an lower bound of A.

5 Dr.B. Surendranath Reddy

1 LUB Axiom and Archimedean property

Definition 1.14. Greatest Lower Bound

Let φ 6= A ⊂ R then β ∈ R is said to be greatest lower bound of A or infimum of A if

1. β is an lower bound of A.

2. If γ is a lower bound of A then β ≥ γ.

and is denoted by β = glb A or β = inf A.

Example 1.15. Let A = (0, 1) then glb A = 0.

Proof. Let β be an lower bound of A. We have to show that 0 ≥ β. Suppose not i.e.
β > 0,
take x = β2 .
Then 0 < x < 1 ⇒ x ∈ A (∵ β ≤ 1 ⇒ x < 1)
β
x−β = 2
− β = − β2
Since β > 0 ⇒ − β2 < 0.
∴x−β <0
∴x<β
Which is an contradiction.
∴0≥β
⇒ 0 is glb of A.

Remark 1.16. 1. For an unbounded set A, we define inf A = −∞.

2. If A = ∅, then inf A = +∞ as every real number is a lower bound for A.

Lemma 1.17. For any subset A of R sup(−A) = − inf A

where −A = {−a|a ∈ A}.

Proof. Let α = sup(−A)

⇒ −a ≤ α ∀ a ∈ A

6 Dr.B. Surendranath Reddy

1 LUB Axiom and Archimedean property

⇒ a ≥ −α ∀ a ∈ A.
− α is a lower bound of A. Let β be a lower bound of A
⇒ β ≤ a∀a ∈ A
⇒ −β ≥ −a ∀ a ∈ A.
⇒ −β is an upper bound of −A.
Since α = sup(−A)
⇒ α ≤ −β
⇒ −α ≥ β
⇒ −α is the glb of A.
⇒ α = − inf A
∴ sup(−A) = − inf A

Definition 1.18. textbfCompleteness Axiom or LUB Axiom

Any non empty subset A of R with an upper bound has the least upper bound.

Theorem 1.19. Archimedean Property

If x, y ∈ R with x > 0 then ∃ n ∈ N such that nx > y.

Proof. We prove by contradiction. Suppose @ n ∈ N such that nx > y then ∀ n ∈

N, nx ≤ y.
Let A = {nx : n ∈ N} then A 6= φ and y is an upper bound of A. By Completeness
axiom A has a least the upper bound say α ∈ R i.e. ∃ α ∈ R such that α = lub A
⇒ nx ≤ α ∀ n ∈ N.
⇒ (n + 1)x ≤ α ∀ n ∈ N
⇒ nx + x ≤ α ∀ n ∈ N
⇒ nx ≤ α − x ∀ n ∈ N
⇒ α − x is an upper bound of A.
Since α is lub A ⇒ α ≤ α − x

7 Dr.B. Surendranath Reddy

1 LUB Axiom and Archimedean property

which is a contradiction to α − x < α. (∵ x > 0 ⇒ α − x < α)

∴ ∃ n ∈ N such that nx > y.

Lemma 1.20. N is not bounded above i.e. ∀ x ∈ R ∃ n ∈ N such that x < n.

Proof. Suppose N is bounded above i.e. ∃ x ∈ R such that n ≤ x ∀ n ∈ N. By

Completeness axiom N has the least upper bound say α.
⇒ n ≤ α ∀ n ∈ N.
⇒ n + 1 ≤ α ∀ n ∈ N (∵ n + 1 ∈ N)
⇒ n ≤ α − 1∀n ∈ N
⇒ α − 1 is an upper bound on N.
Since α = lub N, α ≤ α − 1 which is a contradiction.
∴ N is not bounded above.

Remark 1.21. The above lemma is also one version of Archimedean principle.

Theorem 1.22. Prove that both the version of Archimedean principle are equivalent.

Proof. (1) ⇒ (2). Suppose for given any x, y ∈ R with x > 0, ∃ n ∈ N such that
nx > y.
In particular, for x = 1 > 0, y ∈ R, ∃ n such that n.1 > y ⇒ n > y.
∴ Given any y ∈ R, ∃ n ∈ N such that n > y.
⇒ N is not bounded above.
(2) ⇒ (1). Assume that N is not bounded above.
Let x, y ∈ R with x > 0. We have to prove that there exists n ∈ N such that nx > y.
Suppose nx ≤ y ∀ n ∈ N
y
⇒n≤ x
∀n ∈ N
y
⇒ x
is an upper bound of N, which is a contradiction. Hence there exists n ∈ N such
that nx > y.

8 Dr.B. Surendranath Reddy

1 LUB Axiom and Archimedean property

Theorem 1.23. Density of Q

If a and b are real numbers with a < b then ∃ r ∈ Q such that a < r < b.

Proof. Since a < b, b − a > 0

∴ By Archimedean principal ∃ n ∈ N such that
n(b − a) > 1
nb − na > 1
Take m = k + 1 where k = [na] then m ∈ Z. Since for any x, [x] ≤ x < [x] + 1
⇒ na < [na] + 1
⇒ na < m.
Suppose m ≥ nb then 1 = k + 1 − k
i.e. 1 = m − k ≥ nb − na (k ≤ na ⇒ −k ≥ −na)
⇒ nb − na ≤ 1, which is a contradiction.
∴ m < nb
⇒ na < m < nb
m
⇒a< n
<b
m
∴ ∃r = n
∈ Q such that a < r < b.

Exercise: For any a, b ∈ R with a < b, show that there exists s ∈ R \ Q such
that a < s < b.

9 Dr.B. Surendranath Reddy

2 Sequences

2 Sequences
Definition 2.1. Let X be a non-empty set. A sequence in X is a function f : N → X
and is denoted by (xn ) where xn = f (n).
1
e.g. f (n) = n
is a sequence whose elements are 1, 21 , 13 , 14 , ....

Definition 2.2. A sequence (xn ) is said to be increasing(non decreasing) if xn ≤

xn+1 ∀ n i.e. x1 ≤ x2 ≤ x3 ... and is said to be decreasing(non increasing) if xn ≥
xn+1 ∀ n i.e. x1 ≥ x2 ≥ x3 ≥ ...

Definition 2.3. A sequence (xn ) is said to be monotone if (xn ) is either increasing

or decreasing.

Definition 2.4. A sequence (xn ) in R is said to be bounded if ∃ M ∈ R such that

|xn | ≤ M ∀ n ∈ N, otherwise it is called unbounded.
1
Example: xn = n
is bounded where as xn = n is unbounded.

Definition 2.5. A sequence (xn ) in R is said to be converge to x ∈ R if given

 > 0 ∃ n0 ∈ N such that |xn − x| <  ∀ n ≥ n0 .i.e. xn ∈ (x − , x + ) ∀ n ≥ n0
and is denoted by xn → x.

1
Example 2.6. Show that n
→ 0.

Proof. Let  > 0. We want to find n0 ∈ N such that ∀n ≥ n0 | n1 − 0| = 1

n
< .
1
By Archimedean property ∃ n0 ∈ N such that n0
< .
1 1
Let n ≥ n0 then n
≤ n0
<
1
⇒ n
<
1
∴ n
<  ∀ n ≥ n0
⇒ | n1 − 0| < 
1
⇒ n
→ 0.

1
Example 2.7. Show that xn = n+1
→0

10 Dr.B. Surendranath Reddy

2 Sequences

1
Proof. Let  > 0. We want to find n0 ∈ N such that n ≥ n0 | n+1 − 0| < .
1
i.e. ∀ n ≥ n0 , n+1 <
1
By Archimedean property ∃ n0 ∈ N such that n0
< .
Let n ≥ n0
1 1 1
∴ n+1
≤ n
≤ n0
<
1
⇒ n+1
<
1
∴ n+1
<  ∀ n ≥ n0
1
⇒ | n+1 − 0| < 
1
⇒ n+1
→ 0.

Theorem 2.8. Every convergent sequence in R is bounded.

Proof. Suppose xn → x. Then for  = 1, ∃ k ∈ N such that |xn − x| < 1 ∀ n ≥ k

Now |xn | = |xn − x + x| ≤ |xn − x| + |x|
Therefore,|xn | ≤ 1 + |x| ∀ n ≥ k
Take M = max{|x1 |, |x2 |, ...|xk−1 |, 1 + |x|}
Then, |xi | ≤ M 1 ≤ i ≤ k − 1 and
|xn | < 1 + |x| ≤ M ∀ n ≥ k
⇒ |xn | ≤ M ∀ n
∴ (xn ) is bounded.

Remark 2.9. The converse of the above theorem is false. For example, consider the
sequence xn = (−1)n . As |xn | ≤ 1 for all n, xn is bounded.
Now suppose xn converges to some x ∈ R.
Then for  = 1, there exists k ∈ N such that |xn − x| < 1 for all n ≥ k.
In particular for n = 2k, we get |1 − x| < 1 ⇒ x ∈ (0, 2).
Similarly, for n = 2k +1, we get |−1−x| < 1 ⇒ x ∈ (−2, 0), which is a contradiction.
Hence xn is not convergent.

11 Dr.B. Surendranath Reddy

2 Sequences

Theorem 2.10. A monotone bounded sequence of real numbers converges.

Proof. Let (xn ) ⊂ R be a bounded monotone sequence.

Firstly, we assume that (xn ) is increasing.
Let A = {xn | n ≥ 1}.
Since (xn ) is bounded, A is bounded and sup A exists, say x = sup A.
We will show that xn → x.
Let  > 0. Since x = sup A, x −  is not an upper bound for A.
∴ ∃ k ∈ N such that x −  < xk .
Let n ≥ k. Since (xn ) is increasing xn ≥ xk .
Implies, x −  < xk ≤ xn ≤ x ≤ x + 
⇒ |xn − x| < 
⇒ xn → x.
Now suppose (xn ) is decreasing
⇒ (−xn ) is increasing.
⇒ (−xn ) → sup(−A)
⇒ (−xn ) → − inf(A) (∵ sup(−A) = − inf A)
xn → inf(A)
Therefore in either case, xn converges.

Remark 2.11. If S ⊂ N is infinite then the elements of S can be listed as n1 < n2 <
n3 ... < nk ....

Proof. Since S is infinite, S is non empty, and so by Well Ordering Principle, S has
the least element say n1 .
Also S \ {n1 } is no empty, and again by Well Ordering Principle, S \ {n1 } has the
least element say n2 .
Then n1 < n2 (∵ n1 ≤ n2 and n1 6= n2 ). By repeating this process infinitely many
times, we get that S = {n1 , n2 , · · · , }.

12 Dr.B. Surendranath Reddy

2 Sequences

Definition 2.12 (Subsequence). Let x : N → X is a sequence in X then the map

x restricted to an infinite subset of N is called a subsequence of x i.e. if S < N is
infinite then x : S → X is a subsequence. But S can be listed as n1 < n2 < n3 < ....
∴ subsequence is denoted by (xnk ) where n1 < n2 < n3 < ...
e.g. If xn = x and S=Set of even numbers then xnk = (2, 4, 6, 8, ...) is a subsequence
of xn .

Remark 2.13. What is meant by subsequence (xnk ) converges?

Let yk = xnk then xnk → x if and only if yk → x.
If and only if for  > 0 ∃ p ∈ N such that |yk − x| <  ∀ k ≥ p
Therefore, xnk → x if and only if ∀ > 0, ∃ p ∈ N such that |xnk − x| <  ∀ k ≥ p.

Remark 2.14. Let (xnk ) be a subsequence of (xn ) then nk ≥ k ∀ k ∈ N.

Proof. Since (xnk ) is a subsequence of (xn ), we have n1 < n2 < n3 ... < nk ∈ N.
Clearly, n1 ≥ 1
As n1 < n2 ⇒ 1 ≤ n1 < n2
⇒ n2 ≥ 2
Suppose assume that nk ≥ k
Then nK+1 > nk ≥ k implies nk+1 ≥ k + 1.
Therefore by mathematical induction, nk ≥ k ∀ k ∈ N.

Theorem 2.15. Every sequence of real numbers has a monotone subsequence.

Proof. Let (xn ) be a sequence in R.

Consider S = {n ∈ N|xn > xm for m > n}
(Recall the example of Volcano discussed in the class for the description of the set S)
Then S may be finite or infinite.
Suppose S is finite. Then ∃ N ∈ S such that s ≤ N ∀ s ∈ S.
We can choose n1 ∈ N such that n1 > N and so n1 ∈
/ S.

13 Dr.B. Surendranath Reddy

2 Sequences

⇒ ∃ n2 ∈ N with n2 > n1 such that xn1 ≤ xn2 .

Since n2 > n1 ⇒ n2 ∈
/S
So ∃ n3 ∈ N with n3 > n2 such that xn2 ≤ xn3 .
By repeating this process we get, n1 < n2 < n3 < ... such that xn1 ≤ xn2 ≤ xn3 < ...
∴ (xnk ) is a increasing subsequence of (xn ).
Suppose S is infinite. Since S 6= ∅, by well ordering principle S has the least element
say n1 ,
i.e. n1 ≤ n ∀ n ∈ S.
Applying well ordering principle to S \ {n1 } =
6 ∅ there exists n2 ∈ N, which is least
element of S \ {n1 }
⇒ n2 > n1 and hence xn1 > xn2 .
Again as S \ {n1 , n2 } =
6 ∅, it has the least element say n3 .
∴ n3 > n2 and so xn2 > xn3 .
By repeating this process we get,n1 < n2 < n3 ... such that xn1 > xn2 > xn3 ...
∴ (xnk ) is a decreasing subsequence of (xn ).
∴ In either case there exists a subsequence (xnk ) of xn which is monotone.

Theorem 2.16. Bolzano-Weierstrass Theorem

Every bounded sequence of real number has a convergent subsequence.

Proof. Step1: First we prove that every sequence in R has a monotone subsequence.
(Write the proof of theorem 2.16)
Step2: Next we show that that every monotone bounded sequence converges.
(Write the proof of theorem 2.11)
Let xn be a bounded sequence. Then by step1, there exists a monotone subsequence
xnk .
Since xn is bounded, xnk is also bounded.
Then by step2, xnk converges and hence the result.

14 Dr.B. Surendranath Reddy

2 Sequences

Lemma 2.17. If xn → x and (xnk ) is a subsequence of (xn ), then xnk → x.

Proof. Let  > 0. Since xn → x, ∃ p ∈ N such that |xn − x| <  ∀ n ≥ p.

Let k ≥ p then nk ≥ k ≥ p.
Implies, |xnk − x| <  as nk ≥ p
Therefore,|xnk − x| <  ∀ k ≥ p
⇒ xnk → x

Definition 2.18. Cauchy Sequence

Let (xn ) be a sequence in R, then (xn ) is said to be Cauchy if given
 > 0 ∃ p ∈ N such that |xn − xm | <  ∀ n, m ≥ p.

Theorem 2.19. Every Cauchy sequence is bounded.

Proof. Let (xn ) be a Cauchy sequence.

Then for  = 1, there exists p ∈ N such that |xn − xm | < 1 ∀ n, m ≥ p.
In particular, |xn − xp | < 1 ∀ n ≥ p
Now |xn | = |xn − xp + xp | ≤ |xn − xp | + |xp |
⇒ |xn | < |1 + |xp | ∀ n ≥ p.
Take M = max{|x1 |, |x2 |, ...|xp−1 |, 1 + |xp |}.
Then |xn | ≤ M ∀ n ∈ N
∴ (xn ) is bounded.

Next we prove the important property of Cauchy sequence that if a subsequence

of a cauchy sequence converges, then the sequence also converges.

Theorem 2.20. If (xn ) is Cauchy and xnk → x then xn → x.

Proof. Let  > 0. Since (xn ) is Cauchy ∃ p1 ∈ N such that |xn − xm | < 2 ∀ n, m ≥ p1 .
Also xnk → x so ∃ p2 ∈ N such that |xnk − x| < 2 ∀ k ≥ p2
Take p = max{p1 , p2 } and let n ≥ p. Then

15 Dr.B. Surendranath Reddy

2 Sequences

|xn − x| = |xn − xnp + xnp − x| ≤ |xn − xnp | + |xnp − x|


Since n ≥ p ≥ p1 and np ≥ p ≥ p1 , we get |xn − xnp | < 2

Similarly as np ≥ p ≥ p2 , we get |xnp − x| < 2
 
∴ |xn − x| < 2
+ 2
=
⇒ xn → x.

Theorem 2.21. If (xn ) converges then it is Cauchy.

Proof. Let xn → x and  > 0.

Then ∃ p ∈ N such that |xn − x| < 2 ∀ n ≥ p.
Let n, m ≥ p. Then
 
|xn − xm | = |xn − x + x − xm | ≤ |xn − x| + |xm − x| < 2
+ 2
=
∴ (xn ) is Cauchy.

Theorem 2.22. Completeness of R

Every Cauchy sequence in R converges in R.

Proof. Let (xn ) be a Cauchy sequence. Then it is bounded.

By Bolzano-Weierstrass theorem, there exists a subsequence xnk such that xnk → x.
Since (xn ) is Cauchy and xnk → x then xn → x.
(Write the proof of theorem 2.20).

Theorem 2.23. Nested Interval Theorem

If (In ) is a sequence of closed bounded non empty intervals in R with
I1 ⊃ I2 ⊃ I3 ⊃ ... then ∩∞ ∞
n=1 In 6= φ. If length (In ) → 0 then ∩n=1 In contains precisely

one element.

Proof. As each In is closed and bounded interval, we denote it by In = [an , bn ].

Since In ⊃ In+1 , we get an ≤ an+1 ≤ bn+1 ≤ bn
Therefore, (an ) is increasing and (bn ) is decreasing.

16 Dr.B. Surendranath Reddy

2 Sequences

Now (an ) is bounded above as an ≤ bn ≤ b1 , ∀ n. ∴ an converges to a = sup an .

Similarly, bn converges to b = inf bn .
Now we show that each bk is an upper bound for the sequence (an ).
Let k ∈ N.
If k < n, then an ≤ bn ≤ bk .
If k > n, then an ≤ ak ≤ bk . Therefore, an ≤ bk for all n ∈ N and hence bk is an
upper bound.
Since a = sup an , we get that a ≤ bk for all k.
Implies a is a lower bound for (bn ).
As b = inf bn , we get that a ≤ b.
We now show that ∩∞
n=1 In − [a, b].

Let x ∈ ∩∞
n=1 In

⇒ x ∈ In ∀ n
∴ an ≤ x ≤ b n ∀ n
⇒ limn an ≤ x ≤ limn bn
⇒a≤x≤b
⇒ x ∈ [a, b]
⇒ ∩∞
n=1 In ⊂ [a, b].

Let x ∈ [a, b] ⇒ a ≤ x ≤ b
⇒ an ≤ a ≤ x ≤ b ≤ b n ∀ n
⇒ an ≤ x ≤ b n ∀ n
⇒ x ∈ In ∀ n
⇒ x ∈ ∩∞
n=1 In

∴ [a, b] ⊂ ∩∞
n=1 In

∴ ∩∞
n=1 In = [a, b] 6= φ.

Suppose l(In ) → 0 i.e. bn − an → 0

On other hand bn − an → b − a

17 Dr.B. Surendranath Reddy

3 lim sup And lim inf of an

∴b−a=0
⇒ a = b.
∴ ∩∞
n=1 In = {a}.

3 lim sup And lim inf of an

Consider a bounded sequence (an ). Let tn = inf{xn , xn+1 , ...} = inf k≥n xk
and Tn = sup{xn , xn+1 , ...} = supk≥n xk
Since tn ≤ xk ∀ k ≥ n
⇒ tn ≤ xk ∀ k ≥ n + 1
⇒ tn is a lower bound of {xn+1 , xn+2 , ...}. Since tn+1 is the glb of {xn+1 , xn+2 , ...}
⇒ tn ≤ tn+1 ∀ n
⇒ (tn ) is an increasing sequence.
Similarly, (Tn ) is a decreasing sequence. Since xn is bounded. (tn ) and (Tn ) are
bounded.
∴ lim tn = sup tn and lim Tn = inf Tn .
Define limn→∞ inf an = sup tn = supn≥1 inf k≥n ak and
lim sup an = inf Tn = inf n≥1 supk≥n ak
If (an ) is not bounded above then define lim sup an = +∞.
And if (an ) is not bounded below.
lim inf an = −∞.

Example 3.1. 1. an = (−1)n

⇒ Here tn = inf{(−1)n , (−1)n+1 , (−1)n+2 , ...}
tn = −1 ∀ n
∴ sup tn = −1
lim inf an = −1

18 Dr.B. Surendranath Reddy

3 lim sup And lim inf of an

and Tn = sup{(−1)n , (−1)n+1 , ...} = 1, ∀ n

⇒ inf Tn = 1
lim sup an = 1.

1 2 1 4 1 6
, , , , , , ..., n1 , n+1
n


2. an = 2 3 4 5 6 7
, ...
t1 = inf{ 12 , 32 , 14 , ..., n1 , n+1
n
}=0
t2 = inf{ 32 , 41 , ..., n1 , n+1
n
}=0
tn = 0
lim inf an = sup tn = 0
Tn = 1, ∀ n
⇒ inf Tn = 1
∴ lim sup an = 1.

Theorem 3.2. If (an ) converges then

limn→ inf an = limn→ sup an = limn→ an .

Proof. Let limn→∞ an = a

α = lim inf an , β = lim sup an .
We have to show that α = β = a,
Let  > 0 since an → a, ∃ k ∈ N such that
|an − a| <  ∀ n ≥ k
a −  < an < a +  ∀ n ≥ k.
Since a −  < an ∀ n ≥ k
⇒ a −  is a lower bound of {an |n ≥ k}.
⇒ a −  < tk
⇒ a −  < tn ∀ n ≥ k
and α = sup tn ⇒ tn ≤ α
∴ a −  < tn ≤ α
⇒ a −  < α.

19 Dr.B. Surendranath Reddy

3 lim sup And lim inf of an

Also an < a +  ∀ n ≥ k
∴ a + α is an upper bound of {ak , ak+1 , ...}
⇒ Tk ≤ a + 
⇒ Tn ≤ Tk ≤ a +  ∀ n ≥ k
Since β = inf Tn
⇒ β ≤ Tn ∀n
∴ β ≤ Tk ≤ a + 
⇒β ≤a+
∴a−<α≤β ≤a+
|α − β| < 2
Since  > 0 is arbitrary α − β = 0 ⇒ α = β.
Since α, β ∈ (a − , a + )
⇒ |a − α| < 2 and |a + β| < 2
⇒ a = α and a = β
∴α=β=a
∴ lim inf an = lim sup an = lim an

Theorem 3.3. If (an ) is bounded and lim inf an = lim sup an then
an converges and lim an = lim inf an = lim sup an .

Proof. Let a = lim inf an = lim sup an . Since (an ) is bounded a ∈ R.

Let  > 0 since a = sup tn . a −  is not an upper bound of (tn ) ∃ k1 ∈ N such that
a −  < tk1
⇒ a −  < tk1 ≤ an ∀ n ≥ k1
a −  < an ∀ n ≥ k
Also a = lim sup an = inf = Tn
⇒ a +  is not a lower bound of (Tn ) ∃ k2 ∈ N such that Tk2 < a + 
⇒ an ≤ Tk2 < a +  ∀ n ≥ k2

20 Dr.B. Surendranath Reddy

3 lim sup And lim inf of an

∴ an < a +  ∀ n ≥ k 2
Take k = max{k1 , k2 } then
a −  < an < a +  ∀ n ≥ k
⇒ |an − a| <  ∀ n ≥ k
an → a
∴ lim an = a = lim inf an = lim sup an .

21 Dr.B. Surendranath Reddy

4 Countability

4 Countability
Definition 4.1 (Equivalent Sets). Two sets A and B are said to be equivalent if
there is a one-one correspondence between the sets A and B.
i.e. ∃ a bijective map f : A → B and is denoted by A ∼ B.

Example 4.2. Exercise

Show that the relation ∼ is an equivalence relation.

Definition 4.3. A set X is said to be finite if either X = φ or X is equivalent to

{1, 2, , 3, ...n} for some natural number n.

Definition 4.4. A set X is said to be infinite if @ n ∈ N such that X ∼ {1, 2, ..., n}.

Definition 4.5. A set X is said to be countable if either X is finite or is equivalent

to N .
e.g.1) N is countable as N ∼ N.
2) Z is countable i.e. Z ∼ N.
Define f: Z → N
 2n, n≥1
f (n) =
 1 − 2n, n ≤ 0.
To show f is one-one, let f (n1 ) = f (n2 )
⇒ 2n1 = 2n2 or 1 − 2n1 = 1 − 2n2
⇒ n1 = n2
To show that f is onto, let m ∈ N. Suppose m is odd.
⇒ m = 1 − 2x, x ∈ N
1−m
⇒x= 2
≤0
Then f (x) = 1 − 2(x) = 1 − 2 1−m
2
=1−1+m=m
Now suppose m is even.
⇒ 2|m

22 Dr.B. Surendranath Reddy

4 Countability

m
Then take x = 2
∈Z
f (x) = 2x = 2( m2 ) = m
∴ f is onto.
Z ∼ N and hence it is countable.

Example 4.6. Show that N × N is countable. i.e. N × N ∼ N.

Proof. Given any natural number k ∈ N, ∃ m, n ∈ N such that

k = 2m−1 (2n − 1)
Define f : N → N × N by
f (k) = (m, n).
Let f (k1 ) = f (k2 )
⇒ 2m1 −1 (2n1 − 1) = 2m2 −1 (2n2 − 1)
⇒ m1 = m2 and n1 = n2
∴ f is one to one.
To show onto, let (m, n) ∈ N × N then take k = 2m−1 (2n − 1) and
f (k) = (m, n)
⇒ f is onto.
∴N×N∼N
N × N is countable.

Theorem 4.7. If A and B are countable then A × B is countable.

Proof. Since A and B are countable, ∃ bijective maps f : A → N and g : B → N.

Define h : A × B → N × N by
h ((a, b)) = (f (a), g(b))
then h is well defined.
Let h ((a1 , b1 )) = h ((a2 , b2 ))
⇒ (f (a1 ), g(b1 )) = (f (a1 ), g(b2 ))

23 Dr.B. Surendranath Reddy

4 Countability

⇒ f (a1 ) = f (a2 ) and g(b1 ) = g(b2 )

⇒ a1 = a2 and b1 = b2
⇒ (a1 , b1 ) = (a2 , b2 ).
∴ h is 1-1.
To show onto, let (m, n) ∈ N × N. Since f and g are onto ∃ a ∈ A and b ∈ B such
that f (a) = m and g(b) = n
h ((a, b)) = (f (a), g(b)) = (m, n)
⇒ h is onto.
∴ h is bijective.
A × B ∼ N × N.
Since N × N ∼ N
⇒ A × B is countable.

Theorem 4.8. If An ’s are countable then ∪∞

n=1 An is also countable.

i.e.Countable union of countable sets is countable.

Proof. Since each An is countable, we can represent An ’s as follows

A1 : {a11 , a12 , a13 , ...}
A2 : {a21 , a22 , a23 , ...}
An : {an1 , an2 , an3 , ..}.
Then ∪∞
n=1 An = {aij |(i, j) ∈ N × N}

Now define f : ∪∞
n=1 An → N × N by

f (aij ) = (i, j).

Suppose f (aij ) = f (alm )
(i, j) = (l, m)
i = l, j = m
⇒ f is 1-1.
To show onto, let (i, j) ∈ N × N.

24 Dr.B. Surendranath Reddy

4 Countability

Then aij ∈ ∪∞
n=1 An such that f (aij ) = (i, j) and hence f is onto

⇒ ∪∞
n=1 An ∼ N × N

⇒ ∪∞
n=1 An is countable.

Theorem 4.9. The set of rational numbers Q is countable.

Proof. For n ≥ 1, define An = { m

n
|m ∈ Z}. Clearly ∪∞
n=1 An ⊂ Q Since for any q ∈ Q,
m
we can write q = n
,m ∈ Z, n ≥ 1.
⇒ Q ⊂ ∪∞
n=1 An

Therefore, Q = ∪∞
n=1 An

Now we will show that each An is countable.

Define f : An → Z by
f(m
n
) = m.
Suppose f ( mn1 ) = f ( mn2 )
⇒ m1 = m2
⇒ f is one-one.
k
Let k ∈ Z. Then n
∈ An
∴ f ( nk ) = k
⇒ f is onto.
⇒ An ∼ Z.
Since Z is countable, An is also countable.
Since Q = ∪∞
n=1 An and each An is countable, Q is also countable.

Theorem 4.10. If A is an infinite subset of N then A is countable.

Proof. Let A ⊂ N be an infinite set.

Since A 6= ∅, by well ordering principle there exists a least element say x1 .
Also A \ {x1 } =
6 ∅ again by well ordering principle, let x2 be the least element of
A \ {x1 }.

25 Dr.B. Surendranath Reddy

4 Countability

Then x1 < x2 .
By repeating this same arguments, we get, x1 < x2 < x3 < ... such that xn is the
least element of A \ {x1 , x2 , ...xn−1 }.
We will show that A \ {x1 , x2 , ...} = ∅.
Suppose not. Then ∃ m ∈ A such that m 6= xi ∀ i.
Take S = {k ∈ N|xk > m}.
Since xm ≥ m and m 6= xm , we get xm > m and so m ∈ S.
∴ By well ordering principle S has least element say n
⇒ 1, 2, ..., n − 1 ∈
/ S and n ∈ S
⇒ x1 < x2 < ... < xn−1 < m and xn > m
⇒ x1 < x2 < ... < xn−1 < m < xn .
Since xn is the least element of A \ {x1 , x2 , ..., xn−1 }
⇒ xn ≤ m, which is a contradiction.
∴ A \ {x1 , x2 , ...} = ∅
⇒ A = {x1 , x2 , ...}.
∴ A is countable.

Corollary 4.11. Any subset of a countable set is countable.

Theorem 4.12. (0, 1) is uncountable where (0, 1) = {x|0 < x < 1}.

Proof. Suppose (0, 1) is countable, then (0, 1) = {x1 , x2 , ...}.

We can write each xi in decimal expansion as follows:
x1 = 0.a11 a12 a13 ...
x2 = 0.a21 a22 a23 ...
x3 = 0.a31 a32 a33 ....
Define y = 0.b
11 b22 b33 ...

nn 6
 2, if a = 2;
Where bnn =
 3, if a = 2.
nn

26 Dr.B. Surendranath Reddy

5 Metrics and Norms

⇒ ann 6= bnn ∀ n
Then y ∈ (0, 1) but y 6= xn ∀n
Which is a contradiction.
∴ (0, 1) is uncountable.

Theorem 4.13. R is uncountable.

Proof. We know that (0, 1) is uncountable. Suppose R is countable then (0, 1) ⊂ R

is also countable. Which is a contradiction.
⇒ R is uncountable.

Theorem 4.14. R/Q is uncountable.

Proof. Suppose R/Q is countable then R = Q ∪ (R/Q) is also countable. Which is a

contradiction.
⇒ R/Q is uncountable.

5 Metrics and Norms

Definition 5.1 (Metric Space). Let M be a non empty set. A function d : M ×
M → R is said to be a metric on M if

1. d(x, y) ≥ 0 ∀ x, y ∈ M

2. d(x, y) = 0 iff x = y

3. d(x, y) = d(y, x) ∀x, y ∈ M

4. d(x, y) ≤ d(x, z) + d(z, y) ∀x, y, z ∈ M

The pair (M, d) is called a metric space.

27 Dr.B. Surendranath Reddy

5 Metrics and Norms

Example5.2. Every set M admits a metric, d : M × M → R given by,

 0, if x = y
d(x, y) =
 1, if x 6= y
The above metric d is called discrete metric and (M, d) is called a discrete space.
(Exer: Show that d is a metric)

Example 5.3. Define d : R × R → R by d(x, y) = |x − y|, then d is a metric called

usual metric on R.

Example 5.4. Define d : (0, ∞) × (0, ∞) → R defined by d(x, y) = | x1 − y1 | is a metric

on (0, ∞).
p
Example 5.5. If d is a metric on M then show that ρ(x, y) = d(x, y), σ(x, y) =
d(x,y)
1+d(x,y)
and τ (x, y) = {min d(x, y), 1} are also metrics on M .

Definition 5.6 (Normed Vector Space). Let V be a vector space over a field F .
A norm on V is a mapping ||.|| : V → R satisfying the following:

1. ||x|| ≥ 0 ∀x ∈ V

2. ||x|| = 0 iff x = 0

3. ||αx|| = |α|||x|| ∀x ∈ V, ∀α ∈ F

4. ||x + y|| ≤ ||x|| + ||y|| ∀x, y ∈ V.

Example 5.7. ||.|| : R → R define ||x|| = |x|. Then (R, ||.||) is a normed vector
space.

Example 5.8. On Rn = {(x1 , x2 , ..., xn )|xi ∈ R, ∀ i} define the maps

n
X
||x||1 = |xi |
i=1

28 Dr.B. Surendranath Reddy

5 Metrics and Norms

n
! 21
X
||x||2 = |xi |2
i=1

||x||∞ = max |xi |

1≤i≤n

then ||x||1 , ||x||2 and ||x||∞ are norms on Rn

Definition 5.9 (lp -Space). For 1 ≤ p < ∞, define lp space as the space of all real
P∞
sequences x = (xn )∞ n=1 such that
p
i=1 |xi | < ∞.

i.e. lp = {x = (xn )| ∞ p
P
i=1 |xi | < ∞}

i.e. x ∈ lp iff ∞ p
P
i=1 |xi | < ∞}
1
Define ||.||p on lp as ||x||p = ( ∞ p p
P
i=1 |xi | )

Example 5.10. Show that (l1 , ||.||) is a normed vector space.

∞
X
l1 = {x = (xn )| |xi | < ∞}
i=1
P∞
Proof. ||x||1 = i=1 |xi |.

P∞
1. Since |xi | ≥ 0 ⇒ i=1 |xi | ≥ 0
⇒ ||x||1 ≥ 0
P∞
2. ||x||1 = 0 ⇔ i=1 |xi | = 0
⇔ |xi | = 0 ∀i
⇔ xi = 0 ∀i
⇔ x=0

3. αx = (αx1 , αx2 , ...)

||αx||1 = ∞
P
i=1 |αxi |

= ∞
P
i=1 |α||xi |

= |α| ∞
P
i=1 |xi |

= |α| ||x||.

29 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

4. Let x, y ∈ l1 , then x + y = (x1 + y1 , x2 + y2 , ...)

∞
X ∞
X
||x + y||1 = |xi + yi | ≤ (|xi | + |yi |)
i=1 i=1
∞
X ∞
X
≤ |xi | + |yi |
i=1 i=1

≤ ||x||1 + ||y||1

6 Open sets and Closed sets

Definition 6.1 (Open Ball). Let (M, d) be a metric space. Given x ∈ M and r > 0,
the set Br (x) = {y ∈ M |d(x, y) < r} is called the open ball about x of radius r.

Remark 6.2. 1. If, on a set M , more than one metric is defined say d1 and d2
then the open balls are denoted by Brd1 (x), Brd2 (x).

2. x ∈ Br (x) as d(x, x) = 0 < r

Example 6.3. Let M = R, d(x, y) = |x − y|

Let x ∈ R and r > 0,

y ∈ Br (x) ⇔ d(x, y) < r

⇔ |x − y| < r

⇔ x−r <y <x+r

⇔ y ∈ (x − r, x + r)

∴ Br (x) = (x − r, x + r)

Example6.4. Consider a discrete metric space (M, d) i.e.

 0, if x = y
d(x, y) = then
 1, if x 6= y

30 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

1. Open ball B 1 (x)

2

1
B 1 (x) = {y ∈ M |d(x, y) < }
2 2
= {y ∈ M |d(x, y) = 0}

= {y ∈ M |x = y}

∴ B 1 (x) = {x}
2

2. Open ball B1 (x)

B1 (x) = {yM |d(x, y) < 1}

= {y ∈ M |d(x, y) = 0}

= {y ∈ M |x = y}

∴ B1 (x) = {x}

3. For any 0 < r < 1

Br (x) = {y ∈ M |d(x, y) < r < 1}

= {y ∈ M |d(x, y) = 0}

= {y ∈ M |x = y}

∴ Br (x) = {x} if 0 < r < 1

4. Open ball B2 (x)

B2 (x) = {y ∈ M |d(x, y) < 2}

= {y ∈ M |d(x, y) = 0 or d(x, y) = 1}

= {y ∈ M |x = y or x 6= y}

∴ B2 (x) = M

31 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

5. For any r > 1

Br (x) = {y ∈ M |d(x, y) < r}

= {y ∈ M |d(x, y) = 0 or d(x, y) = 1}

= {y ∈ M |x = y or x 6= y}

∴ Br (x) = M

Hence in a discrete metric space, open balls are either singleton or whole space.

Remark 6.5. Closed ball Br (x) is defined as

Br (x) = {y ∈ M |d(x, y) ≤ r}.

Example 6.6. Exercise

1. If M = R, d(x, y) = |x − y| what is Br (x) ?

2. Let (M, d) discrete space, find B 1 (x), B1 (x), B2 (x)

2

Example 6.7. M = R2 , d-Euclidean metric,

i.e.for any x = (x1 , y1 ), y = (x2 , y2 ) ∈ R2 ,
p
d(x, y) = (x2 − x1 )2 + (y2 − y1 )2 .
what is Br (x) =?

Proof.

(x2 , y2 ) = y ∈ Br (x) ⇔ d(x, y) < r

p
⇔ (x2 − x1 )2 + (y2 − y1 )2 < r

⇔ (x2 − x1 )2 + (y2 − y1 )2 < r2

∴ Br (x) = {(x2 , y2 )|(x2 − x1 )2 + (y2 − y1 )2 < r2 }

which is an open disc with center x = (x1 , y1 ) and of radius r.

32 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

Example 6.8. Let (V, ||.||) be a normed vector space. Define d : V × V → R by

d(x, y) = ||x − y||.

Br (x) = {y |d(x, y) < r}

= {y | ||x − y|| < r}

In particular Br (0) = {y | ||y|| < r}.

Example 6.9. For z = (x, y) ∈ R2 , define,

p
||z||1 = |x| + |y|, ||z||2 = |x|2 + |y|2 and ||z||∞ = max{|x|, |y|}.
Find and draw the graphs of B1 (0) in (R2 , ||.||1 ), (R2 , ||.||2 ) and (R2 , ||.||∞ ).

Example 6.10. Consider M = {(x, y) ∈ R2 |x ≥ 0, y ≥ 0}.

p
For any z ∈ M, ||z||2 = x2 + y 2 , find and draw the graph of B1 (0) in M .

Theorem 6.11. In a normed vector space (V, ||.||), Br (x) = x + Br (0) and Br (0) =
rB1 (0)
Where x + Br (0) = {x + y|y ∈ Br (0)} and rB1 (0) = {ry|y ∈ B1 (0)}

Proof. First we will show that Br (x) ⊂ x + Br (0).

Let y ∈ Br (x) ⇒ ||x − y|| < r.

Let z = y − x ⇒||z|| = ||y − x|| < r

||z|| < r

z ∈ Br (0)

y = x + z ∈ x + Br (0)

Br (x) ⊂x + Br (0)..........(1)

Conversely, let y ∈ x + Br (0) ⇒ y = x + z where z ∈ Br (0)

⇒ ||z|| < r

33 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

⇒ ||y − x|| < r

⇒ y ∈ Br (x)
⇒ x + Br (0) ⊂ Br (x)..........(2)
From (1) and (2), we get
Br (x) = x + Br (0)

Now to show that Br (0) ⊂ r.B1 (0), let y ∈ Br (0).

⇒ ||y|| < r
⇒ || yr || < 1 ⇒ y
r
∈ B1 (0)
∴ y = r yr ∈ r.B1 (0)
⇒ Br (0) ⊂ r.B1 (0).................(3)
Let y ∈ r.B1 (0) ⇒ y = rz where z ∈ B1 (0) ⇒ ||z|| < 1.
∴ ||y|| = ||rz|| = r.||z|| < r
⇒ ||y|| < r ⇒ y ∈ Br (0)
⇒ r.B1 (0) ⊂ Br (0)...........(4)
From (3) and (4),
Br (0) = r.B1 (0)

Definition 6.12 (Bounded Set). Let (M, d) be a metric space. A subset 0 A0 of M

is said to be bounded if A is contained in some open ball i.e. ∃ x ∈ M and r > 0 such
that A ⊂ Br (x)

Definition 6.13 (Diameter of a set). For a subset A of M , the diameter of A is

denoted by diamA and is given by,

diamA = sup{d(a, b) | a, b ∈ A}

Theorem 6.14. A subet 0 A0 of M is bounded iff diamA < ∞

34 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

Proof. Suppose A is bounded then ∃ x ∈ M and r > 0 such that A ⊂ Br (x).

Let a, b ∈ A ⇒ a, b ∈ Br (x)
⇒ d(a, x) < r and d(b, x) < r
⇒ d(a, x) + d(b, x) < 2r
⇒ d(a, b) ≤ d(a, x) + d(b, x) < 2r
⇒ d(a, b) < 2r ∀ a, b ∈ A
⇒ supa,b∈A d(a, b) ≤ 2.r
⇒ diamA ≤ 2r < ∞
Conversely, suppose that diamA < ∞, say diamA = r
We will show that for any x ∈ A, A ⊂ B2r (x)
Let y ∈ A. Since x, y ∈ A ⇒ d(x, y) ≤ diamA = r
⇒ d(x, y) ≤ r < 2r
⇒ d(x, y) < 2r
⇒ y ∈ B2r (x)
∴ A ⊂ B2r (x)
⇒ A is bounded.

Definition 6.15. Let (M, d) be a metric space. A sequence (xn ) in M converges

to x ∈ M if for given  > 0, ∃ N ∈ N such that d(xn , x) <  ∀n ≥ N Or xn ∈
B (x) ∀n ≥ N.

Definition 6.16. Let (M, d) be a metric space. A sequence (xn ) in M is said to be

Cauchy if for given  > 0, ∃ N ∈ N such that d(xn , xm ) <  ∀n, m ≥ N.

Theorem 6.17. Let (M, d) be a metric space. Then every convergent sequence is
Cauchy but not the converse.

Proof. Let (xn ) be a convergent sequence, i.e. xn → x, for some x ∈ M .

Let  > 0.

35 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

 
Since xn → x, for 2
> 0, ∃ N such that d(xn , x) < 2
∀ n ≥ N.
We will show that, d(xn , xm ) <  ∀ n, m ≥ N
Let n, m ≥ N . Then by triangle inequality,

d(xn , xm ) ≤ d(xn , x) + d(x, xm )

 
⇒ d(xn , xm ) < + = 
2 2

⇒ (xn ) is Cauchy.
Converse is not true in general.
1
Consider, xn = n
in (M, d) = ((0, 1], |.|) .
To show (xn ) is Cauchy, let  > 0.
We have to find N such that ∀n, m ≥ N, d(xn , xm ) < .

1 1 n−m
i.e.|xn − xm | = | − |=| |
n m nm
n 1
< = < .
nm m

By Archimedean property, there exists N such that N > 1 .

1 1
Now, for m, n ≥ N , d(xn , xm ) < m
≤ N
< .
Therefore, (xn ) is Cauchy in M.
Now Suppose xn → x, for some x ∈ (0, 1] since x > 0, by Archimedean property, ∃N
1
such that N
< x.
1
Let  = x − N
then  > 0. For this  > 0, as xn → x, ∃ n0 such that
|xn − x| <  ∀n ≥ n0 .
For any n ≥ max{2N, n0 }
|xn − x| = |x − n1 | ≥ |x − 1
2N
| > |x − 1
N
| =
⇒ |xn − x| >  ∀n ≥ n0 , which is a contradiction.
∴ (xn ) does not converge in M .

36 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

Remark 6.18. In a metric space, bounded sequences may not have convergent sub-
sequence.
For example, let (M, d) = (R, discrete metric). Let (xn ) = n then (xn ) is bounded as
{xn |n ≥ 1} ⊆ B2 (0). But (xn ) does not have any converge subsequence.
Suppose (xnk ) is convergent.
⇒ (xnk ) is Cauchy.
⇒ For  = 12 , ∃N such that,
1
d(xnp , xnq ) < 2
∀p, q ≥ N
Which is a contradiction as d(xnp , xnq ) = 1 p 6= q
∴ (xn ) does not have convergent subsequence.

Theorem 6.19. If xn → x in (M, d), then d(xn , y) → d(x, y) for any y ∈ M . More
generally, if xn → x, yn → y in (M, d) then d(xn , yn ) → d(x, y).

Proof. Let y ∈ M . We have to show that d(xn , y) → d(x, y) in R.

Let  > 0 since xn → x in (M, d), ∃N such that d(xn , x) <  ∀n ≥ N .
By triangle inequality, we get

d(xn , y) ≤d(xn , x) + d(x, y)

d(xn , y) − d(x, y) ≤d(xn , x)...............(1)

Similarly, d(x, y) ≤d(xn , x) + d(xn , y)

⇒ d(x, y) − d(xn , y) ≤d(xn , x)......(2)

From (1) & (2), we have,

|d(x, y) − d(xn , y)| ≤d(xn , x) <  ∀n ≥ N

⇒ d(xn , y) →d(x, y) ∀y ∈ M.

Suppose xn → x and yn → y in (M, d).

Let  > 0. Then ∃N1 and N2 such that

37 Dr.B. Surendranath Reddy

6 Open sets and Closed sets


d(xn , x) < 2
∀n ≥ N1

d(yn , y) < 2
∀n ≥ N2
Take N = max{N1 , N2 }.
Then for any n ≥ N

d(xn , yn ) ≤d(xn , x) + d(x, yn )

≤d(xn , x) + d(x, y) + d(y, yn )

⇒ d(xn , yn ) − d(x, y) ≤d(xn , x) + d(y, yn )

Similarly,

d(x, y) − d(xn , yn ) ≤d(xn , x) + d(y, yn )

|d(x, y) − d(xn , yn )| ≤d(xn , x) + d(y, yn )

 
< + =
2 2
∴ d(xn , yn ) → d(x, y)

Definition 6.20. Let (M, d) be a metric space. A set N is said to be a neighbourhood

of x ∈ M if ∃  > 0 such that B (x) ⊂ N .

Example 6.21. Every open ball B (x) is a neighbourhood of x.

Definition 6.22 (Open Set). Let (M, d) be a metric space. A subset U ⊂ M is

called an open set if U is neighbourhood of each of its points.
i.e. for any x ∈ U, ∃  > 0 such that, x ∈ B (x) ⊂ U

Example 6.23. In a metric space (M, d), the set M is open, as for any x ∈ M, B (x) ⊂
M ∀ > 0.
Trivially ∅ is open.

38 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

Example 6.24. In R, every open interval is an open set.

Proof. Let x ∈ (a, b). Then a < x < b.

Take  = min{x − a, b − x}. then  ≤ x − a and  ≤ b − x.
⇒a≤x− & x+≤b
⇒ (x − , x + ) ⊂ (a, b)
⇒ (a, b) is open.
Let x ∈ (a, ∞) take  = x − a then (x − , x + ) ⊂ (a, ∞) ⇒ (a, ∞) is open.
Similarly, (−∞, b) is also open.
Therefore, every open interval in R is open.

Remark 6.25. (0, 1) is an open set but [0, 1) is not open as 0 ∈ [0, 1) but (−, ) *
[0, 1) for any  > 0.

Example 6.26. In a discrete space, every subset is open.

Proof. Let (M, d) be a discrete space and A ⊂ M .

To show A is open, let x ∈ A.
Then x ∈ B1 (x) = {x} ⊂ A and hence A is open.

Theorem 6.27. Let (M, d) be a metric space. For any x ∈ M and

 > 0, B (x) is open. i.e. Every open ball is an open set.

Proof. Let y ∈ B (x) ⇒ d(x, y) < 

Take δ =  − d(x, y) then δ > 0.
We will show that, Bδ (y) ⊂ B (x).
Let z ∈ Bδ (y) ⇒ d(z, y) < δ
⇒ d(z, y) <  − d(x, y)
⇒ d(z, y) + d(x, y) < .
⇒ d(x, z) ≤ d(z, y) + d(x, y) < .

39 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

⇒ d(x, z) < 
⇒ z ∈ B (x)
⇒ Bδ (y) ⊂ B (x)
⇒ B (x) is open.

Theorem 6.28. Any open set U in M can be written as union of open balls contained
in U .
i.e. U = ∪{B (x) | B (x) ⊂ U }

Proof. Since B (x) ⊂ U ⇒ ∪B (x) ⊂ U .

Let x ∈ U . Since U is open ∃  > 0 such that x ∈ B (x) ⊂ U
⇒ x ∈ B (x) ⊂ ∪{B (x)|B (x) ⊂ U }
⇒ x ∈ ∪{B (x)|B (x) ⊂ U }
⇒ U ⊂ ∪{B (x)|B (x) ⊂ U }
∴ U = ∪{B (x)|B (x) ⊂ U }

Theorem 6.29. Arbitrary union of open sets is open, i.e., if (Uα )α∈Λ is any collection
S
of open sets, then Uα is open.
α∈Λ
S
Proof. Let V = Uα and x ∈ V .
α∈Λ
Then x ∈ Uα for some α ∈ Λ.
Since Uα is open, ∃  > 0 such that x ∈ B (x) ⊂ Uα
S
⇒ x ∈ B (x) ⊂ Uα ⊂ Uα = V
α∈Λ
⇒ x ∈ B (x) ⊂ V
Hence V is open.

Theorem 6.30. Finite intersection of open sets is open, i.e. If U1 , U2 , ..., Un are
n
T
open, then Ui is open.
i=1

40 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

n
T
Proof. Let V = Ui and x ∈ V .
i=1
Then x ∈ Ui for all i.
Since Ui is open, ∃ i > 0 such that x ∈ Bi (x) ⊂ Ui
Take  = {1 , ..., n }
Then  ≤ i ∀i
⇒ B (x) ⊂ Bi (x) ⊂ Ui ∀i
Tn
⇒ x ∈ B (x) ⊂ Ui = V
i=1
Hence V is open.

Remark 6.31. Arbitrary intersection of open sets need not be open. For example,
∞ ∞
the sets Un = (− n1 , n1 ) are open in R, but (− n1 , n1 ) = {0} is not open in
T T
Un =
n=1 n=1
R.

Theorem 6.32. If U is an open set of R, then U may be written as countable union

of disjoint open intervals that is U = ∪∞
i=1 In , where In = (an , bn ) and In ∩ Im =

φ ∀ n 6= m

Proof. We first show that for every x ∈ U , there exists maximal open interval Ix
containing x i.e. if x ∈ I ⊂ U then I ⊂ Ix .
Consider Ax = {a | (a, x] ⊂ U } and Bx = {b | [x, b) ⊂ U }
Let ax = inf Ax , and bx = sup Bx and Ix = (ax , bx ).
Since U is open, ∃ > 0 such that (x − , x + ) ⊂ U .
In particular, (x − , x] ⊂ U and [x, x + ) ⊂ U .
⇒ ax ≤ x −  and x +  ≤ bx
⇒ ax ≤ x −  < x < x +  ≤ b x
⇒ ax < x < b x
Now, let Ix = (ax , bx ). Since ax ≤ a < x ⇒ ax < x
⇒ x ∈ (ax , bx ) = Ix ⊂ U

41 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

To show that Ix is maximal, let x ∈ I ⊂ U , say I = (ai , bi ).

⇒ ai < x < bi then (ai , x] ⊂ (ai , bi ) = I ⊂ U
⇒ (ai , x] ⊂ U
⇒ ax ≤ ai
Similarly, bx ≥ bi ⇒ (ai , bi ) ⊂ (ax , bx )
⇒ I ⊂ Ix
Since for every x ∈ U, ∃ Ix such that x ∈ Ix ⊂ U
S
⇒ U = Ix (by Theorem 6.28)
Now we will show that, either Ix ∩ Iy = ∅ or Ix = Iy .
Suppose z ∈ Ix ∩ Iy .
⇒ z ∈ Ix and z ∈ Iy .
Also Ix ∪ Iy is an open interval.
Since x ∈ Ix ∪ Iy and Ix is maximal, we get Ix ∪ Iy ⊂ Ix .
Similarly, Ix ∪ Iy ⊂ Iy
Then Iy ⊂ Ix ∪ Iy ⊂ Ix ⊂ Ix ∪ Iy ⊂ Iy
⇒ Ix = Iy .
∴ U is the disjoint union of open intervals.
S
Since for every Ix , there exists a rational number rx ∈ Ix , U = Ix is countable.

Theorem 6.33. A set U in (M, d) is open iff whenever a sequence xn in M converges

to x ∈ U , then xn ∈ U , for all but finitely many.

Proof. Suppose U is open.

Let xn in M such that xn → x ∈ U .
Since x ∈ U & U is open, ∃  > 0 such that B (x) ⊂ U .
As xn → x, for this  > 0 ∃ N such that d(xn , x) <  ∀n ≥ N
⇒ xn ∈ B (x) ∀n ≥ N
⇒ xn ∈ B (x) ⊂ U ∀n ≥ N

42 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

⇒ xn ⊂ U ∀n ≥ N
⇒ xn ⊂ U for all but finitely many.
To prove converse, suppose U is not open.
⇒ ∃ x ∈ U such that B (x) * U ∀ > 0
In particular for  = n1 , ∃xn ∈ B 1 (x) but xn ∈
/U
n
1
⇒ d(xn , x) < n
and as n → ∞, d(xn , x) → 0 ⇒ xn →d x
∴ By assumption, xn ∈ U for all but finitely, which is a contradiction as xn ∈
/ U ∀n.
Hence U is open.

Definition 6.34. A set F ⊂ M is said to be closed if F c is open. i.e. for every

x ∈ F c , ∃  > 0 such that B (x) ⊂ F c

Example 6.35. 1. M and φ are closed sets as M c = φ and φc = M are open.

2. Every closed interval is a closed set as [a, b]c = (−∞, a) ∪ (b, ∞) and (−∞, b]c =
(b, ∞) are open sets.

Theorem 6.36. An arbitrary intersection of closed sets is closed.

Proof. Let (Fα )α∈Λ be closed sets.

⇒ Fαc is open for each α ∈ Λ.

S c
⇒ Fα is open.
α∈Λ
 c
T
⇒ Fα is open.
Tα∈Λ
⇒ Fα is closed.
α∈Λ

Theorem 6.37. Finite union of closed sets is closed.

Proof. Let F1 , F2 , ..., Fn be closed.

⇒ F1c , F2c , ..., Fnc are open.

43 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

n
Fic is open.
T
⇒
i=1
 n
c
S
⇒ Fi is open.
i=1
n
S
⇒ Fi is closed.
i=1

Remark 6.38. Arbitrary union of closed sets need not be closed.

For example, each Fn = 1 + n1 , 3
 
n ≥ 1 is closed in R
n ∞
1 + n1 , 3 = (1, 3] is not closed in R.
S S  
but Fn =
i=1 n=1

Theorem 6.39. In a discrete metric space, every subset is closed.

Proof. Let A ⊂ M . Since every subset of M is open, Ac is open.

⇒ A is closed.

Theorem 6.40. In a metric space, finite set is closed.

Proof. First we show that every singleton set {x} is closed.

Let y ∈ M \ {x} ⇒ y 6= x ⇒ d(x, y) > 0
Let  = d(x, y) then x ∈
/ B (y) as  = d(x, y) <  is false.
∴ B (y) ⊂ M \ {x}
∴ M \ {x} is open and hence {x} is closed.
n
S
Now let A = {x1 , x2 , ...xn } be a finite subset of M then A = {xi }
i=1
Since each {xi } is closed and finite union of closed set is closed, we get that A is
closed.

Remark 6.41. Sets are not doors as there are some sets which are neither closed nor
open. For example, [0, 1) is neither closed nor open in R.

Theorem 6.42. Given a set F in (M, d), the following are equivalent:

44 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

(i) F is closed.

(ii) If B (x) ∩ F 6= ∅ ∀  > 0, then x ∈ F .

(iii) If a sequence (xn ) in F converges to x ∈ M , then x ∈ F .

Proof. (i) ⇒ (ii) : Assume that B (x) ∩ F 6= ∅ ∀  > 0.

Suppose x ∈
/ F.
Then x ∈ F c and since F c is open, ∃  > 0 such that B (x) ⊂ F c
⇒ B (x) ∩ F = ∅, which is a contradiction. Hence x ∈ F .
(ii) ⇒ (iii) : Let (xn ) be a sequence in F such that xn → x for some x ∈ M .
Let  > 0. Since xn → x, ∃ N such that d(xn , x) <  ∀n ≥ N.
⇒ xn ∈ B (x) ∀n ≥ N
⇒ xn ∈ B (x) ∩ F ∀n ≥ N
∴ B (x) ∩ F 6= φ ∀  > 0
⇒x∈F
(iiI) ⇒ (i) : We will show that F is closed. i.e. F c is open.
Suppose F c is not open. Then ∃ x ∈ F c such that B (x) * F c ∀ > 0
In particular,for  = n1 , ∃ xn ∈ B 1 (x) but xn ∈
/ F c.
n

⇒ xn ∈ B 1 (x) and xn ∈ F
n

⇒ xn → x and xn ∈ F
Therefore x ∈ F , which is a contradiction.
Hence F c is open.

Definition 6.43. Interior Point

Let A ⊂ M then a point x ∈ A is called interior point of A. If ∃  > 0 such that
B (x) ⊂ A. The set of interior points of A is denoted by A◦ or int(A).
int(A) = {x ∈ A | B (x) ⊂ A for some  > 0}

45 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

Remark 6.44. Interior of A is the largest open set contained in A.

int(A) = ∪{U is open|U ⊂ A}

Example 6.45. Let A = [0, 1]. Show that intA = (0, 1).

Proof. Let 0 < x < 1. We will show that x is an interior point. since x ∈ (0, 1) and
(0, 1) is open. ∃  > 0 such that (x − , x + ) ⊂ (0, 1).
∴, x ∈ (x − , x + ) ⊂ (0, 1) ⊂ [0, 1]
∴, x ∈ (x − , x + ) ⊂ A
⇒ x is an interior point of A = [0, 1].
Now, x = 0 & x = 1 are not interior points of A. As (−, )& (1 − , 1 + ) * A
∴ intA = (0, 1)
In general, int([a, b]) = (a, b).

Remark 6.46. If A ⊂ M is open then int A = A.

Example 6.47. If A = { n1 |n ≥ 1} then intA = ∅.

Proof. For any x ∈ A, @  > 0 such that (x − , x + ) ⊂ A as (x − , x + ) contains

irrationals where as A does not. Therefore, int A = φ

Example 6.48. int(N) = int(Z) = int(Q) = int(R/Q) = φ

Proof. Let A = N or Z or Q then for any x ∈ A, and for any

 > 0, (x − , x + ) * A, as (x − , x + ) contains irrational where as A does not.
If A = R/Q then also for any x ∈ A and  > 0 (x − , x + ) * A.
As (x − , x + ) contains rational where as A does not.

Definition 6.49. Closure of set Let A ⊂ M then closure of A is the smallest closed
set containing A and it is denoted by cl(A) or A.
i.e.A = ∩{F is closed|A ⊂ F }

46 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

Remark 6.50. Clearly, A ⊂ A

Definition 6.51. Limit Point Let A ⊂ M then any point x ∈ M is said to be a

limit point of A if every neighbourhood of x contains a point of A other than x.
i.e. (B (x)/{x}) ∩ A 6= φ ∀ > 0

Remark 6.52. The set together with all limit points of A is the closure of A.
i.e. A = A ∪ limit points of A

Example 6.53. 1. A = (0, 1] then A = [0, 1], as 0 is the limit point of A.

2. A = (0, 1) then A = [0, 1], as 0 and 1 are the limit points of A.

3. A = { n1 |n ≥ 1} then A = A ∪ {0}, as ’0’ is the limit point of A.

Theorem 6.54. x ∈ A iff B (x) ∩ A 6= φ ∀  > 0.

Proof. Suppose x ∈ A. Suppose ∃  > 0 such that B (x) ∩ A = φ

⇒ A ⊂ (B (x))c . Since (B (x))c is closed and A is the smallest closed set containing
A we get, A ⊂ (B (x))c .
Since x ∈ A ⇒ x ∈ (B (x))c ⇒ x ∈
/ B (x).
Which is a contradiction.
⇒ B (x) ∩ A 6= φ ∀  > 0.
Suppose B (x) ∩ A 6= φ, ∀  > 0.
B (x) ∩ A 6= φ, ∀  > 0.
Since A is closed and B (x) ∩ A 6= φ, ∀  > 0, we get x ∈ A.

Corollary 6.55. x ∈ A iff there is a sequence (xn ) ⊂ A such that xn → x.

1
Proof. Let x ∈ A ⇒ B (x) ∩ A 6= φ, ∀  > 0. In particular,  = n
∃ xn such that
xn ∈ B (x) ∩ A
⇒ xn ∈ B (x) and xn ∈ A.

47 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

1
⇒ d(xn , x) < n
and (xn ) ⊂ A
⇒ xn → x
∴ ∃(xn ) ⊂ A such that xn → x.
Conversely, let (xn ) be a sequence in A such that xn → x. Given  > 0, ∃ N such that
d(xn , x) <  ∀n ≥ N
⇒ xn ∈ B (x) ∀n ≥ N
⇒ xn ∈ B (x) ∩ A ∀n ≥ N
⇒ B (x) ∩ A 6= φ ∀ > 0
⇒x∈A

Definition 6.56 (Relative Metric). Let (M, d) be a metric space and A ⊂ M then
0 0
A inherits a metric ’d’ by restriction of d i.e. d = d|A and (A, d ) is called relative
0
metric of (M, d). For convenience, (A, d ) will also be denoted by (A, d).

Definition 6.57 (Open ball in (A, d)). For any x ∈ A, open ball about x of radius
 > 0 is denoted by BA (x) and is defined as,

BA (x) ={y ∈ A|d(x, y) < }

= A ∩ {y ∈ M |d(x, y) < }

∴ BA (x) =A ∩ B (x)

Theorem 6.58. Let A ⊂ M then

1. Any set G ⊂ A is open in A iff G = A ∩ U for some open set U in M .

2. Any set F ⊂ A is closed in A iff F = A ∩ C where C is closed in M .

Proof. 1. Suppose φ 6= G ⊂ A is open. Let x ∈ G since G is open in A and

x ∈ G, ∃  > 0 such that x ∈ BA (x) ⊂ G
⇒ G = ∪{BA (x)|BA (x) ⊂ G}

48 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

⇒ G = ∪(A ∩ B (x))
⇒ G = A ∩ (∪B (x))
Then U = ∪B (x) is open in M .
∴ G = A ∩ U , where U is open in M .
Conversely, suppose G = A ∩ U where U is open in M . We have to show that,
G is open in A. Let x ∈ G ⇒ x ∈ A ∩ U ⇒ x ∈ A and x ∈ U .
Since U is open and x ∈ U
→ ∃  > 0 such that B (x) ⊂ U
⇒ A ∩ B (x) ⊂ A ∩ U
⇒ x ∈ BA (x) ⊂ G
⇒ G is open in A

2. Let F ⊂ A is closed then F c ⊂ A is open.

∴ by (1), F c = A ∩ U , where U is open in M ⇒ F = A ∩ U c .
Let C = U c then C is closed in M and F = A ∩ C.
Suppose F = A ∩ C where C is closed in M .
⇒ C c is open open in M and F c = A ∩ C c
⇒ F c is open in A.
⇒ F is closed in A.

Definition 6.59 (Boundary Points). Let A ⊂ M then any point x ∈ M is said to

be a boundary point of A if every ball around x intersects A and Ac , i.e., B (x) ∩ A 6=
∅ & B (x) ∩ Ac 6= ∅ ∀ > 0.
The set of all boundary point of A is denoted by bdry(A)
i.e. bdry(A) = {x ∈ M |B (x) ∩ A 6= φ and B (x) ∩ Ac 6= φ ∀ > 0}

Remark 6.60. 1. bdry(A) = Ā \ A◦

49 Dr.B. Surendranath Reddy

6 Open sets and Closed sets

2. cl(A) = bdry(A) ∪ int(A)

3. M = int(A) ∪ bdry(A) ∪ int(Ac )

Example 6.61. 1. A = [0, 1) then bdry(A) = {0, 1}

2. A = N, then bdry(N) = N

3. bdry(Q) = R

4. bdry(R/Q) = R

Definition 6.62 (Dense set). A set A is said to be dense in M if Ā = M .

i.e. For any x ∈ M, B (x) ∩ A 6= φ ∀  > 0

Example 6.63. 1. Q is dense in R as Q = R.

2. R/Q is dense in R.

Definition 6.64 (Nowhere Dense set). A set A is said to be nowhere dense in M

if int(cl(A)) = φ.
i.e. (A)◦ = φ
The set of natural number N is nowhere dense.

50 Dr.B. Surendranath Reddy

7 Continuity

7 Continuity
Definition 7.1. Let (M, d) and (N, ρ) are two metric spaces. Let f : M → N then f
is said to be continuous at x0 ∈ M if given  > 0, ∃ δ > 0 such that ρ(f (x), f (x0 )) < 
whenever d(x, x0 ) < δ. If f is continuous at each xo ∈ M , then f is continuous on
M.

Remark 7.2. f is continuous at x0 if, f (x) ∈ Bρ (f (x0 )) whenever x ∈ Bδd (x0 ),
i.e. f (Bδd (x0 )) ⊂ Bρ (f (x0 )).

Theorem 7.3. Given f : (M, d) → (N, ρ) the following are equivalent:

1. f is continuous on M .

2. For any x ∈ M , if xn → x in M then f (xn ) → f (x) in N .

3. If E is closed in N , then f −1 (E) is closed in M .

4. If V is open in N , then f −1 (V ) is open in M .

Proof. 1. → 2.
Suppose f is continuous at x. Then for given  > 0, there exists δ > 0 such that
ρ((f (x), f (y)) <  whenever d(x, y) < δ.
Let xn → x in M . Then for δ > 0, ∃ N such that d(xn , x) < δ ∀ n ≥ N
⇒ ρ(f (xn ), f (x)) <  ∀n ≥ N
⇒ f (xn ) → f (x) in N
2. → 3.
Let E is closed set in N . We have to show that f −1 (E) is closed in M . Let (xn ) ⊂
F −1 (E) such that xn → x in M . By 2., f (xn ) → f (x) in N .
(xn ) ⊂ f −1 (E) ⇒ (f (xn )) ⊂ E. Since E is closed & f (xn ) → f (x).
⇒ f (x) ⊂ E

51 Dr.B. Surendranath Reddy

7 Continuity

⇒ x ∈ f −1 (E)
⇒ f −1 (E) is closed.
3. → 4.
Let V is open in N .
⇒ V c is closed in N .
⇒ f −1 (V c ) is closed in M .
c
⇒ (f −1 (V )) is closed in M .
⇒ f −1 (V ) is open in M .
4. → 1.
To show that f is continuous on M . Let x ∈ M and  > 0 then
f (x) ∈ N & Bρ (f (x)) is open in N .
⇒ f −1 (Bρ (f (x))) is open in M .
Let V = Bρ (f (x)) then f (x) ∈ V ⇒ x ∈ f −1 (V ).
Since x ∈ f −1 (V )& f −1 (V ) is open.
⇒ ∃ δ > 0 such that Bδd (x) ⊂ f −1 (V )
⇒ f (Bδd (x)) ⊂ V = Bρ (f (x))
∴ f (Bδd (x)) ⊂ Bρ (f (x))
f is contninuous at x ∈ M .
⇒ f is continuous at x ∈ M
∴ f continuous on M .

 1, if x ∈ Q
Example 7.4. Show that χQ (x) = is no where continuous.
 0, if x ∈
/Q

Proof. Let x ∈ Q then χQ (x) = 1 & χQ (x) ∈ B 1 (1).

2
−1
⇒x∈ χQ (B 1 (1)).
2

⇒ Q ⊂ χ−1
Q (B 1 (1)).
2

Let y ∈ χ−1
Q (B 12 (1)).

52 Dr.B. Surendranath Reddy

7 Continuity

⇒ χQ (y) ∈ B 1 (1).
2

⇒ χQ (y) = 1
⇒y∈Q
⇒ χ−1
Q (B 1 (1)) ⊂ Q
2

∴ Q= χ−1
Q (B 21 (1))

Since Q is not open.

⇒ χ−1
Q (B 1 (1)) is not open.
2

⇒ χQ is not continuous at any x ∈ Q.

Let x ∈
/ Q ⇒ χQ (x) = 0 and χQ (x) ∈ B 1 (0).
2
−1
⇒ x ∈ χQ (B 1 (0)).
2

⇒ R/Q ⊂ χ−1
Q (B 12 (0)).

Also χ−1
Q (B 1 (0)) ⊂ R/Q
2

∴ R/Q = χ−1
Q (B 12 (0)) is not open.
∴ χQ is not continuous at any x ∈
/ Q.
∴ χQ is nowhere continuous.

Example 7.5. Let f : N → R. Then f is continuous.

1
Proof. Let n ∈ N and  > 0 take δ = 2
then Bδ (n) = {n} as
f (n) ∈ Bρ (f (n))
⇒ f (Bδ (n)) ⊂ Bρ (f (n))
∴ f is continuous at 0 n0 .
⇒ f is continuous on N

Definition 7.6. Isometry

A function f : (M, d) → (N, ρ) is called an isometry if f preserves distances
i.e. d(x, y) = ρ(f (x), f (y)) ∀ x, y ∈ M .

Exercise

53 Dr.B. Surendranath Reddy

7 Continuity

1. Show that every isometry f : M → N is continuous. Exercise

2. Prove that a function f : R → N is continuous if and only if f is constant.

Example 7.7. Let f : R → R be a continuous show that {x|f (x) > 0} is an open
subset of R and {x|f (x) = 0} is a closed subset of R.

Proof. {x|f (x) > 0} = {x|f (x) ∈ (0, ∞)} = {x|x ∈ f −1 (0, ∞)}
Since (0, ∞) is open and f is continuous, f −1 (0, ∞) is open.
⇒ {x|f (x) > 0} is open. And {x|f (x) = 0} = f −1 (0).
Since {0} is closed in R, f −1 (0) is closed in R.
⇒ {x|f (x) = 0} is closed.

Definition 7.8. Let (M, d) be a metric space and A ⊂ M be a non empty. Then for
any x ∈ M , distance from x to A is denoted by d(x, A) and is defined as d(x, A) =
inf{d(x, a)|a ∈ A}.

Remark 7.9. If x ∈ A then d(x, A) = 0. But the converse is not true as d(x, Q) =
0, ∀ x ∈ R.

Theorem 7.10. d(x, A) = 0 if and only if x ∈ A.

Proof. d(x, A) = 0 ⇔ inf{d(x, a)|a ∈ A} = 0

1
⇔ for  = n
∃, xn ∈ A such that d(x, xn ) < n1 .
⇔ xn → x where xn ∈ A.
⇔ xn ∈ A.

Example 7.11. Let (M, d) be a metric space and φ 6= A ⊂ M . Then the map
f : M → R defined by f (x) = d(x, A) is continuous.

54 Dr.B. Surendranath Reddy

7 Continuity

Proof. Let x ∈ M and  > 0.

First we show that |d(x, A) − d(y, A)| ≤ d(x, y) for any a ∈ A.
d(x, a) ≤ d(x, y) + d(y, a), ∀y ∈ M
⇒ inf a∈A d(x, a) ≤ d(x, y) + inf a∈A d(y, a)
⇒ d(x, A) ≤ d(x, y) + d(y, A)
⇒ d(x, A) − d(y, A) ≤ d(x, y)
Similarly, d(y, A) − d(x, A) ≤ d(x, y)
⇒ |d(x, A) − d(y, A)| ≤ d(x, y)
Take δ = , suppose y ∈ M such that d(x, y) < δ then

|f (x) − f (y)| =|d(x, A) − d(y, A)|

≤d(x, y) < δ = 

⇒ |f (x) − f (y)| <

∴ f is continuous at x ∈ M .
⇒ f is continuous on M .

Remark 7.12. If f : M → R is continuous then the set E = {x|f (x) = 0} is closed.

Conversely, if E is a closed set in M , then E is the zero of the function f : M → R
given by f (x) = d(x, E) which is continuous.
i.e. E = {x|f (x) = d(x, E) = 0}.

Definition 7.13 (Homeomorphism). We say that two metric spaces (M, d) and
(N, ρ) are homeomorphism if ∃ a bijective map f : M → N such that f & f −1 are
continuous. This map f is called homeomorphism from M to N .

Remark 7.14. If f : M → N is a homeomorphism then f −1 : N → M is a homeo-

morphism.

Example 7.15. 1. Every metric space (M, d) is homeomorphic to itself by iden-

tity map i.e.f : M → M such that f (x) = x is a homeomorphism.

55 Dr.B. Surendranath Reddy

7 Continuity

2. (0, 1] and [1, ∞) are homeomorphic by the map x 7→ x1 .

3. In R, any two intervals that look alike are homeomorphic.

d−c
(define f : [a, b) → [c, d] by f (x) = c + m(x − a) where m = b−a
then f is
bijective and f and f −1 are continuous.)

Example 7.16. Exercise

1. Prove that (N, usual metric) is homeomorphic to { n1 |n ≥ 1}, usual .

2. Check that the relation ”is homeomorphic to” is an equivalence relation on pairs
of metric space.

Example 7.17. (R,usual metric) and (R,discrete metric) are not homeomorphic.

Proof. Suppose these two metrics are homeomorphic then there exists a bijective map
f : R → R such that f and f −1 are continuous. Since {f (x)} is open in discrete space
and f is continuous.
⇒ f −1 (f (x)) is open in (R, |.|) space.
⇒ {x} is open in (R, |.|) which is a contradiction.
∴ (R, |.|) and (R,discrete) are not homeomorphic.

Theorem 7.18. Let f : (M, d) → (N, ρ) be a bijective map then following are equiv-
alent.

1. f is homeomorphism.

2. xn → x ⇔ f (xn ) → f (x).

3. G is open in M ⇔ f (G) is open in N .

4. E is closed in M ⇔ f (E) is closed in N .

56 Dr.B. Surendranath Reddy

7 Continuity

Proof. 1. → 2.
Let f be a homeomorphism and xn → x.
Then f (xn ) → f (x) as f is continuous.
Conversely suppose f (xn ) → f (x).
Then f −1 (f (xn )) → f −1 (f (x)) as f −1 is continuous.
⇒ xn → x.
2. → 3.
By the assumption, f and f −1 are continuous as xn → x ⇒ f (xn ) → f (x).
and xn → x ⇒ f −1 (xn ) → f −1 (x).
Let G be open in M . Since f −1 : N → M is continuous & G is open, we get that
(f −1 )−1 (G) is open in N .
⇒ f (G) is open in N .
Conversely, suppose f (G) is open in N . Since f : M → N is continuous & f (G) is
open.
⇒ f −1 (f (G)) is open in M .
⇒ G is open in M .
3. → 4.
E is closed in M .
⇔ E c is open in M .
⇔ f (E c ) is open in N .
⇔ (f (E))c is open in N .
⇔ f (E) is closed in N .
4. → 1.
To show f is homeomorphism we show that f and f −1 are continuous.
Let E ⊂ N is closed.
⇒ E = f (f −1 (E)) closed in N .
⇒ f −1 (E) closed in M .

57 Dr.B. Surendranath Reddy

7 Continuity

⇒ f is continuous.
To show f −1 continuous. Let F ⊂ M is closed.
⇒ f (E) is closed in N .
⇒ (f −1 )−1 (E) is closed in N .
⇒ f −1 is continuous.
⇒ f −1 is a homeomorphism.

Lemma 7.19. Let f : L → M & g : M → N . If f is continuous at x ∈ L and g is

continuous at f (x) ∈ M then g ◦ f : L → N is continuous at x.

Proof. To show g ◦ f is continuous at x ∈ L. Let (xn ) ⊂ L such that xn → x. Since

f is continuous at x ⇒ f (xn ) → f (x) .
Since g is continuous at f (x) ⇒ g (f (xn )) → g (f (x))
i.e.⇒ (g ◦ f )(xn ) → (g ◦ f )(x).
⇒ g ◦ f is continuous at x ∈ L.

Definition 7.20. Space of Continuous Functions

Let (M, d) be a metric space. Let C(M ) = {f : M → R|f is continuous}. Define for
any f, g ∈ C(M )
(f + g)(x) = f (x) + g(x) and (af )(x) = a(f (x), a∈R
then C(M ) is a vector space over R.

Theorem 7.21. If f, g : M → R are continuous then f ±g, f g, af, min{f, g}, max{f, g}
f
are continuous. Also if g(x) 6= 0 for all x, then g
is continuous.

Proof. Let (xn ) be a sequence such that xn → x . Since f is continuous f (xn ) → f (x)
and g is continuous implies g(xn ) → g(x).
⇒ f (xn ) + g(xn ) → f (x) + g(x)
⇒ (f + g)(xn ) → (f + g)(x)

58 Dr.B. Surendranath Reddy

8 Connectedness

min{f, g}(x) = min{f (x), g(x)}

For any a, b ∈ R, we have min{a, b} = 12 (a + b − |a − b|)
and max{f, g} = 21 (a + b + |a − b|).
Since f and g are continuous f + g, f − g, |f − g| are also continuous.
⇒ min{f, g} = 12 [(f + g) − |f − g|] is continuous.
Similarly, max{f, g} is continuous.
Since (f g)(xn ) = f (xn )g(xn ) → f (x)g(x) = (f g)(x), we get that f g is continuous.
f
If g(x) 6= 0 for all x, then g
= (f )( g1 ) is continuous.

8 Connectedness
Definition 8.1. Connectedness
A metric space (M, d) is called disconnected if there exists non empty disjoint open
set A & B such that M = A ∪ B
i.e. ∃ A 6= φ, B 6= φ, A ∩ B = φ such that M = A ∪ B the pair (A, B) is called
disconnection or separation of M .
If there is no disconnection of M , then M is called connected.

Example 8.2. Let (M, d) be a discrete metric space with at least two elements. Then
(M, d) is disconnected.

Proof. For any x ∈ M, M = {x} ∪ M/{x}.

Here {x} =
6 φ, M/{x} =
6 φ and are disjoint open set.
⇒ M is disconnected.

Definition 8.3. Clopen Set

A set which is both open as well as closed is called Clopen set.

Remark 8.4. φ and M are always clopen and are called trivial clopen sets. If A 6= φ
&A 6= M is clopen, then it is called non trivial clopen set.

59 Dr.B. Surendranath Reddy

8 Connectedness

Theorem 8.5. M is connected if and only if M contains no non trivial clopen sets.

Proof. Assume that M is connected. Suppose M has a non trivial clopen set A. i.e.
A 6= ∅, A 6= M and A is open & closed.
⇒ A and Ac are open and M = A ∪ Ac which is a contradiction.
Conversely, assume that M has no non trivial clopen sets.
Suppose M is disconnected then there exists two non empty disjoint open sets A and
B such that M = A ∪ B
Since Ac = B, we get Ac is open ⇒ A is closed.
∴ A is a non trivial clopen set, which is a contradiction.
Hence M is connected.

Lemma 8.6. Let E be a subset of metric space M . If U and V are disjoint open
set in E then there are disjoint open sets A and B in M such that U = A ∩ E and
V =B∩E

Proof. Since U is open in E. For every x ∈ U, ∃ x > 0 such that

Bx (x) ⊂ U
⇒ E ∩ Bx (x) ⊂ U
Similarly, V is open in E for every y ∈ V ∃ δy > 0 such that
Bδy (y) ⊂ V
⇒ E ∩ Bδy (y) ⊂ V
Since U ∩ V = φ, ⇒ E ∩ Bx (x) ∩ Bδy (y) = ∅.
In particular, we have B x (x) ∩ Bδ y (y) = ∅, ∀x ∈ U, y ∈ V
2 2

Let A = ∪{B x (x) | x ∈ U } and B = ∪{Bδ y (y) | y ∈ V }. Then A ∩ B = ∅.

2   2 
E ∩ A = E ∩ ∪x∈U {B x (x)} = ∪x∈U B x (x) ∩ E
2 2

E ∩ A = U and similarly we get, E ∩ B = V.

Remark 8.7. Suppose E ⊂ R is disconnected.

Then E = U ∪ V, U and V are open in E.

60 Dr.B. Surendranath Reddy

8 Connectedness

⇒ U = E ∩ A, V = E ∩ B for some open sets A and B in E

⇒ E = (E ∩ A) ∪ (E ∩ B)
E = E ∩ A ∪ B ⊂ (A ∩ B)

Definition 8.8. Disconnected subset

Let (M, d) be a metric space. A subset E is said to be disconnected if there exists
non-empty disjoint open sets A and B in M such that E ⊂ A ∪ B.

Theorem 8.9. The closed interval [a, b] is connected in R.

Proof. Suppose [a, b] disconnected. Then there exists two non- empty disjoint open
sets A and B in R such that [a, b] = A ∪ B.
If x ∈ A, then x ≤ b and so A is bounded above.
By LUB axiom, sup A exists, say c = sup A.
We show that a ≤ c ≤ b.
If c < a, then A = ∅, which is not true.
Similarly, if b < c, then B = ∅, which is also not true.
Thus c ∈ [a, b].
Now we claim that c ∈
/ A and c ∈
/ B.
Suppose c ∈ A. Since A is open, there exists  > 0 such that [c, c + ) ⊂ A
 
Then c + 2
∈ A and thus c + 2
≤ c, which is absurd.
Now if c ∈ B, as B is open, there exists  > 0 such that (c − , c] ⊂ B

Then c − 2
∈ B and thus c ≤ c − 2 , which is absurd again.
Therefore c ∈ [a, b], but c ∈
/ A and c ∈
/ B, which is a contradiction.
∴ [a, b] is connected in R.

Theorem 8.10. A subset E of R containing more than one point is connected if and
only if whenever x, y ∈ E with x < y then [x, y] ⊂ E.
i.e. Connected subset of R are precisely intervals.

61 Dr.B. Surendranath Reddy

8 Connectedness

Proof. Suppose E is connected.

Suppose there exists x, y ∈ E with x < y but [x, y] 6⊂ E.
Then ∃ z ∈ [x, y] such that z ∈
/E
⇒ E ⊂ (−∞, z) ∪ (z, ∞).
Let A = (−∞, z) &B = (z, ∞) then A ∩ B = φ,
E ∩ A 6= φ, E ∩ B 6= φ and E ⊂ A ∪ B.
⇒ E is disconnected, which is a contradiction.
Hence if x, y ∈ E with x < y then [x, y] ⊂ E. Conversely, assume that E satisfies the
given condition but E is not connected.
Then there exists two disjoint non-empty open sets A and B such that E ⊂ A ∪ B .
Since E ∩ A 6= φ, E ∩ B 6= φ, let a ∈ E ∩ A and b ∈ E ∩ B.
Then a, b ∈ E. suppose a < b. Then [a, b] ⊂ E ⊂ A ∪ B
⇒ [a, b] ⊂ A ∪ B
⇒ [a, b] is disconnected, which is contradiction.
∴ E is a connected.

Corollary 8.11. R is connected as [x, y] ⊂ R, ∀x, y ∈ R

Theorem 8.12. M is disconnected if and only if there exists a continuous map f :

M → {0, 1}, which is onto.

Proof. Suppose M is disconnected then there exists two non-empty open sets A & B
such that M = A ∪ B. 
 0, x ∈ A
Define f : M → {0, 1} by f (x) =
 1, x ∈ B
Since f is continuous on A & B and A ∩ B = φ, f is continuous on A ∪ B = M . Also
f is onto as A, B 6= ∅.
Conversely, suppose there exists a continuous function f : M → {0, 1} which is onto.

62 Dr.B. Surendranath Reddy

8 Connectedness

⇒ f (M ) = {0, 1}
Take A = f −1 ({0}) and B = f −1 ({1}), then A and B are open sets in M and
A ∩ B = φ.
Also A ∪ B = f −1 ({0}) ∪ f −1 ({1}) = f −1 ({0, 1}) = M
⇒ M is disconnected.

Theorem 8.13. Exercise

M is connected iff every continuous map f : M → {0, 1} is constant.

Theorem 8.14. If A and B are connected then A × B is connected.

Proof. Let f : A × B → {0, 1} be a continuous map.

To show that A × B is connected it is enough to show that f is constant.
Let (a0 , b0 ) ∈ A × B be a fixed point and (a, b) be any arbitrary point in A × B.
We show that f (a, b) = f (a0 , b0 ).
0 0 0
Consider the map fa : B → {0, 1} defined by fa (b ) = f (a, b ), ∀b ∈ B.
Since f is continuous on A × B, we get that fa is continuous on B.
As B is connected, fa is constant.
0 0
Then fa (b ) = k1 , ∀ b ∈ B
0 0
⇒ f (a, b ) = k, ∀ b ∈ B In particulary, f (a, b) = f (a, b0 ) (b, b0 ∈ B) Similarly, the
0 0
map fb0 : A → {0, 1} given by fb0 (a ) = f (a , b0 ) is continuous on A.
Since A is connected, fb0 is constant.
0 0 0
Then fb0 (a ) = f (a , b0 ) = k2 , ∀ a ∈ A
In particular, f (a0 , b0 ) = f (a, b0 ) (a, a0 ∈ A) Hence f (a, b) = f (a, b0 ) = f (a0 , b0 ).
Hence A × B is connected.

Corollary 8.15. 1. R2 is connected. In general, Rn is connected.

2. [0, 1] × [0, 1] is connected.

63 Dr.B. Surendranath Reddy

8 Connectedness

Theorem 8.16. (Continuous Image of a connected set is connected.) Let

f : (M, d) → (N, ρ) be continuous and E be a subset of M . If E is connected then
f (E) is connected.

Proof. Suppose f (E) is disconnected then thereexists g : f (E) → {0, 1} which is a

continuous and onto.
⇒ g ◦ f : E → {0, 1} continuous and onto. Since f is continuous on E and g is
continuous on f (E). Since g is onto, g ◦ f is also onto.
⇒ E is disconnected which is a contradiction. Hence f (E) is connected.

Example 8.17. Show that the intervals that look, different are not homeomorphic.

Proof. We will show that (a, b] and (a, b) are not homeomorphic. Suppose they are
homeomorphic then thereexists f : (a, b] → (a, b) which is a continuous. Let c = f (b)
then a < c < b. Since f is bijective
f ((a, b] \ {b}) = (a, b) \ {c} = (a, c) ∪ (c, b)
Since (a, b] \ {b} = (a, b) connected and f is continuous, f (a, b) is connected, which
is a contradiction to (a, c) ∪ (c, b) is disconnected.
∴ (a, b] and (a, b) are not homeomorphic.

Example 8.18. R2 is not homeomorphic to R.

Proof. Suppose f : R2 → R is homeomorphic. Since R2 \ {(x, y)} is connected but

R \ f (x, y) is not connected, we get a contradiction.
∴ R2 is not homeomorphic to R.

Example 8.19. [0, 1] × [0, 1] is not homeomorphic to [0, 1]

Proof. Let [0, 1] × [0, 1] → [0, 1] is bijective and continuous also.

Take c ∈ [0, 1] then thereexists (x, y) such that f (x, y) = c then f ([0, 1] × [0, 1] \ {(x, y)})
is connected but f ([0, 1] × [0, 1] \ {(x, y)}) = [0, 1] \ c = [0, c] ∪ [c, 1] is not connected.
Therefore, [0, 1] × [0, 1] is not homeomorphic to [0, 1].

64 Dr.B. Surendranath Reddy

9 Totally Bounded

9 Totally Bounded
Definition 9.1. Totally Bounded
A subset A of M is said to be totally bounded if for given  > 0,
∃ x1 , x2 , ...xn ∈ M such that A ⊂ ∪ni=1 B (xi )

Remark 9.2. {x1 , x2 , ...xn } is called as -net for A.

Lemma 9.3. If A is totally bounded, for given  > 0, ∃ y1 , y2 , ...yn ∈ A such that
A ⊂ ∪ni=1 B (yi ).

Proof. Since A is totally bounded ∃ x1 , x2 , ...xn ∈ M such that A ⊂ ∪ni=1 B 2 (xi ).

Without loss of generality, we can assume A ∩ B 2 (xi ) 6= φ ∀ i = 1, 2, ...n. Let
yi ∈ A ∩ B 2 (xi ).Let a ∈ A then ∃ j such that a ∈ B 2 (xj ) then
 
d(a, yj ) ≤ d(a, xj ) + d(xj , yj ) < 2
+ 2
= .
⇒ a ∈ B (yj )
∴ A ⊂ ∪ni=1 B (yi ).

Lemma 9.4. If A is totally bounded if and only if for given  > 0,

∃ A1 , A2 , ...An ⊂ A with diam(Ai ) < , for every i, such that A ⊂ ∪ni=1 Ai .

Proof. A is totally bounded. Let  > 0 then ∃ x1 , x2 , ..., xn ∈ A such that A ⊂

∪ni=1 B 3 (xi ).
Take Ai = A ∩ B 3 (xi )


⇒ A = A ∩ ∪ni=1 B 3 (xi ) = ∪ni=1 A ∩ B 3 (xi ) = ∪ni=1 Ai
and diam(Ai ) = diam(A ∩ B 3 (xi )) ≤ diam B 3 (xi )) ≤ 2( 3 ) < .

Conversely, suppose ∃ A1 , A2 , ...An ⊂ A with diam(Ai ) = 0 <  such that A ⊂ ∪ni=1 Ai

then Ai ⊂ B (xi ) for any xi ∈ Ai .
∴ A ⊂ ∪ni=1 Ai ⊂ ∪ni=1 B (xi ).
⇒ A is totally bounded.

65 Dr.B. Surendranath Reddy

9 Totally Bounded

Example 9.5. In a metric space any finite set is totally bounded.

Proof. Let A = {x1 , x2 , ...xn } take Ai = {xi }. Let  > 0 then A = ∪ni=1 Ai and
diam(Ai ) = 0 < .
⇒ A is totally bounded.

Example 9.6. In a discrete metric space M , a subset A is totally bounded if and

only if A is finite.

Proof. A is finite then A is totally bounded. Let A = {x1 , x2 , ...xn } take Ai = {xi }.
Let  > 0 then A = ∪ni=1 Ai and diam(Ai ) = 0 < .
⇒ A is totally bounded.
Conversely, suppose A is totally bounded then for  = 1, ∃x1 , x2 , ...xn ∈ A such that
A ⊂ ∪ni=1 B1 (xi ). But in discrete space. B1 (x) = {x}.
∴ A ⊂ {x1 , x2 , ...xn }.
⇒ A is finite.

Example 9.7. Every totally bounded set is bounded but not the converse.

Proof. Let A be totally bounded then for  = 1, ∃ x1 , x2 , ...xn ∈ A such that A ⊂

∪ni=1 B1 (xi ). Let r = max{d(x1 , xi )|1 ≤ i ≤ n} + 1.
We show that A ⊂ Br (x1 ). Let a ∈ A ⇒ ∃j such that a ∈ B1 (xj )
∴ d(a, x1 ) ≤ d(a, xj ) + d(xj , x1 ) < 1 + r − 1 = r.
⇒ a ∈ Br (x1 )
∴ A ⊂ Br (x1 )
⇒ A is bounded.

Remark 9.8. Converse of the above example is not true in genral,

e.g. consider the discrete space (R, discrete). Let A = [0, 1] then A is bounded as
A ⊂ B2 (0) but A is not totally bounded as A is infinite.

66 Dr.B. Surendranath Reddy

9 Totally Bounded

Lemma 9.9. Let (xn ) be a sequence in (M, d) and A = {xn |n ≥ 1} be its range.

1. If (xn ) is Cauchy then A is totally bounded.

2. If A is totally bounded, then (xn ) has a Cauchy subsequence.

Proof. 1. Let  > 0. Since (xn ) is Cauchy, ∃ k ∈ N such that

d(xn , xm ) < 2 ∀ n, m ≥ k.
Let A1 = {x1 }, A2 = {x2 }, ...Ak−1 = {xk−1 } and Ak = {xn |n ≥ k} then
A = ∪ni=1 Ai and diam(Ai ) = 0 < , 1 ≤ i ≤ k − 1.
For any xn , xm ∈ Ak ,

d(xn , xm ) <
2

⇒ sup{d(xn , xm )} ≤ < 
2
⇒ diam(Ak ) <.

∴ A = ∪ki=1 Ai such that diam(Ai ) < .

⇒ A is totally bounded.

2. Let A be a totally bounded set. Case1:If A is finite, then there are infinitely
many xn ’s which are equal say to k and hence this elements form as subsequence
which is a convergent and so Cauchy.
Case2: So assume that A is infinite. For  = 1, A can be covered by finitely
many sub set with diameter less than ’1’. Since A is infinite, these is at least
one subset say A1 of A which is infinite with diam(A1 ) < 1. Since A1 ⊂ A, A1
is totally bounded then for  = 21 , A1 is covered by finitely many subsets of A1
with diameter less than 12 . Since A1 is infinite there is atleast one subset of A1 ,
say A2 which is infinite and diam(A2 ) < 21 . By repeating this process, we get
A1 ⊃ A2 ⊃ A3 ... such that each Ak is infinite and diam(Ak ) < k1 .
Let xnk ∈ Ak ∀k. Let  > 0 then by Archimedean property, ∃ m ∈ N such

67 Dr.B. Surendranath Reddy

9 Totally Bounded

1
that m
< . Let p, q ≥ m
1 1
d(xnp , xnq ) ≤ diam(Aq ) < q
≤ m
< .
∴ d(xnp , xnq ) <  ∀p, q ≥ m.
⇒ xk is Cauchy subsequence.

Theorem 9.10. A set A is totally bounded if and only if every sequence in A has a
Cauchy subsequence.

Proof. From above lemma first part follows. Now we will prove second part. Suppose
every sequence in A has a Cauchy subsequence. Let A is not totally bounded then
∃  > 0 such that A can not be covered by finitely many subsets of A of diameter less
than . Therefore, there are infinitely many elements of A which are not in those above
sets then we get a subsequence xnk with these elements for any p 6= q, d(xnp , xnq ) ≥ .
⇒ (xnk ) is not a Cauchy. Which is a contradiction.
∴ A is totally bounded.

Corollary 9.11. Bolzano-Weierstrass Theorem

Every bounded infinite subsets of R has a limit point in R.

Proof. Let A ⊂ R be a bounded infinite set, we get a sequence (xn ) with distinct
elements in R. Since every bounded set is totally bounded.
⇒ A is totally bounded.
∴ (xn ) has a subsequence (xnk ) which is a Cauchy in R.
⇒ (xnk ) converges to some x ∈ R.
⇒ x is a limit point of A in R.

Definition 9.12. Complete Metric Space

Let (M, d) be a metric space then M is said to be complete if every Cauchy sequence

68 Dr.B. Surendranath Reddy

9 Totally Bounded

in M converges in M .
e.g. R is complete.

Example 9.13. Every discrete metric space is complete.

Proof. (xn ) be a Cauchy sequence in discrete space (M, d), for  = 12 , ∃ k ∈ N such
that d(xn , xm ) < 21 ∀ n, m ≥ k.
⇒ d(xn , xm ) = 0 ∀ n, m ≥ k
⇒ xn = xm ∀ n, m ≥ k
⇒ xn = xk ∀ n ≥ k
⇒ |xn − xk | = 0 <  ∀ n ≥ k
⇒ xn → xk
∴ M is complete.

Example 9.14. (0, 1) is not complete as ( n1 ) is Cauchy sequence in (0, 1) but 1

n
does
not converge in (0, 1).

Theorem 9.15. Let (M, d) be a complete metric space and (A, d) is subset of (M, d)
then (A, d) is complete if and only if A is closed.

Proof. Let (A, d) be complete subset of (M, d). Let (xn ) ⊂ A such that xn → x ∈ M .
⇒ xn converges in M and hence Cauchy. Since (A, d) is complete and (xn ) is Cauchy,
converges in A.
⇒ x ∈ A.
⇒ A is closed.
Conversely, suppose A is closed. Let xn be a Cauchy sequence in A.
⇒ (xn ) is Cauchy in M .
⇒ xn → x ∈ M
Since A is closed, x ∈ A.
∴ (A, d) is complete.

69 Dr.B. Surendranath Reddy

9 Totally Bounded

Example 9.16. [0, 1], [0, ∞], N, Z are complete.

Theorem 9.17. Let (M, d) be a metric space. Then the following are equivalent.

1. M is complete.

2. (Nested theorem) Let F1 ⊃ F2 ⊃ F3 ⊃ ... be a decreasing sequence of non

empty closed sets in M with diam(Fn ) → 0 then ∩∞
n=1 Fn 6= φ and contains

exactly one point.

3. (Bolzano-Wierstrass Theorem) Every infinite totally bounded subset of M

has a limit point in M .

Proof. 1. → 2.
Suppose M is complete. Let F1 ⊃ F2 ⊃ F3 ... decreasing sequence of non empty closed
subset of M such that diam(Fn ) → 0. Since each Fn 6= φ. Let xn ∈ Fn ∀n
∴ {xk |k ≥ n} ⊂ Fn ∀n
⇒ diam{xk |k ≥ n} ≤ diam(Fn ) → 0 as n → ∞
∴ diam{xk |k ≥ n} → 0 as n → ∞
⇒ (xn ) is Cauchy and M is complete, (xn ) converges to x ∈ M . Since each Fn is
closed, x ∈ Fn ∀n ∩∞
n=1 Fn 6= φ.

Let x, y ∈ ∩∞
n=1 Fn then d(x, y) ≤ diam(Fn ) → 0 n → ∞

⇒ d(x, y) = 0
⇒ x = y.
2. → 3.
Let A be an infinite totally bounded subsets of M . Since A is infinite and totally
bounded A contains a Cauchy sequence (xn ) of distinct points.
Let An = {xk |k ≥ n} then A1 ⊃ A2 ⊃ A3 .... Since (xn ) is Cauchy, diam(An ) → 0 as
n → ∞ then Ā1 ⊃ Ā2 ⊃ ...
with diam(Ā)n = diam(An ) → 0.

70 Dr.B. Surendranath Reddy

9 Totally Bounded

∴ by 2., ∩∞ ∞
n=1 Ān 6= φ. Let x ∈ ∩n=1 Ān

⇒ x ∈ Ān , ∀n
Since (xn ) ⊂ An ⊂ Ān → xn ∈ Ān
d(xn , x) ≤ diam(Ā) →
⇒ xn → x
∴ x is a limit point of A.
3. → 1.
Let (xn ) be a Cauchy sequence in M . Let A = {xn |n ≥ 1}. Since xn is Cauchy, A is
totally bounded. If A is finite, then there are infinitely many xk ’s which are equal &
hence we get subsequence which converges.
Suppose A is infinite, then by 2., A has a limit point say x where x ∈ M .
∴ For  = k1 , let xnk ∈ B 1 (x) ∩ A/{x}
k

⇒ xnk ∈ B 1 (x) and (xnk ) ⊂ A

k
1
⇒ d(xnk , x) < k
→ 0 as k → ∞
∴ xnk → x.
∴ (xnk ) is a convergent subsequence of (xn ). Since (xn ) is Cauchy, xn → x.
∴ M is Complete.

Example 9.18. Show that (l2 , ||.||2 ) is complete.

Proof. Let (fn ) be a Cauchy sequence in l2 . Let  > 0 then ∃ n0 ∈ N such that
||fn − fm ||2 <  ∀ n, m ≥ n0
We show that for every k ∈ N, (fn (k))∞
n=1 .

71 Dr.B. Surendranath Reddy

9 Totally Bounded

Let p, q ≥ n0
∞
X
|fp (k) − fq (k)|2 ≤ |fp (x) − fq (x)|2
x=1
∞
X
= |(fp − fq )(x)|2
x=1

=||fp − fq ||22 < 

∴ |fp (k) − fq (k)| ≤

⇒ (fn (k)) is Cauchy in R and hence converges.

Let f (k) = limn→ fn (k)
Step 2: We will show that f ∈ l2 . Since (fn ) is Cauchy is bounded. i.e. ||fn || ≤
B ∀n ≥ 1. Fix n0 ∈ N.
Consider
n0
X n0
X
2
|f (k)| = | lim fn (k)|2
n→∞
k=1 k=1
n0
X
≤ lim |fn (k)|2
n→∞
k=1

= lim ||fn ||22 ≤ B 2

n→∞
n0
X
∴ |f (k)|2 ≤ B 2
k=1

As n0 is arbitrary
P∞ 2
k=1 |f (k)| < ∞

⇒ f ∈ l2

72 Dr.B. Surendranath Reddy

9 Totally Bounded

Step 3: We show that fn → f in l2 . Fix n0 ∈ N

n0
X n0
X
|fn (k) − f (k)|2 = |fn (k) − lim fm (k)|2
m→∞
k=1 k=1
n0
X ∞
X
2
lim |fn (k) − fm (k)| ≤ lim |fn (k) − fm (k)|2
m→∞ m→∞
k=1 k=1

= lim ||fn − fm ||22

m→

Since fn is Cauchy for given  > 0 ∃ p ∈ N such that ||fn − fm ||2 <  ∀ n, m ≥ p
∴ For n ≤ p
Pn0 2 2
k=1 |fn (k) − f (k)| < 
P∞ 2 2
k=1 |fn (k) − f (k)| < 

⇒ ||fn − f ||22 < 2

∴ fn → f
∴ l2 is complete.

Example 9.19. Exercise Show that lp is complete for 1 ≤ p < ∞.

Definition 9.20. Let (xn ) be a sequence in a norm a vector space X then the series
P∞
n=1 xn is said to be convergent, if the sequence of partial sums (sn ) converges.

If n=1 xn converges to x then ∞

P∞ P P∞
n=1 xn = x and n=1 xn is sumable to x.

Theorem 9.21. A normed vector space X is complete if and only if every absolutely
sumable series in X is sumable i.e. X is complete if and only if ∞
P
n=1 xn converges
P∞
in X whenever, n=1 ||xn ||∞ converges.

Proof. Let X be a complete. Suppose assume that ∞

P
n=1 ||xn || < ∞.
P∞
To show that n=1 xn converges, let (sn ) be its sequence of partial sum i.e. sn =
Pn
i=1 xi

73 Dr.B. Surendranath Reddy

9 Totally Bounded

n
X m
X
||sn − sm || =|| xi − xi || f or n ≥ m (9.1)
i=1 i=1
Xn
=|| xi || (9.2)
m+1
n
X
≤ ||xi || (9.3)
m+1
P∞ 0
Since ||xn || converges, it’s sequence of partial sum (sn ) converges & hence
n=1
0
Cauchy where sn = ni=1 ||xi ||
P

∴ For  > 0 ∃ p ∈ N such that

0 0
||sn − sm || <  ∀ n ≥ m ≥ p
n
X
||xi || <  ∀ n ≥ m ≥ p (9.4)
i=m+1

Form 9.3 & 9.4

||sn − sm || <  ∀n, m ≥ p
⇒ (sn ) is Cauchy.
Since X is complete (sn ) converges.
⇒ ∞
P
n=1 xn converges.

Conversely, assume that every absolutely summable series in X is complete. Let (xn )
1
be a Cauchy sequence in X then for  = 2k
, ∃ n0 ∈ N such that

1
||xn − xm || < ∀ n, m ≥ n0 (k)
2k
1
⇒ ||xnk+1 − xnk || < k ∀k
2
∞ ∞
X X 1
⇒ ||xnk+1 − xnk || < k
= 1 ∀k
k=1 k=1
2
∞ ∞
X X 1
∴ ||xnk+1 − xnk || < <∞
k=1 k=1
2k

74 Dr.B. Surendranath Reddy

9 Totally Bounded

Let yk = xnk+1 − xnk then ∞

P
k=1 ||yk || < ∞
P∞
⇒ By our assumption k=1 yk also converges. ⇒ It’s sequence of partial sums also
converges.
Pm = i = 1m yi
P

i = 1m (xni+1 − xni )
P
pm =
∵ xnm = xn1 + m
P
i=1 (xni+1 − xni )

∴ xnm = xn1 + pm
as m → is also pm converges & hence (xnm+1 ) is also converges.
∴ (xn ) also converges in X. Since X is complete.

Definition 9.22. Banach Space

A complete normed linear space (X, ||.||) is called a Banach Space.
e.g. R is a Banach space.

Definition 9.23. Contraction Mapping

Let (M, d) ba a metric space. A map f : M → M is said to be contraction if
∃ α, 0 ≤ α < 1 such that d(f (x), f (y)) ≤ α.d(x, y) ∀ x, y ∈ M .
e.g.

x
1. f (x) = 2
x ∈ R then f is a contraction as
|f (x) − f (y)| = | x2 − y2 | = 21 |x − y|
1
Here α = 2
then |f (x) − f (y)| ≤ α|x − y|

2. f (x) = x is not contraction.

Definition 9.24. Fixed Point

A point x0 ∈ M is called a fixed point of a map f : M → M if f (x0 ) = x0 .

Theorem 9.25. Banach Contraction Principle

Let (M, d) be a complete metric space and f : M → M be contraction then f

75 Dr.B. Surendranath Reddy

9 Totally Bounded

has unique fixed point. Moreover given any x0 ∈ M , the sequence of functions
(f n (x0 )) is always converges to a fixed point of f where f 2 (x0 ) = f (f (x0 )), f n (x) =



f f ...f (x0 )

Proof. First we show that for any x0 ∈ M, f n (x0 ) converges. Since M is complete it
is enough to show that (f n (x0 )) is Cauchy. Since f is contraction ∃ some α, 0 ≤ α < 1
such that d (f (x), f (y)) ≤ α.d(x, y) ∀ x, y ∈ M

d f n+1 (x0 ), f n (x0 ) =d f (f n (x0 )), f (f n−1 (x0 ))

≤αd f n (x0 ), f n−1 (x0 )

≤α2 d f n−1 (x0 ), f n−2 (x0 )

≤αn d(f (x0 ), x0 )

=αn .c

For n < m,

d (f n (x0 ), f m (x0 )) ≤d f n (x0 ), f n+1 (x0 ) + d f n+1 (x0 ), f m (x0 )

≤d f n (x0 ), f n+1 (x0 ) + d f n+1 (x0 ), f n+2 (x0 ) + ... + d f m−1 (x0 ), f m (x0 )

≤αn .c + αn+1 .c + ... + αm−1 .c

m−1
!
X
=c αk
k=n
1 − αm−n
=c.αn
1−α
c
d (f n (x0 ), f m (x0 )) ≤ αn (1 − αm−n )
1−α

as n → ∞ αn → 0
∴ d (f n (x0 ), f m (x0 )) → 0 as n → ∞

76 Dr.B. Surendranath Reddy

9 Totally Bounded

∴ (f n (x0 )) is a Cauchy sequence.

Let limx→∞ f n (x0 ) = x. Since f is a contraction, f is continuous.
∴ f (limn→∞ f n (x0 )) = f (x)
⇒ limn→∞ f n+1 (x0 ) = f (x)
⇒ f (x) = x
∴ x is a fixed point of f .
Let x and y are two fixed points of f then f (x) = x and f (y) = y

d(x, y) =d (f (x), f (y))

≤αd(x, y) (∵ f is contraction)

⇒ 1 − αd(x, y) ≤ 0
⇒ d(x, y) = 0
⇒x=y
∴ f has unique fixed point.

Theorem 9.26. (M, d) be a complete metric space and f : M → M be continuous.

If f k is contraction for some k ∈ N then f has a unique fixed point.

Proof. Since f k is a contraction by Banach contraction principle f k has a unique fixed

point say x0 .
i.e. f k (x0 ) = x0
We show that f (x0 ) is also a fixed point of f k .
f k (f (x0 )) = f (f k (x0 )) = f (x0 )
⇒ f (x0 ) is a fixed point of f k . Since f k has a unique fixed point.
⇒ f (x0 ) = x0
∴ f has a unique fixed point of x0 .

77 Dr.B. Surendranath Reddy

10 Completion of a metric/norm space

10 Completion of a metric/norm space

Definition 10.1. A map f : (M, d) → (N, ρ) is called an isometry, if ’f ’ preserves
distance.
i.e. ρ(f (x), f (y)) = d(x, y) ∀x, y, ∈ M .
e.g.Every isometry is continuous.(take δ = )

ˆ is a completion
Definition 10.2. Let (M, d) be a metric space. We say that (M̂ , d)
of (M, d) if
ˆ is complete.
1)(M̂ , d)
ˆ i.e.∃ an isometry f : (M, d) →
2)(M, d) is isometric to some dense subset of (M̂ , d)
ˆ such taht f (M ) is dense in M̂
(M̂ , d)

Example 10.3. Let R is completion of Q as R is complete and i : Q → R is isometry

and i(Q) = Q is dense in R.

Let(M, d) be a metric space then let l∞ (M ) = {f : M → R/f is bounded} is the

set of all bounded real valued function on M .
On l∞ (M ) define k f k∞ = supx∈M | f (x) |, then l∞ (M ) is complete with respect to
k . k∞

Lemma 10.4. Let (M, d) be any metric space, then M is isometric to a subset of
l∞ (M )

Proof. Fix any point a ∈ M then for any x ∈ M ,

definefx (t) = d(x, t) − d(a, t) then

| fx (t) | =| d(x, t) − d(a, t) |

≤ d(x, a) = c

78 Dr.B. Surendranath Reddy

10 Completion of a metric/norm space

| fx (t) |≤ c ∀t ∈ M
⇒ fx : M :→ R is bounded.
fx ∈ l∞ (M )
Now define, ϕ : M → S ⊂ l∞ (M ) by ϕ(x) = fx
where S = {fx : M → R is bounded|x ∈ M }
For any x, y ∈ M , k fx − fy k∞ = supt∈M | fx (t) − fy (t) |

| fx (t) − fy (t) | =| d(x, t) − d(a, t) − d(y, t) + d(a, t) |

=| d(x, t) − d(y, t) |

≤ d(x, y)

For t = x, | fx (t) − fy (t) |= d(x, y)

∴ supt∈M | fx (t) − fy (t) |= d(x, y)
⇒k fx − fy k∞ = d(x, y)
⇒k ϕ(x) − ϕ(y) k∞ = d(x, y)
⇒ ϕ is an isometry from M to S ⊂ l∞ (M )
∴ M is isomorphic to subsets of l∞ (M )

Theorem 10.5. Every metric space has a completion.

Proof. The set S = {fx : M → R is bounded|x ∈ M } is a subset of l∞ (M )

Then S̄ is a closed subset of l∞ (M ) and hence S̄ is complete.
ϕ : M → S̄ defined as
ϕ(x) = fx , is an isometry
then ϕ(M ) = S and ϕ(M ¯ ) = S̄

⇒ ϕ(M ) is dense in S̄.

⇒ S̄ is a completion of M .

Theorem 10.6. If M1 and M2 are completions of M , then M1 and M2 are are

isometric.

79 Dr.B. Surendranath Reddy

10 Completion of a metric/norm space

Proof. Since M1 is a completion of M , ∃ f : M :→ M1 isometry such that F (M ) is

dense in M1
Since M2 is completion of M ∃ g : M :→ M2 isometry such that F (M ) is dense in M2
Let x ∈ M1 since f (M ) is dense in M1 ∃ (xn ) in M such that f (xn ) → x
then we will show that (g(xn )) is Cauchy in M2

d2 (g(xn ), g(xm )) = d(xn , xm )

= d1 (f (xn ), f (xm ))

Since (f (xn )) is Cauchy, (g(xn )) is also Cauchy.

Since M2 is complete (g(xn )) converges in M2 say y
i.e. g(xn ) → y y ∈ M2
Now define ψ : M1 → M2 as ψ(x) = y
where ∃ (xn ) in M such that f (xn ) → x and g(xn ) → y
Now we will show that ψ is an isometry
0
Let x1 , x2 ∈ M1 then ∃ (xn ) and (xn ) in M and y1, y2 ∈ M2 such that
0
ψ(x1 ) = y1 ψ(x1 ) = y2 and f (xn ) → x1 , f (xn ) → x2
0
g(xn ) → y1 , g(xn ) → y2
0
Since f (xn ) →d1 x1 , f (xn ) →d2 x2
0
⇒ d1 (f (xn ), f (xn )) → d(x1 , x2 )

0
∴ d(x1 , x2 ) = lim d1 (f (xn ), f (xn ))
n→∞
0
= lim d(xn , xn )
n→∞
0
= lim d2 (g(xn ), g(xn ))
n→∞

= d2 (y1 , y2 )

= d2 (ψ(x1 ), ψ(x2 ))

0 0
(∵ g(xn ) → y1 , g(xn ) → y2 ⇒ d2 (g(xn ), g(xn )) → d2 (x1 , x2 ))

80 Dr.B. Surendranath Reddy

11 Compactness

⇒ ψ is an isometry.
∴ M1 and M2 are isometric.

11 Compactness
Definition 11.1. A metric space (M, d) is called compact if it is complete and totally
bounded.

Example 11.2. (Heine-Borel theorem)

A subset K of R is compact if and only if it is closed and bounded.

Proof. Suppose K is compact

⇒ K is totally bounded and complete.
Since K is totally bounded in R ⇒ K is bounded in R
Since K closed subset of R which is complete, K is closed.
Conversely, suppose K is closed and bounded.
Since K is closed subset of complete space, K is complete.
Since in R, every bounded set is totally bounded, K is totally bounded.

Example 11.3. In a discrete space, a subset is compact if and only if it is finite.

Proof. Let M is discrete and A ⊂ M .

Suppose A is compact, then A is totally bounded.
⇒ A is finite, as only totally bounded subset of discrete space are finite sets.
Conversely, supposeA is finite.
Since M is discrete A is totally bounded.
Since every subset in M is closed.
⇒ A is closed.

81 Dr.B. Surendranath Reddy

11 Compactness

Since M is complete(∵ every discrete space is complete.) and A ⊂ M is closed

⇒ A is complete.

Example 11.4. [0, 1] is compact where as (0, 1) is not compact as (0, 1) is not closed.

Theorem 11.5. A metric space (M, d) is compact if and only if every sequence in
M has a subsequence that converges to a point in M .

Proof. Suppose (M, d) is compact.

⇒ M is totally bounded and complete.
Let (xn ) be a sequence in M .
Since M is totally bounded, every sequence in M has a Cauchy subsequence.
∴ (xn ) has a Cauchy subsequence (xnk ).
Since M is complete (xnk ) converges in M .
⇒ Every sequence in M has a convergent subsequence in M .
Conversely, suppose every sequence in M has a convergent subsequence in M .
Since convergent ⇒ Cauchy, every sequence in M has a Cauchy subsequence.
⇒ M is totally bounded.
To show M is complete, let (xn ) be a Cauchy sequence in M by assumption, (xn ) has
a subsequence (Xnk ) which converges in M .
Since (xn ) is Cauchy, it also converges in M .
⇒ M is complete and hence M is compact.

Theorem 11.6. If f : (M, d) → (N, ρ) is continuous and K is compact in M then

f (K) is compact in N .
(i.e. continuous image of a compact subset is compact.)

Proof. To show that f (K) is compact.

Let (yn ) be a sequence in f (K).
⇒ yn = f (xn ) where xn ∈ K .

82 Dr.B. Surendranath Reddy

11 Compactness

∴ (xn ) is a sequence in K and K is compact.

⇒ (xn ) has a subsequence (xnk ) such that xnk → x for some x ∈ K
Now,since f is continuous, f (xnk ) → f (x) where f (x) ∈ f (K)
∴ yn has a subsequence (ynk ) which converges in f (K).
⇒ f (K) is compact.

Theorem 11.7. Let (M, d) be compact. If f : M → R is continuous then f is

bounded. Moveover 0 f 0 attains it’s maximum and minimum values.

Proof. Since f : M → R is continuous and M is compact.

⇒ f (M ) is compact in R.
⇒ f (M ) is bounded.
⇒ f is bounded in M .
Let N = supx∈M f (x) and L = inf x∈M f (x)
We will show that N, L ∈ f (M )
Suppose there does not exists any x ∈ M such that f (x) = N .
⇒ f (x) 6= N ∀ x ∈ M then N − f (x) > 0 ∀ x ∈ M
1
then g(x) = N −f (x)
is continuous on M .
⇒ g is bounded.
⇒ ∃ k > 0 such that g(x) ≤ k ∀x ∈ M
1
⇒ N −f (x)
≤ k, ∀x ∈ M
1
⇒ k
≤ N − f (x), ∀x ∈ M
⇒ f (x) ≤ N − k1 , ∀x ∈ M
1
⇒ N− k
is an upper bound.
1
∵ N is least, we get N ≤ N − k
which is contradiction.
∴ ∃x ∈ M such that f (x) = N .

83 Dr.B. Surendranath Reddy

11 Compactness

Similarly, suppose f (x) 6= L ∀x ∈ M

then f (x) − L > 0 ∀x ∈ M
1
then h(x) = f (x)−L
is continuous on M .
⇒ h(x) is bounded.
⇒ ∃ p > 0 such that h(x) ≤ p ∀x ∈ M
1
⇒ f (x)−L
≤ p ∀x ∈ M
1
⇒ p
≤ f (x) − L ∀x ∈ M
1
⇒ L+ p
≤ f (x) ∀x ∈ M
1
⇒ L+ p
is a lower bound for f (x).
1
⇒ L+ p
≤ L which is a contradiction.
∴ ∃ x ∈ M such that f (x) = L
∴ f attains its minimum and maximum.

Theorem 11.8. M is compact if and only if whenever G is a collection of open sets

S
in M with M ⊂ {G/G ∈ G} then there are finitely sets G1 , G2 , ...Gn ∈ G such that
M ⊂ ni=1 Gi
S

(i.e. M is compact iff every open cover of M has a finite subcover.)

Proof. Suppose M is compact then M is totally bounded and complete.

S
Suppose G is a collection of open sets such that M ⊂ {G/G ∈ G}.
We will prove by contradiction. i.e. assume that M can not be covered by finitely
many G’s.
Since M is totally bounded, M can be covered by finitely many sets of diameter< 1.
Since M can not be covered by finitely many G’sthere is at least one set say A1 of
diam< 1 such that A1 can not be covered by finitely many G’s.
(∵ suppose not then Aj ⊂ m
S Sn Sn Sm
i=1 Gji ) ⇒ M ⊂ j=1 Aj ⊂ j=1 i=1 Gji )

In particular, A1 6= ∅ and is infinitely.

Since A1 ⊂ M ⇒ A1 is totally bounded.

84 Dr.B. Surendranath Reddy

11 Compactness

1
⇒ A1 can be covered by finitely many sets of diameter< 2
1
there is at least one set, say A2 with diam(A2 ) < 2
such that A2 can be covered by
finitely many G’s. Continuing in the same way,we get,
1
A1 ⊃ A2 ⊃ A3 ⊃ ... with diam(An ) < n

such that no Ai is covered by finitely many G0 s.

1
Ā1 ⊃ Ā2 ⊃ Ā3 ⊃ ... with diam(Ān ) < n

⇒ diam(Ān ) → 0 as n → ∞ Since M is complete,

T∞
n=1 Ān 6= ∅

Let x ∈ ∞
T
n=1 Ān

⇒ x ∈ Ān ∀n
S
Since M ⊂ {G/G ∈ G}
⇒ ∃G ∈ G such that x ∈ G
Since G is open, ∃  > 0 such that B (x) ⊂ G.
1
Since  > 0 ∃ n such that n
<  (By archimedean property.
1
then diam(An ) < n
< .
⇒ Ān ⊂ B (x) ⊂ G
⇒ An ⊂ G which is a contradiction.
∴ M can be covered by finitely many sets G from G.
⇒ Every open cover has a finite subcover.
Conversely, to show that M is compact.We will show that M is totally bounded and
complete.
Since for any x ∈ M x ∈ B (x) for any  > 0
S
we get M ⊂ {B (x)/x ∈ M }
⇒ {B (x)/x ∈ M } is an open cover for M
By assumption, it has a finite sub cover I.E ∃ x1 , x2 , ...xn ∈ M such that M ⊂
Sn
i=1 B (xi )

⇒ M is totally bounded.

85 Dr.B. Surendranath Reddy

11 Compactness

To show that M is complete, let F1 ⊃ F2 ⊃ F3 ⊃ ... be non empty closed subsets of

M such that,
diam(Fn ) → 0 as n → ∞
Now Fn = ni=1 Fi .
T

Since Fn 6= ∅ for every n

⇒ ni=1 Fi 6= ∅ for every n
T

Suppose ⇒ ∞
T
i=1 Fi = ∅
S∞ c
⇒ i=1 Fi = M
⇒ {Fic }∞
i=1 is an open cover for M .

⇒ ∃ n such that, M ⊂ ni=1 Fic

S

⇒ M c ⊃ { ni=1 Fic }c
S

⇒ ∅ ⊃ ni=1 Fi
T

⇒ ni=1 Fi = ∅ which is a contradiction.

T

⇒ ∞
T
i=1 Fi 6= ∅.

⇒ M is complete.

Theorem 11.9. M is compact if and only if F is any collection of closed sets in

M such that ni=1 Fi 6= ∅ for all choices of finitely many sets F1 , F2 , ...Fn ∈ F then
T
T
{F : F ∈ F} =
6 ∅

Proof. Suppose M is compact. Let F is a collection of closed sets in M such that

Tn
i=1 Fi 6= ∅ ∀ n
T
Suppose {F : F ∈ F} = ∅.
⇒ {F c } = M is an open cover for M .
S

Since M is compact, it has a finite subcover.

⇒ ∃ n such that M ⊂ ni=1 {F c }
S

⇒ ∅ ⊃ ni=1 Fi
T

⇒ ni=1 Fi = ∅, which is contradiction.

T

86 Dr.B. Surendranath Reddy

12 Uniform Continuity

T
∴ {F/ F ∈ F} =
6 ∅
Now to prove M is compact. Let (xn ) be a sequence in M .
Let An = {xk / k ≥ n}.
⇒ A1 ⊃ A2 ⊃ A3 ⊃ ... and each An 6= ∅ ∀n
∴ we have a decreasing sequence of non-empty closed sets,
i.e. Ā1 ⊃ Ā2 ⊃ Ā3 ⊃ ... such that Ān 6= ∅ ∀n
⇒ ni=1 Āi 6= ∅ ∀n
T

By assumption, Ān 6= ∅. Let x ∈ ∞

T T∞
i=1 Āi . Since i=1 Āi is closed. There is a

subsequence (xnk ) such that xnk → x

(∵ x ∈ Ā iff (xn ) ⊂ A such that xn → x )
⇒ M is compact.

Corollary: M is compact iff every countable open cover admits a finite subcover.

12 Uniform Continuity
Definition 12.1. A map f : (M, d) → (N, ρ) is said to be uniformly continuous
if for gien  > 0, ∃ δ > 0 such that whenever x, y ∈ M with d(x, y) < δ then
ρ(f (x), f (y)) < .

Example 12.2. Every Lipschitz function is uniformly continuous.

Proof. Suppose f is Lipschitz. Let  > 0 since f is Lipschitz, ∃ k such that

ρ(f (x), f (y)) ≤ kd(x, y)

Take δ = k
then for any x, y ∈ M with d(x, y) < δ
ρ(f (x), f (y)) ≤ kd(x, y) < kδ = 
⇒ f is an uniformly continous.

Remark 12.3. Uniformly continuous function need not be Lipschitz.

√
Example: f : [0, ∞) → R such that f (x) = x then f is uniformly continuous.

87 Dr.B. Surendranath Reddy

12 Uniform Continuity

But we will show that f is not Lipschitz.

Suppose f is Lipschitz then ∃ k ∈ R such that
| f (x) − f (y) |≤ k. | x − y | for all x, y ∈ [0, ∞)
√ √
| x − y |≤ k. | x − y | for all x, y ∈ [0, ∞)
In particular, for y = 0,
√
| x |≤ k | x | for all x ∈ [0, ∞)
√ √
x x
⇒ x
≤ k for all x ∈ [0, ∞) as x → 0, x
→∞
⇒ k → ∞ which is a contradiction.

Theorem 12.4. Every uniformly continuous function is continuous.

Proof. f : M → N is uniformly continuous.

∴ for  > 0 ∃ δ > 0 such that
ρ(f (x), f (y)) <  whenever d(x, y) < δ
⇒ f is continuous at x ∈ M .
⇒ f is continuous on M .

Remark 12.5. The converse of the above theorem is false i.e. every continuous
function need not be uniformly continuous.
For example, f (x) = x1 , x ∈ (0, ∞) then f is continuous.
But f is not uniformly continuous as n → ∞,
| x − y |→ 0 ⇒| x − y |< δ
1 1
But | f (x) − f (y) | =| f ( ) − f ( )|
n n+1
=| n − n − 1 |

=1

For  = 21 , | f (x) − f (y) |> .

Theorem 12.6. If M is a compact metric space then every continuous map f : M →

N is uniformly continuous.

88 Dr.B. Surendranath Reddy

12 Uniform Continuity

Proof. Let  > 0.Since f is continuous for each x ∈ M ∃ δx > 0 such that,

ρ(f (x), f (y)) < 2
whenever d(x, y) < δx.
∴ {B δx (x)/x ∈ M } is an open cover for M .
2

Since M is compact, ∃ x1 , x2 , , , , xn ∈ M such that

M ⊂ ni=1 Bri (xi ) where ri = δxi
S
2
.
Take δ = min{r1 , r2 , ...rn }.
Let x, y ∈ M such that d(x, y) < δ.Since x ∈ M ∃ j such that x ∈ Brj (xj ). Also

d(y, xj ) ≤ d(y, x) + d(x, xj )

< rj + rj

< 2rj = δxj

⇒ y ∈ Bδxj (xj )

⇒ ρ(f (y), f (xj )) < 2
δxj
Since x ∈ Brj (xj ) ⇒ d(x, xj ) < rj = 2
< δxj

ρ(f (x), f (y)) ≤ ρ(f (x), f (xj )) + ρ(f (y), f (xj ))

 
< + = .
2 2

⇒ f is uniformly continuous.

Remark 12.7. Any continuous function f : [a, b] → R is uniformly continuous.

e.g. f (x) = xn is uniformly continuous on [1, 2].

Theorem 12.8. If f : M → N is uniformly continuous, then f maps Cauchy se-

quence in M to Cauchy sequence in N i.e. if (xn ) is Cauchy in M then f (xn ) is
Cauchy in N .

Proof. Let xn be a Cauchy sequence in M and  > 0 since f is uniformly continuous

∃ δ > 0 such that

89 Dr.B. Surendranath Reddy

12 Uniform Continuity

ρ(f (x), f (y)) <  whenever d(x, y) < δ

Since xn is Cauchy,∃ N > 0 such that,
d(xn , xm ) < δ, ∀n, m ≤ N
⇒ ρ(f (xn ), f (xm )) < , ∀n, m ≤ N
⇒ (f (xn )) is Cauchy in N .

Remark 12.9. If f is not uniformly continuous, then the above result is not true.
e.g.f : (0, ∞) → R such that f (x) = x1 .
( n1 ) is Cauchy in(0, ∞).
But (f ( n1 )) = n is not Cauchy in R.

Theorem 12.10. If f : M → N is uniformly continuous, then f maps totally bounded

sets to totally bounded sets.

Proof. Let A ⊂ M is totally bounded and  > 0.Since f is uniformly continuous

∃ δ > 0 such that, ρ(f (x), f (y)) <  whenever d(x, y) < δ
⇒ f (Bδd (x)) ⊂ B (f (x)), ∀x ∈ M
Since A is totally bounded, and δ > 0, ∃ x1 , ...xn ∈ M such that,
n
[
A⊂ Bδ (xi )
i=1
n
[
⇒ f (A) ⊂ f ( Bδ (xi ))
i=1
n
[
⊂ f (Bδ (xi )))
i=1
[n
⊂ B (f (xi ))
i=1

⇒ f (A) is totally bounded.

90 Dr.B. Surendranath Reddy

13 Equivalent Metrics

13 Equivalent Metrics
Definition 13.1. Let d and ρ are two metric on M . Let i : (M, d) → (M, ρ) and
i−1 : (M, ρ) → (M, d) be identity maps then we say that d and ρ are equivalent if
both i and i−1 are continuous.(i.e. if i is a homeomorphism)

Remark 13.2. 0 i0 is continuous if given  > 0. ∃ δ > 0 such that ρ(i(x), i(y)) < 
whenever d(x, y) < δ. i.e.ρ(x, y) <  whenever d(x, y) < δ.
Similarly, since i−1 is continuous d(x, y) <  whenever ρ(x, y) < δ.

Definition 13.3. d and ρ are called uniformly equivalent, if i and i−1 are uniformly
continuous.

Definition 13.4. d and ρ are called strongly equivalent, if i and i−1 are Lipschitz
maps.

Remark 13.5. Strongly equivalent ⇒ uniformly equivalent ⇒ equivalent.

But the converse is not true.
p
e.g. d(x, y) =| x − y | and ρ(x, y) = | x − y | M = [0, 1]

Example 13.6. Show that d and ρ are uniformly equivalent but not strongly equiv-
alent.

Proposition 13.7. Suppose (M, d) is compact and ρ is another metric space on M

then d and ρ are equivalent iff d and ρ are uniformly equivalent.

Proof. Let d and ρ be equivalent.

⇒ i and i−1 are continuous.Since (M, d) is compact and i : (M, d) → (M, ρ) is
continuous.
⇒ i is uniformly continuous. Also i−1 : (M, ρ) → (M, d) is continuous and onto.
⇒ i−1 (M, ρ) = (M, d)

91 Dr.B. Surendranath Reddy

13 Equivalent Metrics

Since i is continuous (M, ρ) = i(i−1 (M, ρ)) is compact.

⇒ i−1 is uniformly continuous.
⇒ d and ρ are uniformly equivalent.
Conversely, if d and ρ are uniformly equivalent then d and ρ are equivalent.

Theorem 13.8. Let (V, k . k), (W, k . k) be two normed vector space and T : V → W
be a linear map then following are equivalent.

i). T is Lipschitz.

ii). T is uniformly continuous.

iii). T is continuous.(Every where)

iv). T is continuous at 0 ∈ V

v). There is a constant c < ∞ such that k| T x |k≤ c k x k ∀x ∈ V .

Proof. i) ⇒ ii) ⇒ iii) ⇒ iv) is trivially true.

iv) ⇒ v) : Since T is continuous at 0 ∈ V , For  = 1 ∃ δ > 0 such that,
0 δx
k| T (x) |k< 1 whenever k x k< δ. Let x 6= 0 then take x = 2.kxk
then
0 δx δ δ
k x k=k 2kxk
k= 2kxk
k x k= 2
<δ
0
⇒k| T (x ) |k< 1
δx
⇒k| (T 2kxk ) |k< 1
δ
⇒ 2kxk
k| T (x) |k< 1
2
⇒k| T (x) |k< δ
kxk
⇒ k| T (x) |k≤ c k x k where c < ∞
v) ⇒ i) : suppose ∃ c such that,
k| T (x) |k≤ c k x k ∀x ∈ V then
k| T x − T y |k=k| T (x − y) |k≤ c k x − y k

92 Dr.B. Surendranath Reddy

13 Equivalent Metrics

(∵ if x, y ∈ V ⇒ x − y ∈ V ) ⇒
T is Lipschitz.

Theorem 13.9. Any two norms on a finite dimensional vector space are equivalent.

Proof. Let V be a finite dimensional vector space of dimension 0 n0 .Let {x1 , x2 , ..., xn }
be a basis of V then for any x ∈ V ∃ α1 , α2 , ..., αn ∈ R such that
x = ni=1 αi .xi .
P

Define a map f : V → Rn by f (x) = (α1 , α2 , ..., αn ).

We will define a norm ||.|| on V as ||x|| = || ni=1 αi xi || = ni=1 |αi |
P P

then f : (V, ||.||) → (Rn , ||. k1 ) is isometry, as

||x|| = ni=1 |αi | = ||(α1 , α2 , ..., αn ) k1 =k f (x)||1
P

∴ f is an isometry.
Let |||.||| be any other norm on V , then

n
X
|||x||| = ||| αi .xi |||
i=1
n
X
≤ |αi |.|||xi |||
i=1
n
X
≤ max |||xi ||| |αi |
1≤i≤n
i=1
n
X
≤ c1 . |αi | where c1 = max |||xi |||
1≤i≤n
i=1

⇒ |||x||| ≤ c1 .|||x|||......................(1)

Let S = {x ∈ V / k x k= 1}. Since f is isometry and

f (S) = {(α1 , ...αn )/ | αi |= 1} is compact in Rn .
P

⇒ S is compact in V .
|||.||| : V → R is continuous as | |||x||| − |||y|||| ≤ |||x − y||| ≤ c1 ||x − y|| ⇒ |||.||| is

93 Dr.B. Surendranath Reddy

13 Equivalent Metrics

continuous on S.
Since S is compact,|||.||| attains it’s minimum say c2 then
c2 > 0 and |||x||| ≥ c2 x ∈ S.
x
For any x 6= 0, ||x||
∈S
x
⇒ ||| ||x|| ||| ≥ c2 ∀o 6= x ∈ V ⇒ |||x||| ≥ c2 |||x|||...............(2)
From (1) and (2),i and i−1 are continuous.
⇒ ||.|| and |||.||| are equivalent.
Hence any two norms on a finite dimensional vector space are equivalent.

Corollary 13.10. Let V and W be normed vector spaces with V finite dimensional
then every linear map T : V → W is continuous.

Proof. Since V is finite dimensional. Let {x1 , x2 , ...xn } be a basis of V then for any
x ∈ V ,define ||x|| = ni=1 |αi | where x = ni=1 αi .xi .
P P

Let |||.||| be a norm on W . For any x ∈ V ,

Xn
|||T x||| =|||T ( αi .xi )|||
i=1
Xn
= ||| αi T (xi )||| (∵ T is linear)
i=1
n
X
≤ |αi ||||T (xi )||| where c = max |||T (xi )|||
1≤i≤n
i=1
Xn
=c |αi |
i=1

⇒ |||T x||| ≤ c.||x||

⇒ T is continuous

Theorem 13.11. Let D be dense in M and N be complete. Let f : D → N be

94 Dr.B. Surendranath Reddy

13 Equivalent Metrics

uniformly continuous then f extends uniquely to a uniformly continuous map F :

M → N moreover if 0 f 0 is an isometry then so is the extension F .

Proof. Let x ∈ M . since D is dense in M ∃ (xn ) ⊂ D such that xn → x

⇒ (xn ) is Cauchy in D.
since f is uniformly continuous, (f (xn )) is Cauchy in N .
since N is complete,f (xn )) converges say to y in N . i.e. f (xn ) → y
Hence define F : M → N by
F (x) = y = limn→∞ f (xn ) where xn → x
To show F is well defined, let xn → x and yn → x then the sequence x1 , y1 , x2 , y2 ...
also converges. This sequence x1 y1 , x2 y2 , ... is Cauchy D.
⇒ f (x1 ), f (y1 ), f (x2 ), f (y2 ), ... is Cauchy in N .
since N is complete, the above sequence converges to z. In particular, (F (xn )) con-
verges to z and (F (yn )) also converges to z.
⇒ F is well defined.
Now we will prove F is uniformly continuous.
 0
since f is uniformly continuous, for 3
> 0 ∃ x ∈ D such that,
0 0  0 0 0 0
ρ(f (x ), f (y )) < 3
whenever d(x , y ) < δ x , y <∈ D
0
Given any x ∈ M ∃x ∈ D such that
0 δ 0 
d(x, x ) < 3
and ρ(F (x), f (x ) < 3
0 0
(∵ D is dense,D ∩ B δ (x) 6= ∅, Let x ∈ D ∩ B δ (x) ⇒ d(x, x ) < 3δ ).
3 3
0
⇒ ∃ (xn ) ⊂ D such that d(xn , x ) < δ
δ
Let x, y ∈ M with d(x, y) < 3
0 0 0 δ 0 
⇒ ∃ x , y ∈ D such that d(x, x ) < 3
and ρ(F (x), F (x )) < 3

95 Dr.B. Surendranath Reddy

13 Equivalent Metrics

0 δ 0 
and d(y, y ) < 3
and ρ(F (y), F (y )) < 3

0 0 0 0
d(x , y ) ≤ d(x , x) + d(y, y ) + d(x, y)
δ δ δ
≤ + + =δ
3 3 3
0 0
d(x , y ) ≤ δ

0 0 
⇒ ρ(F (x ), F (y )) ≤ 3

0 0 0 0
∴ ρ(F (x), F (y)) ≤ ρ(F (x), F (x ) + ρ(F (x ), F (y )) + ρ(F (y), F (y ))
  
< + + =
3 3 3
δ
∴ ρ(F (x), F (y)) <  whenever x, y ∈ M such that d(x, y) < 3

⇒ F is uniformly continuous.
Now it remains to prove F is an extension of f to prove this we will prove
F/D = f i.e. F (x) = f (x) ∀x ∈ D
Let x ∈ D since xn = x, x converges to x i.e.xn → x.
⇒ f (xn ) → f (x)
⇒ limn→∞ f (xn ) = f (x)
⇒ F (x) = f (x) ∀x ∈ D
⇒ F/D = f
To show uniqueness, Let F, G : M → N which are uniformly continuous and F/D =
f = G/d.
Let x ∈ M then ∃ (xn ) ⊂ D such that xn → x. since F and G are continuous,
F (xn ) → F (x) and G(xn ) → G(x)
since F/D = G/D and (xn ) ⊂ D.
⇒ F (xn ) = G(xn )
⇒ F (xn ) → F (x) and F (xn ) → G(x)
⇒ F (x) = G(x).

96 Dr.B. Surendranath Reddy

13 Equivalent Metrics

∴ F = G on M .
suppose f is an isometry. Let x, y ∈ M
⇒ ∃(xn ), (yn ) ⊂ D such that xn → x andyn → y
⇒ d(xn , yn ) → d(x, y)
⇒ d(x, y) = limn→∞ d(xn , yn ) = limn→∞ ρ(f (xn ), f (yn ))
0 0 0 0 0 0
(∵ d(x , y ) = ρ(f (x ), f ( )) ∀x , y ∈ D)
since f (xn ) → F (x) and f (yn ) → F (y)
ρ(f (xn ), f (yn )) → ρ(F (x), F (y)).
∴ d(x, y) = ρ(F (x), F (y))
⇒ F is isometry.

Theorem 13.12 (Baire Category theorem ). If (Gn ) is a sequence of dense open

subsets in R then ∞
T T∞
n=1 6= ∅. Infact n=1 Gn is dense in R.

Proof. Let x0 ∈ R and I0 be any open interval containing x0 . Since G1 is dense in

T T
R ⇒ I0 G1 6= ∅. Let x1 ∈ I0 G1
Since G1 is open in R ⇒ I0 ∩ G1 is open.
⇒ ∃ open interval I1 such that x1 ⊂ I1 ⊂ I0 ∩ G1
By shrinking I1 (if necessary),we may assume that
diam(I1 ) < 1 and I¯1 ⊂ I0 ∩ G1 ..
Now,G2 is dense in R ⇒ I1 ∩ G2 6= ∅.
Let x2 ∈ I1 ∩ G2 . Since G2 is open.
⇒ I1 ∩ G2 is open.
⇒ ∃ an open interval I2 such that x2 ∈ I2 ⊂ I1 ∩ G2
again by shrinking I2 , we may assume that diam(I2 ) < 1
2
and I¯2 ⊂ I1 ∩ G2
∴ I¯1 ⊃ I¯2 and I¯2 ⊂ I1 ∩ G2 ⊂ I0 ∩ G1 ∩ G2 . By repeating this process, we get a
decreasing sequence of closed and bounded intervals which are non-empty such that
I¯n ⊂ I0 ∩ G1 ∩ G2 ∩ ... ∩ Gn

97 Dr.B. Surendranath Reddy

14 Sequence of functions

Now, by nested interval theorem, we have,∩∞ ¯

i=1 In 6= ∅.

⇒ ∩∞ ¯ ∞
i=1 In ⊂ I0 ∩ (∩i=1 Gn ).

Since ∞ ¯ ∞
T
i=1 In 6= ∅ ⇒ I0 ∩ (∩i=1 Gn ) 6= ∅.

⇒ ∩∞
i=1 Gn intersects every open interval.

⇒ ∩∞
i=1 Gn is dense.

Corollary 13.13. Q cannot be written as the countable intersection of open subsets

of R

14 Sequence of functions
Definition 14.1. Let X be any set and (Y, ρ) be a metric space. Let f and (fn ) be
functions from X to Y . We say that the sequence (fn ) converges pointwise to f on
X, if for each x ∈ X, (fn (x)) converges to f (x) in Y , i.e. for given  > 0 and for
x ∈ X, there exists N such that ρ(fn (x), f (x)) <  ∀n ≥ N.

x
Example 14.2. Let fn (x) = n
x ∈ R.
x
Since for each x ∈ R, limn→∞ fn (x) = limn→∞ n
= 0, fn converges pointwise to f = 0.

Example 14.3. Let fn (x) = nx x ∈ R.

Then fn does not converge as fn (x) → ∞ as n → ∞.

n
Example 14.4. Let fn (x) = x+n
x ∈ R.
Then fn converges pointwise to f = 1.

Example 14.5. Let fn (x) = n2 xn x ∈ [0, 1).

Since fn (x) = n2 xn = n2 enlogx → 0 as n → ∞, fn converges pointwise to f = 0. ( as
xn approaches to zero faster than n2 approaches to ∞.)

xn+1
Example 14.6. Let hn (x) = n+1
x ∈ [0, 1]. If possible find h(x) such that hn (x) →
h(x). Is h0n (x) → h0 (x)?

98 Dr.B. Surendranath Reddy

14 Sequence of functions

n+1
Solution: Since |hn (x)| = | xn+1 | ≤ 1
n+1
→ 0, hn converges pointwise to h = 0.
0 ≤ x < 1, h0n (x) → 0 and for x = 1, h0n (x) → 1.
Here h0n (x) = xn . Then for 
0 0 ≤ x < 1

0
Therefore, hn (x) → g(x) =
1 x = 1.


Since h0 (1) = 0 and g(1) = 1, h0n does not converge to h0 .

Example 14.7. Let fn (x) = (1 + nx )n x ∈ R. Find f (x) such that fn (x) → f (x).
Is fn0 (x) → f 0 (x) and fn (x) → f (x)? Justify.
R R

Solution: limn→∞ fn (x) = limn→∞ (1 + nx )n = ex .

Therefore, fn converges pointwise to f = ex .
Also fn0 (x) = (1 + nx )n−1 .
limn→∞ fn0 (x) = limn→∞ (1 + nx )n−1 = ex = f 0 (x).
R1 R1 h
n+1 i R1
Also 0 fn (x)dx = 0 (1 + nx )n dx = n+1
n
1 + n1 − 1 → e − 1 = 0 f (x)dx

Definition 14.8. Let X be any set and (Y, ρ) be a metric space. Let f and (fn ) be
functions from X to Y . We say that the sequence (fn ) converges uniformly to f on
X, if for given  > 0, ∃N such that ρ(fn (x), f (x)) < , ∀x ∈ X, ∀n ≥ N . It is
denoted by fn ⇒ f .

Remark 14.9. Uniform convergence implies pointwise convergence as N depends

only on , not on x.
nx
Converse is not true, for example, define fn (x) = 1+n2 x2
.
Then fn → f = 0 pointwise.
Suppose fn converges uniformly, then for  = 12 , there exists N such that |fn (x) −
1
f (x)| < 2
∀x ∈ X, ∀n ≥ N .
1
In particular, for n = N and x = N
, we have |fN ( N1 ) − f ( N1 )| = 1
2
< 12 , which is a
contradiction.
Therefore fn does not converge uniformly to f .

99 Dr.B. Surendranath Reddy

14 Sequence of functions

1
Example 14.10. Show that fn (x) = n+x
is uniformly convergent on [0, k] for any
k > 0.
Solution: Clearly fn → f = 0 pointwise.
Let  > 0. We have to find N such that |fn (x) − f (x)| < , for all x ∈ [0, k] and for
all n ≥ N .
1 1
|fn (x) − f (x)| = | n+x − 0| = n+x

If x ∈ [0, k], then n ≤ n + x

1
⇒ |fn (x) − f (x)| < n

By Archimedean principle, ∃N such that N > 1 .

1 1
Then for n ≥ N , |fn (x) − f (x)| < n+k
< N

Therefore, fn converges uniformly to f = 0.

Remark 14.11. 1. fn ⇒ f if and only if for given  > 0, ∃N such that ρ(fn (x), f (x)) <
, ∀x ∈ X, ∀n ≥ N
if and only if supx∈X ρ(fn (x), f (x)) < , ∀n ≥ N
if and only if Mn → 0 as n → ∞ where Mn = supx∈X ρ(fn (x), f (x))

2. To find Mn , let g(x) = ρ(fn (x), f (x)). By equating g 0 (x) to zero and using first
derivative test, we find the maximum value of g(x).
x
Example 14.12. Discuss the convergence of fn (x) = 1+nx2
for x ∈ R.
x
Solution: limn→∞ fn (x) = limn→∞ 1+nx2
=0
Therefore, fn → f = 0 pointwise.
x
Let g(x) = fn (x) − f (x) = 1+nx2
1−nx2
Then g 0 (x) = (1+nx2 )2

g 0 (x) = 0 implies x = ± √1n .

Since g 00 ( √1n ) < 0, g has maximum value at x = √1 .
n

Therefore, Mn = g( √1n ) = 1
√
2 n
→ 0 as n → ∞.
Hence fn converges uniformly to f = 0.

100 Dr.B. Surendranath Reddy

14 Sequence of functions

Exercise: Discuss pointwise and uniform convergence of

x
1. fn (x) = 1 + n
on R.

2. fn (x) = xn on (−1, 1]

3. fn (x) = n2 x(1 − x2 )n on [0, 1]

nx
4. fn (x) = 1+nx
on [0, ∞)

nx
5. fn (x) = 1+n2 x2
on [0, ∞)

6. fn (x) = xe−nx on [0, ∞)

7. fn (x) = nxe−nx on [0, ∞)

Theorem 14.13. Let (X, d) and (Y, ρ) be metric spaces. Let f and (fn ) be functions
from X to Y . If (fn ) converges uniformly to f on X and each fn is continuous, then
f is continuous.

Proof. Let  > 0 and x ∈ X. Since (fn ) converges uniformly to f , ∃N , such that
ρ(fn (x), f (x)) < 3 , ∀x ∈ X, ∀n ≥ N .
In particular, ρ(fN (x), f (x)) < 3 , ∀x ∈ X.

Since fN is continuous, ∃δ > 0 such that ρ(fN (x), fN (x0 )) < 3
whenever d(x, x0 ) < δ.
Let x0 ∈ X such that d(x, x0 ) < δ.

ρ(f (x), f (x0 )) ≤ ρ(f (x), fN (x)) + ρ(fN (x), fN (x0 )) + ρ(fN (x0 ), f (x0 ))
  
< + +
3 3 3
= .

Therefore, f is continuous on X.

101 Dr.B. Surendranath Reddy

14 Sequence of functions

Example 14.14. Let  fn (x) = xn on [0, 1].

0 0 ≤ x < 1

Then fn (x) → f (x) =
1 x = 1.


Since each fn is continuous and f is not continuous at x = 1, by above theorem, fn

does not converge uniformly to f .

n x 2
Example 14.15. Let fn (x) = 1+n 2 x2 on [0, 1].

0 x = 0

Then fn (x) → f (x) =
 1 x 6= 0.

x
Since each fn is continuous and f is not continuous at x = 0, by above theorem, fn
does not converge uniformly to f .

Theorem 14.16. Suppose fn : [a, b] → R is continuous for each n and that fn ⇒ f

Rb Rb
on [a, b]. Then a fn (x)dx → a f (x)dx.

Proof. Since each fn is continuous and fn ⇒ f , f is continuous on [a, b].

Rb
So a f is well defined and
Z b Z b Z b
| fn (x)dx − f (x)dx| = | [fn (x) − f (x)]dx|
a a a
Z b
≤ |fn (x) − f (x)|dx
a
Z b
≤ supx∈[a,b] |fn (x) − f (x)|dx
a

= (b − a)supx∈[a,b] |fn (x) − f (x)|

Since fn ⇒ f , supx∈[a,b] |fn (x) − f (x)| → 0 as n → ∞.

Rb Rb
Therefore, a fn (x)dx → a f (x)dx.

Theorem 14.17. Suppose that fn is a sequence of real valued functions, each having
a continuous derivative on [a, b] and suppose that the sequence of derivatives (fn0 )

102 Dr.B. Surendranath Reddy

14 Sequence of functions

converges uniformly to a function g on [a, b]. If (fn (x0 )) converges at any point x0 ∈
[a, b], then fn converges uniformly to a differentiable function f on [a, b]. Moreover
f 0 = g. That is (fn0 ) converges uniformly to f 0 .

Proof. First we show that fn converges pointwise to some function f on [a,b].

Let c = limn→∞ fn (x0 ).
By fundamental theorem of calculus, for any x ∈ [a, b],
Rx Rx
fn (x) = fn (x0 ) + x0 fn0 (t)dt (∵ x0 fn0 (t)dt = fn (x) − fn (x0 ))
Rx
⇒ limn→∞ fn (x) = limn→∞ fn (x0 ) + limn→∞ x0 fn0 (t)dt
Rx Rx
Since each fn0 is continuous and fn0 ⇒ g, we get limn→∞ x0 fn0 (t)dt = x0 g(t)dt.
Rx
⇒ limn→∞ fn (x) = c + x0 g(t)dt
Rx
Define f (x) = c + x0 g(t)dt. Then fn → f pointwise.
Rx Rx Ra Ra
Since x0 g(t)dt = a g(t)dt + x0 g(t)dt and f (a) = c + x0 g(t)dt,
Rx
we get f (x) = f (a) + a g(t)dt
Since the right hand side of the above equation is differentiable, f is differentiable
and f 0 (x) = g(x).
Therefore, f = g and (fn0 ) converges uniformly to f 0 .
Rx
Also fn (x) = fn (a) + a fn0 (t)dt. Then
Z x
|fn (x) − f (x)| = |fn (a) − f (a) + [fn0 (t) − g(t)]dt|
Za x
≤ |fn (a) − f (a)| + |fn0 (t) − g(t)|dt
a

≤ |fn (a) − f (a)| + (b − a)supx∈[a,b] |fn0 (x) − g(x)|

 
< + (b − a)
2 2(b − a)
= .

103 Dr.B. Surendranath Reddy

14 Sequence of functions

The space of bounded functions: Given a set X, let B(X) be the set of all
bounded real valued functions on X. i.e. B(X) = {f : X → R | f is bounded}. Then
B(X) is a vector space and k.k∞ on B(X) is given by
kf k∞ = supx∈X |f (x)|.

Remark 14.18. If fn → f pointwise in B(X) then fn ⇒ f

because fn ⇒ f implies kfn − f k∞ → 0
⇒ supx∈X |fn (x) − f (x)| → 0
⇒ fn ⇒ f .

Theorem 14.19. If (fn ) is a Cauchy sequence in B(X), then fn converges uniformly

to some f ∈ B(X). Moreover supkfn k∞ < ∞ and kfn k∞ → kf k∞ as n → ∞.

Proof. Since (fn ) is Cauchy in B(X), kfn − fm k∞ → 0 as n, m → ∞.

Then |fn (x) − fm (x)| ≤ supx∈X |fn (x) − fm (x)| = kfn − fm k∞ → 0 as n, m → ∞.
Therefore, (fn (x)) is Cauchy in R and hence converges, say to f (x), for all x ∈ X.
∴ limn→∞ fn (x) = f (x).Then

|f (x)| = | lim fn (x)|

n→∞

≤ lim |fn (x)|

n→∞

≤ lim kfn k∞
n→∞

Since (fn ) is Cauchy in B(X), it is bounded.i.e. ∃C, such that kfn k∞ ≤ C for all n.
⇒ |f (x)| ≤ C and hence f ∈ B(X).
Now we show that fn ⇒ f .

|fn (x) − f (x)| = | lim fn (x) − fm (x)|

m→∞

≤ lim sup|fn (x) − fm (x)|

m→∞

= lim kfn − fm k∞
m→∞

104 Dr.B. Surendranath Reddy

15 The Weierstrass Theorem

Since limm→∞ kfn − fm k∞ → 0 as n → ∞, |fn (x) − f (x)| → 0 as n → ∞.

∴ fn converges uniformly to f .
Also kfn k∞ ≤ C for all n implies supn kfn k∞ ≤ C < ∞.
 
Since kfn k∞ − kf k∞  ≤ kfn − f k → 0, we have kfn k∞ → kf k∞ .
 

15 The Weierstrass Theorem

Let C[a, b] be the space of real valued continuous functions on [a, b].i.e
C[a, b] = {f : [a, b] → R | f is continuous}

Lemma 15.1. There is a linear isometry from C[0, 1] onto C[a, b] that maps polyno-
mials to polynomials.

x−a
Proof. Define σ : [a, b] → [0, 1] by σ(x) = b−a
.
Then σ and its inverse σ −1 = a + t(b − a) are continuous. Also σ is bijective.
Therefore σ is a homeomorphism.
Now define Tσ : C[0, 1] → C[a, b] by Tσ (f ) = f ◦ σ.
Then Tσ is well defined and

kTσ (f )k∞ = kf ◦ σk = supx∈[a,b] |(f ◦ σ)(x)|

= supx∈[a,b] |f (σ(x))|

= supt∈[0,1] |f (t)|

= kf k∞

∴ Tσ is an isometry from C[0, 1] to C[a, b].

105 Dr.B. Surendranath Reddy

15 The Weierstrass Theorem

n
ak tk be a polynomial in C[0, 1]. Then for any x ∈ [a, b]
P
Let P (t) =
k=0

Tσ (P )(x) = (P ◦ σ)(x)

= P (σ(x))
x − a
=P
b−a
Xn  x − a k
= ak
k=0
b−a
n
X ak
= (x − a)k , which is a polynomial inC[a, b].
k=0
(b − a)k

Definition 15.2. Given f ∈ C[0, 1], the sequence (Bn (f ))∞

n=1 of Bernstein’s polyno-

mial for f is given by

Xn  k n
Bn (f )(x) = f xk (1 − x)n−k 0≤x≤1
k=0
n k

Notation: Let f0 (x) = 1, f1 (x) = x and f2 (x) = x2 .

Lemma 15.3. 1. Bn (f0 ) = f0 and Bn (f1 ) = f1

2. Bn (f2 ) = (1 − n1 )f2 + n1 f1 and hence Bn (f2 ) ⇒ f2

n  2

k n x(1−x) 1
xk (1 − x)n−k =
P
3. n
−x k n
< 4n
k=0

4. Given δ > 0 and 0 ≤ x ≤ 1, let F denotes the set of k in {0, 1, 2, ..., n} for
P n
k
which | nk − x| ≥ δ. Then k
1
x (1 − x)n−k ≤ 4nδ 2.
k∈F

106 Dr.B. Surendranath Reddy

15 The Weierstrass Theorem

Proof. 1.
n
X  k n
Bn (f0 )(x) = f0 xk (1 − x)n−k
k=0
n k
n  
X n k
= x (1 − x)n−k
k=0
k

= (x + 1 − x)n = 1 = f0 (x)

n
X  k n
Bn (f1 )(x) = f1 xk (1 − x)n−k
k=0
n k
n
X k n
= xk (1 − x)n−k
k=0
n k
n  
X k n k
= x (1 − x)n−k
k=1
n k
n  
X n−1 k
= x (1 − x)n−k
k=1
k−1
n−1  
X n − 1 j+1
= x (1 − x)n−j−1
j=0
j
n−1  
X n−1 j
=x x (1 − x)n−1−j
j=0
j

= x(x + 1 − x)n−1 = x = f1 (x)

∴ Bn (f0 ) = f0 and Bn (f1 ) = f1 .

107 Dr.B. Surendranath Reddy

15 The Weierstrass Theorem

2.
n
X  k n
Bn (f2 )(x) = f2 xk (1 − x)n−k
k=0
n k
n
k2 n k
X  
= 2 k
x (1 − x)n−k
k=1
n
n  
X k n−1 k
= x (1 − x)n−k
k=1
n k−1
n   n  
X k−1 n−1 k n−k
X 1 n−1 k
= x (1 − x) + x (1 − x)n−k
k=1
n k − 1 k=1
n k − 1
n   n  
X n−1 n−2 k n−k 1 X n−1 k
= x (1 − x) + x (1 − x)n−k
k=2
n k − 2 n k=1
k − 1
n−2   n−1  
n − 1 X n − 2 j+2 n−k−2 1 X n − 1 j+1
= x (1 − x) + x (1 − x)n−j−1
n j=0 j n j=0 j
n−1 2 1
= x + x
n n
 1 1
= 1− f2 (x) + f1 (x)
n n

∴ Bn (f2 ) = (1 − n1 )f2 + n1 f1 .

1 1
kBn (f2 ) − f2 k∞ = k(1 − )f2 + f1 k∞
n n
1 1 1
= k (f1 − f2 )k∞ ≤ → 0 (∵ kf1 − f2 k = sup|x − x2 | = )
n 4n 4

∴ Bn (f2 ) ⇒ f2 .

3.
n  2 n
X k
−x xk (1 − x)n−k = Bn (f2 ) + x2 (1) − 2xBn (f1 )
k=0
n k
1 1
= (1 − )f2 + f1 + x2 − 2xf1
n n
x(1 − x)
=
n

108 Dr.B. Surendranath Reddy

15 The Weierstrass Theorem

Since the maximum value of x(1 − x) on [0, 1] is 14 , we get

n  2

k n
xk (1 − x)n−k = x(1−x) 1
P
n
− x k n
< 4n .
k=0

4. If k ∈ F , then | nk − x| ≥ δ
 2
k
⇒ n − x ≥ δ2
 2
⇒ 1 ≤ δ12 nk − x . Therefore,

X n X 1 k 2 n
k n−k
x (1 − x) ≤ 2 n
−x xk (1 − x)n−k
k∈F
k k∈F
δ k
n 2 n
X 1 k 1
≤ 2
−x xk (1 − x)n−k ≤
k=0
δ n k 4nδ 2

Theorem 15.4 (Weierstrass Approximation Theorem). Given f ∈ C[a, b] and  >

0, there is a polynomial p such that kf − pk∞ < . Hence there is a sequence of
polynomials (pn ) such that pn ⇒ f on [a, b].

Proof. Since C[0, 1] and C[a, b] are isometric and this isometry maps polynomails to
polynomails, it is enough to prove the result for any f ∈ C[0, 1].
So consider f ∈ C[0, 1] and  > 0. Since [0, 1] is compact, f is uniformly continuous.

Therefore, ∃δ > 0 such that |f (x) − f (y)| < 2
whenever |x − y| < δ.

Xn  k n
|Bn (f )(x) − f (x)| = | f xk (1 − x)n−k − f (x)|
k=0
n k
n  
X n k  k 
=| x (1 − x)n−k f ( ) − f (x) |
k=0
k n
n
X n    k 
≤ xk (1 − x)n−k | f ( ) − f (x) |
k=0
k n

109 Dr.B. Surendranath Reddy

16 Equicontinuity

n
Let F = {k | | nk − x| ≥ δ}. Then 1


xk (1 − x)n−k ≤
P
k 4nδ 2
.
k∈F
If k 6∈ F , then | nk − x| < δ ⇒ k
f (x)| < 2 . Therefore, we get
|f ( n −
X n  k 
|Bn (f )(x) − f (x)| ≤ xk (1 − x)n−k | f ( ) − f (x) |
k∈F
k n
X n    k 
+ xk (1 − x)n−k | f ( ) − f (x) |
k n
k∈F
/
1 
≤ 2kf k∞ 2
+
4nδ 2

We want to find N such that ∀n ≥ N , kBn (f ) − f k∞ < 

kf k∞ 
i.e. we want N such that 2N δ 2
+ 2
<
kf k∞
i.e. we want N such that δ 2
<N
kf k∞
By Archimedean principle, there exists N such that δ 2
<N
Therefore, n ≥ N , kBn (f ) − f k∞ < .
Let p(x) = BN (f )(x), then kf − pk < .
Now the sequence pn (x) = Bn (f )(x) converges uniformly to f .

16 Equicontinuity
Definition 16.1. Uniformly Bounded
A collection of real valued functions F on X is said to be uniformly bounded if the
set {f (X)|x ∈ X, f ∈ F } is bounded in R.
i.e. supf ∈F supx∈X |f (x)| < ∞
i.e. supf ∈F ||f ||∞ < ∞

Definition 16.2. Equicontinuous

Let F be a real valued continuous functions on a metric space X then F is called
equicontinuous if given  > 0, ∃ δ > 0 such that
whenever x, y ∈ X with d(x, y) < δ then |f (x) − f (y)| <  ∀ f ∈ F .

110 Dr.B. Surendranath Reddy

16 Equicontinuity

Remark 16.3. Equicontinuous is also known as uniformly equicontinuous or some

times uniform continuous.

Example 16.4. Any finite subset of C(X), C(X) = {f : X → R|f is continuous.} is

equicontinuous where X is compact.
i.e. {f1 , f2 , ...fn } is equicontinuous ∀ n ≥ 1, fi ∈ C(X).

Proof. Let  > 0, since each fi is continuous on X and X is compact.

⇒ each fi is uniformly continuous.
⇒ ∃ δi > 0 such that whenever d(x, y) < δi then |fi (x) − fi (y)| < .
Now take δ = min{δ1 , δ2 , ...δn } then δ > 0.
Let x, y ∈ X such that d(x, y) < δ then d(x, y) < δ ≤ δi ∀ i
⇒ d(x, y) < δi ∀ i
⇒ |fi (x) − fi (y)| <  ∀ i
⇒ {f1 , f2 , ...fn } is equicontinuous.

Lemma 16.5. Let X be a compact metric space if F is totally bounded subset of

C(X) then F is uniformly bounded and equicontinuous.

Proof. Since F is totally bounded it is bounded.

⇒ supx∈X |f (x)| < ∞ ∀ f ∈ F
⇒ supf ∈F supx∈X |f (x)| < ∞
⇒ F is uniformly bounded.
Let  > 0 since F is totally bounded ∃ f1 , f2 , ...fn ∈ F such that for any f ∈ F such

that ||f − fi ||∞ < 3
for some i.
Since {f1 , f2 , ...fn } is equicontinuous, ∃ δ > 0 such that
whenever d(x, y) < δ then |fi (x) − fi (y)| < 3 ∀ i.

111 Dr.B. Surendranath Reddy

16 Equicontinuity

Let f ∈ F and d(x, y) < δ since f ∈ F, ∃ i such that ||f − fi ||∞ < 3 .

N ow |f (x) − f (y)| ≤ |f (x) − fi (x)| + |fi (x) − fi (y)| + |fi (y) − f (y)|
  
< + + =
3 3 3

⇒ F is equicontinuous.

Theorem 16.6. The Arzela-Ascoli Theorm

Let X be a compact metric space and let F be a subset of C(X) then F is compact if
and only if F is closed, uniformly bounded and equicontinuous.

Proof. Let F is compact then F is complete and totally bounded. Since F is complete,
it is closed.
Since F is totally bounded it is bounded.
⇒ supx∈X |f (x)| < ∞ ∀ f ∈ F
⇒ supf ∈F supx∈X |f (x)| < ∞
⇒ F is uniformly bounded.
Let  > 0 since F is totally bounded, ∃ f1 , f2 , ...fn ∈ F such that for any f ∈ F such

that ||f − fi ||∞ < 3
for some i.
Since {f1 , f2 , ...fn } is equicontinuous ∃ δ > 0 such that
whenever d(x, y) < δ then |fi (x) − fi (y)| < 3 ∀ i.
Let f ∈ F and d(x, y) < δ since f ∈ F ∃ i such that,
||f − fi ||∞ < 3 .

N ow, |f (x) − f (y) ≤ |f (x) − fi (x)| + |fi (x) − fi (y)| + |fi (y) − f (y)|
  
< + + = .
3 3 3

⇒ F is equicontinuous.
conversely, suppose F is closed, uniformly bounded and equicontinuous. Since F is a
closed subset of C(X) which is complete.

112 Dr.B. Surendranath Reddy

16 Equicontinuity

⇒ F is complete.
To prove F is compact, it is enough to prove F is totally bounded that is every
sequence in F has a Cauchy subsequence, in F .
Let (fn ) be a sequence in F . Since F is equicontinuous, for given  > 0, ∃ δ > 0 such
that
whenever d(x, y) < δ then |f (x) − f (y)| < 3 ∀ f ∈ F .
In particular, |fn (x) − fn (y)| < 3 ∀ n.
Since X is totally bounded, ∃ x1 , x2 , ...xr ∈ X such that for any x ∈ X ∃ i, 1 ≤ i ≤ r
such that d(x, xi ) < δ.
Since F is uniformly bounded {fn (xi )}∞
n=1 is bounded in R for i = 1, 2, ...r. Hence

there exists a subsequence (fnk (xi )) which converges in R and hence Cauchy.
⇒ ∃ Ni such that |fnp − fnq | < 3 ∀ p, q ≥ Ni
Take N = max{N1 , N2 , ...Nr } then for any p, q ≥ N
⇒ p, q ≥ N ≥ Ni ∀ i
⇒ |fnp (xi ) − fnq (xi )| < 3 | ∀ p, q, ≥ N and ∀ i.
Let x ∈ X then d(x, xi ) < δ for some i.
|fnp − fnq | ≤ |fnp − fnp (xi )| + |fnq (xi ) − fnp (xi )| + |fnq (xi ) − fnq (x)|
Since each fn is uniformly continuous, and d(x, xi ) < δ
⇒ |fn (x) − fn (xi )| < 3 ∀ n

∴ |fnp (x) − fnq | ≤ 3
+ 3 + 
3

∴ (fnk ) is a Cauchy subsequence of (fn ).

∴ F is totally bounded.

113 Dr.B. Surendranath Reddy

17 Infinitely Differentiable Functions

17 Infinitely Differentiable Functions

Let C ∞ (R) is the class of infinitely differentiable functions f : R → R.

i.e. C ∞ (R) = {f : R → R|f has continuous derivatives of all orders.}

= {f : R → R|f k (x) exists and is continuous.}

Lemma 17.1. There is F ∈ C ∞ (R) such that f (x) = 0 for x ≤ 0 and f (x) > 0 for
x > 0.

0 if x ≤ 0

Proof. Define f (x) =
e− x1

if x > 0.
It is clear that f is infintely differentiable every where except possible at x = 0.
Note that for x > 0,
0 1
f (x) = x−2 .e− x = x−2 F (x)
00 1
and f (x) = (x−4 − 2x−3 )e− x .
By induction, for x > 0,
f ( k)(x) = pk (x−1 )F (x)
where pk is polynomial of degree at most 2k and for x < 0, f k (x) = 0
So we have to show that f k (x) exists and is continuous at x = 0.
yn ym
For anyy > 0, ey = ∞
P
n=0 n! > m! for m = 0, 1, 2, ..

∴ For x > 0, 0 < f (x) = e−1 x < m!xm , m = 0, 1, 2, ...

∴ f (x) → 0 as x → 0
f (x)
Similarly, x
→ 0 as x → 0
0
∴ f exists and is continuous at x = 0.
Now, assume that pk exists and is continuous at x = 0 then since F k (x) = pk (x−1 )f (x)
f k (x)
x
= x−1 pk (x−1 )F (x) = qk (x−1 f (x)
where qk is a polynomial of degree at most 2k + 1 and f (x) < (2k + 2)!x2k+2
f k (x)
∴ x
< qk (x−1 )(2k + 2)!x2k+2 → 0 as x → 0

114 Dr.B. Surendranath Reddy

18 The Stone-Weierstrass Theorem

∴ f k+1 (x) exists at x = 0 and is equal to 0.

i.e. f k+1 (x) exists at x = 0 and is equal to 0.
i.e. f k+1 (0) = 0
Similarly, f k+1 (x) = pk+1 (x−1 f (x) → 0 as x → 0.
∴ f k+1 is continuous at x = 0.
∴By induction, f k (x) exists and is continuous at x = 0.

18 The Stone-Weierstrass Theorem

Definition 18.1. Let A be collection of real valued function on some set X (i.e.A =
{f : X → R|f be a real valued fucntion}. We say that ’A’ separates points in X if
given x 6= y ∈ X then ∃ f ∈ A such that f (x) 6= f (y). We say that A vanishes at no
points of X, if given x ∈ X, ∃ f ∈ A such that f (x) 6= 0.

Example 18.2. 1. The single function f (x) = x (i.e.A = {f }). Clearly it sepa-
rates points in [a, b] [∵ if x 6= y then f (X) 6= f (y)] And the single function
e(x) = 1 (i.e. A = {e}) vanishes at no points of [a, b] (∵ if x ∈ X then ∃ e such
that e(x) = 1 6= 0).
∵ Any subalgebra A of C[a, b] containing these two functions separates points
in [a, b] and vanishes at no points of [a, b].

2. For any metric space X, the collection of C(x) separates point in X and van-
ishes at no pint of x.
Since, identity function f (x) = x and constant function e(X) = 1 are continu-
ous.

3. E=Collection of even function in C[−1, 1] fails to separates points in [−1, 1].

Since, x 6= −x but f (X) = f (−x) ∀f even function and vanishes at no points
of [−1, 1] as e(X) ∈ E.

115 Dr.B. Surendranath Reddy

18 The Stone-Weierstrass Theorem

0
4. The set E of odd function in C[−1, 1] separates point in [−1, 1] as identity
0 0
function ∈ E and vanishes at 0 as f (0) = 0 ∀f ∈ E .

5. The set of all functions f ∈ C[−1, 1] such that f (0) = 0. Separates points in
C[−1, 1] as identity function is there in there in the collection and vanishes at
’0’.

Lemma 18.3. Let A be an algebra of real valued function on set X. Suppose that
A separates points in X and vanishes at no point of X. Then given x 6= y ∈ X and
a, b ∈ R ∃ f ∈ A such that f (x) = a and f (y) = b.

Proof. Since, A separates point in X and x 6= y ∈ X ⇒ ∃ g ∈ A such that g(x) 6= g(y).

Also, since, A vanishes at no points of X and x, y ∈ X
⇒ ∃ h, k ∈ A such that h(x) 6= 0 and k(y) 6= 0
u(T ) = g(t)h(t) − g(y)h(t) and
v(t) = g(t)k(t) − g(x)k(t)
Since, A is an algebra, u, v ∈ A
Also, u(x) = g(x)h(x) − g(y)h(x) = (g(x) − g(y)) h(x) 6= 0 and
v(x) = 0, v(y) 6= 0, u(y) = 0
u(t) v(t)
Define f (t) = a. u(x) + b. v(y) then f ∈ A
and f (x) = a. u(x)
u(x)
+ b. v(x)
v(y)
=a+0=a
u(y)
f (y) = a u(x) + b. v(y)
v(y)
=0+b=b

Theorem 18.4. The Stone-Weierstrass Theorem

Let X be a compact metric space and let ’A’ be a sub algebra of C(X). If ’A’ separates
point in X and vanishes at no point X then A is dense in C(x).

Proof. Without loss of generality we assume that A is closed and A is sublattice of

C(X).
Step I : Given any f ∈ C(x), x ∈ X and  > 0 we will show that ∃ g(x) in A with

116 Dr.B. Surendranath Reddy

18 The Stone-Weierstrass Theorem

gx (X) = f (X) and gX (y) > f (y) ∀y ∈ X.

Since, A separates point in X and vanishes at no point of X for any y ∈ X, y 6= x
then ∃ hy ∈ A such that hy (X) − f (X) and hy (y) = f (y).

Let Uy = {t ∈ X|hy (t) > f (t) − }

= {t ∈ X|(hy − f )(t) − }

= (hy − f )−1 (−, ∞)

Since, hy − f is continuous and (−t, ∞) is open.

⇒ (hy − f )−1 (−, ∞) is open.
⇒ Uy is open for all y ∈ X.
Since f (x) is greater than f (x) −  and f (x) > f (y) −  and hy (y) > f (y) − .
⇒ x, y ∈ Vy .
Then this Vy where y 6= x is an open cover for X. Since, X is compact it has finite
subcover. Now let gx = max{hy1 , hy2 , ...hyn }. Since, A is a sublattice of C(X), gx ∈ A
then

gx (x) = max{hy1 (X), hy2 (x), ...hyn (x)}

= max{f (x), f (x), ...f (x)}

= f (x)

For any y ∈ X, ∃ some i such that y ∈ Uyi

⇒ hyi (y) > f (y) − 
Since,gx = max{hy1 .hy2 , ...hyn }
⇒ g( x) ≥ hyi
⇒ gx (y) ≥ hyi (y) > f (y) − 
∴ gx (y) > f (Y ) − 
Step II: Given f ∈ C(X) and  > 0, ∃ h ∈ A such that ||f − g||∞ <  from step I for

117 Dr.B. Surendranath Reddy

19 Inequalities

each x ∈ X ∃ gx ∈ A such that gx (x) = f (X), gx (y) > f (y) − .

LetVX ={t ∈ X|gx (t) < f (t) + }

= {t ∈ X|(gx − f )(t) < }

= (gx − f )−1 (−∞, )

Since, g(x) − f is continuous and (−∞, ) is open, (gx − f )−1 (−∞, ) is open.
⇒ Vx is open ∀ x ∈ X
⇒ {Vx }x∈X is an open cover for X. Since, X is compact, X = ∪ni=1 Vxi .
Let h = min{gx1 , gx2 , ...gxm } since gxi (y) > f (y) −  ∀y ∈ X
h(y) > f (y) −  ∀y ∈ X
⇒ h(y) − f (y) > − ∀y ∈ X
Since, any y ∈ X∃ such thaty ∈ Vxi
⇒ gxi < f (y) + 
⇒ h( y) ≤ gxi (y) < f (y) + 
⇒ h(y) − f (y) <  ∀x ∈ X
⇒ ||h − f ||∞ < 
∴ A is dense in C(X).

19 Inequalities
Lemma 19.1. Cauchy Schwartz Inequality
∞
X
|xi .yi | ≤ ||x||2 .||y||2 ∀ x, y ∈ l2
i=1

Proof. Assume that x 6= 0 and y 6= 0 then

for any t ∈ R, ∞ 2
P
i=1 (txi + yi ) > 0

⇒ ∞ 2 2 2
P
i=1 (t xi + yi + 2txi .yi ) ≥ 0

⇒ t2 ∞
P 2
P∞ 2 P∞
i=1 x i + i=1 y i + 2t i=1 xi yi ≥ 0

118 Dr.B. Surendranath Reddy

19 Inequalities

P∞ 2
P∞ P∞
Let a = i=1 xi , b= i=1 xi yi , c= i=1 yi2
⇒ at2 + 2tb + c ≥ 0 ∀t ∈ R
In particular, for t = − ab

2
a − ab + 2 − ab b + c ≥ 0

b2 2
⇒ a
− 2 ba + c ≥ 0
⇒ −b2 + ac ≤ 0
⇒ b2 ≤ ac
√
⇒ |b| ≤ ac
12 12
⇒ |Σ∞ ∞ 2 ∞ 2
i=1 xi .yi | ≤ Σi=1 (|xi | ) Σi=1 (|yi | )

In particular, for (|xi |), (|yi |) the above inequality is true.

12 12
⇒ |Σ∞ ∞ 2 ∞ 2
i=1 |xi |.|yi || ≤ (Σi=1 |xi | ) . (Σi=1 |yi | )
12 12
⇒ Σ∞ ∞ 2 ∞ 2
i=1 |xi |.|yi | ≤ (Σi=1 |xi | ) . (Σi=1 |yi | )

⇒ Σ∞
i=1 |xi |.|yi | ≤ ||x||2 .||y||2

Example 19.2. Show that ||.||2 is a norm on l2 .

Proof. We will prove that ||x + y||2 ≤ ||x||2 + ||y||2

Since |xi + yi | ≤ |xi | + |yi |.
⇒ ∞
P 2
P∞ 2
i=1 |xi + yi | ≤ i=1 (|xi | + |yi |)

= ∞
P 2
P∞ 2
P∞
i=1 |xi | + i=1 |yi | + 2 i=1 |xi yi |

By Cauchy-Schwartz inequality,
∞
X ∞
X ∞
X
2 2
|xi + yi | ≤ |xi | + |yi |2 + 2||x||2 .||y||2
i=1 i=1 i=1

= ||x||22 + ||y||22 + 2||x||2 ||y||2 + 2||x||2 .||y||2 = (||x||2 + ||y||2 )2

1
⇒( ∞ 2 2
P
i=1 |xi + yi | ) ≤ ||x||2 + ||y||2

||x + y||2 ≤ ||x||2 + ||y||2

∴ ||.||2 is a norm on l2

119 Dr.B. Surendranath Reddy

19 Inequalities

Lemma 19.3. Let 1 ≤ p < ∞ and a, b ≥ 0 then

(a + b)p ≤ 2p (ap + bp ). Consequently x + y ∈ lp whenever x, y ∈ lp .

Proof. Given that a, b ≥ 0 as a, b ≤ max{a, b}

⇒ a + b ≤ 2 max{a, b}

⇒ (a + b)p ≤ 2p (max{a, b})p

= 2p (max{ap , bp })

⇒ (a + b)p ≤ 2p (ap + bp )

Since x, y ∈ lp ,
P∞ p
P∞
⇒ i=1 |xi | < ∞ i=1 |yi |p < ∞
Since |xi + yi | ≤ |xi | + |yi |
|xi + yi |p ≤ (|xi | + |yi |)p
P∞ p
P∞ p
i=1 |xi + yi | ≤ i=1 (|xi | + |yi |)

Since |xi |, |yi | ≥ 0 ∀i

⇒ (|xi | + |yi |)p ≤ 2p (|xi |p + |yi |p )
∞
X ∞
X
∴ |xi + yi |p ≤ 2p (|xi |p + |yi |p )
i=1 i=1
∞
X
p
≤2 (|xi |p + |yi |p )
i=1

<∞

⇒ x + y ∈ lp

Lemma 19.4. Young’s Inequality Let 1 < p < ∞ and 0 q 0 be defined by 1

p
+ 1
q
=1
then for any a, b ≥ 0 we have
ap b q
ab ≤ +
p q

120 Dr.B. Surendranath Reddy

19 Inequalities

1 1
Proof. Since p
+ q
= 1.
1 1 p−1
⇒ q
=1− p
= p
p
⇒q= p−1
>1
⇒ q > 1.
Define f (x) = xp−1 and g(x) = xq−1 ∀ ≥ 0x
then (f ◦ g)(x) = f (xq−1 ) = x(p−1)(q−1) = x.
Similarly, (g ◦ f )(x) = x
. ∴ f & g are inverse to each other.
Consider the graph of two functions y = xp−1 for 0 ≤ x ≤ a and x = y q−1 for
0 ≤ y ≤ b the the area of the rectangle is at most equal to the sum of the areas of
the above two curves.
Z a Z b
p−1
ab ≤ x dx + y q−1 dy
0 0
ap b q
ab ≤ +
p q

Lemma 19.5. Holder’s Inequality Let 1 < p < ∞ and 0 q 0 be defined by 1

p
+ 1q = 1
then for any x, y ∈ lp we have
∞
X
|xi + yi | ≤ ||x||p .||y||q
i=1

Proof. If ||x||p = 0 or ||y||q = 0 then result holds so assume that ||x||p > 0 and
||y||q > 0.
|xi | |yi |
Let a = ||x||p
and b = ||y||p
then a, b ≥ 0.
∴ By Young’s Inequality,
ap bq
ab ≤ +
p q

121 Dr.B. Surendranath Reddy

19 Inequalities

 p  q
|xi |
⇒ . |yi |
||x||p | ||y||q |
≤ 1
p
|xi |
||x||p |
|yi |
+
||y||q |
1
q
|xi | |yi | PN 1  |xi | p 
|yi |
q
⇒ N
P PN 1
.
i=1 ||x||p | ||y||q | ≤ i=1 p ||x||pp + i=1 q ||y||qq
N PN
≤ p1 ||x||
1 p 1 1 q
P
p
p i=1 |x i | + q ||y|| q
q i=1 |yi |
P 
N p 1
Since i=1 |xi | p
= ||x||p where x = (x1 , ...xN ).

N
1 X 1 1 p 1 1 q
⇒ . |xi .yi | ≤ . p .||x||p + . q .||y||q
||x||p .||y||q i=1 p ||x||p q ||y||q
1 1
= +
p q
=1
N
1 X
⇒ . |xi .yi | ≤ 1
||x||p .||y||q i=1
N
X
⇒ |xi .yi | ≤ ||x||p .||y||q ∀N.
i=1
X∞
⇒ |xi .yi | ≤ ||x||p .||y||q .
i=1

Remark 19.6. If p = q = 2 above inequality is same as Cauchy- Schwartz inequality.

Lemma 19.7. Minkowski’s Inequality Let 1 < p < ∞ and 0 q 0 be defined by

1 1
p
+ q
= 1 then for any x, y ∈ lp we have

||x + y||p ≤ ||x||p + ||y||p

Proof. Since x, y ∈ lp ⇒ x + y ∈ lp .
1 1
Let zi = |xi + yi |p−1 we will show that z ∈ lq where p
+ q
=1

122 Dr.B. Surendranath Reddy

19 Inequalities

∞ ∞
X
q
X
q
|zi | = |xi + yi |p−1
i=1 i=1
∞
q
X
= |xi + yi |(p−1)
i=1

1 1 1 p−1
p
+ q
=1⇒ q
= p

⇒ p = (p − 1).q
∞
X ∞
X
q
|zi | = |xi + yi |p < ∞............(1)
i=1 i=1

1
⇒ z ∈ lp , and ||z||q = ( ∞ q q
P
i |zi | )
1
also ||x + y||p = ( ∞ p p
P
i |xi + yi | )

⇒ ||x + y||pp = ( ∞ p p
P
i |xi + yi | ) ...............(2)

∞
! 1q
X
∴ ||z||q = |zi |q
i

1q
= ||x + y||pp
p

= ||x + y||pq

= ||x + y||p−1
p

123 Dr.B. Surendranath Reddy

19 Inequalities

Now consider,

|xi + yi |p = |xi + yi |.|xi + yi |p−1

≤ (|xi | + |yi |).|xi + yi |p−1

≤ (|xi | + |yi |).|zi |

≤ |xi |.|zi | + |yi |.|zi |

∞
X ∞
X ∞
X
⇒ |xi + yi |p ≤ |xi .zi | + |yi .zi |
i i i

By Holder’s Inequality,
P∞
i |xi yi | ≤ ||x||p .||z||q

and ∞
P
i |zi yi | ≤ ||z||p .||y||q

∞
X
∴ |xi + yi |p ≤||x||p .||z||q + ||y||p .||z||q
i

= (||x||p + ||y||p ).||z||q

∴ ||x + y||pp ≤ (||x||p + ||y||p ).||x + y||p−1

p

||x + y||pp
⇒ ≤ ||x||p + ||y||p
||x + y||p−1
p

⇒ ||x + y||p ≤ ||x||p + ||y||p

124 Dr.B. Surendranath Reddy