Functional Analysis by R.

Vittal Rao

Lecture 4: Metric Spaces - June 1, 2012

For real numbers, if x, y ∈ R, we define a distance from x to y as 2 2

Example 4.4 X = R = {xp= (x1 , x2 ) : x1 , x2 ∈ R}. x, y ∈ R .
d(x, y) = |x − y|. We observe:
Metric is defined as ρ2 = (x1 − y1 )2 + (x2 − y2 )2 .
(i) d(x, y) ≥ 0. Equality iff x = y.
(i) ρ2 (x, y) ≥ 0, equality iff x = y.
(ii) d(x, y) = d(y, x).
2
(ii) ρ2 (x, y) = ρ2 (y, x) ∀x, y ∈ R .
(iii) d(x, y) ≤ d(x, z) + d(z, y) ∀x, y, z ∈ R.   12
2
(iii) X 
ρ2 (x, y) =  2
 
|xi − yi | 

Definition 4.1 Metric Space  
Let X be any non-empty set. Let ρ : X×X → R be s.t. ∀x, y, z ∈ X,
 
i=1
  12
2
(i) ρ(x, y) ≥ 0. Equality iff x = y.
X 
= + 2
 
|x − z z − y |


 i i i i 
(ii) ρ(x, y) = ρ(y, x).
 
i=1
  12   12
2 2
(iii) ρ(x, y) ≤ ρ(x, z) + ρ(y, z) (Triangle inequality)
X   X 
2
+ 2
   
≤ |x − z | |y − z |
 

 i i 
 
 i i 

   
then, we say ρ is a metric on the set X. X together with the metric i=1 i=1

‘ρ’ is called a metric space (X, ρ). (by Minkowski)

∴ ρ2 (x, y) ≤ ρ2 (x, z) + ρ2 (z, y).

Example 4.1 (R, d) as specified above is a metric space. ‘d’
2 2
is called the usual metric on R. Hence, ρ2 is a metric on R ; (R , ρ2 ) is a metric space. ρ2 is
2
called usual metric on R .
Example 4.2 Let X be any non-empty set. We define a metric
on X as follows: 2
(
0 if x = y Example 4.5 Similarly, for p ≥ 1, on R , we can define
ρD (x, y) =
1 if x , y   p1
2
Easy to see that ρD is a metric on X. This is called discrete X 
ρp (x, y) =  p
 
|xi − yi | 

metric on X. (X, ρD ) is a metric space. 
 

i=1

Example 4.3 In particular, R with the discrete metric is a Again verify triangular inequality using Minkowski.
2 2
metric space (R, ρD ) [which is different from (R, d)]. ∴ ρp is a metric on R . (R , ρp ) is a metric space.

1
 CHAPTER 4. METRIC SPACES - JUNE 1, 2012

Example 4.6 We can further generalize Example 4.5 as fol- Remark 4.1 Subspace
lows: Let (X, ρ) be any metric space. Let Y be a non-empty subset
n
Let X = R and p ≥ 1. Define of X. ρ is also a metric on Y, and (Y, ρ) is a metric space, called
subspace of the original metric space.
  1p
Xn 
ρp (x, y) =  p
 
|xi − yi | 


 

i=1
Example 4.9 X = {a, b, c}. Define a metric on X as follows:
Again use Minkowski to get triangle inequality. ρ(a, a) = ρ(b, b) = ρ(c, c) = 0
n n
∴ (R , ρp ), (C , ρp ) are metric spaces for any positive integer ρ(a, b) = 1 = ρ(b, a)
n, for any p ≥ 1. ρ(a, c) = 2 = ρ(c, a)
ρ(b, c) = 3 = ρ(c, b)
n n (X, ρ) is a metric space.
Example 4.7 X = R or C .
We define for x, y ∈ X,

ρ∞ (x, y) = max |xi − yi | Example 4.10 X = Set of all metro stations.

1≤i≤n (
metro line distance if they are in the same line
ρ(s1 , s2 ) =
n n
Clearly, ρ∞ is a metric. (R , ρ∞ ) and (C , ρ∞ ) are metric d(s1 , T) + d(s2 , T) if on different lines
spaces. ∴ (X, ρ) is a metric space.

Notation:
For any positive integer n, 1 ≤ p ≤ ∞;
n p
(R , ρp ) is denoted by ln (R) and
n p
(C , ρp ) is denoted by ln (C).

T
Examplen 4.8 X = Collection of all binary o strings of length N Terminus
i.e., X = x = (x1 , x2 , . . . , xN ) : xi ∈ {0, 1} .
Define for x, y ∈ X,
N
X
H(x, y) = |xi − yi |
i=1

i.e., H(x, y) = Number of coordinates of x that differ from that

of y.
This is a metric on X called the Hamming metric. Line B Line A
∴ (X, H) is a metric space.

2
 CHAPTER 4. METRIC SPACES - JUNE 1, 2012

Example 4.11 General Metro Example 4.13 Let X be the set of all 2 × 2 real matrices.
The distance defined similar to Example 4.10. ( ! )
a11 a12
X= A= : aij ∈ R
a21 a22
!
a11 a12
Take any matrix A ∈ X, A = . Clearly any such
a21 a22
 a11 
 
 a 
4
matrix can be thought of as an element  12  ∈ R .
 
 a21 
a22
4
∴ Can invoke all metrics of R to define metric on X.
|a11 | + |a12 | = R1
|a21 | + |a22 | = R2
Define NA = max{R1 , R2 }.
For C, D ∈ X, we define ρ(C, D) = NC−D .

The next step is to move to ‘infinite’ dimensional spaces.

p
ln (R) → lp (R)
p
ln (C) → lp (C) for 1 ≤ p ≤ ∞.

Example 4.14
 ∞
X 
X = x = (x1 , . . . , xn , . . .) : x j ∈ R, |x j |p < ∞, p ≥ 1
Example 4.12 Circular distance j=1

X = [0, 2π) i.e., {x ∈ R : 0 ≤ x < 2π}. For any x, y ∈ X, we define the metric
The usual metric on R, namely, ρ(x, y) = |x− y| is also a metric
on X. We look at another metric on this set. 
∞
 1p
ρ(x, y) = min{|x − y|, 2π − |x − y|}.
X 
ρp (x, y) =  |x j − y j | 
p

y x 
j=1


The fact that this is a metric is obvious, the triangle inequality

0 x y 2π being a consequence of Minkowski’s inequality.
Linear Distance ∴ (X, ρp ) is a metric space.
Circular Distance We shall denote this by lp (R). Similarly, we have lp (C).

3
 CHAPTER 4. METRIC SPACES - JUNE 1, 2012

Example 4.15
X = {x = (x1 , . . . , xn , . . .) : the sequence is bounded, x j ∈ R}
i.e., ∃M s.t. |xn | ≤ M ∀n.
∴ ∃a l.u.b. for the sequence which we denote by Nx .
For any x, y ∈ X, we define ρ∞ (x, y) = Nx−y .
Clearly, this is a metric on X. The metric space (X, ρ∞ ) is
denoted by l∞ (R). Similarly, we have l∞ (C).

Example 4.16 Consider an interval I in R.

X is the collection of all functions
f : I → R s.t. f is bounded. I
i.e., ∃M s.t. | f (x)| ≤ M ∀x ∈ I.
We denote the l.u.b. of f by N f .
For f, g ∈ X, we define ρ( f, g) = N f −g .
Clearly, this is a metric on X. We denote (X, ρ) by BR (I).
Similarly, we have the metric space BC (I).